ON THE ERDÖS-STRAUS CONJECTURE

EUGEN J. IONASCU and ANDREW WILSON

Paul Erdös conjectured that for every n ∈ N, n ≥ 2, there exist a, b, c natural

numbers, not necessarily distinct, so that n4 = a1 + 1b + 1c (see [3]). In this paper
we prove an extension of Mordell’s theorem and formulate a conjecture which is
stronger than Erdös’ conjecture.
AMS 2010 Subject Classification: 11A07, 11A41, 11A51.
Key words: Egyptian fractions, integers, prime numbers.

1. INTRODUCTION

The subject of Egyptian fractions (fractions with numerator equal to one

and a positive integer as its denominator) has incited the minds of many people
going back for more than three millennia and continues to interest mathemati-
2
cians to this day. For instance, the table of decompositions of fractions 2k+1
as a sum of two, three, or four unit fractions found in the Rhind papyrus has
been the matter of wonder and stirred controversy for some time between the
historians. Recently, in [1], the author proposes a definite answer and a full
explanation of the way the decompositions were produced. Our interest in this
subject started with finding decompositions with only a few unit fractions.
It is known that every positive rational number can be written as a finite
sum of different unit fractions. One can verify this by using the so called
Fibonacci method and the formula n1 = n+1 1 1
+ n(n+1) , n ∈ N. For more than
4
three fourths of the natural numbers n, n can be written as sum of only
two unit fractions: the even numbers, and the odd numbers n of the form
2
n = 4k − 1, via the identities 2k−1 = k1 + k(2k−1)
1 4
, and 4k−1 = k1 + k(4k−1)
1
,
k ∈ N. It is clear that if we want to write
4 1 1 1
= + + , a, b, c ∈ N,
n a b c
we can just look at primes n. However, just as a curiosity, for n = 2009 =
72 (41) (multiple of four plus one) one needs still two unit fractions, and there
are only three such representations
4 1 1 1 1 1 1
= + = + = + .
2009 504 144648 574 4018 588 3444
REV. ROUMAINE MATH. PURES APPL., 56 (2011), 1, 21–30
22 Eugen J. Ionascu and Andrew Wilson 2

This follows from the following characterization theorem which is well

known (see [2] and [4]).
Theorem 1. Let m and n two coprime positive integers. Then
m 1 1
= +
n a b
for some positive integers a and b, if and only if there exist positive integers
x and y such that
(i) xy divides n, and
(ii) x + y ≡ 0 (mod m).
In what follows we will refer to the equality
4 1 1 1
(1) = + + .
n a b c
and say that n has a representation as in (1), or that (1) has a solution, if
there exist a, b, c (a ≤ b ≤ c) natural numbers, not necessarily different
satisfying (1). Since n4 > a1 , the smallest possible value of a is n/4. The
biggest possible value of a is  3n 4 1 1 1
4 , for instance 9 = 6 + 6 + 9 .
If n = 4k + 1, k ∈ N, then we can try to use the smallest value first for
a, i.e., a = k + 1:
4 1 3
(2) = + .
4k + 1 k + 1 (k + 1)(4k + 1)
Now, if the second term in the right hand side of (2) could be written
as a sum of two unit fractions we would be done. This is not quite how the
things are in general, but if we analyze the cases k = 3l + r with r ∈ {0, 1, 2},
l ≥ 0, l ∈ Z, we see that there is only one excepted case in which we get stuck:
k = 3l. This because Theorem 1 can be used in one situation: k = 3l + 1
implies 1 + (3l + 1 + 1) ≡ 0 (mod 3). On the other hand, if k = 3l + 2 we
get (k + 1) = (3l + 2) + 1 = 3(l + 1) ≥ 3 and the second term is already a
unit fraction.
In order to simplify the statements of some of the facts in what follows
we will introduce a notation. For every i ∈ N let Ci be defined by
 n + 4i − 1 
Ci := n | (1) has a solution with a ≤ .
4
It is clear that Ci ⊂ Ci+1 and then Erdös-Straus’ conjecture is equivalent
to Ci = N. Thus we obtained a pretty simple fact about the Diophantine
i∈N
equation (1):
Proposition 1. The equation (1) has at least one solution for every
prime number n, except possibly for those primes of the form n ≡ 1 (mod 12).
3 On the Erdös-Straus conjecture 23

Moreover,
N \ C1 ⊂ {n | n ≡ 1 (mod 12)}.
We observe that 12 = 22 (3), a product of a combination of the first
two primes. The first prime that is excluded in this proposition is 13. The
equality (2) becomes
4 1 3
(3) = + .
12l + 1 3l + 1 (3l + 1)(12l + 1)
At this point we can do another analysis modulo any other number as
long we can reduce the number of possible situations for which we cannot say
anything about the decomposition as in (1). It is easy to see that 3l + 1 is
even if l is odd and then Theorem 1 can be used easily with x = 1 and y = 2.
This means that we have in fact an improvement of Proposition 1:
Proposition 2. The equation (1) has at least one solution for every
prime number n, except possibly for those primes of the form n ≡ 1 (mod 24).
In fact,
N \ C1 ⊂ {n | n ≡ 1 (mod 24)}.
Let us observe that 24 + 1 = 52 , 48 + 1 = 72 , which pushes the first prime
excluded by this last result to 73. Quite a bit of progress if we think in terms
of the primes in between that have been taken care off, almost by miracle.
If n = 24k + 1, then the smallest possible value for a is 6k + 1 and at this
point let us try now the possibility that a = 6k + 2 = n+7 4 ,
4 1 7
(4) = + , k ∈ N.
24k + 1 6k + 2 2(3k + 1)(24k + 1)
In the right hand side of (4), the second term has a bigger numerator
but the denominator has now at least three factors. This increases the chances
that Theorem 1 can be applied and turn that term into a sum of only two unit
fractions. Indeed, for k = 7l + r, we get that n = 24k + 1 ≡ 0 (mod 7) if r = 2,
2(3k + 1) + 1 ≡ 0 (mod 7) if r = 3, n + 1 = 2(12k + 1) ≡ 0 (mod 7) if r = 4,
and n + 2 = 24k + 3 ≡ 0 (mod 7) if r = 6. Calculating the residues modulo
168 in the cases r ∈ {0, 1, 5} we obtain:
Proposition 3. The equation (1) has at least one solution for every
prime number n, except possibly for those primes of the form n ≡ r (mod 168),
with r ∈ {1, 52 , 112 }, k ∈ Z, k ≥ 0. More precisely,
N \ C2 ⊂ {n | n ≡ 1, 52 , 112 (mod 168)}.
Let us observe that 168 = 23 (3)(7), 168 + 1 = 132 , and the excepted
residues modulo 168 are all perfect squares. Because of this, somehow, the
first prime that is excluded by this result is 193 = 168 + 25. Again, we have
24 Eugen J. Ionascu and Andrew Wilson 4

even a higher jump in the number of primes that have been taken care of. As
we did before there is an advantage to continue using (4) and do an analysis
now on k modulo 5.
For k = 5l + r, we have n ≡ 0 (mod 5) if r = 1, 3k + 1 ≡ 0 (mod 5) if
r = 3, and 6k + 1 ≡ 0 (mod 5) if r = 4, which puts n ∈ C2 again. Therefore,
we have for r ∈ {0, 2} the following excepted residues modulo 120.
Proposition 4. The equation (1) has at least one solution for every
prime number n, except possibly for those primes of the form n ≡ r (mod 120),
with r ∈ {1, 72 }, k ∈ Z, k ≥ 0. More precisely,
N \ C2 ⊂ {n | n ≡ 1, 72 (mod 120)}.
One can put these two propositions together and get Mordell’s Theorem.
Theorem 2 (Mordell, [5]). The equation (1) has at least one solution
for every prime number n, except possibly for those primes of the form n =
840k + r, where r ∈ {1, 112 , 132 , 172 , 192 , 232 }, k ∈ Z, k ≥ 0. Moreover,
we have
N \ C2 ⊂ {n | n ≡ 1, 112 , 132 , 172 , 192 , 232 (mod 840)}.
Proof. By Proposition 3, n = 168k + 1 may be an exception but if
k = 5l + r, with r ∈ {0, 1, 2, 3, 4} we have n ≡ 1 or 72 (mod 120) only for
r ∈ {0, 1}. These two cases are the exceptions for both propositions and they
correspond to n ≡ 1 or 132 (mod 840). All other excepted cases are obtained
the same way. 
Let us observe that 840 = 23 (3)(5)(7) and the residues modulo 840 are
all perfect squares. Not only that but 840 + 1 = 292 , 840 + 112 = 312 , and
1009 = 840 + 132 is the first prime that is excluded by this important theorem.
While 193 is the 44th prime number, 1009 is the 169th prime. It is natural to
ask if a result of this type can be obtained for an even bigger modulo. We will
introduce here the next natural step into this analysis, which implies to allow
a be the next possible value, i.e., n+11
4 , and we will be using the identities
4 1 11
(5) = + , k ∈ N,
120k + 1 30k + 3 3(10k + 1)(120k + 1)
4 1 11
(6) = + , k ∈ N.
120k + 49 30k + 15 3(5)(2k + 1)(120k + 49)

2. THE ANALYSIS MODULO 11

According to Proposition 4 we may continue to look only at the two cases

modulo 120 and use only the two formulae above. If we continue the analysis
modulo 11 in these two cases we obtain the following theorem.
5 On the Erdös-Straus conjecture 25

Theorem 3. The equation (1) has at least one solution for every prime
number n, except possibly for those primes of the form n = 1320k + r, where

r ∈ {1, 72 , 132 , 172 , 192 , 232 , 292 , 312 , 7(103), 1201, 7(127), 23(47)} := E,
k ∈ Z, k ≥ 0.

Moreover, we have

N \ C3 ⊂ {n | n ∈ E (mod 1320)}.

Proof. If n = 120k + 1 and k = 11l + 1, we see that n ≡ 0 (mod 11) and

so (5) gives the desired decomposition as in (1) right away. If k = 11l + r and
r ∈ {2, 4, 5} the Theorem 1 can be employed to split the second term in (5)
as a sum of two unit fractions. For instance, for r = 5 we have 1 + 3(10(11l +
5) + 1) ≡ 0 (mod 11), so one can take m = 11, x = 1 and y = 30k + 3 in
Theorem 1. Hence we have seven exceptions in this situation:
• r = 0 corresponds to n ≡ 1 (mod 1320),
• r = 3 gives n ≡ 192 (mod 1320),
• r = 6 corresponds to n ≡ 7(103) (mod 1320),
• r = 7 gives n ≡ 292 (mod 1320),
• r = 8 corresponds to n ≡ 312 (mod 1320),
• r = 9 gives n ≡ 23(47) (mod 1320), and finally
• r = 10 corresponds to n ≡ 1201 (mod 1320).
If n = 120k + 49 and k = 11l + r, then for r = 5 we have n ≡ 0 (mod 11).
If r ∈ {3, 5, 6, 8, 9, 10} we can use Theorem 1. The exceptions then are:
• r =0 corresponds to n ≡ 72 (mod 1320),
• r =1 gives n ≡ 132 (mod 1320),
• r =2 corresponds to n ≡ 172 (mod 1320),
• r =4 gives n ≡ 232 (mod 1320),
• r =7 corresponds to n ≡ 7(127) (mod 1320). 
Putting Theorem 2 and Theorem 3 together we get the following 36
exceptions:

12 132 172 192 232 292

312 1201 372 412 432 13(157)
472 2521 19(139) 2689 532 3361
592 3529 612 29(149) 672 712
13(397) 732 5569 17(353) 31(199) 792
832 7561 7681 892 8089 8761
26 Eugen J. Ionascu and Andrew Wilson 6

The residue 1201, the first prime in this list is not really an exception
because of the following identity
4 1 1
(7) = + +
9240k + 1201 2310k + 308 5(9240k + 1201)(15k + 2)
1
+ ,
770(9240k + 1201)(15k + 2)
which shows that 9240k + 1201 ∈ C8 for all k ∈ Z, k ≥ 0. We checked for
similar identities and found just another similar identity for the exception
17(353) = 6001:
4 1 1
(8) = + +
9240k + 6001 2310k + 1540 385(9240k + 6001)(2034k + 1321)
1
+ ,
22(3k + 2)(2034k + 1321)
which shows that 9240k + 6001 ∈ C40 for all k ∈ Z, k ≥ 0.

Theorem 4. The equation (1) has at least one solution for every prime
number n, except possibly for those primes of the form n ≡ r (mod 9240) where
r is one of the 34 entries in the table:

12 132 172 192 232 292

2 2 2 2
31 37 41 43 13(157) 472
2521 19(139) 2689 532 3361 592
3529 612 29(149) 672 712 13(397)
2 2 2
73 5569 31(199) 79 83 7561
7681 892 8089 8761

Moreover, if n is not of the above form, it is in the class C3 , or in C8 if

n ≡ 1201 (mod 9240), or in C40 if n ≡ 6001 (mod 9240).

Proof. We look to see for which values of r ∈ {0, 1, 2, 3, 4, 5, 6} we have

1320(7l + r) + s ∈ {1, 52 , 112 } (mod 168) with s ∈ E (E as in Theorem 3). For
instance, if s = 192 we get that r must be in the set {0, 1, 3} in order to have
1320(7l + r) + 192 ∈ {1, 52 , 112 } (mod 168). These three cases correspond to
residues 192 , 412 and 29(149) modulo 9240. Each residue in E gives rise to
three exceptions. We leave the rest of this analysis to the reader. 
It is important to point out that these residues appear also in [9] but
as a result of an algorithm which is only described there. Their idea is based
on a result proved in [6] which basically uses sufficient conditions to solve the
7 On the Erdös-Straus conjecture 27

equation (1)
n = 4ab(cd − b) − c or cn + 1 = 4ab(cd − b), n, a, b, c, d ∈ N.
The reader can verify that in the first case we have
4 1 1 1
= + +
n ad(cd − b) nad(cd − b) nabd
and for the second condition
4 1 1 1
= + + .
n ad(cd − b) abd nab(cd − b)
Let us denote by δ(r) a divisor of r. Terzi’s program is based on three different
ways of writing the first condition (above) and another way of writing the
second condition
n = 4αβk − δ(α + β), n = 4αβk − 4αδ(α) − β
2
n = 4αβk − δ(4αβ + 1), and n = (4αβ − 1)k − 4αδ(α).
Also, Terzi [9] provides a list of 198 exceptional residues for the modulo 120120.
For two of the exceptions that they have there, 2521 and 9601, we have found
the following identities
4 1
(9) = +
120120k + 2521 30030k + 4004
1 1
+ + , k ≥ 0,
1001(120120k + 2521)(810k + 17) 22(15k + 2)(810k + 17)
and also for k ≥ 0,
4 1
= +
120120k + 9601 2436 + 30030k
1 1
+ + .
14(120120k+9601)(58+715k)(470+5880k) 6(58+715k)(470+5880k)
This shows that the program used in [9] was not exhaustive and the
method was completely different of ours. We have implemented the same idea
into a program, as in [9], and obtained different results than the ones stated.

3. NUMERICAL COMPUTATIONS AND COMMENTS

We observe that the first ten of these residues in Theorem 4 are all perfect
squares. In fact, all 19 squares of primes less than 9240 and greater than 112
are all excepted residues. There is something curious about the fact that all
the perfect squares possible are excepted. This may be related with the result
obtained by Schinzel in [7] who shows that identities such as (7), (8) and
others in this note, cannot exist if the residue is a perfect square. The same
28 Eugen J. Ionascu and Andrew Wilson 8

phenomenon is actually captured in Theorem 2 in [11]. The good news about

Theorem 2, Theorem 3, and Theorem 4, is that the first excepted residues
are all perfect squares or composite and moreover their number is essentially
increasing with the moduli.
With our analysis unfortunately, there are a few other composite and 9
prime residues that have to be excluded. The prime 2521 is only the 369th
prime and it is the first prime that is excluded by this theorem. However, a
decomposition with the smallest a possible is exhibited in the equality

4 1 1 1
= + + ,
2521 636 70588 5611746

which puts 2521 ∈ C6 . The other primes are in the smallest class C as follows

C1 C2 C3 C4 C5 C6
3361, 7681, 8089 3529, 5569 8761 2689 7561 2101, 2521

Clearly, one can continue this type of analysis by adding more primes to
the modulo which is at this point 9240. It is natural to just add the primes in
order regardless if they are of the form 4k + 1 or 4k + 3. We see that Erdös’
conjecture is proved to be true if one can show that the smallest excluded
residue for a set of moduli that converges to infinity is not a prime. One
way to accomplish this is to actually show that the pattern mentioned above
continues, i.e., the number of excluded residues which are perfect squares or
composite is essentially growing as the modulus increases. This is actually our
conjecture that we talked about in the abstract. Numerical evidence points
out that for residues r which are primes, we have 9240s + r ∈ Ck(s,r) with
k(s, r) bounded as a function of s. For example, 9240s + 2521 ∈ C12 for every
s = 1 . . . 100000 and the distribution through the smaller classes is

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12

10852 6444 5332 811 612 277 63 82 6 7 0 5
44.3% 26.31% 21.78% 3.3% 2.5% 1.13% 0.26% 0.34% 0.025% 0.029% 0% 0.021%

