1. In the notation of Equation (9) in Section 10.1 of the text we have al = PI = ° and
a2 = !h = 1, so Theorem 1 implies that the eigenvalues are all nonnegative. If A = 0, then y" = ° implies that y(x) = Ax + B. Then y'(x) = A, so the endpoint conditions yield A = 0, but B remains arbitrary. Hence Ao = ° is an eigenvalue with eigenfunction
yo(x) = 1.
If A = if > 0, then the equation y" + ely = ° has general solution
y(x) = A cos ax + B sin ax,
with
y'(x) = -Aa sin ax + Ba cos ax.
Then y'(O) = ° yields B = ° so A -:1= 0, and then
y'(L) = -Aa sin aL = 0,
so aL must be an integral multiple of n: Thus the nth positive eigenvalue is
and the associated eigenfunction is
nnx Yn(x) = cOS-.
L
2. In the notation of Equation (9) in this section we have al = !h = 1 and a2 = PI = 0, so Theorem 1 implies that the eigenvalues are all nonnegative. If A = 0, then y" = ° implies y(x) = Ax + B. But then yeO) = B = ° and y'(L) = A = 0, so it follows that ° is not an eigenvalue. We may therefore write A = if > 0, so our equation is
550 Cha~er10
y" + dy = 0 with general solution
y(x) = A cos ax + B sin ax.
Now yeO) = A = 0, so y(x) = B sin ax and
y'(x) = Bacos ax.
Hence
y'(L) = Ba cos aL = 0,
so it follows that aL must be an odd multiple of ;r/2. Thus
(2n-l);r
a = ....:___-,--
n 2L
3. If A = 0 then y" = 0 yields y(x) = Ax + B as usual. But y'(O) = A = 0, and then hy(L) + y'(L) = h(B) + 0 = 0, so B = 0 also. Thus A = 0 is not an eigenvalue. If A = d > 0 so our equation is y" + dy = 0, then
y(x) = A cos ax + B sin ax,
so
y'(x) = -Aa sin ax + Ba cos ax.
Now y'(O) = 0 yields B = 0, so we may write
y(x) = cos ax,
y'(x) = -a sin ax.
The equation
hy(L) + y'(L) = h cos aL - a sin al. = 0
then gives
h hL
tanaL = - = -, a aL
so f3n anL is the nth positive root of the equation
hL
tanx = -. x
Thus
Finally, a sketch of the graphs y = tanx and y = hLlx indicates that f3n ::::; (n - l);r for n large.
Section 10.1
551
4. Here al = h > 0, a2 = 131 = 1, and f3z = 0, so by Theorem 1 in Section 10.1 there are no negative eigenvalues. If A = 0 and y(x) = Ax + B, then the equations
hyCO) - y'(O) = hB - A = 0,
y(L) = AL + B = 0
imply h = AlB = -1 I L < O. Thus 0 is not an eigenvalue. If A = d > 0 and
y(x) = A cos ax + B sin ax,
then the condition hyCO) = y'(O) yields B = hAl a, so y(x) = A(acosax+hsinax) a
because the denominator integral here evaluates - by use of the trigonometric identity sirr' A = t(1-cos2A)- to L12.
7. The coefficient en in Eq. (23) of this section is given by Formula (25) with j(x) = rex) = 1, a = 0, b = L, and Yn(x) = sin fJnx. Using the fact that tan fJ" = - fJn, so
L hL
sinfJ cos fJ
__ " = _ n we find that
fJn hL '
JL sirr' fJ"x dx = fL .!..(1_coS2fJnX)dx = '!"[x-_£_sin 2fJnX]L o L 0 2 L 2 2fJ" L 0
552
Chapter 10
- 1 (L L sin f3n f3 ) _ 1 (L L cos f3n f3 ) hL + cos! f3n
- - - --cos - - + cos = -------"-
2 f3n n 2 hL n 2h
IL f3 x L(l-cos f3n)
and sin cn.: dx = Hence the desired eigenfunction expansion is
o L 13
n
1 = 2hL f 1- cos f3n sin f3nx
,,=, 13/1 ( hL + cos' 13,,) L
for ° < x < L.
8. The coefficient en in (23) is given by Formula (25) with j(x) = rex) = 1, a = 0, b = L, and Yn(x) = cos f3"x1 L:
f cos/ f3t dx f ±(1 + cos 2~X)dx [M X+-2~-" sin-2Pr-'X)]
L . 13
-SIn /I
f3n
= -,---''-'-'----""7"
!(L+~sin2/3/1)
2 2/3/1
4sin/3n
= --__.:....:.:.__
2/3/1 + sin 2/3/1
Hence the desired eigenfunction expansion is
1 ~ 4 sin f3n f3nx
= ~ cOS-.
n=' 2f3n +sin2f3n L
9. The coefficient en in (23) is given by Formula (25) with j(x) = rex) = 1, a = 0, b = 1, and Yn(x) = sin f3"x. Using the fact that tan f3n = -f3nl h, so
1 x sin f3nx dx = ~ 1 f3nX sin f3nx . f3ndx = ~ {" u sin u du
/3n /3/1
Section 10.1
553
It follows that the desired expansion is given by
x = 2h(1 + h) f sin fln sin flnx n=l fl,; ( h + cos' fln)
for ° < x < 1.
10. The coefficient c; in (23) is given by Formula (25) with fix) = x, rex) = 1, a = 0, b = 1, and Yn(x) = cos fJnX. Integrations similar to those in Problems 8 and 9 give
1 x cos fJnx dx
en = =
1 cos' fJ"x dx
4(fl" sin fJ" + cos fJ" -1) fJn (2fJ" + sin 2fJ,,)
With this value of Cn for n = 1,2,3,···, the desired eigenfunction expansion is
00
x = Le" cosflnx.
n=l
11. If A = ° then y" = ° implies that y(x) = Ax + B. Then yeO) = ° gives B = 0, so y(x) = Ax. Hence
hy(L) - y'(L) = h(AL) - A = A(hL - 1) = °
if and only if hL = 1, in which case An = ° has associated eigenfunction yo(x) = x.
12. If A = - d < 0, then the general solution of y" - dy = ° is
y(x) = A cosh ax + B sinh ax.
But then j'(O) = A = 0, so we may take y(x) = sinh ax. Now the condition hy(L) = y'(L) yields
h sinh al. = a cosh al..
It follows that fl = aL must be a root of the equation
x tanhx = -. hL
554
Chapter 10
The curve y = tanh x passes through the origin with slope 1, and is concave upward for x < 0, concave downward for x > O. Hence this curve and the straight line y = x/hL intersect other than at the origin if and only if the slope of the line is less than 1 - that is, if and only if hL > 1. In this case, with fJo the positive root of tanh x = x/hL, we have Au =-fJg and Yo(x) = sinhfJox.
13. If A = + d > 0, then the general solution of y" + dy = 0 is
y(x) = A cos ax + B sin ax.
But then yeO) = A = 0, so we may take y(x) = sin ax. Now the condition hy(L) = y'(L) yields
h sin aL = acos aL.
It follows that fJ = aL must be a root of the equation
x tanx = -.
hL
So if fJn is the nth positive root of this equation, then An = a,~ = fJ,; / L2 and the corresponding eigenfunction is Y/l (x) = sin fJ/lX / L.
14. With A = 0, y" = 0, and hence y(x) = Ax + B, we have yeO) = B = 0, so y(x) = Ax. Then the condition hy(L) = y'(L) reduces to the equation hL = A, which is satisfied because hL = 1. Thus Ao = 0 is an eigenvalue with associated eigenfunction yo(x) = x. Together with the positive eigenvalues and associated eigenfunctions provided by Problem 13, this gives the eigenfunction expansion
where tan fJn = fJn. The coefficients are given by
r f(x)x dx _ 1_ rL
Co = " - 3.b xf(x) dx,
1 x2 dx L
r f(x)sinfJ/lx/ L dx
Co = _:o:_-.,-- =
IL sin? PnX / L dx
2 rL
. 2 .b f(x) sin P/lx / L dx,
Lsm P/l
the latter because
Section 10.1
555
[ ]L
L 2 IL 1 L 2x
1 sin f3nxlLdx = -1 (l-cos2f3nxIL)dx = - x--sin~
2 2 2Pn L 0
I ( sin Pn ) L ( 2) L sin 2 f3
= 2" L-L,y,cosf3n ="2 I+cos f3n = 2 n,
15. If Au = 0, then a general solution of y" = ° is y(x) = Ax+ B. The conditions
y(o) + y'(O) = B+A = 0,
y(l) = A+B = °
both say that B = -A, so we may take Yo (x) = x -1 as the eigenfunction associated with Au = 0, If A = +d < 0, then the general solution of y" + dy = ° IS
y(x) = A cos ax + B sin ax,
But y(O) + y'(O) = A+Ba = 0, so A = -Ba, and then
y(I) = Acosa+Bsina = -B(acosa-sina) = 0,
Thus the possible values of a are the positive roots {f3n} of the equation tan x = x, and the nth eigenfunction is YII (x) = 1311 cos 1311 X - sin f3l1x,
17. The Fourier sine series of the constant function f(x) == w for 0< x < L IS
w = 4w I _!_sin mrx,
7r II odd n L
If y = I b; sin nnx I L, then
00 4 4b
E1 Y(4) E1"\' n 7r 11 ' nnx
= L. 4 SIn--,
11=1 L L
Upon equating coefficients in these two series and solving for b.; we see that
() 4wL4 "\' I , nnx
y x = --5 L. -5 SIn--,
EI7r II odd n L
18. By Equation (16) in Section 9,3, the Fourier sine series of f(x) = bx for 0< x < L is
2bL 00 (_1)"+1 . nnx
bx = --I SIn--,
7r 11=1 n L
If y = Ibll sin zerx I L, then
556
Chapter 10
<Xl 4 4b
EI (4) = EI" n Te n • nTex
y L... 4 sm .
n;1 L L
Upon equating coefficients in these two series and solving for b.; we see that
( ) _ 2bL5 ~ (_1)"+1 . nTeX
y x - 5 L... 5 sm .
Elst n;1 n L
19. With A = a\ the general solution of y(4) - a4y = 0 is
y(X) = A cosh ax + B sinh ax + C cos ax + D sin ax,
and then
y'(X) = a(A sinh ax + B cosh ax - C sin ax + D cos ax).
The conditions yeO) = 0 and y'(O) = 0 yield C = -A and D = -B, so now
y(X) = A(cosh ax - cos ax) + B(sinh ax - sin ax).
The conditions y(L) = 0 and y'(L) = 0 yield the two linear equations
A(cosh aL - cos aL) + B(sinh aL - sin aL) = 0, A(sinh aL + sin aL) + B(cosh aL - cos aL) = O.
This linear system can have a non-trivial solution for A and B only if its coefficient determinant vanishes,
(cosh aL - cos aLi - (sinh2aL - sin2aL) = o.
U sing the facts that cosh 2 A - sinh 2 A = 1 and cos' A + sin 2 A = 1, this equation simplifies to
cosh aL cos aL - 1 = 0,
so p = aL = x satisfies the equation
cosh x cos x = 1.
The eigenvalue corresponding to the nth positive root Pn is
Finally the first equation in the pair above yields B = _ coshaL-cosaL sinhaL -sinaL'
Section 10.1
557
so we may take
Yn(x) = (sinhPn -sinpn)( cosh P2x -cos P2x)
-( cosh ji, -cos Pn)( sinh P2x -sin P2x)
as the eigenfunction associated with the eigenvalue An.
20. As in Problem 19, the solution of y(4) - a4y = 0 satisfying the left-endpoint conditions yeO) = 0 and y'(O) = 0 is given by
y(x) = A(cosh ax - cos ax) + B(sinh ax - sin ax).
The right-endpoint conditions y"(L) = 0 and y(3)(L) = 0 now yield the two linear equations
A(cosh aL + cos aL) + B(sinh aL + sin aL) = 0, A(sinh aL - sin aL) + B(cosh aL + cos aL) = O.
This linear system can have a non-trivial solution for A and B only if its coefficient determinant vanishes,
(cosh aL + cos aLi - (sinh2aL - sin2aL) = O.
This equation simplifies to
cosh aL cos aL + 1 = 0,
so P = aL = x satisfies the equation
cosh x cos x = - 1.
The eigenvalue corresponding to the nth root Pn is
Finally the first equation in the pair above yields
B = _ coshaL + cosaL sinh aL + sin aL '
so we may take
558
Chapter 10
Yn (x) = (sinh Pn + sin Pn) ( cosh PiX - cos PiX)
- ( cosh a, + cos Pn ) ( sinh Pix - sin PiX)
as the eigenfunction associated with the eigenvalue An.
21. As in Problem 19, the solution of y(4) - a4y = 0 satisfying the left-endpoint conditions yeO) = 0 and y'(O) = 0 is given by
y(X) = A(cosh ax - cos ax) + B(sinh ax - sin ax).
The right-endpoint conditions y(L) = 0 and y"(L) = 0 yield the two linear equations
A(cosh aL - cos aL) + B(sinh aL - sin aL) = 0, A(cosh aL + cos aL) + B(sinh aL + sin aL) = O.
This linear system can have a non-trivial solution for A and B only if its coefficient determinant vanishes,
(cosh aL - cos aL)( sinh aL + sin aL)
- (cosh aL + cos aL)(sinh aL - sin aL) = O.
This equation simplifies to 2 cosh aL sin aL - 2 cos aL sinh aL = 0, which is equivalent to tanhaL = tanaL. Hence P = aL = X satisfies the equation tanh X = tan x,
and the eigenvalue corresponding to the nth positive root Pn is All = a,; = (PII / L)4.
SECTION 10.2
APPLICATIONS OF EIGENFUNCTION SERIES
1. The substitution u(x,t) = X(x)T(t) yields the separated equations
x"+dx= 0
and T' = -kAT
with separation constant A = a", In Problem 3 of Section 10.1 we saw that the SturmLiouville problem
X" + dx = 0
,
X'(O) = hX(L) + X'(L) = 0
Section 10.1
559
has eigenvalues A" = a,; = p,; I L2 and eigenfunctions
Xn(x) = cos Pix
for n = 1, 2, 3, .. " with {Pn} being the positive roots of the equation tan x = hLlx. The solution of T,; = -kA"T" is then
- (P;ktJ
T" (t) - exp --rr- '
so the resulting formal series solution is
- co (P;ktJ p"x
u(x,t) - ICn exp --2- cOS-.
n=1 L L
The coefficients in the eigenfunction expansion are given by
8. With m = ° the boundary value problem in Example 2 is
UII = a2uxx (0 < x < L, t > 0),
u(O,t) = ux(L,t) = 0,
ul(x,O) = 0,
u(x,O) = bx.
The substitution u(x,t) = X(x)T(t) gives the separated equations
X" +AX = Til + Aa2T = 0.
and the eigenfunctions of the eigenvalue problem
X" +AX = 0, X(O) = X'(L) = °
are of the form
X () . nn:x
n X = Sln--
2L
with n odd, with corresponding eigenvalue All = n2;r2 14L2. This leads readily to the solution
~ . nn:x nn at
u(x,t) = L. cn sm--cos--,
II odd 2L 2L
where c; is the odd half-multiple sine coefficient (of Problem 21 in Section 9.3) given by
JL 8bL sin (2n -1);r
- ~ b . (2n-l)Jrx dx _ 2
C211_1 - L x sm L - ----=-'2 "'-:2:--"
(2n -1) ;r
o
=
8bL( _1)"+1 (2n-li;r2·
9.
(a)
With ,.1,=0, the endpoint-value problem in (19) is X" = 0, X(O) = X'(O) = 0,
which has only the trivial solution X(x) == 0. Thus ,.1,=0 is not an eigenvalue.
(b) With ,.1,= _a2 < 0, the endpoint-value problem in (19) is
X"-a2X = 0,
X(O) = 0, -ma2X(L) = Ac5X'(L).
The differential equation and the left-endpoint condition here give X(x) = sinh ax, and substitution in the right-endpoint condition gives
-ma2 sinhaL = Ac5acoshaL, that is, tanhaL =_.!5_ aL
566
Chapter 10
with k = A5L/ m > O. But the graph y = tanh x lies (aside from the origin) in the first and third quadrants, while the graph y = -k / x lies interior to the second and fourth quadrants. Hence the two cannot intersect, and it follows that there cannot be an eigenvalue of the assumed form A = _a2 < 0, .
10. (a) With 5= 7.75gm/cm3 and E = 2.1012 inEquation(16),thespeedofsoundin steel is
a = ~ ~ 5.08 X 105 ern/sec ~ 11364 mph.
(b) With 5 = 1 gm/cnr' and K = 2.25.1010 in Equation (16), the speed of sound in
If /"'0 = 0 then R(r) = Ar + B. The condition R(O) = 0 gives B = 0, and R(r) = Ar satisfies the condition R(a) - aR'(a) = O. Thus Ao = 0 is an eigenvalue with eigenfunction
Ro(r) = r;
To(t) = 1.
If A = d > 0 then
R(r) = A cos ar + B sin ar
and R(O) = 0 gives A = 0, so
R(r) = sin ar,
R'(r) = a cos ar.
The condition R(a) = aR'(a) yields sin aa = aacos aa, that is,
tanx = x
where x = aa. If {fin} are the roots of this equation, then An eigenvalue with associated eigenfunction
and
( /32ktJ
TJt) = exp -~ .
We therefore obtain a solution of the form
The coefficient formulas given in the textbook follow immediately from Problem 14 in Section 10.1, and finally we obtain u(r, t) upon division of vCr, t) by r.
18. The only difference from Example 3 in the text is that the solution of Equation (37) with
2 2 2
r,;(0) = 0 is T,,(t) = sin n 1l' 2a '; L
19. With the given initial velocity function g(x) with constant value P / 2p& concentrated in the interval L / 2 - e < X < L / 2 + e, the coefficient formula of Problem 18 gives
Section 10.2
569
2L2p . nst . mrs
3 3 2 SIll-SIll--.
nr: a pe 2 L
This gives the s-dependent solution
Because
L . nsts sin ( nne! L)
-SIll-- = ~ 1 as s~O,
nne L mrs/ L
the limit y(x,t) = limy(x,t,s) has the expansion
6--+0
2LP 00 1 . nst . n2;r2 a2t . ntt x
y(x,t) = 2 2 L-2 SIll-SIll 2 SIll--.
;r a P /1=1 n 2 L L
20.
c = /I
2L fL . nnx 2v L2
2 2 2 voSIll--dx = 3 03 2 (1-cosn;r)
nst:a 0 L nn:a
The fundamental frequency is
_ ;r2 a2 _ ;r2 {fl
{OI - -2- - -2 - {OI·
L L P
With
E = 2.1012 dyne/ern',
I = (2.54)4112 ~ 3.47 em",
p = (7.75)(2.54/ ~ 50.00 gm/cm, L = (19)(2.54) ~ 48.26 em,
we calculate
{OI ~ 1578 rad/sec ~ 251 cycles/sec.
Thus we hear middle C (approximately).
570
Chapter 10
SECTION 10.3
STEADY PERIODIC SOLUTIONS AND NATURAL FREQUENCIES
In Problems 1-6 we substitute u(x,t) = X(x)cosmt III
_ 2
Utt - a Uxx
and then cancel the factor cos an to obtain the ordinary differential equation
with general solution
mx . mx
X(x) = Acos-+BsIll-.
a a
(*)
It then remains only to apply the given endpoint conditions to determine the natural (circular) frequencies - the values of m for which a non-trivial solution exists.
1. Endpoint conditions: X(O) = X(L) = 0
With conditions X(O) = 0 in (*) implies that A = 0, so X(x) = sin(mx I a). Then X(L) = sin(mLI a) = 0 implies that mLI a = nx, an integral multiple of n: Hence
. nst a nst ~
the nth natural frequency IS mil = -- = - -.
L L 0
2. Endpoint conditions: X'(O) = X'(L) = 0
The condition X'(O) = 0 gives B = 0 in (*), so we have
mx X(x) = cos-, a
X'( ) ai : oix
so x = --SIll-.
a a
Hence the condition X'(L) = 0 implies that iol.l a is an integral multiple of n: Thus
. nx a nst ~
the nth natural frequency IS mil = -- = - -.
L L 0
3. Endpoint conditions: X(O) = X'(L) = 0
The condition X(O) = 0 gives A = 0 in (*), so we have
X(x) = sin mx, a
'() ca oix
so X x = -cos-.
a a
Section 10.3
571
Hence the condition X'(L) = 0 implies that OJLI a is an odd integral multiple of 1rI2.
Just as in Example 3 of Section 10.2, the conditions X(O) = X"(O) = 0 imply that A = C = 0 in (**), so
X(L) = Bsinh eL +Dsin eL = 0,
a a
X"(L) - e2 (B 'nheL D' eL) - 0
_- Sl -- sm- - .
a2 a a
It follows that
Bsinh eL = Dsin eL = O.
a a
But sinh eLI a "* 0 so B = O. Hence D "* 0 so sin eLla = o. Thus eLI a = nn; an integral multiple of Jr. Therefore the nth natural frequency OJn = e,; is given by
When we substitute the derivatives above and write z = pL we get
A (coshz+cosz)+B (sinhz+ sinz) = 0,
A [(k - mol)(coshz - cosz) - E1p3(sinhz - sinz)] +B[(k-mal)(sinhz-sinz) -Eli(coshz+cosz)] = O.
If Ll denotes the coefficient determinant of these two linear equations in A and B, then the necessary condition Ll = 0 for a non-trivial solution reduces eventually to the equation
Elp3(1 + cosh z cos z) - (k - moi)(sinh z cos z - cosh z sin z) = o.
Finally we substitute p = zlL, M = pL, and
to get the frequency equation
MElz\l + cosh z cos z) = (kML3 - mElz4)(sinh z cos z - cosh z sin z).
We may divide by cosh z cos z to write this equation in the form
MElz\l + sech z sec z) = (kML3 - mElz4)(tanh z - tan z).
If fJn denotes the nth positive root of this equation, then as usual the nth natural frequency is
578
Chapter 10
15. We want to calculate the fundamental frequency of transverse vibration of a cantilever with the numerical parameters
L = 400 em
E = 2· 1012 gm/ em-sec'
I = (1/12)(30 cm)(2 em)" = 20 cm4
p = (7.75 gm/cm3)(60 crrr') = 465 gm/cm.
When we substitute these values and f31 = 1.8751 in the frequency formula
we find that ml :=;:j 6.45 rad/sec, so the fundamental frequency is ml/2:r:=;:j 1.03 cycles/sec. Thus the diver should bounce up and down on the end of the diving board about once every second.
16. When we substitute y(x, t) = X(x)cos ox in the given partial differential equation
and cancel the factor cos ax, we get the ordinary differential equation
El)(4) + P X" - AX = 0
where A = pm2• By solving the characteristic equation
we find the general solution
X(x) = A cosh ax + B sinh ax + C cos f3x + D sin f3x
where
f32 = _ -p-.Jp2 +4AEl . 2El
2 _p+.Jp2 +4AEl
a =------
2El '
The endpoint conditions X(O) = X"(O) = 0 imply that A = C = 0, so
X(x) = B sinh ax + D sin f3x.
Then the conditions X(L) = X"(L) = 0 yield the equations
Section 1 0.3
579
B sinh aL + D sinf3L = 0, dB sinh aL - (iD sin f3L = 0.
The determinant of these two linear equations in Band D must vanish in order that a nontrivial solution exist, so
(d + (i)sinh aL sin f3L = 0.
It follows that sin f3L = 0, so f3L must be an integral multiple of 1(. The definitions of d and (i imply that
Hence if f3n = nn! L, the corresponding value of an is
all =
Then the corresponding value of A is
2 2 (n41(4 J( PL2 J
All = EIallf311 = EI -4- 1- 2 2 .
L n 1( EI
Finally, the nth natural frequency is given by
mil = p;; = n2~2 (1- PL2 J1/2 W.
fp L n21(2EI ~p
17. When we substitute y(x, t) = X(x)cos ox in the given partial differential equation
and cancel the factor cos tot, we get the ordinary differential equation
EIx-4) + P X" - AX = °
where P = AI/ pA and A = pm2• By solving the characteristic equation
we find the general solution
X(x) = A cosh ax + B sinh ax + C cos f3x + D sin f3x
580
Chapter 10
where
f32 = _ -p-.Jp2 +4AE1. 2E1
_p+.Jp2 +4AE1 2E1
The endpoint conditions XeD) = X"(O) = ° imply that A = C = 0, so
X(x) = B sinh ax + D sin f3x.
Then the conditions X(L) = X"(L) = ° yield the equations
B sinh aL + D sin f3L = 0, dB sinh aL - rfD sin f3L = 0.
The determinant of these two linear equations in B and D must vanish in order that a nontrivial solution exist, so
(d + rf)sinh aL sin f3L = 0.
It follows that sin f3L = 0, so PL must be an integral multiple of Jr. The definitions of d and rf imply that
2p2 A
a =-.
E1
Hence if Pn = nit! L, the corresponding value of an is given by
2 = n2Jr2 _~
all L2 pAE
Then a;' PI; = All / E1 gives the equation
that we readily solve for All' The resulting value of the nth natural frequency is
18. Substitution of u(x,t) = X(x)sinwt in the longitudinal bar problem
Section 10.3
581
a2u = a2 a2U (a2 = Ep)
at2 aX2
U(O,t) = 0, AEux(L,t) = Fa sin ex
yields the endpoint problem
m2
X"+-2 X = 0, X(O)=O, AEX'(L)=Fa.
a
Because of the right-endpoint condition, we try X(x) = Bsinmx / a and get
ca oil.
AE·B·-cos- = Fa,
a a
so
Faa
B = ----~~----
AEm cos( to L / A)
The resulting steady periodic solution is
Faa sine oix / a) sin tot
u( x, t) = --"--_;_--'----
AEm cos(mL / A)
19. Substitution of y(x,t) = X(x)sinmt in the transverse bar problem
a2y + a4 a4y = °
at2 ax4
y(O,t) = yx(O,t) = 0, yxx(L,t) = E1 yxxx(L,t) + Fa sin or = °
yields the endpoint problem
X(4) _ p4X = ° (where p2 = to / a2),
XeD) = X'(O) = 0,
X"(L) = El X'"(L) + Fa = 0.
When we impose the fixed-end conditions X(O) = X'(O) = ° on the general solution
X(x) = A coshpx + B sinhpx + C cospx + D sinpx
we find readily that C = -A and D = -B, so
X(x) = A(coshpx - cospx) + B(sinhpx - sinpx).
It remains only to find A and B. But the free-end conditions yield the linear equations
and cancel the factor eiwt, the result is the ordinary differential equation
where
E"(x) - rE(x) = 0
r = (RG - LC oi) + i rd.._LG + RC).
If (a + P02 = y, then the general solution is
In order that e(x ,t) be bounded as x ~ 00 we choose B = 0, and in order that
e(O, t) = Eocos ax we choose A = Eo. Then our steady periodic solution is the real part
SECTION 10.4
CYLINDRICAL COORDINATE PROBLEMS
1. Substitution of u(r,t) = R(r)T(t) in the wave equation
82u 2 (82U 1 8U)
8t2 = a 8r2 +;- 8r
yields the separation
Section 10.3
583
R" +.!. R'
= _ __,_r_ = /L = _a2 •
R
The r-equation has general solution
T(t) = A cosaat + Bsinaat,
and we choose B = 0, so that T'(O) = 0 (because the membrane is initially at rest). The r-equation can be written in the form
which is the parametric Bessel equation of order zero, with continuous solution
R(r) = Jo(ar). In order that the fixed boundary condition R(e) = 0 be satisfied, we
choose a = r" / e, where r" is the nth positive solution of Jo(x) = O. At this point we have product functions of the form Jo (r"r / e) cos(r"at / e) that satisfy the wave equation and the homogeneous boundary conditions, so we form the formal series solution
In order to satisfy the initial position condition u(r, 0) = f (x) is suffices that the {e,,} be the Fourier-Bessel coefficients of the function f(x) given by
2. This is the same as Problem 1, except that the membrane has initial position u(r,O) = 0, sointhet-factor T(t) = Acosaat+Bsinaat we choose A=O sothat T(O)=O. We then get product functions of the form Jo tr.r / e) sin(r"at / e) that satisfy the wave equation and the homogeneous boundary conditions, so we form the formal series solution
( ) ~ J (rnr) . rnat
u r,t = L.Jen 0 - Slll--.
,,=1 e e
In order that the initial velocity condition u, (r, 0) == Vo we satisfied, we want
v = ~ rna. e J (r nr )
o L.J 11 0 ,
Il~ e e
and hence
584
Chapter 10
«: =~. 2 2 2 Je r v.J; (rnr) dr
rna C J1(r,J 0 C
2cvo ern J ( ) dx
= ar~JI(r,Y 1 x 0 x
2cvo [J ( )JYn = _2_cv..::.....o _
= ar,3,J1(rn)2 X 1 X 0 2J()·
«r; 1 r,
(with X = rnr / c)
This gives the desired solution
_ 2cvo ~ Jo (rllr / c )sin(rnat / c)
u(r,t) - -- L.... 2 •
a n=1 rnJI (rn)
3.
(a)
As in Problem 2,
u(r,t) = i>nJo (rnr )sin rnat .
n=1 C C
In order to satisfy the given initial condition we must choose
(with x = rnr / c)
(b) The final formula given in the text for u(r, t) now follows because pa2 = T and J,(x)/x ~ 112 as x ~ o.
4.
(a)
Just as in Example 1 of the text we derive first a formal series solution of the form
u(r,t) = fCn exp(- r~~JJo (rnr).
n=1 C C
In order to satisfy the given initial condition we calculate the Fourier-Bessel coefficient
cn =
2 J"( qo) J(rnr)d
7 2 --2rO- r
C J1 (rn) 0 sits c
Section 10.4
585
(with x = Yllr Ie)
(b) The final formula given in the text for u(r, t) now follows because J1(x)lx ~ 112
as x ~ O.
5.
(a)
We start with the steady-state boundary value problem
(r < c, 0 < z < L)
u(c,z) = 0 u(r, 0) = 0, u(r, L) = ui:
The substitution u(r, z) = R(r)Z(z) yields the equations
rR"+R'+drR = 0
,
Z"- dz = 0
with separation constant A = d. The homogeneous endpoint conditions are
R(c) = Z(O) = O.
If A = d = 0 then r R" + R = 0 implies
R(r) = A + B In r.
We choose B = 0 for continuity at r = 0, so R(r) = A. Then R(c) = 0, so A = 0 also, and hence 0 is not an eigenvalue.
If A = d > 0 then we have the parametric Bessel equation with general solution
R(r) = AJo(ar) + BY(ar).
In order that R(r) be continuous at r = 0 we choose B = 0, so R(r) = AJo( ar). Then
R(c) = aAJo(ac) = 0
586
Chapter 10
requires that Y = ac be a root of the equation
Jo(x) = o.
If an = Ynl c where Yn is the nth positive root of this equation, then
The corresponding function Z(z) of z is
Zn(z) = ~ cosh Ynz + B; sinh Ynz ,
c c
and we choose An = 0 because Z(O) = O. Thus we get the formal series solution
where JO(Yn) = O. To satisfy the condition u(r, L) = Uo, we need (by Eq. (22) in the text)
(with x = Ynr Ie)
This gives the desired solution
6.
(a)
Here we have the same separation of variables
r R" + R' + d r R = 0,
z"-dz= 0
as in Problem 5, but the insulation condition ur(c,z) = 0 implies
If A = d = 0 then r R" + R = 0 implies
R(r) = A + B In r.
Section 10.4
587
We choose B = 0 for continuity at r = 0, so R(r) = A. Then R'(e) = 0, so ~ = 0 is an eigenvalue, and we may take Ro(r) = 1. The equation Z"(z) = 0 implies Z(z) = Az + B, but Z(O) = 0 implies B = 0, so we take Zo(z) = z.
If A = if > 0 then we have the parametric Bessel equation with continuous solution R(r) = A Jo(ar). Then
R'(e) = aAJo'(ac) = 0
requires that r = ae be a root of the equation
Jo'(X) = O.
If an = rnl C where rn is the nth positive root of this equation, then
The corresponding function Z(z) of z is
Zn(z) = ~coshrnz+Bnsinhrnz,
C C
and we choose An = 0 because Z(O) = O. Thus we get the solution
where J~(r,J = O. To satisfy the condition u(r, L) = j(r) we apply the formulas in (24) in the text and choose
Co = ~ I r f(r)dr, Lc .b
« = 2 2 2 fe r f(r)Jo (rnr) dr.
c sinh(rnL I c)Jo (rn) 0 c
(b) If j(r) = Uo (constant), then the coefficient formulas above readily yield
Co = uol Land c; = 0 for n > 0, the latter because
jxJo(x)dx = xJ](x)+C = -xJ~(x)+C.
Hence the series reduces to the solution u(r, z) = uoZl L that one might well guess without all these computations.
588
Chapter 10
7. We want to solve the boundary value problem
(r < 1, Z > 0)
hu(l,z)+ ur(l,z) = 0
u(r, z) bounded as z ~ 00 u(r,O) = uc,
We start with the separation of variables in Problem 5,
rR"+ R'+ drR = 0
,
Z"- dz = 0
and readily see that a = 0 is not an eigenvalue. When we impose the condition
hR(I) + R'(1) = 0
on R(r) = Jo(ar), we find that a must satisfy the equation
hJo(x) + xJo'(x) = 0
that corresponds to Case 2 with n = 0 in Figure 10.4.2 of the text. If {rn} are the positive roots of this equation then
The general solution of Z" = rn2Z is
and we choose Bn = 0 so that Zn(z) will be bounded as z ~ 00. Thus we obtain a solution of the form
w
u(r, z) = L ell exp( -rllz )Jo (rllr)
11=1
where
h.I; (rll) + rllJ~ (rll) = 0,
so rIlJI(rll) = hJo(rll) because J~=-JI' Finally,Eq.(23)inthetextgives
2 2 fe ()
e = rn r u J rnr dr
II 22 2 2 00
e (rll + h )JoCrll) 0 e
Section 10.4
589
(with x = rllr / c)
so
11. When we substitute u(r, t) = R(r) sin ox in the given partial differential equation and cancel the factor sin ox, we get the ordinary differential equation
R" +!R' +(m)2 R = _ Po
r a a2
The associated homogeneous equation is the Bessel equation of order zero with parameter «[a. Hence it follows readily that the solution that is continuous at r = 0 is
R(r) = AJo( w;)_ :~.
The condition R(b) = 0 yields A = Po/w2Jo(mb/a), so it follows that the desired steady periodic solution is
Fo [(wr) (mb)].
u(r,t) = 2 Jo - -Jo - smwt.
to Jo(wb/ a) a a
12. When we substitute y(x, t) = X(x)sin ox in the partial differential equation
and then cancel the sin ox factor, we get the ordinary differential equation
2
2X" X' oi Xx 0
x +x +- = .
g
This is of the form of Equation (3) in Section 8.6 with A = 1, B = 0, C = m2/g, and q = 1, so its general solution is given by
590
Chapter 10
We choose B = 0 for continuity at x = 0, and the condition X(L) = 0 then requires that 2(O~LI g = rn' one of the roots {rn} of the equation Jo(x) = O. Hence the
nth natural frequency of vibration of the hanging cable is
'". ~ ~ ~
13. With w(x) = wx and hex) = h (where wand h on the right are constants) the given partial differential equation
w(x) a2y = ~(W(X)h(X) 8.Y)
g at2 ax ax
(*)
reduces to
When we substitute y(x, t) = X(x)cos ax we get the parametric Bessel equation
2 2 X2X"+XX'+~X = 0 gh
with bounded solution
X(x) ~ A.lo ( ;ih J
The condition X(L) = Yo implies that A = Yo I Jo ((OLI.Jih), so
Jo ( (Ox I .Jih)
y(x,t) = Yo ( G) cos-or.
Jo (OLI ygh
14. With w(x) = wand hex) = hx (with wand h being constants on the right) the partial differential equation in (*) above reduces to
When we substitute y(x, t) = X(x)cos ox we get the ordinary differential equation
Section 10.4
591
io:x X2X"+xX'+-X = 0. gh
This has the form of Equation (3) in Section 8.6 with A = 1, B = 0, C = rJ/gh, and q = 1, so its (bounded) solution is given by
X(x) ~ AJo( 2mHiJ
The condition X(L) = Yo now implies that A = Yo / s, (2OJ~L/ gh), so
Jo (2w.Jxlih)
y(x,t) = Yo ( fTI"::i:) coscat,
s, 2OJ'\j L / gh
15. With w(x) = wx and hex) = hx (with wand h being constants on the right) the partial differential equation in (*) above reduces to
When we substitute y(x, t) = X(x)cos ax we get the ordinary differential equation
2
x2X" +2xX' + w x X = 0. gh
This has the form of Equation (3) in Section 8.6 with A = 2, B = 0, C = w2/gh, and q = 1, so its (bounded) solution is given by
X(x) ~ ~Jl( 2mHiJ
The condition X(L) = Yo now implies that A = yoJi/Jl(2w~L/gh), so
J1 (2w.Jxlih)
y(x, t) == Yo (fTI"::i:) cos tot.
J1 2w'\j L / gh
16. With 2 = if the general solution of the parametric Bessel equation of order ° is
y(x) = AJo(ax) + BYo(ax).
The endpoint conditions yea) = y(b) = ° yield the linear equations
592
Chapter 10
A Jo(aa) + B Yo(aa) = 0, A Jo(ab) + B Yo(ab) = 0
in A and B. In order for there to exist a non-trivial solution for A and B the coefficient determinant must vanish. Hence a must be one of the solutions {rn} of the equation
Jo(ax) Yo(bx) - Jo(bx) Yo(ax) = O.
(#)
With a = rn, A = Yo(rna) and B = -Jo(rna), both conditions above are satisfied and we have the eigenfunction
17. Just as in Problem 1 above, substitution of u(r,t) = R(r)T(t) in the wave equation
yields the separation
R" +!R'
= _ ___,_r_ = A = _a2 •
R
The r-equation has general solution
T(t) = Acosaat+Bsinaat,
and we choose B = 0, so that T'(O) = 0 (assuming, for instance, that the membrane is initially at rest). The r-equation can be written in the form
which is the parametric Bessel equation of order zero. By Problem 16, its solutions satisfying R(a)=R(b)=O are of the form Rn(x) = Yo(rna)Jo(rnX) -Jo(rna)Yo(rnX) with a = r" being one of the positive roots of Equation (#) there. This leads to a formal
series solution of the form
co
u(r,t) = I R,,(x) ( All cosrnat + B" sin rllat) ,
n=1
where the frequency of the nth term is OJn = rna = rn~T / p.
Section 10.4
593
18. We start with the substitution u(r, t) = R(r)T(t) in the heat equation. The result is given in Equations (25) and (26):
?R" +rR' + d?R = 0, T' = -dkT.
The first of these equations, together with the endpoint conditions
R(a) = R(b) = 0,
comprise the regular Sturm-Liouville problem of Problem 16. Hence its eigenvalues are given by an = rn where {rn} are the positive roots of the equation in Eq.(41) in the text. The nth eigenfunction is the function Rn(r) defined in (42). Finally the solution of
'T' , - 2kT·
1 n - -rn n IS
Tn(t) = exp( - rn2 kt),
so we get a solution of the form
00
u(r,t) = Ie" exp( -r;kt )Rn(r).
11=1
19. We want to solve the boundary value problem
(a < r < 1, Z > 0)
u(a,z) = u(b,z) = °
u(r, z) bounded as z ~ 00
u(r,O) = fer).
Just as in Problem 5, the substitution u(r, z) = R(r)Z(z) yields the equations
rR"+R'+drR = 0,
ZIt - dz = °
with separation constant A = d. When we impose the conditions R(a) = R(b) = 0 on the r-equation here, we have the Sturm-Liouville problem of Problem 16, so R(r) must be one of the eigenfunctions {R,,(r)} corresponding to the positive roots {rn} of Equation
(#) there. The general solution of ZIt = rn2Z is
and we choose B; = ° so that Zn(z) will be bounded as z ~ 00. Thus we obtain a solution of the form
594
Chapter 10
00
u(r,z) = L>II exp(-rllz)RII(r)
11=1
with the coefficients {Gil} calculated as in Problem 18.
SECTION 10.5
HIGHER-DIMENSIONAL PHENOMENA
This section provides the interested student with an opportunity to study several applications at greater depth than is afforded by the usual textbook exercises. The problem sets outlined in Section 10.5 can serve as the basis for several fairly substantial computational projects. Because these problem sets and projects are rather heavily annotated in the text, further outlines of solutions are not included in this manual. However, additional discussion - particularly regarding computer implementations - may be found in the applications manual that accompanies the text.
Section 10.4
595
APPENDIX
EXISTENCE AND UNIQUENESS OF SOLUTIONS
In Problems 1-12 we apply the iterative formula
Yn+! = b + r f(t'Yll(t)) dt
to compute successive approximations {yn(x)} to the solution of the initial value problem
Then Y3(1) ~ 0.350185, which differs by only 0.0134% from the Runge-Kutta approximation y(l) ~ 0.350232. As a denouement we may recall from the result of Problem 16 in Section 8.6 that the exact solution of our initial value problem here is