I've been trying to understand how ring morphisms work when dealing with polynomial rings, but I'm having a little bit of trouble understanding how they work.
Let $R,S$ be rings (they need not be commutative nor have a multiplicative identity) and $f:R\to S$ a ring morphism. We know that this induces a ring morphism $f^*:R[x]\to S[x]$ given by $f^*(a_0+a_1x+\cdots+a_nx^n)=f(a_0)+f(a_1)x+\cdots+f(a_n)x^n$. My problem now is: is every ring morphism $F:R[x]\to S[x]$ of the form $F=g^*$ for some $g:R\to S$?
I don't know whether it is or not the case, but if it is I'm having trouble proving it.
My first approach would be that any polynomial can be thought as a sum and product of constant polynomials and $x^k$ for some $k$ and in that sense, then $F(a_0+\cdots a_nx^n)=F(a_0)+\cdots F(a_n)F(x)^n$, so if I knew that $F$ sends constant polynomials to constant polynomials and polynomials of the form $x^n$ to themselves, then I would have the desired result, but this is something I haven't been able to prove yet.
Only thing I know is that for $c\in R$, $F(c)=b_0+\cdots+b_mx^m$ with $b_i\in S$ for all $i$, but I can't deduce from there that $b_i=c$ if $i=0$ and $b_i=0$ otherwise. Same case with $x^n$ .
If this is not the case, however, and I'm trying to prove something false, then how do these morphisms work? Is there a way to more or less understand their behaviour?
Any help is thanked for :)
The result is not true:
Take $R = S = \mathbb{Z}[y]$. Then $R[x] = S[x] = \mathbb{Z}[x,y]$ has an automorphism given by sending $x$ to $y$ and $y$ to $x$. This cannot be of your desired form, as it sends (for example) the polynomial $x$ to something in $S$, but any map of your desired form must send it to something of the form $sx$, where $s \in S$, which does not lie in $S$.
What your approach gives you is a grading on $R[x]$ (indeed, the very prototypical example of a grading). That is: a collection of subsets $R_i$ with $R = \bigoplus R_i$ as additive groups, such that if $a \in R_i$ and $b \in R_j$, then $ab \in R_{i+j}$: the grading is precisely by $x$-degree, so that $R_i = \{ax^i | a \in R\}$. Then each element of $R[x]$ is a sum of one element from each $R_i$, as you've written.
With that, then, the morphisms that do what you ask for are precisely the graded morphisms (with respect to this grading): that is, the morphisms $\varphi: R[x] \to S[x]$ such that $\varphi(R_i) \subseteq S_i$ for all $i$. Your problems, though, are twofold: