For an abelian variety $A_{/\mathbb{Q}}$, its volume $vol(A(\mathbb{R}))$ appears in the conjectured Birch Swinnerton-Dyer formula for the L-series at 1.

I am having trouble in understanding the size of this volume, as a function of the (minimal) equations defining the variety. Hence the question:

Say we have a hyperelliptic curve given by $$y^2 = x^{2g+1}+a_{2g}x^{2g}+\ldots +a_0$$

for simplicity, assume that $a_i\approx M > 0$ for all $i$. What is a good crude estimate for the volume of the Jacobian?

Note: for an elliptic curve this is easy, since there is only one differential and hence only one simple integral.

  • $\begingroup$ If you really want an answer and none is forthcoming you might consider asking on MO. $\endgroup$ – Qiaochu Yuan Dec 26 '10 at 20:54
  • $\begingroup$ I'm afraid it'll be closed for not being research level :/ $\endgroup$ – simplequestions Dec 26 '10 at 23:16
  • $\begingroup$ How do you define the volume of an abelian variety? $\endgroup$ – Kevin H. Lin Dec 27 '10 at 2:02
  • $\begingroup$ $A_{/\mathbb{Q}} \rightarrow A_{/\mathbb{C}}$, the volume is then the determinant of the pairing matrix between a basis for the real differentials and a basis for the real homology (loops in complex homology fixed by complex conjugation). $\endgroup$ – simplequestions Dec 30 '10 at 12:18

I assumed above that all coefficients are positive, but this isn't easiest to compute with, so apply the isomorphism $x \mapsto -x$, and now the coefficients are alternating. The volume is unchanged.

A basis for the differentials is $\{\omega_i : 1\le i \le g\}$, where $$\omega_i = x^{i-1}\frac{dx}{2y}$$

For $M$ large, there is a positive root $\alpha_1$ of size $M$, and the other $2g$ roots are small. We now need to figure out a basis for the real homology (loops fixed by complex conjugation). There are $g-1$ independent loops $\gamma_{j+1}$, the $j$-th loop being from $\alpha_{2j}$ to $\alpha_{2j+1}$ and back around the "hole" (for example, if all roots are real, these are simply the paths where the polynomial is positive). One more independent loop is from $\alpha_1$ to $\infty$ and back (being real, this is the easiest to understand).

The volume of the jacobian of the curve is $|det((\int_{\gamma_j} \omega_i)_{ij})|$. So let's approximate these integrals.

For any $j>1$, since the roots involved are $O(1)$, the loop $\gamma_j$ is close to the origin. In this vicinity the root $\alpha_1$ contributes to $\int_{\gamma_j} \omega_i$ a factor of $M^{-1/2}$, and if we factor it out (recall that we only want a crude approximation), we are left with an integral close to the origin over a function that is relatively small. Hence $\int_{\gamma_j} \omega_i \approx M^{-1/2}$. Note that this is the same for any $i$ since close to the origin $x^{i-1}$ doesn't add much.

For $j=1$, the integral can be approximated as $$\int_{\gamma_1} \omega_i = 2\int_M^\infty \frac{x^{i-1} dx}{2y} \approx \int_M^{2M} \frac{x^{i-1} dx}{\sqrt{x^{2g+1}+\cdot a_0}}$$

The small roots give each the denominator a factor of $M^{1/2}$, so we can approximate further to arrive at: $$M^{-g} \int_M^{2M} \frac{x^{i-1} dx}{\sqrt{x-M}} \approx M^{1/2+i-1-g}$$

We have approximated all entries of the matrix. The top row's largest coefficient is of size $M^{-1/2}$, and hence the largest product of a generalised diagonal is $(M^{-1/2})^g$, since the entries in the other rows are also of size $M^{-1/2}$.

The crude approximation for the volume of the jacobian of a hyperelliptic curve given by a polynomial as above with positive coefficients of size $M$, a large number, is: $$Vol(J(\mathbb{R})) \approx M^{-g/2}$$


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.