13.11.2014 Views

proofs

proofs

proofs

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Cauchy’s rigidity theorem 83<br />

Now let n ≥ 4. If for any i ∈ {2, . . ., n − 1} we have α i = α ′ i , then the<br />

q<br />

corresponding vertex can be cut off by introducing the diagonal from q i−1<br />

to q i+1 resp. from q i−1 ′ to q′ i+1 , with q i−1 q i+1 = q′ i−1 q′ i+1 , so we are done<br />

by induction. Thus we may assume α i < α ′ i for 2 ≤ i ≤ n − 1.<br />

Now we produce a new polygon Q ∗ from Q by replacing α n−1 by the<br />

largest possible angle α ∗ n−1 ≤ α′ n−1 that keeps Q∗ convex. For this we<br />

replace q n by qn, ∗ keeping all the other q i , edge lengths, and angles from Q.<br />

i<br />

α i<br />

Q:<br />

α n−1<br />

If indeed we can choose α ∗ n−1 = α ′ n−1 keeping Q ∗ convex, then we get<br />

q 1 q n < q 1 qn ∗ ≤ q′ 1 q′ n , using the case n = 3 for the first step and induction<br />

as above for the second.<br />

Otherwise after a nontrivial move that yields<br />

Q ∗ :<br />

q 1<br />

α ∗ n−1<br />

q n<br />

q 1 q ∗ n > q 1 q n (1)<br />

we “get stuck” in a situation where q 2 , q 1 and qn ∗ are collinear, with<br />

Q ∗ :<br />

q 1<br />

α ∗ n−1<br />

q ∗ n<br />

q 2 q 1 + q 1 qn ∗ = q 2 q∗ n . (2) q 2<br />

Now we compare this Q ∗ with Q ′ and find<br />

q 2 qn ∗ ≤ q 2 ′ q′ n (3)<br />

by induction on n (ignoring the vertex q 1 resp. q 1 ′ ). Thus we obtain<br />

q 1<br />

q ∗ n<br />

q ′ 1 q′ n<br />

(∗)<br />

≥ q ′ 2 q′ n − q ′ 1 q′ 2<br />

(3)<br />

≥ q 2 qn ∗ (2)<br />

− q 1 q 2 = q 1 qn<br />

∗<br />

(1)<br />

> q 1 q n ,<br />

where (∗) is just the triangle inequality, and all other relations have already<br />

been derived.<br />

□<br />

We have seen an example which shows that Cauchy’s theorem is not true<br />

for non-convex polyhedra. The special feature of this example is, of course,<br />

that a non-continuous “flip” takes one polyhedron to the other, keeping the<br />

facets congruent while the dihedral angles “jump.” One can ask for more:<br />

Could there be, for some non-convex polyhedron, a continuous<br />

deformation that would keep the facets flat and congruent?<br />

It was conjectured that no triangulated surface, convex or not, admits such<br />

a motion. So, it was quite a surprise when in 1977 — more than 160 years<br />

after Cauchy’s work — Robert Connelly presented counterexamples: closed<br />

triangulated spheres embedded in R 3 (without self-intersections) that are<br />

flexible, with a continuous motion that keeps all the edge lengths constant,<br />

and thus keeps the triangular faces congruent.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!