Visualisation de l`écoulement dans un système vis/fourreau. Suivi en
Transcription
Visualisation de l`écoulement dans un système vis/fourreau. Suivi en
N° d’ordre 2008-ISAL-0124 Année 2008 Thèse PhD dissertation Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. Visualization of the flow in a screw/barrel system. Real-time monitoring of polymer/clay mixing. Présentée devant Presented to Institut National des Sciences Appliquées (INSA) de Lyon Pour obtenir le grade de In partial fulfillment of the requirements for the Degree of Docteur Doctor of Philosophy selon l’arrêté du 30/03/1992 according to the French law (decree of 30/03/1992) Dans la spécialité In the area of expertise Matériaux Polymères et Composites Polymer and Composites Materials Au sein de In the frame of École Doctorale Matériaux de Lyon Antonella ESPOSITO Ingénieur des Matériaux Materials Engineer Soutenue le 5 décembre 2008 devant la commission d’examen: Defended on December 5th 2008 in the presence of the examination jury: Rapporteur José Maria KENNY (Professor, University of Perugia, Italy) Rapporteur Serge BOURBIGOT (Professeur, ENSC Lille) Examinateur Jean-François GERARD (Professeur, INSA de Lyon) Examinateur Jean-Jacques FLAT (Ingénieur Docteur, ARKEMA CERDATO) Directeur de Thèse J.-Y. CHARMEAU (Professeur, INSA de Lyon) Directrice de Thèse J. DUCHET-RUMEAU (MdC, INSA de Lyon) Ingénierie des Matériaux Polymères (IMP) – UMR CNRS #5223 Laboratoire des Matériaux Macromoléculaires (LMM) Referee Referee Examiner Examiner PhD advisor PhD advisor a papà Antonio Senza il tuo incoraggiamento non sarei arrivata fino a qui a mamma Ivana Il mio punto di partenza, sei tu... a Feliciano Segui sempre la tua strada Andata e Ritorno sed ad meliora semper e poi a tutte le persone che amo e che mi vogliono bene, e che per questa pura e semplice ragione m’hanno sempre e da sempre sostenuta, incoraggiata, e soprattutto attesa Un pensiero particolare va ai miei nonni, che vedono il tempo fuggire via senza fare i bagagli, per i quali il mio ritorno non è mai troppo presto, e che spero non partiranno mai senza di me. Jesus ALAJARIN MARTINEZ Pierre ALCOUFFE (et le Vietnam) Firas ALMOUSTAFA Jenny ALONGI (e il famoso baffo) Marco AMICI (e Irena) Angelica ANTONIELLI D’OULX (Erasmus) Kamira AOUACHRIA (et la rose du désert) Rafael ARIOLI (et le collier de coquillages) Murat ARLI Maëlenn AUFRAY Ludivine AUGRY (et son papa) France AZZOPARDI Céline BAGUENARD Jean BALCAEN (et la Visiovis) Liubov BARDACH Najoua BARHOUMI (et le voile interdit) Amélie BARON Claire BARRES (et Arianne…) Julien BARRUET Ahmed Ali BASFAR (et la rose rouge) Sylvain BAUDU (et Céline) Ludovic BEAL Pierre BEAUDET Redha BELLA (et Ninou) Ditza BEN-ZION Hynek BENES (et toutes ses Ianas) Mallaouia BENGOUA (Mallou, et l’histoire sans fin du complément labo…) Gwladys BENISTANT (et Alain) Bruno BERAL (SAMPE) Yves BEREAUX (et les randonnées et les crêpes et la musique) Vincent BERLIER Julien BERNARD Sandra BINAULD (et Morgan) Françoise BISCARA (et le squatteur toxicomane) Anne BLOND (et la cabine peinture) Gisèle BOITEUX (mais n’en parlons pas…) Liliane BOKOBZA (JNC15) Dimitris BOLLAS (et Natasa et Nikos et Ppappas) Mélanie BORDES Luigi BOTTA Ghalia BOUKELLAL (JEPO35) Serge BOURBIGOT (soutenance de thèse) M’hamed BOUTAOUS (et la sacoche volée) Séverine BOYER (JEPO35) Nadia BRAHMIA (et Mikaël) Raphaël BRUNEL (et chaque jeudi soir en sortant du labo) Annarita CASCONE (et Pascal) Philippe CASSAGNAU (et tous ses chapeaux) Nelly CHAGNEUX (JEPO35) Mikaël CHAILLY (et Nadia) Aurélia CHARLOT Jean-Yves CHARMEAU (et Odile et Juliette et Mathilde) Thomas CHAUSSEE (JEPO35) Jiang-Feng CHEN (et les cours de chinois et la soupe au chocolat) Sambor CHHAY Antonio CHIECHI (SAMPE Europe) Ecaterina CHILAT (JEPO35) Jae-Won CHO (et la tortue dorée) Dieu Huong CHU Nicolas CINAUSERO (SAMPE France) Sandra CLEMENSON (JEPO35) Floraine COLLETTE (JEPO35) Gaël COLOMINES Yves-Marie CORRE Magali COULAUD (et les pauses café et les "petites" mésaventures…) Cécile COUSTAL (et Séb et tout le reste…) Romain CREAC'HCADEC (JNC15) Agnes CREPET Caroline CREUZET Lizong DAI Céline DAMEZ Emmanuelle DANET (et la marseillaise!) Emilie DEGOUE Guillaume DELAITTRE (JEPO35) Maxime DERANCY (et le bon voisinage) Sandra DIRE Eric DROCKENMULLER (et Francesca) Jannick DUCHET-RUMEAU (et Pascal) Florence DUPASQUIER (et la Thaïlande) Jérôme DUPUY Lama ELIAS Françoise FENOUILLOT Christiane FERRET (et le 1er cour de français) Paola FINOTELLO (Erasmus) Jean-Jacques FLAT (soutenance de thèse) Etienne FLEURY Nadège FOLLAIN Nicolas FORTIN Elsa FRANCHINI (et ses chignons et Abdou) Olivier GAIN (et la moto) Jean-Paul GALLETIER (SAMPE) Arnaud GALLON (et Claire et Mathis) Jocelyne GALY Franck GAUDIN Estelle GAUTIER (et Loïc) Jean-François GERARD (le chef !) Frédéric GILLE (et les scouts et les concerts et la politique et la patate) Nelly GIMENEZ (et Jérôme et Isabelle et le nouveau né) Henri GIRARDY (SAMPE) Emilie GONCALVES Guilene GOOUREY Fabrice GOUANVE Joackym GRANAL (SAMPE) Raffaella GUBBIOTTI (la Raf nazionale) Nadia GUERROUANI (et son pourrimère) Abdoulallye GUEYE Johann GUILLEMINOT (SAMPE France) Chady HAKME Lofti HAMITOUCHE (SAMPE France) Sanna HAVERI (SAMPE Europe) Lina HENAO (et la Colombie et sa bonne humeur… et Michelin) Marcelo HERNANDEZ-AVILA (et le Mexique) Christian HOCHARD (JNC15) Amélie HOUEL (et son pH…) Hikmet HOUICHI (et les randonnées et la boue et la chanson des schtroumpfs et la femme qui parlait au basilic) Ye HUA Marie-Pierre INNOCENTE (et son/mon appartement dans les pentes) Emilie JACQUELOT (et son bébé) Bénédicte JACQUETFRANCILLON (et un bocal de cornichons) Boris JAFFRENNOU Stéphane JEOL James JOHNSON (SAMPE) Christophe KAUFFMANN (SAMPE) José Maria KENNY (le grand chef!) Greet KERCKHHOFS (SAMPE Europe) Magorzata KNAPIK (Nanofun-Poly) Constandinos the first KONSTANTIS (SAMPE) Adrie KWAKERNAAK (SAMPE) Jean LACHAUD (SAMPE France) Sébastien LADET (et la Cécile nationale) Nadir LAHLALI Khalid le maître LAMNAWAR Yoann LAMY Fabrice LANDAIS (SAMPE) Brigitte LATOUR (et le 2ème cour de français) Massimo LAZZARI (NanofunPoly) Jihean LEE (SAMPE USA-Europe) Michael LEMOINE (JEPO35) Pierre LISSAC (JMC) Sébastien LIVI Frédéric LORTIE (le grand… tout court!) Luis LUNA PEREZ (un des cours de français) Abdou MAAZOUZ Sara MAIEZ-TRIBUT (et Laurent et le gros ventre et puis Rémi) Francesca MANGONI (e la mia tesi di laurea) Gilbert MARTIGNAGO Françoise MARTIN (ou Sylvain) Matthieu MARTINEAU (Erasmus) Valérie MASSARDIER Benoît MALLET Françoise MECHIN Andrea MEDICI (et Nanofire et Elsa) Flavien MELIS Julien MESNAGER (et la visite chez Michelin) Marie-Laure MILAN (et le changement de bureau) Cheima MNEKBI Nizar MNIF Mahdi MNIF (et le kidnapping et la Tunisie que j’aime) Maël MOGUEDET (et Véronique et ses deux anges) Saber MOHAMMADI (et sa petite famille) Yannick MOLMERET (JEPO35) Fabrice MONTEZIN (JEPO35) Bruno MOUGIN (et la philo-photographie) Jean-Damien MULLER (et piou-piou) Marc MULLER (SAMPE Europe) Amapola MUNUERA (SAMPE Europe) Loubna NAJEMI (JEPO35) Addisalem NEGASH (cours de français) Van NGO THI THANH Viviane O’BRIEN (et Lyon International) Meigui OU (cours de français) Jean-Pierre PASCAULT (le grand grand chef !) Michela PASQUALETTI (Miky ou Pas ?) Francesca PEDITTO (et Eric) Jean-Marc PELLETIER (et l’Ecole Doctorale) Hervé PERIER-CAMBY (ou Gilbert) Edith PEUVREL-DISDIER (JEPO35) Guillaume PIBRE Emilie PICARD Lionel PICARD (et Hélène) Pascal PICHON Maribel PINEDA (et le Mexique et Rafael et Ximena) Delphine PLATEL Pavel PODANY (SAMPE Europe) Tomasz POKROPSKI (et la Pologne, and a year in the merde) Isabelle POLO (et Evan et son deuxième bout de chou Eléa) Isabelle PONCELET (ma che ?! et sa maman Lucienne, son fils Sylvain, et puis Thiaïs et Poussy et tous les autres animaux…) Simina POPA-NITA Julien PORTAL Daniel PORTINHA Patricia POULARD (SAMPE) Charlyse POUTEAU (PEP et RAID DAHU 2008) Arnaud PREBE (et Céline) Olivier RACCURT (CEA) Corinne RATTON-BENOIT (et ses deux merveilleux enfants) Raphaël REGISSER Fanny RICHARD (JEPO35) Rocio-Noemi RIVAS-ARAIZA Maurice ROCHET (NCT ABG) Steven RODGERS (SAMPE) Laurent ROUGEAU Alain ROUSSEAU (et la responsabilité ATG) François ROUSSET Riccardo RUGGERONE (SAMPE Europe) Pascal RUMEAU (et Jannick) Daniela RUSU (JEPO35) René SAINT-LOUP (et la facette noire de la noblesse française) Xavier SAMAIN (JEPO35) Mara SAPONARO (et le TER LyonGenève) Henry SAUTEREAU (et tous les salons du vin) Adil SBIAI Henrik SCHMIDT (SAMPE Europe) Domenico SCIABOLETTA (et le noir de carbone) Elena SERRANO-TORREGROSA (et le tabagisme passif qui fait du bien) Romain SESCOUSSE (JEPO35) Gérard SEYTRE Keyhan SHAHRIARI (cours de français) Layth SLIMAN (cours de français) Rodolphe SONNIER (et la République Tchèque) Manel SORBA Pavla ŠVIGLEROVA Emilie TAILLON Valeria TAGLIAZUCCA Agnieszka TERCJAKS (et Elena et la boucle est bouclée) Julie TEUWEN (SAMPE Europe) Marie-Claire THIVEND (et l’aide à l’insertion professionnelle) Morgan TIZZOTTI (et Sandra) Hervé TOLLENAERE (molto così) Valentina TORNABENE (e zio Franco Giancane) Alain TRANQUARD (et Gwladys) Christophe TRAVELET (JEPO35) Laurent TRIBUT (et Sara) Matthieu VALLE Pascale VALOT (et la Côte d’Azur) Perrine VAN NIEUWENHUYSE Philippe VAUTEY (SAMPE) Karel VELECHOVSKY (et le Beaujolais) Ruben VERA (et les rayons X) Ronen VERKER (SAMPE Europe) Jérémy VIALE (et Tours) Guy VILLEVIEILLE (et le bouquin italien) Demian VON OSTEN (Erasmus) Paul WILLIAMSON (SAMPE Europe) Aristide WOLFROM (et Question pour un Champion) Nikos ZAFEIROPULOS (Nanofun-Poly) Nathalie ZYDOWICZ Evgeny ZELIKMANN (Nanofun-Poly)… Ero bambina e già se ne parlava: la fuga dei cervelli. In tutta onestà, non credevo potesse succedere anche a me… non in Francia, e soprattutto non così giovane. E invece eccomi qua, alla fine di tre anni gradevolmente sofferti sulla tesi di dottorato, alla soglia di un quarto anno da insegnante/ricercatrice assunta con contratto a tempo determinato, ma espatriata a tempo indeterminato e… senza più cervello (dicono che sia in fuga). C’est vraiment la fin des haricots. Cette thèse a été financée par le Ministère National de l’Enseignement Supérieur et de la Recherche (MNESR) français, mais s’inscrit dans le cadre du réseau européen d’excellence pour les nanotechnologies NanoFun-Poly (Nanostructured and Functional Polymer-based Materials and Nanocomposites). Ce réseau a été constitué en 2004 et est actuellement en train d’évoluer vers un European Centre for Nanostructured Polymers (ECNP). Le financement était destiné à un(e) étudiant(e) ressortissant(e) d’un des pays européens hormis la France : c’est justement grâce à la communication promue au sein du réseau que ce travail a pu voir le jour. TABLE OF CONTENTS INTRODUCTION Chapter I 1 PROCESSING OF NANOCOMPOSITES Melt compounding 3 I-1 --------- MIXING AND PROCESSING ------------------------------------------------------ 6 I-1.1 Melt processing in screw/barrel systems ................................................ 8 I-1.2 Melt conveying in the meter section .................................................... 15 I-1.3 Mixing in the molten state.................................................................... 17 I-1.3.1 Mixing steps ..................................................................................... 17 I-1.3.2 Laminar mixing ............................................................................... 18 I-1.3.3 Mixing of highly viscous fluids by helicoidal screws ...................... 21 I-1.3.4 Mixing of solid particles with a molten polymer ............................. 23 I-1.3.5 Distributive and dispersive mixing .................................................. 26 I-2 --------- MACRO-, MICRO-, NANOCOMPOSITES --------------------------------------- 27 I-2.1 Nanocomposite morphology Agglomeration, aggregation, dispersion and distribution .................... 31 I-2.2 Techniques for morphological analysis................................................ 37 I-2.2.1 Morphological characterizations ex situ ......................................... 38 Transmission Electron Microscopy (TEM) 38 X-Ray Diffractometry (XRD) .................... 41 Other techniques......................................... 44 I-2.2.2 Morphological characterizations in situ ......................................... 48 Winch ......................................................... 50 Laser Doppler Velocimetry (LDV) ............ 50 Particle Imaging Velocimetry (PIV) .......... 52 Antonella ESPOSITO PhD INSA de Lyon (2008) Table of contents 1/5 Ultrasound Doppler Velocimetry (UDV) ... 53 Other techniques ......................................... 54 I-2.2.3 Local probes .................................................................................... 55 Laser Induced Fluorescence (LIF) .............. 55 Concentration field ..................................... 57 Mixing time ................................................ 57 I-2.2.4 Computer simulation........................................................................ 60 Identification of chaotic flow zones ........... 60 Quantification of laminar mixing ............... 61 I-3 ---------- PHOTO-FUNCTIONALIZATION OF NANOFILLERS ---------------------------- 62 I-3.1 Lamellar fillers...................................................................................... 63 I-3.1.1 Structure and chemistry ................................................................... 63 I-3.1.2 Photo-functionalization methods ..................................................... 65 I-4 ---------- A DEEPER INSIGHT INTO THE STATE OF THE ART --------------------------- 67 I-R --------- REFERENCES ------------------------------------------------------------------- 85 Chapter II PHOTO-FUNCTIONALIZATION Lamellar fillers 95 II-1 --------- MATERIALS -------------------------------------------------------------------- 96 II-2 --------- PHOTO-FUNCTIONALIZATION METHODS ------------------------------------ 98 II-2.1 Clay swelling (A) ..................................................................................98 II-2.2 Dry compounding (B) ...........................................................................99 II-2.3 Melt compounding (C) .........................................................................99 II-2.4 Cation exchange processing (D) .........................................................100 II-3 --------- CHARACTERIZATIONS -------------------------------------------------------- 101 II-3.1 X-Ray Diffractometry (XRD).............................................................102 II-3.2 ThermoGravimetric Analysis (TGA) .................................................102 II-3.3 Elemental Analysis (EA) ....................................................................103 II-3.4 Fourier Transform InfraRed (FTIR) spectroscopy .............................104 II-3.5 TGA coupled to FTIR spectroscopy (TGA-FTIR) .............................105 II-3.6 Spectrofluorimetry ..............................................................................106 II-4 --------- REFERENCE MEASUREMENTS ----------------------------------------------- 108 Antonella ESPOSITO PhD INSA de Lyon (2008) Table of contents 2/5 II-4.1 Reference XRD measurements .......................................................... 108 II-4.2 Reference TGA measurements........................................................... 109 II-4.3 Reference EA measurements ............................................................. 111 II-4.4 Reference FTIR spectra ...................................................................... 112 II-4.5 Reference fluorescence spectra .......................................................... 115 II-5--------- EVALUATION OF THE PERFORMED PHOTO-FUNCTIONALIZATION METHODS AND CHOICE OF THE PROTOCOL -------------------------------- 116 II-6--------- OPTIMIZATION OF THE PROTOCOL FOR CATION EXHANGE PROCESSING (D) ----------------------------------- 122 II-6.1 Choice of the fluorescent molecule .................................................... 123 II-6.2 Influence of the fluorescent molecule concentration ......................... 129 II-6.3 Complementary characterizations (C30B 0.25MC RhP) ................... 133 II-6.4 Further general comments about the efficiency of the photo-functionalization ...................................................................... 139 II-7--------- CONCLUSIONS ---------------------------------------------------------------- 144 II-R -------- REFERENCES ------------------------------------------------------------------ 146 Chapter III PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing 149 III-1-------- MATERIALS ------------------------------------------------------------------- 150 III-2-------- PHOTO-FUNCTIONALIZATION PROTOCOL ---------------------------------- 152 III-3-------- CHARACTERIZATIONS -------------------------------------------------------- 152 III-4-------- REFERENCE MEASUREMENTS ----------------------------------------------- 152 III-4.1 Reference XRD measurements .......................................................... 152 III-4.2 Reference TGA measurements........................................................... 154 III-4.3 Reference EA measurements ............................................................. 156 III-4.4 Reference FTIR spectra ...................................................................... 157 III-4.5 Reference fluorescence spectra .......................................................... 160 III-5-------- CHARACTERIZATION OF THE PHOTO-FUNCTIONAL INORGANIC/ORGANIC COMPLEXES Antonella ESPOSITO PhD INSA de Lyon (2008) ----------------------------------------- 160 Table of contents 3/5 III-5.1 Photo-responsive CNa+ 0.25CEC RhP ...............................................160 III-5.2 Photo-responsive C30B 0.25MC RhP ................................................161 III-5.3 Photo-responsive C10A 0.25MC RhP ................................................162 III-5.4 Photo-responsive C15A 0.25MC RhP ................................................164 III-5.5 Comparison of the photo-responsive complexes ................................166 III-6 -------- CONCLUSIONS----------------------------------------------------------------- 172 III-R ------- REFERENCES ------------------------------------------------------------------ 174 Chapter IV PROCESSING Real-time monitoring of mixing 175 IV-1-------- VISIOVIS ----------------------------------------------------------------------- 176 IV-1.1 Original configuration ........................................................................177 IV-1.1.1 Components and utilization ...........................................................179 IV-1.1.2 Advantages and limitations ............................................................182 IV-1.2 Evolutions of the configuration ..........................................................186 IV-1.2.1 From 3D lighting to 2D laser plan ................................................187 IV-1.2.2 Position of the CCD cameras ........................................................192 IV-1.2.3 Optical fiber and in-line spectrofluorimetry ..................................193 IV-1.2.4 Calibration of the detection systems .............................................195 IV-1.3 Actual configuration ...........................................................................202 IV-1.3.1 Objectives .......................................................................................204 IV-2-------- EXPERIMENTAL PROTOCOL -------------------------------------------------- 205 IV-2.1 Acquisition of data ..............................................................................210 IV-2.1.1 Images ............................................................................................211 IV-2.1.2 Videos.............................................................................................211 IV-2.1.3 In-line fluorescence spectra ...........................................................213 IV-3-------- PROCESSING OF THE ACQUIRED DATA-------------------------------------- 215 IV-3.1 Images .................................................................................................215 IV-3.1.1 Standard deviation of image luminosity ........................................217 IV-3.1.2 Discrete Fourier Transform (DFT) of textured images .................218 Antonella ESPOSITO PhD INSA de Lyon (2008) Table of contents 4/5 IV-3.1.3 Validation of data processing ........................................................ 225 IV-3.2 Videos................................................................................................. 231 IV-3.3 Fluorescence spectra .......................................................................... 232 IV-4 ------- SOME EARLY RESULTS ------------------------------------------------------- 233 IV-4.1 Comparison of different photo-active lamellar fillers ........................ 233 IV-4.2 Comparison of different amounts of filler.......................................... 237 IV-4.3 Regulation of the back pressure ......................................................... 242 IV-5 ------- CONCLUSIONS ---------------------------------------------------------------- 246 IV-R ------- REFERENCES ------------------------------------------------------------------ 247 Chapter V VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration 248 V-1 -------- MATERIALS ------------------------------------------------------------------- 249 Model fluid .................................................... 249 Photo-functional fillers .................................. 254 V-2 -------- EQUIPMENT ------------------------------------------------------------------- 257 Configuration................................................. 257 Light source ................................................... 257 Barrel ............................................................. 258 Screw ............................................................. 258 Feeding .......................................................... 259 V-3 -------- PROCESSING OF THE ACQUIRED DATA ------------------------------------- 259 V-4 -------- INTERPRETATION AND VALIDATION OF THE RESULTS -------------------- 260 V-4.1 Real-time sampling ............................................................................ 261 XRD .............................................................. 262 Rheology ....................................................... 264 V-4.2 Computer simulation .......................................................................... 266 V-R -------- REFERENCES ------------------------------------------------------------------ 267 CONCLUSIONS 269 APPENDIX Résumé détaillé en français Antonella ESPOSITO PhD INSA de Lyon (2008) Table of contents 5/5 INTRODUCTION We live in a time which cherishes the extremes – when people dream of having the biggest benefits with the smallest effort. It’s happening to the materials engineers and to the chemists, as well… the smallest the length scale on which they can control the matter, the biggest their hopes of getting incredible performances. The question is: should we get satisfied with a power we have only on a laboratory scale? Polymer/clay nanocomposites are the new Eldorado for materials engineers: the number of scientific papers published on this topic in unceasingly increasing. The progresses accomplished in the research laboratories are pushing the global market of the advanced materials to turn to the world of the nanocomposites, but the reality is that no market segments (or only limited ones) will free up if the scientific progresses won’t be accompanied by a proved economic efficiency. Nanocomposites will be considered economically efficient when it will be possible to process them by conventional tools, following a protocol whose outcomes can be guaranteed on an industrial scale, that is for medium to large production rates. It is time to evaluate whether the existing processing tools can be used Antonella ESPOSITO 1 INTRODUCTION also to obtain this new class of materials, or they should be adapted to satisfy the actual and future technological needs, and how. The answers come from a clear definition of the objectives, the understanding of the mechanisms involved, a deep knowledge of the available tools and the capability of envisioning which is the best solution to capitalize on (or modify) them. This work represents a starting point in this direction. Our objective was to set up a new method to visualize how do lamellar fillers (e.g. clays) – which are susceptible of generating nanocomposite morphologies – behave when mixed with molten polymers in a screw/barrel system. Chapter I will dive deeper into topics such as melt compounding, the mixing mechanisms, the nanocomposites and their processing, the issues about their morphology and its characterization, the notions of aggregation, dispersion, distribution, etc. An extensive review of the state of the art about process monitoring by fluorescence will also be presented. Chapter II will answer to the question “How to visualize a multiscale lamellar filler which doesn’t have any native optical activity? Which procedure to photo-functionalize clays?”. Chapter III will prove that clay photo-functionalization can be performed by cation exchange processing, and the experimental protocol established and optimized in Chapter II is suitable to photo-functionalize several commercial clays. In Chapter IV we present Visiovis, the original tool which symbolizes our answer to the question “How to visualize polymer/clay compounding in a screw/barrel system, whose walls are generally opaque? How to detect multi-scale lamellar fillers, how to process the acquired data, and how to interpret the results?”. In Chapter V we’ll make a point about the progresses accomplished on Visiovis, trying to identify the difficulties we have not yet overcome and suggesting some solutions for further ameliorations, whenever a problem has been encountered. A brief conclusion precedes the Appendix, in which we reported the Matlab codes written purposely for Visiovis data processing. We preferred not to abuse of the Appendix to promote a regular rhythm of reading. Each chapter can be read independently from the others, as we regularly reminded the global context of the study. A multimedia CD-Rom accompanies the manuscript, for all the contributions which cannot be written down or otherwise commented. The reader disposes also of a bookmark, particularly useful because it contains any otherwiseforgotten-or-probably-lost chemical formula and sample nomenclature. Enjoy! PhD INSA de Lyon (2008) 2 Chapter I PROCESSING OF NANOCOMPOSITES Melt compounding Polymer properties directly depend on polymer chemistry. All the polymers that are members of the same “family” are uniquely identified by their identical “chemical origin” (the repeating unit) and approximately have the same physical, mechanical and thermal properties – which differentiate them from the members of the other polymer families. Sometimes, even though they belong to the same family, polymers can slightly differ from each other because some of their properties are influenced by a specific choice of processing route and parameters. Undeniably, significant modifications of the properties of a given polymer can be achieved by its copolymerization in presence of one or more monomers having different chemical structures – by the way, shouldn‟t we consider any copolymer as the member of a brand new family of polymers, rather than Antonella ESPOSITO 3 Chapter I as the modification of an existing one? All of this just to emphasize the fact that the only way left to significantly modify and tailor polymer properties without changing its intimate chemistry is compounding – which is nothing else than adding to the polymer some other components, chemically and/or physically heterogeneous, and dispersing them into the host matrix – typically by melt processing. Additives contribute to the formation of a new material (a polymer composite) which keeps belonging to the same family and nevertheless has different properties with respect to the neat polymer. The properties of a polymer composite can diverge from those of the neat polymer matrix to different extents, according to the chemical, physical and morphological nature of the additives, as well as their physical and/or chemical interactions with the host polymer matrix [1]. Any processing step involving compounding (e.g. the formulation of the masterbatch to produce composite pellets by extrusion) could be crucial for the whole processing of polymer composite materials. The factors which can originate most of the problems encountered during compounding are [2]: the chemistry of the polymer matrix (which can engender a semicrystalline or an amorphous structure and which determines the glass transition temperature, the melt rheological behavior, etc.); the physical-chemical properties of the modifiers (inorganic or organic fillers, pigments, flame retardants, UV radiation absorbers, curing agents, etc.); the percentage of modifier to be added to the mixture; the method chosen for compounding. As a result, quality and performances of polymer composites don‟t exclusively depend on the physical and chemical properties of the ingredients chosen for its receipt, but also on the capability of the processing step to achieve the best dispersion of all the components of the mixture in the entire volume of the processed polymer – to an extent which should allow to assume that any portion of the polymer composite, whatever its location, has the same target properties [1]. Compounding (to tell it differently: the production of a masterbatch, and thus the formulation of composite pellets) typically involves three steps – all of them critical to the achievement of good results [2]: feeding, mixing and pelletizing. PhD INSA de Lyon (2008) 4 PROCESSING OF NANOCOMPOSITES Melt compounding The choice of a proper feeding system is crucial to a point that it often influences the results of compounding in spite of the fact that the subsequent step of mixing may be highly performing. Traditionally, dry fillers are added directly to the molten polymer evolving into the extruder – but it frequently happens that shearing isn‟t sufficient to break filler aggregates and agglomerates. Certainly, it isn‟t accidentally that Dupuy and Bussi [3] patented a new dispersion process for submicron fillers in thermoplastics. The aim was to optimize filler desagglomeration before feeding it into the extruder: the filler is suspended in a liquid phase – which helps desagglomeration – then fed to the system by pumping the suspension at the entrance of the extruder, where pellets are still solid. In relation to mixing, it must guarantee some specific conditions but, primarily, it should assure the achievement of an adequate degree of dispersion of the modifiers into the polymer matrix and a good homogenization of the compound. Mixing is typically performed by an extruder (single- or twin-screw, the latter being co-rotating or counterrotating) or by an internal mixer (continuous or discontinuous). Last step involves the fabrication of pellets from the formulated compound in the molten state and, analogously to the previous steps, may also influence the final results: the equipments for injection molding, for instance, require pellets having all the same dimensions, regularly shaped and homogeneous to assure good performances [2]. If it has been proved that high-quality products obtained by injection molding require highquality1 pellets, nobody could exactly tell whether this condition is necessary and also sufficient, or it is only necessary and thus requires something else to get good products. Indeed, both extrusion and injection molding involve melt processing: the objective of the former is to formulate composite pellets, whereas the latter makes the composite pellets melt again in order to reshape the material in the form of the final product. In addition, both extrusion and injection molding are performed in more or less complex systems (screw/barrel systems). In conclusion, the question we should try to answer to is the following: how do fillers exactly behave when mixed to a molten polymer evolving into a screw/barrel system? 1 We stressed that “high-quality composite pellets” means “regularly-shaped, uniformly size-distributed and homogeneous pellets”. Starting from now, “high-quality” will implicitly and more specifically means “characterized by a good filler dispersion and distribution into the polymer matrix”. Antonella ESPOSITO 5 Chapter I I-1 MIXING AND PROCESSING Up to now, we‟ve just reminded something universally recognized and admitted: mixing is a crucial issue for polymer composites processing. Generally speaking, the combined activities of mixing and stirring are certainly very ancient and maybe represent one of the very first organized activities of the human beings: probably, mixing two or more ingredients to get a final product has always been such an ordinary and almost instinctive doing that rarely captivated our attention. Little by little (and much earlier than the first industrial revolution), these two activities have become much more than a hobby for (processing) engineers. By the way, it is worthy to emphasize that, as any other engineering activity, mixing and stirring have always been subjected to requirements that, of course, engineers couldn‟t really decide to ignore (e.g. some volume and/or dimension limitation, the reutilization of an existing container, the presence of other interfering tool or, occasionally, a specific behavior of the ingredients of the mixture). Anyhow, having one or more requirements to fulfill doesn‟t prevent from making (reasonable) suggestions to ameliorate both the processing equipments and parameters. Sometimes, the simple fact of knowing the advantages and (even more interesting!) the disadvantages of a given processing system is worthy investigating. In recent times, as the oil industry progressed, the increased need for mixing organic (and frequently viscous) fluids introduced a further requirement: the necessity of shearing (not only stirring!) mixtures. As the market expanded and more products derived from the differentiation of the oil industry – that is, as the market of plastics developed – the equipments specifically devoted to highly performing mixing became more and more complex. Today, mixing and processing are more than ever strictly connected. Nowadays, several diversified tools are available to stir a fluid, or to compound it with one or more additives and then to mix everything together: rotating mechanical stirrers and mixers, vibrating systems, pumping or jet systems, equipments operating an external recirculation of the fluid, ultrasounds probes, etc. As a general rule, the stirrers explicitly designed for homogenization are highly performing for the generation of fluid movements, whilst the mixers designed to ameliorate the dispersion of a phase into PhD INSA de Lyon (2008) 6 PROCESSING OF NANOCOMPOSITES Melt compounding another are primarily performing in terms of energy dissipation [4]. Mechanical systems are by far the most common industrial systems for stirring and mixing – especially for plastics. Any other procedure is considered beneficial only when specific requirements appear to be for some reason incompatible with the traditional equipments. Ultrasounds probes, for instance, are extremely efficient to generate intense local fluid movements, to a point that they could be indispensable when mixing should be limited to small portions of the whole volume of processed material (e.g. to facilitate chemical reactions or develop specific morphologies [5][6]), but their use is limited by obvious scaling-up difficulties: that‟s why ultrasound probes haven‟t yet found large application at the industrial scale. Most of the industrial stirring and mixing processes are nowadays performed by means of tools stably joint to a rotating axis, which is itself adjusted in a cylindrical pan. The industry of plastics has developed its own specific equipments, rather adjusting a screw into a cylindrical barrel. Definitely, most industrial mechanical mixing processes involving plastics are performed in screw/barrel systems – which exist in many different configurations having a more or less complex geometry. In a still medium, mixing is controlled by diffusion and conforms to Fick law. To accelerate transfer phenomena and, thus, to accelerate mixing, it is necessary to generate a movement in the volume of the fluid (flow) so that the contact between the ingredients is regularly renewed by the mechanical action exerted by the mixer. Stirring and mixing elements are commonly classed on the basis of the direction(s) in which the fluid is expulsed when exiting the volume swept during rotation. If several stirring and mixing elements can be easily classed as axial, radial or mixed (whether they generate an axial, a radial, or a combination of axial and radial flow), the equipments used by the industry of plastics are often too complex to be unequivocally classed that easy – we previously cited screw/barrel systems. Nonetheless, performing processing requires performing mixing, which in turn greatly depends on the choice of the most suitable equipment, and obviously starts with an optimum design of the mixing elements (or, with reference to the screw/barrel systems, an optimum design of the screw profile). Once the system geometrically optimized, it is necessary to find the optimum processing parameters – rotational speed and residence time, for instance. The best way to make the best choice is to clearly know which are the desired results and how it is possible to obtain them – Antonella ESPOSITO 7 Chapter I in other words, the best choice is based on the comprehension of the phenomena (mixing mechanisms, filler dispersion and distribution) but also on the knowledge of the available processing tools (essentially the screw/barrel systems). I-1.1 Melt processing in screw/barrel systems We already called attention to the fact that compounding is typically performed by melt processing – undeniably, both extrusion and injection molding involve melting. Melting isn‟t a simple phase transition for polymers – on the contrary, it strictly depends on polymer properties. If the polymer is semicrystalline, its fusion forms a low-viscosity phase only in case the processing temperature is higher than the melting temperature of its crystalline portion – temperature which is normally rather higher than glass transition temperature. Conversely, for amorphous polymers, talking about a melting temperature is somehow meaningless: in fact, the polymer starts flowing as soon as the value of the processing temperature gets higher than its glass transition temperature. In any case, as glass transition is a phenomenon which occurs in a range of temperatures rather than at a specific temperature, as a general rule the processing temperature is set to a value considerably higher than glass transition temperature, with the intention of facilitating polymer flowing2. A certain fluidity of the processed material – more or less accentuated, depending on its molecular weight and on the mismatch between processing and glass transition temperature – helps reducing the internal friction between the macromolecular chains (intrinsically due to the fact that polymer is flowing, thus its macromolecular chains are moving with respect to each other) and could avoid the inconveniences of excessively high shear rates (e.g. chain breaking, thus reduction of the average molecular weight). By the way, applying high shear rates to the processed polymer is sometimes necessary to obtain specific (morphological) properties: melting first, then conveying and mixing phenomena are typically accompanied by a shear rate which is not only responsible for any eventual orientation of the macromolecular chains, but can also be involved in the 2 It is well known that polymer viscosity depends on temperature and typically decreases as temperature increases. PhD INSA de Lyon (2008) 8 PROCESSING OF NANOCOMPOSITES Melt compounding distribution, dispersion and orientation of reinforcing additives (fillers). These factors can significantly influence the properties of the final products [7]. Figure I-F1 Schematic representation of the main processing steps for thermoplastic polymers: extrusion (top) followed by injection molding (bottom). Composite materials in which the matrix is a thermoplastic polymer are typically processed by extrusion (the step properly devoted to compounding, i.e. during which the neat polymer is melt-mixed with all the required additives to formulate the masterbatch and then the composite pellets) followed by injection molding (the step which uses the composite pellets previously formulated by extrusion to fabricate the final products), as schematized in Figure I-F1. As a matter of fact, any thermoplastic polymer composite processed by extrusions and injection molding undergoes double melting: a first time for the formulation of the masterbatch, and a second time to be injected into the mould. In the hypothesis that the masterbatch can be considered a high-quality compound and that, hence, the composite pellets are also perfectly homogeneous (i.e. filler particles are completely dispersed and uniformly distributed into the matrix), the critical step which still has to be evaluated, in terms of performances and with particular attention to the properties of the final products, is injection molding. Since the simple fact of filling up a mould causes high gradients of velocity in the injecting channels, any process involving injection risks to be concerned by some phenomena of filler segregation in some parts Antonella ESPOSITO 9 Chapter I of the injected pieces. In the case of pigments, segregation (which essentially means bad distribution3) produces only an effect of color heterogeneity; in contrast, when fillers are meant to reinforce, any morphological phenomena directly affect mechanical properties, thus the piece doesn‟t have the same designed mechanical properties in all its portions. To avoid these problems, one solution could be to ameliorate the moulds by designing them as capable of strategically diminishing local shear [8]; anyway, even with perfect moulds, the main requirement remains a highest quality of the initial masterbatch, of the formulated composite pellets and of the melt compound in the injection screw/barrel system – to be assured by using chemical compatibilizing agents, if necessary [5]. As previously said, the screw/barrel systems for polymer and polymer composite processing are quite complex and may have different geometrical (screw) profiles. The main parameters to describe screw profiles are essentially three4: the nominal diameter (D); the total flighted length (L); the efficient flighted length, i.e. the portion of flighted length which effectively contributes to processing5. The flighted length of a plasticating screw typically consists of three sections – easily identifiable as geometrically different (Figure I-F2): the feed section – having a constant deep channel; the transition section – having a channel of decreasing depth; the meter section – having a constant swallow channel. The feed section transports the pellets received by the feeding system (which are still solid); the transition section operates the plastication process by melting pellets; the meter section (in which pellets are meant to be totally melted) conveys and keeps 3 We‟ll later get deeper into the notions of distribution, dispersion, agglomeration, aggregation… More precisely, screws are normally defined by their geometrical parameters normalized with respect to their diameter – in particular by the ratio L/D (length over nominal diameter). 5 This parameter is valid only for injection screws. Indeed, in the reciprocating injection screws, only a portion of their flighted length continuously contributes to processing: the screw extremity is unavailable when the unit is totally retracted. 4 PhD INSA de Lyon (2008) 10 PROCESSING OF NANOCOMPOSITES Melt compounding mixing the masterbatch6 to assure the best physical, chemical and thermal homogeneity of the evolving compound. Figure I-F2 Scheme illustrating the typical geometrical configuration of a plasticating screw. It is interesting to observe that the flighted length can be divided in three main sections: the feed section, the transition section and the meter section. Please note that the right portion of this scheme corresponds to the entrance of the screw/barrel system, whereas the left portion corresponds to its exit [9]. The ratio between the channel depths in the feed section and in the meter section corresponds to the screw compression ratio – a further parameter characterizing screw profiles. In addition to the abovementioned parameter, the absolute depth of the channel can also be relevant: indeed, a screw having a quite deep channel is highly performing in terms of transport, but requires high values of torque for rotation; conversely, a screw having a quite swallow channel requires higher rotational speeds to convey the fluid at the same transport rate, but also lower values of torque. Besides, the shear rate imposed to the polymer by a screw having a deep channel and rotating at low speed is less than the shear rate generated by a screw having a shallow channel but rotating at high speed. Finally, the relative lengths of feed, transition and meter sections are also responsible of 6 When preparing for injection moulding, the composite pellets formulated by extrusion can be further compounded with an additional amount of neat polymer in order to dilute the initial concentration of the masterbatch and obtain the target concentration for the final products. Antonella ESPOSITO 11 Chapter I screw performances: to improve the homogeneity of the molten compound, for instance, it may be worthy to increase the length of the meter section. Most conventional screws have a square pitch, viz. their pitch is similar to their diameter, and even if most screws have a single flight, it is possible to find screws with a double flight [10]. V tot Figure I-F3 On the left: the vector V represents the apparent total velocity of the barrel with respect to the screw and results from the addition of two components (Vz and Vx), respectively longitudinal and perpendicular to the screw flights. On the right: in the case of injection screws, the vector V must be corrected by considering the translation of the screw inside the barrel [9]. The three steps of melt processing corresponding to each of the aforementioned flighted sections of screws can be modeled by mathematical equations and, under given hypothesis, collected in a unique global model. This work is exclusively aimed to the evaluation of the last step – the one corresponding to the meter section – in which the evolving material is supposed to be fully melted. During screw rotation (the movement which, both in extrusion and injection molding units, is associated to the plastication process), convection and transportation are supported by the drag forces exerted on the matter longitudinally with respect to the flights, and the recirculation flow responsible of mixing is due to the component of the velocity perpendicular to the flights (Figure IF3). The feature which essentially makes the difference between extrusion and injection molding is the fact that, while rotating, reciprocating injection screws also translate longitudinally inside the barrel, whereas extrusion screws execute a simple rotational movement. Thus, in the case of injection molding, the velocity field globally imposed to the material evolving into the screw channels isn‟t purely circumferential, but has also a non-zero radial component (Figure I-F3, on the right). In general, the radial component PhD INSA de Lyon (2008) 12 PROCESSING OF NANOCOMPOSITES Melt compounding of velocity is rather small in comparison to the circumferential components, and the contribution due to the longitudinal translation of injection screws can be neglected: for this reason, the behavior of reciprocating injection screws has been often assimilated – just to simplify models – to the behavior of a single-screw extruder. Figure I-F4 Illustration of the main steps of the plastication process in screw/barrel systems [9]. When passing through the flighted length of the screw, the polymer experiences several effects, for each section generates different solicitations (Figure I-F4): in the feed section, the pellets are packed together and proceed as if they were an elastic solid; in the transition section, the pellets are melted by the heating system around the barrel and with the help of friction and viscous dissipation of energy; in the meter section, the molten compound is homogenized (if necessary) or simply stirred, and conveyed towards the exit of the screw/barrel system. As previously said, the role of the feed section is essentially to bring the pellets into the system and push them towards the transition section. The phenomena occurring in this first portion of the screw/barrel system are controlled by gravity, analogously to what happens in an Archimedean screw – even if for big screws rotating at high speeds the centripetal forces can become relevant, as well. Initially, the polymer pellets aren‟t at all compacted and tend to roll on each other and to rearrange; then, they‟re rapidly Antonella ESPOSITO 13 Chapter I packed together and continue proceeding as if they were a unique and elastic solid. As solicitations increase, the mechanisms become more and more dependent on the friction between the polymer and the barrel surface, as well as between the polymer and the screw surface. The feed section, anyway, is not the section which influences the most screw performances – although an inadequate feeding may have relevant effects on the following steps of melting and pumping. Generally speaking, the sections which could influence the most the performances of a screw/barrel system are the transition and the meter sections. The role of the transition section is to melt the pellets, previously compacted by the feed section. Pure solid conveying ends when a thin film of molten polymer is formed at the interface between the compact solid and the hot barrel: melting is initially controlled by the formation and development of the film of molten polymer parallel to the screw flights; then, as the film reaches a critical value of thickness 7, the action exerted by the screw flights becomes rather scraping and a pool of molten polymer in continuous recirculation is formed by the side of the pushing surface of screw flights. As a consequence, the width of the residual solid packed pellets diminishes and the pool of molten polymer grows till the melting process is complete. With reference to the meter section (melt conveying), it is worthy to underline some more differences between extrusion and injection. In single-screw extruders, it is common to use screw profiles characterized by long meter sections in order to stabilize the production rate of molten polymer and consequently assure a good homogeneity of the masterbatch: this precaution is necessary because extrusion is a continuous process and any variation directly influences the quality of the products. In the case of injection molding, conversely, melting/mixing and injection aren‟t directly connected and can be treated separately – which means that a slight variation of the processing parameters doesn‟t necessarily influence the quality of the final products or, anyway, consequences aren‟t as direct as in the case of extrusion. Moreover, extrusion requires high pressures to convey the molten polymer to the exit of the screw/barrel system, whereas injection molding exploits a total different mechanism to supply the pressure necessary to fill up 7 This critical thickness value normally corresponds to slightly more than the space between the barrel and the crest of screw flights [9]. PhD INSA de Lyon (2008) 14 PROCESSING OF NANOCOMPOSITES Melt compounding the mould. For such reasons, the meter section is relatively less important for injection molding than it is for extrusion – fundamentally because, as previously explained, the main role of the meter section is to assure that melting is complete, that the compound is homogeneous and that the pressure is sufficient to overcome the resistance due to the smaller section at the exit of the system (back pressure). Briefly, in injection screws the meter section can be rather short – surely shorter than a quarter of the overall flighted length and even shorter if the processed polymer is easily melted but highly viscous in the molten state. I-1.2 Melt conveying in the meter section A simple analysis of melt conveying shows that the flow of a Newtonian fluid in a simple rectangular channel (in the hypothesis that the curvature of the surface and the thickness of the screw flight are negligible) is determined by two contributions8 [9]: Q 2 D 2 hN sin cos 2 Dh3 sin p 12 z (I-E1) the contribution of the drag flow, i.e. the flow generated by the velocity of the screw relative to the barrel, in particular by its component parallel to the screw flights; the contribution of the pressure flow. p 0 , the second contribution is nil In the absence of a gradient of pressure z and the global flow rate Q depends only on the drag flow; conversely, in the presence of p p 0 or increases 0 a gradient of pressure, the global flow rate Q diminishes z z as shown in Figure I-F5. 8 Q is the global flow rate, D is the nominal diameter of the screw, h is the flight depth, N is the number of screw revolutions per minute, is the helix angle, p is the pressure, is the viscosity of the fluid. Antonella ESPOSITO 15 Chapter I Figure I-F5 Velocity field in a screw channel having a simple rectangular geometry and in the presence of a positive (a) and negative (b) pressure gradient, respectively [9]. If an increasing back pressure is applied to the system, the second contribution influences more and more the global flow rate Q. If some obstacles are added to make the flow more complex, Q diminishes and the residence time increases, thus mixing is improved and compound homogeneity is ameliorated, as well. When designing screw profiles, all these aspects must be taken into account to choose the optimum geometrical parameters (Figure I-F2). Meter sections design always results from a compromise between flow rate and mixing efficiency (linked to flow complexity and recirculation). Changing screw rotational speed could radically modify the melting process, as well: when speed increases, Q increases and the conversion of mechanical into thermal energy by viscous dissipation is enhanced. However, the majority of polymers have a pseudoplastic behavior – their viscosity in the molten state diminishes if the imposed shear rate increases – which limits the effect of speed rising on viscous dissipation. In reality, increasing screw rotational speed requires longer flighted sections to accomplish complete polymer melting. In order to optimize the equipments for high values of speed or flow rate, the latest trend is to use longer screws; in practice, the best performances can be obtained by simply choosing the lowest rotational speed compatible with all the PhD INSA de Lyon (2008) 16 PROCESSING OF NANOCOMPOSITES Melt compounding other processing conditions such as the time required for a complete production cycle. When production cycles have to be short, processing time must be reduced and the phenomenon which contributes the most to plastication is viscous dissipation; on the contrary, when production cycles can be long, processing time is less constraining and conductive heat transfer from the barrel surface becomes the most relevant factor. I-1.3 Mixing in the molten state Mixing is nothing else than using mechanical energy to develop a velocity field and induce a fluid movement in order to homogenize the concentration field present in a fluid and initially characterized by high gradients. In spite of its universality, mixing is poorly understood and generally difficult to characterize: the first unified treatment of the mixing of fluids from a kinematical point of view has been done by Ottino [11]. It is universally admitted that obtaining a mixture statistically perfect right down to the molecular scale is quite hard and could require intolerably long processing times. Defining the perfect mixing equipment implies the possibility of achieving the absolute homogeneity of property fields such as concentration, temperature, etc. in the entire volume of processed material. In practice, assuring such conditions is considered to be almost impossible: even if the concentration field appears perfectly homogeneous at the macroscopic scale, it could not be true at the microscopic or molecular scale. We should say: it was considered to be almost impossible. Recent progresses in materials science engaged processing engineers in increasingly tough challenges: new classes of materials morphologically structured down to the nanoscale (polymer-clay nanocomposites) are nowadays attracting the interest of both academic and industrial researchers. Henceforth, the new objective will be: obtaining high-quality compounds down to the molecular scale – possibly by means of optimized existing equipments. I-1.3.1 Mixing steps Beek and Miller (1959) described the mixing process of two fluids by four steps, summarized in Table I-T1 [4] with the corresponding mechanism and length scale. Antonella ESPOSITO 17 Chapter I Table I-T1 Mixing steps: mechanisms and relative length scales [4] MIXING STEP Macroscopic Mesoscopic Microscopic Microscopic MECHANISM Dispersion of the fluids into each other thanks to the velocity field Size reduction of the interdispersed parcels of fluid in consequence of turbulence (if present) Formation of microcoils, stretching and enrolling of spirals Molecular interdiffusion of the fluids into each other from the spirals whithin the microcoils LENGTH SCALE Macroscopic cm Taylor mm Kolmogoroff 10-100 m Batchelor 0.5-5 m If spatial scales are easily assignable, temporal scales obviously depend on the processing parameters (especially the rotational speed of the mixing element), as well as on the physical and chemical properties of the ingredients of the mixture. Laminar mixing is based on the mechanisms corresponding to the last two steps, which in this particular case we would rather define as macroscopic mixing (formation and cyclic stretching, cutting, recombining of multilayered structures 9) and microscopic mixing (molecular diffusion at the interfaces of the multilayered structures). For high viscosities and in the absence of turbulence, microscopic mixing is the slowest step, thus diffusion is the phenomenon limiting the whole mixing process. If we decompose the concentration in an average and a fluctuating value, one can affirm that macroscopic mixing corresponds to the ensemble of phenomena contributing to uniform the local average value of the concentration in a given volume of the processed mixture, whilst microscopic mixing corresponds to the phenomena contributing to the reduction of the local fluctuations with respect to the average concentration value. In the presence of fillers, these mixing steps are correlated only to distributive mixing.10 I-1.3.2 Laminar mixing The phenomena controlling flow and mixing directly depend on the conditions in which processing is carried out. In general, one distinguishes two flow regimes (laminar 9 These mechanisms will be better illustrated in the next paragraph (§ I-1.3.2). The difference between distributive and dispersive mixing will be clarified in § I-1.3.5. 10 PhD INSA de Lyon (2008) 18 PROCESSING OF NANOCOMPOSITES Melt compounding and turbulent) separated by a transition regime. It was in 1883 that Reynolds discovered the existence of different flows into a channel: its experiment consisted in visualizing a colored tracer, isokinetically injected in the centre of a transparent tube, while mixing to a fluid flowing at a known flow rate (Figure I-F6). This technique allowed to visualize the different behaviors of the tracer. The parameters which determine the flow regime during processing are grouped into the Reynolds number, representing the ratio between inertial and viscous forces11: Re D (I-E2) water tracer laminar regime water tracer water tracer transition regime water water tracer turbulent regime water Figure I-F6 Reynolds experiment [4]. The laminar flow regime occurs when flow velocity is very low or fluid dynamic viscosity is very high and is facilitated by a small characteristic dimension of the system D. In a laminar flow, the fluid can be imagined as a stack of extremely thin layers (fluid lamellae) which proceed by slipping on each other without crossing. The movement is transferred from a fluid lamella to the adjacent ones by friction. Adjacent lamellae don‟t move at the same velocity: the difference produces balancing opposite drag forces that generate a shear stress [Pa]. The gradient of velocity between two adjacent fluid lamellae associated to the shear stress is called shear rate [s-1]. is the dynamic viscosity, is the density, D is the characteristic dimension of the geometrical system and is the velocity of the fluid. 11 Antonella ESPOSITO 19 Chapter I The ratio shear stress over shear rate is defined as the fluid viscosity . Therefore, viscosity can be considered as a coefficient describing the ability of the fluid to transfer the movement from a lamella to the adjacent ones during lamellar flow. stretching cutting stretching recombining cutting Figure I-F7 Mixing mechanisms in laminar flow: stretching, cutting, recombining [4]. The mixing mechanisms for highly viscous fluids are schematically illustrated in Figure I-F7. Under the solicitations exerted by the rotating screw, the fluid stretches and assumes the shape of lamellar elements. During stretching, the contact surface between adjacent lamellae increases and the distance between them diminishes. The lamellae can also fold, get cut and recombine. This cycle of morphological modifications repeatedly occurs until the characteristic minimum lamellae thickness is achieved, viz. until fluid lamellae are sufficiently thin to consider, at the macroscopic scale, that the compound is homogeneous. Actually, as previously underlined, no mixture should be considered perfect if molecular diffusion hasn‟t rendered it homogeneous also at the microscopic (or even nanoscopic) scale; in practice, this condition has been traditionally neglected as it would have probably required too much time to be accomplished but, with the advent of polymer-clay nanocomposites, all these considerations have to be taken into account and even extended to the molecular scale (one of the mechanisms of clay exfoliation is PhD INSA de Lyon (2008) 20 PROCESSING OF NANOCOMPOSITES Melt compounding thought to be diffusion of polymer macromolecules into clay galleries 12). If stretching is assured by the velocity gradients developed into the system, cutting and recombining of several lamellae are caused by three-dimensional fluctuating phenomena which perturb the flow. An example of such periodic phenomena is the rotation of the screw itself and, more particularly, the regular movement of the flights in the processed volume of fluid. Mixing is aimed to the achievement of lamellae as thin and homogeneous as possible. I-1.3.3 Mixing of highly viscous fluids by helicoidal screws Highly viscous fluids are frequent in industrial applications – in some industrial fields (plastics, food, drug and cosmetic industries) they even represent the most used fluids. The main difficulty associated to the mixing of highly viscous fluids is that the only possible flow regime is the laminar flow, as previously enlightened. In the absence of turbulence, there are no vortexes to facilitate mixing: the movement of the fluid is only assured by the convection movement imposed by the rotating screw as well as, to a negligible extent, by molecular diffusion (Figure I-F8). Figure I-F8 Mixing mechanisms in laminar flow: convection and molecular diffusion [4]. Being aware of the limitations imposed by laminar flow, it isn‟t surprising that the equipments used to process molten polymers are specific for each application and 12 Polymer-clay nanocomposites morphology (intercalated, exfoliated, etc.) will be described in § I-2. Antonella ESPOSITO 21 Chapter I must be specifically designed for. A good design of screw profile is essential, for it is the movement induced by the rotation of the screw that directly determines the quality of mixing. It is desirable that any fluid element spends most of its residence time close to screw flight tips, in order to exploit the action of convection to the maximum extent: that‟s why the mixing elements specifically designed for highly viscous fluids typically sweep a significant volume of fluid13. Helicoidal screws are, essentially, Archimedean screws. They can continuously generate an axial movement of the fluid in both the longitudinal directions (“forwards” and “backwards”), depending on the direction of rotation. The fluid evolves in the space comprised between two adjacent screw flights, the root surface of the screw and the internal surface of the barrel. The average trajectory of any element of fluid can be statistically identified, and its knowledge may greatly help ameliorating the quality of the final products. Each elementary volume of fluid statistically follows the same trajectory and undergoes the same shear history if fluid can be considered Newtonian. Indeed, at a given temperature, the viscosity can be considered an intrinsic property of the fluid only in case it is Newtonian – for the viscosity of non-Newtonian fluids is a local property, which depends on the local conditions of flow (namely, on the shear rate imposed by the rotating screw into the screw/barrel system). The homogenization of a compound when fluids are highly viscous is a process which may be extremely long if compared to the same process performed by the same equipment on a mixture of fluids having a low viscosity. The reason, as observed15, is the absence of turbulence typical of laminar flow: the only mechanism responsible for mixing is convection, and molecular diffusion can be neglected because too slow for the industrial rhythms of production. Mixing time can be expressed in the form of an adimensional number which, in reality, stands for the number of screw revolutions necessary to achieve a mixture with a sufficiently good quality. This observation, along with all the considerations about the mechanism which prevalently influences mixing, 13 Here is the reason why screws occupy as much space as possible inside the barrel, and their diameter is often very close to the diameter of the barrel. This last feature is also due to the difficulty of stirring and renewing the fluid in proximity of the barrel surface. 14 We introduced the notion of viscosity in § I-1.3.2. 15 Find some explanations in § I-1.3.2. PhD INSA de Lyon (2008) 22 PROCESSING OF NANOCOMPOSITES Melt compounding spontaneously lead to the conclusion that mixing time depends on the geometry of the screw/barrel system but also on the rheology of the fluid. Even though – as Reynolds showed16 – mixing can be studied with the help of tracers, the introduction of additives into non-Newtonian highly viscous fluids must be careful, for macromolecular materials (and in particular their rheology) could be highly sensitive to the presence of smaller heterogeneous molecules. In other words, the risk is to alter the observed phenomena… Let‟s consider a screw/barrel system filled with a homogeneous fluid evolving in steady-state conditions. If, at a given time t 0 , a small amount of a second heterogeneous phase is fed to the system, the moieties of such added phase follow the flow established into the system until complete homogenization of the mixture (concentration C 0 ). With mixing going on, the initial concentration is gradually diluted: the function C (t ) tends to zero. The added phase can be considered as a tracer and, by definition, should fulfill the following requirements: it should have the same physical properties (i.e. the same density) of the fluid; it should be inert, i.e. neither react nor be dissolved during measurement. The time necessary for the curve C (t ) to reach its asymptote corresponds to the time spent by the tracer molecules into the screw/barrel system, i.e. the residence time. By diving C (t ) by the total amount of tracer, the normalized concentration represents the Residence Time Distribution (RTD). Since the tracer is supposed to reliably follow the flow, its behavior reproduces the behavior of the fluid. Investigations like this one are commonly used to evaluate the mixing efficiency of screw/barrel systems: what’s about mixing solid particles with a molten polymer? I-1.3.4 Mixing of solid particles with a molten polymer Processing highly viscous fluids is sometimes aimed to the creation of particular morphologies – whether the fluid is compounded or not with fillers or other additives. Indeed, textures in polymers can be due to specific macromolecular arrangements or to the presence of additional heterogeneous phases. One objective of polymer composite 16 See Figure I-F6 in § I-1.3.2. Antonella ESPOSITO 23 Chapter I processing could be, for instance, the alignment of fibers (or any other filler) or the development of a semicrystalline structure with preferentially-oriented crystallites. A system containing both melt (fluid) and solid phases presents some additional stirring problems due to the intrinsic nature of such a kind of mixture. These problems can be summarized as follows: the solid phase has a different density from the melt phase (typically higher): a phase separation by sedimentation is often spontaneous in the absence of movement – thus, stirring is firstly aimed to generate the movement to contrast sedimentation; contrarily to fluids, solids can‟t be deformed and have a completely different mechanical behavior – therefore, the presence of a solid phase generally modifies the rheological behavior of the fluid; matter and energy transfers between the phases may occur and are typically accelerated by stirring and mixing. How solid particles can remain suspended in a flowing fluid? Intuitively, if the particles are sufficiently small or have a density similar to the density of the fluid, they will behave exactly as the fluid – except for the possibility of undergoing deformations. But how to explain that particles having a size up to millimeters and a density lower or higher than the fluid can remain suspended, whereas they would float or sediment if plunged in the same fluid but in the absence of stirring? Actually, even if a solid particle is isolated, the movement of the fluid will exert on it some forces and transport it. It is evident that, for high concentrations, the shocks between particles and the resistance to the movement due to their proximity will delay the process of separation of the particles from each other – that is, dispersing and distributing particles into the fluid won‟t be the same as for lower particle concentrations. The mechanical equilibrium of one spherical particle plunged in a fluid is a topic well known by the experts of fluid dynamics and by processing engineers [4] and the sedimentation and aggregation of relatively small solid particles from liquid suspensions are phenomena frequently encountered in a variety of manufacturing processes. More details about the mixture of solid particles with liquids can be found in specialized books [12][13] and specifically dedicated works [14]. We‟d like to remind that the flow in presence of suspended solid particles can be described by PhD INSA de Lyon (2008) 24 PROCESSING OF NANOCOMPOSITES Melt compounding adimensional parameters such as Reynolds (Rep), Froude (Frp) and Archimedes (Ar) numbers. Once again, we are not particularly interested in the adimensional description of the flow in the presence of solid particles, thus we won‟t give more details; however it is noteworthy to observe that on the basis of such adimensional numbers big particles (1 mm) can follow turbulent flows, are sensitive to gravity effects and rapidly sediment, whereas smaller particles (1 m or less) are insensitive to turbulence and gravity and do not sediment: they follow the flow and behave as if the fluid was homogeneous [4]. It is clear, thus, that particle size represents a key factor drastically influencing not only flow behavior, but also the possibility of characterizing it. Melt mixing with conventional fillers to obtain polymer micro-composites is surely easier to characterize than melt mixing with smaller multiscale fillers (clays) to obtain polymer nanocomposites17. Describing particle breaking and aggregation is possible by choosing a property of interest (e.g. particle size dp) and expressing the number of particles having a given value of such property at the time t and in the point of coordinates ( x, y, z ) as n( , x, y, z, t )d . The distribution n(d p , x, y, z , t ) is the solution of an equation of partial derivatives called “equilibrium of particle populations” which can be written as following (in its general form and for a finite volume V into which the number n can be considered as a constant) [4]: (Vn) Qs ns V [ Dbr Dag ] Qe ne V [ Bbr Bag ] t (I-E3) The distributions B and D respectively indicate the number of particles appearing or disappearing in consequence of breaking (br) of particles with d > dp or aggregation (ag) of particles with d < dp, per unit volume and time18. The possibility of taking into account particle breaking and aggregation implies a potentially more complex analysis of the flow – indeed, in the presence of a filler to be mixed up with a molten polymer, both distributive and dispersive mixing are important to obtain high quality products. If distributive mixing is moderately affected by particle morphology and size (except for sedimentation, concerning mostly big and dense particles), dispersive mixing is much 17 The notions of micro- and nanocomposites will be introduced in § I-2. The hypothesis is that no nucleation, no crystalline growth, no precipitation and no dissolution of the particles occur. 18 Antonella ESPOSITO 25 Chapter I more related to the physical properties of the filler particles, as well as to their chemical compatibility with the molten polymer in which they have to be dispersed. Melt mixing with lamellar fillers to obtain polymer-clay nanocomposites is complex, to the point that many researchers are still looking for a deeper understanding of the mechanisms of clay exfoliation and dispersion during melt compounding [15-18] – which would finally help finding the most appropriate parameters to optimize in order to improve dispersive mixing. I-1.3.5 Distributive and dispersive mixing Aggregation and agglomeration are the names commonly given to the processes during which two particles in suspension collide and remain jointed to each other. These phenomena are prevalently binary, since ternary or high-order collisions are rare. On the other hand, disaggregation and desagglomeration (or more generically breaking) occur as a consequence of the rupture of aggregates and agglomerates. These notions are quite generic but rather adequate for macro- and micro-composites, whose filler particles are big enough and have a sufficiently simple morphology19. In contrast, nanocomposites morphology (in particular polymer-clay nanocomposite morphology) makes everything more complicated and the words aggregation/agglomeration, dispersion/distribution are erroneously considered equivalent – their utilization is sometimes confusing. The same uncomfortable feeling of inadequateness affects also nanocomposite processing. Up to now, we described mixing and processing issues primarily as they suit to conventional polymer composites: after all, the main requirements for a high quality mixing appeared to be a good distributive mixing and an optimized residence time of the compound in screw/barrel systems. They certainly remain crucial for processing control but there are other factors, typically requiring a multiscale approach [19], that have to be considered as well: dispersive mixing and a suitable chemical compatibilization of all the ingredients of the compound are additional key parameters to control nanocomposite processing. Nowadays, the literature provides ample evidences that nanocomposites can 19 Breaking by collision with the rotating screw or the barrel surface, for example, is almost impossible when the particles have a size dp<100 m (unless it happens to aggregates or agglomerates) [4]. PhD INSA de Lyon (2008) 26 PROCESSING OF NANOCOMPOSITES Melt compounding be formed by melt processing, and several reasons make melt processing the preferred method to obtain nanocomposites for commercial purposes. However, while it might be anticipated that melt processing conditions would have an important influence on the morphology of the processed nanocomposite, until recently the literature contained no definitive conclusions about the optimum process conditions [20]. On the contrary, many articles focus on the importance of the chemistry used to modify the surface of the nanofillers without including the role of processing – as stressed by Dennis et al. [15] – or eventually taking into account processing conditions but to a lesser extent, or with less marked results [21]. This is the reason why we decided to focus on nanocomposites processing: to understand which requirements should processing have to obtain high quality nanocomposites. But before that, some morphology issues have to be clarified. I-2 MACRO-, MICRO-, NANOCOMPOSITES We‟ve already reminded that, in general, any intimate association of immiscible phases, capable of conferring to the final compound some specific properties which raw materials didn‟t singularly have, can be considered as a composite. In particular, any composite having a polymer as the prevalent phase is labeled as polymer composite. But a composite can also be defined as the dispersion, whether organized or not, of one or more secondary phases in a primary phase20. Depending on the size of the moieties of secondary phase dispersed in the matrix, composites can be classed in three categories: macro-composites; micro-composites; nano-composites. Macro-composites correspond to traditional composites, reinforced by fibers that can be long or short, continuous or discontinuous and disposed more or less regularly into the matrix. The critical issue to assure a good performance of macrocomposites is, chiefly, the optimization of the interface between the polymer matrix and the secondary phase. This category of composites have their greatest advantage in the improvement of 20 Primary and secondary only refer to the relative amount of the phase of interest in the compound. Antonella ESPOSITO 27 Chapter I the mechanical properties and, in relation with this aspect, it is worthy to remind that the stresses produced by a mechanical solicitation are efficiently transferred from the matrix to the reinforcing phase only if the phases are intimately and strongly connected 21 [22]. Contrarily to macrocomposites (which nowadays still represent the majority of the composites produced at the industrial scale), micro-composites can be thought as a transition between traditional composites and the emerging nanocomposites. In the past, the major interest in using clays for polymer enhancement was to break clay aggregates into clusters so that polymers were reinforced by microsized fillers. The results obtained with clusters, however, weren‟t much better than those obtained for the corresponding macrocomposites: indeed, it is easy to guess that the excellent properties owned by each silicate layer composing clays can‟t be properly exploited with such a morphology [23]. Additionally, the weak interaction between clay platelets22 transform clay clusters in damage initiation sites by locally reducing mechanical resistance, thus facilitating the formation and propagation of cracks [24]. A nano-composite is a composite in which the secondary phase consists of moieties having at least one of their three dimensions of the order of the nanometer – or anyway less than a hundred nanometers [25]. The idea from which the development of nanocomposites started is certainly related to some frustration due to the quite ordinary results of microcomposites but also, someway, to chance. First, as the most interesting properties of clays are largely due to the structure of the individual silicate layers, it was quite spontaneous to consider exploiting each clay platelet rather than clay clusters; in addition, as each clay cluster can contain hundreds (even thousands) platelets, another advantage of disaggregating clays consists in reducing the amount of filler to be added to the polymer matrix to get high-performance composites. As a matter of fact, if the optimum level of filler loading is 60% for macrocomposites [22] and almost as high as 60% for microcomposites [3], for nanocomposites a level of filler loading typically less than 5% wt can remarkably improve the initial properties of the neat polymer matrix 21 Of course, the properties of traditional macrocomposites depend also on other parameters such as the diameter and the length of the fibres, their length distribution, the volume fraction of fibres and the way the reinforcement is arranged into the matrix. 22 The structure of clay minerals is described in § I-3.1. PhD INSA de Lyon (2008) 28 PROCESSING OF NANOCOMPOSITES Melt compounding [24]. Indeed, it is particularly in the domain of low filler loadings that nanocomposites result much more high-performing than traditional composites. Finally, comparing to traditional composites, nanocomposites have the advantage of improving not only the mechanical properties, but also the barrier properties, electrical conductivity and fireresistance potentiality of the matrix without affecting its optical transparency [24][26]. Nevertheless, the effective development of polymer-clay nanocomposites started mostly by chance, when a group of researchers working for Toyota got the very first surprising results by in-situ polymerization of nylon-6: before that, even if some studies had been conducted about polymer-clay nanocomposites, this new class of materials didn‟t have the expected success as the first works didn‟t regularly result in dramatic improvements of the properties [24]. The first patents about the fabrication of nanocomposites from nylon-6 and clays have been registered in the latter part of the „80s [27][28] but the first commercial application of these materials appeared on the market only few years later [24] and the factor which made it possible was unsurprisingly related to a better control of filler dispersion in the matrix. Even nowadays, nanocomposites haven‟t yet find a veritable market segment – not as much as traditional composites, and in spite of their numerous advantages – maybe because several physical and chemical mechanisms involved in nanocomposites formation haven‟t yet been understood and several aspects of nanocomposite processing still have to be optimized [23]. Controlling nanocomposite processing means controlling nanoparticle size distribution, as well as dispersion and distribution into the matrix, and the fact that such a level of control is the critical condition to exploit the exceptional properties of polymer nanocomposites is nothing else than an evidence, today. We previously defined a nanocomposite as a composite containing a secondary phase composed of moieties having at least one of their three dimensions of the order of the nanometer. It is possible to further class nanocomposites according to the number of dimensions which are of the order of the nanometer (< 100nm). Nanocomposites can contain nanoparticles having [25]: one nano-dimension; two nano-dimensions; three nano-dimensions. Antonella ESPOSITO 29 Chapter I When nanoparticles have only one nano-dimension (e.g. a thickness about 1 nm) and a shape factor of at least 25, they have the appearance of thin layers (nanoplatelets). The nanoparticles having such morphology and most commonly used to form polymer nanocomposites are double-layer hydroxides (DLH) and swelling23 clays (smectites). The most important smectite is montmorillonite (MMT), an aluminosilicate constituted of an octahedral alumina sheet stacked between two tetrahedral silica sheets [23][25]. More details about lamellar fillers and MMT structure are available in § I-3.1. Nanoparticles with two nano-dimensions (the third being much bigger) can appear as empty tubular (nanotubes) or bulk fibrous (nanofibers and nanowhiskers) particles, having a diameter less than 100 nm and a shape factor of at least 100 [23]. The group of nanoparticles with three nano-dimensions includes metallic nanoparticles, fullerenes, and other isometric nanoparticles derived from oxides and based on sulphur, selenium, nitrogen, carbon, e.g. magnetite nanoparticles (Fe3O4), quantum dots (CdS, CdSe). These particles are not supposed to act as a mechanical reinforcement but rather contribute to specific properties (magnetic, optical, conductive, etc.). We focused our attention on processing of nanocomposites containing particles having one nano-dimension, i.e. lamellar fillers and, more particularly, swelling clays. The addition of clays to polymers has been proven to greatly improve their mechanical performances, fire retardancy and barrier resistance potentialities, as well as thermal and electrical conductivity, whereas optical properties are preserved. Clays are frequently used to prepare nanocomposites and, nowadays, many reviews are available about this topic [24][26][29-43]. In spite of the fact that the inorganic platelets are generally required to perfectly disaggregate into the organic matrix24, clays are often affected by dispersion and/or distribution problems [44-46]: a method well adapted to monitor the evolution of the morphology during processing would greatly help the development of polymer-clay nanocomposites. 23 Swelling clays are also known as expandable clays. For most applications it is generally believed that the greatest benefits are achieved when the platelets are fully dispersed into the matrix [15]. 24 PhD INSA de Lyon (2008) 30 PROCESSING OF NANOCOMPOSITES I-2.1 Melt compounding Nanocomposite morphology Agglomeration, aggregation, dispersion and distribution At the very beginning, fillers such as talc and calcium carbonate were added to polymers primarily to reduce the cost of the final products and, additionally, to increase their rigidity [3]. Today it is universally admitted that adding microsized fillers to thermoplastic polymers has many other advantages: reduction of thermal deformations, improvement of thermal stability and fire-resistance, tailoring of optical properties, etc. [47]. However, the addition of microsized fillers to polymers frequently reduces their impact strength, since mineral particles represent zones of stress concentration [23]. Any filler participates someway to the improvement of material properties and, in any case, the presence of fillers always alter the rupture mechanisms of the polymer matrix – according to filler shape, dimensions and size distribution, compatibility with the polymer and, most of all, dispersion and distribution into the matrix. In theory, the mechanical reinforcing efficiency of fillers should increase as the average particle size decreases [48] – moreover, reducing filler size should improve matrix rigidity without affecting its resiliency [3]. But the real revolution of nanofillers is the development of an extraordinarily extended interface with the polymer matrix: the available surface dramatically increases when filler diameter is lower than 100 nm (Figure I-F9) [23]. Figure I-F9 Available Surface Area per Unit Volume as a function of filler Particle Diameter in the case of perfectly dispersed spherical filler particles [23]. Therefore, the smallest the particles, the biggest the available surface: here is the reason of the strong interparticle interactions experienced by nanoparticles – to a point Antonella ESPOSITO 31 Chapter I that the formation of aggregates is almost unavoidable. The presence of aggregates has to be carefully avoided: when composites are subjected to mechanical solicitations, the presence of aggregates can cause premature failures and modifies the micromechanical behavior (thus the rupture mechanisms of the polymer) in an unpredictable manner (Figure I-F10) [48]. Even the presence of a few aggregates can greatly (negatively) affect the impact strength and reduce composite performances, annulling the benefic effects of a filler with reduced size [3]. In conclusion, the addition of nanofillers to thermoplastic polymers is certainly a good method to improve their properties, but if the nano-particles are perfectly dispersed and homogeneously distributed into the matrix. Figure I-F10 Comparison of a polymer composite having a heterogeneous (on the left) and a homogeneous (on the right) particle distribution. Influence of the particle distribution on the micromechanical behavior and the fracture mechanisms [48]. The available filler surface directly determines the composite interface which, in turn, controls the quality of the interactions between the filler particles and the polymer matrix and consequently influences the properties of the final products. Somehow, the biggest challenge of nanocomposites could seem similar to the traditional challenges of macrocomposites: being able to perfectly control the quantity and quality of the fillermatrix interface. In reality, nanofiller reduced dimensions make it more complicated. In traditional macrocomposites, the interface is defined as the region starting in the point of the fiber where the properties of the composite are different from fiber bulk properties, and ending in the point of the matrix where the properties of the composite PhD INSA de Lyon (2008) 32 PROCESSING OF NANOCOMPOSITES Melt compounding are exactly the same as those of the neat polymer25. If we consider the interparticle spacing as a function of the particle size and in the hypothesis that the filler particles are perfectly dispersed into the matrix, we realize that at low filler loadings the interface coincides exactly with the volume of the matrix (Figure I-F11) [23]. The interparticle spacing for fillers having an average diameter of 15 nm and a volume concentration of 10%, for instance, is about 10 nm: even if the interface measured only few nanometers, when moving from the filler surface one would observe that the whole matrix has a behavior which is different from the one had in the absence of the filler [23]. Figure I-F11 Interparticle Distance as a function of the Volume Fraction of Nanoparticles for different particle sizes, in the case of perfectly dispersed spherical particles [23]. According to the aforementioned definition, nanocomposites have an interface at least one order of magnitude bigger than traditional macrocomposites: but which is the exact definition of interface in the case of nanocomposites? We could probably affirm that interface in nanocomposites is represented by any portion of the matrix in which the macromolecular chains are immobilized because of the presence of the nanofiller – meaning that one should be able to easily tell the interface from the unmodified neat polymer matrix. However, it frequently happens that the properties of a nanocomposite change continuously from the nanofiller surface towards the neat matrix. Moreover, the relationship between nanocomposite properties and average particle size isn‟t absolutely linear. As for macrocomposites, the interface doesn‟t exclusively depend on the particle 25 Such properties could be the chemical composition, the mobility of the macromolecular chains, the polymerization degree, the polymer crystalline fraction, etc. [23]. Antonella ESPOSITO 33 Chapter I diameter, but also on the quality of the interactions between the filler particles and the polymer matrix – the strongest such interactions, the thickest the portion of polymer whose macromolecules have a reduced mobility: the presence of the nanofiller induces a continuous (rather than discrete) variation of the nanocomposite properties [23]. The relationship between particle size, filler surface and resulting interface – everything gets more complicated if we consider that nanoparticles may have a different shape factor according to the number of nano-dimensions26. The geometrical appearance of a particle modifies the relationship existing between surface and volume (Figure I-F12) [40]. Figure I-F12 Surface area/volume relations for various particle geometries [40]. Nanofillers can have different size, shape factor, chemical composition and, thus, nanocomposites morphology may be very different according to nanofiller properties. But nanocomposite morphology depends also on how the secondary phase is arranged into the primary phase. Filler spatial arrangement can be better defined in terms of [23]: aggregation (or agglomeration), which essentially depends on the physical and/or chemical interactions between filler particles; dispersion, which denotes whether the filler particles are physically separated from each other (i.e. isolated) or not; distribution, which quantifies nanocomposite global homogeneity, because it measures the ability of the filler particles (whether isolated or not) to homogeneously occupy the entire volume of the polymer matrix. 26 We have previously explained how it is possible to class nanocomposites according to their number of nano-dimensions (§ I-2). PhD INSA de Lyon (2008) 34 PROCESSING OF NANOCOMPOSITES Melt compounding Nanocomposites can have good distribution but poor dispersion (Figure I-F13 (a)), poor distribution and dispersion (Figure I-F13 (b)), poor distribution but good dispersion (Figure I-F13 (c)) or good distribution and dispersion (Figure I-F13 (d)) [23]. (a) (c) (b) (d) Figure I-F13 Schematic illustration of the possible aggregation and/or distribution outcomes in composites: (a) good distribution but poor dispersion, (b) poor distribution and poor dispersion, (c) poor distribution but good dispersion, (d) good distribution and good dispersion [23]. It is obvious that, to obtain nanocomposites with good filler dispersion, the first obstacle to surmount is represented by the spontaneous tendency that nanofillers have to aggregate: a choice between avoiding aggregation, assuring disaggregation during nanocomposite processing or – why not? – both of them. The difficulty of this task is due to the fact that, no matter their morphology, all nanoparticles form aggregates stabilized by forces (ionic and Van der Waals interactions, hydrogen bonding, etc.) which sometimes are stronger than the interactions between filler surface and polymer matrix [25]. The interactions between clay platelets, for instance, are stabilized by the anionic attraction of mobile cations localized in the interlayer spacing27: these cations, highly hydrated, render clay galleries highly hydrophilic and hinder clay intercalation by most of the organic macromolecules, prevalently hydrophobic. Thus, sometimes the hydrated 27 Some more details are available in § I-3.1. Antonella ESPOSITO 35 Chapter I (hydrophilic) inorganic cations have to be exchanged for organic (hydrophobic) cations. As organic cationic molecules are bigger than inorganic ones, the exchange induces clay swelling (increase of clay interlayer spacing). More information about MMT structure, cation exchange process and organically-modified clays (organoclays) are available in § I-3. Organoclays can be then compounded with polymers, and several methods exist to produce polymer-clay nanocomposites [26][38]: compounding, independently from the chosen procedure, is supposed to produce intercalation or complete exfoliation (also called delamination) of clay platelets by polymer chains, according to the mechanisms previously proposed [15-17]. But polymer-clay nanocomposites can develop such a complex morphology that it appears impossible to establish a simplified, unambiguous description – as demonstrated by the variety of morphological descriptions present in literature [49]. Biswas and Sinha Ray [29] individuated three categories of polymer-clay nanocomposites: conventional composites (in which clay tactoids exist in their original aggregated state, with no intercalation of the polymer into clay lamellae); intercalated composites (in which the insertion of polymer into the clay structure occurs in a crystallographic regular fashion, regardless of the clay to polymer ratio, and normally only a few molecular layers of polymers are intercalated into clay galleries); exfoliated composites (in which the individual clay layers are separated in a continuous polymer matrix by average distances which depend on loading, typically with a lower content of clays than in intercalated composites). Dennis et al. [15], in contrast, differentiated four possible morphological outcomes: tactoid (clay layers remain unseparated), intercalated (a small amount of polymer matrix diffuses into clay galleries but causes less than 2030 Å of separation between clay layers), intercalated disordered and delaminated (or exfoliated) (polymer further separates clay layers, e.g. 80-100 Å or more) Figure I-F14, on the left. Still different, Schadler [23] distinguished conventional composites with tactoids from intercalated, ordered and disordered exfoliated nanocomposites (Figure IF14, on the right). Eckel et al. [50] stated that polymer-clay nanocomposite morphology can be “simply” classified as exfoliated (or delaminated) (clay platelets are dispersed as discrete sheets with the single platelets having “no observable” association with other silicate platelets), tactoid28 (clay platelets are stacked face to face in clay particulates) 28 Tactoids tend to be less than 100 silicate sheets thick and can range down to a few layers thick [50]. PhD INSA de Lyon (2008) 36 PROCESSING OF NANOCOMPOSITES Melt compounding and agglomerated tactoid29 (bigger clay particulates are made of multiple tactoids). Homminga et al. [51] gave one more version. Finally Liu et al. [44] recently reviewed the apparently delicate issue of the morphology of polymer nanocomposites reinforced by clays: these examples are just meant to point out how complex could be to describe polymer-clay nanocomposite morphology. Well, the techniques to characterize it aren‟t expected to be any more anodyne. Figure I-F14 Schematic illustrations of the terminology used by Dennis et al. [15] (on the left) and Schadler [23] (on the right) to describe polymer-clay composite morphology. I-2.2 Techniques for morphological analysis The difficulties encountered in describing the morphology of polymer-clay nanocomposite affect also the techniques for their morphological analysis. Analogously to several other properties, morphology can be investigated by techniques ex situ (off-line) or in situ (on-line). The latter can be performed just by complex equipments (generally available only in research laboratories) in particular conditions and, sometimes, with the help of model materials. Processing engineers have been looking for new techniques to perform on-line process monitoring since a long time – certainly long time before the advent of polymer nanocomposites. However materials evolve, their applications don‟t 29 The agglomerated tactoids can be several microns large and are comprised of loosely bound tactoids. The agglomerated tactoids can be distinguished from non-agglomerated tactoids by their relatively close proximity to one another relative to the overall dispersion of the clay [50]. Antonella ESPOSITO 37 Chapter I stop to differentiate and progress unavoidably demands the development of more and more innovating techniques of characterization. The methods to assess dispersion [52] and distribution of fillers have to follow, as well. I-2.2.1 Morphological characterizations ex situ The most commonly used techniques for morphological characterization ex situ of polymer-clay nanocomposites are X-Ray Diffractometry (XRD) and Transmission Electron Microscopy (TEM), although some researchers tried also to employ Scanning Electron Microscopy (SEM) and Differential Scanning Calorimetry (DSC). Biswas and Sinha Ray [29] lately proposed a review of some relevant results obtained on polymerclay nanocomposites by XRD, TEM, SEM and DSC. In the next paragraph we rapidly review the main principles of TEM and XRD and we point out their limitations. Transmission Electron Microscopy (TEM). TEM is the sole technique able to visualize morphology and qualitatively estimate nanofiller dispersion by providing an image of the sample in which one can appreciate some specific structural features of the nanocomposite such as the shape, size and local repartition of nanofiller particles, both on the surface and in the volume of the sample (observations can be made on a cross section of the sample, usually obtained by microtome slicing). Contrarily to the photons used for light microscopy, electrons have short wavelengths: typical energy values used to perform electron microscopy are 100-300 keV (corresponding to theoretical optimum resolutions of 0.2 nm), which is comparable to the interatomic distances of the matter. Such a high spatial resolution makes TEM a technique well adapted to the investigation of nanocomposite morphology. Nevertheless, TEM analysis is only possible on samples which are transparent to electron beams, viz. very thin films (thickness < 120 nm) [25]: therefore, it is necessary to cautiously prepare the samples, following a procedure to be chosen according to the nature of the sample itself. Independently on the method used for sample preparation, one should assure that the chosen cross section is representative of the entire nanocomposite. TEM images show dark features representing the fillers on a clear background representing the polymer matrix – since the atoms of most of the PhD INSA de Lyon (2008) 38 PROCESSING OF NANOCOMPOSITES Melt compounding fillers are heavier and stop the transmission of the electrons more than the hydrocarbon macromolecular matrix [25]. As previously stressed, polymer-clay nanocomposites can have several different morphologies typically quite difficult to describe. To simplify the task, let‟s say that during clay intercalation (i.e. while polymer chains diffuse into clay galleries) two morphological outcomes are possible: a structure made of regular stacks of clay platelets alternated to polymer layers (intercalated) or a quite homogeneous but rather irregular morphology (exfoliated or delaminated). Please note that these represent only two morphological models: in reality, polymer-clay nanocomposite morphology is partially intercalated and partially exfoliated, both regular and disordered, and may even contain tactoids in specific portions of the sample. TEM is the best characterization tool to identify all the possible morphological outcomes with all their variety, but can only visualize a reduced portion of the sample: how to be sure that the sample has the same morphology in any of its portions? In other words, how to be sure that the observed sample is really representative of the whole final product? Figure I-F15 Low-magnification (on the left) and high-magnification (on the right) TEM image of the same PS-clay nanocomposite (5% wt) [49]. High-magnification shows an intercalated/ exfoliated morphology, whereas low-magnification reveals small (a) and large (b) tactoids. An example of TEM image is shown in Figure I-F15 [49] but the literature offers plenty of papers in which impressive TEM images of different nanocomposite systems are reported [15-17][45][50][53][54], often substantiated by results obtained by other techniques (XRD, rheology measurements, etc.). Some authors made considerable Antonella ESPOSITO 39 Chapter I efforts to find some quantitative methods to exploit TEM images. Dennis et al. [15], for instance, judged the dispersion level of polymer-clay nanocomposites by counting the number of platelets or intercalates seen in twelve 6.25 cm2 cutouts from a sheet of paper laid over a photomicrograph printed at 130,500 magnification, and averaging them to get a number representing delamination and dispersion for each image: they considered that the higher the number, the better the delamination and dispersion of the sample are. Eckel and coworkers [50] observed that, except for few works, TEM is used only quantitatively to characterize dispersion in polymer-clay nanocomposites and, therefore, tried to assess a quantitative use of TEM images. First of all, they proved a significant dependency of TEM results on sample preparation and, in any case, on the dimensions and orientation of clay platelets: they showed that clay features can be observed easily (and independently from their orientation) only if sufficiently thick (i.e. thicker than the microtomed section); as their thickness approaches the width of a single silicate layer, they are observable only edge on (i.e. only if they are perpendicular to the sample slice). They also recognized that a qualitative description of clay structures is insufficient to adequately describe clay dispersion, and may be misleading for several reasons: a small area may not be representative of the overall structure; the morphology of polymer-clay composites can be comprised of multiple clay structures; clay structures typically have a range of different sizes. For these reasons, Eckel et al. decided to quantify dispersion and considered using several methods: measuring particle size (but they found it tedious and requiring too many samples, thus they excluded it), measuring particle density (they proceeded as Dennis et al. [15]) and measuring the linear intercept distance between particles. A higher particle density at a fixed (or normalized) content of clay indicated a better dispersion. About linear intercept measurements: they placed an array of parallel lines over the micrograph and then divided the total length of the lines by the number of times the lines intersect a clay structure30 – a smaller linear intercept distance indicated a better dispersion. In conclusion, the linear intercept method has several advantages: it is quicker, less tedious and more objective; 30 In samples with highly oriented clay particles, the authors placed the array of lines perpendicularly to the dominant clay axis. PhD INSA de Lyon (2008) 40 PROCESSING OF NANOCOMPOSITES Melt compounding in the case of exfoliated nanocomposites, it does not depend on the length of the silicate platelets; for an aligned exfoliated nanocomposite, it provides a theoretical mean linear intercept distance by dividing the initial clay interlayer spacing by clay volume fraction. It‟s noteworthy that the values obtained with this method tend to be slightly smaller than the theoretical ones (i.e. dispersion is underestimated) because of a stereological error31. A disadvantage of the linear intercept method with respect to the particle density method is the dependence on orientation: clay particles showing a random orientation in the sample plane will have a larger linear intercept distance than clay particles having a preferred in-plane orientation. Vermogen et al. [45] recently observed that more and more researchers are using TEM image analysis to characterize the dispersion of polymer-clay nanocomposites and proposed a novel method to exploit TEM micrographs by image processing and an appropriate statistical analysis. According to some corrections they had to take into account, they could compute thickness average t , length average L , aspect ratio average AR , interparticle distance II and (parallel and perpendicular to the length, respectively) and average particle density per m2 for each classed tactoid. Indeed, the more meticulous TEM analysis gets, the more evident the morphological complexity of polymer-clay nanocomposite becomes: to develop such an exhaustive and methodic procedure, Vermogen and coll. couldn‟t avoid assuming some hypothesis and had to define six classes of tactoids to be able to compare different samples. X-Ray Diffractometry (XRD). If qualitative TEM can visualize all the possible morphological outcomes in polymer-clay nanocomposites (parenthetically, we reported proofs that quantitative TEM is developing but is not yet sufficiently reliable as a standalone tool), another technique appears handier, quicker, cheaper than TEM. In general, XRD is used firstly to establish whether the nanocomposite has or not an intercalated 31 Stereological errors in TEM arise from the image projection of microtomed sections and, thus, increase with thicker slices and smaller particles. This error affects both the linear intercept and the particle density measurements. Antonella ESPOSITO 41 Chapter I morphology and then, if morphology is intercalated, an approximate quantification of clay interlayer spacing32 can be done on the basis of the X-ray diffraction patterns. Figure I-F16 WAXD patterns (a) and corresponding TEM images (b) of three different types of polymer-clay nanocomposite [38]. In a well-ordered layered structure, the basal reflections conform to Bragg law and are expected to generate harmonic series of diffraction peaks. Natural clay minerals such as montmorillonite (MMT) are characterized by an initial interlayer spacing of about 1.2 nm, but modifications by cation exchange with an organic cation bearing a long alkyl chain swell clay galleries and increase the interlayer spacing to a minimum of 1.5 nm up to a maximum of 3.0 nm, according to the length of the alkyl chain and to its configuration33 [25]. Any change in the X-ray diffraction pattern of the nanocomposite sample (peaks appearing, disappearing, shifting, broadening or narrowing, heightening or lowering) may provide valuable information about the interactions between the clay 32 33 This notion will be clarified in § I-3.1, where the structure of lamellar fillers has been rapidly described. XRD measurements are currently performed on crystalline powders to characterize their structure. PhD INSA de Lyon (2008) 42 PROCESSING OF NANOCOMPOSITES Melt compounding and the polymer matrix. Morphological assessments of polymer-clay nanocomposites are thus performed by observing position, shape and intensity of any peak produced by X-ray diffraction from clay platelets in the range 2 = 1-10°: in practice, the presence of one or more peaks indicates that the nanocomposite has an intercalated morphology (with the clay platelets stacked in a crystallographic order and separated by an interlayer spacing which depends on the position of the peak(s)), and when the peaks broaden or disappear, clays are completely exfoliated and isolated platelets are dispersed into the polymer matrix at the molecular level [23][29][38]. An example of XRD patterns (correlated to TEM images) are shown in Figure IF16 [38] but, once again, the literature offers lots of papers in which authors use XRD to characterize polymer-clay nanocomposite morphology [15-17][29][53-58]. It is worthy to remind that intercalated and exfoliated morphologies are only some ideal configurations to make comparisons easier, since the real morphology of polymer-clay nanocomposites is typically a mixture of tactoids having different sizes (the biggest visualizable only by light microscopy [45][54], the smallest also by TEM), intercalated (quantified by XRD), exfoliated (observable only by TEM) clay structures. None of these techniques is stand-alone, since none of them can unmistakably describe the morphological outcomes of nanocomposite. Researchers are gradually getting aware of the limits of the available techniques and do not hesitate to point out the difficulties encountered and the uncertainties not yet explained. Indeed, several ambiguities in XRD data risk to complicate nanocomposite characterizations. Eckel et al. [50], for instance, properly reminded that the basal reflections of lamellar fillers and of the corresponding nanocomposites do not always form harmonic series of diffraction peaks: it‟s quite the opposite regularly happening. This observation is easily explained by mixed-layering: the interlayer spacing between clay platelets is a mixture of two or more types 34. Mixedlayering generates significant uncertainty in the estimation of the interlayer spacing, the classification of the structural ordering and the exact evaluation of morphology. By the way, even in the absence of mixed-layering, the interactions between the lamellar fillers and the polymer are difficult to discern from X-ray diffraction patterns and, in addition, 34 Clay minerals are considered mixed-layered if their peak position (compared to their nominal Bragg position) exceeds 0.75% [50]. Such phenomenon is also known as the Hendricks-Teller effect [142]. Antonella ESPOSITO 43 Chapter I instrumental parameters35, particle size, defect density and strain effects can affect (to a minor extent, surely) the observed peaks and, in particular, their width. Because of the plate-like morphology of lamellar fillers36, preferred orientation can affect XRD results (more intense peaks indicate a higher clay content or a significant preferred orientation), as previously reported [50]. In conclusion, the absence or presence of basal reflections could actually indicate whether exfoliation has occurred or not, but other factors such as clay dilution, preferred orientation and artifactual peak broadening should be taken into account: a dilution of the clay content can otherwise result in the false conclusion that exfoliation has occurred and, conversely, a preferred orientation can result in the false conclusion that exfoliation has not occurred. XRD is perhaps handy (though not uncontroversial) to quantify dispersion, but surely cannot help evaluating distribution. Other techniques. We presented TEM and XRD from a critical point of view: these techniques are certainly limited and cannot be considered sufficient, if performed separately, to characterize polymer-clay nanocomposites. Nevertheless, nowadays they represent the most commonly used techniques and will surely keep being essential for any morphological characterization ex situ. Some researchers keep comparing XRD and TEM either to find out whether the respective results are comparable, or to understand at once which technique is the most effective [49][50][56]. In opposition, many other researchers have started looking for complementary techniques in order to provide some alternative tool for polymer-clay nanocomposite morphological characterization. Clay multiscale structure and the morphology of the resulting polymer-clay nanocomposites rather require a complete panel of analytical techniques covering several length scales. Rheology is increasingly used to evaluate nanocomposite morphology. The most simplistic approach consists in considering that an increase of complex viscosity at low shear rates (flow measurements) or low frequencies (dynamic measurements) indicates a good dispersion of the clay platelets into the polymer matrix. Indeed, clay dispersion in a melt is due to both macroscopic and local shear and results from the stress transfer of the melt medium to the silicate layers. During melt processing, the macroscopic shear 35 36 Instrumental broadening is expected to contribute no more than 0.2 2. Lamellar fillers are described in § I-3.1. PhD INSA de Lyon (2008) 44 PROCESSING OF NANOCOMPOSITES Melt compounding imposed by the mixing tool makes clay agglomerates peel, then the local shear due to the viscosity of the matrix lets polymer chains intercalate in clay galleries [15][45]. The increase of viscosity at low shear rates or frequencies would be related to the hindrance exerted on polymer macromolecules by the presence of the clay platelets, dispersed at the molecular level [45]. In reality, rheology provides more information if the slope of the viscoelastic moduli is also considered (network formation, percolation), and not only at low frequencies and shear rates. Wagener and Reisinger [59] developed a semi-quantitative37 method which, on the basis of the shear thinning exponent, lets compare the extent of delamination of clay platelets stacks. This method is one of the first rheological methods for morphological analysis of thermoplastic polymer nanocomposites containing lamellar fillers. Before Wagener and Reisinger, the literature contained several papers describing melt rheology as a potential method to analyze polymer-clay nanocomposites [60], but no publications disclosing some practical approach to quantify shear thinning with the perspective of comparing clay exfoliation or nanoscale distribution – even though pronounced shear thinning had been already found to be a characteristic feature of truly nano-dispersed polymer-clay composites [61]. Under specific conditions, rheological pseudo solid-like behaviors indicate edge-to-face interactions of the clay platelets with each other or with clay tactoids, which would contribute to the construction and mechanical stabilization of mesoscale card-house structures of silicate layers; at higher shear rates or under the prolonged action of slow shear forces, such structure would get broken in consequence of an increasing alignment of the platelets – which is responsible for the shear thinning effect. Of course any filler – especially for high loadings and no matter its morphology, (i.e. whether lamellar or not) – modifies the rheological behavior of polymer melts [62]: thus, rheology appears as a universal tool for the analysis of filler dispersion in molten polymers even for composites different from polymer-clay composites [63]. Lately, Vermant et al. [64] used again rheology to compare polymer-clay nanocomposites processed by melt mixing: they employed steady-state as well as transient 37 The term semi-quantitative means that there is no unequivocal relation between the shear thinning exponent and the degree of clay exfoliation. It also means that the average number of clay platelets per tactoid for a given nanocomposite cannot be calculated by the shear thinning exponent – better used for direct comparisons of the quality of exfoliation. Antonella ESPOSITO 45 Chapter I nonlinear measurements to better separate the contributions of flow-induced orientation of clay tactoids, and of clay particles networking. They determined the conditions under which the rheological properties are dominated by the presence of clays, then analyzed the low frequency linear viscoelastic behavior by the scaling concepts of fractal theories (in order to get the degree of network formation by clay exfoliation) and assessed the quality of clay dispersion on the basis of the high frequency behavior of the viscoelastic moduli. Contrarily to TEM, the macroscopic samples used for rheology measurements offer an integral “vision” of the composite morphology with increased reliability – this is an obvious advantage in comparison with methods using smaller samples, more prone to microscale heterogeneity. However, melt rheology data are strongly influenced by the flow and shear history of the sample: it is necessary to carefully distinguish thixotropy (which could provide interesting information) from any other effect due to the enhanced dispersion got by the rheometer – this is the reason why establishing a good protocol for rheological measurements could be delicate and surely represents the most difficult part of the whole characterization procedure. In addition, prolonged exposure to heat (during long lasting measurements) may start the thermal degradation of particularly sensitive samples. By the way, Vermant et al. showed that, if a suitable experimental protocol is established, scaling laws for fractal networks can be used to assess clay dispersion in the melt at low clay contents. The percolation threshold can be determined from the clay concentration effects on the elastic moduli, i.e. from a deviation from viscoelastic-liquid to elastic-solid behavior. This percolation value also provides an average particle aspect ratio for the entire nano-composite: above the percolation threshold, the elastic moduli and the onset strain to disrupt the network can be scaled with clay volume fraction to yield a fractal dimension characterizing the network. Changes in the high frequency moduli were also considered as an indication of dispersion. Since then, rheology has been used more and more frequently to characterize the structure and melt behavior of polymer-clay nanocomposites [17][65][66]. TEM isn‟t the only microscopy used for morphological analysis of polymer-clay nanocomposites. A first proof has been given when discussing TEM: clay minerals and their nanocomposites have multiscale structures which require a set of characterization techniques sweeping several length scales. For such reason, even light microscopy could PhD INSA de Lyon (2008) 46 PROCESSING OF NANOCOMPOSITES Melt compounding contribute to morphological analysis – some features, even if definitely undesirable, can be visualized only on a microscopic scale and mustn‟t be neglected [45][51][54]. If TEM is known to provide useful information on nanocomposite morphology, Scanning Electron Microscopy (SEM) has never been considered particularly effective, as reported by Biswas and Sinha Ray [29]. However, some researchers use SEM [51] and try to develop a method to extract quantitative information from the images (as for TEM) to assess filler dispersion – although filler is not necessarily lamellar [67]. Atomic Force Microscopy (AFM) has been recently used as well [68], but the usefulness of this technique is, analogously to SEM, still questionable [50]. Intercalated morphologies in which the distance between regularly stacked clay platelets is more than 70 Å can be observed but are quite rare. In this situation, XRD performed in the typical conditions (2 = 1-10°) becomes useless, since it characterizes only intercalated morphologies with an interlayer spacing comprised between 10 and 40 Å [38]. If Wide-Angle X-ray Scattering (WAXS) is insufficient, or when the polymer separates clay platelets and their spatial distribution isn‟t anymore characterized by a regular, periodic arrangement, researchers use Small-Angle X-ray Scattering (SAXS). This technique exploits the difference of electron density of clay platelets with respect to the polymer matrix – which lets detect the presence of the clay from the lowest-angle fraction of the diffraction pattern. Cser and Bhattacharya [55] published an interesting work about the methods based on X-ray radiation and able to detect clay platelets in polymer-clay nanocomposites. They suggested to be careful about the interpretation of XRD data and showed that the small-angle portion of the XRD scattering curves can be deconvoluted according to particle and reciprocal lattice scattering in order to estimate the ratio of exfoliated to intercalated clay structures. Sinha Rya and Okamoto [38] lately reviewed some results obtained by coupling Wide-Angle X-ray Scattering (WAXS) and Small-Angle X-ray Scattering (SAXS). These techniques are currently used coupled for morphological characterizations [69]. Apparently, Differential Scanning Calorimetry (DSC) has also been successfully used to obtain evidences of polymer intercalation: Biswas and Sinha Ray [29] reported that Giannelis et al. observed a different thermal behavior of intercalated PS-organoclay Antonella ESPOSITO 47 Chapter I in comparison with the bulk PS polymer. The fact is that both the neat polymer and the physical mixture PS-clay clearly exhibited the characteristic glass transition temperature at 96°C, whereas it seems that the intercalated hybrid did not show any transition in the temperature range 50-150°C. They also reported that soon after Krishnamoorti et al. demonstrated that the local and global dynamic behavior of confined polymer chains is markedly different from the bulk and can be easily visualized by DCS, as shown in the case of intercalated PS-clay or PEO-clay (absence of any thermal indication of glass or melting transitions). DSC is rarely used to investigate nanocomposite morphology: this technique is probably more interesting to analyze the behavior of systems in which the presence of lamellar fillers not only induces a particular morphological configuration, but has also a catalytic effect on the formation of the composite. Finally, some researchers used Solid-State Nuclear Magnetic Resonance (NMR) to characterize polymer-clay nanocomposites. VanderHart et al. [70-72] demonstrated that solid-state NMR can be used to quantitatively characterize the dispersion of various organoclays in PS. The direct influence of the paramagnetic Fe3+ (embedded in the aluminosilicate layers of montmorillonite)38 on polymer protons within about 1 nm from the surface of clay platelets creates relaxation sources which, via spin diffusion, significantly shorten the overall proton longitudinal relaxation time. They used such relaxation time as an indicator of clay dispersion in PS and showed that this approach correlated reasonably well with XRD and TEM observations. Indeed, they presented a double choice: a less complete NMR assay of dispersion that is significantly faster than TEM, or a slower but more complete NMR analysis – with sampling times comparable to TEM, information rivaling that of TEM, and a substantial advantage: the fact of performing bulk characterizations of the polymer-clay nanocomposite of interest. I-2.2.2 Morphological characterizations in situ It‟s interesting to observe that, since the beginning, fluid dynamics appears to be a relevant parameter to understand the relationships between mixing and processing: as 38 The structure of montmorillonite is rapidly described in § I-3.1. PhD INSA de Lyon (2008) 48 PROCESSING OF NANOCOMPOSITES Melt compounding soon as technology progressed enough and the available technical resources permitted it, numerous sensors have been conceived and designed to evaluate the performances of mixing devices. In most cases, these new detection systems were preliminary assembled and tested on a laboratory scale: it‟s what happened with the hot-wire anemometry, the Laser Doppler Velocimetry (LVD) and the Particle Imaging Velocimetry (PIV), just to cite some examples. Parenthetically, it is in the field of stirring, mixing and processing that the first codes for computer-aided simulation of processing have been generated. Besides, while progressing in the comprehension of mixing and processing and in the evaluation of the performances of standard devices for quite standard applications, new emerging fields impose further challenging requirements: that‟s what is truly happening with polymer-clay nanocomposites. The pans (or the barrels) are usually cylindrical and must necessarily be made of a material compatible with the processed products: steel is often a good solution, even if plastic39 and glass40 recipients have also been used [4]. In most cases, pans and barrels are not transparent – making any direct visualization impossible – and processing is thus reduced to a mysterious “black box”: one can completely know the pristine state of the processed materials, as well as the properties of the final products, but there‟s no way to know which is the correlation between all these factors (initial properties, process, final properties) and how to control them. The study of the mechanisms involved in polymer melting has been made primarily thanks to experimental investigations: Maddock [73] firstly conceived and tested a technique which, subsequently, let study the flow in the plastication section of extrusion and injection screw. The technique required waiting for the processing system to reach a steady state condition, and then stopping the rotational movement of the screw and rapidly quenching the evolving material by cooling down the screw/barrel system with cold water recirculating around the barrel. The screw and the solidified material were then extracted from the barrel and the block of polymer41 was unrolled and cut – normally one cut per revolution (or half revolution) of the screw. Nowadays, this method is still used (as demonstrated by Moguedet [74]) but not exactly 39 Plastics are light, even if poorly resistant to the internal pressure and to mechanical solicitations. Glass is inert, compatible with almost everything, but brittle. 41 This technique already uses optical methods to better identify the different portions of the sample – typically by covering the pellets with some coloured contrast agent [9]. 40 Antonella ESPOSITO 49 Chapter I practical. Moreover, a proper and complete flow characterization firstly requires the measurement of the local velocity vectors of the fluid. Several techniques can be used to determine the velocity field of a fluid, but most of them require complex equipments and a certain level of know-how which are a strict prerogative of research laboratories. Here we report some information about the existing techniques [4]. Winch. It‟s the easiest and most accessible tool to determine the velocity field of a fluid. This technique is obviously intrusive – the winch must be immersed into the flow – and implies the risk of perturbing the flow stream. Therefore, winch is reserved to the flow characterization in big containers, i.e. when the size of the winch (thus, the induced perturbation) is negligible in comparison to the size of the stirring elements. It essentially consists of a small wheel: once introduced in the flow, the wheel turns and its rotational speed is recorded. By repeating this operation in several points of the flow, the correlation with the local values of velocity is made by means of a calibration curve. Laser Doppler Velocimetry (LDV). Contrarily to the previous technique, LDV is a non-intrusive optical system, whose principle is based on the measurement of the velocity of some particles (naturally present or appositely introduced into the fluid) by means of a laser source. Of course, the particles are meant to be small enough to follow the flow streams without modifying them (negligible drag effects). Velocity is measured as follows: when two coherent beams of the same monochromatic wavelength (laser beams) cross, their intersection produces a network of interference fringes, i.e. a series of alternated bright and dark bands. This network, typically shaped as an ellipsoid of revolution, is the volume of interest for measurement (~ mm3). When tracing particles pass through the volume of interest, they diffuse the laser light in the bright bands but are not detectable in the dark bands. The frequency at which particles diffuse light is a function of the component of their velocity in the plan of the laser beam, as well as of the distance between the interference fringes. If 0 is the laser wavelength and is the angle formed by the laser beams at their convergence, the instant velocity u of a tracing particle traversing the volume of interest can be calculated directly from the Doppler frequency fd of the recorded signal: PhD INSA de Lyon (2008) 50 PROCESSING OF NANOCOMPOSITES Melt compounding fd u y 2 u sin( ) 2 (I-E4) 0 motor optical unit on mobile support laser beams splitter laser source laser beams photomultiplier tube oscilloscope cylindrical pan introduced in a rectangular pan (transparent) covariance analyzer computer Figure I-F17 Laser Doppler Velocimetry (LDV) [4]. As the interference fringes are stationary, the Doppler frequency doesn‟t depend on the direction of the flow. To take into account directional effects, a Bragg cell is added to one of the laser sources so that a constant frequency shift is applied to the interfering beams: as a result, the interfering plans aren‟t anymore immobile but rather pass through the volume of interest with a constant speed. The measured value now corresponds to the sum of the particle velocity and the passage speed. The light diffused by the particles is collected by a photomultiplier that converts luminosity in an electrical signal, which in turn will be processed to get the velocity of the fluid. An oscilloscope is also integrated to the acquisition system in order to check the quality of the acquired signals. The LDV technique provides, in each position, a time-averaged value of a given velocity component and its fluctuating value – Reynolds decomposition of velocity is particularly adapted to detect turbulence or any other instability of the flow. To get the other components of the velocity vector, it is necessary to rotate the laser beams to align Antonella ESPOSITO 51 Chapter I the volume of interest to the desired direction. The main limitation of LDV is the need for a perfectly transparent system (fluid and equipment). Such constraint obliges to use water (or another transparent model fluid) and suitable transparent equipments, whose walls can be traversed by a laser beam without diminishing its power or compromise its optical coherence. Besides, some particular shapes of the transparent pan in which the model fluid evolves cause serious optical problems: cylindrical barrels should rather be externally shaped as parallelepipeds to avoid any divergence of the laser beams due to the curved surface of the cylinder. LDV technique is schematized in Figure I-F17. Particle Imaging Velocimetry (PIV). PIV is a non-intrusive technique which lets acquire, almost instantaneously, the velocity field of a flow in a given geometrical plan. This technique can be defined by comparison with the LDV technique. If the LDV technique provides the average value of a given component of the velocity vector in a given position of measurement, the PIV technique rather provides the instant value of two components of the velocity vector (the two components corresponding to the plan of measurement). However, like LDV, PIV technique also requires a perfectly transparent system (fluid and equipment). The principle of this technique consists in recording the position of small particles formerly introduced into the fluid: by performing two records shifted of short time lapses it is possible to follow the discrete movement of the particles and deduce their speed. Recording is made on optical plans of measurement generated by a laser sheet: a camera is placed perpendicular to the laser sheet and records images in which the particles are visible as bright points. The plans of measurement have to be discretized in multiple small areas: for each area, a vector of velocity is recorded and averaged. This discretization step is important for it partially influences accuracy: if the areas are too large, most gradients of velocity won‟t probably be detected; if they‟re too small, there won‟t be statistically enough tracing particles to perform image processing. The time lapse between two successive images is an important parameter, as well: it is a function of the order of magnitude of the velocity vectors expected to be measured, as well as of the size of the areas in which the plans of measurement are discretized. Once the images acquired, processing is based on the use of a function of crossed correlation to calculate the most probable trajectory of the visualized particles by two-dimensional PhD INSA de Lyon (2008) 52 PROCESSING OF NANOCOMPOSITES Melt compounding Rapid Fourier Transform. This way, it is possible to obtain an instant velocity field on a given plan of the flow; the average velocity field can be obtained by averaging several (sometimes thousands) instant velocity fields. Figure I-F18 illustrates PIV technique. fluid containing a tracer pulsed laser laser sheet function of crossed correlation CCD camera image n image n+1 discretized area velocity vector Figure I-F18 Particle Imaging Velocimetry (PIV) [4]. Ultrasound Doppler Velocimetry (UDV). UDV differs from LDV and PIV for the fact that it doesn’t require a transparent system (neither the fluid nor the equipment) but rather a system in which the sound propagation is good. Most of the liquids and any equipment having metal walls are adapted to such a technique, which has the advantage of being adaptable to real fluids and industrial (non specific) equipments. It essentially consists in using ultrasound probes to emit intermittent sequences of pulsed waves and collect, by the same transducer (i.e. the ultrasound probe switched in mode reception), the echoes produced by small particles introduced into the fluid at a frequency which gets shifted because of the Doppler effect. The frequency shift provides a measure of particle velocity, supposed to be similar to fluid velocity. If the value of frequency shift is completed by the estimation of the total time needed by the emitted waves to go and the echoes to return, it is possible to estimate the distance of the particle with respect to the ultrasound probe. The limitation of UDV technique is the way velocity is measured: the value obtained corresponds to the projection of the particle velocity vector on the Antonella ESPOSITO 53 Chapter I axis determined by the ultrasound probe. Since flows are typically three-dimensional, it could be difficult to extrapolate the real velocity vector from its projection: hence, UDV is better suited to analyze a mono-dimensional rather than a complex three-dimensional flow. UDV is schematized in Figure I-F19. wall 2 flow particle B particle A wall 1 transducer emission reception Figure I-F19 Ultrasound Doppler Velocimetry (UDV) [4]. Other techniques. There exist some other techniques for the characterization in situ of the flow, but they are less common. Lagrangian techniques [11][75] consist in following the trajectory of a single particle plunged in the fluid rather than choosing a position in the flow and wait for the particle to pass in the volume (or plan) of interest. Such particle can be either real (tracer) or virtual (element of fluid assumed to have no volume) depending if the experiment is performed on a pilot equipment or by computer simulation. In the former case, the particle can be tracked by a video apparatus (if fluid is transparent and the particle is big enough) or by any other system capable of detecting some specific radiation (e.g. gamma radiation) eventually emitted by the particle (e.g. Positron Emission Particle Tracking, PEPT). The trajectory is a function of the flow but also of the physical properties of the tracing particle (density, dimensions, shape). PhD INSA de Lyon (2008) 54 PROCESSING OF NANOCOMPOSITES I-2.2.3 Melt compounding Local probes Local probes are used to obtain the local concentration value in a given position of the flow. Probes typically require a preliminary calibration, for concentration cannot be directly measured but is be rather deduced from the variations of specific physical and/or chemical properties of the flowing materials. Formerly, colorimetry was limited to the observation of samples recovered only at the exit of the mixing system [76]. Since then, progresses have been done in terms of local probes quality – mostly thanks to the advent of optical fibers, and especially on account of the constructors of spectrophotometers. Nowadays, accurate results can be obtained by using fluorescent tracers and, recently, a successful employ of infrared spectroscopy has also been reported in the presence of particles absorbing specific ranges of radiation wavenumber. Conductivity is another property commonly exploited to measure mixing time and Residence Time Distribution (RTD). High temporal (ms) and spatial (m) resolution probes can be used when visualization is otherwise impossible (e.g. for equipments with opaque walls). There are also some techniques based on pH measurements: pH probes cover bigger volumes (mm3) and have a slower time-response (s) in comparison with conductivity probes; on the other hand, their response is less affected by bubbles and undesired suspended particles. Finally, a possibility is offered by thermal methods: the concentration field can be deduced from temperature if, for example, the tracing agent is a fluid warmer (or colder) than the observed one, and the probe consists of a platinum resistance. However, the temporal resolution of thermal methods is mediocre (the probe is too big), the utilization of viscous fluids is delicate (their viscosity depends on temperature) and sensitivity is limited by the admissible temperature mismatch between the fluid and the tracer (excessive mismatches could locally modify the flow behavior). Laser Induced Fluorescence (LIF). LIF techniques are advantageous for their resolution and sensitivity. Analogously to most techniques of characterization in situ, LIF is restrained to laboratory applications. Fluorescent probes absorb the radiation at a characteristic wavelength and then emit a fraction of the absorbed energy at a longer Antonella ESPOSITO 55 Chapter I wavelength42. If the fluorescent tracer is adequately diluted, the intensity of its emission is proportional to the incident luminous power P0 as well as to the local concentration of tracer C and to the concentration C averaged on the whole optical pathway throughout the solution. Rhodamines are the most frequently used class of fluorescent tracers: they are typically excited by a radiation having a wavelength of 470-520 nm and they emit at a wavelength comprised between 550 and 600 nm. LIF is then performed by injecting a given concentration of fluorescent tracer in the continuous flow where is focused a lowpowered laser: the emission intensity is then collected perpendicularly to the incident laser by a photomultiplier equipped with a filter. The advantage of LIF is represented by its excellent temporal (ms) and spatial resolution (10 m), along with the fact that it‟s non-intrusive and that the admissible tracer dilution can be relatively low (10-5 or even 10-6 mol·dm-3). The disadvantages are the need for equipments with transparent walls and the fact that it‟s only a local measurement. This last problem has been solved by the introduction of a laser sheet (Plan Laser Induced Fluorescence, PLIF). PLIF is a particularly powerful technique to investigate the concentration field of a fluorescent tracer. This technique was firstly used by Van Cruyningen and coll. (1990) and soon after by Mahouast (1993) before being further developed by Houcine and coll. (1996), Marcant and coll. (1997) [4]. The intensity of fluorescence emission is usually mapped by a CCD camera equipped with a filter, then converted to a concentration field by means of a preliminary calibration of the detection system. CCD cameras permit a direct evaluation of the mixing system in order to optimize its geometrical profile, the flow rate, the injection procedure and the feeding position. The preliminary calibration of the detection system typically proceeds as following: the system is filled up with water (or any model fluid) and residual luminosity (background) is measured Io(x, y, t); the system is emptied and filled up again with a solution of fluorescent tracer with the target mixture concentration C* and the luminosity corresponding to the target mixture is measured I*; 42 Some more information about fluorescence and spectrofluorimetry will be given in § II-3.6. PhD INSA de Lyon (2008) 56 PROCESSING OF NANOCOMPOSITES Melt compounding the system is emptied and filled up again with the fluid of interest, the tracer is injected and the luminosity field of the local emission intensity is measured I(x, y, t). Houcine and coll. (1996) reported that the local instant concentration of tracer can be expressed as: C ( x, y, t ) I ( x, y, t ) I o ( x, y) * C I * ( x, y) I o ( x, y) (I-E11) In some cases, this technique can provide an index to estimate the quality of the mixing process [4]. Such index can be derived from the average and variance values of the concentration field or from the mixing time (more precisely from the RTD). Concentration field. The local average concentration C ( x, y, z ) is obtained by averaging a large number n of local values of concentration C ( x, y, z, t ) measured in a given position. A reduced concentration value C ( x, y, z ) can be useful to indicate (if C * ( x, y, z ) different than 1) the deviation of the mixture from the ideal case of a macroscopically perfect mixture. The temporal variance of concentration is expressed as: 2 ( x, y , z , t ) c 2 1 n 2 C ( x, y , z , t ) C ( x, y , z ) n k 1 (I-E12) and could help characterizing mixing dynamics (high variance values correspond to significant fluctuations of the concentration). An advantage of the variance analysis is that the contributions due to vortexes of different size are additives. Anyway, not all the probes are enough sensitive to detect the spatial and temporal phenomena responsible of the fluctuations of concentration [4]. Mixing time. Mixing time is defined as the time necessary the get a good quality mixture. However, the notion of good quality mixture is relative, subjective and depends on the requirements imposed to the final products and on the techniques used to verify that such requirements have been fulfilled. For laminar flows, a local criterion to get the degree of mixing is the thickness of fluid lamellae (also called striations)43; moreover, if 43 Laminar mixing has been introduced in § I-1.3.2 but a more detailed description is due to Ottino [11]. Antonella ESPOSITO 57 Chapter I the flow is non-chaotic, the number of striations is proportional to the number of screw revolutions and the thickness of the striations is inversely proportional to the number of screw revolutions. Contrarily to turbulent mixing, laminar mixing can be affected by “dead zones”, in which the flow is almost stationary: only molecular diffusion assures some matter exchanges between looping fluid lamellae (the existence of multiple loops can require an infinite mixing time, for it would mean that the cyclic stretching, cutting and recombining mechanisms typical of laminar mixing are not possible). Therefore, the presence of loops can be a problematic issue for polymer processing. There are several methods (experimental and computer aided) to measure mixing time. Experimental methods include global and local techniques. Global techniques are based on the visualization of the mixing process directly into the equipment of interest. Of course, these techniques can be used only for equipments having transparent walls and with transparent fluids: their use is limited to specific laboratory equipments filled with model fluids. Anyway, the possibility of visually detect the zones in which mixing isn‟t efficient (maybe because of loops) represents a great interest. The difficulty is due to the fact that the interpretation of the results obtained by visual technique may depend on the observer and the results themselves (qualitative) can‟t help understanding mixing dynamics. The only way to exploit qualitative information is to use it for comparison of different mixing system and to search a solution for eventual mixing inefficiencies. One possibility of visualization is given by rapid reactive systems whose kinetic depends on the mixing conditions and the main result is a color change of the fluid44. Local techniques require placing a concentration probe in one or more positions of the flow and following the response upon injection of a tracer. The possible response of the local probe can be modeled as shown in Figure I-F20 [4]: 44 In this case, it is important to check that the reaction generating the color change doesn‟t significantly modify the rheological behavior of the fluid, and that the amount of tracer introduced is negligible with respect to the volume of the mixing system. Moreover, the method used to inject the coloring solution into the system must be clearly stated in the experimental protocol, for results can dependent on feeding position and rapidity. PhD INSA de Lyon (2008) 58 PROCESSING OF NANOCOMPOSITES Melt compounding measured property global trend equilibrium time measured property equilibrium time measured property equilibrium time Figure I-F20 Possible responses of a local probe used for a local characterization technique of the flow in order to measure the mixing time necessary to accomplish a good quality mixing [4]. Trend I – typical of a laminar flow – the probe periodically detects the passage of the tracer and its response gets attenuated with time. After a given number of passages, an equilibrium value (corresponding to the macroscopically perfect mixture) is reached; Antonella ESPOSITO 59 Chapter I if the asymptote of the curve is shifted with respect to the equilibrium value, the mixing system is affected by some dead zones. Such a curve can always provide a mixing time. Trend II – typical of a situation in which either the injection or the probe are placed in a dead zone of a laminar flow – the tracer arrives to the probe exclusively by diffusion at the interfaces of the recirculation loops, thus very slowly. Such a response usually cannot inform about the mixing time. Trend III – typical of a turbulent flow – as the mechanisms of turbulent diffusion assure a rapid dispersion of the tracer in the fluid, it is common to observe just one peak on the probe response (sometime no peaks are visible, depending on the position of the probe relative to the position of injection). Turbulence produces signal fluctuations, which are even better resoluted if the time-response of the probe is rapid. These methods (both global and local) are well adapted to quantify mixing when the knowledge of flow mechanisms is partially available. If the system generates several recirculation loops, the result strongly depends on the position of the probe with relation to the position in which the tracer is injected. Thus, it is always preferable to perform first a global visualization (just to understand in which position it would be better to place the probe and inject the tracer) followed by a local technique. I-2.2.4 Computer simulation Computer simulation represents a very useful (and probably the most common) tool to visualize and understand the influence of fluid dynamics [11] on processing. Parenthetically, we‟ve already reminded that it is exactly in the field of stirring, mixing and processing that the first codes for computer simulation have been generated. Here, we just want to give some examples of computer simulation of laminar flows. Identification of chaotic flow zones. We‟ve previously introduced the notion of laminar mixing (§ I-1.3.2) insisting on the fact that thermoplastic polymer processing is never supported by turbulence: nonetheless, there‟s a form of chaotic mixing which may be considered a particular type of laminar mixing and typically occurs when a periodic perturbation is applied to the flow. The effect produced on two fluid elements initially PhD INSA de Lyon (2008) 60 PROCESSING OF NANOCOMPOSITES Melt compounding very close is that the perturbation makes them move away from each other: if the distance of the fluid elements is an exponential function of the number of perturbations (i.e. of time) the flow is considered chaotic. Conversely, the laminar flow is non-chaotic if the distance is a linear function of time. Chaotic flow obviously facilitates mixing, for it separates two fluid elements much faster than non-chaotic flow. Mixing occurring in conditions of chaotic flow is called chaotic mixing. Each fluid element evolving in a zone of chaotic flow could statistically occupy all the available positions – at least over long processing time. This property is used to visualize by computer simulation whether a zone of flow is chaotic or not: a significant cross section is chosen and its intersections with the trajectories of the fluid elements are visualized (Poincaré sections) to identify non-chaotic zones – which are then represented by the absence of points (Figure I-F21). Figure I-F21 Example of Poincaré section obtained by computer simulation and showing two zones of non-chaotic laminar flow [4]. Quantification of laminar mixing. A fluid experiencing laminar flow appears made of layers (fluid lamellae) slipping on each other, as previously described. Two main variables can quantify laminar mixing and be deduced from the analysis of striations45 [4][11]: the thickness of the striations – which provides the distance between adjacent fluid elements and therefore determines which mechanism (between molecular diffusion and convective mixing) is the most probable; 45 Information about striations can be found also in § I-1.3.2 and I-2.2.3. Antonella ESPOSITO 61 Chapter I the surface of contact of adjacent striations – which greatly influences any transfer coefficient and, thus, the kinetics of chemical reactions (if present) between the components to be mixed. Again, laminar flow is considered chaotic when the thickness of the striations decreases exponentially with time, whereas the surface of contact, of course, increases exponentially. As it is difficult to estimate the value of these variables – especially when the test involves three-dimensional flows – computer simulation represents a valuable tool to support experiments. An example of computer-aided quantification of laminar mixing is reported in Figure I-F22. Figure I-F22 Quantification of the thickness of the striations for a flow in a micro-channel: (a) Poincaré section (vertical cross section of the micro-channel) and (b) estimation of the thickness of the corresponding striations [4]. I-3 PHOTO-FUNCTIONALIZATION OF NANOFILLERS We pointed out all the difficulties which can be encountered during polymer-clay nanocomposite processing by melt compounding in geometrically complex screw/barrel systems. The geometrical complexity of the equipments, the morphological issues associated to the multiscale structure of such materials, the disadvantages of the existing PhD INSA de Lyon (2008) 62 PROCESSING OF NANOCOMPOSITES Melt compounding tools for morphological characterizations ex situ – as well as the progress of the modern on-line process monitoring techniques – pushed us to try conceiving a detection system capable of performing real-time monitoring of melt compounding of a polymer with a lamellar filler. Our proposal was preparing us a new challenge: how to characterize in situ polymer-clay nanocomposite morphology, being aware of the difficulties we could encounter? Our detection system should be sensitive to multiscale morphologies and compatible with some specific property of the materials of interest, i.e. lamellar fillers. Which properties do lamellar fillers have? Which possibilities of modification? I-3.1 Lamellar fillers Lamellar or plate-like fillers are so called because of their morphology – they are made of platelets having different composition and spatial arrangement. The dimensions of each platelet typically range in the following scales: few nanometers (thickness), tens of nanometers (width) and from tens of nanometers to few micrometers (length). Having a thickness of few nanometers is enough to class lamellar mineral fillers in the group of nanofillers46. Their reduced dimensions are responsible for a high specific surface (from 100 up to 1000 m2g-1) and their particular morphology confers them a high aspect ratio (from 100 up to 1000). The most widespread lamellar fillers belong to the family of 2:1 (double layer minerals) phyllosilicates (montmorillonite) but other families of lamellar fillers are also used to prepare polymer composites, e.g. polysilicates (magadiite) and Layered Double Hydroxide (hydrotalcites). Lamellar fillers can be natural or synthetic. For our purposes, we selected natural lamellar fillers (cheaper than synthetic) belonging to the family of phyllosilicates (more interesting for commercial application, as their use is already widespread but not yet optimized). I-3.1.1 Structure and chemistry Montmorillonites (MMTs) are naturally-occurring multiscale lamellar mineral fillers which consist of crystalline hydrated aluminosilicates containing two repeating 46 The notion of nanocomposite and nanofiller has been introduced in § I-2. Antonella ESPOSITO 63 Chapter I units: tetrahedral and octahedral sheets. In particular, each MMT platelet consists of one octahedral Al(OH)3 layer stacked between two tetrahedral SiO4 layers – the former containing edge-shared octahedra and the latter corner-linked tetrahedra. If each silicate tetrahedron contains one Si atom in its centre, and all gibbsite octahedrons contain only Al atoms, the overall structure is electrically neutral. However, during the geological processes which lead to the formation of MMTs, some structural atoms are replaced by other atoms having the same size (isomorphic substitution) but lower electrical charge: Si4+ can be replaced by Fe3+ or Al3+, and Al3+ can be replaced by Fe2+ or Mg2+. These substitutions generate a deficiency of positive charge and justify the fact that the stacks of platelets are globally negative. The permanent negative charge associated to the immobilized anions47 is equilibrated by the presence of singly- and doubly-charged mobile cations (Na+, K+ and Ca2+, Mg2+) which act as counterions (i.e. compensating cations). These counterions are principally grouped in the interlayer space (also known as clay gallery) – which separates the platelets and whose dimension (basal spacing d001) depends on the nature of the cations and the degree of clay hydration [23][25] – but some of them are also located on the external surface of the platelets. The surface counterions can be readily exchanged with other inorganic or organic ions in all clay minerals, whilst the interlayer counterions are only accessible in expandable (swelling) clays. MMTs are expandable clays: their interlayer inorganic cations can be replaced by other cations having a higher affinity for the fixed ionic sites on the platelets – without any major change of the crystalline structure but an increase of its basal spacing. The d001 value of MMT in the form of powder may vary between 9 and 60 Å [77] depending on the nature and the concentration of the molecules absorbed between the layers and, in the presence of a polymer matrix, it could even attain a hundred of ångströms. The general formula which describes MMT structure is Al2-xMgx[Si4O10](OH)2NaxnH2O. Silicate platelets represent MMT elementary particles (thickness ~ 1 nm): MMT primary particles include 5 to 10 platelets (size 8 to 10 nm) and MMT aggregates consist of several MMT primary particles stacked together without any preferential orientation (size 0.1 to 10 m), as shown in Figure I-F23 [78]. 