Enskog and Enskog–Vlasov equations with a slightly modified correlation factor and their H theorem Shigeru Takata, Aoto Takahashi Abstract A novel modification of the original Enskog equation is proposed. The modification changes the correlation factor from a function of density to a functional of density in a simple form without series structure. The proposed modification is semi-phenomenological but is generic to be adapted to a given equation of state for non-ideal gases. It is shown that the H-theorem can be established for the Enskog and the Enskog–Vlasov equation with the proposed modification. I Introduction Behavior of ideal gases is well described by the Boltzmann equation for the entire range of the Knudsen number, the ratio of the mean free path of gas molecules to a characteristic length of the system. The kinetic theory based on the Boltzmann equation and its model equations have been applied successfully to analyses of various gas flows in low pressure circumstances, micro-scale gas flows, and gas flows caused by the evaporation and/or condensation at the gas-liquid interface. The extension of the kinetic theory to non-ideal gases would go back to the dates of Enskog [ 1 ] . He took account of the displacement effect of molecules in collision integrals for a hard-sphere gas and proposed a kinetic equation that is nowadays called the (original) Enskog equation (OEE). In the original Enskog equation, there is a correlation factor that represents an equilibrium correlation function at the contact point of two colliding molecules. Although satisfactory outcomes of the OEE, such as the dense gas effects on the transport properties, led to recent developments of numerical algorithms [ 2 , 3 , 4 , 5 ] and their applications to physical problems, e.g., [ 6 , 7 , 9 , 5 , 8 , 10 ] , the intuitive choice of the correlation factor was recognized as causing difficulties in recovering the Onsager reciprocity in multi-component gases and in establishing the H theorem. These demerits triggered off further intensive studies on the foundation of the equation around from late 60’s to early 80’s, see, e.g., [ 11 , 12 , 13 ] and references therein. Among many efforts in the above-mentioned period, Resibois [ 12 ] succeeded to prove the H theorem, not for the original but for the modified (or revised) Enskog equation (MEE) [ 11 ] . Resibois’s work motivated further theoretical researches, e.g., [ 14 , 15 , 16 , 17 , 18 ] , and the H theorem for the MEE was later extended to include isolated systems [ 15 ] and closed systems in contact with a heat bath [ 18 ] ; see also the monograph of Dorfman et al. [ 13 ] . The MEE has thus offered a satisfactory basis to the theoretical study of dense gases. Nevertheless, the mathematical intricacy has been hindering its further applications to various fundamental thermal-fluid-dynamic problems, especially to non-equilibrium boundary-value problems. Indeed, to our best knowledge, no numerical results have been reported so far based on the MEE, except for the case where the density is uniform and accordingly MEE is reduced to OEE [ 2 , 19 ] . Rather recently, Benilov & Benilov proposed an alternative construction of the correlation factor from the semi-phenomenological perspective, which is more flexible to a given equation of state (EoS) for a non-ideal gas [ 16 , 17 ] . However, their construction is based on a series expression taking care of many-body configurations, like the MEE, and requires a truncation at a certain order reasonable for both accuracy and computational cost caused by the rapidly growing complexity. In the present work, we shall focus on a single-component dense gas and newly propose a much simpler correlation factor from the semi-phenomenological perspective that is free from the series structure and from the many-body configuration. Moreover, we establish the H theorem for the Enskog equation with the proposed correlation factor. Hereinafter, we shall call the Enskog equation with the proposed correlation factor the Enskog equation with a slight modification (EESM) to discriminate from the others, since our proposal is a slight modification of the correlation factor in the OEE. Since multi-component systems are not the target of the present work, we will not discuss the issue related to the Onsager reciprocity. The rest of the paper is organized as follows. First, a generic description of the Enskog equation, associated notation, and a novel form of the correlation factor are presented in Sec. II . Next it is shown in Sec. III that there is a function monotonically decreasing in time for the EESM and the H-theorem is established for typical physical settings: (i) a periodic domain, (ii) an isolated domain, and (iii) a closed domain in contact with a heat bath. Some technical calculations that are required to obtain the results in Sec. III are summarized in Appendices A – C . Further supplemental discussions on the boundedness of that function and the extension of the results in Sec. III to the case of the Enskog–Vlasov equation [ 20 ] is presented in Appendices D and E . The paper is concluded in Sec. IV . II The Enskog equation and the novel correlation factor We consider the Enskog equation for a single species dense gas that is composed of hard sphere molecules with a common diameter \sigma and mass m . Let D be a fixed spatial domain in which the center of gas molecules is confined. Let t , \bm{X} and \bm{Y} , and \bm{\xi} be a time, spatial positions, and a molecular velocity, respectively. Denoting the one-particle distribution function of gas molecules by f(t,\bm{X},\bm{\xi} ), the Enskog equation is written as \displaystyle\frac{\partial f}{\partial t}+\xi_{i}\frac{\partial f}{\partial X_{i}}=J(f)\equiv J^{G}(f)-J^{L}(f),\quad\mathrm{for\ }\bm{X}\in D, (1a) \displaystyle J^{G}(f)\equiv\frac{\sigma^{2}}{m}\int{g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{+})f^{\prime}(\bm{X})} \displaystyle\qquad\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}, (1b) \displaystyle J^{L}(f)\equiv\frac{\sigma^{2}}{m}\int{g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})} \displaystyle\qquad\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}, (1c) where \bm{X}_{\bm{x}}^{\pm}=\bm{X}\pm\bm{x} , \bm{\alpha} is a unit vector, d\Omega(\bm{\alpha}) is a solid angle element in the direction of \bm{\alpha} , \theta is the Heaviside function \theta(x)=\begin{cases}1,&x\geq 0\\ 0,&x<0\end{cases}, (2a) and the following notation convention has been used: \displaystyle\begin{cases}{f(\bm{X})=f(\bm{X},\bm{\xi}),\ f^{\prime}(\bm{X})=f(\bm{X},\bm{\xi}^{\prime})},\\ {f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})=f(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{\xi}_{*}),\ f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{-})=f(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{\xi}_{*}^{\prime})},\end{cases} (2b) \displaystyle\begin{cases}{\bm{\xi}^{\prime}=\bm{\xi}+V_{\alpha}\bm{\alpha},\quad\bm{\xi}_{*}^{\prime}=\bm{\xi}_{*}-V_{\alpha}\bm{\alpha},}\\ {V_{\alpha}=\bm{V}\cdot\bm{\alpha},\quad\bm{V}=\bm{\xi_{*}}-\bm{\xi}.}\end{cases} (2c) Here and in what follows, the argument t is often suppressed, unless confusion is anticipated. The convention ( 2b ) will be applied only to the quantities that depend on molecular velocity. It should be noted that ( 1 ) makes sense only when the positions \bm{X} , \bm{X}^{+}_{\sigma\bm{\alpha}} , and \bm{X}^{-}_{\sigma\bm{\alpha}} are all in the domain D , which may restrict the range of integration with respect to \bm{\alpha} and \bm{\xi}_{*} . However, by including the indicator function \chi_{D} \chi_{D}(\bm{X})=\begin{cases}1,&\bm{X}\in D\\ 0,&\mbox{otherwise}\end{cases}, (2d) in the definition of g in such a way that g(\bm{X},\bm{Y})=\mathsf{g}(\bm{X},\bm{Y})\chi_{D}(\bm{X})\chi_{D}(\bm{Y}), (2e) the range of integration in ( 1b ) and ( 1c ) can be treated as the whole space of \bm{\xi}_{*} and all directions of \bm{\alpha} , irrespective of the position in the domain D . The factor \mathsf{g} occurring in ( 2e ) is generically assumed to be positive and symmetric with respect to the exchange of two position vectors: \mathsf{g}(\bm{X},\bm{Y})=\mathsf{g}(\bm{Y},\bm{X}) . There are some varieties of \mathsf{g} in the literature. In the case of the OEE, it is assumed to be the pair correlation function evaluated as a function of the density in the hard-sphere gas in uniform equilibrium. However, its exact expression is not available [ 21 ] and the Carnahan–Starling approximation [ 22 ] is often used instead, e.g., [ 2 , 3 , 4 , 7 , 9 , 19 , 23 ] : \mathsf{g}(\bm{r}_{1},\bm{r}_{2})=\frac{1-\phi/2}{(1-\phi)^{3}},\quad\phi=\frac{\pi\sigma^{3}}{6m}\rho(\frac{\bm{r}_{1}+\bm{r}_{2}}{2}), (3a) where \rho is the density defined by \rho(t,\bm{X})=\int f(t,\bm{X},\bm{\xi})d\bm{\xi}. (3b) In the case of the MEE, \mathsf{g} is given as a series of functionals of the density in the form [ 11 , 13 ] \mathsf{g}(\bm{r}_{1},\bm{r}_{2})=1+\int\frac{\rho(\bm{r}_{3})}{m}\mathcal{V}(\bm{r}_{1}\bm{r}_{2}|\bm{r}_{3})d\bm{r}_{3}\\ +\frac{1}{2}\int\frac{\rho(\bm{r}_{3})}{m}\frac{\rho(\bm{r}_{4})}{m}\mathcal{V}(\bm{r}_{1}\bm{r}_{2}|\bm{r}_{3}\bm{r}_{4})d\bm{r}_{3}d\bm{r}_{4}+\cdots, (3c) where \mathcal{V}(\bm{r}_{1}\bm{r}_{2}|\bm{r}_{3}\cdots\bm{r}_{l}) is the sum of all Mayer graphs of l labelled points. The factor \mathsf{g} in ( 3c ) is a pair correlation function that treats the density dependence in the hard-sphere gas in nonuniform equilibrium and reduces to the one in the OEE when the density is uniform. Although the examples in the previous paragraph are for the hard-sphere gas, it is possible to use different forms of \mathsf{g} from a semi-phenomenological perspective. Indeed, Benilov & Benilov [ 16 , 17 ] have recently proposed the following alternative form \mathsf{g}(\bm{r},\bm{r}_{1})=1+\sum_{l=1}^{\infty}c_{l}\int\Big[\prod_{i=2}^{l+1}\prod_{j=i+1}^{l+1}\theta(\sigma-|\bm{r}_{i}-\bm{r}_{j}|)\Big]\\ \times\Big[\prod_{i=2}^{l+1}\frac{\rho(\bm{r}_{i})}{m}\theta(\sigma-|\bm{r}-\bm{r}_{i}|)\theta(\sigma-|\bm{r}_{1}-\bm{r}_{i}|)\Big]\prod_{i=2}^{l+1}d\bm{r}_{i}, (3d) where c_{l} ’s are constants for adjusting to a given EoS. The form ( 3d ) involves ( 3c ) as a special case and ensures the H theorem as in the case of the MEE. The semi-phenomenological approach thus has an advantage of flexibility in choosing the desired EoS. However, when using ( 3d ), the desired EoS can be recovered only approximately, since the terms become rapidly complicated and the series practically has to be truncated, leaving the first few terms. In the present paper, retaining the aforementioned symmetry, we newly propose the following much simpler form of \mathsf{g} : \displaystyle\mathsf{g}(\bm{X},\bm{Y})=\mathcal{S}(\mathcal{R}(\bm{X}))+\mathcal{S}(\mathcal{R}(\bm{Y})), (4a) \displaystyle\mathcal{R}(\bm{X})=\frac{1}{m}\int_{D}\rho(\bm{Y})\theta(\sigma-|\bm{Y}-\bm{X}|)d\bm{Y}, (4b) where \mathcal{S} is a non-negative function and show that the H theorem is ensured for the Enskog equation equipped with ( 4 ). The specific form of \mathcal{S} will be determined in the connection to the EoS of the gas under consideration from the semi-phenomenological perspective. As is clear from ( 4b ), the key difference from the OEE is that \mathsf{g} is no longer a function of the density but of the functional \mathcal{R} of density representing the number of molecules in the sphere with radius \sigma centered at the position under consideration, which may be interpreted to reflect the exclusion volume effect of a molecule on the probability of the collision partner through \mathcal{S} . We close this section by listing the definitions of macroscopic quantities for later convenience. In addition to the density \rho already given in ( 3b ), the flow velocity \bm{v} (or v_{i} ) and temperature T are defined by \displaystyle v_{i}=\frac{1}{\rho}\langle\xi_{i}f\rangle, (5a) \displaystyle T=\frac{1}{3R\rho}\langle(\bm{\xi}-\bm{v})^{2}f\rangle, (5b) and the so-called kinetic part of the specific internal energy e^{(k)} , that of the stress tensor p_{ij}^{(k)} , and that of the heat-flow vector \bm{q}^{(k)} (or q_{i}^{(k)} ), are defined by \displaystyle e^{(k)}=\frac{1}{2\rho}\langle(\bm{\xi}-\bm{v})^{2}f\rangle(=\frac{3}{2}RT), (5c) \displaystyle p_{ij}^{(k)}=\langle(\xi_{i}-v_{i})(\xi_{j}-v_{j})f\rangle, (5d) \displaystyle q_{i}^{(k)}=\frac{1}{2}\langle(\xi_{i}-v_{i})(\bm{\xi}-\bm{v})^{2}f\rangle, (5e) where \langle\bullet\rangle=\int\bullet d\bm{\xi} . III Main results III.1 Kinetic part of the H function First we shall focus on the so-called kinetic part of the H function [ 24 ] that is defined by \mathcal{H}^{(k)}\equiv\int_{D}\langle f{\ln f}\rangle d\bm{X}. (6) The transformation in this small section does not require the proposed form of the correlation factor and can be found in the literature, e.g., [ 12 , 14 , 15 , 18 ] . Multiply the Enskog equation ( 1a ) by (1+{\ln f}) and integrate the result with respect to \bm{\xi} gives \frac{\partial}{\partial t}\langle f{\ln f}\rangle+\frac{\partial}{\partial X_{i}}\langle\xi_{i}f{\ln f}\rangle=\langle J(f){\ln f}\rangle. (7) By further integrating ( 7 ) with respect to \bm{X} over the domain D , we have \frac{d}{dt}\mathcal{H}^{(k)}+\int_{D}\frac{\partial}{\partial X_{i}}\langle\xi_{i}f{\ln f}\rangle d\bm{X}=\int_{\mathbb{R}^{3}}\langle J(f){\ln f}\rangle d\bm{X}, (8) where the range of integration on the right-hand side has been changed from D to \mathbb{R}^{3} , thanks to the indicator function \chi_{D} occurring in g [see ( 4a )]. No restriction on the range of spatial integration allows the shift and other operations summarized in Appendix A . As is explained in Appendix A , the right-hand side of ( 8 ) can be transformed into the form that \displaystyle\int_{\mathbb{R}^{3}}\langle J(f){\ln f}\rangle d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int\ln\Big(\frac{f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{-})f^{\prime}(\bm{X})}{f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})}\Big)g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\times f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X}. (9) Since x\ln(y/x)\leq y-x for any x,y>0 and the equality holds if and only if x=y , we have the estimate that \int_{D}\langle J(f){\ln f}\rangle d\bm{X}\leq I(t), (10) where \displaystyle I(t) \displaystyle=\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})[f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{-})f^{\prime}(\bm{X}) \displaystyle\quad-f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})]V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X}, (11) and the equality holds if and only if I(t)=0 or equivalently f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{-})f^{\prime}(\bm{X})-f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})=0. (12) It should be remarked that, as is explained in Appendix B.1 , I(t) is eventually reduced to \displaystyle I(t)=-\frac{\sigma^{2}}{m} \displaystyle\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{+})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{+}) \displaystyle\quad\times\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X}. (13) III.2 Collisional part of the H function Next introduce the following function \mathcal{H}^{(c)}(t)=\int_{D}\rho(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx]d\bm{X}. (14) We shall show below that its time derivative is related to I(t) . Namely, using a concise notation r=|\bm{Y}-\bm{X}| , \displaystyle\frac{d}{dt}\mathcal{H}^{(c)}= \displaystyle\frac{d}{dt}\int_{D}\rho(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx]d\bm{X} \displaystyle= \displaystyle\int_{D}\{\frac{\partial\rho(\bm{X})}{\partial t}\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\rho(\bm{X})\frac{\partial\mathcal{R}(\bm{X})}{\partial t}\mathcal{S}(\mathcal{R}(\bm{X}))\}d\bm{X} \displaystyle= \displaystyle\int_{D}\{\frac{\partial\rho(\bm{X})}{\partial t}\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\frac{\rho(\bm{X})}{m}[\int_{D}\frac{\partial\rho(\bm{Y})}{\partial t}\theta(\sigma-r)d\bm{Y}]\mathcal{S}(\mathcal{R}(\bm{X}))\}d\bm{X} \displaystyle= \displaystyle\int_{D}\frac{\partial\rho(\bm{X})}{\partial t}[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial\rho{v}_{i}(\bm{X})}{\partial X_{i}}[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial}{\partial X_{i}}\{\rho{v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]\}d\bm{X} \displaystyle+\int_{D}\rho{v}_{i}(\bm{X})[\frac{\partial\mathcal{R}(\bm{X})}{\partial X_{i}}\mathcal{S}(\mathcal{R}(\bm{X}))+\int_{D}\frac{\rho(\bm{Y})}{m}\frac{\partial}{\partial X_{i}}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial}{\partial X_{i}}\{\rho{v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]\}d\bm{X} \displaystyle+\int_{D}\rho{v}_{i}(\bm{X})\{\int_{D}\frac{\rho(\bm{Y})}{m}\frac{Y_{i}-X_{i}}{|\bm{Y}-\bm{X}|}\delta(\sigma-|\bm{Y}-\bm{X}|)[\mathcal{S}(\mathcal{R}(\bm{X}))+\mathcal{S}(\mathcal{R}(\bm{Y}))]d\bm{Y}\}d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial}{\partial X_{i}}\{\rho{v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]\}d\bm{X} \displaystyle+\frac{\sigma^{2}}{m}\int\rho{v}_{i}(\bm{X})\{\int\rho(\bm{X}_{\sigma\bm{\alpha}}^{+})\alpha_{i}[\mathcal{S}(\mathcal{R}(\bm{X}))+\mathcal{S}(\mathcal{R}(\bm{X}_{\sigma\bm{\alpha}}^{+}))]\chi_{D}(\bm{X})\chi_{D}(\bm{X}_{\sigma\bm{\alpha}}^{+})d\Omega(\bm{\alpha})\}d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial}{\partial X_{i}}\{\rho{v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]\}d\bm{X} \displaystyle+\frac{\sigma^{2}}{m}\int\rho(\bm{X}){v}_{i}(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{+})\alpha_{i}g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{+})d\Omega(\bm{\alpha})d\bm{X} \displaystyle= \displaystyle-\int_{D}\frac{\partial}{\partial X_{i}}\{\rho{v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-r)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]\}d\bm{X}-I(t). (15) As the result, the time derivative of the sum \mathcal{H}\equiv\mathcal{H}^{(k)}+\mathcal{H}^{(c)} is found to satisfy the inequality that \frac{d\mathcal{H}}{dt}=\frac{d}{dt}(\mathcal{H}^{(k)}+\mathcal{H}^{(c)})\leq\int_{\partial D}(H_{i}^{(k)}+H_{i}^{(c)})n_{i}dS, (16a) where \bm{n} (or n_{i} ) is the inward unit normal to the boundary, dS is the surface element of the boundary \partial D , and \displaystyle H_{i}^{(k)}= \displaystyle\langle\xi_{i}f\ln f\rangle, (16b) \displaystyle H_{i}^{(c)}= \displaystyle\rho(\bm{X}){v}_{i}(\bm{X})[\int_{0}^{\mathcal{R}(\bm{X})}\mathcal{S}(x)dx \displaystyle+\int_{D}\frac{\rho(\bm{Y})}{m}\theta(\sigma-|\bm{Y}-\bm{X}|)\mathcal{S}(\mathcal{R}(\bm{Y}))d\bm{Y}]. (16c) The inequality ( 16a ) is the H theorem for the EESM. III.3 Monotonicity in three typical cases We shall discuss the more details about ( 16a ) for three typical cases: the domain D is three dimensional and is (i) periodic [ 25 ] , (ii) surrounded by the specular reflection boundary, and (iii) surrounded by the impermeable surface of a heat bath with a uniform constant temperature T_{w} . First consider the case (i). In this case, the surface integrals over \partial D vanish, thanks to the periodicity, and accordingly we have \frac{d\mathcal{H}}{dt}\leq 0, (17) and the equality holds if and only if ( 12 ) is satisfied. Next consider the case (ii). Then H_{i}^{(k)}n_{i}=0 on \partial D , since \xi_{i}n_{i}f\ln f is odd with respect to \xi_{i}n_{i} . Noting that the specular reflection boundary is impermeable, H_{i}^{(c)}n_{i}=0 holds as well because of v_{i}n_{i}=0 on \partial D by the impermeability. As the result, we have again \frac{d\mathcal{H}}{dt}\leq 0, (18) and the equality holds if and only if ( 12 ) is satisfied. Finally, consider the case (iii). Because of the impermeability, v_{i}n_{i}=0 and thus H_{i}^{(c)}n_{i}=0 on the boundary. Then, thanks to the Darrozes–Guiraud inequality (Lemma C.1 ) [ 26 , 27 , 28 ] , it holds that \int_{\partial D}\langle\xi_{i}n_{i}(\bm{X})f\ln\frac{f}{f_{w}}\rangle dS(\bm{X})\leq 0, (19a) where f_{w}=\frac{C}{(2\pi RT_{w})^{3/2}}\exp(-\frac{\bm{\xi}^{2}}{2RT_{w}}), (19b) with C(>0) is an arbitrary constant. Since T_{w}(>0) is a constant, the multiplication of \ln f_{w} with ( 1a ) and its integration with respect to \bm{X} over the domain D leads to \displaystyle\frac{d}{dt}\int_{D}\langle f\ln f_{w}\rangle d\bm{X}-\int_{\partial D}\langle\xi_{i}n_{i}f\ln f_{w}\rangle dS \displaystyle=\int_{D}\langle J(f)\ln f_{w}\rangle d\bm{X}, (20) where the right-hand side vanishes as is explained in Appendix B.2 [see ( 44 )]. Hence, ( 16a ) is recast as \frac{d\mathcal{F}}{dt}\leq 0, (21a) where \displaystyle\mathcal{F}= \displaystyle RT_{w}(\int_{D}\langle f\ln\frac{f}{f_{w}}\rangle d\bm{X}+\mathcal{H}^{(c)}) \displaystyle= \displaystyle RT_{w}\mathcal{H}+\int_{D}\langle\frac{1}{2}\bm{\xi}^{2}f\rangle d\bm{X}+\mathrm{const}., (21b) and the equality holds again if and only if ( 12 ) is satisfied. [ 29 ] Incidentally, it is readily seen that \mathcal{H}^{(c)} is the counterpart of \int_{\Omega}Q(n)d\vec{r} in [ 30 ] multiplied with m , where \Omega is a periodic domain, and that \mathcal{H}^{(c)} indeed satisfies the compatibility condition with the thermodynamics , (3.6) in [ 30 ] [ \epsilon_{ij} in (3.6) should read \epsilon_{ij}^{-1} ]. Therefore, the present result for the periodic domain is consistent with Theorem III.4 in [ 30 ] . Moreover, the extension of Theorem III.4 in [ 30 ] to the other two cases considered in the present subsection can be made by simply including the contribution from H_{i}^{(k)} and H_{i}^{(c)} in the discussion there. It is seen that the same compatibility condition applies to these two cases as well in order to ensure the H theorem. The details are omitted here. III.4 The Maxwellian at the stationary state for the three typical cases The condition ( 12 ) is rewritten as \ln f(\bm{X})+\ln f_{*}(\bm{X}^{-}_{\sigma\bm{\alpha}})=\ln f^{\prime}(\bm{X})+\ln f_{*}^{\prime}(\bm{X}^{-}_{\sigma\bm{\alpha}}), (22) which implies that such a \ln f is the summational invariant and is restricted to the form \displaystyle f(\bm{X})= \displaystyle\frac{\rho(\bm{X})}{(2\pi RT)^{3/2}}\exp(-\frac{[\bm{\xi}-\bm{v}(\bm{X})]^{2}}{2RT}), (23a) \displaystyle\bm{v}(\bm{X})= \displaystyle\bm{u}+\bm{X}\times\bm{\omega}, (23b) see, e.g., [ 31 , 15 ] . Here, it should be noted that T , \bm{u} , and \bm{\omega} are independent of \bm{X} and that \bm{u} and \bm{\omega} represent the translational and the angular velocity of the flow, respectively. In the case (i), the periodic condition prohibits the rotational mode of velocity, so that ( 23 ) is reduced to f(\bm{X})=\frac{\rho(\bm{X})}{(2\pi RT)^{3/2}}\exp(-\frac{(\bm{\xi}-\bm{u})^{2}}{2RT}), (24) where T(>0) and \bm{u} are uniform scalar and vector that are determined in accordance with the total momentum and energy, thanks to the conservation laws (see Appendix B.2 ). In the case (ii), the specular reflection does not allow the translational velocity \bm{u} , so that ( 23 ) is reduced to f(\bm{X})=\frac{\rho(\bm{X})}{(2\pi RT)^{3/2}}\exp(-\frac{(\bm{\xi}-\bm{X}\times\bm{\omega})^{2}}{2RT}), (25) where a non-zero \bm{\omega} is admitted only when the shape of the boundary is symmetric around \bm{\omega} . Equation ( 25 ) represents a rigid rotation of the fluid. Moreover, T>0 and \bm{\omega} are again a uniform scalar and vector that are determined in accordance with the total momentum and energy, thanks to the conservation law. If the shape of the boundary is not symmetric, the total momentum is not conserved and finally the angular velocity vanishes irrespective of the initial state due to the impermeability of the boundary. Finally, in the case (iii), ( 23 ) is reduced to f(\bm{X})=\frac{\rho(\bm{X})}{(2\pi RT_{w})^{3/2}}\exp(-\frac{\bm{\xi}^{2}}{2RT_{w}}), (26) since the Darrouse–Guiraud inequality (Lemma C.1 ) forces the gas to be at rest and at the same temperature as the heat bath on the boundary. III.5 Relation to the equation of state As is well known, the stress tensor p_{ij} in the case of the Enskog equation is expressed by the sum of the kinetic and the collisional contribution, p_{ij}^{(k)} and p_{ij}^{(c)} ; see ( 50c ), ( 5d ), ( 49b ), and Appendix B.2 . Since the pressure p is defined as the one third of the trace of the stress tensor, it is expressed as \displaystyle p= \displaystyle\frac{1}{3}(p_{ii}^{(k)}+p_{ii}^{(c)}) \displaystyle= \displaystyle\rho RT+\frac{\sigma^{2}}{6m}\int{\int_{0}^{\sigma}}V_{\alpha}^{2}\theta(V_{\alpha})g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+}) \displaystyle\quad\times f_{*}(\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f(\bm{X}_{\lambda\bm{\alpha}}^{+}){d\lambda}d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}. (27) Now consider the infinite expanse of the dense gas in the uniform equilibrium state. In the case, the last integral is reduced to (\sigma^{2}/6)(\rho^{2}RT/m)\int_{0}^{\sigma}\int g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})d\Omega(\bm{\alpha})d\lambda . In the case of the proposed correlation factor ( 4a ) with ( 4b ), g becomes constant such that g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})=2\mathcal{S}(2b\rho),\quad b=\frac{2\pi}{3}\frac{\sigma^{3}}{m}, (28) since the density is constant. As the result, ( 27 ) is reduced to the following EoS for the gas under consideration: p=\rho RT(1+2b\rho\mathcal{S}(2b\rho)). (29) Now we shall present the specific form of \mathcal{S} for two typical equations of state in the literature: the van der Waals equation of state [ 36 ] for non-attractive molecules p=\frac{\rho RT}{1-b\rho}=\rho RT(1+\frac{b\rho}{1-b\rho}), (30) and the Carnahan–Starling equation of state [ 22 ] p=\rho RT\frac{1+\eta+\eta^{2}-\eta^{3}}{(1-\eta)^{3}}=\rho RT(1+\frac{4\eta-2\eta^{2}}{(1-\eta)^{3}}), (31) where \eta={b\rho}/{4} . It is readily seen that the appropriate form of \mathcal{S} for the van der Waals equation of state for non-attractive molecules is \mathcal{S}(x)=\frac{1}{2-x}, (32) while that for the Carnahan–Starling equation of state is \mathcal{S}(x)=\frac{16(16-x)}{(8-x)^{3}}. (33) It should be noted that the collision term of the Enskog equation is not responsible for the attractive part of the EoS. The attractive part is to be recovered by the Vlasov term of the Enksog–Vlasov equation; see Appendix E and, e.g., [ 20 , 8 ] . Therefore the form ( 32 ) applies to the EESM with the Vlasov term for the full version of the van der Waals fluids as well (see Appendix E ). IV Conclusion In the present paper, a novel form of the correlation factor in the collision integral of the Enskog equation has been proposed. The new factor is a slight modification of the counterpart of the original Enskog equation from a function to a functional of the density and still keeps a simplicity without a series structure. The H theorem has been established for the Enskog equation with the proposed modification. The function monotonically decreasing in time has been presented for three typical cases: the domain is (i) periodic, (ii) surrounded by the specular reflection boundary, and (iii) surrounded by the impermeable surface of a resting heat bath. For these cases, that function is shown to be bounded in Appendix D for the van der Waals (without the attractive part) and the Carnahan–Starling equation of state, if it is initially bounded. These results, i.e., the H theorem, the monotonically decreasing function, and its boundedness, have also been extended to the Enskog–Vlasov equation in Appendix E . The specific form of the novel correlation factor is shown to be determined in accordance with the EoS (more precisely the repulsive part of the EoS) under consideration. As two typical examples, the form in the cases of the van der Waals and the Carnahan–Starling equation of state have been presented as well. Although simple kinetic models at the level of relaxation model have been proposed and used for the dense gas in the literature, e.g., [ 32 , 33 , 34 , 35 ] , the present result is expected to activate further numerical applications of the Enskog equation. Indeed, the novel correlation factor is so simple to be implemented in the numerical schemes for the OEE in the literature [ 2 , 3 , 4 , 5 ] . Hence, by a slight modification, those schemes can be used for the EESM and put the numerics of the Enskog equation on the firm ground of the thermodynamical consistency. Finally, although not discussed in the present paper, the fluid-dynamic limit of the EESM recovers the same transport properties as that of the OEE at the level of the Navier–Stokes–Fourier equations, provided that the corresponding EoS is common. The details will be reported elsewhere. Numerical application of the EESM to thermo-fluid phenomena is a natural direction of research stemming from the present work. Another natural direction is to examine the feasibility of the EESM in the case of multi-component systems, in which the Onsager reciprocity should be properly recovered as often mentioned in the literature, e.g., [ 11 , 13 ] . This topic, while important, should be the subject of future work. Appendix A Shift and other operations We summarize the shift and other standard operations that are used in the transformations of the collision integral. There are three types of operation that are standard in the case of the Boltzmann equation as well: (I) to exchange the letters \bm{\xi} and \bm{\xi}_{*} ; (II) to reverse the direction of \bm{\alpha} ; (III) to change the integration variables from (\bm{\xi},\bm{\xi}_{*},\bm{\alpha}) to (\bm{\xi}^{\prime},\bm{\xi}_{*}^{\prime},\bm{\alpha}) and then to change the letters (\bm{\xi}^{\prime},\bm{\xi}_{*}^{\prime}) to (\bm{\xi},\bm{\xi}_{*}) . First, by (III) and (II), \int{\varphi}(\bm{X})J^{G}(f)d\bm{\xi}=\int{\varphi}^{\prime}(\bm{X})J^{L}(f)d\bm{\xi}, (34) holds for any {\varphi}(\bm{X},\bm{\xi}) . Hence, we have \displaystyle\langle\varphi(\bm{X})J(f)\rangle \displaystyle= \displaystyle\frac{\sigma^{2}}{m}\int[{\varphi}^{\prime}(\bm{X})-{\varphi}(\bm{X})]g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}. (35) Next by integrating ( 35 ) with respect to \bm{X} over the domain D , we have \displaystyle\int_{D}\langle\varphi(\bm{X})J(f)\rangle d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{m}\int_{\mathbb{R}^{3}}\int[{\varphi}^{\prime}(\bm{X})-{\varphi}(\bm{X})]g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}d\bm{X}. (36) Here the domain of integration with respect to \bm{X} has been changed from D to \mathbb{R}^{3} , thanks to the factor \chi_{D} occurring in g ; see ( 4a ). This change also allows \bm{\alpha} to take all the directions. Now consider the shift operation by \sigma\bm{\alpha} followed by the operations (II) and (I). Then, the right-hand side of ( 36 ) is recast as \displaystyle\frac{\sigma^{2}}{m} \displaystyle\int_{\mathbb{R}^{3}}\int[{\varphi}^{\prime}_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})-{\varphi}_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})]g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\qquad\times f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X}. (37) and thus, we have \displaystyle\int_{D}\langle\varphi(\bm{X})J(f)\rangle d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int_{\mathbb{R}^{3}}\int[{\varphi}^{\prime}(\bm{X})+{\varphi}^{\prime}_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})-{\varphi}_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})-{\varphi}(\bm{X})] \displaystyle\times g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}d\bm{X}. (38) The substitution of \varphi=\ln f in ( 38 ) yields \displaystyle\int_{D}\langle J(f)\ln f\rangle d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int_{\mathbb{R}^{3}}\int\ln\frac{f^{\prime}_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f^{\prime}(\bm{X})}{f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})}g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X}) \displaystyle\quad\times f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}d\bm{X}, (39) which is the form of ( 9 ). Appendix B Collisional contributions to entropy generation and transport properties B.1 Collisional contributions to the entropy generation The transformation of I(t) from ( 11 ) to ( 13 ) has been done by the repeated operations of the shift, (II), and (III): \displaystyle I(t) \displaystyle= \displaystyle-\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})f_{*}^{\prime}(\bm{X}_{\sigma\bm{\alpha}}^{-})f^{\prime}(\bm{X}) \displaystyle\qquad\qquad\times V_{\alpha}^{\prime}\theta(-V_{\alpha}^{\prime})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X} \displaystyle-\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\qquad\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\qquad\qquad\times[(\bm{\xi}-\bm{\xi}_{*})\cdot\bm{\alpha}]d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\qquad\quad\times\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X} \displaystyle-\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{-}) \displaystyle\qquad\quad\times\bm{v}(\bm{X}_{\sigma\bm{\alpha}}^{-})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{2m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{-})\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X} \displaystyle-\frac{\sigma^{2}}{2m}\int g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{+})\rho(\bm{X})\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X} \displaystyle= \displaystyle\frac{\sigma^{2}}{m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{-})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{-})\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X} \displaystyle= \displaystyle-\frac{\sigma^{2}}{m}\int g(\bm{X},\bm{X}_{\sigma\bm{\alpha}}^{+})\rho(\bm{X})\rho(\bm{X}_{\sigma\bm{\alpha}}^{+})\bm{v}(\bm{X})\cdot\bm{\alpha}d\Omega(\bm{\alpha})d\bm{X}, (40) where V_{\alpha}^{\prime}\equiv(\bm{\xi}_{*}^{\prime}-\bm{\xi}^{\prime})\cdot\bm{\alpha}=-V_{\alpha} has been used at the beginning of the above transformation. B.2 Conservation laws and the collisional momentum and energy transports Consider first the integration of ( 1a ) with respect to \bm{\xi} . Since \langle J(f)\rangle=0 by ( 35 ) with {\varphi}=1 , we have the standard form of the continuity equation: \frac{\partial\rho}{\partial t}+\frac{\partial\rho v_{i}}{\partial X_{i}}=0. (41a) Next consider the integration of ( 1a ) multiplied by \psi=\xi_{j} and \psi=\bm{\xi}^{2}/2 with respect to \bm{\xi} . Then we have \displaystyle\frac{\partial\rho v_{j}}{\partial t}+\frac{\partial}{\partial X_{i}}(\rho v_{i}v_{j}+p_{ij}^{(k)})=\langle\xi_{j}J(f)\rangle, (41b) \displaystyle\frac{\partial}{\partial t}[\rho(e^{(k)}+\frac{1}{2}\bm{v}^{2})] \displaystyle+\frac{\partial}{\partial X_{i}}[\rho v_{i}(e^{(k)}+\frac{1}{2}\bm{v}^{2})+p_{ij}^{(k)}v_{j}+q^{(k)}_{i}]=\frac{1}{2}\langle\bm{\xi}^{2}J(f)\rangle, (41c) where the definitions ( 5d ), ( 5e ), and ( 5c ) have been used. The difference from the Boltzmann equation occurs in that there are contributions from collision term on the right-hand side. Indeed, ( 35 ) with \varphi=\psi leads to \displaystyle\langle\psi J(f)\rangle= \displaystyle-\frac{\sigma^{2}}{m}\int(\psi-\psi^{\prime})g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X}) \displaystyle\qquad\quad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}. (42a) But this expression can be further transformed by using the operations (I) and (II) first and then using the relation \psi + \psi_{*} = \psi^{\prime} + \psi_{*}^{\prime} as \displaystyle\langle\psi J(f)\rangle= \displaystyle\frac{\sigma^{2}}{m}\int(\psi_{*}^{\prime}-\psi_{*})g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f(\bm{X}_{\sigma\bm{\alpha}}^{+})f_{*}(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*} \displaystyle= \displaystyle\frac{\sigma^{2}}{m}\int(\psi-\psi^{\prime})g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f(\bm{X}_{\sigma\bm{\alpha}}^{+})f_{*}(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}. (42b) Note that \psi^{\prime}-\psi=\begin{cases}V_{\alpha}\alpha_{i},&{\displaystyle(\psi=\xi_{i}),}\\ {\displaystyle\frac{1}{2}V_{\alpha}(\bm{\xi}+\bm{\xi}_{*})\cdot\bm{\alpha},}&({\displaystyle\psi=\frac{1}{2}\bm{\xi}^{2})},\end{cases} (42c) holds, thanks to the relation ( 2c ). Hence, once integrated over the domain D , the result is expressed after the shift operation as \displaystyle\int_{D}\langle\psi J(f)\rangle d\bm{X} \displaystyle= \displaystyle-\frac{\sigma^{2}}{m}\int(\psi-\psi^{\prime})g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X} \displaystyle= \displaystyle-\frac{\sigma^{2}}{m}\int(\psi-\psi^{\prime})g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f(\bm{X}_{\sigma\bm{\alpha}}^{+})f_{*}(\bm{X}) \displaystyle\qquad\times V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}d\bm{\xi}_{*}d\bm{X} \displaystyle= \displaystyle-\int_{D}\langle\psi J(f)\rangle d\bm{X}, (43) where ( 42b ) has been used at the last equality. This implies that \int_{D}\langle\psi J(f)\rangle d\bm{X}=0,\quad(\psi=\xi_{j},\frac{\bm{\xi}^{2}}{2}), (44) and accordingly the integration in space over the domain of ( 41a ), ( 41b ), and ( 41c ) leads to \displaystyle\frac{d}{dt}\int_{D}\rho d\bm{X}=0, (45a) \displaystyle\frac{d}{dt}\int_{D}\rho v_{j}d\bm{X}=\int_{\partial D}p_{ij}^{(k)}n_{i}dS, (45b) \displaystyle\frac{d}{dt}\int_{D}[\rho(e^{(k)}+\frac{1}{2}\bm{v}^{2})]d\bm{X}=\int_{\partial D}(p_{ij}^{(k)}v_{j}+q_{i}^{(k)})n_{i}dS, (45c) for three typical cases discussed in Sec. III.3 . In the case (i), the surface integrals in ( 45 ) vanish because of the periodic condition. Accordingly, in addition to the total mass \int_{D}\rho d\bm{X} , the total momentum \int_{D}\rho v_{j}d\bm{X} and total energy \int_{D}\rho[e+(1/2)\bm{v}^{2}]d\bm{X} are constant in time. In the case (ii), because of the specular reflection condition, the surface integral in ( 45c ) vanishes, although that in ( 45b ) does not vanish in general. However, the total mass \int_{D}\rho d\bm{X} and the total energy \int_{D}\rho[e+(1/2)\bm{v}^{2}]d\bm{X} are constant in time. The gas approaches to a resting state or a rigid rotational motion in accordance with the boundary shape (see Sec. III.4 ). In order to see it a little more closely, consider the conservation law of the angular momentum, which is obtained by multiplying ( 41b ) with \epsilon_{ijk}X_{k} , where \epsilon_{ijk} is the Eddington epsilon. After integrating over the domain D , we have eventually \frac{d}{dt}\Omega_{i}=\int_{\partial D}\epsilon_{ijk}p_{jl}^{(k)}n_{l}X_{k}dS, (46) where \Omega_{i}=\int_{D}\epsilon_{ijk}\rho v_{j}X_{k}d\bm{X} is the total angular momentum. In deriving ( 46 ), \int_{D}\epsilon_{ijk}X_{k}\langle\xi_{j}J(f)\rangle d\bm{X} has vanished, since \epsilon_{ijk}X_{k}\xi_{j} is a summational invariant [see [ 31 ] and ( 38 )]. Because of the specular reflection, p_{jl}^{(k)}n_{l} has no components tangential to the boundary. Thus the integrand on the right-hand side vanishes if the position vector \bm{X} of a point on the surface is perpendicular to the tangential direction of the boundary around the axis of rotation. As the result, if the shape of the boundary is symmetric around a certain axis, the total angular momentum \Omega around the same axis does not change in time. If not, at the stationary equilibrium state discussed in Sec. III.4 , the rigid rotation conflicts with the impermeability of the boundary and the flow ought to vanish. In the case (iii), the surface integrals in ( 45 ) do not vanish in general, since there can be the friction and energy exchanges on the impermeable surface of a heat bath. Hence only the total mass is constant in time. In the meantime, there is another way of transformation of \langle\psi J(f)\rangle , which leads to the concept of the collisional contributions to the stress tensor and the heat-flow vector. To see it, combine ( 42b ) and ( 35 ) for {\varphi}=\psi . Then, we have \displaystyle\langle\psi J(f)\rangle \displaystyle= \displaystyle\frac{1}{2}\frac{\sigma^{2}}{m}\int(\psi^{\prime}-\psi)\{g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X}) \displaystyle-g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f(\bm{X}_{\sigma\bm{\alpha}}^{+})f_{*}(\bm{X})\}V_{\alpha}\theta(V_{\alpha})d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}. (47) This form allows further local transformation, thanks to [ 37 , 15 ] , \displaystyle g(\bm{X}_{\sigma\bm{\alpha}}^{-},\bm{X})f_{*}(\bm{X}_{\sigma\bm{\alpha}}^{-})f(\bm{X})-g(\bm{X}_{\sigma\bm{\alpha}}^{+},\bm{X})f(\bm{X}_{\sigma\bm{\alpha}}^{+})f_{*}(\bm{X}) \displaystyle= \displaystyle-\int_{0}^{\sigma}\frac{\partial}{\partial\lambda}[g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f_{*}(\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f(\bm{X}_{\lambda\bm{\alpha}}^{+})]d\lambda \displaystyle= \displaystyle-\frac{\partial}{\partial X_{i}}\int_{0}^{\sigma}\alpha_{i}g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+}) \displaystyle\qquad\qquad\times f_{*}(\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f(\bm{X}_{\lambda\bm{\alpha}}^{+})d\lambda. (48) Then, eventually the right-hand side of ( 41b ) and ( 41c ) can be rewritten as [ 37 , 15 ] \displaystyle\langle\xi_{j}J(f)\rangle= \displaystyle-\frac{\partial}{\partial X_{i}}p_{ij}^{(c)}, \displaystyle\frac{1}{2}\langle\bm{\xi}^{2}J(f)\rangle= \displaystyle-\frac{\partial}{\partial X_{i}}(p_{ij}^{(c)}v_{j}+q_{i}^{(c)}), (49a) where \displaystyle p_{ij}^{(c)}= \displaystyle\frac{\sigma^{2}}{2m}\int{\int_{0}^{\sigma}}\alpha_{i}\alpha_{j}V_{\alpha}^{2}\theta(V_{\alpha})g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+}) \displaystyle\quad\times f_{*}(\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f(\bm{X}_{\lambda\bm{\alpha}}^{+}){d\lambda}d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}, (49b) \displaystyle q_{i}^{(c)}= \displaystyle\frac{\sigma^{2}}{4m}\int{\int_{0}^{\sigma}}\alpha_{i}[(\bm{c}+\bm{c}_{*})\cdot\bm{\alpha}]V_{\alpha}^{2}\theta(V_{\alpha}) \displaystyle\quad\times g(\bm{X}_{\lambda\bm{\alpha}}^{+},\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+})f_{*}(\bm{X}_{(\lambda-\sigma)\bm{\alpha}}^{+}) \displaystyle\qquad\times f(\bm{X}_{\lambda\bm{\alpha}}^{+}){d\lambda}d\Omega(\bm{\alpha})d\bm{\xi}_{*}d\bm{\xi}, (49c) with \bm{c}=\bm{\xi}-\bm{v} and \bm{c}_{*}=\bm{\xi}_{*}-\bm{v} ; see, e.g., [ 37 , 7 ] . Hence, ( 41b ) and ( 41c ) are rewritten as \displaystyle\frac{\partial\rho v_{j}}{\partial t}+\frac{\partial}{\partial X_{i}}(\rho v_{i}v_{j}+p_{ij})=0, (50a) \displaystyle\frac{\partial}{\partial t}[\rho(e+\frac{1}{2}\bm{v}^{2})]+\frac{\partial}{\partial X_{i}}[\rho v_{i}(e+\frac{1}{2}\bm{v}^{2})+p_{ij}v_{j}+q_{i}]=0, (50b) with p_{ij}=p_{ij}^{(k)}+p_{ij}^{(c)},\quad q_{i}=q_{i}^{(k)}+q_{i}^{(c)},\quad e=e^{(k)}, (50c) and the usual form of the conservation laws of mass, momentum, and energy has been recovered, together with ( 41a ). The superscript (c) denotes the collisional contribution to individual quantities. Appendix C The kinetic boundary condition for the heat bath and the Darrozes–Guiraud inequality The kinetic boundary condition on the impermeable surface \partial D of a heat bath can be written generically as \displaystyle f(t,\bm{X},\bm{\xi})=\int_{\bm{\xi}_{*}\cdot\bm{n}<0} \displaystyle K(\bm{\xi},\bm{\xi}_{*}|\bm{X})f(t,\bm{X},\bm{\xi}_{*})d\bm{\xi}_{*}, \displaystyle\quad(\bm{\xi}\cdot\bm{n}>0,\ \bm{X}\in\partial D), (51a) where K(\bm{\xi},\bm{\xi}_{*}|\bm{X}) is a scattering kernel assumed to be time-independent. Assuming that the boundary is at rest, the following properties are conventionally required to hold: [ 28 ] 1. Non-negativeness: K(\bm{\xi},\bm{\xi}_{*}|\bm{X})\geq 0,\quad(\bm{\xi}\cdot\bm{n}>0,\ \bm{\xi}_{*}\cdot\bm{n}<0); (51b) 2. Normalization: \int_{\bm{\xi}\cdot\bm{n}>0}\Big|\frac{\bm{\xi}\cdot\bm{n}}{\bm{\xi}_{*}\cdot\bm{n}}\Big|K(\bm{\xi},\bm{\xi}_{*}|\bm{X})d\bm{\xi}=1,\quad(\bm{\xi}_{*}\cdot\bm{n}<0), (51c) where the integrand in ( 51c ) is the so-called reflection probability density. Equation ( 51c ) implies that the boundary \partial D is impermeable; 3. Preservation of equilibrium: The resting Maxwellian f_{w} characterized by the surface temperature T_{w} and defined by ( 19b ) satisfies the boundary condition ( 51a ), and the other Maxwellians do not satisfy ( 51a ). The diffuse reflection, the Maxwell, and the Cercignani–Lampis condition [ 38 , 27 , 28 ] that are widely used for the Boltzmann equation are specific examples of ( 51 ). Note that the uniqueness in the third property listed above excludes the adiabatic boundary such as the specular reflection condition. On the boundary with the property ( 51 ), the following statement is known to hold. Lemma C.1 . (Darrozes–Guiraud [ 26 , 27 , 28 ] ) If the velocity distribution function f satisfies the boundary condition ( 51 ), then it holds that \int_{\partial D}\langle(\bm{\xi}\cdot\bm{n})f\ln\frac{f}{f_{w}}\rangle dS\leq 0. (52) Here the equality holds if and only if f=f_{w} . Appendix D Boundedness of \mathcal{H} and \mathcal{F} In this Appendix, we will show that \mathcal{H} and \mathcal{F} are bounded. First consider the part \mathcal{H}^{(k)} . The following discussion on this part is along the same line as the proof in [ 39 ] for the Boltzmann equation. We will show that \mathcal{H}^{(k)} is bounded from below. To this end, it is enough to consider the case f<1 , since \mathcal{H}^{(k)}=\int_{D}\int f\ln fd\bm{\xi}d\bm{X} . Separate the range of integration into three parts: \mathcal{D}^{+}=\{(\bm{x},\bm{\xi})|f\geq 1\} , \mathcal{D}^{-}_{1}=\{(\bm{x},\bm{\xi})|f<1,f>\exp(-\beta\bm{\xi}^{2}/2)\} , \mathcal{D}^{-}_{2}=\{(\bm{x},\bm{\xi})|f<1,f\leq\exp(-\beta\bm{\xi}^{2}/2)\} . Here \beta is a constant that is strictly positive. Then, the contribution from the range \mathcal{D}^{+} is non-negative, \mathcal{H}^{(k)}\geq(\int_{\mathcal{D}^{-}_{1}}+\int_{\mathcal{D}^{-}_{2}})f\ln fd\bm{\xi}d\bm{X}. (53) The function f\ln f decreases monotonically from 0 to -e^{-1} as f increases from 0 to e^{-1} , while f\ln f>-f for f>e^{-1} . Therefore, f\ln f>-(\beta\bm{\xi}^{2}/2)f in the range \mathcal{D}^{-}_{1} , while f\ln f\geq-f-(\beta\bm{\xi}^{2}/2)\exp(-\beta\bm{\xi}^{2}/2) in the range \mathcal{D}^{-}_{2} . Applying these estimates to ( 53 ), \displaystyle\mathcal{H}^{(k)}\geq \displaystyle-\int_{\mathcal{D}^{-}_{1}}\frac{\beta\bm{\xi}^{2}}{2}fd\bm{\xi}d\bm{X} \displaystyle-\int_{\mathcal{D}^{-}_{2}}[f+\frac{\beta\bm{\xi}^{2}}{2}\exp(-\frac{\beta\bm{\xi}^{2}}{2})]d\bm{\xi}d\bm{X} \displaystyle\geq \displaystyle-\int_{D}\langle\frac{\beta\bm{\xi}^{2}}{2}f\rangle d\bm{X}-\int_{D}\langle f+\frac{\beta\bm{\xi}^{2}}{2}\exp(-\frac{\beta\bm{\xi}^{2}}{2})\rangle d\bm{X} \displaystyle= \displaystyle-\beta\int_{D}\rho(e^{(k)}+\frac{1}{2}\bm{v}^{2})d\bm{X}-\int_{D}\rho d\bm{X}+\mathrm{const.}, (54) where the last constant depends on \beta . When the domain is periodic or is surrounded by specular reflection boundary, both the total mass \int_{D}\rho d\bm{X} and energy \int_{D}\rho(e^{(k)}+\frac{1}{2}\bm{v}^{2})d\bm{X} are constant in time. Hence \mathcal{H}^{(k)} is bounded from below. When the domain is surrounded by the impermeable surface of a resting heat bath with a uniform constant temperature T_{w} , put \beta=(RT_{w})^{-1} . Then, ( 54 ) implies that RT_{w}(\mathcal{H}^{(k)}+\int_{D}\langle(\bm{\xi}^{2}/2)f\rangle d\bm{X}) , in place of \mathcal{H}^{(k)} , is bounded from below. Next consider the part \mathcal{H}^{(c)} for the van der Waals case ( 32 ) and for the Carnahan–Starling case ( 33 ). For the van der Waals case, \int_{0}^{x}\mathcal{S}(y)dy=-\ln(2-x)+\ln 2\geq 0 , so that \mathcal{H}^{(c)}\geq 0 and is bounded from below. For Carnahan–Starling case, \mathcal{H}^{(c)} is again bounded from below, since \int_{0}^{x}\mathcal{S}(y)dy=16(12-x)/(8-x)^{2}-3\geq 0 . To summarize, \mathcal{H} (or \mathcal{F} ) is bounded from below. Hence, if \mathcal{H} (or \mathcal{F} ) is bounded at the initial time, it is also bounded entirely in time and approaches a stationary value as t\to\infty . Appendix E Extension to the Enskog–Vlasov equation In the case of the Enskog–Vlasov equation, an external force term F_{i}{\partial f}/{\partial\xi_{i}} is added on the left-hand side of ( 1a ), where F_{i}=-\int_{D}\frac{\partial}{\partial X_{i}}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}, (55) and \Phi is the attractive isotropic force potential between molecules. Since the (1+{\ln{f}}) -moment of the external force term vanishes as \langle(1+{\ln f})F_{i}\frac{\partial f}{\partial\xi_{i}}\rangle=\langle F_{i}\frac{\partial}{\partial\xi_{i}}(f{\ln f})\rangle=0, (56) the external force term does not contribute to ( 7 ). Hence, ( 8 ), ( 16a ), and eventually ( 17 ) for the periodic domain and ( 18 ) for the domain surrounded by the specular reflection boundary remain unchanged. Next consider the multiplication of ( 1a ) with (1+\ln({f}/f_{w})) in the case of the domain surrounded by the impermeable surface of a heat bath with a uniform constant temperature T_{w} : \displaystyle\langle(1+\ln\frac{f}{f_{w}})F_{i}\frac{\partial f}{\partial\xi_{i}}\rangle \displaystyle=-\langle({\ln{f_{w}}})F_{i}\frac{\partial f}{\partial\xi_{i}}\rangle \displaystyle=F_{i}\langle\frac{\bm{\xi}^{2}}{2RT_{w}}\frac{\partial f}{\partial\xi_{i}}\rangle \displaystyle=-\frac{\rho v_{i}F_{i}}{RT_{w}}. (57) Since F_{i} is given by ( 55 ), \displaystyle-\int_{D}\frac{\rho v_{i}F_{i}}{RT_{w}}d\bm{X} \displaystyle= \displaystyle\int_{D}\frac{\rho v_{i}}{RT_{w}}\frac{\partial}{\partial X_{i}}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}d\bm{X} \displaystyle= \displaystyle-\int_{\partial D}\frac{\rho v_{i}}{RT_{w}}n_{i}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}dS \displaystyle-\int_{D}\frac{1}{RT_{w}}\frac{\partial(\rho v_{i})}{\partial X_{i}}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}d\bm{X} \displaystyle= \displaystyle\int_{D}\frac{1}{RT_{w}}\frac{\partial\rho(\bm{X})}{\partial t}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}d\bm{X} \displaystyle= \displaystyle\frac{1}{2}\frac{d}{dt}\int_{D\times D}\frac{\Phi(|\bm{Y}-\bm{X}|)}{RT_{w}}\rho(\bm{X})\rho(\bm{Y})d\bm{X}d\bm{Y}, (58) where v_{i}n_{i}=0 on \partial D and the continuity equation ( 41a ) have been used. This result implies that, in place of \mathcal{H} [or \mathcal{F} defined by ( 21b )], \widetilde{\mathcal{H}}\equiv\mathcal{H}+\frac{1}{2}\int_{D\times D}\frac{\Phi(|\bm{Y}-\bm{X}|)}{RT_{w}}\rho(\bm{X})\rho(\bm{Y})d\bm{X}d\bm{Y}, (59a) or \widetilde{\mathcal{F}}\equiv\mathcal{F}+\frac{1}{2}\int_{D\times D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{X})\rho(\bm{Y})d\bm{X}d\bm{Y}, (59b) decreases monotonically in time in the case of the Enskog–Vlasov version of the EESM. The contribution of the Vlasov term to the momentum conservation is written as \langle\xi_{j}F_{i}\frac{\partial f}{\partial\xi_{i}}\rangle=-\rho F_{j}=\rho(\bm{X})\int_{D}\frac{\partial}{\partial X_{j}}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}, (60) for \bm{X}\in D . Since \bm{Y} can be expressed in two ways \bm{Y}=\bm{X}\pm\bm{r} , it follows that \displaystyle\rho(\bm{X})\int_{D}\frac{\partial}{\partial X_{j}}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{1}{2}\int\frac{r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)[\rho(\bm{X})\rho(\bm{X}_{\bm{r}}^{-})\chi_{D}(\bm{X}_{\bm{r}}^{-})\chi_{D}(\bm{X})-\rho(\bm{X})\rho(\bm{X}_{\bm{r}}^{+})\chi_{D}(\bm{X}_{\bm{r}}^{+})\chi_{D}(\bm{X})]d\bm{r} \displaystyle= \displaystyle-\frac{1}{2}\int\frac{r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)\int_{0}^{1}\frac{\partial}{\partial\lambda}[\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})]d\lambda d\bm{r} \displaystyle= \displaystyle-\frac{1}{2}\int\frac{r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)\int_{0}^{1}r_{i}\frac{\partial}{\partial X_{i}}[\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})]d\lambda d\bm{r} \displaystyle= \displaystyle-\frac{1}{2}\frac{\partial}{\partial X_{i}}\int\frac{r_{i}r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)\int_{0}^{1}\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})d\lambda d\bm{r}, (61) where \Phi^{\prime}(x)=d\Phi(x)/dx , which is non-negative because \Phi is the attractive potential. Similarly, the contribution to the energy conservation is written as \displaystyle\langle\frac{1}{2}\bm{\xi}^{2}F_{i}\frac{\partial f}{\partial\xi_{i}}\rangle \displaystyle= \displaystyle-\rho v_{i}F_{i} \displaystyle= \displaystyle\rho(\bm{X})v_{i}(\bm{X})\int_{D}\frac{\partial}{\partial X_{i}}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}] \displaystyle-\frac{\partial\rho v_{i}}{\partial X_{i}}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}] \displaystyle+\frac{\partial\rho(\bm{X})}{\partial t}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}] \displaystyle+\frac{1}{2}\int_{D}\Phi(|\bm{Y}-\bm{X}|)[\frac{\partial\rho(\bm{X})}{\partial t}\rho(\bm{Y})-\frac{\partial\rho(\bm{Y})}{\partial t}\rho(\bm{X})]d\bm{Y} \displaystyle+\frac{1}{2}\int_{D}\Phi(|\bm{Y}-\bm{X}|)[\frac{\partial\rho(\bm{X})}{\partial t}\rho(\bm{Y})+\frac{\partial\rho(\bm{Y})}{\partial t}\rho(\bm{X})]d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}] \displaystyle+\frac{1}{2}\int_{\mathbb{R}^{3}}\Phi(|\bm{r}|)[\frac{\partial\rho(\bm{X})}{\partial t}\rho(\bm{X}^{+}_{\bm{r}})\chi_{D}(\bm{X})\chi_{D}(\bm{X}^{+}_{\bm{r}}) \displaystyle\qquad-\frac{\partial\rho(\bm{X}^{-}_{\bm{r}})}{\partial t}\rho(\bm{X})\chi_{D}(\bm{X})\chi_{D}(\bm{X}^{-}_{\bm{r}})]d\bm{r} \displaystyle+\frac{1}{2}\frac{\partial}{\partial t}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{X})\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}] \displaystyle+\frac{1}{2}\int_{\mathbb{R}^{3}}\int_{0}^{1}\Phi(|\bm{r}|)\frac{\partial}{\partial\lambda}[\frac{\partial\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})}{\partial t}\rho(\bm{X}^{+}_{\lambda\bm{r}}) \displaystyle\quad\times\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})]d\lambda d\bm{r} \displaystyle+\frac{1}{2}\frac{\partial}{\partial t}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{X})\rho(\bm{Y})d\bm{Y} \displaystyle= \displaystyle\frac{\partial}{\partial X_{i}}[\rho(\bm{X})v_{i}(\bm{X})\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y} \displaystyle+\frac{1}{2}\int_{\mathbb{R}^{3}}\int_{0}^{1}r_{i}\Phi(|\bm{r}|)\frac{\partial\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})}{\partial t}\rho(\bm{X}^{+}_{\lambda\bm{r}}) \displaystyle\qquad\times\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})d\lambda d\bm{r}] \displaystyle+\frac{1}{2}\frac{\partial}{\partial t}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{X})\rho(\bm{Y})d\bm{Y}. (62) These results lead to the concept of additional contributions to the stress tensor, the heat-flow vector, and the internal energy from the Vlasov term. If respectively denoted by p_{ij}^{(v)} , q_{i}^{(v)} , and e^{(v)} , they are expressed as follows: \displaystyle p_{ij}^{(v)}= \displaystyle-\frac{1}{2}\int_{\mathbb{R}^{3}}\frac{r_{i}r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)\int_{0}^{1}\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}}) \displaystyle\qquad\times\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})d\lambda d\bm{r}, (63a) \displaystyle q_{i}^{(v)}= \displaystyle\frac{1}{2}\int_{\mathbb{R}^{3}}r_{i}\Phi(|\bm{r}|)\int_{0}^{1}\frac{\partial\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})}{\partial t}\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}}) \displaystyle\qquad\times\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}})d\lambda d\bm{r} \displaystyle+\frac{1}{2}v_{j}\int_{\mathbb{R}^{3}}\int\frac{r_{i}r_{j}}{|\bm{r}|}\Phi^{\prime}(|\bm{r}|)\int_{0}^{1}\rho(\bm{X}^{+}_{\lambda\bm{r}})\chi_{D}(\bm{X}^{+}_{\lambda\bm{r}}) \displaystyle\qquad\times\rho(\bm{X}^{+}_{(\lambda-1)\bm{r}})\chi_{D}(\bm{X}^{+}_{(\lambda-1)\bm{r}})d\lambda d\bm{r} \displaystyle+\frac{1}{2}\rho v_{i}\int_{D}\Phi(|\bm{X}-\bm{Y}|)\rho(\bm{Y})d\bm{Y}, (63b) \displaystyle e^{(v)}= \displaystyle\frac{1}{2}\int_{D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{Y})d\bm{Y}. (63c) Hence, by redefining the stress tensor, heat-vector, and specific internal energy as \displaystyle p_{ij}=p_{ij}^{(k)}+p_{ij}^{(c)}+p_{ij}^{(v)},\quad q_{i}=q_{i}^{(k)}+q_{i}^{(c)}+q_{i}^{(v)}, \displaystyle e=e^{(k)}+e^{(v)}, (64) ( 50a ) and ( 50b ) are recovered and form the usual system of the conservation equations, together with the continuity equation ( 41a ) that remains unchanged. [ 40 ] In the uniform equilibrium state in the bulk, p_{ij}^{(v)} and e^{(v)} are reduced to \displaystyle p_{ij}^{(v)}=-\frac{2\pi}{3}\rho^{2}\int_{0}^{\infty}{x}^{3}\Phi^{\prime}(x)dx\delta_{ij}, (65a) \displaystyle e^{(v)}=2\pi\rho\int_{0}^{\infty}x^{2}\Phi(x)dx, (65b) and thus the attractive part p^{(v)} defined by p^{(v)}=-a\rho^{2},\quad a\equiv\frac{2\pi}{3}\int_{0}^{\infty}x^{3}\Phi^{\prime}(x)dx(>0), (66) is added to the right-hand side of the EoS, e.g., ( 30 ) and ( 31 ). In this way, the attractive part of the EoS, if exists, is recovered by the Vlasov term. Incidentally, as far as x^{3}\Phi(x)=0 for x=0 and x\to\infty , e^{(v)} can be rewritten as e^{(v)}=-a\rho and this is consistent with the equilibrium-thermodynamic relation e=e_{\mathrm{ideal}}+\int(p-T\partial p/\partial T)/\rho^{2}d\rho , where e_{\mathrm{ideal}}=(3/2)RT is the internal energy of ideal monatomic gases. Finally, if \Phi is bounded from below, it holds that \int_{D\times D}\Phi(|\bm{Y}-\bm{X}|)\rho(\bm{X})\rho(\bm{Y})d\bm{X}d\bm{Y}\geq C(\int_{D}\rho d\bm{X})^{2}, (67) where C is a certain constant. Since \int_{D}\rho d\bm{X} is the total mass, it is conserved in time. Thus, the boundedness of \mathcal{H} and \mathcal{F} shown in Appendix D is extended to that of \widetilde{\mathcal{H}} and \widetilde{\mathcal{F}} in the case of the Enskog–Vlasov equation. Acknowledgements. The research of the first author is supported in part by the JSPS Grant-in-Aid for Scientific Research(C) (No. 22K03923). The second author is a JSPS Research Fellowship DC1 and his research is supported by the JSPS Grant-in-Aid for JSPS Fellows (No. 24KJ1450).