Methods of Applied Mathematics, MATH6600 Gregor Kovacic

Homework Problems

1. Under what conditions on the scalars 1 , . . . , n , are the column vectors (1, j , . . . , jn1 )T ,
j = 1, . . . , n, in Rn linearly (in)dependent?
HINT: Start with n = 1 and n = 2. Feel free to use Gaussian elimination and/or properties
of determinants.

2. Legendre polynomials are defined by the formulas

1 dn (x2 1)n
P0 (x) = 1, Pn (x) = , n = 1, 2, . . .
2n n! dxn
(i) For m < n, by repeatedly integrating by parts show that
Z 1
Pn (x) xm dx = 0,
1

and conclude that Z 1

Pn (x)Pm (x) dx = 0 if n 6= m.
1

(ii) Again by repeatedly integrating by parts, show that

Z 1 Z 1
2 (2n)!
Pn (x) dx = 2n 2
(1 x)n (1 + x)n dx,
1 2 (n!) 1
and by repeatedly integrating by parts yet again show that
Z 1
22n+1 (n!)2
(1 x)n (1 + x)n dx = .
1 (2n)!(2n + 1)
Conclude that Z 1
2
Pn2 (x) dx = .
1 2n + 1
(iii) Evaluate the n-th derivative of (1 x)n (1 + x)n using the product rule and set x = 1 to
conclude that Pn (1) = 1.
(iv) Show that

1 X
= Qn (x)un , (1)
1 2xu + u2 n=0

for some n-th order polynomials Qn (x), by expanding the left-hand side in the binomial series
in powers of (u2 2xu), using the binomial formula to expand these powers, and rearranging
the terms. (Of course, this is only valid for |u| < 1 and |x| 1.)

1
HINT: The exact coefficients in Qn (x) are not important, just show that its order is n.
(v) Integrate the left-hand side of the equation
Z 1
1 1 X
dx = Qn (x)Qm (x)un v m
2
1 2xu + u 1 2xv + v 2
1 n=0

to obtain the formula

1 1 + uv X 2
log = un v n ,
uv 1 uv n=0
2n + 1
and so conclude that Z 1
2
Qn (x)Qm (x) dx = n,m ,
1 2n + 1
where n,m denotes the Kronecker delta. Set x = 1 in equation (1), and deduce that Qn (x) =
Pn (x).
(vi) By differentiating equation (1) with respect to u, show the recursion relation

(n + 1)Pn+1 (x) (2n + 1)xPn (x) + nPn1 (x) = 0.

(vii) Differentiate the obvious relation

0
(x2 1) (x2 1)n = 2nx(x2 1)n


(n + 1) times to show that the n-th Legendre polynomial Pn (x) satisfies the differential
equation 0
(x 1)Pn0 (x) n(n + 1)Pn (x) = 0.
 2

(viii) Let C be the constant chosen such that the coefficient of the power xn in CPn (x) equals
1. By assuming the expansion of any n-th order polynomial, P (x) = xn + lower-order terms,
in terms of Legendre polynomials, show that the n-th order polynomial with the minimal
L2 -norm on the interval [1, 1, ], i.e., which minimizes the expression
Z 1
P 2 (x) dx,
1

is precisely the n-th Legendre polynomial Pn (x).

HINT: No need to evaluate C.

3. A linear transformation U on Cn is called an isometry if (U x, U y) = (x, y) for all x, y Cn .

(i) Show that U is an isometry precisely when it is unitary.
(ii) Show that U is unitary precisely when it takes any orthonormal basis into another
orthonormal basis.
(iii) Show that a linear operator A on Cn is self-adjoint precisely when it is diagonal and
real in an orthonormal basis.

2
(iv) The state of a quantum system is defined by a normalized vector x Cn , i.e., such
that kxk = 1 in any allowable basis. A physical variable is represented by a diagonalizable
operator A on Cn all of whose eigenvalues must be real. Conclude from (i) through (iii) that
A must be self-adjoint.

4. Let a 6= 0 be a fixed vector in Rn and A the operator defined by the formula Ax =

kak2 x (a, x)a, where is a scalar.
(i) Is A self-adjoint?
(ii) Find the eigenvalues and eigenvectors of A.
(iii) For what, if any, values of is A positive definite?
(iv) For what values of does A1 exist, and what is A1 ?

5. Suppose that in some inner-product space V , there exist two self-adjoint operators, X
and P , which satisfy the relation XP P X = i. Consider the operator H = P 2 + X 2 .
(i) Show that H is positive, i.e., (Hx, x) 0.
(ii) Show that XH HX = 2iP and P H HP = 2iX.
(iii) Let Ek be the eigenvalues and k the eigenfunctions of the operator H. Evaluate
((XH HX)k , l ) and use (ii) to show that (Ek El )(Xk , l ) = 2i(P k , l ). Similarly,
show that (Ek El )(P k , l ) = 2i(Xk , l ).
(iv) Derive from (iii) that either (Xk , l ) = (P k , l ) = 0 or Ek El = 2, and conclude
that the only way this statement is not contradictory is if the former holds if l 6= k 1 and
the latter if k = l 1.
(v) From the two relations you derived in (iii), derive that (Ek El 2) ((P iX)k , l ) = 0.
Conclude that the second factor fails to vanish only when El = Ek 2, and that applying
the operator P iX to an eigenfunction k results in a multiple of the eigenfunction l
corresponding to the eigenvalue smaller by 2 than that of k , and that you can choose
l = k 1. Likewise, show that applying the adjoint operator P + iX raises the eigenvalue
by 2.
(vi) To avoid contradiction with (i), show that there must exist a lowest eigenfunction, say
0 , such that (P iX)0 = 0. Multiply this equation by P + iX from the left to conclude
that E0 = 1, and further derive that En = 2n + 1.
(vii) For any matrix A Cnn define its trace as the sum of its diagonal elements, tr A = a11 +
+ ann . Show that tr A is the second coefficient in the characteristic polynomial of A and
conclude that it is an invariant independent of the basis. Also show that tr AB = tr BA for
any two matrices A, B Cnn . Using the relation they satisfy, conclude that the operators
X and P cannot be acting on a finite dimensional space.
d
(viii) Let X = x and P = i dx on the space of infinitely-differentiable functions. Deduce
that the normalized eigenfunctions are

2 /2 2 /2 dn x2
0 (x) = c0 ex and n (x) = cn ex e , n = 1, . . . ,
dxn

3
where cn normalization constants. (No need to evaluate those.)
Apart from setting all physical constants to 1, this is the quantum harmonic oscillator.

6. A linear operator P on Rn with the property P 2 = P is called a projector.

(i) Show that I P is also a projector.
(ii) Show that the only eigenvalues that P can have are 0 and 1.
(iii) Show that the image, Im P , and kernel, Ker P , of P have only the 0 vector in common,
and that P thus acts as the identity on the former and as the zero operator on the latter.
Conclude that P can always be diagonalized.
(iv) Show that P can be written in the form
p
X
P = xk zkT , p n,
k=1

where the xk s and zk s are column vectors, the xk s form a(n orthogonal) basis of Im P , the
zk s form a(n orthogonal) basis of (Ker P ) , and (xj , zk ) = jk . (We say that P is a projector
onto Im P along Ker P .) Interpret this result geometrically (i.e., draw a two-dimensional
figure) and reconcile it with the analogous representation of I P .
(v) P is an orthogonal projector if it has the form
p
X
Px = (x, xk )xk
k=1

in some orthonormal basis that contains the vectors xk . Interpret this P in light of (iv)
above, and show that a projector is self-adjoint if and only if it is orthogonal.
(vi) In class, we showed that a diagonalizable operator can be written in the form
n
X
A= j xj zj ,
j=1

where xj and j are the eigenvalues and eigenvectors of A, respectively, and zj are the
eigenvectors of the adjoint A that correspond to the eigenvalues j . Show that if p =
p+1 = = q , the corresponding piece of the matrix A which is multiplied by this
eigenvalue, P = qj=p xj zj , is a projector, and determine what subspace it projects onto and
P
along which subspace.
(vii) Let A be as in part (vi). Show that the resolvent (A I)1 is a rational function of
the parameter with simple poles at the eigenvalues, and that the projectors in (vi) are the
corresponding residues.

7. As you know from problem 2, Legendre polynomials are defined by the formulas
1 dn (z 2 1)n
P0 (z) = 1, Pn (z) = , n = 1, 2, . . .
2n n! dz n
4
(i) Use Cauchys integral formula for derivatives to derive Schlaflis formula
( 2 1)n
I
1
Pn (z) = n+1 d, n = 0, 1, 2, . . . ,
2 i C ( z)n+1
where C is any positively-oriented simple closed contour surrounding the point z.
(ii) Assume the generating function (z, u) for the Legendre polynomials to be of the form

X
(z, u) = Pn (z)un .
n=0

Use part (i), interchange summation and integration, and sum a geometric series to find that
I
1 d
(z, u) = ,
2i C z 21 ( 2 1)u
where the contour C is now taken small enough to only enclose the root, u , of the denomi-
nator that reduces to z as u 0.
(iii) Show that if f (z) = g(z)/h(z), and h(z) has a simple zero at z = z0 , then the residue
of f at z0 equals g(z0 )/h0 (z0 ).
(iv) Calculate the two roots of the denominator in the integral in (ii) and show that they are
distinct, so that the integrand has a simple pole at each. Since the contour C only encloses
the single root u , using the residue theorem, derive that
1
(z, u) = .
1 2zu + u2

8. Let
z X Bn
z
= zn.
e 1 n=0 n!

(i) Show that the radius of convergence of this series is 2.

(ii) Multiply the above series by ez 1 to show that the Bernoulli numbers Bn satisfy the
equation
       
n+1 n+1 n+1 n+1
Bn + Bn1 + Bn2 + + B0 = 0
1 2 3 n+1
for n > 0, where B0 = 1. Conclude that these numbers are rational.
(iii) Show that
z z z z
z
+ = coth ,
e 1 2 2 2
and thus obtain B2n+1 = 0 for n > 0, and

X 22n B2n
z coth z = z 2n .
n=0
(2n)!

5
What is the radius of convergence of this series?
(iv) Replace z by iz, to find

X 22n B2n 2n
z cot z = (1)n z .
n=0
(2n)!

(v) Derive the formula 2 cot 2z = cot z tan z to conclude that

2n
X
n1 2 (22n 1) B2n 2n1
tan z = (1) z .
n=1
(2n)!

Where does this series converge?

9. Consider the sequence of real functions

n
X x2
fn (x) = .
k=0
(1 + x2 )k

(i) Show that fn (x) 1 + x2 for x 6= 0 and fn (0) = 0 0.

(ii) By investigating the remainder f (x) fn (x), show that the sequence fn (x) converges
uniformly on |x| a > 0 for any a > 0, but not near x = 0.
(iii) Use plotting software (such as Matlab) to plot fn (x) for n = 10, 50, and 100. How do
fn behave?

10. Let x not be an integer, and let f (t) = cos xt for < t < . Extend f (t) periodically
in t outside the interval < t < , and then expand it in a Fourier series.
(i) Show that this series is
 
2x sin x 1 cos t cos 2t cos 3t
cos xt = + + .
2x2 x2 12 x2 22 x2 32

Convince yourself that this series represents a continuous function near t = . Setting
t = thus conclude that
 
2x 1 1 1 1
cot x = + + + + .
2x2 x2 12 x2 22 x2 32

This is the so-called partial fraction decomposition of the cotangent.

(ii) Let 0 x q < 1 for some q. Write the above formula as
 
1 2x 1 1 1
cot x = + + + .
x 12 x2 22 x2 32 x2

6
Using the Weierstrass M-test, convince yourself that the series on the right-hand side con-
verges uniformly. Integrate term-by-term between 0 and x to conclude that

x2
 
sin x X
log = log 1 2 .
x n=1
n

Interpret the series on the right-hand side as the limit of its partial sums, and invoke the
continuity of the exponential function to show that
N  
x2 x2
 Y 
sin x Y
= lim 1 2 = 1 2 .
x N
n=1
n n=1
n

Therefore,

x2
Y 
sin x = x 1 2 .
n=1
n
This is the infinite product expansion of the sine function. It can be shown that it is not
only valid for 0 x < 1, but for all complex x.

11. (i) Draw the graph of the function t(u) = u  sin u for 0  1, and convince yourself
that the following statements are true:
(a) t(u) is odd, monotonically increasing, and continuously differentiable for all real u,
(b) t(u + 2) = t(u) + 2,
(ii) Consider the equation
t = u  sin u (2)

(a) Given  1, show that there is a unique function u(t) that solves equation (2), which is
odd and monotonically increasing for all u. If  < 1 show that this function is continuously
differentiable for all t, and if  = 1, it is continuously differentiable at t 6= 2n with integer
n. (What happens at t = 2n with integer n?)
(b) Use part (b) of (i) to show that u(t) can be written as u(t) = t + f (t), where f (t + 2) =
f (t) for all real t.
HINT: To show (a), use monotonicity and continuous differentiablity of t(u).
(iii) If 0  < 1, show that the function f (t) can be expanded in a Fourier sine series,

X
f (t) = cn () sin nt,
n=1

where Z
2
cn () = f (t) sin nt dt. (3)
0

7
Calculate cn () in the following way: Write f (t) = u(t) t, and split the integrand in (3)
into a sum of two terms, the first of which is
2
Z
dn () = u(t) sin nt dt.
0
You will be able to calculate the second term easily, but for the first term, integrate by parts,
use periodicity of u(t), and make a substitution t = t(u) to get
Z
2(1)n+1 2
dn () = + cos nt u0 (t) dt
n n 0
Z
2(1)n+1 2
= + cos nt(u) du
n n 0
Z
2(1)n+1 2
= + cos n(u  sin u) du.
n n 0
(Justify all your steps.) This integral involves the Bessel function Jn (x), given by the formula
2
Z
Jn (x) = cos(nu x sin u) du.
0
Thus, express cn () in terms of algebraic expressions and a Bessel function, and u(t) as the
sum of t and a Fourier series in t with coefficients involving Bessel functions.

12. Use Fourier transforms to solve the boundary-value problem

xx + yy = 0, in < x < , y > 0,

with
= f (x) as y 0,
0 as y ,
by completing the following steps:
(i) Let Z
(x, y) = (k, y)eikx dk, (4)

and derive the equation
yy k 2 = 0. (5)
(ii) Show that the solution of equation (5) consistent with the boundary condition as y
is
(k, y) = C(k)e|k|y .
(iii) Use part (ii) and the boundary condition at y = 0 to derive that
Z
1
C(k) = f ()eik d.
2

8
(iv) From equation (4) and parts (ii) and (iii), derive that
Z Z Z Z
1 k[y+i(x)] 1
(x, y) = d f () dk e + d f () dk ek[yi(x)] ,
2 0 2 0

and conclude that Z

1 yf ()
(x, y) = d.
y 2 + (x )2

13. Let a function f (t) be defined for 0 t 2, and continued periodically elsewhere.
Consider a trigonometric interpolation of this function,
n1
X
f (t) fk eikt . (6)
k=0

(i) Assume that the interpolation (6) agrees with the exact values of the function f (t) at n
equidistant points tj = jt, j = 0, . . . , n 1, where t = 2/n, so that
n1
X
f (tj ) = fk eiktj , j = 0, . . . , n 1. (7a)
k=0

Show that
n1
1X
fl = f (tj )eiltj , j = 0, . . . , n 1. (7b)
n k=0

HINT: Substitute equation (7a) into the right-hand side of equation (7b), and reverse the
order of summation on j and k. Write out tj explicitly and note that if k 6= l, you will be
summing a geometric progression of n-th roots of unity.
(ii) Let n = 2m, and let
m1
X m1
X
2i lj/m
El = f (t2j )e and Ol = f (t2j+1 )e2i lj/m .
j=0 j=0

Using equation (7b), show that

1
fl = El + ei l/m Ol ,


n
and that Elm and Olm can be substituted for El and Ol when m l < 2m. Realizing that
ei l/m = ei (lm)/m , derive that for l = 0, . . . m, we obtain
1
fl = El + ei l/m Ol ,


n

9
and
1
fl+m = El ei l/m Ol .


n
If C(n) is the operation count needed to compute all the Fourier coefficients fl , l = 0, . . . n1,
show that these formulae imply that
n
C(n) = 2C + 2n.
2
If n = 2p , conclude recursively that the operation count equals nC(1) + 2pn, which has order
of magnitude O(n log n). This type of recursive computation comprises the fast Fourier
transform.
(iii) Let a function g(x) be defined for a x b, and continued periodically elsewhere.
Map the interval a x b linearly onto the interval 0 t 2, and derive the discrete
Fourier-transform formulas
n1 n1
X
ik (xj a) 1X
g(xj ) = gk e , gk = g(xj )eik (xj a) , (8)
k=0
n j=0

where xj = a + j(b a)/n and k = 2k/(b a).

(iv) Let n = 2m and let f (t) be defined on t and periodic with period 2. Use
this periodicity to show that the summation in formulas (7) can be made to run from m
to m 1. Likewise, show that formulae (8) for a function g(x) on the symmetric interval
L x L, periodic with period 2L, are equivalent to the equations
m1 m1
X 1 X
g(xj ) = gk eik xj , g(k ) = gk = g(xj )eik xj , (9)
k=m
2m j=m

where now xj = jx and k = k, with x = L/m and = /L. Ignore x outside

the interval L x L and outside the interval m/L m/L, and, after an
appropriate rescaling of g(k ), identify formulas (9) as Riemann sums. Let m, L in
the appropriate order, and formally derive the formulas for the Fourier transform.
Rb
14. Let P (x) be a probability density on the interval a x b, so that a P (x) dx =
Rb
1. Denote the mean of the corresponding distribution as = a xP (x) dx and its second
Rb
moment as 2 + 2 = a x2 P (x) dx, where 2 is its variance. Consider the integral equation
Z b
(x) = f (x) + K(x, y)(y) dy, (10)
a

where K(x, y) = P (y)(x + y ) and f (x) is an arbitrary function.

(i) Show that the solution of equation (10) can be written in the form
Z b
(x) = f (x) + R(x, y; )f (y) dy,
a

10
where the resolvent kernel equals
P (y)
(x + 2 ) [1 + (y )] + y .

R(x, y; ) = 2 2
(11)
1
Verify that, when f (x) = F , a constant, this solution reduces to
x + 2 2
 
(x) = F 1 + .
1 2 2

(ii) From equation (11), find the iterated kernels of the equation (10).
(iii) Using equation (11), find the characteristic values of the equation (10) and its corre-
sponding eigenfunctions n (x), as well as those of the adjoint equation, n (x). Verify that
these really satisfy the two respective equations, and that they are bi-orthogonal, i.e.,
Z b
i (x)j (x) dx = 0
a

for i 6= j.

15. Consider the equation

Z
K(x, y) + F (x + y) + K(x, z)F (z + y) dz = 0, (12a)
x

where n
X
F (x) = k ek x , k > 0, k = 1, . . . , n. (12b)
k=1

(i) Try a solution of equation (12) in the form

n
X
K(x, y) = Kk (x)ek y . (13)
k=1

Show that the functions Kk (x) satisfy the system of equations

n
X k
Kk (x) + e(k +l )x Kl (x) = k ek x . (14)
l=1
k + l

(ii) Let A be the matrix with entries

k
Akl = kl + e(k +l )x , k, l = 1, . . . , n.
k + l
Use Cramers rule to find
det A(k) (x)
Kk (x) = ,
det A(x)

11
where A(k) (x) is the matrix obtained from A(x) by replacing its k-the column by the column
of l el x , l = 1, . . . , n.
(iii) Use the result of part (ii), equation (13), and the relation
d l
l e(l +k )x = e(l +k )x
dx l + k
to obtain
d
K(x, x) = log det A(x).
dx
3
(iv) Let k = ck e8k t , ck > 0, for k = 1, . . . , n. Let
dK(x, x, t) d2
u(x, t) = 2 = 2 2 log det A(x, t),
dx dx
with the t-dependence stemming from k s. Use Maple or Mathematica to evaluate u(x, t)
for n = 1, 2, and 3. Experiment with the values of ck and k and plot u(x, t) over the
x t-plane. You should get one, two, and three hump-like waves, respectively. Larger waves
travel faster, and during collisions, the waves pass through one-another cleanly, retaining
their shapes.

16. Compute the Wronskian of the equation

1
y 00 y 0 + y = 0.
x
What appears to be wrong with it? Explain how this apparent contradiction is resolved.

17. Eulers equation reads

x2 y 00 + a1 xy 0 + a0 y = 0. (15)
Recall that assuming y = xr in equation (15) leads to the indicial equation, a quadratic
equation for r, which gives two roots, r1 and r2 . Equation (15) is a perfect counter-example
used to show what all can go wrong with the solutions of a differential equations at singular
points where the existence theorem does not hold.
(i) Where does Eulers equation (15) have such points?
(ii) Restrict the discussion to x 0 and consider the situations in which the roots r1 and
r2 corresponding to a particular Eulers equation (15) are: (a) r1 = 1, r2 = 2; (b) r1 = 1,
r2 = 1; (c) r1 = r2 = 1; (d) r1 = i, r2 = i; (e) r1 = 1 + i, r2 = 1 i; (f) r1 = 1 + i,
r2 = 1 i. Write down the solutions in each case, draw their graphs, and discuss their
behavior near the origin. Also, explain what goes wrong with the initial-value problem
y(0) = y0 , y 0 (0) = y1 in each case.

18. (i) Compute the second derivative 00 (x) of the function

2 /2
(x) = xn ex ,

12
where n is a nonnegative integer, and show that, at the leading order, this second derivative
behaves as
00 (x) x2 (x), as x .
(ii) Consider the equation for the quantum harmonic oscillator

00 + ( x2 ) = 0. (16)
2
Part (i) leads to the conjecture that the substitution (x) = ex /2 (x) may lead to simple
solutions that decay rapidly as x . Show that the resulting differential equation for
the function (x) is
00 2x + ( 1) = 0. (17)
(iii) Show that x = 0 is an ordinary point of equation (17) and assume a power-series
expansion for (x). Derive the recurrence relation for the coefficients of this power series,
and show that the series terminates if = 2n + 1, where n is a nonnegative integer. The
resulting solutions in this case are Hermite polynomials.

19. Show that the equation  

00 1
w 1+ w0 = 0
2z
has formal solutions of the form

X

z an z n
n=

for any , but that only special give a convergent series and hence genuine solutions.

20. Consider the differential equation

 
001 0 1
w + w + 1 2 w = 0. (18)
2z 2z

(i) Show that equation (18) has a regular singular point at z = 0. Assume a solution of the
form
X

w(z) = z an z n , a0 6= 0,
n=0

and find the indicial exponents 1 and 2

(ii) Determine the series for both independent solutions.

21. (i) Show that the differential equation

z 2 w00 + (1 2a)zw0 + (b2 c2 z 2c + a2 2 c2 )w = 0

has solutions w(z) = z a B (bz c ), where B is a Bessel function of order , either J or Y .

13
(ii) Show that the solutions of equation (18) and also the solutions of the Airy equation can
be written in terms of Bessel functions.

22. Bessels equation of order 1 is

 
1 0
00 1
w + w + 1 2 w = 0.
z z

Assume a solution of the form

X
1
w(z) = J1 (z) log z + z cn z n ,
n=0

and compute the coefficients cn . If you cannot find the general formula, compute at least
the first three of them.

23. Show that the extremals of the functional

Z x1
J= F (x, y, y 0 , y 00 , . . . , y (n) ) dx
x0

among 2n-times continuously differentiable functions are given by the equation

d d2 dn
Fy Fy0 + 2 Fy00 . . . + (1)n n Fy(n) = 0.
dx dx dx

HINT: In your derivation, use the function (x) = (x c + )2n (x c )2n for |x c| <
and (x) = 0 elsewhere.

The problems and their numbers will change, so please make sure that you have downloaded
the latest version of this document.
Be on the lookout for typos!

14