* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Derivation of viscous correction terms for the isothermal quantum
Renormalization wikipedia , lookup
Topological quantum field theory wikipedia , lookup
Basil Hiley wikipedia , lookup
Dirac equation wikipedia , lookup
Quantum decoherence wikipedia , lookup
Wave function wikipedia , lookup
Quantum electrodynamics wikipedia , lookup
Particle in a box wikipedia , lookup
Measurement in quantum mechanics wikipedia , lookup
Quantum dot wikipedia , lookup
Copenhagen interpretation wikipedia , lookup
Probability amplitude wikipedia , lookup
Quantum field theory wikipedia , lookup
Quantum entanglement wikipedia , lookup
Theoretical and experimental justification for the Schrödinger equation wikipedia , lookup
Bell's theorem wikipedia , lookup
Renormalization group wikipedia , lookup
Quantum fiction wikipedia , lookup
Coherent states wikipedia , lookup
Many-worlds interpretation wikipedia , lookup
Orchestrated objective reduction wikipedia , lookup
Quantum computing wikipedia , lookup
Scalar field theory wikipedia , lookup
Hydrogen atom wikipedia , lookup
Quantum teleportation wikipedia , lookup
EPR paradox wikipedia , lookup
Path integral formulation wikipedia , lookup
Interpretations of quantum mechanics wikipedia , lookup
Relativistic quantum mechanics wikipedia , lookup
Quantum key distribution wikipedia , lookup
Quantum machine learning wikipedia , lookup
Quantum group wikipedia , lookup
Symmetry in quantum mechanics wikipedia , lookup
Density matrix wikipedia , lookup
History of quantum field theory wikipedia , lookup
Canonical quantum gravity wikipedia , lookup
Quantum state wikipedia , lookup
Hidden variable theory wikipedia , lookup
Derivation of viscous correction terms for the isothermal quantum Euler model S. Brull ∗† F. Méhats§¶ Abstract The aim of this paper is to compute viscous correction terms for the isothermal quantum Euler system of Degond, Gallego, Méhats (SIAM Multiscale Model Simul., 6, 2007). We derive this model by using a Chapman-Enskog expansion up to order 1. In a last part, we consider a situation where the flow is nearly irrotational in order to get a simplified model. 1 Introduction. This paper is the continuation of a program of work initiated in 2003 in [9]. Extending Levermore’s moments method [16] to a quantum context, P. Degond and C. Ringhofer derived hydrodynamical models from ChapmanEnskog expansions around quantum local equilibriums. Next, after this work, an entropic quantum drift-diffusion model was derived in [8], numerically discretized in [11] and discussed in the context of quantum transport in electronic nanostructures in [4]. In [6], an isothermal quantum Euler system was derived, recovering at a semiclassical asymptotics a model obtained in [14]. Note that extended quantum hydrodynamical models were also derived and discussed in [10, 12, 15, 2, 5]. All these models were obtained –formally– by applying a hydrodynamic or a diffusive scaling to a Wigner equation with ad-hoc relaxation operators ∗ *Mathematiques appliquees de Bordeaux, University of Bordeaux I, 351 cours de la Libération 33405 Talence Cedex, France † [email protected] § IRMAR (UMR CNRS 6625), Université de Rennes, Campus de Beaulieu, 35042 Rennes Cedex (France) ¶ [email protected] 1 and performing the limit as the mean free path converges to zero. In this paper, we are interested in the next order approximation of the hydrodynamic limit. In the classical setting, it is well-known [1] that this leads to the viscous correction to the Euler equations, i.e. to the Navier-Stokes system. Here, we apply the same idea in the quantum setting in order to derive the first order correction to the isothermal Euler equation of [6]. By analogy, we shall call quantum Navier-Stokes system the obtained model. A viscous model has been considered in [13, 3] where existence theorems are proved. It consists in two conservation equations for mass and current coupled with the Poisson equation. Let us shortly present the approach. From the Wigner equation, integrations with respect to the momentum variable p ∈ R3 lead to equations for the first two moments n(t, x) (the mass density) and n(t, x)u(t, x) (the current density), both densities being functions of the space variable x ∈ R3 and the time variable t ∈ R: ∂t n + div(nu) = 0, (1) ∂t (nu) + div(nu ⊗ u + P) + n∇V = 0, (2) where V denotes an applied potential. Of course, this system of equations is not closed, since the pressure tensor P is still expressed in terms of the microscopic Wigner function f (t, x, p): Z dp P= (p − u)(p − u) f (t, x, p) , (3) 3 (2π~) d R where ~ is the dimensionless Planck constant. The Wigner function f (t, x, p) can be seen as a quantum extension of the phase-space Boltzmann distribution function, though f (t, x, p) is not a positive function. In [6], an isothermal quantum Euler system was derived by closing the system of moments as follows. The expression (3) of P is simply replaced by another one in terms of n and nu: Z dp eq P= (p − u)(p − u) fn,nu , (4) (2π~)3 Rd eq where fn,nu is the so-called local equilibrium, depending only on n and nu in a non local (and non explicit) way – see Definition 1. Equivalently, this isothermal quantum Euler model can be obtained as the limit ε → 0 of the following quantum BGK equation in a hydrodynamic scaling: 1 eq ∂t f + p · ∇x f + Θ(V )f = (fn,nu − f ). ε 2 In the present paper, in order to obtain a Navier-Stokes model, f will be eq formally expanded up to first order around the local equilibrium fn,nu as eq f = fn,nu + εf1 . The outline of the paper is the following. In Section 2, the quantum Navier-Stokes model is derived and formulated in Theorem 1. Then, Section 3 is devoted to some simplifications of this model when the flow is assumed to be nearly irrotational: Proposition 1 is a reformulation of the quantum Navier-Stokes model in this situation. 2 Derivation of the model In this section we first recall some backgrounds about the Wigner equation and we present the model derived in the general situation. 2.1 Notations and background. By a density operator, we shall always mean a positive, Hermitian, traceclass operator acting on L2 (R3 ). Let us recall the definition of the first moments of a density operator %, i.e. the mass density n and the current density nu. These quantities are defined by duality, considering scalar test functions φ and vector ones Φ. We set Z ∞ 3 ∀φ ∈ C0 (R ) nφdx = Tr{%φ}, (5) ∀Φ ∈ C0∞ (R3 )3 Z nu · Φ dx = Tr %W −1 [p · Φ] 1 = −i~Tr % Φ · ∇ + (∇ · Φ) (6) 2 Note that an immediate consequence of (5) and (6) is the following property which will be useful later: Z Z i i~ ∀Φ − Tr{%(Φ · ∇)} = nu · Φ dx + n ∇ · Φ dx. (7) ~ 2 In (6), W −1 denote the inverse Wigner transform (or Weyl quantization). Let us recall the definitions of the Wigner transform and the inverse Wigner transform. The Wigner transform maps operators on L2 (R3 ) onto symbols, ie L2 (R3 × R3 ) functions of the classical position and momentum variables (x, p) ∈ R3 × R3 . More precisely, one defines the integral kernel of the 3 operator % to be the distribution %(x, x0 ) such that % operates on any function ψ(x) ∈ L2 (R3 ) as follows: Z %ψ(x) = %(x, x0 )ψ(x0 )dx0 . Then, the Wigner transform W [%](x, p) is defined by: Z iη·p 1 1 W [%](x, p) = % x − η, x + η e ~ dη. 2 2 (8) The Wigner transform can be inverted and its inverse (called Weyl quantization) is defined for any function f (x, p) as the operator acting on ψ(x) ∈ L2 (R3 ) as: Z ip·(x−y) x+y −1 −3 W [f ]ψ(x) = (2π~) f , p ψ(y)e ~ dp dy. (9) 2 We have also the following Parseval property: for any operators ρ and σ in L2 = {% : Tr(%%† ) < ∞}: Z dx dp † Tr(ρσ ) = W [ρ]W [σ] , (10) (2π~)3 where the bar means complex conjugation. We also recall the cyclicity of the trace, where a, b, c are three operators: Tr{[a, b]c} = Tr{[c, a]b} = Tr{[b, c]a}. In quantum statistical mechanics, the Von Neumann entropy is defined by: S(%) = Tr(% ln % − %). (11) In order to deal with isothermal problems, we introduce the quantum free energy defined by: G(%) = T S(%) + E(%) = Tr (T (% ln % − %) + H%) . (12) where H is the Hamiltonian defined by H=− ~2 ∆ + V. 2 (13) For simplicity, the potential V = V (t, x) applied to the system of particles is assumed to be given, independent of the particles. 4 In this paper, we suppose that the particle system interacts with a thermal bath and that the mass n and the current nu are conserved by this interaction. Following [9], [8], [6], we shall model this interaction via a relaxation operator constructed according to a principle of free energy minimization. To this aim, we define the local thermal equilibrium associated to n and nu, at temperature T through the following formal procedure. Definition 1. Let the scalar function n(x) ≥ 0 and the vectorial function nu(x) be given. Consider the following constrained minimization problem: min {G(%) such that % is a density operator satisfying (5) and (6)} . (14) The solution, if it exists, is called the local equilibrium density operator associated to n and nu. Lagrange multiplier theory for the constrained problem (14) (see [9], [6]) shows that there exist a scalar function A and a vector function B, both real valued and defined on R3 , such that this local equilibrium density operator takes necessarily the form: 1 eq %n,nu = exp − H(A, B) , (15) T where H(A, B) is the following modified Hamiltonian: 1 −1 1 2 H(A, B) = W (p − B) + A = (i~∇ + B)2 + A. 2 2 (16) The local equilibrium Wigner function associated to n and nu is the Wigner transform of %eq n,nu : eq fn,nu = W [%eq n,nu ]. 2.2 Derivation of the quantum Navier-Stokes system Let us now introduce the quantum BGK equation, by adding a relaxation operator to the quantum Liouville equation. This operator, denoted by Q(%), describes the collisions between the particles and a surrounding environnement at temperature T . In this paper, we choose this collision operator equal to the simplest quantum BGK operator that can be constructed according to the minimization problem (14): ı~∂t % = [H, %] + ı~ %eq n,nu − % , ε (17) where, for all t, the local equilibrium %eq n,nu (t) is defined thanks to Definition 1 from the density n(t, ·) and the current (nu)(t, ·) associated to %(t) through (5), (6). 5 Let us now expand % around %eq n,nu as follows: % = %eq n,nu + ε%1 . By inserting this expansion of % into the equation (17), we get the relation defining %1 %1 = −∂t %eq n,nu + ı [H, %eq n,nu ]. ~ By using a Wigner transformation of the equation (17), we get the following Wigner-BGK equation ∂t f + p · ∇x f − Θ(V )f = eq fn,nu −f ε (18) with V (t, x + ~2 η) − V (t, x − ~2 η) 0 f (x, p0 )eı(p−p )·η dηdp0 . ~ R6 (19) eq The Wigner transformation of the expansion of % gives f = fn,nu + εf1 . Hence f1 has the expression ı Θ(V )f = (2π)3 Z f1 = −∂t f − p · ∇x f + Θ(V )f and, formally, eq eq eq f1 = −∂t fn,nu − p · ∇x fn,nu + Θ(V )fn,nu + O(ε). (20) Moreover, by considering the first moments of (18) with respect to p, we obtain Z Z dp dp ∂t f + div pf =0 3 (2π~) (2π~)3 and Z ∂t fp dp + div (2π~)3 Z p ⊗ pf dp + n∇x V = 0. (2π~)3 The last equation can be rewritten Z Z dp dp eq + n∇x V = −εdiv p ⊗ pf1 . ∂t (nu) + div p ⊗ pfn,nu 3 (2π~) (2π~)3 6 Then (20) gives formally Z dp eq ∂t (nu) + div p ⊗ pfn,nu + n∇x V (2π~)3 Z eq eq eq = εdiv p ⊗ p ∂t fn,nu + p · ∇x fn,nu − Θ(V )fn,nu dp + O(ε2 ). (2π~)3 The isothermal quantum Navier-Stokes system is simply the model obtained when the O(ε2 ) term is dropped in this equation: ∂t n + div(nu) = 0, Z dp eq ∂t (nu) + div p ⊗ pfn,nu + n∇x V (2π~)3 Z eq eq eq = εdiv p ⊗ p ∂t fn,nu + p · ∇x fn,nu − Θ(V )fn,nu (21) dp , (22) (2π~)3 eq is still defined from n and nu through Definition 1. where fn,nu Let us now simplify this system. Equation (22) contains moments of eq order 2 and 3 of the equilibrium function fn,nu which might not be easy to handle. Denote Z Z dp dp eq eq Πij = pi pj fn,nu , Qijk = pi pj pk fn,nu . (23) 3 (2π~) (2π~)3 Using commutator properties as in [6] and [5], one can express divΠ in terms of n, nu, A and B only. We recall from [6] that: X X X (divΠ)i = ∂j Πij = ∂xj (nui Bj ) + n(uj − Bj )∂xi Bj − n∂xi A. (24) j j j P Similarly, we shall express j,k Qijk in terms of n, nu, A, B and Π only. This is the aim of the following theorem: Theorem 1. The above defined isothermal quantum Navier-Stokes system (21), (22) is formally equivalent to the following system, up to terms of order O(ε2 ): ∂t n + div(nu) = 0, (25) 3 X ∂t (nui ) + ∂xj (nui Bj ) + n(uj − Bj )∂xi Bj + n∂xi (V − A) = εSi (26) j=1 7 where A and B are related to n and nu via Definition 1 and where Si is equal to: Si = − 3 X 3 X ∂xk ∂xj (nui Bj ) Bk − n ∂xj (uj − Bj )∂xi Bj Bk k=1 j=1 − 3 X 3 X ∂xi Bk ∂xj (nuk Bj ) + n∂xj (uj − Bj )∂xk Bj k=1 j=1 − 3 X ∂xk (n∂xi (V − A) Bk ) + k=1 3 X (∂xk (nui ∂t Bk ) + ∂xi ∂t Bk nuk ) k=1 3 X 3 X B2 B2 − n ∂xi ∂t A + − n∂xi Bk ∂xk (V − A) + nuk ∂xi A + 2 2 k=1 k=1 3 X 3 X B2 B2 − ∂xj ∂xj A + nui + ∂xi A + nuj 2 2 j=1 j=1 + 3 X 3 X ∂xj ∂xk (Bk Πij ) + ∂xj Bk Πik + ∂xi Bk Πjk j=1 k=1 − 3 3 3 X ~2 X X ∂xj ∂xk (n∂xi ∂xj Bk ) + ∂xj ∂xj V nui + ∂xi V nuj , 4 j=1 k=1 (27) j=1 the tensor Π being defined by (23). Proof. From (24), we deduce directly that the l.h.s. of (22) can be rewritten as the l.h.s. of (26). To prove the theorem, it remains to consider the r.h.s. of (22), which is Z dp = εg1 + εg2 + εg3 , −εdiv p ⊗ pf1 (2π~)3 where we have set, for i = 1, 2, 3, (g1 )i = 3 X Z ∂t ∂xj eq pi pj fn,nu j=1 (g2 )i = 3 X 3 X Z ∂xj ∂xk dp , (2π~)3 eq pi pj pk fn,nu j=1 k=1 (g3 )i = − 3 X Z ∂ xj eq pi pj Θ(V )fn,nu j=1 8 dp , (2π~)3 dp . (2π~)3 (28) (29) (30) The expressions of g1 , g2 and g3 are given in the following lemmas whose proof is in the appendix. By summing up these expressions, the proof of the theorem is complete. Lemma 1. The quantity g3 defined by (30) can be written (g3 )i = 3 X ∂xj ∂xj V nui + ∂xi V nuj . (31) j=1 Lemma 2. The quantity g2 defined in (29) has the following expression: 3 X 3 X B2 B2 nui + ∂xi A + nuj (g2 )i = − ∂ xj ∂ xj A + 2 2 j=1 j=1 + 3 X 3 X ∂xj ∂xk (Bk Πij ) + ∂xj Bk Πik + ∂xi Bk Πjk j=1 k=1 3 3 ~2 X X − ∂xj ∂xk (n∂xi ∂xj Bk ). 4 (32) j=1 k=1 Lemma 3. Up to O(ε) terms, the quantity g1 defined in (28) has the following expression: (g1 )i = − 3 X 3 X ∂xk ∂xj (nui Bj ) Bk − n ∂xj (uj − Bj )∂xi Bj Bk k=1 j=1 − 3 X 3 X ∂xi Bk ∂xj (nuk Bj ) + n∂xj (uj − Bj )∂xk Bj k=1 j=1 − − 3 X k=1 3 X ∂xk (n∂xi (V − A) Bk ) + 3 X (∂xk (nui ∂t Bk ) + ∂xi ∂t Bk nuk ) k=1 n∂xi Bk ∂xk (V − A) + k=1 3 X k=1 B2 nuk ∂xi A + 2 B2 −n ∂xi ∂t A + . 2 3 (33) The system in the irrotational case In this part we focus on the special case of irrotational flows, which enables to simply notably the system obtained in Theorem 1. Indeed, it has been 9 shown in [6], [5] that when ∇ × u = 0 then the quantity B defined in Definition 1 is nothing but the velocity u itself. In the case of the isothermal quantum Euler equation, i.e. when we make ε = 0 in (26), it has been shown in [6] that when the initial data is irrotational, then the solution remains irrotational for all time. Unfortunately, it is not clear whether this property remains true for the quantum Navier-Stokes equations (25), (26). Instead, we have the following property: if (∇ × u)(t = 0) = O(ε) then we have (∇ × u)(t) = O(ε), which implies B = u + O(ε). In such situation, the model obtained by replacing B by u in the r.h.s. of (26) will only differ from (26) by O(ε2 ) terms, and thus will remain consistent with our theory. Proposition 1. Assume that (∇ × u)(t = 0) = O(ε). Then, for all time, we have (∇ × u)(t) = O(ε) and the following system is formally equivalent to (25), (26), up to O(ε2 ) terms: ∂t n + div(nu) = 0 (34) 3 X ∂xj (nui Bj ) + n(uj − Bj )∂xi Bj + n∂xi (V − A) = εSei (35) ∂t (nui ) + j=1 where Sei = 3 X 3 X ∂xj (∂xj uk Pik ) + ∂xj (∂xi uk Pjk ) + ∂x2j xk (Pij uk ) j=1 k=1 + 3 X ∂xk (nuk )∂xi A − n∂xi (∂t A) − 3 3 ~2 X X ∂xj ∂xk (n∂xi ∂xj uk ) 4 k=1 j=1 k=1 (36) and where P is the following tensor: Z eq P= (p − u)(p − u) fn,nu Rd dp . (2π~)d Proof. If B = u + O(ε), then (25), (26) yield ∂t ui + 3 X uk ∂xk ui + ∂xi (V − A) = O(ε). k=1 So 2 ∂t u2 reads 3 ∂t 3 3 XX X u2 =− (uj uk ∂xk uj ) − uj ∂xj (V − A) + O(ε). 2 j=1 k=1 j=1 10 (37) Let us replace B by u in the expression (27) of Si . The terms depending on V vanish: −∂xk (nuk ∂xi V ) − ∂xk (nui ∂xk V ) + ∂xi (−∂xk V ) nuk − n∂xk V ∂xi uk +n∂xi (uk ∂xk V ) + ∂xk (∂xk V nui ) + ∂xk (∂xi V nuk ) = 0. Let X be defined by X = −∂xk ∂xj nuj ui uk − ∂xk nui uj ∂xj uk − ∂xi uj ∂xj uk nuk − ∂xi uk ∂xj (nuk uj ) + ∂xj nuj uk ∂xi uk + n∂xi uk ∂xj uk uj + ∂xj uk ∂xj uk nui + ∂xj uk ∂xi uk nuj . and consider the terms depending on the stress tensor Π: Y = ∂xj (∂xj uk nui uk ) + ∂xj (∂xi uk nuj uk ) + ∂x2j xk (nui uj uk ) + ∂xj (∂xj uk Pik ) + ∂xj (∂xi uk Pjk ) + ∂x2j xk (Pij uk ). Straightforward calculations lead to X + Y = ∂xj (∂xj uk Pik ) + ∂xj (∂xi uk Pjk ) + ∂x2j xk (Pij uk ). Furthermore, the terms depending on A become ∂xk (n ∂xi A uk ) + ∂xk (nui ∂xi A) + ∂xi (∂xk A) nuk +n∂xk A ∂xi uk + ∂xk (nuk ) ∂xi A − n∂xi (uk ∂xk A) −∂xk (∂xk A nui ) − ∂xk (∂xi A nuk ) − n∂xi (∂t A) = ∂xk (nuk )∂xi A − n∂xi (∂t A). Hence (36) follows. Remark 1. The semi-classical limit of (36) leads to the classical isothermal Navier-Stokes equation. Indeed, in that case, Pij , A and B have the expressions [6] Pij = nT δij , A = −T ln(n), B = u. So we obtain 3 X 3 X 3 X ∂xj ∂xj uk Pik + ∂xi uk Pjk = ∂xj (∂xj ui + ∂xi uj )nT , k=1 j=1 j=1 11 T ∂x n div(nu), n i T div(nu) ∂xi A = −div(nu) ∂xi n, n −n∂xi (∂t A) = −T ∂xi (div(nu)) + 3 X ∂xj (div(u Pij )) = T ∂xi div(nu), j=1 and we get finally Sei = 3 X ∂xj (∂xj ui + ∂xi uj )nT . (38) j=1 Appendix. Proof of Lemma 1. Let f be given. Then we have Z dp \ = − ∂ηi ∂ηj Θ(V )f (t, x, 0). pi pj Θ(V )f 3 (2π~) But from (19), it holds that V (t, x + ~2 η) − V (t, x − ~2 η) ˆ \ f (t, x, η). Θ(V )f = ı ~ Hence Z 1 1 ~ ~ pi pj Θ(V )f dp = −ı∂xi ∂xj V (t, x + η) + ∂xj V (t, x − η)fˆ 2 2 2 2 V (t, x + ~2 η) − V (t, x − ~2 η) + ∂ηj fˆ ~ η=0 So Z pi pj Θ(V )f dp ˆ + ∂x V ∂η fˆ (t, x, 0). = −ı ∂ V ∂ f x η j i i j (2π~)3 Then Z pi pj Θ(V )f dp = −∂xj V nui − ∂xi V nuj , (2π~)3 12 . where Z n= dp f , (2π~)3 Z nu = fp dp , (2π~)3 and (31) follows. Proof of Lemma 2. Recall the definition (23) of Q. We claim that, for all i and j, 3 X B2 B2 nuj − ∂xj A + nui ∂xk Qijk = −∂xi A + 2 2 k=1 + 3 X ∂xi Bk Πjk + ∂xj Bk Πik + ∂xk (Bk Πij ) k=1 − 3 ~2 X ∂xk (n ∂xi ∂xj Bk ). 4 (39) k=1 From (39), we deduce (32) and the proof of Lemma 2 is complete. Let us prove (39). To this aim, we recall a useful commutator lemma that was established in [5]. Let us first introduce some notations. If α = (α1 , α2 , α3 ) ∈ N3 is a multi-index (with N the set of natural integers), we denote pα = pα1 1 pα2 2 pα3 3 , ∂xα = ∂/∂xα1 1 ∂/∂xα2 2 ∂/∂xα3 3 , |α| = α1 + α2 + α3 , α1 X α2 X α3 X X α1 α α2 α3 = = and , γ γ1 γ2 γ3 γ1 =0 γ2 =0 γ3 =0 0≤γ≤α αi where are the binomial coefficients. γi Lemma 4 (Lemma 3.4 of [5]). Let α = (α1 , α2 , α3 ) ∈ N3 and β = (β1 , β2 , β3 ) ∈ N3 be two multi-indices and let λ(x) and µ(x) be any smooth real or complex valued functions. Let us denote [λpα , µpβ ]~ the symbol associated to the commutator of the operators W −1 (λpα ) and W −1 (µpβ ), i.e.: h i [λpα , µpβ ]~ = W W −1 (λpα ) , W −1 µpβ . (40) The following formal expansion holds: b(|α+β|−1)/2c α β [λp , µp ]~ = X 2k+1 ı~ k=0 13 k 1 [λpα , µpβ ]2k+1 , − 4 (41) with α β [λp , µp ]2k+1 = X |γ| (−1) α γ β ζ (∂xζ λ) (∂xγ µ) pα+β−γ−ζ , 0≤γ≤α, 0≤ζ≤β |γ+ζ|=2k+1 (42) where b·c denotes the floor function. Let µ(x) be a given test function. From Lemma 4, we compute: 3 X ∂xk µpi pj pk = k=1 ı |p|2 , µpi pj . ~ 2 ~ (43) Hence, an integration by parts gives Z µ 3 X ∂xk Qijk dx = − k=1 3 Z X eq ∂xk µpi pj pk fn,nu k=1 dxdp (2π~)3 |p|2 dxdp eq , µpi pj fn,nu 2 (2π~)3 2 ~ Z ~ ı dxdp −1 eq − ∆, W (µpi pj ) fn,nu = − W . ~ 2 (2π~)3 ı = − ~ Z Now, using the Parseval property (10) and observing that ~2 B2 −1 − ∆ = H(A, B) − W −B · p + A + , 2 2 where H(A, B) is defined by (16), we get Z 3 X ı ~ −1 eq ∂xk Qijk dx = − Tr − ∆, W (µpi pj ) %n,nu µ ~ 2 k=1 B2 ı −1 −1 eq = − Tr H(A, B) − W −B · p + A + , W (µpi pj ) %n,nu ~ 2 ı B2 = Tr W −1 −B · p + A + , W −1 (µpi pj ) %eq n,nu ~ 2 Z ı B2 dxdp eq = −B · p + A + , µpi pj fn,nu ~ 2 (2π~)3 ~ where we used the cyclicity of the trace and the crucial fact that, by (15), %eq n,nu commutes with the operator H(A, B). 14 i h 2 Let us now compute −B · p + A + B2 , µpi pj . From Lemma 4, we ~ deduce B2 B2 B2 A+ p j + ∂ xj A + pi µ, , µpi pj = ı~ ∂xi A + 2 2 2 ~ and that [B · p, µpi pj ]~ = ı~[B · p, µpi pj ]1 − ıh3 [B · p, µpi pj ]3 , 4 with [B · p, µpi pj ]1 = = 3 X k=1 3 X [Bk pk , µpi pj ]1 ∂xi Bk µ pj pk + ∂xj Bk µ pi pk − Bk ∂xk µ pi pj k=1 and [B · p, µpi pj ]3 = 3 X [Bk pk , µpi pj ]3 = − 3 X ∂xi ∂xj Bk ∂xk µ. k=1 k=1 Hence, −[B · p, µpi pj ]~ = −ı~ − ı~ 3 X ∂xi Bk pj pk + ∂xj Bk pi pk µ k=1 3 X ~2 k=1 4 15 ∂xi ∂xj Bk − Bk pi pj ∂xk µ. We get finally, after an integration by parts, Z µ 3 X ∂xk Qijk dx k=1 Z B2 dxdp B2 eq ∂ xi A + =− p j + ∂ xj A + pi µfn,nu 2 2 (2π~)3 Z X 3 eq dxdp + ∂xi Bk pj pk + ∂xj Bk pi pk µfn,nu (2π~)3 k=1 2 Z X 3 ~ dxdp eq µ − ∂xk ∂xi ∂xj Bk − Bk pi pj fn,nu 4 (2π~)3 k=1 Z B2 B2 ∂ xi A + =− nuj + ∂xj A + nui µdx 2 2 Z X 3 + ∂xi Bk Πjk + ∂xj Bk Πik µdx k=1 − Z X 3 ∂xk k=1 ~2 n∂xi ∂xj Bk − Bk Πij µdx. 4 Therefore, since this expression holds true for all function µ(x), one can identify (39). Proof of Lemma 3. From (24), we deduce (g1 )i = ∂t (divΠ)i = 3 X ∂xk (∂t (nui ) Bk ) + k=1 + 3 X k=1 3 X ∂xk (nui ∂t Bk ) + k=1 3 X ∂xi ∂t Bk nuk k=1 B2 B2 ∂xi Bk ∂t (nuk ) − ∂t n ∂xi A + − n ∂xi ∂t A + . 2 2 (44) But according to (21), (22), we have ∂t n = −div(nu), 3 3 X X ∂t (nui ) = − ∂xj (nui Bj ) − n ∂xj (uj − Bj )∂xi Bj − n∂xi (V − A) + O(ε). j=1 j=1 16 So, neglecting the O(ε) terms, we get the following relations: ∂t (nui )Bk = − 3 X ∂xj (nui Bj )Bk − n j=1 3 X ∂xj (uj − Bj )∂xi Bj Bk j=1 − n∂xi (V − A)Bk , 3 X ∂xi Bk ∂t (nuk ) = k=1 3 X k=1 ∂xi Bk − 3 X ∂xj (nuk Bj ) − n j=1 3 X ∂xj (uj − Bj ) ∂xk Bj j=1 − n ∂xk (V − A) and 3 X 1 1 (nuk )∂xi A + B 2 ∂t n∂i A + B 2 = − 2 2 k=1 which enable to simplify (44) and gives (33). Acknowledgement. The authors were supported by the Agence Nationale de la Recherche, ANR project QUATRAIN. References [1] C. Bardos, F. Golse, D. Levermore, Fluid dynamic limits of kinetic equations. I. Formal derivations. J. Statist. Phys. 63 (1991), no. 1-2, 323–344. [2] J.-P. Bourgade, P. Degond, F. Méhats, C. Ringhofer, On quantum extensions to classical Spherical Harmonics Expansion / Fokker-Planck models, J. Math. Phys. 47 (2006), no. 4, 043302 [3] L. Chen, M. Dreher, The viscous model of quantum hydrodynamics in several dimensions, Math. Models Methods Appl. Sci. 17 (2007) 1065 1093. [4] P. Degond, S. Gallego, F. Méhats, An entropic Quantum Drift-Diffusion model for electron transport in resonant tunneling diodes, J. Comput. Phys. 221 (2007), 226–249. 17 [5] P. Degond, S. Gallego, F. Méhats, On Quantum Hydrodynamic and Quantum Energy Transport Models, Commun. Math.Sci. 5 (2007), no. 4, 887-908. [6] P. Degond, S. Gallego, F. Méhats, Isothermal quantum hydrodynamics: derivation, asymptotic analysis and simulation, SIAM Multiscale Model. Simul. 6 (2007), no.1, 246-272. [7] P. Degond, F. Méhats, C. Ringhofer, Quantum hydrodynamic models derived from the entropy principle, Contemp. Math. 371 (2005), 107– 331. [8] P. Degond, F. Méhats, C. Ringhofer, Quantum energy-transport and drift-diffusion models, J. Stat. Phys. 118 (2005), no. 3-4, 625–665. [9] P. Degond, C. Ringhofer, Quantum Moment Hydrodynamics and the Entropy Principle, J. Stat. Phys. 112 (2003), no. 3-4, 587–628. [10] D. Ferry, J.-R. Zhou, Form of the quantum potential for use in hydrodynamic equations for semiconductor device modeling, Phys. Rev. B 48 (1993) 7944-7950. [11] S. Gallego, F. Méhats, Entropic discretization of a quantum driftdiffusion model, SIAM J. Numer. Anal. 43 (2005), no. 5, 1828–1849. [12] C. Gardner, The quantum hydrodynamic model for semiconductor devices, SIAM J. Appl. Math. 54 (1994), no. 2, 409–427. [13] M. P. Gualdani and A. Jüngel, Analysis of the viscous quantum hydrodynamic equations for semiconductors. Europ. J. Appl. Math. 15 (2004), 577–595. [14] A. Jüngel, D. Matthes, A derivation of the isothermal quantum hydrodynamic equations using entropy minimization, Z. Angew. Math. Mech. 85 (2005), 806–814. [15] A. Jüngel, D. Matthes and J.P. Milisic, Derivation of new quantum hydrodynamic equations using entropy minimization, SIAM J. Appl. Math. 67 (2006), 46–68. [16] C. D. Levermore, Moment closure hierarchies for kinetic theories, J. Stat. Phys. 83 (1996), 1021–1065. 18