√

π 163
Why e Is Almost An Integer
Matthew Baker

√
Abstract: It seems a rather amazing coincidence that the number eπ 163 is
within 10−12 of being an integer (check it out for yourself!). There is, in fact,
a good theoretical reason why this should be so. I will use this assertion as
motivation for discussing some important topics in number theory, including
elliptic curves, modular functions, complex multiplication, and ideal class
groups. The talk is intended to be a rather informal introduction to these
concepts, and should be accessible to just about everybody.
I. Introduction
In the April 1975 issue of Scientific American, Martin Gardner wrote
an article about the six most sensational discoveries of 1974. The whole
article was an April Fools’ Day prank: among the discoveries he reported
were a counterexample to the four-color problem and an artificial-intelligence
computer chess program that determined, with a high degree of probability,
that P-KR4 is always a winning move for white. The article also contained
the following:
In number theory the most exciting discovery of the past
year is that when the
√ transcendental number e is raised to the
power of π times 163, the result is an integer. The Indian √
mathematician Srinivasa Ramanujan had conjectured that eπ 163
is integral in a note in the Quarterly Journal of Pure and Ap-
plied Mathematics. Working by hand, he found the value to be
262537412640768743.999999999999. . . The calculations were te-
dious, and he was unable to verify the next decimal digit. . .
In May 1974 John Brillo of the University of Arizona found an
ingenious way of applying Euler’s constant to the calculations and
managed to prove that the number exactly equals 262537412640768744.
How the prime number 163 manages to convert the expression to
an integer is not yet fully understood.
The reference to the paper of Ramanujan is actually legitimate; he discovered
a number of such ‘near integers’, this one being the most remarkable looking,

1
 √
but of course he knew that none of them were really integers. In fact, eπ 163
is easily seen to be transcendental using the Gel’fond-Schneider Theorem,
proven in 1934, which says that if α and β are algebraic, α 6= 0, 1, β irrational,
then αβ is transcendental.
√
In order to understand why eπ 163 is almost an integer, we will have
to introduce elliptic curves over C and discuss the phenomenon of complex
multiplication.
II. Elliptic Curves over C
A lattice in C is just a set
Λ = {mω1 + nω2 : m, n ∈ Z}, ω1 /ω2 ∈ C − R.
In particular, it is a free Z-module of rank 2, with basis ω1 , ω2 . If we quotient
out C by the lattice Λ, we get a complex torus, which we denote EΛ . The
reason for this notation is that a complex torus is just another name for an
elliptic curve over C.
The name ‘elliptic curve’ seems to imply that EΛ is related to some sort
of algebraic curve. This is indeed the case, and to discuss the connection, we
need to briefly talk about elliptic functions.
Given a lattice Λ, we can construct a function ℘(z; Λ), called the Weier-
strauss ℘-function, which is meromorphic and doubly periodic on C, its poles
occurring only at ω1 , ω2 . Equivalently, we can think of ℘(z) as a function on
the complex torus C/Λ, with a double pole at zero.
If we let g2 = 60 ω∈Λ∗ ω14 , g3 = 140 ω∈Λ∗ ω16 , then ℘(z) satisfies the
P P

differential equation
(℘0 )2 = 4℘3 − g2 ℘ − g3 .
So we get a map
C/Λ −→ EΛ = {(x, y) ∈ C2 : y 2 = 4x3 − g2 x − g3 }
z 7−→ (℘(z), ℘0 (z)).
It turns out that for any lattice, the quantity ∆ = g23 − 27g32 is nonzero.
Conversely, given a nonsingular cubic equation y 2 = 4x3 −ax−b [nonsingular
means that a3 − 27b2 6= 0], one can ask if there is a lattice Λ such that
g2 (Λ) = a, g3 (Λ) = b. The answer turns out to be yes, and this amazing fact
is known as the Uniformization Theorem.

2
 III. Isomorphism Classes of Elliptic Curves
Given any set of mathematical objects, one typically is interested in
studying maps between the various objects. In particular, it is important
to say when two such objects should be considered isomorphic. The maps
we want to look at between elliptic curves are called isogenies. The set of
isogenies between two elliptic curves can be identified with the two sets given
in the following proposition.
Proposition: Given two elliptic curves EΛ1 , EΛ2 , there is a bijection
between:

{holomorphic maps Φ : C/Λ1 → C/Λ2 , Φ(0) = 0}

l
{α ∈ C : αΛ1 ⊂ Λ2 }.

In particular, End(EΛ ) ∼= {α ∈ C : αΛ ⊂ Λ}. It follows that EΛ1 ∼

= EΛ2 ⇔
Λ1 , Λ2 are homothetic, i.e. Λ1 = αΛ2 for some α ∈ C∗ .
Def: If E is an elliptic curve, then

g23
(E) = 1728
∆
is called the -invariant of E.
The following proposition is of fundamental importance, and is quite easy
to prove:
Prop: E ∼
= E 0 ⇐⇒ (E) = (E 0 ).
Now let E = C/Λ be an elliptic curve, with Λ = {Zω1 + Zω2 }. Then
Λ is homothetic to Λτ = {Z + Zτ }, where τ = ω1 /ω2 and, without loss of
generality, Im(τ ) > 0. Therefore the corresponding E and Eτ are isomorphic,
hence (E) = (Eτ ). Abusing notation, we write (τ ) for this common quan-
tity. Viewed in this way, (τ ) defines an analytic function from the upper
half-plane H to C.
A basic fact is that the ‘modular group’ SL(2,Z) acts on H by fractional
linear transformations, and two lattices Lτ and Lτ 0 are equal if and only if

3
for some A ∈ SL(2,Z) we have Aτ = τ 0 . This says that (τ ) is invariant
under the action of SL(2,Z). Roughly speaking, holomorphic functions like
(τ ) with special invariance properties under subgroups of SL(2,Z) are called
modular functions.
In particular, the modular function  is invariant under translation, so
that (τ + 1) = (τ ). Therefore  has a Fourier expansion in the variable q =
e2πiτ . With a little cleverness, one can calculate that the Fourier expansion
(also called the q-expansion) of  is:
1
(τ ) = + 744 + 196884q + . . . ,
q
where the coefficient of q n is an intereger for all n. [Note that the factor
of 1728 in the definition of (τ ) was put there in order to give the above
q-expansion ‘residue one’.]
IV. Complex Multiplication
We now return to the endomorphism ring of an elliptic curve. It is clear
from the definition that Z ⊆ End(E) for all elliptic curves E. For most
elliptic curves, this is it. If End(E) is strictly larger than Z, we say that E
has complex multiplication.
Prop: If E has complex multiplication (CM for short), then there is an
algebraic integer τ 0 of degree 2 with τ 0 ∈ H such that End(E)={Z + Zτ 0 }.
Also, Q(τ ) = Q(τ 0 ). [We say that the endomorphism ring of a CM elliptic
curve is an order in a quadratic imaginary field].
Pf: Exercise.
Main Theorem: Let E be an elliptic curve with CM whose endomor-
phism ring is End(E)={Z + Zτ }. Then the -invariant (E) = (τ ) is an
algebraic integer.
We don’t have time to prove this theorem, but we’ll shortly sketch the
proof of a related result that we’ll also need. First, we need a couple of
definitions:
Def: If R is an order in a quadratic imaginary field, then we set
ELL(R) = { Elliptic curves E: End(E) ∼
= R}/(complex isomorphism)

4
 = { lattices Λ : End(EΛ ) ∼
= R }/(homothety).

Def: If K is a quadratic imaginary field, then a fractional ideal is a Z-module

of rank 2 contained in K which is not wholly contained in R. [One should
think of fractional ideals as regular ideals with ‘denominators’ added].
Prop: Given a quadratic imaginary field K, and a nonzero fractional
ideal a of K, then:
1. a is a lattice in C.
2. End(Ea ) = RK , the ring of integers in K.
So each nonzero fractional ideal a of K gives rise to an elliptic curve with
CM by RK . Conversely, since homothetic lattices give isomorphic elliptic
curves, a and ca give rise to the same element of ELL(R). This suggests that
we look at the group of fractional ideals in K modulo the principal fractional
ideals. If you have encountered algebraic number theory before, you will
undoubtedly recognize this fundamental object as the ideal class group of
RK , denoted CL(RK ). Its order, usually denoted hK , turns out to be finite,
a fact which is very useful and not at all obvious. With the tools we have
already developed, it is not hard to show:
Prop: There is a 1-1 correspondence between the sets CL(RK ) and
ELL(RK ).
We now need one more proposition, which we’ll actually prove:
Prop: Let E be an elliptic curve with CM by RK =ring of integers in K.
Then
[Q((E)) : Q] ≤ hK .

Pf: Let σ ∈ Aut(C/Q). It is easy to see that (E σ ) = (E)σ . But also
End(E σ ) ∼
= End(E) ∼ = RK , so E σ is in one of finitely many C-isomorphism
classes of elliptic curves. Therefore (E)σ takes on at most hK values as σ
ranges over Aut(C/Q). It follows that [Q((E)) : Q] ≤ hK . [In particular,
note that (E) is an algebraic number, which is a special case of our main
theorem].
V. Class Numbers and the Denoument

5
 Now suppose we take a quadratic imaginary field K with hK =1 [which
means that RK is a P.I.D.]. Then there is an elliptic curve E with End(E)
∼
= RK , and the above proposition shows that (E) ∈ Q. Furthermore, by our
main theorem, (E) is an algebraic integer. It is an easy exercise to verify
that this implies that (E) ∈ Z.
So: how many quadratic imaginary fields are there with class number one?
It turns out that there are exactly nine such fields, a fact which was conjec-
tured by Gauss, proven with a few gaps by a German high school teacher
named Heegner in 1952, and reproved rigorously in 1967 independently by
Baker and Stark. The fields are:
√
Q( −d), where d = 1, 2, 3, 7, 11, 19, 43, 67, 163.

For reasons that will soon become apparent, we will focus on the largest of
these: d = 163.
√
From elementary number theory,
√
the ring of integers R K in K = Q( −163)
1+ −163
is generated over Z by τ = 2
. So we know that (τ ) ∈ Z. Now recall
the q-expansion for (τ ):
1
(τ ) = + 744 + 196884O(|q|).
q
√
1+ −163
With τ = 2
, we have
√
q = e2πiτ = −e−π 163
≈ −e−40 ≈ −4 ∗ 10−18 ,
√
so −1
q
= eπ 163 will differ from the integer 744-(q) by roughly (2 ∗ 105 )(4 ∗
10−18 ) ≈ 10−12 .
VI. Remarks
We proved that for E having CM by RK , [Q((E)) : Q] ≤ hK . In
fact, we actually get equality here, and even more is true. K((E)) is a
Galois extension of K with abelian Galois group. The Galois group turns
out to be isomorphic to the ideal class group of RK , and K((E)) is the
Hilbert Class Field of K, the maximal unramified abelian extension of K.
VII. References

6
 1. Gardner, M. Time Travel and Other Mathematical Bewilderments. W.
H. Freeman and Co., 1988.
2. Silverman, J. Advanced Topics in the Arithmetic of Elliptic Curves.
Springer-Verlag, GTM 106, 1994.
3. Waldschmidt, M. et. al. (Eds.) From Number Theory to Physics.
Springer-Verlag, 1992.