In this post we denote the Dedekind psi function as $\psi(n)$ for integers $n\geq 1$. This is an important arithmetic fuction in several subjects of mathematics. As reference I add the Wikipedia Dedekind psi function.

An even perfect number is an even integer $n\geq 1$ satisfying that $\sigma(n)=2n$, where $\sigma(n)=\sum_{1\leq d\mid n}d$ is the sum of divisors function.

Claim. Invoking Euclid–Euler theorem it is easy to prove that each even perfect number satisfies $$4\psi(n)^2-(12n+3)\psi(n)=-9n^2,\tag{1}$$ alternatively $$\psi(n)=\frac{12n+3+3\sqrt{1+8n}}{8}.\tag{2}$$

Question. I would like to know if it is possible to get a characterization for even perfect numbers from $(1)$: what I evoke is if we can to prove that if an integer $n\geq 1$ satisfies $(1)$ then $n$ is an even perfect number. Many thanks.

The case $n=1$ is easy. By contradiction also is easy to rule out/discard that $n$ is a power of two and that $n$ is odd. Thus the only discussion is what happens with numbers of the form $$n=2^{\alpha}m\tag{3}$$ where $\alpha\geq 1$ is integer and $(2,m)=1$ for integers $m>1$.

Our case $(3)$ leads to

$$2^{\alpha+1}m\prod_{\substack{\text{primes }p\\p\mid m}}(p+1)^2-(2^{\alpha+2}m+1)\prod_{\substack{\text{primes }p\\p\mid m}}p(p+1)=2^{\alpha+1}m \prod_{\substack{\text{primes }p\\p\mid m}}p^2,\tag{4}$$ and if I was right from $(4)$ on assumption that $m=p$ is a prime (that is we assume that the number of distinct primes dividing $m$ is $\omega(m)=1$) I get that $p+1=\lambda 2^{\alpha+1}$ with $\lambda=1$ using the congruence $\prod_{\substack{\text{primes }p\\p\mid m}}(p+1)\equiv 0\text{ mod }2^{\alpha+1}$.

Thus if my argument works I need to rule out/discard the case of integers $m$ with more than $1$ distinct primes factors in their prime factorization (I want to remove the case $\omega(m)>1$).

  • $\begingroup$ If it is impossible to remove the case in which $m$ has at least two distinct prime factors in its factorization, or you find a mistake in the cases that I've studied, feel free to add a comment or your feedback. $\endgroup$
    – user759001
    Apr 14, 2020 at 20:09
  • 1
    $\begingroup$ Does applying the quadratic formula to Equation $(1)$ give Equation $(2)$? $\endgroup$ Apr 14, 2020 at 21:16
  • $\begingroup$ I removed my previous comment since I understand that you was asking just about clarification about $(2)$: yes (I think that there aren't typos), the quadratic formula gives $(2)$ from $(1)$. Many thanks for your attention @GeoffreyTrang , feel free to provide me your feedback about any of my unanswered questions in Mathematics Stack Exchange. $\endgroup$
    – user759001
    Apr 16, 2020 at 11:48
  • 1
    $\begingroup$ Stated another way, I believe you're asking for a proof of $\sigma_1(2\,n)=\sum\limits_{d|2\,n} d=4\,n\iff\psi(2\,n)=2\,n \prod\limits_{p|2\,n}\left(1+\frac{1}{p}\right)=\frac{24\,n+3+3\sqrt{1+16\,n}}{8}$ where $(n\land d)\in \mathbb{N}\land p\in \mathbb{P}$? $\endgroup$ Jul 18, 2021 at 15:04
  • $\begingroup$ Now I don't know, I'm in a library, and I did the calculations a year ago. Many thanks @StevenClark $\endgroup$
    – user759001
    Jul 19, 2021 at 8:56

1 Answer 1


If a positive integer $n$ satisfies $$4\psi(n)^2-(12n+3)\psi(n)+9n^2=0\tag1$$ then $n$ is an even perfect number.

Proof :

Solving $(1)$ for $\psi(n)$ gives $$\psi(n)=\frac{12n+3\pm 3\sqrt{1+8n}}{8}$$ There has to be a positive integer $k$ such that $1+8n=k^2$, so $$\psi\bigg(\frac{k^2-1}{8}\bigg)=\frac{12\cdot\frac{k^2-1}{8}+3\pm 3k}{8}=\frac 3{16}(k\pm 1)^2$$ Since $k$ has to be odd, there has to be a positive integer $m$ such that $k=2m+1$ to have $$\psi\bigg(\frac{m(m+1)}{2}\bigg)=\frac 3{4}\bigg(m+\frac{1\pm 1}{2}\bigg)^2$$ There has to be a positive integer $\ell$ such that $m+\dfrac{1\pm 1}{2}=2\ell$, so $$\psi\bigg(\ell(2\ell\mp 1)\bigg)=3\ell^2$$ Since $\psi$ is the multiplicative function with $\gcd(\ell,2\ell\mp 1)=1$, $$\psi(\ell)\psi(2\ell\mp 1)=3\ell^2$$ For $\ell=1$, this does not hold. Suppose here that $\ell\gt 1$ is odd. Then, LHS is even while RHS is odd, which is impossible. So, $\ell$ has to be even, and there has to be a positive integer $a,s$ ($s$ is odd) such that $\ell=2^as$, so $$\psi(2^a)\psi(s)\psi(2^{a+1}s\mp 1))=3\cdot 2^{2a}s^2,$$ i.e. $$3\cdot 2^{a-1}\psi(s)\psi(2^{a+1}s\mp 1)=3\cdot 2^{2a}s^2,$$ i.e. $$\psi(s)\psi(2^{a+1}s\mp 1)=2^{a+1}s^2$$

Suppose here that there are $a,s$ such that $\psi(s)\psi(2^{a+1}s+1)=2^{a+1}s^2$. Then, LHS is larger than RHS, which is impossible.

So, we have to have $$\psi(s)\psi(2^{a+1}s-1)=2^{a+1}s^2$$

Let $\displaystyle 2^{a+1}s-1=\prod_{i=1}^{N} p_i^{b_i}$ where $p_i$ are distinct primes and $b_i$ are positive integers. Then, $$\psi(s)\prod_{i=1}^{N} p_i^{b_i-1}(p_i+1)=\bigg(1+\prod_{i=1}^{N} p_i^{b_i}\bigg)s$$Since RHS is not divisible by any $p_i$, we have to have $b_i=1$ for which we have $$\psi(s)\prod_{i=1}^{N} (p_i+1)=\bigg(1+\prod_{i=1}^{N} p_i\bigg)s$$Suppose here that $N\geqslant 2$. Then, LHS is larger than RHS, which is impossible. Therefore, $N=1$ implies $\psi(s)=s$, i.e. $s=1$. So, $n$ has to be of the form $n=2^a(2^{a+1}-1)$ where $2^{a+1}-1$ is prime. This means that $n$ has to be an even perfect number. $\quad\blacksquare$

  • 1
    $\begingroup$ Many thanks, I'm trying to read the proof in next days before expires the bounty. I would like to dedicate the post, and this characterization (now yours) of even perfect numbers, to your person due your excellence in mathematics. $\endgroup$
    – user759001
    Jul 19, 2021 at 15:10

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.