Published on

Geometric Topology 14

Authors

Geometric Topology 14

Surfaces Without Boundary

We shall continue to model surfaces combinatorially, by means of one or more words. But to allow boundaries we shall allow each letter (or edge) to appear twice or once.

Example: As mentioned previously, the word abcbabcb describes a Mobius band MM whose boundary d=ac1d = ac^{-1} is homeomorphic to a circle. That's because the start of aa and end of c1c^{-1} get mapped to the same point of MM, as do the end of aa and start of c1c^{-1}. If we set ab=x1ab = x^{-1} (equivalent to cutting abx+x1cbabx + x^{-1}cb and then pasting) then:

abcbabx+x1cbxab+b1c1xxac1xxxd.\boxed{a{\color{red}{b}}}cb \sim abx + x^{-1}cb \sim xab + b^{-1}c^{-1}x \sim xac^{-1}x \sim xxd.

We can now 'sandwich' dd, using an insert and two more (cut and paste) substitutions:

xxdxxdu1uuxxdu1 (y=ux    x=u1y)yu1ydu1 (z=yu1    y=zu)zz(udu1).xxd \sim xxdu^{-1}u \\ \sim \boxed{ux}xdu^{-1} \ (y = ux \implies x = u^{-1}y) \\ \sim \boxed{yu^{-1}} ydu^{-1} \ (z = yu^{-1} \implies y = zu) \\ \sim zz(udu^{-1}).

Cuffs

The trio udu1udu^{-1} within a word (in which d,ud,u do not appear elsewhere) is called a cuff. The letter uu helps to distinguish different boundary components, and the presence of a cuff indicates that one disk has been removed from the surface.

Proposition: If a word WW represents a surface M=P^\mathscr{M} = \hat{\mathscr{P}} (with or without existing boundary), then Wudu1Wudu^{-1} represents a surface that is homeomorphic to M\mathscr{M} minus an open disk (assuming that u,du,d do not appear in WW).

Proof: A polygon with boundary Wudu1Wudu^{-1} is shown. Folding the uu edges defines the required homeomorphism between the quotients. \blacksquare

Naked Edges

A problem is that two separate single edges inserted into the boundary of a polygon do not always produce two components. Here they unite into a single curve:

abda1eb1abda1eb1 (abd=x    b1=dx1a)xa1edx1ax1axa1ed.abda^{-1}eb^{-1} \sim \boxed{a{\color{red}{b}}d} a^{-1} eb^{-1} \ (abd = x \implies b^{-1} = dx^{-1}a) \\ \sim xa^{-1}edx^{-1}a \\ \sim x^{-1}axa^{-1}ed.

Since all vertices of x1axa1A1x^{-1}axa^{-1} \sim \mathbb{A}_1 are equivalent, eded is homeomorphic to a circle, so the quotient surface is a torus with only one disk removed. To obtain two circles, one must 'insulate' the edges with cuffs; havind one this, the derivation above yields:

ab(udu1)a1(vev1)b1(x1axa1)(vev1)(udu1).ab (udu^{-1})a^{-1}(vev^{-1})b^{-1} \sim (x^{-1}axa^{-1}) (vev^{-1})(udu^{-1}).

This word represents an orientable surface with χ=36+1=2\chi = 3-6+1 = -2, and each cuff contributes 1-1 (1 extra vertex and 2 edges).

The number rr of cuffs is the number of boundary components (necessarily connected compact 11-manifolds, each homeomorphic to the circle). If the surface is orientable and embedded in R3\mathbb{R}^3, then the boundary is a link and each component is a knot. The latter will not in general be ambient isotopic to a circle (think Seifert surface).