0
$\begingroup$

Suppose we have a polynomial $p$

$$ p = a_0 + a_1x + a_2x^2 + \dots+a_Nx^N = \sum_{i=0}^Na_ix^i. $$

what is the sufficient and necessary condition of $\{a_i:i=0,1,\dots, N\}$ such that $p$ is always positive for all $x\in(0, +\infty)$

(does such sufficient and necessary conditions even exist?)

In other words: what kind of (and necessary) coefficients could make the polynomial always positive on the right side of $\mathbb{R}?$

Guess: what about let $\xi_k$ be all of the roots of $\frac{dp}{dx}=0$, and the condition is $p(\xi_k)$ are all positive, and $a_N>0$?

$\endgroup$
8
  • $\begingroup$ Why the RHS lost its $a_i$s? $\endgroup$ Jun 24 '19 at 11:18
  • $\begingroup$ @user10354138 sry, edited. $\endgroup$ Jun 24 '19 at 11:21
  • $\begingroup$ There is no real need for the denominators $i!$, you can very well work with the coefficients $b_i=a_i/i!$. $\endgroup$
    – user65203
    Jun 24 '19 at 11:46
  • $\begingroup$ @YvesDaoust Yes, indeed. Edited. $\endgroup$ Jun 24 '19 at 11:50
  • $\begingroup$ Your guess involved roost and it is not completely in terms of the coefficients. In terms of roots a N and S condition is that all the roots are negative. But I don't know if this can be written in terms of the coefficients. $\endgroup$ Jun 24 '19 at 11:51
3
$\begingroup$

A necessary and sufficient condition uses Sturm's theorem. You want $a_N > 0$ and $V(0) = V(\infty)$, where $V(x)$ is the number of sign changes of the Sturm sequence of $p$ at $x$.

$\endgroup$
3
  • $\begingroup$ so, basically, the polynomial has to have no roots on $(0, +\infty)$, and $a_N>0$ $\endgroup$ Jun 24 '19 at 12:46
  • $\begingroup$ May I further ask, if using the polynomial in my question, are there any general expression for the Sturm's sequence in terms of those coefficients? $\endgroup$ Jun 24 '19 at 15:19
  • $\begingroup$ $p_0 = p = \sum_{i=0}^n a_i x^i$, $p_1 = p' = \sum_{i=1}^n i a_i x^{i-1}$, $p_{i+1} = - \text{rem}(p_{i-1}, p_i)$: its coefficients are rational functions of the coefficients of $p_i$ and $p_{i+1}$, but I don't know of a general expression. $\endgroup$ Jun 24 '19 at 17:52

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.