PROBLEMS ON SUMS AND INTEGRALS

Contents
1 An example 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 An Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Application to the Riemann Zeta Function . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Problems 3
2.1 Concept Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Putnam-Style Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Problems related to more advanced topics . . . . . . . . . . . . . . . . . . . . . . . . 6

1 An example
1.1 Background
This example is taken from Proofs from THE BOOK, which contains many wonderful tricks of
summation and integrals. It is definitely worth a read before the Putnam.
We have all heard about the Riemann Function
∞
X 1
ζ(s) =
ns
n=1

Our task today is to explicitly find its values at the even positive integers. Evaluating this function at
any other points is extremely difficult: with great effort people finally showed that ζ(3) is irrational,
and little else is known. In 18.112 you can explore how to define this function everywhere, and how
to evaluate this function at the negative integers.

1.2 An Identity
Our method of finding this sum comes from an unexpected place: the following identity about a
trigonometry function
X 1
π cot πx =
x+n
n∈Z

We first note that the sum, as presented, doesn’t quite make sense: it diverges. The correct way of
presenting the sum is
1 X 1 1 1 X 2x
+ + = +
x x+n x−n x x2 − n2
n∈N n∈N

Now the sum converges, and it satisfies two other important properties: it converges absolutely,
and converges uniformly on any compact interval minus the integers. If you are unfamiliar with
these two concepts, I strongly recommend reading Principles of Mathematical Analysis by Walter
Rudin, which contains a nice exposition.
Now we are supposed to prove the identity. Euler did this by a fancy bash; however, a more
beautiful proof was discovered by Herglotz, and properly named Herglotz Trick.

1
Proof. Let f, g be functions equal to LHS and RHS of the equation. We want prove that they are
equal; we achieve this by proving several properties of the function h = f − g.
(1) The function h = f − g is defined everywhere, and continuous everywhere. We only need to
show this on a neighbor of 0, and the rest is analogous. We first note that π cot πx − x1 is continuous
in the neighbor(this can be shown by Taylor series, or straight from definition). Then we note that
the rest of the sum converges uniformly, and thus is also continuous.
(2) The function h = f − g is periodic of period 1. This can be easily seen from the expression.
(3) The function h = f − g satisfy the functional equation
x x+1
2h(x) = h( ) + h( ) (1)
2 2
We can verify this for f and g individually at the non-integer points(note there is an interchange
of summation on LHS, which needs to be justified by absolute convergence). The equations carry
to the integer points by continuity.
Now we use the trick: we note that any periodic, continuous function that satisfy (1) must be
a constant! In fact, |h| must assume maximum value at some point x0 . However, from (1) we find
that if |h| has maximum at x0 , then it must assume maximum at x20 and x02+1 as well. Repeating
the argument, we find that |h| assumes maximum at a series of points approaching 0, thus it is
equal to |h(0)| everywhere. We conclude that h = 0, by noting that |h(0)| = 0.

1.3 Application to the Riemann Zeta Function

Now some of you might be able to see the connection between the identity and the zeta function.
1 X 2x
π cot πx − =
x x2 − n2
n∈N

We consider x in a neighbor of 0. Expanding RHS,

∞
X 2x XX 2
2 2
= − 2j x2j−1
x −n n
n∈N n∈N j=1

We can swap the sum by absolute convergence

∞ ∞
XX 2 2j−1 X
− x = − 2ζ(2j)x2j−1
n2j
n∈N j=1 j=1

Note this is precisely the Taylor Expansion! With one formula, we have captured the values of the
ζ function at all the positive even points!
All is left is to find the Taylor expansion of π cot πx − x1 at a neighbor of 0. But this is very
easy: we can do it by direct differentiation, or we can do it smartly by noting that
1 πx cos πx − sin πx
π cot πx − =
x x sin πx
Thus, if we let
∞
1 X
π cot πx − = aj x2j−1
x
j=1

2
Then the series aj satisfy the relation
j−1
X (−1)i π 2i+1 1 1
aj−i = (−1)j π 2j+1 ( − )
(2i + 1)! (2j)! (2j + 1)!
i=0

A convention is to write
Bj π 2j
ζ(2j) =
(2j)!
Thus the recursion becomes
j−1
X (−1)i 1 1
−2 Bj−i = (−1)j ( − )
(2i + 1)!(2j − 2i)! (2j)! (2j + 1)!
i=0

cleaning up,
j  
X 2j + 1
i
(−1) Bi = −j
2i
i=1

We can calculate some values, just for fun

π2 π4 π6
ζ(2) = , ζ(4) = , ζ(6) = ···
6 90 945

2 Problems
2.1 Concept Problems
1. Compute  
1 1 1
lim √ +√ + ··· + √ .
n→∞ 4n2 − 12 4n2 − 22 4n2 − n2
2. Find
n n
1 XX a
lim .
n→∞ n a + b2
2
a=1 b=1
R∞ −x2 √
3. Show that −∞ e dx = π.

4. Find Z 1
log x log(1 − x)dx
0

5. Exhibit a sequence aij indexed by Z2 such that

XX XX
( aij ) 6= ( aij )
i j j i

with all sums converging.

6. Exhibit a smooth(i.e. infinitely differentiable) function f : R → R such that f (x) = 1 when

x < 0 yet f (x) = 0 when x > 1.(This function is one of the most important gadgets in
Analysis)

3
2.2 Putnam-Style Problems
7. Evaluate the improper integral
1
log(1 − x)
Z
dx.
0 x

8. Determine the value of the improper integral

Z ∞
x
dx.
0 ex −1

R∞
9. (a) Show that that min(a, b) = 0 1≤a (t)1≤b (t) dt for any nonnegative real numbers a, b ≥
0. (What do you think 1≤c (t) means?)
(b) Show that if r1 , . . . , rn are nonnegative reals and x1 , . . . , xn are real numbers then
n X
X n
min(ri , rj )xi xj ≥ 0.
i=1 j=1

10. Evaluate the following:

Z ∞
x3 x5 x7 x2 x4 x6
  
x− + − + ··· 1 + 2 + 2 2 + 2 2 2 + · · · dx.
0 2 2·4 2·4·6 2 2 ·4 2 ·4 ·6

11. Show that Z 1 X

x−x dx = n−n .
0 n≥1

12. RSuppose that f is a function on the interval [1, 3] such that −1 ≤ f (x) ≤ 1 for all x and
3
1 f (x) dx = 0. Determine the largest possible value of
Z 3
f (x)
dx.
1 x

13. Let f : R → R be continuous and satisfy f (x) ≥ 1 for all x. Suppose that

f (x)f (2x) . . . f (nx) ≤ 2018n2019

for every positive integer n and x ∈ R. Must f be constant?

14. A rectangle in R2 is called great if either its width or height is an integer. Prove that if a
rectangle X can be dissected into great rectangles, then the rectangle X is itself great.

15. Compute
X 2k
.
k≥0
52k + 1

4
16. Prove that 1
n  
!
n(n+1)
Y n √
lim = e.
n→∞ k
k=0

17. Let f : R≥0 → R be a strictly decreasing continuous function such that limx→∞ f (x) = 0.
Prove that Z ∞
f (x) − f (x + 1)
dx
0 f (x)
diverges.

18. A rectangular prism X is contained within a rectangular prism Y .

(a) Is it possible the surface area of X exceeds that of Y ?

(b) Is it possible the sum of the 12 side lengths of X exceeds that of Y ?

19. For all n ≥ 1, let

n−1
X sin 2k−1
2n π
an = k−1
.
cos2 2 k
k=1 2n π cos 2n π

Find
an
lim .
n→∞ n3

20. For a, b, c > 0 prove that

1 1 1 4 4 4 12 12 12
+ + + + + ≥ + + .
a b c a+b b+c c+a 3a + b 3b + c 3c + a

21. Evaluate n
Y  1 + xn+1 x
lim .
x→1− 1 + xn
n≥0

22. Suppose that f : [0, 1]2 → R is continuous. Show that

Z 1 Z 1 2 Z 1 Z 1 2
f (x, y)dx dy + f (x, y)dy dx
0 0 0 0
Z 1Z 1 2 Z 1Z 1
≤ f (x, y) dx dy + [f (x, y)]2 dx dy.
0 0 0 0

23. hFor each positive integer k, let A(k) be the number of odd divisors of k in the interval
√ 
1, 2k . Evaluate:
∞
X A(k)
(−1)k−1 .
k
k=1

5
 24. (a) Let f (x, y, z) be a continuous real-valued function on R3 . Suppose that for every sphere
S of radius 1, the integral of f (x, y, z) over the surface of S equals 0. Must f (x, y, z) be
identically 0?
(b) What if f is required to be smooth and of compact support?
25. Let a0 = π/2, and let an = sin(an−1 ) for n ≥ 1. Determine whether a21 + a22 + a23 + · · ·
converges or not.
26. For a positive integer N , let
N
X N + 1/2 − n
fN (x) = sin((2n + 1)x).
(N + 1)(2n + 1)
n=0

Determine the smallest constant M such that fN (x) ≤ M for all N and all real x.

2.3 Problems related to more advanced topics

27. For positive integer n, let p(n) be the number of ways to partition n into the sum of some posi-
tive integers, quotient permutation of the integers. The sequence p(n) starts 1, 2, 3, 5, 7, 11, ....
(a) Prove that
log p(n)
lim √
n→∞ n
exists, and find its value.
(b)(Hard) Let the answer of (a) be C. Prove that
np(n)
lim √
n→∞ eC n
exists.
(c)(Hard)Determine the value of the limit in (b).
28. Prove that if f is a continuous function on the unit circle S 1 parametrized by θ ∈ [0, 2π), θ0
θ0
is an angle such that 2π is irrational, then
n Z 2π
1 X 1
lim f (mθ0 ) = f (θ)dθ.
n→∞ n 2π 0
m=1

Hint: Fourier expansion does not hold here, but can you do something similar?
29. Let f (x) be a smooth function with compact support on R, and fˆ be its Fourier transform
Z ∞
ˆ
f (ξ) = f (x)e−2πxξi dx.
−∞

Prove the Poisson Summation Formula

X X
fˆ(n) = f (n).
n∈Z n∈Z

Using this result, prove that for all t > 0

X 2 1 X − πn2
e−πn t = √ e t .
n∈Z
t n∈Z
This formula is useful in the analytic extension of the Riemann Zeta function.

6
30. Prove the Jacobi Triple Product Formula for |q| < 1
∞ ∞
2
Y X
(1 − q 2m )(1 + ω 2 q 2m−1 )(1 + ω −2 q 2m−1 ) = ω 2n q n .
m=1 n=−∞

Using this, either find the number of integer solutions for 12345654321 = a2 + b2 , or exhibit
the difference between the number of even and odd partitions of n(a partition n = a1 +· · ·+ak
is even iff k is even).

31. A smooth function f : C → C is harmonic if

∂2 ∂2
∆f := ( 2
+ 2 )f = 0.
∂x ∂y

(1) Prove the Mean Value Property of Harmonic functions: for any z ∈ C and r > 0, we have
Z 2π
f (z + eiθ r)dθ = 2πf (z).
0

(2) Find all harmonic functions on C that is real and positive everywhere.
(3)(Hard) Harmonic functions can be similarly defined on any open subset of the plane. Find
all harmonic functions on C\{0} that is real and positive on its domain.

32. (Hard) Let

d
X
f (z) = ci z i
i=0

be a polynomial with strictly positive coefficients(ci > 0 for all 0 ≤ i ≤ d). Let An,k be the
coefficient of z n in f k (z). Prove that there exists some K > 0(possibly depending on f ) such
that for all k > K and 1 ≤ n ≤ dk − 1, we have

A2n,k ≥ An−1,k An+1,k .