International Journal of Mathematical Education in

Science and Technology

ISSN: 0020-739X (Print) 1464-5211 (Online) Journal homepage: www.tandfonline.com/journals/tmes20

Approximations of e and π: an exploration

Philip R. Brown

To cite this article: Philip R. Brown (2017) Approximations of e and π: an exploration,

International Journal of Mathematical Education in Science and Technology, 48:sup1, S30-S39,
DOI: 10.1080/0020739X.2017.1352043

To link to this article: https://doi.org/10.1080/0020739X.2017.1352043

© 2017 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group

Published online: 26 Oct 2017.

Submit your article to this journal

Article views: 17621

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tmes20
INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY, 
VOL. , NO. S, S–S
https://doi.org/./X..

Approximations of e and π: an exploration

Philip R. Brown
Department of Liberal Studies, Texas A&M University at Galveston, Galveston, Texas

ABSTRACT ARTICLE HISTORY

Fractional approximations of e and π are discovered by searching for Received  March 
repetitions or partial repetitions of digit strings in their expansions in
KEYWORDS
different number bases. The discovery of such fractional approxima-
Rational approximations;
tions is suggested for students and teachers as an entry point into continued fractions;
mathematics research. convergents of continued
fractions; number bases
AMS SUBJECT
CLASSIFICATION
Primary -; J;
secondary A; A;
D

1. Introduction
The exploration of the fundamental constants e (the natural base of the logarithm) and
π (the ratio of the circumference of a circle to its diameter) provides an access point for
discovery, research and scholarship in mathematics for students and mathematicians at all
levels, from the high school or college student up to the professional. The history of these
constants also provides a gateway into the history of mathematics.
We aim to demonstrate in this article how mathematical discovery can begin with simple
curiosity. Some theory and techniques, including the theory of continued fractions and the
conversion of repeating decimals to fractions, that are accessible to advanced high school
or beginning college students are learned along the way.
The number e has the decimal expansion 2.71828182845904 …. Hand-held calculators
typically provide 10 significant digits of a number. Therefore, the evaluation of e on a calcu-
lator would usually display 2.718281828. A person who is unaware of the irrational nature
of e might suspect that e = 2.71828. We are thus led to curiosity about the repetition of the
digits ‘1828’. Some mathematicians remember 1828 as the birth date of the famous Russian
author Leo Tolstoy. We now consider the following questions:
r What is the number 2.71828 and is it a good approximation of e?
r How do we find better approximations of e?
r What repetitions can we find for π in its decimal or other number expansions?
r What repetitions can we find in the decimal or other number expansions of 1 or 1 ?
e π

We will explore some of these questions below.

CONTACT Philip R. Brown [email protected]

©  The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License
(http://creativecommons.org/licenses/by-nc-nd/./), which permits non-commercial re-use, distribution, and reproduction in any medium,
provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
 INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY S31

It is a basic fact about real numbers that a repeating decimal corresponds to a rational
number, which can always be expressed as a fraction. Therefore, we need to learn something
about rational approximations of irrational numbers before we begin our discovery relating
to the questions we have asked.

2. Continued fractions
In [1, p.66], the authors Pierre Eymard and Jean-Pierre Lafon complain that although con-
tinued fractions have played a major role in the history of mathematics, they have almost
completely disappeared from teaching, even at universities, despite the fact that they are
essential to an understanding of the approximation of real numbers by rationals. The rea-
son for this, perhaps, is that they are particularly ill-adapted to computational uses.
In [2, p.3], it is stated that algorithms equivalent to the modern use of continued frac-
tions were in use for many centuries before their real discovery. The best known example
is Euclid’s algorithm for the greatest common divisor of two integers, which leads to a ter-
minating continued fraction. The approximate computation of square roots led to some
numerical methods which can be viewed as the ancestors of continued fractions. Certain
algorithms discovered by Indian mathematicians, beginning with Brahmagupta in the sev-
enth century, for the solution of Pell’s equation,1 a type of Diophantine equation, are also
analogous to continued fractions.
A first attempt for a general definition of continued fractions was made by Leonardo of
Pisa (Fibonacci), who described ascending continued fractions in his book ‘Liber Abaci’
that he wrote in 1202 (see [2, p.51–52]).
The real discoverer of continued fractions was the Italian mathematician and astronomer
Pietro Antonio Cataldi, who developed a symbolism for continued fractions and derived
some of their properties in a book published in 1613. His work was based on the earlier
work on the extraction of square roots by his older contemporary, Rafael Bombelli (the
founder of imaginary numbers) [2, p.61–70].
We now introduce continued fractions and explain their relationship to the approxima-
tion of irrational numbers.
If x is a real number, an irreducible fraction qp with q > 0 is said to be a best rational

approximation for x if, for any integer q with 0 < q ࣘ q and any irreducible fraction qp = qp ,
we have
   
 p   p 
x −  < x −  .
 q  q 

Furthermore, qp is said to be a best rational approximation in the strong sense if, under the
same conditions,
   
qx − p < q x − p  .

It follows that, if qp is a best rational approximation in the strong sense, then it is a best
rational approximation.
It is known [1, p.73] that the best rational approximations of an irrational number x in
the strong sense are exactly the convergents of the regular continued fraction for x. We use
S32 P. R. BROWN

the notation

[b0 , b1 , . . . , bn , . . .] = lim [b0 , b1 , . . . , bn ] ,

n→∞

where the bn are natural numbers, possibly with the exception of b0 = 0, and

1
[b0 , b1 , . . . , bn ] = b0 +
1
b1 +
1
b2 +
..
. 1
+ bn−1 +
bn

are the convergents, for a regular continued fraction.

Any irrational number is the limit of a regular continued fraction, and of only one, which
can be derived by means of the following formula (see [1, p.72]). We use the notation x
for the floor of x, i.e. the largest integer less than or equal to x.
Let b0 = x, x = b0 + x11 ; b1 = x1 , x1 = b1 + x12 ; …; bn = xn , xn = bn + xn+1
1
;….
The expansion of the number e as a regular continued fraction is known explicitly
[1, p.78]:

e = [2, 1, 2, 1, 1, 4, 1, 1, 6, . . . , 1, 1, 2n, . . .] ,

with the following sequence of convergents:

8 11 19 87 106 193 1264 1457 2721 23225

2, 3, , , , , , , , , , ,.... (1)
3 4 7 32 39 71 465 536 1001 8544

In the case of the number π, however, the regular continued fraction is not known explicitly.
Using the formula above, with the known decimal digits for π, we obtain [1, p.78]

π = [3, 7, 15, 1, 292, 1, 1, 1, 2, 1, 3, 1, 14, 2, 1, 1, 2, 2, 2, 2, 1, 84, . . .] ,

with the following sequence of convergents:

3 22 333 355 103993 104348 208341 312689

, , , , , , , ,.... (2)
1 7 106 113 33102 33215 66317 99532

3. Fractional approximations of e
Fractional approximations of e are rarely mentioned in the mathematics literature. Eli Maor
mentions in [3, p.37] that the closest rational approximation of e using integers below 1000
is 878
323
. The French mathematician, Charles Hermite, who published the first proof of the
transcendence of e in 1873 gave as a sequel to his proof the approximations 58 21 444 and 21 444
291 158 452

2
of e and e , respectively (see [3, p.189]).
 INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY S33

An approximate value of e can easily be obtained by calculating a partial sum of the

series

∞
1
e= .
n=1
n!

611070150698522592097
For example, the first 22 terms of the series add up to 224800145555521536000
, which evaluates to
a decimal giving correctly the first 20 digits of e:

e ≈ 2.71828182845904523536 . (3)

As mentioned earlier, we can approximate e by the repeating decimal fraction 2.71828.

The method that is usually taught to students to convert a recurring decimal to a fraction
is the following:

x = 2.71828
y = 10x = 27.1828
10000y = 271828.1828
99990x = 9999y = 271801
271801
x= . (4)
99990

An alternative method of converting the repeating decimal to a fraction is to express it as

the geometric series

∞  
1 k
10e = 27 + 1828
k=1
104
1828
= 27 +
104 − 1
1828
= 27 +
9999
271801
e= . (5)
99990

A mentioned in [4, p.15], the fraction 271801

99990
cannot be reduced to a simpler fraction, how-
ever, the slightly smaller fraction 271800
99990
reduces to 3020
1111
= 2.71827, which differs from the
decimal expression for e by one unit in the fifth decimal position.
We can improve the estimate by approximating e by 2.7182818. By means of either of the
methods above the corresponding fraction is found to be 271827919999990
, which cannot be reduced;
however, the smaller fraction 27182790
9999990
reduces to the eleventh convergent 2721
1001
= 2.718281 of
the regular continued fraction expansion for e in Equation (1). It is also the closest rational
approximation of e with integers less than 10,000, as can be verified by means of a com-
puter search of all fractions that have no more than four digits in the numerator and the
denominator.
S34 P. R. BROWN

4. Fractional approximations of π
The value of π with 20 correct decimal digits is

π ≈ 3.14159265358979323846. (6)

We will begin a short historical overview of the calculation of π.

4.1. Historical overview

The letter π was introduced as the notation for the ratio of the circumference to the diame-
ter of a circle by the English mathematician William Jones in 1706 in his publication A New
Introduction to Mathematics. He used the letter π as an abbreviation for ‘periphery’ (of a
circle with unit diameter); see [5, p.145].
The approximation π > 3 18 = 3.125 was known to the Babylonians in 2000 B.C. (see [5,
p.21–22]) and the Egyptians found the approximate value 4×( 89 )2 = 258 81 =3.185 by approximating
the area of a disk by the area of an octagon [5, p.24–25].
The famous Greek mathematician and engineer Archimedes determined 3.14085 ≈
223
71
< π < 227 ≈ 3.14286 by approximating the area of a disk by the areas of inscribed and
circumscribed polygons [5, p.67]. The Alexandrian astronomer Ptolemy who lived in the
second century A.D. used the approximation π < 3 120 17
= 377
120 ≈3.14167, which he had proba-
bly inherited from an earlier Greek mathematician Appollonius of Perga who was about
30 years younger than Archimedes [5, p.74].
Early Hindu knowledge of mathematics was summarized by Aryabhatiya in 499 A.D.
He provided the the following recipe for approximating the value of π: Add 4 to 100, mul-
tiply by 8 and add 62000. The result is approximately the circumference of a circle of which
the diameter is 20, 000 (see [5, p.26]). This translates as π ≈ 3927
1250
= 3.1416. Most likely this
fraction was also obtained by means of polygonal approximation. The Hindus at this time
had at their disposal the superior decimal number system which facilitated more compli-
cated calculations than the Greeks could handle.
Fibonacci (the author of Liber Abaci) calculated the approximation 864 275
= 3.1418 of π
using a 96 sided polygon, as Archimedes had done. However, he had the advantage of
being able to compute approximations of square roots using the new decimal arithmetic
[5, p.84].
The invention of decimal fractions and logarithms facilitated numerical calculations
in Europe from the 1500s onward. The approximation 355 311
≈ 3.141592920 (correct to
six decimal digits) known as the Metius approximation (after the Dutch Mathematician
Adriaen Metius) was discovered by several mathematicians during the sixteenth century
using the Euclidean algorithm and continued fractions [2, p.71]. It is the fourth con-
vergent in the sequence of convergents of the regular continued fraction for π given
in Section 2. This fraction was also reported in 480 A.D. by the Chinese Tsu Ch’ung
[1, p.78].
After Isaac Newton discovered the binomial formula for fractional exponents, he was
able to compute π accurately to 16 decimal places by using an integration formula, intro-
duced by Pierre de Fermat and Blaise Pascal, to calculate the area of a sector of a circle
[5, p.142].
 INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY S35

In 1706, the astronomer Johan Machin derived the formula

   
π 1 1
= 4 arctan − arctan
4 5 239

[5, p.144–145] and then calculated π accurately to 100 decimal places by substituting the
fractions on the right hand side into the Gregor series arctan(x) = x− x33 + x55 − x77 +.... The same for-
mula (Machin’s formula) was used in 1949 by programmers of one of the earliest computers,
the ENIAC (Electronic Numerical Integrator and Computer) at Ballistic Research Labs to
calculate 2037 decimal places of π in 70 hours [5, p.184]. With the advance of computa-
tional power and sophisticated programming, the number of digits of π that can currently
be calculated is over 22 trillion! (see [6]).
In the 1990s the mathematicians David Bailey, Peter Borwein and Simon Plouffe discov-
ered the formula

∞  
1 4 2 1 4
π= − − −
k=0
16k 8k + 1 8k + 4 8k + 5 8k + 6

(now known as the BBP formula) which can be used to calculate any specified hexadecimal
digit of π [1, p.79]. For example [7], an international network of computers was organized
by Colin Percival to calculate the quadrillionth binary digit of π.

4.2. A few approximations of pi

An interesting fact about the decimal expansion of π is that the first zero digit does not
occur until the 32nd position after the decimal point. Many poets have written poems using
the digits of π to determine the number of letters in each word. In the following poem,
which appears in the children’s book Somewhen by David Saul and Danielle Mathieson, the
word ‘nothing’ is used in the 32nd position.
It’s a fact
A ratio immutable
Of circle round and width
Produces geometry’s deepest conundrum
For as the numerals stay random
No repeat lets out its presence
Yet it forever stretches forth
Nothing to eternity

Other poets have used words containing 10 letters to represent the zero digit.
However, in some expansions of π in different bases, zero digits do appear near the
beginning of the expansion. Here are the expansions of π in some other bases:

(binary) 11.00100 10000 11111 10110 10101 00010 00100 00101

(ternary 10.01021 10122 22010 21100 21111 10221 22222 01112

(quarternary) 3.02100 33312 22202 02011 22030 02031 03010 30121

(quintary) 3.03232 21430 33432 41124 12240 41402 31421 11430

(sextary) 3.05033 00514 15124 10523 44140 53125 32110 23012

S36 P. R. BROWN

(heptary) 3.06636 51432 03613 41102 63402 24465 22266 43520

(octal) 3.11037 55242 10264 30215 14230 63050 56006 70163

(nonal) 3.12418 81240 74427 88645 17776 17310 35828 51654

(decimal) 3.14159 26535 89793 23846 26433 83279 50288 41971

(duodecimal) 3.18480 9493B 91866 4573A 6211B B1515 51A05 72929

(hexadecimal) 3.243F6 A8885 A308D 31319 8A2E0 37073 44A40 93822

which can be verified at the website [8]. The two zero digits appearing in the fourth
and fifth positions after the point in the quarternary expansion suggest the approxi-
mation π by 3.021 (in base 4). This corresponds to the decimal fraction 3 + 162 + 641 = 201
64 =
3.140625.
The heptary (base 7) approximation 3.06̇ = 3.1 equals the famous approximation 227 in
decimal arithmetic and the quintary approximation of π 3.032 corresponds to the decimal
fraction

∞  
1 1 k 17 1 377
3 + (3 × 5 + 2) 2
=3+ × =
5 k=1
5 5 24 120

that was given by Ptolemy.

Approximating π in base 9 by 3.124188 produces the decimal fraction

∞ 
 
1 k 75248 104348
3 + (95 + 2 × 94 + 4 × 93 + 1 × 92 + 8 × 9 + 8) =3+ =
k=1
96 531440 33215

which is the sixth convergent in the continued fraction of π mentioned earlier.

We can also examine expansions of π1 in the bases above.

(binary) 0.01010 00101 11110 01100 00011 01101

(ternary 0.02212 10010 21220 22120 21110 12121

(quarternary) 0.11011 33030 01231 30213 02020 02221

(quintary) 0.12434 32434 44234 24132 34230

(sextary) 0.15243 10221 33341 41411 31214

(heptary) 0.21411 55610 23405 33404 21346

(octal) 0.24276 301556 23442 02512 37604

(nonal) 0.27703 78276 74177 47214 17757

(decimal) 0.318309886183790671537767526745

(duodecimal) 0.39A05 82886 B3742 17852 28778

(hexadecimal) 0.517CC 1B727 220A9 4FE13 ABE8F

 INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY S37

The quintary expansion is interesting because the approximation 0.124343 corresponds

to the decimal fraction

∞    
1 1 1 k 1 1848 113
+ (2 × 5 + 4 × 5 + 3 × 5 + 4 × 5 + 3)
4 3 2
5
= 1+ = .
5 5 k=1
5 5 3124 355

355
The reciprocal 113
is the Metius approximation.

4.3. Approximating π by a sum of fractions

Instead of approximating π by one fraction, we can look for approximations of π by a sum of
fractions. For example, the repeating decimal fraction 3.14159 is 313835
99900
. The slightly smaller
fraction 313834
99900
simplifies to 4241
1350
and we find that π − 4241
1350
≈ 0.0001111721. Therefore, π is
approximated by 1350 + 9000 = 27000 = 3.141592. Furthermore, we have
4241 1 84823

84823 = 271 × 313 = 270 × 313 + 313 = 270 × 313 + 27 × 11 + 16

so π is approximated by 313
100
+ 11
1000
+ 16
27000
. The last fraction simplifies to 2
3375
. If we replace
2
it by 3374 = 1687
1
we find that

3141 1
π≈ + ≈ 3.141592768 (7)
1000 1687

which differs from π by one unit in the seventh digit after the decimal point. This is an
approximation for π that can be easily remembered because 1687 is the year in which Isaac
Newton published his Principia!
Using the quarternary expansion for π, we have the following calculation in base
4 arithmetic:

3.3̇ − π = 0.31233000211113131322113033 . . .
 
1 1
= 0.31233 + 9 2.1̇ + 4 0.2020211001322 . . .
4 4
 
1 1
≈ 0.31233 + 9 2.1̇ + 4 0.20 .
4 4

In decimal arithmetic, we, therefore, find that π can be approximated by

 
879 7 8
4− + + .
45 3 × 49 15 × 413

That is
145 7 8 395303839
π ≈3+ − − = ≈ 3.1415926535924
4 5 3×49 15 × 4 13 125829120

which differs from the decimal expansion of π by one unit in the 11th decimal position
after the decimal point.
S38 P. R. BROWN

Another way to approximate π by a sum of fractions is to subtract a fraction from π and

then to find a fraction that approximates the result. For example,

355
π− ≈ 0.266764189 × 10−7 .
113
The corresponding value of the decimal string in base 14 is 0.3A40027C which we can
approximate by 0.3A4. This is the decimal fraction 183
686
. Therefore,

355 1 183
π− − 6 = 0.34066592 × 10−12 .
113 10 686
The decimal string is equivalent to the base 14 string 0.4AAB044, which can be approxi-
mated by 0.4Ȧ. This is the decimal fraction 31
91
. Consequently, we have
   
355 1 183 1 31
π≈ − − 12 (8)
113 106 686 10 91

which differs from π by one unit in the 17th position after the decimal point.
Double precision arithmetic on a computer can represent a number with at most 17
significant digits [9]. Therefore, the expression in Equation (8) can be used as the value for
π for computer calculations.

5. A suggestion for teachers

We suggest that teachers challenge their students with the following sequence of questions:

r find the best approximation for e or π among fractions having at most one digit in the
numerator and the denominator
r find the best approximation for e or π among fractions having at most two digits in
the numerator and the denominator
r find the best approximation for e or π among fractions having at most three digits in
the numerator and the denominator
r find the best approximation for e or π among fractions having at most four digits in
the numerator and the denominator

Depending on the teacher’s goals and the level of the students, the best approximating
fractions above may be discovered by means of a computer search, as a programming exer-
cise, by the calculation of continued fraction convergents (which would require the teacher
to introduce continued fractions), or by the discovery of partial digit repetitions in the num-
ber expansions of e and π (or their reciprocals) in different bases, as we have done in this
article.

Note
1. As mentioned in [10, p.43], the Diophantine equation x2 − D y2 = 1, where D is a non-square
integer, is known as Pell’s equation because Euler mistakenly attributed a solution of it to the
 INTERNATIONAL JOURNAL OF MATHEMATICAL EDUCATION IN SCIENCE AND TECHNOLOGY S39

seventeenth century English mathematician Pell. This is an example of European

mathematicians mis-naming parts of mathematics. Brahmagupta’s method of finding solutions
of this equation was the first major advance in number theory since Diophantus (see [10, p.72]),
who lived in Alexandria in the third century AD.

Disclosure statement
No potential conflict of interest was reported by the author.

References
[1] Eymard P, Lafon J-P. The number π. Providence (RI): American Mathematical Society; 2004.
[2] Brezinski C. History of continued fractions and padé approximants. Berlin: Springer-Verlag;
1980.
[3] Maor E. e: the story of a number. Princeton (NJ): Princeton University Press; 1994.
[4] Brown P. Foundations of mathematics: algebra, geometry, trigonometry and calculus. Dulles
(VA): Mercury Learning and Information; 2016.
[5] Beckman P. A history of Pi. New York (NY): Golem Press; 1971.
[6] y-cruncher – A Multi-Threaded Pi-Program [internet]. Alexander J. Yee; [updated 2017
Sept 11]. Available from: http://www.numberworld.org/y-cruncher/
[7] Bailey DH, Borwein JM. Experimental mathematics: examples, methods and implications.
Notices AMS. 2005;52(5):502–514.
[8] Base Converter [internet]. Université Nice Sophia Antipolis; c1998. Available from:
http://wims.unice.fr/wims/en_tool∼number∼baseconv.en.html
[9] Double-precision floating-point format [internet]. Wikipedia, The Free Encyclope-
dia; [updated 2017 Sept 11]. Available from: https://en.wikipedia.org/wiki/Double-
precision_floating-point_format
[10] Stillwell J. Mathematics and its history. 2nd ed. New York (NY): Springer-Verlag; 2002.