Consider the set of $k$ numbers $\{n+c_1,n+c_2,n+c_3,...,n+c_k\}$ where $c_i$ are constant positive integers and $n$ is a varying positive integer such that $n \leqslant M$. What is the probability that there are $x$ numbers in this set which are prime, where $0 \leqslant x \leqslant k$ ? If this is really hard, is there any answer to cases such as $x=0$ and $x=1$ ?

At $k=1$, we have to consider whether $n+c_1$ is prime or composite. We know that $M < n+c_1 \leqslant M+c_1$. Thus, our probability $P(k,M,c_i)$ would be: $$ P(1,M,c_1) = \frac{\pi(M+c_1)-\pi(M)}{c_1} \sim \frac{(M+c_1)\ln M-M\ln(M+c_1)}{c_1 \ln M \ln(M+c_1)}$$

Is the working for $k=1$ right? Can I directly apply Prime Number Theorem for $k \geqslant 2$ in a similar fashion. I am doubtful as we are dealing with a group of equally interspaced numbers. Any help or ideas are accepted. Thanks in advance!

  • $\begingroup$ Probability wrt what probability distribution on $n$ ? The uniform distribution ? Then it is a matter of counting some particular $l$-uples of primes $\le M$ $\endgroup$ – reuns Jan 8 at 8:13

It's a well known fact that the average distance between primes about as large as $n$ is $\ln n$. So the probability for a number as large as $n$ to be prime is about $1/\ln n$.

Probability that your $k$-th number is prime is therefore:

$$p_k=\frac{1}{\ln (n+c_k)}$$

Probability that your $k$-th number is not prime is:

$$\bar{p}_k=1-\frac{1}{\ln (n+c_k)}$$

For $x=0$ all numbers must be composite so the probability is:

$$P=\prod_{i=1}^k \bar{p}_i$$

For $x=1$, exactly one number has to be prime so the probability is:


It's a little bit more difficult to calculate probability for $x>1$ but it's still a fairly straightforward task.

  • $\begingroup$ This is exactly what I had initially done. But can the prime number theorem directly be applied for this question? We are dealing with a group of numbers with constant distance between them. $\endgroup$ – Haran Jan 8 at 12:59
  • 1
    $\begingroup$ This might work as a heuristic/rule of thumb, but not more strictly: the probabilities are not independent and thus cannot be plainly multiplied together. $\endgroup$ – Mees de Vries Jan 8 at 13:11
  • 1
    $\begingroup$ In particular, if $k = 2, c_2 = c_1 + 2, x = 2$, the twin prime conjecture essentially asks, "is there $c_1$ such that for any $M$ the probability equals 0". $\endgroup$ – Mees de Vries Jan 8 at 13:15

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.