Sie sind auf Seite 1von 53

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

The Basel Problem


Numerous Proofs
Brendan W. Sullivan
Carnegie Mellon University
Math Grad Student Seminar

April 11, 2013

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Abstract
The Basel Problem was first posed in 1644 and remained
open for 90 years, until Euler made his first waves in the
mathematical community by solving it. During his life, he
would present three different solutions to the problem,
P
which asks for an evaluation of the infinite series k=1 k12 .
Since then, people have continually looked for new,
interesting, and enlightening approaches to this same
problem. Here, we present 5 different solutions, drawing
from such diverse areas as complex analysis, calculus,
probability, and Hilbert space theory. Along the way, well
give some indication of the problems intrigue and
applicability.

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

0 Introduction

History
Intrigue
1 Proof: sin x and LHpital
2 Proof: sin x and Maclaurin

Brendan W. Sullivan
The Basel Problem

3
4
5
6

Pf 5

References

Analysis: sin x as an
infinite product
Proof: Integral on [0, 1]2
Proof: L2 [0, 1] and Parseval
Proof: Probability Densities
References

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

History

Pietro Mengoli
Italian mathematician and clergyman (16261686) [5]
PhDs in math and civil/canon law
Assumed math chair at Bologna after adviser Cavalieri died
Priest in the parish of Santa Maria Maddelena in Bologna
Known (nowadays) for work in infinite series:
Proved: harmonic series diverges, alternating harmonic
series sums to ln 2, Wallis product for is correct
Developed many results in limits and sums that laid
groundwork for Newton/Leibniz
Wrote in abstruse Latin; Leibniz was influenced by him [6]

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

History

Problem Statement & Early Work


Mengoli posed the Basel Problem in 1644.
Find the numerical value of:

X
1
1
1
1 1
+
+ =
1+ + +
4 9 16 25
k2
k=1

Wallis, 1655: I know it to 3 decimal places.


Jakob Bernoulli, 1689: Its less than 2. Help us out!
Johann/Daniel Bernoulli, 1721: Its about 85 .
Goldbach, 1721ish: Its between

41
35

= 1.64 and

5
3

= 1.6.

Leibniz, DeMoivre: . . . . . . ???


Part of the difficulty is that the series converges slowly:
n = 102 1 place, n = 103 2 places, n = 105 4 places,
n = 106 5 places

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Intrigue

Euler Emerges!
Since this had stumped so many brilliant minds, Eulers
solution in 1735 (at age 28) brought him immediate fame.
He was born in Basel. (Problem name comes from
publishing location of Jakob Bernoullis Tractatus de
seriebus infinitis, though.)
Studied under Johann Bernoulli, starting 1721.
Was working on it by 1728, calculating partial sums.
Published a more rigorous proof in 1741, and a third in 1755
His techniques inspired Weierstrass (to rigorize his methods
and develop analysis) and Riemann (to develop the zeta
function and the Prime Number Theorem) in the 1800s
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Intrigue

Riemmann function: definition


(s) =

X
1
1
1
1
= s + s + s +
s
n
1
2
3

n=1

for <(s) > 1

Non-series definition:
(s) = 2s s1 sin
where

 s 
2

(z) =

(1 s) (1 s)

tz1 et dt

is the analytic extension of the factorial function to C.


Notice k N (2k) = 0; these are trivial roots.

Riemann hypothesis: <z =


Brendan W. Sullivan
The Basel Problem

1
2

for every nontrivial root.

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Intrigue

Riemmann function: applications


If we knew (s), we could evaluate
Z
X

f (x) = Li(x)
Li(x ) log(2) +

where Li(x) =

R
0

dx
log(x)

and

dt
t(t2 1) log(t)

is over nontrivial roots of (s).

This would help us find the primes function (x) by


 1 

1 
(x) = f (x) f x1/2 f x1/3
2
3
Riemann was likely motivated by Prime Number Theorem:
lim

Brendan W. Sullivan
The Basel Problem

(n)
=1
n/ log n

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Intrigue

Eulers work on
Eulers 1741 proof actually adapted his earlier method to find

k N (2k) =

(1)k1 (2)2k
B2k
2(2k)!

where B2k is the Bernoulli Number, defined by


m1
X m
Bk
Bm = 1
k mk+1
k=0

with B0 = 1 and B1 =

12 .

Interestingly, not much is known about (2k + 1).


Closed form for (3) is open! (Its irrational [1], proven 1979.)
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: History
LHpital published Analyse des infiniment petits in 1696.
Standard calc text for many years, until . . .
Euler published Institutiones calculi differentialis in 1755.
New standard! [7]
Eulers book includes a discussion of indeterminate forms.
He gave no credit to LHpital (who, in turn, gave no credit
to Johann Bernoulli), but he did state and prove the rule,
and provide several striking examples, including
xx x
lim
= 2
x1 1 x + ln x

and

n
X
k=1

k=

n(n + 1)
2

and . . .
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Sketch

1. Write sin x as an infinite product of linear factors


(More on this later . . . )
2. Take log of both sides to get a sum
3. Differentiate
4. Make a surprising change of variables
5. Plug in x = 0 and use LHpital thrice
6. Sit back and smile smugly at your brilliance

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Details





t
t
t
t
1+
1
1+

sin t = t 1

2
2
since sin t has roots precisely at t Z



2y
2+y
sin(y) = y (1 y) (1 + y)

2
2




 4 y2
9 y2
= y 1 y 2

4
9


ln(sin(y)) = ln + ln y + ln 1 y 2 + ln 4 y 2 ln 4 +
Differentiate with respect to y . . .
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Details



ln(sin(y)) = ln + ln y + ln 1 y 2 + ln 4 y 2 ln 4 +
1
2y
cos(y)
2y
2y
=


2
2
sin(y)
y 1y
4y
9 y2
1
1
1
1
1
cos(y)
+
+
+
+ = 2
2
2
2
y 1y
4y
9y
2y
2y sin(y)
2
2
COV: y = ix y = x
1
1
1
1
cos(ix)
+
+
+ = 2 +
1 + x2 4 + x2 9 + x2
2x
2ix sin(ix)

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Details

Use Eulers Formula (hes everywhere!) to find:




1
iz
iz
i e2iz + 1
cos(z)
2 e +e
= 1 iz
=
iz )
sin(z)
e2iz 1
2i (e e


e2x 1 + 2
i e2x + 1

cos(ix)
=
2x
=

2ix sin(ix)
2ix
e
1
2x
e2x 1

=
+
2x
2x x (e
1)
Substitute this back into the previous equation . . .

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Details

+ 2x
2x
x(e
1)

}|
{
cos(ix)
1
1
1
1
+
+
+ = 2 +
1 + x2 4 + x2 9 + x2
2x
2ix sin(ix)

x 1
+
=
2
2x
1)
2x
x (e
2x
2x
xe
e
+ x + 1
=
2
2x
2x e
2x2
z

Plug in x = 0:
LHS is the desired sum,

1
k=1 k2 .

RHS is 00 . LHpital to the rescue!


Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Details
xe2x e2x + x + 1
e2x + 2 2 xe2x

7
2x2 e2x 2x2
4xe2x + 4x2 e2x 4x
3 xe2x
7 2x
e
+ 4xe2x + 2 2 x2 e2x 1
3
7
4 + 4 2 x + 2e2x
3
7
4 + 2
2
7
6
Everything works! Because Euler said so. And it does.
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 1: Summary
1.
2.
3.
4.
5.
6.

Write sin x as an infinite product of linear factors


Take log of both sides to get a sum
Differentiate
Make a surprising change of variables
Plug in x = 0 and use LHpital thrice
Sit back and smile smugly at your brilliance Cite Euler
What a marvelous derivation it was. It boasted an all-star cast of
transcendental functions: sines, cosines, logs, and exponentials. It ranged
from the real to the complex and back again. It featured LHpitals Rule in
a starring role. Of course, none of this would have happened without the
fluid imagination of Leonhard Euler, symbol manipulator extraordinaire.
William Dunham [7]

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: History & Sketch

Eulers (and the worlds) first proof, from 1735.


One of the easiest methods to remember.
1. Write sin x as an infinite product of linear factors
(More on this later . . . )
2. Also find the Maclaurin series for sin x
3. Compare the coefficients of x3
4. Marvel at the coincidence

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: Details
sin(x) = x 1 x2




x2
x2
1

1
4
9

= x


1 1
1
+ x3 1 + + +
+
4 9 16


1
1
1
5
+ x
+
+ +
+
14 19
49
+

(x)3 (x)5
+

3!
5!
3
5 5
= x x3 +
x
6
120

sin(x) = x

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: Details
sin(x) = x 1 x2




x2
x2
1

1
4
9

= x


1 1
1
x3 1 + + +
+
4 9 16


1
1
1
5
+ x
+
+ +
+
14 19
49

(x)3 (x)5
+

3!
5!
3
5 5
= x x3 +
x
6
120

sin(x) = x

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: Details

Thus,

1
3
1 1
+ =
1+ + +
4 9 16
6


Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: Details

Thus,
1+

Brendan W. Sullivan
The Basel Problem

1 1
1
2
+ +
+ =
4 9 16
6

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 2: Summary
1. Write sin x as an infinite product of linear factors
2. Also find the Maclaurin series for sin x
3. Compare the coefficients of x3
Not particularly challenging, so to speak. You could present this
to a Calc II class and be convincing!
Hints at the deeper relationships between products/series.
The actual validity depends heavily on complex analysis, and
would only be officially resolved in the mid 1800s by Weierstrass.
Main Question: Why can we factor sin x into linear terms?
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Analysis: sin x as an infinite product

Weierstrass Factorization Theorem


Let f be an entire function and let {an } be the nonzero zeros of
f repeated according to multiplicity. Suppose f has a zero at
z = 0 of order m 0 (where order 0 means f (0) 6= 0). Then g
an entire function and a sequence of integers {pn } such that
 

Y
z
m
f (z) = z exp(g(z))
Epn
an
n=1

where
En (y) =

(
(1 y)

(1 y) exp

y1
1

y2
2

+ +

yn
n

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

This is a direct generalization of the Fundamental Theorem of


Algebra. It turns out that for sin(x), the sequence pn = 1 and
the function g(z) = log(z) works.
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: History
Published by Tom Apostol in 1983 in Mathematical
Intelligencer. Two page proof, with profile pic:

Cites Aperys 1979 proof that (2), (3)


/ Q. Specifically, a
shorter proof by Beukers [4] uses integrals.
This evaluation has been presented by the author for a
number of years in elementary calculus courses, but does
not seem to be recorded in the literature.
Simple, clear exposition, but no indication of insight.
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Sketch

1. Consider

R1R1
0

1
0 1xy

dx dy

2. Observe it evaluates to (2) by writing fraction as infinite


series and evaluating term by term
3. Evaluate another way by rotating [0, 1]2 clockwise 45
4. Make some trig subs, but nothing advanced
5. Applaud this kind of ingenuity 250 years after the fact

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
Since

1
1r

= 1 + r + r2 + for |r| < 1

1Z 1

1
dx dy =
1 xy
=

1Z 1

(xy)n dx dy

0 n0
1Z 1

XZ

n0 0

xn y n dx dy

0
1

1  n+1 1 n
x
y dy
0
n+1
n0 0
X 1
X
1
=
=
2
(n + 1)
n2
=

XZ

n0

Brendan W. Sullivan
The Basel Problem

n1

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
y
v

v =1u

v=u

1
2

S
B

1
2

x
1

x+y
2
x=uv

u=

1
1
=
1 xy
1 u2 + v 2
Brendan W. Sullivan
The Basel Problem

12

yx
2
y =u+v
v=

x

J = u
y
u

x
v
y
v



1 1
=2
=
1 1

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
v

v =1u

v=u

1
2
S1
S10

S2
1
0
2 S2

S
u
1

12

x
B

x
1
1
dA = 2
dA
1 xy
1 u2 + v 2
S
Z 1/2 Z u
o
1
=4
dv
du
I1
2
2
0
0 1u +v
Z 1 Z 1u
o
1
+4
dv
du
I2
1 u2 + v 2
1/2 0

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
In general,
Z z
0

 z

 
t
1
dt
1
1
1 z
tan
=
tan
=
a2 + t2
a
a 0 a
a

Use this in I1 , where a2 = (1 u2 ) and z = u


Use this in I2 , where a2 = (1 u2 ) and z = 1 u
1/2


u
du

I1 = 4
tan

2
1u
1 u2
0


Z 1
1u
du
I2 = 4
tan1

2
1u
1 u2
1/2
Z

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
Evaluating I1 : Let u = sin du = cos d
Z

1/2

I1 = 4

tan1

/6

=4
0

/6

=4

tan1

du

1 u2
!
sin
cos d
p
p
2
1 sin
1 sin2

u
1 u2

tan1 (tan ) d

Z
=4
0

Brendan W. Sullivan
The Basel Problem

/6

d = 4

1  2 2
=
2 6
18

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details
Evaluating I2 : Let u = cos(2) du = 2 sin(2) d


Z 1
1u
du
1

I2 = 4
tan

2
1u
1 u2
1/2
!
Z /6
1 cos(2)
2 sin(2) d
1
p
p
tan
=4
1 cos2 (2)
1 cos2 (2)
0
s
Z /6
1 cos(2)
d
=8
tan1
1 + cos(2)
0
r
Z /6
2 sin2
2
=8
tan1
d
=
2 cos2
9
0
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Details

Evaluating I1 + I2 :
I1 + I2 =

Brendan W. Sullivan
The Basel Problem

2 2
2
+
=
18
9
6

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 3: Summary

1. Consider

R1R1
0

1
0 1xy

dx dy

2. Observe it evaluates to (2) by writing fraction as infinite


series and evaluating term by term
3. Evaluate another way by rotating [0, 1]2 clockwise 45
4. Make some trig subs but nothing advanced
Can be modified using different transformations:
One converts B to a triangle using tan1 in the uv-transform
Hints at relationship to geometry and trigonometry
Who comes up with these?!

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: History

Textbook proof in Fourier analysis [9]


Dont know first appearance or attribution
Depends on Parsevals Theorem
Parseval (1755-1836) was a French analyst, a shadowy
figure in math history; never elected to Acadmie des
Sciences but proposed 5 times; only 5 papers

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Sketch

1. Consider the the space L2 [0, 1] (C-valued functions)


2. Take complete orthonormal set of exp functions
3. Use f (x) = x and evaluate hf, f i as an integral

4. Apply Parsevals Theorem to find hf, f i as a sum

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Details

Consider the space


2

L [0, 1] =


f : [0, 1] C |

Z
0


|f | < +
2

with inner product


hf, gi =

f g
0

This is a nice Hilbert space.

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Details
Consider the set of functions
S = {en (x) := exp(2inx) | n Z}
Claim: S is an orthonormal basis for L2 [0, 1]
Proof: WWTS hem , en i = 0 when m 6= n and = 1 when m = n
Z 1
hem , en i =
exp(2imx) exp(2inx) dx
0

since cos(2nx) i sin(2nx) = cos(2nx) + i sin(2nx)


Z 1
(1 + 0) 1
exp(2i(m n)x) dx =
=
=0
2i(m n)
0
unless m = n, in which case
Brendan W. Sullivan
The Basel Problem

R1
0

exp(0) dx = 1.

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Details
Parsevals Theorem:

f L2 [0, 1]

hf, f i =

X
nZ

|hf, en i|2

More generally:
Z
R

|f (x)| dx =

Z
R

|F (t)|2 dt

where F (t) is the Fourier transform of f (x).


Proof uses Fourier Inversion Formula and is straightforward.
Parseval was working on Fourier series.
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Details
Lets use f (x) = x. (Why not?)
Z 1
Z 1
1
xx
dx =
x2 dx =
hf, f i =
3
0
0
Z 1
Z 1
1
hf, e0 i =
x exp(0) dx =
x=
2
0
0
Z 1
hf, en i =
x exp(2inx) dx for n 6= 0
0


Z 1
1
1
=
[x exp(2inx)]0
exp(2inx) dx
IBP
2in
0


1
1
1
=
(1 0) +
[exp(2inx)]10 =
2in
2in
2in
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Details
Applying Parsevals Theorem:
X
1
= hf, f i =
|hf, en i|2 =
3
nZ

 2
1
+
2

1
= +
4


2
X

1
2in

nZ\{0}

X
nZ\{0}

1
4 2 n2

X 1
1 1
1
=2
2
2
3 4
4 n
nN

X 1
=
6
n2
nN

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 4: Summary
1. Consider the the space L2 [0, 1] (C-valued functions)
2. Take complete orthonormal set, exp(2inx) for n Z

3. Use f (x) = x and evaluate hf, f i directly

4. Apply Parsevals Theorem to find hf, f i as a sum


5. Everything works out

Euler manages to stick his head in. (Eulers Formula!)


