Let $k$ be a field, $p$-prime number and $P(x,y), Q(x,y)\in k[x,y]$ and consider polynomial $$yQ(x,y^p)-P(x,y^p).$$ I claim that this polynomial is a non-zero polynomial over $k$.
Proof: Consider the following two cases:
1) Suppose that $Q(x,y^p)$ depends on $y$ then $yQ(x,y^p)$ has term $y^{pk+1}$ with $k\geq 1$. But note that $P(x,y^p)$ has not such term. Hence $yQ(x,y^p)-P(x,y^p) \not\equiv 0$.
2) If $Q(x,y^p)$ does not on $y$ then $Q(x,y^p)=F(x)$. Then $yQ(x,y^p)-P(x,y^p)=yF(x)-P(x,y^p)$. But the last polynomial is alzo non-zero because here we also can consider two subcases:
2.1 If $P(x,y^p)$ depends on $y$ then $P(x,y^p)$ has term $y^{pk}$, $k\geq 1$ which $yF(x)$ has not.
2.2. If $P(x,y^p)$ does depend on $y$ then $yF(x)$ has term $y$ which $P(x,y^p)$ has not.
Is this reasoning correct? Would be grateful for any comments.
With the added constraint that $Q(x, y)$ is not identically zero, the reasoning is fine but could be streamlined. It must have at least one non-zero term, which without loss of generality is $a x^b y^c$. Then $yQ(x, y^p)$ has a term $a x^b y^{cp+1}$. But the terms of $P(x, y^p)$ (if there are any - $P$ can be identically zero) are of the form $d x^e y^{pf}$, and we can only have integer solutions to $cp+1 = pf$ if $p = 1$.