5
$\begingroup$

Let $S$ be closed hyperbolic surface with genus $g\geq 2$. Let $Teich(S)$ be the Teichmueller space of $S$. It's well known that $Teich(S)$ is diffeomorphic to a (6g-6)-dimensional cell, where a coordinatization is given by the Fenchel-Nielsen length-twist coordinates $\{(\ell_a, \tau_a) \}_{a\in P}$ associated to a pant decomposition $P$ of $S$.

Question: can we replace the twist parameters with length parameters of other curves, and thereby replace the "length-twist" coordinates with "length-length" coordinates on $Teich(S)$?

Related question: are any formulas known which express the twist differentials $d\tau_a$ with linear combinations of length differentials $d\ell_{a'}$ in Wolpert's formula for the Weil-Petersson symplectic form $\omega_{WP}=\sum_{a\in P} d\ell_a \wedge d \tau_a$ ? I.e. can we express $\omega_{WP}$ using only differentials of length functions?

The same question: if we are given a hyperbolic surface $S'$ with known length parameters relative to a pant decomposition, then what metric properties on the surface $S'$ do we use to identify the twist parameters $\{\tau_a\}_a$?

Remark. This question risks being a duplicate. However we find the answer to the above question unsatisfactory, as indicated by our comments below.

In genus $g=2$ we obtain the following almost canonical collection of six simple closed curves on the surface $S$. The value of the lengths of the green curves do not distinguish between left and right Nielsen twists along the red curves. However the derivatives of the lengths of the green curves do distinguish between left and right Nielsen twists along the red curves. This is similar to how the derivatives of strictly convex functions $f: \mathbb{R}^3 \to \mathbb{R}$ are injective where $Df(x_1)=Df(x_2)$ if and only if $x_1=x_2$. Here we are assuming that the lengths of the green curves are basically convex functions in the Neilsen twist parameters in the red curves.

The lengths of the red curves plus their angles of intersection coordinatize Teichmueller space. Wolpert's work shows the differentials of the lengths of the green curves are parameterized by the angles of geodesic intersections between the red and green curves.

Answer: Consider genus $g=2$. Let $t(a)$, $t(b)$, $t(c)$ be the Nielsen tangent vectors in Teich defined by the red geodesic pant decomposition $\{a,b,c\}$. Let $\{a', b', c'\}$ be the "dual pant". Then I propose that the functions $$\ell_a,~~~~~ \ell_b,~~~~~ \ell_c$$ together with the ``cosines of the angles of intersection" $$d\ell_{b'}(t(a)),~~~~~ d\ell_{c'}( t(b))~~~~~~,d\ell_{a'}(t(c)))$$ are globally well defined coordinates on Teich. Notice the collection is of cardinality $6g-6=6$ for genus $g=2$. So they are not length-length coordinates, but length-"d"length coordinates. Here I'm assuming all of Wolpert's work, especially pp.252 in Wolpert's 1983 paper referenced by Alex Nolte's answer.

Likewise if we use Wolpert's Reciprocity formula $d \ell_a (t(b))=-d\ell_b(t(a))$, then we obtain another "reciprocal" global coordinate system on Teich. The idea is that the angles of geodesic intersection are effective parameters of the Nielsen twist parameter along the pant cuffs.

$\endgroup$
10
  • 1
    $\begingroup$ Nielsen, not Nielson. $\endgroup$
    – abx
    Commented Jan 4, 2022 at 15:47
  • 4
    $\begingroup$ Your first question is answered here: mathoverflow.net/questions/243622/… $\endgroup$ Commented Jan 4, 2022 at 16:18
  • 1
    $\begingroup$ @JHM: I think what's going on there is that the sum of the lengths is the proper function, and must achieve a minimum. The point is that if one curve gets too short, then any of the other curves that intersects it must get long. $\endgroup$ Commented Jan 4, 2022 at 21:54
  • 1
    $\begingroup$ Actually, it is easy to see that if a (finite) collection of curves parameterizes then it must be filling. $\endgroup$ Commented Jan 5, 2022 at 2:31
  • 1
    $\begingroup$ @JHM: In your genus-2 example the lengths of the curves do not distinguish between a hyperbolic metric and its mirror image in the plane in which it is drawn, so I guess one should specify what exactly it means to "determine a hyperbolic metric". This is related to the example in my answer, that shows that the twist parameters can be determined up to a sign, which changes under reflection. $\endgroup$ Commented Jan 5, 2022 at 9:48

