Derivation of quantum
hydrodynamic equations with
Fermi-Dirac and Bose-Einstein
          statistics
   Luigi Barletti (Università di Firenze)
  Carlo Cintolesi (Università di Trieste)
                 6th MMKT
         Porto Ercole, june 9th 2012
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                      Madelung equations
       The theory of quantum fluid equations dates back to 1926,
       when E. Madelung discovered the hydrodynamic form of
       Schrödinger equation:
                           ∂t n + div(nu) = 0,
                                                    √ 
                                                ~2 ∆ n
                                              
                                  1     2
                           ∂t u + ∇|u| + ∇ V −      √   = 0,
                                  2             2     n
                          √
       where ψ =              n eiS/~ and u = ∇S (and m = 1).
                            E. Madelung, Quantentheorie in hydrodynamischer
                            Form. Zeitschr. f. Phys. 40, 322–326 (1926)
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                          Bohm potential
       Madelung equations look like an irrotational compressible Euler
       system with an additional term containing the Bohm potential
                                        √
                                    ~2 ∆ n
                                 −      √ ,
                                    2    n
       named after David Bohm, who based on it his famous
       interpretation of quantum mechanics.
                            D. Bohm, A suggested interpretation of the quantum the-
                            ory in terms of “hidden variables”. Physical Review 85,
                            166–193 (1952)
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                        Wigner functions
       An “kinetic” derivation of Madelung equations can be obtained
       by using the Wigner function
                               Z                      
                         1               ξ            ξ
        w(x, p, t) =               ψ x + , t ψ x − , t e−ip·ξ/~ dξ,
                     (2π~)d/2 Rd         2            2
       and writing the equations for the moments
                    Z                              Z
          n(x, t) = w(x, p, t) dp,       nu(x, t) = p w(x, p, t) dp.
                            E. Wigner, On the quantum correction for thermody-
                            namic equilibrium. Physical Review 40, 749–759 (1932)
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                             Mixed states
       Madelung equations hold for a pure state, described by a wave
       function ψ, but for a mixed (statistical) state the system is not
       formally closed.
       However, the derivation of quantum fluid models for collisional
       systems, necessarily requires a statistical description and,
       therefore, the problem arises of generalizing Madelung
       equations to such a situation.
       The question is not merely academic, because quantum fluid
       models can be of great interest for nanoelectronics.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                Quantum hydrodynamics
       Indeed, a renewed interest for the subject dates back to the half
       of nineties with the work of C. Gardner, who proposed a
       quantum hydrodynamic model based on a local quantum
       Maxwellian obtained from Wigner’s O(~2 ) corrections to
       thermal equilibrium.
                            C.L. Gardner, The quantum hydrodynamic model for
                            semiconductor devices. SIAM J. Appl. Math. 54(2),
                            409–427 (1994)
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                              Quantum entropy principle
       But it is only with the work of Degond and Ringhofer that the
       problem was set on a solid theoretical basis with the
       elaboration of the quantum version of the maximum entropy
       principle (QMEP).
                                            P. Degond, C. Ringhofer, Quantum moment
                                            hydrodynamics and the entropy principle. J.
                                            Stat. Phys. 112(3-4), 587–628 (2003)
       The QMEP has been exploited to generate several quantum
       fluid models of various kind.
A short history     Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                  Quantum fluid models from the QMEP (I)
            • P. Degond, F. Méhats, C. Ringhofer, Quantum energy-transport and
                  drift-diffusion models, J. Stat. Phys., 2005.
            • A. Jüngel, D. Matthes, A derivation of the isothermal quantum
                  hydrodynamic equations using entropy minimization, Z. Angew. Math.
                  Mech., 2005.
            • N. Ben Abdallah, F. Méhats, C. Negulescu. Adiabatic quantum-fluid
                  transport models, Commun. Math. Sci., 2006.
            • A. Jüngel, D. Matthes, J. P. Milisic, Derivation of new quantum
                  hydrodynamic equations using entropy minimization, SIAM J. Appl.
                  Math., 2006.
            • P. Degond, S. Gallego, F. Méhats. Isothermal quantum hydrodynamics:
                  derivation, asymptotic analysis, and simulation, Multiscale Model.
                  Simul., 2007.
            • P. Degond, S. Gallego, F. Méhats, An entropic quantum drift-diffusion
                  model for electron transport in resonant tunneling diodes, J. Comput.
                  Phys., 2007.
A short history    Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                  Quantum fluid models from the QMEP (II)
            • S. Brull, F. Méhats, Derivation of viscous correction terms for the
                  isothermal quantum Euler model, Z. Angew. Math. Mech., 2010.
            • L. B., F. Méhats, Quantum drift-diffusion modeling of spin transport in
                  nanostructures. J. Math. Phys., 2010.
            • L. B., G. Frosali, Diffusive limit of the two-band k·p model for
                  semiconductors. J. Stat. Phys., 2010.
            • N. Zamponi, L. B., Quantum electronic trasport in graphene: a kinetic
                  and fluid-dynamical approach. M2AS, 2011.
            • N. Zamponi, Some fluid-dynamic models for quantum electron transport
                  in graphene via entropy minimization. KRM, 2012.
            • A. Jüngel. Transport Equations for Semiconductors. Springer, 2009.
            • A. Jüngel. Dissipative quantum fluid models. Riv. Mat. Univ. Parma,
                  2012.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                       Different statistics
       Although the QMEP was originally stated for a general (convex)
       entropy functional, nevertheless, it has been applied only to
       Boltzmann entropy.
       Partial exceptions:
            • P. Degond, S. Gallego, F. Méhats, An entropic quantum
              drift-diffusion model for electron transport in resonant tunneling
              diodes, J. Comput. Phys., 2007.
            • M. Trovato, L. Reggiani, Quantum maximum entropy principle for
              a system of identical particles. Phys. Rev. E , 2010.
            • A. Jüngel, S. Krause, P. Pietra, Diffusive semiconductor moment
              equations using Fermi-Dirac statistics. ZAMP, 2011.
       The work we are going to expose is exactly intended to fill this
       gap.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                       The kinetic model
       The starting point is the Wigner equation with BGK collisional
       term
                           p                    g[w] − w
                     ∂t + m   · ∇x + Θ~ [V ] w =          ,
                                                    τ
       where:
            • w(x, p, t), is the Wigner function of the system;
            • Θ~ [V ] = ~i V x + i~           i~
                                                                               
                                 2 ∇p − V x − 2 ∇p                                 ;
            • τ is a typical collisional time;
            • g[w] is the local equilibrium state given by the QMEP.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                       Scaling the Wigner-BGK equation
       After a suitable scaling, we can write the Wigner-BGK equation
       as follows:
                                                                       1
                           (∂t + p · ∇x + Θ [V ]) w =                   (g[w] − w)
                                                                       α
            • α = tτ (hydrodynamic parameter);
                   0
            •  = x ~p (semiclassical parameter).
                   0 0
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                                    QMEP
       We assume that g[w] is given by the QMEP:
       g[w] is the most probable state compatible with the information
       we have about it:
            • g[w] has a constant temperature T
            • g[w] has the same density and current as w:
                            hg[w]i = n = hwi,                    hpg[w]i = J = hpwi.
       “Most probable” means that g[w] minimizes an entropy
       functional H.
       Then, g[w] is chosen as a minimizer of H among all Wigner
       functions that share with w the same moments n and J.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                       Entropy functional
       In terms of density operators, the suitable entropy functional for
       the isothermal case is the free-energy                                                               
                                    1
            H(%) = Tr T % log % + (1 − λ%) log(1 − λ%) + H% ,
                                    λ
       which also incorporates the information on particle statistics:
                                     1,                              Fermi-Dirac
                                  
                                  
                                  
                               λ=    0,                 Maxwell-Boltzmann
                                  
                                  
                                    −1,                          Bose-Einstein
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                      Constrained minimization problem
       The Wigner function g[w] satisfies the QMEP if:
       % = Op (g[w]) is a minimizer of H under the constraints
                        hg[w]i = n = hwi,                     hpg[w]i = J = hpwi.
       Here Op denotes the Weyl quantization, mapping 1-1 Wigner
       functions and density operators.
A short history   Applying the QMEP    The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                            The minimizer
       Theorem
       A necessary condition for g = g[w] to be solution of the
       constrained minimization problem is that Lagrange multipliers
       A and B = (B1 , . . . , Bd ) exist such that
                                                 H(A,B) −1
                                       Op (g) = e T + λ     ,
       where                                                                 
                                                                   |p−B|2
                                      H(A, B) = Op                   2     −A ,
       and A, B have to be determined as functions of n, J from the
       moment constraints.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                      Hydrodynamic limit
       Theorem
       In the hydrodynamic limit α → 0, the solution wα of the
       Wigner-BGK equation tends to the local equilibrium state g[w0 ],
       whose moments n and J satisfy the equations
                              ∂      ∂
                                 n+     Jj = 0,
                              ∂t    ∂xj
                              ∂        ∂ 
                 ∂
                                 Ji +     pi pj g[w0 ] + n     V = 0.
                              ∂t      ∂xj                  ∂xi
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                          Formal closure
                              
         
       The unknown moment pi pj g[w] can be expressed in terms of
       the Lagrange multipliers as follows:
               ∂ 
             ∂               ∂                  ∂
                  pi pj g[w] =     (nui Bj ) +     Bj nuj − Bj + n     A.
              ∂xj              ∂xj             ∂xi                 ∂xi
       This provides a formal (and rather implicit) closure of the
       quantum hydrodynamic equations because A, B are depend on
       n, J through the moment constraints.
A short history   Applying the QMEP   The local equilibrium    Semiclassical equations   Some asymptotics   Conclusions
                                      Difficulties with g[w]
       From now on, we shall simply denote by g the local equilibrium
       state g[w].
       The phase-space function g is a very complicated object,
       involving back and forth Weyl quantization:
                                      (        "                                   !#         )−1
                                                              |p − B|2   A
                   g = Op −1          exp Op                         −                 +λ
                                                                 2T      T
       The only hope we have to get something explicit is expanding g
       semiclassically, i.e. in powers of .
A short history   Applying the QMEP   The local equilibrium    Semiclassical equations   Some asymptotics   Conclusions
                            Semiclassical expansion of g
       Skipping all technical details, we find that g has the following
       expansion
                           g = g (0) + 2 g (2) + O(4 )
       where g (0) is the “classical” distribution
                                                                 1
                                          g (0) =          (p−B)2
                                                                  − TA
                                                       e     2T          +λ
       and g (2) is a complicated expression involving A, B and their
       derivatives.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                     Semiclassical expansion of A and B
       What we really need is the expansion of A, B as functions of n,
       J, as it results from the constraints hgi = n and hpgi = J.
       Using the expansion of g, it turns out that
             A = A(0) + 2 A(2) + O(4 ),                      B = B (0) + 2 B (2) + O(4 ).
       At leading order we obtain
                                                                                     (0)           −1          n
                    A         = T φd                              ,       B (0) = u = J/n,
                                      2        (2πT )d/2
       where
                                                                            +∞
                                                                                  t (s−1)
                                                                      Z
                                  1               1
                        φs (z) = − Lis (−λez ) =                                          dt.
                                  λ              Γ(s)                   0        et−z  +λ
A short history        Applying the QMEP          The local equilibrium       Semiclassical equations   Some asymptotics   Conclusions
                                       Condition for invertibility
                  8
                            FD, d=3
                            MB, d=3
                  7         BE, d=3
                            BE, d=2
                            BE, d=1
                  6
                                                                                                 Plots of φ d for some
                                                                                                            2
                  5                                                                              values of λ and d.
                  4
                                                                                                 For λ < 0 and d ≥ 3,
                  3
                                                                                                 φ d ranges from 0 to
                                                                                                   2 
                  2
                                                                                                 ζ d2 /|λ|.
                  1
                  0
                  −2      −1.5    −1       −0.5        0      0.5         1      1.5     2
                                                     d
       In the BE case, we have to assume n < (2πT ) 2 ζ d2 /|λ|.
                                                          
       Exceeding particles fall in the condensate phase, not described
       by our model.
A short history       Applying the QMEP   The local equilibrium    Semiclassical equations   Some asymptotics       Conclusions
                         A and B: second-order expressions
                        1 h                                     i φ0d −2                    0
                                                                            1  ∂i A(0) 2 φ d2 −3
        A(2) =             ∂i uj (∂i uj − ∂j ui ) − 2∂i ∂i A(0)     2
                                                                         −
                       24T                                        φ0d −1   24      T       φ0d −1
                                                                               2                                2
           (2)          nd    h                       i
        Bi        =         ∂j (∂i uj − ∂j ui )φ0d −1
                       12Tn                      2
       where
                                                                    ∂
                                                        ∂i :=
                                                                   ∂xi
       and
                                                   A(0)                  n 
                                   φ0s := φs                      = φs φ−1
                                                                        d
                                                       T                2   nd
A short history    Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                  Semiclassical hydrodynamic equations (I)
       We can now substitute the expansions
                         A ≈ A(0) + 2 A(2)               and B ≈ B (0) + 2 B (2)
       in the quantum hydrodynamic equations
           ∂      ∂
              n+     (nuj ) = 0,
           ∂t    ∂xj
           ∂            ∂               ∂                  ∂
              (nui ) +     (nui Bj ) +     Bj nuj − Bj + n     (A + V ) = 0.
           ∂t          ∂xj             ∂xi                 ∂xi
A short history    Applying the QMEP   The local equilibrium     Semiclassical equations   Some asymptotics   Conclusions
                  Semiclassical hydrodynamic equations (II)
       We obtain therefore our main result:
          ∂      ∂
             n+     (nuj ) = 0,
          ∂t    ∂xj
          ∂            ∂                  ∂ −1  n      ∂              
             (nui ) +     (nui uj ) + Tn    φd       +n      V + 2 Q(n)
          ∂t          ∂xj                ∂xi 2 nd       ∂xi
                                            0
                            2 n ∂ Rjk Rkj φ d2 −2 2 nd      ∂       0
                                                                            
                          =                       +      Rij      Rjk φ d −1 ,
                            24T ∂xi 2 φ0d           12T      ∂xk        2
                                           −1          2
       where
                                                 d                       ∂        ∂
                               nd = (2πT ) 2 ,                 Rij :=       ui −     uj
                                                                        ∂xj      ∂xi
A short history   Applying the QMEP   The local equilibrium       Semiclassical equations   Some asymptotics   Conclusions
                                 Modified Bohm potential
       The term 2 Q(n) can be identified as a modified Bohm
       potential:
                                  0                         0
                                                                
                    1       n  φ d
                                     −2
                                                n  2 φ d
                                                              −3
           Q(n) = − 2∆φ−1         2
                                        + ∇φ−1         02  ,
                                          
                        d         0          d
                   24   2    nd φd           2    n d     φ d −1
                                     −1                       2                                       2
       since
                                                      √
                                                    1∆ n
                                        lim Q(n) = − √ .
                                        λ→0         6 n
A short history   Applying the QMEP    The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                      Limit properties of φs
       From known asymptotic properties of the polylogarithms we
       obtain:
         lim φs (z) = ez ,                            for z ∈ R and s ∈ R
         λ→0
        φs (z) ∼ ez ,                                 for z → −∞, λ ∈ R and s ∈ R
                         zs
        φs (z) ∼               ,                      for z → +∞, λ = 1 and s 6= −1, −2, . . .
                      Γ(s + 1)
        φs (z) ∼ Γ(1 − s)(−z)s−1 , for z → 0− , λ = −1 and s < 1
A short history    Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                             The Maxwell-Boltzmann limit
       In the M-B limit λ → 0 the semiclassical equations become
                  ∂      ∂
                     n+     (nuj ) = 0,
                  ∂t    ∂xj
                                                                                        √ 
                                                                                    2 ∆ n
                                                                                
                  ∂            ∂                 ∂       ∂
                     (nui ) +     (nui uj ) + T     n+n                          V−    √
                  ∂t          ∂xj               ∂xi     ∂xi                         6    n
                                   2 ∂           
                              =           nRik Rkj ,
                                  12T ∂xj
       (corresponding to the equations found by Jüngel-Matthes (’05) and
       Degond-Gallego-Méhats (’07)).
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                             Vanishing-temperature limit
       The behavior of the semiclassical hydrodynamic equations as
       T → 0 depends dramatically on the sign of λ.
       Let us consider separately the three representative cases λ = 1
       (FD), λ = 0 (MB) and λ = −1 (BE).
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                           T → 0 limit: Fermi-Dirac case
       The FD case is the richest: the limit T → 0 is non-singular and
       yields “completely degenerate fluid” equations:
          ∂      ∂
             n+     (nuj ) = 0,
          ∂t    ∂xj
                                       d
                                           2−d       2
          ∂            ∂                2
                                             d
                                                 Γ d2 d ∂ 2
             (nui ) +     (nui uj ) +                    n     nd
          ∂t          ∂xj                     2π           ∂xi
                                                   √ 
                                     d − 2 2 ∆ n
                             
                         ∂
                    +n         V−                 √
                        ∂xi            d 6           n
                             2
                                                                                        
                             dπ          (d − 2) ∂ Rjk Rkj               ∂         d−2
                    =               2              n             + R ij      R jk n  d
                      12 Γ d2 + 1 d            d       ∂xi 4n d2         ∂xk
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                   T → 0 limit: Maxwell-Boltzmann case
       This is the most singular case, because the limits λ → 0 and
       T → 0 are somehow incompatible.
       The formal limit of the semiclassical hydrodynamic equations
       with BE statistics is only compatible with an irrotational fluid
       (R = 0) and depends on how (0, 0) is approached in the
       parameter plane (λ, T )
       Degond-Gallego-Méhats (’07) show that the fully-quantum fluid
       equations admit a limit, which is given by Madelung equations.
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                  T → 0 limit: Bose-Einstein case, d ≥ 3
       For λ = −1 and d ≥ 3, in the limit T → 0 the fluid is completely
       condensate and, then, this case cannot be considered within
       our description.
A short history   Applying the QMEP   The local equilibrium    Semiclassical equations   Some asymptotics   Conclusions
                  T → 0 limit: Bose-Einstein case, d = 2
       In this case the limit T → 0 is only compatible with an
       irrotational fluid. Moreover, we have to rescale the density as
                                                               n
                                                   N=             .
                                                              2πT
       The resulting limit equations are
                           ∂      ∂
                              N+     (Nuj ) = 0
                           ∂t    ∂xj
                                                                   2
                                                                       
                           ∂           ∂        ∂
                              ui + uj     uj +                  V−    ∆N = 0
                           ∂t         ∂xi      ∂xi                 12
A short history   Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                  T → 0 limit: Bose-Einstein case, d = 1
       In this, last, case we obtain
                       ∂      ∂
                          n+    (nu) = 0
                       ∂t    ∂x
                                                                    √ 
                                                            2 1 ∂ 2 n
                                                         
                       ∂     1 ∂ 2     ∂
                          u+      u +                     V− √          =0
                       ∂t    2 ∂x     ∂x                    2 n ∂x 2
A short history     Applying the QMEP   The local equilibrium   Semiclassical equations   Some asymptotics   Conclusions
                                                Conclusions
            • We derived semiclassical isothermal hydrodynamic
                  equations for Fermions or Bosons.
            • The method exploits Degond and Ringhofer’s Quantum
                  Maximum Entropy Principle and the semiclassical
                  expansion of the maximizer Wigner function.
            • We obtained an Euler-like system with quantum
                  corrections, of order ~2 , involving a modified Bohm
                  potential and the velocity curl tensor.
            • The Maxwell-Boltzmann limit and the T → 0 limit have
                  been investigated.
            • L.B., C. Cintolesi, Derivation of isothermal quantum fluid equations with
                  Fermi-Dirac and Bose-Einstein statistics. J. Stat. Phys. (to appear).
Thank you