18
$\begingroup$

The concept of sparse polynomials has its place, and solvable but irreducible quadrinomial examples such as,

$$x^7-7x^4-14x^3-7=0$$ $$x^8+x^7+29x^2+29=0$$ $$x^9-27x^4-9x^3-9^2=0$$ $$x^{12}-36x^5-12x^3-12^2=0$$

may be intriguing, especially the octic which needs the $29$th root of unity. For trinomials, there are infinitely many quintics, sextics, and octics that are solvable and irreducible, such as,

$$x^5-5x^2-3=0$$ $$x^6+3x+3=0$$ $$x^8+9x+9=0$$

For septics, there do not seem to be any as I asked in this ancient MO question. But there are nice reducible ones such as,

$$x^7+7x+8=0$$

where the sextic factor is solvable. For now, we focus on irreducible septic quadrinomials. Over the past ten years, I've only found TWO that do not arise from binomials. (Edit.) Peter Mueller recently found another two of the same kind,

$$x^7-7x^4-14x^3-7=0\tag1$$ $$x^7-98x^4+686x^2+7^3=0\tag2$$ $$x^7+7x^4−35x+54 = 0\tag3$$ $$x^7+7x^3+21x+50 = 0\tag4$$

In Magma notation, the first one is a $7T2$ while the other three are $7T4$.

Question: Are there in fact infinitely many septic quadrinomials (without scaling) that are solvable and irreducible?


Update: Within a space of 24 hours, thanks to Jeremy Rouse, Joachim Konig, and yours truly, we have added 5 more quadrinomials, all of which have order $42$ so are $7T4$. However, these have a root in $\mathbb{Q}[\sqrt[7]z]$ where $z$ is an integer so arise from binomials.

Let $\alpha=14,$ $$x^7 - 6\alpha\, x^4 + \alpha^2\,x^3 - 2\alpha^2 = 0\tag1$$ $$x^7 - 2\alpha^3\, x^2 + 2\alpha^4\,x - 6\alpha^4 = 0\tag2$$

Let $\beta=21,$ $$x^7 - 10\beta\, x^4 + 70\beta^2\,x - 215\beta^2= 0\tag3$$

Let $m=5,$ $$x^7 + 14m^3\, x^3 - 42m^3\,x^2 - 13^2m^4 = 0\tag4$$

Let $n=57,$ $$x^7 + 14n^2\,x^4 - 2(7n)^3\,x^2 - 13^2(\sqrt7\,n)^4= 0\tag5\\\\$$

with fields $\mathbb{Q}[\sqrt[7]2]\,$, $\mathbb{Q}[\sqrt[7]{3^2\times5^3}]\,$, $\mathbb{Q}[\sqrt[7]5]\,$, $\mathbb{Q}[\sqrt[7]{3^2\times19^3}]$, respectively. Are there any more?

P.S. Incidentally, there are infinitely many irreducible but solvable quadrinomials of $9$th and $12$th degrees,

$$x^9-3abc\,x^4+a^3c\,x^3+b^3c^2 = 0$$ $$x^{12}-3abc\,x^5+a^3c\,x^3+b^3c^2 = 0$$

for arbitrary ($a,b,c$) and roots in $\mathbb{Q}[\sqrt[3]c],$ examples given at the start of the post, but a general form for the septic (if any) seems more difficult.

$\endgroup$
13
  • 4
    $\begingroup$ The Klueners and Malle database is here. $\endgroup$ Commented Mar 28, 2023 at 15:47
  • 2
    $\begingroup$ I found one more: $x^{7} - 5488x^{2} + 76832x - 230496$ has a root in $\mathbb{Q}[\sqrt[7]{2}]$. $\endgroup$ Commented Mar 29, 2023 at 1:51
  • 2
    $\begingroup$ In the same vein, $x^7-84x^4+196x^3-392$ and $x^7+1750x^3−5250x^2−105625$ have a root in $\mathbb{Q}(\sqrt[7]{2})$ and $\mathbb{Q}(\sqrt[7]{5 })$ respectively. $\endgroup$ Commented Mar 29, 2023 at 4:38
  • 3
    $\begingroup$ $x^7 + 7x^3 + 21x + 50$ and $x^7 + 7x^4 - 35x + 54$ are two more examples with Galois group 7T4 which do not arise from binomials. $\endgroup$ Commented Mar 31, 2023 at 15:54
  • 1
    $\begingroup$ A few more polynomials: $2x^7 + 7x^6 - 14x^4 + 25$, $4x^7 + 7x^6 + 19x + 19$, $x^7 + 7x^6 + 259x + 370$, $7x^7 - 42x^2 - 14x + 59$. The first has Galois group 7T2, the other ones 7T4. Except possibly for the last one, these do not arise from binomials. The splitting field of the last one contains the $7$-th roots of unity, so it possibly originates from a binomial. $\endgroup$ Commented Apr 4, 2023 at 11:18

1 Answer 1

3
$\begingroup$

It turns out there are infinitely many solvable septic quadrinomials after all (though with a caveat), and the answer was under my nose all along. For example, recall that, $$e^{\pi\sqrt{163}} =640320^3+743.999999999999925\dots$$

Then the septic,

$$x^7+7\left(\tfrac{1-\sqrt{-7}}2\right)x^4+7\left(\tfrac{1+\sqrt{-7}}2\right)^3x = -640320$$

surprisingly is solvable in radicals. In general, this is my version of Klein's 7th order formula for the j-function (which fittingly is also known as Klein's invariant) and is given by,

$$x^7+7\left(\tfrac{1-\sqrt{-7}}2\right)x^4+7\left(\tfrac{1+\sqrt{-7}}2\right)^3x = \sqrt[3]{j(\tau)}$$

which is solvable in radicals and has a nice closed-form using Ramanujan's theta functions described here. And I found the RHS can have a polynomial version as,

$$x\left(x^6+7\left(\tfrac{1-\sqrt{-7}}2\right)x^3+7\left(\tfrac{1+\sqrt{-7}}2\right)^3\right) = \left(\frac{49n^2+13n+1}{n}\right)^{1/3}\left(\frac{n^2+5n+1}{n^2}\right)$$

solvable for any $n$. This septic can be solved by a rather complicated cubic so I will not include it for now.

One can get rid of the cube root by letting $x = m^{1/3}y$ where $m=\left(\frac{49n^2+13n+1}{n}\right)$ to get the slightly simpler,

$$m^2y^7+7\left(\tfrac{1-\sqrt{-7}}2\right)m\,y^4+7\left(\tfrac{1+\sqrt{-7}}2\right)^3y =\frac{n^2+5n+1}{n^2}$$

So the caveat is this quadrinomial contains a square root. It remains to be seen if there really is an infinite family with just rational coefficients.

$\endgroup$
4
  • 2
    $\begingroup$ For what it's worth, with $y$ a root of the last polynomial in the above answer, and $z$ a root of its complex conjugate polynomial (inside the field generated by $y$), the $\mathbb{Q}(n)$-minimal polynomial of $\sqrt{-7}(y-z)$ is of the form $y^7+ a(n) y^4 + b(n) y^3 + c(n) y + d(n)$, i.e., has only one term more than wanted. Of course that's not in itself an exciting property (solvable septics with five terms can just as well be obtained straightaway from Chebyshev polynomials), but maybe there is an extra trick getting rid of one more term? $\endgroup$ Commented Apr 5, 2023 at 4:50
  • $\begingroup$ @JoachimKönig That is an interesting option! Can you also do the reverse? The quintic version of this is $$x^3(x^2+5x+40) = j(\tau)$$ which is solvable in radicals for j-function $j(\tau)$. Can you convert that into a trinomial with square roots in the coefficients? $\endgroup$ Commented Apr 5, 2023 at 18:39
  • $\begingroup$ I'm not sure that's the right analogy. The polynomial in your comment comes from the icosahedral function, which arises from $t(x):=x^3(x^2+5x+40)$; over the $t$-line, thats branched at three points with ramification indices 2, 3 and 5. Solvable quintic trinomials arise from the fact that the $S_5$ quintic $x^5+tx+t$ has infinitely many specializations with Galois group of order 20, which comes from the fact that the degree-$6$ subfields in its Galois closure are rational fields. But that quintic has ramification indices 2,4,5, i.e., is not directly related to the above. $\endgroup$ Commented Apr 6, 2023 at 3:55
  • $\begingroup$ @JoachimKönig Well, it was worth a shot. By the way, the polynomial version of this is, $$x^3(x^2+5x+40) = \frac{(n^2+10n+5)^3}{n}$$ which is solvable for any non-zero $n$ and has Galois group of order 20. $\endgroup$ Commented Apr 6, 2023 at 5:02

You must log in to answer this question.

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