3
$\begingroup$

Let K, L, M be integers with gcd(K,L,M) = 1. They determine a connected Lie subgroup G = G(K,L,M) of the cubical 3-torus (ℝ/ℤ)3 via

G = {(x,y,z) ∊ (ℝ/ℤ)3 | Kx + Ly + Mz = 0}

(where 0 denotes the identity element of ℝ/ℤ).

G is a 2-torus and inherits a Riemannian metric of everywhere zero Gaussian curvature from (ℝ/ℤ)3, which belongs to a unique conformal equivalence class of Riemann surfaces of genus 1 (or at most two such classes, depending on the choice of orientation of G).

As is well-known, every Riemannian flat 2-torus is conformally equivalent to the quotient ℂ/L for some lattice L = ⟨1, 𝛕⟩ for some 𝛕 in the set

X = {z ∊ ℂ | Im(z) > 0 and |z| ≥ 1 and |Re(𝛕)| ≤ 1/2}.

Such a 𝛕 is determined uniquely unless i ≠ 𝛕 ∊ ∂X, in which case the boundary points 𝛕 and conj(-𝛕) correspond to the same class. Now suppose we have chosen an orientation for G.

In terms of (K,L,M), what is the parameter 𝛕 corresponding to the Riemannian flat torus G ? Equivalently, what is the j-invariant of G ?

$\endgroup$
3
  • 1
    $\begingroup$ I would say that a Riemann surface is a surface with a complex structure, not a surface with a Riemannian metric. $\endgroup$ Commented Jun 2, 2022 at 17:12
  • $\begingroup$ @MikhailBorovoi: I've seen the terminology "Riemannian surface" being used to signify the latter object, which can be mighty confusing, especially since this distinction is not preserved well under translation into various other languages. And yet the terminology gets translated that way, and attempts to translate back to English make the situation even messier :) $\endgroup$
    – M.G.
    Commented Jun 2, 2022 at 17:31
  • 2
    $\begingroup$ Mikhail Borovoi — I am using the fact that any complete connected oriented Riemannian surface is conformally equivalent to a unique Riemann surface. $\endgroup$ Commented Jun 2, 2022 at 18:29

1 Answer 1

2
$\begingroup$

This may be easier than I thought. Assume WLOG that |K| ≥ |L| ≥ |M| with gcd(K,L,M) = 1 as above.

Then the subgroup H ⊂ ℤ3 of integer translations of ℝ3 that preserves the 2-plane P, defined by

P = {(x,y,z) ∊ ℝ3 | Kx + Ly + Mz = 0}

(where 0 denotes 0) is generated by v = (-L,K,0) and w = (0,-M,L):

H = ⟨v, w⟩ ⊂ P

as a subgroup of P, which itself is a subgroup of ℝ3. We can assume the counterclockwise angle in P from v to w is less then pi radians, and this condition will determine the orientation on P.

But the 2-torus in question is the quotient P / H. Therefore the parallelogram generated by v = (-L,K,0) and w = (0,-M,L) is a fundamental domain in P for this group action.

Therefore the parameter 𝜏 is determined a) by cos(angle(1,𝜏)) = cos(angle(v,w)), where

cos(angle(v,w) = -KM / (√(K^2 + L^2) √(L^2 + M^2))

and b) by the condition that |𝜏| = ||v|| / ||w||, where

||v|| / ||w||  =  √(K^2+L^2) / √(L^2+M^2).

The above assumes that either division by zero is a valid operation, or else that the case (K,L,M) = (1,0,0) is excluded a priori as trivial.

$\endgroup$
1
  • $\begingroup$ The $v$ and $w$ don’t actually generate H. Take $(K,L,M)= (3,2,2)$. Then $(0,1,-1)$ is in H but not in the sub lattice generated by v, w. This was pointed out by my student Larsen Linov. So at least you need to divide by the gcd of the entries. What is your proof that these generate? $\endgroup$
    – Ian Agol
    Commented Jun 3, 2022 at 15:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .