I am reposting the second question from here (after clarifying it) on the recommendation of user "GH from MO".

Let $b_1,b_2,\dots$ be an enumeration of $\mathbb Q$.

Question 2: Suppose I define $$G(x,y) = a_0(y) + a_1(y)(x-b_1) + a_2(y)(x-b_1)(x-b_2) + \dots$$ where the $a_k(y)$ are polynomials in $y$ and $g(x) = G(x,x)$.

Suppose the $a_k(y)$ are not identically zero for $k$ large enough. Otherwise, we clearly get polynomials. Is this the only way to get a polynomial? That is, if $g(x)$ is equal to a polynomial function, then is $a_k = 0$ for $k \gg 0$?

  • $\begingroup$ The role of $G(x,y)$ is unclear. Why don't you simply define $g(x)$ as $a_0(x) + a_1(x)(x-b_1) + a_2(x)(x-b_1)(x-b_2) + \dots$? $\endgroup$ – GH from MO Dec 12 '17 at 17:03
  • $\begingroup$ You are right, it is just an artifact of where the problem came from. The $\lambda$ in the previous question was part of the original question but of course if we fix $\lambda$, we might as well take $\lambda = 1$ and scale the $a_i$ accordingly. But then, as you note, we might as well do away with $G(x,y)$ altogether. $\endgroup$ – Asvin Dec 12 '17 at 17:08

The answer is no. There are uniquely determined linear polynomials $a_m(x):=x-c_m$ over $\mathbb{Q}$ such that $g(x)$ is identically zero on $\mathbb{Q}$. Indeed, for such polynomials the condition means that $$ \sum_{0\leq m\leq n-1}(b_n-c_m)\cdot(b_n-b_1)\dots(b_n-b_m) = 0,\qquad n\geq 1. $$ Here, the second product is meant to be $1$ when $m=0$. Now suppose that, for a given $n\geq 1$, the coefficients $c_m\in\mathbb{Q}$ have been determined for $0\leq m<n-1$. Then, the above equation has a unique solution for $c_{n-1}\in\mathbb{Q}$. So we obtained a recursion for the rational numbers $c_0,c_1,c_2,\dots$, and we are done.

  • 2
    $\begingroup$ You should be more ambitious! Your idea can show that, for any function $v$ defined on $\Bbb Q$ (including every polynomial function), there is a unique sequence of monic linear polynomials $a_{m;v}(x) = x-c_{m;v}$ such that $g(x)=v(x)$ for every $x\in\Bbb Q$. $\endgroup$ – Greg Martin Dec 12 '17 at 19:18

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.