4
$\begingroup$

Consider a degree four polynomial $$ f = a_4x^4 + a_3x^3 + a_2x^2 + a_1x+ a_0 \in \mathbb{R}[x] $$ with real coefficients. The discriminant $\Delta_f$ of $f$ is a homogeneous polynomials of degree six in the $a_i$.

If $\Delta_f > 0$ then either $f$ has four real roots or $f$ does not have any real root at all.

What conditions does one have to add to $\Delta_f > 0$ in order to characterize the degree four polynomials with no real roots?

Thanks a lot.

$\endgroup$
6
  • 6
    $\begingroup$ This should work: en.wikipedia.org/wiki/Sturm%27s_theorem $\endgroup$ Commented Nov 21, 2022 at 18:57
  • 4
    $\begingroup$ There is not a unique answer to this: The locus with no real roots is one of the connected components of $\Delta_f > 0$, and there are many different inequalities which one can use to distinguish the $0$ root component from the $4$ root component. Sturm's theorem is one method. Another method is to compute the Hermite Matrix; see the top voted answer at mathoverflow.net/questions/20946/… . $\endgroup$ Commented Nov 21, 2022 at 20:00
  • $\begingroup$ taking $a_4=1$ and no real roots it is a sum of squares of real polynomials, although finding the coefficients may be messy. But it may be bounded below by a sum of squares that is not difficult to find. $\endgroup$
    – Will Jagy
    Commented Nov 21, 2022 at 21:18
  • $\begingroup$ Thank you to all of you. From Speyer's comment it seems that inside the $\mathbb{R}^5$ parametrizing degree four polynomials the locus parametrizing those with no real roots contains a semialgebraic subset of dimension 5. Is this correct? $\endgroup$
    – Puzzled
    Commented Nov 21, 2022 at 21:22
  • 3
    $\begingroup$ For a real polynomial of degree $n$, there are $n-1$ polynomial inequalities among the coefficients that are necessary and sufficient for all the roots to be real. So in your situation, there are two inequalities that cannot both be satisfied in order for $f$ to have no real roots. I wrote an exposition of this on pages 11-13 of Conversational Problem Solving. $\endgroup$ Commented Aug 3 at 22:29

1 Answer 1

10
$\begingroup$

The answer is contained in this wonderful paper:

E. L. Rees. Graphical Discussion of the Roots of a Quartic Equation. The American Mathematical Monthly, Vol. 29, No. 2 (Feb., 1922), pp. 51-55

In which there is a full elaboration of the possibilities for the nature of the roots of a quartic polynomial.

The paper states that given a quartic in reduced (or sometimes called "depressed") form $x^4 + qx^2 + rx + s$, if the discriminant $\Delta$ is greater than zero, then the 4 roots are distinct, and either all real or all imaginary (which is what you claim in your question).

Additionally, if $q < 0, s > \frac{q^2}{4}$, then all the roots are imaginary; if $q < 0, s < \frac{q^2}{4}$, then all the roots are real; if $q \geq 0$ then all the roots are imaginary. So, you need to add the condition $q \geq 0 \lor (q < 0 \land s > \frac{q^2}{4})$ to make the quartic have no real roots (when $\Delta > 0$).

You can convert this condition back to deal with a "full" quartic, i.e., $f = \sum_{i=0}^4 a_i x^i$, by looking up the procedure of converting a quartic to its depressed/reduced form. There are algebraic expressions relating the coefficients of both forms. Hence you will get such a condition in terms of your $a_0, a_1, a_2, a_3, a_4$.

Edit: to be precise, "imaginary" above means non-real, i.e. a complex number $a+bi$ with $b \neq 0$.

$\endgroup$
3
  • $\begingroup$ Amazing that this extra condition does not involve $r$ $\endgroup$ Commented Aug 4 at 7:13
  • $\begingroup$ Are you using imaginary to mean non-real (in which case the answer does agree with the claim in the question) or real multiplied by $i$ (in which case it's stronger than the claim in the question)? $\endgroup$ Commented Aug 4 at 7:35
  • 1
    $\begingroup$ @PeterTaylor great point. I checked the Rees paper again and indeed they use "imaginary" to mean non-real. I'll clarify this. $\endgroup$
    – Maaz
    Commented Aug 4 at 18:35

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .