11
$\begingroup$

Find all quadruplet(s) of non-zero real numbers $ (a,b,c,d) $ such that $ a,b,c$ and $ d$ are roots (of $x$) to the equation $ x^4 + ax^3 + bx^2 + cx + d = 0 $.

My friend found a set of irrational roots that satisfy this condition through experimentation.

And no, I don't know how to solve this question properly.

This question was inspired by another math question in another math platform.

My useless attempt: By Vieta's formula, $ 2a+b+c+d = 0, ab+ac+ad+bc+bd+cd = b, abc + abd + acd + bcd = -c , abc =1 $. That's all....

EDIT: Fixed typo/redundancy.

EDIT2: Changed the word "rational" to "irrational", because I'm a klutz.

$\endgroup$
  • $\begingroup$ What happens if you eliminate $c,d$ by using $abc=1,2a+b+c+d=0$? $\endgroup$ – πr8 May 1 '16 at 13:44
  • $\begingroup$ @πr8 If I'm not mistaken, we arrive at $2b^3a^4 + (b^4 + 2*b)a^3 + 3b^2a^2 + (b^3 - b + 1)a + b=0$ as condition for $a,b$. Playing around it seems that e.g. $b=1$ allows rational solutions for $a$ namely $a=1|-1|\frac12$, then $c=-1|1|2$. Only the last case leads to non-zero $d$. I have not guessed other solutions yet $\endgroup$ – Hagen von Eitzen May 1 '16 at 14:06
  • $\begingroup$ Maple only finds the trivial solution $a = b = c = d = 0$, $a = 1, b = -2, c = 0, d = 0$, $a = 1, b = -1, c = -1, d = 0$ and the irrational solution $a = \alpha, b = -2 \alpha^2 - 2 \alpha, c = \alpha^2, d = 0$, where $2 \alpha^3 - 2 \alpha^2 - 1 = 0$. (It also finds two complex solutions whose components are algebraic of degree $14$ over $\Bbb Q$.) $\endgroup$ – Travis May 1 '16 at 14:38
  • $\begingroup$ Given that the reduced condition in @HagenvonEitzen's answer is quartic (without any obvious simplification available), it may be hard to prove that a list of all solutions is exhaustive. J.-P. Serre's famous problem of determining the rational points of $x^4 + y^4 = 17$ was only resolved in the last 15 or 20 years. $\endgroup$ – Travis May 1 '16 at 14:46
  • $\begingroup$ I have added an Edit to my answer. $\endgroup$ – Jean Marie May 1 '16 at 17:12
12
$\begingroup$

Here is a largely modified version of my initial answer.

(thanks to @Yves Daoust who has pointed out a typo).

In particular, I have used powerful algebraic tools such as Gröbner basis (https://en.wikipedia.org/wiki/Gröbner_basis) and resultants (https://en.wikipedia.org/wiki/Resultant) relying on a computer algebra system (Mathematica) for computations.

Beginning by Vieta's formulas

$$\begin{cases}a+b+c+d&=&-a&(V_1)\\ ab+ac+ad+bc+bd+cd&=& \ b&(V_2)\\ bcd+acd+abd+abc&=&-c&(V_3)\\ abcd&=& \ d&(V_4)\\ \end{cases}$$

Using $(V_1)$ and $(V_4)$, one obtains

$$c=\dfrac{1}{ab} \ \ \text{and} \ \ d=-2a-b-c=-2a-b-\dfrac{1}{ab}$$

Therefore, the two other Vieta's formulas yield two other equations with variables $a$ and $b$ only, under the form:

$$\begin{cases}f(a,b)&=&-1 - 2\,a^2\,b - a\,b^2 - 2\,a^4\,b^2 - a^2\,b^3 - 2\,a^3\,b^3 - a^2\,b^4&=&0 & \ \ \ (1)\\ g(a,b)&=&-a - b + a\,b - 2\,a^3\,b - 3\,a^2\,b^2 - a\,b^3 - 2\,a^4\,b^3 - a^3\,b^4&=&0 & \ \ \ (2)\end{cases}$$

1) First approach: Groebner basis of ((1),(2)).

This is a set of equivalent equations to (1)+(2), simpler in a certain sense, as we will see, at the price, in our case, of degree elevation.

I obtained (as I said, using Mathematica with GroebnerBasis[...] function) the two following equations (factorization of (1') has in fact been made in a second step):

$$\begin{cases}(1') \ \ \ &P_1(b)P_2(b)P_3(b)&=&0 \ \ \text{where}\\ &P_1(b)&=&1+b\\ &P_2(b)&=&1+b+2b^2+b^3\\ &P_3(b)&=&4 - 8\,b + 4\,b^2 + 4\,b^4 - 6\,b^5 - 4\,b^6 + 6\,b^7\\ &&&+ 4\,b^8 - b^{10} - 4\,b^{11} - b^{12} + 2\,b^{13} + b^{14} \ \ \ \text{and}\\ (2') \ \ \ &82012\,a &=& 27660 - 66204\,b - 24260\,b^2 + 20952\,b^3 \\ &&& + 121422\,b^4 + 52670\,b^5 - 94920\,b^6 - 9194\,b^7 \\ &&& + 79461\,b^8 + 45735\,b^9 + 19696\,b^{10} - 34429\,b^{11}\\ &&& - 53258\,b^{12} + 3176\,b^{13} + 35041\,b^{14} + 21135\,b^{15} \\ &&& + 5820\,b^{16} + 809\,b^{17}\end{cases}$$

Two immediate remarks:

  • equation (1') is a polynomial equation in the single parameter $b$,

  • equation (2') determines in a unique manner the value of $a$ once a value of $b$ is given.

Let us concentrate our study on (1'), considering the three cases:

  • $P_1(b)=0$ giving $b=-1$. Using (2'), we deduce that $a=1$, giving
    $c=1/b=-1$, and, as a consequence, $d=-2a-b-c=0$, thus non acceptable.

  • $P_2(b)=0$ is a third degree equation with a unique real solution (this can be established rigorously by studying variations of function $P_2$). This solution is $$b_0=-\dfrac{1}{3}\left(2+\sqrt[3]{\dfrac{25+3 \sqrt{69}}{2}}+\sqrt[3]{\dfrac{25-3 \sqrt{69}}{2}}\right) \approx -1.7548776662.$$

  • $P_3(b)=0$ has no real solution, as can be "seen" either on the numerical values of its roots, all complex non real, or by having a look at its graphical representation as a function. But, till now, I haven't been able to prove it rigorously.

Plugging the numerical value of $b_0$ in (2'), one obtains $a_0=1.$. As a consequence:

$$c_0=\dfrac{1}{a_0b_0}\approx -0.5698402910 \ \ \ \text{and} \ \ \ d_0 = -2a_0-b_0-c_0 \approx 0.3247179572$$ that can be checked as the roots of the function $y=x^4+a_0x^3+b_0x^2+c_0x+d_0$ (see graphical representation below)

This "quatuor" $(a_0,b_0,c_0,d_0)$ is the unique solution of the problem for which $abcd \neq 0$.

Note: we have nothing to check because conditions (1') and (2') (which are equivalent to (1)+(2)) are fulfilled.

2) 2nd approach, using resultants:

A resultant (equal to $0$) is a necessary and sufficient compatibility condition between to parametric polynomials (here $f(a,b)$ and $g(a,b)$) for them to have at least a common root. There are two ways to consider this issue (with notations of "abstract algebra": $K[ a,b ] \approx K [a ] [ b ] \approx K [ b ] [ a ]$).

  • either $a$ is considered as the main variable and $b$ is a parameter. In this case the resultant will generate a condition on $b$, which is exactly the same as condition (1').

  • or, in a reverse way ($b$ the main variable and $a$ the parameter):

$${\left( -1 + a \right) }^4\,a^6\, \left( 1 + 3\,a + 6\,a^2 + 6\,a^3 + 3\,a^4 - 12\,a^5 - 34\,a^6 - 44\,a^7 - 28\,a^8 + 4\,a^9 + 48\,a^{10} + 80\,a^{11} + 80\,a^{12} + 48\,a^{13} + 16\,a^{14} \right)=0 \ \ \ (3)$$

which is an interesting relationship giving solutions $a=1$, $a=0$ (not possible) and no other solution because the third factor (the 14th degree polynomial) has no real roots. The interesting thing in (3) is that $a$ can - theoretically - play the main rôle in the computation of the solution as $b$ has played upwards.

A remark: combining equations (1) and (2):

$$h(a,b):=g(a,b)-a h(a,b)=b\left( a - 1 \right) \,\left( 1 + 2\,a^2\,b + a\,b^2 + 2\,a^3\,b + a^2\,b^2 + 2\,a^4\,b \right)=0 \ \ \ (4)$$

The first two factors $b(a-1)$ account for special values $b=0$ (not considered) and $a=1$, which has been the final value for the solution.

The last factor equal to $0$ is a quadratic equation in variable $b$; it can be solved with $a$ as a parameter (with a certain interval restriction on the values of $a$ for the discriminant to be positive). We will not do any computation: it suffices to say that $b$ is determined as a function of $a$ ...

Here is a graphical representation of the fourth degree polynomial:

$$y= x^4 + x^3 -1.7548776662316001862 x^2 - 0.5698402909809427985 x + 0.32471795721254298472$$

with roots: $x_1 = 1$ , $x_2 = -1.7548776662316001862...$ , $x_3 = -0.5698402909809427985...$ $x_4=0.32471795721254298472...$.

enter image description here

$\endgroup$
  • $\begingroup$ This is a pretty impressive job! ThankYou $\endgroup$ – GohP.iHan May 8 '16 at 11:48
  • $\begingroup$ 1) I have corrected a typo in (*) without any consequence: $b^3+2b^2+b+1$ instead of $b^3+3b^2+b+1$ 2) I will improve the demonstration using a Groebner basis $\endgroup$ – Jean Marie May 8 '16 at 16:04
  • $\begingroup$ Looking forward to it!!! =D $\endgroup$ – GohP.iHan May 9 '16 at 5:47
  • $\begingroup$ I just posted the "final" version, shedding - I think - a completely different light on the issue. $\endgroup$ – Jean Marie May 9 '16 at 6:56
  • $\begingroup$ Isn't $V_2$ missing two terms ? $\endgroup$ – Yves Daoust May 9 '16 at 8:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.