3
$\begingroup$

If $\sigma_k(n)=\sum_{d\vert n} d^k$, denote $$F_1(q)=\sum_{n\geq1}\sigma_1(n)\,q^n \qquad \text{and} \qquad F_3(q)=\sum_{n\geq1}n\cdot\sigma_2(n)\,q^n.$$

QUESTION. Assume the prime $p$ is either $2, 3$ or $5$. Is it true that $F_1(q)-F_3(q)\equiv F_1(q^p) \pmod p$? Are there other primes for which such a congruence holds?

Convention. The congruence should be interpreted coefficient-wise.

Postscript. Sorry, there was a typo in the definition of $F_3(q)$. Please read as $$F_3(q)=\sum_{n\geq1}n\cdot\sigma_{\mathbf{{\color{red}3}}}(n)\,q^n.$$

$\endgroup$
9
  • 1
    $\begingroup$ To confirm, can I interpret this as saying the following two statements: $$p\nmid n: \sum_{d|n} d-nd^2 \equiv 0 \pmod p$$ $$p|n: \sum_{d|n} d-nd^2 \equiv \sum_{d| \frac{n}{p}} d \pmod p$$ The second statement follows immediately since $n \equiv 0 \pmod p$, and the divisors $d$ in the left sum that have a factor of $p$ also vanish. Thus we only need to check the first statement. It's straightforward to check $$p\nmid n: \sum_{d|n} d-nd^2 \equiv 0 \pmod p$$ for $p=2$. Since these $n$ are odd, we have that $d(1-nd)$ has at least one of $d$ or $1-nd$ even. $\endgroup$
    – user196574
    Commented Aug 28, 2023 at 22:35
  • 1
    $\begingroup$ Finally, the statements above are not true for $(n,p) = (4,3)$ and $(n,p) = (4,5)$, so the statement is only true for $p=2$ of the three numbers considered. Please let me know if I'm misinterpreting anything or making any mistakes. $\endgroup$
    – user196574
    Commented Aug 28, 2023 at 22:59
  • 1
    $\begingroup$ Additionally, $(n,p) = (3,p)$ breaks the first statement for all remaining $p$ except $13$, for which $(n,p) = (4,13)$ breaks the first statement. Thus the statement is only true for $p=2$. $\endgroup$
    – user196574
    Commented Aug 28, 2023 at 23:04
  • $\begingroup$ I made silly mistake in the definition of $F_3(q)$, it should be $n\sigma_3(n)$ not $n\sigma_2(n)$. Then, the claims are correct. $\endgroup$ Commented Aug 28, 2023 at 23:24
  • $\begingroup$ Thanks, I should have guessed! $\endgroup$
    – user196574
    Commented Aug 29, 2023 at 0:13

1 Answer 1

6
$\begingroup$

I wrote an answer in the comment section, so I'll rewrite it here.


The claim of

$$F_1(q)-F_3(q)\equiv F_1(q^p) \pmod p$$

is equivalent to the following two cases

$$p\nmid n: \sum_{d|n} d-nd^3 \equiv 0 \pmod p$$ $$p|n: \sum_{d|n} d-nd^3 \equiv \sum_{d| \frac{n}{p}} d \pmod p$$

The second case is trivial, because if $p|n$, we have

$$\begin{split} \sum_{d|n} d-nd^3 & \equiv \sum_{d|n} d \pmod p \\ &\equiv \sum_{d|\frac{n}{p}} d \pmod p\end{split}$$


Thus we only need to check $$p\nmid n: \sum_{d|n} d-nd^3 \equiv 0 \pmod p$$

For $p=2$, this is immediate as it holds for each term in the sum individually. That is, we necessarily have $n \equiv 1 \pmod 2$, and $d-nd^3 \equiv d-d^3 \equiv d(1-d)(1+d) \equiv 0 \pmod 2$.


For $p=3$ and $p=5$, it is worthwhile to write

$$\sum_{d|n} d-nd^3 = \sum_{d|n\text{ and } d<\sqrt{n}} (d-nd^3)+\left(\frac{n}{d}-n\left(\frac{n}{d}\right)^3\right)$$

My claim is that each term in the sum on the right is congruent to $0$ mod $3$ and $5$.

This follows from

$$(d-nd^3)+\left(\frac{n}{d}-n\left(\frac{n}{d}\right)^3\right) = \frac{1}{d^3}(d^2+n)(d^2-nd^4+n^2d^2-n^3)$$

We can now proceed by cases, remembering that $p \nmid n$ and hence $p \nmid d$.


For $p=3$, first note that $d^2 \equiv 1 \pmod 3$.

Then we enjoy

$$\begin{split} \frac{1}{d^3}(d^2+n)(d^2-nd^4+n^2d^2-n^3) &\equiv d (1+n)(1-n+n^2-n^3) \pmod 3 \\ \equiv d(1+n)(1-n)(1+n^2) \pmod 3 \end{split}$$

Note that we have $n \equiv 1 \pmod 3$ or $n \equiv 2 \pmod 3$, and plugging those in gives congruence to $0$.

Thus we have confirmed the $p=3$ case.


For $p=5$, first note that $d^2 \equiv 1 \pmod 5$ or $d^2 \equiv 4 \pmod 5$.

In the first case,

$$\begin{split} \frac{1}{d^3}(d^2+n)(d^2-nd^4+n^2d^2-n^3) & \equiv d(1+n)(1-n)(1+n^2) \pmod 5 \end{split}$$

In the second case,

$$\begin{split} \frac{1}{d^3}(d^2+n)(d^2-nd^4+n^2d^2-n^3) & \equiv d(4+n)(4-n)(1+n^2) \pmod 5 \end{split}$$

It's straightforward to check that $n \equiv 1,2,3,4 \pmod 5$ plugged into the above will vanish.

Thus we have confirmed the $p=5$ case.


We can rule out the claim of $$F_1(q)-F_3(q)\equiv F_1(q^p) \pmod p$$ for all other $p$ by noting that for $n=2$, $$\sum_{d|n} d-nd^3 \equiv -15 \pmod p$$

This is a problem for prime $p>5$, as then we have $p \nmid n$ but $\sum_{d|n} d-nd^3$ is not congruent to $0$.

(As an aside, note that this isn't a problem for $p=2$, since then $p|n$ and it's straightforward to check that it's congruent to $\sum_{d|\frac{n}{p}} d$, as needed. This isn't a problem for $p=3,5$, since then $p \nmid n$ and we clearly have congruence to $0$, as needed.)


EDIT: I realized that in proving the $p=3,5$ cases for $p \nmid n$ that there was an edge case I needed to check. I check this edge case here.

To prove the cases of $p=3,5$, I used

$$\sum_{d|n} d-nd^3 = \sum_{d|n\text{ and } d<\sqrt{n}} (d-nd^3)+\left(\frac{n}{d}-n\left(\frac{n}{d}\right)^3\right)$$

This works whenever $\sqrt{n}$ is not an integer. When $\sqrt{n}$ is an integer, it's a valid divisor, and in that case we'll have one more term

$$\sum_{d|n} d-nd^3 = \sqrt{n}-n(\sqrt{n})^3 + \sum_{d|n\text{ and } d<\sqrt{n}} (d-nd^3)+\left(\frac{n}{d}-n\left(\frac{n}{d}\right)^3\right)$$

So far, I showed that the rightmost sum is congruent to $0$ mod $3$ and $5$, so I only need to check this extra term. This extra term is

$$ \sqrt{n}-n(\sqrt{n})^3 = \sqrt{n}(1-n)(1+n)$$

For the case of $p=3$, $n \equiv 1 \pmod 3$, since $n$ is a square number not divisible by $3$. For the case of $p=5$, $n \equiv 1 \pmod 5$ or $n \equiv 4 \pmod 5$, since $n$ is a square number not divisible by $5$. Plugging these values in confirms congruence to $0$.

$\endgroup$

You must log in to answer this question.

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