1
$\begingroup$

From A248667:

The polynomial $p(n,x)$ is defined as the numerator when the sum $$1 + \frac{1}{nx + 1} + \frac{1}{(nx + 1)(nx + 2)} + \cdots + \frac{1}{(nx + 1)(nx + 2)\cdots(nx + n - 1)}$$ is written as a fraction with denominator $$(nx + 1)(nx + 2)\cdots(nx + n - 1)$$

The first six polynomials:

$$p(1,x) = 1$$ $$p(2,x) = 2 (1 + x)$$ $$p(3,x) = 5 + 12 x + 9x^2$$ $$p(4,x) = 4 (4 + 17 x + 28 x^2 + 16 x^3)$$ $$p(5,x) = 5 (13 + 84 x + 225 x^2 + 275 x^3 + 125 x^4)$$ $$p(6,x) = 2 (163 + 1455 x + 5562 x^2 + 10800 x^3 + 10368 x^4 + 3888 x^5)$$ Let $a(n)$ be the sequence of numbers $m$ for which coefficients of the polynomial $p(m,x)$ are relatively prime.

For example: $$p(3,x) = 5 + 12 x + 9x^2$$ $$\operatorname{gcd}(5,12,9)=1$$

The sequence begins: $$1, 3, 7, 9, 11, 17, 19, 21, 23, 27, 29, 31, 33, 41, 43, 47, 49, 51, 53, 57$$

Let $b(n)$ be the sequence of numbers $m$ for which $$\sum |\mu(p_j+1)|=0$$

Here $$m=\prod p_j^{k_j}$$ where $p_j$ are distinct prime divisors of $m$ and $k_j$ are their powers. Obviously $\mu(n)$ is the Moebius function.

The sequence begins $$3, 7, 9, 11, 17, 19, 21, 23, 27, 31, 33, 43, 47, 49, 51, 53, 57, 59, 63, 67$$

I conjecture that $b(n)$ is subsequence of $a(n)$.

Up to $a(60)=167$ there are only $12$ terms which are not belong to $b(n)$: $$1, 29, 41, 61, 73, 87, 101, 109, 113, 123, 137, 157$$

Is there a way to prove it?

$\endgroup$
1

1 Answer 1

7
$\begingroup$

Counterexample: $463 \in b(n)$ (it's a prime and $464 = 2^4 \cdot 29$ is not squarefree), but $463 \not \in a(n)$ because it's a factor of the GCD of the coefficients of $p(463, x)$.

$\endgroup$
7
  • 2
    $\begingroup$ The latter is equivalent to $\sum_{k=0}^{462} 1/k!$ being divisible by 463, is there any way to see this without computations? $\endgroup$ Commented Apr 8, 2022 at 6:32
  • 1
    $\begingroup$ @FedorPetrov, I computed it. The sequence of primes in $a(n)$ doesn't appear to be in OEIS. $\endgroup$ Commented Apr 8, 2022 at 11:08
  • $\begingroup$ @PeterTaylor, thank you for answer! Can you compute another counterexamples? At least 2 more to add the sequence to the OEIS. By the way, the numbers might be called your last name. $\endgroup$ Commented Apr 8, 2022 at 11:22
  • 1
    $\begingroup$ @Notamathematician, the counterexamples up to 10000 are 463, 1389, 3241, 4167, 5093, 7871, 8797, 9723, but absent a reason for expecting there to be a relationship between $a$ and $b$ I don't really think that the sequence is motivated enough to justify its inclusion in OEIS. $\endgroup$ Commented Apr 8, 2022 at 15:59
  • $\begingroup$ @PeterTaylor, thank you very much! All these numbers are divisible by 463 and the result of division is $b(n)$. $\endgroup$ Commented Apr 8, 2022 at 21:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .