Ferimi Liquid123

Embed Size (px)

Citation preview

  • 8/18/2019 Ferimi Liquid123

    1/15

    CCMS Summer 2007 Lecture Series

    Fermi- and non-Fermi Liquids

    Lecture 3: Fermi-liquid Theory

    Dmitrii L. Maslov

    [email protected]

    (Dated: July 22, 2007)

    1

  • 8/18/2019 Ferimi Liquid123

    2/15

    Notation 1  Here and thereafter, L1 stands for Lecture 1: Basic Notions of the Many-body 

    Physics and L2 stands for Lecture 2: Examples of Interacting Fermi systems.

    I. GENERAL CONCEPTS

    All Fermi systems (metals, degenerate semiconductors, normal He3,  neutron stars, etc.)

    belong to the categories of either moderately or strongly interacting systems. For example,

    in metals   rs   in the range from 2 to 5. (There are only few exceptions of this rule; for

    example, bismuth, in which the large value of the background dielectric constant brings the

    value of  rs  to 0.3 and GaAs heterostructures in which the small value of the effective mass

    –0.07 of the bare mass–leads to the higher value of the Fermi energy and thus to  rs  <  1).

    On the other hand, as we learned from the section on Wigner crystallization, the critical  rs

    for Wigner crystallization is very high –about 100 in 3D and about 40 in 2D. Thus almost

    all Fermi systems occurring in Nature are too strongly interacting to be described by the

    weak-coupling theory (Coulomb gas) but too weakly interacting to solidify. Landau put

    forward a hypothesis that an interacting Fermi system is qualitatively similar to the Fermi

    gas [6] . Although original Landau’s formulation refers to a translationally invariant system

    of particles interacting via short-range forces, e.g., normal He3, later on his arguments were

    extended to metals (which have only discrete symmetries) and to charged particles.

    Experiment gives a strong justification to this hypothesis. The specific heat of almost

    all fermionic systems (in solids, one need to subtract off the lattice contribution to get the

    one from electrons) scales linearly with temperature: C  (T ) =   γ ∗T.   In a free Fermi gas,

    γ ∗ = γ  = (π2/3) ν F   = (1/3) mkF .  In a band model, when non-interacting electrons move in

    the presence of a periodic potential, one should use the appropriate value of the density of 

    states at the Fermi level for a given lattice structure. In reality, the coefficient  γ ∗ can differ

    significantly from the band value but the linearity of  C (T ) in  T   is well-preserved. In those

    cases when one can change number density continuously (for example, by applying pressure

    to normal He3), γ ∗ is found to vary. One is then tempted to assume that the interacting

    Fermi liquid is composed of some effective particles (quasi-particles)   that behave as free

    fermions albeit their masses and other characteristics are different from the non-interacting

    values.

    2

  • 8/18/2019 Ferimi Liquid123

    3/15

  • 8/18/2019 Ferimi Liquid123

    4/15

    Fermi liquid (or single band metal) can be written as

    δE  = 

      d3 p

    (2π)3εpδnp.   (1.1)

    (For the sake of simplicity, I also assume that the system is isotropic, i.e., the energy and

    n  depend only on the magnitude but not the direction of the momentum but the argument

    can be extended to anisotropic systems as well). In this formula,   np   is the distribution

    function of  quasi-particles  which are elementary excitations of the interacting system. The

    variation of  np   occurs when, e.g., a new particle is added to the system. However, these

    quasi-particles are not completely free and, therefore, unlike in the phonon analogy, the

    total energy is not equal to the sum of individual energies. However, the variation of the

    total energy δE  is related to  δnp  via (1.1). In fact, this relation serves as a definition of the

    quasiparticle energy via the variational derivative of  E   :

    εp  =  δE 

    δnp.

    One more –and crucial assumption–is that the quasi-particles of an interacting

    Fermi system are   fermions,   which is not at all obvious. For example, regardless of the

    statistics of individual atoms which can be either fermions or bosons, phonons are always

    bosons. Landau’s argument was that if quasi-particles were bosons they could accumulate

    without a restriction in every quantum state. That means that an excited state of a quantumsystem has a classical analog. Indeed, an excited state of many coupled oscillators is a

    classical sound wave. Fermions don’t have macroscopic states so quasi-particles of a Fermi

    systems must be fermions (to be precise they should not be bosons; proposals for particles

    of a statistics intermediate between bosons and fermions–anyons–have been made). On

    quite general grounds, one can show that quasi-particles must have spin 1/2  regardless   of 

    (half-integer) spin of original particles [3],[4]. (Thus, quasi-particles of a system composed

    of fermions with spin  S  = 3/2 still have spin 1/2.) Therefore, the quasiparticle energy andthe occupation number are the operators in the spin space represented by 2 × 2 matrices ε̂and n̂. As long as one has a system of fermions with doubly (S z  = ±1/2) degenerate states,the entropy on purely combinatorial grounds is

    S/vol = −Tr 

      d3 p

    (2π)3 [n̂p ln n̂p − (1 − n̂p)ln(1 − n̂p)] ,

    which is the same expression in for a Fermi gas. The equlibrium occupation numbers are

    obtained by equating the variation of  S   at fixed volume to zero. As for a Fermi gas, this

    4

  • 8/18/2019 Ferimi Liquid123

    5/15

    gives

    n̂ =  1

    expε̂−µT 

     + 1

    ,   (1.2)

    In contrast to the Fermi-gas, however, ε̂ is a functional of n̂  itself.

    If the system is not in the presence of the magnetic field and not ferromagnetic,

    ε̂αβ  = εδ αβ , n̂αβ  = nδ αβ .

    In a general case, instead of (1.1) we have

    δE  = 

      d3 p

    (2π)3Trε̂ ( p) δ ̂n ( p) ,

    which, for a spin-isotropic liquid, reduces to

    δE  = 2 

      d3 p

    (2π)3Trε ( p) δn ( p) .

    The occupation number is normalized by the condition

    δN  = 

      d3 p

    (2π)3Trδn ( p) = 2

       d3 p

    (2π)3δn ( p) = 0,

    where N   is the total number of  real   particles.

    For T=0, the chemical potential coincides with energy of the topmost state

    µ =  ε ( pF ) ≡ E F .

    Another important property (known as  Luttinger theorem)   is that the volume of the Fermi

    surface is not affected by the interaction. For an isotropic system, this means the Fermi

    momenta of free and interacting systems are the same. A simple argument is that the

    counting of states is not affected by the interaction, i.e., the relation

    N  = 24πp3

    F /3(2π)3

    holds in both cases. A general proof of this statement is given in Ref. [ [3]].

    B. Interaction of quasi-particles

    Phonons in a solid do not interact only in the first (harmonic) approximation. Anhar-

    monism results in the phonon-phonon interaction. However, the interaction is weak at low

    5

  • 8/18/2019 Ferimi Liquid123

    6/15

    energies not really because the coupling constant is weak but rather because the scattering

    rate of phonons on each other is proportional to a high power of their frequency:   τ −1ph-ph ∝ ω5.As a result, at small  ω  phonons are almost free quasi-particles. Something similar happens

    with the fermions. The nominal interaction may as well be strong. However, because of the

    Pauli principle, the scattering rate is proportional to (ε− εF )2 and weakly excited statesinteract only weakly. In the Landau theory, the interaction between quasi-particles is intro-

    duced via a phenomenological interaction function  defined by the proportionality coefficient

    (more precisely, a kernel) between the variation of the occupation number and the corre-

    sponding variation in the quasi-particle spectrum

    δεαβ  =    d3 p

    (2π)3f αγ,βδ (  p,   p

    ) δnγδ (  p) .

    Function f αγ,βδ (  p,   p) describes the interaction between the quasi-particles of momenta    p and

      p (notice that these are both   initial   states of the of the scattering process). Spin indices

    α   and   β   correspond to the state of momentum     p   whereas indices   γ   and   δ  correspond to

    momentum   p. In a matrix form,

    δ ̂ε = Tr   d3 p

    (2π)3f̂  (  p,   p) δ ̂n (  p) ,   (1.3)

    where Tr

    denotes trace over spin indices  γ  and  δ. For a spin-isotropic FL, when  δ ̂ε =  δ αβ δεand  δ ̂n =  δ αβ δn  external spin indices (α  and  β ) can also be traced out and Eq.(1.3) reduces

    to

    δε  = 

      d3 p

    (2π)3f  (  p,   p) δn (  p) ,

    where

    f  (  p,   p) ≡  12

    TrTr f̂  (  p,   p) .

    For small deviations from the equilibrium,  δn (  p) is peaked near the Fermi surface. Function

    f̂  (  p,   p) can be then estimated directly on the Fermi surface, i.e., for |  p| = |  p| = pF . Then  f̂ depends only on the angle between    p and    p. The spin dependence of  f̂  can be established on

    quite general grounds. In a spin-isotropic FL,   f̂    can depend only on the scalar product of 

    spin operators but on the products of the individual spin operators with some other vectors.

    Thus the most general form of  f̂   for a spin-isotropic system is

    ν ∗ f̂  (  p,   p) = F s (θ) Î  +  F a (θ) σ̂ · σ̂,

    6

  • 8/18/2019 Ferimi Liquid123

    7/15

    where  Î   is the unity matrix,  σ̂   is the vector of three Pauli matrices,  θ   is the angle between

      p  and    p, and the density of states was introduced just to make functions  F  (θ) and  G (θ)

    dimensionless. (Star in  ν ∗ means that we have used a renormalized value of the effective

    mass so that  ν 

    = m

    kF /π

    2

    , but this is again just a matter of convenience.) Explicitly,

    ν ∗f αγ,βδ (  p,   p) = F s (θ) δ αβ δ γδ  + F 

    a (θ) σ̂αβ  · σ̂γδ .   (1.4)

    In general, the interaction function is not known. However, if the interaction is weak,

    one can relate  f̂   to the pair interaction potential. The microscopic theory of a Fermi

    liquid [3] shows that the Landau interaction function is related to the interaction vertex

    Γαβ,γδ (  pε,   pε|  p + q, ε + ω,   p − q, ε − ω), which describes scattering of two fermions in ini-

    tials states with momenta, energies, and spin projections     p, ε, α   and     p

    , ε

    , β,   respectivelyinto the final states    p +  q, ε + ω, γ  and    p, ε + ω, δ  via

    f αβ,γδ (  p,   p) = Z 2 lim

    q/ω→0Γαβ,γδ (  pε,   pε

    |  p +  q, ε + ω,   p −  q, ε− ω) ,

    where Z  is the renormalization factor of the Green’s function.

    Example 2  At first order in the interaction,  Γ  is represented by two diagrams, one of which 

    is obtained from the other by a permutation of outgoing lines. If the interaction conserves 

    spin,

    Γαβ,γδ (  pε,   pε|  p + q, ε + ω,   p −  q, ε− ω) = δ αγ δ βδU  (q ) − δ αδδ βγ U  (  p−   p + q ) .

    Using an SU(2) identity 

    δ αδδ βγ  = 1

    2 (δ αγ δ βδ  + σ̂αγ ·σ̂βδ) ,

    this expression for  Γ  reduces to

    Γαβ,γδ (  pε,   pε|  p + q, ε + ω,   p − q, ε − ω) =   δ αγ δ βδ

    U  ( q ) − 1

    2U  (  p−   p + q )

    −12

    σ̂αγ ·σ̂βδU  (  p −   p + q ) .

