slides of viscosity talk in Nancy
Transcription
slides of viscosity talk in Nancy
Viscosity and scale invariance in the unitary Fermi gas Searching for the perfect fluid Tilman Enss (TU München) Nancy, 7 February 2011 [Annals of Physics 326, 770 (2011); arXiv:1008.0007] Outline • shear viscosity and perfect fluidity • transport equations for the unitary Fermi gas • analytical and numerical results, universal high-frequency behavior • scale invariance and universal dynamics Shear viscosity: definition [Schaefer, Teaney, Rep. Progr. Phys. 2009] Measures of Perfection Measures of Perfection Measures of Perfection Measures of Perfection • Viscosity determines shear stress, or friction, of fluid flow: determines shearinstress (“friction”) in fluid flow y determinesViscosity shear stress (“friction”) fluid flow Viscosity determines shear (“friction”) stress (“friction”) fluid flow Viscosity determines shear stress in fluidinflow ∂vx F =Aη ∂vx ∂v x ∂y F = AFη = A η ∂y ∂y Dimensionless measure of shear stress: Reynolds number onless measure of shear stress: Reynolds number Dimensionless measure shear stress: Reynolds number • dimensionless measure ofofshear Reynolds number Dimensionless measure shearofstress: stress: Reynolds number n n Re = ×n mvr Re = × mvr n η × mvr η Re =Re = × mvr η η flow fluid flow fluid property fluid fluidproperty flow flow property property property property property property • [η/n] = ! • [η/n] = ! • [η/n]• s=[η/n] ! =! s • Relativistic Re = × τ T s • relativistic systems: T • Relativistic systems Re = × τsystems sRe = × τ T η • Relativistic systems η • Relativistic systems Re = × τ Tη η ∂vx F =Aη ∂y Shear viscosity: estimates • kinetic theory (Boltzmann equation) for dilute gas: η= 1 3 n p̄ �mfp , �mfp 1 = : nσ η� √ mkB T σ(T ) grows with T • superfluid: ηSF = 0 but phonon contribution [Landau, Khalatnikov 1949] • minimum in between: at which temperature and viscosity? η ∼ p̄ �mfp ≥ � • uncertainty suggests n (careful!) Insights from string theory • conformal field theory (CFT) dual to AdS5 black hole: shear viscosity graviton absorption cross section (~ area of event horizon) CFT entropy Hawking-Bekenstein entropy (~ area of event horizon) • specifically SU(N), N = 4 SYM theory (no confinement, no running cpl) 2 λ = g N is dual to classical gravity: in strong-coupling ‘t Hooft limit η � ≥ s 4πkB [Policastro, Son, Starinets 2001; Kovtun, Son, Starinets 2005] • conjecture of universal lower bound Perfect fluidity [Schaefer, Teaney, Rep. Progr. Phys. 2009] • Definition: “perfect fluid” saturates bound η � = s 4πkB How to qualify? • bound is quantum mechanical ➡ need quantum-mechanical scattering mechanism • bound is incompatible with weak coupling and kinetic theory: �/τ � �qp � kB T , s ∼ nkB : η �qp τ � ∼ � s kB kB ➡ need strong interactions, no good quasi-particles (above Quantum Critical Point: �/τ = CkB T [Sachdev]) 200 4π η hs Helium 0.1MPa Nitrogen 10MPa Water 100MPa 150 100 50 Viscosity bound 0 1 10 T, K 100 1000 viscosity helium, FIG. 2 (color online). for The viscosity-entropy ratio for some [Kovtun, Son, Starinets 2005] common substances: nitrogen and water. The ratio is nitrogen,helium, and water always substantially larger than its value in theories with gravity Perfect Fluids: The co Perfect fluids: the contenders [Schaefer, Teaney 2009] Perfect Fluids: The contenders ct Fluids: The contenders Liquid Helium Liquid Helium QGP (T=180 MeV) Liquid Helium (T=0.1 meV) (T=0.1 meV) Quark-Gluon PlasmaMeV) QGP (T=180 (T=180 MeV) (T=0.1 meV) −6 η = 1.7 · 10 Pa · s T=180 MeV) Trapped Atoms (T=0.1 neV) Trapped Atoms oms (T=0.1 neV) V) η = 1.7 · 10−15 Pa · s η = 5 · 1011 Pa · s Trapped Atoms Consider ratios η/s (T=0.1 neV) min=0.5; 0.8; 0.4 The unitary Fermi gas • two-component Fermi gas with zero-range interactions: LE = � σ=↑,↓ ψσ† � � � g(Λ) † † 2 �∂τ − ∇ ψσ + ψσ ψ−σ ψ−σ ψσ 2m 2 2 2 • renormalized coupling g(Λ) �→ g = 4π� a/m −1 r � k • ultracold atoms: interaction range 0 F � a scattering length • Hubbard-Stratonovich transformation with pair field ϕ � � 1 ∗ L[ψ] �→ L[ψ, φ] = L0 + ψ↑ ψ↓ φ + h.c. − φ φ g ∗ • ϕ is massless at infinite coupling g=∞ (unitarity) roundness of the optical trap and its alignment with both Evaporative cooling (achieved by reducing the laser power) stirrer and the axes of the magnetic potential were cr accompanied by a magnetic field ramp to 766 G on the BEC-side 6 3 deviation from cylindrical symmetry owing to misalignme of the resonance typically produced a BEC of 3 £ 10 molecules . aberrations, or gravity rapidly damped the rotation. The Previous experiments studying the rotation of atomic BECs 27–31 of vortices in sodium was comparatively forgiving, and . Because employed magnetic traps operating at low bias fields optimized before vortices in 6Li2 could be observed. the Feshbach resonance in our system occurs between two high-field seeking states that cannot be trapped magnetically, an optical dipole Observation of vortex lattices trap operating at high magnetic bias fields was necessary. Our set-up 6 2 In experiments with Li close to the Feshbach reson radius of 123 m m (wavelength employed a trapping beam with a 1/e Bloch, Dalibard, and Zwerger: Many-bo BCS-BEC crossover interaction strength between atoms in states j1l and j2l ca 1,064 nm), radially confining the gas with a trap frequency of 59 Hz at via the magnetic field. Thus, it is possible to choos a power of 145 mW. Axial confinement withFUNCTIONS trap frequency SPECTRAL ANDtuned RF RESPONSE OF… magnetic fields to optimize the three steps involved in the n z ¼ 23 Hz was provided by an applied magnetic field curvature fermions in tw a vortex lattice: stirring of the cloud (for 800 ms at a typi that decreased the radial trap frequency to n r ¼ 57 Hz. The aspect states: s-wave frequency of 45 Hz), the subsequent equilibration (typica ratio of the trap was 2.5. In this trap, at a field of 766 G, condensates of 6 0.25in our experiment after and time-of-flight expansion for imaging. 1 £ 10 molecules (the typical number belowTo Tcstay cl rotating c F the cloud) have Thomas–Fermi radii of about 45 m12m radially and 110 mm axially, a peak molecular density of 2.6 £ 10 cm23, a 0.2 chemical potential of about 200 nK, and a characteristic microscopic length scale of 1/k F < 0.3 mm. Here, the Fermi wavevector k F is BCS a non-interacting two-state defined by the Fermi energy (E F) of 0.15 mixture of 6Li atoms of mass m with total atom number N in a 1=3 2 2 mean frequency q; ; harmonic trap of (geometric) ! ! EF ¼ "qð3NÞ dE h̄ dE h̄ dE h̄ BEC 0.1 = − C, (2) = − C, (2) = − this Article we will estimate the interaction "2 k2F =2m: Throughout −1 −1 1 4πm 4π da− the average number of fermionda pairs parameterda1/k s s Fa using4πm s N/2 ¼ 1 £ 106. Here, a is the scattering 0.05length between atoms in a ¼the 1.3. Because this states j1l and At same a field ofquantity 766 G,where 1/k Fthat contact is isthe same quantity that appears where the contact C isj2l: the appears inCgas where the contactinC is the sam strongly interacting, it is difficult to extract a temperature from the the asymptotic distribution. The contactmomentum is the asymptotic momentum distribution. The contactmomentum is the asymptotic di spatial profile. For weaker interactions 0 (at 735 G) the condensate -2 0 2 4 the integral over the system of the contact density,over which the integralfraction over the thewhich contact density, which the integral the system of was system in excess ofof80%, would isentropically connect to 1/k a 32 F T/Tpairs ¼ 0.07. The an FIG. ideal Fermi gas isofall proportional tocrossover the number of pairsisof atoms with to difis proportional to in the number of atoms withare difproportional the number o F (a) Inwith the BEC all the atoms are pairs with difms FIG. are in1:pairs dif- limit, 1: (a) In theatBEC limit, the BEC–BCS atoms in pairs with dif, 1)close occurs together. in the region between 780 G andthat 925 G. (1=k F jajare ferent spins are together close Figure together. Tan’s results reveal that Tan’s results reveal ferent spins close toge (blueinto and redferent balls) spins that are bound together into 3 | Optimized vortex latticesthat in the are BEC–BCS crossover are ferent boundspins together ferent spins (blue and red balls) that are bound into The trapped cloud was rotated about its long axis using a blue[Haussmann et al. 2007] vortex lattice wasthat created at 812 G, expect. the field is wasmuch ramped in 100 FIG. 28.(grey !Color online" Critical temperature the homoge28,29,33 this number isa much larger than oneof might diatomic molecules (greythat circles) form a much Bose-Einstein this that number is larger than one might expect. this number larg at form a Bose-Einstein diatomic molecules circles) form Bose-Einstein . that A two-axis acoustodetuned laser beam (wavelength 532that nm) (BEC-side), 833 G (resonance) and 853 G (BCS-side), where th 4/3 condensate. In some diatomic molecules, the pairs are molecules, the pairs are of the condensate. In some of the diatomic molecules, the pairs are optic deflector generated agas two-beam pattern (beam neous asThe alike function coupling strength. The full line is number inofseparation athe small volume scales like The number in a small volume scales (volume) The number in a4/3small volumfi held for 50 ms. After 2 ms of(volume) ballistic expansion, the magnetic m, gaussian waist instead w ¼ 16 mboxes). m) that was rotated d ¼very 60 mThe 2 as close together (b) unitary limit 2 as beam s). very (b) The unitary limit(highlighted close together (highlighted (b) The unitary limit ramped tomight 735 Ginstead forexpect. imagingof (see text for details). The2 field of ofIn like (volume) one In agives later insteadboxes). of like (volume) one might expect. a later like (volume) as on the result obtained by Haussmann et al. !2007" and T c Q. around the cloud a variable angular frequency image is intermediate between the BEC andsymmetrically BCS conditions. (c) In atthe BCS conditions. (c) In is intermediate between BEC and BCS conditions. (c) isIn880 mm £ 880 mm. The unitary Fermi gas Ultracold atom gases θc = Tc/TF = • unitary Fermi gas is superfluid Physics 2,normal Physics 2, 9 (2009) 9 (2009) pr below T ≈ 0.15 T ne al th et ca pa superfluid di tr Pe • single-particle properties: a spectral function A(k,ε) at Tc tw from self-consistent T-matrix tio derived a virial theorem paper, (that is,Tan a relationpaper, Tan derived2009] a virial theorem (that is,Tan a relationderived a virial the [Haussmann, Punk, Zwerger = 0.16T atpaper, unitarity. The form exact asymptotic results Eqs. !136" 1048 the BCS limit, tiny fraction of the atoms formlimit, very at atoms form very alarge the BCS alarge tinyF fraction of the atoms very large between kinetic and potential energy) for atoms ship between kinetic and potential ship energy) for atoms between kinetic and pot ©!!""#!Nature Publishing Group! ship Cooper pairs (grey circles). A small fraction of the atoms are raction of the atoms are Cooper pairs (grey circles). A small fraction of thelimits atoms are and !11" inand the BCS and BEC are[4]indicated by triangles trapped in a harmonic potential and it also involves trapped in a harmonic potential [4] it also involves trapped in a harmonic potenti close together (Illustration: Alan(highlighted boxes). (Illustration: Alan es).very (Illustration: Alan(highlighted boxes).very close together !green" andthe squares contact.!blue", respectively. The the contact. Stonebraker/stonebrakerdesignworks.com) thedashed contact. line gives an Stonebraker/stonebrakerdesignworks.com) s.com) Tan’s firstignored twoof papers were almost completely ignored Tan’s first two papers were almost completely Tan’s first two2007. papers were a • now two-particle properties: the schematic evolution T*. From Haussmann et al., ie for several until 2008and were n for several years and were not published untilyears 2008and were not published for several years transport! [2, 3]. One contributing factor to the th [2, delayed 3]. One apprecontributing factor to the delayed [2, 3]. One apprecontributing fact smooth unitary limit as as → ±∞. They derive the rate They derive the rate smooth unitary limit as as → ±∞. They derive the rate ciation for work was that he derived his Tan’s univerciation for Tan’s work was that he derived his Tan’s univerciation for work was tia th of change of the total with respect to!130". article with respect to energy per particle of change of the total energy per particle with respect to coupling Eq. Its dimensionless constant sal resultsequation from the many-body Schrödinger sal results from the many-body Schrödinger sal resultsequation from the many-bo 1/3 length while n and the entropy are the inverse scattering the entropy are n and the entropy are length while n and the inverse scattering pa / 2" ageneralized = 0.16k much smaller 1 in novel the regime using novel involvingthan generalized functions by using novel methods !n involving functions by using methods involv ddby Fa ismethods Baym-Kadanoff (2PI) conserving approximation 7 !"# at unitarity, (0) Tσ" (τ X1 X1! ) = h̄ 2 ∇1 ∇!1 (5.8) δX1 X1! δ(τ − τ1 ) , m ! 2ḡ(Λ)−1 δX1 X1! δ(τ − τ1 ) for # = 0 7 (0) ! S" (τ X1 X1 ) = 0 for # = 2 . (5.9) • self-consistent equations for fermionic and bosonic self-energies: !"# (5.8) =0 = 2. (5.9) externctions s func- (5.10) (5.11) nd line vertex The response of the grand potential to the external field in terms of the fermionic Green’s functions Gσ,XX ! = #T ψσ (X)ψσ† (X ! )$ and bosonic Green’s func† tions GB,XX ! = #T ψB (X)ψB (X ! )$ is Σ↑!$# (K) = � Γ(Q)G↓ (Q + K) δΩ = tr[(δUσ )Gσ ] + tr[(δUB )GB ] = Q (0) tr[Tσ" Gσ δh" (τ )] − (0) tr[S" Γ (5.10) δh" (τ ) ] (5.11) (1 + 2h" (τ ))2 Σ↑ (X) = Γ(X)G↓ (−X) where the trace includes the spin sum. In the second line the Green’s is replaced by the vertex FIG.bosonic 1: [color online]function Diagrammatic contributions to the 2 ! ! function = −g̃ self-energies. GB,XX . Hence, we obtain fermionicΓand XX bosonic • Dyson equation: " (0) δΩ ! !X = − Tσ" (τ−1 XX ) GS̃ − σ,X −1 followδhfrom scale invariance. T̃ and the fermionic " " are " (τ ) ! σXX response ↑ and bosonic 0 functions ↑ viscosity " 1 (0) S (τ XX ! ) ΓX ! X . (5.12) + ! " δG 2 σ,XX ! T̃σ" (τ XX(1) + = 2h" (τ )) XX = !#T Π" (τ )ψσ (X)ψσ† (X ! )$ , δh" (τ ) In particular for a static scaling perturbation h"=0 (τ(5.16) )≡h ! δΓXX † correct we S̃recover energy formula 2 ! !the Tan = −g̃ #T Π" (τ[27] )ψBwith (X)ψthe " (τ XX ) = B (X )$ . δh" (τ ) UV regularization, (5.17) 2 2# $ " C dΩ h̄ k nkσ − 4 = 2#H$(0)= 2ε . 3p = #Πii $ = − =2 Note that for the shear viscosity dhcase of the2m k where S" = kσ 0, the viscosity response function can be expressed by (5.13) (0) the L function [47] as T̃" = LT" , and the correlation function reduces well-known form (τ )re= The Kubo(5.15) formula (3.1)tointhe imaginary time canχ"be G (K) = G (K) − Σ (K) !$# Σ (Q) = � G (K)G (Q − K) FIG. 1: [color b online] Diagrammatic ↑ contributions ↓ to the fermionic and bosonic self-energies. K Σ (X) = G (X)G (X) follow from scale invariance. T̃" and S̃" are the fermionic b viscosity response ↑ functions↓ and bosonic δGσ,XX ! = #T Π" (τ )ψσ (X)ψσ† (X ! )$ , δh" (τ ) (5.16) δΓXX ! −1 −1 † ! = −g̃ 2 #T Π" (τ )ψB (X)ψB (X S̃" (τ XX ! ) = b )$ . δh" (τ ) (5.17) T̃σ" (τ XX ! ) = Γ (Q) = a − Σ (Q) • solve by iteration on logarithmic grid (300 & 300 momenta) Note that for thefrequencies case of the shear viscosity where Sradial = 0, the viscosity response function can be expressed by (0) the L function [47] as T̃" = LT" , and the correlation function (5.15) reduces to the well-known form χ" (τ ) = (0) (0) − tr[Tσ" (τ )LTσ" (0)]. Up to now the derivation of the correlation function (5.15) has been completely general and exact, but the challenge remains to evaluate the viscosity response functions T̃" and S̃" within the microscopic model. In the following we shall do this within the T-matrix approxi- • conserving: number, momentum current, scale invariance, ... fulfills Tan energy formula & adiabatic relation exactly (τ ) ≡ h (5.12) correct (0) (0) (0) " Pair spectral function Viscosity in linear response: Kubo formula • viscosity from stress correlations (cf. hydrodynamics): 1 η(ω) = ω � d3 x dt eiωt θ(t) with stress tensor Π̂xy = � p x py p,σ m �� Π̂xy (�x, t), Π̂xy (0, 0) † cpσ cpσ ∂vx (cf. Newton ) ∂y • or, from transverse current correlations [Hohenberg, Martin 1965] ηq 2 χ⊥ (ω, q) = 2 , 2 2 ω + (D⊥ q ) • equivalence via momentum balance ∂t (ρvi ) + ∂j Πij = 0 �� η = D ⊥ ρn Transport equations 7 • Single-particle Green functions: 7 (5.8) (5.8) =0 = = 02 . = (5.9) 2. (5.9) exterexternctions s funcnctions s func(5.10) (5.10) (5.11) (5.11) nd line vertex nd line vertex (5.12) (5.12) (τ ) ≡ h correct τ) ≡ h correct [Enss, Haussmann, Zwerger 2010] !"#Response to shear perturbations: !"# !"# !"# correc correc T!MT T!M !$# !$# !$# Maki-Thompson Aslamazov-Larkin !$# T!A T!A The a The a FIG. 1: [color online] Diagrammatic contributions to the fermionic and bosonic FIG. 1: [color online]self-energies. Diagrammatic contributions to the fermionic and bosonic self-energies. !"# !"# follow from scale invariance. T̃ and S̃(Kubo are the fermionic • follow correlation function formula): η(ω) = and bosonic viscosity responseT̃ functions from scale invariance. and S̃ are the fermionic " " " " and bosonic viscosity response functions δGσ,XX ! T̃σ" (τ XX ! ) = = #T Π" (τ )ψσ (X)ψσ† (X ! )$ , δG δhσ,XX " (τ ) ! T̃σ" (τ XX ! ) = = #T Π" (τ )ψσ (X)ψσ† (X ! )$ , (5.16) δh" (τ ) (5.16) δΓXX ! † = −g̃ 2 #T Π" (τ )ψB (X)ψB (X ! )$ . S̃" (τ XX ! ) = δh (τ )! δΓ"XX † ! = −g̃ 2 #T Π" (τ )ψB (X)ψB (X(5.17) )$ . S̃" (τ XX ! ) = δh" (τ ) (5.17) (0) Note that for the case of the shear viscosity where S" = (0)by 0, thethat viscosity function can be expressed Note for theresponse case of the shear viscosity where S = " (0) the L function [47] as T̃ = LT , and the correlation 0, the viscosity response" function can be expressed by " (0) function (5.15) reduces theLT well-known form χ" (τ ) = the L (0) function [47] as T̃to" = " , and the correlation (0) FIG.modes: 2:FIG. [color online]online] Diagrammatic contributions 2: [color Diagrammatic contributionstotothe the • transport via fermionic and bosonic renormalized viscosity response functions T! and . (a)The The renormalized viscosity response functions T! and S! .S!(a) very efficient description, satisfiesfullconservation laws full fermionic response function T! has Maki-Thompsonand and fermionic response function T! has Maki-Thompson • assumes no quasiparticles: Aslamazov-Larkin contributions, in addition to the bareververAslamazov-Larkin contributions, in addition to the bare (0) (0) T! . (b) The full bosonic response function S! has two tex T! tex . (b) The full bosonic response function S! has two contributions which are equivalent for equal fermion masses contributions which are equivalent for equal fermion (0) masses beyond Boltzmann and populations, in addition to the bare vertex(0) S! . (c) The and populations, in addition to theisbare S! . (c) The viscosity correlation function givenvertex by a skeleton diagram viscositywith correlation function is given byviscosity a skeleton diagram one bare and one renormalized response vertex S! ( S! ( isis give give (5.20) (5.20) SS!!