I'm trying to solve this equation $$x^5+3x^3+x+\frac{1}{x^2}=0.$$ It's possible to notice that if we write $3x^3$ as $2x^3+x^3$, then $$ x^3\left(x^2+2+\frac{1}{x^2}\right)+x^3+\frac{1}{x^2}=0, $$ which equals $$ x^3\left(x+\frac{1}{x}\right)^2+x^3+\frac{1}{x^2}=0. $$ But I have no idea how to continue further.
-
3If WolframAlpha does not find a closed-form solution then there probably is none, and you have to solve it numerically. – Martin R May 19 '23 at 06:37
-
2@MartinR You made it to 100k. Well done! – Bumblebee May 19 '23 at 06:47
-
multiply through by $x^2.$ Now you have a polynomial of degree 7. This polynomial has no rational roots. You should not expect a polynomial of degree 5 or greater to be solvable. – user317176 May 19 '23 at 06:47
-
@MartinR Without determine Galois group, still I think we can not reach to the healthy conclusion . – lone student May 19 '23 at 06:52
-
2The polynomial of seventh order has 1 real root and 6 complex roots. $$1+x^{3}+3x^{5}+x^{7}\simeq\left(x+ 0.711579764439192\right) \left(x+ 0.274401076934264+ 0.903369883711077 ,\mathrm{i}\right) \left(x+ 0.274401076934264- 0.903369883711077 ,\mathrm{i}\right) \left(x- 0.0319403820294392+ 1.62172369651801 ,\mathrm{i}\right) \left(x- 0.0319403820294392- 1.62172369651801 ,\mathrm{i}\right) \left(x- 0.598250577124421+ 0.491251338921359 ,\mathrm{i}\right) \left(x- 0.598250577124421- 0.491251338921359 ,\mathrm{i}\right)=0$$ – gpmath May 19 '23 at 06:52
-
@gpmath, thx, btw, have you got this by wolfram or others solvers? – LeonBrick May 19 '23 at 07:05
-
@MartinR R, yes, it seems like lack of rational roots but how to prove it neatly? – LeonBrick May 19 '23 at 07:05
-
2@lonestudent Unfortunately, the Galois group has size $7! = 5040$ so the roots have no algebraic relations with each other , apart from pairs of complex roots being conjugates of each other. That, along with the absence of rational roots, tells us that the present roots are best computed numerically. – Sarvesh Ravichandran Iyer May 19 '23 at 07:08
-
Maple with command: factor(x^7 + 3*x^5 + x^3 + 1, complex); – gpmath May 19 '23 at 07:09
-
1Glad to see you again @SarveshRavichandranIyer ! Sometimes I've seen types of higher degree polynomials that can be solved by radicals but that WolframAlpha failed to solve. Do you know any input for Wolfram that can directly determine the Galois group? Or is there an input that directly determines whether the algebraic solution exists or not? If there is, it would be really great to know. – lone student May 19 '23 at 07:15
-
@Sarvesh Ravichandran Iyer, cool! May, you, please, describe briefly, how have you count this Galois group? – LeonBrick May 19 '23 at 07:16
-
1@lonestudent I use Magma Calculator and this page to find the Galois group. To find algebraic relations is more trickier : there's no "route" but I try to type variations of the equation into Approach0 and see if others have done anything other than calculus. If not, then I give up (after, of course, trying the rational root theorem, maybe seeing if some obvious substitution might work, etc.). – Sarvesh Ravichandran Iyer May 19 '23 at 07:18
-
1@LeonBrick Kind of difficult in one comment, but there's a branch of mathematics called Galois theory. The "Galois group" of a polynomial is roughly a "measure" of how much the roots "algebraically" depend upon each other. For example, is one of the roots a square of another root? Maybe it's the reciprocal. Maybe it's one more than the other root. And so on. If the Galois group's size is small compared to the factorial of the degree of the polynomial, then typically such polynomials should be easy to factorize. In this case, that's not true. I'll use another comment to elaborate. – Sarvesh Ravichandran Iyer May 19 '23 at 07:20
-
1For example, take the polynomial $x^6+x^4+x^2+27$. In this case, the polynomial might be difficult to factorize, but if $\alpha$ is a root, so is $-\alpha$. So you expect that the substitution $x^2 = y$ will work. It turns out the Galois group size is $48$ which is smaller than $6! = 720$, so the polynomial will become of degree $3$ (after which, it turns out, the roots will all stop depending algebraically on each other). I mentioned in a comment above how the Galois group can be used to detect if expressions factorize or not, and if roots could have algebraic relations with one another. – Sarvesh Ravichandran Iyer May 19 '23 at 07:24
-
@SarveshRavichandranIyer In general I always appreciate that you leave comments that are understandable to non-experts too . – lone student May 19 '23 at 07:26
-
2@lonestudent Thank you, running on your support and belief that I can deliver. – Sarvesh Ravichandran Iyer May 19 '23 at 07:28
-
@LeonBrick The lack of rational roots is easy to prove with the rational root theorem. The lack of radical roots is much more difficult to prove. – Servaes May 19 '23 at 14:12
2 Answers
The rational equation:
$$ x^5 + 3\,x^3 + x + \frac{1}{x^2} = 0 $$
shares the same solutions of the polynomial equation:
$$ \underbrace{x^7 + 3\,x^5 + x^3 + 1}_{p(x)} = 0\,. $$
So, once the seven first-attempt solutions on the unit circle have been calculated:
$$ x_k = \cos(2\,k\,\pi/7) + \text{i}\,\sin(2\,k\,\pi/7) \quad \quad \text{with} \; k = 1,2,\dots,7 $$
it's sufficient to update them about ten times by applying the Aberth-Ehrlich method:
$$ x_k' = x_k - \left(\frac{p'(x_k)}{p(x_k)} - \begin{aligned}\sum_{j=1;\,j\ne k}^7\end{aligned}\frac{1}{x_k-x_j}\right)^{-1} $$
i.e
p[u_] := u^7 + 3 u^5 + u^3 + 1
d = Exponent[# /. u -> v RandomReal[], v] &[p[u]];
x = N@Table[Cos[2 k π/d] + I Sin[2 k π/d], {k, d}];
Do[x[[k]] = x[[k]] - (p'[x[[k]]]/p[x[[k]]] -
Sum[If[j == k, 0, 1/(x[[k]] - x[[j]])], {j, d}])^(-1), {n, 8}, {k, d}]
from which:
$$ \begin{aligned} & x_1 = 0.598251 + 0.491251\,\text{i}\,; \\ & x_2 = -0.274401 + 0.90337\,\text{i}\,; \\ & x_3 = -0.71158 - 7.35225\cdot 10^{-19}\,\text{i}\,; \\ & x_4 = 0.0319404 - 1.62172\,\text{i}\,; \\ & x_5 = -0.274401 - 0.90337\,\text{i}\,; \\ & x_6 = 0.598251 - 0.491251\,\text{i}\,; \\ & x_7 = 0.0319404 + 1.62172\,\text{i}\,. \\ \end{aligned} $$
Of course this doesn't prove whether or not it's possible to express them by radicals.
Totally useless but funny (at least to me)
For the only real root, write $$f(x)=x^7+3 x^5+x^3+1$$ as $$f(x)=-4+25 (x+1)-54 (x+1)^2+66 (x+1)^3-50 (x+1)^4+$$ $$24 (x+1)^5-7(x+1)^6+(x+1)^7$$
Now, consider the above as a series expansion to $O\left((x+1)^{p+1}\right)$ with $p\geq 7$ and use power series reversion to get $$x=-1+\sum_{n=0}^p a_n\,(f(x)+4)^n+O((f(x)+4)^{p+1}$$ Since we want $f(x)=0$ then $$x_{(p)}=-1+\sum_{n=0}^p a_n\,4^n$$
Using $p=8$ which is exactly the polynomial, you should get, as an estimate $$x_{(8)}=-\frac{1075280545512027477}{1490116119384765625}=-0.721609$$
Now, repeat and you will get (exact fraction reprted as decimal numbers) $$\left( \begin{array}{cc} p & x_{(p)} \\ 7 & -0.721609 \\ 8 & -0.718758 \\ 9 & -0.716772 \\ 20 & -0.711791 \\ 30 & -0.711595 \\ 40 & -0.711581 \\ 50 & -0.711580 \\ \end{array} \right)$$
Edit (for more fun !)
Transform the series to their equivalent $[1,n]$ Padé approximants $P_n$, that is to say $$P_n=\frac {-4+ a_{(n)}\,(x+1) } {1+\sum_{p=1}^n b_p\,(x+1)^p }\quad \implies x_{(n)}=-1+\frac 4{a_{(n)}}$$ which is the same as using the first itterate of a Newton-like method of order $(n+2)$.
This will generate the converging sequence $$\left( \begin{array}{c} -\frac{21}{25} , -\frac{309}{409}, -\frac{4245}{5881}, -\frac{58357}{81881} , -\frac{807253}{1134777} , -\frac{3730663}{5243699}, -\frac{155226133}{218150521}, -\frac{2152852853}{3025454937} \end{array} \right)$$
The last one in the list is $$x_{(7)}= -\frac{2152852853}{3025454937}={\large{\color{red}{-0.711579}}}88$$
- 289,558
-
1@TeM. If you knew who was my master, my guru, who is still my God (capital on purpose), you would not be very surprized. Be sure that I appreciate. Cheers :-) – Claude Leibovici May 19 '23 at 11:06
-
@TeM. Just for the fun, assuming that you have 30 seconds to waste, have a look at my edit. – Claude Leibovici May 19 '23 at 13:38
-
Just wondering is \frac{2152852853}{3025454937}=0.71157988 irrational number since the Rational Root Theorem does not consider it as a rational solution? Thanks. – NoChance May 20 '23 at 01:31
-
1@NoChance. This is not the solution (which is irrational. For example $$-\frac{71766252868840089984679014456794928069798468947749}{100854 825355243616352598809693657706050583163486975}$$ igives $100$ exact figures but it is not the solution – Claude Leibovici May 20 '23 at 01:37
-
Thanks you for the immediate reply. This is big! I thought it would be the solution to the degree 7 polynomial (at least from the chart) - Why is it not a solution? – NoChance May 20 '23 at 01:40
-
1@NoChance. Because it can not be rational. Very often in answers, speaking about the solution, I use to write $\color{red}{\text{the "exact" solution}}$. The double quotes make a big difference – Claude Leibovici May 20 '23 at 01:47
-