    Exchanging indices  β   and   γ  and taking the limit   q  →  0  (at this order the vertex does not depend on the energy), we obtain for  f̂ 

    f αγ,βδ (  p,   p) = δ αβ δ γδ

    U  (0) − 1

    2U  (|  p−   p|)

    − 1

    2σ̂αβ  · σ̂γδU  (|  p−   p|) .

    7

  • 8/18/2019 Ferimi Liquid123

    8/15

    [Because this relation holds only at first order of the perturbation theory,  Z  still equals to  1.]

    On the Fermi surface, |  p −   p|  = 2 pF  sin θ/2.  Comparing this expression with Eq.(1.4), we  find that 

    F s (θ) =   ν −1

    U  (0) − 12U  (2 pF  sin θ/2)

      (1.5)

    F a (θ) = −12

    U  (2 pF  sin θ/2) .

    Notice that a repulsive interaction  (U > 0)  corresponds to the   attraction in the spin-exchange 

    channel ( F a

  • 8/18/2019 Ferimi Liquid123

    9/15

    In the second term re-label the variables   p →   p,   p →  p , use the fact that  f  (  p,   p) = f  (  p,   p)and integrate by parts

       d3 p

    (2π)3  pδn  =

       d3 p

    (2π)3m

    ∂ε

    ∂  pδn −

       d3 p

    (2π)3

       d3 p

    (2π)3f  (  p,   p)

     ∂n (  p)

    ∂  p  δn (  p) .

    Because this equation should be satisfied for an arbitrary variation  δn (  p) ,

      p

    m =

     ∂ε

    ∂  p −

       d3 p

    (2π)3f  (  p,   p)

     ∂n (  p)

    ∂  p  .

    Near the Fermi surface, the quasi-particle energy can be always written as

    ε ( p) = v∗F  ( p − pF )

    so that

    ∂ε∂  p  = v∗F  ˆ p =   pF m∗ ˆ p,

    where ˆ p  is the unit vector in the direction of    p, and thus

      p

    m =

     pF  ˆ p

    m∗ −

       d3 p

    (2π)3f  (  p,   p)

     ∂n (  p)

    ∂  p

    Now∂n (  p)

    ∂  p  =

     ∂ε (  p)

    ∂  p∂n

    ∂ε .

    Near the Fermi surface,

    ∂n∂ε

      = −δ (ε − E F )

    and  p

    m =

     pF  ˆ p

    m∗  +

     1

    2ν ∗F 

       dΩ

    4π f  (  p,   p)

     pF  ˆ p

    m∗

    or  p

    m =

     pF  ˆ p

    m∗  +

     1

    2ν ∗F 

       dΩ

    4π f  (  p,   p)

     pF  ˆ p

    m∗  =

     pF  ˆ p

    m∗  +

     1

    2

     pF π2

       dΩ

    4π f  (  p,   p) pF  ˆ p

    .

    Setting |  p| = pF  and multiplying both sides of the equation by ˆ p, we obtain

    1m

     =   1m∗

     + 12

     pF π2

       dΩ

    4πf  (θ)cos θ

    or1

    m∗  =

      1

    m −   pF 

    2π2

       dΩ

    4πf  (θ)cos θ.   (1.7)

    This is the Landau’s formula for the effective mass. Noticing that  f  (θ) = (1/2)TrTr f̂  (θ) =

    (2/ν ∗F ) F  (θ) = (2π2/m∗ pF ) , the last equation can be reduced to

    m∗

    m  = 1 +  

      dΩ

    4πF  (θ)cos θ ≡ 1 + F 1.

    9

  • 8/18/2019 Ferimi Liquid123

    10/15

    Although we do not know the explicit form of  f  (θ) ,  some useful conclusions can be made

    already from the most general form [Eq.(1.7)]. If  f  (θ)cos θ   is negative, 1/m∗ >  1/m →m∗ < m; conversely, if f  (θ)cos θ is positive, m∗ > m. For the weak short-range interaction,we know that  m∗ > m, whereas for a weak long-range (Coulomb) interaction  m∗ < m. Now

    we see that both of these cases are described by the Landau’s formula. Recalling the weak-

    coupling form of  F  (θ) [Eq.(1.5)], we find

    F 1   = 

      dΩ

    4πF  (θ)cos θ =

       dΩ

    U  (0) − 1

    2U 

    2 pF  sin

     θ

    2

    cos θ

    = −   dΩ

    1

    2U 

    2 pF  sin

     θ

    2

    cos θ.

