CHAPTER 11 Partial Differential Equations: Change
CHAPTER 11 Partial Differential Equations: Change
CHAPTER 11 Partial Differential Equations: Change
Change
The Fourier transform method has been included in the section on Fourier integrals for
heat problems (Sec. 11.6).
161
162 Instructor's Manual
Steps of Solution
1. Setting u = F(x)G(t) gives two ordinary differential equations for F and G.
2. The boundary conditions lead to sine and cosine solutions of the latter.
3. A series of those solutions with coefficients determined from the Fourier series of
the initial conditions gives the final answer.
4. u
0.8
= -;;; ( cos t sin x 1
+ 33 cos 3t sin 3x 1
+ 53 cos St sin 5x + )
6. u = 2kl('TTa - a 2 ) (sin a cos t sin x + f sin 2a cos 2t sin 2x
+ ! sin 3a cos 3t sin 3x + )
4'TT (1 cos 2t sin 2x - 1 1 cos lOt sin lOx - + )
8. u =
5 4 36 cos 6t sin 6x + 100
00
0.04 n'TT
10. u = ~ Bn* sin nx sin nt, B n * = - - sin - 2
'1Tn3
n=l
12. u = cek<x -y)
14. f' lx f = gly 2g
2 = 3k, u =
3
cek<x +Y
3
)
II a
16 -F = - -G = k 2 ' F II - k 2 F = 0' G- + k 2 G = 0' hence
F
u = (c 1ekx + c 2 e-kx)(A cos ky + B sin ky).
Taking the separation constant negative, we obtain a similar result. Taking it zero,
we have
u = (ax + b)(cy + d).
18. u = (A cos kx + B sin kx)(C cos ky + D sin ky)
or
or
u = (ax + b)(cy + d).
20. TEAM PROJECT. (c) From the given initial conditions we obtain
Gn(O) = Bn = -L
2 f f(x) sin -n'TTX dx,
L
o L
_
Gn(O) - A.,,Bn
*+ 2Aw(l - COS W'TT) _
- 0.
n'TT(A..,,2 - w2 )
(e) u(O, t) = 0, w(O, t) = 0, u(L, t) = h(t), w(L, t) = h(t): The simplest w satisfying
these conditions is w(x, t) = xh(t)IL. Then
v(x, 0) = f(x) - .xh(O)IL = f i(x),
v/x, 0) = g(x) - xh' (0)/L = g1 (x),
11
Utt - c 2 Vxx = -xh /L.
Instructor's Manual 163
SECTION 11.4. D' Alembert's Solution of the Wave Equation, page 595
Purpose. To show a simpler methooof solving the wave equation, which, unfortunately,
is not so universal as separation of variables.
Comment on Order of Sections
Section 11.12 on the solution of the wave equation by the Laplace transform may be stud-
ied directly after this section. We have placed that material at the end of this chapter be-
cause some students may not have studied Chap. 5 on the Laplace transform, which is
not a prerequisite for Chap. 11.
Comment on Footnote 2
D' Alembert's Traite de dynamique appeared in 1743 and his solution of the vibrating
string problem in 1747; the latter makes him, together with Daniel Bernoulli (1700-
1782), the founder of the theory of partial differential equations. In 1754 d' Alembert be-
came Secretary of the French Academy of Science and as such the most influential man
of science in France.
20. TEAM PROJECT. (b) Fn = sin (n1TX!L), Gn = an cos (cn 27T 2t/L2). The solution
satisfying the initial conditions is
2 2 2
u = -8L3 (
cos c (77")
- t sin -7TX + -13 cos c (377")
- 37TX
t sin - - + )
77" L L 3 L L
For the string the frequency of the nth normal mode is proportional to n, whereas for
the beam it is proportional to n 2
(c) u(O, t) = 0, u(L, t) = 0, ux(O, t) = 0, ux(L, t) = 0, Hence
F(O) =A+ C = 0, C = -A, F'(O) = /3(B + D) = 0, D = -B.
