1

This is a self-answer question of the following problem:

Solve the equation:

$$4x^4 - 36x^3 + 61x^2 + 90x + 25 = 0.$$

See my answer.

user2661923
  • 42,303
  • 3
  • 21
  • 46

3 Answers3

3

The rational root theorem gives $$ 4x^4 - 36x^3 + 61x^2 + 90x + 25=(2x + 1)^2(x - 5)^2. $$ Of course, there is some work to do for it, but it seems to me the first thing to try. So the polynomial factors into linear factors.

One could also compare coefficients with $(2x+a)(2x+b)(x+c)(x+d)$ and would find this factorization, or first consider $(4x^2+ax+b)(x^2+cx+d)$.

Dietrich Burde
  • 140,055
  • Why would this be the first thing to try, if you happen to notice the quartic's quasi-symmetry? Why give preference to an approach that may fail, over an approach that is guaranteed to work? – user2661923 Mar 13 '24 at 19:27
  • Because the "guaranteed way" is much too long, and to find a linear factor often works, and if not, we immediately know that it is a product of two quadratics. – Dietrich Burde Mar 13 '24 at 19:28
  • For what it's worth, it was an unusual accident that polynomial long division was needed. Usually, in a quasi-symmetric quartic equation, there will be two distinct roots for $~u,~$ and therefore $~4~$ directly generated roots for $~x.$ Also, often, if the roots are rational, the corresponding equations in $~u~$ and/or $~x~$ may be reasonably easy to factor. – user2661923 Mar 13 '24 at 19:36
  • Also, for what it's worth, my experiences attempting to derive the quadratic factors of $$a_4x^4 + a_3x^3 + a_2x^2 + a_1x + a_0 = (ax^2 + bx + c)(dx^2 + ex + f)$$ have been troublesome, specifically because the resulting equations in $~a,b,c,d,e,f~$ are not linear. – user2661923 Mar 13 '24 at 19:51
  • 1
    But you can take them over $\Bbb Z$ by the Gauß Lemma, and then a non-linear equation $ad=1$ has only the solutions $1\cdot 1=1$ and $(-1)(-1)=1$. This way the system becomes very easy to solve over the integers, and not troublesome at all. You can try this for your polynomial, and you will find a solution $(4x^2+4x+1)(x^2-10x+25)$ easily. – Dietrich Burde Mar 13 '24 at 19:55
  • Interesting rebuttal. I am assuming that you are referring to Gauss' Lemma, which I just read about, for the first time. If I understand correctly, you are citing the result that a non-constant polynomial in $~\Bbb{Z}[X]~$ is irreducible in $~\Bbb{Z}[X]~$ if and only if it is both irreducible in $~\Bbb{Q}[X]~$ and primitive in $~\Bbb{Z}[X].~$ Since the given quartic is primitive, it may or may not be reducible in $~\Bbb{Z}[X].~$ How can you easily determine whether it is reducible in $~\Bbb{Z}[X]?~$ – user2661923 Mar 13 '24 at 20:20
  • By solving the non-linear equations $ad=4$, $cd+2ef+\cdots =0$ over the integers. This is not troublesome. – Dietrich Burde Mar 13 '24 at 20:21
2

Solve the equation:

$$4x^4 - 36x^3 + 61x^2 + 90x + 25 = 0.$$

This problem was lifted from this Youtube Video. I didn't particularly like the video's approach, so I used my own approach, which I have never seen discussed anywhere. To understand the approach, first see this answer.

Assuming that $~u = x - \dfrac{1}{x},~$ the foundation of the linked answer is that in a symmetric quartic equation, you can (for example) express $~\displaystyle x^2 + \frac{1}{x^2}~$ as $~u^2 + 2.~$

This implies that a symmetric quartic equation like

$$a_4x^4 + a_3x^3 + a_2x^2 - a_3x + a_4 = 0$$

can be converted into a quadratic equation in $~u.$

The point of this self-answer posting is that this method may be extended beyond symmetric quartic equations to quasi-symmetric quartic equations (a description that I made up).

Suppose (for example) you have the quartic equation

$$a_4 + a_3x^3 + a_2x^2 - a_1x + a_0 = 0$$

where

$$\left| ~\frac{a_1}{a_3} ~\right|^2 = \frac{a_0}{a_4}.$$

I refer to such a quartic equation as quasi-symmetric, because you can set $~\displaystyle k = \left| ~\frac{a_1}{a_3} ~\right|,~$ and then attack the quartic equation with the substitution

$$~u = x - \frac{k}{x} \implies u^2 + 2k = x^2 + \frac{k^2}{x^2}.$$

In the posted problem, you have that

$$\left| ~\frac{a_1}{a_3} ~\right|^2 = \left| ~\frac{-5}{2} ~\right|^2 = \frac{25}{4} = \frac{a_0}{a_4}.$$

The remainder of this posting works through the $~\displaystyle u = x - \frac{5}{2x}~$ solution.


First, construct the equation

$$4x^2 - 36x + 61 + \frac{90}{x} + \frac{25}{x^2} = 0.$$

Then, since $~\displaystyle u^2 + 5 = x^2 + \frac{25}{4x^2},$

you have that

$$4(u^2 + 5) - 36u + 61 = 0 \implies 4u^2 - 36u + 81 = 0 \implies $$

$$(2u - 9)^2 = 0 \implies x - \frac{5}{2x} = u = \frac{9}{2} \implies $$

$$2x^2 - 9x - 5 = 0. \tag1 $$

So, (1) above yields two of the roots, and the other two roots can be derived by polynomial long division:

                2x^2 -  9x   -  5
               --------------------------------
2x^2 - 9x - 5 | 4x^4 - 36x^3 + 61x^2 + 90x + 25
                4x^4 - 18x^3 - 10x^2
               ---------------------
                     - 18x^3 + 71x^2 + 90x
                     - 18x^3 + 81x^2 + 45x
                     ---------------------
                             - 10x^2 + 45x + 25
                             - 10x^2 + 45x + 25
                             ------------------

Note
I am deliberately ducking a question that I am not really qualified to answer: was the polynomial long division avoidable? That is, since the quadratic equation in $~u~$ gives a repeated root, does this imply that each of the roots of the original quartic equation is repeated?

user2661923
  • 42,303
  • 3
  • 21
  • 46
1

I am aware of a technique of reducing quartic polynomials using substitution $y=x+\frac{a_2}{4a_1}$. Applying this substitution leads to $4y^4-\frac{121}{2}y^2+\frac{121^2}{64}$ which despite somewhat scary coefficients is easy to solve directly or factor. Reference: Factoring Quartic Polynomials

Vasili
  • 11,629
  • If I understand the linked article, your approach is limited to quartics in $~\Bbb{Q}[x]~$ that factor into quadratics in $~\Bbb{Q}[x].$ Is it possible that a quartic in $~\Bbb{Q}[x]~$ does not yield such a factorization? My approach works with on any quartic that is quasi symmetric, without any consideration of whether any coefficients are rational. – user2661923 Mar 13 '24 at 20:03
  • @user2661923: Yes, your are correct that this technique requires existence of rational roots so I appreciate you sharing your approach which I never encountered before! – Vasili Mar 13 '24 at 20:16