Yahoo Answers is shutting down on May 4th, 2021 (Eastern Time) and the Yahoo Answers website is now in read-only mode. There will be no changes to other Yahoo properties or services, or your Yahoo account. You can find more information about the Yahoo Answers shutdown and how to download your data on this help page.

Is the sum of this infinite series zero?

The series starts:

1 - 1/2 - 1/3 + 1/4 - 1/5 + 1/6 - 1/7 - 1/8 + 1/9 + 1/10 - 1/11 - 1/12 - 1/13 + 1/14 + ...

The sign in front of 1/n is negative if the prime factorization of n has an odd number of factors (counted with multiplicity) and is positive otherwise. For example 12 = 2 * 2 * 3, so the prime factorization of 12 has 3 factors, so it is -1/12 in the sum.

Please provide proof that the infinite sum is zero, or explain why it is not zero.

Update:

@kb: I haven't checked all details of your answer, but it looks good so far. You may be interested in this paper of Borwein et al: http://www.cecm.sfu.ca/personal/pborwein/PAPERS/P2... . In that paper it is mentioned that this problem was proved by Landau (1899) in his thesis to be equivalent to the prime number theorem. I just found a translation of that on line here http://arxiv.org/pdf/0803.3787v2.pdf but haven't tried to read it.

1 Answer

Relevance
  • kb
    Lv 7
    6 years ago
    Favorite Answer

    The key point for me is that this series is formally equal to the infinite product

    Π_(p prime) (1 - 1/p + 1/p^2 - 1/p^3 + ...)

    = Π_(p prime) (1/(1 - (-1/p)), via infinite geometric series

    = Π_(p prime) p/(p+1).

    (This should be easy to prove by truncating the product at some large N, use the distributive property, and then show that the error term between the resulting 'truncated' series and your series goes to 0 as N→∞.)

    ----

    Then, to show that the series equals 0, it suffices to show that

    1/Π_(p prime) p/(p+1) = Π_(p prime) (p+1)/p is diverges to infinity.

    To this end, let P(x) = Π_(p ≤ x) (p+1)/p = Π_(p ≤ x) (1 + 1/p).

    Then, ln P(x) = Σ_(p ≤ x) ln(1 + 1/p).

    ----

    We next show that the related series Σ_(p prime) (1/p - ln(1 + 1/p))

    is convergent.

    We start with (via power series for ln(1 + t) with t = 1/p)

    -ln(1 + 1/p) = 1/p - 1/(2p^2) + 1/(3p^3) - 1/(4p^4) + ... .

    So, 1/p - ln(1 + 1/p) = 1/(2p^2) - 1/(3p^3) + 1/(4p^4) - ...

    Not only is this positive, but this is bounded above by 1/(2p^2),

    from the right side being a convergent alternating series.

    This allows us to apply the Comparison Test to the series

    Σ_(p prime) (1/p - ln(1 + 1/p)):

    Σ_(p prime) (1/p - ln(1 + 1/p))

    < Σ_(p prime) 1/(2p^2)

    < Σ(n = 1 to ∞) 1/(2n^2), which converges, being a multiple of a convergent p-series.

    Thus, Σ_(p prime) (1/p - ln(1 + 1/p)) is convergent.

    Moreover, Σ_(p ≤ x) (1/p - ln(1 + 1/p))

    = Σ_(p prime) (1/p - ln(1 + 1/p)) - Σ_(p > x) (1/p - ln(1 + 1/p))

    = Σ_(p prime) (1/p - ln(1 + 1/p)) + O( Σ(n > x) 1/(2n^2) ), as above

    = Σ_(p prime) (1/p - ln(1 + 1/p)) + O(1/x), via Riemann sums.

    = A + O(1/x), where A = Σ_(p prime) (1/p - ln(1 + 1/p)), which we know converges.

    ----

    Therefore,

    Σ_(p≤x) ln(1 + 1/p)

    = -Σ_(p≤x) (1/p - ln(1 + 1/p)) + Σ_(p≤x) 1/p

    = -(A + O(1/x)) + Σ_(p≤x) 1/p.

    However, Σ_(p≤x) 1/p = ln(ln x) + B + O(1/ln x) for some constant B.

    [See Theorem 4.12 in Apostol's "Introduction to Analytic Number Theorem" for a proof.]

    So, Σ_(p≤x) ln(1 + 1/p) = B - A + ln(ln x) + O(1/ln x).

    Exponentiate both sides, noting that e^t = 1 + O(t):

    exp{Σ_(p≤x) ln(1 + 1/p)} = e^(B-A) e^(ln(ln x)) (1 + O(1/ln x))

    ==> P(x) = C ln x + O(1), where C = e^(B-A).

    Finally, this shows that lim(x→∞) P(x) = ∞, and so the series in question equals 1/Π_(p prime) p/(p+1) = lim(x→∞) 1/P(x) = 0, as required.

    -------

    Fun problem!

Still have questions? Get your answers by asking now.