2 Answers 2

1
$\begingroup$

Though there are never $6g-6$ global length coordinates on Teichmüller space, around any point one can always find $6g-6$ length functions that give local coordinates for Teichmüller space. For such local coordinates, there is in fact an expression of the Weil-Petersen symplectic form analogous to Wolpert's formula in Fenchel-Nielsen coordinates, though it does still depend on twists.

To see the existence of such local coordinates, just take a collection of length coordinates $l_1, ..., l_k$ so that $\Sigma \mapsto (l_1(\Sigma), ..., l_k(\Sigma ))$ is an immersion. Then around any $\Sigma$, there is a subset $l'_{1}, ..., l'_{6g-6}$ so that $\Sigma \mapsto (l'_1(\Sigma), ..., l'_{6g-6}(\Sigma))$ has full rank at $\Sigma$. Here $k$ must be $\geq 6g-5$, and $6g-5$ is sharp, as discussed in the question linked in your question. A systematic way to take $k = 9g-9$ such length functions that has a clear relationship to twist parameters is to take the curves of an embedded pants decomposition, take curves $K_i$ meeting each pants curve in two essential points, and $K_i'$ the Dehn twist of $K_i$ around the pants curve. This appears in Thurston's Work on Surfaces, chapter 7.

In any such local coordinates $(l'_1, ..., l'_{6g-6})$ associated to simple closed curves $\alpha_1, .., \alpha_{6g-6}$, Wolpert computes in "On the Kähler form of the Moduli Space of One-Punctured Tori" (1983) an expression for $\omega$. Denoting by $t_i$ the vector field generated by a twist about $\alpha_i$, $\omega_{jk} = \omega(t_j, t_k)$, and $W_{jk}= \omega^{jk}$, he shows $$\omega = - \sum_{j < k}W_{jk} dl_j \wedge dl_k.$$

$\endgroup$
1
  • $\begingroup$ Wolpert (1983) is interesting, especially pp. 252. But I dont follow your argument that $6g-5$ is sharp. My problem with length functions is that they dont readily distinguish between left and right Nielsen twists. And the lengths are like convex functions in the Nielsen twist parameter, and non monotone. $\endgroup$
    – JHM
    Commented Jan 5, 2022 at 20:18
1
$\begingroup$

Regarding your second question, explicit expressions are known for the absolute value of the twist parameter $\tau_a$ in term of the lengths $\ell_a$ supplemented with the lengths $\ell'_a$ of certain transverse curves. See for instance equation (5.12) in Andersen, Borot, Charbonnier, Giacchetto, Lewański, Wheeler. On the Kontsevich geometry of the combinatorial Teichmüller space. arXiv:2010.11806. In the case that $\ell_a$, $\tau_a$ are coordinates associated to a curve $\gamma$ that separates two distinct pairs of pants, it allows to solve for $\cosh \tau_a$ in terms of the length $\ell_a$ of $\gamma$, the length $\ell'_a$ of $\delta$ and the lengths of the other four boundaries of the pairs of pants. See Figure 19 in their paper:

enter image description here

$\endgroup$
1
  • $\begingroup$ Thanks for precise references. The eq (5.12) is quite elaborate. Buser's "Spectra", section 3.3 has further details. $\endgroup$
    – JHM
    Commented Jan 5, 2022 at 12:57

You must log in to answer this question.

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