2
$\begingroup$

Let $M$ be a 1-skeleton of a triangulation of a sphere with $V$ vertices and $E$ edges.

Definition 1 A polyhedron is a map $M\to \mathbb R^3$ that is affine on edges (and non-degenerate on faces). The space of polyhedra is identified with (apparantly an open subset of) $\mathbb R^{3V}$.

Definition 2 Two polyhedra are called isometric if they induce the same metric on $M$. We will denonote this fact by $[P]=[Q]$

Definition 3 A polyhedron $P$ is called infinitesimally rigid if the kernel (at $P$) of the Jacobian of the natural map $f:\mathbb R^{3V}\to\mathbb R^E$ which sends $P$ to the tuple of square-lenghts of its edges has the minimal possible dimension equal to 6. In other words $P$ has no nontrivial infinitesimal deformations (or more algebraically: Zariski tangent space to the scheme $f^{-1}f(P)$ has dimension $6$ at $P$).

It is a well known result of Gluck that the set of infinitesimally rigid polyhedra $M\to \mathbb R^3$ form an open and dense set in $\mathbb{R}^{3V}$. However it can happen that in the set [P] of polyhedra isometric to a given infinitesimally rigid polyhedron $P$ there is a non-rigid representative. Is this phenomenon non-typical? In other words, is the following true:

For a generic $P\in \mathbb R^{3V}$ (you may chose suitable definition) $[P]$ is 6-dimensional (that is, all polyhedra isometric to $P$ are infinitesimally rigid). To put it differently, after quotienting out by isometries of the euclidean space we have that the moduli space of a generic metric polyhedron is a finite union of points (without nilpotents).

$\endgroup$

1 Answer 1

3
$\begingroup$

In short, no, this cannot happen.

To be more specific. A standard definition for ``generic'' is that the coordinates of $P \in \mathbb{R}^{3V}$ form an algebraically independent set. It is a nice exercise proving that: (i) almost all (in the measure theory sense) points in $\mathbb{R}^{3V}$ are generic, and (ii) every generic polyhedra is infinitesimally rigid.

Bob Connelly proved in his paper Generic global rigidity (see Proposition 3.3) the following: given a generic $P \in \mathbb{R}^{3V}$ and $Q \in [P]$, if $P$ is infinitesimally rigid then so too is $Q$. (His result is stated in the more general setting of bar-joint frameworks and so is more general.) The result you need now just follows from basic differential geometry.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.