Upload
independent
View
1
Download
0
Embed Size (px)
Citation preview
TRANSPORT THEORY AND STATISTICAL PHYSICS
Vol. 32, Nos. 5–7, pp. 639–650, 2003
Analytical Lattice Boltzmann Solutions for
Thermal Flow Problems
Dimitris Valougeorgis* and Stergios Naris
Department of Mechanical and Industrial Engineering,
University of Thessaly, Volos, Greece
ABSTRACT
Analytical solutions based on two 13-bit (hexagonal and square) and
one 17-bit square lattice Boltzmann BGK models have been obtained
for the Couette flow, with a temperature gradient at the boundaries.
The analytical solutions for the unknown distributions functions are
written as polynomials in powers of the space variable and the
coefficients of the expansion are estimated in terms of characteristic
flow quantities, the single relaxation time and the lattice spacing. The
analytical solutions of the two 13-bit models contain some nonlinear
deviations from the thermal hydrodynamic constraints and the
analytical solutions, while the 17-bit square lattice model results
into an exact representation of the nonisothermal Couette flow
problem.
*Correspondence: Dimitris Valougeorgis, Department of Mechanical and
Industrial Engineering, University of Thessaly, Volos, 38334, Greece; E-mail:
120025070_TT32_05-07_R1_080103
639
DOI: 10.1081/TT-120025070 0041-1450 (Print); 1532-2424 (Online)
Copyright & 2003 by Marcel Dekker, Inc. www.dekker.com
+ [1.8.2003–10:43am] [639–650] [Page No. 639] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
I. INTRODUCTION
In the last few years the lattice Boltzmann method (LBM) has beendeveloped into an effective computational scheme for a broad variety ofthermo-fluid physical systems (Chen and Doolen, 1998). Since the appear-ance of the LBM and its predecessor, the lattice gas automata (LGA) someanalytical solutions have been obtained for these methods for two andthree dimensional flows (Cornubert et al., 1991; Henon, 1987; Luo,1997; Luo et al., 1991). More recently, analytical solutions of the distribu-tion functions for Poiseuille and Couette flows are found for the triangularand square LBMmodels (He et al., 1997; Zou et al., 1995). All these resultsallow one to calculate the viscosity from given collision rules, to improvethe implemented boundary conditions and to justify the accuracy to expectfrom the method. Overall analytical LB approaches are enhancing ourunderstanding of the method. Nevertheless no analytical solution hasbeen previously reported for thermal flow problems. It is well knownthat one of the major shortcomings of the LBM is the lack of a satisfactorythermal model for heat transfer problems. One of the methodologies todevelop thermal lattice Boltzmann models is the so-called ‘‘multi-speedapproach’’ (Alexander et al., 1993; Chen et al., 1994). Although thisapproach has been shown to suffer from numerical instability(McNamara et al., 1995), some recent work has provided new alternativesand potential in this approach (Pavlo et al., 1998a; 1986b). Some of thevelocity discretization models studied in previous work include 13-bitmodels for either the hexagonal (Alexander et al., 1993) or the square(Qian, 1993) grids, as well as typical 17-bit square lattices.
In the present work an investigation on the accuracy to expect fromthe aforementioned multi-speed models is attempted. The nonisothermalCouette flow is chosen as a typical thermal flow model problem and ananalytical LBM formulation approach developed earlier (He et al., 1997;Zou et al., 1995) is extended to include heat transfer effects. Analyticalexpressions of the distribution functions are obtained and some guidanceis given for thermal flow applications.