Still somehow rooted in his original proofs using series.
Play around with different functions to get related results, e.g.
f (x) = [0,1/2] =

X
nN

Brendan W. Sullivan
The Basel Problem

2
1
=
(2n + 1)2
8

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: History

Published in Amer. Math Monthly, August 2011


Due to Luigi Pace, Dept of Econ & Stats at Udine, Italy [8]
Inspired by a 2003 note that solves the problem using a
double integral on R2+ via Fubinis Theorem
Probabilists are still studying relationship between
independent Cauchy variables and (2k)

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Sketch

1. Take a quotient of two random variables Y =

X1
X2

2. Suppose X1 , X2 are i.i.d. half-Cauchy distributions


3. Find Pr(0 Y 1) by inspection

4. Find Pr(0 Y 1) by joint density

5. Turn fraction into series, integrate term by term


6. Smirk mildly at the no longer surprising appearances of (2)

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Details
Take X1 , X2 : R R+ random variables. Each is governed by a
probability density function pXi : R+ [0, 1] satisfying
Pr (a Xi b) =

pXi (t) dt
a

X1
.
Define Y = X
2
Claim: Probability density function for Y is
Z
pY (u) =
t pX1 (tu) pX2 (t) dt
0

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Details
Proof: Appeal to joint density, then simplify:
Pr (a Y b) =

Z bt2

pX1 (t1 ) pX2 (t2 ) dt1 dt2


0

at2
Z b

t2 pX1 (t2 u) pX2 (t2 ) du dt2


0

Z bZ

t2 pX1 (t2 u) pX2 (t2 ) dt2 du


a

Throw away the outer integral.

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Details
Assign half-Cauchy distribution to X1 , X2 independently:
pXi (t) =

2
(1 + t)2

Use this in the formula for pY (u) obtained above:


Z
1
4
1

dt
pY (u) = 2
t
2
2
0
1 + t u 1 + t2
 
t=
2
1 + t2 u2
= 2 2
ln
(u 1)
1 + t2
t=0

2
2 ln u
4
ln u
= 2 2
= 2 2

u 1
u 1
Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Details

Integrate this expression from 0 to 1:


Z 1
Z
Pr(0 Y 1) =
pY (u) du =
0

4
ln u
du
2
2
u 1

but Pr(0 Y 1) = 21 , obviously, so


Z
0

Brendan W. Sullivan
The Basel Problem

ln u
2
du
=
u2 1
8

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Details

Simplify the integral by using


2
=
8

Z
0

1
1u2

X
ln u
du =
2
1u

n=0 0

= 1 + u2 + u4 + :
1

X
ln u
1
du =
2n
u
(2n + 1)2
n=0

This is equivalent to the original claim. (Prove it!)

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

Proof 5: Summary
1. Take a quotient of two random variables Y =

X1
X2

2. Suppose X1 , X2 are i.i.d. half-Cauchy distributions


3. Find Pr(0 Y 1) by inspection

4. Find Pr(0 Y 1) by joint density

5. Turn fraction into series, integrate term by term


Similar tricks reappearing: e.g. term-by-term integration
Surprising use of probability, but again no indication of insight.
Can probably be modified to find similar results.
Shows this is still an active problem, 250+ years later.

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

References I
W. Dunham
Euler: The Master Of Us All.
MAA, 1999
R. Apry
Irrationalit de (2) et (3)
Astrisque, 61 (1979) 11-13
T. Apostol
A Proof that Euler Missed: Evaluating (2) the Easy Way
Mathematical Intelligencer, 5 (1983) 59-60
F. Beukers
A note on the irrationality of (2) and (3)
Bull. Lon. Math. Soc., 11 (1979) 268-272
Wikipedia
Pietro Mengoli
http://en.wikipedia.org/wiki/Pietro_Mengoli
MacTutor History of Mathematics
Pietro Mengoli
http://www-history.mcs.st-andrews.ac.uk/Biographies/Mengoli.html

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

References II

W. Dunham
When Euler Met lHpital
Mathematics Magazine, 82 (2009) 16-25
L. Pace
Probabilistically Proving that (2) = 2 /6
Amer. Math Monthly, 118 (2011) 641-643
R. Chapman
Evaluating (2)
http://empslocal.ex.ac.uk/people/staff/rjchapma/etc/zeta2.pdf

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar

Intro

Pf 1

Pf 2

Pf 3

Pf 4

Pf 5

References

THANK YOU

Brendan W. Sullivan
The Basel Problem

Carnegie Mellon UniversityMath Grad Student Seminar