93
C H A P T E R 5
First Integrals
In an introductory course in PDEs, the wave equation
u
tt
D c
2
u
xx
; (5.1)
where c > 0 is a constant wave speed, is introduced. e general solution of (5.1) is
u D F .x ct/ C G.x Cct /; (5.2)
where F and G are arbitrary functions of their arguments. In this solution, there are two waves;
one is traveling right (F .x ct/) and one is traveling left (G.x C ct /). Each of these solutions
can be obtained from the following first-order PDEs
u
t
C cu
x
D 0; (5.3a)
u
t
cu
x
D 0: (5.3b)
PDE (5.3a) gives rise to u D F .x ct /, whereas PDE (5.3b) gives rise to u D G.x C ct/.
We question whether it is possible to show this directly—whether the solutions of (5.3a)
and/or (5.3b) will give rise to solutions of (5.1). For example, differentiating (5.3a) with re-
spect to t and x gives
u
t t
C cu
tx
D 0;
u
tx
C cu
xx
D 0:
(5.4)
It is a simple matter to show that upon elimination of u
tx
in (5.4) we obtain (5.1). e same
follows from (5.3b). In this example, we considered the linear wave equation, where the solution
of a lower-order PDE gave rise to solutions of a higher order PDE. Does this idea apply for
NLPDEs? e following example illustrates the concept.
Consider the following pair of PDEs:
u
x
u
y
D 1; (5.5a)
u
xx
u
4
y
u
yy
D 0: (5.5b)
We ask, will solutions of (5.5a) give rise to solutions of (5.5b)? For example, one exact solution
of (5.5a) is u D 2
p
xy. Calculating all necessary derivatives gives
u
x
D
p
y
p
x
; u
y
D
p
x
p
y
; u
xx
D
p
y
2x
p
x
; u
yy
D
p
x
2y
p
y
; (5.6)