You are on page 1of 7

Exact Solutions to Difference Equation

Models of Burgers’ Equation


Ronald E. Mickens
Department of Physics, Atlanta University, Atlanta, Georgia 303 14

This article gives exact solutions to several difference equation models of Burgers’
equation. The particular cases considered correspond to the diffusion-free, nonlinear
steady-state and linear steady-state situations.

1. INTRODUCTION
The purpose of this paper is to provide exact solutions to a number of differ-
ence equation models of Burgers’ equation [ I ] . This equation
u, + uu, = vu,,,
v = constant.

is an appropriate simplification of the Navier-Stokes equation [2]. However, it


is also the governing equation for a number of one-dimensional flow systems
including weak shock propagation, compressible turbulence, and continuum
traffic simulation 121. The Burgers’ equation is a useful test equation for inves-
tigating various numerical schemes which are then applied to more complicated
systems of partial differential equations [3-51. In particular, finite difference
models of this equation have been used to study the filtering of nonlinear insta-
bilities [6], discretization error in computational results 171, the behavior of
multiple solutions to finite difference schemes [8,9], the relationship of unique-
ness and multiplicity of solutions of the cell Reynolds numbers [ 101, and new
discretization methods for first and second derivatives [ 1 1 , 121.
In our view, the presentation of exact explicit solutions, to a number of dif-
ference equation models of Burgers’ equation, will provide the investigator. of
a new numerical technique for integrating partial differential equations, a mathe-
matical handle that may be used to understand the often unpredictable numer-
ical results which can occur [ 131.

Numerical Methods for Partial Differential Equations. 2 . 123- I29 (1986:


C John Wiley 61 Son\. Inc. CCC 0749- 159X/86:020123-07804.00
124 MICKENS

It. DIFFUSION-FREE CASE

For this case, Eq. ( 1 ) becomes


11, + UU, = 0 ,
with the initial conditions

wheref(x) is a given function. The exact solution to Eq. (2:)is [2]

We can also obtain a special class of solutions to Eq. (2) by the method of
separation of variables, i.e.. we assume

The substitution of Eq. ( 5 ) into Eq. (2) gives the following equations for h(x)
and g ( f )

dg = ffg-,
rlt
rib
- -- -ff ,
dX

where a is the separation constant. These equations may be easily solved to give

where A , and A2 are arbitrary integration constants. Therefore, the method of


separation of variables gives the solutions

We will now construct two finite difference models of Eq. (2). both of which
reproduce exactly on the “lattice” the solutions of Eq. (8).

-
First, construct a lattice

(x,t ) (xi, t,,) = [ ( A s ) k ,( A t ) n ] ,


BURGER'S EQUATION 125

where Ax and At are, respectively, the space and time steps, and ( k , n) are inte-
gers. Consider the following two difference equations models of Eq. ( 2 )

- u;
U;+l

At
+ u ; * ' ( u ;AX& ' - u : ) = 0 , ( 10)

and

u;-I - u;l
At
+ u ; ( u ; : ;Ax- u ; + ' ) = 0 , (9)

Note that these two equations correspond to explicit and implicit difference
schemes for Eq. (2). We will now demonstrate that both of these equations
have solutions that are exactly equal to Eq. (8) on the lattice, i.e.,

u; = u [ ( A x ) k ,(At)rt] . (12)

The mathematical details will only be given for Eq. ( IOl, since the procedure to
be followed applies in a direct fashion also to Eq. ( 1 I ) .
Assume a separation of variables solution to Eq. (10) of the form

The substitution of Eq. (13) into Eq. (10) gives the following equations that
must be satisfied by C" and Dk

where a is the separation constant. These two first-order difference equations


may be easily solved; the solutions are

where A I and A2 are arbitrary integration constants. Therefore,

is a solution to Eq. (10) and, as stated above, the relationship given by Eq. (12)
is satisfied. In other words, on the lattice, the rational solutions of Eq. (10) are
126 MICKENS

exactly equal to the rational solutions of Eq. (2). Similar remarks can be made
for Eq. ( 1 1).
It should be indicated that the local truncation errors for the difference
schemes given by Eqs. (10) and ( 1 1 ) are of O(A.r) + O ( A t ) .

111. NONLINEAR STEADY-STATE CASE

For this case, Eq. ( 1 ) becomes

which is an ordinary, nonlinear second-order differential equation. An exact


solution to Eq. (19) is

where A and B are arbitrary constants. We take the following finite difference
model for Eq. (19)

where h = Ax. Note that a first “integral” of Eq. (21) is


uk+I - uk = ( Y U t + 11dk +p, (22)

where (Y = ( h j 2 u ) and /3 is an arbitrary constant which we take to be negative


and of the form

p = -(;)A’,

where A is a real arbitrary constant.

Equation (22) can be solved exactly [14] and its solution is

UI, = -A[%],

where C is an arbitrary constant and A is defined to be

1 - (hA/2v)
A =
1 + (hA/2v)’
BURGER’S EQUATION 127

If we now write C as

where B is an arbitrary constant, then Eq. (24) becomes

This is an exact solution to Eq. (21).


Note that in the limits h + 0, k + x , with hk = x = fixed, Eqs. (21) and
(27) go over (converge), respectively, to Eqs. (19) and (20).

IV. LINEAR STEADY-STATE CASE

Equation (19) can be linearized to the form

yu, = ut,, y = constant. (28)

This equation also describes the problem of steady-state one-dimensional


convection and diffusion [12], and the steady-state transfer of heat in a one-
dimensional moving medium, without internal heat sources [ 1I].
The exact solution to Eq. (28) is

u(x) = C,+ CZeY‘,


where C,and C2 are arbitrary constants.
We choose the following finite difference model for Eq. (28)

where h = A x . Since Eq. (30) is a second-order, linear difference equation


with constant coefficients, it can be easily solved; doing this gives

where C,and C2 are arbitrary constants. Thus, the difference scheme given by
Eq. (30) is an exact or “best” [15, 161 difference scheme, i.e.,

uk = ld(hk),

for arbitrary h = Ax and values of y.


128 MICKENS

V. SUMMARY

We have presented a number of finite difference models of Burgers’ and


related equations. All of these difference equations have the property that exact
solutions to them can be obtained. It should be pointed out that these finite dif-
ference schemes do not suffer from problems of convergence or instability that
often occur in the numerical analysis of differential equations [ 171. Further-
more, there are no restrictions that need to be placed on either the space (Ax)or
time (Ar) step sizes.
We are currently investigating the construction of finite difference models of
more complicared partial differential equations.

The author wishes to thank M . D . Wiggins and H. A. Neal for several useful discus-
sions. This reszarch was supported in part by grants from DOE and KASA. and a UNCF
Faculty Fellowship.

REFERENCES
J . M . Burgers, “A Mathematical Model Illustrating the Theory of Turbulence,”
Adv. Appl. Mech., 1, 171-199 (1948).
G. B. Whitham, Linear and Nonlinear Waves, Wiley-Interscience, New York,
1974.
W. Shyy, ”A Study of Finite Difference Approximations to Steady-State, Convec-
tion-Dominated Flow Problems,” J . Comput. Phys., 57, 415-438 ( 1985).
J . Caldwell, R . Saunders. and P. Wanless, “A Note on Variational-Iterative
Schemes Applied to Burgers’ Equation,” J . Comput. Phys., 58. 275-281 (1985).
K. W. Morton and M . J . Baines. Eds.. Numerical Method7 Jiv Fluid Dynamics,
Academic, London, 1982.
P. K. Khosla and S. G. Rubin, “Filtering of Non-linear Instabilities,” J . E n g .
Math., 13, 127-141 (1979).
S . - I . Cheng and G . Shubin, “Computational Accuracy and Mesh Reynolds
Number,” J . Compur. Phy5., 28, 315-326 (1978).
W.-J. Beyn and E. Doedel, “Stability and Multiplicity of Solutions to Discretiza-
tions of Nonlinear Ordinary Differential Equations,” SIAM J . Sci. Stat. Compur.,
2, 107-120 (1981).
A. B. Stephens and G. R . Shubin, “Multiple Solutions and Bifurcation of Finite
Difference Approximations to Some Steady Problems of Fluid Dynamics.” SIAM
J . Sci. Stat. Comput., 2, 404-415 (1981).
R. B. Kellogg, G. R. Shubin, and A. B. Stephens, “Uniquensss and the Ccll
Reynolds Number,” SIAM .I. Numer. Anal., 17, 733-739 (1980:.
D. B. Spalding, “A Novel Finite Difference Formulation for Differential Expres-
sions Involving Both First and Second Derivatives,” Int. J . Num. Math. Eng.. 4.
551-555 (1972).
G . D . Stubley, G. D. Raithby, and A . B . Strong, “Proposal for a New Discrete
Method Based on an Assessment of Discretization Errors,” Numer. Heat Tran.fer.
3, 41 1-428 (1980).
A. R. Mitchell and J . C . Bruch, Jr.. “A Numerical Study of Chaos in a Reaction-
Diffusion Equation,” Numer. Methods Partial D@, E q . , 1. 13-23 (1985).
H. Levy and F. Lessman, Finite D(8erenc.e Equcuions, Macmillan, New York,
1961. p. 136.
BURGER’S EQUATION 129

[ 151 R . E. Mickens, “Difference Equation Models of Differential Equations having


Zero Local Truncation Errors,” in Differenrid Equariuns, I. W. Knowles and
R . T. Lewis, Eds., North-Holland. Amsterdam, 1984, pp. 445-449.
[ 161 R. B. Potts, “Best Difference Equation Approximation to Duffing’s Equation,”
J. Ausrral. Math. SOC. Ser. B. 23, 349-356 (1982).
[ 171 F. B , Hildebrand. Finite-D$ference Equations arid Sitnulations, Prentice-Hall.
Englewood Cliffs, NJ, 1968.

You might also like