1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Difficulty with the Erdos-Straus Conjecture

  1. Oct 4, 2003 #1
    The Erdos-Straus Conjecture proposes that:

    For n >= 1, 4/n = 1/a + 1/b + 1/c, has postive integer solutions.
    n = 2k (evens)
    n = 4k+3 (odds)
    n = 4k+1 (odds)

    For 2k I have found the pattern --> if k >= 1, 4/2k = 1/2k + 1/2k + 1/k. This is simple, I know.

    I am having difficulty finding the pattern for 4k+3 (no pattern has yet been found for 4k+1, I know).

    So what is the pattern for:

    4/4k+3 = 1/a + 1/b + 1/c


    Does anyone know this pattern? If you do, please reply with such pattern.

    Thank you
  2. jcsd
  3. Oct 5, 2003 #2


    User Avatar
    Gold Member

    it should be 2k>=1 not k>=1
  4. Oct 5, 2003 #3


    User Avatar
    Gold Member

    maybe: 4/4k+3=1/4k+3+1/4k+3+1/0.5*(4k+3)

    but im not sure this what you are after.
  5. Nov 19, 2004 #4
    The denominator of the last term in loop quantum gravity's equation [0.5*(4k+3)] will never be an integer, so it's not a valid unit fraction.

    The {a,b,c} triplet I came up with for 4/(4k+3) is:

    { 2(k+1), 2(k+1), (k+1)(4k+3) }

    This works too: { k+1, 2(k+1)(4k+3), 2(k+1)(4k+3) }.

    These two triplets have the same source: Substract 1/(k+1) from 4/(4k+3) and you get 1/(k+1)(4k+3). But since you now have an equation of the form

    4/(4k+3) = 1/a + 1/b

    you need to add a third term somehow to "fit" the conjecture. You can do so by splitting one of the right-hand terms as follows:

    1/a = 1/(2a) + 1/(2a).

    I've worked on this conjecture off and on the last 20 years but have never proved it. But I have found many results, including that the smallest {n} that is a counterexample--if one exists--is prime and equal to 1 mod 24.
    Last edited: Nov 19, 2004
  6. Nov 19, 2004 #5
    What a weird conjecture. Anyway 4k+1 with trivial solutions (making the substitution k = k - 0.5) is [tex]\frac {4}{4k+1} = \frac {1}{2(k+0.5)(4k+1)} + \frac {1}{2(k+0.5)(4k+1)} + \frac {1}{k+0.5}[/tex]. However, for n mod 4 = 1 (5, for example) neither this equation or the author above's works. Allan Swett (http://math.uindy.edu/swett/esc.htm) appears to be the most learned in this subject. Perhaps his page will give you some insights. Also, your solution to 1/2k I don't think is entirely correct because egyptian fractions are unique usually.
    Last edited by a moderator: Nov 19, 2004
  7. Aug 22, 2005 #6
    It's not that hard to find three different Egyptian fractions that add up to

    For any fraction of the form 4/(4k+3), compute the difference with the nearest integer fraction:

    4/(4k+3) - 1/(k+1) = 1/((k+1).(4k+3)).

    If you are satisfied with the absence of 1/c you are ready:
    4/(4k+3) = 1/(k+1) + 1/((k+1).(4k+3))

    Otherwise, the problem is to write any of the two fractions at the right hand side as the sum of two other - different - Egyptian fractions. You can use

    1/N = 1/(N+1) + 1/(N.(N+1)) for any N giving two schemes:

    for N = k+1 we get
    4/(4k+3) = 1/(k+2) + 1/((k+1).(k+2)) + 1/((k+1).(4k+3)) and

    for N=(k+1).(4k+3) we get
    4/(4k+3) = 1/(k+1) + 1/((k+1).(4k+3) + 1) + 1/((k+1).(4k+3).((k+1).(4k+3) + 1)).

    Since by the same method, we can easily find solutions for even n, the real problem is in the case that n = 1 mod 4. For all but some cases, the Erdös-Straus conjecture has been proved.

    Except for the case where n = 1^2, 11^2, 13^2, 17^2, 19^2, or 23^2 mod 840. Note that these are all subcases of the case n = 1 mod 4.

    See p.34 of http://www.kevingong.com/Math/EgyptianFractions.pdf for conclusion and more references.
  8. Aug 22, 2005 #7


    User Avatar
    Homework Helper

    Please recheck the problem... For n = 1:
    [tex]4 = \frac{1}{a} + \frac{1}{b} + \frac{1}{c} \leq 3[/tex]
    [tex]\Leftrightarrow 4 \leq 3[/tex]!!!!!
    So it's wrong for n = 1.
    I think it should reads:
    For n > 1, 4/n = 1/a + 1/b + 1/c, has postive integer solutions.
    Viet Dao,
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook