Physica 143A (1987) 261-286 North-Holland. Amsterdam
ON SLIP COEFFICIENTS B.I.M.
TEN
Huygens Laboratorium
BOSCH
OF POLYATOMIC and J.J.M.
GASES
BEENAKKER
der Rijksuniversiteit, Leiden,
The Netherlands
I. KUSCER Department of Physics, University of Ljubljana,
Received
1 December
Yugoslavia
1986
A variational scheme is presented to evaluate field-dependent slip coefficients of polyatomic gases in a plane-parallel geometry. With the assumption of perfect accommodation at the surface and with bulk test functions only, fair agreement is obtained between theory and experiment. Also, it is shown that the present method is equivalent to a previously used variational technique, if bulk test functions are used in both cases. Finally, a comparison is made with Maxwell’s and Eddington’s methods. The equivalence of the latter method with the variational scheme is demonstrated.
1. Introduction Angular-momentum alignment or ‘polarization’ may occur in polyatomic gases out of equilibrium, because of non-spherical interaction of the molecules with each other or with the surfaces of the enclosing vessel. One of the ways to study this is to measure
changes
of transport
coefficients
caused
by a magnetic
field. These so-called Senftleben-Beenakker effects were first observed and analyzed in the hydrodynamic regime, where they are by now well understood’). More recently, experimental attention has also been paid to the near-hydrodynamic regime. Here the mean free path 1 is no longer negligibly small in comparison to the width of the channel b but the boundary layers may still be considered separate. Typically the Knudsen number Kn = l/b is about 0.1. Reliable data are by now available on field-dependent velocity slip*), thermal creep3), the viscomagnetic heat flux4) and the viscomagnetic diffusive flux5). In this paper we wish to perform a systematic study of these field-dependent rarefied-gas phenomena on the basis of kinetic theory. Comparison with 0378-4371 I87 /$03.50 0 (North-Holland Physics
Elsevier Science Publishers Publishing Division)
B .V.
262
B.I.M. TEN BOSCH et al
experimental
results
may then
answer
the question
to what extent
production
of polarization at the surface plays a role in these effects. We will restrict ourselves to pure gases consisting of diatomic diamagnetic molecules. Angular momentum surface
will be treated
classically,
and vibrational
of the walls shall be macroscopically
left-right
uniform,
excitation
ignored.
The
rotationally
invariant
and
symmetric.
The method to be employed is a variational scheme based on the linearized Boltzmann equation and boundary conditions. No fitting to experimental data is invoked, as is the case in previous attempts to describe field-dependent boundary-layer phenomena6 “). I n contrast to the method of ref. 10 (henceforth referred to as II), the variational functional to be used will not be related to flow coefficients but to slip coefficients, which occur naturally in a description of boundary layers. In section 2 the variational principle will be outlined, and Onsager’s reciprocity for slip coefficients shown to hold also in the variational approximation. Section 3 deals with applications of the technique to several field-dependent flow problems and in section 4 a link is established between the two variational methods. Finally, section 5 contains a discussion of Maxwell’s variational
and Eddington’s scheme.
2. A variational
methods,
and
a comparison
is made
with
the
principle
Bearing in mind that the boundary layers do not overlap, we shall idealize the problem by considering a gas filling the half-space above a wall at z = 0. The gas is subjected to a uniform magnetic field B, and constant gradients of temperature and flow velocity, VT and v’ = dvidz with v = (u,, u?), are present in the bulk. follows:
where
f” is a Maxwellian
T,(x> Y) = W) f”
For not too large z the distribution
=
+ q
at rest corresponding
function
to the surface
is linearized
as
temperature
- 7, T,
p(m/2~)“‘*(kT,)~“‘*(4~ZkT,)-’
exp(-EIkT,)
,
(2)
and 4” and 4” are correction terms. Most of the concepts and notation are taken from ref. 11, which is henceforth referred to as I. The symbol (( indicates ‘x-, y-components only’. The unknown functions +“, p = q, n-, satisfy the following equation and boundary condition:
ON SLIP COEFFICIENTS
OF POLYATOMIC
GASES
263
(W+i9+cZ&)+“=-PP,
(3a)
~“(O,c,J)=PO~~(O,c,J), c,>o,
(3b)
where
the microscopic
introduced,
fluxes
Q 4 = (E - zkT)c
as well as the Boltzmann
and time reversal operators 9, Chapman-Enskog-like solution,
collision,
and surface
W” = mc,c,, scattering,
have
been
precession
P,i9 and 0. For z + m, 4’ must approach
a
The 2 x 2 and 2 X 3 matrices of slip coefficients y”” and yV4 defined thereby account for velocity slip driven by either a velocity or a temperature gradient, jump in the boundary respectively. Similarly, yfP accounts for a temperature layer; yfP is a 1 X 2 or a 1 X 3 matrix according to whether p = rr or /3 = 4. All these coefficients are uniquely determined by the stated Boltzmann equation and boundary condition. Let us introduce the excess boundary-layer heat flux A4 according to
(5)
bulk heat flux, both where q,l and qf; are the true heat flux and the extrapolated parallel to the wall. The linear dependence of A” upon the gradients may be specified in terms of yet another set of slip coefficients yip, VT AY=y;;q-T+~;*~~‘r
(6)
with yif” a 2 x 3, and yi” a 2 x 2 matrix. The definition of the slip coefficients introduced above, which seems natural in the present context, differs from the one in I by viscosity and thermal conductivity
Y
TY
-
-
factors.
sty 3
The translation
r;;” = 5;;”
3
is
(7~ d)
(7e, f)
264
B.I.M.
where
the thermal
Here
and henceforth
[(h,(h,)]=
conductivity
BOSCH
and planar
the following
f jdz
TEN
inner
et al
viscosity
products
tensors
are given
by
are used:
jdI-h:/“h,.
the Hilbert
(i,ie)’ = -iZ
in X, and Z, ,
(10)
(c;P)’
in SC+ .
(11)
= cz P
spaces
X,,
CT+ and X%. The adjoints
defining, respectively, i_Y and c, P are
of %,
The heat fluxes in eq. (S), in case of a flow of type /3, contain the factors respectively. Definition (6) thus leads (!P~Ic$~) and -(!Pil(%! +L9?)‘Pa), to the following kinetic expressions for the riP,
We have to assume that integrations over z like in eqs. (5) and (12) do not lead to divergences. To obtain an approximation for some of the slip coefficients, we employ a variational method which has been used by several authors’2m’h) for monatomic gases. For the general background of the method the book by Finlayson”) may be consulted. The extension to polyatomic gases and inclusion of a magnetic field requires some modification. For each slip problem we
ON SLIP COEFFICIENTS
introduce
VP=
a tensor
of bilinear
[(O(f&, +(92 +
variational
OF POLYATOMIC
Yap, CX,p = q, rr,
functionals
-iif))‘t”)l(!S!
+iZ+
265
GASES
c, &)+*+
@)]
mb,Ic*(l- fw4)+I*=,
+ (0(9? The trial functions assumed to satisfy
- i.2)‘ly*(c,(&
+ (2
+ i,Fe)-‘*p))I,=O.
(13)
c#+ and 42 depend upon c, J and z, and in view of (4) are the following boundary conditions at infinity:
( 14b) where the parameters /.L, V, p and u stand in place of, but are not to be identified with, the slip coefficients. With some manipulation of the operators and a partial integration over z, one finds from eq. (13) the first variation of
It follows
that 6Y”’ = 0 if and only if C/J*satisfies
equations with opposite field and with P* instead the limiting behaviour of 4, for z--+ 00 must then +1
4
MC,,
The stationary
-
y”“(-B) value
+ (E - ikT)yq”(-B)
(3),
and +i the same
- (9 - iL!?)‘V
.
with (4),
(16)
of Yap is
Yzf = (O(%! - i.Z)‘Q”lc,(+P To establish 0(9X - i2)-‘Wa
eqs.
of W’. In analogy be
+ (6% + i~))‘!P’))I,=,
the meaning of Yzf, we take the with both sides of eq. (3a). After
.
(17)
[( ( )] inner product of some reshuffling of the
B.I.M.
266
operators
and a partial
integration
TEN BOSCH
et al
over z we obtain
the identity
The asymptotic form (4) of 4’ serves to relate the second term of this expression to slip coefficients, whereas the third term is linked to Yzf, eq. (17). So we get
+ [(
OW”l+” + (% +
iZ))‘W’)]
(18)
The second term on the right-hand side vanishes if Wf or 1v” are substituted for ?P”. Namely, if we take the ( ( ) mner product of E - i kT or mc,, with eq. (3a), three terms drop out and we see that (Py]+“) and (lu”l+“) are constants. Substitution of the asymptotic form (4) then shows that the constants are equal to -( V’f](% + i_Y)‘!Pa) and -(w~](% + iZ))‘~p), respectively. On the other hand, a comparison with eq. (12) shows that in case !P’ = !Pi the double-bracket inner product in eq. (18) is equal to -kTyiP/n. Hence, the following relations are obtained (with i, j, k = x, y and also k = z for p = q):
Y"" rk. rt
y@
rk.\t
=
(19c)
Approximate values for the slip coefficients may now be obtained in the standard way. First one chooses trial functions 4, and C#J*which satisfy the conditions (14) and depend upon a finite set of trial parameters. After the Y”’ are made stationary with respect to variations of the trial parameters, eqs. (19) to approximations +“’ for serve to relate the restricted stationary values, e:, the slip coefficients. We are now going to see that variational approximations for the yup do not violate the Onsager symmetries established in I. First, we write the functionals (13) in an alternative way,
261
ON SLIP COEFFICIENTS OF POLYATOMIC GASES
fwb1l@45)+I.=,
(c,(l-
+
+ (c,(&
+ (% - iZ)-‘ly”)lO(%
+ (O(Z
- iL.?)‘W”Ic,(%
+ iL!?)llyp)Iz_O ,
+ iZ)-‘!P’)
(20)
where the symmetry properties (9)-( 11) were used. The first three terms on the right constitute YP”‘(-B), the dagger denoting the transposition. When we request that
SP(B) = 0
SP-B)
)
=0 )
where the variations are performed within a family of trial functions 4, and &, the same solutions are obtained in both cases. Yet the stationary values of the two restricted functionals differ by a constant term, as seen from eq. (20), P$(B)
= ?fT+(-B)
+ (O(,$! - iZ)-‘ly~Ic,(%
With eqs. (19) the Onsager follow (i, j, k = x, y):
relations
y”gyB) + z-A,,(B)y,4,4(B) = y”;f(-B)
for the
approximate
+ TA,,(-B)y”,4y(-B)
r”;;(B) + TA;,(B)y”,4,9(B) = TA,,(-B)7;;(-B)
The tensor
of Burnett
A’ = - &
(O(%
To reach complete
heat flux coefficients - i~))‘lyyIc,(~
accord
has been
relations
(21)
slip coefficients
)
(224 (224
)
+ i_!Z)‘!PV)
with the Onsager
.
+ iZ)-‘!Pp)
introduced, . (1.62),
(23) (1.76) and (1.79),
268
B.I.M.
we must ignore discrepancy
products
with
approximation justification
TEN
BOSCH
of field-induced
I is not
was made
components
conceptual,
there
for neglecting
et al
however,
of A and the y”‘. since
as well (cf. the remarks
such products
and we will stick to this approximation
implicitly
the
This same
after eq. (1.69)).
is the smallness
The
of field effects”),
in the sequel.
3. Specific examples Three applications of the above variational principle will now be given, and a comparison with experimental results made. The effects to be analyzed are the transverse viscomagnetic slip, the transverse thermal creep, and the transverse viscomagnetic heat flux. Only the simplest possible trial functions will be employed, i.e. those which for any z have the bulk form (14),
c#+ = mc,, *p + (E
- zkT)v - (% - i5!?)‘!P”
+2 = mc,, . p + (E -
,
(244
$kT)a - (2 + i_Y)- ‘!PP .
(24b)
The approximations
for the slip coefficients resulting from ?zf contain inner quantities with the first products involving (9? + i2!?~ ’ and P. Whereas operator can in turn be re-expressed in terms of known matrix elements of 92, a guess must be made for the surface scattering operator to make a comparison with experiment possible. As in II, we assume that the surface is perfectly accommodating, Ph = (2nm/kT)“2(c,Ih)+ which results
3.1.
in considerable
,
(25)
simplification.
The transverse viscomagnetic slip
First we will summarize some experimental results. With B = (0,0, B), viscomagnetic slip has been measured for N, indirectly through an investigation of the pressure dependence of Ly,z IL”“. The tensor L”” relates the gapaveraged flow velocity V to the applied pressure gradient yip, (26) whereas
L””
denotes
the
field-free
coefficient.
In
the
near-hydrodynamic
ON SLIP COEFFICIENTS
regime
Breunese
et al.*) found
L;j: qP “” L --F(Kn)‘%&
OF POLYATOMIC
a scaling
behaviour
GASES
according
269
to
5 (
(27)
) ’
(28) with C, = 4(kTln-m)“*, and 71the field-free viscosity. The functions F and G are responsible for the pressure dependence of the magnitude and the position of the effect. The Lorentzian functions
(2% b) are used here and henceforth. Expressions for the non-sphericity parameter angle so*,, in terms of effective cross sections 9 “2Wand the effective precession G(;,z.z,) ‘) are given in the appendix. The scaling functions F and G were approximated by F=l+n,,Kn+B(Kn*),
G=l+n,Kn+Q(Kn*),
(3Oa, b)
with Knudsen correction parameters nP = 10.7 * 0.5, II, = 2.7 2 0.5. To extract from eq. (27) an expression for the velocity slip coefficient we note that up to order Kn this equation may alternatively be written L;.Z U” - - @“‘“g( L On the other
L”” = where
$
hand, b’(S
6 denotes U” L yx=_ “” L
A comparison
502m)(l
+
according
(n,
-
np)
+ f y”” + Q(Kn’))
+ E 7;:
-
to eq. (1.51)
the two-dimensional
%
Kn
- 7-l
2n,
Wl
-f(502?r))).
(31)
and eq. (7a) we have ,
unit tensor.
(32) Therefore.
+ Q(Kn’).
with eq. (31) confirms
y;T,“, as
(33) the bulk formula”)
(34)
270
B.I.M.
and also yields
an expression
1 6 (c,“%
y;;:”
TEN
BOSCH
et al.
for the slip coefficient,
5o*,)(n,
- n, + 250
(35)
- f( 502?r)))
Let us now turn to the variational scheme. With the bulk trial functions (24) the first term on the right-hand side of eq. (13) vanishes and the last one simplifies,
so that
IcAl - W(mc,,
‘P + (E - ;kT)a
- (% + iZ’)‘+~))+l,_,I
kT + y PI/ *P Putting the derivatives equations
(36)
with respect
to p, V, p and u equal to zero, we get the
(c,,Ic,(l- P@)(mc,,*p -
(3 + iT))‘VT))
(E - ;kTlc,(l
- $kT)o
- PO)((E
kT y ‘lit + (WC,, (O((E
- ikT)v
- (3 - i5!T)‘ly”)lcZ(l
account,
- iZ)-‘ly”lc,(l
X (mc,, .p + (E - $kT)a
- PO)mc,,), - PO)(E
When eqs. (19a) and (36) are taken into being omitted in the rest of this section) * (O(%
(374
- (%! + i_Y)‘TT))+
* P - ( 92 - iZ))‘lv”)[cZ(l
Y
r = 0 ,
- ;kT)) it follows
= 0,
WI
= 0,
(37c)
,_ = 0.
(37d)
that
tildes
(the
- PO)
- (% + i.Z)‘V”))
+ ,
(38)
with P =
(c,,IM - fwmc,,L’ *(c,,Ic,(l
- PO)(%!
+ iZ))‘Tm)
+ ,
(39)
and (E - ikTlc,(lu=
(E-
PO)(%
;kTlc,(l-PO)(E-
+ Z!?)‘W”) ;kT))+
+ (40)
ON SLIP COEFFICIENTS
With the assumption of field-effect
of perfect
coefficients,
OF POLYATOMIC
accommodation
271
GASES
(25) and neglecting
products
we get the result
(41) where
another
2 x 2 tensor
and its fieldless
value
An6 were introduced.
(42) In the appendix approximations are given for q, nYX, An and A;X in terms of effective cross sections. With eqs. (A.4), (A.6) and (A.7) eq. (41) becomes
(43) We see that scaling eq. (35), if 15 np = s
K
according
;+2
15 G(20) n, = 8 G(02T)
1
+p
to eq. (27) is recovered,
since eq. (43) reduces
S(20) E(02n)
to
(444
1’
(44b)
’
just as in eqs. (11.41). With the effective cross sections for N, listed in table I we obtain np = 9.4 and nY = 2.7, which is in satisfactory agreement with the previously quoted experimental values.
TABLEI Effective cross sections and gyromagnetic ratios for CO and N, used in this section. Sources: refs. 19 and 20.
G(20) 6(027r) s(p) 2( 1010) Z(1001) XX 5( 12q) ‘G(:;:, 5( :t’) V
10~” m’ 10~” m’ lo-z0 m.’ lo-*” m.’ loo*’ rn.’ lo-*” m2 lo-*” rn’ lo-‘” m’ 10~LOmL 10” (Ts‘)~’
co
N*
34.7 32.5 2.02 29 30 5.5 49 -0.91 4.1 -1.29
34.6 23.7 1.50 27 30 4.0 48 -0.67 4.4 -1.33
212
3.2.
B.I.M.
The transverse
Experimentally, Hermans et al.‘), of Eggermont and
thermal
TEN
BOSCH
et al
creep
this phenomenon has recently been reinvestigated in CO by who corrected systematic errors present in the original work
et a1.2’). The temperature
gradient
B = (0,B sin 0, B cos 0). The results
was taken
showed
in the z-direction,
a simple
orientational
dependence,
PVx -
PTY “;”
-L=-sin0(H+ZCOS20),
4,
(45)
A
where v, is the bulk creep velocity, q, the perpendicular heat flux and A the field-free thermal conductivity. The functions H and I depend upon B/p and are plotted in fig. 1. As we shall see, the variational representation simplifies if these functions are decomposed according to Vestner’), H=S”Y+A;Y,
I=2A;“y-A2”y,
A;‘(B/p)
= A;‘(2B/p).
(46)
in the kinetic theory for bulk Decompositions as in eqs. (46) occur naturally Fig. 2 shows 6”“ and A;i’ as fitted to the experimental transport coefficients. data. Since the derivation of the expression for the creep coefficient riz is
Fig. 1. Thermal creep in CO at room temperature. The coefficients H and I are explained by eq. (45). The symbols indicate experimental data from ref. 3, whereas the drawn curves show the variational results.
ON SLIP COEFFICIENTS
OF POLYATOMIC
273
GASES
05
pv, -9;
I
0
01
-
Blp
10
1
Fig. 2. Thermal creep in CO. The dashed (46) to the experimental data. The drawn
riz” = p - &
vi’
is quoted**),
* (O(CB - i2)-‘!Py”Icz(1
X(mc,,*p+(E-
100
curves show 6”“ and A;'as obtained from a fit of eqs. curves correspond to eq. (51) and (52).
as for ynm, only the result
analogous
mTlPa
- PO)
+ i2)-‘We))+,
ikT)a-(2
(47)
with P =
(c,,Ic,(l - P@)mc,,);’ - ( cIIIcz(l
PO)(%! + i2Z)‘ly~)
-
,
(48)
and
(E - ikTlcZ(l cr=
(E-
- PO)(%
+ i2))‘P’f)
;kTlc,(l-PO)(E-
;kT))+
+ ’
(49)
perfect Let us consider r,“,” and B = (0, B sin 8, B cos O), again assuming accommodation. By aid of the approximate relations (A.8) and (A.9) eq. (47) reduces to y;;(B)
=
&
A:,(B)
+& (n:,(-B)
- TA’;;-B)) ,
(50)
by eqs. (A. 10) and (A.ll). where the coefficients v,, and A:, are explained To make a comparison with experiment possible we employ approximations
B.I.M.
274
(A.14)-(A.23) cross sections. indeed
TEN BOSCH et al.
for hi,, 14:x and n,, in terms It is found that the variational
be cast into the form of eq. (45),
of quantities involving effective result for the ratio pu,Iqz may
with
(51)
substituted into eqs. (46). Cross sections for CO, needed for the evaluation of angle t,,,. the non-sphericity parameters $’ and $“‘, the effective precession and the quantities At/h and S, are listed in table I. The variational results for H, as curves in figs. 1 and 2. As in the case of I, aVq and A;” are presented viscomagnetic slip, the agreement with experimental data is satisfactory. 3.3.
The transverse viscomngnetic heat flux
This effect has been investigated in CO and N, by Eggermont et al.J). For the gap-averaged heat flux qV due to an applied pressure gradient Vyp = vu: and a field B = (0, B sin 0, B cos B), they found an orientational dependence of the form
Pq, M = cos 0 (J - K sin’ 0) hq TV:’
(53)
In figs. 3 and 4, J and K are plotted versus B/p. In addition, 6’” and AT” resulting from the decomposition J = 6””
+ AT-,
K = 24;”
To obtain
a variational
in subsection
3.1 to get
y 4” II = h
(O(S
-
expression
- i$P)-‘Iy~Jc,(l
ApT,
A;“(B/P)
figs. 5-8 show the
= Ay”(2Bip).
for YT:, we follow the procedure
- PO)(mc,, *p - (2 + iZ))‘lyp))
(54) outlined
+, (55)
with p from eq. (39). With surface, y y: is approximated
the specialization to a perfectly accommodating in terms of bulk transport coefficients,
(56)
ON SLIP COEFFICIENTS
OF POLYATOMIC
275
GASES
20 Id3
co
15
1.0
05,
Fig. 3. The transverse viscomagnetic heat flux in CO at room temperature. The coefficients J and K are explained by eq. (53). The symbols indicate experimental data from ref. 4, whereas the drawn curves show the variational results.
I 1 Fig. 4. The transverse
viscomagnetic
,1111111 10
heat flux in N, at room
mTI Pa temperature.
100 Notation
where use was made of relation (A.24). We then arrive at a formula averaged heat flux in Poiseuille flow in a gap (0 < z < b),
as in fig. 3.
for the
(57)
B.I.M.
276
TEN
BOSCH
et al.
15
Id3
10
05 6
‘?n
1 9
Fig. S. 6”” versus B/p for CO. The dashed curve is obtained from a fit of eqs. (S4) to the experimental data. The drawn curves show the variational result and the contribution from the Burnett heat flux, eq. (60).
03
I
I I I11111
Blp
Fig. 6. 3:”
versus
I 1
B/p
I
IIIIlll 10
for CO. Notation
mTlPo
100
as in fig. 5.
In the first equality the equations (1.50) and (6), and the relation u” = -2u’(O) lb were invoked. As in subsection 3.2 we relate the result to combinations of effective cross sections. With eqs. (A.25)-(A.27) from the appendix one finds that the variational expression (57) may be rewritten in the form of eq. (53), with J and K expressed via eqs. (54), where
ON SLIP COEFFICIENTS
OF POLYATOMIC
GASES
Fig. 7. 6 ” versus
E/p
for N,. Notation
as in fig. 5.
Fig. 8. A,‘a versus
B/p for N?. Notation
as in fig. 5.