Now, if we add 13 to the factors we would have an analysis modulo

120120. It turns out that 2521 is not an modulo 120120 exception since we
have (9) which shows that 120120k + 2521 ∈ C3374 for all k ∈ Z, k ≥ 0.
We found similar identities for the residues 2689, 3529, 29(149), 5569, 31(199),
7561, and 7681 modulo 120120. This suggests that one may actually be able
to obtain Mordell type results for bigger moduli, in the sense that the perfect
squares residues appear essentially in bigger numbers, by implementing a finer
analysis that involves higher classes than C3 . It is natural to believe that this
9 On the Erdös-Straus conjecture 29

might be true, by taking into account that Vaughan [10] showed that

1 2/3
#{n ∈ N | n ≤ m, and (1) does not have a solution} ≤ e−c(ln m) , m ∈ N,
m

for some constant c > 0. This is saying, roughly speaking, that the proportion
of the those n ≤ m for which a writing with three unit fractions of 4/n goes to
1
zero a little slower than m as m → ∞. The first few primes that require a big-
ger class than the ones before are 2, 73, 1129, 1201, 21169, 118801, 8803369, . . . ,
corresponding to classes C1 , C2 , C3 , C4 , C8 , C15 , C27 , . . .. which shows a steep
increase in the size of classes relative to the number of jumps.
In [11], Yamamoto has a different approach from ours and obtains a
lesser number of exceptions at least for the primes involved in Theorem 4. For
each prime p of the form 4k + 3 between 11 and 97, there is a table in [11] of
exceptions for congruency classes r (n ≡ r (mod p)) that is used to check the
conjecture using a computer for al n ≤ 107 . In [3], Richard Guy mentions that
the conjecture is checked to be true for all n ≤ 1003162753. Nevertheless, it
seems that the conjecture has been checked for n ≤ 104 , see [8].
However, with our method we extended the search for a counterexample
from 1003162753 further for all n ≤ 4, 146, 894, 049. For our computations we
wrote a program that pushes the analysis for a modulus of M = 2, 762, 760 =
23 (3)(5)(7)(11)(13)(23). The primes chosen here are optimal, in the sense that
the excepted residues are in number less than the ones obtained by other
options. The first 12 exceptions in this case are 1, 172 , 192 , 292 , 312 , 372 ,
412 , 432 , 472 , 532 , 3361, and 592 . The number of these exceptions was 2299
but it is possible that our program was not optimal from this point of view.
Nevertheless, this meant that we had to check the conjecture, on average,
for every other ≈ 1201 integer. The primes generated, 889456 of them, are
classified according to the smallest class they belong to in the next tables:

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12

380547 228230 128494 61129 50853 17116 8459 9580 1836 1386 547 855
42.8% 25.7% 14.4% 6.9% 5.7% 2% 0.9% 1% 0.2% 0.15% 0.06% 0.096%

C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24
115 124 111 26 10 27 2 4 4 0 0 0
0.013% 0.014% 0.012% 0.003% 0.001% 0.003% 0.0002% 0.00045% 0.00045% 0% 0% 0%

C25 C26 C27

0 0 1
0% 0% 0.0001%
30 Eugen J. Ionascu and Andrew Wilson 10

So far, we have not seen a prime in a class Ck with k > 27. However, the result
obtained in [7] seems to imply that the minimum class index for each prime,
assuming the conjecture is true, should have a limit superior of infinity.

REFERENCES

[1] Abdulrahman A. Abdulaziz, On the Egyptian method of decomposing 2/n into unit
fractions. Historia Mathematica 35 (2008), 1–18.
[2] E.S. Croot III, Egyptian Fractions. Ph. D. Thesis, 1994.
[3] R. Guy, Unsolved problems in Number Theory. Third Edition, 2004.
[4] G.G. Martin, The distribution of prime primitive roots and dense egyptian fractions.
Ph. D. Thesis, 1997.
[5] L.G. Mordell, Diophantine equations. London–New York, Acad. Press, 1969.
[6] L.A. Rossati, Sull’equazione diofantea 4/n = 1/x1 + 1/x2 + 1/x3 . Bolettino della Unione
Matematica Italiana, serie III, Anno IX, 1(1954).
[7] A. Schinzel, On sums of three unit fractions with polynomial denominators. Funct. Ap-
prox. Comment. Math. 28 (2000), 187–194.
[8] A. Swett, http://math.uindy.edu/swett/esc.htm.
[9] D.G. Terzi, On a conjecture by Erdös-Straus. Nordisk Tidskr. Informationsbehandling
(BIT) 11 (1971), 212–216.
[10] R.C. Vaughan, On a problem of Erdos, Straus and Schinzel. Mathematika 17 (1970),
193–198.
[11] K. Yamamoto, On the diophantine equation. Memoirs of the Faculty of Science, Kyushu
University, Ser. A 19 (1965), 1, 37–47.

Received 25 July 2010 Columbus State University

Department of Mathematics
Columbus, GA 31907
ionascu eugen@colstate.edu