1 Eutiquio C Young Partial Differential Equations An Introduction Allyn and Bacon 1972 (240 357) (001 020)

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

.

230 Tlie Heat Equation Chap. 6

We first consider the problem consisting of the nonhomogeneous heat


equation

(5.I) u, - ku'"' = F(x. I) (0 < x < L, t > 0)

together with the initial condition (4.2) and boundary conditions (4.3). We
assume a series solution or this problem in the form

. _ ,, . ll71X
(5.2) 11(x. I) - w 11,,(t) c,111-
" I L

where the coefficients 11., are functions or I. This is actually the eigenfunction
expansion of the yet unknown function 11. in terms of the eigenfunctions (4.8)
of the associated Sturm-Liouville problem (4.6). Thus, the coeftlcients 1111 must
be related to the function u by the formula

(5.3) 11 11 (1 ) = 2 JL u(x, t) s111-~-


. 1171X Jx (n = I, 2, ... )
L 0 L
Now let us a~sume that 11 is twice continuously differentiable for 0 s x s L,
t ;:::.: 0, and that the function Fin (5.1) can be represented in Fourier sine series

(5.4) f(x, t) = L" F 11 (f) sin !25..:.


11=1 L

so that

') J'L ll71X


(5.5) F.,(t) = :. F(x, t) sin - dx (11 = 1, 2, ... )
L 0 L

Then, by differentiating (5.3) with respect tot and using (5.1), we obtain

(5.6) 11;,(1) = ·L2 JL


11, sin
.
L
lliTX
dx
0

= 2k
-L J'L Un
. lliTX
Sin - - dx + F,,(t)
.. L
0

where F 11 (t) is given in (5.5). By using the boundary conditions (4.3), the first
term on the right of equation (5.6) can be integrated by parts twice, to give

2/.: r.L Uu
. 1171X d
sm ·-- x = - k. ( )
·1. 1µ., t
L •O L
where }. 11 1s as given 111 (4.7). Thus, (5.6) leads to the ordinary differential
equation

(5.7) 11:,(t) + k).,,u,.(t) = F,.(t)


Sec. 5 Nonhomogeneous lnitial-Bo1111dary Value Problems 231

for the functions 11", n = I, 2, .... The initial condition (4.2) implies that

(5.8) 11,,(0) =
2 JL f(x)sin
. . l/7L'(
dx = b,, (n = I, 2, .. )
L 0 L

Hence. by solving the initial value problem (5.7), (5.8), we find

(5.9)

for /1 = 1, 2, . . . . Substituting this in (5.2), we thus obtain

(5.10) u(x, t) = f: JJ'


l
11=! O
e-U,,(1-rlF,,(T) ch ( sin
(
nn~ +
L
f
n=I
b,,e-u .. i sin nnx
L

for a solution of the problem ( 5.1 ), ( 4.2), (4.3 ).


To verify that (5.10) satisfies (5.1 ). (4.2), and (4.3), we need show only that the
first term on the right of(5. IO) satisfies (5.1), (4.3), and vanishes at t = 0 because
the second term is just the solution (4.11) of the homogeneous initial-boundary
value problem (4.1), (4.2). (4.3). Thus, set

~ . nnx
(5.11) i:(x, t) = £... u,,(t) sm -
11=1 L
with

(5.12) v,,(t) = [' e-k;.,,u-rlF,,(T) ciT (n ~ 1)


• 0

If we assume that F is continuous for 0 s x s L, t ~ 0, then for any fixed


T we see from (5.5) that
2 rL
IF,,(1)1 s I. \F(x, t)i dx s 2M
L •O

where M denotes the maximum value of IFI in the region 0 s x s L, 0 s


t s T. Hence, from (5.12). we have

Since the series of constants 2M /kl.,, converges, it follows by Weierstrass M-test


that the series (5.11) converges uniformly in the region 0 s x s L, 0 s t :::;: T
for arbitrary T. This implies that v is a continuous function for 0 s x s L,
t ~ 0, and thus satisfies (4.3) and vanishes at t = 0 because each of the terms
of the series does. Under more stringent conditions on F [say, F, Fx, and Fx:x:
are continuous for 0 s x s L, t ~ 0, and F(O, t) = F(L, t) = OJ, we can
232 The Heat Equation Citap. 6

verify that the series (5.11) can be differentiated term by term. Then, by (5. 7)
and (5.4), it follows that
00
nnx
vt - kv.n = L [v;,(t) + knv(t)] sin· ··
11=1 L
00
nnx
= L F,,(t)

SJl1.
11=1 L

= F(x, t)
so that v satisfies (5.1 ).
If we substitute (5.5) for F,, in (5.11) and formally interchange the order of
summation and integration, we can write (5.11) in the integral form

(5.13) v(x, t) = LJ: G(x, t - r; ~)F(~, r) di; dr


where G(x, t; !;) is the Green's function given in (4.16). In this form, it can be
verified that (5.13) gives a solution of the problem (5.1), (4.3) with homogeneous
initial conditions under the less restrictive conditions that F and Fx are con-
tinuous and F(O, t) = F(L, t) = 0. Indeed, from the discussion in Section 4,
we see that the inner integral in (5.13) is a solution of the problem

U1 - kuxx = 0 (0 < x < L, 0 < T < t)


(5.14) u(x, O; r) = F(x, r) (r > 0)
u(O, t; r) = u(L, t; r) = 0 (0 < T :S: t)

Our assertion then follows by the Duhamel's principle (Problem 10, Exercises
6.4). The uniqueness of our solution follows by Theorem 3.2.

Example 5.1. Find the solution of the problem

u, - llxx = (sin x (0 < x < n:, ( > 0)


u(x, 0) = 0 (0 s x s n:)
u(O, r) = u(n:, t) = 0 (! 2: 0)
Solution: By (5.5) we find

Fn(t) = -2r
n:
in . .
0
Sll1 x 5111 nx dx = {ot (11 = 1)
(11 > I)

Hence, by (5.11 ), the solution is

u(x, !) = (J: e-«-t>r dr) sin x

= (e-' + t - 1) sin x
Sec. 5 Nonlto111oge11eo11s /11itiul-Bo11nd11ry Value Problems 233

Example 5.2. Find the solution of the problem


11 1 - llxx = F(x, 1) (0 < x < rr, I > 0)
u(x. 0) = 0 (0 :S _\- :S rr)
11(0, I)= u(n, I)= 0 (I 2: OJ
where F(x, I) = x when 0 :S x :S rr/2, and F(x. I) = rr - x when rr/2 :S x :S rr_
Solution: The coefficients in the Fourier sine series expansion of Fare given by

F,,(t) = 2 Jn/2 x sm
_ nx dx + 2 J" (rr - x) sin nx dx
Tr 0 n rr/2

4 . /17[

2
sm (11 = J, 2, ... )
nn 2
or
4( - 1)11 - 1
F2n-1 (I) = - -- (n = I, 2, __ )
rr(211 - 1)2
(See Example 9.1, Chapter 3.) Hence, by (5.12),

V 2 n_ 1 (1) = -- -
4( - l)n-1
- -
J' exp[-(211 - I)2(r - T)] dT
rr(211 - 1)2 o

= 4(--_I)"_~_ [I _ e-<211-1)11]
n(2n - I )4

and therefore, by (5.11), the solution is


4 oo (-1)11-I
u(x, t) = -- E - ~- [I - e-< 2 n- 011 ] sin(211 - l)x
7r n=I (211 - 1) 4

Next we consider the initial-boundary value problem

(0 < x < L, t > 0)


(5.15) u(x, 0) = 0 (0 < x < L)
u(O, t) = a(t), ux(L, t) = b(t) (t > 0)

which involves nonhomogeneous boundary conditions. This problem can, of


course, be reduced to a problem of the kind considered above, in which the
differential equation is nonhomogeneous and the boundary conditions are
homogeneous. Jn fact, if we introduce the new dependent variable w, defined
by w = u - v, where v(x, t) = a(t) + xb(t), then the problem (5.15) becomes

w, - kw xx = - [a'(t) + xb'(t)]
(5.16) w(x, 0) = -a(O) - xb(O)
w(O, t) = 0, wx(L, t) = 0

which can be treated by the method described above.


234 The Heat Equation Chap. 6

It is possible, however, to apply the method directly to the problem (5.15)


without reducing it to (5.16). That is. we assume a solution of (5.15) in the
form of a series involving the eigenfunctions of the Sturm-Liouville problem

X" +IX= 0, X(O) = 0, X'(L) = 0

which arises if the method of separation of variables is applied to the problem


when the boundary conditions are homogeneous. This problem has the
eigenvalues
2
). = (211 - 1 rr)
" 2 L

with the corresponding eigenfunctions

X (x
11
. (211
. ) = s111 - --
2
1) nL..~ (11 = 0, 1, 2, ... )

Therefore, we assume a solution u in the form

(5.17) u(x, t) = Lro .


u 11 (t) s111 (2n - 1) rrx-
- ·· -
n=I 2 L
with

(5.18) u,,(t) = 2- JL u(x, I) sin ('-) 11 - 1) rrx


- dx (11 = 1,2, ... )
L 0 2 . L
Assuming that u is twice continuously differentiable for 0 c:;; x c:;; L, t :;::: 0, we
obtain, by differentiating (5.18) with respect to t and using the differential
equation,

u;,(t) = - 2 JL u,(x, t) sin (211·· ')- -1) rrx- dx


L 0 - . L
2k
= -·
L
JL uxx(x, t) . (211
.
Sll1
-
2
-
1) nxL dx
-
0

Using the boundary conditions in (5.15), this can be integrated by parts twice
to give

(5.19) u;,(t) + kl.,,u 11 (t) = g,,(t)


2 2
where ). 11 = [(211 - l)rr] /4L and

2k -
g,,(t) = -z_- [,/;,,, a(t) + (-1)'·- 1 b(t)] (11 = 1, 2, ... )

From the initial condition in (5.15) we see that u,,(O) = 0. Hence, by solving
the differential equation (5.19) with the initial condition u,,(O) = 0, we find

(5.20) 11,,(t) = { exp(-k).,,(t - r))g,,(r) dr (11 = 1, 2, ... )


Sec. 5 So11lro111oge11eous lnitial-Bormdary Value Problems 235

Substituting 1his in (5.17), we thus obtain

(5.21) u(x, t) = ~ ~· f'exp(-kX 11 (t - c))g,/1) (h1) sin( 211 2- 1


) nx
11= I l.· () L
The verification that this gives a solution of the problem (5.15) is not so straight-
forward as in problems we encountered before. We shall illustrate the situation
in this case by means of the following example.

Example 5.3. Find and verify the solution of the problem

(0 < x < n, t > OJ


u(x, 0) = 0 (0 < x < rr)
11(0, 1) = 0, uJn. I) = I (t > 0)
11 1
So/111io11: Here we have g,,(t) = (2/rrl(-1 ) - so that by the formula (5.21) our
solution is

u(x, I)=~ 1 ~
1 (-1)"-
1
(L exp(--~ 211
-
2

~) (1= r))dr} sin(~i 1


) x

8 I:.,_ (- I)"- 1
I- e· 211
< -
-
1
>'t1 4
·- sm -
. (211~- - I) x
7f n= I (211 - l )2 . 2
To verify that this gives a solution of the problem, we observe that the function
f(x) = x has the Fourier series representation

x =
8 I: oo ( - 1--)"-I sm
. (211 - 1) x (0 < x < 7!)
7! 11=1 (211 - 1) 2 2
in terms of the eigenfunctions X 11 (x) = sin [(211 - 1l/2 ]x, n = I, 2, . . . . Hence,
we can write our solution above in the form

u(x, I) = x -
8
-
00

I: (-1)"-
-
1
? exp
(-(211
- - - 1)
2
1) sm. (211 - 1) x
7T 11= I (211 - 1)- 4 2

The series on the right converges uniformly in the region 0 :S x :S n, t ;::: 0. so


that 11 is a continuous function there. Further, for 0 < x < 11 and t > 0, the
function u has continuous derivatives of all orders with respect to x and t, which
can be obtained by termwise differentiation. Moreover, it is clear that 11(x, 0) = 0,
u(O, t) = 0, and u,(n, t) = 1. By a straightforward calculation it is readily
seen that u satisfies the differential equation for 0 s x :S n, t > 0. This com-
pletes the verification that 11 is a solution of the given problem.

Exercises 6.4

In Problems 1 through 5, reduce the problem to one involving homogeneous boundary


conditions and then find the solution.
l. u1 - kuxx = 0, 0 < x < L, t > O; u(x, 0) = 0, u(O, I)= u(L, t) =A, A a constant.
2. 11 1 - ku_u = 0, 0 < x < L, t > O; u(x, 0) = 0, ux(O, t) = A, u(L, t) = 0, A a
constant.
236 Tlze Heat Equation Clzap. 6

3. 11, -ku'"' = 0. 0 < x < L, t > O; u(x, 0) = 0, 11(0, t) = sin UJt, u(L, t) = 0, (J) a
constant.
4. u, - kuxx = 0, 0 < x < L, t > O; u(x, 0) = 0, 11(0, t) = 0, uJL, t) = cos UJt.
5. u, - uu + hu = 0, lz a positive constant, 0 < x < rr, t > 0; u(x, 0) = 0, u(O, t) =
0, 11(71, t) = J.
6. Solve the problem
u, - ku'"' = e-x (0 < x < n, t > 0)
u(x, 0) = f(x) (0 ~ x ~ n)

11(0, t) = 0, u«n, t) = 0

(See Problem 7, Exercises 6.1.)


7. Solve the problem
11 1 - kuxx = A cos UJt (0 < x < n, I > 0)
11(x, 0) = f(x) (0 ~ x ~ n)
(A a constant)

(Observe that any function oft satisfies the boundary conditions.)


8. Find the solution of the problem

u, - kuxx = g(x, 1) (0 < x < L, I > 0)


u(x, 0) = 0 (0 < x < L)
u(O, I) = 0, u(L, t) = h(r) (I > 0)

9. Find the solution of the problem

11 1 - kuu = g(x, I) (0 < x < L, f > 0)


u«O, I) = h(I), ux(L, 1) = 0 (t > 0)
u(x, 0) = 0 (0 < x < L)

10. (Duhamel's Principle) Let 11(x, t, T) be the solution of problem (5.14) and set
1 (x, I) == s~ u(x. I - T, T) dT. Show that v agrees with (5.13) and that it satisfies
(5.1), (4.3), and the initial condition r(x, 0) = 0.

6. The Initial Value Problem

We now consider the problem of heat flow in an infinite rod without heat
source, whose lateral surface is thermally insulated. This problem leads to the
initial value problem
(6.1) u, - kuu = 0 ( - Cl) < x < Cl), t > 0)
(6.2) u(x. 0) = f (x) (-oo<x<oo)
for the heat equation. We shall solve this problem by means of the Fourier
transform presented in Chapter 5.
We suppose that the problem (6.1), (6.2) has a solution 11, which has the
property that 11, 11,, ux, and uu are all piecewise smooth in x and t, and absolutely
Sec. 6 The Initial Value Problem 237

integrable in x for - oo < x < oo. Then, according to Theorem 8.1 of Chapter
5, the Fourier transforms of all these functions exist; in particular, we have

(6.3) U(s, t) = J~n J:00


u(x, t)e-i.sx dx

and

(6.4) u(x, t) = /
v 2n
- J00

-oc
U(s l)ei.sx ds
. '

Thus, if we can find the transform (6.3), then the solution of the problem will
be given by the formula (6.4).
To determine the function U, let us differentiate (6.3) with respect to t and
use the differential equation (6.1) to obtain
au J~
00 u,(x, t)e-isx dx
1
- (s, t) = - -,=
ot J2n
(6.5)
k
,/2;
J: 00
uxix, t)e-isx dx

Note that because 11, and 11.u are absolutely integrable on - oo < x < oo for
all s and t > 0, differentiation with respect to t under the integral sign is valid.
Integrating by parts twice the last integral above, we have
au k . . 1 oc
---;;-· (s, t) = -r· (uxe-•sx + isue-•sx):
ot --.J2n i-oo

- ---=
ks2 Joo u(x, . dx
t)e-•sx
J2n -oo
If we further assume that u and u, vanish as lxl --> oo, then equation (6.5)
reduces to
au
-~ - (s, t) = -ks 2 U(s, t)
(6.6)
ct
where we have used (6.3).
We now require for this discussion that the function fin (6.2) be piecewise
smooth and absolutely integrable on ( - oo, oo) so that its Fourier transform
F(s) exists. Then, from (6.3) and (6.2), we have

U(s, 0) = ~l~n J: 00
u(x, O)e-isx dx

(6.7)
,-- J_"'oo J(x)e-isx
'\/ 2n
= F(s)
238 The Heat Equation Chap. 6

Thus. as a function of the variable t, the Fourier tramform of the solution u


satisfies the first-order differential equation (6.6) and the initial condition (6.7).
In this way we have therefore transformed the problem (6. I), ( 6.2) for the func-
tion 11 into the simpler problem (6.6), (6.7) for the Fourier transform U o[ u.
The solution or the differential equation (6.6), which satisfies the initial con-
dition (6.7). is given by
(6.8) U(s, t) = F(s)e-" 21
Consequently. by (6.4) and (6.7), we have

u(x, t) = 11·.·
, 2rr
I"'
"- oc
F(s) exp( - ks 2 t + isx) ds
(6.9)

Now the double integral corresponding to the iterated integral in (6.9) is


absolutely convergent because both f(x) and e-ks 21 are absolutely integrable.
Therefore, \Ve can interchange the order of integration in (6.9) and write

(6.10) u(x, 1) = J~ G(x - (. t)fm d(


00

where

(6.11) G(x, t) = 1· II°" exp(isx - ks 2 t) ds


2rr • _"'
The integral on the right of (6. I I) can be evaluated as follows: By using the
Euler formula
eix = cos x + i sin x
(6. I I) can be written as

(6.12) G(x. t) = II"' e-ks''(cos sx + i sin sx) ds


2rr .. - oc
2 2
where the functions e-ks ' cos x and e-ks ' sin sx are both absolutely integrable
on - oo < s < oo, uniformly in x and t for t > 0. Since e-ks'r cos sx is even
ins and e-ks 21 sin sx is odd. it follows that
oc

so that (6.12) becomes


J - 00
e-ksir sin sx ds = 0

G(x, t) I
2rr
J"'_"' e - ks'r cos sx ds
1 Joo e -ks't cos sx cs/
J[ 0
Sec. 6 The Initial Value Problem 239

Jntroducing the new variable z = s, k1, we obtain

By Problem 10, Exercises 1.6, we find

,_;rr e-x'/4kt

so that

G(x, I) =
1
f"' e-k>'r cos sx ds
rr o
(6.13)

Substituting this in (6.10), we finally obtain

(6.14) u(x, t) = - 11'---- J'"' (-(x _ ()2)·· f(c;). de;"


exp ----
2vrrkt -x • 4kt

for a solution of the initial value problem (6.1), (6.2).


The function

(6.15) G(x - (, t) = I
1
..
(-(x -
exp - - c) ·
2
)
(t > 0)
2,i rrkt 4/.:t

is called the fundamental solution or the Green's function for the heat equation
in the infinite domain - oo < x < oo, t > 0. Physically, the Green's function
(6.15) represents the temperature at a point x at time I due to a concentrated
heat source at a point ¢ at time zero.
It is easily verified from (6.15) that (i) G has continuous partial derivatives
of all orders with respect to x and t for - oo < x < oo, I > 0, and satisfies the
heat equation for all x and I > 0, provided x # ( (see Problem 2, Exercises
6.1); (ii) G is continuous at x = (; (iii) G vanishes as !xi -> oo for t > 0; (iv)
G has the property

f~0 G(x - (, t) d( = 1

and, finally, (v) G is symmetric with respect to x and (. These properties are
analogous to those of a Green's function for an ordinary differential equation.
Now, to verify that (6.14) actually gives a solution of the problem (6.1),
(6.2), it is enough to require that f be continuous and bounded on ( - oo, oo ).
Then, for arbitrary constants L > 0 and t 0 > 0, the integral (6.14) together
with its partial derivatives of any order obtained by differentiating under the
~: ..
240 The Heat Equation Chap. 6

integral sign with respect to x and t converges uniformly in x and t for - L s;


x s; L, t 0 s; t. Therefore, the function 11 and its derivatives of all orders exist
for - oo < x < oo, t > 0, and these derivatives can be found by differentiating
under the integral sign. Since C(x - (, t) satisfies equation (6.1) whenever
x # (, it follows that u does also.
Finally, set z = (~ - x)/(4kt) 112 in (6.14) to obtain

u(x, t) = - l_
Jn
J"'
-oo
f(x + J4kt z)e-= 2
dz

Since f is bounded for all values of its argument, the preceding integral con-
verges uniformly with respect to x and t for - oo < x < oo, t ;:::: 0. Hence,

Jim u(x, t) = - :
t->O"
1
'Ii1n
J 00

-oo
f(x)e-= dz
2

= f(x)
which verifies (6.2). From the boundedness off it follows that u is also bounded.
If we define u(x, 0) = f (x), then u is bounded and continuous for - oo <
x < oo, t ;:::: 0, so that by Theorem 3.4, u is uniquely determined.

Example 6.1. Find the solution of the problem (6.1 ), (6.2) if the function f is given by

/(x) = {~ (x < 0)
(x > 0)

Solution: According to the formula (6.14), the solution is

1
u(x, t) = 2vnkt
{"'
Jo
exp --
(-(x4ki-
- ~) )
2

d<;

Jn terms of the error function,


erf(r) = }
v; Jo
r e-zl dz

which has the property that erf( ro) = 1, this solution can be written as

u(x, t) = ·-·1 + -1 erf ( -_--_


x )
2 2 v4kt
Then, by the Leibnitz rule, it is easily verified that this satisfies the heat equation
(6.1). Also, as t-> o+, u(x, t) _, 1 if x > 0, and u(x, t) _, 0 if x < 0. At
x = 0, we have u(O, t) = I /2, which is in accordance with the theory of Fourier
integral transform, since f has a jump discontinuity at that point.

In the case of the initial value problem


U1 - kuxx = h(x, t) ( - 00 < x < 00' t > 0)
(6.16)
u(x, 0) = 0 (-oo<x<oo)
Sec. 6 The Initial Value Problem 241

which involves the nonhomogeneous heat equation, the method described above
leads to the differential equation

(6.17) U, + ks 2 U = H(s, t)
for the Fourier transform (6.3) of the solution 11. Here.His the Fourier trans-
form of the function h; that is.

(6.18) H(s, I) = -
I
If z h(x, t)e- 1" ' dx
.

, 2n • - x

From (6.3) and the initial condition in (6.16). we see that U(s, 0) = 0. By
solving the differential equation (6.17) with the initial condition U(s, 0) = 0,
we thus find

(6.19) U(s, t) = J~ e-' 1' 11 -r


1
H(s, r) dr

Therefore, according to (6.4). the solution of the problem (6.16) is given by

(6.20) u(x, t) = Tl Jx e"-'. f' e-•s'<t-r H(s, r) dr ds 1

v 2n - oc o
By substituting (6.18) for Hand formally interchanging the order of integrations,
this solution can be written as

(6.21) u(x, t)
J' Jx
0 - ""
G(x - (, t - r)h(~, r) d( dr

where, by (6.13),

G(x - ~, t - r) = ··I
2n
J"'_ oc
exp(is(x - () - ks 2 (t - r)) ds

exp(-(x - ¢) 2 /4k(t - r))


I
2, 1 nk(t - r)
is the Green's function for the problem.
To verify that ( 6.2 J) gives the solution of (6.16), we assume that his continuous
and bounded for all x and t ~ 0. Then, for 0 < r < t, we know from the
previous case that the function

w(x, t; r) = f"' G(x (,I - r)h(¢, r) di;


. - "'
is the solution of the initial value problem
H' 1 - /.;11·.u = 0 ( - oo < x < oo, t > r)
1r(x, r: r) = h(x, r)

Hence, by Duhamel's principle (Problem 8, Exercises 6.5), it follows that (6.21)


is the solution of the problem (6.16).
242 The /feat Equation ('hap. 6

Exercises 6.5

I. \\'rite the solution or the problem

lixx 0 (-oc<x<x,l>OJ
(' · ixl
11(x. OJ (-X < X < X)

2. So\\e the problem


0 ( -· X < x < X. I > 0)

11(.r. 0)
{1 ix1 :s: L)
lo ( x > /_)

and express the solution in terms of the error function

erf x = -') f'X !'- '


2
ds
\ 7l .. 0

3. As in Problem 2, find the solution of the problem

11, - kuu 0 ( - 00 < x < CIJ' I > 0)

11(x, OJ
{A (x < 0)
\B (x > 0)
4. Show that if the function fin (6.2) is odd (or even), then the solution (6. 14) is
also odd (or even) with respect to the \ariable x. Thus, deduce that 11(0, I) = O
(or 11x(O. I) = 0).
5. \'erify the result in Problem 4 \1 ithout resorting to the formula (6. 14) and using
only the fact that a solution of the problem (6. 1), (6.2) is uniquely determined.
6. Find the solution of the problem
( - x < x < x. I > 0)
11(x. 0) 0 (-w<x<x)

and sho11 that (6.21) yields 11(1. 1) ~ t 2 '2.


7. Find the solution of the problem
( - oc < x < :::r:. I > 0)
11(x. 0) 0 (-x<x<oc)

and show that (6.21) ~ields 11(x. r) = 1 -- !'-'.

8. (Duhamel':, Principle) Let 11(.1', I, T) be the solution or the initial value problem
// 1 - kun 0 ( - CIJ < x < x, I > 0)

u(x. 0, r) f(x. r) (- x < x < x. r > 0)


and set

r(x, t) = f~ 11(x, I - r, r) dr

Pro\e that z· is the solution of the problem

r, - kt'.n = f (x, t) (- x < x < x, t > 0)


l'(x. 0) = 0 (-oo < x < oo)
Sec. 7 lnitial-Bo1111dary Value Problems in Infinite Domain 243

9. Find the solution of the problem

(h = const; 1 > 0)
1·(.\", OJ= /Ix) (-:x:· < x < :X:)

and shO\\ that z·(x. 1) = 11(.1·. t)e- 111 , where 11 is given by (6.14).
1
IO. By the sub'.;titution 11(\·. I) = r(x, I )e ", show that Problem 9 can be reduced
to the problem (6. I). (6.2).
11. By using the result of Problem 9 and the corresponding Duhamel's principle for
the differemial equation, obtain the solution of the initial value problem

11 1 - k11_u + '111 = /(x, t) ( - oc < X < 02, t > 0)


u(x. 0) = 0
( - oc < x < oc ; h = canst > 0)

7. Inilial-Boundary Value Problems in Infinite Domain

In this last section of the chapter we shall consider initial-boundary value


problems for the heat equation in the domain 0 < x < oo, t > 0. The solutions
of our problems will be obtained by means of Fourier sine or cosine transform,
as this becomes appropriate for this type of problem. We first consider the
problem

(7.1) u, - ku'-' = 0 (x > 0, t > 0)

(7.2) 11(x, 0) = f (x) (x ?: 0)

(7.3) 11(0, t) = 0 (t ?: 0)
which corresponds to the problem of heat flow in a semi-infinite rod x > 0,
with initial temperature distribution described by the function f(x) for x ?: 0,
and whose end point x = 0 is maintained at zero temperature. To solve this
problem. \\e shall use the Fourier sine transform. This choice is based on the
fact that the singular eigenvalue problem that is related to the problem at hand is

X" + l.X = 0, X(O) = 0 (X bounded as x--> oo)


which has the eigenfunctions XJx) = sin sx corresponding to the eigenvalues
2
; = s , s > 0. Accordingly, we assume that the problem (7.1), (7.2), (7.3)
has a solution u, which has the property that 11 and its partial derivatives up to
the second order are all piecewise smooth and absolutely integrable on 0 <
x < oo for t > 0. Then the Fourier sine transform of 11,

(7.4) U,(s, t) = (n,2) 1/2 f"'


o
.
u(x, t) sm sx dx

exists and

(7.5) u(x, t)
2')1,2
-
Joc UJs, t) sin sx ds
( 7[, 0
244 The Hear Equation Chap. 6

Here, Us does not mean the partial derivative of U with respect to s. Following
the procedure described in the preceding section, we proceed to find the trans-
form Us or our solution. Once this function is found, then our solution is given
by (7.5).
By differentiating (7.4) with respect to t under the integral sign and using the
differential equation (7.1 ), we have

au s (s,
ct
t) = (2):" 1 2
1 J"' u,(x, t) sin sx dx
0

112
k (.;) {" u,.Jx, t) sin sx dx

Arter integration by parts twice, this gives

cU
- -

ct
5
(S, t) k ; (2) 1/2
(llx sin sx su cos sx) '
io
(7.6)
ks 2 (n,2)
-
1/2 J"'
o
u(x, t) sin sx dx

ff we assume as before that u and ux vanish as x --+ oo, then because or the
boundary condition (7.3), the boundary terms on the right of (7.6) vanish,
and by (7.4) we are led to the differential equation

(7.7) cl/
- · (s,5 .
t) + /.:.s 2 U,(s, t)· = 0
.

ct
for the function Us. Let us assume that the function f in (7 .2) is piecewise
smooth and absolutely integrable on (0, oo ). Then, from the initial condition
(7.2) and (7.4), we find

U,(s, 0) = (;
2)1;2 •oc
.t u(x, 0) sin sx dx

(7.8) 2)1/2 J"' j(x) sin sx dx


(-
n. o
= F,(s)
which is the Fourier sine transform of the function f. The solution of equation
(7.7) that satisfies the initial condition (7.8) is

(7.9) U,(s, t) = F 5 (s)e-k 521

Hence, by (7.5) and (7.8), we obtain

u(x, t) = (-2)1/2Jx F (s)e-ks-i, sin sx ds


Ji. 0
5

(7.10)
=
n •o •o
.3 c r"·
f(~)e-ks't sins~ sin sx d~ ds

This gives the solution of the problem (7.1), (7.2), (7.3).


Sec. 7 lflitial-Bo1111dary Value Problems ;,, lfljiflite Domain 245

If we replace the function sins¢ sin sx in the integrand of the integral in


(7.10) by
}[cos s(x - ~) - cos s(x + ¢)]

and formally interchange the order of integration. we can write (7. l 0) in the
form

(7.11) u(x, n= 1
. --
2, nf\1
J, Ir
o
exp(. -(x -
.
4k1
¢)2) - exp(-(x + ¢) I /{¢) d¢
4k1
2
)
.

in which we have used the resu It (6.13 ). In this form, it can be verifled that u
is the solution of our problem under the weaker condition that f is continuous
and bounded for x ~ 0 with f(O) = 0.
We remark that the solution (7.1 I) of our problem (7.1), (7.2), (7.3) can also
be obtained by extending the initial datum fin (7.2) as an odd function for
- oo < x < oo and then solving the resulting initial value problem by the
formula (6.14). [See Problem 4, Exercises 6.5.J Using the fact thatJ(-x) =
-f(x), the solution as given by (6.14) then reduces to (7.11). Observe that this
is precisely the method we employed in Section 6 of Chapter 3 for finding the
solution of an initial-boundary value problem for the wave equation. The
method works because both the wave and the heat equations remain unchanged
in form when the variable x is replaced by - x.
In the same manner, we can find the solution of the problem (7.1 ), (7.2)
with the boundary condition
(7.12) u)O, t) = 0 (t ~ 0)

by extending the function fas an even function for - w < x < oo. In this
case, the solution is given by

(7.13) u(x, t) = _l_ f"' lexp(_-_(x - 02) + exp(:-(x + ¢)2)· l f(() d¢


2,' rrkt 0 l 4kt . 4kt J
We note here that to solve the problem (7. l ), (7 .2), (7 .12) by the method of
Fourier transform, it will be necessary to employ the Fourier cosine transform
of the solution u. This is also the case even in the more difficult problem where
the boundary condition (7. l 2) is nonhornogeneous. We illustrate the pro-
cedure in connection with the problem
u, - ku'-' = 0 (x > 0, I > 0)
(7.14) u(x, 0) = 0 (x ~ 0)
uAO, t) = h(t) (1 ~ 0)

As usual, we suppose this problem has a solution that is piecewise smooth


and absolutely integrable on (0, oo) including its first- and second-order
derivatives. We consider the Fourier cosine transform of u,

uces, l) '')) J/2 J'f) u(x, t) cos sx dx


(7.15) = ( ·~
/., 0
..,...,_
246 The Heat Eq11atio11 Chap. 6

Differentiating this with respect to t and proceeding as in previous cases. we


have

au,
- - (s. I )
ct (-re2)1/2 f"' u,(x, 1) cos sx d.x
0

k (.::
7)1;2 I.("° t) COS dx
Uu(X, SX
.n. •. o
2) l !2 Ioc
k - ('IT.
(u, cos sx + su sin sx)'
'0

112
- ks
2
(~) {''' u(x, t) cos sx dx

Under the assumption that u and llx vanish as x --+ oo, and taking note of the
boundary condition in (7.14), this leads to the nonhomogeneous differential
equation

(7.16)
cu
~- c (s, t) + ks 2 uces, t) = -k (2)1/2
- h(t)
ct n

The solution of this equation that, by the initial condition in (7.14), satisfies
the condition U,(s, 0) = 0 is given by

(7 .17) UJs, 1) = -k (D 112 J~ e-ks"(t-T)h(T)<h


Hence, by taking the inverse transform, we obtain

u(x, t) =
2) 1/2 j'"' uces, 1) cos sx ds
(-
'IT 0
(7.18)
- 21·
__'.: f"" f' e-ks'u-' 1h(T) cos sx dT ds
'IT 0 0

1f we formally interchange the order of integration and use the result in (6.13),
this formula simplifies to the form

(") l/l j'I


2
exp(-x /41,\(t - T))ii(T) dT
(7.19) u(x. I) = - - - , -
'IT. 0 '\•l-T

This can be shown to provide the solution or the problem (7.14) under the
assumption that h is differentiable for t :2:: 0.
We remark that we cannot verify that (7. J 9) satisfies the boundary condition
of (7.14) by simply differentiating (7.19) under the integral sign and setting
x = 0. In fact, such a formal manipulation would lead to the false result
u)O, t) = 0. We illustrate the situation in this case by means of the following
example.
Sec. 7 /11itial-Bo11ndary Value Problems in Infinite Do111ai11 247

Example 7.l. Find the solt1t1on or Liie problem

11, II_\-.\ 0 (_\ > 0, t > 0)


11L1, OJ 0 (x .2: 0)
11)0. I) ( t .2: 01

Solution: B) the formuL1(7.19)11e have

11(.1, I)~
,., C';fl( -- X 2 /4(t -· T) I lfT
·-
1 T: I
,O \f-·r

On setting z x. \ 411 - rl. this becomes


_\" ,,,'l. e~z2
II(\". I) dz
\ 7[ J,. '41
_2

for 11 hich it is readily see11 that

11).\", /) = -
\ 7I

\ nl

4r
11,(x, I)
\ 7'.'{

for x > 0 and 1 > 0. Thus. 11, - u'-' 0. sho11 ing that 11 satisfies the heat
equation for x 0. I > 0. As I > 0 it is clear that u(x. t) ' 0. Now, to
1erify the boundar1 condition 11,!0. 11 = I. 11c perform an integration by parts
to put 11 in the form
I'"/

,r; ( --- e- :')


x I 2x
1,4,
~:2
11(x. II -- ~ e dz
\ X, '\ 4f \ n

e
.. 41
'
2x Jy e :' d::
\ ,, X,'- 41

Then

2
u)x. I)
\ -
which immediately yields 11)0. r I l, since Jg· e-=' d::

Exercises 6.6

I. (a) Obtain the solution formula (7. 11) from (6.14) by extending the function fin
(7.2) as an odd function for - T < x < T ..
(bl Obtain the solution formula (7.13) of the problem (7.1), (7.2). (7.12) by
extending the function fas an el'en function for - 'XJ < x < C1J.
248 The Heat Lq11ation Chap. 6

2. Obtain the solution !7.13) by using the Fourier cosine transform.


3. Find the solution of the problem

0 (x > 0, t > 0)
11(\", 01 :re-x (x 2: OJ
11(0, /) 0 (t 2: 0)

4. Find the solution of the problem

11, - kuu: = 0 (x > 0. I > 0)


(a > 0, x 2: 0)
(! 2: 0)

5. Solve the problem

11 1 - kuu: = 0 (x > 0, I > 0)


11(x, 0) 0 (x > 0)
11(0, I) = A (A a cons tan!; t > 0)

and express !he solution in terms of the error function. Verify that it gives the
solution of the problem.
6. Find the solution of the problem

11, ku'"' = 0 (x > 0, I > 0)


11(.1. OJ= 0 (x 2: 0)
11(0, I) = h(t) (t 2: 0)

7. Find the solution of the problem

11, - kuu = f(x, I) (x > 0, t > 0)


11(x. OJ 0 (x 2: 0)
11)0, t) lz(t) (I 2: 0)

8. Find the solulion of the problem

11 1 - ku'"' + /111 = 0 (h = const > 0)


11(.\, 0) = f(x) (x 2: 0)
11)0. /) = 0 ( f 2: 0)

9. Obtain the solution of Problem 8 from the resull of Problem 9, Exercises 6.5,
by extending the function fas an even function for - oo < x < oo.
10. Find the solution of the problem

11, - kuu + bu 0
(b const > O; x > O; t > 0)
11(x, 0) 0 (x 2: 0)
11(0, f) h(t) (t 2: 0)
Sec. 7 Initial-Boundary Value Problems in Infinite Domain 249

1 l. Find the solution of the problem

u, - kuu + bu = f(x, t) (h = canst > 0)


u(x, 0) = 0 (x ~ OJ
ux(O, t) = h(t) (t ~ 0)

12. By using the Fourier cosine transform, obtain the solution of the problem

l/ 1 - kuu + fll = 0 (x > 0, t > 0)


u(x, 0) = e-x (x ~ 0)
11x(O, t) = 0 (t ~ 0)
13. Solve the problem
11 1 - kuu = 0 (x > 0, t > 0)
11(x, 0) = f (x) (x ~ 0)
u/O, t) - hu(O, t) = 0 (h = const > O; t ~ 0)

Hint: Let v(x, I)= u/x, I) - h11(x, t). See Problem JO, Exercises 6.1.
14. Solve the problem
11 1 - k11xx = 0 (x > 0, t > 0)
u(x, 0) = 0 (x ~ 0)
ux(O, t) - hu(O, t) = g(t) t ~ OJ
15. Solve the problem
111 - kuxx + bu = 0 (x > 0, t > 0)
11(x, 0) = f(x) (x ~ 0)
llx(O, f) - h11(0, t) = 0 (t ~ 0)
16. Solve the problem
2
11 1 - llxx - - llx = 0 (0 < a < x, t > 0)
x
u(x, 0) = 0 (x ~ a)
11(a, t) = A (A constant; t ~ 0)

Hint: Set x11 = c, :: = x - a and use the result in Problem 5.

"··.

You might also like