2

I am looking for a way to find a closed form of the real root of the quintic eq. with integer coefficients: $x^5+3x^4+4x^3+x-1=0$.

According to the numerical calculation the root $x_0\approx 0.4591337233102075...$. I am interested in representing the root via a function to use it in further calculation. Probably there is not any closed form, but what about other ways to get a function which produces the constant? Any suggestions?

Just to be more clear the root appears in the problem I posted a few years ago: Bath towel on the rope: minimize the area of self-intersection of a folded rectangle Thank you for any help.

IV_
  • 6,964
  • 2
    The Galois group of the polynomial is $S_5$ and in particular not solvable, so there is no closed form for any root in terms of radicals. For an exact representation there are various choices; see: https://en.wikipedia.org/wiki/Bring_radical – Travis Willse Jul 12 '23 at 09:25
  • Even the formulas for degree $3$ and $4$ are too complicated , so usually not used. Degree $5$ and more is in general not solvable by radicals , and all approaches for "exact closed forms" in fact need after all numerical methods , so it is not wirth the effort and you can apply the numerical method directly. – Peter Jul 12 '23 at 09:35
  • Any series representation? I'd like to add 'Bath rowel on the rope' problem to OEIS which usually has formula to calculate. – Mikhail Gaichenkov Jul 12 '23 at 09:44
  • “Representing the root via a function to use it in further calculation”. FWIW, a general quintic equation can be solved using the Jacobi theta function. – KStarGamer Aug 16 '23 at 12:39
  • @KStarGamer Yes. And for added info, the Jacobi theta function can also be expressed as quotients of the Dedekind eta function. – Tito Piezas III Aug 16 '23 at 13:42
  • @TitoPiezasIII Yes indeed. – KStarGamer Aug 16 '23 at 13:49

2 Answers2

4

We'll give a closed-form to your quintic root. As Klein showed in his "Lectures on the Icosahedron", there is indeed such a thing though it goes beyond radicals.

Two special functions may be used: the Dedekind eta function $\eta(\tau)$ and the hypergeometric function $_2F_1$. The Bring quintic form will be avoided since that involves $2$nd-deg and $4$th-deg transformations, and to undo those will result in a mess.

Fortunately, there are other solutions. And the nice thing about the Klein-Brioschi result is one just uses nested square roots (unlike the Bring form which needs cube roots).


I. Principal quintic

First reduce the quintic to principal form, and yours is special in that its Tschirnhausen transformation needs only rational coefficients. Given,

$$x^5+3x^4+4x^3+x-1=0$$

Let $(x+1)^2-y=0$. Eliminating $x$ between the two using resultants in WA yields the principal quintic form,

$$y^5 - 8y^2 + 4y - 16=0$$

which has discriminant $d = 2^{23}\times47.$ The prime $47$ will appear in the second method.


II. First Method

We'll use Tyma Gaidash's method which is simpler than Klein's. Given,

$$y^5 - 8y^2 + 4y - 16=0$$

or equivalently,

$$\left(\tfrac{y}{2}\right)^5-\left(\tfrac{y}{2}-\tfrac{1+\sqrt{-31}}8\right)\left(\tfrac{y}{2}-\tfrac{1-\sqrt{-31}}8\right)=0$$

then the five roots are,

$$y_k=-2\sum_{n=1}^\infty e^\frac{(2k+1)\pi i n}5 \frac{(q)^\frac n5}{(p)^{\frac{4n}5-1}} \frac{\left(\frac n5\right)!}{n!\,\Gamma\big(2-\frac{4n}5\big)}\;{_2 F_1}\left(1-n,-\frac n5;2-\frac{4n}5;\frac pq\right)\qquad$$

for $k = 0,1,2,3,4,$ and where $p, q = \frac{1\pm\sqrt{-31}}8,$ with gamma function $\Gamma(n).$ Reversing the Tshirnhausen transformation, then the real root is,

$$x_2 = -1+\sqrt{y_2} \; \approx \; 0.45913372331$$

though one has to choose the correct sign of the square root.


III. Second Method

We can also use the Klein-Brioschi method. The real root is,

$$\quad x_2 = -1+\sqrt{\frac{pm+q}{rm^2-3}}\approx 0.45913372331$$

where,

$$m=-\frac{u^2-5}{w^3}\sqrt{\frac{u^2+2u+5}{-r\;}}$$ $$u =\frac{\eta(\tau/5)}{\eta(5\tau)},\quad w =\frac{\eta(\tau)}{\eta(5\tau)}$$ $$\tau = 2+\frac{_2F_1\big(\frac16,\frac56,1,1-\alpha\big)}{_2F_1\big(\frac16,\frac56,1,\alpha\big)}\sqrt{-1}$$ $$\alpha = \tfrac12\left(1-\sqrt{\frac{r}{r-1728}}\right)$$ and $(p,q,r)$ are the constants, \begin{align} \qquad p &= \frac{128\;}{259^3} (34323438 + 645563 \sqrt{470})\\[4pt] q &= \frac{4}{259} (431 + 32 \sqrt{470})\\[4pt] r &= \frac{320 (583125 + 30692 \sqrt{470})^3}{259^5 (2263 + 156 \sqrt{470})}+1728\\ \end{align} where the formulas ($p = e,\, q = \lambda,\, r= 1/f$) in this Klein-Brioschi post.

P.S. In short, just choose which of the two methods you prefer.

0

I take it as a perfect example for a method using continued fraction see for that Got a factored version of the Taylor's series?

1.You need to transform the coefficients

2.Apply the iteration

3.You have a rational approximation of the root (always useful)

On the other hand see the good answer due to TitoPiezasIII to compare .