Math221: HW# 7 solutions

Andy Royston
November 17, 2005
11.3.13 let x = eu . Then ln x1 = u, x2 = e2u , and dx = e2u du. Furthermore, when x = 0,
u , and when x = 1, u = 0. Hence
Z

x
0

1
ln
x

3

dx =

2u 3

u (e

du) =

e3u u3 du.

(1)

In the last step I used the minus sign to flip the limits of integration. Note that, if we want to
apply the definition of the Gamma function to do an integral, it is necessary that the limits of the
integral go from 0 to . If they dont, its not a Gamma function. Also, its necessary that we
have et in the integrand, where t is the integration variable. In this case, we do not, so we need
to make another change of variables. Let t = 3u du = dt/3. Note that the limits do not change:
Z
Z
Z
1 t 3
1 t 41
1
3u 3
e u du = 4
e t dt = 4
e t dt = 4 (4).
(2)
3 0
3 0
3
0
In this case we can simplify the result, using Gamma(n) = (n 1)!, for n
3
Z 1 
2
1
3!
2
x ln
dx = 4 = .
x
3
27
0

N. Thus

(3)

11.3.16 The initial conditions are x(0) = 1, x(0)

= 0. The Lagrangian is
1
1
L = T V = mx 2 m ln x.
2
2

(4)

The Euler-Lagrange equation is

 
d L
L

=0
dt x
x

d
m
(mx)
( ) = 0
dt
2x
1
x +
= 0.
2x

(5)

This is the second order differential equation that we need to solve. It is of the type ddt2x = F (x),
1
. See formulas (7.14) and (7.15), in section 7 of chapter 8. It can be integrated
with F (x) = 2x
once if we multiply each side by x.
The result is
Z
Z
dx
1 2
1
x = F (x)dx =
+ c = ln x + c, or x 2 = ln x + c.
(6)
2
2x
2
This gives us x.
We still need to integrate again to get x(t). First, however, use the initial conditions
to solve for the integration constant c. At time t = 0 we know that x = 1, x = 0. Plugging these
in,
0 = ln 1 + c = 0 + c

c = 0.

(7)

Then
1
x = ln x = ln
x
2

r
1
1
x = ln ln
x
x
dx
q
= dt.
ln x1

(8)

Note that since we know x < 1, ln x is negative, so ln x = ln x1 is positive. Furthermore, we chose

the minus sign from taking the square root because we know that the particles velocity is negative;
its traveling from x = 1 to the origin. Now, let the time T be the time it takes for the particle to
reach the origin. Then
Z 1
Z T
Z 0
dx
dx
q
q
=
.
(9)
T =
dt =
0
0
1
ln x1
ln x1
To do the x integral, make the change of variable x = eu , so that ln x1 = u, dx = eu du. Also,
x = 0 u = , x = 1 u = 0. Then
Z
T =
0

Z 0
Z
eu du
dx
q

=
=
eu u1/2 du
u
1

0
ln x
Z

1
=
eu u1/21 du = ( ) = .
2
0

(10)

11.5.4 We want to prove

1 3 (2n 1)
(2n)!

=
.
2n
4n n!
Use the fundamental property (p + 1) = p(p) for any p > 0. Then
(n + 1/2) =

(11)

(n + 1/2) = ((n 1/2) + 1) = (n 1/2)(n 1/2).

(12)

Then just keep going:

(n 1/2)(n 1/2) = (n 1/2)((n 3/2) + 1) = (n 1/2)(n 3/2)(n 3/2)
= (n 1/2)(n 3/2)((n 5/2) + 1) = (n 1/2)(n 3/2)(n 5/2)(n 5/2) = . . . .

(13)

Now, since n is some fixed integer, this process must come to an end. Eventually, I have to get
down to (1/2). Then at this point,I cant go any further. (p + 1) = p(p) only holds for p > 0.
However, I do know that (1/2) = . Thus,
1
(n + 1/2) = (n 1/2)(n 3/2)(n 5/2) (3/2)(1/2)( )
2
(n 1/2)(n 3/2)(n 5/2) (3/2)(1/2) .
(14)

Now, observe that this is a product of precisely n terms multiplying the . Multiply and divide
by 2n . In the numerator, distribute a factor of 2 to each term. Then
1
(2n 1)(2n 3)(2n 5) 3 1
, or
(n + ) =
2
2n
1 3 (2n 1)
1
(n + ) =
,
2
2n

(15)

which is the first equality. For the second equality, note that the numerator above is a product of
all odd numbers from 1 to 2n 1. Let us multiply and divide by the product of all even numbers
from 2 to 2n. Then the top will just be the product of all integers from 1 to 2n, otherwise known
as (2n)!:
1 2 3 (2n 2)(2n 1)(2n)
1 3 (2n 1)
=
n
2
2n 2 4 (2n)

(2n)!
.
= n
2 2 4 (2n)

(16)

Now the product of even numbers from 2 to (2n) has n terms. Pull out a 2 from each:
2 4 (2n) = 2n (1 2 n) = 2n n!.

