Journal of Mathematical Analysis and Applications 230, 57]69 Ž1999.

Article ID jmaa.1998.6168, available online at http:rrwww.idealibrary.com on

Stability and Newton’s Law of Cooling in Double

Diffusive FlowU
K. A. Ames†

Department of Mathematical Sciences, Uni¨ ersity of Alabama in Hunts¨ ille,

Hunts¨ ille, Alabama 35899

and

B. Straughan ‡

Department of Mathematics, Uni¨ ersity Gardens, Glasgow G12 8QW, United Kingdom

Submitted by William F. Ames

Received September 25, 1998

We study the system of equations for flow of a viscous heat conducting fluid
which has dissolved in a salt field. Under the approximation, often employed in the
fluid dynamical literature, that the Prandtl number be infinite, we develop a priori
bounds and we show the solution depends continuously on changes in the Newton
cooling coefficient. Q 1999 Academic Press

1. INTRODUCTION

Many models for physical systems are governed by partial differential

equations. Frequently the solution to such a system can depend critically
on a coefficient in an equation or in a boundary condition. When the
coefficient is changed the solution can vary enormously. One needs to
develop, therefore, a stability analysis for dependence of the solution on
the model itself. This class of stability problems has been called continuous
dependence on modeling, or structural stability, see, e.g., Flavin and
Rionero w7x, Ames and Straughan w4x. Into this class of problems one must

U
This work was supported in part by NATO Grant Number CRG 910630.
†
E-mail address: ames@math.uah.edu.
‡
E-mail address: bs@maths.gla.ac.uk.

57
0022-247Xr99 $30.00
Copyright Q 1999 by Academic Press
All rights of reproduction in any form reserved.
58 AMES AND STRAUGHAN

also include changes in the partial differential equations themselves. For

instance, frequently in stability studies in fluid dynamics, one linearizes the
equations and one studies instability according to the linearized system of
partial differential equations. It is very important to know that the solution
found by such a linearization process depends continuously on the change
in the model when the appropriate nonlinearity is added, so that the
linearized analysis does actually yield information pertinent to the nonlin-
ear situation. Continuous dependence on modeling, or structural stability
analyses, have been the focus of much research. For example, the papers
of Ames and Payne w1]3x, Bennett w5, 6x, Franchi and Straughan w8x, Lin
and Payne w12, 13x, Payne w15x, Payne and Song w17x, Payne and Straughan
w18]21x, and Straughan and Hutter w26x, deal with various aspects of
continuous dependence on modeling in a variety of practical situations.
Also, the monographs by Flavin and Rionero w7x and Ames and Straughan
w4x treat many examples of structural stability, and many further references
can be found in these books.
In a paper, Ames and Payne w3x presented an analysis of continuous
dependence on modeling for a viscous fluid and their work concentrates
on establishing continuous dependence on the cooling coefficient. If T
denotes the fluid temperature then Newton’s law of cooling may be written
as

­T
q k T s F Ž x, t . , Ž 1.1.
­n

where ­r­ n denotes the outward normal derivative on the boundary, ­ V,

of the finite volume occupied by the fluid, V. The function F is prescribed
and the coefficient k is the cooling coefficient. An analysis of a similar
problem in porous medium flow is contained in Franchi and Straughan w8x.
Ames and Payne w3x deal with the Navier]Stokes equations for a heat
conducting viscous fluid and their analysis encounters trouble from the
nonlinear convective term Žy ? = . y where y is the fluid velocity. They do
establish continuous dependence on k , but due to the convective nonlin-
earity their analysis is conditional on the size of the initial data. In this
paper we investigate continuous dependence on the cooling coefficient k
in a model for viscous flow which is nonisothermal and also includes a
variable salt field. We employ an approximation often employed in the
fluid dynamical literature, namely, we let the Prandtl number, nrk T , be
infinite. Here n and k T are the kinematic viscosity and the thermal
diffusivity, respectively. This approximation is frequently employed, see,
e.g., Thess and Bestehorn w28x and usually allows more solution structure
to be observed because the convective nonlinearity is removed, although
all other nonlinear terms are retained. In this work this approximation
 STABILITY AND NEWTON’S LAW 59