164 Instructor's Manual
------------
Instructor's Manual 165
6. u = 16
'11"2 ( sin 0.1 'Tl"x e-0.01752.,,. 2 t - 91 sin 0.3'Tl"x e-0.01752(3.,,-) t + - ...
2 )
Uu =L ll'TTX
Bn sin - - e-<cn.,,-/L)2t
L
where
Bn = -
2 L f n'TTX
[f(x) - u1(x)] sin - dx
L o L
2 f L ll'TTX 2
= - f(x) sin - dx +- [(-1ru2 - U1J.
Lo L n'TT
10. u =
2
'TT - ~'TT 9
(cos x e-t
1
+ _!_ cos 3x e- 9 t + - - cos 5x e- 25t + )
25
12. u(x, t) = cos 2x e- 4 t
14. w" = -Ne-=tc2 ; make w(O) = w(L) = 0 to get the function
w(x) = c~a2 [ -e-= - 1 (1 - e-aL)x + J.
1
K'TT ~
16. -L ,,. nBn e-
A 2t
n
n=l
We also show typical applications of the Fourier transform and the Fourier sine trans-
form to the heat equation.
Short Courses. This section can be omitted.
2. A(p) = -
2 1 Jee 2 7T '
cos pv du = - . - e-P = e-P by (15), Sec. 10.8,
1rol+v2 1T 2
and
f
co
2 2
u(x, t) = e-(p+c p t) cos px dp.
0
00
2 2 1 2-2
4. A(p) =
1r(l + p2)
, B(p) = 0, u =-
7T 11 +
O p2
cospx e-c~ t dp
3. We need double Fourier series (easily obtainable from the usual Fourier series) to
get a solution that also satisfies the initial conditions.
,\
2
( c21~2
)'
a2 + b2
(1 1)' 1
2
a )' -2 2a
( a2 + A2 = aa + A2 = O;
(d) The form of the series results as in (b) and the formulas for the coefficients fol-
low from
00
am = 0 because the initial deflection is zero. From (15) and (A), with g(r) = 1 and
amr = s, we obtain
1 '
bm =
fXm
J ~(
1 fXm
) f0
rlo(<X.mr) dr
Instructor's Manual 169
10; f(O) = 1; 1.10801, 0.96823, 1.01371, 0.99272, 1.00436, and we see that the last value
is already correct to 3 significant digits.
18. a 11/21r = 0.6099 (see Table Al in Appendix 5)
Short Courses. Omit the derivation of the Laplacian in cylindrical and spherical coordi-
nates.
I C C
u =- u=-+k.
r2 ' r
10. f(w) = w, An =
2
2n + 1 J 1
wPn(w) dw. Since w = P 1 (w) and the P n(w) are or-
-1
thogonal on the interval -1 ~ w ~ 1, we obtain A 1 = 1, An = 0 (n > 1). Answer:
u = r cos <p. Of course, this is at once seen by inspection.
12. u = -Jr2 P 2 (cos <p) + J = r 2 (-cos 2 <p + }) + i
14. u = 4r3 Ps(cos <p) - 2r 2 P 2 (cos <p) + rP1 (cos <p) - 2
16. f (<p) = COS <p, Uint = r COS <p, Uext = r-:- 2 COS <p, f (<p) = COS 2</> = 2 cos 2 <p - 1,
4 1
2x2 - 1 = fP2 (x) - }, Uint = fr 2Picos <p) - !, Uext = r 3 Picos <p) - r
3 3
18. A 4 = 0, A 5 = 605/16, A 6 = 0, A 7 = -4125/128, A 8 = 0, A 9 = 7315/256, A 10 = 0
170 Instructor's Manual
1
20. Set - = p and consider u(p, e, cp) = rv(r, e, cp). By differentiation,
r
Thus
UPP +- 2pUP -
_1 (2Vr
4p + rVrr) - r-5( + -2) Vrr
r
Vr .
By substituting this and u<l><I> = rv<!><I> etc., into (7) [written in terms of p] and divid-
ing by r 5 we obtain the result.
22. v = r- 2 cos (J sin (J = xyl(x 2 + y 2 ) 2
24. TEAM PROJECT. (a) The two drops over a portion of the cable of length Ax are
-RiAx and -L(ai/at)Ax, respectively. Their sum equals the difference ux+t.x - ux.
Divide by Ax and let Ax - 0.
(c) To get the first equation, differentiate the first transmission line equation with re-
spect to x and use the second equation to replace ix and ixt:
-u= = Rix + Lixt
= R(-Gu - Cut)+ L(-GUt - Cutt)
u 4U-
=- 0
sm -'TTX e-A12t
( .
+ -1 sm
. -
37Tx
- e-A32 t + )
'TT l 3 l '
(e) u = U0 cos ('TTtllVLC) sin ('TTxll)
1+s 2
U(x, s) =- -2 - + c(s)e-sx
s
with c(s) = -1/s 2 as obtained from u(O, t) = 1, U(O, s) = 1/s.
Instructor's Manual 171
.
To complete the separation, we take f(x) = x, obtaining
g t
1 + g = g, g + g = t, g = ce-t + t - 1;
hence
u = x(ce-t + t - 1),
which satisfies u(O, t) = 0. Also, u(x, 0) = x(c - 1). Thus c = 1 and the answer is,
as before,
u = x(e-t +t- 1).
w = f
0
t
f(t - r)
k
2v-:;;:;.a
e-k214,,. dr.
= F(s) ;, { a;o} .
Now apply the convolution theorem.
38. u 32 ( 1 cos 2x e- 4 t + 1
= 1T - ---;; cos 6x e- 3 et + )
4 36
172 Instructor's Manual
40
. u = 400 f 1 sin (2n - 1)7Tx sinh [(2n - 1)7Ty/12]
7T n=l 2n - 1 12 sinh (2n - 1)77"
42.
These include automatic variable step size selection in modem codes, the discussion of
the Runge-Kutta-Fehlberg method, and the extension of Euler and Runge-Kutta methods
to systems and higher order equations.
2. y = sin !1rx. Since the values obtained give y9 = 1.01170 > 1, y 10 comes out com-
plex and is meaningless.
Xn Yn Error X 105
0.1 0.15708 -65
0.2 0.31221 -319
0.3 0.46144 -745
0.4 0.60079 -1301
0.5 0.72636 -1926
0.6 0.83433 -2531
0.7 0.92092 -2991
0.8 0.98214 -3109
0.9 1.01170 -2401
1.0
247
248 Instructor's Manual
6. y = 1/(1 + e-x). The given Verhulst equation is a special Bernoulli equation; see
Sec. 1.6.
Xn Yn Error X 106
0.1 0.524969 10
0.2 0.549813 21
0.3 0.574411 32
0.4 0.598645 42
0.5 0.622407 53
0.6 0.645593 63
0.7 0.668114 74
0.8 0.689890 84
0.9 0.710855 94
1.0 0.730955 104
8. y = tan 2x. Note that the error is first negative and then positive and rapidly increasing,
due to the behavior of the tangent.
Xn Yn Error X 105
10. The error of y(l) is -0.0036, hence comparable to that in Prob. 7. The error of y(2)
is +0.0067, hence twice that in Prob. 7 and of the opposite sign.
12. y = 0, 0.2055, 0.4276, 0.6587, 0.8924; error 0, 0.0101, 0.0221, 0.0322, 0.0409. Hence
the error is about 20% less.
Instructor's Manual 249
14. y = 0, 0.1033, 0.2134, 0.3280, 0.4456, 0.5650, 0.6852, 0.8058, 0.9261; error 0, 0.0011,
0.0022, , 0.0071; about 15% of that in Prob. 11.
16. For instance, x = 0.5, y = 0.632 114 762 (error 0.58 10- 5 ); x = 1,
y = 0.864 660 452 (error 0.43 10- 5 ); hence the error is substantially less, and it is
interesting that it is not increasing: for x = 0.1, , 1.0, it is 0.26, 0.42, 0.52, 0.57,
0.58, 0.57, 0.54, 0.51, 0.47, 0.43 times 10- 5 .
18. Errors -0.004, -0.008, -0.011, -0.015, -0.017, -0.019, -0.021, -0.021, -0.021,
-0.021. For the improved Euler method, the errors times 10 5 are 0.8, 1.6, 2.2, 2.5,
2.5, 1.8, +0.4, -1.9, -5.1, -9.3.
6. Solution y 2 - x2 = 8.
Xn Yn
1.2 3.07246
1.4 3.15595
1.6 3.24962
1.8 3.35261
2.0 3.46410
2.2 3.58330
2.4 3.70945
2.6 3.84188
2.8 3.97995
3.0 4.12311
Content
Euler's method for systems (5)
Classical Runge-Kutta method extended to systems (6)
Runge-Kutta-Nystrom method (7)
X Yi Y2
0 0 4
0.2 0.8 3.2
0.4 1.28 2.4
0.6 1.504 1.664
0.8 1.536 1.0304
1.0 1.43488 0.51712
8. We had to choose x0 *
0 because of the factor 1/x. Those initial values were taken
from Ref. [1] in Appendix 1.
1 0.765198 -0.440051 0
1.5 0.511903 -0.558002 -76
2 0.224008 -0.576897 -117
2.5 -0.048289 -0.497386 -95
3 -0.260055 -0.339446 +3
3.5 -0.380298 -0.137795 170
I II Exact
Xn Yn Yn Yn Error
10. (4S)
0 0 0 0 0
0.2 0.02 0.2 1.21 0.0214 0.0014
0.4 0.0842 0.4420 1.4631 0.0918 0.0076
0.6 0.2019 0.7346 1.7682 0.2221 0.0202
0.8 0.3842 1.0883 2.1362 0.4255 0.0413
1.0 0.6446 0.7183 0.0737
252 Instructor's Manual
12. f(2/3) = (3/2)f(5/3) by (25); now use interpolation in Table A2, etc.
In the present case the errors of the two methods are of the same order of magnitude.
An exact comparison is not possible since the errors change sign in a different fash- _
ion in each method.
4. The values obtained by the Gauss elimination agree with those of the exact solution
of the problem, u(x, y) = x3 - 3xy 2 . Gauss-Seidel would need 14 steps to produce
6S-values or 9 steps for 3S-values.
6. This shows the importance of good starting values; it then does not take long until
the approximations come close to the solution. A rule of thumb is to take a rough es-
timate of the average of the boundary values at the points that enter the linear sys-
tem. By starting from O we obtain
[0.094722 0.101487 0.317994 0.321376] (Step 3)
[0.107407 0.107830 0.324337 0.324548] (Step 5).
8. Uu = 92.86, U21 = 90.18, U12 = 81.25, U22 = 75.00, U13 = 57.14,
u 23 = 47.32, u31 = u 11 , etc., by symmetry.
10. All the isotherms must begin and end at a comer. The diagonals are isotherms u =
25, because of the data obtained and for reasons of symmetry. Hence we obtain a
qualitative picture as follows.
Six of the boundary values are zero, and the two on the upper edge are u13 = u 23 =
\!'312 = 0.866 025. Also, on the right we substitute the starting values 0. With this,
254 Instructor's Manual
With the boundary values and the previous solution on the right, this becomes
-4un + u 12 = 0
un - 4u 12 = -0.866 025 - 0.288 675
-4u21 + U22 =0
U21 - 4U22 = -0.866 025 - 0.288 675.
The solution is
Un = 0.076 98
U21 = 0.076 98
U12 = 0.307 92
U22 = 0.307 92.
Second step. Rows. We can use the previous equations, changing only the right sides:
-4un + u 21 = -0.307 92
un - 4u 21 = -0.307 92
-4u12 + U22 = -0.866 025 - 0.076 98
U12 - 4u22 = -0.866 025 - 0.076 98
Solution:
Un = U21 = 0.102 640, U12 = U22 = 0.314 335.
Second step. Columns. The equations with the new right sides are
-4un + u12 = -0.102 640
un - 4u 12 = -0.866 025 - 0.314 335
-4u21 + U22 = -0.102 640
u 21 - 4u 22 = -0.866 025 - 0.314 335.
Instructor's Manual 255
16. CAS PROJECT. (b) The solution of the linear system (rounded to integers), with
the values arranged as the points in the xy-plane, is
160 170 157 110
138 145 125 75
138 145 125 75
160 170 157 110
Twenty steps gave accuracies of 3S- 5S, with slight variations between the compo-
nents of the output vector.
4. The exact solution of the Poisson equation is u = x 2 y 2 The approximate solution re-
sults from Au = b, where
-4 1 0 1 0 0 4
1 -4 1 0 1 0 10
0 2 -4 0 0 1 8
A= b=
1 0 0 -4 1 0 1
0 1 0 1 -4 1 -20
0 0 1 0 2 -4 -103
where the six equations correspond to Pn, P 21 , P 31 , P 12 , P 22 , P 32 , in our usual or-
der. The components of b are of the form a - c with a resulting from 2(x 2 + y 2 )
and c from the boundary values; thus, 4 - 0 = 4, 10 - 0 = 10, 20 - 12 = 8,
10 - 9 = I, 16 - 36 = -20, 26 - 81 - 48 = -103. The solution of this system
agrees with the values obtained at the Pik from the exact solution, un = 1, u21 =
u12 = 4, u 22 = 16, and u31 = 9, u32 = 36 on the boundary. u41 = u21 + 12 and u42
= u 22 + 48 produced entries 2 in A and -12 and -48 in b.
6. Exact solution u = 9y sin i7Tx. Linear system Au = b, where
-4 1 1 0 0 0 a
1 -4 0 1 0 0 a
1 0 -4 1 1 0 2a
A= b=
0 1 1 -4 0 1 2a
0 0 2 0 -4 1 3a +C
0 0 0 2 1 -4 3a + C
a= -8.54733, c = -V243 = -15.5885. The solution of this system is (exact val-
ues of u in par~ntheses)
un = u21 = 8.46365 (exact iV3 = 7.79423)
u12 = u 22 = 16.8436 (exact 9V3 = 15.5885)
u13 = u 23 = 24.9726 (exact 2iv'3 = 23.3827).
14. Let v denote the unknown boundary potential. Then v occurs in Au = b, where
V
The solution of this linear system is u [5 10 10 16]T. From this and
19
5v/I9 = 100 (the potential at Pn) we have v = 380 as the constant boundary po-
tential on the indicated portion of the boundary.
16. Two equations are as usual:
where the right side is due to the fact that we are dealing with the Poisson equation.
The third equation results from (6) with a = p = q = 1 and b = 1/2. We get
2 [
U22
2
+ ul,5/2 +
3/4
Uo2
2
+ U11 -
3/2
3/2
1/2 U12
J-
-
2
L ___
l
I.
The first two terms are zero and u02 = -2; these are given boundary values. There
remains I
iu11 - 6u 12 = 4.
Our three equations for the three unknowns have the solution
U11 = -1.5, U21 = -1, U12 = -1.
Here, by the central difference formula for the normal derivative (partial derivative
with respect to x) we get
auo. 1
0 = -ax-3 = -(u1
2h 3
- U_1 -)
,J
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ~ ---------
0 0 0 0 0 0 0
0.01 0 0 0 0 0 0.5
0.02 0 0 0 0 0.125 0.866025
0.03 0 0 0 0.031 0.279 1
0.04 0 0 0.008 0.085 0.397 0.866 025
0.05 0 0.002 0.025 0.144 0.437 0.5
0.06 0.001 0.007 0.049 0.187 0.379 0
0.07 0.004 0.016 0.073 0.201 0.236 -0.5
0.08 0.010 0.027 0.091 0.178 0.043 -0.866025
0.09 0.019 0.039 0.097 0.122 -0.601 -1
0.10 0.029 0.048 0.089 -0.065 -0.520 -0.866 025
0.11 0.039 0.054 0.040 -0.140 -0.493 -0.5
0.12 0.046 0.047 -0.002 -0.183 -0.406 0
10. r = k/h 2 = 1, k = 1, 2 steps. The series in Sec. 11.5 gives (with L = 10, t = 2,
bi = 2.58012, b2 = 0, b3 = 0.09556)
u = bi sin O. lx exp (- 7T 2 /50) + b3 sin 0.3x exp ( -97T2 /50).
The values fort = 2 and x = 0, 1, , 10 are (exact values in parentheses) 0 (0),
0.6691 (0.6546), 1.2619 (l.2449), 1.7212 (1.7135), 2.0075 (2.0143), 2.1043 (2.1179),
2.0075 (2.0143), etc. (symmetric).
12. CAS PROJECT. u(O, t) = u(l, t) = 0, u(0.2, t) = u(0.8, t), u(0.4, t) = u(0.6, t),
where
X = 0.2 X = 0.4
t= 0 0.587785 0.951057
0.393432 0.636586 Explicit
t = 0.04 0.399274 0.646039 CN
0.396065 0.640846 Exact (6D)
0.263342 0.426096
t = 0.08 0.271221 0.438844
0.266878 0.431818
0.176267 0.285206
t = 0.12 0.184236 0.298100
0.179829 0.290970
0.117983 0.190901
t = 0.16 0.125149 0.202495
0.121174 0.196063
0.078972 0.127779
t = 0.2 0.085012 0.137552
0.081650 0.132112
Instructor's Manual 259
2. Note that the curve of f(x) is no longer symmetric with respect to x = 0.5. The so-
lution was required for O ~ t ~ 1. We present it here for a full cycle O ~ t ~ 2:
u(x, t) =-
1 fx+ct sin 'TTS ds = - 1 [cos 'TT(x - ct) - cos 'TT(x + ct)].
2 x-ct 2 'TT
From (8) we have kgi = O.lgi = 0.1 sin 0.1 'TTi. Because of the symmetry with respect
260 Instructor's Manual
to x = 0.5 we may list only the following values (with the exact values in parenthe-
ses):
8. Since u(x, 0) = f(x), the derivation is immediate. Formula (8) results if the integral
equals 2kgi.
10. Exact solution: u(x, t) = (x + t)2. The values obtained in the computation are those
of the exact solution. u 11 , u 21 , u 3 i , u 41 are obtained from (8) and the initial condi-
tions uio = (0.2i) 2 , gi = 0.2i. In connection with the left boundary condition we can
use the central difference formula
We see that the error of the last value has decreased by a factor 10, approximately,
due to the smaller step.
Instructor's Manual 261
24. y = 2e-x + x2 + I.
Xn Yn Error X 10 4
0.1 2.8205 -8
0.2 2.6790 -15
0.3 2.5738 -22
0.4 2.5033 -27
0.5 2.4662 -32
0.6 2.4612 -36
0.7 2.4871 -39
0.8 2.5429 -42
0.9 2.6276 -44
1.0 2.7404 -46
Xn Yn Error X 106
0.0 0.000000 0
0.2 0.021333 69
0.4 0.091655 170
0.6 0.221 808 311
0.8 0.425 035 506
1.0 0.717 509 772
Xn Yn Error X 106