Abstract
We establish a connection between the relative Classical entropy and the relative Fermi–Dirac entropy, allowing to transpose, in the context of the Boltzmann or Landau equation, any entropy–entropy production inequality from one case to the other; therefore providing entropy–entropy production inequalities for the Boltzmann–Fermi–Dirac operator, similar to the ones of the Classical Boltzmann operator. We also provide a generalized version of the Csiszár–Kullback–Pinsker inequality to weighted \(L^p\) norms, \(1 \le p \le 2\) and a wide class of entropies.
Similar content being viewed by others
References
Alonso, R., Bagland, V., Desvillettes, L., Lods, B.: About the use of entropy production for the Landau–Fermi–Dirac equation. J. Stat. Phys. 183(1), 27 (2021)
Alonso, R., Bagland, V., Lods, B.: Long time dynamics for the Landau–Fermi–Dirac equation with hard potentials. J. Differ. Equ. 270, 596–663 (2021)
Alonso, R., Bagland, V., Lods, B.: Long time dynamics for the Landau–Fermi–Dirac equation with hard potentials. J. Differ. Equ. 270, 596–663 (2021)
Alonso, R., Bagland, V., Desvillettes, L., Lods, B.: About the Landau–Fermi–Dirac equation with moderately soft potentials. Arch. Ration. Mech. Anal. 244(3), 779–875 (2022)
Bobylev, A.V., Cercignani, C.: On the rate of entropy production for the Boltzmann equation. J. Stat. Phys. 94(3–4), 603–618 (1999)
Bolley, F., Villani, C.: Weighted Csiszár–Kullback–Pinsker inequalities and applications to transportation inequalities. Ann. Fac. Sci. Toulouse Math. (6) 14(3), 331–352 (2005)
Breden, M., Desvillettes, L.: Rigorous study of the equilibria of collision kernels appearing in the theory of weak turbulence. Arch. Ration. Mech. Anal. 235(2), 1151–1176 (2020)
Cáceres, M.J., Carrillo, J.A., Dolbeault, J.: Nonlinear stability in \(L^p\) for a confined system of charged particles. SIAM J. Math. Anal. 34(2), 478–494 (2002)
Cáceres, M.J., Carrillo, J.A., Dolbeault, J.: A generalization of the Csiszár–Kullback inequality. https://www.ceremade.dauphine.fr/~dolbeaul/Teaching/files/SNS-2012/Csiszar-Kullback.pdf
Carrapatoso, K., Desvillettes, L., He, L.: Estimates for the large time behavior of the Landau equation in the Coulomb case. Arch. Ration. Mech. Anal. 224(2), 381–420 (2017)
Cercignani, C.: \(H\)-theorem and trend to equilibrium in the kinetic theory of gases. Arch. Mech. (Arch. Mech. Stos.) 34(3), 231–241 (1983)
Chapman, S., Cowling, T.G.: The Mathematical Theory of Non-uniform Gases. Cambridge University Press, Cambridge (1939)
Desvillettes, L.: Entropy dissipation estimates for the Landau equation in the Coulomb case and applications. J. Funct. Anal. 269(5), 1359–1403 (2015)
Desvillettes, L.: Structure entropique du noyau de collision de Landau. In Séminaire Laurent Schwartz—Équations aux dérivées partielles et applications. Année 2014–2015, pp. Exp. No. XIV, 9. Ed. Éc. Polytech., Palaiseau (2016)
Desvillettes, L.: Some remarks about the link between the Fisher information and Landau or Landau-Fermi-Dirac entropy dissipation. arXiv:2401.02877 (2024)
Desvillettes, L., Villani, C.: On the spatially homogeneous Landau equation for hard potentials. II. \(H\)-theorem and applications. Commun. Partial Differ. Equ. 25(1–2), 261–298 (2000)
Desvillettes, L., Mouhot, C., Villani, C.: Celebrating Cercignani’s conjecture for the Boltzmann equation. Kinet. Relat. Models 4(1), 277–294 (2011)
Dolbeault, J.: Kinetic models and quantum effects: a modified Boltzmann equation for Fermi–Dirac particles. Arch. Rational Mech. Anal. 127(2), 101–131 (1994)
Gilardoni, G.L.: On Pinsker’s and Vajda’s type inequalities for Csiszár’s \(f\)-divergences. IEEE Trans. Inform. Theory 56(11), 5377–5386 (2010)
Jüngel, A.: Entropy Methods for Diffusive Partial Differential Equations, vol. 804. Springer, Cham (2016)
Lu, X.: A modified Boltzmann equation for Bose–Einstein particles: isotropic solutions and long-time behavior. J. Stat. Phys. 98(5–6), 1335–1394 (2000)
Lu, X.: On spatially homogeneous solutions of a modified Boltzmann equation for Fermi–Dirac particles. J. Stat. Phys. 105(1–2), 353–388 (2001)
Lu, X.: On isotropic distributional solutions to the Boltzmann equation for Bose–Einstein particles. J. Stat. Phys. 116(5), 1597–1649 (2004)
Lu, X., Wennberg, B.: On stability and strong convergence for the spatially homogeneous Boltzmann equation for Fermi–Dirac particles. Arch. Ration. Mech. Anal. 168(1), 1–34 (2003)
Nordhiem, L.W.: On the kinetic method in the new statistics and application in the electron theory of conductivity. Proc. R. Soc. Lond. Ser. A 119(783), 689–698 (1928)
Toscani, G., Villani, C.: Sharp entropy dissipation bounds and explicit rate of trend to equilibrium for the spatially homogeneous Boltzmann equation. Commun. Math. Phys. 203(3), 667–706 (1999)
Toscani, G., Villani, C.: On the trend to equilibrium for some dissipative systems with slowly increasing a priori bounds. J. Stat. Phys. 98(5–6), 1279–1309 (2000)
Uehling, E.A., Uhlenbeck, G.E.: Transport phenomena in Einstein–Bose and Fermi–Dirac gases. I. Phys. Rev. 43(7), 552 (1933)
Villani, C.: A review of mathematical topics in collisional kinetic theory. In Handbook of mathematical fluid dynamics, Vol. I, pp. 71–305. North-Holland, Amsterdam (2002)
Villani, C.: Cercignani’s conjecture is sometimes true and always almost true. Commun. Math. Phys. 234(3), 455–490 (2003)
Acknowledgements
I thank Bertrand Lods with whom I had fruitful discussions and Matthieu Dolbeault who suggested to me the proof of the nice inequality (92).
I have no competing interests to declare that are relevant to the content of this article.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Clement Mouhot.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Data sharing not applicable to this article as no datasets were generated or analysed during the current study
Appendices
Appendix A Similar Results in the Bose–Einstein Case
An upper-bound inequality similar to (34) can also be obtained in the Bose–Einstein case. This latter case formally corresponds to taking \(- \varepsilon \) instead of \(\varepsilon \) in our formulas. First define for any \(x \in {\mathbb {R}}_+\) and \(\varepsilon > 0\),
and the Bose–Einstein entropy of \(0 \le f \in L^1_2({\mathbb {R}}^3)\):
Lu proved in [21] that, under the condition
where \(\zeta \) is the Riemann Zêta function and \(T_c\) is called the critical temperature, there exists a unique \(\varepsilon \)-Bose–Einstein statistics \({\mathcal {M}}_{\varepsilon }^{BE,f}\) associated to f - that is a distribution such that \(\log \varphi ^{BE}_{\varepsilon } ({\mathcal {M}}_{\varepsilon }^{BE,f})\) is a linear combination of \(v\mapsto 1\), \(v\mapsto v\) and \(v\mapsto |v|^2\) and sharing the same normalization in \(v\mapsto 1\), \(v\mapsto v\) and \(v\mapsto |v|^2\) as f. We can obtain from Proposition 15 in Appendix B that, denoting \(H^{BE}_{\varepsilon }[f|{\mathcal {M}}_{\varepsilon }^{BE,f}] := H^{BE}_{\varepsilon }(f) - H^{BE}_{\varepsilon }({\mathcal {M}}_{\varepsilon }^{BE,f})\), we have
where the last equality comes from \({\Phi ^{BE}_{\varepsilon }}'' = \frac{{\varphi ^{BE}_{\varepsilon }}'}{\varphi ^{BE}_{\varepsilon }}\) and the change of variables \(x = {\mathcal {M}}_{\varepsilon }^{BE,f}(v) + \tau (f(v) - {\mathcal {M}}_{\varepsilon }^{BE,f})\).
In the following Proposition, we provide a link between the relative entropies to equilibrium of the Bose–Einstein and the classical cases. Although I believe that both inequalities could be obtained, we only present here the “upper-bound” inequality as its proof is rather short. Further work may allow to obtain the lower-bound inequality, with a constant that probably depends on an \(L^{\infty }\) bound on f. This constitutes another reason why we did not investigate this other inequality, as, although \(L^{\infty }\) bounds are natural to use in the Fermi–Dirac context, due to Pauli’s exclusion principle, they are not in the Bose–Einstein one, due to the phenomenon of condensation.
Proposition 12
Upper-bound in the Bose–Einstein case. For any \(\varepsilon > 0\) and nonnegative \(f \in L^1_2({\mathbb {R}}^3) \cap L \log L ({\mathbb {R}}^3) \setminus \{0\}\) which density and temperature satisfy (101), we have
Proof
The starting point of the proof are Equation (102) and the inequality \(|y-z| \ge |\varphi ^{BE}_{\varepsilon }(y) - \varphi ^{BE}_{\varepsilon }(z)|\) for all \((y,z) \in {\mathbb {R}}_+^2\), yielding
Applying the change of variables \(y = \varphi ^{BE}_{\varepsilon }(x)\) and using formula (27), we obtain
Remark that, as \(\log M^{\varphi ^{BE}_{\varepsilon }(f)} - \log \varphi ^{BE}_{\varepsilon }({\mathcal {M}}_{\varepsilon }^{BE,f})\) is a linear combination of conserved quantities (namely, \(v \mapsto 1\), \(v \mapsto v\) and \(v \mapsto |v|^2\)), we have by definition of \(M^{\varphi ^{BE}_{\varepsilon }(f)}\) that
so that
ending the proof. \(\square \)
Proposition 13
Bose–Einstein CKP inequality. Let \(\varpi : {\mathbb {R}}^3 \rightarrow {\mathbb {R}}_+\) be measurable, and \(r \in [1,2]\). We recall the definition (88) of the function \(\Lambda \),
Then for any \(\varepsilon > 0\) and \(0\le f \in L^1_2({\mathbb {R}}^3) \cap L \log L ({\mathbb {R}}^3) \setminus \{0\}\) which density and temperature satisfy (101), assuming that the norms below are finite,
where we denoted \({\mathcal {M}}\equiv {\mathcal {M}}_{\varepsilon }^{BE,f}\) the \(\varepsilon \)-Bose–Einstein distribution associated to f, and \(H^{BE}_{\varepsilon }\) is the Bose–Einstein entropy. When \(r=2\), \(L^{\frac{r}{2-r}}\) shall be understood as \(L^{\infty }\).
Proof
The proof is similar to the one of Proposition 10. Let \(\varepsilon > 0\).
We apply Corollary 17 with \(\Phi (x) \equiv \Phi ^{BE}_{\varepsilon }(x) = \int _0^x \log \frac{y}{1 + \varepsilon y} \, \textrm{d}y\), \(J = {\mathbb {R}}_+^*\) and
with \(\rho , T\) and \(\varepsilon \) such that (101) is satisfied, ensuring the existence of \({\mathcal {M}}_{\varepsilon }^{BE,f}\). In the rest of this proof, we denote for the sake of clarity \({\mathcal {M}}\equiv {\mathcal {M}}_{\varepsilon }^{BE,f}\). Then (114) writes, since \(H_{\Phi ^{BE}_{\varepsilon }} \equiv H^{BE}_{\varepsilon }\) the \(\varepsilon \)-Bose–Einstein entropy, and \(\displaystyle \frac{1}{{\Phi ^{BE}_{\varepsilon }}''}(x) = x(1+\varepsilon x)\) is convex,
We focus on the term with the integral in \(\tau \). From Minkowski’s inequality, it is smaller than
where \(\Lambda \) is defined in (88) and appears thanks to a Taylor expansion of \(\lambda \mapsto \lambda \log \lambda - \lambda \) around 1 like in the proof of Proposition 10, allowing to conclude. \(\square \)
From the previous proposition, we easily deduce the following standard inequalities.
Corollary 14
Standard Bose–Einstein CKP inequalities. For any \(\varepsilon > 0\), \(\alpha \ge 0\) and \(0\le f \in L^1_2({\mathbb {R}}^3) \cap L \log L ({\mathbb {R}}^3) \setminus \{0\}\) satisfying (101),
where we denoted for clarity \({\mathcal {M}}\equiv {\mathcal {M}}_{\varepsilon }^{BE,f}\) the \(\varepsilon \)-Bose–Einstein distribution associated to f.
We prove the above inequalities similarly as we did for Corollary 11. For (105) we apply Proposition 13 with \(r = 1\) and \(\varpi (v) = (1+|v|^2)^{\frac{\alpha }{2}} \, \textbf{1}_{f \le {\mathcal {M}}}\) and notice that \(\Lambda \le 2\) on [0, 1]. We then obtain (106)–(107) by decomposing \(|f-{\mathcal {M}}| = f - {\mathcal {M}}+ 2({\mathcal {M}}-f)_+\), using the fact that f and \({\mathcal {M}}\) share the same normalization in \(v\mapsto 1\), \(v\mapsto v\) and \(v\mapsto |v|^2\), and (105) with respectively \(\alpha = 0\) and \(\alpha =2\).
Appendix B A General Discussion About Entropies and Equilibria
In this section we intend to provide general considerations on the entropy, which are much more general than the scope of this paper, but give a good understanding of the notions we used, and could also be helpful in the study of weak turbulence, where various kinds of unusual entropies can emerge (see [7]). Our setting is laid down quite generally. Consider a measured space \(({\mathcal {E}},{\mathcal {A}},\mu )\), an open interval \(J \subset {\mathbb {R}}\) which closure we denote by \(\bar{J}\), and a function \(\Phi \in {\mathcal {C}}^2(J) \cap {\mathcal {C}}^0(\bar{J})\) such that \(\Phi '' > 0\) on J. Remark that \(\Phi '\) is then a \({\mathcal {C}}^1\)-diffeomorphism from J onto \(\Phi '(J)\).
Entropy. We define the \(\Phi \)-entropy, for any (\({\mathcal {A}}\), Bor\((\bar{J})\))-measurable \(f : {\mathcal {E}}\rightarrow \bar{J}\) such that the following integral makes sense and is finite, by
and we denote by \(E_{\Phi }\) the set of such f. We let the relative \(\Phi \)-entropy of f and g to be
We also define, for (\({\mathcal {A}}\), Bor\((\bar{J})\))-measurable \(f,g : {\mathcal {E}}\rightarrow \bar{J}\), the \(\Phi \)-relative-entropy (which in general differs from the relative \(\Phi \)-entropy) of f and g by
Note that \({\mathcal {H}}_{\Phi }[f|g]\) is possibly infinite, but always well-defined, as
and that \({\mathcal {H}}_{\Phi }[f|g]\) is always nonnegative. The following Proposition 15 gives a quite simple but general result linking entropies, conserved quantities and equilibrium distributions, under a sole existence assumption.
Proposition 15
Let I be a countable set, \((\phi _i)_{i \in I}\) a family of measurable real functions, \((\omega _i)_{i \in I}\) a family of real numbers, and
Assume that \(\Phi '(J) = {\mathbb {R}}\), and that there exists \((\alpha _i(\omega )) \in {\mathbb {R}}^I\) such that \(M^{{\mathcal {F}}}_{\Phi } \in {\mathcal {F}}\), where
Then the following four propositions are equivalent. Let \(g \in {\mathcal {F}}\).
In particular, \(M^{{\mathcal {F}}}_{\Phi }\) is the unique minimizer of \(H_{\Phi }\) under the constraints of the set \({\mathcal {F}}\).
The above proposition actually proves the following (under the assumptions \(\Phi '(J) = {\mathbb {R}}\) and of existence of \(M^{{\mathcal {F}}}_{\Phi }\)). An admissible distribution g is an equilibrium relative to the conserved quantities \(\phi _i\), in the sense of the minimization of the \(\Phi \)-entropy [(iii)], if and only if \(\Phi '(g)\) is a linear combination of the functions \(\phi _i\) [(iv)], if and only if the relative \(\Phi \)-entropy between any admissible distribution f and g is given by (110) [(i)], if and only if the quantity \(\Phi '(g)\) is conserved amongst all admissible distributions [(ii)] - indeed, (ii) is equivalent, assuming the following integrals make sense, to \(\forall \, f_1,f_2 \in {\mathcal {F}}\), \(\int _{{\mathcal {E}}} f_1 \, \Phi '(g) \, \textrm{d}\mu (\zeta ) = \int _{{\mathcal {E}}} f_2 \, \Phi '(g) \, \textrm{d}\mu (\zeta )\).
The reader may notice that the Classical case corresponds to the choice \(\Phi (x) \equiv \Phi _0(x) = x \log x - x\), for which \((\Phi '_0)^{-1} = \exp \), hence \(M^{{\mathcal {F}}}_{\Phi }\) is in this case a Maxwellian, since the conserved quantities (corresponding to the functions \(\phi _i\) in the proposition) are \(v \mapsto 1\), \(v \mapsto v\) and \(v \mapsto |v|^2\).
Moreover, the Fermi–Dirac case corresponds to the choice \(\Phi (x) \equiv \Phi _{\varepsilon }(x) \equiv \int _0^x \log \frac{y}{1 - \varepsilon y} \, \textrm{d}y\), for which \(\displaystyle (\Phi '_{\varepsilon })^{-1}(x) = \frac{e^x}{1 + \varepsilon e^x}\), hence \(M^{{\mathcal {F}}}_{\Phi }\) is in this case a Fermi–Dirac distribution, since again, the conserved quantities (corresponding to the functions \(\phi _i\) in the proposition) are \(v \mapsto 1\), \(v \mapsto v\) and \(v \mapsto |v|^2\). Proposition 1 then comes as a corollary to Proposition 15 with \({\mathcal {E}}= {\mathbb {R}}^3\) endowed with the Lebesgue measure, \(J = (0,\varepsilon ^{-1})\), \(\displaystyle \Phi (x) \equiv \Phi _{\varepsilon }(x) \equiv \int _0^x \log \varphi _{\varepsilon }(y) \, \textrm{d}y\), \(M^{{\mathcal {F}}}_{\Phi } \equiv {{\mathcal {M}}_{\varepsilon }^f}\) and
with \(\rho , T\) and \(\varepsilon \) such that \(\displaystyle \gamma > \frac{2}{5}\) (ensuring the existence of \({{\mathcal {M}}_{\varepsilon }^f}\), as proven in [22]).
Remark 6
The Bose–Einstein case is also recovered with \(\displaystyle \Phi (x) \equiv \Phi ^{BE}_{\varepsilon }(x) \equiv \int _0^x \log \frac{y}{1 + \varepsilon y} \, \textrm{d}y\), for which \(\displaystyle (\Phi '_{-\varepsilon })^{-1}(x) = \frac{e^x}{1 - \varepsilon e^x}\), hence \(M^{{\mathcal {F}}}_{\Phi }\) is in this case a Bose–Einstein distribution (when it exists).
Proof
We start by proving \((i) \iff (ii)\). Let \(f,g \in {\mathcal {F}}\) such that \(g \in J\) \(\mu \)-almost everywhere and \({\mathcal {H}}_{\Phi }[f|g]~<~\infty \). Using a Taylor expansion followed by Fubini’s Theorem, we get
thus proving the announced equivalence. Remark that in the last equality we used the fact that \({\mathcal {H}}_{\Phi }[f|g] < \infty \) to ensure that \(\int _{g \in J} (f-g) \Phi '(g) \, \textrm{d}\mu (\zeta )\) is well-defined. We now show \((iv) \implies (ii)\). Since \(\textrm{Im}({\Phi '}^{-1}) = J\), we do have \(M^{{\mathcal {F}}}_{\Phi } \in J\) \(\mu \)-almost everywhere, and
where the last equality comes from the fact that both f and \(M^{{\mathcal {F}}}_{\Phi }\) belong to \({\mathcal {F}}\). We now focus on \((ii) \implies (iv)\). Assume the existence of \(g \in {\mathcal {F}}\) such that \(g \in J\) \(\mu \)-almost everywhere and
Since we also have \(M^{{\mathcal {F}}}_{\Phi } \in {\mathcal {F}}\), then \(M^{{\mathcal {F}}}_{\Phi } \in J\) \(\mu \)-almost everywhere and we just proved that
allowing to deduce that
Since \(\Phi '\) is increasing, this implies that \(g = M^{{\mathcal {F}}}_{\Phi }\) \(\mu \)-almost everywhere. We now focus on \((iv)~\implies ~(iii)\). Since we already proved \((iv) \implies (ii) \implies (i)\), we have for any \(f \in {\mathcal {F}}\),
thus
Finally, we prove \((iii) \implies (iv)\). Assume \(H_{\Phi }(g) = \underset{h \in {\mathcal {F}}}{\min }\ \; H_{\Phi }(h)\). We just proved that \(H_{\Phi }(M^{{\mathcal {F}}}_{\Phi }) = \underset{h \in {\mathcal {F}}}{\min }\ \; H_{\Phi }(h)\), hence \(H_{\Phi }(g) = H_{\Phi }(M^{{\mathcal {F}}}_{\Phi })\) and \(H_{\Phi }[g|M^{{\mathcal {F}}}_{\Phi }] = 0\). Since we also proved \((iv) \implies (i)\), we know that
hence \({\mathcal {H}}_{\Phi }[g|M^{{\mathcal {F}}}_{\Phi }] = 0\), that is
This implies, since \(M^{{\mathcal {F}}}_{\Phi } \in J\) and \((1-\tau )M^{{\mathcal {F}}}_{\Phi } + \tau g \in J\) \(\mu \)-almost everywhere for any \(\tau \in (0,1)\), and \(\Phi ''>0\) on J, that \(g=M^{{\mathcal {F}}}_{\Phi }\) \(\mu \) almost everywhere. \(\square \)
1.1 Appendix B.1 General Csiszár–Kullback–Pinsker Inequality
The famous Csiszár–Kullback–Pinsker inequality, linking the squared \(L^1\) distance of two probabilities with their relative Classical entropy can in fact be generalized to the whole family of \(\Phi \)-entropies, and for weighted \(L^r\), \(1 \le r \le 2\) distances, by the following Proposition 16, and more specifically, in the context of convergence towards equilibrium, by Corollary 17. Again, such inequalities may be useful in the study of weak turbulence [7].
Proposition 16
Let some \(({\mathcal {A}}\), \(\textrm{Bor}(\bar{J}))\)-measurable \(\varpi : {\mathcal {E}}\rightarrow \bar{J}\) and \(r \in [1,2]\). Then for any \(({\mathcal {A}}\), \(\textrm{Bor}(\bar{J}))\)-measurable \(f,g : {\mathcal {E}}\rightarrow \bar{J}\) such that all terms below are finite, we have
where \({\mathcal {H}}_{\Phi }\) is the \(\Phi \)-relative-entropy defined in (110).
Proof
Let us recall the definition of \({\mathcal {H}}_{\Phi }[f|g]\), that is
We fix \(\tau \in (0,1)\). Let \(p = \frac{2}{r}\in [1,2]\) and \(q = \frac{2}{2-r} \in [2,+\infty ]\), so that \(\displaystyle \frac{1}{p} + \frac{1}{q} = 1\). By Hölder’s inequality, we have
Raising the above inequality to the power p, we obtain
where we used the fact that \(\Vert \cdot ^{\frac{1}{p}} \Vert ^p_{L^q} = \Vert \cdot \Vert _{L^{\frac{q}{p}}}\). Since \(\frac{2}{p} = r\) and \(\frac{q}{p} = \frac{r}{2-r}\), (113) actually writes
If \(\varpi \) is zero \(\mu \)-almost everywhere on \(\{g \in J\}\), then the proposition is trivial. Else, since \(\Phi '' > 0\) on \(\{g \in J \}\) and \((1-\tau )g + \tau f \in J\) on \(\{g \in J\}\) for any \(\tau \in (0,1)\), we know that \(\left\| \Phi ''((1-\tau )g + \tau f) \, \varpi ^2 \right\| _{L^{\frac{r}{2-r}}(g \in J)} > 0\). Since we assumed that the quantity
is finite, we also know that for almost every \(\tau \in (0,1)\) we have \(\displaystyle \left\| \frac{\varpi ^2}{\Phi ''((1-\tau )g + \tau f)} \right\| _{L^{\frac{r}{2-r}}(g \in J)}\) \( < \infty \). For these values of \(\tau \), we then have
Integrating in \(\tau \) yields
Since, again by hypothesis, the integral in \(\tau \) is nonzero (its inverse is finite), we obtain (112). \(\square \)
Remarking that \(\{M_{\Phi }^{{\mathcal {F}}} \in J\} = {\mathcal {E}}\) and that for any \(f\in {\mathcal {F}}\), \({\mathcal {H}}_{\Phi }[f|M_{\Phi }^{{\mathcal {F}}}] = H_{\Phi }[f|M_{\Phi }^{{\mathcal {F}}}]\) (see Proposition 15), we straightforwardly obtain the following corollary.
Corollary 17
Let some \(({\mathcal {A}}\), \(\textrm{Bor}(\bar{J}))\)-measurable \(\varpi : {\mathcal {E}}\rightarrow \bar{J}\) and \(r \in [1,2]\). With the same notations as in Proposition 15, assuming \(M_{\Phi }^{{\mathcal {F}}}\) exists, we have for any \(f \in {\mathcal {F}}\) such that the integral term below is finite,
where \(H_{\Phi }\) is defined in (108)–(109) and \(M_{\Phi }^{{\mathcal {F}}}\) is the equilibrium associated to \(\Phi \) and the set \({\mathcal {F}}\), defined in (111).
Appendix C Technical Results
In this section, we consider \({{\mathcal {M}}_{\varepsilon }^f}\), the Fermi–Dirac distribution associated to some \(\varepsilon >0 \) and \(0 \le f \in L^1_2({\mathbb {R}}^3)\) such that \(1 - \varepsilon f \ge 0\). The existence of \({{\mathcal {M}}_{\varepsilon }^f}\) is provided by assuming \(\displaystyle \gamma > \frac{2}{5}\) (see [22]), where we recall the notation
where \(\displaystyle T_F(\rho , \varepsilon ) = \frac{1}{2} \left( \frac{3 \rho \varepsilon }{4 \pi } \right) ^{2/3}\) is the Fermi temperature associated to \(\rho \) and \(\varepsilon \); and \(\rho , T\) are respectively the density and temperature associated to the distribution f, defined in (12). We also recall the notation, for \(x \in [0,\varepsilon ^{-1})\),
1.1 Appendix C.1 \(L^\infty \) bound for the Fermi–Dirac statistics
In this subsection, we provide an \(L^\infty \) bound on the Fermi–Dirac statistics. The following result is very similar to [2, Lemma A.1].
Proposition 18
Let \(\varepsilon >0\) and \(f \in L^1_2({\mathbb {R}}^3)\) be a nonnegative distribution such that \(1 - \varepsilon f \ge 0\) and \(\displaystyle \gamma \ge \gamma ^{\dag }\), where \(\gamma \) is given by (115) and
Then the quantity \(\varepsilon \Vert \varphi _{\varepsilon }({{\mathcal {M}}_{\varepsilon }^f}) \Vert _{\infty }\) satisfies
Proof
Our proof is based on [22, proof of Proposition 3]. We introduce, for \(s \ge 0\),
It is proven in [22, proof of Proposition 3] that P is continuous and increasing on \({\mathbb {R}}_+\), and that
Let \(\alpha > 0\) and
As, for any \(r \ge 0\), it holds that
we have
that is
We define
Then, whenever \(\gamma \ge \gamma ^{\alpha }\), we have \(\displaystyle t \ge \frac{1}{\alpha ^{2/5} - 1}\), so that \(1 + t \le \alpha ^{2/5} \, t\), implying
Since P is increasing, we deduce that, whenever \(\gamma \ge \gamma ^{\alpha }\), we have
that is
By computing the derivative of \(\displaystyle \alpha \mapsto \frac{\alpha }{\alpha ^{2/5}-1}\), we can minimize \(\alpha \mapsto \gamma ^{\alpha }\) and find that the minimum value is
reached for \(\alpha ^{\dag } = \left( \frac{5}{3} \right) ^{5/2}\), and, combined with (118), this proves (117). \(\square \)
1.2 Appendix C.2 Regularity in \(\varepsilon \) of the Coefficients of the Fermi–Dirac Statistics
In this last subsection, for any \(\varepsilon \ge 0\) and \(0 \le g \in L^1_2({\mathbb {R}}^3) \cap L \log L ({\mathbb {R}}^3)\), we denote by \({\mathcal {M}}_{\varepsilon } \equiv {\mathcal {M}}_{\varepsilon }^{\varphi _{\varepsilon }^{-1}(g)}\) the \(\varepsilon \)-Fermi distribution associated to \(\varphi _{\varepsilon }^{-1}(g)\). In particular, in the limit case \(\varepsilon = 0\), \({\mathcal {M}}_0\) is the Maxwellian distribution associated to g. We also denote \(a_{\varepsilon }, b_{\varepsilon }, \bar{u}_{\varepsilon }\) and \(\rho _{\varepsilon }, u_{\varepsilon },T_{\varepsilon }\) the quantities such that, letting
we have, for any \(v \in {\mathbb {R}}^3\),
and
Lemma 19
Let \(0 \le g \in L^1_2({\mathbb {R}}^3) \cap L \log L({\mathbb {R}}^3)\). Using the notation (121), the application \(\varepsilon \mapsto (\rho _{\varepsilon },u_{\varepsilon },T_{\varepsilon })\) is continuous on \({\mathbb {R}}_+\) and \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\).
Proof
The continuity of \(\varepsilon \mapsto (\rho _{\varepsilon },u_{\varepsilon },T_{\varepsilon })\) on \({\mathbb {R}}_+\) comes by dominated convergence, as it holds for any \(\varepsilon \ge 0\) and \(v \in {\mathbb {R}}^3\) that
and by hypothesis \(0 \le g \in L^1_2({\mathbb {R}}^3)\). Similarly, for any \(\varepsilon > 0\) and \(v \in {\mathbb {R}}^3\) we have
Therefore, for any \(\varepsilon > 0\), we have
The differentiability of \(\varepsilon \mapsto (\rho _{\varepsilon },u_{\varepsilon },T_{\varepsilon })\) on \({\mathbb {R}}_+^*\) then comes by dominated convergence, as \(g \in L^1_2({\mathbb {R}}^3)\). \(\square \)
The following lemmas provide the continuity of \(\varepsilon \mapsto (a_{\varepsilon }, b_{\varepsilon }, \bar{u}_{\varepsilon })\) at the point \(\varepsilon = 0\) and its differentiability on \({\mathbb {R}}_+^*\).
Lemma 20
Let \(0 \le g \in L^1_2({\mathbb {R}}^3) \cap L \log L({\mathbb {R}}^3)\). Using the notations (119)–(120), the application \(\varepsilon \mapsto (a_{\varepsilon }, b_{\varepsilon }, \bar{u}_{\varepsilon })\) is continuous at the point \(\varepsilon = 0\).
Proof
Using the notations (119) and (121), we have
Recall the notation, in this case,
where \(\displaystyle T_F(\rho _{\varepsilon }, \varepsilon ) = \frac{1}{2} \left( \frac{3 \rho _{\varepsilon } \, \varepsilon }{4 \pi } \right) ^{2/3}\). By continuity at the point \(\varepsilon = 0\) of the application \(\varepsilon \mapsto (\rho _{\varepsilon },T_{\varepsilon })\), given by Lemma 19, we have
Thereby, there exists \(\varepsilon ^* > 0\) such that \(\gamma _{\varepsilon } \ge \gamma ^{\dag }\) for any \(\varepsilon \in (0,\varepsilon ^*)\), where \(\gamma ^{\dag }\) is a universal constant defined in Proposition 18. Then, from (117) in Proposition 18, for any \(\varepsilon \in (0,\varepsilon ^*)\),
which vanishes as \(\varepsilon \rightarrow 0\), since \(\gamma _{\varepsilon }\) tends to \(+ \infty \) in this limit. Combining this result, the continuity of \(\varepsilon \mapsto (\rho _{\varepsilon },u_{\varepsilon },T_{\varepsilon })\) at \(\varepsilon = 0\), given by Lemma 19, and Equation (122), we obtain
The continuity of \(\varepsilon \mapsto (\rho _{\varepsilon },u_{\varepsilon },T_{\varepsilon })\) at \(\varepsilon = 0\) being equivalent to the statement
we finally conclude that
Both \(M_{\varepsilon }\) and \({\mathcal {M}}_0\) are gaussian distributions, which coefficients are continuously defined by the above moments, allowing to conclude to the continuity of these coefficients at the point \(\varepsilon =0\). \(\square \)
Lemma 21
Let \(0 \le g \in L^1_2({\mathbb {R}}^3) \cap L \log L({\mathbb {R}}^3)\). Using the notations (119)–(120), the application \(\varepsilon \mapsto (a_{\varepsilon }, b_{\varepsilon }, \bar{u}_{\varepsilon })\) is \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\).
Proof
Since the distribution \(M_{\varepsilon }(\cdot + \bar{u}_{\varepsilon })\) is radially symmetric, so is \({\mathcal {M}}_{\varepsilon }(\cdot + \bar{u}_{\varepsilon })\), hence \(\bar{u}_{\varepsilon } = u_{\varepsilon }\), which, by Lemma 19, is \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\). We then define
so that, for any \(\varepsilon > 0\), it holds that
As in (121), we let \(\rho _{\varepsilon }, T_{\varepsilon } > 0\) be such that
Let us now show that \(\varepsilon \mapsto (a_{\varepsilon }, b_{\varepsilon })\) is \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\). It is proven in [22, proof of Proposition 3] that, letting
the function P is an increasing \({\mathcal {C}}^1\) function from \({\mathbb {R}}_+^*\) to \(\left( \frac{3^{5/3}}{5}, +\infty \right) \), with \(P' > 0\) on \({\mathbb {R}}_+^*\). Therefore it is invertible, and \(P^{-1}\) is also \({\mathcal {C}}^1\). All the more, a dominated convergence argument ensures that \(I_2\) is \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\). It is moreover shown in [22, proof of Proposition 3] that
By Lemma 19, the application \(\varepsilon \mapsto (\rho _{\varepsilon }, T_{\varepsilon })\) is \({\mathcal {C}}^1\) on \({\mathbb {R}}_+^*\), hence so is the application \(\varepsilon \mapsto (a_{\varepsilon },b_{\varepsilon })\), as a composition of \({\mathcal {C}}^1\) applications. \(\square \)
Lemma 22
Let \(0 \le g \in L^1_2({\mathbb {R}}^3) \cap L \log L({\mathbb {R}}^3)\). Using the notation (119), for any \(\overline{\varepsilon } > 0\), there exist \(C>0\) and \(\eta > 0\) such that for any \(\varepsilon \in [0,\overline{\varepsilon }]\) and \(v \in {\mathbb {R}}^3\), we have
and
Proof
We denote
Combining the results of Lemmas 20 and 21, the application \(\varepsilon \mapsto (a_{\varepsilon },b_{\varepsilon },u_{\varepsilon })\) is continuous on \({\mathbb {R}}_+\), from which we deduce that \(\underline{a}\), \(\underline{b}\) and \(\underline{u}\) are finite.
Moreover, as \(M_{\varepsilon } \in L^1({\mathbb {R}}^3)\) for all \(\varepsilon \in [0,\overline{\varepsilon }]\), the application \(\varepsilon \mapsto b_{\varepsilon }\) is (strictly) negative on \([0,\overline{\varepsilon }]\), so that, as \(\varepsilon \mapsto b_{\varepsilon }\) is continuous on \([0,\overline{\varepsilon }]\), we have \(\overline{b} < 0\).
Therefore, for any \(0 \le \varepsilon \le \overline{\varepsilon }\) and \(v \in {\mathbb {R}}^3\), we have
and, since \(|v-u_{\varepsilon }|^2 \ge \frac{1}{2} |v|^2 - |u_{\varepsilon }|^2 \ge \frac{1}{2} |v|^2 - \underline{u}^2\) and \(\overline{b} < 0\),
Letting \(\eta = -\frac{1}{2} \overline{b}\) and \(C = \max \left( \underline{a} + 2 |\underline{b}| \, \underline{u}^2, \, e^{\underline{a} + \left| \overline{b} \right| \underline{u}^2} \right) \) yields the result. \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Borsoni, T. Extending Cercignani’s Conjecture Results from Boltzmann to Boltzmann–Fermi–Dirac Equation. J Stat Phys 191, 52 (2024). https://doi.org/10.1007/s10955-024-03262-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10955-024-03262-3