I'm not a math expert so this may be a trivial question; if $p_i$ is the $i$-th prime, let:

$$S(n) = \sum_{i=1}^n p_i$$

be the sum of the first $n$ primes and

$$P(n) = | \{1 \leq i \leq n \mid S(i) \mbox{ is prime} \} | $$

be the number of the sums $S(i), 1 \leq i\ \leq n$ that are prime. Do we have

$$\lim_{n \to \infty} \frac{P(n)}{n} = 0\;?$$

Where can I find a proof?

EDIT: I generated the sequence $P(n)$ and found it on OEIS: Numbers n such that n is prime and is equal to the sum of the first k primes for some k., but there is not a lot of information there.

  • 2
    $\begingroup$ See also oeis.org/A013916 although there is not much information there, either. $\endgroup$ Jan 5, 2014 at 16:17
  • 1
    $\begingroup$ A first step would be to understand how many of the $S(n)$ are multiples of a given number $d$. Already for $d=3$ I don't know how to determine the number of $S(n)$'s that are multiples of $3$. Such questions seem too hard to answer (although one can of course formulate conjectures). $\endgroup$
    – Lucia
    Jan 5, 2014 at 16:37
  • 1
    $\begingroup$ @TerryTao: The sums $S(n)$ alternate in parity. So $P(n)/n$ is trivially at most $1/2$. But what you say would apply to maybe get strictly below $1/2$. $\endgroup$
    – Lucia
    Jan 5, 2014 at 18:39
  • 3
    $\begingroup$ Yes, good point. If one uses Maynard (and Banks et al.) to ensure at least six primes amongst the $n+h_i$, then we can get S(i) divisible by 3 for infinitely many even i, and so one should indeed be able to get $P(n)/n$ below 1/2 by a modification of this argument (one has to allow the diameter of the admissible tuple to be as large as a small multiple of $\log n$). $\endgroup$
    – Terry Tao
    Jan 5, 2014 at 18:41
  • 1
    $\begingroup$ @TerryTao: I don't understand how one gets a constant below $1/2$ though. My understanding is that the work of Maynard et al does not yet give positive proportion of tuples. I.e. how does one get a positive proportion of values for which $S(n)$ is a multiple of $3$? (Infinitely many is fine, and really goes back to Daniel Shiu.) $\endgroup$
    – Lucia
    Jan 5, 2014 at 18:45

1 Answer 1


I wonder if for questions of this type, the following probabilistic paradigm is useful:

Instead of adding all prime numbers, let us "flip a coin" and include the next prime number with a fixed probability $0<q<1.$

In other words, what can we say about the distribution of primes in the sequence of sums

$$S_{\xi}(n)=\sum_{i=1}^n \xi_i p_i,$$

where $p_i$ is the $i$th prime and $\{\xi_i\}$ is a sequence of i.i.d. Bernoulli random variables? The asymptotic needs to be a.s. with respect to the Bernoulli measure.

  • $\begingroup$ Some users have commented that this is "not an answer". Indeed, it looks more like a question/tentative suggestion. Maybe it makes sense to make it CW? $\endgroup$
    – Todd Trimble
    Feb 5, 2014 at 20:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.