How Euler Did It

by Ed Sandifer
Who proved e is irrational?
February 2006
Most readers will know that the constant e is, indeed irrational, even transcendental. I
remember being asked to prove e was irrational on my written exams for my masters degree. It
is natural, then, to ask who was the first to prove it, and to expect an easy and unambiguous
answer. The answer, though, isnt as easy as we might expect, nor is it entirely unambiguous.
Here is some of what MacTutor [McT] has to say about it:
Most people accept Euler as the first to prove that e is irrational.
Certainly it was Hermite who proved that e is not an algebraic
number in 1873.
Note that MacTutor hedges their attribution a bit. They write Most people accept Euler
as the first , (my italics) and do not commit themselves to the more definite Euler was the
first In this case, Eulers rival is not some earlier mathematician who might have a claim to
the result, but Eulers younger protg Johann
Heinrich Lambert (1728-1777), pictured at the right.
Of Lambert, MacTutor writes:
Lambert is best known, however, for his
work on . Euler had already established in
1737 that e and e2 are both irrational.
However Lambert was the first to give a
rigorous proof that is irrational. In a paper
presented to the Berlin Academy in 1768
Lambert showed that, if x is a nonzero
rational number, then neither ex nor tan x can
be rational.
Note that MacTutor chooses words carefully,
Euler established, not proved. On the other hand,
Lamberts proof satisfies most standards of rigor.
The hedging must be because people doubt the rigor
1

of Eulers proof.
Our purpose in this months column is to look at what Euler did, and to see just how
rigorous Eulers results were.
Euler and Lambert both used the tools of continued fractions to produce their results.
Eulers 1737 article that MacTutor mentions is De fractionibus continuis dissertation [E71].
Though Euler was not the first one to study continued fractions, this article is the first
comprehensive account of their properties. Euler repeats most of the elementary properties of
continued fractions in the last chapter of volume 1 of his 1748 masterpiece Introductio in
analysin infinitorum [E101]. Both of these are available in excellent English translations.
The most general form of a continued fraction is

a+

b+

c+

d+

e+
f + etc.
All the symbols, both Latin and Greek, are taken to be positive whole numbers. The Greek
letters Euler calls numerators and the Latin are the denominators. In practice, the most
interesting continued fractions are those for which all the numerators are 1. Continued fractions
in this form are sometimes called regular.
It is not difficult to show that the regular continued fraction expansion of any rational
number is finite, so to prove that a given number is irrational, it suffices to show that its regular
expansion is not finite. We will show how this works using one of Eulers examples from the
e 1
8591409142295
Introductio. We consider the number
0.8591409142295 =
.
2
10000000000000
Since this number is less than 1, the first denominator, a = 0. Now, Euler inverts the
fractional part and gets
10000000000000
1408590847704
= 1+
.
8591409142295
8591409142295
The next denominator is the integer part of this, so b = 1. Invert the fraction part of this and get
8591409142295
139862996071
= 6+
.
1408590847704
1408590847704

The next denominator is the integer part of this, so c = 6. Continue to take integer parts, and
invert fractional parts, and we get
1408590847704
9950896994
= 10 +
,
139862996071
139862996071

so d = 10.

139862996071
551438155
= 14 +
,
9950896994
9950896994

so e = 14.
9950896994
25010204
= 18 +
,
551438155
551438155

so f = 18.
551438155
1213667
= 22 +
,
25010204
25010204

so g = 22.
Euler stops here, saying If the value for e at the beginning had been more exact, then the
sequence of quotients would have been 1, 6, 10, 14, 18, 22, 26, 34, , which form the terms of
an arithmetic progression. It follows that
e 1
= 0+
2
1+

1
6+

1
10 +

1
14 +

1
18 +

1
22 + etc.

Note that Eulers arithmetic progression doesnt start with the first denominator, but
starts with the 6, after which the denominators increase by 4.
Euler adds, somewhat disingenuously, This result can be confirmed by infinitesimal
calculus.
Since the sequence of denominators clearly increases, and never terminates, this is not a
finite continued fraction. Thus, by the work Euler did earlier, its value cannot be rational. Since
e 1
is not rational, e cannot be rational, either.
2
By similar means, Euler shows that
e=2+

1
1+

1
2+

1
1+

1
1+

1
4+

1
1+

1
1+

1
6+

1
1 + etc.
3

Unless Euler skipped something, the proof is done.

Alas, Euler did skip something, and he hid it in that comment, This result can be
confirmed by infinitesimal calculus. He has only observed that finite calculations lead to a
pattern for the first few denominators, and that the pattern seems to extend indefinitely. He has
not proved it, and he knows he has not proved it.
What could he have been thinking?
Earlier in the chapter of the Introductio, Euler showed how to convert a continued
fraction, whether regular or not, into an alternating series. He showed that if

x = a+

b+

c+

d+

e+

f + etc.

then

x=a+

+
...
b b ( bc + ) ( bc + )( bcd + d + b )

e 1
2
and get something related to one of the well-known series for e? But, if we do that, we get