211
(59) With the effective cross sections (59) yield the variational curves
for CO and N, listed in table I, eqs. (58) and drawn on figs. 3-8. For comparison, figs. 5-8
B.I.M. TEN BOSCH et al
278
also show the curves
and
it is seen
neglected. satisfactory,
that
Whereas
which would
the
result
contribution
on figs. 3, 5 and
in case of N, the deviation
from the Burnett
of 7;:
to the
heat flux alone,
total
6 the variational from experimental
effect
results
cannot
be
for CO look
data is more
than
15% for J (fig. 4). Figs. 7 and 8 indicate that the discrepancy must be attributed to ay”. Since it is known, however, that a quantitatively good description of thermal effects in N, invokes still more expansion functions than the ones taken into account in the evaluation of Sq” I’), it should come as no surprise that the disagreement is larger than for CO, where the employed moment approximations fully suffice. It is concluded that in all cases studied, the variational results compare reasonably well with experimental data, with the exception of 6’” in case of N,. In general, the results have so far shown no discrepancy with the simple assumption of perfect accommodation.
4. Comparison
of two variational
methods
One of the results of the previous section was that the variational method slip coefficients yields the same Knudsen corrections for viscosity as method of II. For the approximate results this correspondence is not evident, and we would like to establish this correspondence analytically. way of example we are going to refer to velocity slip. With some rephrasing other effects could be treated in the same way. We recall the stationary values of the functionals X and Y, eqs. (11.14)
for the selfBy the and
(lga),
(61) where the superscripts TV have been dropped. stationary values are related as
According
to eq. (32) the exact
where terms of higher order in Kn have been neglected. We are going to show that eq. (62) also holds in the variational approximation if only bulk trial functions are used.
ON SLIP COEFFICIENTS
The
functional
for the
problem
OF POLYATOMIC
of plane
Poiseuille
219
GASES
flow is given
by eq.
(II.12),
the outer (1.26),
brackets
(1.30)
denoting
a gap average.
As trial functions
we take (cf. eqs.
and (24))
(64a)
h=
b2n 12 mcll’
-6
+
6
SQ, - jj P
bn
(64b)
with Q,(Z) = P,(l - 2zib). The first expressions on the right-hand sides account for the Poiseuille profile with slip parameters p and p. Upon substitution of eqs. (64) into X two terms cancel, and we are left with
+ in2bqi’ Ic,(l
* (O(mc,, ap - (2 - iY))‘W’I)I
- PO)(mc,,
*p - (3 + i.JZ)‘*~))+
* vi’
.
(65)
In view
of eq. (36), the last two terms on the right-hand side constitute These terms only in2bqi’ ’ Ye qi’, if the terms with v and v are neglected. contribute to second-order field effects, and have been neglected in eqs. (64) anyway. As anticipated, we thus find
The ensuing restricted stationary values equation, and hereby the correspondence mations as well.
XS, and YS, clearly obey the same (62) is established for the approxi-
B.I.M.
280
5. Comparison
with Maxwell’s
TEN BOSCH
and Eddington’s
Apart from the variational method, considerable success in approximating viz.
Maxwell’s
and
et al
Eddington’s
methods
two other schemes have also met with slip coefficients for monatomic gases,
methods24m26).
In this
section
we offer
a
suitable generalization for polyatomic gases and compare the outcome with the variational results obtained earlier. Maxwell’s procedure is based on the assumption that the impinging distribution function may be approximated by its bulk form (4) down to z = 0. Subsequently, the scattering law (3b) is invoked to obtain the outgoing part at the surface. Explicitly, at z = 0 one takes +
mc,,.y”“+(E-~kT)y~“-(~+i~)~‘~“,
c,
(67a)
To obtain approximations for the ynp and yfP, use is made of the exact relations expressing the constancy of shear stress and perpendicular heat flux throughout the boundary layer, (68a) (68b) Substituting eqs. (67) into eqs. (68), one obtains equations which may be easily solved for y@ and yYP. In case of Couette flow (p = V) one obtains
(6% b) with p and u from eqs. (39) and (40), deal with thermal
creep
and take
respectively.
On the other
hand,
if we
$2 for $,“, then
(7% b) where p and u are given by eqs. (48) and (49). In general, it may be concluded that Maxwell’s method yields slip coefficients which equal the corresponding tensors of trial parameters as obtained from the variational scheme. With the assumption of perfect accommodation, eq. (69a) implies
(71)
ON SLIP COEFFICIENTS
OF POLYATOMIC
281
GASES
and 15lr
z5.9,
*p = 8 which
should
scheme
be compared
does
not
even
(7% b)
n, =o,
with eqs. (41) and (44).
give qualitative
agreement
‘shift of the maximum’ is obtained (G = 1 + B(Kn2)). creep is equally bad. From eq. (70a) one obtains
It is seen that Maxwell’s with
experiment
The situation
since
no
for thermal
(73)
s=q = A”4 n
=
; ;
+%a2,)
+
%Tcx,,,)) ,
(744
0,
(74b)
which may be compared with eqs. (50)-(52), and also with fig. 2. It is seen again that Maxwell’s method is by no means satisfactory in dealing with effects related to angular momentum polarizations. Finally, it is noticed that the transverse viscomagnetic heat flux cannot be treated with Maxwell’s method. This is due to the fact that the corresponding slip coefficient y q” does not appear in the distribution function (67). Let us now turn to the Eddington method, which is slightly more elaborate than Maxwell’s In addition to eqs. (68) use is also made of relations between the constant quantities (O(!B - i_!Z)‘t,k(I”[~,(+~ + (% + iZ)-‘+‘)) and the various slip coefficients. These relations are obtained by equating the righthand sides of eqS. (17) and (19). In analogy to Maxwell’s procedure the inner products
are evaluated
+‘=mc
,,-p+(E-
at z = 0, with the distribution ;kT)
a-(
%$+iZ))‘*‘,
function c,
(75a)
WI the p and u being reminiscent of the trial parameters in eq. (14b). Finally, the resulting equations may be solved for p, u and the slip coefficients. As an example, let us treat the transverse viscomagnetic heat flux by Eddington’s method. The relations (67) with j3 = r are augmented with rfi”=
j$
(O(%
- iZ)-‘*cl;flc,(+”
and eqs. (75) with p = 7~ are employed
+ (% + iZ)-1+L7i))lZ=0
,
to solve for p, u and 71”.
(76) It is found
B.I.M.
282
that ~4” is given
by eq. (55),
the a-term
may be neglected).
variational
method
ence has already
TEN BOSCH
with p from eq. (39) (just as in subsection 3.3) Hence, Eddington’s method is equivalent to the
in this particular
been noted
et al
case.
by Loyalka”).
For monatomic Further
gases such equival-
inspection
schemes give identical results for all the slip coefficients, equally successful in a description of slip coefficients
shows that both
and consequently are for polyatomic gases.
Ultimately, this correspondence is due to the fact that in both methods slip coefficients are evaluated through equivalent approximations for the quantities (O(9 - i9))‘+“lc,(+” + (9 + Sir?) ‘JIP))Iz _,,.
Appendix The appendix lists some known relations between bulk transport coefficients and effective cross sections, supplemented by a few new derivations. The non-sphericity parameter $0’~ used in eq. (27) is given byiX) -_(
‘::,“)’
(A.11
@‘” = G(20)G(027r) The labels
correspond
to the basis functions
needed
to describe
viscous
flow’),
G4.2) with W= (mi2kT)“‘c sion angle
and P(f an unknown
scalar factor.
The effective
preces-
.$r,z,i is
(A.3)
with y the gyromagnetic ratio. In the derivation of eq. (43) invoked’“):
from
eq.
(41),
the
following
formulae
are
(A.4)
in standard To evaluate A” and AFX, we proceed (.% 2 iT))‘W” in terms of the relevant basis functions
fashion by (A.2) I*),
expanding
ON SLIP COEFFICIENTS
OF POLYATOMIC
+d502*P;?)
283
GASES
(A.5)
)
and similarly for (6%.+ iZ)-‘@:q. Inserting these evaluating the inner products, we finally get
relations
into
eq. (42) and
(A.61
g!y;)d&12,)(1-f(a,J).
fqx = To derive
(0(.92-
eq. (50) from eqs.
iZ))‘V*“lc,(E
(c,Ic,(% + i~))‘W~)+ where tensor
the viscosity A’ are given
77zx
A’= Relation
=
j+
j+
($mc*
(A.8)
+
n,,
i.Z)‘?P~)
= g
the approximate
n,,(-B)
,
relations
(A.8)
A:, ,
(A.9)
and the translational
thermal
conductivity
(A.lO)
, + iL!?-‘ly4)
via the moment T,,(--B)(cz
we need
= g
= g
- gkT)cl(%!
is derived
(CJS!- iZ))‘W~
- iAT))+
coefficient by
(mc:I(%!
(47)-(49)
(A.7)
.
(A.ll)
approximation”)
- cz) + . . . .
(A.12)
The deleted terms (indicated by dots) contain components of @*O and @Ozr which do not contribute to (A.8). Upon insertion of the expansion (A.12) into the left-hand side of eq. (A.8), the stated relation ensues. Relation (A.9) is obtained in analogous fashion. We expand (3 + iZ))‘?Pz as follows23): (22 + iZ)-‘qf
= A:, -j$$j ( $mc2 - $kT)c,
+. . . .
(A.13)
B.I.M. TEN BOSCH et al.
284
The omitted
terms include
the microscopic rotational which both drop @I”,
heat flux W(J’12ZkT - 1) out from the stated inner
and the Kagan polarization product. With only the quoted term taken into account, eq. (A.9) follows. To reduce eq. (50) to eqs. (51) and (52), we need the approximations given below.
First we have’.“)
Aiz = sin ~h’$‘(g(S,7y) with the field-free
translational
thermal
G(lOO1) + ($)“%( *I=*
conductivity $6)
G(1001)+2(~)“2S(~~‘:~I)+
an effective
precession
(A. 14)
T
+ 2g(25,,,))
(A.15)
@(lOlO)
’
angle (A.16)
and a non-sphericity
-S(
parameter
~~~)G(lOlO)
x 5( 12q)(G( Next,
the Burnett AzX(-B)
+ G( :;;‘)G(
lOlO)G(
coefficient
:;;A)
(A. 17)
1001) - G( :;;;)‘)
is given
by7.23)
= sin O(8 + A2 + cos’ 8(2A,
(A.18)
- A,)) ,
with (A.19)
(A.20) q
12q)
(A.21)
s=-qiGj’ *‘”
=
$ j-q;:‘b”, + q;;“,‘)
‘q;“,;;) + ( ~)“26(1010) G(1001)
+ ( ~)“%(
G(ii”) ;;;;) > G(O2%-)G( 12q) (A.22)
ON SLIP COEFFICIENTS
Finally,
the expression
OF POLYATOMIC
for the viscosity
coefficient
285
GASES
(A.lO)
is”)
= sin1377+02n(g(250211) + cos2 0 (2d5o2,) - @to,,))) .
G-B)
(A.23)
The relation
(A.24) needed in the derivation of eq. (56), may be obtained in the same way as eq. (A.9). To obtain eqs. (58) and (59) from eq. (57), we use the relations7’23) -/‘TX= cos 8 (6 + A, + sin2 8 (A, - 24,))
G,(W)
=
cm
0
oJtM5*2q)
+
WY,,,))
(A.25)
,
(A.26)
,
and”) rl yx
=
ax
f3 r1$02=(
d
50271)
+
sin2
8
( d2502n
> -
2g(
to2?r)))
.
(A.27)
References 1) E. Mazur, J.J.M. Beenakker and I. KuScer, Physica 121A (1983) 430. 2) J.N. Breunese, F.W. Godecke, L.J.F. Hermans and J.J.M. Beenakker, Physica 126A (1984) 82. 3) L.J.F. Hermans, R. Horne and E.L.M.J. van Wonderen, Physica 134A (1985) 209. 4) G.E.J. Eggermont, P.W. Hermans, L.J.F. Hermans, H.F.P. Knaap and J.J.M. Beenakker. Z. Naturforsch. 33a (1978) 749. 5) J.N. Breunese, L.J.F. Hermans and J.J.M. Beenakker, Physica 126A (1984) 66. 6) H. Vestner, Z. Naturforsch. 28a (1973) 869. 7) H. Vestner, 2. Naturforsch. 31a (1976) 540. 8) V.D. Borman, S.Yu. Krylov and B.I. Nikolayev, Zh. Eksp. Teor. Fiz. 76 (1979) 1551; Sov. Phys. JETP 49 (1979) 787. 9) E.A. Mason and E. Mazur, Physica 130A (1985) 437. 10) B.I.M. ten Bosch, J.J.M. Beenakker and I. KuSEer, Physica 134A (1986) 522, to be referred to as II. 11) B.I.M. ten Bosch, J.J.M. Beenakker and I. KuSEer, Physica 123A (1984) 443, to be referred to as I. 12) C. Cercignani and C.D. Pagani, Phys. Fluids 9 (1966) 1167. 13) S.K. Loyalka, Z. Naturforsch. 26a (1971) 964. 14) T. Kline and I. KuSEer, Phys. Fluids 15 (1972) 1018. 15) M.H. de Wit, Thesis, Technical University of Eindhoven (1975). 16) C. Cercignani, Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburgh, 1975).
286
B.I.M.
TEN
BOSCH
et al.
The Method of Weighted Residuals and Variational Principles (Academic 17) B.A. Finlayson, Press, New York, London, 1972). 18) F.R. McCourt and R.F. Snider, J. Chem. Phys. 47 (1967) 4117. W.A.P. Denissen, L.J.F. Hermans. H.F.P. Knaap and J.J.M. Beenakker, 19) B.J. Thijsse, Physica 97A (1979) 467. L.J.F. Hermans and J.J.M. Beenakker, Physica 130A (1985) 505. 20) H. van Houten, P. Oudeman, L.J.F. Hermans and J.J.M. Beenakker, Physica 91A (1978) 21) G.E.J. Eggermont, 345. 22) B.I.M. ten Bosch, Chem. Phys. Lett. 122 (1985) 230. 23) A.C. Levi, F.R. McCourt and J.J.M. Beenakker, Physica 42 (1969) 363. The Internal Constitution of the Stars (Cambridge Univ. Press, Cambridge, 24) A.S. Eddington, 1926) p. 333. Phys. Fluids 14 (1971) 2291. 25) S.K. Loyalka, Z. Naturforsch. 26a (1971) 1708. 26) S.K. Loyalka,