Orientations Have to Agree Across a Glued Edge
A polygonal surface is assembled from faces by identifying their sides. An orientation chosen on one face determines a direction along each of its boundary sides. When two sides are identified, the directions on those sides cannot be chosen independently: they must fit together to give an orientation near the resulting edge. The endpoint map of the identification tells us exactly what compatibility requires.
We will use a reference orientation on each polygon face. For a polygon drawn in the plane, take its counterclockwise orientation as the reference. Its induced boundary direction is then counterclockwise around the polygon. Reversing the face orientation reverses every one of these boundary directions.
For a face \(F\), write \(\sigma_F=+1\) when we use its reference orientation and \(\sigma_F=-1\) when we reverse it. The boundary direction on a side of \(F\) is consequently \(\sigma_F\) times its reference boundary direction. The signs record orientation choices; they are not lengths or geometric measurements.
The Compatibility Rule
Proof. Choose coordinates near an interior point of the glued edge so that the two face pieces lie on opposite sides of a horizontal segment. An orientation of the resulting surface must restrict to the same orientation in a small disk crossing the segment. With the outward-normal-first convention for boundary orientation, the two half-disks induce opposite directions on their common boundary segment: for the upper half-disk with orientation \(dx\wedge dy\), the induced direction is left to right; for the lower half-disk with that same orientation, it is right to left. Thus the chosen boundary directions from the two faces must be opposite after the edge identification.
The chosen boundary direction on \(e\) is \(\sigma_F d_e\). Its image under the identification is \(\sigma_F\varepsilon_e d_{e'}\), while the chosen boundary direction on \(e'\) is \(\sigma_G d_{e'}\). Being opposite means \(\sigma_F\varepsilon_e d_{e'}=-\sigma_G d_{e'}\). Since \(d_{e'}\) is a directed edge, this is equivalent to \(\sigma_G=-\varepsilon_e\sigma_F\). Conversely, if this equation holds, the two induced directions are opposite, so the oriented half-disk pieces fit together across the edge. This proves both directions. \(\square\)
This is a local rule: it checks a neighborhood of an interior point of an identified edge. For a polygonal decomposition of a surface, the corresponding face orientations give an orientation of the whole surface when they satisfy the rule on every paired edge, provided each edge has two incident face sides and each vertex has a disk neighborhood. At an edge interior, the theorem gives the required disk orientation. At a vertex, the disk neighborhood consists of sectors arranged cyclically; the edge conditions make the orientations on consecutive sectors agree, including where the cycle closes. These conditions are the local reason the edge-sign calculation controls orientation in a polygonal surface.
Reading the Sign from Endpoints
To avoid guessing the sign from a drawing, write down the directed sides and the endpoint map. If the initial endpoint of one reference direction maps to the terminal endpoint of the other reference direction, then the sign is \(-1\). If initial maps to initial, it is \(+1\). Then substitute that sign into the compatibility equation.
For each paired side, use the boundary direction given by the reference orientation of its face.
Track where the initial endpoint of the first directed side goes.
Set \(\varepsilon_e=+1\) or \(-1\) from the endpoint map, then require \(\sigma_G=-\varepsilon_e\sigma_F\).
Worked Example: Two Triangular Faces Along One Edge
Let \(F\) and \(G\) be triangular faces, each given its counterclockwise reference orientation. Suppose the paired sides have reference directions \(A_0\to A_1\) in \(F\) and \(B_0\to B_1\) in \(G\), and the identification sends \(A_0\) to \(B_0\) and \(A_1\) to \(B_1\).
The initial endpoint \(A_0\) maps to the initial endpoint \(B_0\), so \(\varepsilon_e=+1\). The compatibility equation is \(\sigma_G=-(+1)\sigma_F=-\sigma_F\). Choose \(\sigma_F=+1\); then \(\sigma_G=-1\). On the side of \(F\), the chosen boundary direction is \(A_0\to A_1\). On the side of \(G\), reversing the reference orientation changes \(B_0\to B_1\) to \(B_1\to B_0\). Under the identification, \(A_0\to A_1\) maps to \(B_0\to B_1\), which is opposite to \(B_1\to B_0\), as required.
Check: The sign equation gives \(\sigma_G=-\sigma_F=-1\). The two chosen boundary directions are opposite after identification, so the local orientation extends across the edge.
A Paired Square Giving the Torus
Let a square have vertices \(P,Q,R,S\) in counterclockwise order. Its counterclockwise boundary directions are \(P\to Q\), \(Q\to R\), \(R\to S\), and \(S\to P\). Pair \(PQ\) with \(RS\) by sending \(P\) to \(S\) and \(Q\) to \(R\). Pair \(QR\) with \(SP\) by sending \(Q\) to \(P\) and \(R\) to \(S\). This is the usual paired-square model of the torus.
Worked Example: Checking the Torus Pairings
For the first pair, the listed directions on \(PQ\) and \(RS\) are \(P\to Q\) and \(R\to S\). The map sends \(P\), the initial endpoint of \(P\to Q\), to \(S\), the terminal endpoint of \(R\to S\). Therefore \(\varepsilon_e=-1\). There is one face, so \(F=G\) and the compatibility rule becomes \(\sigma_F=-(-1)\sigma_F=\sigma_F\). This pairing allows either choice of orientation on the face.
For the second pair, the listed directions on \(QR\) and \(SP\) are \(Q\to R\) and \(S\to P\). The map sends \(Q\), the initial endpoint of \(Q\to R\), to \(P\), the terminal endpoint of \(S\to P\). Hence this edge sign is also \(-1\), and its condition is again \(\sigma_F=-(-1)\sigma_F=\sigma_F\). Both paired edges therefore satisfy the rule when the square is given its counterclockwise orientation.
Check: On \(PQ\), the direction \(P\to Q\) maps to \(S\to R\), opposite to \(R\to S\). On \(QR\), \(Q\to R\) maps to \(P\to S\), opposite to \(S\to P\). Thus both pairs have the required opposite boundary directions.
A Paired Square Giving the Klein Bottle
Keep the same square and the same pairing of \(PQ\) with \(RS\), but pair \(QR\) with \(SP\) using the endpoint map \(Q\mapsto S\), \(R\mapsto P\). This is the paired-square model of the Klein bottle. One pairing has sign \(-1\), while the other has sign \(+1\).
Worked Example: The Sign Contradiction in the Klein Bottle
For \(PQ\) and \(RS\), the directions are \(P\to Q\) and \(R\to S\), and \(P\mapsto S\). The initial endpoint of the first direction maps to the terminal endpoint of the second, so \(\varepsilon_e=-1\). Since these sides belong to the same face, compatibility requires \(\sigma_F=-(-1)\sigma_F=\sigma_F\), which imposes no restriction.
For \(QR\) and \(SP\), the reference directions are \(Q\to R\) and \(S\to P\). The map sends \(Q\), the initial endpoint of \(Q\to R\), to \(S\), the initial endpoint of \(S\to P\). Thus \(\varepsilon_e=+1\). The compatibility equation for this pair requires \(\sigma_F=-(+1)\sigma_F=-\sigma_F\). But \(\sigma_F\) must be either \(+1\) or \(-1\), and neither value equals its own negative. No orientation choice for the square can satisfy this condition.
Check: The first pairing sends \(P\to Q\) to \(S\to R\), opposite to \(R\to S\). The second sends \(Q\to R\) to \(S\to P\), the same as the listed direction on \(SP\), rather than opposite to it. This is exactly the conflict recorded by the \(+1\) sign. Hence the face orientation cannot extend over the Klein-bottle quotient.
A Common Sign Mistake
Common pitfall. When pairing \(PQ\) with \(RS\) in the square examples, it is tempting to list the direction on \(RS\) as \(S\to R\), because that is the direction to which \(P\to Q\) is carried by the map \(P\mapsto S,\ Q\mapsto R\). But \(S\to R\) is not the reference boundary direction on that side: the square’s counterclockwise boundary direction is \(R\to S\).
If we incorrectly use \(S\to R\) as the reference direction, we compare \(P\to Q\), mapped to \(S\to R\), with a falsely listed direction \(S\to R\), and obtain the wrong sign \(\varepsilon_e=+1\). The resulting equation would incorrectly demand \(\sigma_G=-\sigma_F\). With the actual reference direction \(R\to S\), the map sends \(P\to Q\) to its opposite, so \(\varepsilon_e=-1\), and the correct condition is \(\sigma_G=\sigma_F\). The correction is to list each side’s boundary direction from the face orientation before examining the endpoint map.
| Endpoint behavior | Edge sign | Required face signs |
|---|---|---|
| Initial endpoint maps to initial endpoint | \(\varepsilon_e=+1\) | \(\sigma_G=-\sigma_F\) |
| Initial endpoint maps to terminal endpoint | \(\varepsilon_e=-1\) | \(\sigma_G=\sigma_F\) |
The edge sign is defined relative to the reference boundary directions, not by whether the drawn identification looks like a “twist.” Once those directions and the endpoint map are specified, the equation \(\sigma_G=-\varepsilon_e\sigma_F\) gives the compatibility test. The torus square satisfies its edge conditions; the Klein bottle square has a condition that would require a sign to equal its negative.
Check Your Understanding
For each question, first write down the reference boundary directions and the endpoint map.
- Two distinct faces have paired reference directions \(A\to B\) and \(C\to D\), with \(A\mapsto D\) and \(B\mapsto C\). Find \(\varepsilon_e\) and the relation between their face signs.
- For two paired sides of the same face, suppose the edge sign is \(+1\). Can that pairing satisfy the compatibility rule? Explain using the equation.
- In the square, pair \(PQ\) with \(RS\) using \(P\mapsto R,\ Q\mapsto S\). Compute the edge sign using the counterclockwise boundary directions.
- Why is the sign determined by the directions induced from the face orientations, rather than by the endpoint map alone without reference directions?
- For a polygonal surface with disk neighborhoods at vertices, what local requirement must hold at each paired edge for the face orientations to fit together?