3

Solving the cubic $x^3-x^2-2x+1 = 0$.

Using the Cubic Formula I get the following three solutions.

$x_1 = \frac{1}{3} - \frac{1}{3}\left(\frac{7}{2} + \frac{21}{2}i\sqrt{3} \right)^{1/3} - \frac{7}{3}\frac{1}{\left( \frac{7}{2} + \frac{21}{2}i\sqrt{3} \right)^{1/3}} \cong -1.2469796037174670610+2.10^{-20}i$

$x_2 = \frac{1}{3} + \frac{1}{6}\left( \frac{7}{2}+\frac{21}{2}i\sqrt{3} \right)^{1/3}(1+i\sqrt{3}) + \frac{7}{6}\frac{1-i\sqrt{3}}{\left( \frac{7}{2} + \frac{21}{2}i\sqrt{3} \right)^{1/3}} \cong .44504186791262880859 - 3.10^{-20}i$

$x_3 = \frac{1}{3} + \frac{1}{6}\left( \frac{7}{2}+\frac{21}{2}i\sqrt{3} \right)^{1/3}(1-i\sqrt{3}) + \frac{7}{6}\frac{1+i\sqrt{3}}{\left( \frac{7}{2} + \frac{21}{2}i\sqrt{3} \right)^{1/3}} \cong 1.8019377358048382525 + 3.10^{-20}i$

It is clear that the three solutions are all real solutions, but is there a way I can remove the complex components algebraically? My ultimate goal is to describe what the Galois group from this polynomial would like.

Zed1
  • 697
  • 2
    You can remove the complex parts if you are willing to settle for solutions involving trig functions. – Simply Beautiful Art Oct 14 '17 at 14:37
  • 4
    The buzzword is casus irreducibilis. – Angina Seng Oct 14 '17 at 14:38
  • 2
    Since the roots are real, the Galois group is not $S_3$ but $A_3$, cyclic, and in particular abelian. So the field is cyclotomic in the sense of being contained in a field $\Bbb Q(\zeta_n)$, field of $n$-th roots of unity. You can determine $n$ from the discriminant of the polynomial. – Lubin Oct 14 '17 at 14:53
  • Yes! That's what I've been trying to work on. Finding polynomials who Galois group is not $S_n$. – Zed1 Oct 14 '17 at 14:58
  • 1
    You need to start with $n$ such that $3|\phi(n)$, that’s the Euler totient function. Like $\phi(9)=6$. The cyclotomic polynomial $\Phi_9(X)=X^6+X^3+1$, so if $\zeta$ is a root, you have $\zeta^3+1+\zeta^{-3}=0$. Now take $\xi=\zeta+\zeta^{-1}$ and discover a cubic satisfied by $\xi$. – Lubin Oct 14 '17 at 15:07
  • But I should make clear that this approach gives you abelian Galois groups only, not polynomials with $A_n$ for $n\ge4$. – Lubin Oct 14 '17 at 15:18
  • @Lubin, is there a relationship between cubics that are casus irreducibilis and possessing Galois groups of order 3? – Zed1 Oct 15 '17 at 16:19
  • Others can answer this better, I’m sure. But it seems to me that if your cubic polynomial is $\Bbb Q$-irreducible and has three real roots, then this is casus irr., and the Galois group is $A_3$, cyclic. Conversely, an odd-degree normal extension must be totally real, so your polynomial must fall into the irreducibilis case. I think the Wikipedia article on casus irreducibilis at least suggests as much. – Lubin Oct 15 '17 at 19:05

3 Answers3

4

I'm sorry, but I'm not entirely sure if there's a way make a transformation to get the appropriate trigonometric roots. However, there is another way to derive its roots using Euler's formula.


Starting with your cubic$$y=x^3-x^2-2x+1$$First, make the substitution $x=t+t^{-1}$. Expanding and collecting in $t$, we find that$$\begin{align*}\left(t+\tfrac 1t\right)^3-\left(t+\tfrac 1t\right)^2-2\left(t+\tfrac 1t\right)+1 & =\frac {t^6-t^5+t^4-t^3+t^2-t+1}{t^3}\\ & =\frac {t^7+1}{t^3(t+1)}\end{align*}$$Therefore, we have that$$t^7+1=0\implies t=e^{\tfrac {\pi i(2k+1)}7}$$Hence, the solutions are given by$$t+t^{-1}=e^{\tfrac {\pi(2k+1)}7i}+e^{-\tfrac {\pi(2k+1)}7i}=2\cos\left(\frac {2k\pi+\pi}7\right)$$for $k=0,1,2$. And it follows immediately that your three solutions are$$\begin{align*} & x_1=2\cos\left(\frac {\pi}7\right)\\ & x_2=2\cos\left(\frac {3\pi}7\right)\\ & x_3=2\cos\left(\frac {5\pi}7\right)\end{align*}$$

Crescendo
  • 4,089
3

Proof of impossibility

I write this section for the following reasons:

  • A proof of impossibility is much more interesting than solving it in terms of $\cos$.
  • You tagged , which is also largely motivated by the concept of proving something is impossible.
  • Casus irreducibilis is mentioned in the comments.

I will attempt to formalize the problem before proving its impossibility:

Is there an intermediate field $F$ between $\Bbb Q$ and $\Bbb R$ so that $f(X) = X^3 - X^2 - 2X + 1$ splits in $F$ and $F$ is radical over $\Bbb Q$?

Explanation

I will explain how the idea of "solvable by radicals" is captured by the idea of "radical extension".

For example, if we want to find the number $\sqrt[3]{11} \sqrt[5]{\dfrac{7+\sqrt3}2} + \sqrt[4]{1+\sqrt[3]4}$ (example given by Ian Stewart in his third edition of Galois Theory), we might want to break this number down and extend our field $\Bbb Q$ adjoining radicals one by one, i.e. come up with:

$$\Bbb Q \subset \Bbb Q\left(\sqrt[3]{11}, \sqrt3, \sqrt[5]{\dfrac{7+\sqrt3}2},\sqrt[3]4,\sqrt[4]{1+\sqrt[3]4}\right) = \Bbb Q(a,b,c,d,e)$$

What makes this extension radical is the fact that each adjoined element, raised to some power, can be found in the previous field:

  • From $\Bbb Q$ to $\Bbb Q(a)$, we find that $a^3 = 11 \in \Bbb Q$.
  • From $\Bbb Q(a)$ to $\Bbb Q(a,b)$, we find that $b^2 = 3 \in \Bbb Q(a)$.
  • From $\Bbb Q(a,b)$ to $\Bbb Q(a,b,c)$, we find that $c^5 = \dfrac{7+b}2 \in \Bbb Q(a,b)$.
  • From $\Bbb Q(a,b,c)$ to $\Bbb Q(a,b,c,d)$, we find that $d^3 = 4 \in \Bbb Q(a,b,c)$.
  • From $\Bbb Q(a,b,c,d)$ to $\Bbb Q(a,b,c,d,e)$, we find that $e^4 = 1+d \in \Bbb Q(a,b,c,d)$.

Note that a radical extension may not be a normal extension, i.e. an extension where every irreducible polynomial over the base field that has at least one zero in the extension splits in the extension. For example, $\Bbb Q(\sqrt[4]2)$ is not a normal extension of $\Bbb Q$, because $X^4-2$ is a polynomial in $\Bbb Q[X]$ that is irreducible, but only two of its roots are in $\Bbb Q(\sqrt[4]2)$ (the other two being $i\sqrt[4]2$ and $-i\sqrt[4]2$).

Proof

A large portion of this proof is inspired by Theodore Shifrin's proof in his book (see section Bibliography).

If such an intermediate field $F$ exists, according to the definition of a radical extension, write it as a tower of extensions:

$$\Bbb Q \subset \Bbb Q(\sqrt[p_1]{b_1}) \subset \Bbb Q(\sqrt[p_1]{b_1}, \sqrt[p_2]{b_2}) \subset \cdots \subset K \subset K(\sqrt[p]a) = F \subset \Bbb R$$

where $b_{i+1} \in \Bbb Q(\sqrt[p_1]{b_1}, \sqrt[p_2]{b_2}, \cdots, \sqrt[p_i]{b_i})$ for each $i$.

WLOG, each $p_i$ is a prime, or else we factorize the composite number and include more intermediate fields to the tower such that each $p_i$ is a prime.

WLOG, let $a_i$ be not a $p_i$th power, or else we can rightfully remove that field from the tower, as it would just be the same as the previous field.

WLOG, let the polynomial $f(X)$ be irreducible in $K$ but has a root in $K(\sqrt[p]a)$ (or else, find such a point in the tower where the previous field has no roots for the polynomial but the next one does, and cut the tower there).

From the other answers, the polynomial has roots in terms of cosine, and one can see that one can write the other roots in terms of one root, using the multiple angle formulas for cosine. Therefore, a field containing one root must contain all of the roots.

Therefore, the polynomial actually splits in $K(\sqrt[p]a)$.

Let the splitting field of the polynomial over $K$ be $L$. We have: $$K \subset L \subset K(\sqrt[p]a)$$

Then, $[L:K] = 3$ as the polynomial has degree $3$. From Lemma 4.7 of Generalizations of the Casus Irreducibilis (link to download), we know that $[K(\sqrt[p]a):K] = p$. Therefore, $p$ is divisible by $3$. But since $p$ is prime, we have $p=3$, whence $L = K(\sqrt[3]a)$.

Since $a$ is not a $3$th power in $K$, $X^3-a$ is irreducible over $K$ (why?). We will prove that $\omega\sqrt[3]a \in L$. (This portion of the proof is highly motivated by the proof that Ian Stewart gave in his Galois Theory.)

Note that $[K(\omega\sqrt[3]a):K] = [K(\sqrt[3]a):K] = 3$, for their minimal polynomial over $K$ is the same ($X^3 - a$). Any isomorphism from $L(\sqrt[3]a)$ to $L(\omega\sqrt[3]a)$ defines an isomorphism from $K(\sqrt[3]a)$ to $K(\omega\sqrt[3]a)$, so $[L(\sqrt[3]a):K(\sqrt[3]a)] = [L(\omega\sqrt[3]a):K(\omega\sqrt[3]a)]$. Therefore: $$\begin{array}{rcll} [L(\omega\sqrt[3]a):L] &=& \dfrac{[L(\omega\sqrt[3]a):K]}{[L:K]} & \text{tower law} \\ &=& \dfrac{[L(\omega\sqrt[3]a):K(\omega\sqrt[3]a)][K(\omega\sqrt[3]a):K]}{[L:K]} & \text{tower law} \\ &=& \dfrac{[L(\sqrt[3]a):K(\sqrt[3]a)][K(\sqrt[3]a):K]}{[L:K]} \\ &=& \dfrac{1 \times 3}{3} \\ &=& 1 \end{array}$$

Therefore, $L(\omega\sqrt[3]a) = L$, so $\omega\sqrt[3]a \in L$, which contradicts $L = K(\sqrt[3]a) \subset \Bbb R$.

$\blacksquare$

Bibliography

The following two links are kept here although it is no longer used above.


Roots expressed in terms of other roots

From the answers above, we know that the roots can be expressed as $a=2\cos\left(\dfrac\pi7\right)$, $b=2\cos\left(\dfrac{3\pi}7\right)$, and $c=2\cos\left(\dfrac{5\pi}7\right)$.

In this section, I will express the roots in terms of the other roots, leaving us with $9$ expressions (among which $3$ are trivial).

Unless otherwise indicated, the formulas are obtained via multiple angle formulas, and the fact that all roots satisfy $x^3 - x^2 - 2x + 1 = 0$.

Express in terms of $a$:

We note that $c=-2\cos\left(\dfrac{2\pi}7\right)$.

  • $a=a$
  • $b=2\left[4\left(\dfrac a2\right)^3-3\left(\dfrac a2\right)\right]=a^2-a-1$
  • $c=-2\left[2\left(\dfrac a2\right)^2-1\right]=-a^2+2$

Express in terms of $b$:

We note that $a=-2\cos\left(\dfrac{6\pi}7\right)$ and $c=2\cos\left(\dfrac{9\pi}7\right)$.

  • $a=-2\left[2\left(\dfrac b2\right)^2-1\right]=-b^2+2$
  • $b=b$
  • $c=2\left[4\left(\dfrac b2\right)^3-3\left(\dfrac b2\right)\right]=b^2-b-1$

Express in terms of $c$:

We note that $a=-2\cos\left(\dfrac{6\pi}7\right)$, $b=-2\cos\left(\dfrac{4\pi}7\right)$, and $c=-2\cos\left(\dfrac{2\pi}7\right)$.

  • $a=-2\left[4\left(-\dfrac c2\right)^3-3\left(-\dfrac c2\right)\right]=c^2-c-1$
  • $b=-2\left[2\left(-\dfrac c2\right)^2-1\right]=-c^2+2$
  • $c=c$

Solution (allowing complex numbers)

For the sake of completeness, a solution including complex numbers is included.

Let $p(x) = x^3 - x^2 - 2x + 1$.

The discriminant of $p$ is $49 = 7^2$, so the Galois group of the splitting field of $p$ over $\Bbb Q$ is $\Bbb A_3$.

Let the roots of $p$ be $a$, $b$, and $c$.


From Vieta: $$\begin{cases} a+b+c &=& 1 \\ ab+bc+ac &=& -2 \\ abc &=& -1 \end{cases}$$

We find $a^2b+b^2c+c^2a+ab^2+bc^2+ca^2$: $$\begin{array}{rcl} (a+b+c) (ab+bc+ac) &=& -2 \\ a^2b+b^2c+c^2a+ab^2+bc^2+ca^2+3abc &=& -2 \\ a^2b+b^2c+c^2a+ab^2+bc^2+ca^2 &=& 1 \\ \end{array}$$

We find $a^3+b^3+c^3$: $$\begin{array}{rcl} (a+b+c)^3 &=& 1 \\ a^3+b^3+c^3 + 3(a^2b+b^2c+c^2a+ab^2+bc^2+ca^2) + 6abc &=& 1 \\ a^3+b^3+c^3 + 3 - 6 &=& 1 \\ a^3+b^3+c^3 &=& 4 \\ \end{array}$$

We find $a^3b^3+b^3c^3+c^3a^3$: $$\begin{array}{rcl} (ab+bc+ac)^3 &=& a^3b^3+b^3c^3+c^3a^3 + 3(ab+bc+ac)(a^2bc+ab^2c+abc^2) - 3(abc)^2 \\ -8 &=& a^3b^3+b^3c^3+c^3a^3 + 3(ab+bc+ac)(abc)(a+b+c) - 3 \\ -8 &=& a^3b^3+b^3c^3+c^3a^3 + 3(-2)(-1)(1) - 3 \\ a^3b^3+b^3c^3+c^3a^3 &=& -11 \\ \end{array}$$

We find $(a^2b+b^2c+c^2a)(ab^2+bc^2+ca^2)$: $$\begin{array}{cl} & (a^2b+b^2c+c^2a)(ab^2+bc^2+ca^2) \\ =& a^3b^3+b^3c^3+c^3a^3+a^4bc+b^4ca+c^4ab+3a^2b^2c^2 \\ =& a^3b^3+b^3c^3+c^3a^3+abc(a^3+b^3+c^3)+3(abc)^2 \\ =& -11+(-1)(4)+3(-1)^2 \\ =& -12 \end{array}$$


Let $h = a^2b+b^2c+c^2a$ and $k = a^2b+bc^2+ca^2$. From the above, $h+k=1$ and $hk=-12$.

Let $p = a+b\omega+c\omega^2$. The Galois group permutes $a$, $b$, and $c$, changing the expression to $p\omega = c+a\omega+b\omega^2$ and $p\omega^2 = b+c\omega+a\omega^2$.

Therefore, $p^3$ is fixed:

$$\begin{array}{rcl} p^3 &=& (a+b\omega+c\omega^2)^3 \\ &=& a^3 + b^3 + c^3 + 3(a^2b\omega + ab^2\omega^2 + a^2c\omega^2 + ac^2\omega + b^2c\omega + bc^2\omega^2) + 6abc \\ &=& a^3 + b^3 + c^3 + 3\omega h + 3\omega^2 k + 6abc \\ &=& 4 + 3\omega h + 3\omega^2 k - 6 \\ &=& -2 + 3\omega h + 3\omega^2 k \\ \end{array}$$

Let $q = a+b\omega^2+c\omega$. By the same argument, $q^3$ is fixed:

$$\begin{array}{rcl} q^3 &=& (a+b\omega^2+c\omega)^3 \\ &=& a^3 + b^3 + c^3 + 3(a^2b\omega^2 + ab^2\omega + a^2c\omega + ac^2\omega^2 + b^2c\omega^2 + bc^2\omega) + 6abc \\ &=& a^3 + b^3 + c^3 + 3\omega^2 h + 3\omega k + 6abc \\ &=& 4 + 3\omega^2 h + 3\omega k - 6 \\ &=& -2 + 3\omega^2 h + 3\omega k \\ \end{array}$$

Then: $$\begin{array}{rcl} p^3 + q^3 &=& -4 + 3(\omega + \omega^2) (h+k) \\ &=& -4 + 3(-1)(1) \\ &=& -7 \\ \end{array}$$

Also: $$\begin{array}{rcl} p^3 q^3 &=& 4 + 9h^2 + 9k^2 - 6(\omega+\omega^2)h - 6(\omega+\omega^2)k + 9(\omega^2+\omega)hk \\ &=& 4 + 9[(h+k)^2-2hk] - 6(-1)(h+k) + 9(-1)hk \\ &=& 4 + 9[1^2-2(-12)] - 6(-1)(1) + 9(-1)(-12) \\ &=& 4 + 225 + 6 + 108 \\ &=& 343 \end{array}$$

Therefore, $p^3-q^3 = \sqrt{(p^3+q^3)^2-4p^3q^3} = \sqrt{-1323} = 21i\sqrt{3}$.

We obtain $p^3 = \dfrac12[(p^3+q^3)+(p^3-q^3)] = \dfrac{-7+21i\sqrt{3}}2$ and $q^3 = \dfrac12[(p^3+q^3)-(p^3-q^3)] = \dfrac{-7-21i\sqrt{3}}2$.

Therefore:

  • $(1): \quad a+b+c=1$ (by Vieta)
  • $(2): \quad a+b\omega+c\omega^2=\sqrt[3]{\dfrac{-7+21i\sqrt{3}}2}$
  • $(3): \quad a+b\omega^2+c\omega=\sqrt[3]{\dfrac{-7-21i\sqrt{3}}2}$

From $\dfrac13[(1)+(2)+(3)]$: $$a = \frac13 + \frac13\sqrt[3]{\dfrac{-7+21i\sqrt{3}}2} + \frac13\sqrt[3]{\dfrac{-7-21i\sqrt{3}}2}$$

From $\dfrac13[(1)+(2)\omega^2+(3)\omega]$: $$b = \frac13 + \frac13\frac{-1-i\sqrt3}2\sqrt[3]{\dfrac{-7+21i\sqrt{3}}2} + \frac13\frac{-1+i\sqrt3}2\sqrt[3]{\dfrac{-7-21i\sqrt{3}}2}$$

From $\dfrac13[(1)+(2)\omega+(3)\omega^2]$: $$c = \frac13 + \frac13\frac{-1+i\sqrt3}2\sqrt[3]{\dfrac{-7+21i\sqrt{3}}2} + \frac13\frac{-1-i\sqrt3}2\sqrt[3]{\dfrac{-7-21i\sqrt{3}}2}$$

P.S.: $\sqrt[3]{re^{i\theta}} = r^{1/3} e^{i\theta/3}$, where $\theta \in (-\pi,\pi]$.

Kenny Lau
  • 25,049
  • Thank you very much! I greatly appreciate the additional resources you provided. It will be beneficial to my further study. – Zed1 Oct 15 '17 at 21:50
  • @Zed1 Thanks for your appreciation. Please ask me for clarity in the comments / in the chat should there be anything unclear. – Kenny Lau Oct 15 '17 at 21:51
  • @Zed1 I've radically changed the proof and included a new section, as well as two new links in the "Bibliography" section. – Kenny Lau Oct 16 '17 at 00:48
2

Use $$\cos\frac{2\pi}{7}+\cos\frac{4\pi}{7}+\cos\frac{6\pi}{7}=-\frac{1}{2}.$$ Finally we obtain $$x_1=-2\cos\frac{2\pi}{7},$$ $$x_2=-2\cos\frac{4\pi}{7}$$ and $$x_3=-2\cos\frac{6\pi}{7}.$$

Because $$\cos\frac{2\pi}{7}+\cos\frac{4\pi}{7}+\cos\frac{6\pi}{7}=\frac{2\sin\frac{\pi}{7}\cos\frac{2\pi}{7}+2\sin\frac{\pi}{7}\cos\frac{4\pi}{7}+2\sin\frac{\pi}{7}\cos\frac{6\pi}{7}}{2\sin\frac{\pi}{7}}=$$ $$=\frac{\sin\frac{3\pi}{7}-\sin\frac{\pi}{7}+\sin\frac{5\pi}{7}-\sin\frac{3\pi}{7}+\sin\frac{7\pi}{7}-\sin\frac{5\pi}{7}}{2\sin\frac{\pi}{7}}=-\frac{1}{2},$$ which says $$x_1+x_2+x_3=1.$$

Now, $$x_1x_2+x_1x_3+x_2x_3=4\left(\cos\frac{2\pi}{7}\cos\frac{4\pi}{7}+\cos\frac{2\pi}{7}\cos\frac{6\pi}{7}+cos\frac{4\pi}{7}\cos\frac{6\pi}{7}\right)=$$ $$=2\left(\cos\frac{6\pi}{7}+\cos\frac{2\pi}{7}+\cos\frac{6\pi}{7}+\cos\frac{4\pi}{7}+cos\frac{4\pi}{7}+\cos\frac{2\pi}{7}\right)=-2$$ and $$x_1x_2x_3=-8\cos\frac{2\pi}{7}\cos\frac{4\pi}{7}\cos\frac{6\pi}{7}=-4\left(\cos\frac{6\pi}{7}+\cos\frac{2\pi}{7}\right)\cos\frac{6\pi}{7}=$$ $$=-2\left(1+\cos\frac{2\pi}{7}+\cos\frac{6\pi}{7}+\cos\frac{4\pi}{7}\right)=-2\left(1-\frac{1}{2}\right)=-1$$ and use the Viete's theorem.

  • It's a little hard for me to see how to get from OP's question to here? Can you provide any more detail, please? +1 for elegance, but I'm still confused. – G Tony Jacobs Oct 14 '17 at 16:15
  • @G Tony Jacobs I added something. See now. – Michael Rozenberg Oct 14 '17 at 16:37
  • I see. You're not algebraically transforming the OP's solutions into real numbers directly. You're determining real roots via a different method, using Viete's theorem. That's what wasn't clear, in the context of the question as it was asked. Thank you. I still don't see how I would go about applying this technique to a similar problem, but I see why this works. – G Tony Jacobs Oct 14 '17 at 16:42
  • What made it clear that those cosines were the right thing to look at? What intuition led to this? – G Tony Jacobs Oct 14 '17 at 16:42
  • @G Tony Jacobs I thought that it's so and I check it and it turned out. Just I solved one or maybe two problems in math before. :) – Michael Rozenberg Oct 14 '17 at 16:50
  • Yeah, I understand that you have experience. Sometimes it's possible to unpack which elements of that experience led to a particular thought. Sometimes, that's too difficult to do. When it's possible, it's extremely helpful to the student. – G Tony Jacobs Oct 14 '17 at 16:51
  • 1
    @G Tony Jacobs I like cubic equations. For example $\cos\frac{2\pi}{19}+\cos\frac{14\pi}{19}+\cos\frac{16\pi}{19}$ is a root of the equation $8x^3+4x^2-4x-1=0$. Try to prove it and find other root. Good luck! – Michael Rozenberg Oct 14 '17 at 16:59
  • Thank you. I enjoy cubic equations, too; I'll try this. – G Tony Jacobs Oct 14 '17 at 17:05
  • 2
    @GTonyJacobs see page 6 at https://books.google.com/books?id=wt7lgfeYqMQC&pg=PR1&lpg=PR1&dq=reuschle++tafeln+complexer+primzahlen&source=bl&ots=VGZFPrfUBn&sig=MlQ667PqXaQ9rAvLWkG3_F1rwsk&hl=en&sa=X&ved=0ahUKEwiIwtSvm9TQAhUJ-2MKHXJIA_kQ6AEIODAE#v=onepage&q=reuschle%20%20tafeln%20complexer%20primzahlen&f=false – Will Jagy Oct 14 '17 at 18:25
  • @Will Jagy Wow! Thank you very much! I made the typo. My equation is $8x^3+4x^2-12x-7=0$. – Michael Rozenberg Oct 14 '17 at 18:30
  • @GTonyJacobs my list of Gauss cubics http://math.stackexchange.com/questions/2022216/on-the-trigonometric-roots-of-a-cubic/2022887#2022887 then Gauss quintics http://math.stackexchange.com/questions/1996552/any-more-cyclic-quintics then Gauss degree 7 https://math.stackexchange.com/questions/1996552/any-more-cyclic-quintics/2033565#2033565 – Will Jagy Oct 14 '17 at 18:33
  • Michael, see page 26 in Reuschle. His roots, when all real, are sums of double cosine terms. So, beginning with his $x^3 + x^2 - 6 x - 7$ replace $x$ by $2t,$ for $8 t^3 + 4 t^2 - 12 t - 7.$ I bought an inexpensive bound reprint of Reuschle, two oversize paperback books. – Will Jagy Oct 14 '17 at 18:55
  • @Will Jagy Yes I saw. In this world everything has been there once. I found this equation when I thought about the Ramanujan's equation $\sqrt{8-\sqrt{8+\sqrt{8-x}}}=x$ and I found the mine: $\sqrt{4+\sqrt{4+\sqrt{4-x}}}=x$. – Michael Rozenberg Oct 14 '17 at 19:00
  • Michael, see Tito's http://math.stackexchange.com/questions/1065862/something-strange-about-sqrt-4-sqrt-4-sqrt-4-x-x-and-its-friends – Will Jagy Oct 14 '17 at 19:20
  • Will, see also my post here: https://artofproblemsolving.com/community/c6h68986p419836 – Michael Rozenberg Oct 14 '17 at 19:29