Image To PDF 20231228 10.12.05

You might also like

Download as pdf
Download as pdf
You are on page 1of 24
2. Compact surfaces whose points are umbilics 0 ae the first few theorems of this chapter we shall use the efnition of surface given in Chapter II, and assume that each point has a neighbourhood (homeomorphie to an open 2-cell) which can be described by parametric equations r= r(u, v). As our first theorem of differential geometry in the large we shall prove TuxoreM 2.1, ThZonly compact surfaces ‘38 > 2 for which every point is an umbilic are spheres. ; fe x m ue “/ This is an example of a global tale tereteae of the hypo- thesis—viz. the compactness of the surface considered as a set of points in Z,—evidently involves the surface as a whole, A useful technique in proving global theorems in differential geometry is first to establish the result locally in some neighbourhood of an arbitrary point, and then to try to extend the result so that it applies globally. We employ precisely this technique in proving Theorem 2.1. By means of the local differential geometry developed in Chapters II and III we shall prove that in the neighbourhood of any point the surface is either spherical or plane. We then use the property.of compactness to reject one alternative, and show that the surface must in fact be a sphere. ' Let S be a compact surface of class > 2 for which every point ig an umbilic. Let P be any point on 8, and let V be a coordinate neighbourhood of S containing P, in which part of S is represented parametrically by r= r(u,v). Since every point of V is an um lying in V must be a line of curvatur formula, at all points of V, | dN+xdr = 9, (2.1) bilic, it follows that every curve e. Hence, from Rodrigues’ * Kkpy Ka has a maximum at Fy @ < Kp, Kq has minimun at as Gaussian curvature K can” be 2 * Wunpace sur ture We note that a compact surface must Possess a ‘highest point’, and at this point the curvature is necessarily non-negative. From this, it follows that a compact surface cannot haye constant nega- tive curvature, Moreover, a compact surface cannot have constant Zero curvature, for other would contradict the compactness, We shall now prove a theorem about compact surfaces of constant Gaussian curvature, and in view of the previous Tema tks we need consider only the case of constant positive Curvature. aS] dq} THeorem 4.1, The only AW Gurvature are spheres. »\ ht 14C€s OF Const ry z . impact Syrfaces with constant Gaussian ath Let S be a compast “urface with constant positive Gaussian “curvature K. Since § is vompact, there is a point P, at which the maximum value of the Principal curvature is attained, Since the » than the minimum. Frora Hilbert’s lemma it follows that the two Principal curvatures must be eaux equal, 18. at tio point does either principal curvature exceed VK. Hence every point of S is an umbilic, and the theorem now follows from Theorem 2.1, _¢ Ma similar manner we prove * THEOREM 4.2, The oni “L. ture is positive and mean y compact surfaces whose Gaussian curva- Curvature constant are spheres, Let S bea Compact surface of positive Gaussian curvature and constant mean curvature, and denote by ka ky Tespectively the larger and smaller Principal curvatures. Since kg is continuous and S compact, there is a point F, at which x, attains its maximum value. Since the mean curvature is constant, it follows that ky attains its minimum value at Fy. Now we have the relation everywhere. If k, > x, at Ps, then the lemma in section Ka > ky c 3 would apply and we could conclude that Kx 0, contrary to hy Heace we must have ka = 'ypothesis, =: ky = pat P, and hence everywhere ons, This completes thé proof of the theorem, ne Complete surfaces The surfaces considered in the previous theorems w ous ws WETS Pestrigted to be compact. This is a strong restriction which wo ‘a uld exclude, \ ‘ i ise it would contain straight lines which _ 132 Geometry of Surfaces in the Large Wis | om cea al surfaces and many common surfaces e paraboloids. It will be seen that a restriction more suitable for the purpose ‘of differential geometry is to require the surface to be complete in the sense of the theory of metric spaces, We first show that the surface can be regarded as a metrie space in this sense \It will be remembered that a set of points \S carries the structure of a metric space when there is a real-valued function p:Sx S > Re with the properties (i) p(A, B) = 0 if and only if A = B, (ii) p(A, B) = p(B, A), (iii) p(A, C) < p(A, B)+p(B, 0), for all points A, B, O of S. ‘We assume that our surface S is connected so that any two points can be joined by arc-wise connected paths. Ify is any path joining A to B, then this path can be divided into a finite number of seg- ments so that each segment lies entirely in one coordinate neigh- pourhood, and adjacent coordinate neighbourhoods overlap. The length of the segment whose equation relative to a coordinate neighbourhood is, = u(t), 0 = v(t) is given by [ JBes2Pis+ oeyat taken between appropriate limits. Then, the length of y is defined to be the sum of the lengths of its segments. We now define the distance function p by. ; (A, B) is the greatest lower bound of the lengths of all arc-wise connected C} paths joining A to B. It is evident that the distance function so defined satisfies conditions (ii) and (iii) of the metric while condition (i) is satisfied because the first the surface is positive definite. n the surface is said to form a Cauchy. ¢ real number , an integer m. can be es found such that p(,,%) <€ provided p and g both exceed 2g. Evidently, if {2,} converges to a limit x, then the sequence {,} is>\ a Cauchy sequence. ‘The question naturally arises whether every : Cauchy sequence of points converges in the surface. If the surface, is such that every ae sequence converges, then the metric space is said to be ‘complete’ : the following aa le eine shows that not all surfaces are complete. Consider the surface form: OR ea mer space axioms, fundamental form of A sequence of points {z,} 01 sequence when, given a positiv ed by the two-dimensional Gomplete Surfaces I, §5 133 \ Cartestan plane of pairs of real numbers (2,1) when theo), (0,0)is removed. The distance function p is the Euclidean q, mgm ustance function defined by ld, B) = lees) + luo}, | when (24,44), (tg, Yg) are the rectangular coordinates of poin A and B. The sequence of points {(1/n,0)} is easily s i Cauchy sequence which does not converge in the surfac, surface is not complete. It may be noted that the two points (a, 0), (=a,0) (a > 0) cannot be joined by a geodesic (straight line) lying entirely on the surface. In the next section we shall prove that on a complete surface any two points can be joined by a geodesic which lies entirely on the surface. 6. Characterization of complete surfaces In this section we consider three properties each of which can be used to characterize complete surfaces. That these three properties are equivalent was first proved by Hopf and Rinow (1931), but the proof given here follows closely a simplified proof given by de Rham (1952). The properties are: (a) Every Cauchy sequence of points of S is convergent. (6) Every geodesic can be prolonged indefinitely in either direc- tion, or else it forms a closed curve. (c) Every bounded set of points of S is relatively compact. It is evident that. condition (c) implies (a). We now prove that (a) implies (6). Let y be a geodesic which cannot be extended in- definitely. If y is a closed curve, then (6) is satisfied. If y is not @ glosed curve and if P(z) is some point on y, then there is om \Ufumber 1 such that y can be prolonged for distances (measured along y) less than ! but cannot be prolonged for distances era n 1. Consider now the sequence of points {x,} lying 0" 7 a : fc 4\fanees from P along y given by (1—1/n). Evidently {z,} 8 A ong + ‘ to some ,@auchy sequence, which by hypothesis (a) ae, anothet \\ point Q on y whose distance from P is precisely !. If on faomne Cauchy sequence such that p(2,,2,,) > then {25 limit Q’. Now, the sequence 2r,, a1, 2.) 7, %) %»- sequence tending to both Q and Q’. Hence @ exists a unique end point Q distant / from P alos i of S which contains @- so a Cauchy y and there ‘Cansidet now AtQ there 18 oe irection of the uniquely determined a direction t which is soa Bod _Bcodesic —y which starts at @, In this coordina : : : 4 irection (—t), and there is a unique geodesic at @ which has the py to hypothesis. this gives a continuation of y beyond Q, contr It follows that y must satisfy condition (b), 80 We HG eo), (2) implies (6), Since (c) implies (a), we conclude that (c) ied i, In order to prove that the three conditions oe Ey econ remains only to prove that (b) implies (¢). 2.” oo. fg Suppose now that S has the property (6).} Consider a point CORD and geodesic ares which start at a. We defitre that initial vector of a geodesic arc starting at a to be the tangent vector to this o ca which has the same sense as the geodesic and whose length is equal to the length of the geodesic arc. Since S has the property (b), it follows that every tangent vector to S at a, whatever its length, is the initial vector of some geodesic arc starting at a which is tuiquely determined. This arc may eventually cut itself or, if it forms part of a closed geodesic, may even cover part of itself. Let S, be the set of points a of S whose distance from a does not ®xoeed 7. i.e. p(x,a) 0, This follows because it would R>0, now prove that, 136 Geometry of Surfaces in the Large TV, §6 Now let h + 00; {zs} will have at least one point of accumulation x with the property i play) 0 such that if ‘Ay) < s(x), the point y is the extremity of the unique geodesic are of length p(x, y) joining x to y. Moreover, the continuous func- tion s(2) attains a positive minimum value on the compact set Ep, and we take s to be thig minimum, Then, if (6.1) is true fogr = Rand if R < p(a,y) < R-+s, there exists an 2 € Ey such that p(a,z) = R and p(x,y) = p(a,y)—R - perties of the space $ are automatically true for the space S Moreover, the differential-geo trcture. on induces & differential-geometric structure on coe Geometry of Surfaces in the Large oT We now return to the proof of Hilbert’s theorem. | We eee that a surface § exists having the required propertiesand we obtain & proof by contradiction (Consider an arbitrary geodesic Tine on the surface $ and take-dn‘arbitrary point O on this geodesic as origin. If s denotes the are length of this geodesic measured from O, the completeness of SLensures that the geodesic can be continued in both ditections from —oo to al It is possible that the geodesic will ultimately cross itself go that the same point on S will have two different s-values. However, pacar of Sits universal covering surface $’, then differelit values of s will correspond to different points on $7] This follows because or{a surface of negative Gaussian curvaturetwo geodesic arcs cannot enclose a simply connected region [ef. Exercises II, 13). ; At each point ét parameter son the given geodesio(ebnsider the orthogonal geodesic line,and let its are length ¢ be chosen as para- mete®) In this notation(the equation of the given geodesic is t = 0. Now, two of these geodesic arcs at s,, 8, cannot meet on the surfac § to form with the geodesie arc s,s, a simply connected region. © For if this were the case, the sum of the angles of the geodesic triangle so formed would not be less than poets to the results of Chapter IT. We denote a point in the eovering space S’ by the pair of coordinates (5,2), and we observe that roy (s,t) correspond to different points on S’, We now show tha every of S can be represented om the covering surfaces" in this int It follows from Chapter II, section 18, that{the line element of the surface assumes the form dst O(6) dP Suppose now that a point P of the surface § remained uncovered by our construction (see Fig. 9), Join P to O (s = 0, t= 0) by some rectifiable curve y) Then there must be some point Q on y with the property that All points between O and Q can be covered, while points on y arbitrarily near Q on the side of Q remote from O cannot be covered (If Q, lies on y between O and Q it follows from the form of the imetric that the length of the curve 0Q, is greater than or equal to 8, where so, is the s-coordinate of the corresponding point on 8’. The set of values {s9,} is therefore pounded, and we define sq to be the least upper bound of this s¢t, Let R bo the point on the geodesic t= 0 distant sq from 0, and consider the orthogonal geodesics along some interval on the NTE IV, §7 Hilbert’s Theorem 139 geodesic t= 0 which contains R. These geodesics will cover a strip of the surface which certainly contains the point Q, and points beyond @ on the curve y. This gives a contradiction, and(we conclude that every point of the surface.S.can.be covered in this way. ) There is thus a local homeomorphism between points of S and the, (6-1) plane, but this correspondence may not be (1-1) in the large.) LI analy Vi P Fre. 9 Peete tees very anace 8" isthomeomorphia with the (st) lane. )) We’now consider the esyraptotic lines on the surface S. These are given by the differential equation Lds?-+ 2M dsdt+-N dt? = 0. Since K < 0 we conclude that LN—M? < 0 and hence that at each point of S the asymptotic directions are real and different. Hence atach point of 8’ these determine two distinct directions, and similarly at each ot of the (s-t) plane. Since the (s-t) plane is simply connectedAt follows that the differential equation gives rise to two vector fields which can be continued over the whole plane.(The Lipschitz’ condition for the uniqueness of the solution of thé differential equation, is satisfied because we have assumed that S is of cless w,) Thus throughout the whole (s-t) plane there are two systems df asymptotic lines with the property that a curve from each system passes through an arbitrary point. Moreover, since S is free from singplarities, the differential equa- tion has no singularities. It followg from the theorem of Bendixont * + Cf. Bieberbach (1923). 1 oe ( IV, $8 Conjugate Points on Geodesics and its next zero. after », is just v-+na. It follows that if length A B exceeds a, then B will not lie between A.and 4, and the theorem is proved. An analogous result is the following: THHOREM 8,54 is of a geodesic the Gaussian curvature is less than 1/b?, the curve is necessarily of shorter length than neigh- bouring curves along an arc length at least equal to ab. The proof follows easily from the hypothesis K < 1/6, and the fact that the interval between consecutive roots of the equation d?p{dv? = —p/b? is 7b. As this cannot be smaller than the interval between consecutive roots of the previous equation, it follows in this case that if the arc length AB is less than ab, then B will certainly lie between A and A,, thus giving the required resuit. Suppose now the surface S is compact, and has the property that A > 1/a? everywhere. From section 6 it follows that if 4 and Bare any two points on S, there is a geodesic joining 4 to B which is of shorter length than the neighbouring curves. It follows from orem 8.4 that the maximum distance between A and # 151 the are The cannot exceed za. This proves the following: PurorEM 8.6. [fon a compact surface S the curvature everywhere L/a®, the maximum distance between any two points cannot 4.1, Prove that the Gaussian curvature at any point EX EF on the ellipsoid is given by where p is the e of the centre from the tangent plane. Show that ifa > b > ¢, every geodesic are of length greater than rab,¢ cannot be the shortest. distance between its extremities; but every geodesic are of length less than nbo/a is necessarily shorter than the neighbouring curves joining its extremities.

You might also like