Suppose the real number $\lambda \in (0,1)$, and let $n$ be a positive integer. Prove that all roots of the polynomial $$f\left ( x \right )=\sum_{k=0}^{n}\binom{n}{k}\lambda^{k\left ( n-k \right )}x^{k}$$ have modulus equal to $1.$

The Putnam problem 2014 B4 is similar: Show that for each positive integer $n,$ all the roots of the polynomial $\sum_{k=0}^n 2^{k(n-k)}x^k$ are real numbers.

  • 4
    $\begingroup$ The polynomial has "symmetrical" coefficents and can therefore be written as $g(\frac{x+\frac{1}{x}}{2})$ where $g$ is a polynomial. Then the problem is reduced to showing that $g$ has all its roots real and inside $[-1,1]$. Here techniques as in the Putnam question mentioned in the OP would probably be useful. $\endgroup$ – Ewan Delanoy May 3 '18 at 11:38
  • 2
    $\begingroup$ Also posted on MO: mathoverflow.net/questions/299304/… $\endgroup$ – Wolfgang May 3 '18 at 12:19
  • 1
    $\begingroup$ @EwanDelanoy Well, $x+1$ has “symmetric” coefficients but how to write it as a polynomial of $\dfrac12\left(x+\dfrac1x\right)$? $\endgroup$ – Saad May 8 '18 at 9:00
  • $\begingroup$ @AlexFrancisco This is a classical exercise : by induction on $n$, one shows that $x^n+\frac{1}{x^n}$ is of the form $Q_n(x+\frac{1}{x})$ where $Q_n$ is a uniquely defined polynomial of degree $n$. The exact term is "palindromic", by the way. $\endgroup$ – Ewan Delanoy May 8 '18 at 10:09
  • $\begingroup$ @EwanDelanoy Isn't this true only for polynomials of even degree? $\endgroup$ – Servaes May 11 '18 at 14:26

Write the above polynomial in the form: $a+bx+cx^2$... and observe that if we put $f(x)=0$ it is same as putting $x^k f(1/x)=0$ {property of binomial coefficients and the power of lambda} conclude that if $x$ is a root then $1/x$ is also has the same modulus. Hence $|x|=1$


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.