19
$\begingroup$

Let $n \geq 3$ be a natural number. Define the set $X_n$ as the set of natural numbers that appear as the number of real roots an irreducible polynomial of degree $n$ over $\mathbb{Q}$ which is solvable by radicals can have.

Example: In case $n=p$ is a prime, we have $X_p= \{1,p \}$.

Is $X_n$ or the cardinality $|X_n|$ also known for other values?

It would be interesting to see the beginning of the sequence $a_n=|X_n|$ for small values of $n$, maybe it appears in the oeis.

$\endgroup$
6
  • $\begingroup$ what (trivial or nontrivial) things do you know about $X_n$ in general? $\endgroup$
    – Arno Fehm
    May 10, 2020 at 20:20
  • $\begingroup$ @ArnoFehm I dont really know much about $X_n$ except for the stated result for prime numbers. $\endgroup$
    – Mare
    May 10, 2020 at 20:32
  • $\begingroup$ You certainly knew $X_4=\{0,2,4\}$, so is $X_6$ the first case you don't know? $\endgroup$
    – YCor
    May 10, 2020 at 20:36
  • $\begingroup$ @YCor Thanks, yes 6 is the first non-trivial case it seems. $\endgroup$
    – Mare
    May 10, 2020 at 20:38
  • $\begingroup$ GAP might provide for small $n$ the list of $k$ such that $S_n$ has a transitive solvable subgroup that contains a product of $k$ disjoint transpositions (which is a necessary condition for the existence of such a polynomial with $n-2k$ real roots). Nevertheless for $n=6,8$ this yields no obstruction. $\endgroup$
    – YCor
    May 10, 2020 at 20:47

2 Answers 2

14
$\begingroup$

Klueners and Malle have a database of number fields of degree $\leq 19$ that (tries) to include every Galois group and every possible signature. Examining this database shows that for composite $n$ with $4 \leq n \leq 18$, $X_{n} = \{ k : 0 \leq k \leq n \text{ and } k \equiv n \pmod{2} \}$.

EDIT: If $n = 2k$ is even, then $\mathbb{Z}/2\mathbb{Z} \wr \mathbb{Z}/k\mathbb{Z}$ contains elements of order $2$ with any desired even number of fixed points. It seems possible that one can obtain the same thing for $n$'s that are multiples of $3$. The smallest $n$ that does not fall into this category is $n = 25$, and using the classification of transitive subgroups of $S_{25}$, one finds that $X_{25} \subseteq \{ 1, 5, 9, 13, 17, 21, 25 \}$.

$\endgroup$
2
  • 3
    $\begingroup$ So quite surprisingly for primes it tends to be very small and for composite it's as large as possible... until 19. For every such $n$ and $k$ with $n-k$ even you manually checked inside the list the polynomials until finding one with $k$ real roots? or you found a quicker way? $\endgroup$
    – YCor
    May 10, 2020 at 21:12
  • $\begingroup$ Thanks, that motivates to experiment a bit also for some n>=20. I will try my luck a bit to see if the pattern continues. $\endgroup$
    – Mare
    May 10, 2020 at 22:40
7
$\begingroup$

Jensen On the number of real roots of a solvable polynomial includes a proof of:

Loewy’s theorem. Let $K$ be a real number field and $f(X)$ an irreducible polynomial in $K[X]$ of odd degree $n$. If $p$ is the smallest prime divisor of $n$ and the Galois group of $f(X)$ over $K$ is solvable, then $r(f) = 1$ or $n$ or satisfies the inequalities $p ≤ r(f)≤ n−p +1.$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.