Commutative Diagrams and Polynomials

104 Views Asked by At

Recently, considering how algebraic numbers may be defined by the algebraic properties that they satisfy (and in particular, the polynomials of which they are roots), I started to wonder about, for algebraic $x$ and $y$, how one might go about learning about algebraic properties of $x+y$ or $xy$. A question which arose in my thoughts was the following:

Suppose we are given some polynomial $P$ of $n$ variables and $n$ polynomials $f_1,f_2,\ldots,f_n$ of a single variable. Is it always possible to find some non-constant polynomial of one variable, $\tilde{f}$, and a polynomial of $n$ variables, $\tilde{P}$, such that the following holds for all $a_i$: $$\tilde{f}(P(a_1,a_2,\ldots,a_n))=\tilde{P}(f_1(a_1),f_2(a_2),\ldots,f_n(a_n))$$

For instance, in two variables, if we were given that $$P(a_1,a_2)=a_1a_2$$ $$f_1(a_1)=a_1^2$$ $$f_2(a_2)=a_2^3$$ we could solve the desired equations by setting $$\tilde{P}(x,y)=x^3y^2$$ $$\tilde{f}(x)=x^6$$ which, after substitution, would give $$(a_1a_2)^6=(a_1^2)^3\cdot (a_2^3)^2$$ which is clearly true. However, I have no idea how one might prove that, for any $P$ and $f_i$, such $\tilde{P}$ and $\tilde{f}$ must exist and moreover little idea of how to construct them if they do. I am not certain that this statement must be true either, but I doubt that proving a counterexample would be trivial. (Currently, I'm wondering about $P(a_1,a_2)=a_1a_2$ and $f_1(a_1)=a_1^2+a_1$ and $f_2(a_2)=a_2$)

1

There are 1 best solutions below

2
On BEST ANSWER

For simplicity I'll work over $\mathbb R$.

If $P(x,y)=x+y$, $f_1(x)=x^2$ and $f_2(y)=y^2$, there are no polynomial $\tilde f$ of one variable and $\tilde P$ of two such that $\tilde f$ is non-constant and $\tilde f(P(x,y))=\tilde P(f_1(x),f_2(y))$, that is, such that $$\tilde f(x+y)=\tilde P(x^2,y^2).$$ Differentiating with respect to $x$ on both sides we see that $$\tilde f'(x+y)=\tilde P_1(x^2,y^2)2x$$ with $\tilde P_1$ denoting the derivative of $\tilde P$ with respect to its first argument. Letting $x=0$ in this equation we see that $\tilde f'(y)=0$ for all $y\in\mathbb R$. This implies that $\tilde f$ is constant.