    If  U   is repulsive and picked at small  θ  where cos θ   is positive,  F 1   <  0 and  m∗ < m.  This

    case corresponds to a screened Coulomb potential. If  U   is repulsive and depends on  θ  onlyweakly (short-range interaction), the angular integral is dominated by values of  θ  close to

    π,  where cos θ m. In general, forward scattering reduces the effective mass,

    whereas large-angle scattering enhances it.

    E. Spin susceptibility

    1. Free electrons 

    We start with free electrons. An electron with spin  s  has a magnetic moment  µ  = 2µBs,

    where   µB   =   eh̄/2mc   is the Bohr magneton. Zeeman splitting of energy levels is ∆E   =

    E ↑ − E ↓  = µBH  − (−µBH ) = 2µBH.  The number of spin-up and -down electrons is foundas an integral over the density of states

    n↑,↓ = 1

    2

       E F ±µBH 0

    dεν  (ε) .

    The Fermi energy now also depends on the magnetic field but the dependence is only

    quadratic (why?), whereas the Zeeman terms are linear in  H.  For weak fields, one can ne-

    glect the field-dependence of  E F . Total magnetic moment per unit volume–magnetization–is

    found by expanding in  H 

    M   =   µB (n↑ − n↓) =  µB2

       E F +µBH 0

    dεν  (ε) −   E F −µBH 0

    dεν  (ε)

    =  µB

    2

       E F 0

    dεν  (ε) + µBHν F  −   E F 0

    dεν  (ε) + µBHν F 

    =   µ2BHν F .

    10

  • 8/18/2019 Ferimi Liquid123

    11/15

    Spin susceptibility

    χ = ∂M 

    ∂H   = µ2Bν F    (1.8)

    is finite and positive, which means that spins are oriented along the external magnetic field.

    Notice that if, for some reason the bare   g- factor of electrons is not equal to 2,   thenEq.(1.8) changes to

    χ = g

    2µ2Bν F    (1.9)

    2. Fermi liquid 

    In a Fermi liquid, the energy of a quasi-particle in a magnetic field is changed not only

    due to the Zeeman splitting (as in the Fermi gas) but also because the occupation number

    is changed. This effect brings in an additional term in the energy as the energy is related

    to the occupation number. This is expressed by following equation

    δ ̂ε = −µBH · σ̂ + Tr 

      d3 p

    (2π)3f̂  (  p,   p) δn (  p) .

    The first term is just the Zeeman splitting, the second one comes from the interaction. Now,

    δ ̂n = ∂ ̂n

    ∂εδ ̂ε,

    and we have an equation for  δ ̂ε

    δ ̂ε = −µBH · σ̂ + Tr 

      d3 p

    (2π)3f̂  (  p,   p)

     ∂ ̂n

    ∂ε δ ̂ε (  p) .

    Replacing   ∂ ̂n∂ε

     by the delta-function and setting |  p| = pF 

    δ ̂ε ( pF  ˆ p) = −µBH · σ̂ − Trν ∗F    dΩ

    4πf̂  ( pF  ˆ p, pF  ˆ p

    ) δ ̂ε ( pF  ˆ p) .   (1.10)

    Let’s try a solution in the following form

    δ ̂ε = −µB2

      gH · σ̂,   (1.11)

    where g  has the meaning of an effective  g− factor. For free electrons,  g = 2. Recalling thatν ∗F f̂  = F  (θ)

     Î  + G (θ) σ̂ · σ̂ and substituting (1.11) into (1.11), we obtain

    −µB2

      gH · σ̂ = − µBH · σ̂ − Tr 

      dΩ

    F s (θ) Î  +  F a (θ) σ̂ · σ̂

    −µB

    2  gH · σ̂

    .

    11

  • 8/18/2019 Ferimi Liquid123

    12/15

    The term containing F  in the integral vanishes because Pauli matrices are traceless: Trσ = 0.

    The term containing  G   is re-arranged using the identity Tr(σ̂ · σ̂) σ̂ = 2σ̂  upon which weget

    g

    2 = 1

    − g

    2F a0

    ,

    where

    F a0   = 

      dΩ

    4πF a (θ) .

    Therefore,

    g =  2

    1 + F a0.

    Substituting this result into Eq.(1.9) and replacing   ν F    by its renormalized value   ν ∗F ,   we

    obtain the spin susceptibility of a Fermi liquid

    χ∗ =  µ2Bν 

    ∗F 

    1 + F a0= χ

    1 + F s11 + F a0

    ,

    where  χ   is the susceptibility of free fermions. Notice that χ∗ differs from  χ  both because

    the effective mass is renormalized and because the g-factor differs from 2. When  F a0   = −1,the g-factor and susceptibility diverge signaling a ferromagnetic instability. However, even

    if  G0 = 0, χ∗ is still renormalized in proportion to the effective mass. If  m∗ diverges (which

    is a signature of a metal-insulator transition of Mott type),   χ∗ diverges as well. Recall

    that [Eq.(1.5)] in the weak-coupling limit  F a (θ) = − 12

    U  (2 pF  sin θ/2) .  Thus, for repulsive

    interactions  g > 2 which signals a ferromagnetic tendency. This is another manifestation of 

    general principle that repulsively interacting fermions tend to have spin aligned to minimize

    the energy of repulsion. For normal He3, G0 ≈ −2/3.

    F. Zero sound

    All gases and liquids support sound waves. Even ideal gases have finite compressibilities

    and therefor finite sound velocities

    s =

     ∂P 

    ∂ρ .

    For example, in an ideal Boltzmann gas  P   = nT   = ρT/m and

    s =

     T 

    m =

      1√ 3

    vT ,

    12

  • 8/18/2019 Ferimi Liquid123

    13/15

  • 8/18/2019 Ferimi Liquid123

    14/15

  • 8/18/2019 Ferimi Liquid123

    15/15

    at xxx.lanl.gov/abs/comd-mat/9503150.

    [1] D. Yue, L. I. Glazman and K. A. Matveev, Phys. Rev. B  49, 1966 (1994).[2] Mahan,  Many-body physics 

    [3] A. A. Abrikosov, L. P. Gorkov, and I. E. Dzyaloshinski,  Methods of quantum field theory in 

    statistical physics , (Dover Publications, New York, 1963).

    [4] E.M. Lifshitz and L.P. Pitaevskii,   Statistical Physics   II, (Course of Theoretical Physics, v.

    IX), Pergamon Press, Oxford, 1980.

    [5] D. Pines and P. Nozieres, The Theory of Quantum Liquids , Benjamin, New York, 1969.

    [6] L. D. Landau, Zh. Eksp. Teor. Fiz. 30, 1058 (1956)–[Sov. Phys. JETP 3, 920 (1957)]. Reprinted

    in: D. Pines,  The Many-Body Problem , Benjamin, Reading, MA.

    [7] B. L. Altshuler and A. G. Aronov, in   Electron-electron interactions in disordered conductors ,

    edited by A. L. Efros and M. Pollak (Elsevier, 1985), p. 1.

    [8] G. Zala, B. N. Narozhny, and I. L. Aleiner, Phys. Rev. B  64, 214204 (2001).

    [9] E. Abrahams, S. V. Kravchenko, and M. P. Sarachik, Rev. Mod. Phys. 73, 251 (2001).

    [10] B. L. Altshuler, D. L. Maslov, and V. M. Pudalov,

    Physica E  9, 209 (2001).

    15