Define the length of a polynomial $A = a_0 + a_1x+\cdots +a_nx^n \in \mathbb{R}[x]$ to be $$L(A) = \sum_{i=0}^n |a_i|$$, and let $$\lambda (r,s) : = \inf \frac{L(AB)}{L(A)L(B)}$$ where the degree of $A$ is at most $r$, and the degree of $B$ is at most $s$.

Prove that $\lambda(r,r) \leq \frac{1}{2^r}$.

I've tried two ideas, neither of which I could really get to work. The first was an induction argument, in which I was able to successfully show the base case, but couldn't get the inductive step.

The second thought I had was to see if I could just explicitly find polynomials $A$ and $B$ that have length $(1/2)^r$ for any $r$, that would also work to show the infimum is less then that value.

Any hints or thoughts would be greatly appreciated.

  • $\begingroup$ Source of this problem, please? $\endgroup$ – Gerry Myerson Sep 26 '18 at 7:11
  • 1
    $\begingroup$ This is from Dr. David Masser's book titled Auxilliary polynomials, and it's problem 5.5 $\endgroup$ – JonHales Sep 26 '18 at 13:10

For a fixed $r\in\mathbb{Z}_{\geq0}$, let $A(x):=(1+x)^r$ and $B(x):=(1-x)^r$. Then, $$L(A)=L(B)=L(AB)=2^r\,.$$ That is, $$\frac{L(AB)}{L(A)\,L(B)}=\frac{1}{2^r}\,.$$ Hence, $$\lambda(r,r)\leq \frac{1}{2^r}\,.$$

There are some easy-to-prove results. First, $\lambda(r,0)=\lambda(0,r)=1$ for all $r\in\mathbb{Z}_{\geq 0}$. Next, $$\lambda(r,1)=\lambda(1,r)\leq \dfrac{1}{r+1}$$ for all $r\in\mathbb{Z}_{>0}$, where the equality holds, at least, for $r=1$. Finally, for all positive integers $r$ and $s$, we get $$\lambda(r,rs)=\lambda(rs,r)\leq \dfrac{1}{(s+1)^r}\,.$$

  • $\begingroup$ I knew there was something simple, thanks a ton. $\endgroup$ – JonHales Sep 26 '18 at 14:51
  • $\begingroup$ @JonHales Does the book mention any specific formula for $\lambda(r,s)$? It would be interesting to know. $\endgroup$ – Batominovski Sep 26 '18 at 14:53
  • $\begingroup$ It does not, the concept was introduced in the exercises. It might in some later chapters. I'll let you know if it turns up. $\endgroup$ – JonHales Sep 26 '18 at 14:54
  • $\begingroup$ Please do. This is an interesting problem. So, I would greatly appreciate what you'd find out. $\endgroup$ – Batominovski Sep 26 '18 at 14:55
  • $\begingroup$ I asked my professor today during lecture, and he said as far as he was concerned, currently in the literature, there is no known solution for how to calculate $\lambda(r,s)$ for a given $r$ and $s$. Sounds like an interesting research problem! $\endgroup$ – JonHales Sep 28 '18 at 16:50

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.