(17)

Hence, noting 2n 2n = 22n = (22 )n = 4n ,

1 3 (2n 1)
(2n)!
(2n)!
= n n
= n
,
n
2
2 2 n!
4 n!
which is the second equality.
3

(18)

11.7.2 We have
/2

Z
0

Z /2
p
1
3
sin3/2 x cos1/2 xdx.
sin x cos xdx = 2
2
0

(19)

Let 2p 1 = 3/2 p = 5/4 and 2q 1 = 1/2 q = 3/4 and compare with formula (6.4) of the
text. Then
1
2
2

/2

1 5 3
sin3/2 x cos1/2 xdx = B( , ).
2 4 4

(20)

Writing this in terms of Gamma functions,

1 5 3
1 ( 54 )( 34 )
1 5
3
1 ( 54 )( 43 )
B( , ) =
= ( )( ),
5
3 =
2 4 4
2 ( 4 + 4 )
2 (2)
2 4
4

(21)

where in the last equality I used (2) = 1! = 1. Now there is some trickery that can be done. Write
5
1
1 1
( ) = (1 + ) = ( ).
4
4
4 4

(22)

Then
1 5
3
1 1
3
( )( ) = ( )( ).
(23)
2 4
4
8 4
4
The reason I did this is because now we have it in the form (p)(1 p), where p = 1/4. We can
use the mystical result, equation (5.4):

sin (p)

(p)(1 p) =

1
3

( )( ) =
= = 2.
4
4
sin (/4)
1/ 2

(24)

Hence, the whole result is

1 5 3
1
B( , ) = 2 0.55536.
2 4 4
8

(25)

11.7.7 Observe that

Z

/2

d
1

= 2
2
sin

/2

sin2p1 () cos2q1 ()d,

(26)

if p = 1/4, q = 1/2. Hence

Z
0

/2

d
1 1 1
1 ( 14 )( 21 )
( 14 )

= B( , ) =
=
.
2 4 2
2 ( 43 )
2 ( 34 )
sin

(27)

I dont know any way to simplify this further, and neither does Steven Wolfram, so this is the final
answer. Numerically, it turns out to be
4

( 14 )
2.62206.
2 ( 34 )

(28)

11.7.9 First we show

B(n, 12 )
B(n, n) = 2n1 .
2
Start with the trig expression for B(n, n) and use the trig identity as suggested:
Z

/2

B(n, n) = 2

sin

2n1

() cos

2n1

/2

(sin cos )2n1 d

()d = 2

(29)

/2

=2
0

2
1
( sin (2))2n1 d = 2n1
2
2

/2

(sin (2))2n1 d.

(30)

Now lets change variables, = 2. Then d = d/2, and note that the upper limit changes: when
= /2, = . Thus,
2
22n1

/2
2n1

(sin (2))

d =

1
22n1

2n1

(sin )

d =

1
22n1

(sin )2n1 (cos )21/21 d. (31)

Note in the last step that (cos )21/21 = (cos )0 = 1, so I didnt do anything. Now, this looks like
a Beta function except that the upper limit is wrong. But observe that sin is symmetric about
/2 from 0 to . The area under the graph from 0 to /2 equals the area under the graph from
/2 to . Hence, this is also true of (sin )2n1 . Then the integral from 0 to is just twice the
integral from 0 to /2. The result follows:

B(n, n) =

1
22n1

2n1

(sin )

21/21

(cos )

d =

1
22n1

Z
2

/2

(sin )2n1 (cos )21/21 d

1
22n1

1
B(n, ).
2

(32)

From this result, we can derive the duplication formula for the Gamma function. Simply use
the formula
B(p, q) =

(p)(q)
(p + q)

(33)

on each side of the above result. On the left side,

B(n, n) =

(n)(n)
.
(2n)
5

(34)

On the right side

1 (n)( 12 )

(n)
1
)
=
=
.
B(n,
22n1
2
22n1 (n + 1/2)
22n1 (n + 1/2)
1

(35)

Now set them equal and get the result:

(n)(n)

(n)
= 2n1
(2n)
2
(n + 1/2)

(2n) =

22n1 (n)(n + 1/2)

(36)

Finally, let us check this formula for n = 1/4. On the left side,

1
(2n) (2 ) = (1/2) = .
4
On the right side, we will need the result (p)(1 p) = / sin (p). We have
221/41 (1/4)(3/4)
1

2 sin (/4)

1
= 2 = .
2

(37)

(38)

So the formula checks out.

11.11.4 Use Stirlings formula in the numerator and denominator. In the limit n , it becomes
exact.
p

(2n)2n e2n 2(2n) n

(2n)! n

lim
= lim
n 22n (n!)2
n
22n (nn en 2n)2

22n n2n e2n 2n

1
1
= lim 2n 2n 2n
= lim = .
n 2 n e
n
2n

(39)

11.11.5 We have
(n + 32 )
((n + 1/2) + 1/2)

lim
= lim
n
n(n + 1) n
n(n + 1)
r
p
(n + 1/2)n+1/2 e(n+1/2) 2(n + 1/2)
(n + 1/2)n+1/2 1/2 n + 1/2
= lim
= lim
e
n n
n
n
nn+1/2
n
nn e
2n

n+1/2+1/2
n + 1/2
1
= lim e1/2
= e1/2 lim (1 + )n+1 .
n
n
n
2n
6

(40)

For the last limit, put it in a more standard form:

lim (1 +

1 n+1
1
1
1
)
= lim (1 + )n lim (1 + ) = lim (1 + )n
n
2n
2nr n
2n r n
2n
1
1
= lim (1 + )2n =
lim (1 + )2n .
n
n
2n
2n

(41)

Here I pulled off one factor of (1 + 1/2n), and took its limit, which is 1. Then I introduced the
square root so that the remaining limit has the form

x
1
lim 1 +
,
(42)
x
x
(just let x = 2n). This limit is something we all supposedly learned in high school.... The answer
is e. If you just quote this or look it up, thats fine by me, but its actually pretty easy to prove.
The trick is write it as an exponential of a log:

x
1
x
= lim eln ((1+1/x) ) = lim ex ln (1+1/x) .
(43)
lim 1 +
x
x
x
x
Now bring the limit in the exponential; you can do thisbring the limit inside the functionwith
any continuous function. (Thats what it really means to be continuous). Then observe that, using
the Taylor series ln (1 + ) =  + O(2 ), for  = 1/x small,
lim x ln (1 + 1/x) = lim x(1/x + O((1/x)2 ))

= lim (1 + O(1/x)) = 1.

(44)

Therefore, the thing in the exponential goes to 1, and we get

x

1
lim 1 +
= e.
x
x

(45)

Plugging this result into our problem,

(n + 32 )
1
= e1/2 lim (1 + )n+1 = e1/2 e = 1.
lim
n
n
2n
n(n + 1)

(46)

12.1.2 Lets first solve the equation by pedestrian methods:

y 0 = 3x2 y

dy
= 3x2 dx
y

ln y = x3 + c

y(x) = Aex .

(47)

Now for the power series solution, assume

y(x) =

an x n = a0 + a1 x + a2 x 2 + a3 x 3 +

(48)

n=0

for some coefficients an that we need to solve for. Then

3x2 y = 3a0 x2 + 3a1 x3 + 3a2 x4 + + 3an2 xn + .

(49)

y 0 = a1 + 2a2 x + 3a3 x2 + + (n + 1)an+1 xn + .

(50)

The derivative is

Now write out y 0 3x2 y = 0, grouping powers of x.

y 0 3x2 y = a1 + 2a2 x + (3a3 3a0 )x2 + (4a4 3a1 )x3 +
+((n + 1)an+1 3an2 )xn + = 0.

(51)

Now heres why we did this. The above equation has to hold for all x. (Its just the original
differential equation, with our guess solution plugged in). Since the xn are linearly independent
functions for each n = 0, 1, 2, . . ., the only way the whole thing will be zero is if the coefficient of
every power of x is zero. Thus, we get a list of equations:

a1 = 0,
2a2 = 0,
3(a3 a0 ) = 0,
(4a4 3a1 ) = 0,
(5a5 3a2 ) = 0,
(6a6 3a3 ) = 0,
...
((n + 1)an+1 3an2 ) = 0,
...

(52)

In particular, we learn that a1 = a2 = 0, but a3 = a0 . Then a4 a1 = 0 and a5 a2 = 0, but

a6 = 12 a3 = 12 a0 . This pattern continues; the general term tells us that a coefficient an is always
connected to three back, an3 . We will have a7 = a8 = 0, but
9a9 = 3a6

1
1
1
a9 = a6 =
a3 =
a0 .
3
32
321

You can now guess the pattern:

8

(53)

1
a0 ,
a3m1 = a3m2 = 0, m .
(54)
m(m 1) 2 1
Thus we have determined all of the coefficients in terms of a single undetermined one, a0 . This
is expected. We have one undetermined coefficient because we started with a first order equation.
Let us now write down the solution:
a3m =

y(x) =

an x = a0 +

n=0

a3m x

3m

m=1

= a0

X
1 3m
x
= a0 + a0
m!
m=1

X
1 3 m
3
(x ) = a0 ex .
m!
m=0

(55)

Hence we get the correct result, with a0 serving as A, the arbitrary constant.
12.1.7 The differential equation is x2 y 00 3xy 0 + 3y = 0. This is a Cauchy type equation. To get
the solution, try y = xk . Then
x2 k(k 1)xk2 3xkxk1 + 3xk = (k(k 1) 3k + 3)xk = 0
k 2 4k + 3 = (k 3)(k 1) = 0 k = 1, 3.

(56)

Therefore the general solution is

y(x) = Ax + Bx3 .

(57)

Lets try it by power series.

y = a0 + a1 x + a2 x 2 + a3 x 3 + a4 x 4 +
y 0 = a1 + 2a2 x + 3a3 x2 + 4a4 x3 + ,
y 00 = 2a2 + 3 2a3 x + 4 3a4 x2 + 5 4a5 x3 + .

(58)

x2 y 00 = 2a2 x2 + 3 2a3 x3 + 4 3a4 x4 + + n(n 1)an xn + ,

3xy 0 = 3a1 x + (3)2a2 x2 + (3)3a3 x3 + + (3)nan xn + .

(59)

Then

Now write down the differential equation term by term in x:

x2 y 00 3xy 0 + 3y = 3a0 + (3 + 3)a1 x + (2 1 3 2 + 3)a2 x2 + (3 2 3 3 + 3)a3 x3 +
+(n(n 1) 3n + 3)an xn + = 0 (60)
9

Every coefficient must vanish. Look at the general term, which holds for all n:
(n(n 1) 3n + 3)an = (n 3)(n 1)an = 0,

n = 0, 1, 2, . . . .

(61)

If n = 1, 3 then this is satisfied, and there are no restrictions on a1 , a3 . However, for all other n,
we must require an = 0. Hence, the solution is
y(x) =

an x n = a1 x + a3 x 3 .

(62)

n=0

This is the solution we had before with a1 = A, a3 = B.

12.1.8 The equation is (x2 + 2x)y 00 2(x + 1)y 0 + 2y = 0. I actually dont see a way to solve this,
except by power series. So lets find the power series solution first and then check the solutions we
get. y, y 0 , and y 00 are as in equation (58) above. All the terms we need are
x2 y 00 = 2 1a2 x2 + 3 2a3 x3 + 4 3a4 x4 + ,
2xy 00 = (2)2 1a2 x + (2)3 2a3 x2 + (2)4 3a4 x3 + ,
2xy 0 = (2)a1 x + (2)2a2 x2 + (2)3a3 x3 + (2)4a4 x4 + ,
2y 0 = (2)a1 + (2)2a2 x + (2)3a3 x2 + (2)4a4 x3 + ,
2y = 2a0 + 2a1 x + 2a2 x2 + 2a3 x3 + 2a4 x4 + .

(63)

Adding everything up, grouping in powers of x,

(2a1 + 2a0 ) + ((2)2 1a2 2(a1 + 2a2 ) + 2a1 )x
+(2 1a2 + (2)3 2a3 2(2a2 + 3a3 ) + 2a2 )x2 + (3 2a3 + (2)4 3a4 2(3a3 + 4a4 ) + 2a3 )x3
+ + (n(n 1)an + (2)(n + 1)nan+1 2(nan + (n + 1)an+1 ) + 2an )xn + = 0. (64)
The general term actually gives the correct form for all n = 0, 1, 2, . . .. Hence, we require
(n(n 1)an + (2)(n + 1)nan+1 2(nan + (n + 1)an+1 ) + 2an ) = 0, n = 0, 1, . . .
(n(n 1) 2n + 2)an + (2n(n + 1) 2(n + 1))an+1 = 0
(n2 3n + 2)an + 2(n + 1)(n 1)an+1 = 0
(n 2)(n 1)an + 2(n + 1)(n 1)an+1 = 0.

(65)

Now, as long as n 6= 1, we can divide out the (n 1) and obtain

an+1 =

(n 2)
an ,
2(n + 1)
10

n 6= 1.

(66)

For n = 0 we have a0 = a1 . For n = 1, we learn nothing about the as. Equation (65) is
automatically satisfied. Now, for n = 2, the equation says a3 = 0a2 = 0. Then, from this,
1
0 = 0,
a5 a4 = 0, . . . .
(67)
24
That is, all of the an , starting with a3 are 0. Note that we never learn anything about a2 . Summarizing,
a4 =

a0 = a1 = arbitrary const.
a2 = arbitrary const.
an = 0, n 3.

(68)

Hence, y(x) is given by

y(x) =

an xn = a0 (1 + x) + a2 x2 ,

(69)

n=0

where a0 , a2 are arbitrary constants. You can plug in each of these functions, (1 + x) and x2 , and
verify that they are indeed solutions to the differential equation.
12.2.1 Legendres equation is (1 x2 )y 00 2xy 0 + l(l + 1)y = 0. The general series solution is


l(l + 1) 2 l(l + 1)(l 2)(l + 3) 4
x +
x +
y = a0 1
2!
4!


(l 1)(l + 2) 3 (l 1)(l + 2)(l 3)(l + 4) 5
x +
x + .
+a1 x
3!
5!

(70)

When l is an integer, one of these series will terminate after a finite number of terms. These special
solutions are the Legendre polynomials Pl (x). To find them, just see which series is going to stop,
and write down all the nonzero terms. For P2 (x), look at the top series. At the third term, and in
all subsequent terms, there is the factor (l 2), which will vanish. Therefore we are interested in
the solution


2(2 + 1) 2
y(x) = a0 1
x = a0 (1 3x2 ).
(71)
2!
In order to make it the Legendre polynomial P2 (x), we fix the constant a0 by demanding P2 (1) = 1.
This gives
1 = a0 (1 3)

Hence,
11

1
a0 = .
2

(72)

1
P2 (x) = (3x2 1).
(73)
2
For the next one, l = 3, look at the lower series. The third term, and all terms afterwards, have
the factor (l 3). Therefore, we are interested in


5
(3 1)(3 + 2) 3
x = a1 (x x3 ).
(74)
y(x) = a1 x
3!
3
Normalize by P3 (1) = 1:
1 = a1 (1 5/3) = a1

2
3

3
a1 = .
2

(75)

Thus
3 5
1
P3 (x) = ( x3 x) = (5x3 3x).
(76)
2 3
2
Finally, for l = 4, the top series will terminate at the fourth term. The fourth term is given by
l(l + 1)(l 2)(l + 3)(l 4)(l + 5) 6
x.
(77)
6!
You can see this by the recursion relation, formula (2.6) of the text, or just observe the pattern.
The point is that it, and every term after it, contain the factor (l 4), and so vanish. So write

y(x) = a0


4(4 + 1) 2 4(4 + 1)(4 2)(4 + 3) 4
x
x
1
2!
4!
35
= a0 (1 10x2 + x4 ).
3

(78)

Normalizing,

1 = a0 (1 10 +

35
a0
8
) = (3 30 + 35) = a0
3
3
3

3
a0 = .
8

(79)

Hence,
35
1
3
P4 (x) = (1 10x2 + x4 ) = (3 30x2 + 35x4 ).
8
3
8

12

(80)

12.2.2 Observe that in the previous problem, we always used the top series for l even and the
bottom series for l odd. This will always be the case because the general term in the top series has
structure
l(l 2)(l 4) (l + 1)(l + 3) ,

(81)

while the general term in the bottom series has structure

(l 1)(l 3)(l 5) (l + 2)(l + 4) .

(82)

So if l is even, (and remember, the Legendre polynomials are only defined for l 0), then the top
series will eventually terminate, while the bottom does not. On the other hand, if l is odd, the
bottom series will be the one that terminates.
Now, the top series contains only even powers of x, so it is an even function. The bottom series
contains only odd powers of x, so it is an odd function. Therefore we learn in general

Pl (x) = Pl (x), l even;

Pl (x) = Pl (x), l odd.

(83)

Now apply this result to x = 1. There we know, (by definition really), that Pl (1) = 1, for any l,
even or odd. Hence,

1 = Pl (1) = Pl (1), l even;

1 = Pl (1) = Pl (1), l odd.

(84)

This may be summarized as

Pl (1) = (1)l .

13

(85)