allows us to obtain continuous dependence on k regardless of the size of

the initial data.
The problem of incompressible flow in a viscous fluid, accounting for
variable temperature and salt fields is known as double diffusive flow. The
basic equations governing such a situation are
­y i 1 gi
q y jy i , j s y p, i q n D y i q r ŽT , S. , Ž 1.2.
­t r0 r0
­y i
s 0, Ž 1.3.
­ xi
­T
q y i T , i s k T DT , Ž 1.4.
­t
­S
q y i S , i s k S D S, Ž 1.5.
­t
where y i , T, S, p are the velocity, temperature, salt, and pressure fields, r 0
is a constant, g is the gravity vector, k T and k S are the thermal and salt
diffusivities, D is the Laplacian operator, r ŽT, S . is the density field, and
standard indicial notation is employed in conjunction with the Einstein
summation convention. Due to the two diffusion coefficients, two species,
namely, T and S, diffuse during the flow process and hence such flows are
termed double diffusive.
In this work we do not adopt a Boussinesq approximation and we
assume the buoyancy force, i.e., the g i rrr 0 term, is linear in T and S. In
general, this is not so, see, e.g., Mellor w14x, and the references therein,
where nonlinear density laws are necessary to describe various geophysical
processes. In fact, an equation of state employed in oceanography by
Mellor w14, p. 114x is
5 4 2
r s ro q Ý aiT i q Ý biT i Sq S 3r2 q c 4 S 2 ,
is1
ž is0
/ ž Ý ciT i
is0
/
for precise constants r o , a i , bi , and c i . We leave the equation of state, i.e.,
the r ŽT, S . relation, general, although we require it to be essentially
polynomial. While here we specifically study double diffusive flow we could
generalize our result and deal with a multicomponent diffusion system. By
this we mean allowing N different salt fields S1 , . . . , SN in system Ž1.5.. We
would then have N equations replacing Ž1.5. and in Ž1.2. we would need
r s r ŽT, S1 , . . . , SN .. Multidiffusion processes are prevalent in life in in-
dustrial and in biotechnological situations. For example, in a fermentation
reaction, there are many species each undergoing diffusive-chemical flow.
60 AMES AND STRAUGHAN

Mathematical studies of multicomponent flows are predicting novel ef-

fects, e.g., Straughan and Walker w27x show that convection in a triply
diffusive system with a density linear in S1 , S2 but quadratic in T can lead
to the onset of convection occurring simultaneously by stationary convec-
tion and by Hopf bifurcation. While the two forms of convection can occur
for exactly the same set of parameters, the cell sizes in the convection
pattern will be different. It is inevitable that future studies of biotechno-
logical systems will lead to a whole plethora of new effects and so a
generalization of true continuous dependence on modeling results of this
paper to the multicomponent system is worthwhile.
The paper now moves on to study continuous dependence on the cooling
coefficient k and in the final section we indicate how to generalize the
results to other fluid systems.

2. CONTINUOUS DEPENDENCE ON THE

COOLING COEFFICIENT

In this section we study equations Ž1.2. ] Ž1.5. on the domain V = Ž0, Tˆ.
where Tˆ is arbitrary and V is a bounded domain in R 3 with boundary ­ V
smooth enough to admit applications of the divergence theorem. On the
boundary ­ V we have the no-slip condition,

y i s 0, on ­ V = Ž 0, Tˆ . , Ž 2.1.
the temperature field satisfies Ž1.1. with k ) 0, and there is no flux of salt
out of the domain so
­S
s 0, on ­ V = Ž 0, Tˆ . . Ž 2.2.
­n
Throughout, 5 ? 5 and Ž , . denote the norm and inner product on L2 Ž V .
and 5 ? 5 p denotes the L p Ž V . norm for 1 - p although we omit the 2 when
L2 Ž V . is intended.
The next step is to nondimensionalize equations Ž1.2. ] Ž1.5. and if we let
L be a length scale and if we choose T s L 2rk T and U s k T rL L as time
and velocity scales, then in terms of the Prandtl number Pr s nrk T and
the diffusivity ratio s s k T rk S , Eqs. Ž1.2. ] Ž1.5. become
1
Ž y i , t q y i y i , j . s yp, i q D y i q g i r Ž T , S . , Ž 2.3.
Pr
­y i
s 0, Ž 2.4.
­ xi
 STABILITY AND NEWTON’S LAW 61

T , t q y i T , i s DT , Ž 2.5.
s Ž S , t q y i S , i . s D S. Ž 2.6.
In Ž2.3. the expression for r ŽT, S . is also suitably nondimensionalized but
we leave this understood because we wish to leave this function in a
general form.
To study continuous dependence on k we let Žy i , T, S, p . and
Žy iU , T U , SU , pU . be two solutions to Ž2.3. ] Ž2.6. subject to boundary condi-
tions Ž1.1., Ž2.1., and Ž2.2., but for different cooling coefficients k and k U ,
respectively. Define the difference variables Ž u i , u , s, p . by

u i s y iU y y i , u s TU y T , s s SU y S,
U
p s p y p, U
v sk yk, Ž 2.7.

and then in the approximation Pr ª ` these variables satisfy the system,

p , i s D u i q g i r Ž T U , SU . y r Ž T , S . , Ž 2.8.
u i , i s 0, Ž 2.9.
­u
q y iUu , i q u i T , i s D u , Ž 2.10.
­t
­s
s ž q y iU s , i q u i S , i s D s.
/ Ž 2.11.
­t

These equations hold on V = Ž0, Tˆ. and the boundary conditions are

­u ­s
u i s 0, q k Uu q v T s 0, s 0, Ž 2.12.
­n ­n
on ­ V = Ž0, Tˆ.. Initial conditions are given as

T Ž x, 0 . s T0 Ž x . , S Ž x, 0 . s S0 Ž x . , Ž 2.13.
with ŽT, S . and ŽT U , SU . both satisfying Ž2.13. for the same functions T0 , S0
so that
u Ž x, 0 . s 0, s Ž x, 0 . s 0. Ž 2.14.
ŽFrom Ž2.8. and Ž2.9. p can be found at t s 0 by solving a forced Poisson
equation and then u i Ž x, 0. solved from Ž2.8...
By Taylor’s theorem there are ŽT1 , S1 ., ŽT2 , S2 . such that

­r ­r
r Ž T U , SU . y r Ž T , S . s Ž T1 , S1 . u q Ž T2 , S2 . D s, Ž 2.15.
­T ­S
62 AMES AND STRAUGHAN

where Ti g ŽT, T U . or Ti g ŽT U , T ., and S i g Ž S, SU . or Si g Ž SU , S . de-

pending on whether T - T U or T U - T and whether S - SU or SU - S.
We henceforth write ­rr­ T s r T and ­rr­ S s rS where the arguments
ŽT1 , S1 ., etc. are understood.
We now multiply Ž2.8. by u i and integrate. After use of Ž2.12. and Ž2.15.
we derive
5 =u 5 2 s Ž g i r T u , u i . q Ž g i rS s, u i . . Ž 2.16.
Now, without loss of generality, we take < g < F 1 and then we employ the
Cauchy]Schwarz and Holder inequalities in Ž2.16. to obtain

5 =u 5 2 F 5 u 5 5 u 5 6 5 r T 5 3 q 5 s 5 5 u 5 6 5 rS 5 3 . Ž 2.17.
Next, for c1 ) 0 we employ the Sobolev inequality,
5 f 5 6 F C1 5 f 5 H 1 ,

for functions f, where 5 ? 5 H 1 denotes the norm of the space H 1 Ž V .. Thus,

from Ž2.17. we may see that

5 =u 5 2 F  c1 5 u 5 H 1  5 r T 5 3 5 u 5 q 5 rS 5 3 5 s 5 4 4 . Ž 2.18.
´ inequality it follows that
Furthermore, from Poincare’s
l1
5 =u 5 2 G 5 u 5 2H 1 ,
ž 1 q l1 /
where l1 is the first eigenvalue in the membrane problem for V. Thus,
employing this inequality in Ž2.18. we may obtain
5 u 5 H 1 F c 2  5 r T 5 3 5 u 5 q 5 rS 5 3 5 s 5 4 , Ž 2.19.
where
c1 Ž 1 q l1 .
c2 s .
l1

The next step is to multiply Ž2.10. by u , to multiply Ž2.11. by s, and to

integrate each over V to find
d
5 u 5 2 s 2 Ž u i T , u , i . y 2 5 =u 5 2 y 2 k U E­ V u 2
dA y 2 v E­ V Tu dA,
dt
Ž 2.20.
d
s 5 s 5 2 s 2 s Ž u i S, s , i . y 2 5 =s 5 2 . Ž 2.21.
dt
 STABILITY AND NEWTON’S LAW 63

We next employ the arithmetic]geometric mean inequality on the bound-

ary terms in Ž2.20. and then we employ the Cauchy]Schwarz and Sobolev
inequalities on the first terms on the right of Ž2.20. and Ž2.21.. The
outcomes are

d 1 v2
5u 5 2 F c12 5 u 5 2H 1 5 T 5 32 q E­ V T 2
dA, Ž 2.22.
dt 2 2kU
d 1
s 5 s5 2 F c12 5 u 5 2H 1 5 S 5 32 . Ž 2.23.
dt 2

We now require a priori bounds for 5 r T 5 3 , 5 rS 5 3 , 5 T 5 3 , and 5 S 5 3 . If

r ŽT, S . is a general polynomial in T and S with highest order terms of
order n q 1, n G 1, then it is sufficient that we produce bounds for 5 T 5 3n
and 5 S 5 3 n . If r has a more complicated form then we can still proceed
provided we can bound 5 r T 5 3 and 5 rS 5 3 by terms of the form 5 T 5 3 j and
5 S 5 3 j , for various j F n. For example, for the function in Ž1.6. r T contains
a term of the form TS 3r2 while rS contains T 2 S 1r2 . We can easily use the
arithmetic]geometric mean inequality to bound these as necessary. Be-
cause S is a concentration S Ž x, t . G 0. Likewise we can always choose the
temperature scale such that T Ž x, t . G 0, e.g., work in degrees Kelvin.
Now, multiply Ž2.5. by T 3 ny1 and integrate using Ž1.1. and Ž2.1. to find

1 d
HV T 3n
dx s yk T E­ V T 3n
dA q E­ VT 3 ny1
F dA
3n dt

y HV =T ? =Ž T 3 ny1
. dx. Ž 2.24.

By using Young’s inequality we see that

3 ny1
1 3n y 1
F F kT T q
T 3 ny1 3n
3n ž 3n k T / F 3n ,

and then from Ž2.24. it follows that

3 ny1
d 3n y 1
T 3 n dx F
dt
HV ž 3n k T / E­ V F 3n
dA. Ž 2.25.

Integration of Ž2.25. leads to the required bound,

5 T 5 33 nn F 5 T0 5 33 nn q d1 Ž n . s d 2 Ž n . , Ž 2.26.
64 AMES AND STRAUGHAN

where d1Ž n. is the data term,

3 ny1
3n y 1 t
d1 Ž n . s dA dh ,
ž 3k T n / Ho E­ V F 3n
Ž 2.27.

and d 2 Ž n. is the data term defined in Ž2.26.. A similar procedure applied

to Ž2.6. using Ž2.2. leads to

5 S 5 33 nn F 5 S0 5 33 nn s d 3 Ž n . , Ž 2.28.
where the data term d 3 Ž n. is as indicated.
Hence, for a given relation r ŽT, S . we may employ Ž2.26. and Ž2.28. in
Ž2.19. to see there are computable constants k 1 and k 2 , dependent only on
c 2 , d 2 , d 3 , and n Žwhich depends on r ., such that

5 u 5 2H 1 F k 1 5 u 5 2 q k 2 5 s 5 2 . Ž 2.29.
The bound Ž2.29. is employed in Ž2.22. and Ž2.23. together with Ž2.26. and
Ž2.28. to deduce

dF v2
F kF q E­ V T 2
dA, Ž 2.30.
dt 2kU
where the constant k depends only on data and time and is given by

k Ž t . s 12 c12 d 21r2 Ž 1 . q sy1d 31r2 Ž 1 . max  k 1 , k 2 4 , Ž 2.31.

and where F is defined by
2
FŽ t. s u Ž t. q 5 s5 2. Ž 2.32.
Upon integration, from Ž2.30. we obtain

v2 t t
FŽ t. F
2kU
H0 ½ exp Hh k Ž j . d j E­ V T 2
5
dA dh

v2 t t
F
2kU
exp H0 k Ž j . d j H0 E­ V T 2
dA dh . Ž 2.33.

It remains to bound E­ V T 2 dA. To do this multiply Ž2.5. by T to find

d 1
5 T 5 2 s y5 =T 5 2 y k E­ V T 2
dA y k E­ V TF dA
dt 2
1 1
F y5 =T 5 2 y k E­ V T 2
dA q k E­ V F 2
dA,
2 2
 STABILITY AND NEWTON’S LAW 65

where the arithmetic]geometric mean inequality has been used. Thus by

integration,
t t
H0 E­ V T 2
dA dh F ky1 5 T0 5 2 q H0 E­ V F 2
dA dh s d 4 . Ž 2.34.

Hence, using Ž2.34. in Ž2.33. we are led to

2 2
u Ž t. q sŽ t . F d5 v 2 , Ž 2.35.
where the constant d5 depends only on data known a priori, the geometry
of V, and time, and is given by
d4 t
d5 s
2k U exp H0 k Ž h . dh . Ž 2.36.

Inequality Ž2.35. establishes continuous dependence on the cooling

coefficient in the measures 5 u Ž t .5 and 5 sŽ t .5. We also note that by utilizing
Ž2.19., Ž2.26., and Ž2.28. we can obtain a similar continuous dependence
estimate in the measures 5 =uŽ t .5 and 5 uŽ t .5.

3. CONTINUOUS DEPENDENCE ON THE COOLING

COEFFICIENT IN OTHER DOUBLE
DIFFUSIVE FLOWS

In this section we indicate how the result of the last section may be
achieved in other double Žor multi. diffusive flow systems.

3.1. The Brinkman Equations

For flow in a saturated porous medium when the porosity is close to 1
then the Brinkman equations are deemed appropriate. These equations
have been much studied in a variety of contexts, see, e.g., Franchi and
Straughan w8x, Guo and Kaloni w9, 10x, Qin and Kaloni w24x, and Payne and
Song w17x. For the double diffusive flow scenario encountered in Sections 1
and 2 and the references therein, the analogous Brinkman system of
equations, involves again Ž2.4. ] Ž2.6. with the boundary conditions Ž1.1.,
Ž2.1., and Ž2.2. being unchanged. However, the equation which replaces
Ž2.3. is
p , i s D y i y ay i q g i r Ž T , S . , Ž 3.1.
where a is a positive constant involving the viscosity and permeability. The
details of the analogous calculation to that of Section 2 for continuous
dependence on k are easily worked out following the method of Section 2.
66 AMES AND STRAUGHAN

3.2. The Brinkman]Forchheimer Equations

These equations are believed suitable in the situation when the porosity
is again close to 1, but the flow velocity is larger than that for which the
Brinkman equations are deemed valid. These equations are studied for
various problems of stability, existence, convergence, and continuous de-
pendence by Kaloni and Guo w11x, Qin and Chadam w23x, Qin and Kaloni
w25x, and by Payne and Straughan w21x. The Brinkman]Forchheimer system
again comprises Ž2.4. ] Ž2.6. together with the boundary conditions Ž1.1.,
Ž2.1., and Ž2.2.. Equation Ž2.3. is now replaced by
p , i s D y i y ay i y b < y < y i q g i r Ž T , S . , Ž 3.2.
where a is as before and b is now a positive constant. We do not go into
detail over the proof of continuous dependence on k although we observe
it follows as in Section 2 when we use the relation, Payne and Straughan
w21x,

Ž < y U < y iU y < y < y i . u i s 12 Ž < y U < q < y < . u i u i q 12 Ž < y U < y < y < . Ž < y U < q < y < . .
2

3.3. Stokes Flow in the Double Diffusi¨ e Case

We cannot discard the Žy ? = .T and Žy ? = . S terms from Ž2.5. and Ž2.6.
because the temperature and salt gradients may be large enough even
though the velocity field and its gradient are small. This point is stressed in
Payne and Straughan w22x who establish continuous dependence on the
spatial geometry in nonisothermal Stokes flow. Thus, the relevant equa-
tions are Ž2.4. ] Ž2.6. together with the boundary conditions Ž1.1., Ž2.1., and
Ž2.2.. Equation Ž2.3. is replaced by
1
y i , t s yp, i q D y i q g i r Ž T , S . . Ž 3.3.
Pr
In studying continuous dependence on k , estimates Ž2.22., Ž2.23., Ž2.26.
and Ž2.28. follow as before. However, we must develop a different route to
an inequality like Ž2.19.. To do this we observe that the difference
variables satisfy
1
u s yp , i q D u i q g i Ž r T u q rS s . . Ž 3.4.
Pr i , t
We multiply Ž3.4. by u i and we integrate over V to find
1 d
5 u 5 2 q 5 =u 5 2 s g i Ž r T u , u i . q g i Ž rS s, u i .
2 Pr dt
F 5 r T 5 3 5 u 5 6 5 u 5 q 5 rS 5 3 5 u 5 6 5 s 5
F c1 5 u 5 H 1 5 r T 5 3 5 u 5 q 5 r S 5 3 5 s 5 .
 STABILITY AND NEWTON’S LAW 67

If we now use the estimates Ž2.26. and Ž2.28. we may then find constants
k 3 , k 4 such that
1 d
5 u 5 2 q 5 =u 5 2 F c1 5 u 5 H 1  k 3 5 u 5 q k 4 5 s 5 4 . Ž 3.5.
2 Pr dt
Hence after integration and use of the inequality before Ž2.19. we may
obtain
1 l1 t t
5 u5 2 q 5 u 5 2H 1 dh F c1 H0 5 u 5  k 3 5 u 5 q k 4 5 s 5 4 dh . Ž 3.6.
2 Pr ž 1 q l1 /H 0
H1

After use of the Cauchy]Schwarz and the arithmetic]geometric mean

inequalities we may thus produce
l1 t t
H 5 u5
2Ž 1 q l . 0 1
2
H1 dh F k 5 H0 Ž 5 u 5 2
q 5 s 5 2 . dh , Ž 3.7.

where k 5 s c12 ŽŽ1 q l1 .rl1 .max k 32 , k 42 4 .

We may now obtain continuous dependence on v from Ž2.22. and Ž2.23.
by integration to find that

1 t v2 t
5u 5 2 q 5 s5 2 F c12 H0 5 u 5 2H 1 5 T 5 23 dh q U H0 E­ V T 2
dA dh
2 2k
c12 t
q H 5 u5
2s 0
H15S53
2 2
dh

t v2 t
F k6 H0 5 u 5 2H 1 dh q H0 E­ V T 2
dA dh ,
2kU
where
c12 c12
k6 s max d 22r3 q max d 32r3 .
2 w0, Tˆ x 2s w0, Tˆ x

Then using Ž3.7. one finds for a data constant k 7 ,

t v2 t
5u 5 2 q 5 s5 2 F k7 H0 Ž 5 u 5 2 q 5 s 5 2 . dh q U H0 E­ V T 2
dA dh . Ž 3.8.
2k
Continuous dependence on k follows from Ž3.8. after using Ž2.34. and
deriving an inequality like
t
H0 Ž 5 u 5 2
q 5 s 5 2 . dh F k 8 v 2 . Ž 3.9.
68 AMES AND STRAUGHAN

3.4. Multicomponent Diffusi¨ e Flows

Finally we observe that we can extend the results of Section 2 and of
Sections 3.1, 3.2, and 3.3 to the case of multicomponent flows involving N
salt fields of concentrations S1 , . . . , SN . Thus we can establish continuous
dependence on the cooling coefficient for infinite Prandtl number flow,
flow in a Brinkman, or Brinkman]Forchheimer porous material, or in
Stokes flow, even with T and N salt fields. As pointed out in the
Introduction such systems are playing an increasing role in the biotechno-
logical industry, cf., Straughan and Walker w27x, and the references therein.

REFERENCES

1. K. A. Ames and L. E. Payne, Continuous dependence results for solutions of the

Navier]Stokes equations backward in time, Nonlinear Anal. Theory Methods Appl. 23
Ž1994., 103]113.
2. K. A. Ames and L. E. Payne, On stabilizing against modeling errors in a penetrative
convection problem for a porous medium, Math. Models Methods Appl. Sci. 4 Ž1994.,
733]740.
3. K. A. Ames and L. E. Payne, Continuous dependence results for a problem in penetrative
convection, Quart. Appl. Math. 55 Ž1997., 769]790.
4. K. A. Ames and B. Straughan, ‘‘Non-Standard and Improperly Posed Problems,’’ Mathe-
matics in Science and Engineering Series, Vol. 194, Academic Press, San Diego, CA,
1997.
5. A. D. Bennett, ‘‘Continuous Dependence on Modeling in the Cauchy Problem for Second
Order Nonlinear Partial Differential Equations,’’ Ph.D. thesis, Cornell University, 1986.
6. A. D. Bennett, Continuous dependence on modeling in the Cauchy problem for elliptic
equations, Differential Integral Equations 4 Ž1991., 1311]1324.
7. J. N. Flavin and S. Rionero, ‘‘Qualitative Estimates for Partial Differential Equations,’’
CRC Press, Boca Raton, FL, 1995.
8. F. Franchi and B. Straughan, Structural stability for the Brinkman equations of porous
media, Math. Methods Appl. Sci. 19 Ž1996., 1335]1347.
9. J. Guo and P. N. Kaloni, Nonlinear stability problem of a rotating doubly diffusive porous
layer, J. Math. Anal. Appl. 190 Ž1995., 373]390.
10. J. Guo and P. N. Kaloni, Doubly diffusive convection in a porous medium, nonlinear
stability, and the Brinkman effect, Stud. Appl. Math. 94 Ž1995., 341]358.
11. P. N. Kaloni and J. Guo, Steady nonlinear double-diffusive convection in a porous
medium based upon the Brinkman]Forchheimer model, J. Math. Anal. Appl. 204 Ž1996.,
138]155.
12. C. Lin and L. E. Payne, On the spatial decay of ill-posed parabolic problems, Math.
Models Methods Appl. Sci. 3 Ž1993., 563]575.
13. C. Lin and L. E. Payne, The influence of domain and diffusivity perturbations on the
decay of end effects in heat conduction, SIAM J. Math. Anal. 25 Ž1994., 1242]1258.
14. G. L. Mellor, ‘‘Introduction to Physical Oceanography,’’ Springer-Verlag, Berlin, Heidel-
berg, 1996.
 STABILITY AND NEWTON’S LAW 69

15. L. E. Payne, A special class of maximum principles with applications to nonlinear

boundary value problems, in ‘‘Proceedings of the First World Congress Nonlinear
Analysts,’’ Tampa, FL. ŽV. Lakshmikanthan, Ed.., de Gruyter, Berlin, New York, pp.
25]39, 1996.
16. L. E. Payne and J. C. Song, Phragmen]Lindelof ¨ and continuous dependence type results
in generalized heat conduction, Z. Angew. Math. Phys. 47 Ž1996., 527]538.
17. L. E. Payne and J. C. Song, Spatial decay estimates for the Brinkman and Darcy flows in
a semi-infinite cylinder, Continuum Mech. Thermodynamics 9 Ž1997., 175]190.
18. L. E. Payne and B. Straughan, Stability in the initial-time geometry problem for the
Brinkman and Darcy equations of flow in porous media, J. Math. Pures Appl. 75 Ž1996.,
225]271.
19. L. E. Payne and B. Straughan, Analysis of the boundary condition at the interface
between a viscous fluid and a porous medium and related modelling questions, J. Math.
Pures Appl. 77 Ž1998., 317]354.
20. L. E. Payne and B. Straughan, Structural stability for the Darcy equations of flow in
porous media, Proc. Roy. Soc. London Ser. A 454 Ž1998., 1691]1698.
21. L. E. Payne and B. Straughan, Convergence and continuous dependence for the
Brinkman]Forchheimer equations, Stud. Appl. Math., in press.
22. L. E. Payne and B. Straughan, Analysis of errors in the spatial geometry for temperature
dependent Stokes flow, to appear.
23. Y. Qin and J. Chadam, Nonlinear convective stability in a porous medium with tempera-
ture-dependent viscosity and inertial drag, Stud. Appl. Math. 96 Ž1996., 273]288.
24. Y. Qin and P. N. Kaloni, Convective instabilities in anisotropic porous media, Stud. Appl.
Math. 91 Ž1994., 189]204.
25. Y. Qin and P. N. Kaloni, Spatial decay estimates for plane flow in the
Brinkman]Forchheimer model, Quart. Appl. Math. 56 Ž1998., 71]87.
26. B. Straughan and K. Hutter, A priori bounds and structural stability for double diffusive
convection incorporating the Soret effect, Proc. Roy. Soc. London Ser. A, in press.
27. B. Straughan and D. W. Walker, Multicomponent diffusion and penetrative convection,
Fluid Dyn. Res. 19 Ž1997., 77]89.
´
28. A. Thess and M. Bestehorn, Planform selection in Benard]Marangoni convection: l
hexagons versus g hexagons, Phys. Re¨ . E 52 Ž1995., 6358]6367.