47 The maximum layer charge for most 2:1 clay minerals is about 1.00 per formula unit [79]. PhD INSA de Lyon (2008) 64 PROCESSING OF NANOCOMPOSITES Melt compounding Figure I-F23 Clay multiscale structure: elementary particles (each single platelet), primary particles (5 to 10 platelets) and aggregates (several primary particles stacked together without any preferential orientation) [78]. This multiscale structure includes voids at any organization level, which explains the outstanding aptitude of MMTs to swell in water or aqueous solutions [80][81] by reversibly gaining and losing water molecules. MMTs can absorb water by hydration of the compensating cations and/or by capillarity penetration, within clay galleries as well as within the voids associated to inter-particle and inter-aggregate porosity [78]. We won‟t provide much more information about clay minerals and MMTs, for nowadays these lamellar mineral fillers probably represent the nanofiller the most commonly used by researchers – most of the scientific papers currently published about polymer-based nanocomposites contain some results obtained with clays (or at least cite someone else‟s results). Many reviews on polymer-clay nanocomposites are available in the literature. For our purposes, we would just underline that natural MMTs don‟t have any specific optical property, i.e. they have no spontaneous optical activities. As we were planning to conceive a new detection system for real-time monitoring of the melt compounding of polymer and lamellar fillers by fluorescence, we firstly needed to find a suitable photo-functionalization method to render clays photo-active48. I-3.1.2 Photo-functionalization methods There are several ways to modify 2:1 phyllosilicates [82]. Clay functionalization can be carried out by adsorption, whether involving chemical linkages (chemisorption 48 The experimental progression which led us to the establishment of a photo-functionalization protocol suitable to our purposes is reported in Chapter II. Further applications of such protocol will be discussed in Chapter III. Antonella ESPOSITO 65 Chapter I or grafting), electrostatic interactions (compensation of electrical charges) or only weak intermolecular interactions (physisorption) between the adsorbent (clay platelet surface) and the adsorbate (any chemical bearing one or more functional groups). The adsorbent surface may be either the external surface of clay platelets, or the internal surface of clay galleries, or even the border of clay platelets – depending on the electrical charges on the surface of the layers, the nature of the interlayer compensating cations and the sterical hindrance effect due to the difference between the adsorbate molecular size and the pristine d001 spacing of the clay. When the sterical effect is moderate and a liquid phase is present to promote molecular diffusion, the adsorbate may migrate in between the clay platelets, getting trapped into the clay galleries and eventually forcing clay to swell49 (i.e. increasing its basal spacing). The ordinary procedure to adsorb a chemical containing a charge-bearing group (i.e. an ionized molecule) on the internal surface of clay galleries is cation exchange process: the cations born by the adsorbate are then substituted to compensating cations. If the adsorbate is an organic molecule containing a charge-bearing group with a high affinity for clay immobilized anions (e.g. a quaternary ammonium salt), the cation exchange process modifies clay nature (from hydrophilic to hydrophobic) to an extent which depends on the nature of the organic molecule. This procedure generates organically-modified clays (or organoclays), characterized by an improved compatibility with most of the polymers. If the adsorbate bears more than one function (e.g. cationic fluorescent dyes contain charge-bearing and fluorophore-bearing groups), cation exchange process is still possible and suitable to functionalize clays, for example by tailoring their optical properties. Whatever the nature of the adsorbate, whatever the procedure used to functionalize clays and regardless of the driving forces which make it possible, diffusion is critical because it influences molecular mobility. If a suitable photo-functionalization method is found to render MMT photoactive, and if the fluorescence behavior of the photo-functionalized lamellar filler can be controlled and correlated to any possible spatial organizations of the platelets (tactoids, intercalated, exfoliated), performing real-time monitoring of the melt processing in a screw/barrel system by fluorescence would present a double advantage: 49 This concerns only expandable clays. PhD INSA de Lyon (2008) 66 PROCESSING OF NANOCOMPOSITES Melt compounding the reduced dimensions of the tracing particles (even if agglomerated) would consent to the photo-functional clays to follow the flow – which is a characterization of distributive mixing; the environmentally-sensitive optical properties of the photo-functional clays would detect (by spectrofluorimetry) any morphological change down to the molecular scale50 – which would be a characterization of dispersive mixing. I-4 A DEEPER INSIGHT INTO THE STATE OF THE ART Cation exchange process is potentially adequate to confer specific (e.g. optical) properties to any layered (natural or synthetic) mineral filler able to exchange cations by dispersing without dissolving in solutions containing water and/or an organic solvent [83-91]. Bujdak [92] has recently analyzed and critically reviewed most of the older papers dealing with the photo-functionalization of clay minerals and their interactions with cationic organic dyes. López Arbeloa and coworkers [93] focused their review on the interactions between rhodamine dyes and clay layered films. MMT is the most common lamellar mineral filler chosen by researchers to study its interactions with rhodamines [86][88][89][91][93][94] and other organic dyes [90][95-98], but other minerals have also been evaluated (fluor-taeniolite [84], laponite B [94][97][99][100][101], saponite [85][98][102], sepiolite [103], fluorohectorite [97], synthetic mica [98], zeolite [87], kaolinite [102]) as well as other cationic organic dyes (1,1'-diethyl-2,2'-cyanine [85][98], oxazine-4 [90], azobenzene [95], methyl green [103], methylene blue [96][97][101][102], acridine orange [104], thiamine [97]). Fluorescent molecules are fickle but versatile tracers, sensitive to any difference in the surrounding environment (chemical composition, atomic arrangement, molecular configuration, physical confinement, temperature). Against any preconceived opinion, the use of fluorescent molecules isn’t only a prerogative of biologists [105-109]: fluorescent tracers and dyes have already been successfully used in the field of polymer processing – to monitor shear effects on phase segregation in the mixture of two 50 The detection of intercalation and exfoliation by fluorescence is possible only if photo-functionalization targets clay galleries. Antonella ESPOSITO 67 Chapter I miscible polymers by in situ fluorescence quenching51 [110], to evaluate the mixing efficiency of an internal batch mixer [111], to estimate the RTD during extrusion [112], to measure the temperature of molten polymers during extrusion and injection molding [113-116], for real-time monitoring of biaxially stretched polypropylene films [117] and, recently, to assess the size and distribution of fillers in a polymer matrix [118]. Considering the number of available on-line monitoring techniques based on optical properties or detection systems52 and the multiplicity of diversified applications [119] [120], fluorescence offers huge potentialities in more than a research field. Rhodamine dyes (whether in the cationic or non-cationic form) have been used since a long time [83] and are nowadays quite familiar to many research teams. Indeed, they represent the ideal probes to study heterogeneous systems thanks to the strong dependence of their absorption and fluorescence emission on the properties of the host matrix: their optical properties are environmentally sensitive and according to the theory of exciton splitting depend on the arrangement and the spatial configuration of the dye molecules (monomers, dimers, J-aggregates, H-aggregates)53 [86]. The organization of the photoactive species within clay inorganic microstructure depends on the host-guest and guest-guest interactions: the adsorption of rhodamine onto clays sometimes leads to metachromasy in the absorption spectrum (i.e. shift of the main absorption band) [86]. Nile Blue (whether in the cationic or non-cationic form) shows clear confinement effects on the amplitude of the Stokes shift54 and the dynamics of solvation in ethanol, when confined to sol-gel glasses with 50 Å and 75 Å average pore size. In particular, its fluorescence emission undergoes a blue shift for confinements inferior to 7.5 nm [121]. If nano-confined in 5 nm pores of the same sol-gel glass but in a non-polar medium such as dodecane, Nile Blue shows an even more pronounced blue shift, from 664 nm (ethanol) to 648 nm (dodecane) [122]. Nile Blue A Perchlorate (NBAP) has already been used for cation exchange processes on commercial clays [122-124]. 51 For the notion of fluorescence quenching, please see § II-3.6. See § I-2.2.2 and § I-2.2.3. 53 These notions will be found again in Chapter II (PHOTO-FUNCTIONALIZATION – Lamellar fillers). 54 The Stokes shift is the wavelength shift of the emission compared to the absorption peak of fluorescent molecules. A simple explanation and some more details about fluorescence will be given in Chapter II. 52 PhD INSA de Lyon (2008) 68 PROCESSING OF NANOCOMPOSITES Melt compounding 9-anthracenemethanol (hydroxylmethyl anthracene) has been previously used as UV-fluorescent tracer by Cassagnau et al. to study mixing processes in a batch mixer for miscible polymers blends, low-viscosity-ratio miscible blends and immiscible blends [111], and to evaluate the residence time distribution in a twin-screw extruder [112]. Zhang et al. [125] recently developed a similar (but ameliorated55) detection system and tested it to assess the local RTD in co-rotating twin-screw extruders. A group of researchers at the National Institute of Standards and Technology (NIST) at Gaithersburg (Maryland, USA) have been working for a long time on process monitoring by fluorescence, using fluorescent local probes for different applications. Fluorescence-based measurements of temperature profiles during polymer processing [113] was one of the first utilizations they made of fluorescent probes – with the purpose of monitoring the packing and cooling phase of injection molding (the optical sensor was positioned in the wall of the mold cavity and consisted of a sapphire window at the end of a sleeved ejector pin into which an optical fiber had been inserted) [126] [127]. Afterward, they adapted this method of temperature measurement to extrusion (once again via standard instrumentation ports)56 [114] and to the process of biaxially stretched polypropylene (PP) films (in this case the sensor with optical fibers, polarizing elements and lenses was mounted above the film, as it was processed in a tenter frame oven stretching machine) [117] – demonstrating the versatility of their technique. Other research groups inspired by this new characterization technique developed their own instrumentation to monitor polymer processing. Cassagnau et al. [111], for instance, developed a UV-fluorescence monitoring device to evaluate in situ the mixing efficiency of an internal batch mixer. They investigated mixing of both miscible and immiscible polymers, as well as mixing of a molten polymer with an additive having low viscosity (plasticizer). They used an optical probe similar to that conceived by the researchers at the NIST, consisting of a system of optical fibers able to both transmit the UV excitation radiation to the processed polymer and to collect the fluorescence 55 Two probes simultaneously measured RTD in any two different locations of the extruder, providing the possibility of calculating the local RTD in between the two locations by deconvolution methods based on a statistical theory for RTD. 56 For extruders, it‟s typically the standard ½ inch diameter sensor port normally used for temperature and pressure transducers. Antonella ESPOSITO 69 Chapter I emission of the tracer. One optical fiber ( 0.9 mm) carries the excitation light and twelve other fibers ( 0.1 mm) transmit the fluorescence emission to a photomultiplier: the total diameter (1.5 mm) probes a sample area which is large compared to the mixing scale but small compared to the volume of the chamber. As their work was intended to quantify the mixing of different materials in molten EVA, the UV tracer was dispersed in a masterbatch of each material of interest (polymers or plasticizer) and an amount57 of masterbatch (1% wt anthracene) was added to the evolving mixture58. They collected fluorescence intensity vs. time curves (Figure I-F24) at different rotational speeds (5, 30, 50, 120 rpm): all the curves showed several peaks59 whose intensity gradually decreased with time and reached a mean value (asymptote). These curves show trends typical of laminar mixing (trend I in Figure I-F20). Actually, the recorded fluorescence intensity corresponds to the emission integrated on the volume of few millimeters in front of the probe, thus characterizes only the macroscale segregation of the tracer. Figure I-F24 UV-Fluorescence device implemented on a Haake Rheomixer 600 batch mixer by Cassagnau et al.[111]. Example of the curves characterizing mixing, scheme of the chamber equipped with the probe and details of the cross section of the optical fibers. 57 All experiments were carried out with 50g of material (total amount) evolving in the chamber of the batch mixer. The total amount of anthracene used was about 0.7 mg, viz. 15 ppm anthracene in the blend. 58 As the fluorescence curves recorded by this method strongly depend on the experimental conditions of injection of the tracer, they chose to simply drop the tracer on the molten flow stream [111]. 59 The earliest peaks can be associated to a saturation of fluorescence emission because they correspond to a high local concentration of anthracene right in front of the device window [111]. PhD INSA de Lyon (2008) 70 PROCESSING OF NANOCOMPOSITES Melt compounding The mixture of highly viscous miscible polymers was investigated by mixing the masterbatch EVA-tracer in the neat EVA. The mixture of a low viscosity additive with a highly viscous molten polymer can provide information complementary with respect to the information got by torque variation (rather representative of the global behavior of the mixture): Cassagnau and coll. found that torque variation and fluorescence emission do not measure the same scale of mixing. The authors also highlighted that their method (as any other method which uses a masterbatch to introduce the fluorescent tracer in the material of interest) informs only about distributive mixing. The UV-fluorescence monitoring device assembled by Cassagnau and coll. [111] has been adapted by Poulesquen et al. [128], in collaboration with Cassagnau‟s equipe, to twin-screw extruders in order to provide experimental validation of theoretical RTD predictions previously made by Poulesquen and Vergnes [129]. Lately, Cassagnau et al. [112] applied their method to study the mixing mechanisms of liquid/polymer in a twinscrew extruder by fluorescence-based measurement of the RTD. The fluorescence-based methods using on-line instruments equipped with optical fibers to assess local RTD during polymer processing are becoming more and more popular, as confirmed by the increasing number of research groups working on this topic [125]. However, since Maddock [73] divulged the experimental procedure60 which, for the first time, visualized (indirectly) polymer processing and illustrated the mechanisms of plastication, the researchers have been looking for an experimental method to directly visualize polymer melting and mixing into screw/barrel systems. Some researchers kept using the procedure indicated by Maddock. Someone considered equipping the existing screw/barrel systems for extrusion and injection molding with transparent windows only in the zones thought to be critical for processing [130][131], so that images from inside the system could be captured by in-line non-invasive techniques. Certainly, many visual analyses of mixing in screw/barrel systems have been done – including those devoted to the mixing in the meter section61 – frequently performed by numerical simulation [132]. 60 This technique, proposed in the late 1950s, was known as “screw freezing” or “screw crash” technique. Alemaskin et Manas-Zloczower [132] properly remind that, although nowadays in polymer processing single-screw extruders are mostly used as pumping devices for processes such as injection or blow molding, their mixing efficiency should not be underestimated. 61 Antonella ESPOSITO 71 Chapter I Indeed, we‟ve already mentioned62 that numerical simulations provide an opportunity to study the mechanisms of flow and mixing in geometrically complex systems without performing the actual experiments. Anyway, in spite of difficulties, some researchers considered the possibility of building specific screw/barrel systems having a completely transparent barrel; Moguedet, in the manuscript of its PhD thesis [74], quickly reviewed the evolution of the main techniques for flow visualization in screw/barrel systems since the pioneer “screw freezing” technique. By the way, it was in the framework of its PhD that Moguedet (with the help of his collaborators) conceived, designed and assembled a pilot screw/barrel system to visualize the 3D trajectories of a single fluorescent particle plunged in a transparent fluid – the barrel being entirely transparent [74]. More detailed information about the system assembled by Moguedet and coll. (called Visiovis) will be given in Chapter IV – as Visiovis represented also the nucleus of the present work. So far, it is obvious that we‟re particularly interested in: melt processing, in-line process monitoring, flow visualization, fluorescent local probes… indeed, although online monitoring of polymer melt processing doesn‟t necessarily have to be performed by a visual detection system (acquisition of images and/or videos by a CCD camera) or exploiting the fluorescence properties of some local probes, such techniques remain the most used in the literature when visualization supposed to help comprehension. On-line monitoring can be visual without using fluorescence (Ing et al. [133] performed in-line monitoring of particles in a polymer melt during extrusion using a scanning particle monitor, Figure I-F25) and, as proved, UV-fluorescence methods don‟t need visualizing to characterize processing (Lutzen et al. [134] performed in-line monitoring of polymer additives during extrusion using a simple UV spectrometer – to cite one more example). Figure I-F25 A schematic diagram of the scanning particle monitor used by Ing et al. [133] to visualize particles in molten polymers during extrusion for applications to quality control. 62 See § I-2.2.4. PhD INSA de Lyon (2008) 72 PROCESSING OF NANOCOMPOSITES Melt compounding In reality, we are rather interested in real-time monitoring of polymer-clay nanocomposite processing, which implies a real-time monitoring of melt compounding and could be helped by a direct visualization of the flow in screw/barrel systems. When processing polymer nanocomposites, conventional techniques of flow visualization can only characterize distributive mixing – since acquiring images and videos of a mixture containing nanofillers almost makes no sense. Nonetheless, visualization keeps being valuable for nanofillers have specific optical properties (e.g. for fluorescent nanofillers). In this case, a fluorescence emission sufficiently intense could be enough to visualize in situ nanofiller distributive mixing by a CCD camera. Besides, the optical properties of fluorescent nanofillers – if properly controlled – could inform on dispersive mixing (something that macroscale techniques – such as CCD cameras – could never do). Once again, fluorescence isn‟t the only property which can be used to monitor polymer-clay nanocomposite processing: Kummer et al. [135], for instance, are actually developing a method for the in-line characterization of polymer-clay nanocomposites by NIR (and UV/VIS) spectroscopy, in combination with microscopic methods. Even the research group at the NIST developed, in association with the fluorescence techniques, a method based on dielectric spectroscopy to monitor polymer/filler compounding [136]: what‟s more, they correlated the degree of exfoliation to the dielectric properties and the light transmission properties of nylon11/clay nanocomposites probed by an online dielectric slit die they specifically designed for process monitoring [137]. However, in our opinion, fluorescent probes still represent the most suitable tool to characterize clay aggregation and, conversely, dispersion (i.e. dispersive mixing), thanks to the incredible versatility of the techniques based on fluorescence measurements. We‟re not alone in our convictions, since Yilmaz and Alemdar [138] proved that fluorescent surfactants can be used for both controlling and measuring the size or organoclay aggregates. In addition, the proficient research activity carried out since a long time at the NIST confirms the huge potentialities of the fluorescence techniques for process monitoring: we aforementioned that they recently extended the use of fluorescence spectroscopy to real-time monitoring of the morphology developed in polymer-clay nanocomposites during melt processing. We‟ll rapidly review the results divulged by this group of researchers. Antonella ESPOSITO 73 Chapter I The researchers at NIST chose fluorescence spectroscopy for process monitoring because one advantage of fluorescence measurements is that, whilst traditional optical methods require the probing light to be perfectly transmittable throughout the thickness of the sample, fluorescence measurements can be carried out by exciting the sample and collecting its response from the same side. This property made fluorescence attractive to implement a new real-time characterization method on the existing equipment (typically having opaque walls). Nanocomposites are frequently prepared by melt compounding of a polymer with lamellar mineral fillers – usually montmorillonite. Cationic fluorescent dyes are easily adsorbed on clay mineral surfaces by cation exchange process: previous works have shown that the intensity of their fluorescence emission depends on both the nature of the adsorbent clay and the concentration of the adsorbate dye – the latter being correlated to the Cation Exchange Capacity (CEC) of the clay. Natural clay can contain significant amounts of Fe in the octahedral layers: some authors assert that the presence of iron can cause fluorescence quenching – but quenching can also occur at high levels of adsorbed dye (concentration quenching) because of dye-dye interactions, as shown in Figure I-F26 [122]. Figure I-F26 Fluorescence spectra of an organically modified clay containing different amounts of Nile Blue (excitation wavelength 407 nm, integration time 4s) [122]. PhD INSA de Lyon (2008) 74 PROCESSING OF NANOCOMPOSITES Melt compounding The researchers at NIST considered exploiting this property of fluorescence (due to the interactions of fluorescent cationic dyes and clay minerals) to monitor polymerclay nanocomposite processing by means of a new multipurpose instrument, conceived to be mounted at the exit of an extruder to obtain real-time dielectric and fluorescence spectroscopy during melt compounding. Bur et al. [136] described the system and demonstrated its capabilities by presenting, at the Polymer Processing and Engineering Conference in Bradford UK (2003), the results of real-time monitoring of nylon-6 and EVA compounded with two commercial organoclays (Cloisite ® 15A and 30B) and doped with benzoxazolyl stilbene (BOS) at concentrations less than 10-5 % wt. The samples used for testing had been prepared by melt compounding the powdered clays and the dye with the polymer (4% and 2% wt clay) in a twin-screw extruder at 30 rpm. The tests revealed that significant differences in the dielectric dispersion parameters are observed for polymer-clay nanocomposites according to their morphology (aggregated, intercalated or exfoliated). A correlation between the degree of exfoliation of the clays, the dielectric properties (i.e. the Maxwell-Wagner and the relaxations) of the polymer composite and its light transmission were soon after reported by probing a compound of nylon-11 with three commercial organoclays (Cloisite ® 15A, 20A and 30B) [137]. It seems that, at the beginning, the new multipurpose instrument was oriented to dielectric spectroscopy, rather than fluorescence… they actually recorded fluorescence spectra too (Figure I-F27) and tried to correlate their observations with WAXS and TEM results. Figure I-F27 Fluorescence spectra of EVA doped with BOS and compounded with Cloisite ® 15A (left) and Cloisite ® 30B (right) clays [136]. WAXD and TEM showed that compounding EVA with Cloisite ® 15A at 194°C generated an intercalated nanocomposite (clay galleries were expanded by 20%), while Antonella ESPOSITO 75 Chapter I Cloisite ® 30B remained aggregated upon mixing with EVA in the same conditions. They observed that the fluorescence spectra of pure BOS in EVA and of the compound BOS/EVA/C30B were identical to the spectrum for pure BOS dissolved in chloroform: they concluded that the fluorescence behavior of free BOS in EVA/30B composite is not affected by the presence of aggregated clays (Figure I-F27 on the right). Conversely, the intercalated EVA/15A nanocomposite causes a radical change of BOS fluorescence spectra (Figure I-F27 on the left): they interpreted such shift in the spectrum as a result of the local change of electric field experienced by dye molecules in their environment. For an intercalated dye molecule, the difference between an organic environment (the matrix) and an inorganic polar environment (clay galleries) can represent a factor of 1.5 concerning the internal electric field, which could explain a change of the magnitude of its Stokes shift. Clay concentration in the polymer matrix is several orders of magnitude greater than dye concentration, thus large clay surfaces absorb few dye molecules. They had to assume that the dye is associated in the intercalated regions of the clay and that this is the reason of the observed spectral shift [136]. We think that the simple addition of BOS to the polymer-clay compound isn‟t as sensitive as expected to its morphological changes. Fluorescent probes can be used in two ways to study the spatial arrangement of organoclays in polymers [138]: the fluorescent dye can be simply added to the compound as a local probe of the system (extrinsic fluoroprobe) – thus the probe measures physical properties of the host medium such as polarity, viscosity and hydrophobicity; the fluorescent dye can be previously adsorbed on the clay particles and act as an organoclay-bonded label (intrinsic fluoroprobe). Apparently, the use of fluorescent dyes as intrinsic fluoroprobes has been proved to be more sensitive and particularly adapted to monitor clay (des)aggregation [138]. Indeed, soon after they developed a new procedure to monitor intercalation and exfoliation in melt-processed polymer-clay nanocomposites by LIF spectroscopy of a dye used as an intrinsic fluoroprobe [122] and reported some early results obtained with nanocomposites prepared compounding an organoclay with polystyrene and polyamide6 (Figure I-F28, respectively on the left and on the right). The organoclay had been formerly exchanged with Nile Blue A Perchlorate (NBAP) as reported elsewhere [124]. PhD INSA de Lyon (2008) 76 PROCESSING OF NANOCOMPOSITES Melt compounding Figure I-F28 On the left: LIF spectra (excitation wavelength 407 nm, integration time 800 ms) of PA-6 processed with 2% of a photo-functionalized organoclay (NBAP level 10% CEC) at 240°C and two different residence times, compared to a melt dispersion of NBAP in PA-6 as reference (240°C, integration time 1s). On the right: LIF spectra of PS nanocomposites with 2% of the photo-functionalized organoclay at different processing temperatures and residence times, compared to a melt dispersion of NBAP in PS (180°C) (exc. wavelength 407 nm, integration time 6s) [122]. They effectively confirmed that NBAP is a sensitive fluorescence probe if coexchanged into clay galleries with traditional quaternary ammonium salts by cation exchange process: the early results indicated that different morphologies (intercalated and exfoliated) can be distinguished on the basis of the appearance and/or change of the relative intensity of fluorescence peaks as a consequence of the decreased fluorescence quenching due the development of the nanocomposite structure. In other words, they assumed that concentration quenching dominates until clay platelets get physically separated by polymer intercalation – then fluorescence emission appears. After that, the interlayer spacing increases and fluorescence emission grows more intense, until clay platelets get totally exfoliated and the fluorescent dye can be considered as unconfined. As expected on the basis of previous studies about Nile Blue [121], while the interlayer spacing increases the wavelength of the main fluorescence emission peak undergoes a shift – which confirms that fluorescence emission depend on the local nanostructure and nano-confinements below 7.5 nm can cause significant blue shifts. The comparison with other techniques correlated polymer intercalation with emissions centered around 565 nm and clay exfoliation with emissions at 605 nm and higher. Antonella ESPOSITO 77 Chapter I At that moment the group of researchers used the analytical microscale technique developed by Bourbigot et al. [72] and other conventional techniques (TEM and XRD) to validate the results obtained with the novel procedure based on NBAP as an intrinsic fluoroprobe [124]. They characterized the samples of PA-6 compounded with NBAPMMT by a mini-extruder at 240°C during two different processing times (1 and 7 min). All the conventional tools for morphological analysis ex situ revealed that the sample extruded for 1 min had a microcomposite intercalated morphology (and a purple color), whereas the sample extruded for 7 min showed a far more exfoliated morphology, with individual clay platelets uniformly distributed in PA-6 matrix (and a bright red color) (Figure I-F29). One possible explanation for color change is that as the layers separate, the effective NBAP concentration decreases, reducing dye-due interactions which cause fluorescence concentration quenching (Figure I-F26). They hypothesized that the 1 min sample is purple because of a mixture of aggregated clay (blue, as NBAP is a dark blue powder) and disordered intercalated or exfoliated clay (red). LIF measurements showed that the color change is also accompanied by a fluorescence emission at 605 nm (red) which grows in intensity as the clay exfoliates [123]. Figure I-F29 PA-6 compounded with NBAP-MMT (2% wt) by a mini-extruder at 240°C for 1 min (A) and 7 min (B). Sample (A) resulted to be a mixture of aggregated and intercalated clay, whereas sample (B) resulted to be a mixture of disordered intercalated and exfoliated clay (according to TEM, XRD and NMR results, and justified by LIF measurements [123]) [124]. The early results obtained with NBAP [122] and the correlations established between morphology and fluorescent emission of the intrinsic fluoroprobe [124] have PhD INSA de Lyon (2008) 78 PROCESSING OF NANOCOMPOSITES Melt compounding been soon after confirmed again by the same group of researchers [123] with two of the most commonly used organoclays: Cloisite ® 15A and ® 30B. They exchanged the selected commercial organoclays with NBAP molecules following a slightly different cation exchange process [83] to replace, in water-ethanol mixture, 1% of the surfactant (in both the organoclays, the surfactant is an organic cation bearing a long alkyl chain). After that, they compounded 4% wt of the photo-functionalized clays with nylon-11 by a batch mixer instrumented with an optical fiber sensor to collect real-time fluorescence spectra [114]. After mixing C15A-NBAP with the polymer at 195°C for 5 min, they got a morphology significantly but not completely exfoliated, i.e. consisting of a combination of exfoliated, intercalated and aggregated clay. Even worse, the compound contained clay particles visible to the naked eye. The absence of fluorescence emission was thus attributed to the high concentration of dye clusters in clay galleries (i.e. to concentration quenching). By acquiring a fluorescence emission spectrum of the compound right after processing (mixing at 195°C for 10 min), they showed that NBAP doesn‟t experience anymore fluorescence quenching and fluorescence emission gets intense (Figure I-F30): they supposed that, during compounding, dye molecules are released from clay galleries when exfoliation occurs (TEM analyses supported their hypothesis that exfoliation is necessary for NBAP molecules to express fluorescence). Figure I-F30 Real-time fluorescence spectra obtained during mixing of nylon-11 at 195°C with Cloisite ® 15A doped NBAP (on the left) and Cloisite ® 30B doped NBAP (on the right). Left: thirteen spectra over a period of 10 min. Right: eight spectra over a period of 9 min [123]. The change in the spectra over a period of 10 min of compounding showed the development of the fluorescence emission: in particular, two peaks (509 and 605 nm) Antonella ESPOSITO 79 Chapter I changed intensity in opposite directions as a function of mixing time. These changes follow the release of dye (from clay galleries) in the form of concentrated clusters that are subsequently dispersed as mixing proceeds. The peak at 605 nm, initially suppressed by concentration quenching, became more intense as the dispersion of NBAP molecules through the matrix proceeded, i.e. as dye-dye interactions diminished. By plotting the intensities ratio of the peaks at 605 nm and 509 nm (I605/I509) as a function of time, they obtained the graph shown in Figure I-F31 (on the left). They observed that after 10 min exfoliation was incomplete and suggested that the positive slope for long processing times could be explained by two phenomena: a better dispersion of the dye molecules (which minimizes dye-dye interactions) and the progress of exfoliation. Figure I-F31 The ratio of fluorescence intensities vs. the mixing time is plotted for the Cloisite ® 15A doped NBAP mixed with nylon 11 (on the left, 605 over 509 nm) and for the Cloisite ® 30B doped NBAP mixed with the same polymer (on the right, 609 over 504 nm) [123]. The results obtained mixing C30B-NBAP with nylon-11 are different from the results obtained with C15A-NBAP: the intensity of fluorescence emission at 605 nm remained relatively small compared to that at 509 nm, although a shoulder in the curve was observed in the vicinity of 605 nm. Plotting the ratio I605/I509 vs. time (Figure I-F31 on the right) showed that the reshape of fluorescence spectra was smaller compared to the previous case: they concluded that the composite with C30B-NBAP had substantial aggregate and intercalated microstructure that did not allow NBAP molecules to migrate from clay galleries. They confirmed the limited extent of exfoliation by means of realtime dielectric and light transmission measurements [140]. PhD INSA de Lyon (2008) 80 PROCESSING OF NANOCOMPOSITES Melt compounding Recently, the same group of researchers published some more results [139]: in practice, they just transposed the fluorescence-based method previously established to monitor polymer-clay compounding and clay exfoliation in a batch mixer [123] to a twin-screw extruder. They confirmed that twin-screw extruders are much more efficient in producing clay exfoliation than batch mixers. However, even if the composite based on C30B resulted better exfoliated by the extruder in comparison with the batch mixer, the morphological differences between C15A and C30B persisted. Before concluding, it‟s worthy to rapidly describe the optical probe used by the researchers at the NIST to obtain the results here reviewed. Figure I-F32 The standard ½ inch bolt modified by Bur et al. [136] to host an optical probe. The system has been reproduced also by Cassagnau et al. [111]. The optical probe developed in NIST laboratories consists of a bundle of seven 200 m core optical fibers placed into a sleeved standard ½ inch sensor bolt with a sapphire window at its extremity (Figure I-F32). It operates in two possible modes: to measure light transmission through the evolving fluid; to measure fluorescence emission of a dye eventually present in the fluid. In the transmission mode, one of the fibers transmits light from the light source through a focusing lens, then the sapphire window, then the fluid; transmitted light reflects off the far stainless steel surface (please remind that this probe was designed to be mounted on a slit at the exit of an extruder) and reverses its path through the fluid, the sapphire Antonella ESPOSITO 81 Chapter I window and the focusing lens. The reflected light is collected by the other six fibers and transmitted to a photomultiplier detector. For fluorescence measurements, the difference is that the optical fibers which collect the signal are rather connected to a spectrometer. Such a probe could probably be adapted to any existing processing equipment – the only requirement being the availability of an access point to the processing volume. If this is impossible, the dispersion mechanisms associated to melt processing and mixing could be both visualized and characterized in situ by a system such as the transparent Couette flow cell that Mighri and Huneault used to visualize the dispersion of model fluids [141]. Indeed, the use of photo-functionalized clays with such a flow cell would be interesting to definitely confirm the relationship between specific morphologies and the expression of fluorescence by a cationic fluorescent dye confined into clay galleries. Table I-T2 Summary of the most relevant works present in the literature (to be continued 1/3) Ref. [126] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Ref. [114] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [141] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system PhD INSA de Lyon (2008) Bur et al. (1997) DiMethylAmino DiPhenyl Hexatriene / Bis-Pyrene Propane (BPP) PolyEthylene (PE) / PolyStyrene (PS) Injection molding Fluorescence intensity measurements Xenon arc lamp Bifurcated optical fibers probe adjusted in mold ejector pin channels Bur et al. (2001) Bis-Pyrene Propane (BPP), BenzOxazolyl Stilbene (BOS), perylene Poly (MethylMethAcrylate) (PMMA), Poly Carbonate (PC) Extruder Fluorescence-based temperature measurements Laser or Xenon arc lamp Bifurcated optical fibers probe through existing instrumentation ports Polymer pellets (5% of the total amount) coated by solution dye doping Mighri et al. (2001) Drop of model fluids colored with red pigments Poly DyMethylSiloxane (PDMS) (5, 10 and 30 Pa·s) Two counter-rotating concentric cylinders (outer quartz, inner steel) Visual Couette rheology High-resolution digital camcorder + macro lens + digital chronometer 82 PROCESSING OF NANOCOMPOSITES Melt compounding Table I-T2 Summary of the most relevant works present in the literature (to be continued 2/3) Ref. [111] Cassagnau et al. (2003) Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [112] Hydroxymethyl anthracene (9-anthracenemethanol) Ethyl Vinyl Acetate (EVA) + additives (miscible/immiscible/low viscosity) Haake Rheomix 600 batch mixer Mixing efficiency evaluated by intensity of fluorescence emission Mercury lamp (filter 380 nm) Optical fibers probe 1% wt tracer in a masterbatch of the material of interest = 15 ppm total Cassagnau et al. (2005) Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [136] Hydroxymethyl anthracene (9-anthracenemethanol) Ethyl Vinyl Acetate (EVA) + additives (miscible/immiscible) Intermeshing self-wiping co-rotating twin-screw extruder RTD measured by the intensity of fluorescence emission Mercury lamp (filter 380 nm) Optical fibers probe as in [111] 1% wt tracer in a masterbatch of the material of interest = 15 ppm total Bur et al. (2003) Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [137] BenzOxazolyl Stilbene (BOS) Nylon-6, Poly Ethyl Vinyl Acetate (EVA copolymer) Cloisite ® 15A and Closite ® 30B -5 Compound (4% and 2% wt clay, 10 % wt BOS) twin-screw extruder Spectrofluorimetry and dielectric spectroscopy (online) n.a. New multipurpose instrument for on-line process monitoring (slit die) Presentation of the slit die, BOS as extrinsic fluoroprobe! Lee et al. (2004) Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [122] Nylon-11 + Cloisite ® 15A, Cloisite ® 20A and Cloisite ® 30B, Cloisite ® Na Compound (4% wt clay) twin-screw extruder (198°C for 4 min) Dielectric spectroscopy and conventional light transmission (on-line) n.a. Cell equipped with a dielectric sensor and an optical sensor (slit die) Correlation morphology vs. dielectric properties vs. light transmission Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Nile Blue A Perchlorate (NBAP), Methylene Blue (MB) PolyStyrene (PS), PolyAmide-6 (PA-6) Two laboratory-modified organoclays Compounding (2% wt clay) mini-extruder, different temperature and time Spectrofluorimetry (off-line) Several (results obtained at 407 nm) Bifurcated optical fibers probe Spectra at room temperature AFTER extrusion: perspective of using the method for on-line monitoring of nanocomposites DURING extrusion Notes Antonella ESPOSITO Maupin et al. (2004) 83 Chapter I Table I-T2 Summary of the most relevant works present in the literature (3/3) Ref. [124] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Ref. [123] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [74] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [125] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes Ref. [139] Tracer Polymer matrix Fillers Processing Analytical technique Excitation light Detection system Notes PhD INSA de Lyon (2008) Gilman et al. (2004) Nile Blue A Perchlorate (NBAP) PolyAmide-6 (PA-6) Organoclay (high temperature stable trialkyl imidazolium-based cation) Melt compounding (2% wt clay) mini-extruder, different mixing times Laser Induced Fluorescence (LIF) (on-line), XRD, TEM, NMR (off-line) n.a. n.a. Bur et al. (2004) Nile Blue A Perchlorate (NBAP) Nylon-11 Cloisite ® 15A and Closite ® 30B Compound (4% wt clay) batch mixer Spectrofluorimetry Diode laser Power Technologies (407 nm, 30 mW) Optical fibers probe Further cation exchange process (replaced 1% clay surfactant) Moguedet (2005) Sample of fluorescent nylon ( 0.4 mm) Poly DyMethylSiloxane (PDMS) (100 Pa·s) Transparent screw/barrel system modeling meter section (Visiovis) Particle Tracking Velocimetry (PTV) (detection of the brightest point) UV diodes (4) (400 ± 5 nm) CCD cameras (4) More details and the origins of Visiovis in Chapter IV Zhang et al. (2006) Masterbatch pellets (PS + anthracene, concentration 1, 3, 5 or 10% wt) Polystyrene (PS) Co-rotating twin-screw extruder, different screw configurations New in-line fluorescence-based RTD measuring system UV high-pressure mercury lamp (125 W) Bifurcated optical fibers probe + photomultiplier Reference to Hu G.H., Kadri I., Picot C. Polym. Eng. Sci. 39, 930 (1999) Bur et al. (2007) Nile Blue A Perchlorate (NBAP) Nylon-11 Cloisite ® 15A and Cloisite ® 30B, laboratory-modified organoclay Compound (4% wt clay) batch mixer and co-rotating twin-screw extruder Spectrofluorimetry Violet laser Power Technology (407 nm, 30 mW) Optical fiber probe in standard ½ inch instrument port - slit die (extruder) Clay photo-functionalization by a second cation exchange process (replaced surfactant 5% mol C15A and C30B, 1% mol other organoclay) 84 PROCESSING OF NANOCOMPOSITES I-R Melt compounding REFERENCES [1] Canalini G. Principi e scopi del compounding. XVI Convegno-Scuola AIM, 29 maggio – 3 giugno 1994, Gargnano (BS). [2] Del Monte E., Guggiari A. Le tecnologie impiegate nel compounding. XVI ConvegnoScuola AIM, 29 maggio – 3 giugno 1994, Gargnano (BS). [3] Dupuy C., Bussi P. New dispersion process for submicron fillers in thermoplastics. Macromol. Symp. 2001, 169, 103-107. [4] Xuereb C., Poux M., Bertrand J. Agitation et mélange. Aspects fondamentaux et applications industrielles. Paris: Dunod, 2006, 398 p. (L‟Usine Nouvelle, Série Chimie) [5] McAlpine M., Hudson N.E., Liggat J.J. et al. Study of the factors influencing the exfoliation of an organically modified montmorillonite in methyl methacrylate/poly (methyl methacrylate) mixtures. J. Appl. Polym. Sci. 2006, 99, 2614-2626. [6] Lee E.C., Mielewski D.F., Baird R.J. Exfoliation and dispersion enhancement in polypropylene nanocomposites by in-situ melt phase ultrasonication. Polym. Eng. Sci. 2004, 44 (9), 1773-1782. [7] Filippini Fantoni R. Effetti della trasformazione sui materiali polimerici. XVI Convegno-Scuola AIM, 29 maggio – 3 giugno 1994, Gargnano (BS). [8] Chailly M. Influence des traitements de surface de moule dans le procédé d’injection moulage. Application aux défauts d’aspects. Thèse. Lyon: INSA de Lyon, 2007, 216 p. [9] Nunn R.E. The reciprocating screw process. Chapter 3. Injection Moulding Handbook (1983). [10] Tavecchia S. Descrizione generale della pressa ad iniezione. XVI Convegno-Scuola AIM, 29 maggio – 3 giugno 1994, Gargnano (BS). [11] Ottino J.M. The kinematics of mixing: stretching, chaos, and transport. 1st Ed. Cambridge: Cambridge University Press, 1989, 364 p. ISBN 0-521-36878-2. [12] Elimelech M, Gregory J., Jia X. et al. Particle deposition and aggregation. Measurement, modelling and simulation. 1st Ed. Oxford [England]: ButterworthHeinemann, 1995, 441 p. [13] Patton T.C. Paint flow and pigment dispersion. A rheological approach to coating and ink technology. 2nd Ed. New York: John Wiley & Sons, New York, 1979, 631 p. [14] Spicer P.T. Shear-induced aggregation-fragmentation: mixing and aggregate morphology effects. PhD thesis. Cincinnati: Division of Research and Advanced Studies of the University of Cincinnati, 1997, 283 p. Antonella ESPOSITO 85 Chapter I [15] Dennis H.R., Hunter D.L., Chang D. et al. Effect of melt processing conditions on the extent of exfoliation in organoclay-based nanocomposites. Polymer 2001,42,9513-9522. [16] Lertwimolnun W., Vergnes B. Influence de la dispersion sur le comportement rhéologique de nanocomposites polypropylène/argile. Rhéologie 2004, 5, 27-35. [17] Zhao J., Morgan A.B., Harris J.D. Rheological characterization of polystyrene-clay nanocomposites to compare the degree of exfoliation and dispersion. Polymer 2005, 46, 8641-8660. [18] Fedullo N., Sclavons M., Bailly C. et al. Nanocomposites from untreated clay: a myth? Macromol. Symp. 2006, 233, 235-245. [19] Valette R., Bruchon J., Digonnet H. et al. Méthodes d’interaction fluide-structure pour la simulation multi-échelle des procédés de mélange. Mécanique & Industries 2007, 8, 251-258. [20] Lertwimolnun W., Vergnes B. Effect of processing conditions on the formation of polypropylene/organoclay nanocomposites in a twin screw extruder. Polym. Eng. Sci. 2006, 46, 314-323. [21] Lertwimolnun W., Vergnes B. Influence of compatibilizer and processing conditions on the dispersion of nanoclay in a polypropylene matrix. Polymer 2005, 46, 3462-3471. [22] Mallick P.K. Fiber-reinforced composites: materials, manufacturing and design. 2nd Ed. New York: Marcel Dekker Inc., 1993, 586 p. [23] Schadler L.S. Polymer-based and polymer-filled nanocomposites. In: Ajayan P.M., Schadler L.S., Braun P.V. Nanocomposite Science and Technology. 1st Ed. Wiley-VCH Weinheim: Verlag GmbH & Co. KGaA, 2003, 240 p. ISBN 3527303596. [24] Gao F. Clay-polymer composites: the story. Materials Today 2004, November, 50-55. [25] Wautelet M. et coll. Les Nanotechnologies. 1st Ed. Paris: Dunod, 2003, 248 p. [26] Alexandre M., Dubois P. Polymer-layered silicate nanocomposites: preparation, properties and uses of a new class of materials. Mater. Sci. Eng. 2000, 28, 1-63. [27] Okada et al. Composite material and process for manufacturing same. US Patent 4,739,007 (April 19, 1988). [28] Kawasumi et al. Process for producing composite material. US Patent 4,810,734 (March 7, 1989). [29] Biswas M., Sinha Ray S. Recent progress in synthesis and evaluation of polymermontmorillonite nanocomposites. Adv. Polym. Sci. 2001, 155, 167-221. [30] Giannelis E.P. Polymer layered silicate nanocomposites. Adv. Mater. 1996, 8(1), 29-35. PhD INSA de Lyon (2008) 86 PROCESSING OF NANOCOMPOSITES Melt compounding [31] Giannelis E.P. Polymer-layered silicate nanocomposites: synthesis, properties and applications. Appl. Organomet. Chem. 1998, 12, 675-680. [32] Gilman J.W. Flammability and thermal stability studies of polymer layered-silicate (clay) nanocomposites. Appl. Clay Sci. 1999, 15, 31-49. [33] LeBaron P.C., Wang Z., Pinnavaia T.J. Polymer-layered silicate nanocomposites an overview. Appl. Clay Sci. 1999, 15, 11-29. [34] Oriakhi C.O. Polymer nanocomposition approach to advanced materials. J. Chem. Educ. 2000, 77 (9), 1138-1146. [35] Zanetti M., Lomakin S., Camino G. Polymer layered silicate nanocomposites. Macromol. Mater. Eng. 2000, 279, 1-9. [36] Schmidt D., Shah D., Giannelis E.P. New advances in polymer/layered silicate nanocomposites. Curr. Opin. Solid State Mater. Sci. 2002, 6,205-212. [37] Gensler R., Gröppel P., Muhrer V. et al. Application of nanoparticles in polymers for electronics and electrical engineering. Part. Part. Syst. Char. 2002, 19, 293-299. [38] Sinha Ray S., Okamoto M. Polymer/layered silicate nanocomposites: a review from preparation to processing. Prog. Polym. Sci. 2003, 28, 1539-1641. [39] Ahmadi S.J., Huang Y.D., Li W. Synthetic routes, properties and future applications of polymer-layered silicate nanocomposites. J. Mater. Sci. 2004, 39, 1919-1925. [40] Thostenson E.T., Li C., Chou T.-W. Nanocomposites in context. Compos. Sci. Technol. 2005, 65, 491-516. [41] Gacitua W., Ballerini A., Zhang J. Polymer nanocomposites: synthetic and natural fillers. A review. Maderas ciencia y tecnología 2005, 7 (3), 159-178. [42] Jordan J., Jacob K.I., Tannenbaum R. et al. Experimental trends in polymer nanocomposites – a review. Mater. Sci. Eng. A 2005, 393, 1-11. [43] Zeng Q.H., Yu A.B., Lu G.Q. (Max) et al. Clay-based polymer nanocomposites: research and commercial development. J. Nanosci. Nanotech. 2005, 5, 1574-1592. [44] Liu J., Boo W.-J., Clearfield A. et al. Intercalation and exfoliation: a review on morphology of polymer nanocomposites reinforced by inorganic layer structures. Mater. Manuf. Processes 2006, 20, 143-151. [45] Vermogen A., Masenelli-Varlot K., Séguéla R. et al. Evaluation of the structure and dispersion in polymer-layered silicate nanocomposites. Macromol. 2005, 38,9661-9669. [46] Pozsgay A, Fráter T, Százdi L. et al. Gallery structure and exfoliation of organophilized montmorillonite: effect on composite properties. Eur. Polym. J. 2004, 40, 27-36. Antonella ESPOSITO 87 Chapter I [47] Murphy J. Even clay has to work nowadays... Plastics Additives & Compounding, March 2000, 18-23. [48] Zilg C., Dietsche F., Hoffmann B. et al. Nanofillers based upon organophilic layered silicates. Macromol. Symp. 2001, 169, 65-77. [49] Morgan A.B., Gilman J.W. Characterization of polymer-layered silicate (clay) nanocomposites by Transmission Electron Microscopy and X-ray Diffraction: a comparative study. J. Appl. Polym. Sci. 2003, 87, 1329-1338. [50] Eckel D.F., Balogh M.P., Fasulo P.D. et al. Assessing organo-clay dispersion in polymer nanocomposites. J. Appl. Polym. Sci. 2004, 93, 1110-1117. [51] Homminga D., Goderis B., Hoffman S. et al. Influence of shear flow on the preparation of polymer layered silicate nanocomposites. Polymer 2005, 46, 9941-9954. [52] Van S.T., Velamakanni B.V., Adkins R.R. Comparison of methods to assess pigment dispersion. J. Coat. Technol. 2001, 73 (923), 61-70. [53] Médéric P., Razafinimaro T., Aubry T. Influence of melt-blending conditions on structural, rheological, and interfacial properties of polyamide-12 layered silicate nanocomposites. Polym. Eng. Sci. 2006, 46, 986-994. [54] Liu T., Chauhari Tjiu W., He C. et al. A processing-induced clay dispersion and its effect on the structure and properties of polyamide 6. Polym. Inter. 2004, 53, 392-399. [55] Cser F., Bhattacharya S.N. Study of the orientation and the degree of exfoliation of nanoparticles in poly(ethylene–vinyl acetate) nanocomposites. J. Appl. Polym. Sci. 2003, 90, 3026-3031. [56] Vaia RA, Liu W. X-ray powder diffraction of polymer/layered silicate nanocomposites: model and practice. J. Polym. Sci.: Part B: Polym. Phys. 2002, 40, 1590-1600. [57] Yu Z.-Z., Yang M., Zhang Q. et al. Dispersion and distribution of organically modified montmorillonite in nylon-66 matrix.J. Polym. Sci. B: Polym. Phys. 2003, 41, 1234-1243. [58] Gianelli W., Camino G., Dintcheva N.T. et al. EVA-montmorillonite nanocomposites: effect of processing conditions. Macromol. Mater. Eng. 2004, 289, 238-244. [59] Wagener R., Reisinger T.J.G. A rheological method to compare the degree of exfoliation of nanocomposites. Polymer 2003, 44, 7513-7518. [60] Krishnamoorti R., Yurekli K. Rheology of polymer layered silicate nanocomposites. Curr. Opin. Colloid Interface Sci. 2001, 6, 464-470. [61] Krishnamoorti R., Ren J, Silva A.S. Shear response of layered silicate nanocomposites. J. Chem. Phys. 2001, 114 (11), 4968-4973. PhD INSA de Lyon (2008) 88 PROCESSING OF NANOCOMPOSITES Melt compounding [62] Aranguren M.I., Mora E., Degroot J.V. (Jr.) et al. Effect of reinforcing fillers on the rheology of polymer melts. J. Rheol. 1991, 36 (6), 1165-1182. [63] Bar-Chaput S., Carrot C. Rheology as a tool for the analysis of the dispersion of carbon filler in polymers. Rheol. Acta 2006, 45, 339-347. [64] Vermant J., Ceccia S., Dolgovskij M.K. et al. Quantifying dispersion of layered nanocomposites via melt rheology. J. Rheol. 2007, 51 (3), 429-450. [65] Rhoney I., Brown S., Hudson N.E. et al. Influence of processing method on the exfoliation process for organically modified clay systems. I. Polyurethanes. J. Appl. Polym. Sci. 2004, 91, 1335-1343. [66] Xu L., Reeder S., Thopasridharan M. et al. Structure and melt rheology of polystyrenebased layered silicate nanocomposites. Nanotechnol. 2005, 16, S514-S521. [67] Mills S.L., Lees G.C., Liauw C.M. et al. An improved method for the dispersion assessment of flame retardant filler-polymer systems based on the multifractal analysis of SEM images. Macromol. Mater. Eng. 2004, 289, 864-871. [68] Schmidt D.F., Clément F., Giannelis E.P. On the origins of silicate dispersion in polysiloxane/layered-silicate nanocomposites. Adv. Funct. Mater. 2006, 16, 417-425. [69] Kaneko M.L.Q.A., Torriani I.L., Yoshida I.V.P. Morphological evaluation of silicone/clay slurries by small-angle/wide-angle X-ray scattering. J. Braz. Chem. Soc. 2007, 18 (4), 765-773. [70] VanderHart D.L., Asano A., Gilman J.W. Solid-state NMR investigation of paramagnetic nylon-6 clay nanocomposites. 2. Measurement of clay dispersion, crystal stratification, and stability of organic modifiers. Chem. Mater. 2001, 13, 3796-3809. [71] VanderHart D.L., Asano A., Gilman J.W. NMR measurements related to clay-dispersion quality and organic-modifier stability in nylon6/clay nanocomposites. Macromol. 2001, 34, 3819-3882. [72] Bourbigot S., VanderHart D.L., Gilman J.W. et al. Investigation of nanodispersion in polystyrene-montmorillonite nanocomposites by solid-state NMR. J. Polym. Sci., Part B: Polym. Phys. 2003, 41, 3188-3213. [73] Maddock B.H. A visual analysis of flow and mixing in extruder screws. SPE Journal 1959, 15, 383-389. [74] Moguedet M. Développement d’un outil d’aide à la conception et au fonctionnement d’un ensemble vis/fourreau industriel – Application à l’injection des thermoplastiques chargés fibres de verre longues. Thèse. Lyon: INSA de Lyon, 2005, 124 p. Antonella ESPOSITO 89 Chapter I [75] Willneff J., Lüthi B. Particle tracking velocimetry measurements for lagrangian analysis of turbulent flows. Optical 3-D Measurements Techniques VI, 22-26 September 2003, Zurich (Switzerland). Vol. II, pp. 191-198. [76] Latif L., Saidpour S.H. Assessment of pigment distribution in molded samples using an image processing technique. Adv. Polym. Technol. 1996, 15 (4), 337-344. [77] Ryan W. Properties of ceramic raw materials. 2nd edition. New York: Pergamon Press, 1978, 120 p. [78] Le Pluart L., Duchet J., Sautereau H. et al. Rheological properties of organoclay suspensions in epoxy network precursors. Appl. Clay Sci. 2004, 25, 207-219. [79] Osman M.A., Ploetze M., Suter U.W. Surface treatment of clay minerals — thermal stability, basal-plane spacing and surface coverage.J.Mater.Chem.2003, 13, 2359-2366. [80] Burgentzlé D., Duchet J., Gérard J.F. et al. Solvent-based nanocomposite coatings I. Dispersion of organophilic montmorillonite in organic solvents. J. Colloid Interface Sci. 2004, 278, 26-39. [81] Gates W.P. Crystalline swelling of organo-modified clays in ethanol-water solutions. Appl. Clay Sci. 2004, 27, 1-12. [82] Bergaya F., Lagaly G. Surface modification of clay minerals. Appl. Clay Sci. 2001, 19, 1-3. [83] Endo T., Sato T., Shimada M. Fluorescence properties of the dye-intercalated smectite. J. Phys. Chem. Solids 1986, 47 (8), 799-804. [84] Fujita T., Iyi N., Kosugi T. et al. Intercalation characteristics of rhodamine 6G in fluortaeniolite: orientation in the gallery. Clays and Clay Minerals 1997, 45 (1), 77-84. [85] Iwasaki M., Kita M., Ito K. et al. Intercalation characteristics of 1,1'-diethyl-2,2'cyanine and other cationic dyes in synthetic saponite: orientation in the interlayer. Clays Clay Miner. 2000, 48 (3) 392-399. [86] Čapková P., Malý P., Pospíšil M. et al. Effect of surface and interlayer structure on the fluorescence of rhodamine B-montmorillonite: modeling and experiment. J. Colloid Interface Sci. 2004, 277, 128-137. [87] Vietze U., Krauß O., Laeri F. et al. Zeolite-dye microlasers. Phys. Rev. Lett. 1998, 81 (21), 4628-4631. [88] Gemeay A.H. Adsorption characteristics and the kinetics of the cation exchange of rhodamine-6G with Na+-montmorillonite. J. Colloid Interface Sci. 2002, 251, 235-241. [89] Pospíšil M., Čapková P., Weissmannová H. et al. Structure analysis of montmorillonite intercalated with rhodamine B: modeling and experiment. J. Mol. Model. 2003, 9,39-46. PhD INSA de Lyon (2008) 90 PROCESSING OF NANOCOMPOSITES Melt compounding [90] Kaneko Y., Iyi N., Bujdak J. et al. Effect of layer charge density on orientation and aggregation of a cationic laser dye incorporated in the interlayer space of montmorillonites. J. Colloid Interface Sci. 2004, 269, 22-25. [91] Sasai R., Itoh T., Iyi N. et al. Preparation of hybrid organic/inorganic luminescent thin solid films with highly concentrated laser-dye cations. Chem. Lett. 2005, 34 (11), 14901491. [92] Bujdák J. Effect of the layer charge of clay minerals on optical properties of organic dyes. A review. Appl. Clay Sci. 2006, 34, 58-73. [93] López Arbeloa F., Martínez Martínez V., Arbeloa T. et al. Photoresponse and anisotropy of rhodamine dye intercalated in ordered clay layered films. J. Photochem. Photobiol. C: Photochem. Rev. 2007, 8, 85-108. [94] López Arbeloa F., Tapia Estévez M.J., López Arbeloa T. et al. Spectroscopic study of the adsorption of rhodamine6G on clay minerals in aqueous suspensions. Clay Miner. 1997, 32, 97-106. [95] Ogawa M., Ishii T., Miyamoto N. et al. Intercalation of a cationic azobenzene into montmorillonite. Appl. Clay Sci. 2003, 22, 179-185. [96] Kaneko Y., Iyi N., Bujdák J. et al. Molecular orientation of methylene blue intercalated in layer-charge-controlled montmorillonites. J. Mat. Res. 2003, 18 (11), 2639-2643. [97] Czímerová A., Bujdák J., Gáplovský A. The aggregation of thionine and methylene blue dye in smectite dispersion. Colloids Surf. A: Physicochem. Eng. Asp. 2004, 243, 89-96. [98] Miyamoto N., Kawai R., Kuroda K. et al. Adsorption and aggregation of a cationic cyanine dye on layered clay minerals. Appl. Clay Sci. 2000, 16, 161-170. [99] Tapia Estévez M.J., López Arbeloa F., López Arbeloa T. et al. Spectroscopic study of the adsorption of rhodamine 6G on laponite B for low loadings. Clay Miner. 1994, 29, 105-113. [100] Salleres S., López Arbeloa F., Martínez V. et al. Adsorption of fluorescent R6G dye into organophilic C12TMA laponite films. J. Colloid Interface Sci. 2008, 321, 212-219. [101] Schoonheydt R.A., Heughebaert L. Clay adsorbed dyes: methylene blue on laponite. Clay Miner. 1992, 27, 91-100. [102] Ghosh D., Bhattacharyya K.G. Adsorption of methylene blue on kaolinite. Appl. Clay Sci. 2002, 20, 295-300. [103] Rytwo G., Nir S., Crespin M. et al. Adsorption and interactions of methyl green with montmorillonite and sepiolite. J. Colloid Interface Sci. 2000, 222, 12-19. Antonella ESPOSITO 91 Chapter I [104] Garfinkel-Shweky D., Yariv S. Metachromasy in clay-dye systems: the adsorption of acridine orange by Na-saponite. Clay Miner. 1997, 32, 653-663. [105] Vallotton P., Ponti A., Waterman-Storer C.M. et al. Recovery, visualization and analysis of actin and tubulin polymer flow in live cells: a fluorescent speckle microscopy study. Biophys. J. 2003, 85, 1289-1306. [106] Bonnet N., Zahm J.-M. Analysis of image sequences in fluorescence videomicroscopy of stationary objects. Cytometry 1998, 31, 217-228. [107] Parak W.J., Pellegrino T., Plank C. Labelling of cells with quantum dots. Nanotech. 2005, 16, R9-R25. [108] Hao M., Maxfield F.R. Analyzing microdomains in biological membranes using fluorescence techniques. J. Fluoresc. 2001, 11 (4), 287-295. [109] Pinaud F., Michalet X., Bentolila L.A. et al. Advances in fluorescence imaging with quantum dot bio-probes. Biomater. 2006, 27, 1679-1687. [110] Mani S., Malone M.F., Winter H.H. et al. Effects of shear on miscible polymer blends: in-situ fluorescence studies. Macromol. 1991, 24, 5451-5458. [111] Cassagnau P., Melis F., Bounor-Legare V. UV fluorescence monitoring of the mixing of molten polymers in a batch mixer. Polym. Eng. Sci. 2003, 43 (4), 923-932. [112] Cassagnau P., Courmont M., Melis F. et al. Study of mixing of liquid/polymer in twin screw extruder by residence time distribution. Polym. Eng. Sci. 2005, 45, 926-934. [113] Migler K.B., Bur A.J. Fluorescence based measurement of temperature profiles during polymer processing. Polym. Eng. Sci. 1998, 38 (1), 213-221. [114] Bur A.J., Vangel M.G., Roth S.C. Fluorescence based temperature measurements and applications to real-time polymer processing. Polym. Eng. Sci. 2001, 41 (8), 1380-1389. [115] Bur A.J., Roth S.C. Fluorescence temperature measurements: methodology for applications to process monitoring. Polym. Eng. Sci. 2004, 44 (5), 898-908. [116] Bur A.J., Vangel M.G., Roth S. Temperature dependence of fluorescent probes for applications to polymer materials processing. Appl. Spectr. 2002, 56 (2), 174-181. [117] Bur A.J., Roth S.C. Real-time monitoring of fluorescence anisotropy and temperature during processing of biaxially stretched polypropylene film. Polym. Eng. Sci. 2004, 44 (4), 805-813. [118] Aloisi G.G., Costantino U., Latterini L. et al. Preparation and spectroscopic characterization of intercalation products of clay and of clay-polypropylene composites with rhodamine B. J. Phys. Chem. Solids 2006, 67, 909-914. PhD INSA de Lyon (2008) 92 PROCESSING OF NANOCOMPOSITES Melt compounding [119] Finder C., Wohlgemuth M., Mayer C. Analysis of particle size distribution by particle tracking. Part. Part. Syst. Char. 2004, 21, 372-378. [120] Wu M., Roberts J.W., Buckley M. Three-dimensional fluorescent particle tracking at micron-scale using a single camera. Exp. Fluids 2005, 38, 461-465. [121] Baumann R., Ferrante C., Deeg F.W. et al. Solvation dynamics of nile blue in ethanol confined in porous sol–gel glasses. J. Chem. Phys. 2001, 114 (13), 5781-5791. [122] Maupin P.H., Gilman J.W., Harris R.H. (Jr.) et al. Optical probes for monitoring intercalation and exfoliation in melt-processed polymer nanocomposites. Macromol. Rapid Commun. 2004, 25, 788-792. [123] Bur A.J., Roth S.C., Start P.R. et al. Fluorescent probes for monitoring microstructure of polymer-clay nanocomposites. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2004, Vol. 1 (Processing), 1315-1318. [124] Gilman J.W., Maupin P.H., Harris R.H. (Jr.) et al. High throughput methods for nanocomposite materials research. Extrusion and visible optical probes. Polym. Mater. Sci. Eng. 2004, 90, 717-718. [125] Zhang XM, Xu ZB et al. Assessing local residence time distributions in screw extruders through a new in-line measurement instrument. Polym. Eng. Sci. 2006, 46, 510-519. [126] Bur A.J., Thomas C.L. Fluorescence monitoring of polymer injection molding: model development. Polym. Eng. Sci. 1997, 37 (9), 1430-1441. [127] Thomas C.L., Bur A.J. Optical monitoring of polypropylene injection molding. Polym. Eng. Sci. 1999, 39 (7), 1291-1302. [128] Poulesquen A., Vergnes B., Cassagnau P. et al. A study of residence time distribution in co-rotating twin-screw extruders. Part II: experimental validation. Polym. Eng. Sci. 2003, 43 (12), 1849-1862. [129] Poulesquen A., Vergnes B. A study of residence time distribution in co-rotating twinscrew extruders. Part I: theoretical modeling. Polym. Eng. Sci. 2003,43(12),1841-1848. [130] Del Pilar Noriega M. Osswald T.A., Ferrier N. In line measurement of the polymer melting behavior in single screw extruders. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2004, 154-161. [131] Gao F., Jin Z., Chen X. A visual barrel system for study of reciprocating screw injection molding. Polym. Eng. Sci. 2000, 40 (6), 1334-1343. [132] Alemaskin K., Manas-Zloczower I., Kaufman M. Color mixing in the metering zone of a single screw extruder: numerical simulations and experimental validation. Polym. Eng. Sci. 2005, 45, 1011-1020. Antonella ESPOSITO 93 Chapter I [133] Ing. L.D., Balke S.T., Stoev N. et al. In-line monitoring of particles in a polymer melt during extrusion using a scanning particle monitor. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2001, Vol.3 (Special Areas), 375-379. [134] Lutzen J., Van Veen K., Balke S.T. In-line monitoring of polymer additives during extrusion using a UV spectrometer. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2002, Vol. 3 (Special Areas), 1028-1031. [135] Kummer S., Müller J., Fischer D. In- and off-line characterization of nanocomposites by NIR spectroscopy, UV/VIS-spectroscopy in combination with microscopic methods. 4th international symposium on NAnostructured and FUNctional POLYmer-based materials and Nanocomposites (NANOFUN-POLY), 16-18 April 2008, Rome. [136] Bur A.J., Roth S.C., Lee Y.-H. et al. In-line monitoring of dielectric and fluorescence spectroscopy during polymer/filler compounding. Polymer Processing and Engineering Conference. July 2003. Bradford (UK). [137] Lee Y.-H., Bur A.J., Roth S.C. Correlation between degree of exfoliation, dielectric properties, and light transmission of nylon11/clay nanocomposites probed by an online dielectric slit die. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2004, Vol. 1 (Processing), 1279-1283. [138] Yilmaz Y., Alemdar A. Fluoro-surfactant as a tool for both controlling and measuring the size of the organoclay aggregates. Appl. Clay Sci. 2005, 30, 154-164. [139] Bur A.J., Roth S.C., Start P.R. et al. Monitoring clay exfoliation during polymer/clay compounding using fluorescence spectroscopy. Trans. Inst. Meas. Control 2007, 29 (5), 403-416. [140] McBrearty M., Bur A.J., Roth S.C. Variation of electrical properties with exfoliation condition in nanocomposites. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2002, Vol. 3 (Special Areas), 882-886. [141] Mighri F., Huneault M.A. Dispersion visualization of model fluids in a transparent Couette flow cell. J. Rheol. 2001, 45 (3), 783-797. [142] Hendricks S., Teller E. X-ray interference in partially ordered layer lattices. J. Chem. Phys. 1942, 10, 147-167. [143] Lee S.M., Park J.C., Lee S.M. et al. In-line measurement of residence time distribution in twin-screw extruder using non-destructive ultrasound. Korea-Australia Rheology J. 2005, 17 (2), 85-97. PhD INSA de Lyon (2008) 94 Chapter II PHOTO-FUNCTIONALIZATION Lamellar fillers A great portion of this chapter corresponds to the content of a paper recently submitted for publication. As previously highlighted (§ I-2), the addition of clays (hydrophilic by nature) to polymers (mostly hydrophobic) may greatly improve their properties, but in case the particles are perfectly disaggregated, dispersed and distributed into the matrix1. Cation exchange processes are commonly used to improve clay compatibility with polymers, thus facilitating a more intimate mixture of the fillers with the matrix. On-line monitoring of polymer nanocomposite processing is difficult – mainly because of filler reduced dimensions – and in the case of polymer/clay nanocomposites it is even more difficult because of clays multiscale structure2. A suitable, additional modification of clays may confer them peculiar fluorescence properties in order to monitor their exfoliation (hence dispersion) and distribution during processing. Moreover, fluorescent clays could be employed to trace pollution in soils and conceive optical devices, for the interactions of clays with cationic organic dyes may lead to the formation of controlled supramolecular assemblies. 1 2 Agglomeration, aggregation, dispersion and distribution have been introduced in § I-2.1. More details about clay multiscale structure are available in § I-3.1.1. Antonella ESPOSITO 95 Chapter II The main goal of this first part of the work was to find a suitable experimental protocol well adapted to confer peculiar emission properties to commercially available organically-modified clays, in the perspective of using such inorganic-organic opticallyactive complexes for real-time monitoring of polymer-clay nanocomposite morphology during processing. We firstly selected an organically-modified clay (Cloisite ® 30B) and we focused our attention on the photo-functionalization process: in the following chapter, all the steps we went through to find the most suitable protocol (which turned out to be cation exchange processing) and the optimization of the parameters to be used for such protocol (essentially the choice of the fluorescent molecule and its optimum concentration in the exchanging medium) are presented. Clay modification has been performed by several methods with three different fluorescent molecules, viz. by clay swelling, dry and melt compounding (9-anthracenemethanol) and by cation exchange processing (Rhodamine 6G Perchlorate, Nile Blue A Perchlorate). Modified clays have been washed and characterized (both before and after washing, when possible) by XRay Diffractometry (XRD), ThermoGravimetric Analysis (TGA), Elemental Analysis (EA), Fourier Transform InfraRed (FTIR) spectroscopy, TGA coupled to FTIR (TGAFTIR) spectroscopy and spectrofluorimetry. Accurate investigations of the molecular arrangement of the organic guests (fluorescent molecules) interacting with the inorganic host structure (clay platelets and galleries) are necessary to understand the mechanisms of the photo-functionalization process (whether the adsorption involves monomers, dimers or aggregates, and which kind of aggregates), to estimate the quality of the photo-functional inorganic-organic complexes (whether they’re photo-active or not and which kind of information can be deduced from fluorescence emission) and to profitably use them for nanocomposite process monitoring. II-1 MATERIALS MMT-MT2EtOH (Cloisite® 30B, Southern Clay Products, USA), a montmorillonite organically modified with a quaternary ammonium salt (MT2EtOH = methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium chloride), was purchased and used as received. The initial Modifier Concentration (MC) given by the supplier is PhD INSA de Lyon (2008) 96 PHOTO-FUNCTIONALIZATION Lamellar fillers 90 meq/100g of clay and tallow fatty chains have the following composition: 65% C18, 30% C16 and 5% C14. Its photo-functionalization has been performed in the presence of the following Fluorescent Molecules (FM): 9-anthracenemethanol, Nile Blue A Perchlorate, Rhodamine 6G Perchlorate (Sigma-Aldrich, France). Raw materials and the structure of the corresponding relevant chemicals, as well as their molecular weight Mw and their maximum dimension Dmax estimated by molecular modeling (Materials Studio – Accelrys Software Inc.), are listed in Table II-T1. Permuted water has been ion-exchanged right before using it, assuring a quality of 18 M -cm. Any other solvent involved in clay photo-functionalization and washing (benzyl alcohol, tetrahydrofurane and ethanol) was purchased from Sigma-Aldrich and used as received. Table II-T1 Raw materials and the corresponding relevant chemicals MW ACRONYM TRADE or IUPAC NAME SUPPLIER C30B Cloisite ® 30B MMT-MT2EtOH Southern Clay Products (USA) anth 9-anthracene methanol Sigma-Aldrich (France) 208.26 10 Å NBAP Nile Blue A Perchlorate Sigma-Aldrich (France) 417.84 15 Å RhP Rhodamine 6G Perchlorate Sigma-Aldrich (France) 543.01 14 Å CHEMICAL [g/mol] 360.80 Dmax a 29 Å average b stretched c a The maximum dimension of each chemical (D max) has been estimated by molecular modeling (Materials Studio v.4.1.0.0, Accelrys Software). b MT2EtOH includes tallow fatty chains (T) having the following composition: 65% C18, 30% C16, 5% C14. c A molecule bearing one or more long chains may assume different molecular conformations. To estimate MT2EtOH maximum dimension, the hypothesis of linearity of the longest tallow fatty chain has been done. Antonella ESPOSITO 97 Chapter II II-2 PHOTO-FUNCTIONALIZATION METHODS The aim of the photo-functionalization is to graft/adsorb a fluorescent molecule onto the clay silicate layers or at their edges, or to introduce it into the clay galleries. Several methods have been tested in order to photo-functionalize the selected clay: (A) clay swelling in a solution containing the fluorescent molecule; (B) compounding of the clay with the dry fluorescent molecule in a batch mixer; (C) compounding of the clay with the molten fluorescent molecule; (D) cation exchange processing in the presence of the fluorescent molecule. II-2.1 Clay swelling (A) C30B has been introduced in a benzyl alcohol solution containing an amount of 9-anthracenemethanol corresponding (in terms of moles) to twice the initial modifier concentration of the organically-modified clay (2MC). 9-anthracenemethanol is a nonionic molecule and talking about Cation Exchange Capacity (CEC) is actually inexact. Anyway, we dealt with it as if it was a monovalent cationic molecule just to provide a kind of criterion to estimate the amount of fluorescent molecule required for the photofunctionalization. The solution has been sonicated for 10 minutes by an ultrasounds probe and let swell at room temperature for at least 12 hours. As a reference, the same procedure has been done without the fluorescent molecule. This method is schematized in Figure II-F1. Figure II-F1 Schematic of the photo-functionalization method by clay swelling. This method has been performed exclusively with 9-anthracenemethanol. PhD INSA de Lyon (2008) 98 PHOTO-FUNCTIONALIZATION II-2.2 Lamellar fillers Dry compounding (B) C30B and an amount of 9-anthracenemethanol corresponding (in terms of moles) to twice the modifier concentration of the organically-modified clay (2MC) have been introduced in a batch mixer (Haake Rheomix, PolyLab System) and let mix at 40 rpm for 5 to 15 min at 160°C to 200°C – the high temperature being supposed to locally melt the fluorescent molecule and thus help mixing. Once again, we decided to consider 9anthracenemethanol as if it was a monovalent cationic molecule just to provide a kind of criterion to estimate the amount of FM required for the photo-functionalization. This method is schematized in Figure II-F2. Figure II-F2 Schematic of the photo-functionalization method by dry compounding. This method has been performed exclusively with 9-anthracenemethanol. II-2.3 Melt compounding (C) A considerable excess of 9-anthracenemethanol has been melted (m.p.160°C) in an oil bath at 180°C and, while stirring, C30B has been added to the liquid phase in a weight ratio anth/C30B between 1 and 3. Samples have been repeatedly washed by THF and recovered by centrifugation (10 min at 3500 rpm) till a colorless supernatant was obtained (meaning that no more fluorescent molecule could be extracted by washing), then let dry under exhaust hood for several days at room temperature. This method is schematized in Figure II-F3. Antonella ESPOSITO 99 Chapter II Figure II-F3 Schematic of the photo-functionalization method by melt compounding, followed by washing. This method has been performed exclusively with 9-anthracenemethanol. II-2.4 Cation exchange processing (D) C30B has been cation-exchanged with two cationic organic dyes (Nile Blue A Perchlorate and Rhodamine 6G Perchlorate), introduced in two different concentrations (corresponding to 1 and 0.25 times the initial modifier concentration of the organicallymodified clay, denoted 1MC and 0.25MC respectively) in a solution 90/10 permuted water/ethanol. The use of permuted water is always worthy to be preferred in order to get rid of any undesired cation. The correct amount of ethanol to be introduced in the exchanging medium has been determined as the smallest amount of co-solvent necessary to completely dissolve the fluorescent molecule, i.e. in order to get a saturated solution. First, the permuted water has been warmed up to 80°C. Then, the organicallymodified clay has been added to the stirring hot water and let disperse for at least 15 minutes. Stirring speed has been adjusted to the value at which a vortex just started to form, which seems to be the optimal condition to get a homogeneous mixing [1]. Separately, the fluorescent molecule has been dissolved in the required amount of ethanol and the obtained solution has been added to the clay suspension. Once started, the cation exchange process has been performed under mechanical stirring for 24 hours. After additional 24 hours in the still exchanging medium, the photo-functionalized clay has been separated from the liquid by centrifugation (20 min at 4000 rpm), washed with ethanol and recovered anew by centrifugation. The described washing procedure has been repeated until a sufficiently colorless supernatant was obtained (meaning that no more fluorescent molecule could be extracted by washing), then let dry under exhaust hood for several days at room temperature. The method is schematized in Figure II-F4. PhD INSA de Lyon (2008) 100 PHOTO-FUNCTIONALIZATION Lamellar fillers Figure II-F4 Schematic of the photo-functionalization method by cation exchange processing, followed by washing. This method has been performed with both NBA (top) and RhP (bottom). II-3 CHARACTERIZATIONS Characterizations were mainly aimed to verify whether the fluorescent molecules adsorbed or not on clay surfaces, and where they are located, as well as to check the efficacy of the washing procedure – supposed to be selective, i.e. capable of washing out the excess of dye without removing the chemicals cation-exchanged into clay galleries. In addition, characterizations to test the photo-activity of the samples have also been performed (spectrofluorimetry). Antonella ESPOSITO 101 Chapter II II-3.1 X-Ray Diffractometry (XRD) X-Ray Diffractometry (XRD) is widely used to quantify the interlayer spacing of crystalline structures, thus it is well appropriate to assess whether the fluorescent molecules have made clay swell or not by diffusing into their galleries. In order to measure the d001 spacing of the pristine and photo-functionalized clays, the value of 2 angle must vary between 1 and 10°: the d001 distance is then referred as a peak on the intensity vs. 2 plot. A shift of the peak to lower values of 2 indicates an increase of clay interlayer spacing ( d001>0) corresponding to the intercalation of the organic moieties into clay galleries. Wide-Angle X-ray Diffractometry (WAXD) has been performed at room temperature in the range 2 = 0.1-10° and with a scan rate of 0.5° per minute on a Bruker D8 Advance diffractometer with a goniometer having a Bragg-Brentano geometry in – configuration, with a 500 mm focalisation diameter and a cupper anode to generate X rays (33 kV, 45 mA, = 1.54 Å). Ruben Vera (Centre de Diffractométrie Henri Longchambon, Université Claude Bernard, Lyon-1) is kindly acknowledged for having performed XRD measurements. II-3.2 ThermoGravimetric Analysis (TGA) In spite of its high sensitivity to the dimension of clay galleries (and, thus, to the presence of chemical species confined in between the silicate layers), XRD isn’t able to detect any molecule which is adsorbed/grafted onto the external surface of the silicate layers or at their edges. ThermoGravimetric Analysis (TGA) measurements may help completing clay characterizations by investigating the effects of functionalization and washing on the thermal stability and the degradation mechanisms of the samples compared to their references, i.e. the pristine commercial clay and the pure fluorescent molecule. It has been observed [2][3] that peaks corresponding to weight losses between 150°C and 250°C (up to 300°C) typically indicate the presence of organics which could PhD INSA de Lyon (2008) 102 PHOTO-FUNCTIONALIZATION Lamellar fillers be simply physisorbed (or eventually interacting by hydrogen bonding) either on the external surfaces of the silicate layers or at their edges. Such organic fraction either didn’t undergo the cation exchange process, or stayed unconfined, or maybe entered the galleries but remained in a peripheral position. On the other hand, weight losses in between 300°C and 550°C rather correspond to organics intercalated into clay galleries [3-6]. Indeed, organics inside clay galleries are better insulated and thus protected from thermal degradation; in addition, the eventual products of an initial thermal decomposition are slowed down from diffusing to the gas phase by the presence of the silicate layers (barrier effect) [5]. A suited washing procedure is expected to reduce the peaks corresponding to the excess of fluorescent molecule and to any chemical physisorbed at the edges or on the external surfaces of the silicate layers, still preserving the peaks which denote the chemicals intercalated into clay galleries [3]. TGA has been accomplished by performing continuous ramp experiments from room temperature up to 580°C with a heating rate of 10°C/min, in aluminium pans, under nitrogen atmosphere and with a gas flux of 90 mL/min (Q500 Thermogravimetric Analyzer, TA Instruments). II-3.3 Elemental Analysis (EA) Along with TGA, Elemental Analysis (EA) completes (but doesn’t exhaust) the set of characterizations which can be made on a functionalized clay. Indeed, this technique allows estimating the relative amount of any specific element (but the oxygen) present in the sample. By performing EA of each sample before and after washing, and by comparing these results to the ones obtained for the raw materials (essentially the pristine C30B clay), it is possible to evaluate the effects of the photofunctionalization as well as those of the washing step. The limiting factor is that such a technique gives only relative elemental compositions, meaning that the amount of any detected element has to be “normalized” to be interpreted. By burning some milligrams of the sample at 1050°C in the presence of He and O2, the main elements detected by Antonella ESPOSITO 103 Chapter II the CHN analyzer are converted to CO2, H2O and NxOy respectively. Subsequently, the nitrogen oxide is reduced to N2 on cupper at 450°C and all the products are separated by column chromatography. EA of the raw materials and of the samples has been performed giving to the elements to be analyzed the following priority: Na, C, N, H and Si. We chose Si as the reference, since it is the only main constituent of clay platelets (the other element, the oxygen, is not detectable by a CHN analyzer) which is not supposed to vary in consequence of a cation exchange process and/or of a washing procedure. EA measurements have been carried out on a CHN analyzer having the following khatarometer detection limits: 0.10 % N, 0.30 % C and 0.30 % H (uncertainty ± 0.30 %). Normalization has been done in relation to the Si % content detected in pristine C30B. We acknowledge the technical staff at the Service Central d’Analyse CNRS in Solaize for EA characterizations. II-3.4 Fourier Transform InfraRed (FTIR) spectroscopy Fourier Transform InfraRed (FTIR) spectroscopy provides information about the chemical bonds or the molecular structure of a material, whether organic or inorganic. By submitting the sample to an infrared beam in a given range of wave number values, bonding and molecular vibrations are excited at their characteristic frequencies and their spectrum of absorbed or emitted vibrational energies is collected [7][8]. Such a spectrum is sensitive to the nature of the atoms involved in the bond, the nature of the bond itself (single or double covalent bond, hydrogen bond…), the molecular conformation (C=C cis or trans…). By comparing the FTIR fingerprint of each sample to the one obtained for the pristine C30B clay, it is possible to detect the presence of any additional chemical and, eventually, its surroundings (i.e. whether it is adsorbed onto silicate layers or is present as a free excess). PhD INSA de Lyon (2008) 104 PHOTO-FUNCTIONALIZATION Lamellar fillers FTIR spectra have been obtained at room temperature on a Nicolet Magna-IR 550 spectrometer on pellets obtained by mixing the sample with dried KBr (2% wt ca) and pressing the mixture by a hydraulic press (10 tonnes ca). Right after, pellets have been rapidly mounted in a stainless steel disk holder for mid-infrared measurements and spectra have been collected in the wavelength range 400-4000 cm-1 at a resolution of 4 cm-1 using 50 signal-averaged scans. II-3.5 TGA coupled to FTIR spectroscopy (TGA-FTIR) TGA coupled to FTIR spectroscopy (TGA-FTIR) is particularly useful to probe and understand the thermal degradation pathways of organic samples [9], which usually involve the release of several volatile products. This technique greatly helps identifying the combustion products emitted while performing conventional TGA characterizations, leading to a better assignment of the degradation peaks to a specific chemical present into the sample. There are several ways in which the TGA-FTIR data may be displayed: (1) a tabular display of the mass losses and the gases which are observed in each region; (2) the display of individual spectra at various temperatures; (3) the display of spectra at several temperatures in a stacked plot; (4) the display of the infrared absorbance for some peak(s) overlaid on the TGA curve [9]. Some authors [10] may prefer to present the TGA-FTIR results in the form of a Gram-Schmidt plot, which shows information related to the total IR absorbance of the evolved gases in the whole spectral range. TGA-FTIR has already been used to study the thermal decomposition products evolved during the degradation of several commercially available organoclays by Cervantes-Uc et al. [10]. TGA-FTIR characterizations have been performed on a TG 209 F1 Iris coupled to a Tensor 27 BRUKER, with an heating rate of 10°C/min, under a flux of synthetic air (20 M% O2, 80 M% N2, H2O < 3ppm) of 80 mL/min. Spectra have been collected in the wavelength range 600-4400 cm-1 at a resolution of 4 cm-1 using 5 signal-averaged scans. Antonella ESPOSITO 105 Chapter II Pascal Rumeau and the technical staff at the Institut Français du Textile et de l’Habillement IFTH (Ecully) are kindly acknowledged for TGA-FTIR measurements. II-3.6 Spectrofluorimetry Fluorescence is a three-stage process that generally occurs in polyaromatic hydrocarbons or heterocyclic molecules called fluorophores or fluorescent dyes. In the first stage, a photon of a given energy is supplied by an external source (e.g. an incandescent lamp or a laser) and absorbed by the fluorescent molecule, creating an excited electronic singlet state. This excited state exists for a finite time, during which the molecule undergoes conformational changes and is subject to several interactions with its molecular environment. Such processes partially dissipate the energy of the excited state, yielding a relaxed singlet excited state and thus producing a fluorescence emission. Finally, a photon is emitted returning the molecule to its ground state. Due to the energy dissipations in the excited-state lifetime, the emitted photon has a lower energy (longer wavelength) than the absorbed one: the difference in energy (or wavelength) of the two photons is called the Stokes shift. The sensitivity of any fluorescence technique strictly depends on this fundamental parameter, which allows emission photons to be detected against a low background, clearly isolated from excitation photons. In theory, fluorescence is a cyclical process: unless the fluorophore is irreversibly destroyed in the excited state (phenomenon known as photobleaching), it can be repeatedly excited and detected. When performing spectrofluorimetry characterizations, it is first necessary to collect the fluorescence excitation spectrum of the sample and to identify its maximum: although the fluorescence emission spectrum doesn’t depend on the excitation wavelength, the best sensitivity is obtained when the excitation source is at its highest efficiency with respect to a given sample. Under the same conditions, the fluorescence emission spectrum of an isolated fluorophore (monomer) in solution is supposed to be identical in shape to its fluorescence absorption spectrum – the only difference being the Stoke shift. Nevertheless, it may happen that a chemical and/or physical change in the fluorophore environment modifies the shape of its fluorescent emission spectrum. PhD INSA de Lyon (2008) 106 PHOTO-FUNCTIONALIZATION Lamellar fillers Emission spectra are generally more sensitive to any change in the molecular environment than absorption spectra, but the latter ones may be helpful to characterize the configuration and the supramolecular arrangement of the fluorescent dyes. Indeed, fluorophores can assemble in dimers and/or higher-order aggregates: any possible configuration produces a well-defined absorption spectrum and doesn’t assure the same fluorescence emission. J-aggregates (head-to-tail molecular arrangements) typically exhibit a narrow absorption peak (the so-called J-band) which is red-shifted with respect to the monomer band (bathochromic shift), whereas Haggregates (tail-to-tail molecular arrangements) show a broader absorption peak (Hband) which is rather blue-shifted with respect to the monomer band (hypsochromic shift). On the basis of the exciton theory, J-aggregates are fluorescent whereas Haggregates are affected by self-quenching. Interactions, either between two fluorophores or between a fluorophore and the other species in the surrounding environment, can produce environment-sensitive fluorescence emission. Fluorescence quenching is a bimolecular process which reduces (and sometime zeroes) the fluorescence signal intensity without reshaping the emission spectrum: it can result from transient excited-state interactions (collisional quenching) or from the formation of non-fluorescent ground-state species (i.e. H-aggregates). Quenching tends to occur when high fluorophore concentrations are used. Clearly, spectrofluorimetry is an essential tool to check the result of a photo-functionalization process. We proceeded as following: (1) we measured the fluorescence excitation and fluorescence emission spectra to identify the optimum excitation wavelength and estimate the Stoke shift; (2) we recorded several fluorescence emission spectra as a function of different concentrations of fluorophore or photo-functionalized clay in order to determine the value of concentration at which fluorescence quenching is observed; (3) we finally established a calibration curve (maximum fluorescence emission vs. concentration) in non-quenching conditions for each sample. Antonella ESPOSITO 107 Chapter II Spectrofluorimetry has been performed at room temperature on a steady-state FS920 spectrofluorimeter (Edinburgh Instruments, UK) with a high spectral resolution (signal to noise ratio > 6000:1), using ethanol as solvent (for the fluorescent molecules) or suspension medium (for the pristine C30B and the photo-functionalized clays). The spectrofluorimeter covers the wavelength range from 200 to 1670 nm using two detectors: a photomultiplier R928 for UV-Vis scans (up to 870 nm) and a solid InGas TE G8605-23 detector for IR scans. The excitation source is a continuous Xenon Arc lamp (450W) coupled to two Czerny-Turner DMX300X 1800tr/mn monochromators, one for UV excitation (focal length 300 nm) and one for the visible wavelength range (focal length 500 nm). Olivier Raccurt and the head of the Laboratoire des Technologies des Traceurs L2T (Département des Technologies des Nanomatériaux, Commissariat à l’Energie Atomique CEA, Grenoble) are particularly acknowledged for the material and technical support, which allowed us to get spectrofluorimetry characterizations of the samples. II-4 REFERENCE MEASUREMENTS First of all, the fluorescent molecules of interest as well as the pristine C30B clay have been characterized by XRD, TGA, EA, FTIR spectroscopy and spectrofluorimetry in order to get reference measurements. II-4.1 Reference XRD measurements XRD shows that 9-anthracenemethanol has a crystalline peak corresponding to 14.5 Å, which considerably diminishes after melting followed by spontaneous cooling at room temperature, as shown in Figure II-F5a. NBAP has a small peak at 10.2 Å (Figure II-F5b), RhP has two marked peaks at 9.8 Å and 12 Å (Figure II-F5c). None of these peaks could affect the interpretation of XRD measurements on the photo-functionalized clays. C30B has an initial interlayer space d001=17.5 Å (Figure II-F5d). PhD INSA de Lyon (2008) 108 PHOTO-FUNCTIONALIZATION Lamellar fillers Figure II-F5 Reference XRD measurements: (a) 9-anthracenemethanol as received and after melting at 180°C followed by cooling down to room temperature, (b) Nile Blue A Perchlorate, (c) Rhodamine 6G Perchlorate, (d) pristine Cloisite ® 30B. Considering that the maximum molecular dimension Dmax estimated under the hypothesis of linearity of the longest tallow fatty chain (stretched molecule) is 29 Å (Table II-T1), this d001 value is consistent with a paraffinic configuration of MT2EtOH molecules into clay galleries with an angle around 37°, which is actually close to the values previously proposed in the literature [11]. II-4.2 Reference TGA measurements 9-anthracenemethanol undergoes thermal degradation in two steps (240°C and 317°C) and is completely degraded starting from 350°C (Figure II-F6a). NBAP loses more than 30% of its initial weight in the temperature range 200-400°C, following a Antonella ESPOSITO 109 Chapter II complex three-steps mechanisms (higher peak at 261°C followed by lower peaks at 303°C and 341°C) (Figure II-F6b). RhP halves its weight in the temperature range 200500°C: this weight loss mostly corresponds to a single sharp peak at 338°C, as shown in Figure II-F6c. Figure II-F6 Reference TGA measurements: (a) 9-anthracenemethanol as received and after melting at 180°C followed by cooling down to room temperature, (b) Nile Blue A Perchlorate, (c) Rhodamine 6G Perchlorate and (d) Cloisite® 30B in its pristine state and after washing with ethanol (recover by centrifugation 20 min at 4000 rpm). C30B (Figure II-F6d) undergoes two weight losses between 150°C and 500°C. The first weight loss (21% ca at 253°C) corresponds to a partial physisorption of the MT2EtOH molecules at the edges or on the external surface of the platelets, since the peak regularly lowers after washing. The fact that it doesn’t completely disappear means that a portion of MT2EtOH is well intercalated but in a peripheral position with respect to the clay gallery, as reported by Davis et al. [6]: such portion of surfactant cannot be washed away (since it underwent cation exchange) but it is not thermally PhD INSA de Lyon (2008) 110 PHOTO-FUNCTIONALIZATION Lamellar fillers stabilized by the presence of the inorganic silicate platelets, thus it degrades at the same temperature as the physisorbed surfactant. The second weight loss (30% at 360°C) corresponds to the intercalation of the MT2EtOH molecules into clay galleries (this perfectly corresponds to the % weight loss on ignition given by the supplier). The organic fraction for each step of weight loss has been evaluated by identifying, isolating and integrating the main peaks of the TGA derivative curves, calculating their area and deducing the weight percentage as the percentage of the corresponding area, referred to the total amount of weight loss up to 550°C. We haven’t characterized the thermal behaviour of the clay samples for T>580°C because we were solely interested in the interactions of the cationic organic dyes with the clay and no organic can resist to temperatures higher than 600°C. In addition, it has been already shown (and we directly checked) that the next step of the thermal degradation of natural and organicallymodified clays would have been the dehydroxilation of the aluminosilicate framework [2][4][5] starting after 500°C and corresponding to a peak centred around 645°C. The experimental modifier concentration for pristine C30B clay (83 meq/100g) has been calculated on the basis of the organic fraction corresponding to the loss of intercalated chemicals, since this is the only fraction of surfactant which has surely replaced the inorganic cations in clay galleries. The theoretical CEC for each cationic fluorescent molecule (239 meq/100g of NBAP and 184 meq/100g of RhP) has been calculated with relation to 100g, since each single molecule (and thus the whole amount of organic cationic dye) is supposed to be able to replace clay inorganic cations. On the contrary, 9-anthracenemethanol has a zero CEC since it’s a neutral molecule. II-4.3 Reference EA measurements Reference EA values are resumed in Table II-T2. The Si % content detected in the pristine C30B has been chosen as the reference value to get the Normalization Factor (NF) for all the other measurements performed on the photo-functionalized clay samples, whether washed or not. Antonella ESPOSITO 111 Chapter II Table II-T2 Reference EA measurements: chemical composition of the raw materials Na [% wt] theor C30B MT2EtOH anth NBAP RhP c n.a. 0.00 0.00 0.00 0.00 exper 0.23 n.a. n.a. 0.00 0.00 C [% wt] theor c n.a. 73.80 86.43 60.31 61.90 exper 20.18 n.a. n.a. 57.40 b 61.90 b N [% wt] theor c n.a. 3.88 0.00 10.00 5.16 exper 1.07 n.a. n.a. 10.05 b 5.20 b H [% wt] c exper n.a. 13.41 5.76 4.79 5.71 3.98 n.a. n.a. n.a. n.a. theor Si [% wt] exper norm a 21.39 Ref n.a. n.a. 0.00 0.00 - a a The amount of any element estimated by EA is relative and has to be normalized. In the following, the Si % content detected experimentally in the pristine C30B clay will be considered the reference value for the normalization of any other EA measurement. b Source: Sigma-Aldrich. c When possible, the theoretical values of elemental composition have been calculated on the basis of the chemical formula. II-4.4 Reference FTIR spectra Neither 9-anthracenemethanol nor NBAP have been characterized by FTIR spectroscopy – mainly because of the mechanisms of the chosen photo-functionalization protocol but also for the selected experimental conditions (further details and the reason of such a choice are given in the following)3. As expected, the FTIR spectrum of RhP (Figure II-F7a) exhibits, in the region 3150-2700 cm-1, some bands characteristic of carbon- and hydrogen-containing species, assigned to various forms of C-H stretching. It also shows a strong but quite large band above 3000 cm-1 (maximum at 3360 cm-1 ca), which is typical of unsaturated and/or aromatic compounds. Amino groups are also used to dominate the region 3650-3250 cm-1 with a broad band of absorption: the presence of a secondary amino group can also contribute to this portion of the spectrum. The sharp peak in the region 1800-1690 cm-1 may be assigned to the presence of a C=O (carbonyl group), but the resonant C=N bond could also interfere. The narrow peak centred at 1650 cm-1 is indicative of some unsaturation – the absorption intensity being intensified 3 At the end of the chapter we’ll realize that the most suitable photo-functionalization method for clays is the one based on cation exchange processing in the presence of a cationic organic dye (§ II-2.4). In particular, we’ll show that RhP looks to be more efficient than NBAP (§ II-6.1). PhD INSA de Lyon (2008) 112 PHOTO-FUNCTIONALIZATION Lamellar fillers by the conjugation with the aromatic rings. This observation is further confirmed by the presence of a pair of weak peaks at 3095 and 3035 cm-1. The two asymmetric doublets in the region 1615-1495 cm-1 (the first around 1600 cm-1 and the second around 1500 cm-1) are consistent with an aromatic compound and are confirmed by the presence of the weak absorption bands in the region 3130-3015 cm-1 (aromatic C-H stretching) and 850-670 cm-1 (C-H out-of-plane bending of aromatic rings). The strong double peaks in the region 1155-950 cm-1 correspond to the typical in-plane C-H bending vibrations of aromatic compounds. The double peak in the region 1350-1200 cm-1, as well as the following peak around 1190 cm-1, correspond to the C-O stretching. Finally, the hidden band corresponding to the region 3300-3030 cm-1, confirmed by the weak hidden bands between 1430 and 1390 cm-1, are correlated to the fact that the molecule contains an ammonium ion. Figure II-F7 Reference FTIR spectra: (a) Rhodamine 6G Perchlorate and (b) pristine Cloisite ® 30B in KBr pellets. Antonella ESPOSITO 113 Chapter II The FTIR spectrum of C30B (Figure II-F7b) exhibits a broad absorption band in the region 3750-3050 cm-1, certainly related to the fact that the clay is organicallymodified (C-H and O-H stretching of the surfactant molecules) but mostly because of the intrinsic structure of clays (MMTs containing many hydroxyl groups and H-bonded OH stretching are usually characterized by broad absorption bands in the region 35703200 cm-1). Moreover the presence of hydrogen bonds typically produces a further significant band broadening: the overlapping of asymmetric and symmetric (H-O-H) stretching vibrations of H-bonded water is always significant in smectites and any other expandable clay. The narrower peak centred around 3630 cm-1 may be explained by OH stretching, as well: in particular, if relatively sharp features occur between 3670 and 3550 cm-1, the compound is likely to contain non-H-bonded OH groups, and it often corresponds to an alcohol with a sterically-hindered OH group. Anyway, such a feature is also exhibited by certain inorganics and minerals as an indication of the presence of “free” OH groups, either on the surface or embedded within the crystal lattice. The absorption of water is partly hidden under the structural OH stretching band, as well. The set of two peaks in the region 3000-2800 cm-1 (approximately 2930 cm-1 and 2860 cm-1) is typical of compounds containing long linear aliphatic chains, and MT2EtOH molecules actually contain a tallow chain. This is confirmed by the sharp but weak peak at 1471 cm-1. The weak broad peak at 1645 cm-1 could have been due to the fact that MT2EtOH is a nitro organic compound and C30B surely still contain an excess of the salt used in the commercial functionalization; indeed, Madejová [30] showed that this peak actually corresponds to the (H-O-H) bending vibrations of adsorbed water molecules. The strong peak at 1050 cm-1 is explained by the chemical composition of MMT, basically a crystalline silicate (Si-O-Si vibrations). The triple set of peaks in the region 670-400 cm-1 has probably to be assigned to all the possible vibrational modes in the inorganic crystalline lattice. PhD INSA de Lyon (2008) 114 PHOTO-FUNCTIONALIZATION II-4.5 Lamellar fillers Reference fluorescence spectra 9-anthracenemethanol hasn’t been characterized by spectrofluorimetry since it is a neutral molecule and couldn’t undergo cation exchange with inorganic cations: thus, it doesn’t present any interest in relation to the selected photo-functionalization method4. By recording the absorption spectrum of NBAP, we found an optimal excitation wavelength of 627 nm, which is in agreement with the value given by the supplier (λmax = 628 nm). The fluorescence emission spectrum at the optimal excitation wavelength for NBAP in ethanol is shown in Figure II-F8a. Figure II-F8 Reference spectrofluorimetry measurements: absorption and emission spectra of (a) Nile Blue A Perchlorate and (b) Rhodamine 6G Perchlorate in ethanol. It is well known that the maximal fluorescence absorption for RhP is around 532 nm: the fluorescence emission spectrum for RhP in ethanol is shown in Figure II-F8b. RhP molecules have a strong tendency to aggregate: the absorption peak around 490 nm corresponds to the hypsochromic effect due to the formation of non-fluorescent RhP 4 At the end of the chapter we’ll realize that the most suitable photo-functionalization method for clays is the one based on cation exchange processing in the presence of a cationic organic dye (§ II-2.4). Antonella ESPOSITO 115 Chapter II H-aggregates. As expected, pristine C30B doesn’t significantly absorb radiations and, thus, it doesn’t show any fluorescence emission (result not shown). II-5 EVALUATION OF THE PERFORMED PHOTO-FUNCTIONALIZATION METHODS AND CHOICE OF THE PROTOCOL Since diffusion may be a critical phenomenon for clay functionalization (its control can be accomplished by adjusting processing temperature [3]), we firstly tried method (A)5 to exploit the well-known swelling mechanism of clays in a liquid phase – which, exceptionally in this case, should also be a good solvent for the chosen fluorescent molecule. In a previous work [2], the dispersion of organophilic clays in organic solvents has been evaluated by studying the interactions between clay platelets and solvent molecules at different scales: it has been shown that the best results in terms of basal spacing increase for a MMT-MT2EtOH are obtained by free swelling in butyldiglycol, dimethylformamide and benzyl alcohol. Benzyl alcohol is proved to produce the best results at the nanoscale level with the minimum macroscopic free swelling (i.e. the minimum solvent uptake), and that’s the reason why we chose benzyl alcohol for method (A). Figure II-F9 confirms that benzyl alcohol is a good swelling agent for C30B ( d001 = +21.1 Å) as previously found [2] and shows that the presence of a fluorescent molecule may even improve the efficacy of swelling mechanism, as indicated by the raise of the diffraction peak at 38.6 Å. In fact, swelling occurs when the surface tension of the solvent is higher than the surface energy of the clay ( Burgentzlé et al. [2] reported that L benzyl alcohol L solvent is 39.0 mN/m and S clay): S C30B indeed, is 34.5 ± 2.0 mJ/m2. Whether in the presence or in the absence of the fluorescent molecule, swelling resulted in a homogeneous but difficultly dryable gel. Even if some of the drying procedures we tested seemed to be more efficacious than others, method (A) regularly caused troubles with the complete evaporation of the residual solvent, mostly because of the high boiling point of benzyl alcohol (b.p.205°C). 5 Photo-functionalization by clay swelling (§ II-2.1). PhD INSA de Lyon (2008) 116 PHOTO-FUNCTIONALIZATION Lamellar fillers Figure II-F9 XRD evaluation of the results obtained with method (A), with and without 9-anthracenemethanol. The problems encountered with solvent evaporation pushed us to look for an alternative procedure to be performed without the use of any solvent. Method (B)6 clearly proved the undeniable practical advantage of avoiding solvents, notwithstanding it appeared to be unsuitable to our purposes because molecules can’t really diffuse in the absence of a liquid phase. In addition, a predictable mixing inefficiency (due to the fact that dry phases can’t really sustain and transfer shear) caused local increases of temperature and thus a visible degradation of the fluorescent molecule. Method (C)7 was designed to ameliorate both methods (A) and (B) by taking advantage of diffusion/swelling mechanisms but still avoiding any solvent, for which another source of liquid phase was to be found: the melting of the fluorescent molecule. Figure II-F10a shows that the same good swelling results of method (A) can also be obtained by method (C) (d001 = 40 Å, i.e. d001 = +22.5 Å). Although this procedure generally resulted in homogeneous gels (which, once cooled down to room temperature, could be easily reduced to powder), the need for a liquid phase composed of the sole molten dye implied the presence of a considerable amount of fluorescent molecule (Figure II-F10a, diffraction peak at 2 = 6°): we had to find a suitable procedure to selectively wash out the excess. 6 7 Photo-functionalization by dry compounding (§ II-2.2). Photo-functionalization by melt compounding (§ II-2.3). Antonella ESPOSITO 117 Chapter II Figure II-F10 Evaluation of the results obtained with method (C). XRD characterizations (a) of clay swelling after melt compounding (the excess of 9-anthracenemethanol is confirmed by the peak at 2 = 6°) followed by washing with THF and acetone, respectively. These results are substantiated by TGA characterizations of the clay after melt compounding (b) followed by washing with THF (c) and acetone (d), respectively. All the washing procedures we tested have already been previously used to recover clays from a solution [12-17] but centrifugation appeared to be the simplest, fastest and most efficient technique to separate a solid residue from a liquid phase [1][18]. The selection of an appropriate solvent to wash photo-functionalized clay has been done on the basis of its boiling point and of the results of a preliminary solubility test of 9-anthracenemethanol in acetone, methanol, toluene, dichloromethane and tetrahydrofurane (THF). We finally chose acetone (b.p.56°C) and THF (b.p.60°C) and validated our choice by verifying by XRD that these solvents don’t swell the pristine C30B on rapid immersion (results not shown). We voluntarily avoided drying photofunctionalized clay by oven in order to better preserve the fluorescent molecules. Unluckily, XRD and TGA show that the washing procedure resulted in a non-selective PhD INSA de Lyon (2008) 118 PHOTO-FUNCTIONALIZATION Lamellar fillers extraction of the fluorescent molecule (Figure II-F10a): both THF and acetone caused a significant shift of the diffraction peak towards lower values of the interlayer space (20 and 18 Å, i.e. d001 = –20 and –22 Å respectively), nearly back to the d001 value of the pristine C30B clay (Figure II-F5d). This conclusion has been substantiated by TGA measurements (Figures II-F10 (b), (c) and (d)): whatever the solvent used for washing, the derivative curve of the weight loss shows two peaks (273°C and 403°C ca), the second one being shifted to higher temperatures in comparison to the pristine C30B peak (360°C) probably exclusively as a result of some molecular rearrangement of the surfactant into clay galleries during the heating step. 9-anthracenemethanol, which is a non-ionic small molecule (Dmax = 10 Å, Table II-T1), could have penetrated the network of the bigger surfactant molecules (Dmax = 29 Å, see Table II-T1) already present into clay galleries, and eventually be trapped inside it by diffusion hindering or H-bonding (both 9-anthracenemethanol and MT2EtOH contain hydroxyl groups), but the results obtained after washing definitely show that these mechanisms cannot assure a stable photo-functionalization. Method (D)8 proceeds from the cation exchange process traditionally employed to characterize [19] and functionalize clays [20][21], and which is still quite popular to render clays organophilic [7]. More precisely, it has been inspired by some recent works of Maupin, Bur, Gilman et al., a group of researchers at the National Institute of Standards and Technology at Gaithersburg, Maryland, USA [22-24]. As ion exchange is a reversible process, clay functionalization is generally performed in the presence of an excess of surfactant – typically twice the value of the cation exchange capacity of the clay (2 CEC) – in order to force the reaction in the desired direction. However, when the adsorbate is a photo-active chemical and the ion exchange is aimed to the photofunctionalization of the adsorbent, a stricter control on the concentration of surfactant is required in order to avoid saturation (i.e. fluorescence quenching). Moreover, C30B is an organically-modified clay which has already undergone a first cation exchange process: we should better talk of modifier concentration (or residual cation exchange capacity) instead of cation exchange capacity tout court. This means that most of the 8 Photo-functionalization by cation exchange processing (§ II-2.4). Antonella ESPOSITO 119 Chapter II Na+ sites originally available are likely to be no longer exchangeable. On the other hand, it is highly probable that a certain (significant) amount of such sites (corresponding to the residual CEC) is still represented by inorganic cations, for the MT2EtOH molecules are surely big enough to be an obstacle to each other because of sterical hindrance. Indeed, high interlayer packing densities require the aliphatic tails to overlap, causing their inclination with respect to the silicate surfaces (typical angle 37°) [13]. Since the residual value of CEC for C30B clay is less than the original value for the inorganic clay, the introduction of an amount of fluorescent molecule equivalent to the residual valued of CEC into the exchanging medium would be already enough to be in the presence of an excess: we decided, anyway, to perform the first trials of cation exchange by method (D) with an amount of fluorescent molecule corresponding to a value of CEC equal to the modifier (MT2EtOH) concentration (1MC, i.e. 90 meq/100g as indicated by the supplier). We employed method (D) to perform the photofunctionalization of C30B with two organic cationic dyes (NBAP, RhP) having roughly the same maximum molecular dimension Dmax, slightly higher than the corresponding value for 9-anthracenemethanol (15 and 14 Å compared to 10 Å, Table II-T1). If we take a look at the results obtained by spectrofluorimetry on the washed photo-functionalized samples, we realize that the photo-functionalization was successful with both the fluorescent molecules (Figure II-F11). In terms of absorption: NBAP-clay absorbs in the same range as the raw cationic dyes (no significant metachromasy) while RhP-clay, as expected, shows a slight metachromatic effect (the main absorption band is blue-shifted, towards higher energies). Vice versa, in terms of emission: it is NBAPclay emission spectrum which shows a slight shift to higher energies in comparison to raw NBAP, while RhP-clay emission spectrum is absolutely the same as the raw RhP. Both NBAP-clay and RhP-clay show a well pronounced photo-activity. RhP-clay differs from NBAP-clay also because in its absorption spectrum there’s a peak likely corresponding to the formation of J-aggregates ( 548 nm), which are fluorescent according to the theory of exciton splitting and which are probably due to a regular organization of the intercalated dye induced by the presence of a regular array of MT2EtOH molecules. The presence of a peak probably corresponding to the formation PhD INSA de Lyon (2008) 120 PHOTO-FUNCTIONALIZATION Lamellar fillers of non-fluorescent H-aggregates ( 490 nm, see Figure II-F8) is justified by the strong tendency of RhP molecules to aggregate – phenomenon which has been reported in the literature [17] and clearly doesn’t concern NBAP (unless at higher concentrations). A fluorescence absorption spectrum which indicates the presence of monomers and Jaggregates (dimers) into RhP-clay galleries is in perfect agreement with the results of molecular modelling recently obtained by Čapková et al. [17]. In the following9 we’ll show that XRD characterizations of C30B-RhP photo-functional complexes are also in good agreement with the results previously reported in the literature [15][17] (stable basal spacing d001 = 22 Å). Figure II-F11 Evaluation of the results obtained with method (D). Absorption and emission spectra in ethanol of C30B clay photo-functionalized by method (D) with (a) NBAP 1MC and (b) RhP 1MC. Samples have been carefully washed with ethanol in order to assure that the response has to be assigned uniquely to the intercalated dye. We chose ethanol to wash photo-functionalized samples because, on the basis of its surface tension ( L ethanol = 22.8 mN/m < S C30B), it shouldn’t produce significant clay swelling [2]. We further validated our choice by confirming by XRD that ethanol doesn’t produce interlayer (i.e. nanoscopic) swelling – at least not on rapid immersion 9 See § II-6. Antonella ESPOSITO 121 Chapter II (results not shown). Some authors wash functionalized clays by water or ethanol-water mixtures, hot or at room temperature. Indeed, Gates [13] reported that washing with ethanol or other organic solvents removes exceeding intercalated molecules from organoclays, whereas washing with water is only partially effective. Actually, the advantage of ethanol (and ethanol-water mixtures) is that it penetrates the interlayer spacing of organically modified clays more easily than water – thus it better removes any excess of surfactant into clay galleries. The choice of pure ethanol instead of an ethanol-water mixture is justified by the fact that Burgentzlé et al. [2] reported that ethanol may even act as a degelling agent and, in any case, alcohols with short alkyl chains (less than five methylene groups) seem to hinder clay interlayer swelling – which means that washing with ethanol won’t absolutely interfere with the effect strictly due to the functionalization. Analogously to method (C), we voluntarily avoided drying the photo-functionalized clays by oven in order to preserve the fluorescent molecules. The described procedure resulted in a fine powder, which didn’t require further grinding. In conclusion the most practicable, appropriate and efficient method to photofunctionalize a commercial organoclay is method (D), i.e. cation exchange processing of the pristine clay with an organic cationic dye followed by washing with ethanol, recovering by centrifugation and drying under exhaust hood at room temperature. Next step will be the optimisation of the parameters for the chosen photo-functionalization protocol, namely the choice of the fluorescent molecule and its concentration. II-6 OPTIMIZATION OF THE PROTOCOL FOR CATION EXCHANGE PROCESSING (D) We selected two organic cationic (perchlorate) dyes and we performed the photo-functionalization of C30B clay following the chosen protocol – cation exchange processing (D) – with two concentrations of the dye (corresponding to 1 and 0.25 times the modifier concentration of the organically-modified clay, denoted 1MC and 0.25MC respectively). The aim was double: we wished to compare two fluorescent molecules in order to choose the one giving the best results and, at the same time, we intended to find the optimum value of fluorescent molecule concentration to get the most efficient photo-functionalization. Indeed, the spatial arrangement of intercalated dyes typically PhD INSA de Lyon (2008) 122 PHOTO-FUNCTIONALIZATION Lamellar fillers depends on the guest concentration in the exchanging medium. Čapková et al. [17] deduced, by molecular modeling, that fully intercalated samples don’t necessarily need to be fully ion-exchanged and, more interesting, disordered monomers adsorbed at the silicate edges may create a sterical barrier to further dye intercalation, which is a reason good enough to limit the excess of fluorescent dye. Another good reason is that organic cationic dyes spontaneously aggregate, especially at high concentrations10. Moreover, the addition of a small amount of adsorbent (clay) to a dilute dye solution causes similar effects to those observed by increasing the concentration of dye [16] – which means that in the presence of clay the actual concentration is higher than the expected one. We’ve always performed cation exchange processing at 80°C since NBPA and RhP could resist such temperature in solution and, parenthetically, it has been previously proved that this is the optimal temperature for cation exchange processes [3]. II-6.1 Choice of the fluorescent molecule NBPA and RhP differ in the wavelength range of absorption (627 and 532 nm respectively) and fluorescence emission (665 and 553 nm respectively) (Figure II-F8), the elemental composition (C:N percentage ratio is 6.00 for NBAP and 12.00 for RhP), the molecular weight (418 and 543 g/mol respectively, see Table II-T1), the crystalline structure (Figure II-F5) and the thermal degradation behavior (Figure II-F6). They basically have the same maximum molecular dimension Dmax (15 and 14 Å respectively, see Table II-T1) and both have a polyaromatic heterocyclic configuration (based on the benzophenoxazine and xanthene in the case of NBAP and RhP respectively), but the different XRD fingerprints suggest that their spatial configurations (bonding angles and possible bendings) are not the same. 10 See considerations about Figure II-F11 – keeping in mind that, according to the theory of exciton splitting, not all the aggregates are fluorescent. Antonella ESPOSITO 123 Chapter II Figure II-F12 Spatial configuration of the free fluorescent molecules as visualized by molecular modeling: NBAP (left) and RhP (right). NBAP is almost bidimensional, contrarily to RhP molecule. Molecular dimensional values are expressed in ångströms. In fact, by molecular modeling we can confirm that the two molecules are quite different in terms of in-plane and out-of-plane spatial arrangement, as shown in Figure II-F12. This difference in spatial configuration may significantly influence the capability of each fluorescent molecule to enter clay galleries (in spite of the sterical hindrance due to the MT2EtOH molecules), to exchange for the Na+ sites corresponding to the residual CEC, to eventually aggregate and then arrange in crystalline structures. PhD INSA de Lyon (2008) 124 PHOTO-FUNCTIONALIZATION Lamellar fillers Figure II-F13 Evaluation of the results obtained with method (D) and a concentration of organic cationic dye equivalent to 1MC. Comparison of NBAP and RhP samples in order to choose the most efficient fluorescent molecule for clay photo-functionalization. TGA characterizations of (a) NBPA and (b) RhP samples supported by XRD analysis of (c) NBPA and (d) RhP samples. Bold lines represent the results obtained after washing, at the very end of the photo-functionalization procedure. TGA characterizations of C30B clay photo-functionalized with 1MC of NBAP and RhP are shown in Figure II-F13 (a) and (b) respectively, supported by the relative XRD paths (c) and (d). Cation exchange processing (D) with 1MC NBAP produces, in comparison with TGA results for the pristine C30B clay, an additional peak in the area corresponding to physisorbed chemicals (219°C) (Figure II-F13a). This peak of the derivative curve is quite high but is easily removed by washing, witnessing that such a big amount of fluorescent molecule is probably useless to the main objective of photofunctionalization. There’s a relevant difference between this first peak and the second one (288°C): one could associated both of them to physisorbed chemicals, but it is clear that washing procedure doesn’t significantly lower it, demonstrating that it corresponds to chemicals intercalated into clay galleries. The last part of the curve (T>300°C) looks Antonella ESPOSITO 125 Chapter II more complicated because the permanence of the clay in an exchanging medium at 80°C during cation exchange processing has surely produced a rearrangement of the chemicals into the galleries. Furthermore, the presence of aromatics always reshapes TGA curves, giving sometimes rise to a brand new peak at higher temperatures – what could explain the large shoulder between 400°C and 500°C. A probable molecular rearrangement into clay galleries could also be a good explanation for the results obtained by XRD – which are a little controversial with TGA results – taking also into account the Dmax and the molecular configuration of NBAP (Figure II-F12). Figure IIF13c shows that the d001 spacing for a sample of clay recovered right after cation exchange processing and simply dried under exhaust hood is 19 Å ( d001 = +1.5 Å) but, after washing with ethanol and drying again under exhaust hood, it slightly shifts right, going back to the initial value for pristine C30B (17.5 Å)11. Likely, an excess of NBAP could enter clay galleries and push further apart the silicate platelets by forming an additional organic layer which is obviously unstable, since it couldn’t have undergone cation exchange because of the sterical hindrance due to the presence of the MT2EtOH molecules. However, TGA assures that photo-functionalization was effective (peak at 288°C): we conclude that XRD measurements cannot detect the photo-functionalization of C30B clay with NBAP because dye molecules, properly exchanged with Na+ sites immobilized on silicate platelets, are too small (Dmax = 15 Å) in comparison with the MT2EtOH molecules (Dmax = 29 Å), even if the paraffinic configuration reduces their “effective length” to 17.5 Å. Cation exchange processing (D) with 1MC RhP produces, in comparison with TGA results for pristine C30B clay, a double additional peak in the area of physisorbed chemicals (230°C and 258°C) (Figure II-F13b). These two peaks are quite intense – especially the second one – but have been easily removed by washing, witnessing once again that such a big amount of fluorescent molecule is absolutely useless to the main objective of photo-functionalization. The other two peaks (380°C and 425°C) correspond to the region representative of the organics which entered clay galleries (T>300°C) and are insensitive to washing – especially the second one, which can be 11 See Figure II-F5d. PhD INSA de Lyon (2008) 126 PHOTO-FUNCTIONALIZATION Lamellar fillers probably due to the aromatic portion of the intercalated fluorescent molecules. Contrarily to what observed by XRD in the case of NBAP, photo-functionalization with RhP (Figure II-F13d) is clearly detectable by this technique (d001 = 22 Å, d001 = +4.5 Å) and the results are surprisingly consistent (even slightly better) with those previously reported in the literature for the definitely simpler case of pure RhB-MMT complexes [17]. All the RhP molecules which entered clay galleries seem to have properly undergone cation exchange, since they’re not removed by washing and the d 001 value doesn’t decrease. Interlayer spacing could have increased in consequence of a kind of induced supramolecular organization or, more likely, a rearrangement of the paraffinic configuration of MT2EtOH molecules induced by the presence of smaller (but still sterically-hindering) molecules on the absorbent surface. Under the hypothesis that RhP molecules into clay galleries don’t interact with each other because of their statistical alternation with MT2EtOH molecules (which masks them to each other) and that MT2EtOH molecules are the sole responsible for clay interlayer spacing, the presence of small molecules in addition to bigger molecules bearing long tallow fatty chains increases the characteristic paraffinic angle between the tallow chains and the silicate substrate from about 37° to 49° ca, as shown in Figure II-F14. Generally speaking, the presence of smaller molecules increases the “effective length” of longer molecules in relation to silicate layers separation, which actually is one of the goals of traditional clay functionalization in perspective of polymer-based nanocomposite fabrication. A similar mechanism has been evocated to explain the crystalline swelling of organoclays in solution: intercalated organics are solvated by solvent molecules, which surround and support the aliphatic tails, causing a tilt with respect to the silicate surface [13]. Molecular modeling confirmed that small molecules, when intercalated in clay galleries, lead to the tilting of bigger molecules, as shown by Čapková et al. [17] in the case of water molecules interacting with the xanthene planes of Rhodamine B (RhB) into MMT galleries: water molecules can be placed adjacent to the silicate layers and fill the empty spaces between RhB cations. The analogy with this study is possible if we compare water to RhP and RhB to MT2EtOH. Iwasaki et al. [16] reported that dye cations into clay galleries are oriented horizontally ( planes parallel to the silicate surfaces) at low loadings, tilt towards the perpendicular direction with an increase in dye concentration Antonella ESPOSITO 127 Chapter II and finally stand vertically at high loadings. In this study, since the chosen pristine clay is an organically-modified clay, we don’t observe any difference in the orientation of the dye molecules probably because their configuration is influenced by the presence of MT2EtOH molecules. Thus, it is possible to vary the concentration of organic cationic dye and always obtain the same tilted configuration. Anyway, the invariable tilting angle obtained seems to be directly dependent on the behaviour of the dye cations: Pospíšil et al. [25] showed by molecular modeling that rhodamine B cations intercalated in a clay host structure from a diluted solution may organize in a monolayer arrangement of tilted monomers with a basal spacing within 21-25 Å with a tilting angle in the range 40-60°, which is in good agreement with the value found in this work (49°). Fujita et al. [15] reported a similar behaviour for the basal spacing of rhodamine/ taenolite complexes obtained with a molar ratio organic/inorganic varying from 0.086 to 0.46 (d001 nearly constant at 21.7 to 22.2 Å). Figure II-F14 Possible explication for the increase of pristine C30B interlayer spacing (a) following photo-functionalization with a cationic organic (b), under the hypothesis of a synergistic effect due to the presence of both small and bigger organic molecules cation-exchanged within the silicate platelets. In conclusion, both NBAP and RhP are suitable to photo-functionalize C30B organically-modified clay. However, RhP looks to be more efficient (the TGA peak which is washing insensitive corresponds to a higher temperature) and, additionally, photo-functionalization with RhP appears to be readily detectable both by TGA and XRD. Finally, RhP has a wavelength range of absorption and a fluorescence emission spectrum which is more interesting for an eventual application with Visiovis, the brand PhD INSA de Lyon (2008) 128 PHOTO-FUNCTIONALIZATION Lamellar fillers new equipment we’re developing in our laboratories12. Thus, we’ll focus our attention on the results obtained with RhP but, as a matter of fact, we’ll keep comparing them with the results obtained with NBAP. II-6.2 Influence of the fluorescent molecule concentration When dealing with fluorescence phenomena, fluorophore concentration is a key parameter because of the risk of quenching and the tendency to form supramolecular assemblies, which occasionally is a desired result but in any case has to be controlled, somehow. This is the reason why we estimated necessary to evaluate the influence of the concentration of fluorescent molecule on the results of clay photo-functionalization. As previously mentioned13, we’ve firstly performed cation exchange processing in the presence of an amount of fluorescent molecule equivalent to 1MC – which turns out to surely be an excess of cationic organic dye. We then realized that performing method (D) with such a concentration of perchlorate is efficacious but not efficient, since TGA measurements revealed that a large amount of fluorescent molecule doesn’t enter clay galleries but rather physisorbs on clay platelets, and is readily washed away by ethanol (see Figure II-F13). Moreover, we’ve already underlined that any excess physisorbed on clay silicate platelets may create a sterical barrier to further intercalation: therefore, we decided to diminish the amount of fluorescent molecule to a quarter of the former amount, i.e. 0.25 times the modifier concentration of C30B (0.25MC). TGA characterizations of the samples obtained by method (D) with 0.25MC of NBAP and RhP are shown in Figure II-F15 (a) and (b) respectively, supported by the relative XRD paths (c) and (d). Cation exchange processing (D) with 0.25MC NBAP produces, in comparison with the results obtained with 1MC NBAP, a smaller additional peak in the area of physisorbed chemicals (253°C) (Figure II-F15a). This peak is lowered by washing but doesn’t completely disappear: the same observation has already been done for pristine C30B, which undergoes a first weight loss at 253°C 12 13 Visiovis configuration and evolutions will be detailed in § IV-1. See considerations about method (D) in § II-1.5. Antonella ESPOSITO 129 Chapter II (Figure II-F6d)14. Therefore, it is confirmed that the small residual peak at 253°C in washed samples does effectively represent a portion of the MT2EtOH molecules well intercalated into clay galleries but in a peripheral position – which explains why it is impossible to completely remove it by washing. The second weight loss occurs at 310°C and has to be assigned to the chemicals intercalated into clay galleries, since the peak turns out to be perfectly washing insensitive. By the way, this second peak is the only one to be truly washing insensitive, since the last portion of the derivative curve (T>325°C) looks complicated, probably because of some molecular rearrangement of the chemicals into the galleries. The hypothesis of a molecular rearrangement, in this case, isn’t substantiated by XRD results (Figure II-F15c), because the interlayer spacing undergoes only an insignificant decrease after washing with ethanol (from 20 to 19 Å, d001 = –1 Å), unlike the samples obtained with 1MC NBAP (Figure II-F13c). The fact that the d001 spacing in the 0.25MC NBAP exchanged clay remains almost constant whereas it was shown that it decreases in the 1MC NBAP exchanged clay is the first evidence that the concentration of fluorescent molecule influences the efficiency of the photo-functionalization by method (D). One could expect that, whatever the fluorescent molecule, an increase of its concentration (from 0.25MC to 1MC) would promote clay swelling by shifting the XRD peak towards smaller values of 2 (corresponding to greater value of the interlayer spacing) or by heightening it (corresponding to a greater fraction of properly photo-functionalized sample, since the intensity of the diffraction peak for a given crystalline population is directly proportional to its mass fraction in the sample [8]). Indeed, we can observe that an increased concentration of NBAP seems to slightly interfere with the cation exchange: when the concentration of fluorescent molecule lowers, the peak shift is barely higher (20 Å for 0.25MC vs. 19 Å for 1MC), the consequences of washing are slightly reduced ( d001 = –1 Å for 0.25MC vs. d001 = –1.5 Å for 1MC, Figure II-F15c vs. Figure II-F13c). The difference is not enormous, but at least we can conclude that an increase of concentration of fluorescent molecule doesn’t produce any amelioration. 14 For further details about TGA measurements performed on C30B, go back to § II-4.2. PhD INSA de Lyon (2008) 130 PHOTO-FUNCTIONALIZATION Lamellar fillers Figure II-F15 Evaluation of the results obtained with method (D) and a concentration of organic cationic dye equivalent to 0.25MC. Comparison of NBAP and RhP samples is made to confirm the choice of the fluorescent molecule for clay photo-functionalization. TGA characterizations of (a) NBAP and (b) RhP samples, supported by XRD analysis of (c) NBPA and (d) RhP samples. Bold lines represent the results obtained after washing, at the very end of the photo-functionalization procedure. Cation exchange processing (D) with 0.25MC RhP produces, in comparison with the results obtained with 1MC RhP, only one additional peak in the area of physisorbed chemicals (243°C) (Figure II-F15b). This peak is lowered by washing but doesn’t completely disappear: the same observation as 0.25MC NBAP samples can be done. The second peak (315°C) may appear ambiguous but it could be explained in a similar way than the previous one, since its behavior in relation to clay washing is analogous. Finally, the last weight loss (406°C) corresponds to the intercalated chemicals and is absolutely washing insensitive. Contrarily to 0.25MC NBAP samples and unexpectedly with reference to the observations done by Čapková et al. [17], no concentration effects appear on XRD results when the amount of organic cationic dye in the exchanging Antonella ESPOSITO 131 Chapter II medium varies (Figure II-F15d): an increased concentration of RhP neither augments the d001 spacing (22 Å, the same value obtained with 1MC RhP), nor hinders the cation exchange mechanism because of the sterical hindrance effects due to the excess of dye adsorbed on the silicate surfaces. Furthermore, the lowered concentration of dye in the exchanging solution doesn’t lead to the presence of two broad partially overlapping diffraction maxima corresponding to two different phases (and, consequently, two different molecular arrangements of the intercalated organics) [17]: for both values of concentration, RhP excess doesn’t enter clay galleries and is efficiently removed by washing, as confirmed by TGA (Figure II-F13b and Figure II-F15b). Clearly, the intercalation mechanisms are influenced by the presence of the MT2EtOH molecules, which reduce the possibilities of several molecular arrangements of the dye. In summary, clay photo-functionalization by cation exchange processing (D) with high concentrations of fluorescent molecule is possible (no significant fluorescence quenching seems to occur, as confirmed by spectrofluorimetry results in Figure II-F11) but unnecessary, since most of the fluorophore excess is washed away and wasted. All the evidences collected so long indicate that cation exchange processing (D) with a smaller amount of an organic cationic dye like RhP (e.g. a relatively small percentage of the modifier concentration, viz. 0.25MC RhP) is sufficient to render the organoclay photo-active (result not shown). XRD reveals the highest crystalline swelling, moreover TGA and EA assure that photo-functionalization was successful even in the absence of a significant surfactant excess. The global shape of the TGA derivative curve for the 0.25MC RhP sample (Figure II-F15b) looks pretty different from the curve obtained with 1MC RhP (Figure II-F13b) uniquely because of the different ratio of dye to MT2EtOH molecules. In the following paragraph we’ll present some complementary characterizations (FTIR spectroscopy and TGA-FTIR) of C30B clay photo-functionalized 0.25MC RhP. Such results, obtained before and during the thermal degradation of the exchanged clay, could eventually support the observations previously made according to XRD and TGA characterizations. PhD INSA de Lyon (2008) 132 PHOTO-FUNCTIONALIZATION II-6.3 Lamellar fillers Complementary characterizations (C30B 0.25MC RhP) As the intercalation of RhP cations into clay galleries is based on a simple cation exchange reaction, the host-guest interactions have a non-covalently-bonded nature [17][25]: a comparison of FTIR spectra for the host inorganic structure (C30B, reported from Figure II-F7b), the guest organic dye (RhP, reported from Figure II-F7a) and the photo-functionalized clay (C30B-RhP) should show that the main absorption bands of the pristine clay and those of the polyaromatic heterocyclic fluorophore coexist in the FTIR fingerprint of the photo-functional inorganic/organic complex. Indeed, in the FTIR spectrum of the clay photo-functionalized 0.25MC RhP (Figure II-F16), one can distinguish the features of both C30B and RhP. Figure II-F16 FTIR spectrum of C30B photo-functionalized 0.25MC RhP compared to the reference FTIR measurements on pristine C30B and RhP (for further details about reference measurements, see Figure II-F7 and related comments15). First of all, the broad absorption band in the region 3750-3050 cm-1 (which is due both to the intrinsic structure of the clay and to the fact that C30B is organicallymodified – please see Figure II-F7b and related comments15) is clearly modified by the contribution of the sharp peak at 3360 cm-1 due to the polyaromatic dye (see Figure IIF7b and related comments15): in the FTIR spectrum of the photo-functionalized clay, 15 Explanations in § II-4.4. Antonella ESPOSITO 133 Chapter II the peak around 3630 cm-1 (“free” OH groups) is still present, but its intensity is definitely reduced in comparison with the intensity of the peak around 3360 cm -1 (unsaturated and/or aromatic compounds). As previously reported by Pospíšil et al. [25] the absorption band at 3360 cm-1 is slightly smoothed and broadened in comparison with the original band in the FTIR spectrum of the free fluorescent molecule. The set of two peaks corresponding to the vibrations of long linear aliphatic chains (approximately 2930 and 2860 cm-1, see Figure II-F7b) are also preserved, but their relative intensity is lowered as well. The two weak peaks at 1645 and 1471 cm-1 previously detected in the pristine C30B spectrum are hidden – in the spectrum of the photo-functionalized clay – by the contribution of RhP bands (see Figure II-F7b): RhP fingerprint is clearly recognizable in the region 1800-1400 cm-1, even if the intensity of all the peaks is scaled down. Finally, the absorption bands previously assigned to crystalline silicate (Si-O-Si) and to the vibrations of the inorganic crystalline lattice (Figure II-F7b) remain unchanged. Cervantes-Uc et al. [26] have already analyzed the thermal degradation of commercial Cloisite® 30B by TGA-FTIR, carrying out their measurements at 10°C/min under dry nitrogen atmosphere. The Gram-Schmidt plot16 showed two main peaks (297 and 427°C) corresponding to the main discharges of volatile products: the first peak corresponded to a spectrum containing the absorption bands for O–H stretching, aliphatic C–H stretching, CO2, C–O stretching from alcohol groups and probably carboxylic acid; the second one (definitely more intense than the first one) corresponded to a spectrum in which the intensity of the absorption bands for CO2 and aliphatic C–H stretching increased and the intensity of the signal for carbonyl stretching decreased. By the way, they also reported the FTIR spectra of the gases evolved at 413 and 500°C, since therein they found some interesting features. In the spectrum at 427°C they found that the intensity of the absorption band corresponding to methyl and methylene groups increased whereas CO2 signal disappeared (no more emission of carbon dioxide), and some absorption bands which were not clearly defined at 297 and 413°C got a little 16 As previously reported, the Gram-Schmidt plot deals with the total IR absorbance of the evolved gases in the whole spectral range (§ II-3.5). PhD INSA de Lyon (2008) 134 PHOTO-FUNCTIONALIZATION Lamellar fillers sharper. They also observed that in the spectrum at 500°C almost all the bands vanished. The only absorption band observed in all the spectra up to 500°C was a band at 3020 cm-1 ca (C–H stretching in double bonds). In summary, they report that C30B degradation products showed bands corresponding to water and/or alcohols and/or carboxylic acid (O–H stretching in the 4000-3300 cm-1 range, O–H bending at 1701 and 1518 cm-1), aliphatic compounds (2990-2800 cm-1) and carbon dioxide (2364 and 2324 cm-1). The band at 1518 cm-1 probably reveals the presence of alkenes (especially if confirmed by another band at 3020 cm-1): indeed, the presence of alkenes has been reported by several authors [26][27] and could be justified by some degradation processes based on the Hoffmann elimination reaction, which is the most probable degradation mechanism of organically modified clays containing hydroxyl groups [26]. Table II-T3 Chemical species evolved during TGA analysis of the raw materials and of C30B photo-functionalized 0.25MC RhP before (bw) and after washing (aw), identified by FTIR spectroscopy. Comparison with some data from the literature [26][27]. RhP C30B C30B [26] C30B [27] C30B RhP 0.25MC bw C30B RhP 0.25MC aw T [°C] Chemicals 275 557 290 350 600 297 413 427 H2O, ethyl chloride, oxidized compounds (3738, 3250, 1744, 1515, 1067 cm-1) H2O, CO2, CO 210 255 375 550 375 560 720 H2O, CO2, CO, aldehyde (C10), ketone (C10) or carboxylic acid (C17) H2O, CO2, CO H2O, CO2, alkanes, alkenes, aldehydes, carboxylic acids, amines H2O, CO2, alkanes, alkenes H2O, alkanes, alkenes, alcohols N,N-dimethylacetamide, linear aldehydes (C7-C12), linear and branched alkanes (C9-C21), alcohols (C8-C16), alkene alcohols (C9-C16), chloro-alkanes (C14-C16), alkenes (C13-C18) H2O, CO2, CO, ethyl chloride, aldehydes (C6) or ketones H2O, CO2, CO, aldehydes or ketones H2O, CO2, CO H2O, CO2, CO, N-methylformamide H2O, CO2, CO The presence of small signals in the 800-650 cm-1 range would represent some chlorinated compounds, which could likely come from the thermal decomposition of the exceeding portion of surfactant which didn’t undergo cation exchange process [26][27]. Antonella ESPOSITO 135 Chapter II In support of TGA results, Edwards et al. [27] rather used solid phase microextraction techniques to analyse the thermal degradation products evolved during the processing of organically-modified layered silicates at a given temperature (210°C). The data reported by Cervantes-Uc et al. [26] and Edwards et al. [27] about the thermal degradation of Cloisite ® 30B are summarized in Table II-T3 to help comparing with the results we obtained by TGA-FTIR, listed in the same table. The parameters of TGA-FTIR analysis chosen for each sample of interest are listed in Table II-T4, along with the total IR absorbance of CO2 and CO. Table II-T4 TGA-FTIR parameters (sample initial weights, TGA temperature ranges and the corresponding weight losses, temperatures at which gases have been discharged and analysed by FTIR spectroscopy, total amounts of evolved CO2 and CO) used to characterize C30B photofunctionalized 0.25MC RhP before (bw) and after washing (aw), as well as the raw materials. Initial weight TGA program [g] T range [°C] RhP 12.20 C30B pristine 16.03 C30B RhP 0.25MC bw 17.58 C30B RhP 0.25MC aw 12.73 a Weight loss FTIR analysis T [°C] [%] 200 320°C 23.0 275 320 640°C 76.6 557 200 450°C 450 800°C 200 280°C 280 450°C 450 800°C 200 450°C 450 620°C 620 850°C 16.0 12.8 9.2 16.9 23.4 11.9 11.1 8.3 290 / 350 600 255 375 550 375 560 720 Total IR absorbance b [u/mg] CO2 CO 1220.0 103.5 315.0 30.4 584.9 56.3 419.5 46.8 a TGA programs are essentially linear ramps performed with a heating rate of 10°C/min. b The total IR absorbance of CO2 and CO has been calculated by integrating the corresponding peaks in space and time, thus u abs cm s The thermal degradation of RhP produces almost 4 times more CO2 and 3.5 times more CO than the pristine C30B: this could be justified by the chemical composition of RhP and pristine C30B determined by EA (Table II-T2)17, since RhP 17 Reference EA measurements are reported in § II-4.3. PhD INSA de Lyon (2008) 136 PHOTO-FUNCTIONALIZATION Lamellar fillers contains 3 times more carbon per unit of weight than C30B. Therefore, it is obvious that the amounts of CO2 and CO produced by thermal degradation of C30B photofunctionalized 0.25MC RhP before washing are higher than the same values for pristine C30B, and that such values slightly decrease after washing (aw). If we consider, now, the chemical species evolved during TGA analysis (Table II-T3), we realize that the only compound which could help evaluating the effects of the photo-functionalization and of the washing procedures is ethyl chloride: any other product of the thermal degradation of RhP (essentially H2O, CO2 and CO) is a common combustion product which could not make the difference. Only Edwards et al. [27] detected, even in commercial C30B, some chloro-alkanes (having much longer aliphatic chains). Of course, we’re aware that the degradation products of a photo-functionalized (organically-modified) clay could consist of several complex molecules and that, by the way, ethyl chloride could react with other degradation products and produce some more complex compounds before its detection: that’s why we’re presenting these results as some complementary characterizations. About pristine C30B, TGA-FTIR results are in perfect agreement with the results previously reported in the literature [26][27] (Table II-T3): among the degradation products of commercial grade C30B one can find (in addition to water, carbon dioxide and carbon monoxide) some alkanes and/or alkenes with several substituting groups (mainly hydroxyl, carboxyl and amine) and always some C=O bond (characteristic of aldehyde, ketones, acetamide and formamide). Let’s focus a little longer on ethyl chloride emissions. It’s interesting to observe that ethyl chloride – which is one of the degradation products of the free fluorescent molecule – has been detected in the spectrum of the volatile products emitted at 255°C by C30B photo-functionalized 0.25MC before washing (bw) but not after washing (aw): this confirms that the first peak of the derivative TGA curve for this sample (Figure IIF15b) effectively corresponds to a portion of physisorbed dye which didn’t enter clay galleries and, thus, didn’t undergo cation exchange process. The fact that no more emission of ethyl chloride is detected in the washed sample (aw) assures that the washing procedure was efficacious and that the sample (aw) doesn’t contain any more exceeding dye, which excludes that the residual peak at 243°C (still visible after Antonella ESPOSITO 137 Chapter II washing) could be assigned to a washable excess. Starting from 550°C only small, residual organic molecules evolve from the burning samples. Contrarily to what observed by Cervantes-Uc et al. [26] in the case of pristine C30B, we detected CO2 and CO emissions from the thermal degradation of C30B photo-functionalized 0.25MC RhP before (bw) and after washing (aw) up to 850°C. A direct correlation between TGA and TGA-FTIR results can be done with the help of Figure II-F17. Figure II-F17 Direct correlation between simple TGA characterizations and TGA-FTIR results for (a) RhP, (b) pristine C30B, (c) C30B photo-functionalized 0.25MC RhP before washing (bw) and (d) C30B photo-functionalized 0.25MC RhP after washing (aw). PhD INSA de Lyon (2008) 138 PHOTO-FUNCTIONALIZATION II-6.4 Lamellar fillers Further general comments about the efficiency of the photo-functionalization Let’s take a look at the complete series of TGA curves measured for the samples obtained by cation exchange processing (D) with both the concentrations (1MC and 0.25MC) of NBAP and RhP (Figure II-F18). Figure II-F18 Global comparison of the complete series of samples obtained by cation exchange processing (D) with two different concentrations (1MC and 0.25MC) of (a) NBAP and (b) RhP. Bold lines represent the thermal behavior of the samples after washing with ethanol and drying under exhaust hood (bw = before washing, aw = after washing). Antonella ESPOSITO 139 Chapter II What’s interesting to observe is that: All the washed photo-functionalized clays are more stable than (or at least as stable as) the pristine C30B clay washed with ethanol, which means that photofunctionalization doesn’t significantly affect thermal resistance; All the washed clays (pristine C30B included) are more stable than the respective unwashed counterparts; Both the fluorescent molecules result thermally stabilized for T>300°C in consequence of their insertion into clay galleries; With both NBAP and RhP, the most stable clay is the one which has been photo-functionalized with 1MC of organic cationic dye and then washed; With both NBAP and RhP, the less stable clay is the one which has been photo-functionalized with 1MC of organic cationic dye but hasn’t yet been washed. Stabilization by washing has already been highlighted in the literature [28] and is justified by the removal of the surfactant excess – which typically is organic and, thus, less stable than clay with respect to temperature. The stabilization of organic chemicals due to the presence of inorganic clays has been already reported and commented [5][6]. The experimental values of MC obtained for all the photo-functionalized clays after washing (Table II-T5) confirm once more that both NBAP and RhP are suitable to photo-functionalize C30B organoclay. Moreover, it is clear that even a smaller concentration of fluorescent molecule (0.25MC) – whatever the fluorescent molecule – is sufficient to increase the value of MC and to assure a certain degree of CEC recovery (167% and 55% for NBAP and RhP, respectively). By the way, in Table II-T5 one can also observe that, for any given concentration of fluorescent molecule and in the same cation exchange processing conditions, NBAP looks to be more efficient than RhP in terms of MC[ meq / 100 g ] and, thus, in terms of CEC recovery as well. PhD INSA de Lyon (2008) 140 PHOTO-FUNCTIONALIZATION Lamellar fillers Table II-T5 Experimental values of MC (derived from TGA results) and CEC recovery for the clay photo-functionalized by cation exchange processing (D), with both NBAP and RhP at both concentrations (1MC and 0.25MC), after washing with ethanol. Tmax a [°C] intercalated chemicals C30B pristine MC MC exper [g/100g] b MC exper [meq/100g] c CEC recovery d [%] 30 83 Ref 70 251 + 202 58 222 + 167 58 190 + 129 25 129 + 55 360 288 345 435 310 400 380 425 C30B NBAP 1MC C30B NBAP 0.25MC C30B RhP 1MC C30B RhP 0.25MC 406 a Tmax is the temperature corresponding to the max weight loss rate in Derivative ThermoGravimetric (DTG) curves. Only the peak(s) assigned to intercalated chemicals are considered in order to measure MC (Modifier Concentration). b Each experimental value of MC expressed in [g/100g] has been evaluated by identifying, isolating and integrating the DTG peak(s) of interest, calculating their area and deducing the % wt as the % area [reference for % = total amount of organic loss up to 550°C]. c Each experimental value of MC expressed in [meq/100g] has been calculated on the basis of the corresponding value expressed in [g/100g] and does already include the contribution due to MT2EtOH molecules (83 meq/100g). In summary: MC[ meq / 100 g ] MC[ g / 100 g ] 1000 Mw MC[ meq / 100 g ] sample MC[ meq / 100 g ] RhP, NBAP MC[ g / 100 g ] RhP, NBAP 1000 M wRhP , NBAP MC[ meq / 100 g ] C 30 B 83[ meq / 100 g ] d CEC recovery corresponds to the variation (expressed as an incremental percentage) of the value of MC expressed in [meq/100g] in comparison with the reference value for the pristine C30B (83 meq/100g) as a consequence of the photo-functionalization. Briefly: CEC re cov ery Antonella ESPOSITO MC [ meq / 100 g ] 100 MC [ meq / 100 g ] C 30 B MC [ meq / 100 g ] sample MC [ meq / 100 g ] C 30 B 100 MC [ meq / 100 g ] C 30 B 141 Chapter II This observation doesn’t really controvert our previous statement (based on TGA and XRD results) that RhP looks to be more efficient than NBAP 18: the differences in MC and CEC recovery for NBAP and RhP can be easily justified by the differences in molecular configuration of the two organic cationic dyes. Indeed, we estimated by molecular modeling that RhP molecules occupy a Van der Waals (VdW) volume of 423 Å3, which is almost double in comparison with the VdW volume occupied by NBAP molecules (297 Å3). The VdW volume of a molecule is the volume within the VdW surface, which corresponds to the surface that intersects with the VdW radii of the atoms in the molecular structure. By considering all the experimental evidences – whether directly collected by XRD and TGA, or deduced by TGA results and then supported by molecular modeling – we can conclude that the efficacy of the photofunctionalization of an organically-modified clay depends on the processing conditions, the nature and concentration of the organic cationic dye, but also its molecular dimension, configuration and occupied volume. Two organic cationic dyes implied in clay photo-functionalization cannot be compared solely on the basis of their maximum molecular dimension (Dmax): the volume occupied is also relevant, for it may be the main parameter responsible for the tilting angle observed by XRD (see Figure II-F14 and related comments)19. Table II-T6 shows the chemical composition (normalized on the basis of the Si % detected by EA in the pristine C30B, Table II-T2) of the photo-functionalized clays, compared to the chemical composition of the pristine C30B clay. The N:C ratio is a parameter which could help evaluating the extent of MC increase and CEC recovery as a consequence of the addition of fluorescent molecules into clay galleries during the cation exchange process. In fact, MT2EtOH has a lower N:C ratio (0.05) in comparison with NBAP (0.17) and RhP (0.08), meaning that the N:C balance in the pristine C30B is shifted towards C. By inserting NBAP or RhP molecules into clay galleries, one could expect that the N:C balance shifts towards N, i.e. increases (particularly in the case of NBAP-exchanged samples) with respect to the initial value. Analogously, the N:Si ratio 18 19 The choice of RhP has been largely vindicated in § II-6.1 See § II-6.1. PhD INSA de Lyon (2008) 142 PHOTO-FUNCTIONALIZATION Lamellar fillers can be used to assess the addition of fluorescent molecules into clay galleries. Right after cation exchange processing, the N:Si ratio should increase (Si % content doesn’t change and fluorescent molecules should bring additional N atoms); after the washing procedure, it should decrease without recovering the initial value – meaning that the excess of fluorescent molecule has been removed but a portion of the nitro organic species inserted into clay galleries is residual. Finally, N:C and N:Si variations at each processing step let estimating its efficiency at a glance whereas their global values (bold values, Table II-T6) highlights the effect of the whole photo-functionalization process. Table II-T6 Chemical composition of the samples photo-functionalized by cation exchange processing (D), with both NBAP and RhP at both concentrations (1MC and 0.25MC), before (bw) and after washing (aw) with ethanol. C30B (MC) pristine C [% wt] norm N [% wt] norm H [% wt] norm Norm Factor (NF) a N:C b N:C variation [%] N:Si c N:Si variation [%] a C30B NBAP 1MC bw aw C30B NBAP 0.25MC bw aw C30B RhP 1MC bw aw C30B RhP 0.25MC bw aw 20.18 1.07 3.98 Ref 23.83 2.39 3.43 0.84 19.05 2.81 2.23 1.06 22.40 1.60 4.00 0.97 18.60 1.46 3.52 1.06 24.43 1.37 3.35 0.74 20.84 1.54 2.91 1.00 23.11 1.33 3.89 0.93 19.83 1.28 3.43 1.03 0.05 0.10 0.15 0.07 0.08 0.06 0.08 0.06 0.06 - + 100 + 200 + 40 + 60 + 20 + 60 + 20 + 20 0.05 0.16 0.12 0.08 0.06 0.11 0.07 0.07 0.06 - + 220 + 140 + 60 + 20 + 120 + 40 + 40 + 20 The Norm Factor (NF) is employed for EA data normalization and is obtained by comparing Si % wt of each sample with Si % wt of the pristine C30B clay (Table II-T2). NF Si % wt sample Si % wt C 30 B b N:C ratio reference values (from Table II-T2): MT2EtOH (theoretical 0.05), C30B (experimental 0.05), NBAP (theoretical 0.16, experimental 0.17), RhP (theoretical and experimental 0.08). c N:Si ratio reference values (from Table II-T2): MT2EtOH (theoretical NBAP (experimental ), RhP (experimental ). ), C30B (experimental 0.05), From Table II-T6 one can see that N:C variation is positive for all the samples and diminishes as the concentration of fluorescent molecule decreases (in the case of NBAP, for example, the N:C variation is +200 % for 1MC and +60 % for 0.25MC). If Antonella ESPOSITO 143 Chapter II we compare the two organic cationic dyes, NBAP looks to be more efficient than RhP in shifting the N:C balance towards N, which definitely agrees with our previsions: 0.25MC of NBAP is sufficient to produce the same N:C balance shift as 1MC of RhP. Washing shifts N:C balance further towards N if ethanol removes a significant amount of exceeding surfactant, which may contain not only the free fluorescent molecules, but also some residual MT2EtOH molecules physisorbed on silicate layers, as previously demonstrated by TGA (amount of physisorbed MT2EtOH molecules in commercial C30B 21% wt, Figure II-F6d). Interestingly, the clay photo-functionalized by cation exchange processing (D) with 0.25MC RhP is the only sample showing no significant N:C variation before and after washing with ethanol: this means that no significant excess is removed, thus the photo-functionalization has been performed in its optimum conditions (minimum concentration of the most efficient fluorescent molecule). About N:Si variation: it is a positive value for all the samples and diminishes without zeroing after washing (a global zero N:Si variation would have meant that the whole photofunctionalization process hasn’t produced any modification of the chemical composition of the pristine C30B clay). II-7 CONCLUSIONS The goal of the work presented in this section was to obtain photo-functional (in other words, photo-active) inorganic-organic complexes from a commercially available organoclay (Cloisite ® 30B) with the perspective of using them for real-time process monitoring of polymer-based nanocomposites containing lamellar fillers (essentially clays) by means of Visiovis, a brand new experimental equipment we’re developing in our laboratory. Several photo-functionalization methods have been tested with different concentrations of few fluorescent molecules. All the experimental evidences collected by XRD, TGA, EA, FTIR spectroscopy, TGA-FTIR and fluorescence spectroscopy indicate that the best photo-functional inorganic-organic complexes can be obtained by performing a classical cation exchange process of C30B with 0.25MC RhP (Rhodamine 6G Perchlorate) in a water/ethanol mixture at 80°C, followed by washing with ethanol, recovering by centrifugation and drying at room temperature under exhaust hood. TGA PhD INSA de Lyon (2008) 144 PHOTO-FUNCTIONALIZATION Lamellar fillers showed that RhP is the most efficient organic cationic dye: photo-functionalization with RhP appears to be readily detectable by both TGA and XRD. In addition, RhP-based complexes have been proved to absorb and emit fluorescence in a wavelength range which is more interesting for the application we’re developing in our laboratories – which will be the topic of future works and papers but has recently been overviewed in a technical journal [29]. Spectrofluorimetry indicates that, whatever the initial concentration of RhP in the exchanging medium, cations are mostly adsorbed into clay galleries in the form of monomers and/or fluorescent J-aggregates (probably dimers) since the presence of the MT2EtOH molecules seems to limit the tendency of RhP molecules to aggregate and probably induces some spatial regularity. Clay photofunctionalization at higher concentrations of fluorescent molecule is still possible (no significant fluorescence quenching occurs, as confirmed by spectrofluorimetry for C30B photo-functionalized 1MC RhP) but unnecessary: most of the fluorophore excess is washed away and wasted. Cation exchange process with a smaller amount of organic cationic dye (e.g. a relatively small percentage of the initial modifier concentration MC) is sufficient to render C30B organoclay photo-active. Indeed, any excess of fluorescent dye preferentially adsorbs at the silicate edges in a disordered configuration and likely creates a sterical barrier to further intercalation: these are the reasons why we preferred to limit the excess of fluorescent dye to 0.25 times the initial modifier concentration – which completely fulfils the requirements of our study. Finally, XRD showed that the intercalation of smaller molecules (RhP cations) into organically-modified clay galleries likely produces a rearrangement of the paraffinic configuration of previously-exchanged bigger molecules (MT2EtOH cations) and increase the tilting angle from 37° up to 49°. Antonella ESPOSITO 145 Chapter II II-R [1] REFERENCES Schmidt D.F. Polysiloxane/layered silicate nanocomposites: synthesis, characterization, and properties. PhD Thesis. Itaka, NY: Cornell University, 2003, 300 p. [2] Burgentzlé D., Duchet J., Gérard J.F. et al. Solvent-based nanocomposite coatings I. Dispersion of organophilic montmorillonite in organic solvents. J. Colloid Interface Sci. 2004, 278, 26-39. [3] Le Pluart L., Duchet J., Sautereau H. et al. Surface modifications of montmorillonite for tailored interfaces in nanocomposites. J. Adhes. 2002, 78 (7), 645-662. [4] Xie W., Gao Z., Pan W.-P. et al. Thermal degradation chemistry of alkyl quaternary ammonium montmorillonite. Chem. Mater. 2001, 13, 2979-2990. [5] Bellucci F., Camino G., Frache A. et al. Catalytic charring – volatilization competition in organoclay nanocomposites. Polym. Degrad. Stab. 2007, 92(3), 425-436. [6] Davis R.D., Gilman J.W., Sutto T.E. et al. Improved thermal stability of organically modified layered silicates. Clays Clay Miner. 2004, 52 (2), 171-179. [7] Xi, Y.; Frost, R.L.; He, H. Modification of the surfaces of Wyoming montmorillonite by the cationic surfactants alkyl trimethyl, dialkyl dimethyl, and trialkyl methyl ammonium bromides. J. Colloid Interface Sci. 2007, 305, 150-158. [8] Shen Z. Nanocomposites of polymers and layered silicates. PhD Thesis. Australia: Monash University, 2000, 308 p. [9] Wilkie C.A. TGA/FTIR: an extremely useful technique for studying polymer degradation. Polym. Degrad. Stab. 1999, 66, 301-306. [10] Cervantes-Uc J.M., Cauich-Rodríguez J.V., Vázquez-Torres H. et al. Thermal degradation of commercially available organoclays studied by TGAFTIR. Thermochim. Acta 2007, 457, 92-102. [11] Alexandre M., Dubois P. Polymer-layered silicate nanocomposites: preparation, properties and uses of a new class of materials. Mater. Sci. Eng. 2000, 28, 1-63. [12] Pozsgay A., Fráter T., Százdi L. et al. Gallery structure and exfoliation of organophilized montmorillonite: effect on composite properties. Eur. Polym. J. 2004, 40, 27-36. [13] Gates W.P. Crystalline swelling of organo-modified clays in ethanol-water solutions. Appl. Clay Sci. 2004, 27, 1-12. PhD INSA de Lyon (2008) 146 PHOTO-FUNCTIONALIZATION [14] Lamellar fillers Endo T., Sato T., Shimada M. Fluorescence properties of the dye-intercalated smectite. J. Phys. Chem. Solids 1986, 47 (8), 799-804. [15] Fujita T., Iyi N., Kosugi T. et al. Intercalation characteristics of rhodamine 6G in fluortaeniolite: orientation in the gallery. Clays Clay Miner. 1997, 45 (1), 77-84. [16] Iwasaki M., Kita M., Ito K. et al. Intercalation characteristics of 1,1'-diethyl-2,2'cyanine and other cationic dyes in synthetic saponite: orientation in the interlayer. Clays Clay Miner. 2000, 48 (3) 392-399. [17] Čapková P., Malý P., Pospíšil M. et al. Effect of surface and interlayer structure on the fluorescence of rhodamine B-montmorillonite: modeling and experiment. J. Colloid Interface Sci. 2004, 277, 128-137. [18] Schmidt D.F., Clément F., Giannelis E.P. On the origins of silicate dispersion in polysiloxane/layered-silicate nanocomposites. Adv. Funct. Mater. 2006, 16, 417-425. [19] Lagaly G. Characterization of clays by organic compounds. Clay Miner. 1981, 16, 1-21. [20] Bergaya F., Lagaly G. Surface modification of clay minerals. Appl. Clay Sci. 2001, 19, 1-3. [21] Cowan C.T., White D. Adsorption by organo-clay complexes. Trans. Faraday Soc. 1958, 54, 691-697. [22] Maupin P.H., Gilman J.W., Harris R.H. (Jr.) et al. Optical probes for monitoring intercalation and exfoliation in melt-processed polymer nanocomposites. Macromol. Rapid Commun. 2004, 25, 788-792. [23] Bur A.J., Roth S.C., Start P.R. et al. Fluorescent probes for monitoring microstructure of polymer-clay nanocomposites. Proceedings of the Society of Plastics Engineers: Annual Technical Conference. ANTEC 2004, Vol.1 (Processing), 1315-1318. [24] Gilman J.W., Maupin P.H., Harris R.H. (Jr.) et al. High throughput methods for nanocomposite materials research. Extrusion and visible optical probes. Polym. Mater. Sci. Eng. 2004, 90, 717-718. [25] Pospíšil M., Čapková P., Weissmannová H. et al. Structure analysis of montmorillonite intercalated with rhodamineB: modeling and experiment. J. Mol. Model. 2003, 9, 39-46. [26] Cervantes-Uc J.M., Cauich-Rodríguez J.V., Vázquez-Torres H. et al. Thermal degradation of commercially available organoclays studied by TGA-FTIR. Thermochim Acta 2007, 457, 92-102. [27] Edwards G., Halley P., Kerven G. et al. Thermal stability analysis of organo-silicates, using solid phase microextraction techniques. Thermochim. Acta 2005, 429, 13-18. Antonella ESPOSITO 147 Chapter II [28] He H., Duchet J., Galy J. et al. Influence of cationic surfactant removal on the thermal stability of organoclays. J. Colloid Interface Sci. 2006, 295, 202-208. [29] Esposito A, Balcaen J, Duchet-Rumeau J, Charmeau JY. Visiovis: monitoring nanofiller dispersion/distribution in molten polymers. JEC Composites Magazine 2008, 41, 67-71. [30] Madejová J. FTIR techniques in clay mineral studies. Vib. Spectrosc. 2003, 31, 1-10. PhD INSA de Lyon (2008) 148 Chapter III PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing A great portion of this chapter corresponds to the content of a paper recently submitted for publication. The aim of photo-functionalization is to graft/adsorb a fluorescent molecule onto clay silicate layers or at their edges, or to introduce it into clay galleries. In the previous section1 we reported all the details about the set of methods we tested in order to photofunctionalize a given organically-modified clay (Cloisite ® 30B) and we concluded that the most practicable, appropriate and efficient method is cation exchange processing in a solution 90/10 permuted water/ethanol at 80°C containing a certain amount of the organic cationic dye of interest [1]. Then we performed cation exchange processing of the same organically-modified clay with two different organic cationic dyes (Nile Blue A Perchlorate, Rhodamine 6G Perchlorate) introduced in two different concentrations in the exchanging medium (i.e. 1MC and 0.25MC) in order to determine the most efficient fluorescent molecule and its optimum concentration (Rhodamine 6G Perchlorate in a concentration equivalent to 25% of the initial modifier concentration, i.e. 0.25MC) [1]. 1 Chapter II (PHOTO-FUNCTIONALIZATION – Lamellar fillers). Antonella ESPOSITO 149 Chapter III In this section we’ll report the results obtained using the procedure previously optimised to prepare other photo-functional inorganic-organic complexes from four commercially available clays (Cloisite ® Na+, Cloisite ® 30B, Cloisite ® 10A, Cloisite ® 15A). The results obtained with Cloisite ® 30B have been discussed in the previous section: we’ll recall them just to make the comparison with the other commercial clays easier. Among the selected commercial clays we included also Cloisite ® Na+ since the same photo-functionalization method performed in the absence of organics may help understanding what’s going on when clay galleries already enfold an organic surfactant: in other words, we selected Cloisite ® Na+ to get a kind of reference for the other clays. III-1 MATERIALS One sodium montmorillonite (Cloisite ® Na+) and three organically-modified clays (Cloisite ® 30B, Cloisite ® 10A and Cloisite ® 15A) have been purchased from Southern Clay Products (USA) and used as received. MMT-Na+ (CNa+) is a natural montmorillonite containing 4-9% ca of moisture and no traces of organic modifiers. MMT-MT2EtOH (C30B) is a natural montmorillonite organically modified with MT2EtOH (methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium chloride) in an initial Modifier Concentration (MC) of 90 meq/100g of clay. MMT-2MBHT (C10A) is a natural montmorillonite organically modified with 2MBHT (dimethyl, benzyl, hydrogenated tallow, quaternary ammonium chloride) in an initial MC of 125 meq/100g of clay. MMT-2M2HT (C15A) is a natural montmorillonite organically modified with 2M2HT (dimethyl, dehydrogenated tallow, quaternary ammonium chloride) in an initial MC of 125 meq/100g of clay. Both tallow (T) and hydrogenated tallow (HT) fatty chains have the following composition: 65% C18, 30% C16 and 5% C14. The selected clays have been photo-functionalized with the organic cationic dye previously chosen, (Rhodamine 6G Perchlorate, Sigma Aldrich). The raw materials and the corresponding relevant chemicals, as well as their molecular weight Mw and their most stable free molecular configuration obtained by energy minimization in a molecular modeling environment (Materials Studio, Accelrys Software Inc., Module Discover), are listed in Table III-T1. Permuted water has been ion-exchanged right before using it, assuring a PhD INSA de Lyon (2008) 150 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing quality of 18 M -cm. Ethanol has been purchased from Sigma-Aldrich and used as received. Table III-T1 Raw materials and the corresponding relevant chemicals CNa+ MC TRADE or IUPAC NAME CHEMICAL meq 100g MW Most stable configuration in the free state a Cloisite ® Na+ MMT-Na+ – 90 b n.a. n.a. Cloisite ® 30B 90 C10A Cloisite ® 10A MMT-2MBHT 125 C15A Cloisite ® 15A MMT-2M2HT 125 RhP Rhodamine 6G Perchlorate 184 d C30B MMT-MT2EtOH 360.80 average c 382.80 average c 527.60 average c 543.01 a The most stable molecular configuration of each chemical in its free state has been determined by minimizing its energy in a molecular modeling environment (Materials Studio v.4.1.0.0, Accelrys Software – Module Discover – Smart Minimizer, medium convergence level, 5000 max interactions). b In the case of CNa+, we report (and we’ve worked referring to) the typical value of Cation Exchange Capacity (CEC) of a natural sodium MMT (90 meq/100gr of clay). c MT2EtOH, 2MBHT and 2M2HT include tallow (T) or hydrogenated (i.e. saturated) tallow (HT) fatty chains having the following composition: 65% C18, 30% C16, 5% C14. d This value has been calculated by considering that the whole amount of cationic organic dye is able to replace clay inorganic cations [1]. Antonella ESPOSITO 151 Chapter III III-2 PHOTO-FUNCTIONALIZATION PROTOCOL An efficient photo-functionalization protocol and its optimum conditions have been established in the previous section of this work: to understand the reasons of our choices about photo-functionalization protocol, we recommend the lecture of Chapter II (PHOTO-FUNCTIONALIZATION – Lamellar fillers). Here we just remind that such protocol (detailed in § II-2.4) was originally designed to exchange organically-modified clays – which is the reason why we referred to the initial Modifier Concentration (MC) in order to determine an optimum concentration of Rhodamine 6G Perchlorate. In the following, however, we are going to use the same protocol to photo-functionalize Cloisite ® Na+, a clay which is completely inorganic and doesn’t contain any modifier in its pristine state: in this case, we’ll simply refer to its Cation Exchange Capacity (CEC) – therefore, the optimum concentration of fluorescent molecule will be 0.25CEC rather than 0.25MC. III-3 CHARACTERIZATIONS Analogously to the criterions used for sample characterizations in the previous section, here again characterizations are mainly aimed to verify whether the fluorescent molecules adsorbed or not on clay surfaces, and where they are located. Of course, once again we tested the photo-activity of the samples by spectrofluorimetry. III-4 REFERENCE MEASUREMENTS All the pristine commercial clays (CNa+, C30B, C10A, C15A) as well as the free organic cationic dye (RhP) have been firstly characterized by XRD, TGA, EA, FTIR and spectrofluorimetry in order to get reference measurements. III-4.1 Reference XRD measurements In the previous section2 we reported that RhP shows two marked peaks at 9.8 Å and 12 Å and we underlined that its peaks couldn’t affect the interpretation of any XRD measurement performed on the photo-functionalized C30B – which has an initial interlayer spacing of 17.5 Å. This is factual also for C10A and C15A, which show an 2 See § II-4.1. PhD INSA de Lyon (2008) 152 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing initial interlayer spacing of 19.2 and 32 Å respectively. On the contrary, a particular attention will be paid to the interpretation of XRD results obtained for the photofunctionalized CNa+, since the initial interlayer spacing of this hydrophilic clay is 12.7 Å (Figure III-F1). In order to exclude any misunderstanding with the interpretation of XRD characterizations for the washed photo-functional inorganic-organic complexes, we judged necessary to include, in the panel of the reference XRD measurements, the complete set of characterizations for the selected commercial clays (CNa+, C30B, C10A and C15A) after one step of the same washing procedure employed for the photofunctionalized clays, i.e. rapid immersion in ethanol, manual stirring, recovering by centrifugation and drying under exhaust hood for several days at room temperature. Figure III-F1 Reference XRD measurements of the selected commercial clays: (a) CNa+, (b) C30B, (c) C10A, (d) C15A. Bold lines represent the samples after washing. This single washing step performed on the pristine clays allows to take into account the effects due to the immersion in the exchanging medium, independently Antonella ESPOSITO 153 Chapter III from the effects due to the organic cationic dye. After the washing procedure, the commercial clays have an interlayer spacing of 12.4 Å (CNa+), 17.5 Å (C30B), 19.2 Å (C10A) and 25.7 Å (C15A) (bold lines in Figure III-F1): it is interesting to observe that ethanol has no influence on the interlayer spacing of the hydrophilic clay and, with reference to the organophilic MMTs, it doesn’t modify the molecular arrangement of the surfactant into the galleries but in the case of C15A ( d001 = 6.3 Å). These results will be crucial for the interpretation of any other XRD measurement. XRD reference measurements for the commercial clays before and after washing are shown in Figure III-F1. The XRD signature of RhP has been shown in Chapter II (Figure II-F5c) [1]. III-4.2 Reference TGA measurements TGA (weight vs. temperature) and DTG (weight derivative vs. temperature) curves for the RhP have been commented in Chapter II [1]: the initial sample weight halves in the temperature range 200-500°C and the derivative curve shows a single sharp peak centered at 338°C (Figure II-F6c). CNa+ contains no traces of organics thus its thermal characterization produces nearly flat TGA and DTG curves (Figure III-F2a), as expected: the main weight loss would have been the dehydroxilation process, viz. the loss of structural water from the crystalline lattice at higher temperature. We previously reported3 that pristine C30B undergoes two weight losses between 150°C and 500°C, the first one (21% at 253°C) partially corresponding to physisorbed MT2EtOH and the second one (30% at 360°C) corresponding to intercalated MT2EtOH (Figure III-F2b). DTG curve for C30B after washing with ethanol helped to realize that the first weight loss included also a portion of MT2EtOH molecules well intercalated but in a peripheral position with respect to the clay gallery (15% of the whole amount of intercalated surfactant). Such observation is also factual for C10A clay (Figure III-F2c) which, in its pristine state, undergoes a first weight loss (13% of sample weight = 37% of surfactant, at 207°C) corresponding to a peak considerably (not completely) removed by washing with ethanol (bold line). After this first weight loss, C10A clay undergoes a second weight loss (10% of sample weight = 26% of surfactant, at 285°C) – absolutely washing 3 See § II-4.2 and [1]. PhD INSA de Lyon (2008) 154 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing insensitive – and a third weight loss (12% of sample weight = 36% of surfactant, at 368°C), which slightly changes upon washing with ethanol, thus indicating a probable rearrangement of the surfactant into the clay galleries. This last peak (408°C), better revealed by washing, corresponds most likely to the aromatic portion of the intercalated molecules, since it is well known that the presence of aromatics can produce additional peaks at higher temperatures in DTG curves [1]. (aw) (bw) Figure III-F2 Reference TGA measurements of the selected commercial clays: (a) CNa+, (b) C30B, (c) C10A, (d) C15A. Bold lines represent the samples after washing with ethanol. Finally, C15A (Figure III-F2d) shows a complex thermal behavior in its pristine state, since its DTG curve presents a broad peak in the temperature range 150-450°C – which probably result from the convolution of at least four peaks, as determined by a multi-peaks fit, giving four shallow humps centered at the following temperature: 252°C, 312°C, 343°C and 413°C. The DTG curve gets much simpler after washing, since removing the excess of surfactant (indeed only 6% of sample weight, as visualized by the different weight loss at 500°C for C15A before and after washing, i.e. 41% (bw) Antonella ESPOSITO 155 Chapter III minus 35% (aw), as shown in Figure III-F2d) allows to better visualize the two weight losses (23% at 312°C, 12% at 401°C) corresponding to intercalated 2M2HT molecules. In the case of C15A clay, no significant “edge effects” seem to affect the intercalated 2M2HT molecules: the first peak (252°C) is almost completely removed by washing, and even a deeper analysis performed by multi-peak fitting and integration of the peak area confirmed that it entirely corresponds to the excess of surfactant physisorbed at the platelet edges or on the external clay surfaces (5.4% ca). It is interesting to observe that the “edge effects” – as visualized by TGA characterizations – seem to depend on the molecular weight of the surfactant molecules: anyway, this doesn’t mean that bigger molecules are better confined in between clay platelets. It simply means that the residual peak assigned to the peripherally-intercalated surfactant molecules is, in the DTG curve of a washed sample, better masked by the main intercalation peak if the surfactant molecules have a higher molecular weight – which actually corresponds to a higher mass of intercalated organic and, thus, to a higher relative weight loss rate. III-4.3 Reference EA measurements Reference EA values are resumed in Table III-T2. The Si % content detected in each of the pristine clays has been chosen as the reference value to get the Normalization Factor (NF) for any other EA measurement performed on the photofunctionalized clay samples, whether washed or not. As expected, the chemical composition of the pristine commercial clays follows some clear trends. Na % content is relatively high in pristine (unwashed) CNa+ and diminishes in the commercial organically-modified clays because of the substitution of Na+ for N+ cations (N % content follows the opposite trend). C % content is not significant in CNa+ (agreeing with the fact that CNa+ contains no traces of organics) but dramatically increases in the commercial organically-modified clays, following a trend which is coherent with the molar weight of the respective surfactants molecules, and which is obviously opposite to the trend of the relative Si % content. The effects of washing with ethanol are well recognizable by comparing the C % content of the commercial organoclays before (bw) and after washing (aw). PhD INSA de Lyon (2008) 156 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing Table III-T2 Reference EA measurements: chemical composition of the raw materials before (bw) and after washing (aw). Na [% wt] bw CNa + C30B C10A C15A RhP aw 3.08 0.23 0.21 780 ‰ 0.00 C [% wt] bw aw N [% wt] bw aw 2.45 < 0.30 < 0.30 < 0.10 < 0.10 < 0.10 20.18 15.36 1.07 0.88 0.13 27.64 21.56 1.20 0.96 0.40 32.64 27.58 0.91 0.82 b b n.a. 61.90 n.a. 5.20 n.a. H [% wt] Si [% wt] a bw aw bw aw 1.41 3.98 4.86 6.31 5.71 c < 0.30 3.32 3.79 5.28 n.a. 25.90 21.39 18.99 17.81 0.00 26.44 22.02 20.18 18.91 n.a. a The amount of any element estimated by EA is relative and has to be normalized. In the following, normalization of any value obtained for each photo-functionalized sample will be done in relation to the Si % content detected in the corresponding pristine clay. b Source: Sigma-Aldrich. c Theoretical value calculated on the basis of the chemical formula. III-4.4 Reference FTIR spectra The FTIR spectra of RhP and C30B have already been shown and commented in the previous section (Figure II-F7) [1]. However, we propose once more the spectrum of C30B, so that it will be easier to make the comparison with the other clays. The FTIR spectra of the selected commercial clays (Figure III-F3) share some features which are common to any MMT clay – whether organically-modified or not: a broad absorption band in the region 3650-3200 cm-1, assigned to H-bonded OH stretching, mostly due to the intrinsic structure of clays (hydroxyl groups) and to the structural water (which hydrates both the exchangeable interlayer cations and the immobilized cations); a strong peak at 1050 cm-1, assigned to the Si-O-Si vibrations, explained by the chemical composition of MMT clay (basically a framework of crystalline silicates); a set of three peaks in the region 670-400 cm-1, assigned to all the possible vibrational modes in the inorganic crystalline lattice. The organically-modified clays (C30B, C10A and C15A) additionally present a set of two peaks in the region 3000-2800 cm-1 (approximately 2930 and 2860 cm-1) which is typical of compounds containing long linear aliphatic chains – MT2EtOH (C30B, Figure III-F3b), 2MBHT (C10A, Figure III-F3c) and 2M2HT (C15A, Figure Antonella ESPOSITO 157 Chapter III III-F3d) do contain one tallow (or hydrogenated tallow) chain 4. This is confirmed by the sharp but weak peak around 1470 cm-1. Figure III-F3 Reference FTIR measurements for the pristine clays in KBr pellets: (a) CNa+, (b) C30B, (c) C10A and (d) C15A. The weak peak around 1635 cm-1 is due to the structural water, always present in natural clays. The narrower peak centred around 3630 cm-1 and present in all the spectra but the one of C15A (Figure III-F3d) may be explained once again by OH stretching (non-H-bonded OH groups), often corresponds to an alcohol with a sterically-hindered OH group (which may be the case of C30B, Figure III-F3b) and could indicate, in some inorganics and minerals, the presence of “free” OH groups either on the surface or within the crystalline lattice. A summary of the peaks is given in Table III-T3. 4 The chemical formula of MT2EtOH, 2MBHT and 2M2HT are shown in Table III-T1. PhD INSA de Lyon (2008) 158 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing Table III-T3 Reference FTIR data for the commercial clays in their pristine state: positions and probable assignment a of the absorption bands (spectra in Figure III-F3). Position [cm-1] CNa+ C30B C10A C15A (a) (b) (c) (d) ~ 3633 3445 3436 3429 3417 3065 and 3035 2928 2854 2364 and 2339 1637 or 1630 1473 or 1468 1417 OH stretching of structural hydroxyl groups (non H-bonded and “free” OH groups) and/or OH stretch of primary alcohols OH stretching of structural hydroxyl groups (Hbonded OH groups) and/or OH stretching of Hbonded water (overlapping asymmetric and symmetric H-O-H) C-H vibrations (together with 1650 cm-1) C-H stretching of alkylammonium cations ( asymmetric CH2) C-H stretching of alkylammonium cations ( symmetric CH2) Atmospheric CO2 OH deformation of water b as(C-H) bending of methylene groups in (CH3)3N+(CH2)nCH3 cations and/or C=C-C stretching of aromatic rings 885 s((CH3)-N) bending of methyl groups C-H in-plane bending of aromatic rings and/or CN stretching of tertiary amines Si-O-Si stretching (longitudinal mode) Si-O-Si in-plane stretching and/or C-N stretching of primary amine Al2OH deformation and/or OH deformation of inner hydroxyl groups AlFeOH deformation 849 AlMgOH deformation 801 Si-O-Si stretching of quartz and silica 726 703 –(CH2)n– rocking (n Si-O-Si vibrations 625 Coupled Al-O and Si-O out-of-plane vibrations 525 Al-O-Si (octahedral Al) deformation 466 Si-O-Si bending and deformation 1216 or 1203 1118 1046 918 a Probable assignments 3) of methylene groups Assignment has been effectuated on the basis of Madejová et al. [2][3] and Coates [4] observations. b This peak may also be explained by some olefinic unsaturation, usually corresponding to a relatively narrow weak-to-moderate peak at 1650 cm-1, whose frequency lowers upon conjugation with another double bond or an aromatic ring [4]. Antonella ESPOSITO 159 Chapter III III-4.5 Reference fluorescence spectra It is well known that the maximum absorption for RhP is around 532 nm: the absorption and fluorescence emission spectra for RhP in ethanol have been reported in the previous section5 [1]. Unsurprisingly, none of the selected commercial clays absorbs radiation and produces a fluorescence emission in its pristine state (results not shown). III-5 CHARACTERIZATION OF THE PHOTO-FUNCTIONAL INORGANIC/ORGANIC COMPLEXES In the previous section we could realize that XRD and TGA effectively represent a set of powerful techniques, which can be easily employed in a complementary way to characterize natural and organically-modified clays. Therefore, we continue using XRD coupled to TGA measurements in order to characterize the photo-functional inorganicorganic complexes obtained with the same protocol previously used for Cloisite ® 30B. III-5.1 Photo-responsive CNa+ 0.25CEC RhP The double set of characterizations for the photo-functional inorganic-organic complex CNa+ 0.25MC RhP before and after washing with ethanol are shown in Figure III-F4. The XRD pattern (Figure III-F4a) indicates that the organic cationic dye has entered and swollen clay galleries, since the d001 value has considerably increased, reaching a value of 17.2 Å ( d001 = +4.8 Å). This result is entirely due to intercalated and properly exchanged molecules, since clay washing with ethanol doesn’t affect the measured d001 value. TGA curves (Figure III-F4b) confirm that the photo-responsive CNa+ complexes are absolutely washing insensitive: the whole amount of fluorescent molecule used for the photo-functionalization process (25% of the CEC of the pristine clay) entered clay galleries since no excess got physisorbed on clay platelets (no weight losses occur up to 300°C). 5 See § II-4.5. PhD INSA de Lyon (2008) 160 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing Figure III-F4 Characterizations of the photo-functional inorganic-organic complex CNa+ 0.25CEC RhP, before and after washing with ethanol. XRD (a) is substantiated by TGA results (b). Bold lines represent washed samples. III-5.2 Photo-responsive C30B 0.25MC RhP XRD and TGA results for the photo-functional inorganic-organic complex C30B 0.25MC RhP before and after washing with ethanol have already been reported and commented in the previous section6 and in a submitted paper [1]. Here we just remind that, in agreement with some other results found in the literature [5][6], right after the photo-functionalization we measured an interlayer spacing d001 = 22 Å ( d001 = +4.5 Å), which resulted to be affected neither by the RhP concentration in the exchanging medium, nor by the following clay washing (Figure III-F5a). We also observed that cation exchange processing of C30B with 0.25MC RhP produced a peak in the area of physisorbed chemicals (243°C) (Figure III-F5b), which is lowered by washing but 6 § II-6.2. Antonella ESPOSITO 161 Chapter III doesn’t completely disappear because it partially corresponds to 2M2EtOH molecules properly exchanged but intercalated in a peripheral position, therefore less protected from thermal degradation. In spite of its maximum temperature (315°C), we could explain the second peak in a similar way, since its behavior in relation to washing is analogous. We finally concluded that the last weight loss (406°C) corresponds to the peak assigned to intercalated chemicals, since it is washing insensitive. Figure III-F5 Characterizations of the photo-functional inorganic-organic complex C30B 0.25MC RhP, before and after washing with ethanol. XRD measurements (a) are substantiated by TGA results (b). Bold lines represent washed samples. III-5.3 Photo-responsive C10A 0.25MC RhP XRD and TGA results for the photo-functional inorganic-organic complex C10A 0.25MC RhP before and after washing with ethanol are shown in Figure III-F6. The photo-functionalization process of C10A with 0.25MC RhP produces, in comparison with the TGA results for the pristine clay, an additional peak in the area of physisorbed PhD INSA de Lyon (2008) 162 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing chemicals (194°C), which is efficaciously removed by washing and surely corresponds to an excess of RhP molecules – since it has never been observed in the pristine organoclay, whether before or after one-step washing (Figure III-F2c). The following peak (235°C) is lowered by washing but it doesn’t completely disappear, analogously to the peak observed at 207°C (and lowered to 225°C by washing) for the pristine C10A (Figure III-F2c). This analogy confirms that such peak must be assigned to the 2MBHT molecules properly exchanged but intercalated in a peripheral position. Figure III-F6 Characterizations of the photo-functional inorganic-organic complex C10A 0.25MC RhP, before and after washing with ethanol. XRD measurements (a) are substantiated by TGA results (b). Bold lines represent washed samples. The peak at 292°C (Figure III-F6b) is also considerably lowered by washing and it reasonably corresponds to the peak already observed in the pristine organoclay at 285°C (Figure III-F2c) and previously assigned to the first weight loss of intercalated chemicals. The main difference is that the peak for the photo-functional complex isn’t Antonella ESPOSITO 163 Chapter III exactly shaped as the peak of the pristine C10A – it rather shows a certain tendency to split up in a double peak after washing. By the way, on the basis of a simple analogy with the peak observed in the pristine C10A, one could hypothesize that a significant portion of the RhP molecules entered clay galleries but wasn’t able to undergo cation exchange process, probably because of the sterical hindrance effect due to 2MBHT molecules. Indeed, Paul et al. [7] showed that the absence of OH groups in the molecular structure of the surfactant reduces interlayer packing and molecular density, leaving more vacancies which could be occupied by smaller molecules (RhP molecules, in this case). However, we haven’t observed any shift of the XRD peak (22.6 Å before and after clay washing as shown in Figure III-F6a, d001 = +3.4 Å with respect to the pristine clay), however this would have been the sole evidence confirming that some molecules entered clay galleries without undergoing cation exchange (we remind that an efficient washing procedure is supposed to remove any unexchanged excess, whether unconfined or confined, and that removing a confined excess usually causes a decrease of the interlayer spacing) [1]. With the help of the evidences collected after the photofunctionalization we can conclude that the previous assignment of the peak at 285°C for the pristine C10A is probably inexact and should be rather assigned exclusively to the aliphatic portions of the surfactant molecules properly exchanged but in a peripheral position, i.e. partially confined. Consequently, the last peak at 374°C (Figure III-F6b), which looks less finely-shaped than the corresponding peak in the pristine organoclay (Figure III-F2c) probably because of a higher complexity of the molecular arrangement into clay galleries, is the peak corresponding to intercalated aromatic chemicals. III-5.4 Photo-responsive C15A 0.25MC RhP XRD and TGA results for the photo-functional inorganic-organic complex C15A 0.25MC RhP before and after washing with ethanol are shown in Figure III-F7. The photo-functionalization of C15A seems to produce the most complex XRD patterns, with two populations of crystallites having different interlayer spacing before washing and a single population having an interlayer spacing d001 = 25 Å after washing (Figure III-F7b). Indeed, these results shouldn’t surprise, since the cation exchange process has been performed on the pristine commercial clays in their as-received state (i.e. without PhD INSA de Lyon (2008) 164 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing further purification and/or washing): we’ve already shown that even a single-step washing procedure with ethanol may considerably change the molecular arrangement of the surfactant into clay galleries by removing any excess (whether unconfined or confined), which sometimes correspond to a significant change of the XRD pattern as highlighted in the case of the pristine C15A (Figure III-F1d). Thus, it isn’t surprising that the photo-functionalization process increased the interlayer spacing of both the crystalline phases detected in the pristine commercial organoclay, shifting the first peak from d001 = 32 Å to 37 Å ( d001 = +5 Å) and the second peak from d001 = 20.3 Å to 22 Å ( d001 = +1.7 Å), the latter value (22 Å) being in agreement with the d001 values obtained with C30B (Figure III-F5a) and C10A (Figure III-F6a). Figure III-F7 Characterizations of the photo-functional inorganic-organic complex C15A 0.25MC RhP, before and after washing with ethanol. XRD measurements (a) are substantiated by TGA results (b). Bold lines represent washed samples. Antonella ESPOSITO 165 Chapter III Apart from these observations, it is noteworthy that the unmodified but washed C15A showed an initial value d001 = 25.7 Å (Figure III-F1d, bold line) and that such value hasn’t really changed upon the cation exchange process followed by washing (d001 = 25 Å, bold line in Figure III-F7a). It is in such kind of situations that complementary characterization techniques (namely TGA) become useful to better understand what’s going on. Figure III-F7b shows that the photo-functionalization of C15A with 0.25MC RhP produces, in comparison with the TGA results for the pristine C15A (Figure IIIF2d), a strong additional peak at 285°C in the area of physisorbed chemicals. Such peak decreases upon washing and surely corresponds to an excess of RhP molecules. After washing, the shape of the DTG curve (bold line in Figure III-F7b) looks quite complex and in any case different from the corresponding curve for the pristine C15A after the single-step washing procedure (bold line in Figure III-F2d). The sample undergoes four weight losses whose peaks can’t be easily resolved. Photo-functionalized C15A presents such a complex structure that neither XRD nor TGA can univocally detect whether the fluorescent molecules are intercalated or not into the clay galleries and, consequently, none of these techniques can explain the molecular configuration of the surfactant and of the cationic dye into the sample. Spectrofluorimetry will prove that all the inorganicorganic complexes we obtained are effectively photo-active and will likely inform about the molecular configuration of the photo-functionalized clays – at least of each complex compared to the others. III-5.5 Comparison of the photo-responsive complexes As previously observed in Chapter II, EA completes the set of characterizations which can be made on functionalized clays. Indeed, this technique allows estimating the relative amount of any specific element (but the oxygen) present in the sample. By performing EA of each photo-functionalized sample before and after washing, and by comparing these results to the ones obtained for the relative pristine commercial clay, it is possible to get some other information about the effects of the photo-functionalization and those of the washing procedure. The limiting factor is that EA can give only relative chemical compositions, meaning that the amount of any detected element has to be PhD INSA de Lyon (2008) 166 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing “normalized”: normalization of EA data has been done with reference to the Si % wt content detected in each of the pristine commercial clays. Table III-T4 EA measurements: chemical composition of the photo-active inorganic-organic complexes (0.25MC RhP) compared to the composition of the pristine clays. Na a [% wt] pristine + RhP + C a [% wt] pristine + RhP 2.45 0.84 < 0.30 C30B < 0.10 ≤ 0.5 ‰ 15.36 C10A 0.13 < 0.1 ‰ 21.56 C15A 0.40 < 0.1 ‰ 27.58 CNa 9.07 19.25 24.23 27.73 N a [% wt] pristine < 0.10 0.88 0.96 0.82 H a [% wt] Si [% wt] + RhP pristine + RhP pristine + RhP 0.60 1.24 1.34 1.13 < 0.30 3.32 3.79 5.28 1.80 3.32 3.70 4.86 26.44 22.02 20.18 18.91 23.68 22.12 20.71 19.54 a All the values are normalized on the basis of a Norm Factor (NF), obtained by comparing the Si % content of each washed sample with the Si % content of the corresponding pristine clay after washing: NF Si % wt sample ( aw) Si % wt pristine ( aw) The chemical composition of all the photo-active inorganic-organic complexes, compared to the composition of each of the pristine commercial clays after washing with ethanol, is listed in Table III-T4. Once again, EA values follow some clear trends. Na % content further decreases after the second cation exchange process, becoming less than few hundreds parts per million in the washed photo-active complexes (whatever the pristine commercial organoclay). C % content dramatically increases for CNa+, since the photo-functionalization is the very first organic modification of the sodium MMT. By the way, C % content slightly increases also for the organoclays – to an extent which depends on several factors, e.g. the initial surface coverage, the sterical hindrance effect due to the surfactant, its molecular arrangement, the interlayer packing and molecular density into clay galleries – following the expected trend. It is worthy to observe that the variation of C % wt adsorbed by the organoclays seems to be inversely proportional to the molar weight of the surfactant ( C% = 3.89, 2.67 and 0.15 for C30B, C10A and C15A, respectively). Interestingly, N % content increases as well, and not exclusively for CNa+ (which is in all circumstances supposed to show the most remarkable variations): this proves that performing a second exchange process of an organoclay with some organic cationic molecules smaller than the first Antonella ESPOSITO 167 Chapter III organic surfactant allows to better take advantage of the intrinsic CEC value of the organic host (CEC recovery) and to increase the surface coverage. We could even think to estimate the extent of CEC recovery by evaluating such increase of the N % content: for all the organoclays we observe an increase of about 40%, with a slight trend inversely proportional to the molar mass of the surfactant ( N% = 0.36, 0.38 and 0.31 for C30B, C10A and C15A, respectively) meaning that the CEC recovery is probably less efficacious in the presence of highly sterically-hindering surfactant molecules. By comparing the FTIR7 fingerprint of each photo-functional inorganic-organic complex to the spectrum obtained for the corresponding commercial clay in its pristine state, it is possible to detect the presence of any additional chemical and, eventually, its surroundings (i.e. whether it is adsorbed onto silicate layers or it is a free excess). As already highlighted in the previous section8, the intercalation of RhP cations into clay galleries is based on a simple mechanisms of cation exchange reaction, thus the host-guest interactions have a non-covalently-bonded nature: a comparison of the FTIR spectra for the host inorganic structures (pristine commercial clays, Figure III-F3), the guest organic chemical (RhP, Figure II-F7) and the photo-functional inorganicorganic complexes (bold lines in Figure III-F8) should show that the main absorption bands of the pristine clays and those of the organic cationic dye coexist in the FTIR fingerprints of the photo-active clays. Indeed, in the FTIR spectra of the photofunctional inorganic-organic complexes, one can distinguish the features of the pristine commercial clays (thin-lined curves in Figure III-F8) but one can also recognize some absorption bands characteristic of RhP in the region 1800-1400 cm-1. The global shape of the absorption band associated to the O-H stretching doesn’t significantly change, but the relative intensities of the two peaks therein seem to inverse. In the case of C15A 0.25MC RhP (Figure III-F8d), the photo-functionalization followed by washing acts also as a purifying procedure since the inorganic-organic complex shows, in comparison with the pristine commercial clay, a signature in which the features typical of clays (i.e. the absorption bands assigned to Si-O-Si and to the inorganic crystalline lattice) are 7 8 The FTIR spectroscopy technique has been briefly introduced in § II-3.4. See § II-6.3. PhD INSA de Lyon (2008) 168 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing better recognizable. This is not surprising, as we previously showed by XRD (Figure III-F1d), TGA (Figure III-F2d) and EA (Table III-T2) that C15A is the organoclay with the highest excess of surfactant in its pristine commercial state. Of course, the FTIR spectrum collected for CNa+ (Figure III-F8a) is the one in which the effects of the photo-functionalization are the most evident – one can even detect the presence of weak absorption bands due to RhP around 3000 cm-1, which can’t be seen in the spectra of the photo-functionalized organoclays because of the double set of peaks assigned to the long linear aliphatic chains. Figure III-F8 FTIR spectra of the washed photo-active inorganic-organic complexes in KBr pellets: (a) CNa+ 0.25CEC RhP, (b) C30B 0.25MC RhP, (c) C10A 0.25MC RhP and (d) C15A 0.25MC RhP. As already stressed in Chapter I, fluorescent molecules are fickle but versatile tracers, sensitive to environmental changes concerning chemical composition of the host Antonella ESPOSITO 169 Chapter III matrix, atomic arrangement and molecular configuration of the dye molecules, physical confinement, temperature, etc. and rhodamines are known to be a probe well adapted to study heterogeneous systems, thanks to the strong dependence of their absorption and emission spectra on the host material. Indeed, apart from a change in the chemical composition of the environment, the presence of a host matrix may produce a physical confinement of the dye molecules and induce the formation of a particular type of aggregates (monomers, dimers, H-aggregates or J-aggregates) [5]. Figure III-F9 Spectrofluorimetry characterizations (absorption and emission spectra) of the washed photo-functionalized inorganic-organic complexes in ethanol: (a) CNa+ 0.25CEC RhP, (b) C30B 0.25MC RhP, (c) C10A 0.25MC RhP and (d) C15A 0.25MC RhP. Thin lines represent the spectra of pure RhP. Prior to perform spectrofluorimetry, samples have been carefully washed with ethanol in order to detect only the response of intercalated dye molecules. Spectrofluorimetry is threefold useful to characterize photo-functional inorganicorganic complexes, because: PhD INSA de Lyon (2008) 170 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing absorption spectra can inform on the configuration and the supramolecular arrangement eventually induced by the host inorganic structure on the guest organic moieties (fluorophore-fluorophore interactions); fluorescence emission spectra are good revealers of host-guest interactions (which can be expected to produce some shift or reshape of the emission band); fluorescence calibration (intensity of fluorescence emission vs. concentration) may help understanding which kind of interactions, whether fluorophore-fluorophore or host-guest, are prevalent (a balance on the side of the latter could mean that there are less fluorophore-fluorophore interactions, thus preventing fluorescence quenching even at the highest concentrations). Figure III-F9 clearly proves that all the inorganic-organic complexes processed with the described method are photo-active, meaning that photo-functionalization was successful for all the selected commercial clays. The shape of RhP absorption and emission bands doesn’t significantly change upon intercalation of the cations into clay galleries, with the exception of C10A inorganic-organic complexes (Figure III-F9c). Both the absorption and emission spectra of C10A 0.25MC RhP show a slight hypsochromic shift with respect to pure RhP, and the absorption spectrum of the photofunctional complex doesn’t contain any peak indicating the presence of non-fluorescent H-aggregates – contrarily to pure RhP, whose molecules have a strong tendency to form aggregates, as previously reported and commented [1]. The specificity of C10A 0.25MC RhP complexes may be due to the chemistry of C10A surfactant (2MBHT, Table III-T1). 2MBHT contains a benzyl group, which could favorably interact with the polyaromatic heterocyclic structure of RhP cations and favor the absorption of RhP in form of monomers rather than dimers or higher-order aggregates. The fact that the surfactant bears an aromatic group could slightly modify the fluorescence mechanisms also because fluorescence phenomena are characteristically associated to the presence of aromatic compounds. However, such an investigation of the fluorescence mechanisms (absorption of photons and relaxation phenomena) would require the utilization of more specific and advanced analytical tools, as well as a deeper knowledge and some skills that the authors haven’t yet developed. Moreover, a detailed investigation of the physics responsible for the optical behavior (absorption of radiation and fluorescence emission) Antonella ESPOSITO 171 Chapter III of the photo-functional inorganic-organic complexes presented in the previous and the present chapters would be out of target with respect to the main objective of the work, i.e. monitoring distributive and dispersive flow in a transparent screw/barrel system (Visiovis) during polymer-clay melt compounding. III-6 CONCLUSIONS The goal of the work presented in this chapter was to obtain some other photoactive inorganic-organic complexes from commercially available organoclays (Cloisite ® 10A and Cloisite ® 15A) and one natural (unmodified) clay purchased from the same provider (Cloisite ® Na+), using the same photo-functionalization protocol established in the previous chapter and already (successfully) used to prepare photo-active Cloisite ® 30B: cation exchange process with 0.25MC (or 0.25CEC, in the case of natural clay) RhP. The main application we envisioned for these additional photo-active inorganicorganic complexes is always the same (being also the reason why we started looking for a photo-functionalization protocol): the real-time process monitoring of polymer-clay nanocomposites by Visiovis, the equipment we are developing in our laboratory and which is accurately described in Chapter IV [9]. However, there are several potential applications for this class of photo-functional inorganic-organic complexes: in addition to a diversified and extensive use for real-time monitoring of any other process for claybased nanocomposites (we discussed the potentialities of fluorescence techniques in Chapter I), these photo-active clay minerals could be used to produce microlasers (as Vietze et al. [8] reported about zeolite), to trace pollution in soils, to obtain a new class of pigments which, exploiting the same advantages of nanocomposites, could have the same performances of traditional microsized pigments but at much lower contents, etc. In this chapter we reported and commented the results of the characterizations (XRD, TGA, EA, FTIR and fluorescence spectroscopy) done on the processed clay minerals, as well as a rapid comparison of the four photo-functional inorganic-organic complexes currently available for Visiovis experiments. All the collected experimental evidences show that our photo-functionalization protocol with RhP seems to be well adapted to any clay mineral, whether organically-modified or not, and whatever the nature and the amount of the pre-existing surfactant (if present). The comparison of the photo-active PhD INSA de Lyon (2008) 172 PHOTO-FUNCTIONAL COMPLEXES Cation exchange processing Cloisite ® Na+ with the photo-active organoclays witnesses the difficulties encountered to characterize the photo-functional inorganic-organic complexes when some organic surfactant is already present into clay galleries – especially if its excess is considerable, as in the case of Cloisite ® 15A. Indeed, even if the interactions of clay minerals with organic cationic dyes (and, in particular, the adsorption of the dye molecules into clay galleries by cation exchange process) represent a topic developed in the literature since a long time (see Chapter I for a hint of state of the art), few studies have dealt with the interactions of dyes with clay minerals already modified by some other organic species. Nonetheless, the behavior of organoclays is attracting more and more attention by the industry and, now more than ever, the possibility of tracing organoclays and monitoring their behavior appear of great interest. Antonella ESPOSITO 173 Chapter III III-R [1] REFERENCES Esposito A, Raccurt O., Charmeau JY, Duchet-Rumeau J. Photo-functionalization of an organically-modified clay. J. Colloid and Interface Sci. Submitted for publication, 2008. [2] Madejová J., Komadel P. Baseline studies of the Clay Minerals Society source clays: infrared methods. Clays Clay Miner. 2001, 49 (5), 410-432. [3] Madejová J. FTIR techniques in clay mineral studies. Vibrat. Spectr. 2003, 31, 1-10. [4] Coates J. Interpretation of infrared spectra, A practical approach, in Encyclopedia of Analytical Chemistry, R.A. Meyers (Ed.), Chichester: John Wiley & Sons Ltd, 2000, pp. 10815-10837. [5] Čapková P., Malý P., Pospíšil M. et al. Effect of surface and interlayer structure on the fluorescence of rhodamine B-montmorillonite: modeling and experiment. J. Colloid Interface Sci. 2004, 277, 128-137. [6] Fujita T., Iyi N., Kosugi T. et al. Intercalation characteristics of rhodamine 6G in fluortaeniolite: orientation in the gallery. Clays Clay Miner. 1997, 45 (1), 77-84. [7] Paul D.R., Zeng Q.H., Yu A.B. et al. The interlayer swelling and molecular packing in organoclays. J. Colloid Interface Sci. 2005, 292, 462-468. [8] Vietze U., Krauß O., Laeri F. et al. Zeolite-dye microlasers. Phys. Rev. Lett. 98, 81 (21), 4628-4631. [9] Esposito A, Balcaen J, Duchet-Rumeau J, Charmeau JY. Visiovis: monitoring nanofiller dispersion/distribution in molten polymers. JEC Composites Magazine 2008, 41, 67-71. PhD INSA de Lyon (2008) 174 Chapter IV PROCESSING Real-time monitoring of mixing The experimental device presented in this chapter has been recently overviewed in an article came out in June 2008 (JEC Composites Magazine n°41) [1]. Polymer-clay nanocomposites became truly marketable only some time after the registration of the first industrial patents [2]: the ability to properly and reproducibly disperse nanofillers in a polymer was, and still is, considered to be a strict requisite for their profitable commercialization. Although inorganic moieties are generally required to perfectly disaggregate, disperse and distribute into the organic matrix, clays are often affected by distribution and/or dispersion problems of their elementary and primary particles [3]. We’ve already insisted on the fact that a method capable of monitoring the evolution of the morphology during processing would greatly help the development of polymer-clay nanocomposites. Indeed, as previously said (§ I-2.1), if filler aggregation depends essentially on the physico-chemical interactions of its particles, filler dispersion and/or distribution into the matrix is directly influenced by processing efficiency. The morphology of polymer composite materials changes throughout the manufacturing Antonella ESPOSITO 175 Chapter IV process but seems primarily influenced, in the case of thermoplastic matrices, by the extrusion of formulated composite pellets and the injection molding of final objects. Such steps of the manufacturing chains are often performed by means of devices which have a complex geometry and, thus, are particularly difficult to monitor. In this chapter we’ll present a novel and innovating equipment entirely designed, assembled and developed in our laboratories over a period of about 5 years – including a little more than two years of a previous PhD project (Maël Moguedet) and the whole duration of three years of the following PhD project (Antonella Esposito). As suggested by the name itself, Visiovis is a tool exclusively devoted to processing visualization. Why processing? Because it consists essentially of a screw/barrel system (French vis = English screw) in which – and this is the real innovation – the barrel is entirely made of a transparent material, which permits to continuously visualize (visio-) the flow from all the possible directions of observation. This property encouraged the exploration of all the potentialities of such a novel tool – and that’s the topic of this chapter. In particular, we slightly changed its original configuration to adapt it to the real-time investigation of mixing between a polymer matrix and a lamellar inorganic filler and, thus, to in-line monitoring of nanocomposite processing which still is, as previously observed, the crucial factor for polymer-clay nanocomposites. The photo-functional inorganic/organic complexes (which are the topic of Chapters II and III) have been prepared expressly to be used as photo-active lamellar fillers for Visiovis experiments. IV-1 VISIOVIS Visiovis has been entirely assembled in our laboratories for the first time in 2003. This tool was originally designed to visualize the 3D trajectories of a single fluorescent particle plunged in a transparent fluid [4][5]. During these last three years, we further developed such an original equipment in order to adapt it to the analysis of nanofiller dispersion/distribution in viscous media (e.g. molten thermoplastic polymers or uncured thermoset resins) flowing in a geometrically complex system which is comparable, in the case of Visiovis, to the meter section of typical screw/barrel systems of industrial devices for extrusion and injection molding. PhD INSA de Lyon (2008) 176 PROCESSING Real-time monitoring of mixing IV-1.1 Original configuration Visiovis is an experimental device designed and assembled by Maël Moguedet during his PhD at the Ecole Supérieure de Plasturgie in Oyonnax1 (2002-2005) with the precious help of the technician Jean Balcaen. The “nucleus” of Visiovis is the assembly composed of a transparent barrel hosting a screw, but the real advantage – essential for a complete in-line 3D visualization of the speed cartogram of a fluid evolving in the space in-between a helicoidal screw and its cylindrical envelop – is the transparency of the barrel. Indeed, transparency is the property which chiefly rendered the Visiovis project innovative and interesting for in-line monitoring of the processing of polymers and polymer-based composites. The total transparency of Visiovis barrel allowed Moguedet and coworkers [4] to employ – in an absolutely innovative way – a technique which is relatively common in the field of fluid dynamics. They specifically designed Visiovis with the perspective of using it for Particle Tracking Velocimetry (PTV), a technique allowing to originate the speed cartogram by following the discrete movements of few particles plunged in a fluid over a finite period of time – under the hypothesis of low particle concentration, which is the condition permitting to follow the movements of each single particle. The principle of PVT is schematized in Figure IV-F1 [4]. The main objective of Moguedet’s work was to find a new method to visualize the entire 3D trajectory of a single particle plunged in a fluid and then to compare the results obtained by Visiovis with the results obtained by computer simulation. Indeed, computer simulation is very common in fluid dynamics since it is capable of deducing information generally not accessible via traditional experiments. On the other hand, a considerable effort has been done by several groups of researchers to find a practical method capable of collecting “real” data (although concerning model materials on pilot implantations) to be compared to the “predicted” ones and to be later “extrapolated” to 1 In February 2005 the Ecole Supérieure de Plasturgie (which up to that moment was a private institution) was integrated by INSA Lyon and, nowadays, the whole center in Oyonnax (laboratories and educational facilities) represents the Site de Plasturgie INSA – in particular, the laboratory and all the technical staff and researchers working therein are now labelled Laboratoire des Matériaux Macromoléculaires (LMM). Antonella ESPOSITO 177 Chapter IV the real, industrial systems. In particular, Moguedet and coworkers had been inspired by some works in which industrial extruders were equipped with one or more windows and (at least) one camera: they found the idea interesting but underlined the limits of such systems and looked for a better solution. Acquiring photos and videos on industrial systems equipped with one or more windows is certainly a good method to visualize the melting mechanisms of polymer pellets during processing – particularly because such observations can be directly correlated to other information acquired by conventional captors for temperature and pressure; however, Moguedet and coworkers were looking for something more complete (3D rather than 2D), more precise (quantitative rather than qualitative), more specific (the ultimate aim was to position a particle in the space and then to reconstruct 3D speed cartograms). Doubtless, that was a first hard challenge. t1 z x tn Lighting for tracer excitation z x y x t1 Camera shifting to follow the particles continuously in time y x tn Flow direction Figure IV-F1 Scheme of the principle of PVT technique, chosen by Moguedet and coworkers to design and develop Visiovis [4]. PhD INSA de Lyon (2008) 178 PROCESSING Real-time monitoring of mixing IV-1.1.1 Components and utilization In its original configuration (Figure IV-F2), Visiovis is composed of a squarepitched2 screw adjusted in a transparent barrel made of poly methyl methacrylate (PMMA), deliberately treated against UV radiation3 and assuring a transparency of about 80% for 44 rpm, torque > 380 nm [4]. The screw is actuated by an electrical motor (speed max = max = 9 N·m). An aperture allows to fill the system with the fluid of choice, to introduce the tracing particle and to empty the circuit at the end of the experiments; a tube connected in close circuit permits to virtually prolong the duration of each experiment by making the fluid continuously circulate into the system. The screw/barrel system is surrounded by a mobile framework supporting all the other components, in particular four CCD cameras (Basler A301F), equipped with yellow filters4 and able to record up to 80 images/s with a resolution of 640 × 480 pixels and a depth of 8 bits (256 grey levels). Two of the cameras are aligned horizontally and two vertically, face-to-face on opposite sides of the screw/barrel system – configuration which, thanks to the mobility of the framework, allows to follow the tracing particle at each instant with at least two cameras, meaning that at any time it is possible to deduce its 3D coordinates. Four UV diodes (emission 400 ± 5 nm) are placed around the screw/barrel system, assuring a 3D lightening. A calibrator, consisting of two perpendicularly crossing matrices of perpendicular filaments made of fluorescent nylon (5mm × 5mm), is fixed at one end of the screw/barrel system and allows to measure and then adjust any difference in the orientation, translation and/or rotation, zoom and/or focus of the four cameras. The images are recorded thanks to an acquisition circuit composed of two PCs and an external clock (to assure the synchronization of the cameras and to set the acquisition time). Visiovis geometrical parameters [4] are listed in Table IV-T1 and compared to the typical design parameters of the meter section of industrial screw/barrel systems for extrusion and injection molding [6] (Figure IV-F3). 2 A screw is square-pitched when its pitch is fairly similar to its external diameter (Table IV-T1). The choice of PMMA treated against UV radiation is justified by the original external light source (four UV diodes). 4 The yellow filters cut-off the excitation light coming from the diodes and make the cameras visualize only the fluorescence emission of the tracing particle. 3 Antonella ESPOSITO 179 Chapter IV UV diodes Electrical motor Aperture PMMA barrel Calibrator CCD camera Screw flight Tube for closed circuit Mobile framework Figure IV-F2 Visiovis in its original configuration – as it was designed, assembled and used by Moguedet and coworkers [4]. Table IV-T1 Visiovis geometrical parameters [4] compared to the typical design parameters of the meter section of industrial screw/barrel systems for extrusion and injection molding [6]. Geometrical parameters Visiovis Meter section Barrel diameter Db 40 mm 120 mm Screw root diameter Ds 30 mm 110 mm 40 mm 120 mm 5 mm 5 mm 34.6 mm 109 mm 20° 18.37° 6.9 21.8 Curvature 0.44 0.15 Torsion 0.16 0.05 Tan( ) 0.33 0.33 250 mm n.a. Screw pitch Channel depth Channel width Screw angle (measured in the middle of the channel) Aspect ratio Screw length PhD INSA de Lyon (2008) 2 Ph H W L 180 PROCESSING Real-time monitoring of mixing Ds H 2 Ph Db Figure IV-F3 Schematic of a typical screw/barrel system and the parameters which characterize its geometry [1]. Moguedet and coworkers [4] chose poly dimethylsiloxane (PDMS) as the model fluid – a transparent silicone oil having a viscosity of about 100 Pa·s and characterized by a Newtonian rheological behavior. As a tracing particle, they used a small fragment of the same filament employed for the calibrator, made of fluorescent nylon (diameter 0.4 mm). This small particle has a density similar to those of the PDMS fluid, which has allowed to presuppose that the particle was going to perfectly follow the flow lines. The acquisition of data in Moguedet’s work was simplified by the fact that it was aimed to the detection of a single point showing the highest luminosity in comparison with the deep black of the background – and of course this single point corresponded to the particle of fluorescent nylon. For this reason, Moguedet and coworkers designed an in-line image processing which noticeably reduced the amount of collected data: once realized that the fluorescent point representing the detected particle occupied an area of 4×4 pixels on each image acquired by the cameras, they designed a simple software (C++ environment) capable of acquiring exclusively the image corresponding to an area of 60×60 pixels centered on the most luminous point visualized by the cameras. Successively this image, the 3D coordinates of the most luminous point and the date and time of the acquisition (reduced to a binary code) were recorded in a bitmap format. The performances of such acquisition system were reduced to the acquisition of 7 images/s on average. A sample of the data acquired with the Visiovis in its original configuration is shown in Figure IV-F4. Antonella ESPOSITO 181 Chapter IV Figure IV-F4 Typical format of the data acquired by Moguedet and coworkers [4] with the Visiovis in its original configuration. This example represents a series of five images, acquired consecutively by a single camera. One can see, in each picture, the detected fluorescent particle and (encircled at the bottom) the date and time of acquisition encoded in a binary format. If the acquisition of data in Moguedet’s work was facilitated by the detection of the brightest point associated to the single fluorescent particle, the processing of such data was definitely more complicated. In addition to a double real-time data processing executed in a C++ environment (selection of the 60×60 pixels area centered around the brightest point and reduction of the noise intrinsically associated to the CCD cameras), a labor-intensive processing of the acquired data was necessary to finally reconstruct the 3D trajectories of the single fluorescent particle. This second part of data processing was realized in the Matlab environment and included image filtering, image adjusting on the basis of the information given by the calibrator, position adjusting on the basis of the refraction index of the air (atmosphere), the PMMA (barrel) and the PDMS (model fluid), as well as any other correction imposed by the optical effects due to the fact that the barrel represents a cylindrical surface (which interfere with any optical acquisition). More details about the whole data processing can be found in the PhD manuscript by Moguedet [4]. IV-1.1.2 Advantages and limitations As already mentioned, since the beginning the Visiovis project presented a great interest in the field of fluid dynamics and, indirectly, of polymer processing. However, Moguedet and coworkers had to cope with several problems before achieving their objectives. If the filament made of fluorescent nylon solved several of their problems (individuation of a proper tracing particle and realization of the calibrator for the correct PhD INSA de Lyon (2008) 182 PROCESSING Real-time monitoring of mixing alignment and regulation of the cameras), the idea of a totally transparent barrel hosting a rotating screw rapidly became a problem much more difficult to be solved. Indeed, the barrel couldn’t have been made of glass because glass stops UV radiation; couldn’t have been made of some special glass (e.g. quartz) because, no matter how good its optical properties would be, it’d always have been too fragile to support the radial pressure originated by the movement of the screw and by the presence of the fluid (moreover, a barrel entirely made of quartz would have been excessively expensive). They realized soon that the only cheap and practicable solution was a barrel made of plastic. PMMA seemed to represent a rapid and cheap solution, primarily thanks to its transparency – but, of course, PMMA cannot tolerate high temperature (even if the glass transition temperature of PMMA is around 100°C, any temperature increase may render the barrel enough malleable to get susceptible of deformation under the action of the internal pressure). In addition, the fact that the barrel is made of plastic imposes to pay a particular attention to the chemical compatibility of any fluid introduced into the system (or any solvent used to wash it) with the material used for the barrel: only inert oils can be chosen as the model fluid, and only solvents which neither swell nor dissolve PMMA can be used for cleaning. By the way, the barrel wasn’t the only source of problems. If a tracing particle was already available, Moguedet had still to find a suitable model fluid, which should have had the following properties: (1) being a macromolecular fluid; (2) resisting UV radiation without degrading; (3) being optically transparent (to UV radiation but also to visible radiation); (4) being inert and, more specifically, being compatible with PMMA; (5) being capable of flowing at room temperature (since any experience with Visiovis had to be realized at room temperature). After having tried (unsuccessfully) to fill the closed circuit of Visiovis with a solution of water and poly ethylene oxide (PEO), they finally found a compromise in poly dimethylsiloxane (PDMS), a transparent and inert silicone oil, commercially available in different molar weight and, thus, different viscosities. At least, for the purposes of Moguedet’s work, an advantage of the fact that PDMS is Newtonian is that this property authorized to perform experiments at relatively low rotational speeds, as Newtonian fluids aren’t sensitive to shear and, consequently, Antonella ESPOSITO 183 Chapter IV their speed cartogram is supposed to stay unchanged even when the screw speed is increased. The independence of the flow speed profile from the rotational speed induced Moguedet to choose the slowest speed available for the screw (1 rpm), in order to optimize the acquisition procedure which, in addition, was slowed down by real-time data processing5. Height (mm) Length (mm) Width (mm) Figure IV-F5 Three-dimensional trajectory of a single fluorescent particle in the channel of the screw/barrel system of Visiovis (screw rotational speed 1 rpm) [4]. The fact that points are missing for all the positions close to the cylindrical surface of the barrel witnesses the limitations due to refraction. The experiments performed with this first version of Visiovis have been positive in terms of concretization of the project and demonstration that this original tool works as expected. These first measurements could produce the approximate 3D trajectory of a particle plunged in a transparent Newtonian fluid evolving into the screw/barrel system (Figure IV-F5). In particular, they revealed the presence of two speed comportments – slow close to the barrel surface and rapid close to the screw surface. These observations, along with the trajectory of the particle, have been successfully confirmed by computer 5 Some information about the real-time data processing during acquisition can be found in § IV-1.1.1. For more details, please refer to Moguedet’s work [4]. PhD INSA de Lyon (2008) 184 PROCESSING Real-time monitoring of mixing simulations (Figure IV-F6). More in detail, a particle introduced into the system and starting its progression close to the axis of the system, proceeds rapidly and then moves towards the surface of the barrel, thus its progression slows down [4]. By the way, the system still presents several optical problems (once more due to the refraction at the cylindrical surface of the barrel – coupled to the fact that the four UV diodes create an isotropic-like lightening which is prone to reflection and diffraction phenomena). These residual optical problems caused an intensive loss of observed data any time that the particle got close to the surface of the barrel. The first results obtained by Visiovis and confirmed by computer simulations have been published in 2004 [7] and the same group of researchers actually keeps developing the initial model [8]. Experimental radial position Simulated final radial position Extrapolation of the simulation Barrel internal surface Radial position (mm) Screw root surface Time (s) Figure IV-F6 Evolution of the radial position of the particle in the channel of the screw/barrel system of Visiovis (screw rotational speed 1 rpm) [4]. Experimental observations correlate well with the results of simulation. Antonella ESPOSITO 185 Chapter IV IV-1.2 Evolutions of the configuration It was Moguedet himself who firstly proposed, at the end of his PhD manuscript [4], some possible modifications of Visiovis: the replacement of the four CCD cameras with some more recent “intelligent cameras”, capable of performing an automatic image processing for the detection of the brightest point without employing the resources of the computers’ processors – which could ameliorate the performances of the acquisition of data; the realization of an additional computer-controlled system for the automatic translation of the mobile framework, so that the cameras could follow the advancement of the tracing particle – which could allow to follow more than one single particle; the replacement of the cylindrical barrel by another barrel with an internal cylindrical surface and an external rectangular shape – which would significantly reduce the problems caused by refraction. Our objective is to adapt Visiovis to the real-time monitoring of the dispersion and distribution of lamellar fillers in a polymer matrix during nanocomposite processing. To achieve our objective, we unavoidably had to modify Visiovis original configuration because of some additional difficulties linked to the multiplicity of the tracing particles to be followed and to their reduced dimensions, as well. Anyway, we haven’t necessarily followed the lines suggested by Moguedet and coworkers – meaning that many other possibilities of evolution, ameliorations and diversification are still left. In fact, we passed through several changes of the configuration, trying to adapt step-bystep this tool to our objectives. The easiest and cheapest changes we could attempt on Visiovis concerned: (1) the position of the CCD cameras, (2) the form of the lightening for fluorescence excitation and (3) the addition of the most sensitive instrument for the characterization of the fluorescence behavior of the photo-functional fillers previously described6 – a spectrofluorimetry relied to the system by an optical fiber. 6 See Chapter II for more details about the photo-functionalization of lamellar mineral fillers and Chapter III for the description of the photo-functional inorganic/organic complexes realized with the perspective of using them with Visiovis. PhD INSA de Lyon (2008) 186 PROCESSING Real-time monitoring of mixing IV-1.2.1 From 3D lighting to 2D laser plan At the real beginning of our experimental work – even before the development of a photo-functionalization method for lamellar fillers – we were quite determined to keep Visiovis in its original configuration, thus we tried to find a suitable photoluminescent nanofiller to continue investigating the flow behavior of the same Newtonian matrix (PDMS) filled with several particle (not just a single tracing particle) evolving in the same geometrical system (i.e. without modifying the screw profile, since we already had some information about the existing one). Indeed, we found a rather good solution for our problems: a cheap, commercially available phosphorescent pigment (GT5700, GloTech Inc., New Zealand) consisting of an alkaline rare-earth aluminate, easily excitable by UV radiation, white or any visible light (240-440 nm), emitting in the yellow/green range (520 nm), with an intense and persistent emission (>12 hours, certified by the provider on the basis of the measuring protocol DIN 67510). Two images of this pigment taken by light microscopy are given in Figure IV-F7 and their excitation and phosphorescence spectra are shown in Figure IV-F8 (data provided by GloTech Inc.). 100 m 100 m Figure IV-F7 Two images of GT5700 particles taken by light microscopy (magnification 20x). Antonella ESPOSITO 187 Chapter IV Figure IV-F8 Absorption and phosphorescence spectra of the pigment GT5700. GT5700 chemical, physical and luminescent properties are resumed in Table IVT2. This photoluminescent pigment is employed for many applications, including brush painting, spray painting, candle making and glass moulding. Table IV-T2 Main chemical, physical and luminescent properties of GT5700 pigment. Chemical properties Composition Alkaline Rare-Earth Aluminate Insoluble in Water, Alkalis and Organic Solvents Decomposition Acids Physical properties Appearance Specific Gravity Particle Size Distribution (Laser Granularity) Yellowish 3.6 g/cm3 45-55 m (D50) Luminescent properties Excitation Excitation Wavelength Peak Value Glow Color Glow Duration UV radiation, white or any visible light 240-440 nm 520 nm Yellow-Green > 12 hours We performed a test of visualization of the GT5700 pigment by Visiovis and we actually got some interesting results: the intense brightness due to the phosphorescence emission is effectively sufficient to visualize the particles with the CDD cameras. We could easily distinguish the dark zones (pure PDMS) from the zones containing the PhD INSA de Lyon (2008) 188 PROCESSING Real-time monitoring of mixing phosphorescent particles. In addition, we could clearly visualize any difference in the spatial distribution of the particles during mixing, as shown by Figure IV-F9. barrel differences in focus screw flight pure PDMS GT5700 particles barrel barrel Figure IV-F9 Some examples of the test of visualization of the GT5700 pigment by Visiovis. Isotropic-like lightening and excitation by the four UV diodes. However, the test of visualization of the GT5700 pigment highlighted once more the great limitations of the system in its actual configuration and of the chosen pigment, as well. As previously reported, Moguedet and coworkers [4] were aware of the optical problems created by the cylindrical PMMA barrel and by the isotropic-like lighting of the four diodes. Their acquisitions and measurement greatly suffered from something they mainly recognized as a refraction phenomenon. The images collected during our test of visualization of the GT5700 pigment (Figure IV-F9) clearly show that refraction isn’t the only optical phenomenon affecting Visiovis in its original configuration: most of the visualization problems come from a strong reflection of the light by the surface of the barrel – to a point that no particle can be visualized in the upper and lower portions of the screw/barrel system, i.e. in the channel sections perpendicular to the cameras. Moreover, as the particles are phosphorescent, once they’re excited they all emit at the same time, in any point of the visualized portion of the screw/barrel system: this means that the cameras should properly (viz. correctly focused) visualize all the particles, even if they’re placed on different focalization plans. This is obviously impossible, as shown by the differences in focus visible on the examples in Figure IV-F9. Antonella ESPOSITO 189 Chapter IV The same images shown in Figure IV-F9 are presented again in Figure IV-F10, but in their negative version: the negative version of such images, in which the only contrast is given by a bright phosphorescence emission on a dark background, greatly helps the evaluation of the visualization limits of Visiovis in its original configuration. Two other negative images are added to support the observations done on the basis of the first two images. It is unambiguous that the zones close to the barrel surface are “critical” for visualization – besides, this area is probably the most interesting, since it could help understanding the dependency of the particle distribution on the radial position, i.e. on the distance from the axis of the screw/barrel system (relative distance from the barrel surface and/or from the screw root surface). barrel differences in focus screw flight pure PDMS GT5700 particles barrel barrel barrel “critical” zones Figure IV-F10 Some examples of the test of visualization of the GT5700 pigment by Visiovis. The negative version greatly helps the evaluation of the test images. The first two images (upper side) are the same images already shown in Figure IV-F9. PhD INSA de Lyon (2008) 190 PROCESSING Real-time monitoring of mixing Taking into account the problems encountered by Moguedet and coworkers in relation with isotropic-like lightening and the refraction phenomena at the cylindrical surface of the barrel, and being aware of the fact that changing the shape of the barrel would have required a deep reorganization of Visiovis configuration, we estimated that the first thing to do was rather to modify the lightening system. Therefore, we planned to fabricate an optical system which, coupled to a laser source, would create a thin sheet of light – a virtual, optical section of Visiovis screw/barrel system. In the presence of such a bidimensional lightening, the fact that GT5700 pigment is phosphorescent not only has no more interest, but appear even inappropriate: indeed, the main interest of a 2D lightening is the possibility of visualizing exclusively the particles included in the thin sheet of light – which requires that the particles are excited by the laser sheet and produce a fluorescent response exclusively when excited, whereas the glow duration of the phosphorescent pigment means that the particles emit even if they aren’t anymore excited. Although GT5700 pigment has good performances when visualized by Visiovis, abandoning its usage hasn’t been too much disappointing for several reasons: the pigment isn’t a lamellar nanofiller – the particles have an almost unitary shape factor, which means that they are spherical (see the D50 value in Table IV-T2); the pigment has a density (3.6 g/cm3, Table IV-T2) which is inadequate for the PDMS (0.97 g/cm3) – a filler which is much denser than the matrix is more prone to sedimentation by gravity; the pigment is phosphorescent – a property which, as previously explained, is no more appropriate for the lightening system we were planning to assemble. The new lightening system will have to be chosen on the basis of several criteria: (1) the commercial availability and the price of the laser source; (2) the wavelength of laser emission – for the optimum excitation of the photo-functionalized filler; (3) the fact that the PMMA barrel must be transparent (but also resistant) to the light emitted by the laser. More details about the new lighting system will be given in § IV-1.3. Antonella ESPOSITO 191 Chapter IV IV-1.2.2 Position of the CCD cameras We’ve already highlighted that the zones close to the barrel surface are “critical” and may be particularly interesting for the visualization of particle distribution during mixing. We’ve also stressed, when presenting Visiovis and its geometrical parameters, that our screw/barrel system may be compared to the meter section of the industrial screw/barrel systems for extrusion and injection molding – thanks to the fact that most industrial devices has a meter zone characterized by a square-pitched screw and a shallow channel (see Figure IV-F3 for a schematic representation of the meter section of a screw/barrel system, as well as Table IV-T1 for a comparison of Visiovis geometrical parameters with the typical design values of the meter section of industrial devices). Once understood that the screw profile adjusted in our transparent barrel has a shallow channel (5mm deep), one realizes how much important is an accurate visualization of the zones close to the barrel surface. The choice of a new lighting system, consisting of a thin laser sheet creating an optical slice of the system, is absolutely coherent with this new criterion of visualization. By making the laser sheet pass exactly by the axis of the screw/barrel system, we’re planning to visualize the longitudinal section of the channel, in which the fluid and its filler are supposed to mix up (see Figure IV-F11). laser sheet Figure IV-F11 Schematic representation of a laser sheet passing by the axis of the screw/barrel system of Visiovis. Only the particles lying on the optical plan lightened by the laser sheet are excited and have a detectable fluorescence emission (similarly to PIV, see Figure I-F18). PhD INSA de Lyon (2008) 192 PROCESSING Real-time monitoring of mixing It is obvious that, in such a configuration, the camera placed on the same side of the laser source and the one placed on the opposite side (viz. the cameras lying onto the optical plan created by the laser sheet) become totally useless. Only one camera – those placed on the plan perpendicular to the laser sheet – continues accomplishing its task of visualization (the camera on the opposite side being useless as well, because the data collected by this second, perpendicular camera would be redundant). In other words, the fact that we changed the lightening system forced us to change the position of the CCD cameras, as well. Since the only interesting position for visualization was, at the present, the plan perpendicular to those traced by the laser sheet, we estimated that the best thing to do was to align the four cameras axially, alongside the screw/barrel system, right in face of the visualized longitudinal section of Visiovis channel (Figure IV-F12). laser source mobile framework cameras laser sheet Figure IV-F12 Position of the CCD cameras in relation to the laser sheet passing by the axis of the screw/barrel system of Visiovis. IV-1.2.3 Optical fiber and in-line spectrofluorimetry The first test of visualization performed with the luminescent GT5700 pigment and the original configuration of Visiovis allowed us to estimate the accuracy of our system – limits that for the moment we could not ameliorate, since we weren’t planning to change either the CCD cameras or the screw/barrel system itself. Thanks to the Antonella ESPOSITO 193 Chapter IV images acquired during the test of visualization, we calculated that the CCD cameras can detect approximately 83 m per pixel – a sensitivity which could be probably acceptable for the phosphorescent pigment (D50 = 45-55 m as reported in Table IV-T2) but surely not enough for lamellar fillers, which have a multiscale structure (Figure IF23) with a minimum dimension of 1 × 100 nm (isolated clay platelets) and a maximum dimension of about 10 m. More specifically, the typical dry particle size of the commercial clays selected to perform photo-functionalization is described by the following distribution7: 10% of the particles measure less than 2 m, 50% less than 6 m and 90% less than 13 m. It is evident that the sole cameras aren’t adequate and anymore sufficient for a proper visualization of the complex phenomena occurring during nanocomposite processing. This is the reason why, in addition to the visual detection performed by the CCD cameras, we decided to equip Visiovis with another detection system – a spectrometer connected to an optical fiber probe able to collect the intensity of fluorescence emission during polymer/clay mixing. Spectrofluorimetry is a technique sensitive to phenomena occurring at a different scale in comparison with the CCD cameras – the latter performing a global in-line monitoring of mixing, the former providing more specific and space-restrained information. The main advantage of this additional technique is that fluorescence is extremely sensitive to several properties of the environment in which the tracing molecule is positioned – including temperature, pH, chemical composition, molecular arrangement and physical confinement – as previously stressed in Chapters II and III. Of course, we had to check the visualization limits and the sensitivity of both the CCD cameras and the spectrometer with the photo-active complexes previously prepared by photo-functionalization of commercial clays. These tests of visualization will also be useful to calibrate both the detection systems (cameras and spectrometer) for any future experience with Visiovis. The results of calibration will be discussed in the following paragraph. 7 These values are provided by the supplier. Percentages are expressed by volume. PhD INSA de Lyon (2008) 194 PROCESSING Real-time monitoring of mixing IV-1.2.4 Calibration of the detection systems Before performing any significant experience with Visiovis, we had to assure a satisfactory calibration of the old and new detection systems with the photo-active lamellar fillers prepared ad hoc for our visualization tool. The best calibration would be enough accurate to allow a direct correlation between the concentration of fluorescent molecule and the luminosity detected for each pixel of the CCD cameras, or the intensity of the fluorescence emission detected by the spectrometer. Such a scheme of correlation would allow to obtain a real-time concentration cartogram of the optical sections created by the laser sheet longitudinally with respect to the screw/barrel system. Unluckily, such a fortunate calibration would require many measurements and would surely be complicated for the following reasons: (1) when penetrating the fluid, the intensity of the planar laser sheet decreases (the higher the concentration of tracer, the quicker it decreases); (2) the initial concentration of the injected masterbatch is well known, but unfortunately there’s no way to predict its spatial and temporal evolution during mixing; (3) we haven’t yet designed an efficient method to get some samples of the fluid evolving in the screw/barrel system, in order to confirm by some other technique the results obtained by the cameras and the spectrometer. Thus, for the moment we could only perform a kind of “qualitative calibration” of the cameras and of the spectrometer with the first photo-functional inorganic/organic complex prepared by cation exchange process: the photo-active filler based on Cloisite ® 30B (C30B 0.25MC RhP)8. Even though any calibration is quantitative by nature, we hazarded called it “qualitative” simply because, in our case, we performed calibration just to understand whether the photo-active fillers are detected or not, and which is the optimum concentration to be used for any future experience made on Visiovis. The calibration of the detection systems required the preparation of a certain amount of mixtures of the photo-active lamellar filler with the transparent model fluid. We prepared 11 mixtures having a controlled concentration (0%, 0.001%, 0.0025%, 0.005%, 0.01%, 0.025%, 0.05%, 0.1%, 0.25%, 0.5%, 1%, where percentages are to be 8 More information about the preparation of this photo-active lamellar filler can be found in Chapter II. Antonella ESPOSITO 195 Chapter IV intended by weight) starting from an initial 1% wt mixture and proceeding by dilution. The initial mixture has been prepared by a mixer equipped with a 60 mm dilacerator disk – which is supposed to facilitate mixing by breaking the eventual aggregates and better dispersing the filler into the matrix – rotating at 1000 rpm for 20 min (Disperser TurboTest Rayneri 33/300P). We mixed the 10 Pa·s PDMS9 (Siliconöl M10000, Carl Roth, Germany) with a proper amount of the photo-active lamellar filler C30B 0.25MC RhP10. These controlled mixtures have then been poured in 4.5 mL disposable PMMA cuvettes11 (Rotilabo ® Plastibrand Elumal-Küvetten) specific for spectrofluorimetry, assuring a perfect optical permeability (wavelength range 300-900 nm, std dev ≤ 0.004 extinction units starting from 320 nm). All the mixtures are shown in Figure IV-F13. Figure IV-F13 Controlled-concentration mixtures (C30B 0.25MC RhP in PDMS) prepared for the “qualitative calibration” of Visiovis detection systems. 9 We used this silicone oil for all our experiences with Visiovis. The reasons why we chose a 10 Pa·s PDMS instead of a 100 Pa·s PDMS (as Moguedet and coworkers did [4]) deal only with practical aspects and will be given in the following. 10 We decided to firstly focus on C30B 0.25MC RhP because it was the first photo-active lamellar filler ready to be used and, parenthetically, it showed a good photo-activity since the first tests of visualization. 11 We purposely chose such special cuvettes since they are made of the same material as Visiovis barrel. PhD INSA de Lyon (2008) 196 PROCESSING Real-time monitoring of mixing The calibration procedure for the spectrometer essentially consisted in recording a fluorescence emission spectrum for each mixture (Figure IV-F14) and the calibration of the CCD cameras in taking an image of each cuvette (Figure IV-F15). Figure IV-F14 “Qualitative calibration” of the spectrometer connected to the optical fibre: test performed with 11 mixtures (C30B 0.25MC RhP in PDMS) having different concentrations (from 0% to 1%). Concentration is expressed in percentages by weight. Integration time 3s. The results of the calibration for the spectrometer (fluorescence emission spectra, Figure IV-F14) are unsurprising and, somehow, reassuring. Pure PDMS isn’t excited by the laser sheet and doesn’t produce any parasite fluorescence phenomenon, as expected. A concentration of 0.001% wt of C30B 0.25MC RhP doesn’t show any significant fluorescence response, as well. This could seem weird if compared to the extremely low concentrations of fluorescent dye used for traditional tracing experiments (only some parts per million); indeed, we should keep in mind that Visiovis experiences are quite different from traditional tracing experiments, since: we’re not using a pure fluorescent dye to trace an homogeneous fluid in which the dye is perfectly soluble – we’re rather introducing a photo-functional lamellar filler Antonella ESPOSITO 197 Chapter IV in a macromolecular fluid, we’re not sure that mixing will be successful and to which extent and, in any case, the fluorescent molecules are supposed to be intercalated into clay galleries, thus clay platelets could engender a barrier effect for fluorescence excitation and emission, analogously to what happen during thermal degradation or gas permeation; the device we coupled the spectrometer to is pioneering – we’re not using a fluorescence microscope or any other commercially available equipment to observe the fluorescence behavior of the photo-active lamellar filler in its environment, thus we risk to have to cope with some additional optical limitations probably unknown to the people who perform traditional tracing experiments. As the concentration reaches a value of 0.0025% wt of C30B 0.25MC RhP, we observe a first slight fluorescence response around 600 nm – a value of wavelength which is not exactly the same observed for pure RhP (553 nm, Figure II-F8b) and for the photo-functional complex (553 nm, Figure II-F11b) in ethanol. This shift may be due to the medium in which the spectra have been collected (PDMS for the calibration and ethanol for the other characterizations) but also to the global environment in which the measurements have been performed, viz. to the fact that the spectra in Figure IV-F14 have been recorded by a spectrometer and collected by an optical fiber placed in front of the PMMA tub, whereas the spectra in Figure II-F8b and II-FB11b have been recorded by a commercial spectrometer in standard measurement conditions. For the values comprised between 0.01 and 0.1% wt, the fluorescence response seems to be perfectly proportional to the concentration of C30B 0.25MC RhP. Starting from the concentration value equal to 0.1% wt, a saturation-like phenomenon seems to occur and the intensity of the main emission peak stop increasing (on the other hand, it doesn’t start decreasing either, as usually observed in case of fluorescence saturation). The results obtained for the calibration of the CCD cameras (images taken in the dark, planar laser source irradiating from top to bottom Figure IV-F15) could a priori confirm the results obtained for the calibration of the spectrometer, but the eventuality of obtaining different (i.e. complementary and/or less precise) results shouldn’t be surprising since, as previously explained, the two detection systems aren’t sensitive to PhD INSA de Lyon (2008) 198 PROCESSING Real-time monitoring of mixing the same phenomena and cover two different length scales (macro-scale global view for the CCD cameras, micro- and/or nano-scale local view for the spectrometer). First of all, the images confirm that pure PDMS isn’t excited by the laser sheet and doesn’t produce any parasite fluorescence phenomenon. On the other hand and unsurprisingly, the CCD cameras look less sensitive to fluorescence emission than the spectrometer – observation which justifies our choice of adding another detection system to Visiovis12. Indeed, nothing appears on the collected images until the concentration reaches a value of at least 0.01% wt, afterward the quality of the acquired images gets better and better up to a concentration value of 0.1-0.25% wt. When the concentration gets higher than 0.25% wt, the local saturation of the fluorescence emission causes a remarkable degradation of the image quality – as well as a considerable loss of data, particularly evident for the highest concentration value (1% wt). Besides, another optical problem arises when the concentration is too high: the mixture becomes less and less transparent and the penetration depth of the laser sheet rapidly decreases: a further good reason to be cautious about concentration issues. It’s worthy underline that the optimum excitation requires the planar laser source to penetrate to a depth at least equal to the maximum flight depth: as the channel of Visiovis profile is constant and swallow like most meter sections (5 mm), this requirement is easily fulfilled (Figure IV-F15). This constraint would have been certainly stricter if Visiovis had a profile similar to the feed section (constant but deeper channel) or to the transition section (variable channel depth)13 – one more reason to be careful about acquiring and interpreting data on the very first portion of Visiovis – containing the connection between the transition and the meter sections. Indeed, this portion is primarily concerned by concentration and laser penetration issues: the highest concentration (the lowest laser penetration depth) is observed right after the injection of the tracing masterbatch, viz. where screw channel is variable (and surely deeper). 12 13 See § IV-1.2.3. The typical screw profile and its different sections are described in Chapter I (see Figure I-F2). Antonella ESPOSITO 199 Chapter IV L A S E R pure PDMS 45 mm 0.001 % 0.0025 % 0.005 % 0.01 % 0.025 % 0.05 % 0.1 % 0.25 % 0.5 % L A S E R 1% L A S E R critical laser depth Figure IV-F15 “Qualitative calibration” of the CCD cameras: test performed with 11 mixtures (C30B 0.25MC RhP in PDMS) having different concentrations (from 0% to 1%). Laser source: from top to bottom. Percentages are expressed by weight. PhD INSA de Lyon (2008) 200 PROCESSING Real-time monitoring of mixing In conclusion, the choice of the correct concentration value of the photo-active lamellar filler to be mixed with PDMS by Visiovis is quite complicated and must be done on the basis of several parameters – but typically the main criteria for choosing are the detection limits of the CCD cameras and of the spectrometer, as well as the laser penetration depth. An acceptable compromise must be found between an insufficient concentration for the visualization by the CCD cameras and an excessive concentration which would cause a saturation of the fluorescence emission and, therefore, a loss of information. The best compromise would be to accept that the data recorded by the spectrometer are slightly under their optimum of detection so that the images are acquired by the cameras in their optimum conditions of detection. Therefore, the best is to assure a concentration of about 0.05-0.1% wt. By the way, as previously underlined, concentration issues about Visiovis are always more complicated than expected: indeed, even if the initial concentration of the injected masterbatch is well known, it is impossible to predict its spatial and temporal evolution during mixing: one can only predict that, after the injection of a tracing masterbatch into the pure PDMS used to fill the closed circuit of Visiovis screw/barrel system, mixing will be accompanied by a dilution of the initial concentration – but such dilution won’t be constant in time and, even worse, it will have a complicated spatial dependence reflecting the mixing efficiency of the system. This means that: (1) injecting an initial masterbatch which is too concentrated will surely hinder the visual detection by the CCD cameras at the very first moments of the experience because of luminosity saturation, as shown in Figure IV-F15 (whereas the fluorescence detection won’t be affected because Figure IV-F14 doesn’t show any decrease of the fluorescence emission intensity, but rather a plateau at the highest concentration values); (2) injecting an initial masterbatch which isn’t enough concentrated will almost certainly make the fluorescence detection less clear (the emission peak won’t be at its maximum intensity, as shown in Figure IV-F14) and, as a consequence of the dilution, will force us to stop the experience sooner than in the previous case (the concentration would drop sooner under the detection limits), but at least the absence of an initial luminosity saturation (Figure IV-F15) would allow us to collect all the visual data by the cameras. Antonella ESPOSITO 201 Chapter IV On the basis of some early tests of visualization with the photo-active lamellar fillers, we observed that the efficiency of visualization by Visiovis seemed to decrease with the degree of homogeneity of the mixture evolving into the screw/barrel system – in other words, the most interesting moments of any experience performed by Visiovis seem to be the very first ones. For such reason, we judged necessary to assure a correct visualization of the first part of any future experience and, after several trials, we found that the best compromise is to inject (in pure PDMS) 10 mL of an initial masterbatch having a concentration of photo-active lamellar filler of 0.25% wt. More details about the experimental protocol will be given in § IV-2. IV-1.3 Actual configuration After the modifications we operated, Visiovis consists of some old components in their previous configuration, some old components in a new configuration and some new components. The screw/barrel system (which can be considered as the “nucleus” of Visiovis) is the main old component kept in its original configuration. Analogously, the electrical motor, the aperture (to introduce the fluid and the tracers), the tube (for close circuit) and the mobile framework – all these components stay unchanged in their initial configuration. On the other hand, the CCD cameras are now disposed differently on the mobile framework (they’re aligned axially, alongside the screw/barrel system)14, their yellow filters have been replaced by new filters (better adapted to the new light source), the diodes have been substituted by a green laser source ( = 532 nm, nominal power 20 mW CW15, Figure IV-F16 (a) and (b)), the acquisition circuit has been simplified (no more external clock and D-latch memories). The new components are: an optical system which creates, from the linear laser source, a bidimensional laser sheet (Figure IV-F16 (c) and (d)); an electromechanical output transducer (a trigger, basically a switch), which couples the image acquisition to the screw rotation (the cameras are controlled by the movement, since they acquire one image per screw revolution); the spectrometer (USB2000+ miniature, Oceanoptics), interfaced to the screw/barrel system by an optical 14 15 More details and the reasons for such change of configuration are available in § IV-1.2.2. Continuous Wave. PhD INSA de Lyon (2008) 202 PROCESSING fiber (600 m Real-time monitoring of mixing with a resolution of 2.5 nm) positioned in front of the channel of the screw/barrel system, on the opposite side of the cameras and perpendicularly to the laser sheet; a tap in the middle of the tube for close circuit, facilitating the draining of the system after each experiment. The calibrator has been simply removed. (a) (c) (b) (d) Figure IV-F16 Thin laser sheet ( = 532 nm, nominal power 20 mW CW) passing by the axis of the screw/barrel system. A global picture of Visiovis in its actual configuration, accompanied by some more detailed pictures of its actual components, is shown in Figure IV-F17 [1]. Antonella ESPOSITO 203 Chapter IV Laser sheet @ 532 nm Aperture Electrical motor CCD cameras Screw/barrel system Tube for close circuit Spectrometer Trigger Optical fiber Figure IV-F17 Visiovis in its actual configuration – after the modifications we made to the lighting system, the change of position of the CCD cameras and the addition of the optical fiber and the spectrometer [1]. IV-1.3.1 Objectives Once the modifications made and the detection systems calibrated, we could plan how to perform experiments on Visiovis in its actual configuration and, in particular, we could finally consider the following questions: Which model materials would it be better to use to perform the visualization experiences (in other words, which model fluid and which photo-active lamellar filler)? How should we prepare the masterbatches to be injected into the screw/barrel system from the apposite aperture? Which method should we use to inject the masterbatches? PhD INSA de Lyon (2008) 204 PROCESSING Real-time monitoring of mixing How will we exploit the experimental data collected by the cameras and the spectrometer? Shouldn’t we conceive a system which would allow us to validate the results eventually obtained by Visiovis (e.g. sampling and coupling to other characterization techniques)? Aware of the multiplicity of problems to be solved and questions to be answered to, we realized that the objectives with the highest priority were, at present: (1) planning a correct experimental protocol able to give some interesting results and (2) performing some early visualization tests on the freshly-reconfigured Visiovis, in order to prepare the way to the future experiments. With these targets in mind, we tried to answer to as many questions as possible – anyway, some of them will rather remain a perspective – and we attempted to suggest some realistic solutions. IV-2 EXPERIMENTAL PROTOCOL As previously announced, our optimization vocation starts with finding a first, realistic and practicable experimental protocol which would allow exploiting the freshly reconfigured Visiovis to obtain some interesting and – of course – interpretable results. We’ll firstly describe the experimental protocol used to perform the visualization tests: the interpretation of the experimental results is an issue to be considered soon after. With reference to the model materials to be used for the visualization test (model fluid and photo-active lamellar filler): as we haven’t yet found an appropriate substitute for the transparent PDMS, we decided to continue using such macromolecular viscous fluid as the model matrix (Siliconöl M1000016, Carl Roth, Germany); about the photoactive lamellar filler, we decided to perform some rapid visualization tests of the four 16 We used Siliconöl M10000 (10 Pa·s) rather than Siliconöl M100000 (100 Pa·s) because, contrarily to Moguedet and coworkers [4], the way we were going to use Visiovis required to change the model fluid after each experiment – whereas Moguedet and coworkers could use the same fluid for longer time. The need for frequently changing the fluid is a key parameter for the feasibility of a given experimental protocol: the steps of filling and empting the screw/barrel system are the slowest and the most delicate, as all the air bubbles must be carefully evacuated before performing any visualization activity – indeed, the presence of air bubbles significantly affects optical phenomena. Antonella ESPOSITO 205 Chapter IV photo-functional inorganic/organic complexes prepared by cation exchange process 17 of commercial clays (Cloisite ® Na+, Cloisite ® 30B, Cloisite ® 10A and Cloisite ® 15A) to be sure that they’re correctly detected by the cameras and the spectrometer. The results of the visualization test performed on the four photo-functional complexes is a little bit surprising (not all the photo-active lamellar fillers can be efficiently visualized by the cameras) but, after all, it shouldn’t astonish that much: as previously commented, fluorescence is a sensitive but fickle technique, about which one can never be confident of getting for sure some good results. Indeed, several reasons could explain the fact that three of the four photo-active lamellar fillers can be perfectly detected (C30B 0.25MC RhP, C10A 0.25MC RhP and C15A 0.25MC RhP) whereas one cannot (CNa+ 0.25CEC RhP). A first reason could be the chemical composition of the commercial clays used to prepare the photo-functional complexes: the photo-active lamellar fillers which can be easily visualized by both the cameras and the spectrometer are the ones prepared from organoclays – contrarily to the one which doesn’t show any visible fluorescence, rather prepared from a natural clay. However, such explication doesn’t really persuade, since the characterizations of the dry photo-functional inorganic/organic complexes by spectrofluorimetry didn’t reveal any relevant difference in the fluorescence emission of the four samples. Another possible reason must be searched, then, in the interactions of the photo-active lamellar fillers with the silicone oil: the presence (or, we should better say, the absence) of an organic surfactant in clay galleries could play a significant role in the formation of a positive (negative) interaction of the photo-active filler with the PDMS, assisting (hindering) the fluorescence emission by the RhP cations. Definitely, finding an explication isn’t that easy. We’ll just observe that some difference has been visually detected even before performing the visualization test by Visiovis: when preparing the masterbatches with the four photo-active lamellar fillers, we noticed that the aspect of the mixture prepared with CNa+ 0.25CEC RhP was different (visibly less homogeneous, meaning a less intimate mixture) with reference to 17 More details about the photo-functionalization method are available in Chapter II. On the other hand, Chapter III deals with the characterization of the four photo-functional inorganic/organic complexes. PhD INSA de Lyon (2008) 206 PROCESSING Real-time monitoring of mixing the other mixtures18, as shown by Figure IV-F18. In any case and whatever the reason for such behavior, this supplementary visualization test made us exclude one of the four photo-active fillers: no further experiments will be performed with CNa+ 0.25CEC RhP. zoom zoom Figure IV-F18 Visual comparison of the masterbatches prepared mixing Siliconöl M10000 with two of the four photo-active lamellar fillers (CNa+ 0.25CEC RhP and C30B 0.25 MC RhP) to perform a preliminary visualization test by Visiovis. The aspect of the mixture prepared with CNa+ 0.25CEC RhP (left) is visibly different from the aspect of the mixture prepared with C30B 0.25MC RhP (right) at the same concentration (0.1% wt). With reference to the method used to prepare the masterbatches: all the mixtures have been prepared by a mixer equipped with a 60 mm dilacerator disk rotating at 1000 rpm for 20 min (Disperser TurboTest Rayneri 33/300P) – exactly the same method used to prepare the controlled mixture for the calibration of the detection systems. We mixed the selected PDMS (Siliconöl M10000) with a proper amount of the photo-active lamellar filler in order to assure a concentration of photo-active lamellar filler of 0.25% wt, as determined by the calibration of the detection systems (§ IV-1.2.4). It is worthy to observe that, after a certain time, the masterbatches prepared with the described method undergo decantation and a significant portion of (not all) the filler sediments. Undoubtedly, this is not a good sign – a suspension of nanosized particles is 18 Parenthetically: as we’ve also prepared some photo-functional inorganic/organic complexes by cation exchange processing with Nile Blue A Perchlorate (Chapter II), we tried to prepare some mixtures of PDMS with CNa+ 0.25CEC NBAP and CNa+ 1CEC NBAP and we compared them to the mixtures of PDMS with C30B 0.25MC RhP and C30B 1MC RhP (the homologue complexes, but functionalized with Rhodamine 6G Perchlorate). We noticed the same differences shown in Figure IV-F18. Antonella ESPOSITO 207 Chapter IV supposed to be stable, since gravity effects should be negligible for very fine particles. The dry photo-active lamellar fillers are certainly characterized by a size distribution19 including more or less fine particles – this would explain the fact that only a portion of filler undergoes sedimentation. In reality, when a lamellar filler is mixed with a polymer matrix and if these two components of the mixture have a high affinity, a certain degree of “spontaneous” exfoliation of the filler (and, thus, an induced reduction of the average particle size) can be observed. In this case, a partial sedimentation of the masterbatches made us early foresee that no driving forces exist for our photo-active lamellar fillers to spontaneously exfoliate into the selected PDMS – neither related to the chemistry of the mixture, nor produced by the shearing effects of mixing. Such forethought will be later proved by rheology measurements20. For the moment we just underline that, if GT5700 pigment was too dense (3.6 g/cm3, Table IV-T2) to avoid sedimentation when mixed up with PDMS (0.97 g/cm3), our photo-active lamellar fillers are less denser but certainly not perfectly compatible – at least in terms of density – with the selected model fluid. Indeed, the clays used to prepare the photo-active lamellar fillers have a density of 2.86 g/cm3 (Cloisite ® Na+), 1.98 g/cm3 (Cloisite ® 30B), 1.90 g/cm3 (Cloisite ® 10A) and 1.66 g/cm3 (Cloisite ® 15A)21. Therefore, their sedimentation is unavoidable (maybe just slower) in the absence of a massive exfoliation. With reference to the method used to inject the masterbatch into the screw/barrel system: we conceived two different modes of injection but only one practical procedure (the only available at the moment). The first mode of injection could be used to model an extrusion step – mixing up pure polymer pellets (which gradually melt) with the dry filler, to formulate composite pellets to be successively used for the fabrication of the final objects. This first injection mode requires the preparation of a three-layer “unmixed masterbatch” composed of a layer of dry filler stacked between two layers of PDMS (total volume 10 mL, equivalent concentration 0.25%wt as previously indicated), to be injected into the system via a 20 19 We couldn’t characterize their particle size distribution because of the tiny amounts of sample we could produce by cation exchange process with the fluorescent molecule (rather expensive). 20 See § V-4.1. 21 Even if the photo-functionalization process could have slightly changed such values, they still represent a good reference for comparisons. PhD INSA de Lyon (2008) 208 PROCESSING Real-time monitoring of mixing mL syringe (previously cut at its extremity to avoid shearing). We tested this injection mode only once, then we had to suspend it because of some problems due to the model materials (poorly compatibles) and the geometry of the screw/barrel system (the actual profile of Visiovis screw has a poor mixing efficiency – unsurprisingly, as we showed that it rather models a meter zone22). Briefly, the filler couldn’t properly be mixed up with the PDMS and, consequently, showed a strong tendency to sediment during – and particularly after – the experiment. Sedimentation resulted in serious problems about the complete purge of the system. The second mode of injection is more adapted to model a portion of the injection molding devices (the meter zone, of course) – melting again the composite pellets previously formulated by extrusion and using the molten mixture for the fabrication of the final objects. This second injection mode requires the preparation of a masterbatch having a good quality of mixing (following the method of the disperser) and a volume of 10 mL, to be injected in the system via the same 20 mL syringe previously described (Figure IV-F19). This is the mode of injection we focused our attention on. Figure IV-F19 Method used to inject the masterbatch (here, the second injection mode is shown) into the screw/barrel system: via the appropriate aperture, using a 20 mL syringe previously cut at its extremity to avoid shearing. In summary, the experimental protocol to test the novel configuration of Visiovis is the following: 22 The geometrical parameters of Visiovis screw/barrel system are listed in Table IV-T1. Antonella ESPOSITO 209 Chapter IV Preparation of the masterbatch. PDMS (Siliconöl M10000) is mixed up with a suitable amount of photo-active filler (target concentration 0.25% wt) then intensive mixing is performed by a mixer equipped with a 60 mm dilacerator disk rotating at 1000 rpm for 20 min (Disperser TurboTest Rayneri 33/300P). We prepared a masterbatch for all the photo-active lamellar fillers that looked suitably detectable (C30B 0.25MC RhP, C10A 0.25MC RhP and C15A 0.25MC RhP). Preparation of the syringe for the masterbatch injection. A standard 20 mL syringe is previously cut at its extremity to avoid shearing while injecting, then it is filled up with 10 mL of the previously prepared masterbatch. Injection of the masterbatch into the system. Once the syringe prepared, the screw/barrel system filled up with neat PDMS (Siliconöl M10000) and purged of all the air bubbles, the room light switched off, the spectrometer zeroed for the black background, the screw rotation set at about 20 rpm (corresponding to a voltage of 15V) and the laser sheet switched on, the syringe is plunged vertically in the aperture and the injection is rapidly achieved in the lowest accessible point. IV-2.1 Acquisition of data Once the injection of the masterbatch into the system executed as previously described, the experiment is officially started and experimental data are automatically acquired: one image per screw revolution is recorded by each CCD camera thanks to the trigger, and a fluorescence emission spectrum is regularly recorded every three seconds by the spectrometer thanks to the optical fiber. The images are stocked sequentially in a *.bmp file named by a code composed by the date and the hour of data recording. At the end of the experiment, each sequence of images can be used to reconstruct the corresponding video. The fluorescence spectra are recorded individually in *.txt files. PhD INSA de Lyon (2008) 210 PROCESSING Real-time monitoring of mixing IV-2.1.1 Images Figure IV-F20 shows a series of consecutive images (top to bottom) acquired by one CCD camera (the first in the progression of the fluid, i.e. the closest to the point of injection – which parenthetically is the camera showing the portion of screw profile in which there’s the transition from the compressing to the pumping zone23) in the dark, after the injection of a masterbatch of C30B 0.25MC RhP (left), C10A 0.25MC RhP (middle), C15A 0.25MC RhP (right). Only the first images are shown, thus these series represent just the beginning of mixing (6 screw revolutions, corresponding to a lapse of time of 18 seconds ca). It is clear that any interpretation of such images would be extremely subjective – unless a suitable image processing is found to “translate” the qualitative information given by the serial images in, at least, some quantitative trends. Obtaining quantitative and absolutely reliable results won’t probably be easy – maybe won’t even be possible – and our present objective is actually to extract some general but provable information. By the way, the possibility of a direct visualization of the mixing progression in any point of the screw/barrel system is innovative and certainly original. IV-2.1.2 Videos As previously said, the series of consecutive images shown in Figure IV-F20 can be also used to reconstruct videos. The same reflections made about the single images (any interpretation would be extremely subjective and certainly biased, somehow) can be transposed to the videos reconstructed from the complete sequences of images. However, no processing algorithms are available to get some quantitative information directly from a video – one would rather use the sequence of images which compose it to perform any quantitative analysis – therefore talking about video processing is here meaningless, as our first source of data are the images. 23 More details about a typical screw profile for injection moulding devices – melting zone, compressing zone and meter zone – are available in Chapter I. Antonella ESPOSITO 211 Chapter IV C30B 0.25MC RhP 0.25% wt C10A 0.25MC RhP 0.25% wt C15A 0.25MC RhP 0.25% wt Figure IV-F20 Three series of consecutive images (top to bottom) acquired by one of the CCD cameras in the dark after the injection of a masterbatch containing C30B 0.25MC RhP (on the left), C10A 0.25MC RhP (in the middle) and C15A 0.25MC RhP (on the right). PhD INSA de Lyon (2008) 212 PROCESSING Real-time monitoring of mixing IV-2.1.3 In-line fluorescence spectra Figure IV-F22 shows some examples of in-line fluorescence-spectra regularly recorded by the spectrometer, interfaced with Visiovis by means of an optical fiber which, as previously explained24, is placed perpendicularly to the laser sheet, in front of the screw/barrel series but on the opposite side of the CCD cameras. The emplacement of the optical fiber is better shown in Figure IV-F21. Figure IV-F21 Two pictures illustrating the position of the optical fiber used to interface the spectrometer with Visiovis and, thus, to regularly collect in-line fluorescence spectra. The fluorescence spectra shown in Figure IV-F22 have been recorded during the visualization tests performed by injection of a masterbatch of C30B 0.25MC RhP (top), C10A 0.25MC RhP (middle), C15A 0.25MC RhP (bottom). The fluorescence emission – whose spectrum is recorded every 3s (integration time 3s) by the spectrometer 25 – cannot support the information supplied by the images and the videos (the phenomena associated to the detection are absolutely different and occur at a totally different length scale) but could eventually complete it. In fact, as previously argued in § IV-1.2.3, the sole CCD cameras are neither adequate nor sufficient for a correct visualization of the phenomena occurring during nanocomposite processing. Spectrofluorimetry, on the other hand, is a technique capable of giving some information which are, surely, spacerestrained, but which could imply a deeper assessment of clay exfoliation mechanisms. 24 25 The actual configuration of Visiovis has been described in § IV-1.3. See § IV-1.2.3. Antonella ESPOSITO 213 Chapter IV Photo-active lamellar filler Laser 532 nm Photo-active lamellar filler Laser 532 nm Photo-active lamellar filler Laser 532 nm Figure IV-F22 In-line fluorescence spectra acquired by Visiovis (details in the text). From top to bottom: C30B 0.25MC RhP, C10A 0.25MC RhP and C15A 0.25MC RhP. PhD INSA de Lyon (2008) 214 PROCESSING IV-3 Real-time monitoring of mixing PROCESSING OF THE ACQUIRED DATA The serial images acquired by each CCD camera (an example is shown in Figure IV-F20) are useful to evaluate visually the temporal evolution of nanofiller distribution in the volume of fluid comprised in between the screw flights, the screw root surface and the barrel surface – in other words, they are useful to estimate the efficiency of the visualized screw profile in terms of distributive mixing. Moreover, images can be used to reconstruct videos which straightforwardly show such temporal evolution – certainly, videos represent a pure qualitative result, but somehow they could help understanding mixing dynamics. We’ve previously avowed that talking about some processing method to get quantitative information from a video is actually meaningless, whereas for serial images (time-related sequence of images) it is possible to conceive some procedure to “translate” the qualitative information in a quantitative trend. With the precious and irreplaceable help of Jean Balcaen we implemented two Matlab programs which process the serial images acquired in presence of the photo-active lamellar fillers on the basis of, respectively, (1) the integral standard deviation of the luminosity of the images, and (2) the Fourier transform of textured images. All the Matlab functions we developed are available in the Appendix. In relation to the collected fluorescence emission spectra, we estimated that their simple visualization as a function of processing time is already a valuable information, thus for the moment we haven’t searched through the possibility of further processing. IV-3.1 Images The images shown in Figure IV-F20 as an example of the data acquired by Visiovis have been taken by one of the four CCD cameras – more specifically by the first camera in the progression of the fluid, i.e. the closest to the point of injection of the masterbatch. Indeed, as previously explained, this first camera shows the portion of the screw profile in which the transition from the compressing to the pumping zone occurs. This zone is certainly interesting; nevertheless, the corresponding visualized volume has a shape which is too complex and inadequate for the image processing we conceived – Antonella ESPOSITO 215 Chapter IV which is actually more adapted to the analysis of a regular rectangular area. Thus, the images in the field of the first camera have been collected, but they won’t be used for any further image processing other than the reconstruction of the videos. The first channel section useful for image processing (i.e. entirely visualized by a single camera) falls in the field of the second camera and, just to begin, we decided to focus on one channel section – as a result, the first and second cameras will be largely enough to perform preliminary visualization experiences by Visiovis. When introducing the topic of the acquisition of data, we said that the images are stocked sequentially in a single *.bmp file named by a code composed by the date and the hour of data recording26. The preliminary procedure for image processing is, thus, to extract each single image from the unique *.bmp file recorded by Visiovis acquisition circuit. This step can be executed by using the Matlab function decoupe.m (available in the Appendix). After that, as a biggest portion of each image is completely black 27 and we’re rather interested to the channel section, a further cutback of the serial images is necessary, in order to reduce the area of the images to be processed just to the rectangle around the channel section: the Matlab function performing this operation is fenetre.m (available in the Appendix). Right after these two basic steps, Visiovis images are ready to undergo the two aforementioned processes28. Figure IV-F23 shows an example of the input (on the left) and output (on the right) images for the function fenetre.m. fenetre.m Figure IV-F23 Example of an input (left) and output (right) image for the function fenetre.m. 26 See § IV-2.1. The black portions of the images correspond to the flight silhouette and the body of the screw. 28 See § IV-3. 27 PhD INSA de Lyon (2008) 216 PROCESSING Real-time monitoring of mixing IV-3.1.1 Standard deviation of image luminosity The image processing performed by the Matlab function ecrtype.m (the M-file is available in the Appendix) is based on the integral standard deviation of the luminosity of Visiovis images. For each picture of a given sequence (input), this function calculates the local values of the luminosity29 and their standard deviations, integrates the values of standard deviation on the whole surface of the picture and then plots the logarithm of the integral standard deviation of the luminosity as a function of the image number in the sequence (i.e. the number of screw revolutions, as each camera acquires one image per screw revolution). Of course, the temporal dependence can be deduced by the rotational speed of the screw – in other words, since the screw turns at 20 rpm ca, one revolution takes 3s and, consequently, the cameras acquire one image every 3s. Interpretation’s key. In theory, the smaller the integral standard deviation of the luminosity is, the more homogeneous the mixture and the more efficient the distributive mixing will be. Limitations. The integral standard deviation will never be smaller than a certain value because of to the difference of the mixture (never completely black) in comparison with the screw profile (always completely black). Moreover, this image processing is not morphology sensitive: two images with different textures may give the same results in terms of standard deviation of luminosity. An example of the plot which can be obtained by this image processing based on the integral standard deviation of the luminosity of the images is shown in Figure IVF24 [1]. Several successive passages of the masterbatch containing the photo-active filler in the field of the camera are easily recognizable. The fact that the first peak rapidly lowers indicates that the masterbatch is gradually diluted by the neat PDMS and 29 We just remind, here, that Visiovis cameras acquire 8bit images, i.e. Visiovis images are represented by a colormap having 28=256 grey levels. Each pixel can assume one of these 256 grey levels, and each level corresponds to a different light intensity (luminosity). Antonella ESPOSITO 217 Chapter IV mixes up with the matrix. In this example, the fluid evolving into the screw/barrel system effectuates 2.5 recirculations in the closed circuit, as it passes three times in the visual field of the acquiring camera. Figure IV-F24 Typical result obtained by the image processing based on the integral standard deviation of the luminosity of Visiovis images – output of the Matlab function ecrtype.m. As the screw rotational speed is 20 rpm ca, the temporal dependence can be derived by multiplying the number of screw revolutions by 3 (one screw revolution every 3s). IV-3.1.2 Discrete Fourier Transform (DFT) of textured images The image processing performed by the Matlab function normft.m (the M-file is available in the Appendix) is motivated by the fact that the Fourier transform (FT) can codify a textured image by the frequencies of repetition of its elementary textural units (e.g. spirals or twisted fluorescent volutes on a homogeneously black background). As any image can be expressed as a function of two discrete spatial variables, the FT of an image is the sum of complex exponentials having different amplitudes, frequencies and PhD INSA de Lyon (2008) 218 PROCESSING Real-time monitoring of mixing phases, and indeed such representation plays a critical role in a broad range of image processing applications (including enhancement, analysis, restoration and compression). If f ( x, y ) is a function of two spatial variables x and y , then we can define the two-dimensional Fourier Transform of f ( x, y ) by the following relationship: F( where ( x F( x , , y x y ) x j f ( x, y)e y xx e j yy dxdy (IV-E1) ) are frequency variables associated to the spatial variables ( x, y ) . , y ) , which can be called the frequency-domain representation of f ( x, y ) , is a complex-valued function periodic in both usually only the range x , x and (period 2 ); for such reason, y of the function is displayed. F (0,0) is the sum y of all the values of f ( x, y ) and, thus, is often called the constant component of the FT. The inverse two-dimensional FT is given by f ( x, y) 1 4 2 F( x x , y )e j xx e j yy d x d y (IV-E2) y Roughly speaking, this expression proves that f ( x, y ) can truly be represented as a sum of an infinite number of complex exponentials (i.e. sinusoids) having different frequencies. The amplitude and the phase spectra of the contributions at the frequencies ( x , y ) are given by F ( x , y ). Just to give an example: let’s consider a function f (m, n) which equals 1 within a rectangle and 0 everywhere else in the plan (Figure IV-F25a): Figure IV-F25b shows the plot of the magnitude spectrum of its FT F ( plot is F (0,0) . The plot shows that F ( m , n m , n ) . The peak at the centre of the ) has more energy at high horizontal than at high vertical frequencies: this means that f (m, n) horizontal cross sections are narrower pulses than vertical cross sections (narrow pulses have more high-frequency content Antonella ESPOSITO 219 Chapter IV than broad pulses). Please note that small dimensions in the Euclidean space correspond to high frequencies in the Fourier space, and vice versa (m vs. n rectangle dimensions). (a) (b) Figure IV-F25 A simple rectangular function (a) and the amplitude of its Fourier transform represented as a mesh plot (b). Another common way to visualize the FT is to display the log F ( m , n ) as an image (Figure IV-F26): the fact of using the logarithm helps to bring out more details of the FT in regions where F ( m , n ) is very close to 0. Figure IV-F26 The logarithm of the Fourier transform of a simple rectangular function (Figure IV-F25a) represented as an image. Figure IV-F27 shows two additional examples of the FT amplitude spectra for simply-shaped functions (a tilted rectangle, on the left, and a circle, on the right). These additional examples show that the FT is sensitive not only to the eventual presence of image textures, but also to their position and orientation (i.e. to the image symmetry). PhD INSA de Lyon (2008) 220 PROCESSING Real-time monitoring of mixing Figure IV-F27 Two additional examples of Fourier transform for simply-shaped functions. The FT is sensitive to the symmetry of the image, i.e. to the position and orientation of its textural features. Actually, the amplitude spectra given by the FT are invariant to translation – a simple translation of a given textural feature doesn’t modify the amplitude but only the phase of the sinusoidal contributions. On the other hand, the TF is sensitive to rotation. Let’s consider a sinusoid of period T having a given initial orientation (Figure IV-28a): its FT is represented by two peaks aligned in the same direction, corresponding to the frequencies 1 T and 1 T . A rotation of an angle makes the FT analogously rotate, without changing its global appearance (Figure IV-28b). Figure IV-F28 Sinusoidal surface of period T parallel (left top) and tilted (left bottom) with respect to the x axis, and the corresponding FT (on the right). Antonella ESPOSITO 221 Chapter IV Evidently, we took into account these properties (which are typical of the FT) when we conceived the Matlab function normft.m for Visiovis image processing. Working on a computer with digital images (i.e. constituted by pixels) requires using a form of the transform which is known as the Discrete Fourier Transform (DFT). There are two principal reasons for using this form: (1) the input and output of the DFT are both discrete, which makes it convenient for computer manipulations; (2) there is a fast algorithm for computing the DFT, known as the Fast Fourier Transform (FFT). The DFT is defined for a discrete function f (m, n) that is nonzero only over a finite region 0 m M 1 and 0 n N 1 (which is exactly the case of a digital image). The two- dimensional M-by-N DFT and inverse M-by-N DFT relationships are given by: M 1N 1 F ( p, q) f (m, n)e j 2 pm M e j 2 qn N where m 0n 0 2 f (m, n) 2 qn j pm j 1 M 1N 1 F ( p , q )e M e N where MN p 0 q 0 p 0,1,..., M 1 (IV-E3) q 0,1,..., N 1 m 0,1,..., M 1 n 0,1,..., N 1 (IV-E4) The Matlab built-in functions fft, fft2 and fftn implement the FFT algorithm for computing the one-dimensional, two-dimensional and N-dimensional DFT respectively. As we were going to deal with images (two-dimensional discrete functions), we were mainly interested in the 2D-DFT Matlab built-in function. The DFT of a digital image gives a spectrum of all the frequencies comprised between a maximum frequency and a minimum frequency 30 m in MAX . The maximum frequency which can be associated to a textured digital image is the frequency intrinsically generated by the regular presence of pixels (maximum of details), whereas the minimum frequency is related to the physical image dimensions31 (minimum of details). The scheme in Figure IV-F29 clarifies the idea of image maximum and minimum frequencies. 30 In other words, the DFT associates a power value (i.e. a value of squared amplitude) at each frequency comprised between the maximum and the minimum frequency. 31 If the image is rectangular, it is related to its biggest dimension. PhD INSA de Lyon (2008) 222 PROCESSING Real-time monitoring of mixing MAX 480 pixels m in 640 pixels Figure IV-F29 The maximum and minimum frequencies which can be associated to a digital image (the dimensions of Visiovis images are 640 × 480 pixels and each pixel visualizes an area having a maximum dimensions of about 83 m, as previously said in § IV-1.2.3). Knowing the nature of the information that a FT can give about a textured digital image, the image processing performed by the Matlab function normft.m is aimed to the codification of the Visiovis images by the frequencies of their elementary textural units (namely, spirals or twisted fluorescent volutes in a homogeneously black matrix). The higher the main coding frequencies, the finer the texture of the digital image: an image which is finely textured is coded by high frequencies, since the highest frequencies are able to represent the finest details. We saw how the FT highlights any regular, repetitive structure which appears in a digital image considered as a 2D signal, function of the two spatial coordinates; we underlined, as well, that the FT is sensitive to image symmetry and that the minimum frequency depends on the maximum physical dimension of the image. Indeed, Visiovis images are rectangular (their biggest dimension is the diagonal of the rectangle, whereas their smallest dimension is their height, i.e. 480 pixels) and only circles are perfectly symmetric in all the directions, which means that only circles can be frequentially isotropic. These observations let guess that a preliminary processing step is necessary to get rid of any dimensional incongruity before calculating the FT of our images. This is the reason why, before doing any other action, the anamorphosis of each picture of a given series is operated, thus eliminating any problem of aspect ratio of the rectangular images and “forcing” Visiovis images to be frequentially isotropic. The Matlab function performs then the FT of each anamorphous image, calculates the logarithm of the squared norm log F ( p, q) , subtracts the noise previously calculated on a reference image, calculates the average amplitude for each Antonella ESPOSITO 223 Chapter IV frequency and plots the averaged values as a function of the frequency. Finally it associates, to each image of a given sequence, the most representative frequency (i.e. the mean frequency weighted by the intensities, in other words the frequency statistically most probable) and plots such values as a function of the number image in the sequence (i.e. the number of screw revolutions, as each camera acquires one image per screw revolution). The temporal dependence is deduced by the screw rotational speed – as the screw turns at 20 rpm ca, one revolution takes 3s and, thus, the cameras acquire one image every 3s. Interpretation’s key. Theoretically, a zero frequency should correspond to the perfect mixture homogeneity; on the contrary, high frequencies reflect the presence of regular and fine textural units. In case of perfect miscibility. The lower the representative frequency, the more homogeneous the mixture, the more efficient the distributive mixing will be. In case of reduced miscibility. The higher the representative frequency, the finer the texture, the more efficient (though incomplete) the distributive mixing. Limitations. It is actually impossible to obtain a truly zero frequency since images have, by definition, finite dimensions (indeed, only an infinite image could have a zero minimum frequency). Hence, if a zero frequency is obtained, it should be certainly interpreted as a relative value. Moreover, the contrast between the mixture (never completely black) and the screw profile (always completely black) produces a Heaviside (i.e. a step-like) function, which corresponds to an artifactual permanent texture of the image. An example of the plot which can be obtained by this image processing based on the Discrete Fourier Transform of textured images is shown in Figure IV-F30 [1]. Once again, several successive passages of the masterbatch containing the photo-active filler in the field of the camera are easily recognizable, but the physical meaning of the observed phenomena is not the same as the previous image processing (§ IV-3.1.1). By PhD INSA de Lyon (2008) 224 PROCESSING Real-time monitoring of mixing the way, the fact that the first peak rapidly lowers still indicates that the masterbatch is gradually diluted by the neat PDMS and mixes up with the matrix. Figure IV-F30 Typical result obtained by the image processing based on the Discrete Fourier Transform (DTF) of textured Visiovis images – output of the Matlab function normft.m. As the screw rotational speed is 20 rpm ca, the temporal dependence can be derived by multiplying the number of screw revolutions by 3 (one screw revolution every 3s). IV-3.1.3 Validation of data processing As Moguedet and coworkers [4] developed an analytical model to describe the flow of a viscous Newtonian fluid in the helical rectangular channel of a screw [7] (validated by three-dimensional finite elements calculations in the Matlab environment), we solicited Yves Béreaux to collaborate to validate our methods for data processing. By adapting the Matlab code previously developed, Yves Béreaux generated a sequence of images showing the deformation of a blob of tracing masterbatch into the helical rectangular channel of Visiovis screw/barrel system. Under the hypothesis that the blob is initially spherical, Béreaux represented, in the longitudinal section of the screw/barrel system, a white circle on a black background: as the model is supposed to reproduce the trajectories of a particle in the screw channel, it also allows to trace its position in a Antonella ESPOSITO 225 Chapter IV given longitudinal section of the screw channel as a function of time. By modeling the initial spherical blob as a given number of particles disposed to form a filled circle, Béreaux could reconstruct the deformation of the blob reproducing by computer simulation the trajectories of the particles but rather tracing their reciprocal distances in conditions of pure recirculation, i.e. under the hypotheses that: molecular diffusion can be neglected32; the global flow rate is nil and the mixing of the blob of tracing masterbatch with the neat fluid occurs in a single screw channel, viz. in the space confined between two adjacent flights, the screw root surface and the barrel surface33. Some of the images obtained by this method are shown in Figure IV-F31 (reproduced with the permission of the author). We validated only the second method for data processing – the one based on the DFT of textured images – since the first one, based on the standard deviation of image luminosity, is more intuitive and its principle is easier to understand. In addition, the hypothesis made about the absence of molecular diffusion makes the images generated by computer simulation rather inadequate to be compared with the images acquired by Visiovis: a comparison of Figure IV-F20 and Figure IV-F31 reveals that the obtained images cannot represent the gradients of luminosity (in terms of brightness and fuzziness of the white spirals on the black background) which, on the contrary, are unavoidable in any image acquired during the experiments. Indeed, this difference could affect also the validation of the second method (it actually does, as we’ll show later on), but certainly to a lesser extent. The result obtained by the DFT method on the numerical images designated as the reference is shown in Figure IV-F32 (a): this curve has to be compared to the first part of the graph in Figure IV-F30 (reported in Figure IV-F32 (b) to ease the comparison), as the curve in its entirety actually represents an experiment which includes three passages of the tracing blob in front of the camera (recirculation operated by the tube for closed circuit34). 32 This hypothesis is deduced from the absence of fuzziness from numerical images: Figure IV-F20 shows that this hypothesis is unrealistic for the system observed by Visiovis. Even if slow, molecular diffusion should not be neglected in real systems. Besides, high concentrations of the tracing masterbatch produce a significant diffusion of the laser source, which definitely is the main cause of dizziness (see § IV-4.2). 33 These conditions could be approximately reproduced experimentally by duly tuning the back pressure: indeed, back pressure facilitates the recirculation flow and decreases the global flow rate (§ IV-4.3). 34 A description of the actual configuration of Visiovis has been given in § IV-1.3 (Figure IV-F17). PhD INSA de Lyon (2008) 226 PROCESSING Real-time monitoring of mixing 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 …30 …35 …40 …45 …50 Figure IV-F31 Sequence of numerical images generated by Yves Béreaux (by means of a code previously developed [7] and suitably adapted to our requirements), showing the deformation of a spherical blob of tracing masterbatch in Visiovis screw channel, under the hypothesis of pure recirculation and absence of molecular diffusion. These images validated the data processing based on the DFT of textured images. Images reproduced with the author’s permission. Antonella ESPOSITO 227 Chapter IV (a) (b) Figure IV-F32 Comparison of the typical results obtained by the processing based on the DTF method (Matlab function normft.m) applied to (a) the images generated by numerical simulation (Figure IV-F31) and (b) the images acquired during Visiovis experiments (Figure IV-F20). The comparison validates data processing but shows some slight divergences – surely based on the different origin of the processed images (more details in the text). The only difference in Figure IV-F32 is that the average representative frequency as a function of the image number for the images generated by computer simulation (a) is characterized by two slope values, whereas the homologous curve obtained for the images acquired by Visiovis (b) presents a slope which varies continuously and assumes at least three different values. By the way, one could deliberately decide to fit the curves obtained by processing acquired images (Figure IV-F32 (b)) with the model revealed by the curve obtained with numerical images (double-slope curve, Figure IV-F32 (a)) – so that two main parameters35 could be deduced and used to compare different systems. A further validation of the DFT method for Visiovis data processing is provided by the frequency distribution obtained for each numerical image of the reference series (Figure IV-F33)36. A visual explication of the features appearing in the graph of Figure IV-F33 is given in Figure IV-F34. At the beginning, the peak represents the minimum 35 Indeed, it is the results obtained applying the DFT data processing to the ideal case of numerical images that proved that mixing could be probably described by two parameters: the slope values of the curve showing the average representative frequency as a function of the number of screw revolutions. 36 The color scale is just an arbitrary scale which shows only the relative “importance” of all the possible frequencies present in each image of the reference series. PhD INSA de Lyon (2008) 228 PROCESSING Real-time monitoring of mixing frequency which could ever be due to morphological features – related to the maximum dimension of the blob (i.e. initial diameter) (see Figure IV-F34 (a)). Since the blob gets stretched and thinner, the white circle is gradually transformed in a lengthening lamella and its characteristic dimension decreases (as expected for a laminar flow, the thickness of the striations decreases): the width of the features in the frequency distribution graph increases (Figure IV-F34 (b)) as shown by the boundary identified by line C. Moreover, as the lamellae stretch and form spirals because of recirculation, the statistical number of intersections observed in any cross section increases and their frequency of repetition increases as well: an additional component in the frequency distribution appears (Figure IV-F34 (c)) producing the peak represented by line B. Indeed, any regular distribution of white features (spiral intersections) on a black background produces, in the frequency domain, a Dirac impulsion whose position varies with the distance (and thus, indirectly, with the number and thickness) of the spirals – in particular, as their reciprocal distance decreases (i.e. as their number increases and their thickness decreases), the DFT gives a peak which shifts towards higher frequencies (Figure IV-F34 (d)): line B bends towards higher frequencies. Concurrently, a further phenomenon occurs: the augmentation and thinning of the lamellae gradually expand the percentage of occupation of the black background by the white features. A better occupation of the space by repeating features is visualized, in the DFT frequency distribution, by an increased intensity of the peak associated to the minimum frequency (see Figure IV-F35), corresponding to line A. This phenomenon helps understanding why, in the graphs of Figure IV-F32, the average frequency diminishes in spite of the appearance of higher-frequency features in the DFT domain and of their further shifts towards higher frequencies: like any averaged value, the representative frequency depends on the whole distribution. An increased intensity of line A (Figures IV-F33 and IV-F35) indicates an improvement of global distributive mixing (repartition of the striations), whilst the appearance and gradual shift of peaks at higher frequencies (lines B and C) represent an improvement of local distributive mixing (striation thickness). Antonella ESPOSITO 229 Chapter IV A B C Figure IV-F33 Frequency distribution obtained for the numerical images (Figure IV-F31) by the DFT method (§ IV-3.1.2). The explications given in the text are illustrated in Figure IV-F34. 1 intercept initial blob (a) 1 intercept thinner striation decreasing width increasing width (b) 2 intercepts (c) 7 inter. increasing frequency shift towards higher frequencies (d) Figure IV-F34 Visual explication of the features in the graph of Figure IV-F33. PhD INSA de Lyon (2008) 230 PROCESSING Real-time monitoring of mixing A B C low frequencies image number high Figure IV-F35 Three-dimensional visualization of the frequency distribution shown in Figure IV-F33. Note that line B undergoes just a frequency shift, whereas line A increases in intensity. IV-3.2 Videos Once the single images extracted and isolated from the sole *.bmp file recorded by Visiovis acquisition circuit (§ IV-3.1), the obtained images can be considered (since they are!) as the frames of a temporal sequence and, thus, used to reconstruct a video. This operation can be executed by running the Matlab function video.m (available in the Appendix). This function also generates a montage of the video, i.e. a panel visualizing at a glance all the frames used for the reconstruction. The montage is particularly interesting when the video is reconstructed from the windows cutback around the channel section, as shown on the example in Figure IV-F36. Antonella ESPOSITO 231 Chapter IV Figure IV-F36 Example of montage of a video reconstructed from the windows cutback around the channel section – one of the output of the Matlab function video.m. Acquisition done by the second camera in the dark after the injection of a masterbatch containing C30B 0.25MC RhP. IV-3.3. Fluorescence spectra The easiest way to visualize the regularly acquired in-line fluorescence emission spectra as a function of processing time is to convert the temporal sequence of (intensity of fluorescence emission vs. frequency) curves in a 3D shaded surface plot to show the results on a rectangular region delimited by the processing time [s] and the wavelength [nm], as shown in Figure IV-F22. The Matlab function used for the visualization of the fluorescence spectra is called spctr.m and is available in the Appendix. If the images (and the results of image processing) can be used to evaluate the distributive mixing into the screw/barrel system, the fluorescence spectra may rather inform about dispersive mixing – since the emission signature of the fluorescent dye PhD INSA de Lyon (2008) 232 PROCESSING Real-time monitoring of mixing used to photo-functionalize clay is sensitive to any change of its molecular environment due to clay intercalation and/or exfoliation, and such morphological changes are a direct consequence of dispersive mixing. Manifestly, all the detection systems the Visiovis is equipped with are complementary and useful to characterize nanocomposite processing. IV-4 SOME EARLY RESULTS Conscious that the development of a tool for the visualization of nanocomposite processing, along with the conception of brand new detection and/or characterization systems, are far from being easy, we kept testing Visiovis with different systems and in different conditions (when and if possible), hopeful to recognize its advantages and its limitations. Keeping in mind that any test, by definition, may give negative results (or, even, may not give any result), we performed some experiments in order to evaluate: the behavior of three different photo-active lamellar fillers – namely C30B 0.25MC RhP, C10A 0.25MC RhP and C15A 0.25MC RhP37; the possibility of using Visiovis to perform conventional tracing experiences – that is, injecting a masterbatch containing a given amount of the pristine commercial clay and a smaller amount of the same clay previously rendered photo-active; the influence of the back pressure on the visualization executed by Visiovis – thanks to an additional van capable of regulating the flow in the closed circuit. IV-4.1 Comparison of different photo-active lamellar fillers While dealing with the modification of Visiovis configuration and planning how to perform the first visualization tests, we dwelled on clay modification since we needed to prepare a suitable photo-active lamellar filler to be used with Visiovis – coherently with the main objective of our work. In Chapter II we detailed the actions we went through to establish an efficacious (and efficient) photo-functionalization protocol and, in Chapter III, we characterized four inorganic/organic complexes that we succeeded to 37 More details about the preparation and the characterization of the three photo-active lamellar fillers are available in Chapters II and III. Antonella ESPOSITO 233 Chapter IV render photo-functional by cation exchanging four commercial clays with a rhodamine dye. However, we remind that during the tests performed to establish an appropriate experimental protocol for Visiovis in its new configuration38, we realized that CNa+ 0.25CEC RhP couldn’t be used in the visualization conditions imposed by Visiovis. In conclusion, for the moment we disposed of three different photo-active lamellar fillers, namely C30B 0.25MC RhP, C10A 0.25MC RhP and C10A 0.25MC RhP. It is with these photo-active lamellar fillers that we performed our first visualization experiences – respecting the experimental protocol previously established. The videos showing the three experiences are available in the multimedia CD-Rom accompanying the PhD manuscript. As expected, images and videos are captivating but really can’t help interpreting the results of the visualization experiences. Figure IV-F37 and Figure IV-F38 show the trends revealed by the image processing previously described and based, respectively, on the integral standard deviation of the image luminosity (§ IV-3.1.1) and on the DFT of textured images (§ IV-3.1.2). In all the curves of Figure IV-F37 the passages of the masterbatch containing the photo-active filler (0.25% wt) in the field of the camera are easily recognizable. During the first passage (which occurs more or less at the same time after the injection of the masterbatch – unsurprisingly, since the pumping effect of the screw is the same for all the samples) the C30B 0.25MC RhP appears less homogeneously distributed than the other samples, as its integral standard deviation is higher meaning that there are intense variations of luminosity in the visualized channel section. We can also observe that the first peak for all the samples is, actually, a double peak: this reflects the fact that the fluorescent masterbatch doesn’t proceed as a block but is stretched, plied and deformed by the action of the screw, thus at the moment of passing in the field of the acquiring camera (which is the second one, thus closer to the middle of the screw rather than to the point of injection) several of its portions arrive with a slight temporal shift. Anyway, this curve doesn’t inform about the actual morphology of the mixture at the right instant of the acquisition (the other image processing will probably do). After this first passage, 38 See § IV-2. PhD INSA de Lyon (2008) 234 PROCESSING Real-time monitoring of mixing the standard deviation of luminosity of course decreases, then increases again at the moment of the second passage in front of the camera, after a complete recirculation. As expected, the second peak is lower than the first one, witnessing the homogenization of the masterbatch with the neat PDMS. One could notice that at the first passage the most similar behaviors were those of C10A 0.25MC RhP and C15A 0.25MC RhP; at the second passage, on the other hand, C30B 0.25MC RhP got closer to C10A 0.25MC RhP whereas C15A 0.25MC RhP makes the difference. First of all, in the case of C30B and C10A fillers, even the second peak seems to be double while the second peak of C15A is clearly single and broad; besides, the whole curve of C15A is constantly lower than the other two curves. These early results don’t show any major difference in distributive mixing for different photo-active lamellar fillers, even if some slight details can anyway be perceived. Figure IV-F37 Processing based on the integral standard deviation of luminosity. Comparison of different photo-active fillers (C30B 0.25MC RhP, C10A 0.25MC RhP, C15A 0.25MC RhP) following the experimental protocol previously established (rotational speed 20 rpm ca). Antonella ESPOSITO 235 Chapter IV Figure IV-F38 Processing based on the DFT of textured images. Comparison of different photo-active fillers (C30B 0.25MC RhP, C10A 0.25MC RhP, C15A 0.25MC RhP) following the experimental protocol previously established (rotational speed 20 rpm ca). Inside: y-zoom of the background plot. 75 is the arithmetical mean value of frequency (reference max value 150). In all the curves shown in Figure IV-F38 the two passages of the masterbatch in the visual field of the camera are easily recognizable as well, and occur at the same moment as it has been previously detected (Figure IV-F37). In the frequency-domain, the image processing produces an additional feature whose explanation is in the processing itself. A point having a frequency of 75.00 is present in all the curves and is the same for all the photo-active fillers: indeed, it doesn’t mean anything as it just represents the image chosen as the reference (i.e. the image just before the arrival of the very first portion of the fluorescent masterbatch in front of the acquiring camera) and, thus, it corresponds to the sole image which is really, completely black (for the background numerical noise is spotlessly subtracted). Therefore, we zoomed to the upper portion of the curves and we showed them again in the inner graph in Figure IV- PhD INSA de Lyon (2008) 236 PROCESSING Real-time monitoring of mixing F38. It is rather upsetting (even if probably expected) to find, in these curves, most of the features previously observed and commented about Figure IV-F37. We’ve already explained, however, that the physical reasons of such observed phenomena aren’t the same – and this can be directly deduced from the fact that the two image processing don’t derive from the same intuition. We preferred think, then, that no significant difference exist between the mixtures processed with the three different photo-active lamellar fillers – at least not in the conditions offered by Visiovis (matrix, screw profile, speed, temperature). At least, these early results have the merit of illustrating Visiovis potentiality. With reference to the in-line fluorescence spectra: the data acquired during these experiences with the three photo-active lamellar fillers have been previously shown in Figure IV-F22. Effectively, no significant differences can be detected by comparing the fluorescence emission spectra either – apart the behavior of C15A 0.25MC RhP, which looks slightly different. Finally, we could just observe that the three photo-active fillers seem to be more similar in terms of “morphology of the mixture” (Figure IV-F38) than in terms of “distribution into the channel section” (Figure IV-F37). IV-4.2 Comparison of different amounts of filler In preparation of the “qualitative calibration” of Visiovis detection systems 39 we announced that we were planning to focus essentially on the first photo-active lamellar filler produces by the photo-functionalization protocol described in Chapter II. The results obtained by comparing the three different photo-active lamellar fillers we disposed of convinced us to continue testing the system only (at least for the moment) with C30B 0.25MC RhP. Manifestly, the factors influencing these early results are several and not always easy to control. We should remind, here, that owing to the calibration, we found that the optimum concentration of the photo-active lamellar filler in the masterbatch for the injection in Visiovis screw/barrel system is 0.25% wt. We decided, however, to test the visualization capability of Visiovis with masterbatches 39 The calibration of the detection system has been detailed in § IV-1.2.4. Antonella ESPOSITO 237 Chapter IV containing different amounts of filler. As we didn’t want to change the optimum concentration of photo-active filler, we simply prepared three masterbatches following the procedure previously described but adding, besides the photo-active filler, a given amount of the corresponding pristine clay. Apart from this, we didn’t change the experimental protocol. Briefly, we tested the system with the following masterbatches: Siliconöl M10000 + C30B 0.25MC RhP (0.25% wt) = total amount of filler 0.25% wt Siliconöl M10000 + C30B 0.25MC RhP (0.25% wt) + C30B (0.75% wt) = total amount of filler 1% wt Siliconöl M10000 + C30B 0.25MC RhP (0.25% wt) + C30B (2.75% wt) = total amount of filler 3% wt We must admit that these tests were risky: the calibration of the system already revealed that an excessively high concentration of filler (whether photo-active or not) can cause some optical problems, since as the concentration increases the mixture with PDMS becomes less and less transparent and the penetration depth of the laser sheet rapidly decreases. Nevertheless, we thought that it could be worthy trying. Figure IVF39 shows the trends revealed by both our image processing approaches. The integral standard deviation of luminosity doesn’t show any particular trend in relation with the presence of the pristine clay and to the fact that we made its amount vary. The only remarkable points would be the shape of the first peak obtained for the injection of C30B/C30B 0.25MC RhP 2.75/0.25% wt (i.e. the masterbatch containing the highest total amount of filler), as well as the shape of its second peak, which is more similar to a single than to a double peak. Otherwise, adding an amount of pristine clay 3 times higher than the amount of the photo-functionalized clay (C30B/C30B 0.25MC RhP 0.75/0.25% wt) doesn’t significantly change the shape of the curve – which is a rather good conclusion in the future eventuality of using Visiovis to perform real tracing experiences, viz. experiences in which only a fraction of the filler acts as a tracer. The same observations can be made about the results of the image processing based on the DFT of textured images (Figure IV-F39, bottom). PhD INSA de Lyon (2008) 238 PROCESSING Real-time monitoring of mixing Figure IV-F39 Processing based on the integral standard deviation of luminosity (top) and on the DFT of textured images (bottom). Comparison of three masterbatches containing different total amounts of filler (0.25%, 1% and 3% wt) but always the same amount of the photo-active filler C30B 0.25MC RhP (0.25% wt). Injection executed following the standard experimental protocol (rotational speed 20 rpm ca). Antonella ESPOSITO 239 Chapter IV We previously stated that no significant differences could be perceived upon the addition of different amounts of pristine clay to the initial masterbatch (0.25% wt photoactive filler). Indeed, the fact that nothing seems to change (or even any little change we could discover) may have an explanation which cannot be found uniquely in the curves. This is a typical example of situation in which, even if we’re aware that Visiovis images and videos can’t really provide quantitative information on our experiences, we realize that they efficiently support the interpretation, e.g. by suggesting the causes which could possibly be attributed to ambiguous features present on the curves. 1st camera 2nd camera (a) (b) (c) Figure IV-F40 Selected images acquired by the first (on the left) and the second camera (on the right) after the injection of C30B/C30B 0.25MC RhP 0.00/0.25% wt (a), 0.75/0.25% wt (b) and 2.75/0.25% wt (c), respectively. A zoom is made on the processed image window. PhD INSA de Lyon (2008) 240 PROCESSING Real-time monitoring of mixing The images and the videos reconstructed for the visualization experiences we’re trying to interpret actually reveal the significant consequences of adding an increasing amount of pristine clay to the standard masterbatch – containing the optimum amount of filler. Here we reported some selected frames (Figure IV-F40) to support our comments, but the videos are also available in the CD-Rom accompanying the PhD manuscript. The uncertainties claimed at the beginning of this paragraph (and justified by the observations made during the calibration of the detection systems) concretized, and with the help of the visualization experiences here presented we could finally proved them. The concentration of filler in the tracing masterbatch has to be attentively controlled and reduced, when possible, for the highest the concentration of filler, the worst the optical clarity of the system will be. (a) (b) (c) Figure IV-F41 In-line fluorescence spectra acquired after the injection of C30B/C30B 0.25MC RhP 0.00/0.25% wt (a), 0.75/0.25% wt (b) and 2.75/0.25% wt (c), respectively. Antonella ESPOSITO 241 Chapter IV By the way, at least theoretically, the in-line fluorescence spectra aren’t affected by such concentration issues: when a tracing experiment has to be performed, in which a higher concentration of filler is supposed to be used, spectrofluorimetry is a valuable alternative to the image acquisition and processing (Figure IV-F41). IV-4.3 Regulation of the back pressure The last visualization test we performed on Visiovis was aimed to verify whether our tool is capable or not to detect the effect of the application and regulation of a back pressure to the system. In reality, any screw/barrel system is subjected to an intrinsic back pressure due to the presence of a restrained section at the exit of the system. The back pressure represents an obstacle to the flow – it increases the residence time of the fluid into the screw/barrel system and is partially responsible of the recirculation within the screw channel. Sometimes, the back pressure has to be increased to accentuate these phenomena and consequently ameliorate the quality of mixing. Keeping in mind the relevance of such processing parameter, we performed three tests to visualize the effects of an eventual variation of the back pressure on mixing. We preferred varying this parameter (instead of the rotational speed of the screw and/or the viscosity of the model fluid) since, for several reasons, we estimated that it was the only factor capable of influencing the mixing process in the conditions imposed by Visiovis. Indeed, PDMS is a Newtonian fluid40 and the screw profile we dispose of41 isn’t the best for mixing. To increase the back pressure (and also to be able to regulate it), we equipped the existing tube for closed circuit of an additional van, which doesn’t alter the circuit if fully opened but can almost completely arrest the flow when fully closed. Besides, any intermediate position is also possible. With such additional van, we could perform three tests in the presence of different back pressures, namely: (1) a test in which the van is fully open (100% OPEN), (2) a test in which the van is half open (50% OPEN) and (3) a 40 Newtonian fluids are characterized by a value of viscosity which doesn’t depend on the shear rate. We remind, here, that Visiovis geometrical parameters are comparable to the typical design parameters of the meter section of industrial devices. The meter section of a screw isn’t, by definition, the most suitable for mixing. 41 PhD INSA de Lyon (2008) 242 PROCESSING Real-time monitoring of mixing test in which the van is fully closed (100% CLOSE). All these tests have been carried out, as usually, by injecting a masterbatch (PDMS Siliconöl M10000 + 0.25% wt C30B 0.25MC RhP) prepared as previously described42. Figure IV-F42 show the results of the tests 100% and 50% OPEN. We couldn’t show the results of the three tests all together, for the test 100% CLOSE lasted 9 times longer than the other tests. The longer duration of the last test is quite obvious, as 100% CLOSE means that the van almost stops the flow and accentuate to the maximum extent the recirculation within the screw channel, as the corresponding video shows (available in the CD-Rom accompanying the PhD manuscript). It’s interesting to observe that, when the van is half open (50% OPEN), the first peak of the curve (log integral std dev of luminosity vs. time) doesn’t really change, but the second one looks smoothed and the whole curve is markedly lowered. However, the fact that the van is half opened must have produced only a slight increase of the back pressure, because the residence time of the fluid into the screw barrel system is basically the same (the position of the peaks hasn’t changed). The same comments can be made for the curve (average representative frequency vs. time). On the contrary, when the van is fully closed (100% CLOSE) (Figure IV-F43) the back pressure is at its greatest value and the prevalent phenomenon is the recirculation within the screw channel: the fluid proceeds very slowly and has enough time to mix up with the neat matrix before arriving in the field of the acquiring camera. This is the reason why the curves for the last test are completely different than the curves for the former ones. The limits of mixing are achieved when the curves reach their asymptote. Note that, apart from the time scale, these graphs have the same scale as the graphs shown in Figure IV-F42 to facilitate the comparison. For all the performed tests, the fluorescence spectra reflect the same behaviors observed by the cameras. 42 See § IV-2. Antonella ESPOSITO 243 Chapter IV Figure IV-F42 Processing based on the integral standard deviation of luminosity (top) and the DFT of textured images (bottom), respectively. The masterbatch contains 0.25% wt of a photoactive filler (C30B 0.25MC RhP). The position of the van for the regulation of back pressure is: fully open (100% OPEN) and half open (50% OPEN). Rotational speed 20 rpm ca. PhD INSA de Lyon (2008) 244 PROCESSING Real-time monitoring of mixing Figure IV-F43 Processing based on the integral standard deviation of luminosity (top) and the DFT of textured images (bottom), respectively. The masterbatch contains 0.25% wt of a photoactive filler (C30B 0.25MC RhP). The position of the van for the regulation of back pressure is: fully close (100% CLOSE). Rotational speed 20 rpm ca. Note the duration of the test (45 min). Antonella ESPOSITO 245 Chapter IV IV-5 CONCLUSIONS In this chapter we presented Visiovis, an original and innovative tool suitable for visualizing viscous fluids flowing in a geometrically complex system – more precisely a screw/barrel system. After a brief summary of Visiovis origins – who assembled it, for which applications and in which initial configuration – we went through all the steps we had to traverse to suitably modify its configuration in order to adapt it to our new needs. Indeed, we were planning to adapt the existing tool to the analysis of nanofiller dispersion/distribution mechanisms in molten thermoplastic polymers or, eventually, in uncured thermoset resins. To attain our objectives, we certainly had to change Visiovis original configuration – but we decided to do it steadily for both practical and economic reasons. The choice of gradually but incessantly change Visiovis configuration made us work in a situation in constant evolution, being ceaselessly faced to new and unexpected problems to be solved. Notwithstanding, we dared developing two detection systems and tried to exploit the acquired data as much as possible. In this chapter we described how we performed the visualizations, how the CCD cameras and the spectrometer are integrated on Visiovis, how we collected and processed the experimental data. All the results shown in this chapter required a lot of work and sometimes didn’t result as expected. Our main objective was, realistically, to increase Visiovis potentialities and to suggest a further, possible employment for a tool which has already demanded large efforts. Have we fulfilled such requirements? In the next and last chapter we’ll rapidly summarize the technical progresses already achieved on Visiovis and we’ll discuss of some possible further ameliorations. PhD INSA de Lyon (2008) 246 PROCESSING Real-time monitoring of mixing IV-R REFERENCES [1] Esposito A, Balcaen J, Duchet-Rumeau J, Charmeau JY. Visiovis: monitoring nanofiller dispersion/distribution in molten polymers. JEC Composites Magazine 2008, 41, 67-71. [2] Gao, F. Clay-polymer composites: the story. Materials Today 2004, November, 50-55. [3] Liu, J., Boo W.-J., Clearfield A. et al. Intercalation and exfoliation: a review on morphology of polymer nanocomposites reinforced by inorganic layer structures. Mater. Manuf. Processes 2006, 20, 143-151. [4] Moguedet M. Développement d'un outil d'aide à la conception et au fonctionnement d'un ensemble vis-fourreau industriel – Application à l'injection de thermoplastiques chargés fibres de verre longues. Thèse. Lyon: INSA de Lyon, 2005, 124 p. [5] Esposito A, Charmeau JY, Duchet-Rumeau J. Analyse des mécanismes de dispersion de nanocharges dans un polymère fondu. Conséquences sur la morphologie de nanocomposites obtenus par injection. CR des 15èmes Journées Nationales AMAC sur les Composites (JNC15), 361. Marseille, 2007, 1216 p. ISBN: 978-2-87717-090-1. [6] Agassant J.-F., Avenas P., Sergent J.-P. et al. La mise en forme des matières plastiques. 3rd Ed. Paris: Tec & Doc Lavoisier, 1996, 640 p. ISBN: 9782743000165. [7] Béreaux Y., Moguedet M., Raoul X. et al. Series solutions for viscous and viscoelastic fluids flow in the helical rectangular channel of an extruder screw. J. Non-Newtonian Fluid Mech. 2004, 123 (2-3), 237-257. [8] Béreaux Y., Charmeau J.Y., Moguedet, M. A simple model of throughput and pressure development for single screw. J. Mater. Process. Technol. 2009, 209 (1), 611-618. [9] Ottino J.M. The kinematics of mixing: stretching, chaos, and transport. 1st Ed. Cambridge: Cambridge University Press, 1989, 364 p. ISBN 0-521-36878-2. Antonella ESPOSITO 247 Chapter V VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration In the previous chapter we presented Visiovis – a visualization tool thoroughly designed, assembled and developed in our laboratories to monitor, model and optimize polymer melt processing in screw/barrel systems. Lately, we intended adapt such a tool to monitor mixing during compounding of molten polymers with inorganic fillers – namely, to monitor polymer-clay nanocomposite processing. We have amply stressed how real-time monitoring techniques would facilitate the work of materials engineers – since in situ characterizations are generally less time-consumptive, less labor-intensive and more cost-efficient than ex situ characterizations. Nonetheless, the intensive work required to conceive, develop, test and then validate any new characterization technique shouldn’t be underestimated. Sometimes, the way which leads to the set up of a new Antonella ESPOSITO 248 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration characterization method isn’t exactly a highway, but does rather look like a tortuous path which requires the collaboration of several experts from different domains and, manifestly, more than a few years of hard and meticulous working. From time to time, it is required and may be useful to make a point about the progresses accomplished (ameliorations or just simple evolutions) – which doesn’t exempt from recognizing any possible mistake or identifying the limitations of the system in its actual configuration. The main objective of this chapter is explicitly to make a point on the technical progresses made during the last three years on Visiovis – giving an instant picture of the actual situation but also some suggestions for further amelioration, whenever a problem has been encountered. V-1 MATERIALS As previously underlined, real-time process monitoring often constrains to use pilot equipments and model materials – in particular, when the objective is to visualize the flows in geometrically-complex equipments, process engineers frequently have to choose model fluids. As our purpose was to monitor polymer-clay melt processing, we had to found both a model fluid and some model fillers having specific and suitable optical properties (optical inertness for the fluid and optical activity for the fillers). Model fluid. The choice of poly dimethylsiloxane (PDMS) as the model fluid has been largely vindicated in Chapter IV. Indeed, the first reflex was of course to make profit of the choices previously made by Moguedet and coworkers [1] and eventually adapt them by taking into account the new requirements. This is the fundamental reason why we chose PDMS: the same kind of fluid had been previously and successfully used for Visiovis experiments. However, silicone oil with lower viscosity (10 Pa·s rather than 100 Pa·s) would have been easier to handle (that is to say, it could make it easier to fill up and empty the screw/barrel system). Nevertheless, whilst PDMS represented an acceptable compromise for Moguedet and coworkers, it didn’t really fulfill all the additional requirements we needed to obtain polymer-clay nanocomposite morphologies by melt compounding. We have previously Antonella ESPOSITO 249 Chapter V called attention to some problems encountered during the preparation of masterbatches (PDMS + photo-active lamellar fillers) for the calibration of Visiovis detection systems (§ IV-2): the photo-active lamellar filler CNa+ 0.25CEC RhP couldn’t be visualized by the CCD cameras, and the aspect of the mixture prepared with such filler was different in comparison with the other mixtures (Figure IV-F18). We then evocated two possible reasons for these evidences: (1) the chemical composition of the commercial clays used to prepare the photo-functional complexes, and (2) the interactions of the photo-active lamellar fillers with the silicone oil. As the first explication didn’t really persuade us, we got more and more convinced that the problem is essentially related to the chemical composition and molecular arrangement of the chosen PDMS fluid (Figure V-F1). To confirm our suspicions, we carried out some rheological measurements, whose results will be shown soon after (§ V-4.1). Meanwhile, we searched through the literature to find out any similar observation reported about the interactions of PDMS and lamellar mineral fillers. Figure V-F1 Chemical formula of poly dimethylsiloxane (PDMS), the silicone oil we used as the transparent, viscous model fluid for Visiovis experiments. The macromolecular backbone consists of alternating Si and O atoms (contrarily to the majority of organic polymers, based on C atoms) and the only accessible groups are methyl groups (–CH3). Schmidt [2] and Paquien [3] dealt with filled polysiloxanes during their PhD research work: the former synthesized, characterized and evaluated the properties of polysiloxane/layered silicate nanocomposites, whereas the latter investigated mostly the rheological properties and the filler dispersion of PDMS/silica slurries. Paquien et al. [4] published soon after the results obtained by dynamic mechanical measurements and TEM on fumed silica/PDMS suspensions and focused their work on: (1) the modulation of the interactions between silica particles and PDMS through a controlled silylation of filler surface; (2) the effect of the procedure used to graft the silanol groups on the silica surface; (3) the effect of silica volume fraction on dispersion. They finally established a PhD INSA de Lyon (2008) 250 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration relation between the silica grafting ratio, the aggregate size and the rheological properties of the suspensions. Indeed, it is well known that silica particles can establish favorable interactions with polysiloxane macromolecular networks: Schaer et al. [5] have even developed a model for the description of silica particles dispersion in silicone polymers, taking into account the particle structure (porosity and density as a function of size), the penetration of PDMS into the silica particles, the bound formation between PDMS and accessible silanol sites, as well as the erosion of silica agglomerates – first in intermediate fragments, then in aggregates of a few hundreds of nanometers. Some works are also available about poly dimethylsiloxane reinforced by other fillers, e.g. barium titanate particles [6] and mica flakes [7]. Recently, lamellar mineral fillers (clays) have become quite popular also as a reinforcing agent of polysiloxane polymers, coherently with the general trend registered about other carbon-based polymers (§ I-2). Burnside and Giannelis [8] reported the first melt-processed layered silicate/poly dimethylsiloxane nanocomposites synthesized by delamination of the silicate particles in the PDMS matrix, followed by cross-linking. Afterwards, the number of works about the synthesis and properties of silicone rubber/clays nanocomposites in the literature noticeably augmented [9-18]. Wang and coworkers [19] openly proposed organic MMT as a substitute for aerosilica in liquid silicone rubber systems. However, if compounding lamellar mineral fillers to polysiloxane networks is nowadays interesting several research groups, it is worthy highlighting that nobody has ever reported about a PDMS matrix with the molecular structure of an inert silicone oil (Figure V-F1). Most of the time clays (whether natural or organically-modified) are added to silanol-terminated (hydroxyl-terminated) PDMS [8-10][17], vinyl-terminated PDMS [10][13], –NH2 and – PEO terminated PDMS [17] and some other PDMS matrices containing at least one reactive site in their repeating unit, for example SiCH3CH=CH2O [18]. The principal objective is, apparently, to introduce the filler in the liquid precursor, disperse it, then add the cross-linking agent (e.g. TEOS and/or tin 2-ethylhexanoate or whatever else) and cure the rubber composite at room [8-10][12] or higher temperatures [13][18][19]. Contrarily to Burnside and Giannelis [8], clay intercalation is typically accomplished in appropriate solutions [13][14] or with the addition of a dispersing aid, such as small amounts of chloroform [9] or even distilled water [10] (only for PDMS–OH). Though, Antonella ESPOSITO 251 Chapter V this strategy doesn’t always work as expected1. Moreover, even if probably performed, the results of rheological measurements on such systems are rarely reported: the authors rather perform mechanical tests [9][10][12][13][15][18][19] in order to evaluate the performances of cured samples (as the main purpose is the reinforcement of the polymer matrix), supported by XRD2, swelling tests, TGA, TEM and sometimes SEM, AFM, permeability measurements, IR spectroscopy. The control of the interactions between polymer macromolecules and the surface of some filler is, in the case of polysiloxanes, more complex than in the case of common carbon-based polymers. Several authors have faced the difficulty of pointing out which are the factors influencing the extent of clay exfoliation in polymer composites (whether the compatibility of clay modification with polymer chemistry, or the processing route and parameters, or both of them, and in which proportion) [20-26]. Takeuchi and Cohen [10] showed that the reinforcement in PDMS elastomers can be attributed to the anchoring of the hydroxyl end-group of polymer chains to the silicate surface of the fillers and that the mechanical properties of the obtained networks cannot be superseded without further chemical modification of the system. They could enhance the properties of their PDMS elastomers only if the networks were formed from the hydroxylterminated precursor. LeBaron and Pinnavaia [12] confirmed that organoclay could readily intercalate linear PDMS molecules terminated by hydroxyl groups, even though they observed little or no intercalation with analogous molecules terminated by methyl groups: this would mean that the interactions of terminal silanol groups with the internal surface of clay galleries represent an essential step of swelling and intercalation mechanisms of clays. Kaneko and Yoshida [18] recently highlighted that, hitherto, few researchers have reported that different kinds of clay have different behaviors when compounded to PDMS matrices. Even if they assure that exfoliation can be achieved in high-molar-mass PDMS matrices without solvent assistance or high shearing, they also admit that reinforcing PDMS rubbers by layered silicates is far more complicated than 1 Takeuchi and Cohen [10] observed that attempted network synthesis using just water as a dispersing aid, whilst efficacious for hydroxyl-terminated PDMS, was unsuccessful for vinyl-terminated PDMS. In this latter case, they rather added a buffer solution (pH=7) since the addition of water is suspected to prevent the hydrosilylation cross-linking reaction from occurring. 2 Kaneko et al.[17], for instance, evaluated the morphology of silicone/clay slurries by SAXS and WAXS. PhD INSA de Lyon (2008) 252 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration reinforcing them by silica or carbon black, mostly because of their unique morphology and interactions with the matrix. They observed that the main driving force for polymer intercalation into clay galleries results from the enthalpic contribution generated by the establishment of many favorable polymer-filler surface interactions and that, in the case of PDMS networks, essentially depends on the insertion of terminal segments of PDMS chains containing specific groups (e.g. Si(CH3)OH groups) into the interlayer spacing. Such end-groups could effectively interact with the silanol sites present on clay platelet surfaces (especially at the platelet edges, indeed) by hydrogen bonds. This is the reason why some authors found that the addition of small amounts of water occasionally helps clay exfoliation in PDMS–OH matrices [10]. However, in the majority of conventional organoclays, most of the –OH polar sites aren’t anymore available after cation exchange process: that’s the reason why Kaneko and Yoshida reported that clay agglomeration was more evident in PDMS-organoclay composites than in the composites containing unmodified clay. This observation contradicts somehow our reflections about PDMS masterbatches containing CNa+ 0.25CEC RhP: at least theoretically, the masterbatches containing the photo-active filler produced from natural clay should present a better dispersion than all the other masterbatches prepared from C30B 0.25MC RhP, C10A 0.25MC RhP, C15A 0.25MC RhP. Indeed, rheological measurements (§ V-4.1) showed no significant differences between the four photo-active lamellar fillers: none of the masterbatches showed the typical behavior of a properly dispersed polymer-clay nanocomposite (the principles of morphological characterization by rheology have been reported in § I-2.2.1). PDMS-clay interactions are maybe more complex than expected. The evidences discussed so far let presume that, in the case of polysiloxane-clay composites, a simple compatibilization of the polymer and the filler particles based on the degree of hydrophilicity/hydrophobicity of the compounded ingredients isn’t enough to assure the formation of nanocomposite morphologies. This difficulty has been only recently pointed out and, unluckily, represents the main limitation of our visualization tool and the associated characterization techniques. The problem is that Visiovis in its actual configuration (§ IV-1.3) isn’t compatible with reactive fluids, liquid precursors, solvents used as dispersing aids, etc. because of the material used to fabricate the barrel (PMMA) and, for the reasons evocated in the previous chapter, it cannot sustain high Antonella ESPOSITO 253 Chapter V temperatures. As a result, a better choice of the model fluid would be possible only if a PDMS containing different terminal or lateral groups could be synthesized, so that the polymer chains could favorably interact with the photo-active lamellar filler but remain inert with respect to the PMMA internal surface of Visiovis barrel. Actually, the quest for such an ideal transparent fluid could be quite difficult, as Schmidt et al. [16] recently published a paper about the origins of silicate dispersion in polysiloxane/layered nanocomposites and clearly showed that the factors influencing clay dispersion are numerous and not always simple to control. They prepared polysiloxane composite samples from a variety of matrices (with respect to both chemical functionality and molecular weight) and lamellar mineral fillers (natural and synthetic, hydrophilic and hydrophobic) just to determine the origins of silicate dispersion in a generic polysiloxane matrix. They found that, in the case of organoclays, the presence of an appropriate number of long (C12-C18) ammonium-bound alkyl chains is essential, as well as the presence of sufficient amounts of polar functional groups. They also concluded that, generally speaking, an otherwise incompatible polymer can be made compatible with a given filler by the inclusion of the appropriate number of dispersion-enhancing functional groups either at the chain-ends or elsewhere in the polymer. It is worthy to report the observation they made that, to the date of their contribution (i.e. 2006), polymer-clay nanocomposites had generated over a thousand publications, whilst only a handful dealt with polysiloxane. The publication by Kaneko et al. [17] came out soon after (2007) and roughly confirmed Schmidt and coworkers’ results. In particular, Kaneko and coll. reminded that, as previously found by LeBaron and Pinnavaia [12], the comparison of PDMS–SiOH with PDMS–Si(CH3)3 showed that, with an equivalent molar mass and in the presence of the same organoclay, the first matrix produced intercalation while the latter did not. Photo-functional fillers. On the basis of the previous considerations, if a proper functionalization of PDMS (modification of its terminal groups or synthesis of a novel repeating unit containing dispersion-enhancing lateral groups) is obtained, as a result of the experimental activity of the last three years four photo-active lamellar fillers are nowadays available to perform visualization experiments by Visiovis: CNa+ 0.25CEC RhP, C30B 0.25MC RhP, C10A 0.25MC RhP, C15A 0.25MC RhP. Any choice about PhD INSA de Lyon (2008) 254 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration both the procedure and the parameters used for clay photo-functionalization has been amply discussed and vindicated in Chapter II, while the results of the characterizations performed on the photo-functional inorganic-organic complexes have been reported and interpreted in Chapter III. If no good alternatives for the choice of the model fluid can be found (since the requirements imposed to the polymer matrix are numerous and quite strict), another possibility would be to purposely prepare a new class of photo-functional complexes – eventually but not necessarily by cation exchange process – in which the photo-activity would be once again assured by the RhP moieties inserted into clay galleries3, but the chemical compatibility with the PDMS matrix would be guaranteed by a specifically designed surfactant. This solution is inspired by the analogous problem encountered by some researchers who wanted to obtain exfoliated PP-clay nanocomposites by melt processing without making use of any dispersion-enhancing additives. Wang et al. [27], for instance, observing that many research efforts had been focused on dispersing MMT in PP (and that such efforts failed because of the absence of strong interactions between clays and highly apolar polyolefins), proposed to overcome the hurdle of formulating complex compounds by a compatibilization method which can be certainly generalized to other systems: synthesizing ammonium-terminated polymer chains (belonging to the same family but eventually shorter than those present in the polymer matrix) and using them to perform cation exchange processing of clay mineral fillers. This idea has been afterwards proposed by Schmidt [2] to obtain PDMS-modified clays by cation exchange processing a MMT-Na+ with short ammonium-terminated PDMS chains resulting from the acidification of a commercial amino-terminated PDMS. In such a situation – that is, when the surfactant for clay modification is insoluble in most of the solvents capable of swell clays – some other procedure for cation exchange processing must be used and the choice of the experimental protocol can be long and labor-intensive: Schmidt developed a melt exchange technique, but the work of Ma et al. [14] let also guess that some other method can be found, for instance choosing a different solvent (or a mixture of solvents) 3 This condition must be assured because, if another fluorescent dye is selected, the lighting source should be consequently changed. Antonella ESPOSITO 255 Chapter V as the exchanging medium4. The same idea has been developed by Li et al. [15]: they modified clay by siloxane surfactants and dispersed it into polymethylsilsesquioxane (PMSQ) by solution intercalation. Afterwards, they compared the system with the same matrix compounded with a commercial organoclay (Cloisite ® 15A). As a confirmation of what previously reported, they concluded that organoclays modified by carbon-based surfactants are not suitable to prepare PMSQ-clay nanocomposites. If no good alternatives for the choice of the fluid or the photo-functionalization of the lamellar mineral fillers are found, there’s yet one alternative: synthesizing photoactive silica fillers. The photo-functionalization of silica particles could be envisaged as: (1) a photo-functionalization ex situ, including a modification of the silica particles right after their synthesis (and typically introducing the fluorescent cationic dye on their external surface) or (2) a photo-functionalization in situ, viz. performed at the same time of the synthesis (which could eventually be designed to obtain a core-shell morphology, in which the silica shell would enclose a fluorescent RhP core). Ow and coll. [28], for instance, recently described highly fluorescent and photo-stable core-shell nanoparticles (size range 20-30 nm) obtained by a modified Stöber synthesis5, which are monodisperse in solution and resulted 20 times brighter and more photo-stable than their constituent fluorophore. They synthesized the particles for biological applications (labeling of macromolecules for bioimaging experiments), but nothing lets imagine that such a class of particles couldn’t be used for applications in the material field. However it is obvious that, if we consider using photo-active core-shell silica particles (whether the fluorophore encloses or is enclosed by silica) for Visiovis experiments, even tough 4 To be honest, during the last year of PhD research activity, we started performing some trials of cation exchange process of Cloisite ® Na+ with a commercial low molecular weight NH2+-PDMS-NH2+. We first tried to reproduce the experimental protocol proposed by Schmidt [2] (melt cation exchange process) with and without the fluorescent cationic dye but then, as we didn’t estimate this procedure sufficiently “clean” for fluorescence applications (Schmidt assumed that the excesses of hydrochloric acid and/or surfactant were negligible, but fluorescence doesn’t tolerate the presence of contaminants or excesses which could affect absorption and emission), we rather started adapting the protocol set up in Chapter II to a surfactant insoluble in water (such as the NH2+-PDMS-NH2+ provided by Degussa). We tested the cation exchange process protocol described in § II-2.4 with Cloisite ® Na+ and a proper amount of NH2+-PDMS-NH2+ in toluene, with and without RhP: however, we won’t present any preliminary result here, since we haven’t yet optimized the parameters of the protocol and we haven’t yet completed the required characterizations. 5 The Stöber synthesis of colloidal silica was first described in 1968 and is nowadays largely used to obtain monodisperse nano- to micro-sized silica particles. Van Blaaderen et al. first reported the covalent incorporation of organic fluorophores into Stöber colloidal silica and the synthesis of fluorescent silica nanoparticles in the hundreds of nanometers size range [28]. PhD INSA de Lyon (2008) 256 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration the interactions between PDMS and filler particles will be probably enhanced, the information retrieved would be less rich than in the case of photo-active lamellar fillers. We amply discussed about clay morphology and the differences between distributive and dispersive mixing during polymer-clay compounding (Chapter I): we also proposed an experimental protocol which, performed on Visiovis, could inform about distribution (thanks to the CCD cameras and the image processing) and probably dispersion (thanks to spectrofluorimetry) of photo-active clays in the transparent fluid modeling a molten polymer. As the core-shell silica particles are spherical and would likely be stabilized by the favorable interactions established with the matrix, only distributive mixing could be visualized. V-2 EQUIPMENT Configuration. As described in Chapter IV, the actual configuration of Visiovis is the result of some reasoned changes justified by: the complex geometry of the screw/barrel system; the selection of the area of interest (the volume of fluid comprised between two adjacent screw flights, the internal surface of the barrel and the screw root surface); the set up of a planar light source for fluorescence excitation (laser sheet); the intrinsic difficulties associated to data acquisition and processing. All these factors have equally contributed to the modification of configuration detailed in Chapter IV. Some of them are strictly correlated to each other and interdependent. As several critical problems have been encountered about the compatibility of the model fluid with the photo-active lamellar fillers (§ V-1) and would need to be promptly fixed, we believe that Visiovis configuration shouldn’t be considered as a main concern. Light source. We mentioned that some of the factors which determined Visiovis actual configuration are strictly correlated to each other and result interdependent: light source is one of the constrained parameters. The first aspect limiting the choice of the light source is, obviously, economic: powerful, monochromatic and perfectly collimated laser sources can be quite expensive. In our case, the choice has been helped by the fact that the cationic dye selected to perform clay photo-functionalization (Rhodamine 6G Antonella ESPOSITO 257 Chapter V Perchlorate) is one of the most common fluorophores, well known by several groups of researchers and regularly used by biologists (see § I-4): this consideration shouldn’t be underestimated, as it obviously implies that the laser sources adapted for RhP excitation (as well as the required filters to separate the emission from the excitation source) are commercially available and (only) reasonably expensive. We’ve repeatedly stressed that RhP represents an ideal probe to study heterogeneous systems thanks to the dependence of its absorption and fluorescence emission on the properties of the matrix: therefore, we suggest to persevere as long as possible with such fluorophore – a choice which will be automatic in case the photo-active clays (whose photo-functionalization protocol has been described and optimized in Chapter II, and whose characterizations are discussed in Chapter III) will be used as they are, but which should be rather imposed if any other solution is found (e.g. Stöber synthesis of core-shell colloidal silica particles, § V-I). Barrel. The transparent PMMA barrel represents, undoubtedly, the main core of Visiovis. Its transparency symbolizes the practical interest of process visualization and, therefore, corresponds to the aspect which should be absolutely preserved. On the other hand, the fact that the barrel is made of plastic have represented, since the beginning, one of the most serious limitations of Visiovis (§ IV-1.1.2). Of course, the best solution would be to find a material (probably a special glass) totally transparent and capable of tolerating high temperatures and high radial pressures; in the absence of such an ideal solution, we suggest to replace the actual barrel with another PMMA barrel having the same internal surface (cylindrical) but a parallelepipedic external surface. This wouldn’t fix the problems due to temperature and pressure, but at least would ameliorate the optical quality of the visualization. Some accesses for real-time sampling would also represent a valuable amelioration and would allow a further exploitation of the actual screw/barrel system (see § V-4). Screw. Far from being a problem, the screw profile rather offers a big margin of modification. In Chapter I we stressed that the industry of plastic gradually developed its own equipments and progressively multiplied the number of applications requiring specific processing tools. Nowadays, the trend of the market concerning the processing tools for the industry of plastic tremendously developed the relation supply-demand, to PhD INSA de Lyon (2008) 258 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration a point that the possible configurations of the same processing tool are as various as its possible applications. Screw profiles are customizable – sometimes several units having specific functions (mixing, homogenizing, etc.) can be composed to form a single screw profile perfectly adapted to a specific application. In relation to the choice of the screw profile, Visiovis doesn’t limit imagination – the only feature directly influenced by the choice of the screw profile is the data acquisition and processing. Feeding. We previously detailed the protocol for the experiments performed by Visiovis and, in particular, we described the procedure used to inject the masterbatch into the system: “the syringe is plunged vertically in the aperture and the injection is rapidly achieved in the lowest accessible point” (§ IV-2). At the moment we performed the experiments, we had no alternatives to inject the tracing masterbatch into Visiovis screw/barrel system. Indeed, as reported by Cassagnau et al. [29]6 and confirmed by our own experience, the feeding system may have a deep influence on the results of tracing experiments: furthermore, the actual feeding system of Visiovis (Figure IV-F19) is far from being practical and handy. The utilization of a syringe allowed us to perform experiments and collect some early results – that’s surely a good point – however we suggest, as a perspective easy to implement and probably rather inexpensive, to equip Visiovis with a permanent, firmly positioned feeding system (eventually shaped as a syringe) which should finally reduce any manipulation inaccuracy inevitably produced by the operator and, thus, assure a better reproducibility of the observed phenomena. V-3 PROCESSING OF THE ACQUIRED DATA In Chapter IV we related Visiovis evolutions and its actual configuration, as well as the possibilities of data acquisition and processing. We described both the categories of experimental data provided by Visiovis (i.e. the images and the fluorescence spectra) but we rather focused on the images, viz. we set up two methods for image processing: 6 Cassagnau et al. [29] developed a UV-fluorescence monitoring device to evaluate in situ the mixing efficiency of an internal batch mixer (§ I-4). They observed that the fluorescence curves recorded by this device strongly depended on the experimental conditions of injection of the tracer, thus they decided to simply drop the tracer on the molten flow stream. Antonella ESPOSITO 259 Chapter V the first based on the standard deviation of image luminosity (§ IV-3.1.1), the second based on the Discrete Fourier Transform of textured images (§ IV-3.1.2). We justified our choices and we validated the selected methods for image processing (in particular the second one) by computer simulation (§ IV-3.1.3), nonetheless a direct and precise correlation between observed and simulated phenomena has yet to be established7, and the data acquisition and processing methods set up during the last three years represent a tremendously interesting starting point. Moreover, if we take into consideration the problems encountered with the model materials (in particular, the lack of compatibility between the model fluid and the photo-active lamellar fillers prepared from commercial clays), it is clear that Visiovis (and the described data collection and processing) has not yet been completely exploited – one more reason to suggest looking for a more suitable set of model fluid and photo-active fillers before changing Visiovis configuration8 or searching for new methods for data processing. In addition, the choice of a set of model materials more prone to clay intercalation and or exfoliation would finally allow to take advantage of the complementary information provided by spectrofluorimetry. V-4 INTERPRETATION AND VALIDATION OF THE RESULTS The interpretation and validation of the results obtained by Visiovis is certainly priority – if compared to any possible change of configuration, or to the quest of some new method for data processing. After having introduced the aforementioned methods for data processing (§ IV-3.1.1 and IV-3.1.2), we provided a preliminary validation of the results obtained by such methods, as well as of their most probable interpretation (§ IV-3.1.3). Indeed, data processing and the interpretation of the ensuing results cannot be considered complete without an accurate and systemic experimental validation, attained by correlating Visiovis experimental results with the evidences provided by other (more conventional) characterization techniques such as XRD, TEM and rheology (§ I-2.2.1). We are absolutely aware of the importance of such validation and, actually, we planned to perform it but, as previously stressed, the choice of methyl-terminated PDMS as the 7 8 We’ll come back on this topic in a following paragraph (§ V-4.2). About this topic, please refer to § V-2. PhD INSA de Lyon (2008) 260 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration model fluid restricted considerably our prospects – directly (electron microscopy cannot be performed on oils or uncured resins) or indirectly (the properties of the model fluid aren’t specifically restraining with respect to XRD and rheology which, on the contrary, contributed to the assessment of a visible lack of compatibility of the photo-active fillers with the PDMS matrix). There’s no need to spell out that the best way to correlate Visiovis results (in situ) with the information provided by traditional techniques (ex situ) would be to perform a systemic real-time sampling of the compound evolving into the screw/barrel system and characterize it by XRD, TEM and rheology. Manifestly, real-time sampling represents something priority for Visiovis development, as well. V-4.1 Real-time sampling As explained when discussing about the calibration of Visiovis detection systems (§ IV-1.2.4), we haven’t yet designed a practical and handy method to get some samples of the compound evolving in the screw/barrel system – nonetheless, we acknowledge its priority. For the moment, just to get a hint of the possible correlation between Visiovis and the other conventional techniques for morphological characterization, we performed some measurement tests on the masterbatches prepared for Visiovis experiments (§ IV2) in their initial state by XRD and rheology. We couldn’t even consider performing electron microscopy on such masterbatches because, as previously reported, electron microscopy imposes a few but strict conditions on the nature and physical aspect of the samples (§ I-2.2.1). One of such conditions – probably the strictest for silicone oil – is that the samples have to be observed in vacuum as the air molecules would significantly scatter the electrons. Environmental electron microscopy could probably represent a possible alternative to conventional (under high vacuum) electron microscopy (it allows even hydrated samples to be viewed in low-pressure wet environments) but, unluckily, we didn’t dispose of an environmental electron microscope during this PhD research activity and, in any case, the characterization would have been surely not conventional. Antonella ESPOSITO 261 Chapter V XRD9. Schmidt et al. [16] pointed out that, because of the lack of X-ray contrast (i.e. differences in electron density) between the silicone and the silicate (as compared to layered silicate nanocomposites with carbon-based polymers), the XRD pathways of PDMS-clay nanocomposites often present low intensities of the diffraction peaks and low signal-to-noise ratios – which limits the use of XRD to materials containing at least 10% wt of clay, even if smaller amounts of clay are likely to produce better dispersion. We definitely confirm the observations reported by Schmidt et al. [16]: indeed, we performed XRD measurements on a number of masterbatches prepared with all the available photo-active lamellar fillers, alone or with the corresponding pristine clays, by different methods of compounding (manual stirring, mixing by a disperser TurboTest Rayneri 33/300P, ultrasound probe) and with different clay contents (1%, 3%, 5% wt)… whatever the clay content, whatever the nature of the filler, whatever the procedure used for compounding, we got silent XRD pathways. Two examples of silent XRD pathways are shown in Figure V-F2 (bold lines): they have been both obtained for a masterbatch containing 0.25% wt of the photo-functional complex C30B 0.25MC RhP and 0.75% wt of the pristine C30B (total amount of clay 1% wt). In particular, Figure V-F2 (a) shows the pathways for the masterbatch obtained by manual stirring, whereas Figure V-F2 (b) shows the pathways for the same masterbatch prepared by the disperser (20 min at 1000 rpm). Probably, we would have erroneously concluded that clay particles were perfectly exfoliated into the polymer matrix (see considerations previously made about erroneous interpretations of silent XRD pathways § I-2.2.1) if we didn’t dispose of the following evidences – rather supporting the fact that clays were distributed into the PDMS, but no dispersive mixing had occurred: the masterbatches appeared initially (i.e. right after mixing) homogeneous, but after some time started settling – meaning that a considerable amount of clay particles remained agglomerated and, under the action of gravity, drifted down to the bottom of 9 The experimental protocol for XRD measurements is the same used for the photo-active lamellar fillers, described in § II-3.1. PhD INSA de Lyon (2008) 262 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration the pan, whereas a minor fraction of smaller clay particles 10 appeared stably suspended just because insensitive to gravity11; to confirm our conjecture, we centrifuged the aforementioned masterbatches (5 min at 10000g, i.e. 15500 tr/min) in order to remove as much PDMS as possible, so that the sediment (the sample) got enriched of silicate filler to the maximum extent. In the case of the centrifuged samples (containing essentially the clay powder wetted by an insignificant amount of PDMS) we got an XRD pathway showing a single low-intensity peak roughly corresponding to the interlayer spacing of the dry filler (Figure V-F2, thin lines). This evidence represents a further confirmation of the fact that methyl-terminated PDMS may have some advantageous properties in terms of visualization (transparency, inertness) but certainly isn’t a suitable fluid to model carbon-based molten polymers compounded with clays. Figure V-F2 XRD pathways of a masterbatch containing 0.25% wt of photo-functional complex (C30B 0.25MC RhP) and 0.75% wt of the corresponding pristine clay (C30B), before and after centrifugation (bold and thin lines, respectively). Two methods having different compounding efficiencies are compared: (a) manual stirring and (b) mixing by a disperser TurboTest Rayneri 33/300P (20 min at 1000 rpm). 10 Clay fillers can be characterized by more or less large size distributions. We affirm that a minor fraction of smaller clay particles appeared stably suspended because, in spite of decantation, the masterbatches kept their colored appearance and we never observed a clear separation of colored (red/rose photo-active filler) sediment from a clear (transparent PDMS) supernatant. 11 Antonella ESPOSITO 263 Chapter V Rheology further confirmed the inadequateness of methyl-terminated PDMS, as we are going to show later on. Rheology. The literature largely confirms that rheology is a valuable tool for the analysis of filler dispersion in molten polymers, as reported in § I-2.2.1. Undoubtedly, rheology can provide useful information in a variety of different measurement modes, e.g. flow, dynamic, transient measurements. As we had already collected some negative evidences from the XRD characterization of the masterbatches (see previous section), we estimated unnecessarily time-consumptive to perform a complete set of rheological characterizations just to confirm several times the same, negative result. Therefore, we proceeded following only the most simplistic approach: we performed some dynamic measurements in order to check whether the complex viscosity at low frequencies got increased by the presence of a percolating filler and, additionally, whether the moduli G’ and G’’ crossed in correspondence of a percolation threshold – indicating a transition from the purely viscous liquid behavior (typical of unfilled methyl-terminated PDMS) to a solid-like behavior as the frequency decreases. The behavior of the samples (that is, the behavior of methyl-terminated PDMS) is particularly exasperating for rheological characterizations: the specific properties of methyl-terminated PDMS [3] make of it a fluid which flows incredibly easy and can perfectly wet most of the surfaces12 – namely, the metallic plate surface of a cone-plate geometry of a rheometer. These extraordinary properties represent a serious obstacle to the formation of a regular and stable meniscus in the gap between the cone and the plate of the rheometer – indeed, the fluid gradually flows and the amount of measured sample diminishes, producing erroneous results. This is the reason why the first measurement tests, performed on a stress-controlled AR1000 rheometer with the largest available cone-plate geometry (35 mm ), were unsuccessful – in spite of the remarkable sensibility of the equipment (even to low-viscosity fluids) and of an optimized choice of the geometry. Resolute to get at least one measurement for confirmation, we performed few further tests on a strain-controlled ARES rheometer with a Couette geometry (in order to prevent the sample from escaping the measurement volume) – unsurprisingly, regardless of the reduced sensibility of the equipment with 12 Paquien [3] offered a clear and complete summary of the typical properties of polysiloxanes. PhD INSA de Lyon (2008) 264 VISIOVIS TECHNICAL PROGRESSES Problems and suggestions for further amelioration respect to the stress-controlled rheometer, these measurements were at least correct. Nonetheless (and unsurprisingly as well), such measurements confirmed that the filler didn’t get dispersed into the matrix (no increase of the complex viscosity has ever been observed for any of the aforementioned masterbatches13) and distribution was probably inhomogeneous and certainly unstable (G’ and G’’ never crossed). The results obtained by dynamic rheological characterizations performed on the same masterbatches chosen to show XRD silent pathways (Figure V-F2), as well as the homologous curves for the neat PDMS, are shown in Figure V-F3. Figure V-F3 Dynamic rheological behavior of the neat PDMS (a) and of the masterbatch containing 1% wt of filler (0.25% wt C30B 0.25MC RhP and 0.75% wt C30B) prepared by (b) manual stirring and (c) mixing by the disperser (20 min at 1000 rpm). 13 Just for rheological measurements, we even prepared a masterbatch containing 10% wt of clay! Antonella ESPOSITO 265 Chapter V V-4.2 Computer simulation As a final point, we would like to express a few considerations about computer simulation. In Chapter IV we affirmed having used computer simulation to justify and partially validate the image processing applied to Visiovis experimental data. We hope we have been sufficiently clear to convince about the pertinence of supporting the real experiments with the results of simulations. Here we would only like to stress that, in order for such correlation to be correct and complete, the experimental conditions and the simulation equations and parameters must be as close as possible to each other (or, better, as coherent as possible with each other): this is possible only if both experiments and simulations are developed at the same time and, in particular, in the total respect of the limitations imposed by the counterpart. The perspective of fixing Visiovis feeding system (§ V-2), for instance, will surely help the correlation with computer simulation. The experimental protocol could then be adjusted for the results to be as repeatable as possible and as coherent as possible with the results of the simulation (§ IV-3.1.3). On the other hand, computer simulation could be adapted to reproduce the flow behavior exactly in the same geometrical plan created by Visiovis laser sheet 14, and the images obtained by computer simulation could be rendered a little more “realistic” by applying a standard image treatment to increase fuzziness before applying the image processing procedures described in § IV-3.1.1 and IV-3-1-2. A lot of work left, quite a lot of courage needed! 14 Indeed, the computer simulation performed by Yves Béreaux (§ IV-3.1.3) represents the screw channel as it appears if observed in the direction perpendicular to the screw flight surface, whereas Visiovis laser sheet enlightens a plan which passes by the axis of the screw/barrel system, thus is parallel to the axis and form an angle equal to the screw helix angle of the screw (§ IV-1.2.1). So far, computer simulation and experiments show the same phenomena – if the screw helix angle is neglected in terms of visualization. PhD INSA de Lyon (2008) 266 VISIOVIS TECHNICAL PROGRESSES V-R Problems and suggestions for further amelioration REFERENCES [1] Moguedet M. Développement d'un outil d'aide à la conception et au fonctionnement d'un ensemble vis-fourreau industriel – Application à l'injection de thermoplastiques chargés fibres de verre longues. Thèse. Lyon: INSA de Lyon, 2005, 124 p. [2] Schmidt D.F. Polysiloxane/layered silicate nanocomposites: synthesis, characterization, and properties. PhD Thesis. Cornell University, 2003, 300 p. [3] Paquien J.N. Etude des propriétés rhéologiques et de l’état de dispersion de suspensions PDMS/silice. Thèse. Lyon: INSA de Lyon, 2003, 270 p. [4] Paquien J.N., Galy J., Gérard J.-F. et al. Rheological study of fumed silica-polydimethyl siloxane suspensions. Coll. Surf. A: Physicochem. Eng. Aspects 2005, 260, 165-172. [5] Schaer E., Guizani S., Choplin L. Model development for the description of silica particles dispersion in silicone polymer. Chem. Eng. Sci. 2006, 61, 5664-5677. [6] Khastgir D., Adachi K. Rheological and dielectric studies of aggregation of barium titanate particles suspended in polydimethylsiloxane. Polymer 2000, 41, 6403-6413. [7] Osman M.A., Atallah A., Müller M. et al. Reinforcement of poly(dimethylsiloxane) networks by mica flakes. Polymer 2001, 42, 6545-6556. [8] Burnside S.D., Giannelis E.P. Synthesis and properties of new poly(dimethylsiloxane) nanocomposites. Chem. Mater. 1995, 7 (9), 1597-1600. [9] Wang S., Long C., Wang X. et al. Synthesis and properties of silicone rubber/organo montmorillonite hybrid nanocomposites. J. Appl. Polym. Sci. 1998, 69, 1557-1561. [10] Takeuchi H., Cohen C. Reinforcement of poly(dimethylsiloxane) elastomers by chainend anchoring to clay particles. Macromol. 1999, 32, 6792-6799. [11] Burnside S.D., Giannelis E.P. Nanostructure and properties of polysiloxane-layered silicate nanocomposites. J. Polym. Sci. B: Polym. Phys. 2000, 38, 1595-1604. [12] LeBaron P.C., Pinnavaia T.J. Clay nanolayer reinforcement of a silicone elastomer. Chem. Mater. 2001, 13, 3760-3765. [13] Osman M.A., Atallah A., Kahr G. et al. Reinforcement of poly(dimethylsiloxane) networks by montmorillonite platelets. J. Appl. Polym. Sci. 2002, 83, 2175-2183. [14] Ma J., Xu J., Ren J.-H. et al. A new approach to polymer/montmorillonite nanocomposites. Polymer 2003, 44, 4619-4624. [15] Li B.-Y., Ma J., Liu H.-Y. et al. Siloxane surfactant-modified clay and its effect in reinforcing the laminate of polymethylsilsesquioxane. J. Appl. Polym. Sci. 2006, 100, 3974-3980. Antonella ESPOSITO 267 Chapter V [16] Schmidt D.F., Clément F., Giannelis E.P. On the origins of silicate dispersion in polysiloxane/layered-silicate nanocomposites. Adv. Funct. Mater. 2006, 16, 417-425. [17] Kaneko M.L.Q.A., Torriani I.L., Yoshida I.V.P. Morphological evaluation of silicone/clay slurries by small-angle/wide-angle X-ray scattering. J. Braz. Chem. Soc. 2007, 18 (4), 765-773. [18] Kaneko M.L.Q.A., Yoshida I.V.P. Effect of natural and organically modified montmorillonite clays on the properties of polydimethylsiloxane rubber. J. Appl. Polym. Sci. 2008, 108, 2587-2596. [19] Wang J., Chen Y., Jin Q. Organic montmorillonite as a substitute for aerosilica in addition-type liquid silicone rubber systems. Macromol. Chem. Phys. 2005, 206, 25122520. [20] Dennis H.R., Hunter D.L., Chang D. et al. Effect of melt processing conditions on the extent of exfoliation in organoclay-based nanocomposites. Polymer 2001,42,9513-9522. [21] Zhang Q., Wang Y., Fu Q. Shear-induced change of exfoliation and orientation in poly propylene/montmorillonite nanocomposites. J. Polym Sci B: Polym Phys 2003, 41, 1-10. [22] Rhoney I., Brown S., Hudson N.E. et al. Influence of processing method on the exfoliation process for organically modified clay systems. I. Polyurethanes. J. Appl. Polym. Sci. 2004, 91, 1335-1343. [23] Wang K., Liang S., Du R. et al. The interplay of thermodynamics and shear on the dispersion of polymer nanocomposites. Polymer 2004, 45, 7953-7960. [24] Homminga D., Goderis B., Hoffman S. et al. Influence of shear flow on the preparation of polymer layered silicate nanocomposites. Polymer 2005, 46, 9941-9954. [25] Fedullo N., Sclavons M., Bailly C. et al. Nanocomposites from untreated clay: a myth? Macromol. Symp. 2006, 233, 235-245. [26] McAlpine M., Hudson N.E., Liggat J.J. et al. Study of the factors influencing the exfoliation of an organically modified montmorillonite in methyl methacrylate/poly (methyl methacrylate) mixtures. J. Appl. Polym. Sci. 2006, 99, 2614-2626. [27] Whang Z.M., Nakajima H. Manias E. et al. Exfoliated PP/clay nanocomposites using ammonium-terminated PP as the organic modification for montmorillonite. Macromol. 2003, 36, 8919-8922. [28] Ow H., Larson D.R., Srivastava M. et al. Bright and stable core-shell fluorescent silica nanoparticles. Nano Lett. 2005, 5 (1), 113-117. [29] Cassagnau P., Melis F., Bounor-Legare V. UV fluorescence monitoring of the mixing of molten polymers in a batch mixer. Polym. Eng. Sci. 2003, 43 (4), 923-932. PhD INSA de Lyon (2008) 268 CONCLUSIONS The interest for polymer/clay nanocomposites blew a few decades ago, but the topic is still having a great vogue, since clay nanocomposites offer greatest hopes for a dramatic improvement of several properties. Definitely, it’s time to adapt the know-how developed about real-time monitoring to the processing of polymer nanocomposites! During the last three years we worked on a project which was started in 2003 and which is far from being over: it is well known that the work necessary to conceive, develop, test and validate new characterization techniques is intensive and time-consumptive. We are conscious that there are some difficulties which haven’t yet been overcome, but we are also glad to affirm that we found an adequate solution to more than one problem. We solved a first technical problem by establishing and optimizing a protocol for the photo-functionalization of commercially available organically-modified clays: we selected one of the most commonly used organoclays and, on the basis of several tests performed with different fluorescent molecules in different experimental conditions, we concluded that the best photo-functional inorganic-organic complexes can be obtained by cation exchange processing with an amount of Rhodamine 6G Perchlorate equal to Antonella ESPOSITO 269 CONCLUSIONS 25% of clay surfactant, in a 90/10 water/ethanol mixture at 80°C, followed by washing with ethanol, recovering by centrifugation, drying at room temperature under exhaust hood. Clay photo-functionalization with higher concentrations of fluorescent molecule is still possible, but unnecessary and trespassing the main objective of the work: just rendering clays photo-active. Any variation of the concentration of fluorescent molecule in the exchanging medium could modify the mechanisms of absorption (e.g. formation of aggregates and adsorption at silicate edges, preventing any further intercalation). The intercalation of smaller molecules into the galleries of a clay previously modified by bigger molecules of surfactant produces a rearrangement of the paraffinic configuration and increases the tilting angle from 37 to 49°. Thanks to the aforementioned protocol, today we dispose of four photo-functional inorganic-organic complexes which, of course, have been prepared to be used with Visiovis, but could be much more versatile. Moreover, we observed that the photo-functionalization followed by careful washing doesn’t affect thermal resistance – on the contrary, a second exchange process with smaller molecules produces some CEC recovery and improve surface coverage. We succeeded in adapting an existing tool (Visiovis) to the real-time monitoring of polymer/clay distributive and dispersing mixing in the molten state, which could help understanding the key factors for the processing of polymer/clay nanocomposites. We set up and calibrated two complementary detection systems (cameras and spectrometer), we showed how to collect experimental data (images, videos and fluorescence emission spectra), we implemented some algorithms for data processing (two image processing based on the integral standard deviation of image luminosity and the Discrete Fourier Transform of textured images), we succeeded in performing some preliminary tests and suggested a possible interpretation of the results – trying to identify the limitations of the systems and to justify and/or explain the problems encountered. Finally, we made a point about the progresses accomplished, giving an instant picture of the actual situation but also some suggestions for further amelioration – a better choice of the model fluid or a better reciprocal compatibilization of the selected fillers and matrix, a slight change of the barrel external surface, quite a lot of opportunities about the screw profile, a more accurate feeding system and the conception of a real-time sampling method which would allow a direct correlation with more conventional characterization techniques. PhD INSA de Lyon (2008) 270 APPENDIX ---------------------------------------------------------------------function []=decoupe(fichvis,name0,deb) % % % % % % % % % % % % % % Compilation by Jean Balcaen Revision by Antonella Esposito INSA de Lyon © 2008 This function extracts a sequence of isolated frames from the images stocked in a single *.bmp file recorded by Visiovis and named by a code composed by the date and the hour of data recording. Example of command line: decoupe('Tue_Jul_24_17h36m00s2007_87','Manip16\Cam87\16-87',0); fichvis = name of the *.bmp file (input) name0 = path to be given to the extracted frames (output) deb = first number of the sequence (output index) name=['C:\MATLAB6p5\',fichvis,'.bmp']; aref=imread(name); s=size(aref) s0=s(1)/480 it0=s0; for it=1:s0 it0=it0-1; a0=aref((it-1)*480+1:it*480,:); nameit0=['C:\MATLAB6p5\',name0,'_',num2str(it0+deb+1),'.bmp'] imwrite(a0,nameit0,'bmp'); end ---------------------------------------------------------------------function []=fenetre(imdec,numdeb,numfin,x0,y0,x1,y1); % Compilation by Jean Balcaen % Revision by Antonella Esposito Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 1/6 % % % % % % % % % % % % % % % % INSA de Lyon © 2008 This function operates a cutback of the sequential images output of the function decoupe.m in order to reduce the area of interest to a rectangle around the channel section. Example of command line: fenetre('Manip16\Cam87\16-87_',1,80,240,370,585,445); imdec = path previously given to the extracted frames (input) numdeb = first number of the sequence (input index) numfin = last number of the sequence (input index) x0 = abscissa of the top left corner of the rectangle y0 = ordinate of the top left corner of the rectangle x1 = abscissa of the bottom right corner of the rectangle y1 = ordinate of the bottom right corner of the rectangle for x=numdeb:numfin x name=['C:\MATLAB6p5\',imdec,num2str(x),'.bmp']; a1=double(imread(name))/255; a=a1(y0:y1,x0:x1); name=[imdec,'ftr',num2str(x),'.bmp']; imwrite(a,name,'bmp'); end ---------------------------------------------------------------------function []=ecrtype(ftr,nameref,deb,fin); % % % % % % % % % % % % % % % % % % Compilation by Jean Balcaen Revision by Antonella Esposito INSA de Lyon © 2008 This function performs the image processing based on the integral standard deviation of the luminosity of Visiovis images, previously isolated by the function decoupe.m and cutback by the function fenetre.m Example of command line: ecrtype ('Manip16\Cam87\16-87_ftr','Manip16\Cam87\16-87_ftr13',14,80); ftr = path previously given to the extracted cutback frames (input) nameref = path of the image chosen as a reference (image completely black at the beginning of the sequence) deb = first number of the sequence (input index) fin = last number of the sequence (input index) namer=['C:\MATLAB6p5\',nameref,'.bmp']; a0deb=double(imread(namer)); a0deb=a0deb/max(max(a0deb)); a0deb=1-a0deb; for x=1:fin-deb X=x+deb Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 2/6 name=['C:\MATLAB6p5\',ftr,num2str(X),'.bmp']; a0=double(imread(name)); a0=a0.*(1-a0deb)+mean(mean(a0.*(1-a0deb)))*a0deb; a1=a0-mean(mean(a0)); a1=(a1.^2/mean(mean(a0))^2).^0.5; a(x)=mean(mean(a0)); end; figure (1) plot(a,'o-') figure(2) imagesc(a0) dlmwrite([ftr,'ecrtype.txt'],a,'\t') ---------------------------------------------------------------------function []=normft(ftr,nameref,deb,fin,freq); % % % % % % % % % % % % % % % % % % % % Compilation by Jean Balcaen Revision by Antonella Esposito INSA de Lyon © 2008 This function performs the image processing based on the Discrete Fourier Transform (DFT) of the textured numerical images acquired by Visiovis, and previously isolated by the function decoupe.m and cutback by the function fenetre.m Example of command line: normft ('Manip16\Cam87\16-87_ftr','Manip16\Cam87\16-87_ftr13',14,80,150); ftr = path previously given to the extracted cutback frames (input) nameref = path of the image chosen as a reference (image completely black at the beginning of the sequence) deb = first number of the sequence (input index) fin = last number of the sequence (input index) freq = value of frequency estimated adapted to the processed images and used for the calculation of the Fourier Transform (150) name=['C:\MATLAB6p5\',ftr,num2str(deb),'.bmp']; adeb=double(imread(name)); namer=['C:\MATLAB6p5\',nameref,'.bmp']; a0deb=double(imread(namer)); a0deb=a0deb/max(max(a0deb)); a0deb=1-a0deb; A0=fftshift(fft2(adeb)); nA0=(A0.*conj(A0)); %.^0.5; sn=size(nA0); xx=zeros([sn(1)*sn(2),1]); yy=zeros([sn(1)*sn(2),1]); a=zeros([freq,fin-deb]); nomb=zeros([freq,fin-deb]); Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 3/6 for x=1:fin-deb X0=x+deb name=['C:\MATLAB6p5\',ftr,num2str(X0),'.bmp']; a0=double(imread(name)); a0=a0.*(1-a0deb)+mean(mean(a0.*(1-a0deb)))*a0deb; A=fftshift(fft2(a0)); nA1=(A.*conj(A)); %.^0.5; [cx,cy]=find(nA1==max(max(nA1))); nA=nA1-nA0; s=size(nA); for x1=1:s(1) for y1=1:s(2) X=2*((x1-cx(1))/s(1))+0.01; Y=2*((y1-cy(1))/s(2))+0.01; R=min(s)/2; if (abs(X)>=abs(Y)) t=Y/X; X=X*cos(atan(t)); Y=Y*sin(atan(t)); end; if (abs(Y)>=abs(X)) t=X/Y; X=X*sin(atan(t)); Y=Y*cos(atan(t)); end; r=round(R*(X^2+Y^2).^0.5)+1; vie(x1,y1)=r; if (r<freq) a(r,x)=a(r,x)+nA(x1,y1);%*r; nomb(r,x)=nomb(r,x)+1; end; end; end; end; figure(4) imagesc(vie) af=a./(nomb+1); saf=size(af) rx=1:saf(1); af=flipud(af); laf=log(af-min(min(af))+1); laf=laf-min(min(laf)); for x=1:fin-deb maf(x)=mean(laf(:,x).*rx')/mean(laf(:,x)); end; laf=laf-mean(mean(laf))+2; laf=laf.*(sign(laf)+1)/2; af=flipud(af); figure(1) surf(laf) Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 4/6 shading interp colormap jet hold on contour3 (laf,20,'r') figure(2) pcolor(laf) shading interp colorbar dlmwrite([ftr,'normft.txt'],maf,'\t'); figure(3) plot(maf(1:end),'o-') ---------------------------------------------------------------------function []=video(nameim,num); % % % % % % % % % % % % % % Compilation and revision by Antonella Esposito INSA de Lyon ® 2008 This function reconstruct a video from a sequence of frames (whether cutback by the function fenetre.m or not) and generates a montage of the video – a panel visualizing at a glance all the frames used for reconstruction (the montage is useful only for the windows cutback around the channel section). Example of command line: video('Manip16\Cam87\16-87_',80); nameim = path of the frames to be used for reconstruction (input) num = total number of frames nameref=['C:\MATLAB6p5\',nameim,'1.bmp']; for x=1:num x name=['C:\MATLAB6p5\',nameim,num2str(x),'.bmp']; [multim(:,:,:,x),map]=imread(name,'bmp'); imshow(multim(:,:,:,x),map); tmp=getframe; vid(:,x)=tmp; end m=montage(multim) saveas(m,'montage','emf') movie2avi(vid,'video','fps',10,'compression','none','colormap',gray) info=aviinfo('video') Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 5/6 ---------------------------------------------------------------------function [a]=spctr(nom,deb,fin) % % % % % % % % % % % % % % Compilation by Jean Balcaen Revision by Antonella Esposito INSA de Lyon ® 2008 This function visualizes the fluorescence emission spectra regularly acquired by Visiovis as a 3D shaded surface plot on a rectangular region delimited by the processing time [s] and the wavelength [nm]. Example of command line: spctr('exp1_00',1,112); nom = path and filename of the sequence of spectra (input) deb = number of the first spectrum fin = number of the last spectrum for x=1:fin-deb nm=[nom,num2str(deb+x),'.txt']; a0=dlmread(nm,'\t',[18,1,2065,1]); xaxis(1,x)=x*3; a(:,x)=a0(:,1); end yaxis=xlsread('yaxis'); surf(a,'XData',xaxis,'YData',yaxis) shading interp xlabel('Time [s]') set(gca,'XDir','reverse') ylabel('Wavelength [nm]') zlabel('Fluorescence emission intensity [a.u.]') ---------------------------------------------------------------------- Antonella ESPOSITO PhD INSA de Lyon (2008) Appendix 6/6 Résumé détaillé en français MISE EN ŒUVRE DE NANOCOMPOSITES. Mélange en voie fondue. Les propriétés des matériaux polymères dépendent directement de leur chimie. Le seul moyen de changer les propriétés d’un matériau polymère sans changer sa nature chimique la plus intime est le compoundage – l’introduction d’additifs chimiquement et/ou physiquement hétérogènes et leur dispersion dans la matrice polymère d’accueil, qui se fait typiquement en voie fondue. Les facteurs responsables de la majorité des problèmes liés au compoundage sont : la chimie de la matrice polymère d’accueil ; les propriétés physico-chimiques des additifs ; le taux d’additifs dans le mélange ; le procédé de mise en oeuvre choisi pour le compoundage. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 1/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. La qualité et les performances des matériaux composites à matrice polymère dépendent certainement des propriétés intrinsèques des matériaux dont ils sont constitués, mais aussi de la capacité du procédé de mise en œuvre à obtenir la meilleure dispersion possible de tous les composants du mélange dans la totalité du volume du polymère transformé – afin que l’on puisse assumer que toute échantillon de matériau composite, quelle que soit son origine, ait les mêmes propriétés souhaitées. Traditionnellement, le compoundage s’articule en trois étapes, toutes essentielles pour obtenir de bons résultats : l’alimentation, le mélangeage et la granulation. On remarquera que l’extrusion et l’injection – les procédés de mise en œuvre des matériaux plastiques les plus diffusées – comportent toutes les deux une étape de mélange en voie fondue : l’objectif de l’extrusion est la formulation des granulés composites, qui seront ensuite refondus lors de l’injection moulage pour enfin obtenir des pièces ayant la forme souhaitée. On remarquera également que l’extrusion et l’injection partagent la même complexité géométrique des outils de transformation, car toutes les deux sont réalisées à l’aide de systèmes vis/fourreau ayant des profils de vis adaptés à chaque application et à chaque matériau. La question que l’on se pose est donc la suivante : comment se comportent les additifs renforçants (les charges) quand ils sont mélangés à un polymère fondu transformé dans un système vis/fourreau ? Pour que le procédé de mise en œuvre soit performant, le mélange en voie fondue doit l’être aussi – ce qui signifie que les performances globales sont influencées aussi bien par le choix de l’outil de transformation (conception et sélection du profil de la vis) que par son utilisation (optimisation des paramètres de mise en œuvre). Le meilleur choix repose sur la définition des résultats que l’on veut obtenir et de comment peut-on les obtenir – autrement dit, le meilleur choix dérive de la compréhension des phénomènes (c’est-à-dire les mécanismes de mélange, dispersion et distribution des charges) mais aussi de la connaissance et de la maîtrise des outils de transformation dont on dispose à l’heure actuelle (les systèmes vis/fourreau). Les systèmes vis/fourreau pour la transformation des matériaux polymères et de leurs composites sont identifiés par le diamètre nominal de la vis (D) et par la longueur de sa portion filetée (L), cette dernière étant composée de trois sections géométriquement bien distinctes : la section d’alimentation (chenal à profil constant et plutôt profond), la section de transition (chenal à profil variable et profondeur décroissante) et la section de pompage (chenal à profil constant et profondeur réduite). Nous nous sommes intéressés seulement à la section de pompage, dans laquelle le polymère est entièrement fondu et qui transporte et mélange davantage la matière pour en assurer l’homogénéité physique, chimique et thermique jusqu’à son injection dans le Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 2/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. moule. Mélanger consiste à utiliser l’énergie mécanique pour développer un champ de vitesse et induire ainsi du mouvement au sein du fluide transformé, de façon à ce que tout champ de concentration à gradient initialement élevé soit homogénéisé. Malgré le caractère universel du mélange, les phénomènes qui le dominent sont encore méconnus et pas tout à fait compris et maîtrisés. S’il est vrai que les aspects cinématiques et les différentes échelles (temporelle et spatiale) du mélange de deux fluides ont été analysés et décrits, on ne peut tout de même pas ignorer que les récents progrès réalisés par la science et le génie des matériaux n’arrêtent pas de lancer des défis de plus en plus ardus aux ingénieurs procédés. Les nanocomposites à matrice polymère et à base d’argile attirent encore (et depuis quelque temps, désormais) une attention croissante, tant de la part du monde académique que de l’industrie, imposant ainsi des critères d’homogénéité du mélange plus stricts et étendus jusqu’aux plus petites échelles. Pour que ces avancements technologiques soient également intéressants du point de vue économique, l’objectif à se donner aujourd’hui est la réalisation de mélanges homogènes sur plusieurs échelles et, surtout, jusqu’à l’échelle moléculaire – si possible en se servant des outils de transformation traditionnels, opportunément optimisés. Traditionnellement les écoulements sont étudiés grâce à des expériences de visualisation – plus précisément, depuis que Reynolds (1883) découvrit l’existence des différents régimes d’écoulement (laminaire, turbulent et de transition) en visualisant le comportement d’un traceur coloré injecté isocinétiquement au centre d’un cylindre transparent dans lequel un fluide transparent s’écoulait avec un débit connu. Le régime d’écoulement laminaire, en particulier, est observé lorsque les fluides s’écoulent à faible vélocité, et d’autant plus facilement s’ils sont aussi hautement visqueux. Comme les polymères (thermoplastiques) à l’état fondu possèdent généralement des viscosités élevées, et comme le volume mis à disposition pour l’écoulement dans les systèmes vis/fourreau est réduit, le seul régime d’écoulement que l’on puisse observer pour un polymère fondu dans le chenal d’une vis d’extrusion ou d’injection est le régime d’écoulement laminaire : les mécanismes responsables du mélange distributif, donc, se réduisent à une séquence d’étirement, découpage et recombinaison de lamelles de fluide qui glissent les unes sur les autres sans se croiser et qui, en s’étirant, deviennent plus minces et facilitent la diffusion moléculaires entre couches voisines. Pour que le mélange soit efficace, l’approche traditionnelle exige que les lamelles de fluide soient assez fines et reparties de façon homogène dans la totalité du volume de matière transformée : évidemment, ces critères ne suffisent plus lorsque l’on s’attaque aux nanocomposites à matrice polymère et à base de charges lamellaires (la diffusion des macromolécules de polymère à l’intérieur des galeries étant considérée comme Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 3/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. un des mécanismes responsables de l’exfoliation, c’est-à-dire du mélange dispersif). D’ailleurs, si les expériences de visualisation à l’aide de traceurs colorés ont – depuis désormais plus d’un siècle et suite à l’exemple donné par Reynolds – contribué à la compréhension des mécanismes d’écoulement et de mélange entre deux fluides, rares sont les applications de telles méthodes (ou d’autres méthodes similaires) à l’étude de la distribution et dispersion de charges au sein d’un fluide. Cette observation est d’autant plus légitime car il s’agit, ici, d’étudier le mélange d’un polymère fondu avec des charges lamellaires (qui ont une structure complexe multi échelle et aucune propriété optique native remarquable) dans un système ayant une géométrie complexe (système vis/fourreau). La question que nous nous sommes posée ensuite est donc la suivante: comment visualiser le procédé de mise en œuvre des nanocomposites à matrice polymère et à base d’argile, sachant que ni les outils de transformation traditionnels (à parois opaques) ni les argiles (optiquement inertes) ne permettent la visualisation ? Le fait que les particules du renfort puissent se briser ou s’agréger davantage pendant le mélange, ne fait que rendre l’analyse de l’écoulement encore plus complexe. Lorsqu’il s’agit d’étudier simultanément l’écoulement d’un polymère fondu et son mélange avec des charges lamellaires, le mélange dispersif (fortement lié aux propriétés physico-chimiques des particules inorganiques, c’est-à-dire à leur taille, leur morphologie et leur compatibilité avec la matrice organique) devient, évidemment, aussi important que le mélange distributif (dépendant plutôt de la géométrie de l’outil de transformation, ainsi que des paramètres de la mise en œuvre). Les notions d’agrégation et d’agglomération (et aussi de dispersion et de distribution) sont plutôt génériques, certes, mais suffisamment précises pour décrire les macro- et micro-composites, dans lesquels les particules du renfort sont assez grandes et ont une morphologie assez simple. En revanche, la morphologie des nano-composites à base d’argile peut être complexe et certainement plus variée – de ce fait, une utilisation équivalente des mêmes notions crée souvent de la confusion et se révèle parfois erronée. La littérature montre clairement que des morphologies nanocomposites peuvent être obtenues par mélange en voie fondue de polymères avec des charges lamellaires : nombreux sont les travaux qui ont été consacrés à la modification de la chimie de surface des argiles (compatibilisation) mais, mises à part les prédictions d’une probable influence des paramètres de mise en œuvre sur les morphologies obtenues, aucun travail n’a encore propose de conclusions univoques à ce sujet. En relation aux avantages des nanocomposites par rapport aux composites traditionnels : la littérature d’aujourd’hui abonde de revues et travaux expérimentaux plus ou moins pointus célébrant les améliorations présumées Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 4/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. que l’on obtiendrait grâce à des morphologies nanostructurées. Oui, mais à quelle échelle de production arrive-t-on à assurer le contrôle d’une morphologie nanocomposite ? La révolution opérée par les morphologies nanocomposites repose sur la création d’une interface exceptionnellement grande que les nanocharges mettent à disposition pour établir des interactions avec la matrice – d’où la présence de fortes interactions entre les particules mêmes et, donc, la tendance à former des agrégats. Les premières utilisations des argiles avaient pour objectif de renforcer la matrice organique par introduction d’agrégats inorganiques de taille micrométrique. Les résultats obtenus, pourtant, n’étaient pas manifestement meilleurs par rapport aux résultats déjà garantis par les composites traditionnels : les argiles sous forme d’agrégats n’ont pas d’avantages par rapport aux charges micrométriques traditionnelles, car l’intérêt majeur leur dérive seulement des feuillets élémentaires, et ces derniers ne peuvent être exploités que si l’on disperse parfaitement les argiles au sein de la matrice polymère. C’est pour cela que les nanocomposites polymère/argile n’ont fait leur apparition sur le marché que bien après les premiers brevets sur leur fabrication ; pour cette même raison, ces matériaux innovants peinent à trouver leur marché de niche… une nouvelle méthode pour le suivi en ligne des évolutions morphologiques des mélanges polymère/argile pendant le procédé de mise en œuvre pourrait contribuer de façon significative au développement des nanocomposites à base de charges lamellaires. La morphologie des nanocomposites à base de charges lamellaires est difficile à caractériser et, comme le montrent les discordances parfois présentes dans la littérature, presque autant difficile à décrire. Le niveau technique atteint de nos jours mets à notre disposition de nombreuses techniques de caractérisation, aussi bien ex situ que in situ. Les techniques ex situ les plus communes sont la Diffraction des Rayons X (DRX) et la Microscopie Electronique à Transmission (MET), mais d’autres techniques ont été utilisées – la Microscopie Electronique à Balayage (MEB), la Calorimétrie Différentielle à Balayage (DSC), la rhéologie, la Microscopie à Force Atomique (AFM), les techniques de diffusion des rayons X (Wide-Angle et Small-Angle X-ray Scattering, WAXS et SAXS), la Résonance Magnétique Nucléaire (RMN) à l’état solide. Chacune de ces techniques possède des avantages et des désavantages ; le principal désavantage commun à toutes les techniques de caractérisation ex situ reste, en tous cas, l’impossibilité de corréler de manière directe et univoque la morphologie obtenue au procédé de mise en œuvre par lequel il a été possible de l’obtenir. Bref, les techniques ex situ ne démystifient pas vraiment le procédé de mise en œuvre qui, au contraire, demeure inconnu, tel qu’une « boite noire ». Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 5/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. Les techniques in situ ont été traditionnellement développées pour des études de dynamique des fluides, plutôt que pour des analyses morphologiques de fluides mélangés à des additifs (ou bien à des charges). Comme cela avait été le cas pour Reynolds, c’est le fait d’avoir visualisé ce qui se passe pendant le procédé de mise en œuvre (Maddock, 1959) qui a indiqué le chemin aux chercheurs. Mais Maddock venait tout juste de lancer une idée, car son système ne permettait pas vraiment le suivi en ligne : le procédé devait être stoppé pour pouvoir récupérer les informations. Les techniques in situ telles que le moulinet, la Vélocimétrie Doppler par Laser (LDV), la Vélocimétrie par Imagerie de Particules (PIV), la Vélocimétrie Doppler par Ultrasons (UDV) ont été généralement développées à l’échelle des laboratoires de recherche, qui disposent d’équipements et de savoir-faire adaptés. Mais d’autres techniques in situ ont été développées, et ont rencontré un discret succès même en dehors des laboratoires de recherche fondamentale : c’est le cas des techniques qui se servent de sondes locales comme les pigments (colorimétrie), les fluorophores, les traceurs ayant une signature infrarouge reconnaissable ou des propriétés spécifiques de conductivité, les sondes sensibles aux variations de pH, etc. Ces techniques (et en particulier la Fluorescence Induite par Laser, LIF) sont typiquement utilisées pour mesurer des valeurs locales (moyennées dans le temps) de concentration, afin d’estimer le temps de mélange, ou encore pour calculer la Distribution des Temps de Résidence (RTD) du matériau à l’intérieur de l’outil de transformation. Les sondes employées représentent rarement des additifs intéressants pour la qualité du mélange : en général elles n’apportent aucune propriété – mise à part, évidemment, celle qui permet le suivi. On est donc obligé d’accepter le risque que la présence d’un matériau sonde puisse modifier les écoulements et, par conséquent, fausser la visualisation… Une dernière possibilité pour la caractérisation morphologique des écoulements lors des procédés de mise en œuvre des polymères est issue de la simulation numérique effectuée grâce aux ordinateurs, qui ont une puissance de calcul de plus en plus développée. La simulation numérique permet, en effet, de « visualiser » les zones de mélange chaotique et d’estimer, de façon quantitative, le mélange laminaire, tout en évitant la difficulté des parois opaques des outils de transformation… ceci dit, toute simulation devrait être validée par l’expérience. Ce travail de thèse s’inscrit dans le contexte que nous venons de décrire : les récents progrès accomplis par les ingénieurs et les chimistes poussent le marché à se tourner vers les matériaux nanocomposites polymère/argile, mais aucun marché significatif ne pourra s’ouvrir si aux avantages technologiques n’est associé aucun avantage économique. Le seul moyen de rendre cela économiquement avantageux est d’adapter les outils de transformation déjà Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 6/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. disponibles à la mise en œuvre de cette nouvelle famille de composites – étape qui ne pourra pas être franchie à moins de passer par la compréhension des phénomènes responsables des résultats que l’on voudrait obtenir, la profonde connaissance des outils dont on dispose actuellement et la capacité de prévoir des modifications qui pourraient les adapter aux nouveau objectifs que l’on se donne. Motivés par ce contexte, nous avons essayé de répondre tout d’abord aux questions techniques que la problématique soulève : comment visualiser des charges lamellaires inorganiques ayant une structure multi-échelle et optiquement inertes (1) pendant leur mélange avec un polymère fondu visqueux réalisé à l’aide d’outils à parois opaques (2) ? Nous avons surmonté la première difficulté technique en rédigeant et puis en mettant en place un protocole de photo-fonctionnalisation d’argiles commerciales. La différence principale par rapport aux méthodes classiques de traçage (LDV, PIV) et aux méthodes qui se servent de sondes locales plus traditionnelles (LIF) est que, dans notre cas, ce sont les charges elles-mêmes qui font office de sondes – autrement dit, tous les ingrédients contribuent activement aux propriétés finales du mélange. De plus, en contrôlant le comportement en fluorescence des charges lamellaires photofonctionnalisées et, surtout, en corrélant leur émission aux possibles morphologies (agglomérée, intercalée, exfoliée), le suivi en ligne du mélange polymère/argile aurait un avantage double : les dimensions réduites des argiles photo-fonctionnalisées (même si agrégées, elles ne mesurent pas plus d’une dizaine de microns) en feraient des traceurs optimaux pour visualiser le mélange distributif sans perturber l’écoulement ; les propriétés optiques des argiles photo-fonctionnalisées pourraient être, selon la nature de la molécule fluorescente sélectionnée, sensibles aux changements de l’environnement et fournir par spectrofluorimétrie des renseignements à propos du mélange dispersif. Par rapport à la deuxième difficulté technique : nous avons repris une maquette froide (Visiovis) qui avait été assemblée auparavant dans nos laboratoires et qui présentait l’avantage considérable d’être constituée d’un système vis/fourreau dont le fourreau est entièrement transparent, mais qui avait été initialement conçue pour d’autres objectifs (Moguedet, 2005), et nous l’avons adaptée et développée davantage pour qu’elle nous permette de visualiser le mélange d’un polymère fondu avec les argiles précédemment photo-fonctionnalisées. PHOTO-FONCTIONNALISATION. Charges lamellaires. Le but principal de cette première partie du travail de thèse étant de trouver une méthode de photo-fonctionnalisation qui puisse rendre les charges lamellaires optiquement actives, nous avons d’abord sélectionné une argile commerciale organiquement modifiée (Cloisite ® 30B) Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 7/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. parmi les plus citées dans la littérature et trois molécules fluorescentes (9-anthracenemethanol, Nile Blue A Perchlorate et Rhodamine 6G Perchlorate), et nous avons testé plusieurs méthodes de photo-fonctionnalisation : (A) le gonflement libre de l’argile dans une solution contenant la molécule fluorescente (9-anthracenemethanol), (B) le mélange en voie sèche de l’argile avec la molécule fluorescente (9-anthracenemethanol) à l’aide d’un malaxeur, (C) le mélange en voie fondue de l’argile avec la molécule fluorescente (9-anthracenemethanol) et enfin (D) l’échange cationique de l’argile avec une molécule fluorescente ionique (Nile Blue A Perchlorate et Rhodamine 6G Perchlorate). Les échantillons ainsi modifiés ont été caractérisés (avant et après lavage, si possible) par Diffractométrie des Rayons-X (DRX), Analyse ThermoGravimétrique (ATG), Analyse Elémentaire (AE), Spectroscopie InfraRouge par Transformée de Fourier (IRTF), ATG couplée IR (ATG-IR) et spectrofluorimétrie. La détermination de la configuration des molécules fluorescentes suite aux interactions avec la structure inorganique qui les accueille est indispensable pour comprendre le mécanisme de photo-fonctionnalisation (évaluer s’il s’agit de l’adsorption de monomères, dimères ou agrégats d’ordre supérieur), pour juger la qualité des complexes inorganiques-organiques photo-fonctionnels obtenus (c’est-à-dire pour vérifier qu’ils soient effectivement photo-actifs et comprendre quelles informations peut-on déduire à partir des mesures de fluorescence) et utiliser les charges photo-fonctionnalisées de manière efficace et pertinente pour le suivi en ligne du mélange polymère/argile. En général, l’objectif de la photo-fonctionnalisation est de greffer/adsorber une molécule fluorescente (espèce optiquement active) sur les feuillets de silicate, aux bords de ces mêmes feuillets, ou encore de les introduire à l’intérieur des galeries comprises entre les feuillets. Les techniques de caractérisations ont été utilisées pour : vérifier que la photo-fonctionnalisation a bien eu lieu ; essayer de comprendre où est-ce que les molécules fluorescentes ont été adsorbées (à l’extérieur ou bien à l’intérieur des espaces interfoliaires) ; prouver l’efficacité du lavage suivant la photo-fonctionnalisation ; tester l’activité optique des complexes inorganiques-organiques obtenus. La méthode (A) était censée faciliter la diffusion des molécules fluorescentes dans les galeries – grâce à la tendance prononcée que les argiles ont à gonfler dans un milieu aqueux ou solvaté (exceptionnellement, dans notre cas, le milieu responsable du gonflement devait être aussi un bon solvant pour la molécule fluorescente). Que ce soit avec ou sans la molécule fluorescente, les gels que nous avons obtenus étaient à chaque fois bien gonflés et homogènes, mais difficiles à sécher : même si quelques unes des techniques de récupération/séchage que nous avons essayées étaient plus efficaces que d’autres, nous ne sommes jamais parvenus à éliminer complètement le solvant résiduel. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 8/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. Les problèmes rencontrés lors de l’élimination complète du solvant résiduel nous ont forcés à chercher une méthode alternative, qui ne demande aucun solvant. En effet, la méthode (B) nous a clairement montré tous les avantages de l’absence de solvants ; pourtant, elle n’est pas adaptée à la photo-fonctionnalisation des charges lamellaires, car en absence d’une phase liquide (quelle qu’elle soit sa nature) les molécules ne peuvent pas vraiment diffuser – d’ailleurs le mélange n’est pas non plus efficace, car une phase sèche n’est pas bien capable de transférer le cisaillement imposé par le malaxeur. La méthode (C) a été donc mise en place en prenant inspiration des avantages des deux autres méthodes : nous voulions profiter des mécanismes de diffusion, sans pour autant utiliser un milieu aqueux ou solvaté – nous avons donc choisi de remplacer les solvants par la phase liquide obtenues par fusion de la molécule fluorescente. Bien que cette méthode nous ait donné des gels équivalents (en termes de gonflement) à ceux obtenus par la méthode (A), le besoin d’un excès remarquable de molécule fluorescente ne correspond pas à des critères économiques. Nous avons constaté que, malheureusement, une modification réalisée avec une molécule neutre (9-anthracenemethanol) n’est pas stable – au contraire, elle est tout à fait réversible et ne résiste pas aux lavages. La méthode (D) dérive des techniques d’échange cationique traditionnellement utilisées pour caractériser et fonctionnaliser les minéraux argileux et qui, en particulier, sont souvent adoptées pour modifier le caractère hydrophile des argiles en organophile. L’idée d’effectuer un deuxième échange avec une molécule fluorescente ionique est inspiré des travaux récents d’un groupe de chercheurs du National Institute of Standards and Technology (NIST) à Gaithersburg, Maryland (USA). Cette méthode – que nous avons réalisée avec deux molécules fluorescentes à deux différentes concentrations dans le milieu d’échange – s’est révélée encore une fois la plus efficace pour la fonctionnalisation des charges lamellaires gonflantes – que ce soit avec le Nile Blue A Perchlorate ou la Rhodamine 6G Perchlorate. Nous avons ensuite décidé d’optimiser les paramètres pour l’échange cationique (D) et de définir la nature et la concentration de la molécule fluorescente la plus efficace. Pour ce faire, nous avons tout d’abord comparé les résultats obtenus en échangeant la même argile avec deux concentrations différentes de la même molécule fluorescente – correspondantes à 100% et 25% du taux de surfactant présent dans les galeries de la Cloisite ® 30B (dénommées 1MC et 0.25MC, respectivement). Effectivement, lorsqu’il s’agit de fluorescence, la concentration de fluorophore devient vite un factor primordial à cause du risque de quenching (diminution, voir disparition totale de l’intensité émise) et de la tendance spontanée des molécules fluorescentes Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 9/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. (surtout ioniques) à s’assembler en structures supramoléculaires, lesquelles peuvent parfois être désirables mais doivent être tout de même maitrisées. Nous avons montré que l’échange cationique en présence d’un excès de molécule fluorescente est possible mais pas nécessaire, car une fraction considérable de molécules fluorescentes ne pénètre pas dans les galeries et, même si c’est le cas, ne subit l’échange et est donc éliminée lors des lavages. En conclusion, nous avons retenu une concentration optimale correspondante à 25% du taux de surfactant présent dans les galeries de l’argile organiquement modifiée. Par rapport aux choix de la molécule fluorescente: sur la base de plusieurs évidences expérimentales, nous nous sommes plutôt orientés vers la Rhodamine 6G Perchlorate – laquelle s’est révélée non seulement plus efficace en termes d’échange cationique et d’expansion des galeries des argiles, mais aussi mieux détectables par les différentes techniques d’analyse employées (DRX et ATG, plus particulièrement). En plus, les plages d’absorption et d’émission de la Rhodamine 6G Perchlorate nous sont apparues mieux adaptées aux caractéristiques techniques de l’outil de visualisation pour lequel nous avions entrepris cette démarche de photofonctionnalisation. Nous avons aussi constaté que, indépendamment de sa concentration, la Rhodamine 6G Perchlorate est responsable d’un gonflement qui correspond toujours à une distance interfoliaire de 22 Å : nous avons émis l’hypothèse (supportée par des cas similaires présents dans la littérature et concernant l’absorption de molécules de solvant dans les galeries d’argiles organiquement modifiées) que les molécules fluorescentes diffusent dans les galeries et s’intercalent entre les chaînes carbonées initialement présentes, en remplissant ainsi le sites ioniques qui n’ont pas été occupés à cause de l’encombrement stérique, de façon à entourer la base des chaînes aliphatiques, à les supporter et à forcer l’angle initial de la configuration paraffinique à subir une augmentation (de 37° à 49°). Le remplissage des sites ioniques vacants par les molécules fluorescentes accroît la fraction de matière organique stablement adsorbée dans les galeries et, par conséquent, rend le « recouvrement » des surfaces inorganiques plus efficace, améliorant ainsi l’efficacité globale de l’échange cationique et regagnant une Capacité d’Echange Cationique (CEC) plus importante. Ce phénomène se vérifie chaque fois que les galeries d’argile se trouvent à accueillir deux espèce chimiques de taille différente : l’efficacité d’occupation du volume globalement disponible au sein des galeries en est améliorée. Le recouvrement de la CEC est le seul aspect qui pourrait éventuellement rendre le Nile Blue A Perchlorate préférable par rapport à la Rhodamine 6G Perchlorate – du moins il a le mérite de nous avoir révélé que l’efficacité des échanges cationiques ne dépend pas seulement de la Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 10/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. concentration de surfactant et de la température, mais aussi de la nature chimique, de la taille maximale (Dmax), de la configuration et du volume occupé par ses molécules. Enfin, en comparant par ATG les différents échantillons (avant et après lavage) obtenus avec les deux molécules fluorescentes et les deux concentrations, nous avons remarqué que : toute argile photo-fonctionnalisée lavée est plus stable (ou, du moins, autant stable) que la Cloisite ® 30B lavée – ce qui signifie que la photo-fonctionnalisation ne diminue pas la stabilité thermique des argiles ; toute argile photo-fonctionnalisée est stabilisée par le lavage (y comprise la Cloisite ® 30B commerciale) ; les deux molécules fluorescentes sont stabilisées, pour T>300°C, grâce à la protection thermique (propriété barrière) des feuillets de silicate ; dans tous les cas, l’argile la plus stable est celle qui a été échangée avec 1MC de molécule fluorescente, et puis lavée ; dans tous les cas, l’argile la moins stable est celle qui a été échangée avec 1MC de molécule fluorescente, mais pas lavée. COMPLEXES PHOTO-FONCTIONNELS. Echange cationique. Nous avons donc mis au point et optimisé un protocole de photo-fonctionnalisation pour des charges lamellaires commerciales organiquement modifiées (Cloisite ® 30B) sélectionnées parmi les plus utilisées : il s’agit d’un deuxième échange cationique réalisé dans une solution 90/10 eau permutée/éthanol à 80°C en présence d’une quantité de molécule fluorescente ionique égale à 25% du taux de surfactant présent initialement dans les galeries de l’argile. En prévision des applications que nous comptions faire des complexes photo-fonctionnels ainsi obtenus, nous avons préféré la Rhodamine 6G Perchlorate au Nile Blue A Perchlorate. En se basant sur le protocole précédemment rédigé, nous avons ensuite réalisé d’autres complexes photo-fonctionnels et, ici, nous présentons et comparons les résultats obtenus à partir de quatre argiles commerciales (Cloisite ® Na+, Cloisite ® 30B, Cloisite ® 10A et Cloisite ® 15A). En particulier, dans cette sélection nous avons inclus aussi une argile commerciale 100% inorganique (Cloisite ® Na+) parce que nous estimions que le même procédé réalisé en absence de molécules organiques dans les espaces interfoliaires ferait une bonne référence pour mieux comprendre certains mécanismes d’adsorption et d’échange cationique. Les autres argiles ont été sélectionnées sur la base de leur fréquence d’utilisation (on les retrouve souvent dans les publications scientifiques) et de leur surfactant : la Cloisite ® 30B contient une longue chaîne Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 11/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. aliphatique et deux courtes chaîne terminées –OH, la Cloisite ® 10A contient une longue chaîne aliphatique et un cycle aromatique, la Cloisite ® 15A contient deux longue chaîne aliphatiques. L’argile hydrophile et ces argiles organophiles présentent des distance interfoliaire variables : 12.4 Å (Cloisite ® Na+), 17.5 Å (Cloisite ® 30B), 19.2 (Cloisite ® 10A), 25.7 (Cloisite ® 15A). Après avoir caractérisé les matériaux dans leur état initial (les argiles et la Rhodamine 6G Perchlorate), nous avons comparé les résultats obtenus pour la série de quatre argiles photofonctionnalisées. L’analyse élémentaire nous a fourni toutes les tendances attendues par rapport aux pourcentages atomiques de Na, C, N, H et Si. En particulier, pour les argiles commerciales pas encore fonctionnalisées mais lavées, nous avons observé que la variation en pourcentage de C adsorbé est inversement proportionnelle à la masse molaire du surfactant – autrement dit, le pourcentage de sites Na+ remplacés lors d’un échange cationique diminue d’autant plus que l’encombrement stériques des molécules de surfactant augmente. Cette observation peut paraître logique mais devient essentielle pour comprendre la notion de « recouvrement » et pour une meilleure exploitation de la CEC d’une argile. En effet, l’analyse élémentaire nous montre que le pourcentage atomique de N augmente pour toutes les argiles après photo-fonctionnalisation : cette augmentation prouve qu’un deuxième échange cationique avec une petite molécule (plus petite que le surfactant utilisé pour le premier échange cationique) permet de mieux exploiter la CEC de l’argile et améliore donc le recouvrement de la surface inorganique par les molécules organiques. Par spectrofluorimétrie nous avons vérifié que les quatre complexes inorganiquesorganiques sont effectivement photo-actifs. Nous avons remarqué un léger déplacement hypsochromique seulement dans les spectres d’absorption et d’émission du complexe obtenu à partir de la Cloisite ® 10A – ce qui pourrait être lié à la présence d’un cycle aromatique dans la structure des molécules de surfactant. D’autres caractérisations en fluorescence, plus développées et mieux adaptées, seraient nécessaires pour comprendre ce genre de phénomènes ; l’objectif de la photo-fonctionnalisation dans ce travail de thèse était principalement de rendre les argiles optiquement actives, et nous l’avons atteint. PROCEDE DE MISE EN ŒUVRE. Suivi en ligne du mélange. Les procédés de mise en œuvre tels que l’extrusion et l’injection moulage sont réalisés à l’aide d’outils de transformation ayant des géométries plutôt complexes et, par conséquent, leur suivi en ligne est particulièrement difficile. Nous présentons, ici, un outil original et innovant (la Visiovis) intégralement conçu, assemblé et développé dans nos laboratoires sur une période totale d’environ cinq ans – un peu plus que deux ans dans le cadre d’un précédent travail de Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 12/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. thèse (Maël Moguedet, Thèse de Doctorat, 2005) et les trois ans prévus pour le travail de thèse qui fait l’objet de ce manuscrit (Antonella Esposito, Thèse de Doctorat, 2008). Comme son nom l’indique, la Visiovis est un outil consacré à la visualisation des procédés de mise en œuvre. Pourquoi ? Parce que cette maquette comprend un système vis/fourreau qui a la particularité d’être constitué d’un fourreau transparent – ce qui permet de visualiser les écoulements en temps réel et dans toutes les directions. La Visiovis avait été initialement conçue pour visualiser les trajectoires 3D d’une seule particule fluorescente plongée dans un fluide transparent : il s’agissait, donc, d’un simple problème de dynamique des fluides. Au cours de ces trois dernières années, nous avons légèrement modifié sa configuration initiale pour l’adapter au suivi en ligne du mélange polymère/nanocharges (plus en particulier, pour la caractérisation de la dispersion et distribution de charges lamellaires dans un polymère fondu) – ce qui pourrait nous renseigner sur le procédé de mise en œuvre des nanocomposites par mélange en voie fondue. Il faut dire que Moguedet et ses collaborateurs conçurent le système vis/fourreau de la Visiovis pour que ses paramètres géométriques soient semblables aux valeurs caractéristiques des sections de pompage des vis d’extrusion et d’injection industrielles. Après avoir évalué les avantages et les désavantages du système dans sa configuration initiale, nous l’avons modifié et y avons intégré un spectromètre afin d’en augmenter davantage les potentialités. En résumé, la Visiovis se compose actuellement de : un système vis/fourreau : une vis à pas carré (pas 40 mm, diamètre à cœur 30 mm) est ajustée dans un fourreau en PMMA (diamètre 40 mm) avec une profondeur du chenal de 5 mm ; un moteur électrique qui actionne la vis (vélocité max 44 tr/min, couple max 9 N m) ; une ouverture qui permet d’introduire le fluide et le mélange maître avec le traceur ; un tuyau qui connecte la sortie à l’entrée et réalise un circuit fermé ; un support mobile autour du système pour installer ce qu’il faut pour la visualisation ; quatre caméras CCD (Basler A301F) alignées le long du fourreau : enregistrement de 80 images/s, résolution 640 480 pixels avec une profondeur de 8 bits (256 niveau de gris) ; une source laser pulsée ( 532 nm, puissance nominale 20 mW CW) ; un système optique qui crée une nappe laser (bidimensionnelle) à partir de la source laser pulsée (monodimensionnelle) ; un transducteur électromécanique (essentiellement un interrupteur) qui connecte la rotation de la vis aux caméras et pilote ainsi l’acquisition d’images (une image par tour de vis) ; Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 13/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. un spectromètre (USB2000+, Oceanoptics) interfacé avec le système vis/fourreau via une fibre optique (diamètre 600 m, résolution 2.5 nm) positionnée à l’opposée des caméras ; un robinet situé au milieu du tuyau de recirculation, qui facilite la vidange. Une fois que la configuration de la Visiovis a été modifiée et que les charges lamellaires ont été photo-fonctionnalisées, nous avons effectué une calibration « qualitative » des systèmes de détection (caméras CCD et spectromètre) – le but étant de vérifier d'abord que les complexes photo-fonctionnels aient une émission en fluorescence suffisamment intense pour être captée et, ensuite, d’estimer la concentration d’argile photo-fonctionnalisée nécessaire pour une détection optimale. La calibration des systèmes de détection a été effectuée avec le complexe à base de Cloisite ® 30B. Pour ce faire, nous avons réalisé plusieurs mélanges maîtres à base de Siliconöl M10000 (huile silicone transparente – essentiellement du PDMS terminé méthyle – viscosité 10 Pa s) et ayant des concentrations d’argile photo-fonctionnalisée connues et contrôlées (0%, 0.001%, 0.0025%, 0.005%, 0.01%, 0.025%, 0.05%, 0.1%, 0.25%, 0.5%, 1% en poids). Les mélanges maîtres ont été préparés avec un disperseur TurboTest Rayneri 33/300P équipé d’un disque dilacérateur (diamètre 60 mm) tournant à 1000 tr/min pendant 20 min. Le choix de la concentration optimale d’argile photo-fonctionnelle à introduire dans le mélange maître (qui sera ensuite injecté dans le système vis/fourreau préalablement rempli de PDMS propre) n’est pas simple et doit se baser essentiellement sur les limites de détection des caméras CCD et du spectromètre, mais aussi sur la profondeur de pénétration de la nappe laser (qui doit être au moins égale à la profondeur du chenal de la vis, c’est-à-dire au moins 5 mm). Sur la base de toutes les évidences collectées lors de la calibration, nous avons trouvé que le meilleur protocole pour effectuer les expériences Visiovis est d’injecter, dans le système vis/fourreau de la Visiovis préalablement rempli de fluide modèle transparent, 10 mL d’un mélange maître préparé avec 0.25% en poids d’argile photo-fonctionnalisée et du même fluide modèle. N’ayant pas d’autres fluides macromoléculaires transparents à proposer en alternative au PDMS, nous avons continué à utiliser l’huile silicone comme matrice modèle. En ce qui concerne les charges photo-actives : nous avons testé les quatre complexes photo-fonctionnels pour vérifier leur activité optique et, curieusement, nous avons constaté que tous les complexes étaient correctement visualisées par les caméras CCD sauf celui à base de Cloisite ® Na+. Pour cette raison, nous avons poursuivi les expériences seulement avec les complexes photo-fonctionnels obtenus à partir des argiles organiquement modifiées (Cloisite ® 30B, 10A et 15A). Par rapport à la modalité d’injection du traceur dans le système vis/fourreau : nous avons conçu deux méthodes, mais nous en avons utilisé seulement une (la seule possible, pour l’instant). La première méthode est plus adaptée à Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 14/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. la modélisation du procédé d’extrusion (mélange en voie fondue de granulés de polymère pur avec une charge sous forme de poudre, afin de formuler des granulés composites) et demande la préparation d’un « mélange maître non homogénéisé » composé de trois couches : une couche de poudre sèche entre deux couches de PDMS pur (volume total 10 mL, taux de poudre 0.25% en poids). La seconde méthode (celle que nous avons utilisée) est plus adaptée à la modélisation de la phase de pompage du procédé d’injection moulage (deuxième fusion et homogénéisation des granulés composites, précédemment formulés par extrusion) et demande la préparation d’un mélange maître ayant une bonne qualité (en termes de distribution et dispersion des charges), d’un volume total de 10 mL, avec un taux de charges de 0.25% en poids. Toutes les expériences ont été réalisées en injectant le mélange maître dans le système vis/fourreau via l’ouverture et grâce à une seringue de 20 mL, coupée à son extrémité de manière à éviter tout cisaillement non contrôlé. La vitesse de rotation de la vis a été réglée à 20 tr/min et les expériences ont été réalisées dans le noir. Une fois le mélange maître injecté dans le système, l’acquisition des données est automatique : pilotées par l’interrupteur, les caméras acquièrent une image par tour de vis ; le spectromètre, de son côté, est programmé pour enregistrer, via la fibre optique, un spectre d’émission de fluorescence toutes les trois secondes. Les deux systèmes de détection sont indispensables, car seules les caméras ne suffisent pas à obtenir une caractérisation correcte et complète du procédé de mise en œuvre des nanocomposites : la spectrofluorimétrie, sensible à des phénomènes qui intéressent une échelle spatiale plus petite par rapport aux caméras, pourrait donner quelques informations complémentaires – notamment concernant la dispersion (intercalation et/ou exfoliation) des charges lamellaires. Les images acquises avec la Visiovis peuvent être utilisées pour reconstruire des vidéos (information qualitative qui permet d’évaluer directement et visuellement l’évolution temporelle de la distribution des nanocharges dans le volume de fluide compris entre deux filets, la surface de la vis et la paroi du fourreau), mais nous avons aussi essayé d’en extraire une information quantitative en mettant au point deux traitements d’image qui peuvent être exécutés dans un environnement Matlab. Les deux traitements se basent, respectivement, sur (1) l’intégrale de l’écart type de luminosité des images et (2) la Transformée de Fourier (TF) d’images texturées. Pour chaque image d’une séquence donnée, le traitement (1) calcule les valeurs locales de l’écart type de la luminosité sur toute la surface de l’image et trace la courbe du logarithme de l’intégrale de cet écart type en fonction du numéro qui identifie les images de la séquence. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 15/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. (1) INTEGRALE DE L’ECART TYPE DE LA LUMINOSITE Interprétation. Théoriquement, plus faible est l’intégrale de l’écart type de la luminosité, plus homogène est le matériau visualisé et donc plus efficace est le mélange distributif. Limites. L’intégrale de l’écart type de la luminosité ne sera jamais inférieure à une valeur seuil à cause de la différence entre le mélange visualisé (jamais complètement noir) et le profil de la vis (toujours complètement noir). De plus, ce traitement ne détecte pas la morphologie des images texturées : deux images ayant différente textures pourraient donner les même résultats en termes d’écart type de luminosité. La Transformée de Fourier (TF) est un algorithme qui peut codifier une image texturée en utilisant les fréquences avec lesquelles ses unités texturales élémentaires se répètent (dans notre cas, il s’agit de spirales ou volutes claires – produites par fluorescence – sur un fond noir). Comme nous avions à faire avec des séries d’images numériques (constituées de pixels), nous avons naturellement utilisé la Transformée de Fourier dans sa version Discrète (DFT). La DFT d’une image numérique donne le spectre de toutes les fréquences comprises entre une fréquence maximale max et une fréquence minimale min. Ces deux fréquences peuvent être attribuées à deux propriétés intrinsèques de toute image numérique : la présence d’une matrice régulière de pixels (niveau de détail max) et la dimension maximale de l’image (niveau de détail min), respectivement. Une image finement texturée sera codifiée par des hautes fréquences dans le domaine de Fourier, puisque les plus hautes fréquences peuvent décrire les plus petits détails. Pour chaque image d’une séquence donnée, le traitement (2) effectue l’anamorphose de l’image, calcule le logarithme de la norme au carré, soustrait le bruit d’une image vide choisie comme référence, calcule l’intensité moyenne de chaque fréquence entre la minimale et la maximale et trace la courbe des valeurs moyennes en fonction de la fréquence. Ensuite, à chaque image il associe la fréquence la plus significative (autrement dit, la fréquence la plus probable du point de vue statistique, que nous appellerons fréquence représentative) et trace la courbe de ces dernières valeurs en fonction du numéro qui identifie les images de la séquence. Si le premier traitement d’image est intuitif et immédiat, nous avons préféré valider le deuxième en l’appliquant à une série d’images numériques synthétiques obtenues avec l’aide de Yves Béreaux, un des collaborateurs de Maël Moguedet à l’époque de son travail de thèse. Nous avons ainsi rapproché la modélisation numérique aux expériences de laboratoire – ce qui ouvre d’autres perspectives pour les deux secteurs. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 16/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. (2) TRANSFORMEE DE FOURIER D’IMAGES TEXTUREES Interprétation. Théoriquement, une fréquence nulle indiquerait un mélange parfaitement homogène, tandis que des hautes fréquences indiqueraient la présence d’unités texturales fines et régulières. Si la miscibilité est parfaite… Plus la fréquence représentative est basse, plus homogène est le matériau visualisé et donc plus efficace est le mélange distributif. Si la miscibilité est partielle… Plus la fréquence représentative est haute, plus la texture est fine et donc plus efficace (encore que incomplet) est le mélange distributif. Limites. En réalité, on ne pourra jamais atteindre des fréquences nulles car toute image numérique, par définition, a des dimensions finies (seulement une image infinie peut avoir une fréquence nulle). Par conséquent, toute éventuelle fréquence nulle devrait être interprétée en tant que valeur relative. De plus, le contraste entre le matériau visualisé (jamais complètement noir) et le profil de la vis (toujours complètement noir) génère une fonction de Heaviside (c’est-àdire une fonction marche) qui correspond à une texture artificielle toujours présente dans les images. Conscients du fait que le développement d’un outil pour la visualisation des procédés de mise en œuvre des nanocomposites – surtout quand il est associé à la conception de nouveaux systèmes de détection et caractérisation en ligne – n’est pas une mince affaire, nous avons continué à tester le système en ayant comme objectif d’en reconnaître les limites mais aussi les potentialités. Nous avons réalisé quelques expériences pour avoir des résultats préliminaires qui puissent aider à développer davantage cet outil. En particulier, nous avons cherché à évaluer : les différences de comportement des trois charges lamellaires photo-fonctionnelles, préalablement modifiées par échange cationique à partir de Cloisite ® 30B, 10A et 15A ; la possibilité d’utiliser la Visiovis pour des expériences de traçage conventionnelles (c’est-à-dire en injectant un mélange maître chargé principalement d’argile non modifiée, mais aussi d’une petite quantité de la même argile photo-fonctionnalisée) ; la façon dont la Visiovis visualise un changement de contre-pression. Les résultats préliminaires n’ont révélé aucune différence majeure en termes de mélange distributif entre les différents complexes photo-fonctionnels testés, et l’absence de différences (observée par tous les système de détection et avec tous le traitements d’images) nous a fait émettre l’hypothèse (successivement confirmée par DRX et rhéologie) qu’aucune des argiles ne peut établir des interactions fortes avec la matrice PDMS – ce qui se traduit en une réduction de l’efficacité du mélange distributif et dispersif, aggravée par le fait que le profil de la Visiovis n’est pas le profil le plus adapté pour le mélange. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 17/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. De ce fait, nous avons continué à travailler de préférence avec le complexe photo-actif à base de Cloisite ® 30B, et notamment nous avons préparé trois mélange maître pour essayer de vraies expériences de traçage : un mélange maître contenant 0.25% en poids d’argile photo-fonctionnalisée et 0% en poids d’argile non modifiée (taux total d’argile 0.25% en poids) ; un mélange maître contenant 0.25% en poids d’argile photo-fonctionnalisée et 0.75% en poids d’argile non modifiée (taux total d’argile 1% en poids) ; un mélange maître contenant 0.25% en poids d’argile photo-fonctionnalisée et 2.75% en poids d’argile non modifiée (taux total d’argile 3% en poids). Nous avons constaté un problème majeur lié à l’augmentation du taux d’argile : le fait que le fluide visualisé devienne de plus en plus opaque. Par conséquent, nous avons observé que la profondeur de pénétration de la nappe laser diminue en dessous du seuil critique, et que les particules d’argile diffusent beaucoup plus la lumière d’excitation et diminuent ainsi la clarté optique du système de visualisation. Ce genre de phénomènes est invisible aux traitements d’images – d’où l’intérêt à coupler tous les systèmes de détection. En revanche, les spectres en fluorescence ne sont pas affectés par l’augmentation de concentration et continuent à détecter seulement la portion d’argile photo-fonctionnalisée. Pour évaluer les effets de la contre-pression, nous avons installé une vanne au milieu du tuyau qui réalise le circuit fermé, à la sortie du système vis/fourreau. Nous avons utilisé cette vanne pour régler le débit et, plus en particulier, nous avons effectué trois expériences avec : la vanne complètement ouverte (ce qui correspond au système non perturbé) ; la vanne ouverte à moitié ; la vanne complètement fermée (ce qui correspond à un arrêt complet du débit). Tout système vis/fourreau est sujet à une contre-pression intrinsèque, due simplement au fait que la sortie soit représentée par une section réduite par rapport au diamètre du fourreau. La contre-pression agit contre le débit et, donc, fait obstacle à l’écoulement – ce qui fait augmenter le temps de résidence et facilite la recirculation dans le chenal de la vis. Parfois, on joue sur la contre-pression pour améliorer la qualité du mélangeage. Nous avons observé que le fait de fermer à moitie la vanne de la contre-pression ne change pas la forme globale des courbes obtenus avec les traitements d’image, mais produit quand même des changements remarquables en termes de lissage et décalage verticales des courbes. En revanche, quand la vanne de la contre-pression est complètement fermée, les courbes changent radicalement d’aspect et les temps s’allongent considérablement à cause de la recirculation. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 18/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. BILAN DE LA MISE AU POINT DE LA VISIOVIS. Problèmes rencontrés et suggestions d’amélioration. Pour conclure, nous avons fait le bilan des progrès déjà accomplis et des améliorations qui devraient être encore apportées au système pour que la Visiovis soit vraiment un outil utile à la visualisation et, ensuite, à la compréhension et à la maîtrise du procédé de mise en œuvre des nanocomposites par mélange en voie fondue. Fluide modèle. Nous avons constaté que le PDMS terminé méthyle n’est pas forcement compatible avec les argiles commerciales organiquement modifiées – censées être compatibles avec des polymères carbonées (plus nombreux et donc plus répandus que les polysiloxanes). Dans la littérature nous avons trouvé quelques confirmations du fait que disperser des charges dans n’importe quel polysiloxane est certainement un défi plus difficile à vaincre par rapport aux polymères carbonées. Equilibrer la balance hydrophilie/hydrophobie n’est pas suffisant pour rendre une charge compatible avec un polysiloxane : un contrôle plus strict de la chimie de surface des charges, mais aussi de la chimie des macromolécules qui constituent la matrice, est essentiel. D’autre part, la fourreau en PMMA n’est pas compatible avec des fluides réactifs, des liquides précurseurs, des solvants (même en très faibles quantités), et ne peut pas être chauffé : le cahier des charges concernant le fluide modèle idéal pour la Visiovis est lourd et pas facile à respecter. Nous suggérons tout de même de songer à une modification des chaînes PDMS afin d’améliorer la compatibilité entre les charges et la matrice sans pourtant interagir avec la surface en PMMA du fourreau. Charges photo-fonctionnelles. Dans l’hypothèse de trouver un fluide modèle mieux adapté que le PDMS terminé méthyle, à l’heure actuelle on disposerait déjà de quatre charges lamellaires photo-fonctionnelles – préparées à partir de la Cloisite ® Na+, de la Cloisite ® 30B, de la Cloisite ® 10A et de la Cloisite ® 15A. Si aucun autre fluide n’est adapté pour remplacer le PDMS terminé méthyle, nous suggérons de préparer une nouvelle série de complexes photofonctionnels contenant la Rhodamine 6G Perchlorate comme espèce optiquement active, et un surfactant spécifiquement conçu pour assurer la compatibilité avec la matrice PDMS. Nous avons déjà commencé à travailler en cette direction, mais notre travail n’a pas encore abouti par manque de temps. Enfin, une dernière possibilité est offerte par la synthèse de billes de silice ayant une morphologie cœur-écorce et qui pourraient envelopper la rhodamine. Dans ce cas, la Visiovis visualiserait seulement le mélange distributif et le spectromètre ne serait plus essentiel. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 19/20 Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. Configuration générale de la Visiovis. Nous n’envisageons pas de modifications de la configuration générale de la Visiovis. Nous n’avons pas remarqué de sérieux problèmes par rapport à la position des caméras, à la position de la fibre optique, à la forme et position de la nappe laser… La source d’excitation devrait être changée seulement si un autre fluorophore est sélectionné pour remplacer la Rhodamine 6G Perchlorate – mais nous suggérons vivement de ne pas changer de fluorophore. Le fourreau en PMMA représente en même temps une grosse limite mais aussi le plus grand avantage de la Visiovis : nous souhaiterions suggérer de reconstruire le fourreau (toujours en PMMA si nécessaire) avec des parois externes carrées – ce qui faciliterait la visualisation et éliminerait tout problème optique lié à la réflexion sur une surface courbée. Par rapport au profil de la vis : nous invitons à tester plusieurs profils de vis afin d’en évaluer les performances. Enfin, nous conseillons de concevoir un dispositif fixe pour une alimentation plus fiables du traceur à l’intérieur du système vis/fourreau et pour une meilleure répétabilité des phénomènes observés. Exploitation des données acquises. Le rapprochement que nous avons opéré, avec la collaboration de Yves Béreaux, entre la simulation numérique et les expériences de laboratoire, représente un point de départ extrêmement intéressant pour d’autres futurs développement des deux secteurs. Néanmoins, la Visiovis n’a pas encore été exploitée à la hauteur de toutes ses potentialités, principalement à cause des problèmes de compatibilité rencontrés entre le fluide modèle et les complexes photo-fonctionnels à base d’argile. Nous encourageons à poursuivre le travail en utilisant les mêmes systèmes de détection et les mêmes traitements d’image. Prélèvement d’échantillons et validation des résultats. Toute nouvelle technique de caractérisation se doit d’être validée par les résultats obtenus sur un même échantillon avec des techniques de caractérisation traditionnelles. Pour ce faire, il est nécessaire de pouvoir effectuer des prélèvements d’échantillons pendant les expériences de visualisation. Nous n’avons pas encore réalisé un dispositif qui permette le prélèvement en ligne d’échantillons, mais nous en reconnaissons l’importance. Pour l’instant, nous avons caractérisé les mélanges maître (avant l’injection dans la Visiovis) par DRX et rhéologie : nous avons rencontré un certain nombre de problèmes qui nous ont empêché d’effectuer une corrélation directe et fiable entre les résultats obtenus avec la Visiovis et ceux obtenus avec les techniques de caractérisation traditionnelles. Nous en avons déduit seulement une confirmation des remarques faites auparavant à propos du manque de compatibilité entre les charges photo-fonctionnelles et la matrice PDMS. Cet aspect de validation est sans doute prioritaire pour le futur développement de la Visiovis. Antonella ESPOSITO Thèse INSA de Lyon (2008) Résumé détaillé en français 20/20 FOLIO ADMINISTRATIF THESE SOUTENUE DEVANT L'INSTITUT NATIONAL DES SCIENCES APPLIQUEES DE LYON NOM : ESPOSITO (avec précision du nom de jeune fille, le cas échéant) DATE de SOUTENANCE : 5 décembre 2008 Prénoms : Antonella TITRE : Visualisation de l’écoulement dans un système vis/fourreau. Suivi en ligne du mélange polymère/nanocharges. NATURE : Doctorat Numéro d'ordre : 2008-ISAL-0124 Ecole doctorale : Matériaux de Lyon Spécialité : Matériaux Polymères et Composites Cote B.I.U. - Lyon : T 50/210/19 / et bis CLASSE : RESUME : L’addition de nanocharges aux polymères à l’état fondu semble pouvoir en améliorer considérablement les propriétés, pourvu que cette deuxième phase soit parfaitement désagglomérée, dispersée et distribuée dans la matrice. Les charges lamellaires (argiles) et leur nanocomposites attirent depuis quelque temps une attention croissante, tant de la part du monde académique que de l’industrie, imposant ainsi des critères d’homogénéité du mélange polymère/charges plus stricts et étendus sur plusieurs échelles. D’autre part, la morphologie des nanocomposites à base de charges lamellaires est difficile à caractériser et presque autant difficile à décrire, étant souvent affectée par des problèmes de dispersion et/ou distribution. Une nouvelle méthode pour le suivi en ligne des évolutions morphologiques des mélanges polymère/argile au cours du procédé de mise en œuvre pourrait contribuer de façon significative au développement et à la commercialisation de cette catégorie de nanocomposites. L’objectif de ce travail de thèse est de poser les bases pour le développement d’une méthode de caractérisation morphologique en temps réel qui puisse permettre de mettre en évidence et comprendre les mécanismes de dispersion/distribution de charges dans un milieu visqueux (polymère thermoplastique fondu ou résine thermodurcissable non réticulée) en écoulement dans un système ayant une géométrie complexe (zone de pompage des ensemble vis/fourreau pour l’extrusion et l’injection), en démystifiant enfin les outils de transformation. Nous avons repris une maquette froide (Visiovis), assemblée auparavant dans le laboratoire du Site de Plasturgie INSA à Oyonnax (Maël Moguedet, Thèse INSA, 2005), présentant l’avantage considérable d’être constituée d’un fourreau entièrement transparent, et nous l’avons adaptée et développée davantage afin de visualiser le mélange d’un polymère modèle (PDMS) avec des argiles préalablement rendues photo-actives. Nous nous sommes donnés les moyens pour de telles expériences de visualisation en entreprenant plusieurs essais originaux de photo-fonctionnalisation des argiles avec différentes fluorophores, et en mettant finalement au point un protocole de photo-fonctionnalisation par échange cationique de montmorillonites avec la Rhodamine 6G Perchlorate ; ce protocole a été ensuite utilisé pour rendre optiquement actives quatre argiles commerciales parmi les plus récurrentes dans la littérature. Nous avons caractérisé les complexes photoactifs ainsi obtenus par DRX, ATG, analyse élémentaire, spectroscopie FTIR, ATG couplée FTIR, spectrofluorimétrie. Après avoir calibré les systèmes de détection dont nous avons équipé la Visiovis (caméras CCD et spectromètre, ce dernier interfacé avec le système vis/fourreau par fibre optique), nous en avons exploré le potentialités et les limitations de visualisation dans une configuration d’éclairage planaire par nappe laser, similairement aux méthodes déjà largement utilisées pour le suivi en ligne des écoulements (PLIF, PVT, PIV). Des méthodes pour le traitement des données acquises ont été suggérées et évaluées. MOTS-CLES : polymère fondu, PDMS, systèmes modèles, compoundage, mélange, argiles, fonctionnalisation, extrusion, injection moulage, suivi en ligne, nanocomposite, morphologie, dispersion, distribution, fluorescence, laser, visualisation des écoulements Laboratoire(s) de recherche : Laboratoire des Matériaux Macromoléculaires (LMM) Ingénierie des Matériaux Polymères (IMP) – UMR CNRS #5223 – INSA de Lyon Directeur(s) de thèse: Jean-Yves CHARMEAU Jannick DUCHET-RUMEAU Président de jury : Composition du jury : Rapporteurs : José Maria KENNY, Serge BOURBIGOT Examinateurs : Jean-Jacques FLAT, Jean-François GERARD Directeur(s) de thèse : Jean-Yves CHARMEAU, Jannick DUCHET-RUMEAU RAW MATERIALS RAW MATERIALS M w = 208.26 M w = 208.26 g/mol Dm ax = 10 Å g/mol Dm ax = 10 Å 9-anthracenemethanol 9-anthracenemethanol M w = 417.84 M w = 417.84 g/mol Dm ax = 15 Å g/mol Dm ax = 15 Å Nile Blue A Perchlorate Nile Blue A Perchlorate M w = 543.01 M w = 543.01 g/mol Dm ax = 14 Å g/mol Dm ax = 14 Å Rhodamine 6G Perchlorate Rhodamine 6G Perchlorate M w = 360.80 M w = 360.80 g/mol MC = 90 g/mol meq 100 g MC = 90 Dm ax = 29 Å Dm ax = 29 Å MT2EtOH Cloisite ® 30B MT2EtOH Cloisite ® 30B M w = 382.80 M w = 382.80 g/mol meq = 125 MC g/mol meq = 125 MC 100 g 100 g 2MBHT Cloisite ® 10A 2MBHT Cloisite ® 10A M w = 527.60 M w = 527.60 g/mol MC = 125 g/mol meq 100 g MC = 125 2M2HT Cloisite ® 15A SAMPLE NOMENCLATURE Cation exchange processing (D) 2M2HT Cloisite ® 15A SAMPLE NOMENCLATURE Cation exchange processing (D) Example: C30B 1MC RhP Example: C30B 1MC RhP pristine clay concentration of the fluorescent molecule (fraction) pristine clay concentration of the fluorescent molecule (fraction) fluorescent molecule C30B C10A C15A CNa+ MC CEC NBAP RhP meq 100 g Cloisite ® 30B Cloisite ® 10A Cloisite ® 15A Cloisite ® Na+ Modifier Concentration Cation Exchange Capacity Nile Blue A Perchlorate Rhodamine 6G Perchlorate fluorescent molecule C30B C10A C15A CNa+ MC CEC NBAP RhP Cloisite ® 30B Cloisite ® 10A Cloisite ® 15A Cloisite ® Na+ Modifier Concentration Cation Exchange Capacity Nile Blue A Perchlorate Rhodamine 6G Perchlorate meq 100 g