16
$\begingroup$

What was the primary motivation for the study of the amoebas?

$\endgroup$
  • $\begingroup$ Hah, that article is really interesting. But does it really have anything to do with tropical geometry? $\endgroup$ – rschwieb May 2 '14 at 17:47
  • 3
    $\begingroup$ Maybe this Notices of the AMS will help: What is an Amoeba?. I'm surprised this isn't cited in the Wikipedia article. $\endgroup$ – Dave L. Renfro May 2 '14 at 17:51
  • 2
    $\begingroup$ Here's a quite readable article. Amoebas are related to tropical geometry. arxiv.org/abs/math/0403015 $\endgroup$ – Fredrik Meyer May 2 '14 at 18:06
  • 1
    $\begingroup$ @Fredrik Well, tropical geometry gives a lot of reasons to study tropical curves; tropical curves are, of course, degenerate amoebae — (which explains the relation of tropical geometry to usual one, but) that doesn't quite explain why «real» (i.e. not degenerate) amoebae are interesting... $\endgroup$ – Grigory M May 9 '14 at 15:29
  • 1
    $\begingroup$ AFAIK, one area where amoebae appear is (random) dimer configurations (see papers of Kenyon &co — e.g. Dimers and Amoebae or (perhaps, better) Lectures on Dimers). $\endgroup$ – Grigory M May 9 '14 at 15:34
4
+250
$\begingroup$

I heard they came from studying Laurent series in several variables. Consider Pascal's triangle of $\binom{n}{k}$ for $n \geq k \geq 0$. Can we find a generating series? Perhaps

$$ \sum_{n \geq k \geq 0} \binom{n}{k} x^k y^n = \sum_{n \geq 0}(x+1)^n y^n = \frac{1}{1 - (x+1)y}$$

This suggest maybe trying an alternative that makes $x$ and $y$ more symmetric:

$$ \sum_{n \geq k \geq 0} \binom{n}{k} x^k y^{n-k} = \sum_{n \geq 0}(x+y)^n = \frac{1}{1 - (x+y)}$$

This only works if $|x + y| < 1$, since we get convergence. Probably we also need $|x| < 1$ and $|y| < 1$ for the first sum to work.


Using Cauchy integral formula, we can recover Laurent series from this rational function. So, if $f(z) = a_0 + a_1 z + a_2 z^2 + \dots $ then

$$ a_n = \oint \frac{dz}{2\pi i z} \times z^{n+1}f(z)$$

Could this formula work in two dimensions. Maybe we can recover the binomial coefficients like so:

$$ \binom{m+n}{m} = \oint \frac{dz}{2\pi i z} \oint \frac{dz}{2\pi i z} \cdot z^{m+1}w^{n+1} \cdot \frac{1}{1- (z+w)}$$

Here $\times, \cdot$ are just multiplication over complex numbers.

This is a double contour integral and whether the "pole" $z+w =1$ lies inside the contour is related to amoebas, I think.

Then you have to solve the equation $|z+w| = |e^{2\pi i \theta} + e^{2\pi i \phi}|\leq 1$ and these are related to the sides of the triangle.

arXiv:math/0701039 - How to compute $\sum \tfrac{1}{n^2}$ by solving triangles (Mikael Passare)


EDIT I found the statement in Discrimants, Resultants and Multidimensional Determinants by Kapranov, Fomin and Zelevinsky.

Chapter 6, Corollary 1.6 Let $f(x)$ be a Laurent polynomial. All the components of the complement $\mathbb{R}^k - \log (Z_f)$ to the amoeba of $f$ are convex subsets in $\mathbb{R}^k$. They are in bijective correspondence with the Laurent series expansions of the rational function $\tfrac{1}{f(x)}$.

That is what I did here, is work through different expansions of $\tfrac{1}{1-(x+y)}$ and understand why the equation amoeba $\{ (\log|z|, \log|w|) : z+w = 1 \}$ makes an appearance.

$\endgroup$

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.