An Egyptian fraction is a sum of positive unit fractions, usually required to be distinct. The
famous Rhind papyrus, dated to around 1650 BC, contains
a table of representations of as Egyptian fractions for odd
between 5 and 101. The reason the Egyptians
chose this method for representing fractions is not
clear, although André Weil characterized the decision as "a wrong turn"
(Hoffman 1998, pp. 153-154). The unique fraction
that the Egyptians did not represent using unit fractions
was 2/3 (Wells 1986, p. 29).
Egyptian fractions are almost always required to exclude repeated terms, since representations such as
are trivial. Any positive rational
number has representations as an Egyptian fraction with arbitrarily many terms
and with arbitrarily large denominators, although
for a given fixed number of terms, there are only finitely many. Fibonacci proved
that any positive fraction
can be represented as a sum
of distinct unit fractions (Hoffman 1998, p. 154).
An infinite chain of unit
fractions can be constructed by repeatedly splitting the term with largest denominator using the identity
|
(1)
|
for integers .
Martin (1999) showed that for every positive rational number, there exist Egyptian fractions whose largest denominator
is at most
and whose denominators form a positive
proportion of the integers up to
for sufficiently large
. Each positive fraction
with
odd has an Egyptian fraction
in which each denominator is odd
(Starke 1952, Breusch 1954; Guy 1994, p. 160). For integers
, there is a
-term representation with
as
(Vose 1985).
For a fixed positive rational number , let
be the number of subsets
with
. Conlon et al. (2025) proved
that
|
(2)
|
where
|
(3)
|
and
is the binary Shannon entropy,
with
. The positive
parameter
is the unique solution of
|
(4)
|
The constant increases strictly with
and tends to 1 as
. In particular,
. Liu and Sawhney (2024) independently
obtained the case
. The error term is in the exponent, and the target
is held fixed as
.
Liu and Sawhney (2024) also proved that, for every fixed and all sufficiently large
, a subset
with
|
(5)
|
contains a subset whose reciprocals sum to 1. They established the sharp size threshold: also suffices. The constant
cannot be decreased, since the integers in
larger than
have total reciprocal sum less than 1 for each fixed
and all sufficiently large
.
Their denominator bound states that every rational number
with integers
, for sufficiently
large
, has a representation by distinct unit
fractions whose largest denominator is at most
|
(6)
|
For a related extremal question, let be the least positive integer
for which 1 has no representation by
distinct unit fractions with least denominator
and largest denominator
at most
. Liu and Sawhney (2024) showed that
|
(7)
|
A minimum number of terms can be found by finite exhaustive search. For a positive
remainder
represented by
distinct unit fractions in increasing order of denominators, the first denominator
satisfies
, where
is the ceiling function
and
is the floor function. Recursion therefore gives
a finite search for each fixed
; testing successive
eventually finds a shortest representation. The smallest possible
largest denominator can likewise be found by enumerating
subsets of
for successive
(Eppstein). There are also a number of algorithms
(including the binary remainder method,
continued fraction unit fraction
algorithm, generalized remainder method, greedy
algorithm, reverse greedy algorithm,
small multiple method, and splitting algorithm)
for decomposing an arbitrary positive fraction
into unit fractions. According to Martin (1999,
p. 3641), Fibonacci published an algorithm for
constructing unit fraction representations in 1202,
and this algorithm was subsequently rediscovered by
Sylvester.
Taking the fractions 1/2, 1/3, 2/3, 1/4, 2/4, 3/4, ... (the numerators of which are OEIS A002260,
and the denominators of which are copies of the integer
), the unit fraction representations
using the greedy algorithm are
|
(8)
| |||
|
(9)
| |||
|
(10)
| |||
|
(11)
| |||
|
(12)
| |||
|
(13)
| |||
|
(14)
| |||
|
(15)
| |||
|
(16)
| |||
|
(17)
|
The number of terms in these representations are 1, 1, 2, 1, 1, 2, 1, 2, 2, 3, 1, ... (OEIS A050205). The minimum denominators for each representation are given by 2, 3, 2, 4, 2, 2, 5, 3, 2, 2, 6, 3, 2, ... (OEIS A050206), and the maximum denominators are 2, 3, 6, 4, 2, 4, 5, 15, 10, 20, 6, 3, 2, ... (OEIS A050210).
The Egyptian fractions for the fractional parts of various constants using the greedy
algorithm are summarized in the following table, where denotes Catalan's constant.
| constant | OEIS | Egyptian fraction
for |
| A006487 | 3, 13, 253, 218201, 61323543802, ... | |
| A118325 | 2, 5, 32, 1249, 5986000, 438522193400489, ... | |
| A069139 | 2, 5, 141, 68575, 32089377154, ... | |
| A006525 | 2, 5, 55, 9999, 3620211523, 25838201785967533906, ... | |
| A006526 | 3, 29, 15786, 513429610, 339840390654894740, ... | |
| A110820 | 2, 13, 3418, 52016149, 153922786652714666, ... | |
| A118323 | 2, 3, 13, 176, 36543, ... | |
| A117116 | 2, 9, 145, 37986, 2345721887, ... | |
| A118324 | 2, 6, 38, 6071, 144715221, ... | |
| A001466 | 8, 61, 5020, 128541455, 162924332716605980, ... | |
| A006524 | 4, 15, 609, 845029, 1010073215739, ... |
Graham proved that infinitely many fractions in a certain range can be represented as a sum of unit fractions with square denominators (Hoffman 1998, p. 156).
The Erdős-Straus conjecture asserts that the Diophantine equation
|
(18)
|
has a solution in positive integers ,
, and
for every integer
, without requiring distinct denominators.
Similarly, Sierpiński (1956) conjectured that
|
(19)
|
can be solved under the same conditions (Guy 1994).
The harmonic number is never an integer except for
. According to Hoffman (1998, p. 157), Taeisinger
proved this result in 1915, and Kürschák proved in 1918 the more general
result that a sum of at least two reciprocals of consecutive
positive integers never equals an integer.
Equally spaced denominators can give a unit
fraction; for example,
.
Nontrivial sets of integers are known whose reciprocals sum to small integers. For example, there exists a set of 366 positive integers (with maximum 992) whose sum of reciprocals is exactly 2 (Mackenzie 1997; Martin). A similar set of 453 small positive integers is known whose reciprocals sum to 6 (Martin).
OpenAI (2026) announced that every rational number with integers
can be expressed using
distinct unit
fractions, with unrestricted denominators. The
claimed order is optimal in the worst case over the numerators
for a fixed denominator
. The assertion is a worst-case bound and does not require
every individual fraction to have that many terms.
Bloom and van Doorn (2026) give a human exposition and simplification, obtaining the explicit bound
for the number of distinct unit fractions needed.
They also prove double-exponential growth for two related quantities. If
counts representations of 1 by exactly
distinct positive unit fractions,
with increasing denominators, and
is the least integer at least
2 absent as a denominator from every such representation,
then
and
as
.