17
$\begingroup$

Consider three functions $f, g$ and $h$ on a smooth curve $X$ over $\mathbb{C}$. I have found the following equality: $$\sum (res(f\frac{dg}{g})\frac{dh}{h}-res(f\frac{dh}{h})\frac{dg}{g})=0.$$

Here the sum is over complex points of $X$, $res$ means residue of the corresponding meromorphic form on $X$. The most important point is that object $$res(f\frac{dg}{g})\frac{dh}{h}$$ should be considered as an element of the module of absolute Kahler differentials $\Omega^1_{\mathbb{C}/\mathbb{Q}}.$

For $f=1$ it gives the Weil reciprocity law, since $$res(\frac{dg}{g})\frac{dh}{h}-res(\frac{dh}{h})\frac{dg}{g}=ord(g)\frac{dh}{h}-ord(h)\frac{dg}{g}=$$ $$=\frac{d\{g,h\}_W}{\{g,h\}_W},$$ where $\{g,h\}_W$ stands for the Weil symbol. So, $$0=\sum (res(\frac{dg}{g})\frac{dh}{h}-res(\frac{dh}{h})\frac{dg}{g})= =\sum\frac{d\{g,h\}_W}{\{g,h\}_W}=\frac{d\prod \{g,h\}_W}{\prod \{g,h\}_W}.$$ From this it follows that $\prod \{g,h\}_W$ is algebraic complex number. Since the product of Weil symbols depends continously on the functions, it should be constant. But for $g=h=1$ it is equal to one. This finishes the proof of the Weil reciprocity law.

Question 1: How one can prove this formula? I have found quite a long proof with Newton polygons, following ideas of Askold Khovansky.

Question 2: Can one interprete it as some kind of a reciprocity law for surfaces?

Question 3: I am almost sure that it is known. I would be very grateful for any reference.

$\endgroup$
9
  • $\begingroup$ The sum of interest is a difference between expressions of the form $\sum_P {\rm{res}}_P(\omega)\omega'$ for rational 1-forms $\omega$ and $\omega'$, but such a sum vanishes because it is $\omega'$ times the scalar $\sum_P {\rm{res}}_P(\omega) = 0$. Am I misunderstanding something? $\endgroup$ – user27920 Jul 10 '14 at 0:05
  • $\begingroup$ The argument "by continuity" (I suppose "by holomorphicity") on non-zero functions is a bit hand-wavy, since to make sense of it one should restrict to $g$ and $h$ inside a fixed space such as functions with poles controlled by a divisor, but then the nature of the definition of the Weil symbol has delicate behavior when pole-orders change within that space. So it isn't clear if this really gives a proof of Weil reciprocity just from the residue theorem alone. $\endgroup$ – user27920 Jul 10 '14 at 0:16
  • 1
    $\begingroup$ I think the expressions outside of the residue are supposed to be evaluated at the points, i.e. the sum is $\sum_P \mathrm{res}_P(\omega) \omega'(P)$. This alone does not make sense, but the difference should make sense (like in the Weil reciprocity). This forces $\mathrm{res}(f d\mathrm{log}(h))\mathrm{ord}(g) = \mathrm{res}(f d\mathrm{log}(g))\mathrm{ord}(h)$ at every point. If you take $f=1/x$, $g=x$ and $h = x+x^2$, the expression still doesn't make sense at $x=0$. $\endgroup$ – Pavel Safronov Jul 10 '14 at 5:56
  • 1
    $\begingroup$ There are higher analogs of Weil reciprocity called Suslin reciprocity laws (due to Milnor for $X=\mathbb{P}^1$, Bass-Tate and Suslin in general). They state that the composite $K^M_n(k(X))\rightarrow \prod_{x\in X} K^M_{n-1}(k(x))\rightarrow K^M_{n-1}(k)$ is trivial ($K^M$ are the Milnor $K$-theory groups). For $n=1$ this is "sum of the residues of a rational function is zero" and for $n=2$ this is Weil reciprocity. You can take the $n=3$ Suslin reciprocity which would involve 3 functions and apply the differential symbol, but it would be some statement in $\Omega^2_{\mathbb{C}}$. $\endgroup$ – Pavel Safronov Jul 10 '14 at 6:02
  • $\begingroup$ About the argument "by continuity": it is hand-wavy, but I think, one can make sense out of it by adding parameters. $\endgroup$ – Daniil Rudenko Jul 10 '14 at 7:52
7
$\begingroup$

Look at so-called Parshin reciprocity law, for example Mazin' approach http://www.math.utoronto.ca/mmazin/thesis.pdf

I guess that is exactly what you want.

If I am not mistaken, your version follows from the standard version of Weil reciprocity law for the functions $f/g,f/h$, in the spirit of logarithmic differentials (cf. http://www.math.toronto.edu/askold/osaka.pdf) If not, it is very strange.

$\endgroup$

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.