Capillary discharge sources of hard UV radiation
Transcription
Capillary discharge sources of hard UV radiation
INSTITUTE OF PHYSICS PUBLISHING PLASMA SOURCES SCIENCE AND TECHNOLOGY Plasma Sources Sci. Technol. 11 (2002) A64–A68 PII: S0963-0252(02)39384-8 Capillary discharge sources of hard UV radiation C Cachoncinlle, R Dussart, E Robert, S G¨otze, J Pons, S R Mohanty, R Viladrosa, C Fleurier and J M Pouvesle Gremi-espeo, Universit´e d’Orl´eans, 14 Rue D’Issoudun, BP 6744, 45067 Orl´eans Cedex 2, France E-mail: christophe.cachoncinlle@univ-orleans.fr Received 26 September 2001, in final form 2 July 2002 Published 19 August 2002 Online at stacks.iop.org/PSST/11/A64 Abstract We developed and studied three different extreme ultraviolet (EUV) capillary discharge sources either dedicated to the generation of coherent or incoherent EUV radiation. The CAPELLA source has been developed especially as an EUV source for the metrology at 13.4 nm. With one of these sources, we were able to produce gain on the Balmer-Hα (18.22 nm) and Hβ (13.46 nm) spectral lines in carbon plasma. By injecting 70 GW cm−3 we measured gain-length products up to 1.62 and 3.02 for the Hα and Hβ, respectively. Optimization of the EUV capillary source CAPELLA led to the development of an EUV lamp which emits 2 mJ in the bandwidth of the MoSi mirror, per joule stored, per shot and in full solid angle. The wall-plug efficiency is 0.2%. Stability of this lamp is better than 4% and the lamp can operate at a repetition rate of 50 Hz. 1. Introduction 1800 0963-0252/02/SA0064+05$30.00 © 2002 IOP Publishing Ltd 1600 Intensity (a.u.) Two fields of scientific and technical applications mainly drive the research on sources capable of emitting a significant level of photon fluxes in the spectral domain of the extreme ultraviolet (EUV). The first one is the biological and living science research for which the generation of photon in the so-called ‘water window’ is of crucial interest. A compact EUV source [1], driven by an electrical discharge, could easily be applied at very low cost to biological imaging in living cells. As illustrated in figure 1, the wavelength range from these sources spans over the ‘water window’ region of EUV or ‘soft-x-ray region’ where water is less absorbing than carbon. The second field of application is EUV microlithography. In the area of microlithography and materials processing, the mercury vapour discharge lamp has been providing very low cost illuminating sources for the last two decades. More recently, UV laser sources, at wavelengths as short as 157 nm, have been tested for sub-micrometre lithography for large volume manufacturing of integrated circuits. As a consequence of the well-known ‘Moore’s law’, new generation lithography (NGL) will require shorter wavelengths in the EUV region for sub-0.1 µm features. During the last few years tremendous efforts have been made to realize LαCVI 3.4 nm CV 4.0 nm 1400 CV 3.5 nm 1200 1000 800 600 400 2.0 2.5 3.0 3.5 Wavelength (nm) 4.0 4.5 Figure 1. Carbon spectra in the ‘water window’ obtained with the ablative capillary discharge developed in this experiment. and develop new EUV sources for the illumination system. Europe, Japan and the United States have already initiated efforts to solve intricate problems of EUV lithography like mask, metrology, photoresists and ‘alpha’ tool. Significant progress has been made in all areas of technological issues concerning EUV lithography. In contrast, none of the presently available EUV sources really matches the technical and financial requirements of the illuminating device for commercial tools. Much effort is still needed through both Printed in the UK A64 Capillary sources of UV radiation research and development programmes on the various EUV sources to fulfil the specifications required, especially in terms of lifetime and average power at reasonable cost. Various types of EUV sources already exist. One of them is a laser-produced plasma EUV source [1–3] in which a powerful laser is focused on a target to convert energy into EUV radiation. The efforts, for many years, towards developing a powerful EUV source based on this technique have led now to the production of high fluxes in the EUV spectral range. The brightest EUV source for scientific applications is obtained from the synchrotron machines. Nevertheless, a synchrotron machine cannot be suitable for commercial usage due to its huge size and cost. On the other hand, EUV sources based on electrical discharges are of great interest from the commercial viewpoint since they are simple, compact, costeffective, and they can operate with a very good ‘wall-plug’ efficiency. Depending on geometric and physical properties, these sources can emit coherent [4–7] as well as incoherent [1–3, 9, 10] radiation. A few years ago, compact EUV lasers driven by electrical discharges were demonstrated [11]. 2. The use of capillary discharges for the production of coherent and incoherent EUV photon sources To produce significant fluxes of high-energy photons, the electrical discharge has to be confined in a very small active volume in order to reach a very high power density and a high plasma temperature. Two methods are mainly used to obtain the power density required for emitting plasma. The first one is based on the z-pinch effect and converts the electrical energy into EUV radiation. The second one uses a capillary tube. The latter technique can be achieved with or without gas to produce a very high temperature plasma. When used under vacuum operation conditions, the emitting plasma is created from the ablation of the capillary wall. This technique using recombination schemes in carbon plasma has already demonstrated a gain in the EUV spectral domain. But the lifetime of vacuum operating discharge is limited since the diameter of the capillary rapidly increases after long operation times and thus leads to a significant reduction in the input power density. In contrast, when the capillary is filled with different gases, various spectral emissions resulting from transitions between high lying multi-charged ion states could be obtained in the whole EUV spectral domain with significant output power from shot to shot up to kilohertz operation. In all cases, the current flowing through the capillary channel should be strong enough to produce high a electronic temperature ranging from a few tens of electron volts up to several hundreds of electron volts. Since the current is proportional to the square root of the capacitance which stores the energy and varies with the inverse of the square root of the inductance, one has, to reach a current of tens of kiloamperes, to work with pulse forming line as fast as possible, i.e. of lowest inductance. Usually, capillary discharges work with small capacitances but charge to high voltage to store a few joules. Once the breakdown occurs inside a capillary channel, in our case of the capillary filled with low-pressure gas, the current begins to flow mostly along the axis whereas in the case of operation under sufficient vacuum condition, the current flows along the wall causing ablation of the capillary material. But in both cases, hot and dense plasma rapidly fills the whole volume of the channel. As the EUV light emitted by the plasma has to be collected, a more convenient means in this geometry is to use hollow electrodes. By doing so, one can easily view the plasma along the capillary channel. The geometry of cathode and anode is important for convenient flow of the charged particles, because hot spots can severely damage electrodes by reducing drastically the lifetime of the source. In fact, the whole set-up is, in principle, very close to a pseudo-spark discharge in which a dielectric capillary would have been added for better confinement of the plasma. The high degree of ionization of the plasma is obtained as a result of the very high power density injected into the gas. Tens of gigawatts per cubic-centimetre can easily be reached in such capillary discharge sources. Amplification of spontaneous emission (ASE) can be expected from such fast and energetic discharges. EUV ASE, the so-called soft x-ray laser, can occur following different pumping schemes: collisional [11], recombination [12] and charge exchange [13, 14] process. Using capillary discharge drivers, higher gain-length products have been achieved in the collisional scheme [6, 11]. According to this scheme, the electronic temperature has to be high enough to ensure population inversion between states of highly ionized atoms. In the recombination scheme, inversion of population is realized only when the electrons are cold enough to encounter recombination. According to this scheme, the proximity of the wall in small diameter capillaries can help fast cooling of the electronic population. Finally, charged exchange process involves ions of different ionization degrees in collision to lead to the required inversion of population. 3. The ablative capillary discharge in carbon plasma The configuration of the capillary discharge used in our experiment for probing the recombination scheme is shown in figure 2. We used a classic RLC circuit as the electrical driver [10]. A high voltage pulse triggered the discharge. The energy stored ranged from 40 to 80 J. Two different materials were used for the capillary of 1 mm in diameter: a polyethylene one and a polyacetal one. Only the polyethylene capillary can be used at very high power density up to 70 GW cm−3 since Faraday cage Optical path of the beam Capillary polyimide filter Pinhole Ø=50 µm Detection part Instrumentation box – 40 kV Figure 2. Experimental set-up. A65 C Cachoncinlle et al the polyacetal capillary breaks more easily. The capillary was evacuated using a turbo molecular pumping unit. A Rogowski coil is used to monitor the current. An EUV fast photodiode was used to record time evolution of the photon pulse. Using appropriate spectral filters, a reliable estimation of the pulse energy can be derived from the photodiode signal. End-on images of plasma were obtained by putting a 50 µm diameter pinhole between the source and the image plane. A flat field spectrometer (Jobin & Yvon PGMPG500) coupled to a gated micro-channel plate connected to an intensified CCD camera was employed to record time resolved (10 ns) spectral measurements of the carbon plasma from the water window region (3 nm) up to 40 nm. Three gratings having different spectral resolutions were used. Time-resolved pinhole imaging [15] of the plasma reveals that hotter plasma, which is highly emissive in the soft x-ray spectral domain, is concentrated on the axis of the capillary channel. But no magnetic compression had been observed. The visible light, corresponding to colder plasma is, as expected, localized in an annular sheet at the periphery, close to the capillary wall. Time-resolved spectroscopy of the carbon plasma shows a strong emission of the Balmer-Hα (18.22 nm) and Hβ (13.46 nm) lines in the initial stages of the discharge before the current reaches the maximum. Then carbon lines from lower ionization degrees dominate the spectrum. The electronic density has been estimated by Stark broadening measurement on a CIV atomic line at 580.29 nm. The values obtained in the plasma column range from 2 × 1018 to 1 × 1019 cm−3 . Estimation of the time evolution of electronic temperature was performed by measuring the ratio of the intensity of atomic lines in the spectral domain 17–19 nm. The temperature reaches a maximum value of 80 eV. Time integrated measurement of gain has been performed on different lines. As illustrated in figure 3, the spectral lines of CV at 16.7 nm increases linearly with the length of the plasma. Only the two CVI lines Hα (18.22 nm) and Hβ (13.46 nm) exhibit a non-linear behaviour with respect to the capillary channel length. If we interpret this behaviour as an indication of ASE, the light intensity, I (l), should increase following the classical law I (l) = S(exp(gl) − 1) (1) where l is the length of the capillary, g the gain value, and S the source function. Numerical optimization of the experimental data sets according to equation (1) gave a gain-length product of 1.62 and 3.02 for the Hα and Hβ, respectively Interpretation in terms of ASE is strongly supported by experimental data [14] and plasma modelling results. Two different numerical codes have been used to simulate the time evolution of the population of different energy levels. Both codes, FLY and CADILAC, predicted inversion of population of the H-like level of carbon. According to the results of these codes the inversion of population occurred only 100 ns after the current began to increase. This value agrees completely with our experimental observations. 