(τ The b The the pa pa the bare b b bare UV di UV di which which The The 100 T=10 T= 5 T= 2 T= 1 T=0.5 T=0.151 1 < d(t) [h n] 10 0.1 ∼ W τη−1 ω2 + −2 τη Cη ∼√ mω 0.01 0.001 0.1 1 10 100 1000 < t [EF / h] Viscosity spectral function [Enss, Haussmann, Zwerger 2010] Transport peak [Enss, Haussmann, Zwerger 2010] • fit to Drude peak and tail: ω2 + −2 τη pressure p=2¡/3 peak weight W 10 1 0.1 0.1 1 T [TF] �3/2 Cη ω(ω + τ −1 ) + √ mω ω 2 + τ −2 visc. scattering time od [<h / EF] peak weight W [EF n] η(ω) = W τη−1 10 10 classical 2.77 T1/2 scattering time od 1 0.1 • near Tc: �/τη ∼ 0.5 EF ∼ 3 kB T decay rate larger than energy, no good quasiparticles! 1 T [TF] 10 Contact coefficient • generically, short-distance (UV) behavior depends on non-universal details of interaction potential −1 kF ) • for zero-range interaction ( r0 � this becomes universal: at most two particles within distance r0 , all others far away (medium) • two-particle density matrix for r0 < r � � −1 kF : many-body few-body � �2 � r † r r r 1 1 † 3 d R ψ↑ (R + )ψ↓ (R − )ψ↓ (R − )ψ↑ (R + ) = C − 2 2 2 2 r a � • Tan contact C: probability of finding up and down close together (property of strongly coupled medium) [Tan 2005; Braaten, Platter 2008; Zhang, Leggett 2009; Combescot et al. 2009; Haussmann et al. 2009; Schneider, Randeria 2010; Son, Thompson 2010; Werner, Castin; Braaten 2010...] formula. Subject Areas: Atomic and Molecular Physics A Viewpoint on: Verification of Universal Relations in a Strongly Interacting Fermi Gas J. T. Stewart, J. P. Gaebler, T. E. Drake and D. S. Jin Phys. Rev. Lett. 104, 235301 (2010) – Published June 7, 2010 Contact coefficient (2) Some of the most vexing present-day problems in physics center on understanding the many-body properties and phases of strongly interacting fermions. Part of the difficulty arises from the fact that while the behavior of interacting fermions is understood in certain welldefined regions of parameter space, this understanding p often breaks down4 in the correlated strongly interactp→∞ ing regime. Achieving this understanding of strongly interacting fermions would have wide-ranging implications: the systems of interest include high-temperature superconductors, high-density quark matter, and ultracold atomic fermion gases. The latter is the subject of a et al.John 2010] recent experimental investigation[Stewart by John Stewart, Gaebler, Tara Drake, and Debbie Jin at JILA in the US, appearing in Physical Review Letters[1]. Stewart et al. succeeded in confirming the validity of certain nonperturbative theoretical formulas, known as the Tan relaFIG. 1: Schematic plot of the typical momentum distribution tions [2], which describe the properties of fermions with n(k ) of a fermion gas with short-range interactions, showshort-range interactions. The verification of these relaing the characteristic behavior of n(k ) approaching a constant tions shows that cold-fermion experiments are particvalue below unity at k → 0 and exhibiting a rapid variation ularly suited as a testing laboratory for studying the near the Fermi wave vector kF (that would be a sharp disgeneral problem of strongly interacting fermion systems continuity at T → 0 for a Fermi liquid). At asymptotically large k, as shown theoretically by Tan and confirmed experi[3, 4]. mentally by Stewart et al., one has n(k ) ∼ C/k4 (characterizThe JILA experiments studied gaseous, trapped ing the fermions as they undergo collisions, seen in the inset), potassium-40 atoms in two different internal states, with the contact C governing thermodynamic quantities like spin-up and spin-down (Fig. 1, inset). The effective inthe total energy of the gas. (Illustration: Alan Stonebraker) teractions between these states are experimentally adjustable via a Feshbach resonance—a resonance of a many-body system in which a bound state is achieved scattering length that is much larger than the interpartiunder certain conditions—that can be tuned with a magcle spacing (| a|3 n >> 1, with n the fermion density), denetic field. For this experiment, the detailed structure spite the fact that the effective range r0 of the interatomic of the Feshbach resonance scattering is less important • determine C: C lim n = p • intuitively: absorb external perturbation with large energy/momentum far away from coherent peak of a single particle ➡ need to hit 2 particles close together to give energy+momentum to both ➡ absorption rate ~C • access strong coupling and arbitrary temperature via perturbation theory 3 [Enss, Haussmann, Zwerger 2010] �3/2 C √ η(ω → ∞) = 15π mω • viscosity sum rule 2 π � ∞ 0 0.1 4 • analytical expression for tail Contact coefficient C [kF ] Viscosity tail Tan contact C visc. tail 15/Cd 0.01 0.1 �2 C dω [η(ω) − tail] = p − 4πma provides non-perturbative check [Enss, Haussmann, Zwerger 2010; cf. Taylor, Randeria 2010] 1 T [TF] 10 20 5 < d/s [h / kB] 10 2 1 phonon classical full Kubo ∼ T 3/2 classical limit phonon contribution ∼ T −8 0.5 Tc 0.2 0.1 0.2 0.5 1 2 T [TF] Shear viscosity/entropy of the unitary Fermi gas 5 10 20 [Enss, Haussmann, Zwerger 2010] Scaling Flows and Dissipation in the Dilute Fermi Gas at Unitarity 25 1.5 !η/s" 1.0 0.5 0.0 0.2 0.4 0.6 0.8 T /TF Fig. 11 Trap average !!s " = !" /s" extracted from the damping of the radial breathing mode. The data points were obtained using equ. (91) to analyze the data published by Kinast et al. [9]. The thermodynamic quantities (S/N) and E0 /EF were taken from [22]. The solid red and blue lines [fit: Schaefer, Chafin 2009; show the expected low and high temperature limits. Both exp: theory curves include relaxation time Kinast et al. 2004, Luo et al. 2009] effects. The blue dashed curve is a phenomenological two-component model explained in the text. shear viscosity for trapped atoms Scale invariance • Unitary Fermi gas: particle spacing only length scale, scale invariance � d3 x Π̂ii (x, t) = 2Ĥ = 2(T̂ + V̂ ) ➡ 3p = 2� [Ho 2004] � � � � � 1 d3 x dt eiωt θ(t) Π̂ii (x, t), Π̂jj (0, 0) ≡ 0 • bulk viscosity ζ(ω) = 9ω [Nishida, Son 2007] • pressure 3p̂ = • response of Green’s functions to scale transformation: Ward identities [Enss, Haussmann, Zwerger 2010] ωT [Πii ] = (� + µ)G(p, �) − (� + ω + µ)G(p, � + ω) ωTbos [Πii ] = (Ω + ω + 2µ)Γ(p, Ω) − (Ω + 2µ)Γ(p, Ω + ω) • WI provides solution of transport equation: conserving approximation Beyond viscosity: universal dynamics • viscosity determines several correlation functions at unitarity (small q): [cf. Taylor, Randeria 2010] mom. conserv. Im �[Πxy , Πxy ]�q,ω = ω η(ω) charge conserv. T Im �[Ji , Jk ]�q,ω L Im �[Ji , Jk ]�q,ω Im �[ρ, ρ]�q,ω (generally) q2 = η(ω) ω � at unitarity � 2 q 4 = η(ω) + ζ(ω) ω 3 q4 4 = 3 η(ω) ω 3 � qi qk qi qk L �[Ji , Jk ]� = 2 �[Ji , Jk ]� + δik − 2 q q � �[Ji , Jk ]�T (generally) (generally) (at unitarity) Conclusion and Outlook • are cold atoms (unitary Fermi gas) the perfect fluid? 20 ➡ most perfect real non-relativistic fluid known (factor ≈7 above string theory bound) d/s [<h / kB] 10 phonon classical full Kubo 5 2 1 0.5 ➡ strongest contender with quark-gluon plasma 0.2 0.1 0.2 0.5 1 2 T [TF] 5 10 20 • transport calculation of η(ω) beyond Boltzmann (tail, large scattering rate) • conserves scale invariance and fulfills sum rules 10 d(t) [<h n] 100 1 • universal dynamic properties of unitary Fermi gas T=10 T= 5 T= 2 T= 1 T=0.5 T=0.151 0.1 0.01 0.001 0.1 1 10 t [EF / <h] 100 1000