Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Cylinder
Author(s): B. W. Martin and A. Payne
Source: Proceedings of the Royal Society of London. Series A, Mathematical and Physical
Sciences, Vol. 328, No. 1572 (May 2, 1972), pp. 123-141
Published by: The Royal Society
Stable URL: http://www.jstor.org/stable/78086
Accessed: 14/05/2010 21:45
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=rsl.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the
Royal Society of London. Series A, Mathematical and Physical Sciences.
http://www.jstor.org
Proc. B. Soc. Lond. A. 328, 123-141 (1972)
Printed in Great Britain
NOMENCLATURE
Subscripts
P pertaining to the node at which variable value is being iterated
N, S, E, W pertaining to nodes neighbouring node P to the north, south, east and
west respectively
INTRODUCTION
Ta. (= 2QS2R2b3/v2(RI+2))
Tangential flow developmentfor laminar axial flow 125
increasesnonlinearly but monotonicallyfrom 1708 for narrowannuli to 3020 when
R1/R2 0.5 and the outer cylinder is stationary. Coles's (I967) suggestion that the
effect of R1/R2on the stability boundary can be suppressedby a suitable choice of
variables is well-supported by the measurements of Taylor (I923), Lewis (1928),
Donnelly & Fultz (I960) and Coles (I965) over the range 0.5 < R1/R2 < 0.942.
For all superimposed axial velocity distributions the tangential velocity com-
ponent is subject to an axial entrance length, as illustrated in figure 1. At the en-
trance only filtid in contact with the rotating surface acquires a circumferential
velocity; the remaining fluid is swept downstream before tangential momentum
can diffuseacrossthe annulus.With increasingdistance downstreamthe tangential
boundary layer thickens eventually to fill the gap. In a sufficientlylong annulusthe
tangential profile further develops to approach Couette flow; when thus fully de-
veloped the tangential velocity distribution is independent of the axial coordinate.
The effect of axial flow on the stability of the fully developed tangential profile
has been theoretically treated by Goldstein(X937), Di Prima (196o), Chandrasekhar
(I960, i962) and Krueger &Di Prima (I964). Measurementsare reportedby Cornish
(1933), Fage (I938), Kaye & Elgar (I958), Donnelly & Fultz (i96o), Yamada (1962),
Snyder (I962), Astill (i96i, I964) and Flower, MacLeod & Shahbenderian (I969).
By hot-wire velocity measurements in fully developed axial flow Kaye & Elgar
(I958) were the first to identify the four flow regimesshown by domain in the Ta-Re
map in figure 2. These are laminar flow with an without vortices and turbulent
flow with and without vortices, such that Ta, increases with Re but Re, diminishes
with increasing Ta from 2300 at zero Ta.
In many practical applications the annulus will be of insufficient length for fully
developed tangential flow to occur. Astill (i96i, I964) found from smoke patterns
of the flow that in laminar developing tangential flow Taylor vortices are generated
within the tangential boundary layer, originating as oscillating ripples near the
126 B. W. Martin and A. Payne
rotating core. The wave motion progresses across the gap as the fluid moves down-
stream finally curling over to form opposed pairs of trapezoidal vortices as in figure 3.
For given Re, the point of origin of the ripples moves upstream with increasing Ta
while for given Ta it moves downstream with increasing Re. By noting the point of
origin of the first discernible ripple in an annulus of R1/R2 = 0.727 Astill developed
turbulent
turbulent
laminar )9 vortices
laminar
with
vortices
PREDICTION PROCEDURE
TABLE1
0 ~~a?, b?, c?, d?,
RV 1 1/2 12 0 (3)
o/R R2 KR2 _paV218Z (4)
0 I/pR2 1 -wR (5)
outer boundary
N N
L1R2
Like others of its kind, the finite-difference procedure of Gosman et al. focuses
attention on a finite number of points distributed over the solution domain as the
nodal points of a grid. An algebraic relation is sought between the value of 0 at a
typical node P of the grid in figure 4 (i) and its four immediate neighbours N, S, E
and W. This relation is obtained by double integration with respect to Z and B
of the differential equation (2), over a control volume or 'cell' (denoted by the
dotted lines in figure 4 (i)) which encloses P. The walls of the cell are supposed to be
mid-way between the grid lines, which need not be equally spaced. The resultant
finite-difference equation can be expressed in the following successive-substitution
form
p = CN\N+CS S+CEVE+CW W+D, (6)
where the Cs are positive coefficients which express the combined influence of
convection and diffusion and involve the values of a., b6, co, do, ~/, Z and R at
the N, S, E and W grid nodes. The term D resulting from the integration of Rdc
represents the total source of entity 0 in the cell.
The present finite-difference method differs from its predecessors either in the
generality of its framework or, following the work of Courant, Isaacson & Rees
(I95Z), Blair, Metropolis, Taub & Tsingou (I 957), Barakat & Clark (I966) and Spald-
ing (I966), in employing a unidirectional upwind-difference scheme; this gives rise
Tangential flow developmentfor laminar axial flow 129
to initial-value iterative solutions less prone to divergence and instability than
conventional backward, forward and central difference schemes, particularly for
high Reynolds number. Thus the gradient of vorticity is with backward reference to
the direction of flow. The present method also ensures that each finite-difference
equation obeys the relevant 0-conservation equation for each cell. The iteration
procedure is a Gauss-Siedel point method whereby freshly updated values are
employed immediately they become available. To minimize truncation errors the
grid lines are staggered to be closest together near the entry and solid boundaries
where gradients are steepest. The computer program is written in Fortran IV using
grid meshes 27 x 27.
(b) Treatment of boundary conditions
To complete the solution of an elliptic second-order partial differential equation,
boundary values must be specified which extend completely round the perimeter
of the solution domain in figure 5. Thus for equation (2) and table 1, values of R V,
(wlR)S,3>p, (RV) p
(o)IR)p (o,/R)p
t R~~P ~ RP
(RV)p (RV)p
(wl.R)N, /r p, (R V) p
inner
defining equations
(17), (15), (12) (8) (10) (17), (15), (12) (8) (I ) (10), (15), (12) (I0
(9) (9) (I 1) (9) (10) (10)
(12) (18), (14), (16) (13)1 (12) (18), (14), (16) (13) (|12) (10), (14), (16) (13)
o/R and if (or their first derivatives) must be supplied at all nodal points on the
boundaries. As already stated, solutions are obtained for the cases where the axial
velocity distribution (a) is fully developed at entry, (b) develops from a uniform
entry profile, (c) remains uniform throughout. (For this case wIR and tk are fixed
throughout the field by equations (10) and (11) below respectively; only RV is
iterated.) Expressions for w/R and #/ are therefore the same (i) at the entry boundary
for cases (b) and (c), and (ii) at the exit boundary for cases (a) and (b). Special treat-
ment of the finite-difference equation is required for the generally unknown vorticity
at the solid boundaries. The argument that wIRis readily calculable from standard
equations only applies to case (c), where an unreal axial profile is imposed; for cases
9 Vol. 328. A,
130 B. W. Martin and A. Payne
(a) and (b) the argument is premature because its presupposes that w/lRis indepen-
dent of the developing tangential flow. This is discussed further belo w.
The well-known profile for fully developed laminar axial flow in a concecntric
annulus can be expressed in terms of mean velocity W by
Jttfollows fron conatinuity considerations that, for axisymmetric flow U is then zero.
For case (a) at entry and cases (a) arid (b) at exit we may therefore write
For cases (b) and (c) at entry and case (c) at exit, where Wp- W,
au aw
U =
(X)p 0' (10)
and ?Ifp
__ Rp-R2
1 .1
pW 2 (11)
Since at the entry no fluid has acquired a tangential velocity it follows that for all
three cases
(R V)p = 0. (12)
The tangential fully developed Couette flow at the exit leads for all three cases to
In the case of the two solid boundaries in figure 5, if the inner rotating surface is
defined as the streamline
fp = O, (14)
the outer stationary surface is readily shown from continuity for all cases to be the
streamline given by
?Wp _R2-R 2
2 (15
pTW 2of (ti
The corresponlding boundary values of (R V)p for all cases result from application of
T'angentialflow developmentfor laminar axial flow 131
the zero-slip condition; equation (12) must be satisfied at the outer boundary, while
for the inner boundary
(RV)p = QR. (16)
Because the vorticity distribution along the solid boundaries is unknown, the
successive substitution equation (6) for iterating (/BR)p at a node adjacent to these
boundaries, as in figures 4 (ii) and (iii), is modified to the following implicit forms,
which make no direct appeal to the boundary values. For nodes adjacent to the
outer boundary we write for cases (a) and (b)
in. (4). For 1 > 0.001 this contribution is insignificant for the range of rotational
speeds over which Taylor vortices are absent.
For the limiting case of a stationary inner cylinder, case (b) compputations of the
developing axial velocity profile for RJ/B2 = 0. 8 were compared with the correspond-
ing predictions of Sparrow & Lin (I964). These workers linearized the momentum
equation by neglecting the second derivative D2VWV/Z2and applied a difference
procedure whereby the axial velocity results from subtracting a 'difference velocity'
from the fully developed value. Over the range 0.002 < I < 0.15 for which da is
correlated, the two solutions are found in an unpresented comparison to agree
within + 5 %. In the immediate entry region (I < 0.001) very large velocity gradielnts
appear to contribute errors to both solutions; within the practical limits of mesh
size significant truncation errors in computational procedures then become un-
avoidable. In subsequent computations advantage was taken of axial flow in-
dependence of developing tangential flow (a one-way uncoupling) by not iterating
values of ifp and (oBR)p for the bulk of case (a) computations; instead the inlitial
values given by (8) and (9) were used throughout iterations along the Z axis.
Before considering the effect of axial velocity distribution on the developing
Tangential flow developmentfor lamtnar axial flow 133
tangential flow, the reliability of the present numerical procedure for predicting the
latter may be gauged by reference to figure 6. This shows a comparison of successive
tangential velocity profiles for case (c) of axial plug flow and the corresponding pre-
dictions of Astill et al. (i968) by direct solution of the Navier-Stokes equations,
IiN1.0 .I
0O.9 9
F_ t~~~~~
0 50 1000
I9 ; 0 1
. ~1.0~-~~~~~~~~~~~~~
~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
.O
. .. 4
6and 0 b --foranl
Be,
2 =.54cm (a); 1700. -, 0, et0 c a sent
0
'0
0.7- %s0
0 50' 100
- axial position (cm)
~~~~~~~~axial
FIGUPRE
FIGuR.E 7.6. Comparisonof dimensionlessvelocity ((V
JQRR1) profiles in an annulus
prof-iles
VIS2RD R/2=0of
of
J2067 7
b = 1.54 cmRe,for17100. -,peetaayi
=064and castilet(a);.... ( as968); 0 rsn
canlsis
(CaS (c).rmntldt f sile a.( 6
case(c) ~~~
*, 03prmna aao sle l T6)
wherein the assumption of an average axi'al velocity across the annulus implies
uncouplingof tlheinteraction between the axial and tangential flows.,The'Reynolds
number is 17002b/R, = 0.482 and R/JR2=0.674. For each profile zero ve-locity
coincides with the axial position (measured in centimetres from the annulus
entrance) of the corresponding predictions. The two sets of predictions are in
extremely close accord.
lFigure7 com.pares present tangential velocity predictions along the annulus for
cases (a), (b) and (c), where Re- =1200 and b/R, (as in figure 6) is 0.482, and suggests
that interaction. between the flows influences more noticeably the tangential velocity
prIofile-than the axial velocity distributions. This is somnewhat clearer in the corre-
spondingpresentationof 60 as a function of I in figure8, wherethe effect of assumi'ng
a uniform axial profile is to reduce 6f,9at low I and near the walls below the valu'es for
9-2
134 B. W. Martin and A. Payne
developing or developed axial flow. This is because the latter involve smaller a V1aR
and greater residence times for fluid particles near the wall to transfer tangential
momentum than does uniform axial flow. Away from the walls the situation is
reversed; a uniform axial profile underestimates the developing or developed near-
parabolic values and at large I this effect dominates the tangential profile.
0.2 .
0 0.05 0.10
length parameter, I 2Z/(Re) b
FIGuRE 8. Comparison of development of dimensionless displacement thickness in an annulus
of RlIR2 0.674 for Re 1200. ---, Present analysis, case (a); ......, case (b);
case (c); , experimental data of Astill et al. (I968).
Differences in &#for cases (a) and (b) are relatively small, developing axial flow
leading to tangential displacement thicknesses up to 7 greater than for developed
0
axial flow for an intermediate range of 1. The two solutions predictably converge as
1 -+. 0 and as the axially developing flow approaches the full development which is
9900 achieved at I 0. 15. The above comments are largely reinforced by the
measurements of Astill et al. for axially developed flow included in figure 8, agree-
ment with corresponding predictions tending to improve with tangential flow de-
velopment. Discrepancies near the rotating surface, also noted by Astill et al.
particularly at low Re, are ascribed partly to non-linearities in the hot-wire calibra-
tion curve at the low fluid velocities used. It may be that the probe orientation neces-
sary to present the hot wire to the oncoming fluid results in operation in its own
wake owing to the presence of the tangential velocity component. The discrepancies
are unlikely to be attributable to predictive errors.
0.3
0.2 12 a 9
tia
0.1 -
S 0.12 - 9 0
0~~~~~~
0 ~ ~ ~ 0
,0 iaS? ... 10
0.04A
FIGURE9. Influence of axial Reynolds number upon growth of SO. Present analysis for annuli
of R1lR2 = 0.5; axial component fully developed, 0, Re = 100; El, Re = 1700.
FIGURE 10. Influence of axial Reynolds number upon growth of 80. Present analysis for
annuli of R1/R2 = 0.5, axial component fully developed, 0, Re = 100; *, Re = 150;
x, Re 300;E, Re = 500; Fl, Re = 1700.
This is demonstrated in figure 9 for case (a), where computed values of 80 are
plotted against I for R/BR2 = 0.5 and Reynolds numbers of 100 and 1700. That 6d is
independent of Re values below 1700 for I > 0.015 is more clearly evident in the
enlargement of the entry region in figure 10, which includes data for intermediate Re
values of 150, 300 and 500. Computed points for Be = l000 are omitted because of
their coincidence with those for 500 and 1700. For I < 0.015, only the departure of
9-3
136 B. W. Martin and A. Payne
points for Reynolds numbers of 100 and 150 from values for Re = 1700 is significant,
by percentages of 39 and 26 respectively at 1 = 0.0002. The solution for Re = 1700
may therefore be regarded as uniquely determining 40 when 1700 > Re > 300 if
I > 0.002. For 300 > Re > 100, the solution for Re = 1700 is valid provided that
Z > 0.6b. Similar unpresented results for case (b) and extreme R1/R2 of 0.05 and
0.98 suggest that the above conclusions are then equally valid provided that
Z > 0.7b is substituted for Z > 0.6 in the previous sentence.
0.5 000000
-~~ ~ 0 0 0 00
WC
o 0l
0g Ep
X Drz
3~~~~~~~~~~~~0 Q0ooo oo oo
0.1
60*0
(21)
where c and m are functions of R1/R2 readily determined by cross-plotting from
figure 11 and unpresented predictions for seven further values of R1/X2 and illu-
strated in figure 12. The rate of increase of both functions diminishes rapidly at
large R1/R2, but neither can be simply expressed as an algebraic function of this
parameter. Parallel unpresented computations for case (b) equally result in a power-
law relation between &tand I with values of c indistinguishable from those for case
(a) but with the slightly lower values of m for all R1/R2 shown in figure 12. These are
alsofor Re= 1700.
Ta ngential flow development for lamitnar axial flow 137
Values of 8o derived from figure 12 are within 3 %accuracy of the individually
computed solutions; as before, the limits of power-lawvalidity are 0.002 < I < 0.15
for300 < Re < 1700andZ > 0.7b,1 < 0.15for 100 < Re < 300. Thecorrespondence
of the predicted power-law dependency with the measurementsof 4 reported by
Astill. et al. (i )68) for Ta < 6 x 103 at Re = 1200 for case (a) and RJ/R2 0.674 is
shown in figure 11. In view of the marginaldifferencesbetween case (a) and case (b)
predictions, Astill's (I964) mean measurements of 6*for the latter case over the
range 1.38 x 1O4 < Ta < 3.12 x 1O5atRe= 1700andR1/R2 0.727arealsoincluded
in figure 11; these further confirmthe predictedpower-lawrelation. The accuracyof
computation i:nthis instance is shown more clearlythan figure 11 permits in table 2,
whereinAstill's (i 964)measurementsare comparedwith da computedfrom figure 12
for cases (a),and (b). The latter and more valid.comparison indicates agreement
within 3%.
0.8 --03
e0.4- t (31
0.3
0~~~~~~~~~~~02
S
* 0 ~ 0.4 0.8
~~~~~~~~~~~0.23
radius ratio R1JR2
FIGuRE 12. Values of coefficient c and exponent mnfor 0.05 < R1/R2 < 0.98.
m for case (a); - - - -, m for case (b); ....... c for both cases (a) and (b).
TABLE 2
8-
C4
0 0.4 0.8
radius ratio, R1/R2
FIGURE 13. Variationof coefficientn with R1fR2 for developedand
developingaxial flow.
Figure 13 shows n as a function of R1/R2 over the range 0.05 < R1/R2 < 0.98,
the relation being equally valid for cases (a) and (b). As R1/R2 -> 1.0, n approaches
an asymptote of about 1075. Because this tendency of n to independence from
radius ratio is shared by m as R1/R2 increases for both developed and developing
axial flow, over the range of R1/R2 of practical interest above 0.5 the stability cri-
terion expressed by (22) is almost uniquely determined by the length parameter 1.
The increasingly weak dependence of Ta, on RJ/R2 between values of 0.5 and
0.98 as I increases is illustrated in figure 14 for cases (a) and (b). Also included are
Astill's (I964) case (b) measurements of vortex transition for R1/R2 _0.727. To
within + 20 ? of all computed values in the range
0.5 <RR2< 0.98, 0.01 I < 0.15,
Tangential flow developmentfor laminar axial flow 139
the demarcation between the stable and unstable regions in figure 14 is represented
by
1150bRe117
Ta= =1 150 (24)
for both developed and developing axial flow. As with (21), the predictions in figure
14 and equation (24) are valid for 300 < Re < 1700 and also for 100 < Re < 300
provided that Z > 0.7b. As 1 increases above 0.15 the onset of fully developed
R1JR2 = 0.98
R1JR2 = 0.5 {\
unstable
X
k Ta5
a1050
stable \X
0.01 0.1
length parameter, 1 2ZIb(Re)
FIGURE 14. Astill's stability criterion using present analysis to predict 80. , case (a);
case, (b); *, experimental vortex transition data according to Astill (i964),
I/R2- 0.727.
Couette flow may be expected to render Ta, independent of Z and therefore only a
function of Re, as indicated in figure 2. However, since Astill's empirical stability
criterion was obtained only for R1/R2 = 0.727, it is conceivable'that Q2R613/V2 in (1)
is a function of R1/R2 rather than the quoted constant of 576. Though in the light
of previous experience with Taylor stability'phenomena such dependency is
unlikely to be significant, at least for larger R1/R2, it would nevertheless involve an
additional function of R1/R2 for Tah as given by (22) a'nd (24) and further research
is needed over the whole range of R1/B2to resolve this uncertainty.
CONCLUSIONS
The authors wish to acknowledge the support given by the Science Research
Council and the assistance rendered by Dr W. M. Pun. They also wish to thank Mrs
S. M. Price for her help in the preparation of this paper.
REFERENCES
Astill, K. N. 196i Modes of adiabatic flow in the entrance region of an annulus with an
inner rotating cylinder. Ph.D. thesis, Massachusetts Institute of Technology.
Astill, K. N. i964 Studies of the developing flow between concentric cylinders with the inner
cylinder rotating. Trans. A.S.M.E. J. Heat Transfer (C) 86, 383.
Astill, K. N., Ganley, J. T. & Martin, B. W. i968 The developing tangential velocity profile
for axial flow in an annulus with a rotating inner cylinder. Proc. R. Soc. Lond. A 307, 55.
Barakat, H. Z. & Clark, J. A. I966 Analytical and experimental study of transient laminar
natural convection flows in partially filled containers. Proc. 3rd Int. Heat Trans. Conf.
Clhicago,vol. II, paper no. 57, 152. New York: A. I. Chem. E.
Blair, A., Metropolis, N., Taub, A. H. & Tsingou, M. I957 A study of a numerical solution to
a two-dimensional hydrodynamic problem. Physics and mnathematics,LA-2165 (TID-4500,
13th ed., rev.). United States Atomic Energy Conmnission.
Chandrasekhar, S. 196O The hydrodynamic stability of viscid flow between co-axial cylinders.
Proc. natn. Acad. Sci., U.S.A. 46, 137.
Chandrasekhar, S. i962 The stability of spiral flow between rotating cylinders. Proc. R.
Soc. Lond. A 265, 188.
Coles, D. i965 Transition in circular Couette flow. J. Fluid Mech. 21, 385.
Coles, D. i967 A note on Taylor instability in circular Couette flow. Trans. A.S.M.E.
(J. appl. Mech.) (E) 34, 529.
Cornish, R. J. I933 Flow of water through fine clearances with relative motion of the
boundaries. Proc. R. Soc. Lond. A 140, 227.
Courant, R., Isaacson, E. & Rees, M. I952 On the solution of non-linear hyperbolic differen-
tial equations by finite differences. Commun. Pure appl. Math. 5, 243.
Di Prima, R. C. I960 The stability of a viscous fluid between rotating cylinders with an
axial flow. J. Fluid Mech. 9, 621.
Donnelly, R. J. & Fultz, D. I96o Experiments on the stability of spiral flow between
rotating cylinders. Proc. natn. Acad. Sci. U.S.A. 46, 1150.
Fage, A. I938 The influence of wall oscillations, wall rotation and entry eddies on the-break-
down of laminar flow in an annular pipe. Proc. R. Soc. Lond. A 165, 51-3.
Tangential flow developmentfor laminar axial flow 141
Flower, R. J., Macleod, N. & Shahbenderian, A. P. I969 The radial transfer of mass and
momentum in an axial fluid stream between co-axial rotating cylinders. Chem. Engng.
Sci. 24, 637.
Goldstein, S. I937 The stability of viscous fluid flow between rotating cylinders. Proc.
Carmb.phil. Soc. 33, 41.
Gosman, A. D., Pun, W. M., Runchal, A. K., Spalding, D. B. & Wolfshtein, M. ig69 Heat and
mass transfer in recirculating flows. London: Academic Press.
Kaye, J. & Elgar, E. C. I 958 Modes of adiabatic and diabatic fluid flow in an annulus with an
inner rotating cylinder. Trans. A.S.M.E. 80, 753.
Krueger, E. R. & Di Prima, R. C. I964 The stability of a viscous fluid between rotating
cylinders with an axial flow. J. Fluid Mech. 19, 528.
Lowis, J. W. 1928 An experimental study of the motion of a viscous liquid contained
between two co-axial cylinders. Proc. R. Soc. Lond. A 117, 388.
Schlichting, H. Ig60 Boundary-layer theory. New York: McGraw-Hill Book Co.
Snyder, H. A. I962 Experiments on the stability of spiral flow at low axial Reynolds num-
bers. Proc. R. Soc. Lond. A 265, 198.
Spalding, D. B. I966 Notes on the solution of the Navier-Stokes equations for steady two-
dimensional turbulent flow by finite-difference techniques. Cambridge, Massachusetts:
Northern Research and Engineering Corporation.
Sparrow, E. M. & Lin, S. H. I964 The developing laminar flow and pressure drop in the
entrance region of annular ducts. Trans. A.S.M.E. (J. Basic Eng.) (D) 86, 827.
Sparrow, E. M., Monro, W. D. & Jonsson, V. K. I964 Instability of the flow between rota-
ting cylinders: the wide gap problem. J. Fluid Mech. 20, 35.
Taylor, G. I. 1923 Stability of a viscous fluid contained between two rotating cylinders.
Phil. Trans. R. Soc. Lond. A 223, 289.
Walowit, J., Tsao, S. & Di Prima, R. C. I964 Stability of flow between arbitrarily spaced
concentric cylindrical surfaces including the effect of a radial temperature gradient.
Trans. A.S.M.E. (J. appl. Mech) (E) 31, 585.
Yamada, Y. I962 Resistance of a flow through an annulus with an inner rotating cylinder.
Bull. Jap. Soc. Mech. Engrs. 5, 302.