* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Chapter 7
History of geometry wikipedia , lookup
Steinitz's theorem wikipedia , lookup
Riemannian connection on a surface wikipedia , lookup
Dessin d'enfant wikipedia , lookup
Cartesian coordinate system wikipedia , lookup
Integer triangle wikipedia , lookup
Euler angles wikipedia , lookup
Multilateration wikipedia , lookup
History of trigonometry wikipedia , lookup
Differential geometry of surfaces wikipedia , lookup
Rational trigonometry wikipedia , lookup
Trigonometric functions wikipedia , lookup
Line (geometry) wikipedia , lookup
Pythagorean theorem wikipedia , lookup
CHAPTER 7 Polyhedral curvature As I mentioned earlier, the differences between spherical/elliptic geometry, Euclidean geometry, and hyperbolic geometry can be characterized in terms of curvature. Having done this, we will have a much more general and complete collection of geometries. To get you thinking, note that the Euclidean plane is flat, the sphere is round, and the hyperbolic plane is wrinkly (see [Bonola, p 133]). Things at least look that way through our Euclidean eyes. As a preview, I can show you how well curvature unifies geometry, and believe me, it unifies amazingly well. Recall that the angle sum of a triangle was different in spherical, Euclidean, and hyperbolic geometry. For a triangle with angles α, β, and γ and area A, we saw that in spherical geometry (206) α + β + γ = π + A, in Euclidean geometry (207) α + β + γ = π, and in hyperbolic geometry (208) α + β + γ = π − A. Eventually, we will define something called Gauss curvature, and it will turn out that the Gauss curvature for the unit sphere is K = 1, for the Euclidean plane is K = 0, and for the hyperbolic plane is K = −1. With this, we can combine these three formulas into one with (209) α + β + γ = π + KA. We’re obviously looking at something quite fantastic. In fact, this last formula is starting to reveal something, that was not obvious before. The deviation of the angle sum of a triangle from π is not so much a function of area, but of the amount of curvature inside the triangle. 122 1. WHAT DID DESCARTES SEE? 123 The Gauss curvature can be thought of as a rate of curvature. It is a rate of another kind of curvature, which I’ll call Cartesian curvature (I’m making up this name), and this is what we’ll focus on first. As an example, the unit sphere has Gauss curvature K = 1. As a rate, this can be interpreted as one unit of Cartesian curvature per unit area. Since the unit sphere has area 4π, that’s how much Cartesian curvature it has. If we use C for Cartesian curvature, then the Gauss and Cartesian curvatures are related by (210) K= C , A or (211) C = KA. Our angle sum formula, could then be written (212) α + β + γ = π + C, where C is the amount of Cartesian curvature inside the triangle. 1. What did Descartes see? In general, I will use the word curvature to describe how something is not flat. Figure 1. You can fold paper without stretching, crinkling, or tearing it. A piece of paper is flat. It has a Euclidean geometry1. You can fold it (as in Figure 1) without much trouble, but you really can’t smoothly bend it into the shape of a sphere. Gauss curvature and Cartesian curvature (I haven’t defined either one) both measure the inability to flatten a surface. 1That is, flat paper has the same geometry as the Euclidean plane 1. WHAT DID DESCARTES SEE? Figure 2. Cut this out, and tape it together to make a cube. 124 1. WHAT DID DESCARTES SEE? 125 You can take a pair of scissors and some tape, cut out Figure 2, and tape it together to build a cube. After you start taping, it is no longer possible to flatten it back out without tearing or pulling the tape off. A cube, therefore, must have some curvature, and it’s not in the edges (the folds). Therefore, it must all be in the corners. Descartes found a simple, and amazingly useful, way of measuring this curvature. At each corner of the cube, the corners of three square disks come together. Each angle of the square measures π2 radians, and the three angles around each vertex sum to 3π 2 radians. Now, the corners of a cube will not lay flat, because their geometry is not Euclidean. In the Euclidean plane, where four squares coming together would lay flat, there is always 2π radians around every point. The corners of a cube don’t lay flat, because they are missing π2 radians. Descartes, therefore, called this an angle defect of π2 radians. The term angle defect, I think, is the common term used today, but actually, Descartes called it the angulum externum, which might be translated from Latin as the external angle (I’m thinking that he didn’t write in English very often) [Federico, p 44]. Let’s do some math. We have a cube, and we know the angle defect at each corner. It’s π2 . There are eight corners, so the total of the angle defects is (213) 8· π = 4π. 2 We’ve seen this number before. It’s the area of the unit sphere, and more importantly, it’s the total amount of Cartesian curvature on the sphere. Let’s do that again for something else. A regular tetrahedron (just the surface as always) is also called a triangular-based pyramid. It is made up of four equilateral triangles. How many vertices does it have? 4. What is the angle defect at each vertex? 2π − 3 · π3 = π. What is the total angle defect? 4 · π = 4π Is a pattern developing? Conjecture 3. For any polyhedron that is a topological sphere, the total angle defect will always be 4π. If this conjecture is true, then it seems likely that the total Cartesian curvature we talked about before is the same as the total angle defect. 2. PROOF OF CONJECTURE: TOTAL ANGLE DEFECT FOR A SPHERE IS 4π 126 2. Proof of Conjecture: Total angle defect for a sphere is 4π Suppose we have a polyhedron that is a topological sphere. In the context of angle defects, we will always assume that the faces are flat. Furthermore, we will assume that all the faces are triangular disks. Since the faces are all flat triangular disks, we know that the angle sum is π radians. Since this is a sphere, we also know that χ = f − e + v = 2. The number of faces is f . If we add up all the angles from all the faces, then we get πf . The number of corners is the same as the number of vertices v. If every vertex were flat, there would be angles totalling 2π surrounding each vertex, and all these angles would add up to 2πv. If there is an angle defect at any corner, then the angles that are there subtracted from 2π would be the angle defect. We can’t compute these individually, because we don’t have enough information for that. We can, however, compute all the angle defects at once. All the 2π’s minus all the angles of all the faces is the total angle defect. That is, the total angle defect is (214) total angle defect = 2πv − πf = (2v − f )π. This is a sphere, so every edge must be shared by exactly two faces. Since every face has three edges, and each edge is used twice, we know that (215) e= 3f . 2 Now, using the fact that χ = 2, we know that (216) 2 = f − e + v, and so substituting equation (215) into this, we get (217) 2=f− 3f f +v =v− . 2 2 Now, multiply this by 2. We get (218) 4 = 2v − f. Substitute this into equation (214) to get (219) total angle defect = (2v − f )π = 4π. How could that work so nicely? 3. WHAT DESCARTES WROTE 127 Now, we’ve got the theorem for a triangulation. For a decomposition into polygons, we simply add diagonals to get a triangulation. The diagonals simply subdivide the existing angles, so the sum of the angles stays the same, and we get the same result. This completes the proof, so we now have the following. Descartes’ Theorem. For a polyhedron that is a topological sphere, the total angle defect is always 4π radians. 3. What Descartes wrote The stuff we typically see in mathematics has been rewritten and redefined a number of times, and we rarely get to see where they came from. Here’s a look. Federico says that Descartes wrote the original manuscript sometime in the early 17th century. After he died, his papers were collected, accidentally dropped in a river, and hung to dry. They were kept for awhile, and Leibniz apparently got a chance to see them, and he copied some stuff out of it. Leibniz’ copy is shown in the book. It looks mostly like a bunch of scribbling. The first sentence is recognizable [Federico, p 11], “Progymnasmata de Solidorum Elementis excerpta ex Manuscripto Cartesii.” The translation given by Federico is [Federico, p 43], “Preliminary Exercises on the Elements of Solids Taken from a Manuscript of Descartes.” In the middle of the fourth line, begins the statement of Descartes’ theorem [Federico, p 44]: Sicut in figura plana omnes anguli externi, simul sumti, aequales sunt quatuor rectis: ita in corpore solido omnes anguli solidi externi, simul sumti, aequales sunt octo solidis rectis. Per angulum externum intelligo curvaturam seu inclinationem planorum ad invicem, quam metiri oportet ex angulis planis angulum solidum comprehendentibus. Nam illa pars qua aggregatum ex omnibus angulis planis unum angulum solidum facientibus, minus est quam quatuor anguli recti planum facientes, designat angulum externum solidum. I can’t read Latin (I believe that’s what this is), but I can match up a lot of the words from the translation given by Federico [Federico, p 44]. As in a plane figure all the exterior angles, taken together, equal four right angles, so in a solid body all the exterior solid angles, taken together, equal eight solid right angles. By exterior angle I mean the 4. ANGLE SUMS OF TRIANGLES ON CONES 128 mutual bending or inclination of the planes, which is to be measured with the help of the plane angles which comprise the solid angle. For the part by which the sum of all the plane angles forming a solid angle is less than the four right angles which form a plane, designates the exterior solid angle. Our very important number, 4π, is there in Liebniz’ handwriting, “aequales sunt octo solidus rectis.” That’s “equal eight solid right angles” 4. Angle sums of triangles on cones I plan on coming back to extensions of Descartes’ theorem, but first, I want to talk about how it sheds light on the geometries of Gauss/Bolyai/Lobachevski, Euclid, and Riemann (“The Big Three geometries”). We could say that the point of Hilbert’s axiom system is to precisely characterize the properties of a straight line. The existence of non-Euclidean geometries illustrate the fact that the concept of a straight line isn’t as simple as we might have thought. The differences between the Euclidean and non-Euclidean geometries lie well outside of our practical experiences. One property that lines in all these geometries have is that they minimize distance. In particular, given two nearby points, the curve joining the points with the shortest possible length is a line segment. Again, this is true in all of the Big Three. On a curved surface (and flat ones as well) the term geodesic (“gee uh DEZ ick”) is used for curves that minimize distance. The lines of spherical, Euclidean, and hyperbolic geometry are all geodesics, and we’ll even think of the geodesics of a surface as the lines for the geometry belonging to that surface. In my mind, there is no mathematical distinction between the words line and geodesic. Take out a piece of paper, and mark two points A and B on it, as in Figure 3. With a ruler, draw a straight line through the two points. We know the segment between A and B is the shortest curve joining A and B, because the geometry of paper is Euclidean, and the shortest distance occurs along a straight line. Now roll the paper into a cylinder as in Figure 3. Of all the curves between A and B that lie on the cylinder, the shortest distance must be along the straight line you drew, which is now curved with the surface, because we haven’t changed any of the small distances measured along the paper. 4. ANGLE SUMS OF TRIANGLES ON CONES 129 B B A A Figure 3. We can roll up a piece of paper to make a cylinder. The line on the flat piece of paper becomes a curve on the cylinder that minimizes distances, so the line is a geodesic. All geodesics on the cylinder can be determined this way. Most of the geodesics on the cylinder turn out to be helixes (I guess you’re supposed to say “helices”). The others can be seen as degenerate helixes. Vertical lines are geodesics, and horizontal circles are also geodesics. Since the geodesics are the lines for the geometry of the cylinder, the things we’ll call triangles will have three sides that are segments from geodesics (i.e., geodesic segments). For a relatively small triangle, we know that when flattened, the angle sum is 180◦ or π radians. The measures of the angles don’t change when we roll the paper into a cylinder, so the angle sum will still be 180◦ . The local geometry of the cylinder is Euclidean. The global geometry is different, of course. We have lines that are circles, for example. We also have that pairs of points can have more than one line through them. Can you see any examples? This is an example of taking one geometry, and creating a new one with different global structure by changing its topology. Since it is easy for us to work in Euclidean geometry, I think it’s important that we can make a rich variety of geometries out of paper (at least theoretically). Basic Principle 5. For any surface we can make out of paper, it is easy to find geodesics. 5. THE CARTESIAN CURVATURE OF CONES 130 5. The Cartesian curvature of cones Most of what I’ve come to understand about geometry has come from playing with cones. They have a Euclidean geometry almost everywhere, so they are easy to work with, but there is curvature in the vertex, so we can see how curvature affects the behavior of geodesics (i.e., you can see how the geometry works). Because of this, I believe that an introduction to geometry should start with cones. Figure 4. Plans for a cone and two geodesics on the cone. 5.1. Exercises. –1– We can make a cone by removing a wedge from a piece of paper, and taping the edges together. Cut out Figure 4, and tape the edges together to make a cone. (There are extra copies of some of these figures at the end of this print out.) –2– The piece of paper you started with was flat, so there were 2π radians around the vertex. How much did you remove? What was the angle defect? –3– I’ve drawn a pair of geodesics on your cone. Do these geodesics behave more like the lines of the sphere, the Euclidean plane, or the hyperbolic plane? 5. THE CARTESIAN CURVATURE OF CONES 131 We saw earlier that it looked like angle defects and Cartesian curvature should be the same. Let’s go ahead and make that explicit. Definition 1. If we remove a θ-radian wedge to form a cone, then the Cartesian curvature at the vertex of the cone is defined to be θ. At non-vertex points, the Cartesian curvature is zero. On the sphere, lines seem to turn towards each other. For example, there were 2-gons (bounding lunes) on the sphere. If you start at one angle of a 2-gon and move away, the two sides move away from each other. They eventually start getting closer together, however, and meet again at the other angle. This could be described as the lines turning towards each other, one to the left and the other to the right. Consider one line on the sphere (i.e., a great circle). Does it really make sense to say it turns to the left or right? This cone provides another interpretation of what happens on the sphere. The lines don’t turn towards each other so much as the space between them gets smaller. On the cone, we actually went in and removed a chunk of the space (the wedge). On the hyperbolic plane, lines seem to move away from each other, and this made it much easier for lines to be parallel. We can cause geodesic behavior like this on a cone by adding a wedge. Figure 5. A cone with an added wedge. Cut along the wavy lines to separate the two pieces. Cut along the dotted lines. Tape the extra wedge along the dotted lines. Cut out Figure 5, including the extra wedge. Cut along the slit and tape the wedge in. You should now have a saddle-shaped cone with an extra 45◦ around the 5. THE CARTESIAN CURVATURE OF CONES 132 vertex. If removing a wedge gives you a positive angle deficit, then it makes sense to think of adding a wedge as creating a negative angle defect. Definition 2. If we add a θ-radian wedge to form a (saddle-shaped) cone, the Cartesian curvature is defined to be −θ. Basic Principle 6. We will associate positive Cartesian curvatures with elliptic geometries (e.g., spherical geometry) and negative Cartesian curvatures with hyperbolic geometry. In the two examples we’ve seen, the cuts are straight lines. This isn’t necessary, but it’s usually the easiest way. The cut, therefore, runs along a geodesic, and except at the vertex, the geometry is Euclidean everywhere along the cut. If another geodesic crosses the cut, we’ll see four angles, and the vertical angles must have to be congruent. When we make our models, this is one of the rules we’ll follow. Basic Principle 7. If we cut and tape along straight lines, any geodesic crossing the cut will form four angles with the cut. We need to make sure that vertical angles are congruent. Figure 6. Some geodesics on a cone. 6. ANGLE SUMS OF TRIANGLES WITH POSITIVE CURVATURE 133 In Figure 6, note that I have four lines meeting at right angles. Two of the lines hit the cut at right angles, and they will come together when you cut and tape. These will come together to form a single geodesic. We have, therefore, three geodesics coming together to form a triangle with three right angles. The angle sum is 3π , which is 2 π more than what we would expect in the Euclidean plane. We might suspect that 2 there must be Cartesian curvature totalling π2 inside the triangle, and we would, of course, be correct. α ϕ ϕ θ γ β Figure 7. This is a triangle with positive Cartesian curvature θ inside. 6. Angle sums of triangles with positive curvature We’ve seen that the presence of curvature in the interior of a triangle affects the triangle’s angle sum. Furthermore, this happens in a nice way. In this section, I’m going to have you derive formulas for the angle sum of a triangle on a cone. In Figure 7, I’ve drawn a fairly typical triangle around the vertex. I want to relate the angle sum of the triangle and the Cartesian curvature θ. Let’s just go straight to the computations. In Figure 8, I’ve subdivided the figure into triangles. 6. ANGLE SUMS OF TRIANGLES WITH POSITIVE CURVATURE α α1 2 134 ϕ 3 2 1 4 ϕ θ 5 7 6 γ2 γ1 β1 β2 Figure 8. In this subdivision, all the triangles are Euclidean. 6.1. Exercises. –1– All of the triangles that look like triangles in Figure 8 are Euclidean triangles. In radians, what is the following sum? (220) α1 + α2 + β1 + β2 + γ1 + γ2 + ϕ + ∠1 + ∠2 + ∠3 + ∠4 + ∠5 + ∠6 + ∠7 + θ. –2– –3– –4– –5– –6– What is ∠4 + ∠5 + ∠6 + ∠7 + θ? Therefore, what is α1 + α2 + β1 + β2 + γ1 + γ2 + ϕ + ∠1 + ∠2 + ∠3? What is ϕ + ∠3? What is ∠1 + ∠2 in terms of θ? You should now have enough information to find α + β + γ in terms of θ. What do you get? You have proven the following theorem. Theorem 14. For a triangle around the vertex of a cone with positive Cartesian curvature θ, the angle sum is given by the formula α + β + γ = π + θ. 8. ANGLE SUM OF A TRIANGLE α 135 ψ ψ 3 4 ϕ 1 2 ϕ γ θ 6 5 β Figure 9. The figure corresponding to Figure 8 7. Angle sum of a triangle with negative curvature The picture looks a little different with negative Cartesian curvature, but we get a similar result. 7.1. Exercises. –1– Look at Figure 9 Following a strategy similar to the one used in the previous section, you should be able to find the angle sum α + β + γ in terms of θ. Do the math. Theorem 15. For a triangle around the vertex of a cone with negative Cartesian curvature −θ, the angle sum of the triangle is α + β + γ = π − θ. 8. Angle Sum of a Triangle The theorems of the last two sections say the same thing. We therefore have the following. 8. ANGLE SUM OF A TRIANGLE 136 Theorem 16. For a triangle around the vertex of a cone with Cartesian curvature θ, the angle sum of the triangle is α + β + γ = π + θ. The number θ may be positive or negative. 8.1. Exercises. –1– Is this last theorem true for a Cartesian curvature of θ = 0? What kind of surface is this? 9. THE ANGLE SUM OF A POLYGON 137 Figure 10. Copy of Figure 4. Figure 11. Copy of Figure 5. Cut along the wavy lines. Cut along the dotted lines. Tape the extra wedge along the dotted lines. 9. The angle sum of a polygon The principle theme lately is that the amount of curvature inside a triangle determines that triangle’s angle sum. This holds equally for the total amount of Gauss curvature (which we haven’t defined), Cartesian curvature, and angle defects. In fact, I claim that these are all the same thing. Another theme is that the angle sum of 9. THE ANGLE SUM OF A POLYGON 138 Figure 12. Copy of Figure 6. a triangle is intimately tied to the kind of geometry we are considering (hyperbolic, Euclidean, or elliptic). Before we continue on, I would like to talk more about the angle sum of a triangle. There is no doubt that a triangle’s angle sum captures something fundamental about geometry, it’s not quite the right thing, however. Let me direct your attention back to Descartes’ statement of Descartes’ theorem. Descartes said, [Federico, p 44] As in a plane figure all the exterior angles, taken together, equal four right angles, so in a solid body all the exterior solid angles, taken together, equal eight solid right angles. “Eight solid right angles” is 4π steradians in modern terminology. A steradian is a unit of measure for solid angles such as the corner of a cube or the vertex of a cone. I’ll talk about them later. The second part of this sentence is Descartes’ theorem. Descartes also refers to another theorem, Theorem 17. In a plane figure, all the exterior angles, taken together, equal four right angles (i.e., 2π radians). This is a lower-dimensional version of Descartes’ theorem, and it suggests a better object of focus than the angle sum of a triangle. That’s what we’ll do now. 9. THE ANGLE SUM OF A POLYGON 139 Bad!! Figure 13. This pentagonal disk is not convex. There is at least one pair of points, where the segment joining them lies partly outside of the disk. In Euclidean geometry, the angle sum of a triangle is π radians. A polygonal disk is convex, if any two points in the disk can be connected by a segment that stays inside the disk. The pentagonal disk shown in Figure 13 is not convex, and a pair of points that cannot be connected by a segment inside the pentagon is shown. Figure 14. A pentagonal disk subdivided into three triangular disks. It is relatively easy to find the angle sum of a convex polygon. For example, the convex pentagon shown in Figure 14 has been subdivided into three triangles. Now, the sum of the nine angles of the three triangles must be exactly the same as the sum of the five angles of the pentagon. Right? The angle sum of the pentagon, therefore, must be 3π. We can play this same game with any convex polygon. 9.1. Exercises. –1– What is the angle sum of a convex n-gon? –2– What is the angle sum of the non-convex pentagon in Figure 13? 10. TRIANGULATING POLYGONAL DISKS 140 –3– Is convexity a topological invariant? That is, does a topological deformation always preserve convexity? You can draw in diagonals for the pentagon in Figure 13 to subdivide it into three triangles, but it is far from clear that you can do this with just any non-convex polygon. I believe that it’s possible in general, but I’m not absolutely sure. If we define a diagonal to be a segment from one vertex to another non-adjacent vertex that lies entirely within the polygonal disk, could you prove that any polygonal disk has a diagonal? Again, I think this is true, but I can’t think of a way to prove it. I also don’t care about it enough to look it up. What I do care about, however, is finding a formula for the angle sum of any polygonal disk (convex or not), and I also want you to wrestle with triangulations at least once. Concepts in both geometry and topology have seemed to me more and more to distill down to questions about triangulations. 10. Triangulating polygonal disks As I’ve just said, I believe that you can always triangulate a polygonal disk with diagonals. If you could prove that, that would be great. The argument we’re going to pursue, however, is sufficient for our needs, and I think the argument itself is more general. Theorem 18. Polygonal disks can always be triangulated. In particular, we can subdivide any polygonal disk into a finite number of triangular disks with straight edges. We will come up with a scheme that will produce a triangulation. This triangulation will not be the most efficient possible, but it will make the proof easy. The basic idea is contained in Figures 15 and 16. Given a polygonal disk, we lay it out in the xy-plane, and impose a system of grid lines. This is shown in Figure 15. At each vertex, we add a horizontal and a vertical grid line (unless they’re already there) to ensure that each vertex is at an intersection of grid lines. This is shown in Figure 16. In Figure 16, the polygonal disk has been subdivided into smaller polygonal disks by the new grid lines. Some of these disks are rectangular. They’re all triangles or convex quadrilaterals. For each quadrialateral, we only have to add a diagonal to subdivide it into a triangle. 10. TRIANGULATING POLYGONAL DISKS 141 Figure 15. A grid system is imposed on the polygonal disk. Figure 16. Additional grid lines are added to ensure that vertices of the polygonal disk correspond to intersections of grid lines. C B A Figure 17. Dotted lines indicate the shortest distance between the heavy line segments. I will want to use the notion of the shortest distance between two line segments. In Figure 17, we have segments A, B, and C. The shortest distance between segments A and B occurs between the upper endpoints on each. The shortest distance between 10. TRIANGULATING POLYGONAL DISKS 142 segments B and C occurs between an interior point of B and an endpoint of C. Note that except for parallel line segments, the shortest distance between two segments cannot be between two interior points. OK. Let’s do the proof in general. Suppose we have a polygonal disk lying in the xy-plane. The shortest distance between adjacent edges is zero. The shortest distance between non-adjacent edges must be strictly positive. Suppose there are n edges (and therefore n vertices). Pick an edge. There are n − 3 edges that are not adjacent to this edge, so there are n − 3 pairs of non-adjacent edges including this one. We can do this n times, but we would be counting every pair twice, so there are (221) n · (n − 3) 2 possible pairs of non-adjacent edges total. The exact number isn’t important, but there surely are only finitely many pairs. Therefore, we have a finite set of positive minimum distances between pairs of non-adjacent edges. We can say, therefore, that there is a pair of non-adjacent edges a positive distance apart, and no other pair is closer together. Let’s say that the shortest distance between the closest pair of non-adjacent edges is , and is positive. Is it possible for an by square to contain parts of two non-adjacent edges? √ √ Well, yes. The distance between opposite corners is 2 > . Dividing by 2 makes the square small enough. Therefore, if we impose a system of grid lines with spacings less than √2 , then we know for sure that each grid square will contain a part of one edge, parts of two adjacent edges, or no part of any edge. If we add horizontal and vertical grid lines through each of the vertices, then we will have grid squares and grid rectangles, but we will know that vertices of the polygonal disk are only at the corners. OK. We now have the polygonal disk subdivided into small polygonal disks. If we can triangulate each of the small polygonal disks, then we have a triangulation of the original polygonal disk. 11. THE ANGLE SUM OF AN N -GON 143 All of the grid rectangles that lie completely outside of the polygonal disk are not needed. We don’t care about them. Some of the grid rectangles lie completely inside the polygonal disk. We can triangulate these rectangles with a diagonal line. All of the other rectangles have an edge or two running through them. We need to address each possible case. 10.1. Exercises. –1– If a grid rectangle has a single edge running through it, what shape can the resulting small polygonal disk take? (I think there are six cases and three kinds of shape (i.e., triangular, quadrangular, etc.)) Draw a picture of each case. –2– If a grid rectangle has two edges running through it, it must have a vertex at one of its corners. What shapes can we get for the small polygonal disks this way? Draw the cases. (I think there are six cases here, as well. It’s possible that two small disks are generated.) This completes the proof of the theorem. Topologists speak of something called the Hauptvermutung. This is a German word, and many German words are compound words. Haupt means something like primary and vermutung can mean supposition. The Hauptvermutung, therefore, is the big conjecture asking if “everything” has a triangulation. In a few cases, it isn’t true, but it seems to be true most of the time. You’ve just proven the Hauptvermutung for flat polygonal disks. 11. The angle sum of an n-gon You showed earlier that the angle sum of a convex n-gon is π(n − 2). In one fairly general case, you saw that a non-convex pentagon satisfied this same formula. This alone should give us great confidence that this formula holds for all n-gons, so let’s prove it. Suppose we have an n-gonal disk and a triangulation of it. This might look something like Figure 18. Since this is a disk, we know that the Euler characteristic is χ = f − e + v = 1. It will be convenient to separate the vertices and edges in the triangulation into different types. We are most interested in the original vertices of the n-gon. I’m going to call these, the vertices like A or D, the corner vertices. There 11. THE ANGLE SUM OF AN N -GON 144 A C B D Figure 18. This is a triangulation of a polygonal disk. are also new vertices along the outer edges, like C, which I’ll call edge vertices. The new vertices inside the n-gonal disk, like B, will be called interior vertices. Let vi be the number of interior vertices, ve the number of edge vertices, and vc the number of corner vertices. We have that (222) v = vi + vb + vc. We actually know what vc is, since these are the corners of our n-gon. We know that vc = n. We’ll also differentiate between two types of edges. The edges along the boundary of the n-gonal disk will be called boundary edges, and the edges lying in the interior will be called interior edges. There are eb boundary edges and ei interior edges. We have that (223) e = ei + eb . Along the boundary of the n-gonal disk, there are ve + vc = ve + n vertices and eb edges, and (224) ve + n = eb , because the boundary is just a simple closed curve. Since the Euler characteristic is χ = f − e + v = 1 in terms of f , vi , vb, vc , ei , and eb , we have (225) 1 = χ = f − (ei + eb ) + (vc + ve + vi ) = f − (ei + eb) + (n + ve + vi). 11. THE ANGLE SUM OF AN N -GON 145 We’re after the angle sum of the n-gon. At each of the corner vertices, like A and D, the interior angles are subdivided into smaller angles. In Figure 18, there are three at A and five at D. If we add up the smaller angles at each of the corner vertices, then we will have the angle sum of the n-gon. This is the ultimate goal. Let’s call the angle sum Σ. The sum of all the angles of all the triangles in the triangulation is πf . We have angles surrounding each of the interior vertices, like B in Figure 18, totaling 2π. Each of the new vertices on the edges have angles around them, but not all the way around. The new vertices on the edges are interior to the edge, so these vertices only have π radians around them. We can now subtract all of the new angles from the all the angles in the triangulation to get Σ. (226) Σ = πf − 2πvi − πve. One last piece of information will make this into an algebra problem. We have f triangles and each of these triangles has three edges. We must account for 3f edges. Some of these edges are used twice and some only once. Since each of the interior edges is shared by two triangles, they are counted twice in 3f . The boundary edges are only used once, so (227) 2ei + eb = 3f. Now, it’s just a matter of mashing these equations together. Getting rid of the v’s in equation (226) works. Note that there is a ve + n in equation . According to equation (224), this is equal to eb, and equation (11) becomes (228) 1 = f − ei − eb + eb + vi . Solving this for vi gives (229) vi = 1 − f + ei . 12. EXTERIOR ANGLE SUMS 146 Now, we can substitute this and equation (224) into (226) to get Σ = πf − 2πvi − πve = π ( f − 2vi − ve ) (230) = π ( f − 2(1 − f + ei ) − (eb − n) ) = π ( 3f − 2ei − eb + n − 2 ) = π( n−2 ), the last step coming from equation (227). This establishes the following theorem. Theorem 19. (Euclidean geometry) The angle sum of an n-gon is π(n − 2). The n-gon need not be convex, but it must bound a disk. Some assumptions that I haven’t stated explicitly are as follows. An n-gon is a simple closed curve that consists of n straight line segments lying in a (Euclidean) plane. It is a topological fact that any simple closed curve lying in the plane will bound a disk. In particular, there are closed polygonal paths that intersects themselves, like a figure-eight curve. We will not call such a thing an n-gon or a polygon. 12. Exterior angle sums Earlier, I mentioned Descartes’ statement, “As in a plane figure all the exterior angles, taken together, equal four right angles, . . .” With Theorem 19, it is easy to prove this fact. First, let’s talk about exterior angles. For cones, the angle defect (or Cartesian curvature) is 2π − θ, where θ is the radian measure of the wedge removed. Descartes called this the anguli externi or external angle. The external angle is a measure of how much the cone is not flat. If we go down one dimension, a vertex of a polygon is a place where the polygon is not straight. Right? The angles are bent, and the edges are straight. If an angle of a polygon measured 2π radians, then we would know that it was not a straight 3 angle, because a straight angle measures π radians. In this case, we would say that the angle defect is π3 radians, because that’s how far the angle is from being straight. Descartes called this the external angle. 12. EXTERIOR ANGLE SUMS 147 Definition 3. If an angle of a polygon measures θ radians, then the external angle, or the angle defect, is π − θ. Note that if θ is greater than π, then the angle defect is negative. We have a nice formula for the angle sum of an n-gon. The exterior angle sum has even a nicer formula giving some indication that it is a more fundamental quantity. These last exercises will establish this fact. 12.1. Exercises. –1– You proved in the last section that the angle sum of an n-gon is π(n − 2). We would find the angle defect of each of the n angles by subtracting the measure of the angle from π. What is the sum of the angle defects? (Hint: You could grind this out. The angles measure θ1 , θ2 , . . ., θn . The angle sum P P is θi and the angle defect sum is (π − θi ).) –2– State a good assignment ending theorem. In this section, we showed that in Euclidean geometry the angle sum of any polygon is π(n − 2), where n is the number of sides. Even better, you showed that the external angle sum of any polygon is 2π no matter how many sides it has. The internal and external angle sums of polygons are generalizations of the angle sum of a triangle, and we’ve seen how the angle sum of a triangle is different in different geometries. In particular, if the Cartesian curvature inside of a triangle is C, then (231) α + β + γ = π + C. The external angle sum of this triangle is (232) (π − α) + (π − β) + (π − γ) = 3π − (α + β + γ) = 3π − (π + C) = 2π − C. As we’ll see, the internal and external angle sums of polygons are nice in other geometries as well, and they look very much like equations (231) and (232). Once we have this nailed down, we’ll be able to understand the behavior of geodesics on polyhedral surfaces pretty well. 13. ANGLE SUMS AROUND THE VERTEX OF A CONE 2a 2b 1a 3a 3b 1b 8a 8b θ 4a 4b 148 7b 7a 6b 6a 5a 5b Figure 19. An octagon around the vertex of a cone with C = θ. 13. Angle Sums Around the Vertex of a Cone In Figure 19, there is an octagon around the vertex of a cone with angle defect θ. There is an extra copy of this figure at the end of this chapter. You can cut this out and tape the dotted edges together. The Cartesian curvature inside the octagon is C = θ. The angle sum of the octagon is (233) ∠1a + ∠1b + ∠2a + ∠2b + · · · + ∠7a + ∠7b + ∠8a + ∠8b. Figure 19 shows a triangulation of the octagonal disk on the cone. There are eight triangles in this triangulation, therefore, all the angles in the figure sum up to 8π. Eight of these angles are at the vertex of the cone. These must add up to 2π − θ, where θ is the measure of the wedge that is removed to make the cone. The angle sum of the octagon, therefore, must be (234) Σ = 8π − (2π − θ) = 6π + θ. Each of the external angles is π minus the internal angle, so the external angle sum must be (235) 8π − Σ = 2π + θ. The external angle sum of an octagon in the Euclidean plane is 2π, and on this cone, we get 2π + θ. 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS 149 13.1. Exercises. –1– Consider a random hexagon around the vertex of this same cone (angle defect of θ). Find the exterior angle sum for this hexagon. –2– Make a conjecture about the exterior angle sum of a polygon around the vertex of this cone. 14. Interior and exterior angle sums of polygons Instead of working on cones, we can just as easily prove this in some generality. Let’s do that. The theorems are as follows. Theorem 20. For a polygon with total Cartesian curvature C inside, the interior angle sum is π(n − 2) + C, where n is the number of sides, and the exterior angle sum is 2π − C. C C B B C D A E F F Figure 20. A pentagon with four points of non-zero Cartesian curvature inside. Look at Figure 20. There is an extra copy at the end of this chapter.You can cut this out, if you want, and tape the dotted edges with the same labels together to see what this looks like. For example, the two segments AB should be taped together. The points A, B, D, and E are interior points of the pentagon, and they have nonzero angle defects. At B, the angle defect, and Cartesian curvature, is negative, but that doesn’t really affect the computations. Let θA , θB , θD , and θE be the angle defects at these interior points. 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS 150 14.1. Exercises. –1– Since both segments AB are taped together, there is really only one edge there. The same goes for BC, CD, and EF . Compute the Euler characteristic for this triangulation. –2– Is this non-flat pentagonal thing a disk? Does that agree with your answer to the previous problem? –3– Using f , ei , eb, vi , ve , and vc for the number of faces, interior edges, boundary edges, etc., what are the values of each of these for the triangulation of Figure 20? –4– I’m not giving you the specific angle measurements, just things like θA . The angles around A must sum up to 2π − θA . Similarly for the angles around B, D, and E. What do these four quantities add up to? That is, what is the sum of all the angles around the vertices A, B, D, and E? –5– The sum of all the angles of all the triangles is 14π, since f = 14. The interior angle sum is part of this. The angles in the provious problem are also. The rest are the angles around the edge vertices. Let Σ be the interior angle sum of this pentagon. Compute Σ in terms of θA , θB , θD , and θE . –6– What is the exterior angle sum of this pentagon? I’ll do the general proof. It’s very similar to the argument in the last assignment. The main difference is that we have an angle defect at each interior vertex to keep track of. Here’s the proof. Suppose we have a triangulation of an n-gon, as in Figure 20. Actually, we’re interested in the n-gonal disk formed after the dotted edges are taped together. There are ei interior vertices, let’s say that ei = m. At each of these interior vertices, we may, or may not, have non-zero angle defects. Let’s say that the angle defects are θ1 , θ2 , . . . , θm . Some of these angle defects may be negative or zero. The sum of the angles around each interior vertex would be 2π, if this disk were flat, but here these angles add up to 2π − θj . Around each edge vertex, the angles add up to π, since these are straight angles. The rest of the angles of the triangulation are at the corner vertices, so all together, these add up to the interior angle sum of the n-gon. Let’s call the angle sum Σ. Since all the angles of all the triangles in the triangulation add up to πf (f is the number of faces), we need to subtract off the angles around the interior and edge vertices to 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS 151 get Σ. This gives us (236) Σ = πf − [(2π − θ1 ) + (2π − θ2 ) + · · · + (2π − θm )] − πve = πf − 2πm + (θ1 + · · · + θm ) − πve . The total amount of Cartesian curvature (the sum of the angle defects) inside the n-gon is C = θ1 + · · · + θm , and ei = m. We can rewrite this last equation as (237) Σ − C = πf − 2πvi − πve. What we have here is a disk, even though it’s not flat. The Euler characteristic must be χ = 1. This tells us that (238) f − ei − eb + vi + ve + n = 1. Around the boundary of the n-gon, the number of edges and vertices must be the same. Therefore (239) ve + n = eb . Finally, each face contributes three edges, but this counts the interior edges twice. Therefore, (240) 3f = 2ei + eb. OK. Equations (237), (238), (239), and (240) look exactly like equations (226), (11), (224), and (227), except we have Σ − C in (237), and we had Σ before. Exactly the same computations, therefore, will lead you to the equation (241) Σ − C = π(n − 2). To compute the exterior angle sum, we subtract each of the n interior angles from π. This means (242) πn − Σ = πn − C − π(n − 2) = 2π − C. This proves the theorem. In Euclidean geometry, a 2-gon should have an interior angle sum of 0. On a polydral surface, if a 2-gon contains Cartesian curvature totalling C, then the angle sum should be C according to Theorem 20. On the following pages there are figures (Figures 21 and 22) that can be cut out showing two 2-gons on the cube. Both contain two vertices, which each have an angle deficit of π2 for a total of C = π, so 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS 152 their angle sums should be π. The interior angles on both 2-gons are right angles, so this is indeed the case. 14.2. Exercises. –1– On the surface of a cube, the angle defects at each vertex are π2 . What is the interior and exterior angle sum of a quadrilateral that contains two vertices? –2– Since the cube is a topological sphere, it’s not really clear what the inside of a quadrilateral is. The quadrilateral divides the cube into two regions (usually one is smaller than the other). For problem 13, we could say that the region containing the other six vertices is the interior. Compute the interior and exterior angle sums with this interpretation. –3– In my proof of Theorem 20, did I do a lot of algebraic computations? Why? 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS Figure 21. If you cut out and tape this cube together, you will see a pair of geodesics forming a 2-gon. 153 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS Figure 22. Here’s another 2-gon on the cube. 154 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS 2a 2b 8a 1a 1b 8b θ 3a 3b 7b 7a 4a 6b 6a 4b 5a 5b Figure 23. This is an extra copy of Figure 19. Cut out figure and tape along dotted edges. 155 14. INTERIOR AND EXTERIOR ANGLE SUMS OF POLYGONS C C B B 156 C D A E F F Figure 24. This is an extra version of Figure 20. Cut out the figure and tape the dotted edges together.