4. The CAPELLA source Under a French national project PREUVE that focuses on the realization of a EUV lithography at 13.4 nm, we have developed an EUV source CAPELLA (Capillary EUV Lamp for Lithography Approach). CAPELLA is based on gas-filled capillary pulsed electric drivers, i.e. a fast high voltage pulsed forming line has been used to produce the gas breakdown instead of direct capacitive discharge. A high voltage pulse (15 kV) of tens of nanoseconds was applied to water-cooled electrodes. The diagnostics surrounding the CAPELLA were the same as those mentioned in the preceding section for the ablative capillary discharge. Most of the investigations have been carried out on a small capillary channel of 10 mm length and 1 mm diameter. Such a geometry is suitable for incoherent sources that are generally required for NGL applications. Such incoherent photon sources are obviously simpler to operate than the coherent ones. Flowing gas to fill the capillary at pressures from 0.2 to 2 mbar assures sufficient density for the production of a high flux of energetic EUV 2250 5000 2025 4500 a 1800 Intensity (a.u.) Intensity (a.u.) 4000 3500 3000 2500 c 2000 1500 b 1350 1125 900 675 1000 450 500 225 0 0.00 0.18 0.36 0.54 0.72 0.90 1.08 1.26 1.44 1.62 1.80 Capillary length (cm) d 1575 e 0 0.00 0.18 0.36 0.54 0.72 0.90 1.08 1.26 1.44 1.62 1.80 Capillary length (cm) Figure 3. The spectral intensity from carbon plasma versus capillary length. (a) Hα line: exponential fit; (b) Hα line: linear fit; (c) CV 16.7 nm line: linear fit; (d) Hβ line: exponential fit; (e) Hβ line: linear fit. A66 Capillary sources of UV radiation photons. Time resolved spectroscopic measurements were carried out for different gases (nitrogen, oxygen, neon, argon, krypton and xenon) and a mixture of gases with helium. In the EUV wavelength spectral domain, the brighter emissions were obtained in krypton (8–11 nm) and xenon (10–15 nm) [16]. The xenon filled capillary is recognized as the best candidate for sources dedicated to NGL applications because of the presence of a wide fluorescence around 13.5 nm that matches the bandwidth of the classical Mo–Si mirror. This mirror is known to reach 0.69 in reflectivity at 13.4 nm in a 2% of maximum bandwidth. From time-resolved pinhole images of the xenon plasma, there clearly appears to be a z-pinch effect at the time of the high level of EUV radiation production. The observed characteristic timescale of the colliding plasma is in complete agreement with our calculation. We first used the very simple and well-known ‘snowploughs’ model, which only needs three parameters: the current, the pressure and the radius of the capillary channel. It is predicted that the maximum of compression of the plasma should occur between 40 and 60 ns in our experimental conditions. A more sophisticated code, based on the conservation of energy, has been developed [17] to gain more physical understanding of the phenomenon, since the compression of the plasma is consequently stopped by the increased pressure. This code predicts a maximal compression at round 45 ns followed by another compression (85 ns) since the time of current flow through the capillary is longer than the characteristic time of the z-pinch effect. The luminous spike observed in figure 4 corresponds to the first maximum compression of the plasma column. It is probably shorter than 5 ns which is the limit of the time resolution of the gated detector. It should be pointed out that, in contrast to other similar observations [18], this ultra-fast luminous spike observed on the photodiode signal should not be attributed to a laser effect since the spectroscopic measurements, run in optimal experimental conditions, did not show any specific increase in intensity of one particular spectral line in the whole spectral domain. Furthermore, the temporal behaviour of the resistance of the plasma column, which should vary as the square of the radius of the plasma, shows that the power transferred to Xenon : 1 torr Diam : 1.2 mm Length : 10 mm 5000 4 Current Intensity [a.u.] 3 4000 Diameter Intensity 3000 2 2000 1 1000 0 0 20 40 60 80 100 120 140 160 0 180 Time [ns] Figure 4. Discharge current waveform together with the time evolution of the plasma diameter and intensity of radiation as observed from pinhole image. Plasma diameter [a.u.] 6000 plasma (in joules) presents the same fast picking behaviour. It is obvious that the EUV light emitted by this plasma is simply proportional to the electrical power injected at each time. Different experimental parameters were varied to optimize the wall-plug efficiency of the xenon source at 13.4 nm. The pressure, measured at the inlet of the capillary, is found to be optimal at 2 mbar, but there exists a strong pressure gradient along the capillary channel due to its very small radius. The xenon flow was controlled by a mass flowmeter. A turbomolecular pumping unit evacuated the capillary. A fast photodiode evaluated the energy radiated. The light was filtered through a EUV zirconium foil 1600 Å thick. From the spectra recorded by the spectrometer, we estimated that no more than 16% of the total energy measured by the photodiode is in the bandwidth of a classical Mo–Si mirror (2% bandwidth at 13.4 nm). Under these experimental conditions the source emits about 2 mJ ‘in band’ per joule stored per shot and in full solid angle. Thus, at the optimum of input power density, the wall-plug efficiency of conversion into ‘in band’ EUV energy is 0.2%. Such high wall-plug efficiency is one of the highest reported in the literature for the discharge techniques [3] and is of course much higher than the wall-plug efficiency of laser plasma EUV sources [1]. The source has already been tested at low repetition rate of 50 Hz leading to an average power of 0.1 W in the bandwidth of the mirror. The shot to shot fluctuation in EUV radiation from this source is less than 4%. From these experiments we can expect to have sources of a few watts at very low cost by using a proper switching element and a small size cooling device. 5. Conclusion From these studies, it appears that the capillary discharges are good candidates for coherent and incoherent sources in the soft x-ray spectral domain. They show exceptionally high efficiency at a very low cost. We have already developed and studied different types of capillary sources either dedicated to the so-called ‘soft-x-ray laser’ or to EUV metrology. In the radiation source based on the ablative capillary discharge we are able to observe a gain in carbon plasma on the Balmer-Hα (18.22 nm) and Hβ (13.46 nm) spectral lines. By injecting 70 GW cm−3 power inside a polyethylene capillary of 1 mm diameter, we measured the indication of ASE with small gain-length products of 1.62 and 3.02 for Hα and Hβ, respectively. The temperature inside the channel was estimated to be 80 eV from spectroscopic measurements and the electron density was of the order of 1019 cm−3 . Optimization of the EUV capillary sources filled with xenon led to the development of an EUV lamp (CAPELLA) dedicated to the metrology for NGL. The lamp emits about 2 mJ ‘in band’ per joule stored per shot and in full solid angle. The wall-plug efficiency of conversion into ‘in band’ EUV energy is 0.2%. The shot to shot fluctuation of this lamp is less than 4%. The low flux (0.1 W) lamp is operated at repetition rate of 50 Hz. Upgrading of this lamp towards higher fluxes is under progress. In the near future we would expect a low cost, high flux (tens of watt) source with high repetition rate (1 kHz) with improved wall-plug efficiency. A67 C Cachoncinlle et al Acknowledgments This work was partially financially supported by the TMR network contract No ERBFMRXCT 980186 and the ‘minist`ere de l’industrie’ within the French Project PREUVE. References [1] Silfvast W 1999 IEEE J. Quant. Electron. 35 700 [2] Hansson B, Rymell L, Berglum M and Hertz H 2000 Microelectr. Eng. 53 667 [3] Lebert R, Bergmann K, Schriever G and Neff W 1999 Micro and Nano Engineering 46 465 [4] Shin H, Kim D and Lee T 1994 Phys. Rev. E 50 1376–82 [5] Liu Y, Meminaro M, Tomasel F G, Chang C, Rocca J J and Attwood D T 2001 Phys. Rev. A 63 3802 [6] Frati M, Seminario M and Rocca J J 2000 Opt. Lett. 25 1022 [7] Tomasel F G, Rocca J J, Shlyaptsev V N and Macchietto C D 1997 Phys. Rev. A 55 1437 [8] Ellwi S, Juschkin L, Ferri S, Kunze H-J, Koshelev K and Louis E 2001 J. Phys. D: Appl. Phys. 34 336 A68 [9] Fiederowicz H, Bartnik A, Daido H, Choi I, Suzuki M and Yamagami S 2000 Opt. Commun. 184 161 [10] Hong D, Dussart R, Cachoncinlle C, Rosenfeld W, G¨otze S, Pons J, Viladrosa R, Fleurier C and Pouvesle J-M 2000 Rev. Sci. Instrum. 71 19 [11] Rocca J, Shlyaptsev V, Tomasel F, Cort´azar O, Hartshorn D and Chilla J 1994 Phys. Rev. Lett. 73 2192 [12] Wagner T, Eberl E, Frank K, Hartmann W, Hoffmann D and Tkotz R 1996 Phys. Rev. Lett. 76 3124 [13] Kunze H-J, Koshelev K, Steden C, Uskov D and Wiesschebrink H 1994 Phys. Lett. A 193 183 [14] Ellwi S S, Andreic Z, Pleslic S and Kunze H J 2001 Phys. Lett. A 292 125 [15] Dussart R, Hong D, G¨otze S, Rosenfled W, Pons J, Viladrosa R, Cachoncinlle C, Fleurier C and Pouvesle J-M 2000 J. Phys. D 33 1837 [16] Robert E, Blagojevic B, Dussart R, Mohanty S, Idrissi M, Hong D, Viladrosa R, Pouvesle J-M Fleurier C and Cachoncinlle C 2001 (Santa-Clara: SPIE) [17] Miyamoto T 1984 Nucl. Fusion 24 337 [18] Niimi G, Hayashi Y, Nakajima M, Watanabe M, Okino A, Horioka K and Hotta E 2001 J. Phys. D: Appl. Phys. 34 1