Schweitzer 2018
Read the problems at the official link. The 2018 Schweitzer problems are © the János Bolyai Mathematical Society, so we don't reproduce them here. Open the 2018 problems (PDF) from the University of Szeged page.
The topic labels, hints and solutions below are our own, written from scratch: our solutions, not official ones. If you spot a mistake, please use “Report a problem” at the bottom of the page.
P6 Number theory: Gaussian periods and fourth powers modulo 13
- Hint 1. The nonzero fourth powers mod 13 are $1, 3, 9$, and they form a subgroup. So it is enough to show that every prime factor $p\ne13$ of the value is $1$, $3$ or $9$ mod 13.
- Hint 2. Let $\zeta=e^{2\pi i/13}$. Sums of $\zeta^h$ over $h$ in a coset of $\{1,3,9\}$ are the four roots of the quartic.
- Hint 3. Work modulo a prime ideal above $p$, where $\zeta\mapsto\zeta^p$ acts like raising to the $p$-th power.
Our solution (not official)
Write $f(x)=x^4+x^3+2x^2-4x+3$.
Step 0: what must be shown. The group $(\mathbb Z/13)^\times$ is cyclic of order 12, so its fourth powers form the subgroup $H=\{1,3,9\}$ of order 3. Also $0=0^4$. The value $f(a)$ is never $0$ (by the rational root test, an integer root would divide 3, and $f(\pm1),f(\pm3)$ are $3,9,117,87$). If $13\mid d$, then $d\equiv0$ is a fourth power. Otherwise $d$ is a product of primes $p\ne13$ dividing $f(a)$, and since $H$ is closed under multiplication, it suffices to prove: every prime $p\ne13$ that divides some $f(a)$ is in $H$ modulo 13.
Step 1: the roots of $f$. Let $\zeta=e^{2\pi i/13}$. The number 2 generates $(\mathbb Z/13)^\times$, so the cosets of $H$ are $2^jH$, $j=0,1,2,3$. Put $$\eta_j=\sum_{h\in H}\zeta^{\,2^jh}.$$ For $r$ prime to 13, the automorphism $\sigma_r:\zeta\mapsto\zeta^r$ permutes the $\eta_j$ (it sends the coset $2^jH$ to $r2^jH$), and it fixes each $\eta_j$ exactly when $r\in H$. So the coefficients of $\prod_j(x-\eta_j)$ are fixed by every $\sigma_r$; they are rational algebraic integers, hence integers. Integers are pinned down by approximations to within $\tfrac12$, and a numerical evaluation gives $$\prod_{j=0}^3(x-\eta_j)=x^4+x^3+2x^2-4x+3=f(x).$$ The same kind of computation gives the discriminant $\Delta=\prod_{i<j}(\eta_i-\eta_j)^2=3^2\cdot13^3$. In particular the four $\eta_j$ are distinct.
Step 2: Frobenius. Let $p\ne13$ be a prime with $p\mid f(a)$, and let $\mathfrak P$ be a maximal ideal of $\mathbb Z[\zeta]$ containing $p$. For $z=\sum c_i\zeta^i\in\mathbb Z[\zeta]$, the binomial theorem in characteristic $p$ and Fermat’s little theorem give $$z^p\equiv\sum c_i^p\zeta^{ip}\equiv\sum c_i\zeta^{ip}=\sigma_p(z)\pmod{p}.$$ Since $0=f(a)=\prod_j(a-\eta_j)$ lies in the prime ideal $\mathfrak P$, we have $a\equiv\eta_j\pmod{\mathfrak P}$ for some $j$. Then $$\sigma_p(\eta_j)\equiv\eta_j^{\,p}\equiv a^p\equiv a\equiv\eta_j\pmod{\mathfrak P}.$$ If $p\notin H$ (mod 13), then $\sigma_p(\eta_j)=\eta_k$ with $k\ne j$, so $\eta_k-\eta_j\in\mathfrak P$, hence $\Delta\in\mathfrak P$ and $p\mid\Delta=3^2\cdot13^3$. That forces $p=3$, but $3\in H$. So in every case $p$ is $1$, $3$ or $9$ modulo 13, as needed.
How we checked it: script: the four periods multiply out to the quartic; its discriminant is $3^2\cdot13^3$; every prime factor of $f(a)$ for $|a|\le3000$ is $0$, $1$, $3$ or $9$ mod 13