The well-known lattice Boltzmann evolution equation is given by
f�, iðxþ e�, i�t, tþ �tÞ � f�, iðx, tÞ ¼ �1
�½ f�, iðx, tÞ � f
ð0Þ�, i ðx, tÞ� ð1Þ
where f�, iðx, tÞ is the distribution function of the particle of type ð�, iÞ atposition x and time t, f
ð0Þ�, i ðx, tÞ is the corresponding equilibrium function
of the particle, e�, i are the unit velocity vectors along the specified direc-tions and � is the single relaxation time, which controls the rate at whichthe system relaxes to the local equilibrium. All quantities in Eq. (1) are in
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 640] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
640 Valougeorgis and Naris
nondimensional form (Sterling and Chen, 1996). The choice of fð0Þ�, i ðx, tÞ is
critical in thermal lattice Boltzmann models. To accurately simulatehydrodynamic phenomena Eulerian and Navier–Stokesian descriptionsof real fluids must be fully recovered. In two dimensions this may beachieved by requiring that the moments of f
ð0Þ�, i ðx, tÞ satisfy the relations
(McNamara and Alder, 1993)X�, i
fð0Þ�, i ¼ n ð2aÞ
X�, i
e�, �, i fð0Þ�, i ¼ nu� ð2bÞ
X�, i
e�, �, ie�, �, i fð0Þ�, i ¼ nu�u� þ n"��� ð2cÞ
X�, i
e�, �, ie�, �, ie�, �, i fð0Þ�, i ¼ nu�u�u� þ n"ðu���� þ u���� þ u����Þ ð2dÞ
X�, i
e2�, ie�, �, ie�, �, i fð0Þ�, i ¼ ðu2 þ 6"Þnu�u� þ ðu2 þ 4"Þn"��� ð2eÞ
where Greek subscripts indicate Cartesian components. For a thermalfluid taking into account the symmetry of the moments under exchangeof any pairs of indices there are 13 such constrains in two dimensions(26 constraints in three dimensions). This suggests the need for at least13 different particle velocities in order to guarantee the linear indepen-dence of the left hand side of Eq. (2) and the full recovery of the thermalNavier–Stokes equations up to the fourth order terms. Typicalequilibrium function has the polynomial form (Pavlo et al., 1998b)
fð0Þ�, i ¼ n½A� þ B�ðe�, i � uÞ þ C�ðe�, i � uÞ
2þD�u
2þ E�ðe�, i � uÞ
3
þ F�ðe�, i � uÞu2þ G�u
4þH�ðe�, i � uÞ
2u2 þ I�ðe�, i � uÞ4� ð3Þ
where the coefficients are functions of the local density n ¼P
�, i f�, i andthe internal energy 2n" ¼
P�, i f�, iðe�, i � uÞ
2, while the bulk velocity isdefined by nu ¼
P�, i f�, ie�, i. The form of expression (3) is based on a
Taylor expansion of the Maxwellian equilibrium distribution in the localvelocity u keeping terms up to the fourth power. The coefficients of therelaxation distribution (3) are obtained in such a manner to remove dis-crete lattice effects and consequently the resulting relaxation distributionis not the Maxwellian.
The Couette thermal-flow problem under investigation consists of afluid contained between two plates, the upper one moving with velocity u0
AQ2
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 641] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
Analytical LBM for Thermal Flow Problems 641
and the lower one is stationary, while a temperature difference existsbetween the boundaries. The x and y components of the velocityu ¼ ðux, uyÞ and normalized energy profiles must satisfy
uxð yÞ ¼ u0y ð4aÞ
uyð yÞ ¼ 0 ð4bÞ
and
"ð yÞ ¼ yþBr
2u20yð1� yÞ
� �ð4cÞ
respectively, where y is the normalized distance from the lower plateand the Brinkman number Br is the product of the Prandtl and Eckertnumbers.
II. ANALYTICAL SOLUTIONS OF THE
13-BIT MODELS
First the 13-bit hexagonal lattice is considered. This model isconsisting of one rest particle,
e�i ¼ 0, ð5aÞ
for � ¼ 0, and two nonzero speeds for which
e�i ¼ � cos�ði � 1Þ
3, sin
�ði � 1Þ
3
� �, ð5bÞ
for � ¼ 1, 2 and i ¼ 1, 2, 3, 4, 5, 6. Taking into account the constrainsmentioned above one can easily solve for the unknown coefficients ofthe equillibrium distribution function. One possible solution is(Alexander et al., 1993):
A0 ¼ 1�5
2"þ 2"2, A1 ¼
4
9"�
4
9"2, A2 ¼ �
1
36"þ
1
9"2,
B1 ¼4
9�4
9", B2 ¼ �
1
36þ1
9",
C1 ¼8
9�4
3", C2 ¼ �
1
72þ
1
12",
D0 ¼ �5
4þ 2", D1 ¼ �
2
9þ2
9", D2 ¼
1
72�
1
18",
E1 ¼ �4
27, E2 ¼
1
108, F1 ¼ F2 ¼ 0:
ð6Þ
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 642] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
642 Valougeorgis and Naris
At this point we note that using the above set of estimates for thecoefficients only the zeroth, first, and second moments of the imposedconstraints corresponding to Eqs. (2a), (2b), and (2c) respectively aresatisfied, while for the third moments, given by Eq. (2d), cubic deviationsare present. Actually none of the possible solutions satisfy exactly therequired constraints.
Now suppose that there is a solution f�, iðx, tÞ of Eq. (1) that exactlyrepresents the Couette flow with a temperature gradient between theboundaries. The solution must contain the following properties:
f�, iðx, tÞ is time independent denoted by f�, iðxÞ ð7aÞ
f�, iðxÞ is a function of one space variable denoted by f�, ið yÞ ð7bÞ
X�, i
f�, ið yÞ ¼ n ð7cÞ
X�, i
f�, ið yÞex, �, i ¼ nuxð yÞ ¼ nu0y ð7dÞ
X�, i
f�, ið yÞey, �, i ¼ 0 ð7eÞ
X�, i
f�, ið yÞðe�, i � u
2Þ
2¼ n"ð yÞ ¼ n yþ
Br
2u20yð1� yÞ
� �ð7fÞ
Equations (7a) and (7b) are due to the fact that the particular flow underinvestigation is steady and fully developed. Equations (7c–7f ) are derivedusing the definitions of the local density, the x and y componentsof velocity and the internal energy respectively supplemented by thewell-known analytical velocity and temperature profiles given in Eqs. (4).
Using properties (7a) and (7b) for � ¼ 1, 2 and i ¼ 0, 1, 4, whichcorrespond to the rest particle and the two particles with motion alongthe x-axis, Eq. (1) may be written as
f�, ið yÞ ¼ f�, ið yÞ �1
�ð f�, ið yÞ � f
ð0Þ�, i ð yÞÞ: ð8Þ
Hence for � ¼ 1, 2 and i ¼ 0, 1, 4 we obtain f�, ið yÞ ¼ fð0Þ�, i ð yÞ. To find the
remaining distribution functions we note that the equilibrium distribu-tions are functions of powers of y up to y3 through linear dependenceon the x-component of the velocity. Thus, the following form of theremaining unknown distribution functions is suggested:
f�, ið yÞ ¼ nða�, i þ b�, i yþ c�, i y2þ d�, i y
3Þ, ð9Þ
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 643] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
Analytical LBM for Thermal Flow Problems 643
for � ¼ 1, 2 and i ¼ 2, 3, 5, 6. The 32 unknown coefficients areestimated by implementing evolution Eq. (1) in all eight velocitydirections accordingly. For example for � ¼ 1 and i ¼ 2 we have
f1, 2ð yþ �Þ ¼ f1, 2ð yÞ �1
�ð f1, 2ð yÞ � f
ð0Þ1, 2ð yÞÞ, ð10Þ
where � is the vertical spacing between the lattice rows. The expressionsfor f1, 2ð yÞ and f
ð0Þ1, 2ð yÞ given by Eqs. (3) and (9) respectively, are sub-
stituted in Eq. (10) and then the left hand side of Eq. (8) is expandedusing Taylor series. Equating terms of equal powers in y in the resultingequation leads to an algebraic system of linear equations to be solved forthe unknown coefficients. Applying the same procedure to all directionsfor which the distribution function is unknown and solving the systemssymbolically yields
a1, 2 ¼ a1, 5 ¼ A1 � b1, 2�� � c1, 2�2� � d1, 2�
3�,
a1, 3 ¼ a1, 6 ¼ A1 � b1, 3�� � c1, 3�2� � d1, 3�
3�,
a2, 2 ¼ a2, 5 ¼ A2 � 2b2, 2�� � 4c2, 2�2� � 8d2, 2�
3�,
a2, 3 ¼ a2, 6 ¼ A2 � 2b2, 3�� � 4c2, 3�2� � 8d2, 3�
3�,
b1, 2 ¼ �b1, 5 ¼B1
2u0 � 2c1, 2�� � 3d1, 2�
2�,
b1, 3 ¼ �b1, 6 ¼ �B1
2u0 � 2c1, 3�� � 3d1, 3�
2�,
b2, 2 ¼ �b2, 5 ¼ B2u0 � 4c2, 2�� � 12d2, 2�2�,
b2, 3 ¼ �b2, 6 ¼ �B2u0 � 4c2, 3�� � 12d2, 3�2�,
c1, 2 ¼ c1, 5 ¼C1 þ 4D1
4u20 � 3d1, 2��,
c1, 3 ¼ c1, 6 ¼C1 þ 4D1
4u20 � 3d1, 3��,
c2, 2 ¼ c2, 5 ¼ ðC2 þD2Þu20 � 6d2, 2��,
c2, 3 ¼ c2, 6 ¼ ðC2 þD2Þu20 � 6d2, 3��,
d1, 2 ¼ �d1, 5 ¼E1
8u30, d1, 3 ¼ �d1, 6 ¼ �
E1
8u30,
d2, 2 ¼ �d2, 5 ¼ E2u30, d2, 3 ¼ �d2, 6 ¼ �E2u
30:
ð11Þ
Putting these results into Eq. (9) we obtain analytical expressions forall 13 distribution functions. These analytical expressions are substitutedfinally into Eqs. (7c–7f ) to findX
�, i
f�, i ¼ n ð12aÞ
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 644] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
644 Valougeorgis and Naris
X�, i
ex, �, i f�, i ¼ nyþ1
3ny�ð2� � 1Þ�2 ð12bÞ
X�, i
ey, �, i f�, i ¼ 0 ð12cÞ
X�, i
ðe�, i � uÞ2
2f�, i ¼ n"�
1
3ny2�ð2� � 1Þ �
2
3n"�ð2� � 1Þ
� ��2 ð12dÞ
It is noticed that the conservation of mass and y-momentum equationsare satisfied exactly, while there is a second order discrepancy in thex-momentum and energy equations. It is seen that the derived analyticalsolutions although represent an exact solution of the LB model, Eq. (1)do not represent an exact solution of the nonisothermal Couette flowproblem. The analytical solution depends on � and �. It is seen that as �goes to zero or � approaches 1/2 the exact solution is recovered.
Next the 13-bit square lattice is considered. This model has one restparticle
e�, i ¼ 0 ð13aÞ
for � ¼ 0 and three nonzero speeds for which
e�, i ¼ e�ðcos�i, sin�iÞ ð13aÞ
for � ¼ 1, 3 and i ¼ 1, 2, 3, 4 where e1 ¼ 1, e3 ¼ 2, �i ¼ ði � 1Þ�=2 and
e�, i ¼ e�ðcos�i, sin�iÞ ð13aÞ
for � ¼ 2 and i ¼ 1, 2, 3, 4 where e2 ¼ffiffiffi2
p, �i ¼ ði � 1=2Þ�=2. Using the
constraints given by Eqs. (2) and the above set of discrete velocities wefind the coefficients of the equilibrium function (3) to be
A0 ¼1
2ð2� 5"þ 4"2Þ, A1 ¼
2
3ð"� "2Þ, A3 ¼
1
24ð�"þ 4"2Þ,
B1 ¼1
3ð2� 3"Þ, B2 ¼
"
4, B3 ¼
1
24ð�1þ 3"Þ,
C1 ¼1
3ð2� 3"Þ, C2 ¼
1
8, C3 ¼
1
96ð�1þ 6"Þ,
D0 ¼ �5
4þ ", D1 ¼
1
3", D2 ¼
1
8ð�1� 2"Þ, D3 ¼
1
24",
E1 ¼1
3, E2 ¼
1
8, E3 ¼
1
96, F1 ¼ �
1
2, F2 ¼ �
1
8,
G0 ¼1
4, H1 ¼ �
1
6, H3 ¼
1
96:
ð14Þ
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 645] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
Analytical LBM for Thermal Flow Problems 645
The coefficients, which are not included in the above equations, aretaken equal to zero. Higher order terms have been added to the expres-sion of the equilibrium function and as a result in this case the first fourmoments of the equilibrium, given by Eqs. (7a–7d), are recovered.However, still it is not possible to find a solution to satisfy the fourthorder constraints given by Eq. (2e).
