While studying, I found the following problem:

Let $f, g \in F[t]$. Prove that $\exists p \in F[x, y], p \neq 0 : p(f(t), g(t)) = 0$

I'd thank any hints that point me in the right direction.

  • 1
    $\begingroup$ Try starting with $n=\operatorname{lcm}(\operatorname{deg}f,\operatorname{deg}g)$, $p(x,y)=x^{n/\operatorname{deg}f}-x^{n/\operatorname{deg}g}$. $\endgroup$ – Alexander Gruber Dec 16 '14 at 4:59

I think this question is more related to elimination theory than to linear algebra: this is about finding a polynomial that vanishes on the parametric curve $t\mapsto(f(t),g(t))$, thus "eliminating" the parameter $t$.

However I think the following linear-algebra inspired proof can be given. Consider the usual subspaces $F[t]_{<d}$ of polynomials in$~t$ of degree${}<d$ and similarly $F[x,y]_{<d}$ of polynomials in$~x,y$ of total degree${}<d$. Then $\dim(F[t]_{<d})=d$ and $\dim(F[x,y]_{<d})=\binom{d+1}2$. For the given $f,g\in F[t]$ let $L_{f,g}:F[x,y]\to F[t]$ be the substitution map $p\mapsto p[x:=f,y:=g]$, which is clearly $F$-linear; any nonzero element $p\in\ker(L_{f,g})$ will answer our question. Now with $m=\max(\deg(f),\deg(g))$ and any $d$, the restriction of$~f$ to $F[x,y]_{<d}$ has its image contained in $F[t]_{<md}$. Since $\binom{d+1}2$ grows quadratically with$~d$, one will have $\binom{d+1}2>md$ for sufficiently large$~d$; indeed this happens as soon as $d\geq2m$. For such $d$ the restriction of $L_{f,g}$ to $F[x,y]_{<d}$ cannot be injective for dimension reasons, completing the proof.


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.