Perhaps we can apply this alternating series formula to the coefficients we got for

x = 1

1
1
+
etc.
7 7 71

which does not seem related to any other well-known series for e.
So, that wasnt how he did it. If we go back to E-71, we get more clues. In fact, he
writes:
In the preceding sections, where I have converted the number e (whose
logarithm is 1) together with its powers into continued fractions, I have only
observed the arithmetic progression of the denominators and I have not been
able to affirm anything except the probability of this progression continuing to
infinity. Therefore, I have exerted myself in this above all: that I might
inquire into the necessity of this progression and prove it rigorously. Even
this goal I have pursued in a peculiar way.
Indeed his solution comes from a most surprising direction, differential equations. It lies
in a form of an important differential equation called the Ricatti equation:

ady + y dx = x
2

4 n
2 n +1

dx
4

Euler claims that if we substitute p = ( 2n + 1) x 2n+1 , then the equation transforms into
adq + q2 dp = dp
I have found that
q=

a
1
+
1
p 3a +
1
p 5a +
7
a
1
p
+ ... +
p
( 2n 1) +
p

1
x

2n
2 n +1

This continued fraction terminates after n ratios, but if n is taken to be one of Eulers
infinite numbers, then the continued fraction goes on forever.
This is a great leap. If Euler did show this previously, then I couldnt find where he did
it. The most likely places to look would be in E-28 and E-31, where Euler does other series
analyses of the Ricatti equation, but I cant see it there, or in any other papers on differential
equations that Euler wrote before he wrote E-71. It was a great mystery to me.
On the other hand, the variables in the equation adq + q2 dp = dp separate to give
adq
= dp
1 q2
which, in turn, integrates to give
a
1+ q
log
= p+C.
2
1 q
The constant C can be taken to be zero with the initial conditions q = and p = 0. A bit of
algebra gives that
2p
q +1
1
ea =
= 1+
q 1
q 1
Just a little while ago, though, we got a continued fraction expansion for q, which we can
substitute into this last expression to get

2p
a

= 1+

a p
1
+
3a
1
p
+
5
a
1
p
+
p 7 a + etc.
p
Various values of p and a give continued fractions for various expressions involving e.
For example, p =1 and a = 2 gives
5

e=e =e
1

= 1+

which is equivalent to the expansion for

21
2

2
1+

1
6+

1
10 +

1
14 + etc.

e 1
that Euler had observed earlier.
2

Euler works through a few other substitutions to derive his other observations, then
writes,
Truly everything found above follows from these formula, by which we
have expressed e and its powers as continued fractions. That is, the
necessity of the progressions only observed earlier is now proved.
So, we complete the path from Eulers continued fraction solution to the Ricatti equation
to the irrationality of e, but we cant be very satisfied with that solution of the differential
equation. I looked at a good sample of Eulers earlier work, and cant find where Euler might
have discovered this solution. I was about to give up and admit Eulers claim to having proved
the irrationality of e had a great big hole in it.
So, I was about to throw in the towel and say Eulers claim to proving the irrationality of
e was kind of weak. I wasnt quite ready to let go of it, when I had one of those right under my
nose experiences. There it was in the last five paragraphs of E-71. Euler gives, in considerable
detail, his proof that the continued fraction solves the Ricatti equation. We wont go into much
detail; the interested reader can find the details in the Wyman and Wyman translation [E71],
starting in paragraph 31. Briefly, he starts with a regular continued fraction in which the
denominators form an arithmetic series. It looks like
s =a+

1
(1 + n) a +

1
(1 + 2n) a +

1
(1 +3 n) a +

1
(1 + 4n) a + etc.

He uses his identities from early in the paper to rewrite this as a ratio of power series
involving n and a, then shows that the power series in the numerator of the ratio is related to the
derivative of the power series in the denominator. This gives him a differential equation, which,
three pages later, he transforms into the Ricatti equation he wanted.
Its right. Its complete, and it works. Id been fooled when Euler suggested that he had
already shown the relation between the continued fraction and the differential equation. Euler
really did prove that e is irrational, and he probably regarded it as the main point of this paper.
6

Were ready to close this months column. There was probably a shorter path from the
question Who proved e is irrational? to the conclusion Euler, but this path shows some of the
details of how we learned the story. We hope youve enjoyed the adventure.
References:
[C]

[E71]
[E101]

[S]
[StA]

Chabert, Jean-Luc, et. al., A History of Algorithms: From the Pebble to the Microchip, Translated from the
French edition Histoire dlalgorithmes. Du caillou la puce. Editions Belin, Paris, 1994, by Chris Weeks,
Springer, Berlin, 1998.
Euler, Leonhard, An Essay on Continued Fractions translated by Myra F. Wyman and Bostwick F.
Wyman, Math Systems Theory 18 (1985) 295-328 (1985)
Euler, Leonhard, Introductio in analysin infinitorum, 2 vols., Bosquet, Lucerne, 1748, reprinted in the
Opera Omnia, Series I volumes 8 and 9. English translation by John Blanton, Springer-Verlag, 1988 and
1990. Facsimile edition by Anastaltique, Brussels, 1967.
Sandifer, Edward, Euler and Pell, How Euler Did It, MAA OnLine,
http://www.maa.org/news/howeulerdidit.html, April 2005.
The number e at http://www-groups.dcs.st-and.ac.uk/~history/HistTopics/e.html, MacTutor History of
Mathematics Archive, University of St. Andrews, Scotland, as revised September, 2001.

Ed Sandifer ([email protected]) is Professor of Mathematics at Western Connecticut State

University in Danbury, CT. He is an avid marathon runner, with 33 Boston Marathons on his
shoes, and he is Secretary of The Euler Society (www.EulerSociety.org)

How Euler Did It is updated each month.

Copyright 2006 Ed Sandifer