Introducing Eqs. (7a) and (7b) we obtain f�, ið yÞ ¼ fð0Þ�, i ð yÞ for � ¼ 1, 3
and i ¼ 0, 1, 3. In this case the remaining unknown distribution functionstake the form
f�, ið yÞ ¼ nða�, i þ b�, i yþ c�, i y2þ d�, i y
3þ e�, i y
4Þ ð15Þ
for � ¼ 1, 2, 3 and i ¼ 2, 3, 4, 6, 7, 8. Following the same procedure asbefore we find
a1, 2 ¼ a1, 4 ¼ A1 þ c1, 2�2�ð�1þ 2�Þ þ e1, 2�
4�ð�1þ 8� � 12�2Þ,
a2, 1 ¼ a2, 3 ¼ A2 � b2, 1�� � c2, 1�2� � d2, 1�
3� � e2, 1�4�,
a2, 2 ¼ a2, 4 ¼ A2 � b2, 2�� � c2, 2�2� � d2, 2�
3� � e2, 2�4�,
a3, 2 ¼ a3, 4 ¼ A3 þ 4c3, 2�2�ð�1þ 2�Þ þ 16e3, 2�
4�ð�1þ 8� � 12�2Þ,
b1, 2 ¼ �b1, 4 ¼ �2c1, 2�� þ 4e1, 2�3�ð�1þ 3�Þ,
b2, 2 ¼ �b2, 4 ¼ �B2u0 � 2c2, 2�� � 3d2, 2�2� � 4e2, 2�
3�,
b2, 1 ¼ �b2, 3 ¼ B2u0 � 2c2, 1�� � 3d2, 1�2� � 4e2, 1�
3�,
b3, 2 ¼ �b3, 4 ¼ �4c3, 2�� þ 32e3, 2�3�ð�1þ 3�Þ,
c1, 2 ¼ c1, 4 ¼ D1u20 þ 6e1, 2�
2�ð�1þ 2�Þ,
c2, 1 ¼ c2, 3 ¼ ðC2 þD2Þu20 � 3d2, 1�� � 6e2, 1�
2�,
c2, 2 ¼ c2, 4 ¼ ðC2 þD2Þu20 � 3d2, 2�� � 6e2, 2�
2�,
c3, 2 ¼ c3, 4 ¼ D3u20 þ 24e3, 2�
2�ð�1þ 2�Þ,
d2, 1 ¼ �d2, 3 ¼ ðE2 þ F2Þu30 � 4e2, 1��,
d2, 2 ¼ �d2, 4 ¼ �ðE2 þ F2Þu30 � 4e2, 2��,
d1, 2 ¼ �d1, 4 ¼ �4e1, 2��,
d3, 2 ¼ �d3, 4 ¼ �8e3, 2��,
e2, 1 ¼ e2, 3 ¼ ðG2 þH2 þ I2Þu40,
e3, 2 ¼ e3, 4 ¼ G3u40,
e2, 2 ¼ e2, 4 ¼ ðG2 þH2 þ I2Þu40,
e1, 2 ¼ e1, 4 ¼ G1u40:
ð16Þ
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 646] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
646 Valougeorgis and Naris
Once the 13 distribution functions are estimated they are substitutedback into Eqs. (7) to find that all properties are fully satisfied. As a resultin the analytical solutions, based on the 13-bit square lattice model, allflow characteristics are recovered and the LB evolution equation issatisfied. Hence the solution is an exact representation of the thermalCouette flow problem and it is valid for any relaxation time � and latticespacing. The only pitfall is that the Navier–Stokes equations are notfully recovered since the fourth order constraints are not satisfied.This drawback is circumvented in the next session by proposing a 17discrete velocity model.
The obtained results of the two 13-bit models are indicative for theaccuracy to expect implementing the 13-bit hexagonal and square lattice.It is seen that the accuracy of the 13-bit square lattice is improvedcompared with the accuracy of the 13-bit hexagonal lattice. No remarkshowever, can be made regarding stability issues of the two models.Actually previous stability analysis performed on the two models(Pavlo et al., 1998b) has shown that the 13-bit hexagonal model ismore stable than the 13-bit square model.
III. ANALYTICAL SOLUTIONS OF
THE 17-BIT MODEL
The 17-bit model introduced here is a straightforward extension ofthe 9-bit model used in isothermal problems. It is consisting of one restparticle,
e�, i ¼ 0, ð17aÞ
for � ¼ 0, and four nonzero speeds for which
e�, i ¼ e�ðcos�i, sin�iÞ, ð17bÞ
for � ¼ 1, 3 and i ¼ 1, 2, 3, 4 where e1 ¼ 1, e3 ¼ 2, �i ¼ ði � 1Þ�=2 and
e�, i ¼ e�ðcos�i, sin�iÞ, ð17cÞ
for � ¼ 2, 4 and i ¼ 1, 2, 3, 4 where e2 ¼ffiffiffi2
p, e4 ¼ 2
ffiffiffi2
p, �i ¼ ði � 1=2Þ�.
The analytical formulation has been described extensively in theprevious section and so we will be brief here presenting only the new
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 647] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
Analytical LBM for Thermal Flow Problems 647
material. The coefficients of the 17-bit equilibrium distribution functions,expressed by Eq. (3) are
A0 ¼1
2ð2� 5"þ 4"2Þ, A1 ¼
2
3ð"� "2Þ, A3 ¼
1
24ð�"þ 4"2Þ
B1 ¼1
3ð2� 3"Þ, B2 ¼
1
4", B3 ¼
1
24ð�1þ 3"Þ
C1 ¼1
3ð2� 3"Þ, C2 ¼
1
12ð2� 3"Þ,
C3 ¼1
96ð�1þ 6"Þ, C4 ¼
1
384ð�1þ 6"Þ
D0 ¼ �1
8ð5þ 7"Þ, D1 ¼
1
3ð�1þ 4"Þ,
D2 ¼ �1
2", D3 ¼
1
48ð1� "Þ, D4 ¼ �
1
32"
E1 ¼ �1
3, E2 ¼ �
1
24, E3 ¼
1
48, E4 ¼
1
384,
F1 ¼1
2, F2 ¼ �
1
8, F3 ¼ �
1
16,
G0 ¼1
4, G2 ¼
1
24, G4 ¼ �
1
96,
H1 ¼ �1
6, H2 ¼ �
1
24, H3 ¼
1
96, H4 ¼
1
384
ð18Þ
Again the coefficients, which are not included in the above expres-sions, are taken equal to zero. The set of equilibrium functions resultingfrom the above constants, unlike the ones obtained by the two 13-bitmodels, satisfy all 13 constraints given in Eq. (2).
The distribution functions for � ¼ 1, 3 and i ¼ 0, 1, 3 are equal to thecorresponding equilibrium distributions. The unknown distributionfunctions for the remaining directions have a polynomial form identicalto Eq. (15). The unknown coefficients for � ¼ 1, 3 with i ¼ 2, 4 and � ¼ 2with i ¼ 1, 2, 3, 4 are given by Eq. (16), while the coefficients for � ¼ 4and i ¼ 1, 2, 3, 4 are given by
a4, 1 ¼ a4, 3 ¼ A4 � 2b4, 1�� � 4c4, 1�2� � 8d4, 1�
3� � 16e4, 1�4�,
a4, 2 ¼ a4, 4 ¼ A4 � 2b4, 2�� � 4c4, 2�2� � 8d4, 2�
3� � 16e4, 2�4�,
b4, 1 ¼ �b4, 3 ¼ 2B4u0 � 4c4, 1�� � 12d4, 1�2� � 32e4, 1�
3�,
b4, 2 ¼ �b4, 4 ¼ 2B4u0 � 4c4, 2�� � 12d4, 2�2� � 32e4, 2�
3�,
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 648] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
648 Valougeorgis and Naris
c4, 1 ¼ c4, 3 ¼ ð4C4 þD4Þu20 � 6d4, 1�� � 24e4, 1�
2�,
c4, 2 ¼ c4, 4 ¼ ð4C4 þD4Þu20 � 6d4, 2�� � 24e4, 2�
2�,
d4, 1 ¼ �d4, 3 ¼ 2ð4E4 þ F4Þu30 � 8e4, 1��,
d4, 2 ¼ �d4, 4 ¼ �2ð4E4 þ F4Þu30 � 8e4, 2��,
e4, 1 ¼ e4, 3 ¼ ðG4 þ 4H4 þ 16I4Þu40,
e4, 2 ¼ e4, 4 ¼ ðG4 þ 4H4 þ 16I4Þu40: ð19Þ
The resulting analytical distribution functions satisfy exactly allproperties described by Eqs. (7) and the LB Eq. (1) and can be consideredas an exact representation of the thermal Couette flow. The analyticalsolution is independent of � and �.
IV. CONCLUDING REMARKS
In conclusion we have developed analytical solutions of the thermalCouette flow with a temperature difference at the boundaries usingthree different discrete velocity models. The analytical solutions for thediscretized distribution functions must satisfy all 13 thermal hydrodynam-ic constraints in order to ensure full recovery of the Navier–Stokesequations, the flow characteristics and the profiles of the macroscopicsolutions. It is seen that the solutions based on a 17-bit square lattice fulfilsall above requirements. The solution based on the 13-bit square latticealthough represents exactly the thermal flow problem does not recoverthe fourth order constraint. Finally the solution based on the 13-bithexagonal model posses a second order error in the thermal flow solutionsand a third order deviation in the third order constraints.
ACKNOWLEDGMENT
This work has been partially supported by the AssociationEURATOM-Hellenic Republic (Contract ERB 5005 CT 990100).
REFERENCES
Alexander, F. J., Chen, S., Sterling, J. D. (1993). Lattice Boltzmannthermohydrodynamics. Phys. Rev. E 47(4):R2249–R2252.
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 649] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
Analytical LBM for Thermal Flow Problems 649
Chen, S., Doolen, G. D. (1998). Lattice Boltzmann method for fluidflows. Annu. Rev. Fluid Mech. 30:329–364.
Chen, Y., Ohachi, H., Akiyama, M. (1994). Thermal lattice Bhatnagar–Gross–Krook model without non-linear deviations in macro-dynamic equations. Phys. Rev. E 50(3):2776–2783.
Cornubert, R., d’Humieres, D., Levermore, D. (1991). A Knudsen layertheory for lattice gases. Physica D 47(6):241–259.
He, X., Zou, Q., Luo, L. S., Dembo, M. (1997). Analytical solutions ofsimple flows and analysis of nonslip boundary conditions for thelattice Boltzmann BGK model. J. Stat. Phys. 87:115–136.
Henon, M. (1987). Viscosity of a lattice gas. Complex Systems 1:763–789.Luo, L. S. (1997). Analytic solutions of linearized lattice Boltzmann
equation. J. Stat. Phys. 88:913–926.Luo, L. S., Chen, H., Chen, S., Doolen, G. D., Lee, Y. C. (1991).
Generalized hydrodynamic transport in lattice-gas automata.Phys. Rev. A 43:7097–7100.
McNamara, G., Alder, B. (1993). Analysis of the lattice Boltzmanntreatment of hydrodynamics. Physica A 194:218–228.
McNamara, G., Garcia, A. L., Alder, B. J. (1995). Stabilization ofthermal lattice Boltzmann models. J. Stat. Phys. 81:395–408.
Pavlo, P., Vahala, G., Vahala, L. (1998a). Higher order isotropic velocitygrids in lattice methods. Phys. Rev. Lett. 80(18):3960–3963.
Pavlo, P., Vahala, G., Vahala, L., Soe, M. (1998b). Linear stabilityanalysis of thermo-lattice Boltzmann models. J. Comput. Phys.139:79–91.
Qian, Y. H. (1993). Simulating thermohydrodynamics with lattice BGKmodels. J. Sci. Comput. 8(3):231–242.
Sterling, J., Chen, S. (1996). Stability analysis of lattice Boltzmannmethod. J. Comput. Phys. 123:196–206.
Zou, Q., Hou, S., Doolen, G. D. (1995). Analytical solutions of the latticeBoltzmann BGK model. J. Stat. Phys. 81:319–339.
Received September 15, 2001Revised June 25, 2002Accepted June 25, 2002
120025070_TT32_05-07_R1_080103
+ [1.8.2003–10:43am] [639–650] [Page No. 650] f:/Mdi/Tt/32(5-7)/120025070_TT_032_005-007_R1.3d Transport Theory and Statistical Physics (TT)
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
650 Valougeorgis and Naris