1: \section{Luo model: LBE for non-ideal dense gases}
2: The model of Li-Shi Luo starts from Boltzmann equation for dense gases
3: \cite{luo:unified-98}, also known as the Enskog equation:
4: \begin{equation}
5: \frac{\partial f}{\partial t}+\xi.\nabla f+{\bf a}.\nabla_\xi f = \Omega
6: \end{equation}
7: where ${\bf a}$ and $\Omega$ are the
8: acceleration and collision operators. With the expansion of
9: $\Omega$ in a Taylor series, using
10: a single relaxation time approximation, and assuming the
11: fluid to be
12: isothermal and incompressible, Luo obtains the
13: following equation
14: \begin{subequations}
15: \label{eq:Luo-Enskog}
16: \begin{align}
17: \frac{\partial f}{\partial t}+\xi.\nabla f+{\bf a}.\nabla_\xi f = -\frac g \lambda
18: (f-f^{eq})+J \\
19: J = -f^{eq}b\rho g(\xi-{\bf u}).\nabla ln(\rho^2g)
20: \end{align}
21: \end{subequations}
22: where $b$ is the second virial coefficient
23: of the equation of
24: state for the hard-sphere system, and
25: $g$ is the radial distribution function.
26: The equilibrium distribution function is discussed
27: by Luo in the continuum form
28: \begin{equation}
29: \label{eqn:int-fcont-Luo}
30: f^{eq}=\frac{\rho}{(2\pi RT)^{D/2}}\exp\left[-\frac{(\xi-{\bf u})^2}{2RT}\right]
31: \end{equation}
32: but is used in the discretized form
33: (\ref{eq:equilibrium-distribution-function-standard}).
34: \\\\
35: Solving (\ref{eq:Luo-Enskog}) to first
36: order in time we have
37: \begin{align}
38: \label{eqluo:LBE}
39: f(x+\xi\delta_t,\xi,t+\delta_t)-f(x,\xi,t) =
40: &-\frac 1\tau\left(f(x,\xi,t)-f^{eq}(x,\xi,t)\right) \nonumber \\
41: &+J(x,\xi,t)\delta_t-{\bf a}.\nabla_\xi f(x,\xi,t)\delta_t
42: \end{align}
43: with a
44: force term, ${\bf a}.\nabla_\xi f$, written
45: \cite{luo:unified-98}
46: \begin{equation}\label{eqluo:force}
47: {\bf a}.\nabla_\xi f = -\rho\omega(\xi)\xi_T^{-2}\left[(\xi-{\bf u})+
48: \xi_T^{-2}(\xi.{\bf u})\xi\right].{\bf a}
49: \end{equation}
50: The auxiliary variable
51: $\xi_T$ is equal to $\sqrt{R T}$ ($\frac{c}{\sqrt{3}}$
52: in the $D2Q9$ model), and the weight
53: weight $\omega(\xi)$ denotes the weight $w_i$ for
54: the momentum $\xi=e_i$.
55: The force term $J$ is proportional to
56: the gradient of the function
57: of density $\rho^2 g$, which as in the Swift model is
58: evaluated by discrete differences over lattice
59: points. The interpretation is also that of a potential
60: term.
61: Equations (\ref{eqluo:LBE}) and (\ref{eqluo:force})
62: together with
63: (\ref{eq:equilibrium-distribution-function-standard}),
64: (\ref{eq:macro-conservation}),
65: (\ref{eq:macro-momentum}), and a choice of a lattice,
66: define Luo's
67: Lattice Boltzmann model.
68: \\\\
69: To proceed further we want to evaluate
70: moments with respect to momentum
71: of the two ``forces'' (the interior force $J$,
72: and the exterior force through the acceleration $a$).
73: The forcing term of
74: eqn.\eqref{eqluo:force} is
75: \begin{equation}
76: F_i = -w_i\rho\left[\frac{1}{\Upsilon^{(2)}}
77: (e_i-{\bf u})+\frac{(e_i.{\bf u})}{\Upsilon^{(4)}
78: }e_i\right].{\bf a}
79: \end{equation}
80: and satisfies the following constraints:
81: \begin{subequations}\label{eqluo:sumFg}
82: \begin{align}
83: \label{eqluo:sumF}
84: \sum_iF_i & = 0 \\
85: \label{eqluo:sumFe}
86: \sum_i(F_ie_{i\alpha}) &=-\frac{\rho}{\Upsilon^{(2)}}
87: \sum_i(w_ie_{i\alpha}e_{i\beta}).{\bf a}
88: =-\rho a_\alpha \\
89: \label{eqluo:sumFee}
90: \sum_i(F_ie_{i\alpha}e_{i\beta}) &=
91: -\left[-\frac{\rho u_\gamma}{\Upsilon^{(2)}}
92: \sum_i(e_{i\alpha}e_{i\beta}w_i)
93: +\frac{\rho}{\Upsilon^{(4)}}\sum_i\big(({\bf e_i}.{\bf u})e_{i\alpha}e_{i\beta}e_{i\gamma}w_i\big)
94: \right].{\bf a} \nonumber\\
95: &=\rho(u_\alpha a_\beta-u_\beta a_\alpha)
96: \end{align}
97: \end{subequations}
98: The interior force
99: term $J_i$ satisfies the following constraints:
100: \begin{subequations}\label{eqluo:sumJg}
101: \begin{align}
102: \label{eqluo:sumJ}
103: \sum_iJ_i &= -b\rho g\left(\sum_i(f_i^{eq}e_{i\alpha})-\sum_i(f_i^{eq}u_\alpha)
104: \right).\nabla ln(\rho^2g) \nonumber\\
105: &= -b\rho g(\rho u-\rho u).\nabla ln(\rho^2g) = 0
106: \end{align}
107: \begin{align}
108: \sum_i(J_i e_{i\alpha}) &=
109: -b\rho g\left(\sum_i(f_i^{eq}e_{i\alpha}e_{i\beta})
110: -\sum_i(f_i^{eq}e_{i\alpha})u_\beta\right)
111: \partial_{\beta} \log (\rho^2g) \nonumber\\
112: \label{eqluo:sumJe}
113: &= -b \Upsilon^{(2)}\partial_\alpha(\rho^2g)
114: \end{align}
115: \begin{align}
116: \sum_i(J_i e_{i\alpha}e_{i\beta}) &=
117: -b\rho g\left(\sum_i(f_i^{eq}e_{i\alpha}e_{i\beta}e_{i\gamma})
118: -\sum_i(f_i^{eq}e_{i\alpha}e_{i\beta})u_\beta\right).\nabla ln(\rho^2g) \nonumber\\
119: \label{eqluo:sumJee}
120: &= b[u_\alpha u_\beta {\bf u}.\nabla -
121: \Upsilon^{(2)}(u_\alpha\partial_\beta+u_\beta\partial_\alpha)]
122: (\rho^2g)
123: \end{align}
124: \end{subequations}
125:
126:
127:
128:
129: %%%%%%%%%%%%%%%%%%%%%%%%%%%
130: %%%%% Chapman-Enskog %%%%%
131: %%%%%%%%%%%%%%%%%%%%%%%%%%%
132: \subsection{Chapman-Enskog analysis}\label{sec:Luo-CE}
133: In this section we will derive the macroscopic equations with the Chapman-Enskog
134: method. The earlier presented method of successive approximations can be considered
135: a special case of the more general and more flexible Chapman-Enskog method.
136: First, we apply the Taylor expansion to equation
137: (\ref{eqluo:LBE})
138: \begin{equation} \label{eqluo:chapmann1}
139: \sum_{n=1}^\infty\frac{\epsilon^n}{n!}\D_t^n f_i({\bf x}, t) =
140: -\frac g\tau\big(f_i({\bf x},t)- f_i^{eq}({\bf x},t)\big)+J_i\epsilon-F_i\epsilon
141: \end{equation}
142: where $\epsilon=\delta_t$ and $\D\equiv(\partial_t+e_{i\alpha}\partial_\alpha)$.
143: We now introduce the following expansion:
144: \begin{subequations}
145: \begin{align}
146: f_i=&\sum_{n=0}^\infty\epsilon^nf_i^{(n)} \\
147: \partial_t =& \sum_{n=0}^\infty\epsilon^n\partial_{nt} \\
148: \label{eqluo:spatial1}
149: \partial_\alpha =& \partial_{0\alpha}
150: \end{align}
151: \end{subequations}
152: The meaning of the expansion of $f$ in powers of $\epsilon$ is clear. The meaning of the formal expansion of the time derivatives in derivatives with respect to different ``times'' is that $f$ is assumed to have support in the frequency domain on bands that are separated by powers of $\epsilon$. We will look for an equation involving a ``slow time'', that is low-frequency components of $f$. Such components are Fourier transforms of $f$, where sufficienty fast oscillations are averaged out. A rigorous theory of these procedures can be given in homogenization theory \cite{Piretal}. In general one can also assume there to be fast and slow spatial variables (small and large scales).
153: For the present problem it is however enough to assume one spatial scale, hence \eqref{eqluo:spatial1}
154: \\\\
155: With $\epsilon^2$ order accuracy, equation \eqref{eqluo:chapmann1}
156: can be rewritten:
157: \begin{align}
158: \label{eqluo:chapmann2}
159: f_i\equiv f_i^{(0)}+\epsilon f_i^{(1)}+\epsilon^2 f_i^{(2)}=f_i^{eq}-
160: \frac\tau g\left(\epsilon\D_t+
161: \frac{\epsilon^2}{2}\D_t^2\right)f_i+\frac\tau gJ_i\epsilon-\frac\tau gF_i\epsilon
162: \end{align}
163: Group the right-hand side of the eqn.\eqref{eqluo:chapmann2} by the order of $\epsilon$.
164: \begin{align}
165: \label{eqluo:chapmann3}
166: f_i^{(0)}+\epsilon f_i^{(1)}+\epsilon^2 f_i^{(2)}=
167: &f_i^{eq}-\epsilon\frac\tau g\D_t\big(f_i^{(0)}+\epsilon f_i^{(1)}+\epsilon^2 f_i^{(2)}\big)
168: \nonumber \\
169: &-\epsilon^2\frac{\tau}{2g}\D_t^2
170: \big(f_i^{(0)}+\epsilon f_i^{(1)}+\epsilon^2 f_i^{(2)}\big) \nonumber \\
171: &+\epsilon\frac\tau gJ_i-\epsilon\frac\tau gF_i
172: \end{align}
173: To successive orders in
174: $\epsilon$
175: we find:
176: \begin{subequations}\label{eqluo:dis-fg}
177: \begin{align}
178: \label{eqluo:dis-f0}
179: \epsilon^0:& \hspace{1cm} f_i^{(0)}=f_i^{eq} \\
180: \label{eqluo:dis-f1}
181: \epsilon^1:& \hspace{1cm} f_i^{(1)}=-\frac\tau g\D_{t_0}f_i^{(0)}+\frac\tau gJ_i- \frac\tau gF_i\\
182: \label{eqluo:dis-f2}
183: \epsilon^2:& \hspace{1cm} f_i^{(2)}=-\frac{\tau}{2g}\D_{t_0}^2f_i^{(0)}-
184: \frac\tau g\partial_{t_1}f_i^{(0)}-\frac\tau g\D_{t_0}f_i^{(1)}
185: \end{align}
186: \end{subequations}
187: The goal of the calculation is to project on the hydrodynamic
188: modes of density and momentum. Equation (\ref{eqluo:dis-f0})
189: shows that the zeroth order distribution function agrees
190: with the equilbrium distribution function, which shares
191: the same density and momentum. The higher order
192: distribution functions therefore obey
193: the following constraints:
194: \begin{eqnarray}\label{eqluo:sum-fng}
195: \sum_i f_i^{(n)} &=& 0 \qquad(n>0) \\
196: \label{eqluo:sum-fne}
197: \sum_i (f_i^{(n)}e_{i\alpha}) &=&
198: 0 \qquad (n> 0)
199: \end{eqnarray}
200: Summing equation \eqref{eqluo:dis-f1} over $i$, using equations \eqref{eqluo:sumF}, and
201: \eqref{eqluo:sumJ}, gives
202:
203: \begin{equation}\label{eqluo:continum1}
204: \partial_{t_0}\rho+\partial_{0\alpha}(\rho u_\alpha) = 0
205: \end{equation}
206: To second order in
207: $\epsilon$, we sum equation \eqref{eqluo:dis-f2}
208: over $i$, substituting the first order solution \eqref{eqluo:dis-f1}
209: which gives
210: %%%%%%%%%%%%%%%%%%%%%%
211: \begin{equation}
212: 0=-\frac 1 2\left[\sum_i\D_{t_0}J_i-\sum_i\D_{t_0}F_i\right]
213: +\left(\frac 1 2-\frac g\tau\right)\sum_i\D_{t_0}f_i^{(1)}-\partial_{t_1}\sum_if_i^{(0)}
214: \end{equation}
215: %%%%%%%%%%%%%%%%%%%%%%
216: or
217: %%%%%%%%%%%%%%%%%%%%%%
218: \begin{equation}
219: \partial_{t_1}\rho = -\frac 1 2\partial_\alpha\left(\rho a_\alpha -
220: b\Upsilon^{(2)}\partial_\alpha(\rho^2g)\right)
221: \end{equation}
222: %%%%%%%%%%%%%%%%%%%%%%
223: Combining the first and the second order results for $f_i$ by
224: $\partial_t=\partial_{t_0}+\epsilon\partial_{t_1}$ and recalling that $\epsilon=\delta_t$,
225: we have a continuity equation with an (unphysical)
226: density diffusion term:
227: \footnote{The author of \cite{luo:unified-98}
228: claims there is no such density diffusion term,
229: but we find that it is there.}
230:
231: %%%%%%%%%%%%%%%%%%%%%%
232: \begin{align}\label{eqluo:continuity}
233: \partial_t\rho + \partial_\alpha(\rho u_\alpha) =
234: -\frac{\delta_t}{2}\partial_\alpha\left(\rho a_\alpha
235: -b\Upsilon^{(2)}\partial_\alpha(\rho^2g)\right)
236: \end{align}
237: %%%%%%%%%%%%%%%%%%%%%%
238: The first order of the
239: momentum equation can be produced by multiplying equation \eqref{eqluo:dis-f1} with
240: $e_{i\beta}$, then summing over $i$:
241: %%%%%%%%%%%%%%%%%%%%%%
242: \begin{align*}
243: 0 &= \sum_i\D_{t_0}(f_i^{(0)}e_{i\beta}) + \sum_i(J_ie_{i\beta})- \sum_i(F_ie_{i\beta})
244: \end{align*}
245: %%%%%%%%%%%%%%%%%%%%%%
246: By using equations \eqref{eqluo:sum-fng}, \eqref{eqluo:sumJe},
247: and \eqref{eqluo:sumFe}, we have
248: %%%%%%%%%%%%%%%%%%%%%%
249: \begin{align}\label{eqluo:momentum0}
250: \partial_{t_0}(\rho u_\beta) + \partial_\alpha(\rho u_\alpha u_\beta)
251: &= -\Upsilon^{(2)}\partial_\beta\rho-
252: b\Upsilon^{(2)}\partial_\beta(\rho^2g) +\rho a_\beta \nonumber\\
253: &=-\partial_\beta P+\rho a_\beta
254: \end{align}
255: %%%%%%%%%%%%%%%%%%%%%%
256: where
257: %%%%%%%%%%%%%%%%%%%%%%
258: \begin{align*}
259: P=\Upsilon^{(2)}\rho(1+b\rho g)
260: \end{align*}
261: is a non-ideal pressure. Equation \eqref{eqluo:momentum0}
262: is hence Euler's equation.
263: The second order momentum equation is
264: %%%%%%%%%%%%%%%%%%%%%%
265: \begin{align*}
266: 0 &= \partial_{t_1}\sum_if_i^{(0)}e_{i\beta}
267: +\frac 1 2\D_{t_0}\sum_i(J_ie_{i\beta}-F_ie_{i\beta})
268: +\left(1-\frac{g}{2\tau}\right)\D_{t_0}\sum_i(f_i^{(1)}e_{i\beta})
269: \end{align*}
270: %%%%%%%%%%%%%%%%%%%%%%
271: Subtituting the equations \eqref{eqluo:sum-fne}, \eqref{eqluo:sumFg}
272: and \eqref{eqluo:sumJg} into above we have
273: %%%%%%%%%%%%%%%%%%%%%%
274: \begin{equation}\label{eqluo:momentum1}
275: \partial_{t_1}(\rho u_\beta) =
276: \left(\frac{g}{2\tau}-1\right)\partial_\alpha\Pi_{\alpha\beta}^{(1)}-
277: \frac 1 2\D_{t_0}\sum_i(J_ie_{i\beta}-F_ie_{i\beta})
278: \end{equation}
279: %%%%%%%%%%%%%%%%%%%%%%
280: where the first-order momentum flux tensor
281: $\Pi_{\alpha\beta}^{(1)}$ is $\sum_if_i^{(1)}e_{i\alpha}e_{i\beta}$
282: and therefore
283: %%%%%%%%%%%%%%%%%%%%%%
284: \begin{align}
285: \partial_\alpha\Pi_{\alpha\beta}^{(1)}
286: &=\partial_\alpha\sum_if_i^{(1)}e_{i\alpha}e_{i\beta} \nonumber \\
287: &=-\frac{\tau}{g}\partial_\alpha\sum_i\D_{t_0}\underbrace{f_i^{(0)}e_{i\alpha}e_{i\beta}}
288: _{\Pi_{\alpha\beta}^{(0)}}
289: +\frac{\tau}{g}\partial_\alpha\sum_i(J_ie_{i\alpha}e_{i\beta}-F_ie_{i\alpha}e_{i\beta})
290: \end{align}
291: %%%%%%%%%%%%%%%%%%%%%%
292: where $\Pi_{\alpha\beta}^{(0)}$ is the zeroth-order momentum flux tensor.
293: Furthermore
294: %%%%%%%%%%%%%%%%%%%%%%
295: \begin{align}\label{eqluo:Pi0}
296: \D_{t_0}\Pi_{\alpha\beta}^{(0)}
297: =& \partial_{t_0}\left(\Upsilon^{(2)}\rho\delta_{\alpha\beta}+\rho u_\alpha u_\beta\right)+
298: \Upsilon^{(2)}
299: \big[\partial_\alpha(\rho u_\alpha)\delta_{\alpha\beta}+
300: \partial_\alpha(\rho u_\beta)+
301: \partial_\beta(\rho u_\alpha)\big]
302: \nonumber \\
303: =& \Upsilon^{(2)}
304: \big[\partial_{t_0}\rho+\partial_\alpha(\rho u_\alpha)\big]\delta_{\alpha\beta}+
305: \partial_{t_0}(\rho u_\alpha u_\beta)+
306: \Upsilon^{(2)}
307: \big[\partial_\alpha(\rho u_\beta)+\partial_\beta(\rho u_\alpha)\big]
308: \nonumber \\
309: =& \partial_{t_0}(\rho u_\alpha u_\beta)+
310: \Upsilon^{(2)}
311: \big[\partial_\alpha(\rho u_\beta)+\partial_\beta(\rho u_\alpha)\big]+
312: \Order(\delta_t)
313: \end{align}
314: %%%%%%%%%%%%%%%%%%%%%%
315: Substituting back the above results into \eqref{eqluo:momentum1}, we have,
316: %%%%%%%%%%%%%%%%%%%%%%
317: \begin{align}\label{eqluo:momentum2}
318: \partial_{t_1}(\rho u_\beta)
319: =& \Upsilon^{(2)}\frac{2\tau-g}{2g}\partial_\alpha
320: \big[\partial_\alpha(\rho u_\beta)+\partial_\beta(\rho u_\alpha)\big]+
321: \frac{2\tau-g}{2g}\partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)+
322: \nonumber \\
323: & \frac{2\tau-g}{2g}\partial_{t_0}\sum_i(J_ie_{i\beta}-F_ie_{i\beta})+
324: \frac 1 2\partial_\alpha\sum_i(J_ie_{i\alpha}e_{i\beta}-F_ie_{i\alpha}e_{i\beta})
325: \end{align}
326: %%%%%%%%%%%%%%%%%%%%%%
327: Combining the first and the second order results, equations \eqref{eqluo:momentum0} and
328: \eqref{eqluo:momentum2}, for $f_i$ by $\partial_t=\partial_{t_0}+\epsilon\partial_{t_1}$
329: and recalling that $\epsilon=\delta_t$, we then have the
330: Navier-Stokes level
331: momentum equation,
332: %%%%%%%%%%%%%%%%%%%%%%
333: \begin{align}\label{eqluo:momentum3}
334: \partial_t(\rho u_\beta) + \partial_\alpha(\rho u_\alpha u_\beta) =
335: -\partial_\beta P
336: +\nu\partial_\alpha\big[\partial_\alpha(\rho u_\beta)+\partial_\beta(\rho u_\alpha)\big]
337: +\rho a_\beta +\mathcal{E}_\beta
338: \end{align}
339: %%%%%%%%%%%%%%%%%%%%%%
340: where the kinematic viscosity is
341: %%%%%%%%%%%%%%%%%%%%%%
342: \begin{align*}
343: \nu =& \Upsilon^{(2)}\frac{2\tau-g}{2g}\delta_t
344: \qquad \left(\,= c^2\frac{2\tau-g}{6g}\delta_t
345: \hbox{in the $D2Q9$ model}\right)
346: \end{align*}
347: %%%%%%%%%%%%%%%%%%%%%%
348: and various extra viscous-order terms
349: are grouped together in $\mathcal{E}_\beta$:
350: %%%%%%%%%%%%%%%%%%%%%%
351: \begin{align}
352: \mathcal{E}_\beta
353: =& \delta_t\frac{2\tau-g}{2g}\partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)+
354: \nonumber \\
355: & \delta_t\frac{2\tau-g}{2g}\partial_{t_0}\sum_i(J_ie_{i\beta}-F_ie_{i\beta})+
356: \delta_t\frac 1 2\partial_\alpha\sum_i(J_ie_{i\alpha}e_{i\beta}-F_ie_{i\alpha}e_{i\beta})
357: \nonumber \\
358: =& \delta_t\frac{2\tau-g}{2g}\partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)+
359: \delta_t\frac{2\tau-g}{2g}\partial_{t_0}
360: \left[\frac{c^2}{3}\partial_\beta(\rho^2 g)-\rho a_\beta\right]+
361: \nonumber \\
362: & \delta_t\frac 1 2\partial_\alpha
363: \left\{b[u_\alpha u_\beta{\bf u.\nabla}-
364: \Upsilon^{(2)}
365: (u_\alpha\partial_\beta+u_\beta\partial_\alpha)](\rho g^2)+
366: \rho(u_\alpha a_\beta+u_\beta a_\alpha)\right\}
367: \end{align}
368: %%%%%%%%%%%%%%%%%%%%%%
369: This momentum equation is not Galilean invariant
370: when density gradients are present, and
371: contains some unphysical terms. If we assume that Mach number
372: is small ($M$ formally of the same order as $\epsilon$
373: in the expansion),
374: then the extra term $\partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)$ in $ \mathcal{E}_\beta$
375: can be further simplified as:
376: %%%%%%%%%%%%%%%%%%%%%%
377: \begin{align}\label{eqluo:Ebeta1}
378: \partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)
379: =& \partial_\alpha\big\{u_\alpha\big[\partial_{t_0}(\rho u_\beta)-
380: u_\beta\partial_{t_0}\rho\big]+u_\beta\partial_{t_0}(\rho u_\alpha)\big\}
381: \nonumber\\
382: =& \partial_\alpha\big\{-u_\alpha\partial_\gamma(\rho u_\beta u_\gamma)
383: +u_\alpha u_\beta\partial_\gamma(\rho u_\gamma)
384: -u_\beta\partial\gamma(\rho u_\alpha u_\gamma) \nonumber\\
385: & -(u_\alpha\partial_\beta P+u_\beta\partial_\alpha P)
386: +\rho(u_\beta a_\alpha+u_\alpha a_\beta)\big\}+\Order(\delta_t)
387: \end{align}
388: %%%%%%%%%%%%%%%%%%%%%%
389: Terms of cubic order in Mach number
390: ($\Order({\bf u}^3)$)
391: can now be neglected, and
392: we simplify to
393: %%%%%%%%%%%%%%%%%%%%%%
394: \begin{align}\label{eqluo:Ebeta2}
395: \partial_\alpha\partial_{t_0}(\rho u_\alpha u_\beta)
396: =& -\Upsilon^{(2)}\partial_\alpha(u_\alpha\partial_\beta\rho+u_\beta\partial_\alpha\rho)
397: -b\Upsilon^{(2)}
398: \partial_\alpha[u_\alpha\partial_\beta(\rho^2g)+u_\beta\partial_\alpha(\rho^2g)]
399: \nonumber\\
400: & +\partial_\alpha[\rho(u_\beta a_\alpha+u_\alpha a_\beta)]
401: +\Order(\delta_t) + \Order({\bf u}^3)
402: \end{align}
403: %%%%%%%%%%%%%%%%%%%%%%
404: which allows the extra terms in the momentum equation to be
405: rewritten as
406: with
407: \begin{align}
408: \mathcal{E}_\beta =
409: \nu\partial_{\alpha}(u_{\alpha}\partial_\beta(\rho^2g))-
410: \delta_t\frac{2\tau-g}{2g}\partial_{t_0}(\rho a_\beta)+
411: \delta_t\tau\partial_\alpha[\rho(u_\beta a_\alpha+u_\alpha a_\beta)]
412: \end{align}
413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
414: \subsection{Summary}
415: \begin{itemize}
416: \item This model is correct to first order (inviscid equations),
417: but incorrect to second order (viscous equations).
418: The continuity equation contains density diffusion
419: terms, and the momentum equation various unphysical
420: terms, somewhat similar to the ones appearing in the
421: model of Swift and collaborators.
422: \end{itemize}
423: