Let $f_1,f_2,\ldots,f_n\in \mathbb C[z_1,\ldots, z_n]$ be such that the quotient ring $$A:=\mathbb C[z_1,\ldots, z_n]/(f_1,f_2,\ldots,f_n)$$ is finite dimensional (in other words, it's a zero-dimensional complete intersection). I've heard that such a ring is always a Frobenius algbera when equipped with the counit $$ \varepsilon(g):= \int_{|f_1|=\ldots=|f_n|=1} \frac{g(z_1,\ldots,z_n)}{\prod_i f_i(z_1,\ldots,z_n)}dz_1\ldots dz_n, $$ and I'd like to check that claim. By definition, $A$ is a Frobenius algebra if the bilinear form $\langle g,h\rangle:=\varepsilon(gh)$ is non-degenerate. So I would like to know whether:

If $g\in \mathbb C[z_1,\ldots, z_n]$ is such that $$\int_{|f_1|=\ldots=|f_n|=1} \frac{g(z_1,\ldots,z_n)h(z_1,\ldots,z_n)}{\prod_i f_i(z_1,\ldots,z_n)}dz_1\ldots dz_n=0$$ for every $h\in \mathbb C[z_1,\ldots, z_n]$, does is then follow that $g$ is in the ideal generated by $f_1,f_2,\ldots,f_n$?


Yes, this is discussed in page 659 of "Principles of Algebraic Geometry" by Griffiths and Harris. They call it the local duality theorem.

  • $\begingroup$ Ok. I'll check it out. Is it easy to say what kind of ingredients come into the proof? $\endgroup$ – André Henriques Oct 23 '14 at 21:52
  • 1
    $\begingroup$ @AndréHenriques The main steps are roughly: 1) Reduce the global duality theorem (the statement in your question) to the local case around a single zero. 2) Prove it for the case $f_i=z_i^{k_i}$. 3) Deduce the general form by using the transformation formula (which says that if you take the residue with respect to $Af$ then the residue changes by factor of $\det (A)$.) $\endgroup$ – Gjergji Zaimi Oct 23 '14 at 22:14

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.