You are on page 1of 309
PRINCETON MATHEMATICAL SERIES Editor ‘Marston Morse, H. P, Robertson, A. W. Tucker 1, THE CLASSICAL GROUPS THEIR INVARIANTS AND REPRESENTATIONS By Hermann Weyl 2, TOPOLOGICAL GROUPS By L, Pontrjagin 3. AN INTRODUCTION TO DIFFERENTIAL GEOMETRY WITH USE OF THE TENSOR CALCULUS By Luther Pfabler Eisenhart DIMENSION THEORY By Witold Hurewies and Henry Wallman (In Press) THE ANALYTICAL FOUNDATIONS OF CELESTIAL MECHANICS By Aurel Wintner (In Preparation) AN INTRODUCTION TO DIFFERENTIAL GEOMETRY WITH USE OF THE TENSOR CALCULUS By LUTHER PFAHLER EISENHART DOD PROFESSOR OF MATHEMATICS PRINCETON UNIVERSITY PRINCETON PRINCETON UNIVERSITY PRESS LONDON : HUMPHREY MILFORD : OXFORD UNIVERSITY PRESS 1940 Preface Since 1909, when my Differential Geometry of Curves and Surfaces was published, the tensor caleulus, which had previously been invented by Ricei, was adopted by Einstcin in his General Theory of Relativity, and has been developed further in the study of Riemannian Geometry and various generalizations of the latter. In the present book the tensor calculus of euclidean 3-space is developed and then generalized so as to apply to a Riemannian space of any number of dimensions. The tensor calculus as here developed is applied in Chapters III and TV to the study of differential geometry of surfaces in 3-space, the material treated being equivalent to what appears in general in the first eight chapters of my former book with such additions as follow from the introduction of the concept of parallelism of Levi-Civita and the content of the tensor calculus. Of the many exercises in the book some involve merely direct appli- cation of the text, but most of them constitute an extension of it. Tn the writing of the book I have received valuable assistance and criticism from Professor H. P. Robertson and from my students, Messrs. Isaac Battin, Albert J. Coleman, Douglas R. Crosby, John Giese, Donald C. May, and in particular, Wayne Johnson. The excellent line drawings and half-tone illustrations were conceived and executed by Mr. John H. Lewis. Princeton, September 27, 1940 Lutuer PrautEr EXseNHART. Contents Cnarrer I CURVES IN SPAC ssncx108N Pace 1, Curves and surfaces. ‘The summation convention. at 2. Length of a curve. Linear element. 8 3. Tangent to a curve. Order of contact. Osculating plane n 4. Curvature. Principal normal. Circle of curvature 16 5. Binormal. Torsion 19 6. The Frenet formulas. The form of a curve in the neighborhood of a point. : 5 7. Intrinsic equations of a curve........ oa . : 31 8. Involutes and volutes of a curve 9, The tangent surface of acurve. ‘The polarsurface. Osculating sphere... 38 10. Parametric equations of a surface. Coordinates and coordinate curves in a surface. “ JL. Tangent plane to a aurface 50 12, Developable surfaces. Envelope of a one-parameter family of surfaces... 53 Cuaprer I TRANSFORMATION OF COORDINATES, TENSOR CALCULUS 13, Transformation of coordinates, Curvilinear coordinates se 8 M4, The fundamental quadratic form of space 70 15, Contravariant vectors, Scalars : w 16. Length of a contravariant vector. Angle between two vectors......... 80 17. Covariant vectors. Contravariant and covariant components of a veetor....... : 83 18, Tensors. Symmetric and skew symmetric tensors 80 19. Addition, subtraction and multiplication of tensors. Contraction... 94 20. The Christoffel symbols. The Riemann tensor 98 21. The Frenet formulas in general coordinates. . 103 22, Covariant differentiation. ..... 107 23. Systems of partial differential equations of the first order. Mixed pyYMCMS.. vevdeeeetenee 4 Cuarrer IIL INTRINSIC GEOMETRY OF A SURFACE 2%. Linear clement of a surface. First fundamental quadratic form of a surface. Vectors in a surface oe 12 25. Angle of two intersecting curves in a surface. Element of area. 9 28, Families of curves in a surface. Principal directions : 138 2%. The intrinsie geometry of a surface, Isometric surfaces. . 48 28. The Christoffel symbols for a surface. The Riemannian curvature tensor, The Gaussian curvature of a surface eres ix 40. 41. 4s. 46. a7. 49. CONTENTS Differential parameters . Isometric orthogonal nets. Isometric coordinates - Isometric surfaces . Geodesies : . Geodesic polar coordinates. Geodesie triangles . Geodesic curvature. . The vector asociate to a given vector with respect to a eurve. Paral- lelism of vectors. . . Conformal correspondence of two surfaces . Geodesic correspondence of two surfaces. Cuaprer IV SURFACES IN SPACE The second fundamental form of a surface............+ . The equation of Gauss and the equations of Codazzi Normal curvature of « surface, Principal radii of normal curvature. Lines of curvature of a surface. Conjugate directions and conjugate nets. Isometrie-conjugate nets Asymptotic directions and asymptotic lines. Mean conjugate directions, The Dupin indicatrix... . Geodesic curvature and geodesic torsion of a curve. . Parallel vectors in a surface. Spherical representation of a surface. ‘The Gaussian curvature of a surface. ‘Tangential coordinates of a surface... . Surfaces of center of a surface. Parallel surfaces Spherical and psoudospherical surfa Minimal surfaces... . Bibliography. boevttetestersstetenenes Index... 155 161 166 170 180 186 194 201 205 212 218 222 228 231 . 237 243, 249 252 268, 277 CHAPTER I Curves in Space 1, CURVES AND SURFACES. THE SUMMATION CONVENTION Consider space referred to a set of rectangular axes. Instead of denoting, as usual, the coordinates with respect to these axes by 2, y, 2 we use z', 2, 2°, since by using the same letter z with different super- scripts to distinguish the coordinates we are able often to write equa- tions in a condensed form. Thus we refer to the point of coordinates z', ', x as the point x‘, where 7 takes the values 1, 2,3. We indicate a particular point by a subscript as for example zi, and when a point is a general or representative point we use 2‘ without a subscript. When the axes are rectangular, we call the coordinates cartesian. In this notation parametric equations of the line through the point zj and with direction numbers u’, u’, w* are* (4) gearitue (@ = 1, 2,3). This means that (1.1) constitute three equations as i takes the values 1, 2,3. Here Z is a parameter proportional to the distance between the points 2{ and 24, and ¢ is the distance when ut are direction cosines, that is, when (1.2) Zw = 1, which we write also at times in the form Ylaf = ‘An equation of a plane is (1.3) az! + a2 +asz* +a = where the a’s are constants. In order to write this equation in con- densed form we make use of the so-called summation conrention that when the same index appears in a term as a subscript and a superseript this term stands for the sum of the terms obtained by giving the index *Sce C.G., p. 85. A reference of this type is to the author's Coordinate Geom- etry, Ginn and Company, 1939, 1 2 CURVES IN SPACE [Ca 1 each of its values, in the present case the values 1, 2,3. By means of this convention equation (1.3) is written (1.4) az'+a=0. This convention is used throughout the book. At first it may be troublesome for the reader, but in a short time he will find it to be preferable to using the summation sign in such cases. A repeated index indicating summation is called a dummy index. Any letter may be used as a dummy index, but when a term involves more than one such index it is necessary to use different dummy indices. Any index which is not, a dummy index and thus appears only once in a term is called a free index. Since two intersecting planes meet in a line, the two equations (1.5) azita=0, bai +b=0, that is, aye’ + aye” + age? +a =0, bit! + den” + ba? +b = 0, are equations of a line, provided that the ratios a/b: , a:/be , a4/bs are not equal; if these ratios are equal the planes are coincident or parallel according as a/b is equal to the above ratios or not.” An equation (1.4) is an equation of a plane in the sense that it picks out of space u two dimensional set of points, this set having the property that every point of a line joining any two points of the set is a point of the set; this is Euclid’s geometric definition of a plane. In like manner any functional relation between the coordinates, denoted by (1.6) JG’, 2°, 2°) = 0, picks out a two dimensional set of points, by which we mean that only two of the coordinates of a point of the locus may be chosen arbitrarily. The locus of points whose coordinates satisfy an equation of the form (1.6) is called a surface. Thus (1.7) Dake! + 2aja’ +b 7 where the a’s and } are constants, is an equation of a sphere with center at the point —a; and radius r given by = Daya — bt 7 *C.G., pp. 100, 101. +C.G, p. 128, Ex. 14, §l] CURVES AND SURFACES 3 Whenever throughout this book we consider any function, it is under stood that the function is considered in a domain within which it is continuous in all its variables, together with such of its derivatives as are involved in the discussion. Since an equation which does not involve one of the coordinates does not impose any restriction on this coordinate, such an equation is an equation of a cylinder. Thus (1.8) S@', a) =0 is an equation of a cylinder whose generators, or elements, are parallel to the a-axis, each generator being determined by a pair of values satisfying (1.8). If 21, zi are two such values, the generator is defined by the two equations deat, steal, these being a special form of (1.5) in this ease. It does not follow that when all three coordinates enter as in (1.6) that the surface is not a cylinder, but that if it is a eylinder the generators are not parallel to one of the coordinate axes. Later (§12) there will be given a means of determining whether an equation (1.6) is an equation of a cylinder. ‘Two independent equations (1.9) Jia’, 2’, 2) = 0, fila’, 2°, 2°) = 0 define a curve, a one dimensional locus, for, only onc of the coordinates of a point on the locus may be chosen arbitrarily. A line is a curve, its two equations being linear, as for example in (1.5). A curve, being one dimensional, may be defined also by three equations involving a parameter, as (1.10) a= sd, which are called parametric equations of the curve. These are a gen- eralization of equations (1.1). The functions f‘ in (1.10) are under- stood to be single-valued and such that for no value of ¢ are all the first derivatives of f' equal to zero; the significance of this requirement appears in §3. If g(u) is a single-valued function of u, and one replaces f in equations (1.10) by ¢ = ¢(u), there is obtained another set of parametric equa- tions of the curve, namely (i) 2! = few) = of. Since all the first derivatives @ are not to be zero for a value of u, 4 CURVES IN SPACE [Cu 1 te there is the added condition on y that * + 0; this means that the equation ¢ = g(u) has a unique inverse” ‘rhus the number of sets of parametric equations of a particular curve is of the order of any function satisfying the above condition. When, in particular, one of the coordinates, say 2°, is taken as parameter, the equations are (1.12) H=f@, 2=fe), #=2, the forms of the f’s depending, of course, upon the curve. From the form of (1.12) it follows that the curve is the intersection of the two cylinders whose respective equations are the first two of (1.12). When all the points of a curve do not lie in a plane, the curve is said to be skew or twisted. The condition that a curve with equations (1.10) be a plane curve, that is, all of its points lie in a plane, is, as follows from (1.4), that the functions f* be such that (1.18) afi +a=0, that is af + af +afi+a=0, where the a’s are constants. Differentiating equation (1.13) three times with respect to ¢ and denoting differentiation by primes, we ob- tain the three equations (1.14) af’=0, af” =0, af” = In order that the a’s be not all zero, we must. havet seg gh | (1.15) pepe Toa =0. yen pee "| Conversely, we shall show that if three functions f“(i) satisfy this condition, constants a, and a ean be found satisfying (1.13); and conse- quently that the curve 2" = f‘(t) is plane. If (1.15) is satisfied there exist quantities b;, ordinarily funetions of ¢, such that (1.16) uf’ =0, vf =0, bf” = O04 * Fine, 1927, 1, p. 55. References of this type are to the Bibliography at the end of the book. 4C.G., p. 14. 1C.G., p. 16. a) CURVES AND SURFACES 5 Assuming that the 6’s in these equations are functions of t and differ- entiating the first two of these equations with respect to ¢, the resulting equations are reducible to (1.16’) vs =0, vif’ =0. The b’s in the first two of (1.16) are proportional respectively to the cofactors of the elements of the last row in the determinant (1.15).* ‘The same is true of the b”s in (1.16’). Consequently we have by bbs bbe bs If we denote the common value of these ratios by ¢”(, we have on integration be = ae’, where the a’s are constants. Substituting in the first of (1.16) and discarding the factor e*, we obtain af On integrating this equation with respect to f, we obtain an equation of the form (1.13). Hence we havet =0. [1.1] A curve with equations (1.10) ie a plane curve, if and only if the Sunetions f° satisfy equation (1.15). When for the curve with the equations (1.17) g=a, taf, c= ct the expressions for are substituted in an equation of a plane (1.4), we obtain a cubic equation in ¢ for each of whose roots the corresponding point of the curve lies in the given plane. The curve (1.17) is called a twisted cubic. When a curve meets a general plane in n points, it is called a twisted curve of the n™ order. *C.G, p. 104, { In the numbering of an equation or equations, as in (1.3), the number pre- ceding the period is that of the section and the second umber specifies the par- ticular equation or equations. ‘The same applies to the number of a theorem but in this ease brackets are used in place of parentheses. This notation is used throughout the book. 6 CURVES IN SPACE [Ca I From (1.17) we have that the projections of the curve upon the coordinate planes have the respective equations de — alr) =0, 2 = 0, (1.18) ez —a(c’=0, 2 =0, ay — day =0, 2 =0. ‘These eurves are shown schematically a8 follows for positive values of the ¢’s: ti Fie. 1 From (1.18) it is seen that, the curve is the intersection of the three cylinders whose equations are the first of cach pair of equations (1.18), the generators of these eylinders being parallel io the z*-, 2 and 2-axes respectively. 14 will be found that many formulas and equations eau be put in simpler form by means of quantities e:j. aud ¢™ defined as follows: are (0 when two or three of the indices have the same values; (1.19) os} _, |1 when the respective indices have the values 1, 2, 3; & 2,3, Lor 3, 1, 25 —1 when the respective indices have the values 1, 3, 2; 3,2, 1or2, 1,3. Consider, for example, the two equations azei=0, ay = from which it follows that* eoDiiat at vei yg (1.20) ay20205 = *C.G,, p. 104. $1) CURVES AND SURFACES 7 Thus, denoting by 1/r the factor of proportionality, we have ray = erjet'y = ensa?y? + eme’y? = 22y' — oy, and consequently (1.20) may be written ra; = esna’y*. Applying this process to the last two of equations (1.14) we have ras = conf fe, and when these expressions for a; are substituted in the first of equations (1.14) we obtain (1.21) end Pf" = 0, which is equation (1.15), as one verifies by forming the sum indicated by the summation convention as each of the indices independently takes the values 1, 2, 3. EXERCISES 1, Parametric equations of a line normal to the plane (1.4) and passing through a point of the curve (1.10) are @ Xf = f'® + au, where u is a parameter; as (and u take all values equations (i) give the coordi~ nates of points on the cylinder whose generators pass through points of the curve and are normal to the plane (1.4); when in equations (i) we put 1 “Ser (+ Eas), the resulting equations are parametric equations of the projection of the curve (1.10) on the plane (1.4). 2. The equation Caca(tate + tats + tats) — creates + ba + ta)a* + cicat® — cicacstitits = 0 is an equation of the plane through three points of the twisted eubie (1.17) with parameters f; , t:andt: 3. The plane Bexesa%x! — Bcyesar* + eyes — creaesa = 0 for cach value of the constant a meets the curve (1.17) in three coincident points at the point t = a. 4. Thore pass through a given point zi in space three planes with equations of the form of the equation of Ex. 3, and if a1, az, as denote the corresponding values of a, we have get 32% atata=—, antmataa =, e a 8 CURVES IN SPACE (Cu. I from this result and Ex. 2 it follows that an equation of the plane through the three points in which these three planes meet the cubic (each in three coincident, points) is Bea(aiat — zjz*) + crca(e* — 2) = 0, which plane passes through the point zf . 5. Four planes determined by a variable chord of the cubic (1.17) and four fixed, points of the eubic are in constant cross-ratio. 6, The curve meacost, z=asint, 2 = bsin 2% is the intersection of a circular cylinder and a hyperbolic paraboloid. 7. Determine f(t) so that the eurve waacost, e=asint, x= f(t) shall be plane; what is the form of the curve? 8, By means of the quantities ei and e'* ‘one has ay a aw Qn ax Gas | = anon aes = eft argday dae, au dy an and, if the determinant is denoted by a, then end = are Oni Ont « 9, Show that x eapnoutn = Bix dst — Budse, where = 1 or 0 according asi = j ord # fj; from this rosult it follows that DX Ceiincem + enact + cnsneqn) = 0. 7 2. LENGTH OF A CURVE. LINEAR ELEMENT Consider a curve with the equations (2.1) f= sf, and the are of the curve between the points Py and P, for which the parametric valucs arc f and é respectively. Consider also inter- §2] LENGTH OF A CURVE, LINEAR ELEMENT 9 mediate points P, Pz, --- for which the values of the parameter are th, fa, +++. The length l of the chord PxPay2 is given by VE Ga) ~ FOF VE @F Gan — 4) e+ Giltus—t) 0<6<1, A q where § the second expression for i; following from the mean value theorem of the differential calculus, where the prime denotes the derivative. As the number of intermediate points Ps increases indefinitely and each ls approaches zero, the limit of the sum of the iy’s is the definite integral STE { Vena" By definition this is the length of the are PyP,. If then s denotes the Jength of the are from the point of parameter é to a representative point of parameter ¢, we have (2.2) =f of Stba. This gives s as a function of f; we denote it by (2.3) s=¢, where ¢ involves to also. From (2.2) we have (2.4) ds? = (dx'? + da’)? + (ax’y D dex‘ de’, 7 where dzé = ¢ dt. As thus expressed ds is called the element of length, or linear element, of the curve. As remarked in §1 there is a high degree of arbitrariness in the choice of a parameter for a curve. In what follows we shull often find that it adds to the simplicity of a result, if the are s is taken as parameter. From (2.2) we have [2.1] For a curve with equations (2.1) the parameter t is the length of the curve measured from a given point, if and only if (2.5) ef 7 - 10 CURVES IN SPACE (Cu. I It is evident that s is defined by (2.2) except when (2.6) Since it is understood now, and in what follows, that we are considering only real functions f* of a real parameter, that is, one assuming only real values, there is no real solution of (2.6) other than f* constant, that is, the locus is a point. If we admit complex functions of a complex parameter, the curves for which (2.6) hold are called curves of length zero, or minimal curves, There are eases in which it is advisable to consider such curves, but unless otherwise stated they are not involyed in what follows. Consider a curve defined in terms of the are as parameter. Let P and P of coordinates 2‘ and # be points for which the parameter has the values s and s + ¢ By Taylor’s theorem we have Lerea. = 1,23). (2.7) aap thee + 8 Here and in what follows an 2 with one or more primes means that the are s is the parameter and the primes indicate derivatives with respect ; 3 la* . to s; if the parameter is other than s, we write & and similarly for bigher derivatives. Tn this notation (2.5) is (2.8) Dev a1 7 Differentiating this equation with respect to s, we have 29) Date If we denote by I the length of the chord PP, it follows from (2.7), (2.8) and (2.9) that eso & a2 aaa Dea a 4 baa e+ From this result it follows that as P approaches P along the curve the ratio of the chord to the are ¢ approaches unity as limit, §3] TANGENT TO A CURVE MW 3. TANGENT TO A CURVE. ORDER OF CONTACT. OSCULATING PLANE The quantities 2° — 2‘ in (2.10) are direction numbers of the line through P and P, and (2° — 2')/Lare its direction cosines. From (2.7) and (2,10) it follows that Since by definition the limiting position of the line through P and B as P approaches P along the curve is the tangent to the curve at P, we have [3.1] When for a curve x‘ are expressed in terms of the length of the arc from a given point as parameter, the quantities x" are direction cosines of the tangent at a point x", If 2' for a curve are expressed in terms of a parameter ¢, the quantities e are direction numbers of the tangent. Thus the tangent is not defined if all of these quantities are zero, which possibility was excluded in $1. As a result of this theorem we have as parametric equations of the tangent to a curve at a point 2° Ga) Xeata"d, where X‘ are coordinates of a representative point on the tangent and dis the distance from the point 2‘ to the point X‘* We define positive sense along the tangent as that for which d in (8.1) is positive, this means that a half line drawn from the origin parallel to the tangent makes with the coordinate axes angles whose cosines are 2“. This same convention applies to any line associated with a curve when diree- tion cosines of such a line are given in terms of quantities defining the curve.f *C.G., p. 85. + Here we define sense by means of direction cosines, which means that a line has two sets of direction cosines, differing in sign. ‘This is not the convention adopted in C. G., pp. 77, 78. 12 CURVES IN SPACE [Cu. 1 Af we denote by a‘ the direction cosines of the tangent, we have from the above result and (2.2), when the parameter ¢ is any whatever, das (3.2) "Tae aa The plane through a point of a curve and normal to the tangent at the point is called the normal plane at the point; an equation of this plane is (3.3) z (X' = sa! 9, where a! are given by (8.2).* ; Parametric equations of any line through the point 2" of a curve are BA) Xisai tui, - where u‘ are direction cosines of the line. The square of the distance of the point P(z') from this line is given byt (3.5) When the point P is a point of the curve, its coordinates being given by (2.7), the expression (3.5) becomes c= fe" wae we + 3 Qh we at ue (3.6) + F at = ad +. J Hi ah = whet FH (Gu ~ 2h whet s- fr. Thus d is of the order of ¢ unless the u’ are proportional to 2"; since in the latter case both of these sets of argnt sare by theorem [3.1] direction cosines, it follows that ué = ex’, where ¢ is +1 or ~1, and that the distance of a point on the curve nearby a’ is of the second, or "OG, p. 9 1C.G, p. 96. 3] ORDER OF CONTACT. OSCULATING PLANE, 13 higher, order. ‘The distance is of the third order if also 2°” are pro- portional to 2”, and of the (n + 1)" order if 2°", 2°”, ..., 2 are proportional to x”. In general, n = 1 and the contact of the tangent with the curve is said to be of the Jirst order; for n > 1, the contact is of the n™ order. If the equations of the curve are in terms of a general parameter ¢, as (2.1), we have sinee tis a function of ¢ dz’ _ da’ dt a _ bat (“y dx’ &t 8) Eade &~ ele) tae Hence the tangent at a point t for which ¢ satisfies for some value of n (> 1), if any, the equations do da da a cz dct a (8.8) has contact of the n** order. By definition the osculating plane of a curve at a point P(x’) is the limiting position of the plane determined by the tangent at P und a point P of the curve as P approaches P along the curve. Since the plane passes through P its equation is of the form (8.9) a(X' — 2!) = where a; being direction numbers of the normal to the plane must be such that (3.10) Equations (3.9) and (3.10) express the condition that the tangent at P lies in the plane.* Substituting in (8.9) for X* the expressions (2.7) for z' and making use of (3.10) we obtain Lier Lae = (3.1) ai(§e tye et ) =0. ‘As P approaches P, that is, as . approaches zero, we have in the limit ” @.12) ag’ = 0. +O. G, p. 120, 14 CURVES IN SPACK (Cu. T Jn order that equations (8.9), (8.10) and (8.12) be satisfied by a’s not all zero, we must have [xiao Xt w Xt (3.13) | e a | = em(X! — 2a’ at” which is an equation of the osculating plane at the point 2°. When the curve is defined by (2.1) in terms of a general parameter ¢, it follows from (3.7) that an equation of the osculating plane is Ivacd vag YAoe | | ae! ac a | (3.14) | @ a @ i= | é2 dt dst | | dé dt dé® | If the tangent at a point has contact of higher order than the first, equation (3.14) is satisfied ideutieally. In this ease from (3.11) and (3.7) it follows that an equation of the osculating plane at a point for which the tangent has contact of order n — 1 is Xi-g a When a curve is plane and its plane is taken for the plane 2° = 0, equation (3.13) is equivalent to 2° = 0, that is, the osculating plane of a plane curve is the plane of the curve. Conversely, when all the oseulating planes of a curve coincide, the curve is a plane curve since all of the points of the curve lie in this plane. EXERCISES 1, The eurve t=eacost, 2=asint, = bt lies on a circular cylinder (see Fig. 2); find the direction cosines of the tangents to the eurve and show that the tangent makes aconstant angle with the generators of the cylinder; the curve is called a "C.G., pp. 115, 121, $3] OSCULATING PLANE 15 Fic. 2, Cireular helix nition a cylindrical helix is a curve lying on a cylinder and which generators under the same angle; if this constant angle is denoted It, oo), =f, Be cote f VGRFFP are parametric equations of a cylindrical helix; is any cylindrical helix so defined? 3. By definition a conical helix is a curve lying on a cone which meets all the generators of the cone under the same angle; if this angle is denoted by 0, x! = f*(t) are equations of @ conical helix, if the fauctions * satisfy the conditions anf) + aa(f)? + a(f)? = 0, Dar = cso V OFF VErr where the a’s are constants not all of the same sign; is any conical helix so defined? 16 CURVES IN SPACE (Cx. 1 4, Find an equation of the oseulating plane of the twisted cubie (1.17) and compare the result with Ex. 3 of §1. 5. The distance of a point P on a curve from the osculating plane at a nearby point P is of the third order at least in the are PP; and for any other plane through P not containing the tangent at P the distance is of the first order; discuss the case of planes containing the tangent at P. 6. Phe curve with the equations BaZa ey" + UO - 50, am Sa ayy - HO +10, yy 20, ‘f() is any funetion of ¢, and primes indicate differentiation with respect to the parameter f, is a minimal enrve, and any minimal eurve is 80 defined; discuss the case when /() = ct? + ext -+ es, where the c’s aro constants, 7. If at every point of a curve the tangent has contact of the second order with the curve, the latter is a straight line, 8. In terms of the are ¢ a8 paranreter equations of the circular helix (Ex. 1) are e 1 z= a cos eeasin Vere’ tach oseulating plane of the helix aneets the circular cylinder on which it lies in an ellipse. 9, The curve seasint(, z= asintcoal, zt = acost is a spherical curve, that is, lies on a sphere; its normal planes pass through the center of the sphere; the curve has a double point at (a, 0, 0), and the tangents to the curve at this point are perpendicular to one another. 10. Find an cquation of the osculating plane of the curve si=acost+bsint, st=asint+beost, 2° = esin 2; find also two equations of the form (1.9) as equations of the eurve. 4, CURVATURE. PRINCIPAL NORMAL. CIRCLE OF CURVATURE Let P and P be two points on a curve C, As the length of the are between these points, and A@ the angle of the tangents at P and P, that is, the angle between two half-lines through any point and having as As ap- the positive senses of the two tangents. ‘The limit of |x proaches zero, measures the rate of change of the direction of the tangent at P. This limiting value, denoted by «, is called the curvature of C at P, and its reciprocal, denoted by p, the radius of curvature; from 54} CURVATURE. PRINCIPAL NORMAL 7 their definition it follows that « and p are non-negative. When C is a plane curve, the definition of curvature here given is that usually given in the differential calculus. In order to derive in terms of s an expression for « in terms of the quantities defining a curve C, we consider the auxiliary curve P with the equations ax 4.1 xia", an) ds which in consequence of (2.8) is a curve upon the sphere of unit radius with center at the origin. The radius of the sphere at any point of T is parallel to the tangent to C at the corresponding point, that: is, the point with this tangent. The curve I’ is called the spherical indicatriz of the tangents to C. If we denote by o the are of I', it follows from (4.1) and (2.4) that ex bx r_pdede' is (4.2) do’ = x de age From the definition of x we have reece Since Ad is the length of the are of the great circle between the points on the unit, sphere corresponding to P and P on C, the limit of 2 is o unity in consequence of the result at the close of §2 and the fact that A@ and Ao have a common chord. Consequently as follows from (4.2) Saree 7 dst de" For a straight line, with equations (1.1) in which tis the are, one has x = 0, which is evident also geometrically from the fact that the tangent to a straight line at cach point is the line itself. In order to obtain the expression for « when the equations of the curve are in terms of a general parameter f, we observe that from (2.3) we have aii ate (4.4) bo? ie ov (4.3) K where primes denote differentiation with respect to é, and from (2.2) (4.5) HEP lel = EHH 18 CURVES IN SPACE (Cu. f the second following from the differentiation of the first. Substituting in (4.3) from (3.7) and making use of (4.4) and (4.5), we obtain (4.6) From (2.9) it follows that the line through the point 2‘ of a curve ae and with direction numbers a is perpendicular to the tangent at the point, and thus is one of an endless number of normals to the curve at the point. If we define quantities 6° by da _ as (4.7) “aa 7 it follows from (4.8) that g' are direction cosines of the positive sense of this normal.* Its equations are (4.8) X= att Bd, where d denotes the distance of the point X‘ from the point 2° of the curve. This normal is called the principal normal of the curve at the point. When the expressions (4.8) are substituted in the equation (3.13) of the osculating plane, the equation is satisfied for all values of d, in consequence of (4.7), that is, the principal normal at a point lies in the osculating plane at the point. Hence the osculating plane at a point of a skew curve is the plane determined by the tangent and principal normal of the curve at the point. The cirelc in the oxeulating plane with center at the point 4s) Xadapiadsly and of radius p is called the cirele of curvature of the curve at the point a‘ and its center the center of curvature of the curve for the point 2". Evidently this cirele and the curve have a common tangent at 2‘, EXERCISE 1, When a curve is defined in terms of a general parameter ¢ by (1.10), the direction cosines 6° of the principal normal are given by B= Fh ~ 99"), where ¢(#) is defined by (2.3), and primes denote differentiation with respect to t. * See the statement about positive sense after equations (3.1). $1 BINORMAL. TORSION 19 2. For a cylindrical helix, as defined in §3, Ex. 2, oO = cso VP PEP, B= 0, PR), where ¢ is -+1 or —1 20 that « is positive Let P be a point of a curve; a cirele C with a common tangent to the curve at P is determined by requixing that it pass through another point Q of the curve; the limiting cirele as Q approaches P along the eurve is the circle of curva- ture of the given curve at P. 4, Find the function (0) so that the eurve a =feo sintd, ats feo costd, = feo tantdt shall be a curve of constant curvature. 5. Determine the form of the function ¢(t) so that the principal normals to the curve a Basin, 2 = lt) are parallel to the 2tz*plane. 6. The circle of curvature of a curve at a point of the curve has contact of the second order with the curve; every other circle whieh lies in the osculating plane and is tangent to the curve has contact of the first order; accordingly a circle of curvature is called an osculating circle of the curve. 7. Find equations of the surface consisting of the principal normals of a eireu- lar helix (see §3, Ex. 1), and show that the locus of the center of curvature is a eireutar helix. 8, Find the coordinates of the center of curvature of the curve zt=acos, at=asin’, 2? = acos2t. 5. BINORMAL. TORSION ‘The normal to a curve at a point P which is normal to the osculating plane at P is called the binormal at P. Evidently it is perpendicular to the tangent and to the principal normal at P, From (8.13) it follows that. 20 CURVES IN SPACE {Ca. 1 are direction numbers of the binormal. In order to find direction co- sines of the binormal we make use of the identity [pe Dee [Eee Dae” (1) | * ‘The reader should verify that this is an identity whatever be the quanti- ties involved without any use of the primes as indicating derivatives. When in particular the quantities x" and 2*”’ have the meaning ascribed to them in §2, it follows from (2.8), (2.9) and (4.3) that the left-band member of (5.1) is equal to x’. Hence direction cosines 7‘ of the bi- normal and the positive sense along the latter are defined by p(n a ~ ®), Yea eo = pla” a” — a 2"), (5.2) These expressions may be written in the form (6.3) vf = ole” ah” — 2" a), with the understanding that i,j, & take the values 1, 2, 3 cyclically. Hence equations of the binormal are (5.4) Xf=ai ted. The significance of the choice of sign in (5.2) is seen when we observe that the expressions (3.2), (4.7) and (5.2) for a’, 6’, and y‘ respectively are such that jo oe | (5.5) ie 6? 6) = 41, | 7 Fy | as is readily verified sinee the right-hand member of (5.1) is equal to «°. The result, (5.5) means that the positive directions of the Langent, principal normal and binormal of a curve at cach point of a eurve have the mutual orientation of the 2’, 2°= and z-axes respectively (Fig. 3).* *C.G, p. 162. $5] BINORMAL. TORSION 2 From the equations Das! ZDely' = 0, Lsiy' = 0, yy De® a1, De and from (5.5) it follows that, each element in the determinant equation (5.5) is equal to its cofactor.* ‘This result may be written (5.7) af = Bly — By’, Bi = yok — va’, 7! = a’pt — 6’, as i, j,k take the values 1, 2, 3 cyclically. (5.6) 1 Fra. 3 From the definition of the binormal it follows that the binormals of a plane curve are the normals to the plane at points of the curve, and consequently have the same direction at all points of the curve. For a skew curve the direction of the binormal changes. If A@ is the angle of the positive directions of the binormals at two points of parameters sand s + As, the limit of # as As approaches zero measures the rate 3 ‘ of change of the direction of the binormal at the point of parameter s, and consequently the rate of change of the orientation of the osculating *C.G, p. 161, 22 CURVES IN SPACE (Ca. 1 plane. This limit is called the torsion of the curve and is denoted by r. Sometimes the curvature, as defined in §4, and the torsion are called the first and second curvatures respectively of the curve. In order to obtain an expression for 7, we introduce the spherical indicatrix of the binormal, that is, the curve defined by (5.8) Xay. Evidently this is a curve upon the unit sphere with center at the origin, and such that the radius of the sphere to any point of the curve is parallel to the positive binormal to the given eurve at the point with the same value of the parameter s. The linear element of the indicatrix je given by aw=>D dy! ay ag. w ds ds By an argument similar to that used in §4 we have 5 rapa (5.9) HL In order to find expressions for a, we differentiate with respect to # the equations vy = 4, Lay =0, and obtain 61) Deo, L(e% + 7%) =0. From (4.7) and (8.2) we have (5.11) ce = from which and the third of (5.6) it follows that the second of (5.10) reduces to Let =0. From this equation and the first of (5,10) we hav in consequence of the second set of equations (5.7) that es proportional to 8’, and from (5.9) that the factor of proportionality *C.G,, p. 104. $5] BINORMAL. TORSION 23 is ror —r. Thus far ris defined by (5.9) to within sign; we choose the sign so that we have ay’ (6.12) a We are now in position to obtain an expression for in terms of the derivatives of 2°. In fact, if we differentiate (5.3) with respect to s, the result may be written in consequence of (5.12) (5.13) From the definition of torsion it follows that 7 is zero for a planc curve. Conversely, if 7 is zero at every point of a curve, the latter is plane in accordance with theorem [1.1] and equation (5.13). At points of a curve, if any, for which the determinant in (5.13) is zero the osculat- ing plane is said to be stationary. EXERCISES 1, When a curve is defined in terms of a gencral parameter ¢ by (1.10), the direction cosines ¥# of the hinormal are given by oa Fe AF), e as i, j, & take the values 1, 2, 3 cyclically, where y(t) is defined by (2.8) and primes denote differentiation with respect to (; also the torsion of the curve is given by pop op venga wre pen pee ga 2. ‘The curvature and torsion of a circular helix, ae defined in §3, Ex. 1, are constants, namely a -b Tate e+e 24 CURVES IN SPACE (Ca. I 3. For a cylindrical helix, as defined in §3, Ex, 2, 0b (8H = py ee (PRE from which result and that of §4, Ex. 2 it follows that. 7 = —ex cot 8, that ia, «/r is a constant. 4, Find the curvature and torsion of the eurve mae, taet stm /2t. 5, The curvature and torsion of the curve Pea@-f), a3, = atte) are given by eer Ball + 6. Find the points of the curve of §3, Ex, 9 at which the torsion is equal to zero. 7. When two curves are symmetric with respect to a point, or a plane, their curvatures at corresponding points are equal and their torsions differ in sign. 8. A necessary and sufficient condition that the cirele of curvature have con- tact of the third order with the curve at a point is that at the point 7 = 0, * =o (sce $4, Bx. 6); at such a point the circle is said to euperosculate the curve. 9. If @ and y are the angles made with a fixed line in space by the tangent and binormal respectively of a curve, sineds _ « singde + 40. A necessary and sufficient condition that the principal normais of a curve are parallel to a fixed plane is that the curve be a eylindrical helix. Ll, If the eurve a¥(s) is a cylindrical helix so also is the curve with the equations xtm eat — fa, 12. ‘The curvature and torsion of the curve a = [1 sintd, = fro costdt, af = frovo a are given by 1_wity ""Fisete from this result it follows that the curves for which « or 7 is a constant can be found by quadratures, 36) THE FRENET FORMULAS 25 6. THE FRENET FORMULAS. THE FORM OF A CURVE IN THE NEIGHBORHOOD OF A POINT Tf the second set: of equations (5.7), that is, 8 as i, j, k take the values 1, 2, 3 cyclically, are differentiated with respect, to s, the result is reducible by (5.11), (5.12) and (5.7) to a8 = seyigh — ah? , ‘ OE = aty’s* — v6") + olp'a! — sta’) = —(ca! +7), Gathering together this result, (5.11), and (5.12), we have the following set of equations fundamental in the theory of skew curves and called the Frenet formulas: dat ‘ dg* @y Saw, B= Go! +279, wal — bol ‘ or = Bt, On replacing A* in the second set of equations (6.1) by | 2” (see equations (4.7)), the resulting equations are reducible to (6.2) x Wat ae Bi wey’, Differentiating with respect to ¢ and making use of (6.1), we obtain gue f. (6a) 2 = 3 oF & ) et - [é (er) + aly We observed following equation (6.6) that the positive directions of the tangent, principal normal and binormal at each point of a curve have the mutual orientation of the coordinate axes. If then we take for coordinate axes these lines at a point Py of a curve and measure the arc from the point, we have at Po 4) a=, p=, v=, where 8 are Kronecker deltas defined by (6.5) af = 1 or 0 according as ¢ = j ori ¥ j. By Maclaurin’s theorem we have that the coordinates x‘ of any point on the curve are given by (66) a= Oe HR oF + EC st ER sth, 8 i 26 CURVES IN SPACE [Cx. I where a subscript zero indicates the value of the quantity at Py. From (8.2), (4.7) and (6.4) we have @") = 81, @ Oy = abs From these results, (6.2), and (6.3) we have from (6.6) for this choice of coordinate axes 2 —Mg 1 (du) ee wa po tives 2 koe de\ o, Mfde_ os _ a) (6.7) roses (Seta x whey It follows from these equations that in the neighborhood of a point, if any, at which x = 0 the curve approximates a straight line. Also if « # 0 the curve with s increasing crosses the oseulating plane at the point, from the positive to the negative side when 7 > 0 and vice-versa when + < 0; in the former case the curve is said to be left-handed and in the latter right-handed at the point. At a stationary point, that is, when 7 = 0, the curve remains on the same side of the osculating plane in the neighborhood of the point (provided dz/ds # 0), since in this case the sign of 2° does not change with s for sufficiently small values of s. These each of equatio esults follow in fact when we consider only the first terms in s (6.7), that is, the approximate curve xoT0 3 a8 (6.8) 2 in which xp and 79 are constants. ‘This curve is a twisted cubic, whose projections upon the coordinate planes are shown in Fig. 1, the a’-, a, and 2'-axes being respectively the tangent, principal normal, and binormal of the curve at the point of the given curve which is the origin of the coordinate system used. The coordinates X* of a point in the osculating plane to a twisted curve are given by (6.9) X= a! + ua’ + v8! for suitable values of u and v. We raise the question of determining wand v as functions of sso that the locus of the points of coordinates 46) THE FORM OF A CURVE 7 X* given by (6.9) shall be an orthogonal trajectory of the osculating planes, Differentiating (6.9) with respect to s and making use of (6.1), we obtain . IX! . . . Since o are direction numbers of the tangent to the desired locus, as ax‘ , : A wand v must be such that “ Ge we proportional to +’, that is, the co- efficients of a’ and 8‘ must be zero. If we introduce the parameter ¢, defined by few this gives the following conditions to be satisfied: dui _ at, tun. (6.10) Differentiating the second of equations (6.10) with respeet to ¢ and substituting from the first of (6.10), we obtain @o 1 (6.11) oe From the theory of linear ordinary differential equations it follows that the general solution of equation (6.11) may be obtained by quadratures. When such a solution has been obtained and substituted in the second of equations (6.10), w is given directly. Hence we have [6.1] The orthogonal trajectories of the osculating planes of a skew curve can be obtained by quadratures, EXERCISES 1, For a plane curve the Freuet formulas are da’ agi i a Gg as @ = 1,2), and equations of the curve are (i) ge fcoseas, ae in ¢ ds, 28 CURVES IN SPACE (Cu. 1 v= fet from equations (i) it follows that ¢ is the angle which the tangent to the curve makes with the z-axis. 2. When all the osculating planes of a curve have a point in common, the curve is plane. 3, The locus of the centers of curvature of a twisted curve of constant curva- ture is an orthogonal trajectory of the osculating planes of the curve, and is a curve of constant curvature. 4. A tangent to the locus of the centers of cur perpendicular to the corresponding tangent to C. where ture of a twisted curve C is it coincides with, or is per pondicular to, the principal normal to ( only at points for which + = 0, okt =0. 5. When for a cylindrical helix (sec §3, Ex. 2) the generators of the cylinder are parallel to the z-axis, then a? = cos @, and from the Frenet formulas it follows that. ae f=5, “=a @ w=0, Ifa function ¢ is defined by a! = sin cose, a? = sin Osin ¢, then ioe Bi=-esinc 6 =ecosc, x= esind—, ds where ¢ is +1 or —1 so that « is positive, and yeesin’, 7 = —excot é (see §5, Ex. 3). 6, For a curve for which 7/x = ¢, where c is a constant, it follows from the Frenet formulas that yi = cat +5, where bf are constants, from which it follows that 2 ab’ = const., that is, the curve makes a constant angle with the lines of direction numbers bf, and hence is a cylindrical helix. 7. ‘The equations Hema f (iat — ter as i, j, k take the values 1, 2, 3 cyclically, where a is a constant, and 7! are fune- tions of a single parameter such that 2 7'y! = 1, are equations of a eurve of torsion 1/a and ‘are direction cosines of the binormal. Does it follow from §6) THE FRENET FORMULAS 29 this result that any curve on the unit sphere can serve as the spherical indicatrix of the binormals of a curve of constant torsion? 8. If Cis a curve of constant torsion, for the associated curve with the equa- tions: Es ftas 7 the curvature is constant. 9. When two twisted curves are in one-to-one correspondence with tangents at corresponding points parallel, the principal normals at corresponding points are parallel, and also the binormals; two curves so related are said to be deducible from one another by a tranaformation of Comboseure. 10. The equations @ B= a tala, where a isa constant, are equations of a curve € whose points are on the tangents to the curve a‘(s) and at the constant distance a from the corresponding points of contact; the arc 8, the direction cosines a‘ of the tangent, and the curvature % of C are given by is [ Viera, ae #4 Vitae’ o(iteer! y pot, ee a+ ae Q+aee)’ the tangents to @ are parallel to the corresponding osculating planes of the given curve. _ 11. In order that the curve Cin Ex. 10 be a straight line it is necessary and 1 7 sufficient that + = 0 and a? + — = ce*/*, where c is an arbitrary constant, If we put c= at, we have 1 vo fede =} [pte = corte From this result and Ex. 1 we have geal, gta {vi — ele ds, from which and equation (i) of Ex. 10 for i = 1 we have that the locus @ is the axis. The curve is called the (ractriz, In terms of ¢, the angle which the tangent makes with the z-axis, equations of the curve are sin? al= —acose, Ae de = o + tan e) — sino] 08 @, eT cog g = 7 Hog (s00 «+ tan ) — sin o} 30 CURVES IN SPACE [Ca 1 12, A curve whose principal normals are the principal normals of another curve is called a Bertrand curve; if x‘ and 2 are the coordinates of corresponding points on the respective curves C and @ and s and 3 corresponding arcs, one has a= ot + Apt, a aah = (1 th) at + pF — rho, m7 ) af + of — thy Since #* = ef" by hypothesis, where ¢ is +1 or ~1, it follows that A is a constant; denoting by & the angle between the osculating planes of C and its conjugate C, one has a’ = cos w a! + sin w 74; 3! = ¢ (~sin wa! + cos w 74); from the Frenet formulas for @ it follows that « is a constant and ds hr as ~ sine Also % and 7 for __sinto we Se, thus ¢ is to be chosen so that & is non-negative. 13. A curve for which ax+br=4, here a and } are constants different from zero, is a Bertrand curve; the equa- tions (see §5, Ex. 12) ne fo sintd, = fro costd, t= [ov a, where ¥(¢) is any function of ¢ and Maas /OE,, Vie ee a+ ee and b being constants different from zero, are equations of a Bertrand curve. 14, A circular hetix is a Bertrand curve; it has an infinite number of conju- gates, each lying on @ circular cylinder with the same axis as that of the given helix, 15, A nocessary and sufficient condition that the osculating planes of a Bert- rand curve and of its conjugate coincide is that the curve be a plane curve; any curve parallel to the given curve is a conjugate curve. 16. If Cis a curve of constant torsion, the curve with equations sa atso(E fora), where a and b are constants, is a Bertrand eurve. 17. The binormals of a curve are the binormals of another curve, if and only if the given curve is plane. $7] INTRINSIC EQUATIONS OF A CURVE 31 18, In order that the principal normals of a curve C be the binormals of a curve G, it is necessary and sufficient that @ = a(n? + 7), where a is a constant; then equations of C are B= i + apt; curves C satisfying the condition (i) can be found by quadratures (sce §5, Ex. 12) 7, INTRINSIC EQUATIONS OF A CURVE When equations (6.3) are differentiated successively with respect to sand in each case the derivatives of a‘, g‘, and 7! are replaced by their expressions from the Frenet formulas (6.1), we find that each derivative of x! is expressible linearly and homogeneously in a, 8’, and y‘, the coefficients being functions of x, 7, and their derivatives of various orders with respect to s. Consequently the coefficients of further terms in (6.7) as derived from (6.6) involve only the values of x, 7, and their derivatives for s = 0, because of the particular values (6.4) at the origin in the coordinate system used. Hence, if for two curves the functions x and 7 of sare the same functions, the expressions for 2’ for each curve relative to the axes consisting of the tangent, principal normal, and binormal of each curve at the point s = 0 are the same. Since either set of axes can be brought into coincidence with the other by a rigid motion, we have: [7.1] Two curves whose curvature and torsion are the same functions re- spectively of the are are congruent. From this it follows that a curve is determined to within its position in space by the expressions for « and 7 in terms of s. Consequently (7.1) K=fi@), 7 = fle) are equations of the curve. Since they are independent of the coordin- ate system used, they are called intrinsic equations of the curve. From the manner in which equations (6.7) were obtained, it follows that « and 7 derived from these equations by means of (4.3) and (5,13) are power series in s, the coefficients being values of x, 7, and their derivatives evaluated for s = 0. Consequently if we have any two equations (7.1) in which f,(s) is a non-negative function of s, the cor- responding equations (6.7) are equations of the curve for which (7.1) are the intrinsic equations. Although this method of obtaining x‘ as functions of s gives the equations of a curve for given functions fy and 32 CURVES IN SPACE iC. [ fx in equations (7.1), it gives these equations as infinite series, We shall consider another approach to this problem which may in certain cases lead to finite expressions for 2‘. ‘Dhe three sets of quantities a’, 6°, and y' as 7 takes the values 1, 2, 3 are seen from (6.1) to be solutions of the following system of ordinary differential equations where « and 7 are given functions of 8: dw _ ds If we have three sets of solutions, u’, v’, w'; u’, v°, ws u,v’, w° of these equations, which may be denoted by u’, o', w‘ for i = 1, 2, 3, and put. (7.8) wid + o'v’ + w'w! = c? Gj = 12,3) we find by differentiation that in consequence of (7.2) the c’s are con- stants. We define quantities by ~ = — (ku + rw), 0. du (7.2) be aiw", where the a’s are constants, and seek under what conditions these a’s can be chosen so that (7A) esau, pi=ap', (7.5) wal + pip’ + 7'y’ = 34 (7 = 1, 2,3), where (7.6) = 1 or O according as i = j ori # j. ‘This choice is made in order tbat a’, 8°, and 7’ shalll be direction cosines of three mutualy perpendicular vectors. Substituting from (7.4) in (7.5) and making use of (7.3) we have (7.7) ajaic™ = If the determinant | o | is different from zero, 44 cam = 0 is an equation in homogeneous coordinates 21 , 2 , 3 of a non-degenerate conic in the plane. With respect to this conic a}, a}, a3; a1, a3, a3; a}, a3, a§ satisfying (7.7) are the coordinates of the vertices of a sclf- polar triangle, and consequently an endless number of sets of a} sati fying (7.7) can be found.* For such a set of af the quantities a4, 84, y! defined by (7.4) are solutions of equations (7.2), that is, we have equa- tions (6.1), the signs of a} having been chosen so that equation (5.5) holds. * Veblen and Young, 1910, 1, p. 282. §7] INTRINSIC EQUATIONS OF A CURVE 33 If then we define z* by (78) aa [tas for the curve so defined s is the are and a’ direction cosines of the tan- gent. Also from (7.8) and (6.1) we obtain by differentiation equations (4.7) and (6.2), from which it follows that « and 7 arc the curvature and torsion of the curve, and 8° and 7‘ direction cosines of the principal normal and binormal respectively. Thus we have shown that three sets of solutions of equations (7.2) for given functions « and 7 of s lead by quadratures (7.8) to a curve for which x and 7 are the curvature and torsion, provided that the de- terminant | c'’ | is not equal to zero. That there are sets of solutions of equations (7.2) satistying this condition follows from the theory of such sets of equations, namely that there exists a unique solution for a given set of initial values of u, v and w.* Since ec” are constants, it follows that one has only to choose the initial values of the three sets of solu- tions so that it shall follow from (7.3) that the determinant of the c’s is not equal to zero. EXERCISES 1. A solution of the equation constitute a solution of equations (7.2); and any solution of (7.2) is expressed in terms of three sets of solutions by (7.4) for suitable values of the constants aj - 2. If u, v, w are solutions of equations (7.2) such that u? +o? + wt = 1, the quantities ¢ and w defined by uti itv, l-w u-i are solutions of the Riceati equation *Goursat, 1924, 1, vol. 2 p. 368. 34 CURVES IX SPACE (Cu 1 3. The general integral of an equation of Riceati dt ° + M0 -+ Ne, ds where L, M, N are functions of s, is of the form where a is an arbitrary constant, and P, Q, R, S are functions of 8, 4. From theorem [7.1] and §5 Ex. 2 it follows that a necessary and sufficient condition that a curve be a circ helix is that its eurvature and torsion be constant; show also by means of Ex. 1 that this condition is sufficient. 5, Establish the statement made about the number of solutions of equations (7.7) by purely algebraic methods, 8. INVOLUTES AND EVOLUTES OF A CURVE As shown in §3 the equations @.1) X= a + ua’ are parametric equations of the tangents to the curve C defined by 2‘ as functions of the are s. For a particular tangent u is the distance between the points 2’ and X". If wis replaced in equations (8.1) by a function of 3, the resulting equations are equations of a curve I whose points lie on the tangents to the given curve, Differentiating equation (8.1) with respect to s, one has in consequence of (3.2) and (6,1) ‘ (2) =. (: + ) af + aus’. Since i are direction numbers of the tangent to I, if the latter curve Is is to be such that its tangent at each point is perpendicular to the tangent to C through the point, it is necessary and sufficient that (8.3) from which it follows that (8.4) users $8) INVOLUTES OF A CURVE 35 where c is an arbitrary constant. Hence there is an infinity of such curves I, each defined by 5) Xt aah t ( — s)n! for a particular value of c. They are called the involutes of the given curve. From (8.4) it follows that the length of the segment of any tangent to the curve determined by two involutes has the same value, the difference of the c’s of the two involutes. Fre. 4. Involute of the circular helix of Fig. 2 When a curve C is defined in terms of a general parameter ¢, the determination of s requires a quadrature (2.2), and then the involutes are given dircetly by (8.5). An involute when c — sis positive may be described mechanically as follows: Take a string of length ¢, fasten one end at the point of the curve for which s = 0 and bring the string into coincidence with the curve; when the string is unwound from the curve and is kept taut, the other end point describes the involute as is seen from (8.5). From (8.2) and (8.4) we have ax’ ds (8.6) = (c — 8)xs*. Hence we have [8.1] A curve has an infinity of involutes; a tangent to an involute at a Point X* is parallel to the principal normal to the curve at the corresponding Point x, 36 CURVES IN SPACE {Cx. 1 If F is an involute of a curve C, we say that C is an evoluie of I’, that is, it is a curve whose tangents are normal tol’, Suppose then that we start with a curve of coordinates z', and seck its evolutes. Since the points of an evolute lie in the normal planes to the curve, its equations are of the form 7) Xi = af + upi + oy, where w and v are functions of s to be determined. From these equa- tions in consequence of (3.2) and (6.1) we have by differentiation with respect to s aX yy (du sy (dv : es) mae t (4 or ot (= ur) Since these quantities are direction numbers of the tangent to an evolute, they must be proportional to X* — 2‘, that is, to u8* + vy’, which as follows from (8.7) are direction numbers of the line joining the points X‘ and x‘, Consequently we must have wu = 1/« and du a o (+0) =u(Z-w). When this equation is written in the form we sce that its integral is (8.9) 2 = tanto), where by definition (8.10) o= f 1s, and cis an arbitrary constant, Substituting these results in (8.7), we obtain (gu) xiad4l@itionw@tor. For each value of ¢ these are equations of an evolate, Consequently @ curve has an infinity of evolutes. Since the curve is an involute of each evolute, we have $8) EVOLUTES OF A CURVE 37 [8.2] A curve C has an infinity of evolutes; the principal normal to an evolute is parallel to the tangent to C at the corresponding point. From equations (8.11) and (4.9) it follows that the points of all of these evolutes corresponding to a given point on the given curve lie on the line, called the polar line, parallel to the binormal and through the center of curvature for the given point on the curve. Moreover from (8.9) it follows that » + ¢ is the angle which the line joining a point on the curve to the corresponding point on an evolute makes with the osculating plane of the curve at this point, Hence we have 18.3] When each of the normals to a curve C which are tangent to an evolute 8 turned through the same angle about the corresponding tangent to C, the normals in their new position are tangent to another evolute of C. From (8,11) and (8.10) it follows that it is possible to choose ¢ so that the points of the corresponding evolute lie in the osculating planes only in ease w is a constant, in which ease + = 0, that is, when the curve is plane. In this exceptional case this evolute is the locus of the centers of the circle of curvature and is a plane curve, except when the given curve is a circle, in which case the locus is the center of the circle. This evolute is the one which in the differential calculus is called the evolute of the plane curve. However, a plane eurve has an infinity of evolutes, as ¢ in equations (8.11) takes all possible values. From the form of these equations it follows that these evulutes lie on the cylinder whose generators are the normals to the plane of the given curve at points of the evolute in the plane, that is, the evolute for which e = —w. Moreover, from the remark preceding theorem [8.3] it follows that the tangents to each evolute make constant angles with the generators of this cylinder and consequently these evolutes are cylindrical helices (see §3 Ex. 2). Hence we have [8.4] A plane curve other than a circle has an infinity of evolutes, each of which is a helix of the cylinder whose right section by the plane of the curve is the plane evolute of the curve. EXERCISES 1. The involutes of a circular helix (§8, Exs. 1, 8) are plane curves, which also are involutes of circular sections of the circular cylinder upon which the helix lies, 2. For an involute (8.5) for which ¢ — 8 is positive the are, direction cosines of the tangent, principal normal, binormal, and the eurvature and torsion are, 38 CURVES IN SPACE (Cx. 1 given by G-oe te 3. A necessary and sufficient condition that the involutes of a twisted curve be plane curves is that the ourve be a cylindrical helix (see §6, Ex. 6). 4, Find the evolutes of the curves of §5, Exs. 4 aud 5. 5. For an evolute (8.11) the are, direction cosines of the tangent, principal normal, and binormal, and the curvature and torsion are given by t= [[EQ)+ieera}eeroa, af = cos (w + o)pt + sin (w+ ody’, BF = - eal, #8 = el— sin (w+ opt + 003 (w+ e)r4, ex cos (w+ ¢), — «sin (w +e) see (w+) [i () + tan weal ds \« x where ¢ is -+1 or ~1 80 that Xis positive. 9. THE TANGENT SURFACE OF A CURVE, THE POLAR SURFACE. OSCULATING SPHERE When for any curve the two parameters s and u are eliminated from the three parametric equations of its tangents, namely (1) 2i(s) + uai(s), we obtain a single equation in the X'. Consequently (see §1) the locus of points on the tangents to a curve is a surface, called the tangent surface of the curve, and each of the tangent lines is called a generator of the surface, When the curve is defined in terms of a general param- eter ¢ equations of the tangent surface are Ld at’ where now u is not the distance of the point X* from the point 2° as it is in (9.1). (9.2) Xec()tu 99] TANGENT SURFACE OF A CURVE 39 When s and w in (9.1) are given particular values, equations (9.1) give the coordinates of a point on the surface. ‘Thus the locus is two dimensional, which is another proof that the locus is a surface. Tf in (9.1) we replace w by 2 function of s, say g(s), the resulting equations are parametric equations of a curve on the surface. In particular, if We put «= ¢ — s, where c is a constant, then, as follows from §8, the curve for each valuc of ¢ is an involute of the given curve. Consequently Fra. . Tangent surface of the cireular helix of Fig. 2 all the involutes of a curve lic on its tangent surface. They are the curves which intersect the generators of the surface at right angles, that is, the involutes are the orthogonal trajectories of the generators. Since there is only one set. of orthogonal trajectories of a set of lines, we have [9.1] The orthogonal trajectories of the generators of the tangent surface of 4 curve are the involutes of the curve. According as u in equation (9.1) has a positive or negative value the Point lies on the portion of the tangent drawn in the positive direction 40 CURVES IN SPACE [Cu 1 from a point on the curve or in the opposite direction. Hence the surface consists of two parts, or sheets, one part consisting of ail the points for which « 2 0, the other of the points for which u $ 0. Thus the curve forms a common boundary of the two sheets. In order to get an idea of the form of the surface in the neighborhood of the curve, we recall from §6 that in the neighborhood of the point Py (s = 0) the curve approximates the twisted eubic (9.3) where « and 7» are the curvature and torsion of the given curve at the point Po, the tangent, principal normal, and binormal at Py being the coordinate axes. Noting that s is the are of the given curve but not of the cubic, we have that equations of the tangent surface to the curve (9.3) are se ‘Ko T Gl @4) Nasty Ma o(§ + su), Xa (5 + eo) In this coordinate system the plane z' = 0 is the plane normal to the given curve and to the cubie at Pp, and cuts this tangent surface in the curve F for which u = —s. From the second and third of equa- tions (9.4) it follows that equations of this plane section are 2 : xt = THe xt = MT0S" 2° 3 On eliminating s from the second and third equations we sec that the curve is a. semi-cubical parabola, with the negative half of the principal normal for cuspidal tangent. Since this is the ease at every ordinary point of the curve, that is, every point at which neither « nor 7 is zero, we have P 4 [9.2] The tangent surface of a curve consists of two sheets which are tangent to one another along the curve, and thus form a sharp edge, namely the curve. The eurve is called the edge of regression of the surface. An iden of the form of the surface in the neighborhood of the eurve may be had from Figs. 5 and 6. ‘The line with the equations (9.5) X= 2 + p8' + wy" for each value of s is the polar line, as defined in §8, corresponding to the point 2” on the curve for this value of s, and the parameter a is the 99] THE POLAR SURFACE 41 distance of a. point on the line from the corresponding center of curva- ture, namely the point 2‘ + 8%. Since equations (9.5) involve two parameters, s and u, it follows that the totality of the polar lines of a curve constitute a surface; it is called the polar surface of the curve. We shall show that this surface is the tangent surface of another curve defined by (9.5) when w is replaced by a suitable function of s. Fra. 6 If wis any function of s, we have from (9.5) in consequence of (3.2) and (6.1) dx? ‘dp i du i 6 dp a (9.6) ds (B+ e+ as oT )Y ‘These ure direction numbers of the tangent to the curve (9.5) for wa given funtion of s. If the polar line with respect to the given eurve is to be the tangent to the curve (9.5) for w equal to some function of s, the direction numbers must. be proportional toy’, This requirement is ldp satisfied, if and only if u = — as” Hence we have 9.3] The polar surface of a curve is the tangent surface of the curve with the equations (9.7) Xoo te -_7y Consider the sphere 2 with center at the point of coordinates (9.7) for a given value of s and passing through the corresponding point 2° on the curve, that is, the point for this value of s. Evidently the eirele of curvature for the point 2 lies on 2, and the square of radius R of 3 2 isp? + ¢: a) . We shall show that ¥ has contact of the third order with the curve at P, that is, if R + 6 denotes the distance between the center of © and a point P on the curve such that the are PP is s, then 5 is of the order of s*. In order to establish this result we make use of equations (6.7). The center of the sphere corresponding to the point {Cx. CURVES IN SPACE the polar developable with normal pla Fig. 7, A twisted curve so] OSCULATING SPHERE 43 of the curve at the origin of the coordinate system used in (6.7) is at ‘ 1dp the point (0, ert *). Hence we have (Reh ma 4 a4 (“18 rds On substituting for 2 from (6.7) we find that 2R3 + 8? is equal to an expression in the fourth and higher powers in s as was to be proved. Hence we have [9.4] The sphere through an arbitrary point P of a skew curve and with center given by (9.7) has contact of the third order with the curve al P. This sphere is called the osculating sphere of the curve at the point P. Hence we have [9.5] The polar line for a point of a curve is tangent to the locus of the center of the oseulating sphere of the curve at the corresponding point. ‘The previous results are represented in Fig. 7 in which the curve is the locus of the points M, M,, Ms, --- ; the points C, C,, C2, --- are the corresponding centers of curvature; the planes MCN, MCN, - re normal to the curve; the lines CP, €,P,, --. are the polar lines; and the locus of the points P, P;, --- is the edge of regres- sion of the polar surface. A curve all of whose points lie on a sphere is called a spherical curve. The normal planes to such a curve pass through the center of the sphere. ‘The coordinates of the eenter are given by equations (8.7), where w and v must be such functions of s that ak = 0, From (8.8) it follows that u = 1 = p and that dp de (98) Pe w=0, Pa or= From the first of these equations we have v = ~¢ and hence the coordinates of the point are given by (9.7). When this value of v is substituted in the second of (9.8), we have the following condition to be satisfied by p and 7: d (\dp (9.9) ert F 18) oo, 44 CURVES IN SPACE [Cu. 1 From (9.7) it follows that the square of the distance between the points : Xtand at isp + (: A) , which is a constant when the condition (9.9) ; is satistied. Conversely, when the condition (9.9) is satisfied it follows from (9.7) Thus, when equation (9.9) is satisfied, all the normal planes to the curve pass though a fixed point, which is at the same distance from the points of the curve. Hence we have: [9.6] A necessary and sufficient condition thal a curve be spherical is that its intrinsic equations salisfy the condition (9.9). Also we have proved incidentally that [9.7] When all the normal planes to a curve have a point in common, the curve is spherical. EXERCISES 1. The tangont surface of the cubic (1.17) is an algebraic surface of the fourth order, 2. The osculating plane of the curve (9.3) at the point ¢ = 0 is the plane z* = 0, which mects the tangent surface (9.4) in the generator and in the parabi zt = 26/3, 2° = Kys/6, whose curvature Is $x ; thus the osculating plane at 2 point P of a curve meets the tangent surface in a generator and in a curve whose curvature at P is three-fourths of the curvature of the curve at P. 3, The polar surface of a plane curve is a cylinder, whose right scetion is the plane evolute of the curve. 4. Any sphere which contains the cirele of curvature for a point P of a non- spherical twisted curve and which is not the osculating sphere at P has contact of the second order with the curve at P. ‘he angle between the radius of the osculating sphere of a twisted c ata point P and the locus of tite center of the sphere is equal to the angle betwe: the radius of the circle of curvature and the locus of its eonter. 6. When a twisted curve is spherical, the center of curvature for a point is the orthogonal projection of the center of the sphere upon the osculating phe. 7. The only spherical curves of constant curvature are circles. 10. PARAMETRIC EQUATIONS OF A SURFACE. COORDINATES AND COORDINATE CURVES IN A SURFACE Equations (9.2) and (9.5) of the tangent surface and polar surface of a curve respectively are particular cases of three equations of the form (0.1) =Sf',wv) @ = 1, 2,3). §10) PARAMETRIC EQUATIONS OF A SURFACE 45 Suppose now that one has three such equations where j* are one-valued functions of two variables u' and wu’, If the functions f* are such that it is possible to eliminate u’ and u? from these equations and obtain a single equation (10.2) F(a’, xo, 2°) = 0, then in accordance with the definition of a surface in §1 equations (10.1) are equations of a surface. This idea was used in §9 in establishing that equations (9.2) are equations of a surf: In order to determine whether three equations (10.1) in which the f's are functions of the w’s are equations of a surface, we define quanti- ties A” thus jar" af? spe au aut 0.3) Ae = OFS) GJ=12350% 9, aw, 2) aft ag ae aut that is, A” is the jacobian of f and f’ with respect to uw’ and w. If A” is identically equal to zero, then there is a functional relation be- tween 2! and 2’, thatis, g:(2', 2’) = 0." If also A = 0, there is a fune- tional relation y,(2', 2*) = 0, and thus the locus is a curve and not. a. surface (see §1). It is readily shown that if A” = A™ = 0 then A™ Hence we have [10.1] Three equations (10.1) are equations of a surface, if the jacobian ‘matrix jart art a aut aut at (10.4) toe of of of Cr | is of rank two, that is, not all of the determinants of order two are identically zero, It is to be observed, as just remarked, that if two of these determinants are identically zero, the third also is, and the rank is less than two (see Ex. 6). The defi wu and w as 1 of asurface by three equations in Lerms of two variables in (10.1) was introdueed by Gauss.} Formerly a surface * Fine, 1927, 1, p. 257. 4 1827, 1. 46 CURVES IN SPACE [Cu 1 was defined by 2 single equation (10.2), until Monge used the particular form (10.5) = f(a, 2°), ‘The latter had advantages over the form (10.2) in the investigation of certain types of surfaces. However, the method of Gauss is in many respects superior to both the othor methods. It is customary to refer to equations of the form (10.1) as parametric equations of a surface. In particular, the method of Monge is equivalent to the following equations: (10.6) =n, @-2, «= fz,2), where now z' and 2” are the variables u' and wu", Another interpretation of theorem [10.1] is that wv’ and u’ are inde- pendent. variables, and thus that the locus is two dimensional. When in equations (10.1) w’ aud w’ are given particular values, these equations give the coordinates of a point in the surface, as viewed from space in which the surface lies, that is, the enveloping space. However, one may consider the situation in the surface itself without reference to the enveloping space, and say that w’ and u? are coordinates in the surface of a point in the surface. An example of this is the use of latitude and longitude as coordinates in the surface of the carth, It should be remarked that, if one eliminates u' and u’ from equations of the form (10.1) and obtains an equation of the form (10.2) it may be that equations (10.1) apply to only a portion of the surface with the resulting equation (10.2) (see Ex. 3). However, in consequence of theorem [10.1] it follows that. twa of the equations (10.1) ean he solved for u' and u’, namely two for which the corresponding A“ is not iden- tically zero, and when these values are substituted in the third one ob- tains an equation similar to (10.5). This equation defines the same surface or portion of a surface as equations (10.1), and thus for the coordinates 2! of a point in the surface wand w are uniquely defined. When in equations (10.1) u* is given a constant value and u' varies, the locus is a curve, as viewed from the enveloping space, its equations being of the form (1.10) with u' as parameter; moreover, it is a curve in the surface. There is an infinity of such curves in the surface, one for each value of u’; we eall them the coordinate curves u° = const., and also the w'-coordinate curves. In like manner there is an infinity of coordinate curves 12 = const., and called also the w-coordinate curves. They are the analogue of the lines parallel to the coordinate axes in the plane referred to a cartesian system. When the plane is referred to polar $10) COORDINATE CURVES IN A SURFACE 47 Coordinates, they are the analogue of the lines through the pole of the system, and of the circles with center at the pole Gee Fig. 8), The surface with equations of the form (10.7) @sfw), =f, a fui, w) is a cylinder whose equation of the form o(2', ') = 0 is obtained on the elimination of u! from the first two of equations (10.7); in this case the curves u! = const. are the generators of the cylinder, and the character of the curves u* = const, depends upon the form of the funetion Fic. 8. Coordinate curves in a surface In general the coordinate curves are not straight lines but curves, Accordingly some writers call u? and u® curvilinear coordinates in the surface, but we call them simply coordinates in the surface, If in equations (10.1) we replace uf by (10.8) uw = eu’), the resulting equations are equations of a curve, as viewed from the enveloping space. Since this curve is in the surface, an equation of the form (10.8) is an equation of a curve in the surface expressed in terms of the coordinates u! and 2 in the surface, as is also an equation of the form 5 (10.9) ou’, uw) = 0, 48 CURVES IN SPACE (en. 1 For example, for a cylinder (10.7) the equation (see §3, Ex. 2) POL, aw) — cote | VF EG Faw =0 is an equation of a curve in the cylinder which intersects the generators under the constant angle 6, that is, the curve is a cylindrical helix. Also cach of the equations Ce as 7 takes the values 1, 2, 3 and the c’s are constants, is an equation of the family of curves in which the corresponding planes 2° = e' in space cut the surface with equations (10.2) If in (10.1) we substitute for «’ and u* the expressions (10.10) ue = pu", uw”) (a = 1, 2), (10.11) we obtain another set. of parametric equations of the surface, sa} (10.12) v= fil ul” @ = 1, 2, 3) Consequently there is great generality in the choice of two coordinates in berms of which a surface is defined by means of equations of the form (10.1). We refer to (10.10) as a transformation of coordinates in the surface. If the jacobian (10.11) is identically zero, this moans that there is 2 functional relation between the g's, say F(¢', ¢*) = 0, and then from (10.10) it follows that the coordinates u’* for « = 1, 2 apply only to the curve F(u', u’) = 0 aud not to the whole surface. Another signifi- cance of the condition (10.11) is that equations (10.10) can be solved for w’*,* say (10.13) ul = yl (ul, wv), which equations give the coordinates u'* of a point of coordinates u” in the w-system. Equations (10.13) define the inverse of the transforma- tion (10.10). It should be remarked that even if (10.11) holds, there may be values of u’* for which the jacobian is equal to zero. For such values there is * Fine, 1927, 1, p. 334, $10} COORDINATE CURVES IN A SURFACE 49 not a unique inverse (10.13), and special consideration has to be given to such eases. However, the discussion which follows presupposes that for the domain considered the condition (10.11) holds. Clearly for such a domain no two of the w'- or u’-coordinate curves can intersect. From (10.10) it follows that the coordinate curves u* = ¢* in the uwsystem have the equations eta, uw’) in the w-system, and from (10.13) that the coordinate curves w'* = c’* in the w-system have the equations oul, wv) = * in the w-system. However, for a transformation (10.14) wag(u), v= (uw) or (10.15) w= lu), eu") the net of coordinate curves, that is, the two families of eurves a) = const. and wv = coust., is not changed but the coordinates are changed, as is readily verified. EXERCISES 1, The equations zi=asinulcosw, a= asinutsinw, 2! = a cos ut, where @ ia a constant, are parametric equations of a aphere of radius, What are the coordinate curves ut = const, and u? = eonst.? 2. A surface with the equations @ teutcosu ata ulsinut, 2 = glu) ig the surface generated wen the plane curve with equations a= (2), at=0 is revolved about the 2°-axis; such a surface is called a surface of revolution, What, are the coordinate curves u! = const, and u? = const.? The latter are called the meridian curves; by what change of coordinates can the equations of Ex, 1 be given the form (i)? 8, The equations wl) (ay a;)(a% ~ ay) 50 CURVES IN SPACE [cH f in which the a’s are constants, and 4, j, k take the values 1, 2, 3, cyclically, are equations of a central quadric; it is an ellipsoid when a > a? > a1 > ut > ai > 0, an hyperboloid of one sheet when a > ut > a2 > 0 > a2 > 0, an hyperboloid of two sheets when a > 0 > a1 > ut > a> w. 4. A surface which is the locua of a line perpendicular to a fixed line, called the axis, and satisfies a further condition is called a right conoid; equations of a right conoid are m= ueosu, at=utsinut, 2 = ou); when g(ut) = a cot u? + 6, where a and b are constants, the conoid is a hyperbolie paraboloid. 5. Find equations of a right conoid whose axis is the 2*-axis and which contains the (att | oe ne oe Te 6. When the rank of the jacobian matrix (10.4) is two and one of the deter- minants of the second order is identically zero, the surface is a cylinder, ti. TANGE! XT PLANE TO A SURFACE ‘The tangent at a point P to a curve upon a surface qu.) a= fiw, uw) @ = 1, 2,8) is called a tangent to the surface at P, It is evident that there are an infinity of tangent lines to a surface at a point. We shall show that ordinarily these lines lic in a plane, called the tangent plane to the surface at the point. If we define a curve through a point 2° by the equations (11.2) =e) (a = 1, 2), we have from (11.1) da! _ af! dul | aft dai a ae dt" eat di” ‘Hence equations of the tangent at 2! are (see §3) 4, } 2 de and i from these equa- d where Lis a parameter. Eliminating 05% il] TANGENT PLANE TO A SURFACE bi tions, we obtain pane Xx-2¢ Xie o£, (11.3) | out aul aut | = 0. | af’ af? af* aut au aut For particular values of u' and w’, that is, at a point on the surface, this is an equation of a plane. Since the equation is independent of the functions ¢' and ¢’ in (11.2), it follows that this plane contains all the tangent lines to the surface at the point, and consequently is an equation of the tangent plane at the point. At points of the surface, if any, for which the cofactors of the elements of the first row in the determinant (11.3) are simultaneously equal to zero, the equation (11.3) is not defined. Such points are called singular points of the surface, and all other points ave called ordinary points. Henee we have (11.1] The tangents at an ordinary point on a surface to the curves on the surface through the point lie in a plane; when the surface is defined by para- metric equations (11.1), equation (11.8) te an equation of the tangent plane at the point 2. For the tangent surface to a curve with the equations (9.1) the equa- tion (11.3) reduces to x'- | 8 which by means of (5.7) is = (Xt = ayy Since this equation does not involve w, the tangent plane at one point of the surface is tangent to the surface at each point of the generator through the given point. Comparing this equation with (3.13), we have in consequence of (4.7) [11.2] The tangent plane to the tangent surface of a curve is the same at all points of a generator; il is the osculating plane of the curve al the point where the generator is tangent to the curve, 52 CURVES IN SPACE [Cu. I In order to obtain an equation of the tangent plane to the polar sur- face of a curve, we note that in consequence of equations (6.1) we have from (9.5) ax" au 2b wr )6" — ery’, ‘t Hence in this case equation (11.8) is reducible to "57 Xiaa xia! BY & B | =0, io 2 2 ¥ ¥ yt where in this equation X" are current coordinates. In consequence of (6.7) this equation reduces to = (X= 2a’ =0. From the form of the equation and the fact that it does not involve wu’ we have [11.8]. The tangent plane to the polar surface of a curve ix the same at all points of a generator; it is the normal plane to the curve at the point whose polar line is this corresponding generator of the polar surface. In order to find an equation of the tangent plane to a surface defined by a single equation (11.4) Se, 2, 2) we assume that 2" for the surface ave expressed in. the form (11,1) in terms of two coordinates u' and x’. When these expressions are sub- stituted in (11.4) the resulting equation is an identity in u! and, wv and consequently the derivatives of this identity with respect to wv’ and u* are equal to zero. Hence we have > of af* at aw! 0, of From these equations it follows that the quantities 3°, are proportional to the cofactors of the clements of the first row in equation (11.3).” Consequently an equation of the tangent plane to the surface (11.4) *C.G., p. 104, $12} DEVELOPABLE SURFACES 53 at a point 2° is 1.) rt“ 7 Ot EXERCISES 1, Find an equation of the tangent plane to the sphere with equations of the form in §10, Ex. 1, and show therefrom that the tangent plane at a point is normal to the radius of the sphere at the point. 2, Find an equation of the tangent plane at a point of a cylinder with equa- tions (10.7), and show that the tangent planes at all points of a generator are the same plane. 3. Find an equation of the tangent plane at a point of the cone a2"* ++ asz** + a;z*" = 0, where the a’s are not all of the same sign, and show that the tangent planes at all points of a generator are the same plane. 4. For a surface with the equation (10.5) an equation of the tangent plane is ve my iy Bee (Xt = 28) + (XP = at) 5 (KF - 2) a this result follows also from (10.6) und (11.8). 5. The tangent plane to the right conoid (see §10, Ex. 4) m= woosut, 2t=utsinw, 2 = asin ut at a point not on the 2*-axis meets the conoid in the generator through the point of tangeney and in an ellipse. 6. The tangent planes at points of a generator of the right conoid aim ulcosut, ste ulsinw, a Vianwt meet the plane 2? = 0 in parallel lines, 7. ‘The normal to the tangent plane to a surface at the point of tangeney is called the normal to the surface at the point; the normals to any right conoid (Gee §10, Ex. 4) at points of a generator are one family of rulings of a hyperbol paraboloid. 8. A surface with equations of tho form ateulcosut, zt = utsinul, at = oful) + aut, where a is a constant, is called a helicoid; the coordinate curves ut = const are circular cylindrical helices; find an equation of the tangent plane. 9. The distance of a point Q ona surface from the tangent plane to the surface at a nearby point P is of the second order at least in comparison with the length of the are PQ. 12, DEVELOPABLE SURFACES. ENVELOPE OF A ONE-PARAMETER FAMILY OF SURFACES _ From the form of the equation (11.3) of a tangent plane to a surface it follows that ordinarily this equation involves both of the parameters u' and u’, and consequently in general there is a double infinity of 54 CURVES IN SPACE (Cu, I tangent planes to a surface. This is the case, for example, with a sphere, the tangent plane at a point being normal to the radius of the sphere at the point, as one sees geometrically (see §11, Ex. 1). How- ever, we have seen that the tangent planes along a generator of the tangent surface of a curve coincide, this tangent planc for any generator being the osculating plane of the curve at the point of tangency of the given generator. Likewise the tangent planes to a cylinder, or a cone, along a generator are the same as is evident geometrically, Hence the tangent planes to a tangent surface of a curve, toa cone, or toa cylinder involve only one parameter, and consequently these surfaces are the envelopes of a one-parameter family of planes. They are called developable surfaces, since any such surface can be developed upon a plane, that is, rolled out without stretching or contracting any part of it, Iv is evident that this can be done with a cylinder or a cone. We desire to show that with the exception of cylinders and cones every developable surface is the tangent surface of some curve. We consider first; the more general problem of finding the envelope of a one-parameter family of surfaces. ‘An equation (12.1) fa, 2, 2;w) = 0 involving a parameter u' as well as the 2’s is an equation of a one- parameter family of surfaces, each of which is defined by (12.1) when wd is assigned a particular value. Consider now the curve of intersec- tion of the surfaces (12.1) and f(z’, 2°, 2°; w! + Au’) = 0 for particular values of u' and Au’. This curve is the curve of intersection also of (12.1) and the surface Sa, #, fc, #, By) =0. There is such a curve unless f(x’, 2°, 2°; u') = g(a’, 2, 2°) + wu’), in which case the preceding equation does not involve 2‘, We assume. therefore, that u' does not enter in this manner. As Au! approaches zero this curve approaches a limiting curve defined by (12.1) and of 12.2) F =0. (12,2) al The curves defined by (12.1) and (12.2) as w' takes all values are called the characteristics of the family of surfaces (12.1). They form a surface E, called the envelope of the surfaces (12.1). Bach surface of the family and the envelope have one of the characteristics in common, and, as We een §12] ENVELOPE OF A FAMILY OF SURFACES 55 shall show, are tangent to one another along this characteristic; that is, at each point of the characteristic the surface and envelope have the same tangent plane. Let the coordinates 2‘ of a point on the envelope be expressed in terms of w! and some second parameter w’, say 2° = f(u', wu’). With this choice of coordinates the coordinate curves u' = const. are the char- acteristics. When these expressions for 2‘ are substituted in (12.1), the resulting equations are identities in u' and 12, and consequently the derivatives of this expression with respect to u' and wu’ are equal to zero. In consequence of (12.2) we have af axt Laat = ° From these equations it follows that the cofactors of the elements of the first row in the equation (11.3) of the tangent plane to the nvelope are proportional to a and consequently an equation of the tangent plane to the envelope at the point 2" is a La-2)b=o. This is an equation of the tangent plane to surface (12.1) for a given value of w!, as follows from (11.8). Since uf does not enter in 22, it follows that at each point of a characteristic the corresponding surface (12.1) and the envelope have the same tangent plane. We consider now in addition to equations (12.1) and (12.2) the equation of (12.3) a We denote by (12.4) a= fiw) the common solution, if any, of equations (12.1), (12.2) and (12.3), Equations (12.4) are equations of a curve on the envelope of the sur- faces (12,1) since they are solutions of (12.1) and (12.2), and for a value u' = a' the corresponding point of the curve (12.4) is a point of the characteristic ui = a'. We wish to show that at this point the curve and the characteristic have a common tangent line. If the expressions (12.4) are substituted in (12.1) and (12.2) the re- 56 CURVES IN SPACE. (Cu. 1 sulting equations are identities in u', and consequently their derivatives with respect to u' are equal to zero. In consequence of (12.2) and (12.3) the result of differentiating (12.1) and (12.2) with respect to wi is (12.5) cer no DSL sr =o, fOr 7 Bu! Bert where the prime indicates differentiation with respect to u', Ata point of the curve (12.4) the quantities f” are direction numbers of the tangent to the curve at the point. A characteristic u' = a’ is the intersection of the surfaces defined by (12.1) and (12.2) for this value of u'. From (11.5) it follows that the tangent planes to these two surfaces at the point 2} = f(a’) ave re- spectively a ae a pa-og-0 Barat =o ‘Taken together these are equations of the line of intersection of these two planes, that is, the tangent at 2 to the character 7 tie ul = al. From these equations and (12.5) it follows that f” are direction numbers of the tangent. at zi to the characteristic. Consequently the charac feristies are tangent to the curve (12.4), and accordingly the latter is the envelope of the characteristies. It is ealled the edge of regression of the envelope £ of the surfaces (12.1). Gathering together these results we have {12.1] The envelope of a one-parameter family of surfaces is a surface which is tangent to each surface of the family along a curve, the characteristic corresponding to the particular surface; the characteristics are tangent to a curve, the edge of regression of the envelope. When all the characteristics have one and only one point in common, that is, when equations (12.1), (12.2) and (12.3) admit a common solu- tion only for a particular value of w, this point is a degenerate cdge of regression. We apply these results to a one-parameter family of planes with the equation (12.6) aa’ +a =0, where a; and a are functions of a parameter u' with the understanding that a; are not proportional to a set of constants ¢;, that is, that the planes of the family are not parallel.* In this case equations (12.2) *C.G,, p. 101, sl2} DEVELOPABLE SURFACES 37 and (12.3) are respectively (12.7) ai ta'=0, afz'+a"=0, where the primes denote differentiation with respect tou’, rom (12.6) and the first of (12.7) it follows that the characteristics are straight lines. If the determinant. A, defined by ‘a1 ap ay | (128) Az=lat a al, ay ay ay is not identically zero, equations (12.6) and (12.7) admit a unique solu- tion (12.4), and by theorem [12.1] the envelope is the tangent surface of the curve (12.4), unless the functions s‘(u’) in (12.4) are constants, If f'(u') are constants e’, then from (12.6) we have a = —a.e', in which case all of the planes of the family pass through the point c’. When all the planes (12.6) have a point in common, the envelope is a cone, unless all the planes have a line in common. In the latter case the line of intersection of the planes (12.6) and the first of (12.7) must be in- dependent of u'; that. is, that linc and the one with equations (12.7) must coincide. ‘The condition for this is that the matrix a 2 a @ (12.9) a, 42 43 a ay ay a3 a’! be of rank less than three.* Hence we have [12.2] The envelope of a one-parameter family of non-parallel planes (12.6) for which A, defined by (12.8), is not identically zero is the tangent surface of a curve, or a cone; in the laller case a = —aye', where the c's are constants, and c' are the coordinates of the vertex of the cone. We consider next the case when A = 0. In this case there are quanti- ties h', h*, A such thatt (12.10) War=0, hia =0, hia’ = 0. Differentiating the first two of these equations, we obtain in consequence of the third (12.11) ha: =0, ha, = 0. *O.G., p. 125, 7C.G, p. Ne. 58 CURVES IN SPACE (Ca. 1 From these equations and the first two of (12.10) it follows that aj are proportional to the corresponding a; , and consequently a, are propor- tional to constants ¢;, the factor of proportionality being a function of wu’, in which case the planes (12.6) are parallel. Since this case has been excluded from this discussion, it follows that the h’s are constants so that equations (12.11) do not exist. Since a; are direction numbers of normals to the planes,* it follows from the first of (12.10) that all of the planes are parallel to a line of direction numbers hi. Consequently the envelope is a cylinder, or all the planes have a line in common. ‘The former case arises when the three equations (12.6) and (12.7) do not have a common solution, that is, when the matrix (12,9) is of rank threet. As temarked above, the planes pass through a line when the rank of the matrix is less than three. Hence we have [12.3] When for a family of non-parallel planes aye‘ + a = 0 the deter minant (12.8) is identically zero, the envelope of the planes is a cylinder or all the planes have a line in common according as the rank of the augmented matrix (12.9) és three or less than three. We seek now a necessary condition upon a function f(a", 2%, 2") in order that the surface (12.12) fa’, 2,2) = shall be a developable surface. Assume that the surface is defined by the equations x’ = g'(u', u°), where u! = const. are the generators and = const. another sct of coordinate curves. When these expressions for z‘ are substituted in (12.12), the resulting equation is an identity in uw and u’, and consequently any a0, ino, (12.18) fe aa 7. and the summation convention is applied. An equation ont of the tangent plane to the surface (12.12) at the point 2‘ is (12.14) (X' — xf = 0. Since equation (12.14) does not involve uv’ by hypothesis its derivative with respect to w’ is equal to zero, Therefore, in consequence of the where f, *C.G., p. 92 tC.G., p, 126, §12] DEVELOPABLE SURFACES 59 second of (12.13) we have (12.15) (x=)? where f ata?” Since u' = const. are the generators of the surface by hypothesis, their equations are given in accordance with (12.1) and (12.2) by (12.14) and a inp) a Ale - 2A = 0, which in consequence of the first of (12.13) reduces to aa? pi (12.16) (X= 2) On comparing equations (12.15) and (12.16) with (12.13), we see that (12.17) (X' = xfs = Hi, where ( is a factor of proportionality. In order that the three equations (12.17), and (12.14) be consistent, it is necessary* that the equation |fa fe fia St (12,18) fu fo So fr =o Ja fas fa fo ff fs 0 be satisfied in consequence of (12.12) or identically. Conversely, if this condition is satisfied and x* = ¢*(u', u*) are equa- tions of the surface, that is, satisfy (12.12) identically, when these expressions for x‘ are substituted in (12.18) the resulting equation is an identity in u' and uw’. Hence there exist functions a‘ and a of u! and w such that (12.19) afi; +of;=0, a'fr=0. If any one of the quantities f, is equal lo zero identically, then (12.18) » is satisfied identically and the surface is a cylinder whose generators are "C.G,, p. 139, 60 CURVES IN SPACE [Cu £ parallel to a coordinate axis. There remains to consider the case when. each f; is not identically zero, and we write the equation (12.14) of the tangent plane to the surface (12.12) in the form yt 2 fe afe _ xf (12.20) X'+X i +X ih The quantities fe/fi , fo/fi, and 2'fs/f: arc functions of ul and ut, their form depending upon the functions g‘(u', u’). In order that the surface (12.12) be developable it is necessary and sufficient that there be co- ordinates u" and w” in the surface such that fo/fi, Jo/fi, and a'fi/fr shall be functions of one of them, say «". ‘This means that these three quantities must be functions of the same function, say ¥(u', u’), in which case uw” = y(u', u’) and w” is any other function of u’ and wv? such that the jacobian of this function and y is not identically zero. If this condition is satisfied, the surface is developable and the eurves ¥(u', wv) = const. are its generators. Hence a sufficient condition that the surface be developable is that when (12.19) are satisfied the jacobian of each pair of the quantities fo/ft , fa/f:, and 2'fi/f: with respect to u’ and w’ be identically zero. The jacobian of the first two of these quanti- ties is / ia teh 2 idan — bid 2B Cider = fei) 2, idan — Fat) | which is equal to* fi fe Ss ' act ar’, on! | i, A Ae Alfio Aran fu 5a | H i ax’ oa Liza fixe Size | In consequence of (12.19) and (12.13) this determinant is equal to zero. Because of this result and (12.13) the jacobian of 2'fi/f, and each of the quantities fo/f: and fa/J: can be shown to be equal to zero. Hence we have [12.4] A necessary and sufficient condition that f(x’, x*, x’) = 0 be an equation of a developable surface is that equation (12.18) be satisfied in consequence of the equation f(x", 2°, x*) = 0 or identically. + Fine, 1904, 1, p. 505. §12] DEVELOPABLE SURFACES 61 Consider the equation (12.21) é x where the e’s are constants. Evidently any solution of this equation satisfies (12.18) identically. Also (12.21) is the condition that the normals to such a developable surface ba perpendicular to a line with direction numbers c’. Hence we have [12.5] A necessary and sufficient condition that a surface f(z', 2°, 2°) = 0 be a cylinder is that f be a solution of an equation (12.21) tn which the c's are constants, EXERCISES 1. The envelope of the plancs normal to a curve is the polar surface (see §11); the polar surface is also called the polar developable. 2. The envelope of the plane normal to the principal normal to a curve at a point of a curve is ealled the rectifying developable of the curve; equations of its characteristies are Xt = ait (rat — wyi)t, where 1 is a parameter, and equations of its edge of regression are 3. If cach of the generators of a developable surface, other than @ cone or cylinder, is revolved through the same angle about tho tangent to an orthogonal trajectory of the generators at tho point of interscetion, the locus of the resulting lines is a developable surface whose edge of regression is an evolute of the given trajectory. 4. The edge of regression of the family of planes (1 ~ utjat + il + ute? + Qua? + flu) = where i = 4/=T and w is the paramoter, is a minimal enrve; the envelope is called an isotropic developable surface. 5. Find the edge of regression of the developable surface which envelopes the hyperbolic paraboloid az* = x'2* along the curve in whieh the paraboloid is cut by the paraboli 6. The curvature and torsion of the edge of regression of the family of planes agz' + a = 0, where the a’s are functions of a parameter u, are given by = “ oe 62 CURVES IN SPACE (Cu T a” ay’ as” al” where |\% = mu @ a, a a; | Js 7 + fal oa a a io tl] | a a2 a |, Del in a tht | | @22oy a a al ay ay | 1 | primes denoting differentiation with respect to w. 7. For a family of spheres Law adi - a) - @ =0, in which the a’s are functions of the are of the curve € of centers, the edge of regression consists of two parts with corresponding points symmetric with respect to the corresponding osculating plane of C, unless @ @(1 ~ a’) — pil ~ (aa')'}) = where p is the radius of curvature of C; when the condition (i) is satisfied, the edge is a single curve, Cr, its points lying in the osculating planes of C, and the spheres of the family aro the osculating spheres of Ci . CHAPTER II Transformation of Coordinates. Tensor Calculus 13. TRANSFORMATION OF COORDINATES. CURVILINEAR COORDINATES In the preceding chapter x'(i = 1, 2, 3) denote cartesian coordinates of space referred to rectangular axes. ‘There are many such coordinate systems, and the relation between two such systems 2’ and 2’ is given by equations of the form (13.1) at = aja! +p bt The coordinates «"' refer to a set of rectangular axes whose origin in the a-system is b‘, and aj, aj , a} are the direction cosines of the 2”-axis with respect to the z-system. These direction-cosines satisfy the conditions* (13.2) Laat = sn, DL atat where the quantities 6,, and 6*, called Kronecker deltas, are defined by (13.3) 6; and 6* = 1 or 0 according as j = k orj # k. Moreover the determinant of the quantities a} , that is, Jat a: a} (13.4) lai|=lat a as lat a: a} is equal to +1.t If we denote by a// the cofactor of aj in the determinant (13.4) divided by the determinant, then (13.5) aia = 63, ajar’ = 8}, where 4 , also called Kronecker deltas, are defined by (13.6) 81 = 1 or 0 according asi = j ori = ji. *0.G., pp. 161-164. tC. G., p. 162, 64 TRANSFORMATION OF COORDINATES (Cx, IL ‘The second set of equations (13.5) is equivalent to the statement that the sum of the products of the elements of the j" column and the co- factors of the corresponding elements of the i column is equal to the determinant or zero according as t = j ort # j,.* If then we multiply equation (13.1) by ar’ and sum with respect to i, we obtain (43.7) ai(e’ — b) = alain? = dhe! = 2, Hence the loci 2* = const. are parallel planes for which the quantities a;* are direction numbers of the normals to these planes in the z-system.+ However, if the first set of equations (13.2) for j # k are not satisfied, the axes of the z’-system are not mutually perpendicular, as follows from the remark following equations (13.1). Hence in this case the 2’-system is an oblique system of coordinates. We consider next the case, when the determinant (13.4) is equal to zero. In this case equations (13.1) in the r’s are consistent only when the three determinants obtained on replacing the elements of a column in the determinant (12.4) by 2' — bi, 2? — B, 2° — b respectively are equal to zero.t Consequently in this case the quantities 2” are defined only at points of the locus with these equations, and thus are not co- ordinates for the space. Equations (13.7) define the inverse of the transformation (13.1). Moreover, equations (13.1) define the inverse of the transformation (13.7) as follows from (13.5). Polar coordinates constitute another type of coordinates frequently used in space, particularly in astronomy. With reference to a cartesian coordinate system 2* polar coordinates 2" are defined by (13.8) z= 2" sina” cos x”, The inverse of this transformation is (13.9) 2" = va we, a =cos* VE ro 7 From these equations it follows that 2!" is the distance of the point P(x’) from the origin O of the z-system; 2” is the angle which the line OP makes with the z-axis; and 2” is the angle which the projection of OP on the z'z-plane makes with the positive z'-axis. From (13.9) it follows that the loci 2” = const. are spheres with O *C.G, p. 1. $C. G., pp. 92-98. $C. G., pp. 124, 413) TRANSFORMATION OF COORDINATES 65 as center; the loci x” = const. are right circular cones each of which has its vertex at O and the z-axis for axis; and the loci 2’ = const. are planes through the z-axis. We call these surfaces the coordinate surfaces in the a’-system. In like manner the coordinate surfaces of the 2/-system defined by (13.1) and (13.7) are planes, which as shown are mutually orthogonal only in case equations (13.5) are satisfied. Equations (13-1) and (13.8) are particular cases of equations (13.10) a = g(a", 2%, 2"), where the ¢’s are one-valued functions of «”, «”, 2. For any such functions ¢', these are equations of transformation to a general set of coordinates x", 2”, 2”, provided that the functions ¢* are independent. If they were not independent there would be one, or two, relations of the form F(¢', ¢’, ¢') = 0, and in that case (£3.10) are not equations of a transformation of space; they have meaning only at points of the locus with this equation, or equations. A necessary and sufficient condition that the functions g' be independent is that their jacobian, namely ag de! | i O27 aay [2e! ag ag | _ sn de! ag ag : x’ | ac aa? ax"? ax’i ax’® ag ag ax"? aa’ | be not identieally zera* (see (1.19)). When this condition is satisfied, there may be particular values of 2! for which the jacobian is zero, but in general this is not the case, that is, about a point 2” for which | = | 0 there is a domain for which 13 this inequality holds. For such values of 2/‘ equations (13.10) can be solved for 2"; we denote such a solution by (13.12) alt = (al, 2, 24 These equations define the inverse of the transformation (13.10). For equations (13.1) the jacobiun (13.11) is the determinant (13.4). When this determinant is not equal to zero, the aj‘ exist and equations * Fine, 1927, 1, p. 253. + Fine, 1927, 1, p. 253. 66 TRANSFORMATION OF COORDINATES — [Cu II (13.7) are the inverse of equations (13.1) and give the coordinates 2’* for a point in terms of the 2’s of the point. For equations (13.8) the jacobian (13.11) is found to be equal to (@"Y’ sin 2. Since this quantity is not identically zero, polar coordi- nates apply to all points of space, but they are not uniquely defined when 2" = 0 or when sin 2 = 0, that is, at 0 or on the 2*-axis. Thus at O, x’) = 0 and 2” and 2” can take any values, that is, all the co- ordinate surfaces 2”-const. and x’-const. pass through 0. For any point on the z-axis, x” can take any value, that is, all the coordinate planes 2'* = const. meet in the 2°-axis. In these cases the coordinate surfaces are degenerate, thatis, x” = 0 is a point, and 2” = 0, x, a line. But for all other points equations (13.9) hold and are the equations of the inverse. When the expressions (13.12) for 2” are substituted in (13.10) we have x! o'e oo) =0, which are identities in the 2’s, as follows from the definition of (13.12). Since they are identities, the left-hand member does not vary with any of the 2’s, and consequently the derivative with respect to each 2° is equal to zero, Hence we have aa! _ ae! ag" ax Oy" aah” where k is a dummy index. Since 2’ and 2’ for i # j are independent, the left-hand member is +1 or 0 according as i = j ori # j. Accord- ingly these equations may be written « _ da* ax’ (13.13) be Se where 4{ are defined by (13.6). Sinco # and j take the values 1, 2, 3 there are nine equations in the set (13.13). If we consider the expressions (13.10) for 2 substituted in (13.12), we obtain analogously to (13.13) az! aak ag (13.14) Sat api When the jacobian of the transformation (13.10), namely (13.11) denoted by | z |, is multiplied by the jacobian of the inverse (13.12), one obtains in consequence of (18.13) a determinant in which each $13] TRANSFORMATION OF COORDINATES 67 element of the main diagonal is +1 and every other element is zero. Hence we have | ax) jax \ex"| "| ox | (13.15) that is, either jacobien is the reciprocal of the other. Thus the relation between equations (13.10) and (13.12) is such that either is the set of equations of the inverse transformation of the other. If in equations (13.13) we give # a fixed value and letj take the values . FP ol ax" 1, 2, 3, we have three equations of the first degree in ae? jx and a Solving these equations for these quantities, we obtain (see Ex. 1) ae! | an: ac’. gt _ cofactor of in| alt aa! Or Although we started this section, interpreting x‘ as cartesian co- ordinates in space, and took equations (13.10) as equations of a trans- formation from such coordinates to any other, in deriving the properties of the transformation (13.10) and the inverse (13.12), no use has been made of the fact that x‘ were cartesian coordinates, Therefore all the results of this section apply equally well when equations (13.10) give the relation between the coordinates of any two general systems what- ever in space. Thus, if x* are any set of coordinates, equations (13.1) and (13.8) define a transformation of coordinates; but the geometric interpretation of the new coordinates in these two eases given above applies only to the case when 2‘ are cartesian coordinates. We consider now in connection with a transformation (13.10) a second transformation (13.17) a= ver, 2%, 2%, (13.16) + «| aa’ |. A . it being understood that the jacobian | £5,| is not identically zero. When these expressions for 2’ are substituted in equations (13.10), the resulting equations denoted by (3.18) a= sie, a, ty define a transformation from 2‘ into 2’, called the product of the trans- formations (13.10) and (13.17). Since oat aah aa aa” ani? 68 TRANSFORMATION OF COORDINATES (Cu. 15 it follows from the rule of multiplication of determinants that | az | | ax | || aa’) law) = laa] | ae" |? and hence that the jacobian of the product of two transformations of coordinates is the product of the jacobians of these transformations, We say that transformations of coordinates have the group properly by which we mean that the product of any two such transformations is a transformation of coordinates. If x are cartesian coordinates, each of equations (13.12) for a partie ular value of an 2’ is evidently an equation of a surface in space, and for three particular values of x”, 2* and x these are equations of three surfaces in space intersecting in the point with these coordinates in the a'-system, as discussed above in the case of linear equations (13.1) and polar coordinates (13.8). In like manner for particular values of cartesian coordinates 2, equations (13.10) are equations in the co- ordinates «'* of the planes through the point x‘ parallel to the coordinate planes 2 = 0. If x‘ and 2” are any coordinates whatever, equations (13.12) for particular values of x", 2”, 2” are equations of three surfaces, defined in terms of the coordinates x*, which pass through the point (x”, 2”, x"); and similarly for equations (13.10). Another way of stating the above remarks is that in any coordinate system each of the equations (13.19) for particular values of the constants ¢ is an equation of a surface for which one of the coordinates is a constant for all points of the surface, the values of the other two coordinates determining a particular point on the surface. ‘Thus equations (13.19) are equations of surfaces which are the analogue of planes parallel to the coordinate planes of a rec- tangular coordinate system. As the constant c’ in any one of these equations takes on a continuum of real values, the equation is an equation of a family of surfaces. Thus each of these equations is an equation of an endless number of coordinate surfaces, In the x'- system defined by (13.10) equations of these coordinate surfaces are e(x", 2”, x") = ce. There passes one and only one surface of each family through each point in space for which the jacobian of the trans- formation from cartesian coordinates to the coordinates in question is not zero. At points where the jacobian vanishes the new coordinates are not: uniquely defined, as remarked in the case of equations (13.8). §13] CURVILINEAR COORDINATES 69 Any two of equations (13.19) for particular values of the ¢’s are equations of the curve of intersection of the corresponding coordinate surfaces. Along such a curve the remaining coordins parameter; thus ’ is a parameter for (any) one of the curves 2° and 2° = ¢’, We call these curves coordinate curves, and, in particular, an a coordi nate curve one for which x alone varies, and thus will serve as a param- eter. Those curves ara the analogues of lines parallel to the coordinate axes in cartesian coordinates. Since these curves are not in general straight lines, the corresponding coordinates are called curvilinear. From the above discussion it follows that whatever be the coordinates ‘in general the locus defined by an equation f(2’, 2’, 2°) = 0 isa surface, and the locus defined by two independent equations f(z’, 2’, 2") = 0, f(z’, 2, 2) = 0 isa curve. This is equivalent to the definition of a surface as a two-dimensional locus, and of a curve as a one-dimensional locus (sce §1). EXERCISES 1, It follows from (13.5) that and that af is the cofactor of aj? in the determinant | a}‘| divided by the determinant, 2. From (13.3) and (13.6) it follows that Bie = OL, and consequently 44 is the cofactor of 8:, in the determinant | 64; | (see (13.16)) 3. From (13.6) it follows that ais 4. From (13.4) and (1.19) it follows that ernaiazat = Ja} | eine 5 etitalamaz = | ai | em 5. By giving to i and j different values one verifies that the cofactor of aj in the determinant (13.4) is given by de emma at. 6. Determine the coordinate surfaces x’ = const, for each of the following transformations, in which #/ are cartesian, and find the points for whieh the iacobian is equal to zero: @ atm tx! cos 2, at = ate’? sin 28, a? = (at)? ~ (a')2y “(a a! ayer an p a ' ii, Gi) oe Gi 70 TRANSFORMATION OF COORDINATES (Cu. If where i, j, k take the values 1, 2, 3 in eyclie order, that is 1, 2, 332, 3, 1; 8, 1, 2, and where at > a? > a? > 0. 7. Discuss the coordinate curves for each of the systems in Ex. 6 and also for (18.8). 14. THE FUNDAMENTAL QUADRATIC FORM OF SPACE We have seen in §2 that for a curve in space defined by equations of the form (14.1) 2 =f, the 2’s being cartesian, the differential of length of the curve is given by a= (a y de. 7 \dt Accordingly we say that the element of length, or linear clement, ds of space is given by (14.2) ds’ = (da')* + (dz*)* 4- (da')*, by which we mean that, when the differentials de‘ from (14.1) are substituted in (14.2), the expression for ds obtained therefrom is the differential of are length of the curve (14.1). The right-hand member of (14.2) is called the fundamental quadratic form of space. In terms of any other coordinate system 2” we have from (13.10) ax" (143) de = dx”, and from (14.2) 4) wo > (%, aw (25, ax’) =a dxdz", oat where (14.5) from which it follows that aj, = a4. Since the coordinate system 2” is any whatever, it follows that in any coordinate system o‘ the fundamental quadratic form is (14.6) ds’ = as; da' de’, where ai; is symmetric in the indices i andj, that is, ac; = aj. When, in particular, the coordinates 2° are eartesian, we have (14.2), and con sequently (14.7) a3; = 853, ds? = 85; dx‘ dx’, §l4] FUNDAMENTAL QUADRATIC FORM OF SPACE a In the ease of polar coordinates (13.8) we have from (14.5) (48) an =1, 0 an = (e"), ae = (2%) sin’ 2” ai; = U fort # 7. Hence in polar coordinates the fundamental form is (14.9) ds = (de! + (2)? (de)? + sin? 2? (de), We desire now to find the relation between the coefficients aj; and aj; of the fundamental form in any two coordinate systems x‘ and 2”, Since the element of length ds does not depend upon a coordinate system, the fundamental forms in any two coordinate systems are equal. Hence we have aydxide? = aly de'® de" Substituting for dz‘ and dz! from equations of the form (14.3), we have ae a? ot \ sik aelt (0 25 = oi) ae dz! = 0. Since this equation must hold for arbitrary values of the dx”s, and the expression in parentheses is symmetric in k and 1, it follows that , act aa? (14.10) Ost = OG Se gl” In order to obtain this result, one takes de” = 0, dx’ = dx” = 0, and gets equation (14.10) for k = 1 = 1, Jn like manner, as one takes every other two of the differentials equal to zero, one gets (14.10) for k = 1 = 1, 2,3. In order to obtain the remaining three equations (14.10) for k # J, one takes one of the differentials equal to zero at a time, and the others not equal to zero, That there are only three of these equations follows from the fact that in any coordinate system au: = au, as previously shown. hon If we multiply (14.10) by 22; 4" and sum with respoet to and J, we have on making use of (13.13) ay 2@ 2", de" a2 ar’ aa! "Gah am ~ OF Ga Oak Bx” Oe™ < 4j548m = Ohm 4 Hence, on changing indices, we have the following equations (4.11 = aly 9% Ox" ) ay = Ob Se Sai 72 TRANSFORMATION OF COORDINATES {Cu. IL connecting the a’s and a's which ate equivalent to (14.10), but in ‘inverse form. If we denote by a’ and a the determinants of the quantities aj; and as respectively, we have from the rule for the multiplication of deter- minants (14.12) a = fate) = | ass we { ax | El =a aa" a | | jax"! ax!* Bx | ~ For the values 8;, of the a’s, that is, when the coordinates arc cartesian, the determinant is -+1. From this result and (14.12) we have [14.1] The determinant of the coefficients of the fundamental form of euclid- ean space in any coordinate system is positive. Since a ¥ 0, functions a are defined uniquely by (14.1) In fact, on solving these equations for a in the manner which led to (13.16) from (13.13) we have . cofactor of aa; in a a : (14.14) a™ Since a;; = ajx, it follows from (14.14) that a = a", that is, the quanti- ties a* are symmetric in the indices, in this case superscripts. In particular, when ac; = 6:3, we have a” = 6" (see §13, Ex. 2). If a denotes the determinant of the quantities a“, it follows from (14.13) that the product of the determinants a and 4 is equal to the determinant for which each element of the main diagonal is -+1, and every other element is zero. Hence a) 1 (14.15) a={a"|= a For any other coordinate system 2” the coefficients ai; of the funda- mental form are given by (14.10). By means of equations of the form (14.13) functions a’* are uniquely defined. We shall show that the following relations hold between the functions a’ and a¥':; a Ox" Ga? ax dat” (14.16) dis Tn fact, if we take the equations aan = dh, $14] FUNDAMENTAL QUADRATIC FORM OF SPACE 73 which are of the form (14.13), and substitute for ais expressions of the form (14.11), we obtain If now we multiply by uz and sum with respeet to Z, we have a1 Ox" Bx" yaa _ arg, a? _ gy 02" _ ax” dat Ga! "aah * ant Bot aah Since = is equal to 64 ae , We can write the above equations in the r form er Oa" Bx" 7) an" (« Qaf aor 2 ss) Ga 7% (we say equations, because j and h being free indices there are 9 of these equations). For a fixed value of j as h takes the values 1, 2,3 we have three equations, linear and homogeneous in the three expressions in ax’ | Ir | parentheses as m takes the values 1, 2,3, Since = 0, we have on changing indices Maa! MOF Oe dim — 84 = 0. oat Oe Multiplying by a’ and summing with respect to m, we obtain finally (14.16). EXERCISES: 1. Show that ¥/@n dz! is the differential of length of an coordinate eurve; apply this result to the ease when the 2's are polar coordinates 2 Show that when in (14.13) ai; = 0 for i ¥ J, then 1 e = * » atiad GI. o 3. Show that in any coordinate systom 2 the expression auy(d2!)? + 2aya det dz* + an(dz8)t : fs the square of the differential of length of a curve on any coordinate surface ¥ = const, the curve bejug defined by « ,« 4. When the coordinates 2! are cartesian, the coefficients af) of the funda- _ Mental form in the coordinates 2’ for a transformation (13.1) are given by ay = Dat 74 TRANSFORMATION OF COORDINATES (Cu. IT and the determinant a’ of these quantities is given by as | a5 consequently the coefficients ai; are constants. 5. Show that the fundamental quadratic forms of space in the coordinates 2’ of §13, Ex, 6 are respectively @) (Ce)? + (@) (a2)? + (de)] + (ett) (d's); Ly t= 2)" — 2) ae") 4O waa — aa a)’ Gi) where j and k are the numbers 1, 2, 3 other than i. 15. CONTRAVARTIANT VECTORS. SCALARS From the equations (13.12) of a transformation of coordinates, namely (15.1) xis Mala, 2), we have aa 15.2) 1h OE aah (15.2) del! = de ‘Thus whatever be the coordinates in two systems in transforming from one system to the other differentials undergo a linear homogeneous transformation, the coefficients being in general functions of the co- ordinates. In this sense a transformation of coordinates induces a linear homogencous transformation oe differentials of the coordinates, JF equations (15.2) are multiplied by 3 7, and summed with respect to J, we have in consequence of (13.13) aa! yy _ an’ ax" ga Ot = Sa gaa U2! = Sha! = de’, as the inverse of (15.2). Equations (15.8) are, in fact, equations (14.3), which were obtained from the inverse of (15.1). If in (15.3) we replace " da! by de! = * dz! obtained from (13.17), we have (15.3) (15.4) drt = =, ee da = oe dl” Hence the set of induced transformations has the group property (sce §13). If ¢' im (15.1) are cartesian coordinates, the differentials dx’ are direc tion numbers of a direction in space, the direction being that of the line 515) CONTRAVARIANT VECTORS 75 segment with end points 2‘ and 2‘ + dz\* ‘The differentials dz” given by (15.2) determine the same direction in the z’-system, since a direc tion is independent of a coordinate system. However, whereas when the 2’s are cartesian coordinates the direction determined by fixed dx’ is the same everywhere in space, the corresponding values of dr” given by (15.2) depend upon the point at which the direction is considered, unless av, 3ah = Obs the a’s being constants. From these equations we obtain in this ex- ceptional case by integration equations (13.1), in which case the co- ordinates 2” are cartesian or oblique according as equations (13.2) hold or not. Consequently, although for any transformation of cartesian coordinates into other coordinates by equations not of the form (13.1) differentials dx’ determine a direction at a point, they are not direction numbers of a line in space in the sense that the same values of dz’ determine (he direction of the line at every point of it, as is the case with differentials of cartesian coordinates. In §3 we saw that for a curve with equations (15.5) =f), where the coordinates x‘ are cartesian, the quantities ey ae FO= | evaluated at a point of the curve are direction numbers of the tangent vector at the point. If for a coordinate system x” we put. vy _ dal vo-%, we have in consequence of (15.2) Hay = ep 2 (15.6) BO = 80 55° ‘Thus the quantities £(1) determine at each point of the curve its tangent veetor in the 2/-system. We consider now three functions of cartesian coordinates x‘ which we denote by ‘(z). ‘The values of these functions at each point in space *O.G, pw 76 TRANSFORMATION OF COORDINAT (Cu. TL may be taken as direction numbers of a vector at the point. Such a set of vectors is called a veclor-field. If we put (15.7) da’ = pr’, where p is any function of the 2’s, these values of the differentials at each point of space are also direction numbers of the vector determined by X* at the point. Consider for any other coordinate system 2’ the functions of the 2s denoted by )""(z") and defined by an" (15.8) v@) a v(@) when the 2’s in the right-hand members are replaced by their expressions in the ”’s which define the transformation of coordinates. When we compare these equations and (15.7) with (15.2), we see that (15.9) dz! = pr(2’), that is, the quantities “(2") define in the 2s the veetor-field defined by 2‘(@) in the 2's. Just. as equations (15.3) were obtained from (15.2), £0 equations (15.8) are equivalent to (15.10) YO =e), and thus equations (15.8) and (15.10) are reciprocal in character. For a third coordinate system x”! we have analogously to (15.10) ax" NG@) = ae (aly From these two sets of equations we have fey OF. yrtean, OF wie) SE = wna) Be. a If we multiply this equation by e and sum with respect to i, we have in consequence of (13.14) (a!) a thei O2 NM) = NO) Fre Hence the equations (15.10) have the group property, and consequently in any two coordinate systems x’ and x" the quantities \‘(z) and \"‘(2’), which in the respective systems determine at each point the vector of vector-field, are in the relation (15.10). The quantities x’ and \’' are NE CONTRAVARIANT VECTORS W7 called the components in the z-system and 2/-system respectively of a contravariant vector. As thus defined there is a contravariant vector at each point in space, that is, a vector-field. When the coordinates are cartesian, they are direction numbers of the vector; their geometric significance in any other coordinate system is shown in $16. From the foregoing discussion it is seen that a contravariant vector is entively defined by its components in any one coordinate system, and then its components in any other system are determined. Hence one may assign arbitrary functional expressions to X° in the a-system. For example, one may take d' as constants, and then in general the components in another coordinate system are not constants. If the coordinates 2’ are cartesian and the d‘ are constants, all the vectors are parallel, since they have the same direction numbers. But in a general coordinate system constant components do not define parallel vectors. ‘At times in the consideration of a geometric problem one arrives at equations of the form (15.10) connecting certain quantities \‘ and X’* in any two coordinate systems. In this case we say that the geometric entity thus defined analytically is a contravariant vector whose com- ponents are d' and 2" in the respective coordinate systems. If functions \ and 2" defined at points of a curve and not throughout space, satisfy equations of the form (15.8), \‘ are said to be the compo- nents of a contravariant veetor at points of the curve; an example is afforded by equations (15.6). Consider the differential equations 1 2 2 (15.11) ee where A‘ are the components of a contravariant vector. From the theory of such equations it follows that their integral is given by two equations of the form (15.12) A@,ea)=a, fle’, where the e’s 2) =o, re arbitrary constants,* and that f; and f2 are two inde- Pendent functions which are solutions of the equation of (15.13) ox! For each pair of values of c: and ¢: equations (15.12) are equations of a curve. Through a point in space for which the functions fi and fo * Fine, 1927, 1, pp. 322, 325. 78 TRANSFORMATION OF COORDINATES (CH. Il are single-valued there passes one and only one curve of the family. Such a two-parameter family of curves is called a congruence. From the discussion of equations (15.7) and (15,9) it follows that in any other coordinate system 2" the equations (15.11) are do" det _ a" yi = ye Tyee and that their integral consists of the equations in the x”s obtained from (15.12), when the 2’s are replaced by the functions of the 2s which define the transformation of coordinates. From (15.11) and (15.13) we have oh 46 fe gat = pads = 0, Bar’ =o. From these equations it follows that when the 2’s are cartesian dr’, and consequently \’, are direction numbers of the tangent vector at a point to the curve of the congruence (15.12) through the point. Hence in any coordinate system X‘ are the components of such a tangent vector. If fis any function of x’ and f’ the function of 2” obtained from f when x’ are replaced by the functions of 2’ which define the transformation of coordinates, we have (15.14) Sa’, a 2) = fa", 2”, 2"). Either of these functions is called the (ranaform of the other. From this equation we have af _ af’ aa! On 82% ort From this result and (15.10) we have (15.15) © Of yu 8a! af! a2! sus af se yy a" (15.16) Bat — "ae aa age 7" age =” ga Hence the transforms of the solutions of an equation (15.13) are solu- tions of the equation ao ea Any function f(z’, 2’, 2*) and its transform in any other coordinate system define in their respective coordinate systems an entity called a scalar, Whenever in considering a problem one arrives in two coordi- nate systems at quantities which are transforms of one another in the (15.17) X SCALARS a) sense of (15.14), one says that the entity so defined is a scalar. For example, since either of the first and last members in (15.16) is a trans- form of the athor, wa say that vf ica sealar, meaning that jis transform in any other coordinate system is the analogous expression, in this case af’ “ N ox"i EXERCISES 1, For a linear transformation 2 = ajz’/, where the a's are constants, the coordinates are components of a coutravariant veetor in the two coordinate systems; in consequence of Euler's theorem the most general transformation for which this is true is when gé in (13.10) are homogeneous funetions of the first degree in 2" 2. Show that. in any coordinate systems 2; , 0,0; 0, 28, , 0; 0, 0, x8 ponents of the tangents to the coordinate curves, the subscript of a indicating that it applies to the curve of parameter 2‘, where i has the value of this sub- seript; also that these components may be written w Ma iene where the subscript h! for h = 1, 2, 3 denotes the vector, the 8s are defined by (13.6), and the ¢'s are non-zero functions of the 2’. . Show that in any other coordinate system 2” the eomponents of the veetors (i) of Ex. 2 are given by ‘ aa’ Mil eh gee 4. What are the components in any coordinate system 2” of the vector whose components in the 2's are where j‘ are any functions of z', 2*, z' which involve z! at least? 5. Three contravariant. vectors of components dj; in whieh hl for k= 1, 2, 3 indicates the veetor, are independent, that is, thore are no functions «; of the 2's sueh that @ arty + ashy) + adj, = 0, if and only if the determinant | d4) | is not identically zero. Show that in this ease any vector NM! is expressible in the form dt = DING, + OARS, + AA}, = OA) for suitable values of the 6's as functions of the z's. _ 6. Show that when the determinant in Ex. 5 is equal to zero there exist fune- tions ay, az, as such that equations (i) hold; discuss the geometric meaning of the cases when all the a's are different from zero and when one of them is equal to zero. 80 TRANSFORMATION OB COORDINATE (Cn. UW 7. Given a contravariant vector \¥ in any coordinate system 2‘; show that, a, if ¢' and gare independent solutions of the equation Mt an = 0, andg*is asolution ae ay : of the equation x5 = 1, the jacobian | © | 0; also that in the coordinate = (2) the components of the given veetor are 0, 0, 1 system 2/¢ defined by 2” 16. LENGTH OF A CONTRAVARIANT VECTOR. ANGLE BETWEEN TWO VECTORS Let \‘ be the components of @ contravariant vector in any coordinate system z' and consider the quantity as;\')’, where a.; are the coefficients of the fundamental form (14.6). In consequence of (15.10) and (14.11) for any other coordinate system 2” we have, using (13.14), , x! aa' ax" aa! ant’ = afntyn O2 a" (16.1) ayn = a ant ax” et ax’ = aN" N Ohi = aur N'. Hence for any contravariant vector \' the expression a;;\‘X’ is a scalar (see §15). In cartesian coordinates a,;\‘)’ assumes the form >) \‘X‘, as follows from (14.7). At each point it is the square of the length of the Jine-segment whose orthogonal projections upon the coordinate axes are X; that is, Mf are the rectangular components of the vector. Henee we have {16.t] Jf \' are components in any coordinate system of a contravariant vector and a; are the coefficients of the fundamental form in this coordinate system, the quantity a,;'d’ is a scalar; it is the square of the length of the line-segment whose rectangular components are the corresponding d's in a cartesian coordinate system. Hence in any coordinate system a set of components d define at each point in space a vector whose length and direction are determined by \.. Presently we shall sce what the geometric significance of the components 2° is in any coordinate system. When, in particular, \‘ are such that (16.2) ay = 1, we say that the vector is a unit vector. In this case the components din a certesian system are direction cosines of the vector at each point. For two vectors Aj and 2; we have (16.3) aagdidy = anda, $16] ANGLE BE WEEN TWO VECTORS 8L as follows directly when we proceed as in the case of (16.1). ‘Thus @:MiM is a scalar. If the coordinates are cartesian, in which case the components are direction numbers, this sealar is > Aj. If this quantity is divided by VS nin’) (Soa), the resulting expression is the cosine of the angle of the two vectors.* Hence we have: [16.2] If Xi and dj are the components of two contravariant vectors in any coordinate system, the angle (S 180°) between the two vectors at a point is given by aN (16.4) cos 8 = V (ajXini)(aurXbr’) As @ corollary we have: 116.3] A necessary and. sufficient condition that at each point the contra variant vectors dj and X; be perpendicular is that the equation (16.5) arr = 0 be an identity in 2*. . When the veetors \j and 3 are not perpendicular at every point in space, that is, when (16.5) is not an identity, they are perpendicular at each point of the surface whose equation in the 2’s is given by (16.5). For an 2*-coordinate curve dz‘ = 0 for ¢ ~ k; consequently 4: for k fixed andi = 1, 2, 3 are components of the contravariant vector tangent tothe curve. From (16.4) we have that the angle at a point between the tangents to the 2‘ and 2'-coordinate eurves is given by i 55:54 Ger V (0is5:5) (Gimdt57) Vann Hence we have [16.4] In any coordinate system the cosine of the angle between an x" coordinate curve and an 2’-coordinate curve at any point is equal to the value of ais/V axa; at the point, If at a point a;; = 0 ford, 7 = 1, 2,3 and i ¥ J, the three coordinate curves are mutually perpendicular, and consequently at the point the tangent planes to the three surfaces are mutually perpendicular. In this case we say that the coordinate surfaces through the point are orthogonal to one another at the point. If this situation exists at every *C.G, p. 86. 82 TRANSFORMATION OF COORDINATES LCx, 1 point, we say that the coordinate surfaces form a triply orthogonal family of surfaces. Hence we have [16.5] The coordinate surfaces for a given coordinate system form a triply orthogonal system, if and only if the coefficients a.,(i # j) of the funda mental quadratic form in this coordinate system are equal to zero identically. Suppose now that we consider any vector ‘and find the angle which it makes with each of the vectors (16.6) ¥,0,0; 0, », 0,0,>, which are tangent vectors to the z'-, 2”-, and 2*-coordinate curves re- spectively. If we denote these respective angles by @, for ¢ = 1, 2, 3, we have from (16.4) - aU (16.7) ai 008 6 = HIN. , vi VajNv Since the lengths of the vectors (16.6) are uN, Van’, Vas, the sum of the projections of these lengths upon the line of the veetor \ is in consequence of (16.7) that is, V/a,;\°V which is the length of the vector \°. ‘This means that the veetor d' at a point P is the diagonal from P of the parallelopiped whose sides are the lengths of the veetors (16.6). Hence we have [16.6] For a contravariant vector \* in any coordinate system the geometric significance of the components d' is that the length of the vector at any point P is the diagonal of the parallelopiped whose edges are line segments, wilh P as initial point, tangential lo the coordinale curves al P and of the respective lengths Jai’. When the coordinate curves through a point are mutually perpen- dicular, the parallelopiped is rectangular, but only when the coordinates are cartesian are X‘ the lengths of orthogonal projections of the length of the vector upon the tangents to the coordinate eurves at the point EXERCISES 1, When space is referred to polar coordinates (13.8), the coordinate surfaces form a triply orthogonal system. S17] COVARIANT VECTORS 83 2. Determine the character of the coordinate surfaces 2 = const. for the transformation Ps ateosr?, a= 2%sinz, 2 = 2, and show that they form a triply orthogonal system; the coordinates 2‘ are called cylindrical. 3. If Mand a are unit contravariunt veetors perpendicular to one another, (anja — ayeais)M aii at = 1. 4. If wé and y! are contravariant vectors perpendicular to a contravariant, yeetor M4, 80 also up! + vv! for values of u and v are the compont contravariant veetor perpendicular to X°; for two sets of values a: , 0: and ws, U2 such that Usuaassmin’ + (uve + uaviagimte’ + vivzaizvivé = 0 the vectors uy! + vv and uae! + var are perpendicular. 5. If dir are the components of three mutually perpendicular variant veetors, where h for h = 1, 2, 3 denotes the vector and 7 the co then Dial =a 5 17, COVARIANT VECTORS. CONTRAVARIANT AND COVARIANT COMPONENTS OF A VECTOR Given any function f(z’, 2’, x’) in any coordinate system, we have of _ af Bel! oz! x? at! contra~ ponents, (17.1) where /’ is the transform of the given f for the transformation of the 2’ into any other coordinates 2”. When we compare these equations with (15.10) we see that x and x are not components of 2 contravariant veetor in their respective coordinate systems. However, they do belong toa new class of functions di and Xj of the 2’s and 2”s respectively (with indices as subscripts) related thus rox"! (17.2) M=NT- ) ) ag TF one has two sels of functions so related and multiplies these equate tions by a and sums with respect to i, one obtains 1x" Ox "act ax" 1 rep yt (17.3) a No = M, which shows that the relation is reciprocal, as was shown to be the case with contravariant vectors (see Ex. 1). 84 TRANSFORMATION OF COORDINATES (Cu, II ‘Two sets of functions d, and dj of the z's and 2”s related as in (17.2) are said to be the components in their respective coordinate systems of a covariant vector, there being a vector at each point of space. We observe that the indices of a covariant vector are written as subscripts, whereas for a contravariant vector they are written as superscripts, and that the partial derivatives enter in different manners in (15.10) and (17.2); but that in each case the dummy index applies to one coordinate system and the free index to the other. As remarked in the case of contra- variant vectors, a covariant vector is completely determined by its com- ponents in one coordinate system, and its components in any other system are determined by the above equations, Also whenever one has, no matter how derived, two sets of functions satisfying equations (17.2) or (17.3), one concludes that the entity under consideration is a covariant vector, For example, from (17.1) it follows that for any function f of the 2’s the derivatives a and the derivatives with respect to 2" of the transform of f are components in their respective systems of a covariant vector; this covariant vector is called the gradient of f. Hence we have (17.1) The gradient of a scalar is a covariant vector. If \‘ and X’ are components of a contravariant vector in coordinate systems 2’ and 2", we have from (14.10) and (15.8) , ax"* an” = ae art (17.4) ayo 22 = ayn! 2% NO, By = Ga! Se We note that these equations are of the form (17.3), which means that the linear combinations a,,\* of the components of the given contra- variant vector in the e-system, and the linear combination aj," in the x'-system are components in their respective systems of a covariant veetor. In this sense we have {17.2] If \* are the components of a contravariant vector, a;;\° are the com- ponents of @ covariant vector. From (14.16) and (17.3) we have Man ag! aly! = git 8" aa) ae! gi7] CONTRAVARIANT AND COVARIANT COMPO) Comparing this result with (15.8) we have [17.3] If ds and dj are the components of a covariant vector in their respec tive courdinate systems, ad; and ald are the components in these respective systems of a contravariant vector. When we apply this theorem to the covariant vector of components ai’, we have in consequence of (14.13) a® aj? = 8? = 0, that is, we obtain the veetor \‘ from which the covariant vector was derived in accordance with theorem [17.2]. Similarly, if we start with a covariant vector dz, find the corresponding contravariant vector by theorem [17.3], and then from it the corresponding covariant vector by theorem [17.2], we obtain the original vector A; . In view of the above results we say that X* and d; are the contra- variant and covariant components respectively of the same vector, if (17.6) Me = ajar’, either of which set of equations, as we have seen, implies the other. When the coordinates are cartesian, in which case a; = 3:;, a = 8", the corresponding contravariant and covariant components are equal (see Ex. 6) and are direction numbers of the vector, as shown in §15. In consequence of the first of (17.6) and (14.13) we have (17.7) aA; = a%ara rt = Had = apr' Since the last of these quantities is a scalar, as shown in §16, the first quantity is a sealar. From this result and theorem [16.1] we have [17.4] The square of the length of a vector whose covariant components are Ys is equal to the scalar ad; . Tf dsj: and doi; are the covariant components of two vectors, and Xi, and i; their respective contravariant components, by 2 procedure similar to that usod in (17.7) we obtain (17.8) a Ryders = ae AIM] - From this result, equation (17.7), and theorem [16.2] we have {17.5] The angle @ (S 180°) between the vectors at a point of two vectors ‘whose covariant components are dy; and dai; is given by Nara (17.9) cos § = - oes : Va iid wea) 86 TRANSFORMATION OF COORDINATES (Ca, If ‘The geometric significance of the contravariant components of a vec- tor is stated in theorem [16.6]. Now we derive the geometric signifi- cance of the covariant components. In consequence of the first of (17.6) we may write (16.7) in the form 17.10) V ajed’ 0* cos 8 = ; ‘ : Vai where 6; is the angle which the vector at a point makes with the 2'-coor- dinate curve through the point. The left-hand member of (17.10) is the length of the orthogonal projection upon the tangent to the 2-coordinate curve of the length of the vector X‘ at the point. Hence we have [17.6] The geometric significance of the covariant components dx of a vector is that at each point P d;/+/ax is the length of the orthogonal projection of the vector upon the iangent at P to the x"-coordinate curve through P. In §§15 and 16 we introduced the concept of a contravariant vector and derived properties of such vectors. At the beginning of the present section we defined the concept of a covariant vector. When one com- pares equations (15.10) and (17.2) giving the relations between the components of the two types of vectors in two coordinate systems, one observes that they are essentially different and might conclude that the two vectors are different entities. However, it has been shown with the aid of the coefficients a,, of the fundamental form that the two entities.are in fact identical, but that it is their determining components which have different geometric significance. Were it not for the existence of the fundamental form aj; dz da’, we should be compelled to treat contravariant and covariant vectors as different geometric entities. There are types of geometry in which no such form occurs as part of the theory, and in these geometries a dis- tinction is made, Consider in connection with the above remarks a covariant vector \; From (17.2) and (15.3) it follows in consequence of (13.13) that Dede’ = Xj de", that is, \sdx' is a scalar. Hence the equation (17.11) dz = 0 is of the same form in any coordinate system. If dr! and dex! are two sets of differentials satisfying this equation which are not proportional, it follows from the equ: da dyer §17| CONTRAVARIANT AND COVARIANT COMPONENTS — 87 and (17.11) that any solution of the latter is of the form* dx Hence at cach point in space every direction dx‘ satisfying equation (17.11) is in the plane determined by the point and by the directions dx’ and dex". Consequently we say that a covariant vector determines a plane at each point in space. It is not true that every equation of the form (17.11) admits an integrating factor, that is, a function ¢ of the z’s such that (see Ex. 9) (17.12) tridr' = dg, where y is some function of the 2’s. But when such an integrating factor exists, = const. is an integral of the equation, and we have as equivalent to (17.14) the equation cy dyt* + Cy dye’, 26 agi pad = 0. This is the condition that cach set of differentials satisfying the original equation determines at a point a tangent to the surface ¢ = const. through the point. Hence when the equation (17.11) admits an inte- grating factor, the planes determined by the vector 2; are the tangent planes to a family of surfaces, Suppose now that we invoke the metric properties of space based upon the fundamental form aj; dx‘ dx’, and Ghat we replace X; in (17.11) by aj, obtaining ay de’ = 0. In consequence of theorem [16.3] we have that the directions dz‘ satis- fying equation (17.11) are perpendicular to the vector \°. Consequently at each point the given d; are the covariant components of the normal to the plane determined by \;._ When furthermore the equation admits an integrating factor, \; are the covariant components of the normal to the surfaces ¢ = const. Hence we have [17.7] When a covariant vector is the gradient of a function g, or its com- Ponents are proportional to the components of a gradient, the vector-field consisis of vectors normal to the surfaces p = const. EXERCISES , J: Show that equations (17.2) possess the group property, as defined in §13 | after equation (13.18). *C.G., pp. 115-116, 88 TRANSFORMATION OF COORDINATES (Cu. 11 2. IfX5, wc and N's, ui are the components in the z-system and 2’-system respec- tively of a contravariant and covariant vector, Mui = N'iy; , that is, Muyis a scalar. 3. When 2! are polar coordinates, the covariant components of the vector with contravariant components \¢ are » » “eo Gana 4. If for two sets of quantities u,(z) and u;(z") we have @ BML where the Xj; and Ajj for k = 1, 2, 3 are the components of three independent, contravariant vectors (see §15, Fx. 5), then the a; and 4; are the components in their respective coordinate systems of a covariant vector. Does this follow ‘equations (i) hold for fewer than three contravariant vectors? 5. If Mi for k = 1, 2, 8 are three independent contravariant veetors (see $15, Ex. 5), and N4 is the cofactor of Aji in the determinant | 4, | divided by the determinant, then i! are the components of » covariant veetor fur cach value of A, 6. In order that corresponding contravariant and covariant components of x vector be equal in the coordinate system z‘, the contravariant components A! must satisfy the equations BML» G) (aaj — 84) = 0, that is, the rank of the determinant | a;; — 4;; | must be less than 3. For equa- tions (i) to hold for every vector the coordinate system must. he cartesian, 7. From theorems {16.6] und [17.6] it follows that \s/4/ai; is equal to the sum of the orthogonal projections upon the tangent to the t/-coordinate curve of th lengths of the vectors (16.6). 8. Show that the equations dat _ dt _ det uM ory Ms in one system of coordinates do not transform to equations of the same forim in another system. Compare with equations (15.11). a 9. From equations (17.12) one has 55, = ¢A, from whieh one obtains Oy, ot agi tag NS noting that equations obtained from these by interchanging i and j must hold, and multiplying these equations by ey and summing with respect to ¢ and j, one obtains an in OMe ene as a necessary and sufficient condition that an equation (17.11) shall admit a integrating factor. 3 $18) TENSORS 89 18. TENSORS. SYMMETRIC AND SKEW SYMMETRIC TENSORS The equations (14.10) connecting the coefficients of the fmdamental quadratic form in two coordinate systems are of the type (18.1) Vile’) = dale) 2 oe There is a similarity between these equations and (17.3) in the sense that the indices are subscripts in both eases and that the derivatives occur in similar manner. We say that functions bj: and bi; of the 2”s and z’s respectively related as in (18.1) are the components in their respective coordinate systems of a covariant tensor of the second order, and that a covariant vector is a covariant tensor of the first. order, in each case the order being equal to the number of subscripts, Also we refer to the subscripts in each ease as covariant indices. Similarly equations (14.16) are of the type (at) = pii(q) 22" ax"! (18.2) O(a’) = o%(a) at at” There is a similarity between these equations and (15.8) in the sense that the indices are superscripts in both cases and that the derivatives occur in similar manner, We say that functions / and 6” related as in (18.2) are the components in their respective coordinate systems of a contravariant tensor of the second order, and that a contravariant vector is a contravariant tensor of the first order. Also we refer to the super- scripts in each case as contravariant indices. If \° and w are the components of a contravariant and a covariant, vector respectively, it follows from (15.8) and (17.3) that mot ye on™ aah Mat = Nas og agli ‘These equations are of the type aa! 18.3 (a) = wi (x) Om. Oe (18.3) whe) = Via) ZS We say that functions b”, and b'; are the components in their respective Coordinate systems of a mized tensor of the second order. The com- Ponents have one contravariant and one covariant index, and so we say that the tensor is contravariant of the first order and covariant of the first order. 90 TRANSFORMATION OF COORDINATES {Cu. IT From (18.1) we have 1 ax! ax! _ aa’ aa? an" an! (1s) "oak gam opm aa" aa Ozm = bydhbL = bin. In like manner, we have from (18.2) and (18.3) (18.5) and (18.6) a These results show the reciprocal character of the equations of tensors of the second order, that is, all coordinate systems are on a par. Equations (17.3) and (18.1) are particular cases of the equations soi tin (48.7) Direct = Bevsgs oct the b’s and b”s being functions of x's and x”’s respectively and having m indices (subscripts), where m is any positive integer. In this case the b’s and b”s satisfying these equations by means of the transformation equations are said to be the components in the x-system and 2’-system. respectively of a covariant tensor of the m*® order. Equations (15.8) and (18.2) are particular cases of the equations tm D2!" al" galt Qe" an? | ag'm? ~ (18.8) Berm = Bt the b’s and bs being functions of 2’s and '’s respectively and having m indices (superscripts), where m is any positive integer. The b’s and b”s satisfying these equations are said to be the components in the a-system and 2’-system respectively of a contravariant tensor of the m'” order. Equations (18.3) are a particular case of the equations sesnm 08 onl aut an” a ax ar” ax? (18.9) eae the b’s and b’’s being functions of 2’s and x's respectively and having m upper indices and 7 lower indices, where m and n are positive integers. ‘The b’s and b’’s satisfying these equations are said to be the components $18] TENSORS 91 in the a-system and z/-system respectively of a mized tensor of order m + n, contravariant of order m and covariant of order n. We observe that in each set of equations (18.7), (18.8) and (18.9) the number of sets of partial derivatives entering in the equations is equal to the order of the tensor, but the way in which they enter depends upon whether the indices are contravarian! (superscripts) or covariant (subscripts). Tn the case of a scalar, as defined in §15, there aye no such sets of derivatives in the one equation (15.14) which expresses the equality of a function and its transform. Hence a scalar is called a tensor of order zero. When one applies to a tensor of any order and type the processes used to obtain (18.4), (18.5), and (18.6) from (18.1), (18.2), and (18.3) respectively, one obtains equations which are the inverses of (18.7), (18.8), and (18.9) respectively. This shows that the equations giving the relations between the components of any tensor in two coordinate systems are reciprocal in character, and thus that no one coordinate system has a preferred position in the definition of a tensor From the above definition of tensors it follows that one may choose arbitrarily the components of a tensor in one coordinate system, and then the components in any other coordinate system are determined by (18.7), (18.8) or (18.9) (see Ex. 1). Frequently in the consideration of a geometric problem we deal with a geometric entity and find that its (analytical) components in two coordinate systems are related as in (18.7), (18.8) or (18.9). ‘Then we say that we are dealing with a tensor. For example, at the beginning of this section we observed that equations (14.10) are of the type (18.1), and so we say that the coefficients of the fundamentak quadratic form are components of a covariant tensor of the second order, or briefly that @;, is a covariant tensor of the second order. We call it the covariant metric tensor, because, as we have st it enters into the determination of arc lengths, magnitudes of vecti and of angles. In like manner, we call a the contravariant metric tensor. Because of the linear homogencous character of equations (18.7), (18.8) and (18.9) and their inverses we have 8.1] Tf all the components of a tensor are equal to zero in one coordinate system, they are equal to zero in every system. Such a tensor is called a zero tensor. From the form of equations (18.7), (18.8) and (18.9) it is clear that the relative order (position) of the indices plays a role in these equations, It may be, however, that in the case of certain tensors when two contra- Variant (or covariant) indices are interchanged the new component is 92 TRANSFORMATION OF COORDINATES (Cu, IT equal to the original one. In this case the relative order of these particular indices is immaterial. From the form of equations (18.7), (18.8) and (18.9) it follows that if this is true for certain indices in one coordinate system it is true for the corresponding indices in every system. For example, suppose that Dejagsy--otn = Gneisyss-nm, then from (18.7) we have , Ox"! ax" . aat™ Meatats-sig = Dageyey ote ~ (18.10) az"! ax" "ani = Drarirassorae When the relative order of two or more indices is immaterial, we say that the tensor is symmetric with respect to these indices. When the relative order of all the indices is immaterial, the tensor is said to he a symmelric lensor. Thus the metric tensors a;; and a’ are symmetric tensors. A general teusor of the second order, whether contravariant or covariant, has 9 different components, whereas a symmetric tensor has only 6 different components. When for a tensor two components obtained from one another by the interchange of two particular indices, either contravariant or covariant, differ only in sign, the tensof is said to be skew-symmetric with respect to these indices. It can be shown that if a tensor is skew symmetric in any two indices in one coordinate system, it has this property in every system. For example, if the tensor be.»....2, i8 skew-symmetric in the first two indices we have (18.10) with a minus sign in the second and third members of these equations, When a tensor, whether contra- variant or covariant, is skew-symmetric with respect to every pair of indices, it is called a skew-symmetric tensor. . From equation (14.12) and the definition of a scalar in §15 it follows that @ is not a scalar. When one has two functions b’ and b of the x"’s and 2’s such that bia!) = b¢ we say that b is a relative scalar of weight p. ‘Thus the determinant a of the components ay; of the covariant metric tensor is a relative scalar of weight 2, and from (14.15) and (14.12) it follows that the determinant of the contravariant metric tensor is a relative scalar of weight —2. A relative scalar of weight 1 is called a scalar density. In similar manner if the b”s and 6’s are such that instead of equations p in | (18.7), (18.8) and (18.9), we have equations with the factor lw $18} =SYMMETRIC AND SKEW. YMMETRIC T SORS 93, the right-hand members of these equations, we say that the bs and b's are the components in their respective coordinate systems not of tensors but of relative tensors of weight p of the order and type determined by the character of their indices. In consequence of (13.15) and the above observation concerning the reciprocal character of equations (18.7), (18.8) and (18.9) it follows that the corresponding equations for relative tensors are reciprocal. | ax For a transformation of coordinates for which the jacobian & is positive, we have from (14.12) (18.11) Va = Val % an" If the jacobian is negative, by a change of the sign in one of the equa- tions of the transformation the resulting jacobian is positive. Hence there is no loss in generality in understanding that equation (18.11) holds. ‘Thus +/a is a scalar density. From (18.11) and the definition of relative tensors we have: [18.2] If b51:..7% are the components of a relative tensor of weight p, then "/a*? are the components of a tensor. EXERCISES 4. If the Kronecker deltas 8} are taken as the components of a mixed tensor in one coordinate system, the components in every other coordinate system are of the same kind; that is, 6} is a mixed tensor of the second order. 2. If bi; and e* are components of covariant and contravariant tensors respec tively, the quantities 5;;c/* are components of a mixed tensor of the second order, and bisc' is a sealar. 3. If bis and eg; are two symmetric tensors such that bse — daca + bac — bucsy = 0, then ci = pbis, where p is a scalar. 4. How many functions are required to define a skow-symmetrie covariant, tensor of the second order? 5. For any skew-symmetric tensor all the components having two, or more, indices alike are equal to zero. 6. If by; is a skew-symmetric tensor and di is a contravariant veetor, then bs = 0; conversely, if by! = 0 for an arbitrary vector, bj; is a skew symmetric tensor. . | ax | ani Oxi Oat 7. Since eam 9) 7 8 aan Bal aa are the components of a relative tensor of weight —1; ¢** are components of a Telative tensor of weight +1 (see §13, Ex. 4), the quantities ecm 94 TRANSFORMATION OF COORDINATES, (Ca. IL 8. The quantities are the components of covariant third order, when the jacobian (13.11) is positive, 8. The rank of the determinant | b,; | of a tensor by, is called the rank of the tensor; show that the rank is invariant under any trausformation of coordinates. }0. When the rank of a covariant tensor bj is three, the cofactors of bs, in the determinant (b;;| are components of a relative contravariant tensor of weight two. 11. When the tank of a covariant tensor b1, i8 two, there exist two relative contrayariant vectors Mand i each of weight one, such that the cofactor of bj) in the determinant | b;;| is equal to dw; when Bi; is symmetric, M and wi are the same vectors. 19, ADDITION, SUBTRACTION AND MULTIPLICATION OF ‘TENSORS, CONTRACTION From the form of equations (18.7), (18.8) and (18,9) it follows that the sum or difference of two tensors of the same type and order is a tensor of the same type and order. The same is true of any linear homogeneous combination of tensors of the same type and order, the coefficients being constants or scalars. If we take two tensors of any type and order, and form all possible products of a component of one tensor and a component of the other, we obtain a tensor whose order is the sum of the orders of the two given tensors. ‘The number of contravatiant, or covariant, indices is equal to the sum of the numbers of contravariant, or covariant, indices of the given tensors, For example, we have from (18.9) as et OF On Ox” On Oc” (19.1) Oey = been Se SF Gat Bgl Ba? and thus b%ic‘n are components of a mixed tensor of order 5, contra- variant of order 3 and covariant of order 2. ‘This process is general, so that by multiplying the components of any number of tensors we obtain a tensor, culled the product or outer product of the given tensors, which is contravariant and covariant of orders which are the respective sums of the contravariant and covariant, orders of the tensors multiplied. Another process for obtaining a tensor from a given tensor, or 2 product of tensors, is called contraction. We have used this process in obtaining the covariant components of a contravariant vector in equa- tions (17.4). Thus the product ai,\° of the metric tensor a;; and the vector \* is a’ mixed tensor of the third order, contravariant of the first $19] CONTRACTION 95 order and ‘covariant of the second order. Each quantity a,,\' of this tensor for particular values of j is the sum of three components of this mixed tensor, and from (17.4) it follows that these quantities are com- ponents of a covariant vector; that is, by the summation of one contra- variant and one covariant index we have obtained a tensor of order 2 less, the contravariant and covariant orders being each one less than for the original tensor. This process is called contraction. It applies also to any mixed tensor, and we have [19.1] By the contraction of any contravariant index with any covariant index there is obtained a tensor of order one less contravariant and one less covariant, For example, ax’? oat ax aa! a2” da Ox ax’ Oa’? Ax’ (19.2) : i gt OR" ax" Oa™ _ az du"? ax* ax” ‘md te oF ga pale x? Bx" ax!* When, in particular, contraction is applied to the product of two tensors, the resulting tensor is called an inner product of the two tensors. The process of contraction may be applied in more than one way, and more than once. ‘Thus from (19.2) we have oF O45! = bitin + When, in particular, we apply contraction twice to the product tensor a;°x', we obtain the scalar a;;\'n’ which by theorem [16.1] is the square of the length of the vector \*, This is a particular case of the following theorem, which is a consequence of theorem {19.1}: [19.2] When as the result of contraction of one or more pairs of indices there remain no free indices, the resulting quantity is a scalar. An application of contraction, which is used frequently in tensor cal- culus, is what is called lowering » contravariant index by means of the covariant metric tensor ay; and raising a covariant index by means of the contravariant metric tensor a‘. This process was used in obtaining the covariant components of the vector \’, and the contravariant com- Ponents of the vector 4. In carrying out this process it is important 96 TRANSFORMATION OF COORDINATES ix that the position of the index affected be not ambiguous. For example, we have the following tensors derived from the tensor ise ; aoa Wy = aibins bby = albany bat =a bins bm, = alain 5 BY = aan 5 DIY = aaa bie In similar manner we have ik be = ab; Ban’ = asd imb™5 Damp = Air jndyd™. We remark that this process is reversible. ‘Thus from the first of (19.3) we have inde = ind bin = Indije = bik y which is the tensor from which h!,, was obtained At times in order to indicate the position from which an index has been raised or lowered a dot is placed in the original position of the index; thus the first of (19.3) would be b!.. This notation emphasizes the position of an index. Instead of referring to the quantities b’jx, bi‘s, --» , b! in (19.3) as different tensors, we shail say that the quantities of each set are a sct of components of the same tensor, Any set of components deter- mines the tensor, but in different manner according to the character of the indices, as was seen in §17 to be the case with the contravariant. and covariant components of a vector. When we apply this process to the tensor aj; itself, we obtain a'aay = fa" = a", and hence we refer to a;; and a” as the covariant and contravariant components respectively of the metric tensor. . Let now bi{» and b/?/" be functions of «’ and ©” respectively, such that biL ws! and b:2X" are components in their respcetive coordinate systems of a tensor, and X' and \’* components of a contravariant vector. From this hypothesis it follows that , ij 1 20" ax"? ax* ax” ax! ax’? ax” BIBENM = bE n! 222 28% Oa Oe” _ yas ye Dal Oa’? a” “ Mim “at Bah at Gxt Mad ax ae that is, =0. at) — beim = pize — pi, 20"? ax" act ax! ax”) 4 ax* da! ax’? ax’* az’, Tf these equations hold for every contravariant vector \‘, and conse- quently for any three independent vectors, the quantity in parentheses $19] CONTRACTION 97 is equal to zero, and hence biim is a mixed tensor of the fifth order. By a similar argument we have [19.3] If a set of functions bie" and BEY are such that Bi. p XP and BYNM. aN for any ps and qx are components of a tensor, where \' and X"* are components of an arbitrary veclor in these respective coordinates, then the given functions are com- ponents of a tensor. 2m of at and x! respectively A similar theorem holds if \' is replaced by any arbitrary tensor, and a covariant, or contravariant, index is contracted with a contravariant, or covariant, index of {the given functions. Indeed, it suffices to take ‘a tensor which is the product of distinct arbitrary vectors, in which case the result follows by repeated application of Theorem [19.3]. This is sometimes called the quotient law of tensors. EXERCISES 1. Show that it follows from the identity big = B(bay + ja) + Albis — B58) that any covariant tensor of the second order is the sum of a symmetric and a skew-symmetric tensor; is the same true of a contravariant tensor of the second order? 2. UE bay = basis + bass» Where bitsy is symmetric in the indices and bsi 4; is skew-symmetric, then bay dt dz — byryy dé di. 3. If byAAM is a scalar for Xi an arbitrary contravariant vector, then biy ++ by: are components of a tensor; if b;; are symmetric in the indices, then b,j are the components of a tensor. 4. Tf bjMuing is a scalar for arbitrary vectors M, u‘, and vz, bh; is a tensor. 5. If bie de! def dz* = 0 for arbitrary values of the differentials, then basa + b2ar + bars + bias + baer + bara = 0; what is the condition when bse is symmetric in i and j? 6. If bY are components of a tensor, and ab‘ + cbi# = 0, where a and ¢ are sealars, bf is cither a symmetric or skew symmetric tensor. 7. When the coordinates are cartesian, the various components in (19.3) with the same values of the indices are equal. 8. Show that [asi Gbi2 atria Gigiy Aisin Gigia| = CéitatsCirizia lly [Gisiy isis isis and consequently these determinants are the components of a covariant tensor of the sixth order which is the outer product of the tensor ei (§18, Ex. 8) with itself. 98 TRANSFORMATION OF COORDINATES (Cu. IL 20, THE CHRISTOFFEL SYMBOLS. THE RIEMANN TENSOR At times equations involving the first derivatives of the components of the metrie tensor are given simpler form by moans of the following symbols: 0x) Wi = (r+ Se 3), AY ayes (20.2) {i} = ali, Hl. are symmetric in 7 Observe that from their definition [i,k] and { andj. The symbols defined by (20.1) and (20.2) are called the Chris- toffel symbols of the first and second kinds respectively.* We now derive equations involving these symbols which are of frequent use. From (20.2) and (14.13), namely 203) atan = 0, we have (20.4) a {i = stfii, BL = Gis Also from (20.1) we have Bas oat (20.5) Uj, Al + [Aj a. Differentiating (20.3) with respect to z', we have aa" i aa, 7. Multiplying by a’, summing with respect to k, and substituting from (20.5), we have is ich at , St = mal alt 208 = a ail, Hl + Ls, D, from which we have, in consequence of (20.2), my (orf 4 afi}, xt *The oe of these symbols as defined by Christoffel, 1869, 1, p. 49, were {i i] and {f 43 hut we have adopted the aboye forns because they are in keeping, with the summation convention. §20} THE CHRISTOFFEL SYMBOLS 99 If we denote by A” the cofactor of a;;in the determinant a = | a; | and apply to this determinant the rule for the differentiation of a determinant, we have the last expression being a consequence of (14.14). From this result and (20.5) we have aa ax* = an 28 = alee, 1+ Li, fl) = mo{i}, and consequently aloe Va _ fi 20. eva (20.7) 3a ikl” where since 7 is repeated the summation convention applies, that is, the right-hand member is the sum of three symbols. Dummy indices often occur in expressions involving Christoffel symbols, We now find the relation between the Christoffel symbols of the second kind in two coordinate systems. To this end we differentiate with respect to <” the equations aa! ae (208) pe = Ou Gale G4,P, 7 = 1,2,3), and obtain. ayy _ ai; ax! ax’ a ac a’ ae! ax ark ax’? ax’? ar’ i\ ax"? are arn't " ax!@ ax ax! By suitable changes of free and dummy indices in the above equation, we have the following equations Gry _ dau; ax aa’ ax” (& ae ax? aa ) aa’? ~ “axt a's Gx" Ba "aa da’ean'® "axl? Bx? ar" Baye _ dain ax’ ax’ ak act Pe ae oe! ae’ ant ax’? ar’e ax * 9G anvean * ax” ax"? axe If from the sum of these equations we subtract (20.9) and divide the resulting equation by 2, we have in consequence of (20.1) ac! ac’ axt ax a! 20. y= an ax”? dard’ (20.10) [pg, rl’ = tii, ax"? ax’ Ox | ax’ agian? where [pq, r]' is formed with respect to the tensor a/; . 100 TRANSFORMATION OF COORDINATES 1Ca. 1. From (14.16) we have If we multiply the Teft- and right-hand members of equation (20.10) by at 2 and d aa” consequence of @. 2) , tospeetively, and sum with respect to 7, we have in ax’ ax’? ac’a” which reduces to h\ act ae awe t {i} 2s ant = ‘These are the relations which we set out to obtain. On comparing (20.10) and (20.11) with (18.7), (18.8) and (18.9), we see that neither (20.11) (ij, k) nor u are components of a tensor (see Ex. 5). If we differentiate equation (20.11) with respect to x”, and subtract the resulting equation from the one obtained therefrom on interchanging the indices q and r, we obtain (2,(4) —as(8) a a aa \ikf ~ aa® aif) an an" aa! ax? ee ow) ij bah F ~ are ant ax’, = (218) = 30% \pr - aa” 4 (ey ae -{¥ ad pr) acon \pal aa! ax"** On substituting for the second derivatives in this equation their ex- pressions from equations of the form (20.11), we obtain a ax’ ax’ ant aa (20.12) ‘tt Sale gala agin = "par Zia {e+ eG} {aie where by definition asfh\_ a ‘ani lik) ~ (20.18) Rn §20] THE RIEMANN TENSOR 10t (I being a dummy index indicates summation), and where Rpg is the similar expression in the Christoffel symbols formed with respect to a's; . Tf equations (20.12) be multiplied by i and summed with respect. ier to h, we have ax‘ ax’ aa aa! 4 ee ph, Oe. axe) aa* ant (20.14) Rl ye = Rin 75 50% ga From the form of these equations it follows that R’:j,, which are called Riemann symbols of the second kind, are the components of a tensor contravariant of the first order and covariant of the third order (see (18.9). It is called the Riemann tensor of the fourth order. From (20.13) it follows that this tensor is skew-symmetric in the indices j and k. The quantities Rij which are defined by (20.15) Rue = anki, Rin = a Reese are called Riemann symbols of the first kind. ‘They are the components of the Riemann tensor with all four indices covariant. When the coordinate system is cartesian, the coefficients of the funda- mental form are constants, namely 6,;. In this case the Christoffel symbols of either kind are zero, as follows from (20.1) and (20.2). From (20.13) it follows that in this ease the components of the Riemann teusor are equal to zero, and by theorem [18.1] that the components: are zero in every coordinate system. Hence we have [20.1] The Riemann tensor of euclidean 3-space is a zero lensor. This does not mean that the Christoffel symbols in any coordinate system are equal to zero, but that the functions (20.13) of these symbols are equal to zero. If g:; is any symmetric covariant. tensor such that the determinant g of gi; is not zero, that is, (20.16) 9 = || #0, quantities g’’ are defined uniquely by (20.17) oGn = 5. As in the case of a“ in §14 it can be shown that g’ is a symmetric con- travariant tensor. Similarly to (20.1) and (20.2) we can define Christoffel symbols by means of giz and g” in which case it is advisable at times to use the 102 TRANSFORMATION OF COORDINATES: (Ca. 1 notation f jet to indicate this fact. Likewise a Riemann tensor in terms Lak} of these symbols can be defined by (20.13). Ordinarily the Riemann tensor so defined is not a zero tensor (see §23). EXERCISES 1. Show that Bac; Ouse Bae gat = Like — Ui Bh. 2. Show that if'aj; = Ofori #9 f| _ 1 alogass AY 1 a dow ass vif 2 ot wf? azi {:} 1 bay {ay - GikA, ~ Bas: Oat” \ik whore a repeatad index does not indieate summation and the notation (i, j, # #) means that no two of i, j, & are equal 3. When 2 are potar evordinates (see (14.8)) all the equal to zero except. the following {i}={ihe Gabon hese {3} = coz, {a} = sin 2 cos 2, 4, When af are cartesian, one has from (20.11) stoffel symbols are O28 en, ci Tobe age POT. Show that from these equations one may derive equations of the form (20.11) in general coordinate systems 2" and 2", 5, For any linear transformation (13.1) the Christoffel symbols of the second kind are related 2s components of a mixed tensor of the third order. 6. Show that for Christoffel symbols formed with respeet to any tensor gsi satisfying (20.16) a fh\ & [hl ig. . 9 {i = agi fh ~ a (thi, U + 17, hl). 7. Using (20,15), (20.13) and Ex. 6 show that h ® fun = 35 teh 2 t+ {i} te, a) — {i} Ww, » 1a) aa! {i} ax! §2l] FRENET FORMULAS LN GENERAL COORDINATES — 103 where the symbols [ij, KH] are formed and that in terms of gi; we have 1( 90 | Bq, Pow mas) h respect to any gi; satisfying (20.16); a Bux 9\ ariast + data ~ Oxidzt ~ az Oxt + g'™ (lis, m)lUk, hl — (ak, m){lj, Al). & Show that it follows from Ex. 7 (ii) that @ Rise = Raye = —Roay = Raw 5 and Gi) Racin + Rigas + Rig = 0. 9. Show that Rh = 0. 10. From (20.14) it follows that for the quantities Rj defined by Re = Ria = oR icin one has dat bus Re ae ane and consequently Ry; are the covariant components of a tensor of the second order, which is called the Ricci tensor;* from (20.13) and equations analogous to (20.7) one has R Blogvg_ 2 fh + t ne 1) 8 log Vo " Or'dxi ant Vig anf Gf Vi, on where g is defined by (20.16) u.Tf ik and ie aro Christoffel symbols formed with respect to sym= Wa wy, metric tensors ai; and gi; such that| aj;| 40 and | gi; | #0, then the quantities anh ~ {it are components of a tensor of the third order. ae 21, THE FRENET FORMULAS IN GENERAL COORDINATES If the coordinates 2" are cartesian and 2‘ are any general coordinates, equations (20.11) become avian * Vf at ac * See 1904, 2, p. 1234. 104 TRANSFORMATION OF COORDINATES (Cu. If since the as are constants in the this case and consequently the Christoffel symbols {3} are all zero. a We consider now a curve with the equations SW. When these expressions are substituted in the equations «° = g(x”, 2”, 2”) con- necting the two sets of coordinates x‘ and «”', we obtain equations of ¢ the curve in terms of 2‘ and ¢, The derivatives & are given by as! ‘ as! ax’ dx? i de? (21.2) from which it is seen that — are the components of the contravariant. dt dz! . K . vector whose components —- in cartesian coordinates x’' are direction dt numbers of the tangent to the curve. Differentiating equations (21.2) with respect to £ we have da! ata! de de”, ae’ de” de ~ ax'ian® dt dt" aa db? which in consequence of (21.1) may be written in the form fil dat dx! _ d’x az! Ce a +i fdt dt de a Since for cartesian coordinates 2” the symbols {7 yup Bre wero, ie my d be given the same form ax the leftehand member of (21.8), namely a", [5\' de da de kif dt at” (15.6) and (15.10), we have [21.1] For a curve defined im terms of any coordinaics 2 the quantitics me i ad bt dx" are components of @ contravariant vector, meaning that Viki dt dt in Many coordinate system the components have this form. Consequently on comparing this result with If the are s is the parameter we have from (21.2) and (8.2) da de aft * (21.4) where a’ are components in the «”s of the direction cosines of the tangent to the given curve, and consequently of the unit vector tangen! §21) FRENET FORMULAS IN GENERAL COORDINATES 105 to the curve. From equations (21.4) and (15.10) it follows that a’ defined by ‘ gett ds are the contravariant components in the 2’s of the unit vector tangent to the curve. This result is in keeping with the observation in §15 that when a coordinate system is cartesian the contravariant components of a unit vector are direction cosines of the vector. Accordingly if 8” and y"" are direction cosines of the principal normal and binormal of a curve defined in cartesian coordinates 2", that is, their contravariant components in these coordinates, the contravariant components of these respective unit vectors in any coordinate system 2’ are given by (21.5) 15 dae" (21.6) a agi? In cartesian coordinates 2!‘ we have from (4.7) (21.7) <> = «8", where « is the curvature of the curve. In any coordinates x‘ in conse- quence of (21.6) and (21.7) we have from (21,3) del, [ilar det as: T \jkf ds ds "8° which beeauise of (21.5) may be written daw’, jt aus (21.8) dst {fe ds KB. If we differentiate equations (21.6) with respect to s, and make use of the Frenet formulas (6.1) and equations (21.1) and (21.6), we obtain Bath gg? 1a! del ds 7 GR ty +o aera ds a 5 aa" ax! da ~ Goal 9 BE = we Eh Oe, = (wa + ry!) — oo et we 106 TRANSFORMATION OF COORDINATES {Cu f From these results and (21.8) we have 21.2] When a curve is defined in terms of general coordinates x‘, the Frenet formulas are da | Jil saat _ tet {ite a 19) aw {; jy OSE = ~Coa! + 099, ds we fil ft ag + ne aR” ° where a’, B', andy‘ are the contravariant components of unit vectors having the directions of the tangent, principal normal, and hinormal. respectively of the curve, being the arc of the curve, and x and 7 the curvature and torsion respectively. Consider now the equations of motion in cartesian coordinates 2” met av (21.10) mae = Waar of a particle of mass m in a field of potential V. From Theorem [21.1] it follows that in general coordinates the left-hand members of these equations are the expressions given in Theorem [21.1] multiplied by m, and that they are contravariant components of a vector. Consequently the right-hand members in general coordinates must appear as contra- variant components of a vector. In §17 it was shown that in any coordinate system av are covariant components of a vector and that 1 2V are its contravariant components. Honce in goneral coordinates x’ equations (21.10) are dat | fi \ de’ de® av (21.11) n(@ +{5)% ea) = 8 Sa Another way of arriving at this result is to note that fi \ da! aa! ig OV mn (Se + Gel a wea at are the components of a contravariant vector. From (21.10) we have that the components of this vector in cartesian coordinates are equal §22] COVARIANT DIFFERENTIATION 107 to zero, and consequently they are equal to zero in any coordinate system, and one obtains (21.11) in any coordinate system. EXERCISES 1, In any coordinate system the equations ae {ipa 0 dst“ \SKI ds ds ~ are the differential equations of the straight lines iu space, 2. When 2 are polar coordinates equations (21.11) are (see §20, Ex. 3) eat de\? . av m| — — (=) — asin e(=) |=, de a dt oat [a 2detde (ay) 1 av m +4 — sin 2t eos 2 - de a dt dt a “ean” da | 2de'dz dxt da* 1 ave mf {2422 te ae . [se Bat dt 7F ral (et sin wt) Ae 3. LEA‘) are components of a contravariant vector at points of a enrve 2'(s), the quantities @ ° i as + (id as are components of a contravariant vector at point of the curve. 4. If M(s) in Ex. 8 are such that the quantities (i) are identically zero, the veetors Mare parallel, that is, the components of the vectors in any cartesian system are constants, 22. COVARIANT DIFFERENTIATION By theorem [17.1] we have that the partial derivatives of a scalar, that is, a tensor of order zero, are the components of a covariant vector, that is, a tensor of the first order. In this section it will be shown that this is the ouly case in which the derivatives of a Lensor are components of a tensor, but at the same time we shall find expressions involving derivatives of a tensor which are components of a tensor. We consider first a contravariant, vector d‘, and differentiate with Tespeet to 2’ the equations + = ye Ott (22.1) _ a with the result an an’? ar'® ax’ an Ot Oa” 22. ee belle (22.2) ae an aa aa’y aa’ an" ae!" 108 TRAN SFORMATION OF COORDINATES [Ca IL In consequence of (20.11) the right-hand member of this equation is equal to ty oat _ {2 )\ ark ark yf ax ~ \nkf ax aa”, an”? py) art fi) aan’ - a me (B+ 8 a )s x ~ \ak _ (x? So \z ac fila =e {raf ) ax ax ~ Yep ® 8 Hence equation (22.2) may be written Maa (2 a ae RY) a a aa Batis ete iat) are a Consequently, if we define 4, by Ie 22.4) x ne (22.4) 7 x n+ fi. and similarly \”,,, equation (22.3) is ayn, 2H a2" az’? ari” Hence by (18.6) \‘; and \’”,, are the components in their respective coordinate systems of a mixed tensor of the secoud order. We say that the component \‘; is obtained from the contravariant veetor \‘ by covariant differentiation with respect to x’. Tf we differentiate the equations aa de az’ with respect to 2” and make use of (20.11), we obtain ac’ ax’ (22.5) ee where : (22.6) >. (i, J) and similarly for \’,,- Consequently d,, are components of a covariant tensor of the second order, which is said to be obtained from the co- variant. vector A; by covariant differentiation. s2] COVARIANT DIFFERENTIATION 109 Consider next the covariant tensor b;; , and differentiate with respect. to 2” the equations 22.1) Oe = bu ax'? aa’e Making use of equations of the form (20.11), we obtain Doo _ aby Ox" ax’ ax* 4b LY’ axé art m) ax’? ark ax'e ar't ax * nt pr} ax - Vi an”? ar” az! ({1\! ax! a" ar! + aye Ge Saat {i ax =): from which by means of (22.7) and suitable choice of dummy indices we have motel {i} ax” prf 0" \or abi; 4 — tnd? ax’ ax’ aa* art OM i jk) aa”? ax" aa" i {n\ fh Ae and U'pey for the left-hand member of the above equation, we have that bé;,, are components of a covariant tensor of the third order, which we say is obtained from bs; by covariant differentiation. By proceeding in like manner one may show that Hence, if we put (22.8) bin = (22.9) ve oe + wy ‘ + wld il, and Zé a 1" vf h (22.10) Via e+ aah (ih—s fa are mixed tensors of the third order, which we say are obtained by covariant, differentiation By referring to (22.4), (22.6), (22.8), (22.9) and (22.10), one observes that in every case covariant differentiation is indicated by a covariant index preceded by a comma; that each expression contains the deriva- tive of the original tensor with respect to 2‘, where k is this index; and 10 TRANSFORMATION OF COORDINAT ss [Cn that corresponding to each contravariant index there is added a term involving a Christoffel symbol, and corresponding to each covariant index there is subtracted such a term. All of these expressions are particular cases of the following general rule for covariant differentiation: BEI, = (22.11) (22.12) 2 may) + (hy. we see from (22.8) that (22.1) aja = 0. ‘Also when equations (20.6) are compared with (22.9), we see that (22.14) ay = 0. By §18 Ex. 1, 8) are the components of a mixed tensor. From (22.10) it follows that (22.15) of, = 0. Hence we have {22.1] The tensors ai;, a’, and 8} behave as constants in covariant difer- entiation. When the Christoffel symbols are formed with respect to any tensor gi; such that the determinant g # 0, the above results covariant differentiation hold equally well. However, it is advisable in such a case to use the term covariant differentiation based upon gi; - Thus we should say that the results of the first part of this section involve differentiation based upon the metrie tensor as; of spact Since we shall have occasion later to use covariant differentiation not based upon the metric tensor of space, it is understood in what fol- lows that we are dealing with properties of covariant differentiation based upon a general tensor gi; . From the form of equations (22.11) it follows that the covariant derivative of the sum, or difference, of two tensors of the same order and type is equal to the sum, or difference, of the covariant derivatives 22] COVARIANT DIFFERENTIATION aa) of the two tensors. We consider next the covariant derivative of the product of two tensors, and in particular the following: em = ee) ~ e (om dh} + 604 A }) +b; (¢ M {ie} +e “{, My = Em + bist ans which is the same as the rule for ordinary differentiation of a product. Since a tensor formed by multiplication and contraction is a sum of products, we have in particular (22.17) (dip )am =O Dijam TH Biz ms ‘The results for these particular cases illustrate the following general theorem: (22.16) [22.2] Covariant differentiation of the sum, difference, outer and inner product of tensors obeys the same rules as ordinary differentiation. If we differentiate covariantly the tensor d.,; defined by (22.6), we have a (a _ fry ‘adr 1 -(™- “Ny 1) ( w(t) (2- ni) art Vat jks a (2) - a fh\ afi) ans ft ax* \aw) ~ ax \aif ~ axe \ikf ~ ax! \5k, (ots) eee lad) Since a (ar) _ a (ars (22.18) age () =a ( we have in consequence of (20.13) Maik = (22.19) Deie - Nesey = MR je + In like manner for a tensor b;; we have (22.20) benat — Dean = dajR ar + bok er, and in general fgg (22.21) Bey ceratt — Prysconnate = Lo Pry racadeas ete Rett 2 TRANSFORMATION OF COORDINATES [Cu I Thus far we have been dealing with covariant tensors. We shall now find the corresponding results for contravariant and mixed tensors. Instead of proceeding directly to do so we make use of the fact that g'.; = 0 by an argument similar to that which led to Theorem [22.1]. ‘Thus, considermg \‘ as the contravariant components of the co- variant vector \;, we have Mie = (G'Ad, 1 = GM) = Ait From this result, (22.19), and (20.15) we have Min — Mas = 9" Arie — Aged) = 9g "Rize = GgN”Rmiie » From equations (ii) of §20 Ex. 7 and the second set of (20.15), we have 9 Reaie = —O" Rinse = —Rimie Consequently (22.22) Mie — Mag = NR nie Similarly it ean be shown that BM ie — OM ay = OR ne — BR ase From results of this type and (22.21) we have the following general formula: wp WE Ue aa bey cite (22.23) 0 it follows from an equation of the form (14.12) that g’ > 0. If the quantities gi; > 0 are not equal to 6;; , it follows from the theory of algebraic quadratic forms that there exist linear transformations* with constant coefficients by means of which the quadratic form is trans- formed into (14.2). Henee we have [23.3] A symmetric covariant tensor gi; for which the determinant is positive is the metric lensor of euclidean 3-space, if and only if the Riemann tensor Sormed with respect to gi; is a zero tensor. Although this result has been derived for the case when n = 3, there is nothing in the proof which limits the result to this value for n, that is, it applies to any case for n > 1. ‘Although euclidean space of any dimensions n is characterized by the property that there exist coordinate systems for which the Christoffel symbols are equal to zero, we shall show that for any space there exist coordinate systems for which at a given point the Christoffel symbols are all equal to zero. 7 fact, if the space is referred to a general co- ordinate system x" and Va jee denote the values of the Christoffel symbols for this coordinate system at the point 2}, and coordinates 2” are de- fined by * Seo Bocher, 1907, 1, Chapter i0. §23] PARTIAL DIF ERENTIAL EQUATIONS 121 then the point z’ = 0 is the point 2{, and at this point oe ant { i\ Bah, ne Yap ope ax’? Ox! PA}o Substituting these values in (20.11), we find that {rn} = 0 at the point, as was to be proved. Hence we have (23.4] Even when the Riemann tensor based upon a tensor gi; is nol a zero tensor, a coordinate system can be found such thal al a given point all of the Christoffel symbols in this coordinate system are equal to zero. In this coordinate system at the origin, that is, 2” = 0, the covariant derivative of a tensor is the ordinary derivative. In particular, we have bho ar" a ye 1 none = oh = aalaga” \pr ax" 32" \paf? as follows from (20.13). From the: equations we have Re jae +R ansg FR par = Oe The loft-hand members of these equations are components of a tensor. Since they are equal to zero at the point i one coordinate system, they are equal to zero at this point in any coordinate system. Moreover, this result can be obtained at each point of the space, and consequently we have [23.5] For any space and in any coordinate system (23.18) Bim + Resim + Rimes = 0. Since the covariant derivatives of g" are equal to zero, we have Be sam = (Rain) m =O Ra im « From this result and’ (23.18) it follows, since the determinant of ¢ is not equal to zero, that (on changing indices) (23.19) Rijtim + Rijime + Reiman = 0. Equations (23.18) and (23.19) are called the Bianchi identities. They hold in any space of any dimensionality and in any coordinate system.* * See 1002, 1, p. 351. 122 SFORMATION OF COORDINATES (Cx, If EXERCISES 1, Find the mixed system of differential equations of the type (23.1) for which az! + bat, v= aziz, w= Bat) + ale)? is the general solution, a and b being arbitrary constants; what is the system Eq in this case? 2. When the funetions yf in (23.1) are linear and homogeneous in the 6’, the equations of the sets £1, --- , Bw have this property; and in order that the equations (23.1) have a solution, it is necessary and sufficient that there exist ive integer N (S$ m) such that the rank of the matrix of the sets of equa~ , Ex ism — g (q 2 1), and that this be the rank also of the equa- tions of the sets Bi, «++, Ewa. 3. Consider the Riemann tensor Ras based upon a tensor gi, where i, j = 1, «++, n; because of equations (i) of §20, Ex. 8 there are nz = n(n — 1)/2 ways in which the first pair of indices are like the second pair, and na(na — 1)/2 ways in whieh the first aud second pairs are different, and consequently there arc na(na + 1)/2 distinct components as regards equations (i); however, there are n(n — 1)(n — 2)(n — 3)/4t equations of the form (ii) of §20, Fx. 8; hence there are n(n? — 1)/12 distinet components of the tensor. 4. For w = 3 there are 6 distinel components of the tensor Ryju, and there are 6 cquations Rik = g Rese 5 show that the solutions for sx of these equations are given by R Riga = gale + ge Ra — gaRa — gaRa + 2 (ging: ~ 995K); where 5. Sinee Ria, one has from (23.18) on contracting for i and 1 Rim — Ryne + Rimes = OF if this equation is multiplied by gi* and summed with respect to j and &, one obtains 1 aR Bes 9 Oem CHAPTER III Intrinsic Geometry of a Surface 24, LINEAR ELEMENT OF A SURFACE. FIRST FUNDAMEN’ 1 QUADRATIC FORM OF A SURFACE. VECTORS IN A SURFACE Consider upon a surface with the equations (24.1) a= fiw, v) @ = 1, 2,3), a being cartesian coordinates in the space in which the surface is tm- bedded, a curve defined by expressing w! and wu! as functions of a param- eter f, thus (24.2) ut = o%(t) (a = 1, 2).* Looked upon as a curve in space, its element of length ds is given by (see §2). (24.3) ds* = dx de. From (24.1) we have 6 of ee (24.4) de’ a oF au’, where as follows from (24.2) a _ dg (24.5) du = at, Substituting from (24.4) in (24.3) we have (24.6) ds? = gag du® du’, where the quantities gap are defined by ac’ ant 24, (aa) “= aut auP* * In general in what follows Latin indices take the values 1, 2, 3 and Greek indices the values 1, 2. 123 124 INTRINSIC GEOMETRY OF A SURFACE [Cu, IE The expression for ds given by equation (24.6) is called the linear element of the surface in terms of coordinates u' and wu, The right-hand member of (24.6) is called the first fundamental quadratic form of the surface. Although the linear element was derived in seeking the element of length of a curve on the surface looked upon as a curve in the euclidean 3-space in which the surface is imbedded, the significance of the linear element of the surface is that onec it has been obtained for a surface the are of any curve (24.2) is given by Lap ‘ _ de dat =f a ee In particular, for the coordinate curves u’ = const. and u' = const. the respective elements of length are given by (24.8) dst = gu(du'’, ds} = gaa(du')’. Consequently gs: > 0 and go2 > 0 unless the respective coordinate curves are minimal curves (see §2). Hence, if we are dealing with real surfaces and real coordinates, we have that gu and ga: are positive functions, ‘A change of coordinates to a new set wu", uv is defined by means of equations (24.9) = vu", u”) (a = 1,2), rovided the jacobian a » is not identically zero. From (24.9) p au wr) we have (24.10) dul = ae du (ay = 1,2). When these expressions are substituted in (24.6) we obtain (24.11) ds = g'y3 du" du®, where «aye (24.12) Os = gop 28, 24 “Fault aul® From (24.7) it follows that ges = gsa; that: is, the g’s are symmetric in the indices. Since equations (24.12) are of the form (18.1) we say that gos are the components of, or that gas is, the covariant metric tensor of the surface; it is symmetric and of the second order, §24] FIRST F NDAMENTAL QUADRATIC FORM 125 If we denote by g the determinant of the g’s, that is (24.13) 9 = | Geel, we have from (24.7), analogously to (5.1), (24.14) 9 = Idea | ~ (APP + (APY + (AY, where A‘ are defined by (10.3). On the understanding that we are dealing only with real quantities it follows from (24.14) that g is never a negative quantity. It is equal to zero only in case (24.15) A™ = A" = A™ = 0. These equations cannot be identities in the w’s, as follows from theorem (10.1). However, equations (24.15) may be satisfied by certain values of w and v’. The points having such values for coordinates are called singular points, as remarked in §11; they may be isolated points, or constitute one or more singular curves on the surface. Only ordinary points, that is, non-singular points, on a surface are considered in this book, unless the contrary is stated. Since g ~ 0 functions g® are uniquely defined by (24.16) am = 855 where (24.17) 8% = 1 or 0 according as a = y ora # ¥. In fact, equations (24.16) are equivalent to 24.18) aM, gtagi=—%8, gta, (418s) PHB, Pad 7 og ‘As in §14 it can be shown that it follows from (24.12) that ag _ re OU" au (24,19) gf =O” oe au? where g’™ bear to gis the relation (24.18). Thus g® is a contravariant tensor; it is called the contravariant metric tensor of the surface and is of the second order. We observe that equations (24.12) and (24.19) are of the same form as (14.10) and (14.16) respectively, which refer to the fundamental amadratic form of euclidean 3-space in two general systems of coordi- nates 2‘ and 2%, 126 INTRINSIC GEOMETRY OF A SURFACE [Cn, III As stated in §10, the use of two coordinates uw" in the definition of a surface, as in equations (24.1), was introduced by Gauss. He used p and q as coordinates and wrote (24.6) in the form ds’ = Edp’ + 2F dpdq + Gag’. This notation, at times with u and v used in place of p and gq, has been followed gencrally, but in what follows we use (24.6), since it enables one to write equations in simpler form with the use of the summation convention, From (24.4) it follows that at each point of the surface differentials du’ and du’ determine in the enveloping space a direction tangential to the surface (see §11) for which dx‘ are direction numbers. Accordingly we say that at each point of a surface differentials du* determine avector in the surface. Moreover, since equations (24.10) for a trans- formation of coordinates in the surface are of the form (15.10), we say that du® are the contravariant components of the vector at each point. One would expect from geometric considerations that for a curved sur- face vectors with the same components du” at different points would have different directions as viewed from the enveloping space. This expectation is verified analytically by (24.4), since ordinarily the quan- tities 2°, are functions of the ws, and consequently vary from point to point in the surface. Consider two sets of functions A°(u', uw’) and A’/*(w, w”) in two co- ordinate systems u” and u’ in the surface related as follows eye ou" (24.20) aa which are equivalent to au’ 24.21 We MS (24.21) Td in view of the identities au" dul? _ ae ou we oe (24.22) pa or = oy, oe aa On these being analogues of (13.13) and (13.14). Since (24.20) and (24.21) are of the form (15.10) and (15.8) respectively, we say that \* and X'* are the components in their respective coordinate systems of a contravariant vector in the surface, there being one such vector at each point of the surface. §24] VECTORS IN A SURFACE 127 Tf we define quantities &* by : «ae (24.23) a it follows from (24.20) that , sy Oa (24.24) Be sties 2 ang 2 recta . Fi ‘The quantities 2 and 2%; are direction numbers in the enveloping space ue aul of the tangents to the coordinate curves u” = const. and u' = const. respectively in the surface, that is, the w'-coordinate curves and the u'-coordinate curves. Since & are linear homogeneous combinations of these direction numbers, they are direction numbers of a line tangent to the surface, as follows from the results of §11, and consequently are the components in the enveloping space of the contravariant vector whose components in the surface are \%. Two sets of functions da(u’, uw’) and AL(u", uw follows which are related as “ , , (24.25) de =) ee Meee 2 (cither set of equations being equivalent to the other in consequence of (24.22)) are the components in their respective courdinate systems of a covariant vector in the surface (see §17). From the results of §19, which apply to any number of coordinates, it follows that if d* are components of a contravariant, vector in the surface, then (24.26) Re = Gash” are the components of a covariant vector, since gag is a covariant tensor. Also if \. are the components of a covariant vector, then (24.27) M = gy are the components of a contrayariant, or, since g® js a contra- variant tensor. As in $17, we say that two sets of functions \* and\. of u* related as in (24.26) and (24.27) are the contravariant and covariant components respectively of the same vector in the surface. The geomet- tic significance of these components is shown in §25. From (24.23), (24.7), and (24.27) we have (24.28) Do g's! = NN gap = 97, g" AsGag = 97, Me = OT, Nee < 128 INTRINSIC GEOMETRY OF A SURFACE [Cu, III The first member of these equations is the square of the length of the vector as viewed from the enveloping space (sce §16 following (16.1)). From §19 it follows that gash’ and g*"AgA; are scalars. Hence we have [24.1] For a vector whose contravariant and covariant components are A" and Aq respectively each of the scalars gagh"X’ and garg is equal to the square of the length of the vector. As a corollary we have [24.2] A necessary and sufficient condition that \" and dq are the contra- variant and covariant components of a unit vector in the surface is that (24.29) ges = 1, gM rady = 1. E: RCISES 1, For a sphere with equations of the form in §10, Ex. 1 the coefficients gas of the linear element are given by gi = 4%, gi = 0, gaz = at sin wt, 2. When equations of the tan S(ud), where ut is the are, are w: gent surface of a eurve with equations zt = ten in the form (sec (8.5)) the curves ut = const. are the orthogonal trajectories of the generators, and in consequence of (2.8), (2.9), and (4.3) one has gu = ut — ul), g2=0, gaat 3. For any surface of revolution, when the equations are given in the form in §10, Ex. 2, gizl+e% gu=9, gu =u, where the prime indieates the derivative: 4, Theequations js uw wm ulcos ut, at = usin ut, 2 = a cosht™ s are equations of the surface of revolution of the eatenary 2t = a eosh = about the z-axis (see §10, Ex. 2); this surface is called the calenwid. In this case . ala =~ s0 thal for the coordinates w't, ut where ut = V/uiT = a =a gil giz 0, ger at + (wy? Does this change of coordinates change the coordinate curves? §25) ANGLE OF TWO INTERSECTING CURVES 129 5. For a right conoid with equations of the form in §10, Ex. 4, gi=1, g2=0, gaa u te” 6. Ona surface with the equatio: rsuleosu, ra utsinw, 2 = (ut) + aw, where ais a constant, the curves ut = const, are circular helices (see §3, Ex. 1), and gialte", gasae, ga ul + at; such a surface is called a kelicoid; all the eurves u? = const. are congruent plane alled the helicoidat parameter. 7. When in Ex. 6 o(ut) = 0, the helicoid is a right-conoid as follows from §10, Ex. 4, and onal oe 0, oe oF wy this surface for any non-zero value of a is ealled a skew helicoid. due 8. Along any curve on a surface the quantities 7 are the contravariant com- ponents of the unit tangent vector. 25. ANGLE OF TWO INTERSECTING CURVES IN A SURFACE. ELEMENT OF AREA We have seen in §24 that values of du* determine at each point of a surface a vector tangent to the surface whose components dz‘ in the enveloping space are given by (24.4). From the standpoint of the en- veloping space the direction cosines « of this tangent vector are ae (soe §3). In consequence of (24.4) and (24.6) we have an! aya dx! out 25.1 au —_ C5) ON Sgandhit dt Since dz are direction numbers of such a tangent to the surface, ‘ and the quantities 2%, are functions of the w's, it follows from (244) that the direction of a vector of components du* depends not only upon these components but also upon the point of tangency. However, values of the differentials du* determine a direction at each point of the surface, For a curve on the surface defined by equations of the form (24.2) we can find from (25.1) the expressions for «' in terms of ¢, giving the direction cosines in the enveloping space of the tangent at any point of 130 INTRINSIC GEOMETRY OF A SURFACE [Cu. HL the curve. Equations (24.5) give values in the surface of the compo- nents du* of this vector. Suppose now that at a point on a surface we have two veetors in the surface of components d,u* and dix“, and we denote by aj and os the direction cosines of the tangent vectors so determined in the en- veloping space. An angle @ between these tangents is given by cos @ = Dalal = - which because of (24.7) becomes (25.2) cos @ = — pa fu, V Gan du" dy wi) (gya dau daw) In terms of ai and a2 we have sin? @ — cos? @ = ‘ tt Lalai Laial! 1 oath let ef? det atl ai aif jal aif lat al] =Taaal T haa atl That atl at ai! jab at! © ab at From (25.1) we have act diu® az’ dyu"} jaz aa4| |diu' di’ jaf af] fou die out de | jan dull [ihe ha | 3 = =| i"} . fab ai! |aa! dau" da! daut| | aa! aa!) dau! da du* das Out des}, |du? due|!chs chs | From this result and (24.14) we have Jaa’ du ‘Gs Gs} sinkoe gh idw dul) (dee “des | * See remark following (6.1); also C. G., p. 80. §25) ANGLE OF TWO INTERSECTING CURVES 131 Thus far @ is one of two angles whose sum is 360°, and the formula (25.2) does not distinguish between these two angles, ‘This ambiguity is removed, if we take a de = da V Gea din? dw) (Gy daw daut) ‘The significance of this choice is shown later. From (25.2), and also from (25.3), it follows that the angle between tangents of components du“ and du“ depends not only upon these components but also upon the point of tangency unless gas are con~ stants. We show later that gas are constants only in case of developable surfaces, and then only for particular sets of coordinates. Tf two curves defined by equations u* = or(h), — u" = @F (4) have a@ point in common, for values of 4 and f& at the point we can obtain from (25.2) and (25.3) the angle @ between the tangents to the curves at the point of intersection, that is, the angle of the two inter- secting curves. In order to find at a point the angle w of the coordinate curves u? = const. and u! = const., for the directions in which w! and u’ respectively are increasing, we take (25.3) sino = Vg (5.4) dw >0, dv®=0, dat de® > 0, and obtain from (25.2) and (25.3) (25.5) coso= 9 sing = fo Vouge Vonge Thus w is the angle formed by the vectors at the point in the directions in which u' and w° respectively are increasing and such that 0 0, du’ = 0 must be rotated to be brought into coincidence with the tangent vector to the curve in the direction for which du? > 0. This is the sense in which the first of these vectors is rotated through the angle » to be brought into coincidence with the vector du’ = 0, du® > 0, and may be called the positive sense of rotation. We shall show that the angle 9 through which the vector diu® must be rotated in the positive sense to be brought into coincidence with the vector dyu* is given by (25.3). Wesay that this is the angle which the vector dyu® makes with the vector diu*. In fact, if 6,, and 62, denote the angles made at a point by the vectors dyu* and d,u“ with the vector du’ > 0, du’ = 0, it follows from (25.6) that VG det ggdiu’ = dy gy dew In VA Gag dru" dy v8) (Gye dgu% dou’) - The right-hand member of this equation reduces to the right-hand member of (25.3). If now 01, < 04, 8 = 0, — O15, and if, > 2, 6 = 2r — 6, + 62, ; in either case the left-hand member of the above equation is sin @ For a transformation of coordinates (24.9) we have from (24.12) that at, w) aqui, uw) }? where g’ and g are the determinants of g’g and gag respectively. Also for this transformation of coordinates equation (25.3) becomes sin (62 — Or (25.8) g = ol {dia Idvu” du” -\ ds las “de sin 0 = { , mo Bayt dau® las as §25] ANGLE OF TWO VECTORS 133 the plus or minus sign to be used according as the jacobian is posi or negative. If we denote by w’ the angle between the coordinate curves u'* = const., and if @ denotes this angle in the wsystem, we have from (25.5) sin = + sino’, This means that positive sense with respect. to the u-system and u’- system is the same or opposite according as the jacobian is positive or negative. Hence we say that a transformation of coordinates is positive or negative according as the jacobian is positive or negative. However, if a transformation is negative, and if in the equations of the transforma- tion one replaces u”' or uw by —u” or ~u, the sign of the jacobian is changed, and the new transformation is positive. This change does not change the coordinate curves u’* = const., but merely the positive sense on one family of coordinate curves. Hence there is no loss of generality, if we understand in what follows that a transformation is positive unless there is a statement to the contrary. Consider now two vectors of contravariant components Aj, and dj). If at a point on the surface we take differentials diu* and dyu" given by dy =, dau = 4, we have from (25.2) and (25.3) cos 8 = - V (Gasd5, Ni (gas 3 4) (25.9) wi ee sind = Vg — MUM NM V Goa dS, MGA) ® being the angle whieh the vector Aj) makes with the vector A¥ + We have, on changing dummy indices, Goad iNe) = $009 ing” Deis = GF Ariyhaiada = J MrpaAs 5 and MIM — MAL = g!rtag” dale — grin"? raIa = Og — GY)Oudae — Mieden) From these results, (25.9), (24.18) and theorem [24.1] we have 9? Mtadaia VENER AusAgia — AnaAgn Yi VG Natadste) GG daly dala) cos 6 = (25.10) sin @ = 134 INTRINSIC GEOMETRY OF A SURFACE cer meets From the first of equations (25.9) and (25.10) we have [25.2] A necessary and sufficient condition that al a point the vectors df; and XS, be perpendicular is GaphTiN) = AvgNS, = MiAata = 9 Aadays Ania and Aria being the covariant componente of the respective vectors. When, in particular, we denote by 6: and 62 the angles which the vector A“ makes with the vectors \’, 0 and 0, d’ respectively, that is, the angles which the vector A“ makes with tangents to the curves u’-const, and u = const, respectively, we have from (25.9) Ges aa Aan, 08 Oy = Vangasd“X* Von2gusX° re (25.11) cos 0) = If then in the tangent; plane at a point P, we lay off the vectors \*; 4,0; 0, \%, and note that the lengths of the last two vectors are W/gy and +/gw X* respectively, we have from (25.11) that the projection of these lengths upon the line of the vector A* are respectively GaN oad’ Vash Vass ‘The sum of these two vectors is ~/g.gdX4, that is, the length of the yeetor \*. Again from (25.11) we have that the projections of the vector A“ upon the tangents to the curves «? = const. and u’ = const, at P are respectively gaa \* goad" Von Von Noting that the numerators of these expressions are the covariant components of the vector A", we have (see theorems [16.6] and [17.6}) [25.3] At any point P of a surface the vector \* is the diagonal from P of the parallelogram whose sides are segments of the tangents at P to the curves u’ = const. and wv = const. of the respective lengths ~/gu Mand ~/gn 0} and the projections of the vector upon these tangents are the covariant com- ponents of the vector divided by ~/gu and ~/ gz respectively. We define two sets of numbers és and e® as follows: (25.12) en =en =e? =e" =U, ew §25] CONTRAVARIANT AND COVARIANT COMPONENTS 135 For a transformation of coordinates in the surface we have au" au _ au" aut © Fal Gu ~ © Ful Gul ay OH Ou aul u) aut aul jul au" ~ a(uityw’) aul jul” If the transformation is pos (25.13) _vi=vad atu’, w’) aus w)? ve, we have consequently aut aut en VI = 8 VI Si ait In like manner it can be shown that Hence we have [25.4] The quantities (25.14) fas = Cas Gy are the components under a positive transformation of a skew-symmetric covariant and contravariant tenaor of the second order respectively. From (25.14) and (25.12) we have (25.15) egy = eg = BF. Also the second of equations (25.9) may be written a (25.16) sing= = ei V aad) aM) Hence we have [25.5] A necessary and sufficient condition that a unit vector Mi, makes a right angle with a unit vector Xf, is that (25.17) cap fidhy = 1. 136 INTRINSIC GEOMETRY OF A SURFACE [Cu, III Consider a unit vector \* and the vector whose covariant components Ha are given by (25.18) Ha = ean. Since u.\" = 0 it follows from theorem [25.2] that u_ is perpendicular to the veetor A%. Its contravariant components are given by wo = gus = ged” = GN — GX) Vg = "9" — G0") Va rs = ors. From (25.18), (25.19), and (25.15) we have (25.19) Hite = Xe = 1, that is, 2 is a unit vector, and (25.20) ead" = cag" y = Me = 1. Hence by theorem [25.5] we have [25.6] The quantities uq defined by (25.18) are the covariant components of the unit vector which makes a right angle with the unit vector \*. Consider next upon a surface the curvilinear quadrilateral formed by the coordinate lines, the vertices being the four points P(u', wu’), Qe + dw, wv), Ra’, w + dv’), S@ + du’, w + du’). To within terms of higher order the lengths of the sides PQ and PR are ~/g,: du! and 1/gm du’, and the sides opposite them have these respective lengths to within terms of higher order. Since the distances of Q and # from the tangent plane to the surface at P are of the second order in com- parison with PQ and PR (sce §11, Ex. 9), it follows that when Q, 2, and § are projected orthogonally upon the tangent plane to within terms of higher order these projections are the vertices of a parallelogram of sides gu du’ and ~/gu du’. The area of this parallelogram is equal to sino Von dul Von dul = v/g du du’, the latter expression following from (25.5). Accordingly we have that the element of area de of a surface is given by (25.21) do = Vg du' du’, §25] ELEMENT OF AREA 137 by means of which may be found by integration the area of any portion of a surface by evaluating the definite integral f f Vg du! du! for appropriate limits. EXERCISES 1. Find the angle of the coordinate curves of a helicoid defined in §24, Ex. 6. 2. When the coordinate curves on a surface are the curves in which the surface is intersected by the planes z = const. and 2? = const., the parametrie equations of the surface are ger, Pea, = set 2), the ear clement is dst = (1 + pi) dat + 2pyprdetda? + (1 + p3ddz®, af where pa = gga’ and the angle w of the coordinate curves is given by 3. When the coordinates of the enveloping space are general coordinates 2, equations (24.7) are replaced by oxi axi 08 = TF a 5,8? where a,, is the covariant metrie tensor of space. 4. When the coordinates of the enveloping space are general coordinates 24, the form of equations (24.23) is not changed. 5, Find the element of area of a sphere of radius a as defined in parametric form in §10, Ex, 1 (see §24, Ex. 1), and usc the result to find the area of the sphere. 6. If 1p and G2, denote the augles at a point on a surface which the vectors du and dau* make with the vector du! > 0, du! = O at the point €08 (09 — Big) = C08 9 SiN (O2y — iq) = Bin 8 where cos @ and sin @ are given by (25.2) and (25.3). In particular, 1 du* 08 (w — 66) = =F gaa, sin (w 8) = V Von os where w and 0p are given by (25.5) and (25.6). 7. From (25.15) it follows that 28, ab 138 INTRINSIC GEOMETRY OF A SURFACE [Cu, II 8. From (25.12) and (25.14) it follows that 1 @ 0 gary 958 = Gers» cap g*’ g = oo Gi) 8 Gay 958 = eb» cap 27 = € (iii) 9 eyaesp = Orb dager (iv) eT 4g = €g5g% 26. FAMILIES OF CURVES IN A SURFACE. PRINCIPAL DIRECTIONS Upon a surface defined in terms of coordinates u* an equation (26.1) ful, @) - 0 is an equation of a curve, as remarked in §10. If this curve is defined also by equations u* = 9), in terms of a parameter /, and these expressions are substituted in (26.1) the resulting equation being an identity in ¢ does not vary with ¢ and consequently the derivative with respect to ¢ is equal to zero; that is, af de aus dt From this equation and (24.5) we have that differentials du" determining the tangent at any point of the curve satisfy the equation 8 aye = (26.2) aut du = 0. Thus one of the differentials may be chosen arbitrarily (+ 0), and the other is given by (26.2) when the values of w! and w for the point are substituted in 2, aut Consider an equation (26.3) fu, wv) = 6, where ¢ is an arbitrary constant. For each value of-¢ equation (26.3) is an equation in the coordinates w" of a curve in a surface defined by equations of the form (24.1). ‘Through cach point of the surface for which f is single-valued there passes 2 curve of the family of curves (26.3). In fact, when the curvilinear coordinates u', u* of a point are §26] FAMILIES OF CURVES IN A SURFACE 139 substituted in (26.3), a value of ¢ is uniquely determined, and evidently the equation (26.3) in which c is given this value is an equation of a curve through the point. An equation (26.4) fw’, ’) = is an equation of tho same family as (26.3), if fr is a function of f, that is, i(w, w) = F(f(u', w)). For, in this case a locus of points for which fis a constant is also a locus for which fi is a constant. Conversely, if every curve of the family (26.4) is a curve of the family (26.3), then fi must be a function of f. For, as follows from (26.2) the differentials du determining the tangent to the curve of each of the families (26.3) and (26.4) are given by Sf aye = A yt a ed 0, I If then each curve of one family is to coincide with a curve of the other family, we must have from which it follows that f and f; are functions of each other. Consider next the differential equation (26.5) M.du" = 0, where M; and Mz are functions of the ws. In accordance with the theory of such equations there cxists a function ((u', x), called an integrating factor,* such that a function f is defined by (26.6) ie m. (a= 1,2), and thus (Me du® = df. Consequently fw, w) = 6 where ¢ is an arbitrary constant, is aun inte- gral of (26.5). If there is another integral of (26.5), say fi(u', w) = «1, corresponding to another integrating factor t,, it follows from (26.6) * Kine, 1927, 1, p. 292. 140 INTRINSIC GEOMETRY OF A SURFACE (Cu, IIL and similar equations in fi and 4; that the jacobian of f and fi with re- spect to u' and u’ is equal to zero, and consequently fi is a function of and in view of the results of the preceding paragraph, ) is an equation of one, and only one, family of curves on the surface. We say that this family of curves is the family of integral curves of the equation (26.5). We seok now the differential equation whose intogral curves aro the orthogonal trajectories of the integral curves of equation (26.5). If we substitute in equations (26.5) Me, dé = —pM, this equation is satisfied whatever be p. If then we substitute these expressions in place of dyu' and dyu* in the equation (26.7) Gos dou" du? = 0, which from (25.2) is the condition that at a point of the surface the vectors du" and du* are perpendicular, we obtain on discarding the factor p (M2 ~ gepMi) du? Hence we have [26.1] The orthogonal trajectories of the integral curves of an equation M,du" = 0 are the integral curves of the equation (26.8) (gn Mz — gM) du! + (92M: — gxMi) du’ = 0. As a corollary we have (setting M; = 0 and M, = 0 successively) [26.2] The orthogonal trajectories of the coordinate curves w> = const. are the integral curves of the equation (26.9) giz du! + gaa du® = 0; and of the curves u® = const. they are the integral curves of (26.10) gu du + gn du? = 0. If f(u’, w) is any function, for a change of coordinates in the surface we have af _ af’ au” aus du’ aus’ where f’ is the transform of f (see §15). From (24.25) it follows that f. are covariant components of a vector, the gradient of f(§17)- out §26} FAMILIES OF CURVES LY A SURFACE 141 From this result and (26.6) one has that the quantities M, in equation (26.5) are covariant components of a vector. Since its contravariant components are gM. , we have gas9°"M, du’ = 53M, du’ = My dt = 0. Hence from (25.2) we have that M. in equation (26.5) are the covariant components of the field of normals to the intogral curves of the equa- tions (26.5), these normals being tangent to the surface at the cor- responding points of the integral curves. Consider next a differential equation of the first order and second degree of the form (26.11) dag du* du? = an(du')’ + 2ay du‘ du? + ap(du’) = 0, where audz2 — aj2 #0. This equation is equivalent to the two equations (a2 + Vaiz — aun) du! + and 0, (26.12) ee (az — Vale — auan) du’ + andu? = 0. We scek the condition that the integral curves of these equations, and consequently the two families of integral curves of equation (26.11), are orthogonal to one another, that is, form an orthogonal net (see §25). These integral curves are real or conjugate imaginary according as Qa — ayy is a negative or positive quantity. If we denote the dif- forentials in the two equations (26.12) by du and diu* respectively, solve these equations for du” and dyu’, and substitute in (26.7), we have [26.3] The two families of integral curves of an equation dag du du’ = 0 form an orthogonal net, if and only if (26.13) Gude — 2Wgud2 + gudn = 0. Under a change of coordinates the equation (26.11) becomes (26.14) ai; du’ du® = 0, where au® au" > ow aul (26.15) Oya = Cos 5 gyn) — M8 = O98 Ba Bab * Thus dag are the components of a covariant tensor of the second order, and there is no loss in generality in assuming that it is a symmetric tensor (see §19, Ex. 2). In consequence of (24.18) the condition (26.13) may be written (26.16) 9 deg = 0. 142 INTRINSIC GEOMETRY OF A SURFACE [Cu. Ht ‘The left-hand member is a scalar quantity, whose vanishing is the condi- tion that the integral curves of equation (26.11) form an orthogonal net. We consider now the equations (26.17) (aug ~ 7 gaa) = 0, from which it follows that quantities ’ are determined to within a common factor for cach root 7 of the determinant. equation (26.18) {dp — 7 Gap | = 0. It is evident, that equations (26.17) do not admit a common solution other than \” = 0 unless r is a root of equation (26.18). Since dag and ges are covariant tensors, it follows that in any other coordinate system the left-hand member of equation (26,18) is equal to , vy (acu, ur)? wat gag| (ee) {ace ~ gee ( aw, 1B) Since the jacobian is different from zero, we have in the coordinates w'* an equation of the form (26.18), and since r is unaltered by a change of coordinates, it is a scalar. Moreover, the equations (26.17) trans- form into , ry yg au’ aut (ar = row) Noe Gut which are equivalent to (a4; — rgys)\" = 0, where that is, each set. of quantities \” satisfying (26.17) for suitable values of v are the contravariant components of a vector. When the determinant (26.18) is expanded, one obtains the quadratic equation (26.19) (guge — gis)r* — (anges + aeegn — 2aiegre)r + (Gudee — ais) = 0. The discriminant of this equation, namely (anger + angi — 2arzqra)* — 4(guige2 — gis)(Qnder — ain), reduces to (26.20) (auger — aaagss)” + Aang — ngrz)(arege2 — anagr). Since the roots of equation (26.18) are scalars, if they are real and distinct or equal in one coordinate system, the same is true in any co- 426] PRINCIPAL DIRECTIONS 143 ordinate system. Suppose then that the coordinate curves form a real orthogonal net. For such a system the expression (26.20) reduces to (axugan — e0g)° + Aafeguges , which is non-negative since gi aud gos are positive (§24). Hence the roots of (26.18) are real. They are distinct, unless the above expres is equal to zero, In this ease that is, a4g = dag, in which case r = 4, and equations (26.17) are identities, and thus do not determine quantities °. We consider now the case when dag are not the same multiple of gas, that is, the tensor aug is not proportional to the tensor g.,.._We denote by r: and 7 the roots of equation (26.18), and by Mj, and X42) the corre- sponding vectors; thus we have (26.21) (dag — rigag)Nii = 0, (dag — rxgag)X3i = 0. If we multiply these respective equations by A¥ and Af; , sum with re- spect to a in cach case and subtract the resulting equations, on changing indices suitably we get (72 — 11) gashtiN4i Since r: # 7 we have the first. of the equations (26.22) gasdiiNt1 = 0, GaghiiNEr = 0, the second being a consequence of the first and either of (26.21). From the first of (26.22) it follows from theorem [25.2] that the two vectors Ai, and Ag, are orthogonal, that is, the two vectors at each point are perpendicular. ‘These two vectors at each point are the tangents to the curves of an orthogonal net on the surface. In order to obtain the differential equation of which these curves are the integral curves, we replace \? by du® in equations (26.17) and eliminate r from the two equations, as @ takes the values 1, 2. ‘The result is aig du? ony du? (26.23) " . | agdu® gey du? which upon expansion is (@ugiz — augu) du + (auger — angus) du’ du? (20,24 28) + (auager — daxgis) du 144 INTRINSIC G fOMETRY OF A SURFACE (Cu, TU Since these integral curves form an orthogonal net, when they are the coordinate curves we have in this coordinate system the first of the equations (26.25) g2=0, a2 =0, and the second follows from the fact that in this coordinate system the coefficients of du" and du™ in (26.24) are equal to zero, since (26.24) must be satisfied separately by du! = 0 and du = 0, and gu and gx: are different from zero. In this coordinate system the directions of the integral curves of equation (26.11) are given by ad tn au = * Vv a’ and consequently these directions are equally inclined to the curves u? = const. as follows from (25.7). As a result of the foregoing discus- sion we have [26.4] If aus is a tensor not proportional to the fundamental tensor gap , the integral curves of the equation (26.23) form a real orthogonal net, and bisect the angles of the integral curves of the equation dae du® dé = 0. We call the directions determined by (26.17), that is, the directions of the integral curves of equation (26.23), the principal directions for the tensor ang. Another property of these directions and the significance of the roots of equation (26.18) follow from the consideration of the expression = ap"? (26.26) 1 Ata point the value of r depends not only upon the point, but also upon the direction \* at the point, ‘The maximum and minimum values of rata point are given by the directions for which x = 0, that is, Grad"N'dash? — ay:N™N'Gagd? = 0. By means of (26.26) this is expressible in the form (26.17), and we have {26.5} Ala point of a surface the maximum and minimum values of the expression (daph*N*)/(gasd"X*) are given by the principal directions for the tensor dag , these directions being perpendicular. §26] PRINCIPAL DIRECTIONS 145 EXERCISES upon a sphere the two families of curves which bisect the angles be- tween the meridians and parallels, and then find the linear element of the sphere when these are the coordinate curves (sec §10, Ex, 1 and §24, x. 1) 2. By definition a curve upon a surface of revolution which meets the meridians under constant angle is a loodromic curve; the equation of loxodromic curves is (sce $10, Ex. 2 and §24, Ex. 3) is 1 fe Vit ot du + bu tos where b, ¢ are constants, ~b being the cotangent of the constant angle. 3. On the skew helicoid (see §24, Ex. 7) sleuteosu at = utsinw, 2 = aut the integral curves of the equation (dust = (ae? (duty? = form an orthogonal net. 4. An angle 0 botweon the integral curves of the equations Mig du* = 0, Mig duX = 0 is given by 0 MaMa V (9? Mig Mg) (9? M, 5. Ry means of Q418) and (25,14) equati 03 0 Ms) QRS) may be writ capg?™My dub = 0, 6. Derive the result expressed by equation 6.16) by showing that this scalar is equal to zero when the integral ettrves of (26.11) form an orthogonal net and are the coordinate curves. 7. For the paraboloid aut cos ut, at = butsin wt, a = dul® (au costut +b sintut), where a and b are constants, find an equation of a curve on the surface such that the tangent planes along the curve make a constant angle with the 2tz*-plane; find the edge of regression of the developable surface which is the envelope of the tangent planes to the surface along one of these curves 8. For a surface S$ with equations of the form a= ef) cos(ut + ut, eMo(ut), ut) sin(ut + uw), where h is a constant, the curves ut st, lie on the quadrie cones 146 INTRINSIC GEOMETRY OF A SURFACE [Cx. UL and cut the generators of the cone under constant angle, that is, they are conical helices (sce §3, Ex. 3). Such curves are also ealled spirals, and the surface S is called a spiral surface, 9. The first fundamental form of the surface $ of Ex. 8 is ds? = ot [(72 + fe + gt) dul” + QA! + f+ hyp") du dt + (ME + ft + htyt) dut*l, where a prime indicates a derivative with respect to u'; the orthogonal traje tories uw = const. of the curves ui = const, can be found by quadratures and the linear element is reducible to the form dst =o’ %du® | g(w't) du’), w't being a suitable function of w. 27. THE INTRINSIC GEOMETRY OF A SURFACE. ISOMETRIC SURFACES Tn §§24~-26 it has been shown that the metric properties of a surface, that is, lengths of curves, angles between intersecting curves, and areas, are expressible completely by means of the first fundamental form of the surface. It is true that these results have been obtained by the consideration of these quantities from the standpoint of the enveloping euclidean space. In this sense the metric properties of the surface are induced by the euclidean metric of the enveloping space. However, once the formulas for the measurement of length, angle, and area have been found in terms of the first fundamental form of the surface, there- after these metric formulas may be used without considering the surface as imbedded in space. For example, since the first fundamental form of a sphere of radius @ is (sce §24, Ex. 1) (27.1) ds’ = a((du'Y + sin? w(u’)’), the coordinate curves being meridians and parallels of latitude, all the motrie properties of the sphere are obtainable by the use of the formulas of §§24-26 applied to the form (27.1). In the next chapter we consider geometric properties of a surface in- volving its shape as viewed from the enveloping space and we find that they are expressible in terms of the first fundamental form and another quadratic differential form, called the second fundamental form of the surface. In order to distinguish between the properties expressible entirely in terms of the first form from those expressible only in terms of the two forms, we say that the intrinsic geometry of a surface consists of the properties expressible in terms of the first form alone. When two surfaces are such that there exists a coordinate system on 427] ISOMETRIC SURFACES 147 each in terms of which the first fundamental forms of the two surfaces are identical, the intrinsic geometry of the two surfaces is the same. This means that so far as measurement. on the two surfaces is concerned there is no difference in the two surfaces, no matter how different the surfaces may appear to be as viewed from the enveloping space. Two such surfaces are said to be isometric. From (24.12) it follows that, if the fundamental forms of two surfaces are identical for one coordinate system on each, they are identical when any and the same transforma- tion of coordinates is applied to the two surfaces. An example of isometric surfaces is afforded by the eatenoid and the skew helicoid (see $24, Exs. 4, 7). From geometrical considerations it follows that a cylinder and a cone are isometric with the plane, since either may be rolled out upon a plane and thus brought into coincidence with a portion of the plane. In the case of a cylinder this is shown analytically in Ex. 1, the coordinate curves u’* = const. on the surface corresponding to a rectangular cartesian system in the plane. In §12 it was noted that in general the tangent planes to a surface constitute a two parameter family, but that thereis a group of surfaces ” whose tangent planes constitute a one parameter family. Such a sur- face is called a developable surface, since it can be rolled out upon a plane, just as a cylinder or cone can be. In view of this property it is evident that a developable surface is isometrie with a plane. It was shown also that with the exception of cones and cylinders a developable surface is the tangent surface of a curve, the tangent planes Lo the surface being the osculating planes of the curve. In §24, Ex. 2 it was stated that the first fundamental form of the tangent surface of a eurve is — wie du? + du®, wu being the are of the curve. Since this expression does not involve the torsion r of the curve, it follows that the tangent surfaces of the curves which have the same first intrinsic equation « = fi(u') but different second intrinsie equations 7 = fa(u’) (see (7.1)) are isometric. In particular, when + = 0 we have the tangent surface of the given twisted curve isometric with the plane, the correspondence being be- tween points of the surface, and points of the tangents to a plane curve with the same intrinsic equation « = fi(u’) as the given twisted curve. Since the tangent plancs to the surface are the osculating planes of the curve of which the surface is the tangent surface, when the developable surface is rolled out upon the plane, the tangents of the curve become the tangents to the plane curve of the same curvature « and the principal we 148 INTRINSIC GEOMETRY OF A SURFACE [Cu. Ill normals of the former become the normals to the plane curve in its plane. With respect to a rectangular coordinate system in the plane the direction cosines of the tangent a* and of the normal §* are such that = a" des* a dp" (27.2) wa, Soap, = as follows by processes for the plane which gave the Frenet formulas for a twisted curve; they follow also from (6.1) for r = 0. In §12 we showed that any developable surface other than a cylinder or a cone is the tangent surface of some curve C, it heing the envelope of the osculating planes of C. For a point x‘ on C the coordinates of any point in the corresponding osculating plane are given by (see §6) (27.3) Xi = a! + ua’ + 28%, for suitable values of u and ». When the developable surface is rolled out upon a plane, the curve C becomes a plane curve P. The tangent and principal normal to @ at a point go into the tangent and normal to T at the corresponding point, and consequently the point X* given by (27.3) goes into the point of coordinates (27.4) X* = 8 + wa® + 06°, where X* and #* are cartesian coordinates and where #°, «’, and 8“ are appropriate functions of s, which is the same for both curves, and u and y are the same numbers as in (27.3). Differentiating equations (27.4) with respect to s, and making use of (27.2), we have (07.8) wa (ie +au wet +(% +e) Tf u and v are such that (27.6) # —w+1=0, # + xu then X* are constants, that is, such values of u and v determine in each osculating plane of the curve a point such that all these points go into the same point in the plane when the surface is developed upon the plane. Equations (27.6) being the same as (6,10), whose solutions u and » determine an orthogonal trajectory of the osculating planes, we have [27.1] When a developable surface, other than a cone or cylinder, is de- veloped upon a plane, all the points of an orthogonal trajectory of the tangent planes to the surface go into one and the same point of the plane. §28] CHRISTOFFEL SYMBOLS FOR A SURFACE 149 EXERCISES 1. When equations of any cylinder are given in the form Bef), 2 fu), osu, the first fundamental form is (f+ FP) (dud)? + (du. In terms of parameters u’! and u’? defined by aie f Viet iia, wie wt the form is (du’!)* + (du’?)*, What are the coordinate curves in this cas 2, Ona cylinder with equations as in Ex. | the helices, thatis, the curves w meet the generators under constant angle, are defined by a f Vie + iB dul + but +e = 0, where a, b, ¢ are constants, 3. Given a surface of revolution with equations as in §10, Ex. 2 and a right conoid with the equations (see §10, Fx. 4) wisinw?, 2 = y(u); zis ufleosu', at = in order that a surface of revolution and a right conoid be isometric with the meridians of the former corresponding to rulings on tho latter, it is necessary and sufficient that w’? = f(u?), and Vitetdut = dit, wt pyr = sow; show that it follows that the surface of revolution is a catenoid (eee §24, and the right conoid is a skew helicoid (see §24, Ex. 7) 4. When the polar developable of a curve (seo §12, Ex, 1) is developed upon a plane, the points of the curve go into one, and the same, point of the plane 28. THE CHRISTOFFEL SYMBOLS FOR A SURFACE. THE RIEMANNIAN CURVATURE TENSOR. THE GAUSSIAN CURVATURE OF A SURFACE. From the definition (20.2) of Christoffel symbols of the second kind, and (24.18) it follows that the Christoffel symbols formed with respect to the first fundamental form of a surface are (no indices being summed) fe\_1 aa lea _ ages {a} ~ 2g (ow gue t9°8 ays — 208 set), B\_ 1 (4 Oe _ 9, Mon as (28.1) {e}= ( $09 Fa $00 Gis Tt Bow ) 2g aus Ou’ 1 Ja {ohm b (on tte ~ 5) (a, B = 1, 25.0 = 8) 150 INTRINSIC GEOMETRY OF A SURFACE [Cu, HI From (20.5), (20.6) and (20.7) we have (28.2) in all of which the summation convention appli From the expression in §20, Ex. 7 for the covariant components of the Riemann tensor it follows that the tensor is skew-symmetric in the first two indices and also in the last two indices. Consequently for a surface (28.3) Roose = Rasy = Hence every non-zero component is equal to Rye, or to its negative. From the form (ii) of Ex. 7 in §20, we have LO Foe _ Fon _ Fon Ran = 3\ aioe bu? — dul + ool }0)- HE} When a surface is isometric with the plane, there necessarily exist. upon it orthogonal nets with respect to which as coordinate curves , Riaz = Ran = — Rone = — Rin. (28.4) gu = . ge = 0, the net in the plane consisting of lines parallel to the coordinate axes. Hence Riz = 0 in this coordinate system and consequently in every’ coordinate system by theorem [18.1]. Conversely. if the Riemann tensor for a surface is a zero tensor, it follows from theorem [23.3], which applies to any n-space for n > 1, that there exist coordinate systems on the surface with respect to which the above equations hold. Hence we have (28.11 A surface is isometric with the plane, if and only if the Riemann tensor is a zero tensor. ‘The quantity K defined by Riaz (28.5) K=*m (28.5) 9 $28) GAUSSIAN CURVATURE OF A SURFACE 151 is called the Gaussian curvature of a surface, and also the tolal curvature of a surface. As thus defined it is an intrinsic property of the surface ag was shown by Gauss." The geometric significance of the Gaussian curvature of a surface as viewed from the enveloping space is treated in §46. From (28.5), (28.3) and (25.14) we have (28.6) Resa = Keases and consequently (see §25, Ex. 7) (28.7) K = bRasee". Hence from theorems [25.4] and [19.2] it follows that (see Ex. 8) [28.2] The Gaussian curvature of a surface is a scalar. From (28.2) and analogously to theorem [22.1] we have [28.3] The covariant derivatives of gan and of g°° are equal to zero. We leave it as an exercise to show that in consequence of the third of equations (28.2) [28.4] The covariant derivatives of és and of €* are equal to zero. Although we have defined K to be the Gaussian curvature of a surface with the first fundamental form gg du* du’, it is advisable also to speak of K as the curvature of the quadratic form gaz du du’. In §26 we considered the integral cinves of the equation dasdu® du? = 0 and pointed out that these curves are real only in case the determinant @ = aude — az is non-positive, When this determinant is negative, which is the condition that there be two distinct families of real curves, we say that the form dag du* du’ is indefinite just as we say that the first fundamental form of a surface is definite since g > 0. When ais nega~ tive, it is possible to find real coordinates a* such that eg du dub = t da! di’, where ¢ is some factor. In this case @* = constant are the integral curves. We wish to consider particularly the case when { = 1. In this case the curvature of the right-hand member is zero, and consequently this case can arise only when the curvature of the left-hand number is zero since curvature is a scalar. * 1827, 1, p. 236, 152 INTRINSIC GEOMETRY OF A SURFACE [Cn IIL When = 1, we factor the left-hand member of the above equation and write the equation as follows ae\( vi a + moves ca) an (via au + mt vna an We replace this equation by the two # (vam at + BEM aH ae) = aa, a (van dy + i) = dit, Van where ¢” and ¢™ are to be determined. Evidently they are integrating factors of their respective equations. Consequently y is such that heva-~2(¢ wives an (28.8) 4 Va) — 2 ( aut » . + Oy a hat, eae From these two equations one can obtain 5% and 5%. ‘Their condition of integrability is necessarily satisfied by the fact that the eurvature of the above form is zero, Hence u ean be found by a quadrature and then a and aby further quadratures from (28.8). Weaccordingly have [28.5] Whon a quadratic form day du du? is indefinite and ite curvature ts zero, real coordinates wi’ and i’ can be found by quadratures in lerms of which the form is equal to dit’ da. When the form is definite, a and @ as derived by the above processes are conjugate imaginary. {f then a’ and a are replaced by a + it and a — ia? respectively we have the following theorem: [28.6] When a quadratic form des du* du® is definite and its curvatisre is zero, real coordinates i’ and w can be found by quadratures in terms of which the form is (da')* + (aa). As a corollary of this theorem we have [28.7] Upon a developable surface cartesian coordinates can be found by quadratures, that is, the coordinates are cartesian when the surface is developed upon a plane. §28) GAUSSTAN CURVATURE OF A SURFACE 153 EXERCISES 1. When the coordinate curves on a surface form an orthogonal net, the cor responding Christoffel symbols are (see §20, Ex, 2) Ge egy 20) ou ac) ~~ Bg5y dub — oe GA), a) _ 2108 VG00 rT from which and (28.4) it follows that rg, gn) | tT (agu\* , Ogu Aga) | 1 [ (Age)? | dre Age aaa aa) tiga (ae + ou au + an But tie out =} 2 (1 ton) 2 (Lao © BNE) BWANA gigs OH)” WV gran OF 2. The components of the Ricci tensor (see §20, Ex. 10) for a surface are given by 7 Ryags =~" Rav, g 185 o and the scalar curvature K of the surface by 1 Rasy =} sat gy = Ee 5a Rag = = 3, For a sphere of radius a with the equations of the form in §10, lows from Ex. 1 and §24, Ex. 1 that fuga a 4, For a surface of revolution with the equations a= ulcosu, at=usinu, 2 = elu), it follows from Ex, I and §24, Ex. 3 that Run _ __ ¢'e" Re on WU ot 5. For a surface with the first fundamental form a{cos? o(dut)? + sin? o(du?)4), where w is a function of ut and u? ae de = atsinweoso( 2% 8), Ring = atsin w eos w( 3 — he 154 INTRINSIC GEOMETRY OF A SURFAC (Cx, TIT 6. The integral curves of the equation gag du du? = 0 are conjugate imaginary curves of length zero, that is, minimal curves (see §3). 7. ‘The equations to wt iw — w) wut 1 geo a ear l-we lfua! where ais a constant, i= +/—1, and u! and u? are conjugate imaginary, are para- metric equations of a sphere of radius a; the coordinate curves are minimal lines. 8. In consequence of the remark following (28.3) au’ au aw aw’? du du Ow ov Joo © | ate, 8) consequently K is a scalar. 9, From equations analogous to (20.15) and from (28.3) and (28.5) it follows that Repys Rg = K Fags ~ 8957). 10. From §20, Ex. 7 and (28.5) it follows that ke} Ca aq 1 ”) 2VG LIM on V9 OM v7 g Out wo Lagu ga vate avai) and in consequence of (25. IV on _ OVD 606.6 ou out Von tino aan _ AV 02 oop 2 aw aut 1 | de ~ ql auow Kk a _ + oi (ii) 11, When the equations of a surface are of the form in §25, Ea. 2 the Christoffel symbols have the values { sh = Gielen ars where g= 14/3} +J2, and Sov = Sa55,- §29] DIFFERENTIAL PARAMETERS 155 12, From the definition (§22) of the covariant derivative based upon gap of the contravariant components \* of a vector, namely @ and from (28.2) one obtains “a a a) ae (VOX) = V0 The scalar \*,, is called the divergence of the vector X* (see §22, Hix. 2). 12, For the covariant derivative Xq,s of the covariant. components Xa of avector one has that 1 (am _ ar Prag 2 2 (OM _& Yo rate ms) is a scalar, called the curl of the vector Xg ; when the curl is equal to zero, the veetor is a gradient (see §17). 14, If Mand Xe are the contravariant and covariant components of @ unit vector, one has A*ahq = Aa6\* = 0, from whieh it follows that Ata = wp, das = Mave» where n* is the vector perpendicular to the given veetor, and yp is some vector. 15. When the coordinates are chosen so that an indefinite quadratic form reduces to 2axs du! du! the curvature of the form is given by 1 a Ro ay iain low az. 29. DIFFERENTIAL PARAMETERS If ¢ is any function of wi and u, EA are the covariant components of a vector, the gradient of g, as observed in §26. Consequently the quantity Aw defined by ct 8¢ oe (29.1) Au per is a sealar (see theorem [19.2]). It is the square of the length of the gradient # by theorem [24.1]. Also 28 OP: depo (29.2) 8 (29.2) Arter, 2) = 9 ai? where gi and ge are any functions of u' and 1’, is a scalar. 156 INTRINSIC GEOMETRY OF A SURFACE (Ca, IIT Consider also the quantity 1 A(gr, ee) _ 08 Ben Bye 29 x aa ee he (293) Orler, 2) V9 ala, wt) © Bue bub Tn consequence of theorem [25.4] this quantity is a scalar for a positive transformation, The sealars Avy, Ar(y: , 2); Oiler , v2) are called differential parameters of the first order, involving as they do derivatives of the first order. Tn order to give a geometric interpretation to these differential parameters, we note that for any function ¢ of u’ and u’ one has oe (29.4 “ du® = 0, (29.4) ae where du* are the contravariant, components of the tangent vector at a point of cach curve of the family y(u', u*) = const. From (29.4) and theorem [25.2] it follows that this tangent vector and the vector x at each point are perpendicular. For two families of curves gi(u', u*) = a and yaw’, u?) = ¢ the angle @ which the normal vector See at a point makes with the normal vector 24 at the point, and consequently an angle between the curves at the point, is given by (29.5) cosp = CL) ging = On oe)_ Agr digs Var digs as follows from (25.10), (29.2), aud (29.8) (see Ex. 5). From (29.5) we have [29.1] Two families of curves o(u', w) = const. and gr(u', u2) = const. form an orthogonal net, if and only if As(¢1 , g2) = 0. From (29.5) it follows that the following identity holds between the differential parameters of the first order: (29.6) (Anon , ¢2))* + rer, @2))* = Ares Awe. When, in particular, we take g: = w', gz = u’, we have from (29.1), (29.2), and (24.18) in the coordinate system u* (29.7) $29] DIFFERE NTIAL PARAMETERS 157 Since differential parameters are scalars, when Ayu’, Aw’, Ai(u’, u*) are written out in another coordinate system u’*, we have pe Out and 29.8) f<-gt (29.8) T= 9 a ay? that is, equations (24.19). Since gu, gs» and g are positive quantities when uv! and u? are real coordinates (sce §24), it follows from (29.7) that Ay is a positive scalar when ¢ is a real function. Let g be a solution of the differential equation (29.9) Ay = 1, and ya solution of the equation (29.10) Ane, ») = 0, in which ¢ is the given solution of (29.9). By theorem (29.1] the curves y = const. and ¥ = const. form an orthogonal net. If they are taken as the parametric curves w' = const. and w? = const. respectively, it follows from (29.7) that giz = 0, g’ = g'ug’», and #8 = Au! = Ae guigaz and consequently the linear element is (29.11) ast = (dul? + gis(dul)? From this result it follows that along a curve ¥ = w? = const., ds = du”, and thus the distance along this curve from a curve g = ¢ toa curve ¢ = cris equal toe: — 4. Since this distance is the same along all curves ¥ = const., we say that the curves ¢ = const. are parallel curves. Since Ar has been shown to be positive for a real function g, if Aw is a positive function of g, say (29.12) Aw = Fy), a real function f(g) is defined by di to = | Tae From (29.1) one has for any function f(y) ave) = (y dre. 158 INTRINSIC GEOMETRY OF A SURFACE [Cu, III Hence for the above function f(g) one has Ayf(g) = 1. Accordingly ‘S(e) = const, are parallel curves, and therefore ¢ = const. are. Hence we have 29.2] If o is any real function of real coordinates u* and Aw = Fe), the curves ¢ = const. are parallel curves. For example, when the linear clement is ds = gi(w')(du'y + gn(av’y’, 1 Ane = Fla * follows from (29.7) and consequently the curves u! = const. are parallel. The quantities gas, defined by (29.18) = oe _ defy . Poe = Suzaub uv \aBf? being the second covariant derivatives of g, are the components of a symmetric covariant tensor of the second order. Consequently the quantity Asp defined by (29.14) dw = 9M ea0 is a scalar, It is called the fundamental differential parameter of ¢ of the second order; it involves derivatives of the second and first orders. The differential parameters Ap and Asp were introduced by Lamé* for space referred to general coordinates in his study of physical prob- lems. They were introduced in the study of the geometry of a surface by Beltrami.t In order to write Asp in another form, we note from the definition (22.4) of the covariant derivative of a contravariant vector that the covariant derivative of the vector g*” £2 9g.) is given by a du a per a {7 ap 9 e0) + 9) onl ih By theorem [28.3] the left-hand member of the above equation is equal to gees. From this result and (29.15) it follows that (29.14) can be written (29.15) (9 ea), dag = 2, 0% 0) +9 e0 {8 } + 1859, 1, pp. 5, 17. 1 1864, 1, pp. 359, 365. §29) DIF SRENTIAL PARAMETERS: 159 In consequence of this result and (28.2) we have (29.16) de = Wie 2 (Vase), which because of (24.18) may be given the form = gn 2? be _ 4, 98 (29.17) dep = F di (maim au ae) a C ane ut q vo aut vo EXERCISES 1, Show that a Far Me = 20% ea%.6r- 2. Show that one —on {a} enue 3. The following are differential parameters of the second order: Ardiy; Ailes diy), Orly) Aiy)y Aidile, W), di(dre, 49); Or(Are, A1¥). 4. Iffand g ure any functions of u! and ut, FV wpe FF a asa (5h) a F2 5 pauls + (RY ave, 85 89 4 Af 80) 4 oye yy 4 2 20 aaw td 54) ae, we) + oy 2h 28 aut ML = Fa 5a aw +(j bere ay ao. a, a, f= Loe +r haw +250 aula ) ean ae 5. Ata point of a curve o(u!, u*) = 0 the unit normal vector 5, //aiemakes 5 due a right-angle with the unit tangent vector“ when the latter is chosen in direc de tion s0 that by theorem [25.6] ag aut aye V8 0 a If for two intersecting eurves o(u!, ut) = 0 and y(ul, uw) = 0 one has respectively ae aut ay daub 28 wn Ai tom fae says aur VEO gt aye VNC gs the expression (29.5) for sin @ is equal to (25.8) (see (25.15). 160 INTRINSIC GEOMETRY OF A SURFACE [Cu. HI 6. Ona surface of revolution (sce §10, Ex. 2) the curves ut = const., that is, the orthogonal trajectories of the meridian curves, are parallel curves (sec §24, Fx. 3), 7. Fora helicoid, as defined in §24, Ex. 6, the helices are parallel curves, and their orthogonal trajectories are the integral curves of the equation a aut + aw = arp Mt tae = 0; find the linear element of the surface when the helices and their orthogonal trajectories are the coordinate curves, and show that any given helicoid is iso- mnetric with some surface of revolution. 8, When tl ‘sts upon a surface an orthogonal coordinate net such that the quantities gu: and gze are functions of ut alone or u® alone, the surface is isometric with a surface of revolution. 9. Ifgisasolution of thc equation Ay = 0, the eurvesy = const. are imaginary, as follows from (29.7). 10. Of the differential parametors Aw, 4i(e, ¥), Ave, Orly, ¥) the last is the only one which changes sign (but not magnitude) when a transformation is negative. 11. When the coordinate net is orthogonal, Lie Vain V Gan from this result and (29.7) it follows that if u* are solutions of the equation Agu ( # a); @ Aue, Aw) = 2da(Are — 1) other than those for which Aig = 1, then (6 4a). In this case the fundamental form may be writte i) ds? = cos? 8 du!* + sin? 6 du? ; if ¢ is any solution of (i) such that Ay # 4, the function ¥ of the orthogonal trajectories y = const. of the eurves » = const. can be chosen so that the funda- mental form dst = cost 6 det + sin? 6 dy? 12. When the linear element is the form (ii) of Fx. 11 and one effects the trans. formation of coordinates wewl pu, ute uu, the linear element becomes @ dst = (du')? + 2 cos. dul! du’? + (du'), where «(= 26) is the angle of the new coordinate curves. A net with respect to which as coordinate the linear element is of the form (i) is called a Tekebychef §30] ISOMETRIC ORTHOGONAL NETS 161 net; a necessary and sufficient condition that the coordinate curves form such a net is { ay =0 (a #8). When the linear element is gag du* du and one effects the transformation wie | Vow, [ Vane, the new coordinate eurves form a Tehebychef net; the determination of all such nets is equivalent to the solution of equation (i) of Ex. 11. 30. ISOMETRIC ORTHOGONAL NETS. ISOMETRIC COORDINATES When the coordinate curves on a surface form an orthogonal net such that gs: = gee = ¢, in which case the linear element of the surface is (30.1) ds = E(du™ + du), the elements of length of the u'- and u*-coordinate curves are t du’ and du’ respectively. Hence the coordinate curves divide the surface into small squares to a first approximation. Such a net is called an isometric orthogonal net and u* isometric coordinates. From (29.7) it follows that a necessary and sufficient condition that two families of curves ¢(u’, u*) = const. and ¥(u', u”) = const. form an isomctric nct and that g and ¥ be isomctric coordinates is that (30.2) Aw = Av, Ailey, ¥) = 0. In this case the linear clement of the surface is (30.3) ds = (dg + ay’), where f is the reciprocal of the common value of the (wo members of the first of equations (30.2). When the linear element is in the form (30.3) it follows from (29.17) that Ag = 0 (« = 1,2). It will now be shown conversely that cach real solution of the equation 4:6 = 0 determines an isometric orthogonal net. If gis a sulution of the differential equation of the sevond urder 428 = 0, we have from (29.16) 0.4) A Vaicten) +2 Wise. 162 INTRINSIC GEOMETRY OF A SURFACE (Cu. IL From this equation it follows that a function y is defined by (30.5) Vo" = eee, Vartee= —M = tu, aut oa since (30.4) expresses the condition of integrability of equations (0.5). Tf these equations are multiplied by g,: and g,2 respectively and the resulting equations are added, we have 909" O.0 = a Gna — gaa) The left-hand member of this equation reduces toy. For the values = landy = 2 these equations are reducible, in consequence of (24.18), to (30.6) oa=VG0vs, 92 = —Vo eva. Expressing the condition of integrability of these equations, we have 2, (Vag"va) = 0, ut which by (29.16) is equivalent to the condition day = 0. Tf equations (30.5) are multiplied by ¢, and ¢.2, and the resulting equations added, we have V0 0% 0.00.8 = eda os, which in consequence of (29.1) and (25.14) may be written (30.7) Aw = “onde. Likewise, if equations (30.6) are multiplied by ¥.2 and y.1 respectively and the resulting equations are subtracted, we have (30.8) oath = Aad. From these two equations we have the first of equations (30.2). Again, if equations (30.5) are multiplied by va and ¥2, and the resulting equations are added, we have the second of equations (30.2). Hence we have (30.1] Any real solution ¢ of the equation A,0 = 0 and the function ¥ obtained by quadrature from the corresponding equations (30.5) are is0- metric coordinates of an isometric orthogonal nel. If now we define coordinates i and a by (30.9) e=fhw), y= fiw), §30] ISOMETRIC COORDINATES 163 the linear element (30.3) becomes (80.10) ds = E(ffdul + fra”). ‘Thus the linear element is no longer of the form (30.1), but the coordi- nate curves are the same as before (see (10.14). For the form (30.10) Ut 30.11 a oo am Ui where U; and U: are functions of u' and w’ respectively. Conversely, if for an orthogonal net on a surface the quantities gi: and gzz are in the relation (30.11), the linear element is of the form. (80.12) ds = @(Uidu"® + Ufdu®). If then we define coordinates u’* by weft, ut= fv, the coordinate curves are the same as before, but the linear element is ds = @(du™ + du™). Hence we have [30.2] When for an orthogonal coordinate net, the condition (30.11) is satisfied, the net is isometric, and the isometric coordinates can be obtained by quadratures. If any funtion of g, say f(y) is a solution of the equation 420 = 0, in which case f(y) is one of a pair of isometric coordinates, we have from (29.16) (30.13) Aref'(v) + dwf"(y) = 0, where the primes indicate derivatives with respect to y. When this equation is written in the form Ay fe)’ we have that, if Ax/Aw is a function of y, say F(y), then the function S(%) obtained by two quadratures from (30.14) $e) — cFente fide! aivnee is such that f(y) is an isometric coordinate, Hence we have [80.3] A necessary and sufficient condition that a family of curves ¢ = const. und (heir orthogonal trajectories form an isometric net is that Meg = 0 or that the ratio of Awe and Ay be a function of y. 164 INTRINSIC GEOMETRY OF A SURFACE [Cu, III From me it follows that fn may be written in the form Pi WOVI I G0) + Hence a function y is defined by 30.15) SOVO Mea = 0a, MW)V90 ea = and y is such that the second of (30.2) is satisfied, that is, the curves ¥ = const, are the orthogonal trajectories of the curves ¢ = const. and form with the latter an isometric net. Equations (30.15) follow also from (30.5) when ¢ is replaced by f(). Consequently in place of (30.6) we have 018) Fler = Vota, fen = —Vod he From (30.15) and (30.16) we have similarly to (30.7) and (30.8) 2 eva9" ga) = 0. + ie Sadie = Heaba = y(cy Sets that is, in place of the first of (80.2) we have (30.17) Aw = f?()Aw. From this result, equations (29.7), and (30.14) we find that the linear element is (30.18) ds= as (ag? + Bf Aerlareras ayy, Tf a and 22 are isometric coordinates of an isometric orthogonal net, that is, if the linear element is of the form (30.1) and functions g and y are defined by (30.19) ¢ + ty = fu + iw), 9 — = flu F iv’), where f and fy are conjugate functions, we have dg + dv = f'f(du” + au’). Consequently the linear element is ds = ip (de? + dy’), and hence ¢ and y are isometric coordinates of 2 real isometric orthogonal net, different from the give en one since g and ¥ are not functions of a and u° respectively, or of u* and u' respectively. $30) ISOMETRIC COORDINATE N 165 When the linear clement is of the form (30.1), equations (30.5) become be _ aut ant? that is, the Cauchy-Riemann equations,* the integrals of which are given by (30.19). Hence we have [80.4] When one isometric orthogonal net is known for a surface, all other such neis may be obtained direcily by equations of the form (30.19). EXERCISES 1. The meridians and their orthogonal trajectories on a surface of revolution (see §24, Ex. 3) form an isometric net. 2. The rulings on a skew helicoid and th isometric orthogonal net (see §24, Ex. 3. When the coordinate curves of a surface form an isometric orthogonal net, the curves r orthogonal trajectories form an wi+ut= const, ut — ut = const. are the bisectors of the angles of the coordinate net, and form an isometric orthog- onal net. 4, When on a surface two families of curves » = const. and y = const. form an isometric orthogonal net sueh that aw = Aw = S(y) the surface is isometric with a aurface of revolution 5. For a plane referred to cartesian coordinates the equations wp iu define an isometric orthogonal net consisting of two families of circles. 6. For a central quadrie with parametric equations of §10, Ex. 3 ea = 0 (a, B= 1,230 8), henee the coordinate curves form an isometric orthogonal net, 7. The equations [ay — a tee {/ tue, a a “(Lb aru) + au) , se paw baw), ares * Fine, 1927, 1, pp. 408, 409. , 166 INTRINSIC GEOMETRY OF A SURFACE [Cu, III in which the a’s are constants are equations of a paraboloid. For the coordinate system ua 0 ton = as)u® — ua(l au) (ya — 4) gw 0 (= 1,3 Bs and consequently the coordinate curves form an isometric orthogonal net, 31. ISOMETRIC SURFACES In this section there is derived a necessary and sufficient condition that two surfaces S and S’ be isometric (see §27), their respective first fundamental forms being 1.1) Gestu'du’, — giasdu’*du®, The surfaces are isometri if there exist two independent equations B12) elu, w) = eu" w®), vu w) = veut, wl establishing a one-to-one correspondence between points of S and S’ such that by means of (31.2) either of the quadratic forms (31.1) is transformed into the other. Since differential parameters are scalars, it follows that a necessary condition that S and S" be isometric is that (B13) Aw= de’, AG, = AG), AW = AW, where the differential parameters on the left and right are formed with respect to the respective forms (31.1). Conversely, the conditions (31.3) are sufficient conditions that S and .S’ be isometric. In fact, if the curves ¢ = const., ¥ = const. are taken as coordinate on S and g! = cous, ¥ = const. on 8, in consequence of (29.7) the respective quadratic forms may be written in the form Ade — 2a(¢, pdgdy + Aedy Aid — Ate, ¥) 20i(y', v') de’ dy’ + Aig’ ‘ay? Aly’ Ay = ABW) . Hence when there exist two independent equations (31.2) such that (31.3) are satisfied, the surfaces S and S' are isometri Since by theorem [28.2] the Gaussian curvature is a sealar, a ncecssary condition that surfaces § and S’ with the fundamental forms (31.1) be isometric is that. (31.4) Kw, uw’) = Ku, uw), §31) ISOMETRIC SURFACES 167, where K and K’ are the Gaussian curvatures of S and S' respectively. We now seek a necessary and sufficient condition that S and 8’ be isometric. We consider first the case when and K’ are equal constants, that is, (31.5) Rim = ag, Rim = ag’. If a = 0, each surface is isometrie with the plane by theorem [28.1], and consequently with the other. In order to discuss the case a + 0 we remark that equations (31.5) are equivalent to (81.6) Ross = 4(arg38 — Jason), Rags = O(arg"s8 — Paid’ sy)» We apply to this case the discussion of equations (23.12) and (23.13) eet ara" of ions pt = 2 for the funetions u* of w", wand the functions p3 = 3". Because of (81.6) the corresponding equations (23.16) are satisfied in consequence of Gor = Ges EV that is, equations (24.12). ‘These equations, three in number, are the set. Ey, and all the scts Ey, --- are satisfied because of the set Eo. Since there are six functions u*, pj and three equations in the set Eo, we have in consequence of theorem [23.2] [31.1] Two surfaces of equal constant Gaussian curvature are isometric, the equations giving the one-to-one correspondence involving three arbitrary constants, In this connection it must be remarked that. such correspondence is limited to domains for which theorem [23.2] applies. When K is not a constant, we consider in addition to (31.4), the equation (31.7) AK = Aik’ We note that Aik ~ 0 for K real (sce §29). If then equations (31.4) and (31,7) are independent, these equations establish 2 correspondence between S and 8! which is isometric, if and only if (81.8) A\(K, 4K) = Ai(K', AK), = AK = AjAiK'. If equations (81.4) and (1.7) are not independent, and they are to be Satisfied ’we must have 1.9) aK = f(K), aK’ = f(K), 168 INTRINSIC GEOMETRY OF A SURFACE [Cu, III where f is some positive function of K and K’ respectively. In this case we may take for equations (31.2) equations (314) and (31.10) ALK’, unless GL) AK =fi(K), AiK’ = filK’). If equations of the form (31.11) do not exist, then equations (31.4), (31.9), (31.10) and (31.12) Ai(K, deK) = AL(K’, AK"), AiasK = ALAIK’ constitute a necessary and sufficient condition that $ and S’ be isometric. We consider finally the case when both (31.9) and (31.11) are satisfied, including the possibility /a(K) = fa(K’) = 0. If fi(K) = 0, the ratio of AoK and AK is a function of K, and by theorem [30.3] the curves K = const. and their orthogonal trajectories ¥ = const. form an isometric net, the function ¥ being obtained by a quadrature (see §30). Further- more, the respective quadratic forms are by (30.18) 1 lind 1 aiNdKe B11) 7 (aK? 4 cfntnes gy, 10 (aK? + efsinae’ ayy and when 4K = 0, A:K' = 0, by (30.3), they are i 1 S(K) F(R’ In either case it is seen from (31.13) and (31.14) that the equations (31.14) (dK? + dy’), (dK” + dy”). K=K, y= WV +4, where a is an arbitrary constant, define the isometric correspondence of the two surfaces, We have thus treated all possible cases and as the result have [31.2] Given two surfaces whose Gaussian curvatures are not constant; it can be determined directly, that is, without quadratures, whether the surfaces are isometric. When (31.9) and (31.11) are not both satisfied, the equations deter- mining the correspondence are given dircetly, whereas in the case when (31.9) and (31.11) are both satisfied, the determination involves # quadrature. Tn this case the correspondence can be effected in an infinity of ways. What this means geometrically follows from the observation that from the forms (31.13) and (31.14) it is seen that S and $31) ISOMETRIC SURFACES 169 S' are isometric with a surface of revolution (see §24, Fx. 3). When one considers a surface of revolution from the standpoint of the enveloping space, one sees that a surface of revolution is isometric with itself in an infinity of ways, each such correspondence being given by a suitable rotation of the surface about its axis. It should be remarked that for any two surfaces in isometric corre- spondence such correspondence is established only for the domains for which the equations given above, such as (31.4), (81.7), and (31.10), are independent in each of the cases considered. EXERCISES 1. When the linear clement of a surface is written in the form @ dst = dul + gar du", one has al Oe i) wat, aw oe Vi _£ from which it follows that AQut, daw!) = —K ~ (Ant)? Hence for any funetion ¢ the equation Arty, Asp) = —K = (Ase)* is an identity 2. If equations (31.4) and (31.9) are satisfied, for the functions « and o’ defined by _ aK . aK’ WM Jim the equation (31.10) reduces to Ag = Ayo’, and the equation Ai(e, Ay) = Ai(o’, Aye’) is a consequence of (31.10) and (31.4) (see Ex. 1). 3. ‘Phe equation (i) of Ex. 1 is the linear clement of a surface isometric with a surface of revolution if gas is x function of w alone (see §24, Ex. 3). In order that the surface have constant Gaussian curvature gs; must be a solution of the third of equations (ii) of Ex. 1, in which K is a constant. Tf K 0, there are the two cases . 1 = : @ K-1, Von = boos +esin™ @ a a “a 1 _ wl uw (ii) Ke- m2 = beosh » +e sinh ~ ge’ Vae= Deosh = +e sinh =, where U and ¢ are arbilrary constants. 170 INTRINSIC GEOMETRY OF A SURFACE [Cu. IT 4. For a surface with the lincar element du det Ty aah aye where a is a constant, the Gaussian curvature is equal to 4a. 32. GEODESICS ‘The Christoffel symbols { a and (ry in two coordinate systems u” \ and w’* respectively in a surface are in the relation out a\ au? au” _ J r\! aut 32. 1 i. ve a“ (2.1) aurou> + {ge aul aul? ul au’ ‘This result follows from (20.11), which result is general and applies to any transformation of coordinates in any number of variables, provided only that the determinant of the covariant tensor with respect to which the Christoffel symbols are formed, in the present case gag, is not equal to zero. For any curve on the surface defined by u” as functions of s, and for any transformation of coordinates, we have du™ du du® —du® _ au du (32.2) de ~ but ds’ ds ~ but de ‘Thus A* defined by «a (32.3) we are contravariant components of the tangent vector to the curve, and it isa unit vector since from (24.6) de dul? 99 ds ds Differentiating the second set of equations (32.2) with respect to s and making use of (32.1), we have Pull _ ut dul (Ry au" * a’ ow) dul dul” ‘ds Gu’# ds ww) Bu ~ \erf du au) “ds “ds” (32.4) from which it follows that aut {aha du" _ au" (82.5) ds? ds “ds DPV du da’ Byf ds “ds ~ bu wi) ae §32| GEODESICS 171 If we differentiate (32.4) with respect to s, we obtain uk du? gag du* du? du’ _ 2008 Ge de + Gur ds de ‘de From this result and (28.2) we have dut a\ dub du’\ du® _ (626) 908 ie F {s} & fe) : If then we put (changing dummy indices) du fa\didw _ (32.7) ie +{z) wa! it follows from (32.5) that «* ace the contravarianl components of a vector at each point of the curve, which may be a zero vector (that is, u*=0). From (32.6) it follows that if the vectors 2“ at points of a curve are not: zero vectors they are perpendicular to the tangents at the cor responding points; this case is considered in §34. We consider now the curves at each point of which the vector y* is a zero vector, that is, the curves for which w" as functions of s are solutions of the equations dub, fa\ db duit (32.8) a +{eh i de These curves are called geodesics, Before considering geodesies on a general surface, we observe that, if the surface is a plane and the coordinates are cartesian, equations (82.8) 28 reduce to ae = 0, the integral of which is uw = a" + 0%, when the a’s and b’s are constants. Hence the geodesies of a plane are straight lines, and conversely any straight line is a geodesic. ‘The reader should compare the results in this section concerning geodesics on any surface with the properties of straight lines in the plane, From the form of equation (32.8) it follows that isometric surfaces ($27) have the same equations of geodesics. In particular, the geodesies on a developable surface are such that they become straight lines when the surface is rolled out upon a plane. A characteristic property of geodes- ies on a surface as viewed from enveloping space is shown in §44, 172 INTRINSIC GEOMETRY OF A SURFACE = [Cn. TIE A solution of equations (32.8) is determined by initial values of u* and that is, the values when s = 0. For such values we obtain ds from (32.8) the corresponding initial values of , and from the result of differentiating (32.8) corresponding initial values of all higher deriva~ tives of u*. In aceordance with the theory of the existence of integrals of ordinary differential equations, tbe corresponding integral of equa- tions (82.8) is given by (32.9) weds (2 ns +3(4 ee +h (GE) et for values uf s for whieh (he series converge, the subscript 0 indicating initial values. If we differentiate the equation (82.10) gape au? we obtain (32.6), and consequently any integral of equations (32.8) satisfies (32.10). Hence, in order that ¢ in a solution (32.9) shall be the are, it is necessary and sufficient to choose the initial values (#4), so that (82.4) be satisfied. Since the values (2x), determine the direction of the geodesic at the initial point us , we have the following fundamental theorem: {82.1] Through each point in a surface and in any given direction there passes a unique geodesic. From (82.8) and (32.6) we have [82.2] The coordinate curves u" = const. for a = 1 or 2 are geodesics, if and only if (32.11) ‘eh 0 (a = Lor 2; 6 = 2or 1). *liwe put = pt, these equations and we {eh jeer = 0, which are equa- tions (82.8), are of the form discussed by Darbous (see §23). §32| GEODESICS 173 When the coordinate curves form an orthogonal net, equations (32.11) reduce to 99H = 0 (cee §28, Ex. 1). Hence we have [32.3] When the coordinate curves on a surface form an orthogonal net, a necessary and sufficient condition that the curves u* = const. be geodesics is that gs , where B # ee, be a function of u* alone. ‘Thus for example, the meridians on a surface of revolution are geodesics (see §24, Ex. 3), as are the rulings on a right conoid (see §24, Ex. 5). Returning to the consideration of the series (32.9), and the remarks preceding (32.9), we see that if we put du” 32.12 a* = s(“ (32.12) ’ “( 0? we have te gt tata © bala? + ott uw us = a + aa 4 owe + ct +, where a, b,c, --- , are functions of uj, and uj. ‘These series are con~ vergent for values of a! and a in absolute value less than some fixed quantity. Since the jacobian of u* with respect to w*(a, 8 =1, 2) for a* 0 is equal to +1, these series may he inverted giving i" as power series in u' — up and u’ — uj, which are convergent so long as u* — ug in absolute value are less than some fixed quantity.* For such values of u* — uf the values of a" are uniquely determined, and consequently there passes only one geodesic through the points ug and u". Morcover, from (32.4) and (32.12), it follows that the length of the are of the geo- desies between these points is given by F = (Gap). Hence we have [32.4] Through two sufficiently near points on a surface there passes one and only one geodesic.+ In order to find the rate of change with respect to s along a geodesic of the angle % which the geodesic makes with the curves u? = const., we differentiate with respect lo s equation (25.7) written in the form aa Viz dat" fre Fy @ = tan” * See Goursat, 1924, 1, vol. 1, p. 474. $ See Darbuux, 1889, 1, p. 408, 174 INTRINSIC GEOMETRY OF A SURFACE | [Cu. LL Making use of (28.2) and (24.6), the resulting equation is reducible to dh [a de _aidu ds ds ds — ds dst 1 f2 {2}) a due a (82.13) ta (5.42) - Sra {2\- 9S aa!) ds de ds =e du" Pe B dul du\ v4 2 \ dut as \ dst ds ds laf ds’ where ¢3 are defined in (25.14). This result holds for any curve on a surface. When the curve is a geodesic, the above equation reduces in consequence of equations (32.8) to de 5 Vo 2a 1 au es (32.14) ds t gu (la ds Consider now in a surface an orthogonal net of coordinate curves for which the curves u? = const. are geodesics, By theorem {32.3] the coordinate w' can be chosen so that the linear element is (32.15) ds = du” + gudr’*. From this result it follows that the length of the segment of a curve uf = const, between the curves u! = ¢ and 1’ = eis given by [atoms Since this length does not depend upon x’, the lengths of the segments of all the geodesics u* = const. between any two orthogonal trajectories are equal. In consequence of this result and theorem [32.1] we have the following theorem of Gauss:* [82.5] Given any curve C upon a surface and the geodesics orthogonal to C; when equal lengths are measured from C along these geodesics, the locusof their end. points is an orthogonal trajectory of the geodesics. The curves thus defined are called geodesic parallels to the curve C. From the discussion of equations (29.9) and (29.11) it follows that the curves there called parallel are geodesic parallels. From theorem (29.2), and the above discussion we have [32.6] A necessary and sufficient condition that a family of curves ¢ = const. be geodesic parallels is that Ae = F'(g), where P(g) is any pos * 1827, 1, p. 24. $82] GEODESICS 175 function of ¢; is the length of the geodesics measured from one of the curves @ = const, if and only if dip = 1. Suppose then that P; and P, are points of a surface through which there passes only one geodesic, and that we take this for the curve a? = 0 of a system of geodesic parallels 1° = const. and for u' = const. their orthogonal trajectories, so that we have the linear element (32.15). An equation of any other curve which passes through P: and P2 is of the form 1? = y(w'), and the length of the are P:P2 of this curve is given by s= {, Vit guy? de’, a where wi and 14 (> ui) are the values of w' at the points P; and Ps respectively. Since the are PiP; of the given geodesic is v3 — uw}, and za being positive the quantity 1 + guy” is greater than one, we have the following fundamental theorem: {32.7] If two points in a surface are such that only one geodesic passes through them, the length of the segment of the geodesic is the shortest distance in the surface between the two points. If one has a solution g(w’, 1’, a) of the equation (82.16) Aw = 1 such that ae involves the constant a, when the solution is substituted in (32.16) and the resulting identity is differentiated with respect to a, we have ag) _ (32.17) AL ( % a) =0. Consequently for each value of a the curves dp + (32, = (32.18) aa where b is a constant, are geodesics, being the orthogonal trajectories of the eurves y= const., which hy theorem [32.6] are geodesie parallels For a particular point ug the components du* of the tangent to a geodesic P through the point are given by acta du® = Ofor each value of a. Con- versely, if du“ are given, the corresponding value of a is determined by 176 INTRINSIC GEOMETRY OF A SURFACE [Cu. HIT this equation sinee a geodesic is uniquely determined by a point and a direction. ‘Then b is determined by the equation aglub, m8, a) _ aa . Hence all of the geodesics in a surface are given by (32.18), and we have [32.8] If y(u', x, a) isa solution of the equation Aye = 1 such that e involves the constant a, the equation fey da for all values of the constant b is the finite equation of the geodesics of the surface, and the arc of the geodesics is measured by ¢. ‘A surface possessing an orthogonal parametric net with respect to which the linear clement is (32.19) ds® = (Ai + As)(Bidu" + Bidu”), where A, and Bg are functions of u" alone is exlled a surface of Liowille. For such a surface equation (32.16) is reducible to 1 (aey 1 fae\? A=- 3, As. B 9) B (32) | We seek the solution of this equation such that each member of this equation is equal to a constant a, and find that ¢ is given by the two quadratures on f Vat ad + | Va oat. Hence by theorem [32.8] an equation of the geodesies is (32.20) du? = 2b, and we have [32.9] The geodesics on a surface of Liouville, when its linear element is given in the form (82.19), can be found by two quadratures. $32] GEODESICS 77 ‘lose this section with the derivation anew of the equations of s, using the property of geodesics in theorem [32.7]. To this end we consider a curve C with the equations (32.21) ut = sf", and two points P; and P. on the eurve with the respective values t and tz of the parameter f. Any curve in the neighborhood of C and passing through the points P1 and Ps is defined by (32.22) w= PQ + wd for a sufficiently small absolute value of the constant ¢, and for functions (0) such that (32.23) oh) = wh) = 0. The are s of C between P,P is given by = / : Vacakfi dt and the are s; of Ci by ' Lt AP) ae (32.24) (32.25) y= f oS beet fi beat f+ ek f! + ee”) at, where the primes indicate differentiation with respect to t. In order that the are s of C be the minimum of the ares of the curves through P: and P2 in the neighborhood of C, it is necessary that the derivative of 8, with respect to € be zero for e = 0, that is, "(00 a 1 Oe a = = dt = 0. [Geet =) Since it is understood that the derivatives involved are continuous in the interval 4, f, on integrating by parts the second term in the integrand, we obtain in consequence of (32.23) te afd d ve) at (6 0 fe Vd = 0. [ Naps ~ at age Since this equation must hold for arbitrary functions #* such that (32.23) is satisfied, we have the equations of Euler (32.26) ‘ ee = a 178 INTRINSIC GEOMETRY OF A SURFACK = [Cu If From the definition (82.24) of ¢ we have, noting that f*is u’, aut 1 agp, du? du? ap gush _ 9 att bg Daud a gasp ds? ft ds at a Substituting these expressions in (82.26), we obtain ds Bul 4 Aaa du’ du? _\ apr dul? du” _ | dub de 90 Ge aur dt a Bou a a a which is reducible by means of (28.2) to ds bl e de dl du de a(S + Nosh at di 7 ae an} Since g ¥ 0, it follows that as dul | [8 \duidu _ du? de (32.27) ae tf a di a as ai When the parameter ¢ is the arc these equations reduce to (32.8). EXERCISES 1, The great circles on a sphere are its geodesics, 2. A nccessary and sufficient condition that thero exist upon a surface a family of geodesics whose orthogonal trajectories also are goodesies is that the surface be isometrie with the plane. 3. ‘The geodesies on a cylinder are hieliees (seo §27, Ex. 2) 4, When the linear clement of a surface of revolution is written in the form (see §24, Ex. 3) @ ds = (1 +9") du* + wt? du, ad from the second of equations (32.8) one obtains by integration u!® a ¢, where cis an arbitrary constant. From this result and (i) it follows that ut, dut ds ds ds §32] GEODESICS 179 vite From this result one obtains that ut = -be f uVato dwt + d, where dis e a constant, is an equation of the geodesics, 5. When the linear element of a surface of revolution is written in the form dst = du? + y2(ut) du’*, aut = te +4, i Vere where c and d are constants, is an equation of the geodesics in the surface, and consequently in any surface isomotrie to it. 6. If a family of geodesies and their orthogonal trajectories on a surfaee form an isometric net, the surface is isometrie with a surface of revolution 7. A necessary and sufficient condition that 1? = g(ut) be an equation of a geodesic is that ong (Lh 2!) ye \ ? - {he 7 ({o} a{ oh de + (ea - {i} +{i} where the primes indicate differentiation with respect tou’. 8. A necossary and sulficient condition that » be a solution of the equation Aig = 1 is that ds! — de! be a perfect square. 9. Ife = a6 + 02, where @ and 02 are functions of ut and ut, are solutions of the equation Aig = I for all values of the constant a, the curves 1 = const. are minimal curves, and the ettrves @ = eonst. are geodesic parallels 10. When the linear element of a spiral surface is written in the form (sce §26, Ex. 9) ds? = {dul + Uru) duel, the equation a = Ladmits the solution e*y(u!), where y is any solution of the equation y/? +4 = 1; by the integration of this equation one obtains all the geodesics in the, wurfaee. U1. The orthogonal trajectorics of the curves 6(w', w*) = const. aro integral curves of the equation (sce theorem [26.1]) 9°90, du! — 90,» du® = 0. The integral e(u!, w*) = const, of this equation is given by te ft det Dat Fp 89 Bar = Ia au V9 ae Vg where ¢ is an integrating factor. If the eurves 0 = const. are geodesics, by a Suitable choice of ¢ the function ¢ is such that Aw = 1, Henee ¢ is given by B Lm a coy 0?" Oars 0, By means of §25, Ex. 8 the rightzhand member af this equation redneos to & Axe a j 180 INTRINSIC GEOMETRY OF A SURFACE [Cu. III and consequently is given by the quadrature vo a0 20 1 L on aw - 0% % a8) Ig 7 wA( aye ond Vai 12. If the integeal curves of an equation Ma du* = 0 are geodesies, their orthog- onal trajectories are given by the quadrature 678.9% 0a 7. f Vi GPA a du! — My du’) ey? Mg du? ° one nae) : g Vo MMs V 9% MoM dy 13, If 2 = ¥(ut, uf, a), where a is a constant, is a first integral of the equa- tion of Ex. 7, it follows from theorem [32.8] and Ex. 12 that the finite equations of the geodesics is 8 fg bony) du + (ou + omy) de oa Von + gab + gee 14. Surfaces isometric with surfaces of revolution and the quadrie surfaces (see §30, Exs. 6 and 7) are surfaces of Liouville. 15. A surface of constant Gaussitin curvature is a surface of Liouville, in con- sequence of theorem [31.1] and §31, Ex. 3. 16. For a surface of Liouville with the linear clement (32.19) the angle 0 which a geodesic makes with the curves ut = const. is given by Ax sin® Hy) — As cos? +a = 0, where a is the constant appearing in (82.20). 33. GEODESIC POLAR COORDINATES. GEODESIC TRIANGLES In accordance with theorem [32.1] through a point P of a surface there passes a geodesic in cach direction. Consider a domain about P such that no two geodesics through P meet again within the domain. When then the geodesics through P are taken for the parametric curves 1? = const. and their orthogonal trajectories the curves u’ = const. the linear clement of the surface for the domain under consideration is of the form (32.15) by a suitable choice of the parameter u’. If iis replaced by «’ plus a constant, the form (32.15) is unaltered. It follows from the discussion following (32.15) that u! may be chosen so that it is the dis- tance along each geodesic from the point P. Thus each eurve u! = const. is the locus of a point at a constant distance from P, this distance being measured along the geodesics through P. Tn this sense the curves a’ = const. are geodesic circles. Since at P x = 0, it follows from (24.7) that gx = 0 at P, as is also gz = 0. The former result follows also from the fact that the are of a curve u! = const. between two geodesics u? = 0 and uw = a1 is $33) GEODESIC POLAR COORDINATES 181 given by [ V Gu dv’, and this quantity approaches zero in the limit as. ut lb approaches zero, In like manner the angle at P between these geodesios is given by oo f Von de lim 2 = im win sine the right-hand member being obtained by the use of differentiation to evaluate the indeterminate form of the left-hand member. Con- sequently a necessary and sufficient condition that the coordinate u* be the angle made at P with the geodesic u° = 0 by the coordinate curvesa? = const. is that ee) aut are called geodesic polar coordinates, because they are analogous to polar coordinates in the plane. Hence we have In this case the coordinates 133.1] A necessary and sufficient condition that the coordinates in terms of which the linear element is (33.1) ds = duY + goo du” be geodesic polar coordinates is that = a i) = (33.2) Gabon 0, (VG) wt, Consider, for example, the sphere with the equations = asin“ coi, at aasinesina, 2° = acon! a a ¢ ‘The curves u* = const. are the great circles through the point (0, 0, a). 1 Now gu = 1, g2 = 0, gn = a? sin? %., which satisfy the conditions of theorem [33.1]. From §28, Ex. 1 we have that for the linear element in the form (33.1) the Gaussian curvature (28.5) of the surface is given by -— 1 bon (33.3) ke Me. If K = 0, Vm = aw + b, where a and b do not involve w. In order that the conditions (33.2) be satisfied, we must have a = 1, b=0, thatis, (33.4) ds = du® + ubdu®, 182 INTRINSIC GEOMETRY OF A SURFACE [Cu, IIL This is the linear element of the plane in polar coordinates, and conse- quently, the surface is isometric with the plane. Conversely, if a surface is isometric with the plane, its linear element can be given the form (33.4). ‘Thus we have another proof of theorem [28.1]. If K ¥ 0 for a surface referred to polar geodesic coordinates, it follows from (83.3) and (33.2) that (Pa (33.5) _ -@ Mi) -- CP) Ky e i) aut? Jutao’ aut where Ko is the value of K at the pole of the coordinate system. From (33.2) and (33.5) we have Hence the perimeter and area respectively of a geodesic circle of radius u! are given by a 4 3.0)” Vandit = 2" (u' - px + -), ul pte _ at can [" Vanda = 2(u'- Aa + ). We consider next. the integral of the absolute value of the Ganssian curvature over a geodesic triangle, that is, a triangle whose three sides are geodesics, the triangle being of such size that no two geodesics through a vertex meet again within or on the triangle, and such that that K has the same sign within and on the triangle. Such a triangle is shown in Fig. 9. We choose a geodesic polar coordinate system with pole at P. and with the sides PP; and PP; as the geodesics u? = 0 and a? = a. In consequence of (33.3) the integral is (33.8) rae ff K Vaden =-ef FV G8 aad, ant? where ¢ is +1 or ~1 according as K is positive or negative. Integrating with respect to u' between the limits 0 and w’, we have, in consequence of 3.2), reef ( - ov Gn) au. §33) 183 ESIC TRIANGLE From (32.14) we have along the geodesic P:Ps = 2 Van 4,2 ay = 2 VG ae, and consequently (39) T= e(f ae i) = dat B+y—2). Wesay that I defined by (33.8) measures the lolal curvature of the triangle. From (33.9) it follows that for a geodesic triangle such that no two geodesics through at least one of the vertices meet again within or on the triangle, and such that the Gaussian curvature has the same sign at all points of the same, the total curvature of the triangle is equal to the excess over 180° of the sum of the angles of the triangle or to the deficit P, 7 Pav] Be. 9 from 180° according as the Gaussian curvature is positive or negative. If for a geodesic triangle geodesics through each vertex meet again within the triangle and thus there does not exist a polar geodesic system with one of the vertices as pole which system applies to the whole triangle, it is possible to subdivide the given triangle into a number n. of smaller geodesic triangles, for each of which a polar geodesic system holds, in consequence of theorem [32.4]. Equation (83.9) applics to each of these triangles, and when the corresponding equations are added one obtains ¢ times the sum of all the angles of the n triangles minus nz. The situation at all the vertices of the smaller triangles except at the vertices of the given triangle is the same as when a plane triangle is subdivided into n small triangles. In this case the sum of all the angles of nx, and since the sum of those at the vertices is 7, it follows that the suin of all the others is (n — 1). Hence in the case of a geodesic triangle divided into n geodesic triangles when we subtract (2 — 1)r 184 INTRINSIC GEOMETRY OF A SURFACE [Cu. TIL which is the sum of the angles not at the vertices of the given triangle we obtain the right-hand member of equation (33.9). Hence we have the celebrated theorem of Gauss* [83.2] The total curvature of a geodesic triangle such that at all points K has the same sign is equal to the excess over 180° of the sum of the angles of the triangle or the deficit from 180° according as K is positive or negative. As a consequence of this theorem we prove the following: [33.3] Two geodesics on a surface of negative curvature cannot meet in two points and enclose a simply connected area. For suppose two geodesics through a point A meet again in a point B, and enclose a simply connected area, as in Fig. 10. In consequer.ce of theorem [32.4] it is possible to find two points @ and D on the two geodesics sufficiently near to A so that there is a unique geodesic through Cand D, thus forming two geodesic triangles ACD and BCD. The total curvature of these two triangles is equal to 2¢ mimus the angles C D Fic. 10 at A,B, Cand D. But the sum of the angles at Cand D is 2x, and thus the total curvature of the two triangles is a negative quantity, whereas it ig understood to be positive, as defined in (32.8). Hence the theorem is proved. It follows from theorem [32.6] that if 6 is a solution of Ay@ = 1, the curves @ = const. are geodesic parallels, and @is the are of the orthogonal geodesics measured from one of the parallels. If then we take two curves Cy and C; not geodesically parallel, their gcodesie parallels form a net on the surface. If u' and w denote the geodesic distances of these parallels from Cy and Cy respectively, they are solutions of 410 = 1, and in the coordinate system u" we have from (29.7) (33.10) fe Moy, gy y From these equations and (25.5) we have cos @ we sin? w’ * 1827, 1, p. 246, $33) GEODESIC POLAR COORDINATE 185 where w is the angle of the two coordinate families of geodesic parallels. Hence the linear element of the surface is (33.11) ag = tl + 20s waul du? + dui " sin? . A similar result follows, if we take for coordinate curves geodesie circles with centers at two points F', and F on the eurface. Conversely, equations (33.10) follow from (33.11), and consequently when the linear element of a surface is of the form (33.11), the co- ordinate curves are geodesic parallels which may be two families of geodesic circles. If we put (33.12) w=hub+uy, vu? acurve wu” = const. or u” = const. is the locus of a point such that the sum or difference of the geodesic distances from C; and C2, or the points Fy and F’, as the ease may be, is a constant. In the latter ease these curves are analogous to ellipses and hyperbolas in the plane. Ac- cordingly they are called geadesie ellipses and hyperbalas not only in this case, but also when the distances are measured from curves C1 and C2. From (33.12) and (83.11) we find that in terms of the coordinates u/* the linear clement of the surface is Ci (83.18) ds = {duly +7" sin®§ Conversely, by means of (33.12) the linear element (33.13) is trans- formed into (33.11), and consequently when the linear clement of a . 4 + : 1 surface is expressible in the form (33.18), that is, when “+ = =1, nt gn the coordinate curves are geodesic ellipses and hyperbolas. As in the case of confoeal ellipses and hyperbolas in the plane, we have [38.4] A set of geodesic ellipses and hyperbolas form an orthogonal net EXERCISES 1, The angles of any two families of geodesic parallels on a surface are bisected by the corresponding geodesic ellipses and hypcrbolas. 2. In order that a coordinate system of geodesic ellipses and hyperbolas on a surface form an isometric orthogonal net, it is necessary and sufliciont that the linear element be of the form dst = (Ur 4 Us)(du 4 dus), where U, and U, are positive functions of ut and u? respectively. 186 INTRINSIC GEOMETRY OF A SURFACE [Cu. IIT 3, When the plane is referred « a system of confocal ellipses and hyperbolas, whose foci are at the distance 2a apart, the linear element can be written aut eh 4, Anccessary and sufficient condition that an orthogonal coordinate net on a surface be a system of geodesic ellipses and hyperbolas is that Ur Us a+ Ou ae 2 dst = (ut? — ut) (a where 7; and U, are positive functions of ut and 1? respectively. 5. If an orthogonal coordinate net is to be a system of geodesic ellipses and hyperbolas in two ways it is necessary and sufficient that there be two sets of positive functions Usg(a = 1, 2), where Ua: and Us: are functions of ut and ut respectively sueh that @ — + in this case the linear clement of the surface is reducible to the form in Hx, 2. 6. If the conditions (i) of Ex. 6 are satisfied, it follows that tat 1= Ua v2 which for (x constant is of the form (i); consequently, if an orthogonal net con sists of a system of geodesic ellipses and hyperbolas in two ways, it is such a, system in an endless number of ways, 7. When two families of geodesics on x surface mect under constant angle and their orthogonal trajectories are the curves u* = const., the angle» in (33.11) is, a constant, and consequently the surface is isometric with the plane. 8. When gz: in theorem [33.1] is 9 function of 1! alone, it follows fram (38.2) and (33.3) that for a snrface of constant Gaussian curvature one has (sec §31; Ex. 3) 1 w o Kea) ga = atoint @ a ss 1 ape co) K=- 5, gm=atsinht. a a 9. If C denotes the circumference of a geodesic cirele of radius a, the Gaussian 3(Qra — C) curvature K at the center of the circle is given by lim a as follows bm from (33.6). 34. GEODESIC CURVATURE We returm to the consideration of equations (32.7) when yu" is not a zero vector, that is, when the curve is not a geodesic. From (32.6) it §34] GEODESIC CURVATURE 187 follows that this vector is perpendicular to the tangent vector to the curve, that is, to the unit vector ae We replace u* in (32.7) by ds kyu and obtain (34.1) au (hae du" oo whore now it is understood that 4" is the unit vector which makes a right angle with the unit vector He , that is, (see theorem (25.51), (34.2) Since u“ in (34.1) is a unit vector, the absolute value of x, is the length of the vector whose components are the left-hand members of equations (84.1). We call the vector 1” the curvalure veclor of the curve at a point, and k, the geodesic curvature of the curve. From (34.1) and (34.2) we have _ op [ae (et je au m= Va [ee (@ + {2 ds i) du (dub 1 \ du’ du” (34.3) -& Ge + tod be *)| at (bl phew au = (Oe + Wal as ds)” For a geodesic x, = 0 as follows from (4.1) and (82.8). Conversely, if x, = 0, it follows from the first form of (34.8) that du, fa\du'du’ _ du® _ a ‘s we (am?) where ¢ is a factor to be determined. In consequence of (32.6) we have . du? ae that ¢ = 0, since 908-9, Gg =) Hence we have [84.1] A necessary and sufficient condition that a curve be a geodesic is that the geodesic curvature of the curve be zero. When the coordinate curves form an orthogonal net, for a curve u* = 1 const. we denote the geodesic curvature by ka and take ae = # : 0 that is, we take the tangent vector in the direction of u’ increasing. By means of §28, Ex. 1 we have from (34.3) 8: = — 12s vou (84.4) ket Yn me 188 INTRINSIC GEOMETRY OF A SURFACE (Cu. IIL : In like manner for a curve w = const., if we take@% = — 1, the ds Von geodesic curvature «1 is given by (34.5) ga = — 12108 Von Von 3 ~ ide Ls dae From (34.2) and the choige 9 = - (in the frat ease and 9% = 1. + . « we ~ pin the second ease it follows that in the fst case n#is positive, and in the second case 4’ is positive. Hence in the first case the normal y* is tangent to the curve u' = const. and in the direction in which wv’ is in- Fre. 1 creasing, and in the second case it is tangent to the curve u? = const. and in the direction in which w' is increasing, Consider now any curve C on the surface, and two nearby points P and Pion. At P and P; draw the unique geodesics g and g; tangent to C, and denote by Q their point of intersection and by Ay the angle under which they meet, as shown in Fig. 11. We shall prove the following theorem: [34.2] The limit of the ratio x as P; approaches P along the curve is the geodesic curvature of C at P. In order to prove this theorem we take the curve C as the curve u' = 0 of an orthogonal coordinate net. The tangent geodesic at P makes with the curve u® = ¢(>c;) the angle 87/2 and with the curve wv’ = «1 the angle + a, hence At) = a — =/2. From (33.9) applied to the 34) GEODESIC CURVATURE 189 geodesic triangle QP:R we have Tee(ap 4% ta—n) = olay + an). As P, approachs P along the curve, approsehes zero and consequently Ay dd. sae gi : 5% approaches — 2, the latter being given by (32.14). Hence we have lim Y = — 7 du __ 1 Oleg Von ds Vin ow ds gn and thus in consequence of (34.5) the theorem is proved. Tn view of this result geodesic curvature of a curve on a surface is a generalization of eurvature of a plane curve, and from theorem [34.1] we have that a gcodesic is a generalization of a straight fine in the plane. ‘The geometric significance of geodesic curvature as viewed from the enveloping euclidean space is established in §44, We shall show in what manner the quantity x, fora curve ow, u*) = const., is expressible in terms of differential parameters of ¢. To this end we obscrve that for the coordinate system used in establishing equation (34.4) we have from (29.7), (29.2), and (29.16) 1 av ge 2 Aad On Ow : 1 _ aul = Tu, Vo —- ‘Gn Gee By means of these expressions equation (34.4) can be written in the form _ fae 2 .)] w=-[Seh ta(e 2), Tn any other coordinate system the curves u’ = const. are defined by an equation g(u', u°) = const., and consequently we have the theorem of Beltrami:* {84.3] The geodesic curvature of a curve ¢ = const. is given by L (34.6) G=- [$e +A CG +). Vane Vaio. By means of . 16) the quantity in brackets is equal to a (va 0" vee) Vo Vitenea * 1865, 1, p. 83. 190 INTRINSIC GEOMETRY OF A SURFACE [Cu. TIL and consequently (34.6) may be written in the following form due to Bonnet :* (31.7) w--) 2 (var “en, (Vg! Vp ere In particular, if kq1 and yp denote the geodesic curvatures of the curves u® = const. and u' = const. respectively of any coordinate system, we have (see (24.18) 1a (x ae a) Out (34.8) vo vote (3 # @) 1 (0 Gas “Vg (ore Voje0 oe Von): of which (344) and (34.5) are particular cases when the coordinate curves form an orthogonal net. Equation (32.13) gives the rate of change with s of the angle 0) which a curve makes with the curves u* = const. By means of this result the geodesic curvature of the eurve, as given by (34.3), is expressible in the form do VG ayer (34.9) oat =: Koa With the aid of this equation we establish an important result due to Bonnet. To this end we use the formula of s namely (34.10) I(& - 3) a! dy! = fu du + uy du’, In this formula the surface integral is applied to a simply connected portion of the surface, and the curvilinear integral to the contour C, the positive sense of the latter being such that as a point describes C in this sense the portion of the surface is on the left. It is understood that #1, «2 and their first derivatives are finite and continuous within and on the contour. From (34.10) and (84.9) we have a4. [ate [uae = ~ [YO (ih a + (hae) 34, Wee ti = 3 Bae \) ata. * 1860, 1, p. 166. + Fine, 1927, 1, p. 397. §34] GEODESIC CURVATURE 191 Since the left-hand member of this equation is independent of the co- ordinate system wu’, the same is true of the right-hand member. By (25.21) the element of area in any coordinate system isde = Vg die du’. If then the coordinate system is such that the linear element is of the form (88.1), one finds by means of §28, Ex. 1 that the quantity in brackets in, (34.11) multiplied by 1/Vg reduces to — ovis, Gan Ow as the reader should verify. From this result and (33.3) it follows that in any coordinate system the integrand of the right-hand member of (4.11) is K do. If. the contour C consists of arcs of curves forming a curved polygon with exterior angles 01, --- , 0p, the first integral in (84.11) is equal to 2n ~ 34. Honce we have the Gauss-Bonnet thearem:* [34.4] For a simply connected portion S of a surface for which the Gaussian curvature K is finite and continuous and the geodesic curvature ky of the contour C is finite and continuous (34.12) f ky ds + ff KV/q du dt = On ~ So, where 0, +--+, Op are the exterior angles at the vertices of the contour C if any. When the coordinates u* undergo any transformation, the quantities Ha in the right-hand member of (34.10) are seen to be the covariant. components of a vector. ‘he integrand in the left-hand member of (34.10) is equal to (ue, — .2)du'du’, In terms of the quantities «”* defined in (25.14) equation (34.10) may be written (34.13) f pedi = f f naa Vo de ad, ° If A“is any vector, then pq defined by (34.14) ba = e4aX9 are the covariant components of a vector since ¢_ are the covariant components of a tensor (see theorem [25.4]). When the expressions from (84.14) are substituted in the left-hand member of (34.10), we *See 1848, 1, p. 131. 192 INTRINSIC GEOMETRY OF A SURFACE [Ca, I] obtain in consequence of (25.14), (28.2) and §28, Ex. 12 ie (ear) ~ 2, (oor) aud = I f Loe vara ae -[fuvi Va du ae. Hence another form of the formula of Green is [[ revi aut = fart ae = = [van as, where at (34.16) P= 9 Fj thus yy are the covariant components of the unit vector which makes a right-angle with the unit vector by theorem [25.6], and whose di- rection depends upon the sense of rotation (§25). EXERCISES 1. The parallels on a surface of revolution are curves of constant geodesic curvature, 2. A small circle on sphere has constant geodesic curvature. 3. When a surface has an orthogonal net such that the curves of one family are geodesics, and thoao of the other family have constant (s 0) geadosie enrva- ture, the surface is isometric with a surface of revolution, 4. For a family of loxodromic curves upon a surface of revolution (see §26, Ex. 2), that is, curves which make the same constant angle with the meridians, the geodesic curvature of all these curves is the same at their points of inter- section with a parallel 5. For a surface with the linear clement gee = 8 tat (i+ Uy” jon of u* alone, xg = Us, xo = Uy, where the primes indicate differentiation with respect to the argument, that is, the coordinate curves have constant geodesic curvature. Conversely, when the curves of a coordi orthogonal net have constant geodesic curvature the linear clement is reducible to the above form. 6. If tho curves of one family of an isometric orthogonal net have constant geodesic curvature, the curves of the other fainily have the same property. where Us is a fant $34] GEODESIC CURVATURE 193 7. The geodesic curvature of the integral curves of the equation My dut = is given by L Siete) Vo a Vo? My Ms, 8. Derive the theorem of Gauss (33.2 from theorem [34.4]. 9. From (34.8) and (25.5) one has fo ve (cas Vad — ove] xa). 10. From the discussion of (34.11) it follows that xe LP 2 (wala) 2 (vo(2r)], Vala ga ul) ae g, iat) f this should be verified by means of (28,5), (28.2) and (20.18). 11. From §28, Ex. 10 and (34.8) one obtains the formula of Liouville 1 Bw eo ata a Vie) + 2 (Von) | 12, When the coordinate curves form an orthogonal net, the formula of Liou- ville may be written a i 2. atom Ki in qf and equation (34.9) may be written (sce §28, Ex. 1) dy ‘ Ka = Fy t £08 Bosgr — sin Dox 18. From (29.2) and (29.14) one has Aly, ¥) do = | | [G*e2¥),8 — vargl do, [Jotove-[f from which and (34.15) it follows that [fatonaet ff vrede= - [ve eared 14, From (29.16) and §28, Ex, 12 one has f [aevoaw au = [ [ EOVi0%, op) dae = f [oeadaVadu au, and from this result und (84.15) [fovea [ourteat 194 INTRI IC GEOMETRY OF A SURFACE (Cu. HIT 35. THE VECTOR ASSOCIATE TO A GIVEN VECTOR WITH RESP! ( TO A CURVE. PARALLELISM OF VECTORS Let @ be any curve upon a surface defined by u" — f%(s), where s is the are of C, and let “be the components of a family of unit vectors one at each point of C, \* being functions of s. Since the vectors are unit vectors, (35.1) gasd"W = 1 Differentiating this equation with respect to s and making use of (28.2), we obtain (35.2) oan (Se ate <} ae =0. The components \’* of these vectors in any other coordinate system u!* are given by mn ou yr a nou (35.3) x ie aa Differentiating the first of these equations with respect to s and making use of => we may write the resulting equation in the form ae ay” du y cay Tw Gh Sr oe Ge or (oh ar) Hence, if we put A" ye falal _ ds vB} ds : it follows from (35.4) that »* are the contravariant components of a yector, and from (35.2) that the vector v* at a point is perpendicular to the vector \* at the point. Following Bianchi* we call »* the vector associate to \“ with respect to the curve C. (35.5) When in particular the vector A” is the tangent vector a v® isthe vector xu in equation (34.1), and we have [35.1] The vector associate to the tangent vector with respect to a curve 18 the geodesic curvature vector of the curve. Following McConnellf we call the left-hand member of (85.5) the * 1922, 1, p. 161. +1981, 1, p. 179-181; see also Synge, 1926, 2. $35) VECTOR ASSOCIATE TO A GIVEN VECTOR 195 intrinsic derivative of \* for the curve Cand denote it by Bs , that is, , Bd" _ ant a\ dub (35.6) _-=— 7 a as ds + {ah ds ‘Thus the intrinsic dorivative of a contravariant vector for a curve is a contravariant vector. In like manner by means of (32.1) it can be shown that the intrinsic derivative of a covariant vector A defined by de De _, friar’ de a de is a covariant vector. és ds Tf \* and d, are the contravariant and covariant components of a vector defined at all points of a surface, then for a curve one has (35.7) no De _, du? (35.8) _ In like manner if one has a tensor a$1""j2" defined at all points of a surface, at points of a curve the quantities (35.9) are the components of a tensor of the same order. The results con- cerning in differentiation follow from equations (32.1), just as the results concerning covariant differentiation in §22 are a consequence of the corresponding equations (20.11). Hence for intrinsic differentiation the rules of the ordinary caleulus as regards the differentiation of the sum, difference and product of quantities apply (see §22). Also just as the covariant, derivative of a scalar is the ordinary derivative of the scalar, so the intrinsic derivative of a scalar is the ordinary derivative. From the above results and theorems [28.3] and [28.4] we have [85.2] For any curve Bas bg"? Beas e* 35.10) 0s? = a = be (35.10) is 7% és , bs , és From (25.16) we have that the angle 9 made with the unit vector a tangent to a curve C by a unit vector A“ at all points of C and not tangent to C is given by . du* (35.11) sin 9 = eg N. 196 INTRINSIC GEOMETRY OF A SURFACE [Cu, III In consequence of (35.10), (84.1) and (35.5) the intrinsic derivative of equation (35.11) is reducible to 0 a ° (35.12) cos = tap (so + “ “). When C is a geodesic, that is, when x, = 0, the condition that the angle @ be a constant is a = 0, that is, either »° is a zero vector (#° — 0), or the vector » is tangent ; to C. In the latter case we should have from (35.2) gag\* ae = 0, the 2 intrinsic derivative of which is gag a, that is, »% is normal to C. Hence »* = 0 and we have [35.3] The unit vectors \* at points of a geodesic make equal angles with the geodesic, if and only if at _ A yy fala’ (35.13) w= a +X iG a 0. When the surface is a plane, in which case a geodesic is a straight line, the vectors \“ are parallel in the euclidean sense. Accordingly it is a natural generalization to say that the family of vectors \* satisfying (35.13) are parallel with respect to the geodesic involved. Although we have introduced the notion of the parallelism of vectors at points of a geodesic, we make a further generalization and say that given uny curve, Where wu are functions of the are, the functions \” satisfying (35.13) are the components of a family of parallel vectors with respect lo the curve; this concept is due to Levi-Civita (see §45). From the theory of differential equations of the form (35.13) it follows that any such family of parallel vectors is completely determined by the values of the components at some point of the curve, that is, by initial values of A* for the value of s at the point. Since equations (35.13) involve the equations of the curve, it follows that if one has two curves intersecting at two points P; and P2 , and finds with respect to each of the curves the families of parallel vectors having the same values at Py, in general the vectors at P2 will be different for the two families. Thus parallelism as just defined is relative to a curve. When the surface is a plane referred to cartesian coordinates, equations (35.13) reduce to a = 0, that is, ” are constants, and parallelism is absolute, Ut is, $35) PARALLELISM OF VECTORS 197 it is not relative to a curve. In this case the vectors are parallel in the euclidean sense. Since for 4° = ae equations (5.13) are equations of geodesies, we have that the tangents to a geodesic are parallel with respect to the curve, and in this sense geodesies may be called the straightest lines on the surface. In this connection it is interesting to observe that for the domain about a point P within which no two geodesics through P meet again coordinates u* can be chosen in terms of which equations of the geodesics through P are given by (32.12), that is, w= (a) "ds Jo’ where u” is a given set of coordinates on the surface. Thus in this coordinate system equations of the geodesics through P are of the form of equations of straight lines in the plane referred to cartesian coordi- nates. The equations of geodesics not through P do not have this simple form in this coordinate system. Since equations (35.13) involve only the first fundamental form of a surface, parallelism is an intrinsic property of a surface. As a conse- quence we have [35.4] If n° are the components of a family of parallel vectors with respect to a curve on a surface, they are the components of vectors parallel with respect to the corresponding curve on a surface isometric with the given surface. If \“ are the contravariant components of a unit vector-field such that (35.14) A%3 = 0, the vectors of the field at two points P; and P; are parallel with respect to every curve through the two points and thus are absolutely parallel. In particular, for the plane referred to cartesian coordinates equations (35.14) reduce to x = 0, that is, 47 = const, The conditions of integrability of equations (35.14), namely the Ricci identities (see (22.22) (35.15) Moy — Abas = — NM Rg = ~ Ng Rosy, are satisfied identically for a plane, or any surface isometric with the plane. In this case equations (35.14) are completely integrable, that is, 198 INTRINSIC GEOMETRY OF A SURFACE | [Cu. ILL a field of absolutely parallel vectors is determined in any coordinate system by initial values of \*. Conversely, in order that there be a field of absolutely parallel unit vectors in a surface, its contravariant components must be solutions of equations (35.14), and consequently from (35.15) one has that \ Rasy = 0, which equations are equivalent to (see (28.3)) Rane = 0, Ruz = 0. Hence Riz = 0 and we have [35.5] A necessary and sufficient condition that there be a field of abso- Iutely parallel unit vectors in a surface is that the surface be isometric with the plane. In the plane the angle of two vectors not at the sume points is by definition the angle between either vector at its point of application and a vector at this point parallel to the other vector. In consequence of theorem [35.5] this definition does not apply to a general surface. However, we can speak of the angle of two vectors relative to a curve, this angle being the angle between either vector at its point of applica- tion and a vector at this point parallel to the other vector with respect to the curve. ‘The angles thus formed at each point are equal, these being the angles referred to in Ex. 2. We return to the consideration of equations (35.5) when »* # 0, and recall that the vector » is perpendicular to the veclur AS. We write equations (35.5) in the form a fala oN +X (35.16) és ds \vBs ds 7? where now v* is the unit vector which makes a right angle with the vector A". In this case the absolute value of r is the length of the associate vector, Differentiating intrinsically the equation gad"? = 0, we have in consequence of (35.10) and (35.16) ao? 6. eed" = +7 = 0. Sine v* is a unit vector and —\* makes a right angle with it, we have oF oT *, On substitution in the above equa- analogously to (35.16) §35) PARALLELISM OF VECTORS 199 tion we find that 7’ Hence we have “ 8 1 falda « —=-rr +e Vel he When, in particular, \“ is the unit tangent vector to a curve, »“ is the curvature veetor of the curve, and ris the geodesic curvature. Conse- quently we have y og (35.17) (35.6] If * and u* ave the contravariant components of the unit tangent vector and unit curvature vector of a curve on a surface, then a ylalae = at” {op a = aunt bus _ du® | yfa\ du? a SE en SE a ht +e Nal as ” These are the Frenet formulas of a curve in a surface (see §21), If the unit vectors \“(s) are parallel with respect to a curve C, the angle @ which \* makes with a unit vector u“(s) at cach point of C is given by (see (25.16)) (35.18) sin 0 = exp’. Taking the intrinsic derivative of this equation and making use of the first of (25,9) and (35.13), one obtains do %) M (up — eee ~) = 0. (m ds °° s In consequence of Ex. 2 the expression in parentheses is independent of the choice of the parallel vectors A“, and consequently dO LH ds “Ge yi 68 When, in particular, 2* is the unit tangent vector to C, this equation becomes, in consequence of theorems [35.1] and [25.5], (35.19) g = Koy where x, is the geodesic curvature of C. Hence we have* 35.7] If a vector undergoes a parallel displacement along a curve C the arc-rate of change of the angle which the vector makes with the curve 4s the negative of the geodesic curvature of C. * See 1927, 2, p. 136. 200 INTRINSIC GEOMETRY OF A SURFACE (Cx. TIL Consider a smooth closed curve C in a surface, enclosing a simply connected region, and a vector \ ata point P of the curve. Let this veetor be displaced parallel to itself around C in the positive sense, and denote by i its final position at, P, and by g the angle X makes with d. ‘This angle is equal to its rotation relative to the tangent to C plus the rotation of the tangent about C, namely 2m. In consequence of (35.19) we have (35.20) on [fs ds = ff Kaj gui ae, 8 the last expression being a consequence of theorem [34.4]. We observe that this result is independent of the position of the point P on C and of the veetor A. EXERCISES 1, ‘The covariant components bz of a set of unit. vectors parallel with respect to ‘a curve are solutions of the equation ay { yy aut ds“ \aaf ds 2. For two families of unit vectors Af and X§ parallel with respect to a curve the angle between the vectors at a point is the same for all points of the curve 3. A necessary and sufficient condition that the unit tangent vectors to the curves u* = const, for a = 1 or 2 be parallel with respect to a curve C is that the latter be an integral curve of the equation ol ae @) {aha (8# a) wut veetors to a fumily of geodesic parallels (§32) at points of ics are parallel with respect to 4. The unit tas interseetion with cach of the orthogonal geod: the geodesic. 5, When the coordinate eurves of a surface form a 'Tchebychef net (§29, Ex. 12), in whieh ease the linear clement is du!* + 2 cos du' du? + du”, the tangents to the parametric curves of cither family at their points of meeting with a curve of the other family are parallel with respect to the latter. 6. Let P:, Pz, Ps be the vertices of a geodesic triangle on a surface, and 1, 6, 0 the interior angles at these respective points; when the tangent vector at P, to the geodesie Ps is transported parallel to itself around the triangle in the direction PPal’sP1 it makes the angle ~ 6; — 6; — with its original diree- tion at Py. 7. If A*(s) is a solution of equations (8) is any funetion of s, then [rae (et + (th) (2 5.13) and one puts A* = w%e(s), where §36) CONFORMAL CORRESPOND 201 is independent of ¢(s), and this is the condition that a family of non-unit vectors be parallel with respect to the curve, 8, Show that theorem [34.2] is a consequence of equation (35.19). 9. Given a field of unit vectors \* and the eurves C to which these vectors are tangent; from (35.18) one has w Mad = wot, wha! = —Kpd*, where p# is the unit vector at each point P of a curve C normal to Cand xy is the geodesic curvature of Cat P. Since wus = 1, Mus = 0, one has (see §28, Ex. 14) wig = wads from which and the second of equations (i) it follows that Mrp = — *) and con- sequently wie = — Ky. 36. CONFORMAL CORRESPONDENCE OF TWO SURFACES In §27 we defined as isometric two surfaces such that there exists on each a coordinate system in terms of which the first fundamental forms of the two surfaces are identical. For two isomcirie surfaces corre- sponding lengths are equal, and corresponding angles are equal as follows from (25.2) and (25.3). ‘The converse is not true, that is, the equality of lengths is not a consequence of the equality of angles. We consider now (he case when two surfaces are in one-Lo-one correspond- ence such that corresponding angles are equal. In this case the corre- spondence is said to be conformal. Given two surfaces S and 5 expressed in terms of coordinates u® and 4", a one-to-one correspondence is established between points of the two surfaces by equations of the form ut — o@, @), provided the jacobian of the g's with respeet to the @’s is not identically zero. If then we change the coordinates a* on S into coordinates u* by means of the above equations, corresponding points on the two surfaces have the same coordinates. In terms of these coordinates we denote by gag and gas the components of the first fundamental tensors of S and § respectively The orthogonal trajectories of the curves u’ = const. on the two surfaces are the integral curves of the respective equations (see [26.2]) die Since these trajectories correspond, we must have (36.1) gis da + gon? = 0, Ai? = 0. da _ Oe gia Jen 202 INTRINSIC GEOMETRY OF A SURFACE (Cu. Til In like manner from the equations (36.2) gid + ged’ =0, gudw + grdv’=0 of the orthogonal trajectories of the curves u* = const. we must have Ju _ de gu gn Hence we have the three equations (36.3) Gea _ 2 (a, B = 1,2), Gas where r is a function of a? and uw’. Consequently (36.4) as =r ds*, Conversely when (36.8) is satisfied it follows from (25.2) and (25.3) that any two corresponding angles on the two surfaces are equal, and we have [86.1] A necessary and sufficient condition that two surfaces having two corresponding nets of curves as coordinate curves be in conformal correspond ence with points having the same coordinates corresponding is that the first fundamental forms of the two surfaces be proportional. We say that the correspondence is direct or inverse according as corre- sponding angles have the same or opposite sense. In §30 it was shown that the first fundamental form of a surface is expressible in the form (36.5) P(de* + aut) and that in (36.8) wt =f ta), ul — iu? = fw F a’), where f and fo are conjugate functions, the first fundamental form is e or If ‘Thus the equations w'* = u" determine a conformal correspondence of the surface upon itself, and from theorem [30.4] it follows that the most general correspondence is defined by (36.6). Hence we have s of real coordinates u’* defined by (see (30.19) (36.7) ((du") + (du). [36.2] When a pair of isometric coordinates u* of a surface are known, the most general conformal correspondence of the surface with itself is obtained by making the point (u', u*) correspond to the point (wu, uw), where w* are defined by equations (36.8). $36) CONFORMAL CORRESPON DENCI 208 Also we have [36.3] When for each of two surfaces « pair of isometric coordinates u™ and u!* are known, the equations u!* =u define a conformal correspandence of the surfaces, and by means of equations of the form (36.6) applied to one of the surfaces all the conformal correspondences of the two surfaces are obtained. ‘Thus a conformal correspondence can be established between any two surfaces and in a great variety of ways, but it is not possible to establish an isometric correspondence of any two surfaces, that is, the particular conformal correspondence for which r = 1 (see §31). If we denote by 6 and 65 the angles made at a point by a curve with the curves u? = const. and u” = const. on a surface with the funda- mental forms (36.5) and (36.7) respectively, it follows from (25.6) that dul du Vai (ay du‘, du Verret COS Hy , SiN A = which in each case is an abbreviated way of writing two equations. From these expressions we have, since e“ = cos @ + isin 6, Me du! tid gto = C2 _ du tidw oe dw ide? From these equations according as the first or second signs are used in (36.6) we have aa) iw)’ sure MARY awagiey _ Sw! (36.8) — (8) fl = iat)’ falw If & and 8) are the angles made by a second curve with the curves u’ = const. and u® = const. respectively, it follows from the first of (36.8) that 4 — 6 = 8 — 4, and from the second that — = ~ (6 — 6;). Consequently we have [36.4] In the conformal correspondence of a surface with itself defined by (36.6), angles have the same or opposite sense according as the first or second signs are used in (36.6). From the foregoing discussion it follows that conformal correspond- ence is intimately related to functions of a complex variable. In treat- ments of the latter there is extensive study of conformal correspondence in the plane (see Exs. 1, 2). 204 INTRINSIC GEOMETRY OF A SURFACE (Cx. 1 EXERCISES 1, For the plane referred to cartesian coordinates the equation 2 aint wm igt defines a conformal correspondence of the plane upon itself such that the lines re = const. correspond to two families of circles all passing through the origin with the centers of the cireles of each family on one of the coordinate axes; also corresponding points are on a line through the origin and their distances r and # from the origin are in the relation 17 = ¢* 2, In terms of coordinates u* in the plane defined by was athist, uta ot mist, ian coordinates, an equation of any real circle not passing of the form where 2 are cart through the origin @ autut + but + cut + d = 0, where @ und d are real numbers, and b and ¢ are conjugate imaginaries, The equations bya + be “a bai by define a conformal correspondence of the plane upon itself such that circles correspond to circles or straight lines, 3. For the amit sphere with the equations, Cwatt “but! ww | 2 Faw ge ut and w being conjugate imaginaries, equation (i) of Ex, 2 is an equation of a circle on the sphere, and equations (ii) of Ex. 2 define a conformal correspondence of the sphere with itself in which cireles correspond to circles. 4, When in Ex, 3 the second of equations (ii) in laced by a at = a — at + ay" the correspondence of the sphere with itself is isometric, and defines a rotation of the sphere into itself, when the a's are chosen go that the correspondence is real. 5. The equations of Ex. 3 and a gle +), Be for the plane define a conformal corresponde that corresponding points are an the line with the equati x Xt x1 wee” if w) 37| GEODESIC CORRESPONDENCE, 205 where X‘ are cartesian coordinates in space; thus corresponding points lie on @ line through the point P (0, 0, 1), and consequently a line in the plane corresponds to the circle on the sphere which is its interscetion by the plane determined by P and the line; this correspondence is called the stereographic projection of th sphere on the plane 6. When iwo surfaces 8 and S are referred to coordinates ut and g = 9, cor responding elements of area (sce §25) are equal, and the correspondence is said tobe equivalent. Ifg ~ gand onc effcets the change of coordinates w/t = ylu!, uw), w? = u? on S, where g is defined by : — Vai ~ Vale 28, e then g/(w', uw) = g(u’t, w’), and consequently an equivalent correspondence between S and 5 is defined by the equations wu’! 7, Let S and S be two surfaces such that the points with the same coordinates us correspond, and let geg and gap be (he first fundamental tensors of Sand S respectively; if the correspondence is not conformal, on cach surface the integral curves of the equation obtained from equation (26.24) when deg are replaced by Goa form an orthogonal net, as follows from theorem [26.4]; if the correspondence is conformal, any orthogonul net on one surface corresponds to an orthogonal net on the other surface. wi wt = 37. GEODESIC CORRESPONDENCE OF TWO SURFACES In this section we consider the relation betwecn two surfaces S and § in one-to-one correspondence such that to each geodesic on either surface there corresponds a geodesic on the other, corresponding points having the same coordinates, as shown in §36. In this case the two surfaces are said to be in geodesic correspondence. Since the are of a curve is peculiar to the curve and thus cannot be tuken as a common parameter of corresponding curves on the surfaces Sand §, it is necessary to express the equations of the geodesies in terms of some common parameter , that is, u" as functions of t. In consequence of (82.27) we have as equations of the geodesics on the surface S in terms of a general parameter ¢ au (eu { tu aur ae (fat 1 a day ray EEE ee ae) ae a ae ae) <0 In like manner equations of geodesics on 3 are da (@e (XT da du) _ at du fi tu da) (37.2) (Ge +f aa) a\ae *\erfa a) ~ where the Christoffel symbols {2 are formed with respect to the fundamental tensor gag of 5. Since it is understood that the coordinate 206 INTRINSIC GEOMETRY OF A SURFACK [Cx II curves on S and § are such that the points on the two surfaces with the same coordinates correspond and that the parameter fis the same for both surfaces, on subtracting equations (37.1) from (37.2), we obtain du! _ du wan / (7.8) (4 ci, ~ OF a) a = 0, where a sft Aah fe {ey If we denote by { ay } and { by } the Christoffel symbols for a coor- dinate system wv, it follows from equations (32.1) and analogous equa- tions for S that ag, Ou Ow gn aut By sala 5h =O yh? which are equivalent by (24.22) to ad = ag, ae He au ” Sule au aue Hence, a3, are the components of a mixed tensor, contravariant of the first order and covariant of the second order (see §20, Ex. 11), sym- metric in the indices 8 and y as follows from (87.4). Equation (37.3) must be satisfied identically; otherwise we should have an equation of the first order and third degree satisfied by all the geodesics; this would t through each point there would pass at most three geodesics. ‘Thus it could not be true that a geodesie passes through a given point in any given direction. In order to obtain the conditions upon the tensor af, so that equation (37.3) shall be an identity, we observe that this equation may be written in the form ae pew y duh du" dué (88 aby — Bay) “Te ae ae =0 for a # wand that the latter is ansidentity for a = p. If these equa- tions are to be satisfied identically, we must have (sce §19, Ex. 5) at ah, — blag, + 8Fa%, — das, + fats — sas = 0. Contracting for w and ¢, that is, putting » = ¢€ and summing with respect to ¢, and noting that 6; = 2 and that aj, is symmetric in 6 §37) G DESIC CORRESPONL 207 and y, we obtain as a necessary condition Bas, = GF a5 + by ape. Since aj, is a tensor, a’, is a covariant vector, which we denote by 3a, (see §19). From this result and (37.4) we have (37.5) {e| = {eh +a 1 + Sag. ¢ af orf + Be + 8545 When these expressions are substituted in (37.2), the resulting equation reduces to (37.1), and consequently (37.5) are sufficient as well as necessary conditions that S and 5 be in geodesie correspondence. Contracting equations (37.5) for @ and 8, we have in consequence of the third of equations (28.2) _. V3 aur Vo From (25.8) it follows that g/g is a scalar, and consequently a, is the 3a,. . so vont ve wo hays gradient, that is a, = °F, where e = log’. Hence we have [37.1] A necessary and sufficient condition that twa surfaces with funda- mental tensors gen and Jap be in geodesic correspondence is thal their Christoffel symbols be in the relation (37.6) {s} = {sg} + iF oy + B90; the function o is equal to i log 2. g If we denote by R‘ys the Riemann tensor for gas , we find from (37.6) and (20.13) (87.7) Fog = Raa + 85 eo, — 85 908, where 87.8) or = Ptr — CBP: sy being the second covariant derivative of y based upon gap - From (28.5) and (28.3) we have (87.9) Rims = g “Rea = K(85905 — 83901). Contracting for ¢ and 8, we obtain, since 8 = 2, 87.10) Rpy = —Kgpr- 208 INTRINSIC GEOMETRY OF A SURFACE (Cx, Contracting (37.7) for ¢ and 8, we have (87-11) Fes, = Ray + 85 from which, (87.10), and a similar equation for Rg, we have (87.12) er = Kon — Kas, , where K is the Gaussian curvature of 5. We consider, in particular, the case when K is a constant. From (87.8) and (37.12) we have (87.13) Finding the covariant derivative based upon the tensor gag, we have 714) aay = ay 0.9 + tate, — RyGaas — K From equations similar to (28.2) we have #0 ne) 4 oa {} aur = arf * 9 ey)? from which on substituting from (37.6), we obtain Gasy = 2Gas Py + Gay Pa + Gay es + Substituting this expression in (37.14) and replacing the gas by their expressions from (37.13), we obtain 7a) °°" = 20.0 Psy + 9 Pre + Py e.09) — Wa O9 Py : — K(2ge5 1 + 9a Pa + Gra e.a) — Kydes from which it follows that Ptr — Pore = K Guy 0.3 ~ Jas 0) ~ Ryas + Riser - From the appropriate Ricci identity (22.19) and (87.9) we have Gaby — Prd = Rey = 9K 65gar — 55905) = K (Gar 9.8 — Gos Px): and consequently Kas — Kaiter = 0. From these equations we have that K is a constant, since the deter- minant of gag is not equal to zero, Hence we have the theorem of Beltrami:* *1869, 2, p. 282. 337] GEODESIC CORRESPONDENCE 209 [37.2] The only surfaces in geodesic correspondence with a surface of con~ stant Gaussian curvature are surfaces of constant Gaussian curvature. We consider in particular the case when K = 0 and assume that the coordinates u* are cartesian. In this case the covariant derivatives of g in (37.15) are ordinary partial derivatives. If then we put ¢ = —} log 0 and thus understand that 0 > 0, equation (37.15) reduces in this case to a0 (7.10) due aud aur = and from (37.13) we have 1 Ges = ike (268,43 — 9,29,a)- ‘The integral of (37.16) is 8 = aagu*u? + byut® + 6, where dug, ba and ¢ are constants, By suitable transformations of the cartesian coordinates this expression for @ is reducible to one or the other of the forms* O= au +e, 6 = au + 2b +e, depending on the values of ag. ‘The expressions for gag for these ex- pressions for @ are respectively _ a(aru™ + 0), —aau'w, (aul + 0) Kau" + of (87.17) Gu, gre, Gee and 7.18) duns dis, den = COME Os abn <0 Kau" + 2bu® + cy From these expressions we have respectively fee 0 .. Ragu +c)” Rau + 2bu? + 0) The constants and the domains of the variables must be such that > 0. Also both gu and gz must he positive. This means for (37.18) that K is negative, since 6 is necessarily positive. *C.G, p. 227. g 210 INTRINSIC GEOMETRY OF A SURFACE, (Cx, IL We return to the consideration of the general case and assume that the surfaces $ and 8 are referred to the orthogonal net on S which cor- responds to an orthogonal net on 5 (see §36, Ex. 7). In this coordi- nate system equations (37.6), on putting = —} log @, reduce to (see §28, Ex, 1) 9 108 Jaa 108 dew _ fee 2 fee _ 9 tog Vex aut ~aue Gut 8g 1 ee Sut (8 # a). Expressing the condition of integrability of the first two of these equa- & log @ aul awe and Us are functions of w! and 1° respectively, both having the same sign since @ must be positive. ‘Then from the first two of the equations tions, we obtain = 0 and consequently @ = U,U2, where U, we have Jou = Ua, (6 # a). When these expressions are substituted Up in the third of the above equations we find that Gea = (Ua — Us)oa(u"). Hence by a suitable choice of the coordinates without changing the coordinate lines the fundamental forms of S and 8 are reducible to the respective forms * (37.19) (Uy — Ula" + aii”, ( - (e + Consequently we have the theorem of Dini* (see §32): [37.3] A necessary and sufficient condition that a surface be in geodesic correspondence with another surface is that it be a surface of Liowille. EXERCISES 1, When the equation (37.3) is written (Gr) + ees ~ a(S) Gt Ge ato (Gi) ~#(az) =° the condition that it be an identity is 4, = 0, iy = ats — ah, = 2ate ~ age = which are equivalent to a5, = 4 (Opa5, + 5505.) = dpa, + S705. * 1870, 1, p. 278. $37] GEODESIC CORRESPONDENCE 211 2. The results of Ex. 1. follow also from the equation of §32, Ex. 7. 3. Since the first of the forms (37.19) is not altered when U; and Usare replaced by Ui + cand U2-+ a, where ais a constant, the second of (37.19) may be replaced by 1 1 du ae (& +a Gt zt tau; ea) 4. For a surface with the first fundamental form (ve du +9 (3) we), w 1 1 . the correspondence defined by ut = =, u? = © isa geodesis correspondence of the (ut — surface with itself. §, From (37.7) and (37.11) one finds that Ways = Wages where Weays = Reaya + Ofhpy — BIR and similarly for f7*,,,.. Thus the tensor Wé,,; has the same components for two surfaces in geodesic correspondence. Determine under what conditions it is a zero tensor - 6, From (37.6) one has fj, = 113,, where } Hale} +0()) By) and similarly for 1i3,. Thus the quantities 1, are equal for two surfaces in geodesic correspondence. In terms of these quantities equations (37.1) and (87.2) become du! (aut dul dur dut (dul , due dur a (Get at a) at (ue rT a) CHAPTER IV Surfaces in Space 38. THE SECOND FUNDAMENTAL FORM OF A SURFACE In the preceding chapter we considered a surface in space defined by three equations (88.1) =F) @ = 1, 2,3), where 2! are cartesian coordinates. Applying the euclidean metric we found that lengths of arcs and angles between curves in the surface are expressible in terms of the first fundamental tensor gas defined by ox (38.2) fea = x ane ae (a, B = 1, 2).* Thus the euclidean metric of space induces a metric on a surface ex- pressible in terms of the tensor gag. ‘'hroughout, the preceding chapter we considered the geometry of a surface in terms of this tensor, and without any further reference to the character of the surface as viewed from the enveloping space. From this point of view any two de- velopable surfaces are equivalent to one another, and to a plane, that i have the same metric properties. Also any surface isometric with a surface of revolution is equivalent to the latter. However, as viewed from the enveloping space two such surfaces may be entirely different in form, It is this question of the geometry of a surface as viewed from the enveloping space which is the subject of the present chapter. This study involves the consideration of the form of a surface in the neighborhood of an ordinary point and its position relative to the tangent plane at the point. In §10 quantities A” were defined by — WSF) (38.3) AY em ray and it was shown in §11 that the tangents to all the curves through av ordinary point x‘, that is, one for which the three quantities A‘ are not * In what follows Latin indices take the values 1, 2, 3 and Greek indices the values 1, 2. 212 38) SECOND FUNDAMENTAL FORM OF A SURFACE — 213 all equal to zero, lie in a plane containing this point, called the tangent plane, and that an equation of the plane is ADAM ER BAB 4 where #' are current coordinates. In consequence of the identity — A" = 0, (88.4) g = | gee | = gugs ~ gis = (A“) + (A®)* + (A")’, established in §24, the quantities X* defined by ae! oat 1 1 loud aw 38.5) xia Lyee , ‘ Vin ~ Salat at oe bel where i, j, & take the values 1, 2, 3 cyclically, are such that (38.6) Dxixiss. 7 Accordingly an equation of the tangent plane is (38.7) Tx - 2) =0, 7 and X‘ are direction cosines of the normal to the surface at the point 2’, that is, the normal to the tangent plane at the point. From (88.5) it follows that* (38.8) Dixie (@=1,2, 7 aus and using (38.4) one may verify that x! ax? ax" dul dul Gul (38.9) ac! aa” a? | = Vt. oe aut he | xox x) We have from (24.4) and (24.8) that, *C.G., Theorem [26.4]. oat . + . = are the direction cosines 1 Vane In (88.5) and (38.9) /g means the positive square root; this convention is Always used. i 5 a 214 SURFACES IN SPACE (Cu, IV of the vector from a point w* on a surface tangent to the curve wv = const. and in the direction in which wu’ is increasing; and similarly for 1 ax' Von 8e surface is an orthogonal net, equation (38.9) expresses the fact that the and the eurve u! = const. Henco if the coordinate net on a as ar’ positive tangent vectors of components a? be and the normal vector X’ have the same mutual orientation as the 2'-, 2*-, and z'-axes.* If the coordinate curves do not form an orthogonal net, we consider the X Tt zie g Fra, 12 result of applying a positive transformation so that the new coordinate curves shall be the given curves w const. and their orthogonal ajectories, the equations of such a transformation being of the form = g(u', v), uw? = wv. Then equation (38.9) transforms into an equation of the same form in consequence of (25.13) and positive seuse in (he surface is unaltered as shown in §25. ‘Henec the mutual orie 1 ari 1 br" ‘ oo ayes, —— SX! is as shown in Fig, 12 what- Vou au" /gz2 au? ever be the coordinate curves. *C. G., p. 162. tion of the vectors $38) SECOND FUNDAMENTAL FORM OF A SURFACE 215 For a curve in the surface through two nearby points x and 2° we have pa yt (a0 1(@a'\ 2 (88.10) a wi (BY h(E) tee where a subscript zero indicates the value of a quantity at 2‘ and s is the are from the point 2°. Since 2* are functions of the coordinates u*, we have csit) 2% = 32 du® @a Fa! duédw’ , ax! Put . ds au" ds’ de ~ purdw de de" due de®” Consequently because of (38.8) the distance p of the point P,(2) from the tangent plane (88.7) at P(z) is given by* @s.12) p= B(dos HO) + B, where B indicates the sum of terms of the third and higher orders in s, and where by definition 2! (38.13) dog = EX iano In consequence of (38.8) equations (38.13) are equivalent to (38.14) dag = 2 X'ziea, where ai, is the second covariant derivative of 2‘, that is, _ &xi aa! { a ~ autau® ~ aur lass” Since z‘ being functions of u“ are scalars as regards a change in the coordinates uw“, as are also X‘, it follows from (38.14) that dag are the components of a symmetric covariant tensor of the second order, being the sum of three covariant tensors each multiplicd by a scalar. It is ealled the second fundamental quadratic tensor of the surface, and : dag du® du’ is called the second fundamental quadratic form of the surface. ; It is shown in §39 that a surface is completely characterized by its first sand second fundamental forms and thus that the distinction between | two isometric surfaces as viewed from the enveloping space is hased upon their respective second fundamental forms. Two surfaces which have the same first fundamental form but different second fundamental forms are said to be applicable. *C.G., p. 95. (38.18) « 216 SURFACES IN SPACE [Cu. IV Gauss,* who introduced the second fundamental form, used the ex- pression Ddp* + 2D’ dpdg + D'’ dy’, as have many subsequent writers. We have adopted the expression dag du“ du*, since it enables one to write equations in simpler form with the use of the summation convention, From (38.13) and (38.8) we have _ ex ar aX! aa (38.16) dag = due due Ou ous” Since X‘ are scalars as regards a change of coordinates u", it follows again from (38.16) that das are the components of a symmetric covariant tensor of the second order. If we differentiate equations (88.2) with respect to u7 we obtain xi ox Fa’ ax’ _ Aes bus Our Bub TAM Bw AUr Bue Bu” If we permute the indices «, 6, 7 cyclically twice, we obtain the two equations Fx! xt aa’ ac! gay © sive aus +2 aso ont = ane? z, Axi at Fa! Or! _ Bre Bu au Bu" 7 Gu" Bu* Bur aut Subtracting from the sum of these two equations the preceding one, and making use of the definition (20.1) of the Christoffel symbols of the first kind, we obtain aa’ az dus due dur From this result and (88.15) we have (see (20.4)) Siig Oh = fas. al - peel las, 1 - 9m {5} =0 Comparing these equations with (88.8), we have that avs is some multiple of X*, say z'ng = fesX*, Multiplying by X* and summing with respect to 7, we have from (38.6) and (38.14) hag = dag , and conse- quently the following equations of Gauss:f (38.18) = dug X*. * 1827, 1, p. 234. 1827, 1, pp. 233, 234. (38.17) = [o8, 7]. & TAI. FORM OF A SURFACE 217 988) |= SECOND FUNDAME : ‘We wish now to find expressions for i . From (38.6) it follows that ax? x Ee Ge =o woh ats ax! Seti from which it is seen that oe for each value of a are direction numbers of a vector orthogonal to the normal. Hence each such vector is a : ‘ . a linear homogencous function of the two tangential vectors i . Thus we have where the expressions for (% are to be determined. In order to find these expressions, we multiply both sides of the above equation by = and sum with respect to 4, and in consequence of (38.16) and (38.2) obtain day = Fe Ie Multiplying both sides of this equation by g” and summing with respect to 7, wo obtain because of (24.16) t = deg”. Consequently the desired equations are ax* ox* (38.19) = = day? ,. (38.19) au" box a8 EXERCISES 1, When dag = 0, it follows from (38.19) that the surface is a plane. 2. When fy de a0), ou Gn i axe bx itfollows from (88.10) that =~ = ~ 157,. From theeonditions of integrability tribal, of these equations it follows that is a constant, since g #0. Hence X* . and the surface is a sphere. sont line I ae! ae, ¥ 3. If a straight line lies entirely on a surface, “j= 0 along the line, ron wo 218 SURFACES IN SPACE (Cx. IV (8.11) and (38.13) it follows that the line is one of the integral curves of the equa- tion das dus dut = 0, 4. For a surface of revolution with the equations geweosut, a= ulsinu, 2 = (ut), one finds that x x xen TOC —¢'sinw,1 OR ” 0, we" du, dia, dp = SEE Vite" 5. For a helicoid with the equations (see §24, Ex. 6) wm uleosu mteutsinut, 2 = y(u!) + au, one finds that asin ut — ute’ cos ut, — (acos # + uly’ sin ut), w X1, Xt, Xo = yaar , Vel +e) + at dy dy, day = wig = ute! : +6) +a 6. Two points x! and 7! are symmetric with respect to the point x4 , if and only if 2 4 2° = 224. Two surfaces symmetric with respeet to a point are isometric, and their second fundamental forms diller only in sign. 7. For any surface Aw = 1X8, Ag(ai, of) =~ XEXE Ga where the differential parameters are formed with respect Lo gug (sec §29). 39. THE EQUATION OF GAUSS AND THE EQUATIONS OF CODAZZT In order that tensors gag and dag be the fundamental tensors of @ surface it is necessary that the conditions of integrability of the equa- tions (88.18) be satisfied, that is, (39.1) Blaby ~ Bars = Ta aay » as follows from equations of the form (22.19). By means of (38.19) these equations of condition are reducible to : OF (deny — days) X' = (ayy — dap, ~ Rana 5 §39] EQUATION OF GAUSS. EQUATIONS OF CODAZZI 219 since Ray = of"Reapy (ee (20.15)). If these equations are multiplied by # and summed with respect to 7, we get, by means of (38.2), (38.8), and (24.16), the first of the following equations, and the second, when the above equations are multiplied by X* and summed with respect to i: (g9.2) daydss — depdys ~ Riasy = 0, daiiy ~ day.o = 0. Because of (28.3) the first set of equations (39.2) which are not satisfied identically are equivalent to (39.3) udm ~ dig = Rin. From this result and (28.5) we have the equation of Gauss* Ring _ dudn — dis 9 Guges — gir Since des is a symmetric tensor, the second set of equations (39.2) consists of the two equations (39.5) daa, — dag.a = 0 (a # 8); they are known as the equations of Codazzi because they are equivalent to equations derived by him.} Equivalent equations had been derived earlier by Mainardi.t When the expressions for the covariant deriva- tives in (89.5) are written out, these equations become Odon _ Adog _ 3 3 (89.6) a ae {3} + dhs {3} =0 (a ¥ 8). ‘When equations (38.19) are differentiated covariantly with respect tow’, on noting that the covariant derivative of gag is zero by theorem [28.3] and making use of (38.18), one obtains (39.4) K= i é _— ” a" i t Xa =—-9 (dona et avd) Since X" are scalars, the condition of integrability of (38.19) is X’as = ‘4a. This condition is satisfied because of the second set of (39.2), fthat is, the Codazzi equations. Having shown that the first and second fundamental tensors of a * 1827, 1, p. 234. t 1869, 3, p. 275. $1856, 1, p. 395. 220 SURFACES IN SPACE [Cu. IV surface satisfy the equations (39.3) and (39.5), we inquire whether two tensors satisfying these equations are the fundamental tensors of a surface. ‘This means that there must. exist solutions 2* and X' of equa- tions (38.18) and (38.19) which satisfy the equations (38.2), (38.6) and (88.8). In order to answer this question we put ari (39.7) me in terms of which equations (38.18) and (38.19) are r Y a OR 8 pi dagX', oh + de Opa _ oi _ (398) ip = PH Gua = den BS, and cquations (38.2), (38.6) and (88.8) are (39.9) Lipipp= os, UX = 1, EX p= 0. Equations (39.8) and (39.9) constitute a mixed system of the type con- sidered in §23. ‘Ihe conditions of integrability of equations (39.8) are satisfied because of (39.3) and (39.5), this being in fact the manner in which equations (39.3) and (39.5) were obtained. Equations (39.9) are the sct Ey in the terminology of $28. When thcy are differentiated it is found that the resulting equations are satisfied in consequence of (39.8), and consequently there are no sets Zi, E2, --- , in the terminology of §23, to be satisfied. Equations (39.9) constitute. six conditions upon the nine functions p, and X‘, and thus by theorem [23.2] for a mixed system the solution of the mixed system involves three arbitrary con- stants. When such a solution is given the equations of the surface are given by the quadratures (39.10) re fv dus +0, as follows from (39.7), where b* are three additional constants. From the second and third of (39.9) it follows that X" are the direction cosines of the normal to this surface. We give an interpretation of the six arbitrary constants involved in the present problem. We observe that if x* and X* constitute a solu- tion of equations (38.2), (38.6), (38.8), (38.18), and (38.19), so also do the quantities @ and X* defined by (39.11) @sajai tb, X= alx, where the a's and b’s are constants and the a’s are subject to the six conditions $39] EQUATION OF GAUSS. EQUATIONS OF CODAZZI 221 (39.12) Gj, k = 1, 2, 8). From the form of the first set of equations (39.11) and from (39.12) it follows that two surfaces defined by 2‘ and #' may be obtained from one another by a rotation and a translation, that is, by a motion. Since these equations of motion involve the same number of arbitrary con- stants as the general solution of the present problem, it follows from the above discussion that any tivo surfaces with the same tensors gas and dag are transformable into one another by a motion in space. Hence we have [39.1] Two sets of functions gog and dag which satisfy the equations of Gauss and Codazzi and such thal g > 0 are the components of the first and second fundamental tensors of a surface which is determined to within a motion in space. EXERCISES 1, When a surface is defined by an equation 2* = (e', 2%) (See §25, Ex. 2), one finds that 2 ~ = mt Tuy Te XX, = 1 dn, du, de = Fa, Vit pit et Vi+pi+pi where an ar > Pam jue grays (a= 3,2). 2. For aright conoid with the equations (see §10, Ex. 4) Baweosuw, x= wsinu 2 = o(u), one finds that sin uty’, —cos uty’, ul Xt, X3— es oeere Ve diy diay dao = Verba? the locus of the normals to the surfaco along a generator is a hyperbolic Paraboloid. 3. For a coutral quadrie with the equations in §10 Ex, 3 one finds that / axle; — was — ¥#) * “fe aaa)’ 222 SURFACES IN SPACE Eon. ay where i, j, & take the values 1, 2, 3 cyclically, and 1 faasas ut fea fs wat Gara ew) OMe dun 0. 4, Fora paraboloid with the equations of §30, Ex. 7 one finds that Xi, Xa, Xs ail? Va =a 1/03 (a, = a3)(u® = ‘ doa <1 4% “arth aut) (@~ 8 d= Oy a Va, a Va — a 0 Faw), -Vaa, where A= Viator = ashe = ail fasts — asda = aa. 40. NORMAL CURVATURE OF A SURFACE. PRINCIPAL RADII OF NORMAL CURVATURE Consider any curve C upon a surface, Its tangent vector Ea at any point P is perpendicular to the principal normal and binormal to the curve and the normal vector X* to the surface. These three vec- tors are shown in Fig. 13 lying in the plane of the paper, and it is under- stood that the tangent vector to the curve at P is normal to the plane of the paper and is directed toward the reader; the line PR represents the line of intersection of the plane of the paper and the tangent, plane to the surface at P. Denote by @ (see Fig. 13) the angle made by the normal vector X‘ with the principal normal, whose components are denoted by °. Hence on making use of (4.7) we have (40.1) cox = 5 X's! = p DX wie where p is the radius of curvature of Cat P. In consequence of (38.11), (88.8), (38.13), and (24.6) we obtain from this equation 0S @ _ degdu" dub (40.2) Pp Gapdu* dub Thus °° ® is equal to the ratio of the values of the second and first fundamental quadratic forms of the surface for the direction of the curve at P, that is, for differentials du* giving the direction of the tangent to the curve at P. Since the right-hand member of (40.2) is completely determined by $40) NORMAL CURVATURE OF A SURFACE 223 the values of the differentials du", it follows that the quantity “ is the same for all curves through P having a common tangent at P, denoted by PT. Since by hypothesis p is a positive quantity, it follows that according as cos & is positive or negative for one of these curves it is the same for all of the curves. For a curve whose direction at P satisfies the equation (403) das du" du? = 0, cither 1/p = 0 at P or the osculating plane at P is the tangent plane to the surface at P as follows from (40.2). i x Fra. 13 Consider, in particular the plane curve in which the surface is cut by the plane determined by the normal to the surface at P and the tangent PT called the normal section of the surface for the direction PT. In this case & is 0° or 180° according as the principal normal to C and the normal X* have the same or opposite directions. If we denote by pn - the radius of curvature of this plane curve at P, assumed to be positive, we have (40.4) cose Pp Pn where ¢ is +1 or —1, according as the principal normal to C and the normal X" have the same or opposite directions, G 224 SURFACES IN SPACE (Cu. IV Hence we have the theorem of Meusnier:* [40.1] The center of curvature al a point P for a curve on a surface is the projection upon ils oseulating plane of the center of curvature of that normal section of the surface which is tangent to the curve at P. If now we define a quantity R for a direction du* at a point u* by 5 1 _ cos _ dagdu*du® (08) Roe Gadde it follows from (40.4) that the absolute value of R is the radius of curvature of the normal section of the surface for the given direction, and that. R is positive or negative according as the normal of the plane section at «* and the normal X‘ to the surface have the same or opposite directions. The quantity & defined by (40.5) is called the radius of normal curvature of the surface at a point «* for the direction du*. A normal section of a plane being a straight line, 1/R = 0 in this case. A normal section of a sphere being a great, cirele R is equal to the radius of the sphere. In both these cases the value of R does not vary with the direction du*. Conversely, if R is not to vary with du’, it is necessary and sufficient that das = 0, or that deg be proportional t0 gas; in the former case the surface is a plane, and in the latter a sphere (see §38, Exs. 1 and 2), In order to find the directions at a point on a surface for which 1/R. is a maximum or minimum, we note that equation (40.5) is of the form (26.26). Consequently these directions are given by the values of du satisfying the equations (40.6) (ij jen) det = as follows from (26.17) and the discussion following (26.26), and the corresponding valucs of 1/R are the roots of the determinant equation Ll | dag — = (40.7) [das — pao} = 0. Thus the directions sought are the principal directions for the tensor dag i these directions are called the principal directions of normal curvature. They are given by the equation | dhe du" deqadu® | (40.8) |gradu gaa dtu" * 1776, 1, p. 489. $40] PRINCIPAL RADIL OF NORMAL CURVATURE 225, as follows from (26.23), and they are real by theorem [26.4]. When theorem [26.5] is applied to this case, we have [40.2] At each point of a surface, other than a plane or a sphere, equation (40.8) determines the two directions for which the normal curvature is a maximum or minimum, these directions being perpendicular. The values of R for these directions, called the principal radii of normal curvature, are the roots of the equation (40.7), that is (40.9) (dude — diz)R* — (dugex + dagu — 2dgu)R + g = 0. If these roots are denoted by p: and pa, it follows from (40.9) and (39.4) that. 2 (40.10) AL dada = dh _ Pipe a dy — gud; net dads _ gat a From (40.10) we have the following theorem of Gauss:* (on) K,=t+le ai pe [40.3] The curvature K at a point of a surface is equat lo the reciprocal of the product of the principal radii of normal curvature at the point. 1 pe of the surface at a point. Equation (40.11) gives its expression in terms of the two fundamental tensors of the surface. Ordinarily the principal radii at 2 point are unequal, but there may be points at which they are equal. Such a point is called an umbilical point, From (40.10) and (40.11) it follows that [40.4] A necessary and sufficient condition that a point u® be an umbilical point is that at the point ‘The quantity + + 1, denoted hy Kn, is called the mean curvature a (40.12) (og cas)? ). 5 ade = From geometric considerations it follows that every point on a sphere is an umbilical point. For this reason an umbilical point is sometimes called a spherical point, The centers of curvature of the principal normal sections of a surface * 1827, 1, p. 231. 226 SURFACES IN SPACE fCu. FY at a point, that is, the normal sections in the principal directions, are given by (40.13) ai =a! + p.X%, wht pax! These centers are called the principal centers of curvature for the point. When the Gaussian curvature K is positive at a point P of a surface, the principal radii have the same sign. Since in this case dudz — dz > 0, as follows from (40.10), the second quadratic form deg du du’ has the same sign at a point for all values of du’, there being no real values of du* for which the quadratic form is equal to zero. Consequently it follows from (38.12) that points in the neighbor- hood of P lie entirely on one side of the tangent plane. A surface at all points of which K is positive is called a surface of positive curvature. When K < 0 ata point P, the principal radii differ in sign. In this case equation (40.3) defines two real directions at P for each of which R is infinite. From theorem [26.4] we have that the bisectors of the angles between these directions are the principal directions at P. Since the principal radii differ in sign and since the second fundamental form for a point can change sign only at the directions given by (40.3), it follows that in the neighborhood of P part of the surface lies on one side of the tangent plane and part on the other, these parts being separated by the two branches of the curve in which the surface is cut by the tangent plane at P, the tangents at P to these branches having the directions given by (40.3). A surface ut ull points of which K <0 is called a surface of negative curvature. From the considerations of this and the preceding paragraph it follows from geometric considera- tions that ellipsoids, hyperholoids of two sheets and elliptic paraboloids are surfaces of positive curvature, and that hyperboloids of one sheet and hyperbolic paraboloids are surfaces of negative curvature. When K = 0 at a point, the directions given by (40.3) coincide, since in this case dude — diz = 0, and for this direction R is infinite. In this case the coefficient of 45° in (38.12) is a perfect square and the question of whether the distance phas always the same sign depends on the terms of higher order. Consider, for example, a cylinder whose cross- section has a simple inflection ata point P. K = 0 for all points of the surface, and the surface lies on both sides of the tangent plane along the generator through P. EXERCISES 1. If a segment equal to twice the radius of normal curvature for a direction PY ata point # ona surface is laid off from P on the normal to the surface in the #0) PRINCIPAL RADII OF NORMAL CURVATURE, 227 appropriate direction, and a sphere is described with the segment as diameter, the circle in which the sphere is met by the osculating plane of any curve through P and with the direction P7 is the circle of eurvature of the eurve, 2. When the coordinate curves of a surface are such that giz = die = 0, 1/R is a maximum or minimum in the direction of the coordinate curves, and lid) 1 _ de om Gn? in this coordinate system the points, if any, for which digex — dag = 0 are umbilical points, 3. Show analytically that each point of a sphere is an umbilical point. 4. ‘The tangent to the merdians and parallels of a surface of revolution are the principal directions at each point of the surface; the prineipal radii of a surface with the equations of §10, Fx. 2 are given by 7 1 a (ee)? pp utd gt” (gee §24, Ex. 3 and §38, Ex. 4); p2 is the segment of the normal between the point of the surface and the axis of the surface, 5. Let Po denote the center of normal curvature in the direction bisecting the principal dircetions at a point P of a surface, and P, and P; the centers of normal curvature in dircetions equally inclined to this biscetor; then P, P: , Po, Pstorm a harmonic range. 6. Let Ri, Ra, +++, Re denote the radii of normal curvature for m (> 2) diree- tions such that the angle of two adjoining directions is 2x/m; then i 1) _1/a,1 aati’ i) i+2). 7. From §30, Ex. 6 and §39, Ex. 3 it can be shown that the points ' oak (aa a3) | o-0 a= a of ahyperboloid of two sheets; and that there are no umbilical points on @ hyper- boloid of one sheet. 8. The principal radii of normal curvature of the surface of revolution of a Parabola about its directrix arc in constant ratio 9. The surface of revolution of a circle of radius a about a line in the plane of the circle at the distance &(> a) from the center of the circle is called a lorus; the Gaussian curvature of the surface at a point P is positive, zero, or negative ac- cording as the distance of P from the axis is greater than, equal to, or less than b, 228 SURFACES IN SPACE. [Cu. I 41. LINES OF CURVATURE OF A SURFACE The integral curves of the differential equation (40.8), that is, the equation (ugie — dixgus) du + ugar — deegus) dut dub + (digsx — dagne) du* = 0, are called the lines of curvature of a surface. We have seen that they form a real orthogonal net, and that the tangents to the two curves of the net at a point are the principal directions of normal curvature at the point. We now give another characteristic property of the lines of curvature. The normals to a surface at points x‘ of a curve C form a ruled surface, whose equations in parametric form are (41.2) B= at + 1X, (41.1) the quantities x, X* being expressed in terms of the are s of C, and ¢ being a second parameter, namely the distance of the point 2° from the point 2°. In order that the ruled surface be the tangent surface of a curve I’, and thus a developable surface, the normals to the given surface being the tangents to I’, the equation = f(s) of the curve I’ must be such that & are proportional to X*, From (41.2) it follows that this condition is (41.3) - ef it iF x tt tx! = hx’ Multiplying these sauation by X‘ and summing with respect to #, we have in consequence of (38.6) and (38.8) that dt 4 at (41.4) as and consequently (41.3) reduce to (41.5) a = + 1X Lo, ds In order that: the normals to the surface along a curve shall form & cone, it is necessary that the equation f = f(s) be such chat the point @* defined by (41.2) be the vertex of the cone, in which case ae =0. In this case we have (41.3) with A = 0, and from (41.4) ¢ = const., and again we have equations (41.5). §43] LINES OF CURVATURE OF A SURFACE 229 When equations (41.8) written in the form (41.5) (& +t x) a 4 aur) ds : are multiplied by x and summed with respect to 4, we obtain in conse- quence of (38.16) aut (41.6) (Gap = dos) “Ge Since these equations are equivalent to (40.6) with 1 = R, and the equations (40.8) are a consequence of (40.6), we have [41.1] A necessary and sufficient condition that the normals to a surface along a curve form a developable surface is that the curve be a line of curva~ ture; when the developable surface is the tangent surface of @ curve, equa~ tions of this curve are (41.2) where t is the principal radius of normal curvature of the line of curvature concerned; the developable surface is a cone, if and only if the corresponding principal radius is a constant along the curve. The proof of th theorem ix not. complete until one considers the possibility of the normals forming a cylinder, that is, a =0. In “this case we have, using (38.16), > ax’ dX* _ sp ax AX" du® Tau" ds“ au* au® ds du 0 (a= 1,2 og Ge = 0 (a= 1,2). A curve for which these two equations hold is evidently an integral curve of equation (40.8), and consequently is a line of curvature. In order that the above equations for a = 1, 2 be consistent we must have dndex — diz = O along the curve. Hence from (39.4) and theorem [28.1] either the surface is a developable surface and the curve is a generator, or K = Oat least at all points of the curve. Since the condition ax = 0 holds along each generator of a develop- able surface, we have [41.2] The lines of curvature of a developable surface are its generators and their orthogonal trajectories. In order that the coordinate curves be lines of curvature, it is neces- sary and sufficient that (41.7) dy gx — da gu =0, de gx — da gx = 0, 230 SURFACES IN SPACE {Cn. IV as follows from equation (41.1). Consequently gx = dz = 0, unless the two fundamental forms are proportional, which is the case only when the surface is a plane or sphere (see §88, Exs. 1,2), Hence we have [41.3] A necessary and sufficient condition that the lines of curvature be coordinate, that ts, the coordinate curves, on a surface other than a plane or sphere is that (41.8) diy = ga = 0; any orthogonal net on a plane or sphere satisfies these conditions. When the lines of curvature are coordinate, equations (38.19) re- duce to (41.9) (a = 1,2;anot summed), in consequence of (41.8) and (24.18), the principal radii being given by (41.10) Lid 1 de Pr gu Po G2 When the lines of curvature are coordinate, the Codazzi equations (89.6) reduce to (see §28, Ex. 1) adn _ 5 (@ +t) ag at 2\ ou au a1ay m an Oe Adee _ 1 (du, das Agee Adz _ 1 (du, dm\ dom _ 9, au o(# + i) oul In consequence of (41.10) these equations are expressible in the form afi 1/1 _ 1\alog gn _ m3) +3(3 A) oe = 8 (2 411 _ 1) a logge _ aut \p/ " 2\e a) at EXERCISES (41.12) 1. When a line of eurvature is a plane curve and the normals to the surface along the curve lic in the plane of the curve, the developable surface of the normals consists of the tangent lines to the plane cvolute of the curve. 2. The lines of curvature of the tangent surface of a curve are the tangents t the curve and itsinvolutes. 3. ‘The meridians and parallels of a surface of revolution are the lines of eurva- ture of the surface (sco §24, Ex. 3 and §38, Ex. 4). $42] CONJUGATE DIRECTIONS AND CONJUGATE NETS 231 4. The coordinate curves on a central quadrie with the parametric equations in §10, Ex. 3 are the lines of curvature of the quadric (see §30, Ex. 6 and §39, Ex. 3). 5. The coordinate curves on a paraboloid with the parametrie equations in §30, Ex. Zare the lines of curvature of the paraboloid (sce §39, Ex. 4). 6. The lines of eurvature of a spiral surface as defined in §26, 1x. 8 ean be found by quadratures. 42. CONJUGATE DIRECTIONS AND CONJUGATE NETS. ISOMETRIC-CONJUGATE NETS Consider a curve C on a surface and the tangent planes to the surface at points of the eurve, The envelope of this one parameter family of tangent planes is a developable surface (§12). ‘Through each point of the curve there passes a generator of the developable surface. The generator of the surface and the tangent to the curve at the point are said to have conjugale directions. In order to find an analytic expression for conjugate directions we consider the equation of the tangent plane (42.1) rxa@ — «') = 0, where 2° are current coordinates, and X‘ and 2° are functions of the are along the curve. On differentiating (42.1) with respect to s we have in consequence of (38.8) (42.2) This equation and (42.1) are equations of the generator of the de- velopable through the point x’. The quantities ° — 2‘ are direction numbers of the generator, and they are proportional to = bu‘, where i éu* are differentials in the direction conjugate to the tangent to the curve at the point (sec (24.4)). In consequence of this observation and equations (38.16) we have from (42.2) (42.3) degdu® bu? = 0, “ which gives the relation between the differentials du” for C and du" for its conjugate direction. From the form of (42.3) it is seen that con- jugacy of directions is reciprocal, that is, if a direction is conjugate to a given direction, the latter is conjugate to the former. Comparing equations (42.3) and (26.7), we observe that the second fundamental tensor bears to conjugacy the same relation which the first fundamental tensor bears to orthogonality. 232 SURFACES IN SPACE (Cu. IV The direc (42.4) g(u', wu) = const. tions conjugate to each of the curves at each point of the curve are given by (see (26.2)) de _ 4, 8) 5 90 _ gy 8°) yt = (42.5) (a jan de a) bul + (a 5am de a) ao = 0. Looking upon this equation as a differential equation in du“, we have that its integral curves and the curves ¢ = const. form a net, such that a curve of each family passes through a point of the surface and at the point their directions are conjugate. Such a net is called a conjugate net. From (42.3) it follows that the curves conjugate to the curves u? = const. are the integral curves of the equation (42.6) dydul + ddu” = 0. In order that they be the curves u' = const. we must have dz = 0, and conversely. Hence we have [$2.1] A necessary and sufficient condition that the coordinate curves form a conjugate net is that dia be zero. From this theorem and theorem [41.3] we have 142.2} The lines of curvature of a surface form a conjugate net, and the only conjugate net which is an orthogonal net. From equations (38.18) it follows that when the coordinate curves form a conjugate net each of the coordinates 2° is a solution of an equa- tion of the form a0 1 80 2 OO (42.7) oe ee where in general a’ and a’ are functions of u*. Conversely, if x for ¢ = 1, 2, 3 are linearly independent solutions of an cquation of this type, then et aa Out du! aut du® — dul du?) aa! aa? ax? | 0 ut aut out , az! a? aa au aut ae $42) CONJUGATE DIRECTIONS AND CONJUGATE NETS 233 from which and from (38.5) and (38.13) it follows that dz = 0. Hence we have (42.31 If f(u', uw) for i = 1, 2, 8 are three linearly independent solutions of an equation of the form (42.7), the coordinate curves form a conjugate net on the surface with the equations 2° = fi(u', uv’) When the lines of curvature are coordinate, we have, since gz = 0, a at 1 But aut In this case not only are 2* solutions of an equation of the form (42.7) but also the sum of their squares, that is >, xr’, is a solution as one verifies by substitution, and conversely. Hence we have the theorem of Darboux:* [42.4] A necessary and sufficient condition that the coordinale curves be the lines of curvature on the surface defined by three linearly independent solutions of an equation of the form (42.7) is that the sum of the squares of these solutions is also a solution Darbouxt has applied this result to the proof of the following theorem: [42.5] When a surface 8 is transformed into a surfare Sy hy an inversion, the lines of curvature of S are transformed into the lines of curvature of Si, . An inversion, or a Lransformation by reciprocal radii, is defined by {eae Ge @) ES SNe ale! = ah)’ where c is a constant, and the point z@ is the center of the transforma- : tion. From (42.8) it follows that Eile — she — 2) @ — ad(e — at) = (42.9) 2p) — gi ? jt ei = 7) Tt SS Dai Say! DiGi = zi = 2) 7 Which shows the reciprocal character of the transformation, Suppose now that 2’ are solutions of an equation (42.7), then 2} given by (42.8) are functions of u* which are solutions of the equation in 6, resulting from the substitution (42.10) a : = (xi — 28)(et — 28) * 1887, 1, p. 136, £1887, 1, p. 208. 234 SURFACES IN SPACE [Cu. IV in equation (42.7), as follows from the second of equations (42.9), x3 being constants. If ;2%x' isa solution of (42.7), then as follows from the first of (42.9) and (42.10) c‘ is a solution of the equation in 6, which consequently is of the form (42.7). Since (42.7) admits unity for a solution, it follows from (42.10) that the equation in 6; admits the solution 2, xiz}, and theorem [42.5] follows in consequence of theorem [42.4]. For a surface of positive or negative curvature, dudas — diz , denoted by d, is not equal to zero. Consequently quantities d® are uniquely defined by (42.11) a dg, = 85; in fact a _ de 2 de ds (42.12) a2, a, =%. If then we define quantities Ay and A,(y, ¥) by (42.13) Rig =d¥ aps, By, 0) = d%oabs, these are scalars for a transformation of coordinates u*, since it follows from (42.11) as in §14 that d® is a symmetric contravariant tensor, From (42.13) we have uw! =a" =%, kw ds (42.14) Bul, 2) = d? = and consequently in terms of any net ¢ = const., ¥ = const., thesecond fundamental form may be written Aivdg’ — 2Ai(y,W dedy + Aiedy* Aivdy — Be” We define also the quantity As by an equation analogous to (29.16), namely (42.15) a = fod Bee 7 ae 2 (Ved dg), where ¢ is +1 or —1 according as K is positive or negative for the portion of the surface under consideration. If then g is any real solu- $42) ISOMETRIC-CONJUGATE NETS: 235 tion of the equation Ax = 0, a real function ¥ is obtained from (see (30.5) (42.18) Ved tpn = Hs Ved dp = Wa by a quadrature. Proceeding as in §30, we find from these equations that ea = ev/ed dy, va = —e ved dys, and Ay= chy, Aly, ¥) = 0, so that the second fundamental form (42.15) becomes as (ag + dy’). In ease the coordinates are such that the second fundamental form is (42.17) A(du® + ¢ du”), the coordinate curves are said to form an isometric-conjugate nel. When for a conjugate net on a surface dy _ filu') da felt)? new coordinates can be chosen in terms of which the second fundamental form assumes the form (42.17). The coordinates are called isometric- conjugate when the second fundamental form is of the particular form (42.17). Proceeding as in §30, we have analogously to theorem [30.3] [42.6] A necessary and sufficient condition that a family of curves ¢ = const. and their conjugate trajectories form an isometric-conjugate net is that Bop = 0 or that the ratio of Bap and Ay is a function of e. (42.18) EXERCISES 1. The evordinate curves on a surface defined by @ ai = fi(ut) + faut) form a voujugate net. Buch of the curves ut = const, may be oblaiued from Uke curve 2 = f{(u!) by a suitable translation, and the curves u! = const. by asuitable translation of the curve 2! = f§(u!); and the tangents to one coordinate family at Points of a curve of the other family are parallel. A surface with equations of the 236 SURFACES IN SPACE [Ca IV form (i) is called a surface of translation, and the curves 2° are called the generating curves 2. The coordinate curves form a conjugate net on the surface ‘Fi(uY) and xi = si(ut) Fw) + glee) 3. The meridians and parallels on a surface of revolution form an isometrie- conjugate net (see §88, Fx. 4.) 4, The lines of curvature on a central quadric and on a paraboloid form an iso- metric-conjugate net (see §39, Exs. 3 and 4). 5. The coordinate curves form an isometrie-conjugate net on the surface filutler(u), at z Si(wer(u?}, 2 = ga(ut). 6. When the coordinate curves of a surface form an isometric-conjugate net en? 1 a and the coordinates are isomotvie-conjugate, then K=—, Tfone puts > = —=, onjugs 7 oes 7, thon du = V9, day = M9, ana the equations of Codazzi (39.6) are ° ? a, vo {2} (=| jy OB asl ~ Yash (a, 8 =1,2;8 4a) 7. When the coordinate curves on a surface are isometric-eonjugate and the coordinates are isometric-conjugate, equations (42.16) reduce to eb. Whene — I -aeolution of these equations ie + ip — (wt + iw’), and whene~ —1, es fila +) $ faut — w), v= fall + ut) — fa(ut — wd) 8. When a plane is subjected to un inversion it is tranformed into a plane or « sphere according as the given plane passes through the ceuter of the inversion or not ; when a line is subjected to an inversion it is transformed into a line or acirele cording as the given line passes through the center of the inversion or not. 9. From (42.8) one obtains ot So dada 7 Dae dxj (Week ep r and consequently a transformation by reciprocal radii is conformal. 10. When a cone of revolution is subjected to a transformation by reciprocal radii whose centor is not on the cone, the lines of curvature on the new surface are circles, and this surface is the envelope of a family of spheres having puint in common, the spheres being the transforms of the tangent planes to the cone. 11. A necessary and sufficient condition that a conjugate net on a surface be ‘Vehebychef net (see §29, Ex. 12) is that the surface be a surface of translation. §43] ASYMPTOTIC DIRECTIONS AND ASYMPTOTIC LINES 237 43. ASYMPTOTIC DIRECTIONS AND ASYMPTOTIC LINES. MEAN CONJUGATE DIRECTIONS. THE DUPIN INDICATRIX A direction at a point in a surface which is self-conjugate is called asymptotic. Frovi equation (42.3) it follows that the asymptotic directions at each point of a surface are given by (43.1) dos du* du® = 0. ‘The integral curves of this differential equation of the first order and called the asymptotic lines of the surface. Accordingly is a curve whose tangent at any point is an asymp- ction at the point. From (43.1) it follows that the asymptotic directions upon a surface are conjugate imaginary, real and distinct, or real and coincident according as did — diz is positive, negative, or equal to zero, From this result and (39.4) we have [43.1] The asymptotic lines are conjugate imaginary, real and distinct, or real and coincident on a surface, or any portion of a surface, for which the Gaussian curvature is positive, negative, or equal to zero. From theorem [26.4] applied to equation (43.1) and equation (40.8) we have [43.2] On a surface, or portion of a surface, for which the Gaussian curva ture is not equal to zero, the lines of curvature bisect the angles formed by the asymplotic lines. When a straight line lies entirely in a surface, the tangent plane to the surface at each point of the line contains the li From the definition of asymptotic lines it follows that (see §38, Ex. 3): [43.3] When a straight line lies entirely in a surface, it is an asymptotic line of the surface. In §41 we saw that the generators of a developable surface are also lines of curvature. From (40.1) and (40.2) we have 148.4] The oseulating planes of a curved asymptotic line are tangent to the surface along the curve. From equation (43.1) we have [43.5] A necessary and sufficient condition that the asymptotic lines be coordinate is that (43.2) dy = de = 0. 238 SURFACES IN SPACE [Cu. 1 From this result and (40.11) it -_ res the asymptotic lines form an orthogonal net, if and only if Kn = — +> “= 0. A surface for which the mean curvature Kn is equal to toro ts called & minimal surface; the property of such a surface which justifies the term minimal is established in §50. From (40.11) it follows also that when the minimal curves (§2) on such a surface are coordinate, that is, gun = gz = 0, then diz = 0. Hence we have [43.6] On a minimal surface the asymptotic lines form an orthogonal net and the minimal lines a conjugate net; each of these properties characterizes a minimal surface. From theorem [43.5] and equations (38.18) it follows that, asymptotic lines are coordinate, the quantities 2° as functions of and u’ are solutions of two equations of the form Ca 36 vo 36 06 <3 = 4 ay, ao, = dn tan <, aw "5 iat 8 aut jut = aga FO as vhen the 1 (43.3) where in general the a’s are functions of u*. By an argument similar to that which led to theorem [42.3] we have [43.7] When too equations of the form (43.3) admit three linearly inde- pendent solutions f'(u', u’) for i = 1, 2, 8, the coordinate curves are the asymptotic lines on the surface with the equations x = fi(u’, uw’). When equations (43.2) are satisfied, and we put 1/p = dye/+/q, then the Gaussian curvature is given by (43.4) and the Codazzi equations (39.6) reduce to dlog Vp _ [2\ a log Vp (43.5) a Bp ae = {bh by means of equations (28.2), namely (43.6) a toe va = {2 }. The conditions of integrability of (43.5) are (43.7) 24h} = a3}. $43] ASYMPTOTIC LINES 239 from which and (43.6) we have also af2\_afi (43:8) aul {on} = adh}: Conversely, if three real functions gag such that g > 0 satisfy the condition (43.7) and these functions and the function » determined to within sign by the quadrature (43.5) satisfy the equation (43.9) Bae 9 then gas, du = de = 0, diz = +/Q/p satisfy the equations of Gauss and Codazzi, and in accordance with theorem [39.1] we have [43.8] If @ real tensor gag for which g > 0 satisfies the equation (43.7), and ges and a function p determined lo within sign by (43.5) satisfy the equation (43.9), then gag is the first fundamental tensor of a surface of negative curvature upon which the asymptotic lines are coordinate; this surface is determined in space lo within a motion and a symmetry with respect to a point, The possibility of symmetry with respect to a point, (see §38, Ex. 6) is due to the fact that the sign of p is not determined by (43.5). When the lines of curvature of a surface are coordinate, and the angles which a pair of conjugate directions at a point make with the tangent to the curve u* — const. at the point are denoted by 6 and 6’, that is (sce (25.7), Ge du tang /qa 8 (43.10) tana = 4/2, fe the equation (42.3) with d = 0 may be written in the form (43.11) tan @ tang! = —, pi where . (43.12) tis i.2 pL gn Pz Grae From this equation we obtain poco 6 + p: tand PL . (43.13) tan (6 — 6) = If we equate to zero the derivative with respect to tan 6 of the right- hand member of this equation, we find that the values of tan 6 for 240 SURFACES IN SPACE [Cu. IV which this derivative is zero are -+-~/p2/p, , and that for tan @ = ~/p:/piy when | pi | > | p2| then tan (@ — 6) is a minimum. This direction is real only at a point for which K is positive, and from (43.11) we have that in this case tan 6’ = — tan @, that is, the two. conjugate directions are equally inclined to the lines of curvature. Conversely if the latter condition holds, it follows from (43.11) that tan? @ = ps/o. Hence we have [43.9] A necessary and sufficient condition that there exist a pair of real conjugate directions at a point P in a surface which make equal angles with the tangents to the lines of curvature at P is that the Gaussian curvature be positive at P; the angle 8 of one of these directions is given by tan 0 = Valo. ; an angle between these directions is the minimum of the angles between conjugate directions at P. From (43.10) it follows that these conjugate directions are given by a y/tam, Eft de gap’ Bu gz pr" For both of these directions the radius of normal curvature is given by Rade tnd _ tn uu? + dn du®™ 2 in consequence of (43.12). ‘Thus the radius of normal curvature is the mean of the principal radii of curvature. Accordingly we call these directions the mean-conjugate directions, en the lines of curvature are coordinate and @ denotes the angle which a direction ata point. P makes with the enrve a? = const. at P, the equation (40.5) may be written in the form (43.14) nt 2 (43.15) h = od sin’ Ps pe : in consequence of (25.6) and (43.12). This equation is called the equa- tion of Euler. From it we have that the angle @ which an asymptotic direction makes with the curve u® = const. at the point is given by (43.16) tan?9 = —%, a This result follows also from (43.11) and the definition of asymptotic directions as self-conjugate. We have remarked in §40 that at a point of a surface at which K > 0 the principal radii have the same sign, and that & for any other normal $43] THE DUPIN INDICATRIX 241 section has the same sign. If in the tangent: plane at such a point P, we take P for the origin of a cartesian coordinate system, and the tangents to the directions of the principal radii and m for the 2- and yraxes respectively, and lay off on each line through /’ and in both direc- tions from P line segments of length / & |, where J is the correspond- ing radius of normal curvature, an equation of the locus of the end points of these segments is found from (48.15) to be 2 2 (43.17) 2+ fol © [eel ‘Thus the locus is an ellipse, whose principal axes are the principal direc- tions at P, and from (43.11) it follows that conjugate diameters of the ellipse* are conjugate directions on the surface. “ag Cae | Fic. 14 When K < 0, the principal radii , and p: differ in sign, and certain values of R are positive and others are negative. In this case the locus of the end points of the segments y/| |, as shown in Fig. 15, consists of the two conjugate hyperbolas with the equations 2 a 3 (43.18) task e+oe ot 2 The asymptotes of these hyperbolas are given by ; + = 0, which 1 pe equation is equivalent to equation (43.16), that is, the asymplotes are the asymptotic directions at the point, which accounts for the name of these directions, In this case also equation (43.11) gives conjugate diameters of the hyperbolast as well as conjugate directions on the surface. According as K is positive or negative at a point the conic defined by equation (43.17) or (43.18) is called the Dupin indicatriz at the point. *C.G., p. 192. +C.G., p. 192. 242 SURFACES IN SPACE [Cu. IV Hence a point of a surface at which K is positive or negative is some- times called elliptic or hyperbolic. EXERCISES 1. The asymptotic lines on a surface of revolution can be found by quadratures (see §38, Ex. 4). 2. When the coordinate curves on a surface form an isometrie-conjugate net, the asymptotic lines can be found by quadratures. 3, The coordinate curves on a surface of translation with the equations alm flu), a = p(w), = falut) + ei(ut) are plane curves, the planes of one family being perpendicular to those of the other family, and the curves form an isometric-eonjugate net. 4, ‘The catenoid (see §24, Ex. 4) and the skew helicoid (see §24, Ex. 7) are ap- plicable minimal surfaces 5, From (43.8) and theorem [43.7] it follows that a necessary condition that two equations of the form (43.3) admit three linearly independent solutions is that Ban _ Pax ut ~ but 6. For asurface of constant negative curvature ~ 1 /a*veferred to its asymptotic Jines the coordinates can be chosen so that gu = gn = a, giz = a? cose, wherew is a solution of the equation Be aulaut = sine. 7, The integral curves of an equation aagdutdu? = 0 form a conjugate net, if and only if auidas + aoadis — 2aredie = 0, 8. ‘The surface with the equations actu tu), at = belut ut), at = Bouter, where a, b, care constants, is a hyperbolic paraboloid referred to its rulings, that is, its asymptotic lines; find the equations (43.8) in this ease 9. The surface with the equations oot out wy lave tw a’ bc Tule where a, }, care constants, is a hyperboloid of one shect referred to its rulings, that is, its asymptotic lines; find the equations (43.3) in this ease. 10. A necessary and sufficient condition that the asymptotic lines on a surface form a Tehebychef net (see §29, Ex, 12) is that the surface have constant Gaussian curvature. 44] GEODESIC CURVATURE, 213 11, For a minimal surface referred to its lines of curvature one has from equa tions (41.12) that gn sai, gar = aa, where Uy and Uz are functions of ut and u? respectively; hence the lings of eurva- ture on a minimal surface form an isometrie net (see (30.11). 12. Given two surfaces S and S with the respective seeond fundamental forms dog dut du and dag due du; a necessary and sulficient condition that the asymptotic lines on S correspond to a conjugate system on 8, corresponding points being those with the same values of won the two surfaces, is that dy dex + dn du — dhe die = 05 when this condition is satisfied the asymptotic lines on § correspond to a con- jugate system on 13. On the two surfaces of Ex. 12 the integral curves of the equation dia du radu | da du ® dag du form & eonjug: net on each surface (see theorem [26.4)). 14, From (43.14), 40.10) and (40.11) it follows that the radius of normal curvature in cither mean-conjugate direction is given by 199" dog 2 Finda — di when this expression for R is substituted in (10.5) the integral curves of the re~ sulting equation are the mean-conjugate curves of the surface. 15, When Ue mean-conjugate curves of a surface of positive curvature are coordinate dy _ de —=- dis = 0. mae 44. GEODESIC CURVATURE AND GEODESIC TORSION OF A CURVE Consider any curve C upon a surface and the tangent plane to the surface at a point P of C, Project orthogonally upon this tangent plane the portion of the curve in the neighborhood of P and let C’ denote the resulting plane curve, The curve C’ is a normal section of the projecting cylinder and C is a curve upon this cylinder, and con- sequently theorem [40.1] may be applied to the cylinder to determine the relation between the radius of curvature p of C andr of ("This relation is (44.1) 244 SURFACES IN SPACE (Cu. IV where y is the angle which the principal normal to C makes with the tangent plane to the surface at P, that is, with the normal to C’ in this plane. From this result it follows that the normal to the osculating plane of C at its center of curvature for the point P mects the tangent plane to the surface at P in the center of curvature of C’ for the point, P, and by theorem [40.1] this same normal meets the normal plane section of the surface at P and tangential to C in the center of eurva ture of the normal section. In Fig. 13 C, is the latter point, and C, and C, the centers of curvature of C and C’ respectively. In order to find the coordinates of the center of curvature of C’, we make use of the components X* of the unit normal vector to the pro- jecting cylinder. Since this normal is perpendicular to the tangent to C and lies in the normal plane, we have rae Urxveo 7 ds 7 ‘These conditions are satisfied by (44.2) x i, j, & take the values 1, 2, 3 ly, from which we havi x 1. By the choiec (44.2) the positive sense given to the is such that the tangent to C, that is, the vector om the vector X‘, and the vector X‘ have the same mutual orientation as the zi, 2, and 2’-axes*, since dx! dx’ de’ ds ds ds (44.3) yop pol xO x? If we put (44.4) year dr _ dut or! aur ds ds due? face of the ti ee then x” and r are the components in the agent vectors d x de’ x > ctiv and “TE respeetiy ly (see (24.23)). From the discussion following *C.G., p, 162, $44) GEODESIC CURVATURE 245 (38.9) we have from (44.3) that the veetor 1“ makes a right-angle with the vector we. Hence from (25.19) and the first of (44.4) we have (44.5) ‘The angle y in (44.1) is equal or supplementary to the angle between the principal normal to @ and the vector X" according as the latter is in the direction PC, in Fig. 13 or in the opposite direction, Hence from (44.1) we have (44.6) fa 1D Rei rps where B* are direction cosines of the principal normal to C at P, and where ¢ is +1 or —1 according as the angle between the principal normal and the veetor X* is acute or obtuse. In consequence of (4.7), (44.5), and (38.17) we have ee, ty itde 7 0 Ge aur a, due > ast (aa dub du’. art dat "ds “ jue \aubout ds ds” dub dst ae du’ dut éu x0 he a (co ie ee le where [8¢, a]-are Christoffel symbols of the first kind formed with re- spect to the tensor gap. Since g™fyé, 8] = {3} by (20.2), by means of the identities of §25, Ex. 8, the above equation may be written with a change of indices in the form C du*(d? uw? B\ du’ du’ 44.7 © we egg M (SE ae). (44.7) r os (SE + {8 ds ds Comparing this result with equation (34.3) we have that to the geodesic curvature x, of the curve C, defined intrin: §34, and that «, is positive or negative according as the vector the direction PC, in Fig. 1, or in the opposite direction. Accordingly the center of curvature of the projected curve C’ for its point of contact 246 SURFACES IN SPACE [Cz. Ty. with the given curve C is called the center of geodesic curvature of C for this point. In view of the above results we have [44.1] The normal ta the oxculating plane of a enrve Con a surface at the center of curvature for a point P of C meets the tangent plane to the surface at P in the conter of geodesic curvature of C for the point P, and the length of the line segment with end points P and the center of geodesic curvature is equal to the reciprocal of the absolute value of the geodesic curvature of C for the point P. From the above discussion, equation (44.1), and theorem [34.1] we have [44.2] A necessary and sufficient condition that a curve be a geodesic on a surface is that its principal normal at each point of the curve be normal to the surface at the point. When a curve C is defined by an equation g(u', u*) = 0, or as an integral curve of a differential equation M, du“ = 0, its geodesic curva- ture x, can be found directly (see (34.7) and §34, Ex. 7), From (40.2) we have that os? may be found directly. Also from (44.1) we have (44.8) m= S28, > since y — & — Tor . — & according ase is +1 or —1. Henee p and @ can be found directly. We shall show that the torsion r of the curve also may be obtaincd in terms of & and the fundamental tensors of the surface. Consequently the intrinsic equations in space of the curve C can be found directly. From the definition of & in §40 it follows that (44.9) sind = Do X*y', where 7’ are the direction cosines of the binormal of C. Differentiating equation (44.9) with respect to s, and making use of the Frenet formulas (6.1) and equation (40.1), we obtain (44.10) cos a(t - ) =Dy¥ ax = ~dag” ae Dy ds va aur’ the last expression following from (38.19). Since the binormal is in the plane of the vectors X* and X*, we have vy = aX' + 6X’, 44] GEODESIC TORSION 247 where a and b are to be determined. The vector y/ makes with the 3 and @ — m respectively (see Fig. 13). On multiplying the above equations by X* and summing with respect to i, and also by X' and summing with respect to 7, we find that a = sin, b = — cos &. Consequently vectors X‘ and X' the angles @ — in @ X* — cos & X*, ve When these expressions are substituted in the right-hand member of (44.10), the result is reducible, in consequence of (38.8) and (44.5), to eg du® du® 5 ac! ax" 608 B dead Odea Ge Ge De aut aur = EOS Bas where = dee aut du "7 03989 Ts ds Gan) 1 (@i.ga = diagu) du” V9 goad? due + (digs — deagns) du! du + (ages ~ dergiz) dee”), sequently (44.10) reduces to . (da cosa( 2 —rt ") =0, When cos & 0, that is, when the curve is not an asymptotic line, we have (44.12) ratty. We consider now the exceptional ease when the curve is an asymptotic line, and assume that the asymptotic lines are coordinate, that. is, dy = dx = 0. With this choice the direction cosines of the tangent a’ and of the binormal y‘ of a curve u® = const. are given by ‘ 1 ox! yi = ex! Vou du where ¢ is +1 or —1 as the case may be. From (5.7) we have for the direction cosines of the principal normal (% - Van as t, j, k take the values 1, 2, 3 cyclically. @ 248 SURFACES IN SPACE [Cn Vv From the Frenet formulas (6.1) we have = dy oe “Von the last expression following from (38.19). Substituting the above ex- pressions for 8, we obtain in consequence of (38.9) In like manner the torsion of an asymptotic line u' = const. ean be shown to be equal to ~/—K. Hence we have the theorem of Enneper: [44.3] The square of the torsion of an asymptotic line al a poind of a surface is equal to the absolute value of the Gaussian curvature at the point; the lorsions of the two asymptotic lines through a point differ in sign. In both of the above cases r = 1, as follows from (44.11), Con sequently equation (44.12) holds alsoin this case since & is constant for an asymptotic Tine. Since ry is a function only of a point and a diree- tion at a point as follows from (44.11), the quantity r ~ “is the same for all curves ut a point having a common tangent. It is equal to r for any curve for which 4 is a constant, and in particular for the geodesic through the point in this direetion. Accordingly 7) is called the geodesic torsion of a curve. From (44-11) and (41.1) we have [44.4] A necessary and sufficient condition that the geodesic torsion of « curve be zero at a point is that the curve be tangent to a line of curvature at the point. EXERCISES 1. A pline curve on asusface is a geodesic, if and only if the Lange phines to the surface along the curve are perpendicular to the plane of the curve, 2. On the reetifying developable of a twisted curve (see §12, Ex. 2) the given curve isa geodesie, and thus becomes a straight line when the developable surface is rolled out upon a plane. 3. Straight lines on a surface are the only geodesi 4. At each point of an orthogonal net on 2 surfa two curves of the net differ only in sign. 5, A yeodesie line of curvature is a plane curve, and a’plane guode: line of curvature. 6. When two surfaces mect under a constant angle, the geodesic torsions of the curve of interscetion with respect to the two surfaces are equal; if the curve of intersection is a line of curvature of one of the surfaces, it is for the other also; asyinptotie lines. the geodesic torsions of the 45) PARALLEL VECTORS IN A SURFACE 249 if two surfaces intersect along a curve which is a line of curvature of each, they intersect under constant angle. 7, When a surface is met by a plane or a sphere under constant angle, the curve of intersection is a line of curvature of the surface. 8. The geodesic torsion of acurve is given by where 0 is the angle made by the curve and the line of eurvature at the point for which p: is the radius of normal curvature. 9. A necessary and sufficient condition that the curves uw’ be straight linos, and thus that the surface by const, ona surface ruled surface, is that » {i} 10, Upon a surface straight lines are the only plane asymptotic lines. 11. Show that ii sata (242) from this it follows that any family of parallel pls orsect 9 mininial surface in curves which together with their orthogonal trajectories form an isometr du 45. PARALLEL VECTORS IN A SURFACE In §35 we gave an intrinsie definition of parallelism in a surface. We now interpret parallelism from the view-point of the space in which the surface may be considered as imbedded. Denoting by £ the eom- ponents in space referred to cartesian coordinates x* of the unit vectors X* at points of a curve C, we have i 2 Ox (45.1) & » jue’ Differentiating with respect to s, we obtain de ant ae! yy Pat a? ds ds aut aurau® ds (45.2) ‘ = (24 a {g} a oe ds Bf ds / aut the last expression being a consequence of (38.18). Since >, ¢’' = 1, we have analogously to (35.5) that the vector 9‘ associate to the vector # in space is given by (45.3) ae at. 250 SURFACES IN SPACE [Cu 1V The vector 7‘ is in general different from the associate vector of A“ in the surface (see theorem [45.1] and Ex. 3). If then equation (45.2) is multiplied by oe and summed with respect to 7, one obtains in i consequence of (38.2), (38.8), and (45.3) 45. oe (Sal dl (45. rts an( +X {a} ae). If the vectors at points of a curve are parallel as viewed from the enveloping space, their components in this space are constant, that is dé ds and, since g ~ 0, the quanti = 0. Hence the right-hand member of (45.4) is equal to zero, os * satisfy (35.13). ‘The same result follows if Dn! = , = 0, that is, if the vectors associate to the vectors 7 gare normal to the surface, Conversely, if equations (35.13) are satisfied, either the associate vector 9’ is a zero veetor, or it is normal to the surface. Hence we have [45.1] A necessary and sufficient condition that vectors along a curve in a surface be parallel with respect to the curve is that the vectors be parallel as viewed from the enveloping space or that the associate vectors in this space be normal to the surface. We have remarked that the tangents lo a curve are parallel with respect to the curve, if and only if the latter is a geodesic. In this ease the veetor 7’ in (45.8) is the principal normal (see (4.7)), and consequently theorem [44.2] is a particular ease of theorem [45.1] As a consequence of theorem [45.1] we have [45.2] If two surfaces are tangent to one another along a curve, vectors parallel with respect to this curve in either surface are parallel also in the other surface. From this theorem we have that if the vectors \" at points of a curve € in a surface S are parallel with respcet to C, they are parallel with respect to C in the developable surface which is the envelope of the tangent planesto Sat the points of C, andconversely. In consequence of theorem [35.4], when this developable surface is rolled out upon the plane the set of parallel vectors go into vectors which are parallel in the euclidean sense, If then one bas such a developable surface and C’ is the curve into which C goes, in order to find geometrically a set of vectors parallel with respect to C, one has only to take a set of vectors §45] PARALLEL VE! ‘ORB IN A SURFACE 251 parallel in the euclidean sense at points on C’, and when the develop- able is rolled back over the surface the resulting vectors are parallel with respect to C. The notion of parallelism of vectors in a Riemannian space of any number of dimensions is due to Levi-Civita.* He introduced the notion of infinitesimal parallelism which in terms of a surface imbedded in a euclidean space is as follows: Given two nearby points Py and P, of a curve in a surface corresponding to values s and s + ds of the are of the curve, and vectors \“(s) and A’(s + ds) = v%(s) + aN" as, ds terms of higher degree in ds being neglected; by definition the vector at Py is parallel to the veetor \" at Po, if as viewed from the enveloping space it makes the same angle with an arbitrary vector tangential to the surface at Po as the vector A* at Py does. Such a tangential vector an a is given by a® fae The components of the vector at P, as a vector of the enveloping space are given by dr! ar a +(° ga tds GX) as, which follows from (45,2) and (35.5). The condition that the two angles mentioned be equal is Of an! jae je + As 5 OE P+ Gds= ae ds In consequence of (38.2) and (38.8) this equation reduees to a" gaov")o ds = 0. Since the a’s are arbitrary, and the determinant g > 0, we have vi that is, the vector \ must satisfy equations (35.12) at, Py, and cor sequently depends upon the direction PoP; on the surface. Having arrived at this result, Levi tat used equations (35.13) to define a family of vectors \“ parallel with respect to a curve. x) ds = 0. 0 EXERCISES 1, When the coordinates 2‘ of the enveloping space are any whatever, the equations (45.3) aro aw a fil ae ast {ih as”? * 1017, 1. $1917, 1, p. 179. 252 SURFACES IN SPACE [Cx. IV where the Christoffel symbols {i are formed with respect to the metric tensor ay of space. 2. A nceessary and sufficient condition that vectors in a surface parallel with respect to a eurve C be parallel as viewed from the enveloping space is that the direetion of the veetor at cach point of @ be conjugate to C. 3. A necessary and sufliciont condition that the associate vector with respect toa curve C of a tangent veetor to a surface coincide with the associate vector with respeet to C of the tangent veetor as looked upon us a veetor in the surface is that the latter be conjugate to the curve C. 4. In order that the normals toa surface along a eurve C be parallel, it is neees- sary tht d)dss ~ d}, = O along C; in this case C is an asymptotic line. 5, Show by means of equations (38.19) that a necessary and sufficient condition that the vector associate to Lhe normal veetor to a surface with respect to a curve Con the surface he tangent to C is that C be a Tine of enrvature. 6. Show by means of equations (38.19) that a necessary and sufficient condition that the veetor associate to the normal veetor to a surface with respeet to x curve C on the surface be perpendicular to C is that the curve be an asymptotic Tine. 46. SPHERICAL REPRESENTATION OF A SURFACE. THE GAUSSIAN CURVATURE OF A SURFACE ‘The Gaussian curvature of a surface has been derived as an intrinsic property of a surface and it has been shown that it is equal to the reciprocal of the product of the principal radii of normal curvature, when the surface is considered as imbedded in space. These results are due to Gauss, who has also given an interpretation of curvature as a gencralization of the concept of curvature of a curve. In arriving at this result Gauss* introduced the concept of the spherical representation of asurface. In accordance with this concept one takes the unit sphere, that is, the sphere of unit radius with center at the origin, and makes correspond to each point P of the surface the point Pin which the sphere is met by the radius parallel to and with the same sense as the normal X* to the surface at P. Thus to a point of coordinates 2‘ corresponds the point of coordinates z‘, where (46.1) xt We denote by dé the linear element of the sphere, written = Sagtag = FOX OX yeaa! = hogan at (46.2) d= Lae dé 0 aut aut du? du® = hog du* du’, * 1827, 1, p. 226. 46) SPHERICAL REPRESENTATION OF A SURFACE 253 has being thus defined. In consequence of (38.19) the quantities has are expressible in the form art art txt (46.3) hap = ON dda” D Y But Bub 7 Bub Bu"? which in consequence of (24.7) and (24.18) reduces to (46.4) hag = day dps 9”. In order to give these expressions another form, we have hy = du dy g® + dads g® = duds g” + 9 (dw dis — dir du) = dudag” + g°(diz — didn) = duKw — guK, in consequence of (40.10) and (40.11). Proceeding in like manner we have hap = depKm ~ 09K. ly the linear element, af the spherieal representation may be written in the form (46.6) d® = Kn dag du* du’ — Kgag du* du’, and in consequence of (40.5) in the form 7 g — (Kn _ ) 2 (46.7) ds -(% K ds. Af h denotes the determinant of the quantities hag , we have from (46.4) and the expression (40.10) for K (46.8) Vh = KV 9, where ¢ is +1 or —1 according as K is positive or negative at the point under consideration. _ If we denote by X‘ the vector normal to the sphere at the point 2’, we have analogously to (38.5) ax! ax* < 1 {aut ut (46.9) x= ) Vhi ax! ax’ ad ond as i, j, k take the values 1, 2, 3 cyclically. Substituting from (38.19) and making use of (38.5) and (46.8), we have gi Ldn ang” | i _ ¢ |duthe| |g"g" Vil day day) 9° ~ K ddl |y!g 254 SURFACES IN SPACE (Cu. LY which reduces in consequence of (15.5) and (39.4) to (46.10) Xi = ox*, where ¢ is +1 or —1 according as K > Oor K <0. From this result it follows that according as a point of a surface is elliptic or hyperbolic, that is, K > 0 or K < 0, the vectors X‘ and X' have the same or opposite sense, that is, the positive sides of the tangent planes to the surface and to the unit sphere are the same or opposite. This is seen also from equations (41.9) when the lines of curvature are coordinate, since for an elliptic point the vectors ae us . ant have the same sense as the corresponding vectors due both have the 1 opposite sense, whereas for a hyperbolic point one has the same sense and the other opposite sense, If then we have a bounded portion of a surface such that K has the same sign at all points and on the boundary C, and we denote by © the contour of the corresponding bounded portion of the unit sphere, it follows that as a point P describes the curve C the corresponding point P on the sphere describes the curve C in the same or opposite sense according as K is positive or negative. The areas of two such portions of the sphere and surface are given by f Via aut = ff KV Gg dui du’, f vad ac respeetively, the limits of integration being the same for tlie two inte- gr From this result it follows that the limit of the ratio of these two integrals as the portion of the surface shrinks to a point (and con- sequently also the portion of the sphere) is equal to the value of eK at the point. When K = 0 at all points of a bounded portion of a surface and on the boundary, this portion of the surface is developable. Since the normals to a developable surface at all points of a generator are parallel, it follows that the spherical representation of € for a developable is the segment of a curve, and consequently of zero area. Hence the above result applies to the case K = 0 also, and we have the following theorem of Gauss*: [46.1] The limit of the ratio of the area of the spherical representation of bounded portion of a surface and the arca of this bounded portion as the latter shrinks to a point is equal to the absolute value of the Gaussian curva- ture at the point. * 1827, 1, p. 226, $46) GAUSSIAN CURVATURE OF A SURFACE 255 We consider now further consequences of the preceding results. Sine the radius of normal curvature of a surface at a point depends upon the direction of the normal section unless the surface is a plane or a sphere ¢ §40), it follows from (46.7) that the correspondence between a uurface other than a plane or a sphere and its spherical representation is conformal (sce §36), if and only if Ky, = 0, that is, when the surface . . . 1 is a minimal surface. For a sphere 1 + = £1, where ais the R a radius of the sphere; in this case ds Hence we have [46.2] A surface and its spherical representation are in conformal cor- respondenee, if and only if the surface is a sphere or a minimal surface. In both of these cases to an orthogonal net on the surface corresponds an orthogonal net on the unit sphere, From (46.5) it follows that the coordinate curves on a surface and the unit sphere are orthogonal nets only in case K,, = Oor dy = 0. Henee we have [46.3] The lines of curvature of a surface are represented on the unit sphere by an orthogonal net; this is a characteristic property of the lines of curva- ture of a surface which is not a minimal surface nor a sphere. This theorem is also a corollary of the following theorem: [46.4] The tangents to a curve in a surface and to its spherical representa- tion at corresponding points are parallel, if and only if the curve is a line of curvature. In order to establish this theorem we assume that the surface is referred to a coordinate system such that the given curve is a curve u” = const. and that the coordinate curves form an orthogonal net. For the given 5 ar’ . a s°*- and 2, must be proportional, being dircetion eS aut OM" oul a _ numbers of parallel lines by hypothesis. Hence from (38.19) we have ~d,g" = 0. Since g® = 0 for an orthogonal coordinate net and 9” 0, we have dy = 0 and the theorem is proved. In what follows we derive for a surface various results which are expressible in terms of the two tensors des and hag. Equations (41.5)’ express the condition that the normals to a surface along the curve : . du* ‘whose unit tangent vector has components iE form a developable curve the quan (surface. If these equations are multiplied by x and summed with ‘ ' 256, SURFACES IN SPACE (Cu. IV respect to i, the resulting equation in consequence of (38.16) and (46.3) is (46.11) (op = than) = @ = 1,2). Eliminating / from these two equations, we obtain idiadu" daadu™} |Inaiu” Ioedu®| which in consequence of (46.5) is equivalent to the equation (40.8) of the lines of curvature, as was to be expected. In order that quantities“ be determined by (46.11), ¢ must be a solution of the determinant equation | das ~ thas | = 0. In $41 preceeding theorem [41.1] it was shown that ¢ is one of the principal radii of normal curvature (sce Ex. 7). Hence the principal radii p: and pp are roots of the equation (hates — his)? — (dutex + dehy — 2dvhu)R (46.12) + Gide — Hence we have ot gy = tiles teh Piola _ 9 spa, (46.13) pipe where the quantities h® are defined uniquely by 1 hy = 85, that is (see (24.20), has he ea (46.14) Wea WB oR WE From (38.16), (46.10), and (46.3) it follows that the coefficients of the second fundamental form of the spherical representation are given by = yy eX! ax" © Tau" dub $46) SPHERICAT. REPRES TATION OF ASURFACE 257 From this result, (46.10), and equations of the form (38.18) we have for the spherical representation &xt — axt [7 ‘ (46.15) awaie— aur \apf ~~ MX where the Christoffel symbols are formed with respect to the form (46.2). By means of equations (38.19), that is, oxi (48.16) aun Tui? equations (46.15) may be given the form ee on OX” tayo ae fa = — Ras X*. (46.17) Bu and le If we differentiate (46.16) covariantly based upon gas and make use of (88.18), we obtain ax’ ax’ fy «i OX" Py i -— = — dacag” F, — 0 dacdspX', dueau® ~ Gut {ah doce 6" 543 — 9° daxtes on noting that the covariant derivative of g™ is equal to zero. In consequence of (46.16) and (46.4) these equations may be written fe ) a On! aura’ * (v3 Fees) 0 5 From these equations and (46.17) we have (oss ofg} 0) aa ‘ Multiplying by - and summing with respect to 4, we obtain 2 dang + dye {3 —dy {3} =0, = = hap X'. that is, ade, Srl y\ (46.18) oot dan {\ + deer If we multiply these equations by d, as defined by (42.11), and sum With respect to v, we obtain (46.19) {i} =a? a - oda {% 258 SURFACES IN SPACE [Cx 1V ‘The left-hand member of equations (46.15) is the second covariant derivative of the scalars X" based on has. We indicate such covariant derivatives by placing a bar over the index of covariant differentiation ; thus we write (46.15) m the form (46.20) Xiag = — hagX. Differentiating covariantly with respect to u” and noting that hes,; = 0, we obtain ax? Xlaay = — has 7 :aBy 8 our Applying the appropriate Ricci identities (22.19) to this covariant differentiation, namely ax* = ig Baar X's; — X! na where Réggy ate of the form (20.13) in terms of the symbols {ah we obtain ‘ = haa OX aur "7 Ou ub . axt ae - Multiplying these equations by x and summing for 7, we obtain in consequence of (46.3) (46.21) haghse — haghye — heeR ary = Reasy 5 which expresses the fact that the Gaussian curvature of the sphere is equal to +1, In terms of this covariant differentiation of de, based upon fag equa- tions (46.18) are expressible in the form (46.22) hang + day (3) - (3) =0, from which it follows that (46.23) dari — dag.» = 0. We now derive an important set of formulas. Since ax. 0, $46] 9 SPHERICAL REPRESEN’ TATION OF A SURFACE — 259 and the equations i (ys AX* aX’) _ x (x ‘ou Xa) = where > indicates the sum as i, j, & that the values 1, 2, 3 cyclically, are identities, it follows that 4 AX? ax* axe 24 5 OX" = ay , ao” Ae * oe on Fa or toa Si ae where i, j, k take the values 1, 2, 3 eyelically, and the a's are to be determined. Multiplying these equations by ax and summing with respect to i, the resulting left-hand member is equal to zero identically, and we have = Gathia + Garhon « Again multiplying the equations (46.24) by 3 x! (6 # a) and summing with respect to é, we have in consequence of (46.9) and (46.10) the = aahig + darks 5 where the sign on the left is + or — according as @ = 1,8 = 2ora = 8 = 1. From these two sets of equations we obtain one ag Bg tee Ie m=, 2 n=, = Vk vi vh vi and consequently equations (46.24) are : ( xi8X xe aX! _ _t (- a en w= evi ow oul Vk oy i due? (829) xt x _ eax! @( OX a a? 2) avin oh ae a Wh aul ant aus? as i, j, k take the values 1, 2, 8 cyclically, where eis 4-1 or —1 aecording as the Gaussian curvature is positive or negative. EXERCISES 1, The spherical representation of the lines of curvature of a surface of revolu- tion is an isometric orthogonal net. 2. The osculating planes of a line of curvature and its spherical representation are parallel (sec §6, Ex. 9). 3. If «and x denote the ra ii of curvature of a line of curvature and its sphericul 260 SURFACES IN SPACE (Cu. IV representation, and x, and i the radii of geodesic curvature of these respective curves, then eds = Rds, nds = ig di, 4, The spherieal representation of a plane line of eurvature is a cirele. 5. The tangents to a curve on a surface and to its spherical representation at corresponding points are orthogonal, if and only if the curve is an asymptotic line. 6. The angles botween the asymptotic lines at a point on a surface and their sphcrieal representation are equal or supplementary according as the Gaussian curvature of the surface is positive or negative at the point. 7. The prineipal directions for the tensor hag (see §26) are the principal diree- tions of normal curvature; find the relation between rin (26.17) and R for this ease. 8, When the asymptotic lines are coordinate, it follows from (46.18) that fa}-{2}-2{2} (Gl--{s} (B}--(G) exe. 9. When the coordinate curves form a conjugate net, it follows from (46.18) that eee fil -E2{G} oxo. 10. The asymptotic lines on 2 minimal surface form an isometric orthogonal net, a8 do also their sphorieal representation. 11. From (46.4) one obtains a3 = danedssh™® and from (16.5) gag = (01 + pr)das — prpsiias « 12, A necessary and sufficient condition that the linear element of a surface referred to a conjugate net be expressible in the form ds? = p*(indue” — 2hixdu'dut + hasdu*”) is that the Gaussian curvature of the surface be positive and that the coordinate conjugate net be the mean-conjugate net: (see $43). 13. ‘Theorem [46.4] and Ex. 5 are equivalent respectively to Exs. § and 6 of §45. 47. TANGENTIAL COORDINATES OF A SURFACE From the oquation of the tangent plane to a surface at the point 2‘, namely (47.0) x where # are current coordinates, we have that the algebraic distance W of the origin from the tangent plane is given by (47.2) W= DX. §47| TANGENTIAL COORDINATES OF A SURFACE 261 Looking upon this equation as an identity in u*, we have in conse- quence of (38.8) ; aw _ pax! (47.3) et ee xt For cach point P on a surface the quantities X‘, a and ox are the components of independent. veetors, as follows from (38.6) and (47.4) Accordingly the direction numbers 2‘ of the line segment OP, where O is the origin, are expressible in the form ai = aX’ + 6" ox ous In order to determine a and 6 we multiply these equations by X* and a sum with respect to i, and again by ox and sum with respect to 4, with the result, in consequence of (47.2), (47.3), and (46.3), ow a=W, hag = Oy If the second set of these equations is multiplied by 2°, where the latter are defined by (46.14), and summed with respect to 8, we obtain wre an hog h?™ — bY. Hence we have bk pryt po OX! OW (47.5) af = Wxt 4 ne Tf these equations are differentiated covariantly based on hes, we have in consequence of (46.20) and the fact that the covariant deriva- tive of his equal to zero ae! 0. x Wee se, CU aa 7 7 Gus ze) ae ou _ 3 Wu Cth aus W3% 262 SERFACES IX HACK (Cn. EV From these equations we have (see (38.2) and (38.16) Pe ox = ax! Gos = E(w Baa + he aa) (WO aa tS Ws) ag = Whe + a +h WasaWaa, and _ pax! OX a zs) deg = z — “(wx ae tos Wai) = —(W.a8 + has). Conversely, given any four functions X* and W of coordinates u such that >: X‘X' = 1, we have (47.4), and equations (46.20) follow from the definition (46.3) of has and the discussion in §46. rom (47.6) we have «Oa Ee ya oO Consequently X* are direction cosines of the normal to the surface for which 2’ as functions of u* are given by (47.5). Since (47.2) follows from (47.5), W is the algebraic distance of the origin from the tangent plane at the corresponding point. Four such functions X‘ and W are called tangential coordinates of the surface. ‘There can be no relation between the X’s and W of the form a;X! + bW = 0, where the a’s and b are constants, For, if there were, it would follow from (47,8) and (46.20) that dag = 0, and consequently the surface for which x! are given by (47.5) would he a plane (see §88, Ex. 1), in whieh case the X’s would be constants. Hence we say that X‘ and W are linearly independent (constant coefficients) and we have [471] Four linearly independent (constant coefficients) functions X‘ and W of coordinates u* for which 3); X'X* = 1 are tangential coordinates of @ surface whose equations x* = f'(u', u*) are given by (47.5), and whose two fundamental tensors are given by (47.7) and (47.8). If one has any four functions 6', 6°, 6°, 6* linearly independent (con- stant, coefficients) and one puts (47.9) wat @ = 1,2,3), where ¢? = 6” + 6” + 6", then X' and W satisfy the conditions of theorem [47.1]. Consequently from four linearly independent (con- stant coefficients) functions of u' and u’ one obtains the tangential 847] TANGENTIAL COORDINATES OF A SURFACE 263 coordinates of four different surfaces according as one chooses three of the functions to define X" by (47.9), that is, as one chooses the three functions which are direction numbers of the normal to the surface. When the coordinate lines on a surface form a conjugate net, the tangential coordinates X' and W are solutions of an equation of the form ad « 08 (47.10) oe as follows from (46.20) and (47.8). Conversely, if X* and W are four linearly independent (constant coefficients) solutions of an equation (47.10) such that 3), X‘X' = 1, it follows from equation (47.10) and (46.20) that (« - {a}) ON Oth)! = 0 From these equations we have a= {ip b= hu, and consequently for the surface with equations (47.5) we have die = 0, as follows from (47.8). Hence we have [47.2] Four linearly independent (constant coefficients) solutions X', W of an equation of the form (47.10) such that y.X'X' = ‘1 are the tangential coordinates of a surface upon which the coordinate curves form a conjugate nel. If one has any four lincarly independent (constant coefficients) solutions 6', --- , 6 of an equation of the form (47.10), the fmetions X', W defined by (47.9) are solutions of an equation of the form (47.10) and satisiy the conditions of theorem [47.2]. Hence we have [47.3] Four linearly independent (constant coefficients) solutions of an equation of the form (47.10) determine four surfaces upon each of which the coordinate curves form a conjugate net; cach surface is determined by which Uhree of the four solutions are direction numbers of the normals to the surface. When the asymptotic lines upon a surface of negative Gaussian curva- ture are coordinate, that is, di; = dx = 0, the tangential coordinates X‘ and W are solutions of two equations of the form 7 vo 60 06 ve 86 06 GTA) Sos an ge tang tO, Sie = a Han 5, + bd, 264 SURFACES IN SPACE (Cu. IV a8 follows from (46.20) and (47.8). By an argument similar to that which led to theorem (47.2} we obtain [47.4] Four linearly independent (constant coefficients) solutions X*, W of two equations of the form (47.11) such that 0: X'X* = 1 are the langen- tial coordinates of a surface upon which the coordinate curves are the asymptotic lines. Also by an argument similar to that which led to theorem [47.3] we obtain [47.5] When two equations of the form (47.11) admit four lincurly inde- pendent (constant coefficients) solutions, these solutions determine four surfaces upon each of which the coordinate curves are the asymptotic lines; each surface is determined by which three. of the four solulions are direction numbers of the normals to the surface From (47.6) and (47.8) we have ox! tv -e By (47.12) Sue =~ Aol ax* our” For a surface of negative Gaussian curvature referred to its asymptotic lines we have (47.13) dy = dn die = ho, where from (46.13) p? = — pipe. In consequence of (46,14) equations (47.12) are in this case expressible in the form ou’ _ eb ( ax’ ox? out VAN Ou OE)? (47.14) ac! __e(,,,0X"_ ox" au VANE ek Bui)? which in consequence of (46.25) are equivalent to ct wm (47.15) at aut? where 7, j, k take the values 1, 2, 3 cyclically. If then we define func- tions »* by (47.16) v= VX, §47) TANGENTIAL COORDINATES OF A SURFACE 265 these equations become where i, j, k take the values 1, 2, 3 cyclically. For the values (47.13) equations (46.23) reduce to (718) log Vio = {i} - {Rh} Palo Vip = {3 - {i} Analogously to (28.2) we have alog Vi _ fp 47.1 — (47.19) au 28}? and consequently (47.18) are equivalent to alogva_ fF (47.20) ar = haf Tn consequence of this result and equations (46.15) we have from (47.16) (47.21) or (2 vo i) v4, aul aut \\/p but au from which it follows that the conditions of integrability of (47.17) are satisfied, Conversely, we have [47.6] If v', »*, »* are any threc lincarly independent (constant coefficients) solutions of an equation of the form a) 7.22) sn = M, (47.22) audut where d is a function of u*, the coordinate curves on the surface whose coordinates are given by the corresponding quadratures (47.17) are the asymptotic lines, and the Gaussian curvature of the surface is equal to -V/ (Le For, the conditions of integrabilit y of (47.17) are satisfied by solutions of (47.22). If then we define X* by (47.16) with p = So; v's’, equations (47.17) are reducible to (47.15), from which we obtain ast aX! rae axt yy re aX" = hy sa he = Vie, 7 dul dul 7 On? uF du! Ou? 266 SURFACES IN SPACE [Cx. IV the last being a consequence of (46.9) and (46.10)., From this result and (46.13) it follows that K = —1/(D>:%')*. Equations (47.17) are called the formulas of Leliewore.* EXERCISES 1, On the envelope of the family of planes D Wi t+ Udet + Ur + Us 0, where Uj, Us are funetions of wt, and U}, Uz of 2, the coordinate eurves form a conjugate net of plane eurves. 2. On the envelope of the family of planes cos utzt + sin uta? + cot uta + Ur + Ur = 0, where U, and Uz are functions of u' and w respeetivel plane lines of curvature. 3. The conditions of integrability of equations (47.12) are equivalent to equa-, tions (46.23). 4, In consequence of (47.19) equations (46.23) are equi , the coordinate ew'ves are alent to 8 dey 6 deg 4 tex 6 ass [73 Gab [5 ays ab Va bt Vi + Sa {ah+ Sia} Vi (3) - ale =o 5. When a surface is referred to its asymptotic lines, the equ reduce to ns of Ex. 4 a de 5) ~-, log —= = —2. (A x a). aun BR a8) 6. A necessary condition that the coordinate curves on the unit sphere are the representation of the asymptotic lines on a surface is that a {i} . when this condition is satisfied, the coordinate curves on the sphere are the represcutntion of the asymptotic lines on a family of surfaces, whieh are homo- thetie transforms of one another, and the equations z* = /*(w', w2) of the surfaces can be found by quadratures, 7, In order that equations of the form (47.11) admit four linearly independent (constant cocfticients) solutions it is necessary that 3 a jue YT jy 8, ‘The most general right conoid is defined by equations (47.17) for the values vou ptm ut ote oft), * 1989, 1, p. 126. 547] TANGENTIAL COORDINATES OF A SURFACE 267 9. When a surface is referred to its asymptotic lines, a necessary and sufficient condition that the lines ut = const. be straight, and thus that the surface be ruled, is that the normals to the surface along each curve ut = const. be parallel to a plane, as ean be shown by means of equations (47.17) 10. When the asymptotic lines in one system on asurface are represented on the sphere by great circles, the surface is a ruled surface. 11, When the equations of the unit sphere are of the form of §28, Ex. 7 with a = 1, the coordinate curves are the usymptotie lines of the sphere and equation (47.21) is aro awa (+ wat = -%, of which the general integral is = Hee) + ever) T+ eo —¥ nw whereg and y are arbitrary functions of u‘ and u! respectively 12. Find the equation (47.21) when the surface is a hyperboloid of one sheet; when a hyperholie paraboloid. (See §43, Exs. 8, 9) 13, When the coordinate curves on the unit spher a (2) _yfi\f2 ai aaa atisty the condition they represent the asymptotic lines on a surface whose total curvature is of the form bole) + ye 14, When the coordinate curves on the unit sphere forin an orthogonal net necessary and sufficient condition that the curves uw? = const. be circles, that is, curves of constant geodesic curvature, is that 13 — hog Vi Vig 8 vi 15. Whew upon the nnit sphere the eurves of one family of an orthogonal iso metrie net are cireles, so also ave the curves of the other farnily (see $34, Ex. 6) 16. When the equations of the unit sphere are Quit, Quty ul? tt = +ut td the coordinate curves are circles whose planes are afuls—1)=0, at uted — 1) = 0, and they form an orthogonal net, since the linear element is, 4(dut? + du’) Oe ue DE 268, SURFACES IN SPACE [Cn IV ‘These curves are the spherical representation of the plane lines of curvature of surfaces for which where U' and Up are arbitrary functions of ut and w respectively. 17. When the equations of the unit sphere are xs, x xe VE at sin ot, VT = a sinh, cos wt + a cosh 02 cosh u? + a cos ut , where ais a constant (| a| <1), the coordinate curves are cireles whose planes are V1 = a 2 = tanut (2 ~ a) = 0, and they form an orthogonal net, since the linear element is _ = edu + du) © (cosh wt + a cos w! ag ‘These curves are the spherical representation of the plane lines of curvature of surfaces for which wa Mice + Ud cosh @ + a cos ub" where U, and U, are arbitrary funetions of u! and u? respectively 18. When a Oin Ex. 17, the eurves u? = it. ona surface with this spherical representation Liv in pavalfel plies and the planes uf Uw eaves ue eunst envelope a eylinder, 48. SURFACES OF CENTER OF A SURFACE, PARALLEL SURFACES: In §41 it was shown that the normals to a surface S along the lines of curvature C, and C; through a point P on a surface form two develop- able surfaces D, and D;, and that the coordinates of the points P, and Pz on the edges of regression of these surfaces corresponding to P, that is, on the normal to § at P, are given by (48.1) gaat aX! a = 2h + oX', where p; and p, are the principal radii of norma] curvature at P. The surfaces S, and S», which are the loci of the points P; and Ps respec- tively, are called the surfaces of center of the surface S. When S is # sphere, in place of two surfaces of center, there is only the center of the sphere. Also it is evident geometrically that the normals to & surface of revolution at points of parallel form a right cireular cone 38 SURFACES OF CE TER OF A SURFACE 269 with vertex on the axis of the surface, and consequently, the axis is one of the surfaces of center (see Exs. 1, 2,3). In what follows we consider only the case where neither of the surfaces $, and S: is degenerate. ‘The edges of regression 1', and 1’, of the developable surfaces D, and Dz lie on the respective surfaces Sy and S:. Since the normal at P is tangent to Ty at P; and to P's at Pe, this normal is a common tangent to S, and S:, and consequently the normals to S are common tangents of the surfaces $: and $,, Since the generators of Dy are tangent to Sz, it follows that D; is the envelope of the tangent planes to S: along, a curve T's, and that at P: the directions of I and P° are conjugate as follows from the definition of conjugate directions in §42. In like man- ner the developable D, envelops S; along a curve I and at P; the direc- tions of ['; and Ty are conjugate. The tangent planes to D, und Ds slong their common generator, namely the normal at P, are perpendicular, since they are determined by this normal and the tangents to C, and C: at P. This tangent plane to Dy is the osculating plane of TP; at P:, and is perpendicular to the tangent plane to S; at P,, since the tangent plane to Dz is this tangent plane to S,. Hence I is a geodesic on $; by theorem [44.2], and similarly T2 is a geodesic on Sz. Accordingly we have [48.1] The edges of regression of the developable surfaces consisting of the normals to a surface along the lines of curvature of one family are geodesics on the surface which is the locus of these edges; the developable surfaces which consist of the normals to the surface along the lines of curvature of the other family envelope this surface of center along the curves conjugate to the edges; the osculating planes of the edges on one surface of centers are tangent to the other surface of centers. We now obtain in an analytical manner the results which have just been deduced geometrically. We assume that the surface S$ is referred to its lines of curvature, and that p: and pz are the radii of principal curvature for the directions of the respective curves u’ = const. and u’ = const. at a point. Accordingly we have Tid lim 4 (48.2) , pa gt pe gee = on Making use of (41.9). we obtain from the first of (48.1) azi _ aps y _ a ae ae +1 au” (48.3) 270 SURFACES LN SPACK, (Cu. IV Fig. 16. A surface and its two surfaces of center 548] SURFACES OF Ch eR OF A SURFACE 271 from which one finds for the components g's of the first fundamental tensor of the surface of centers S; (ay yap do (am 84) os — (8), a= gage #2 — (Ga) +O Hence the linear element of S; may be written : (48.5) dst = dpi + ( - n one’, m from which it follows that the curves u" = const., that is, the edges of regression of the developable surfaces D,, are geadesies (see (32.15), and their orthogonal trajectories are the curves for which p; = const. From (48.3) we have 35, 2%, 2 si gv 7 0 (= 3, 2), from which it follows that &, are direction numbers of the normal to 8), a result which we obtained pi viously from geometrical considerations. Henee the divection cosines of the normal to S1 are given by i a ar (48.6) xia 8%, Vou ou where « is +1 or —1so that there shall hold for 8; the results analogous to (38.9). By substitution it is found that this means that « is +1 . om . or —1 according ax ( - 1) °! is positive or negative. pe au If dig denotes the second fundamental tensor of S,, one finds from equations analogous to (38.13) on making use of (41.9) and (38.17) VGu p dy = ~aV@ a (48.7) dy = 8 (: - a) age _ 6p) ge 2Van pf ok gu the last expression being a consequence of (41.12). Since djs second part of theorem [48.1] is established anew. In like manner from the second of equations (48.1) we obtain as the fundamental tensors of the second surface of centers S: 0, the on _ (ae) p2\" by Ope Ops 8.8) gu = (3) + (: > ") guy 2 = Fh ae (48.9) dio HO, ait wi Vga Ou" 272 SURFACES EN SPACE [eu 1. where the normal X/ to Se is given by (48.10) Xp= From the foregoing results we have by (40.10) that the total curva~ tures Ky and Kz of S: and Sz respectively are given by opt Lae (p2 = 1)" Ope au (48.11) Ki = -~— Ky = — We inquire next under what conditions the normals to a surface S are normal to @ second surface 3. ‘The equations of such a surface 3 are given by (48.12) Baal + ext where ¢ is to be determined so that ri OE xe = that is, fie , yi, ax) x (e. F aig + ex) =o. Tt follows from these equations in consequence of (38.2) and (47.4) that ¢ is a constant, and that for any constant { these equations are satisfied. Hence we have [48.2] If segments of constant length are laid off along the normals to a surface from points of the surface, the locus of their other end points is a surface with the same normals as the given surface. ‘Two surfaces in such relation are said to be parallel. It is evident that there is an endless number of surfaces parallel to a given surface, that the lines of curvature correspond on all these surfaces in consequence of theorem [41.1], and that the family of parallel surfaces have the same surfaces of center. Tf Gap and dag denote the first and second fundamental tensors of the surface 3 with equations (48.12), we obtain in consequence of (38.16) and (46.3) pF a: gue But 7 Yee ~ dag + Phas, (48.13) - > at’ ax" Le oe dag — thea. & 0 548] PARALLEL SURFACES 273 Since the surfaces S and S have the same surfaces of center, it follows from (48.1) and similar equations for 5 that the principal radii of normal curvature of 3 are given by (48.14) A=o- pap k If Sis a surface of constant total curvature 1/a’, we have G+ (2+) =a" When in particular ¢ = ka, this equation reduces to iji (1 atay ta and we have the following theorem of Bonnet*: [48.3] Among the surfaces parallel to a surface of constant total curvature L/a’, there are two of constant mean curvature +1/a respectively at the distances + a from the given surface. Also we have conversely [48.4] Among the surfaces parallel to a surface of constant mean curva~ ture one has constant positive total curvature and another constant mean curvature, From (48.3) it follows that the normals to the surface S are tangents to the curves v= const. on $1, which from (48.5) are seen to be geo- desics. We shall prove the converse theorem [48.5] The tangenls to the geodesics on any surface are normal to a family of parallel surfaces. We assume that the surface is referred to a family of geodesics u? = const, and their orthogonal trajectories, and write the linear element in the form (see (32.15) (48.15) ds? = dul” + gad”. | The tangents to the geodesics have equations of the form (48.16) from which we have da! = OF ; (48.17) ext eat x 4 x A + (tae + fea). * 1853, 1, p. 437. 274 SURFACES IN SPACE (Cu. IV In order that the tangents to the geodesics be normal to a surface (48.16) + must be such that Dy 22 dé! = 0 forall values of du and du, Since tof ‘ Pag as.is) 32% 2 yp ae! ae! > att a 7 Ou! Gut 7 But ut = gn, from which we have feat fata! oO 9 ee te 1 or or = 0, nd (48.19) 7 au! au 7 au! aut aut ‘one finds that the conditions upon 7 are ar or T+ 5a 7% 5a = Consequently (48.20) ree-u, where ¢ is an arbitrary constant. Each value of ¢ determines a partic- ular one of the family of parallel surfaces normal to the tangents to the geodesics 1? = const. on S, and the theorem is established. ‘The given surface S is one of the surfaces of center of these parallel surfaces. In order to find the other surface of center 3 we note from theorem [48.1] that (48.16) are equations of 5 provided that 7 is such that d#* are direction numbers of the tangent to the edge of regression of the envelope of the tangent planes to S$ along a curve conjugate to the geodesics u? = const. Since this tangent must lie in the correspond- ing tangent plane to S which is also the corresponding osculating plane . . ox* of the edge on S, the direction numbers of whose normal gre a an! rmust be such that > 2%, dz’ = 0. Tn consequence of (48.18), (48.19), and 2 of ant oogtat re ae re ax ne? out A Be Ow Out zr ax! xt 7 He Die which result from (48.18), we have = (v0 ea 1_ _alog Vgzz r aut Hence (48.21) $48] PARALLEL SURFACES 275 and equations of 5 are _~— 1 a 4 log V/am du aut (48.22) From (34.5) it follows that r given by (48.21) is the radius of geodesic curvature of the curves u' = const. and consequently (48.22) gives the centers of geodesic curvature of these curves. If then we say that the surface S defined by equations (48,22) is the surface complementary to the surface S as determined by the family of geodesics u? = const. on 8, we have [48.6] A complementary surface of a surface S is the locus of the centers of geodesic curvature of the orthogonal trajectories of the geodesics on S which determine the complementary surface. It follows from the above definition that either surface of centers of a given surface is a complementary surface of the other. Moreover, from equation (48.5) of the linear element of the surface of center Sit follows that the geodesics which determine the other surface of center S; as a complementary surface of S: are the orthogonal trajectories of the curves p, = const, on S;. Also from the linear element of the surface Sp, namely, dg o(a - a) dul + dp, ps which follows from (48.8), one has that the surface 8; is the complemen- tary surface of S; which is determined by the geodesies orthogonal to the curves p» = const. on S:. Hence we have the following theorem of Beltrami*: [48.7] The centers of geodesic curvature of the curves p. = const. on 8 and of p = const. on S2 are corresponding poinis on Se and Ss respectively, corresponding points being the points of tangency of a common tangent lo the two surfaces. EXERCISES 1, One of the surfaces of conter of a surface of revolutis other is the surface of revolution of i surface, 2. For the envelope of a one-parameter family of spheres one of the surfaces is the axis, and the volute of the meridian eutve of the } Of center is the curve of centers of the spheres. * 1865, 1, p. 18. 276 SURFACES IN SPACE (Cu. IV 3. In order that the surface of center 8; be a curve, it is necessary and sufficient that 91,932 — 913 = 0; it follows from (48.4) that in this case p: is a function of u* alone. 4. The surfaces of center of a helic: (see §38, Ex. 5) are helicoids with the same axis and the same parameter as the given helicoid. 5. The surface Vent ui oa a whut abt ute ab ute’ has the following properties: the coordinate curves are plane lines of eurvature} pi = wt, p2 = —uh; the surface is algebraic of the fourth order; the surfaces of center are focal conics. 6. A necessary and sufficiont condition that the asymptotic lines correspond on the two surfaces of center of a surface is that there be a functional relation be- tween the principal radii of tho surface. When there is such a funetional relation the surface is called a surface of Weingarten. 7, Equations of the lines of curvature on S; and S2 are respecti y yr Sp: 3, 2 a [8p \* jx Op, Bpy ju a 80 804 py +[2(%) 1g. Ger 20s Os 4 52 ln — nn | ene pF aul aut” pr aut) © ps Sut but pt pF a PF ut? = 0, rat 2 au 8p. 860 2, [01861 Ope gee (Boe ph auta™ T | gt awa * gt ou) * a nceessary and sufficient condition that the lines correspond on the two surfaces is that p1 — ps = a, where a is a constant, in which ease Ki = Ks = 1/a%, and the asymptotic lines on the two surfaces eorrespond. . 1 8, Show that, in consequence of (41.12), fo: = 3, Ata: = — pia ee differential parameters arc formed with respect to (48.4); then from (34.6) it fol- lows that the radius of geodesic curvature of the curves a: = const. on S; is equal to ps ~ px, from whieh follows theorem [48.7] 9. The surfaces parallel to a developable surface are developable surfaces 10. The surfaces parallel to a surface of revolution are surfaces of revolution, 11, Lines of curvature on two parallel surfaces are the only corresponding conjugate nets. 12, A necessary and sufficient condition that th correspond to a conjugate system on a parallel surface is that the twosurfaces he surfaces of constant, mean curvature in the relation of theorem [48.4]. From (48.20) and (48.21) it follows that the principal radii of normal curva- » where the asymptotic lines on a surface $49] SPHERICAL SURFACES 277 ture of the paralle] surfaces normal to the tangents to the geodesics ut = const. of a surface S with the linear clement (48.15) are given by Weg pene 9 tox Von but 14, If $ is a surface applicable to a surface of revolution, the tangents to the curves corresponding to the meridians of the surface of revolution are normal to family of surfaces of Weingarton as follows from Exs. 6 and 13. 15. From theorem [26,4] it follows that the left-hand member of equation (41,1) of the lines of curvature of a surface is an indefinite quadratic form (see §28). When the lines of curvature are coordinate, this form divided by 4/9 is reducible by (41.10) to 1 1 ~ Cy by means of (41.12) and §28, Ex. 15 the curvature of this form is reducible to ____ 2oipe (oes) Vange (ox ~ on)? (us, wt) hence the curvature is equal to zero for a Weingarten surface and in consequence of theorem [28.5] the lines of curvature on a surface of Weingarten can be found by quadratures. 49. SPHERICAL AND PSEUDOSPHERICAL SURFACES A surface whose Gaussian curvature K is a constant not zero is called a surface of consiant curvature. According as K is positive or negative the surface is called spherical or pscudospherical. We consider first spherical surfaces and put K = 1/a’, When such a surface is referred to a family of geodesics u* = const. and their orthogonal trajectories, the linear element is of the form (49.1) ds? = du™ + gz du™. From (28.5) and §28, Ex. 1 we have that gee is such that o Von 1 (49.2) NE = — 5 Vom: The integral of this equation is 1 1 (49.3) Von = ou’) os = + Wut) sin S 278 SURFACES IN SPACE (Cu. IV From (34.5) we have that the geodesic curvature xe of the coordinate curves u' = const. is given by alog Va (49.4) fa ON 922 If, in particular, the coordinate curves are chosen so that the curve u’ = 0 is a geodesic, the curves w = const. the geodesic parallels to this geodesic, and the curves u’ = const. their geodesic orthogonal trajectories, it follows from (49.4) for the curve u' = 0 that ¥(w”) = 0 in (49.3). Hence by a suitable choice of the coordinate u* the linear clement becomes 1 (49.5) dé = du +2 cos? dul, where ¢ is a constant. From this it follows that e(w: — ui) is the arc of the geodesic u' = 0 between the curves u* = u3 and u* = uf. From the foregoing discussion it follows that all spherical surfaces of the same Gaussian curvature are isometric, and consequently all these surfaces have the same intrinsic properties. However, there is a dis- tinetion between spherical surfaces of the same curvature as viewed from the enveloping space. This is seen, in particular, when we con- sider spherical surfaces of revolution A surface of revolution with the 2”-axis for axis of revolution is de- fined by the equations (49.6) e=deos@, wv=asind, ot), the function ¢ determining the character of the rotated curve. In terms of these coordinates the linear element is (49.7) ds = (1 + ¢'(u')®) da” + a" da™. When we compare this equation with (49.5), we have du? = (14+ oY) da, a from which it follows that eta) = f vita = aa* = f 4/1 3s Consequently equations of a surface of revolution with the linear element (19.5) are in du. a uw sg ccos—cosw’, — x* = ¢ cos—sinu’, a a (49.8) —— e-[Yr~% $49] SPHERICAL SURFACES 279 There are three cases to be considered according as ¢ is equal to, greater than, or less than a. Case 1°. ¢ = a. Now equations (49.8) are : . 1 wo a _ i (49.9) 2 =acos\cosw, 2 =acosSsinv, 2 = asin 2 a a which are seen to be equations of the sphere of radius a with center at the origin. We observe before taking up the other two cases that the expression for z* in (49.8) is an elliptic integral and consequently in each case 2° is expressible by means of appropriate elliptie functions. 17. A spherical surface of revolution of the hyperbolic type Casz 2°. ¢ > a. From the expression in (49.8) for x* it follows that 1 : sin’ & = and consequently Vi Since 2° is periodic, the surface consists of a succession of like parts or zones each of which is bounded by minimum parallels of radius ¥/c? — a®, the greatest, parallel of each zone being of radius c, as shown in Fig. 17. The angle which the tangent to a meridian curve makes with the plane «* = 0 is given by ‘These spherical surfaces are said to be of the hyperbolic type. 280 SURFACES IN SPACK | Gaz 1. Casz 8°. ¢ @ and the surface is applied upon the sphere the zone of the surface not only covers the corresponding zone of the sphere, but there is an overlapping; whereas when ¢ < a the zone of the surface fails to cover the zone of the sphere. The form (49.5) was obtained from (49.3) by taking o(u*) = c, ¥(’) = 0. For any constant values of g and y the expression (49.3) can be written in the form mn (49.10) Von = ¢ cos (¢ + v), where b and ¢ ure arbitrary constants. According as we take b = 0, — 1/2, or — 7/4 wo get the following respective forms of the linear element, 1 @ ds = du +0 cos du, 1 (49.11) (i) dé = du" + é sin® Sau", (ii) ds’ = du + 2 cost (¢ - 3) au. a 4, We have seen that for first of these forms the curve u' = 0 is a geo- desic. For the form (ji) w’ and cu’/a are polar geodesic coordinates by theorem [33.1], and for the form (iii) the curve u' = 0 has geodesic » curvature t/a, For a pseudospherical surface of curvature —1/a® one has in place of (49.3) 1 2 (49.12) Joe = plat) cosh = +002) sinh 5 282 SURFACES IN SPACE [Cu, IV When (i) the curve u' = 0 is a geodesic, (ii) the coordinates u', eu’/a are polar geodesic, or (iii) the curve u' = 0 has geodesic curvature — 1/a, the linear element has the following forms respectively: : @ dat = dul* +c cosh? ~ du, (9.18) Gi) de = du +e sinh? di, (ii) ds = du® + be die", Any ease other than these for which ¢(u’) and ¥(u*) are constants may be obtained by taking for ~/gz. either of the values ¢ cosh (« + ) orc sinh (@ + ), where b and ¢ are constants. By change of the co- ordinate u', the corresponding linear elements are reducible to (i) or (ii), Hence the forms (49.13) are general for the case when g(u*) and ¥(u*) are constants ‘The forms (49.13) are linear elements of surfaces of revolution whose equations are given by (49.6) where for the respective forms we have @ : u c cosh -—, x a Go.) (i) a! = esinh ©, @ Gi) oth set, ae [y — en ae In considering these three cases in detail we remark that. the integrals in (i) and (ii) are clliptic and consequently in each of these cases 2° is expressed by means of appropriate elliptic functions. 1 Case (i). The maximum and minimum values of sinh* = are a/c and 0, respectively and consequently the maximum and minimum values of the radius a' are 1/a? + ¢ and ¢ respectively. At points of a maximum parallel the tangents to the meridian curve are perpendicular to the axis of rotation and at points of a minimum parallel they are parallel to this axis as follows from the value of dz*/da'. Since 2° is periodic the surface consists of a succession of spool-like zones, see Fig. 19, the maximum parallels being cuspidal edges. ‘These pseudospherical surfaces are said to he of the hyperbolic type. 549] PSEUDOSPHERICAL SURFACES 283 Case (ii). In order that the surface be real, c? < a’, a restriction not: necessary in either of the other cases. If we pute = asin @, the maximum i and minimum values of cosh? are esc’# and 1 respectively, and the corresponding values of the radius a’ are acos @ and 0. The tangents to the meridians at points of the maximum circle are perpendicular to the axis of rotation and at points for which a — 0, the tangents make the angle @ with the axis, The surface is made of a succession of zones Fic. 19 Fie. 20 Fie. 19. A pseudospberical surface of revolution of the hyperbolic type Fic. 20, A pscudospherical surface of revolution of the elliptic type similar in shape to hour glassos. Fig. 20 represents such a zone, the maximum parallel being a cuspidal edge. These pseudospherical surfaces are said to be of the elliptic type. Cas (iii). If we make the substitution sin@ = ‘ 1 in (49.14) (ii), we obtain aleos 6 — log (ese 6 + eat. 6]. 284. SURFACES IN SPACE [Cn. IV From this result we have that 22 angle which the meridian makes with the axis of rotation. One finds that the length of the segment of a tangent to a meridian from the point of contact to the axis of rotation is a, and consequently the meridian curve is a tractrix (see §6, Ex. 11). These pseudospherical surfaces are eaid to be of the parabolic type. They are called pseudospheres. See Fig. 21. = cot @, and consequently @ is the Fic. 21. A pseudosphere (pseudospherical surface of revolution of the parabolic type) From (49.4) we find that the geodesic curvature of the parallels of the surfaces with the linear elements (49.13) are given by the respective expressions : : @ -lionb”, Gi) -Jeoth%, Git) -2, ae ah a Since no one of these expressions can be transformed into any other, if u! is replaced by u' + ¢, where c is a real constant, it follows that two $49] SPHERICAL SURFACES 285 pseudospherical surfaces of revolution of difforent. types are not appli- cable to one another with merdians in correspondence, whereas we have just found this to be the case between different types of spherical surfaces of revolution. The results (49.11) and (49.13) constitute another proof of the first part of theorem [31.1], and more particularly we have [49.1] The linear clement of any surface of constant curvature 1/a* or — 1/a? is reducible to the respective forms (i), (ii), Gii) of (49.11) or (49.18) according as the coordinate geodesics “are (i) orthogonal to a geodesic, (ii) pass through a point, or (iii) are orthogonal to a curve of constant geodesic curvature. We proceed now to the consideration of surfaces of constant curvature when the lines of curvature are coordinate. ‘To this end we assume that the Gaussian curvature of the surface is equal to e/a’, where c is +1 as —I according as the surface is spherical or pseudospherical. Tf such a surface is referred to an isometric-conjugate net (see §42), we have du deg _ 1 15 wae Bae, (49.15) vo vo a When the expressions for du, da: in (49.15) are substituted in the Codazzi equations (39.6), the resulting equations are reducible, in consequence of (28.1) and (28.2), to Ogu: \gaz Oguz Agu gu (& - — 2) $n out fea 9912 nu out aut aut + dou _ , 8a _ 9 Aw gu 82 4 9, 802) _ $a (& na? 7) + oe 2 (ee aut 22 2a) Since g ¥ 0, these msi are equivalent to ag. 2 (on ~ eg) ++ 20 5 a =0, (49.16) a a je (On — ee) — 2 0P = These equations are satisfied by (49.17) gu —egm — const, gs = hence we have [49.2] The lines of curvature on a surface of constant Gaussian curvature form an isometric-conjugate net. 286 SURFACES IN SPACE (Cu, IV For the case of spherical surfaces the equations (49.17) are satisfied by (49.18) gu = @ cosh’ 6, giz = 0, gaz = a” sinh? 4, and from (49.15) we have (49.19) du = dg = asinh @cosh 8, dy = 0. When these values are substituted in the Gauss equation (40.10), where Kis given by (28.5) and §28, Ex. 1, one obtains a6 49,20) (49.20) ae = + ao —_ + sinh 6 cosh @ = 0. au Conversely, in consequence of theorem [39.1] we have [19.3] For cach solution of equation (19.20) the quantitics (19.18) and (49.19) determine a spherical surface, the lines of curvature being coordinate. For the case of pscudospherical surfaces the equations (49.17) are satisfied by (49.21) gu =a" cos’ 6, ge = 0, gue = and from (49.15) we have (49.22) dy = dz = asin Ocos 6, dz = 0. When these values are substituted in the Gauss equation, one obtains ao Fo 19.23 “sy — sin 9 cos 6 = 0. (1928) au a Hence we have [49.4] For cach solution of equation (49.23) the quantities (49.21) and (49.22) determine a pseylospherical surface, the lines of curvature being coordinate. EXERCISES . The lines of curvature and the asymptotic lines on a surface of constant Gaussian curvature can be found by quadratures (see §48, Hx. 15). 2. When the linear clement of a pscudospherical surface is in the form (49.13) (iii), the surface defined by at "ut Hegt- is pscudospherical, and the tangent planes to the two surfaces at corresponding points are perpendicular. §49] PSEUDOSPHERICAL SURFACES 287 3. The helicoids an ke geveosw, 2=wsinw, = f — 4 ldu! + hut, eke where a, h, and kare constants, are spherical surfaces. 4. A helicoid whose generating curve is a tractrix is a pseudospherical surface ; it is called a surface of Dini 5. For a pseudospherical surface defined by (49.21) and (49.22) the ment of the spherical representation is ear ele- dj! = sin? @ du’ + cos? 0 du’ 6. The surfaces of center of x pscudospherical surface are applicable to a catenoid. 7. On a surface of constant curvature the area of a geodesic triangle is ina constant ratio to the difference between the sum of the angles of the triangle and two right angles (see theorem [33.2)). 8. When upon a surface there is more than one family of geodesies which together with their orthogonal trajectories form an isothermal system, the surface is of constant curvature, 9. When the equations of the sphere with center at the origin and radius a are written in the form zisasinutcosut, 2t=asinwtsinw, 2 = a cos ut, the linear element is @ ds? = aX(dul® + sin? ut du2*), The equation of any great circle of the sphere is of the form (ii) A sin u! cos u? + B sin uw! sin u? + C cos ul = 0, where A, B, C are constants not all zero; this is an equation of the geodesics on any surface of Gaussian curvature 1/a? in terms of the coordinate system for whieh the Tinear clement is (1). 10. When in equation (ii) of Ex. 9 w and ut are expressed as functions of general parameter ¢ and this equation is differentiated twice with respect to t, and A, B, and C are eliminated from the three equations, one obtains BLO coy yale te] ELH cous (AV oo a | ae “ual alae west Va) | since this is the equation (87.1) for the case when the linear element is (i) of Ex.9, this iy au analy tic proof of Uhe fact that equation Gi) is an equation of the geo- desics of the surface with this lincar element. 11. When the coordinates on a pseudospherical surface are sueh that the linear element is ds? = at(du 4 sinhtut du2*), an equation of the geodesics is @ Atanh ut cos ut + B tanh wt sin wt + C = as may be verified by the process used in Ex, 10. 288 SURFACES IN SPACE [Cu. IV 12, When equation (ii) of Ex. 9 is divided by cos u! and the terms multiplying A und B in this equation are equated to z and y and similarly the terms multiply- ing A and B of cquation (i) of Ex. 11, these equations define the correspondence between the surface and the plane in accordance with which geodesics on the sur face are represented by straight lines in the plane. 13. When the coordinates on a pseudospherical surface are such that the linear clement is of the form (iii) of (49.13), an equation of the geodesics is A@ens'i* + cw") + Bowe +0 = 0, as may be verified by the process used in Ex. 10;in this case the equations gee, yaar determine a conformal representation of the surface npon the plane geodesic in the surface is represented in the plane hy a circle with the z-axis or by a line perpendicular to this axi uch that any center on 50. MINIMAL SURFACES In $43 a minimal surface was defined as one whose mean curvature is equal to zero. In this section we establish a property of minimal surfaces which accounts for their name, and then derive other properties of such surfaces. Consider a surface S in space defined by the equations (50.1) v= si, wv’). Consider upon this surface a simply connected region R with curve C as contour and for this region consider the integral (50.2) I J J Lau du’, where L is a function of the 2’s and their first derivatives w'(u', 2°) be three arbitrary functions such that (50.8) (u,v) = 0 along C. ‘Then (50.4) Site, where ¢ is an infinitesimal, define another surface 3 nearby S and con- taining the curve C. When these expressions are substituted in the funetion L in (50.2), we have for 3 the corresponding integral - By eats Oy 6 80") andar’, (50.5) 7- [1(+ bet oad be) aut ae, $50] MINIMAL SURFACES 289 In order that J shall be a minimum for all the surfaces passing through C it is necessary that the derivative of T with respect to ¢ be equal to zero for ¢ equal to zero. From (50.5) we have that this condition is [J (Gee + sat on) a =o where 2, = 24. This equation may be written ip the form [Po (Se~ sew aet,) eet + Lf ge (ot gut) a= 0. In consequence of (50.3) the second of these integrals is equal to zero and, since the first integral must be equal to zero for arbitrary functions a’ subject to the condition (50.3), we have 5 a ab (50.6) aut dat, ~ 3 These are the generalized equations of Euler. When these conditions are satisfied the surface S is said to be an extremal for the integral (50.2). We consider now the particular case when the integral (50.2) is the integral of area, namely, (50.7) I= ff Vig aul at. In this case the equations (50.6) are a avg _ ava (508) ou* dai, oat where g is the determinant | gop | and (50.9) Ges = Dy ciat'g. Now we have = VG Gaia + oe » @ ¥ a) = V9 97 290 SURFACES IN SPACE [Cx. IV From these equations we have in consequence of (28.2) wit) = valores ff} + 25 (org) - ota) = V9 0" Tay = V9 0" dayX', the last expression being a consequence of (38.18). From (50.9) it follows that ./g does not involve onsequently the second term of (50.8) is identically zero. Hence for the equations (50.8) to hold we must have g*’day = 0, which by (40.11) is the condition that the mean curvature be cqual to zero. Hence we have [50.1] A minimal surface is an extremal for the integral of area. The determination of whether this is the minimum area for a given contour involves derivatives of higher order and is an important problem in the caleulus of variations. However, any surface satisfying theorem {50.1] is called a minimal surface, Lagrange using the equation of a surface in the Monge form (see §10) derived as the differential equation of minimal surfaces (see §39, Ex. 1) (1+ Piru — 2pmre + (1 + pire = 0; in this notation the left-hand member is the expression for mean curva~ ture. Lagrange raised the question of finding the minimal surface for a given contour. Plateau gave a physical realization of this problem by means of a glycerine film, it heing a consequence of surface tension of such a film that the surface would assume the form having a minimum area. Accordingly, this problem proposed by Lagrange is now known as the Plateau problem. The mathematical solution of this problem for a given contour has been the subject of study by mathematicians up to the present time. From (40.11) it follows that when minimal curves on a minimal surface are coordinate, these curves form a conjugate net. In this case gu = 92 = dx = 0. From the first two of these equations it follows from (28.1) that (60.10) {ho} > {fh} From these results we have from (38.18) that. 5 = 0 and con- pat ue $50] MINIMAL SURFACES 291 sequently (50.11) a = Ui + Ui, where Uj and U; are functions of uj and uz respectively. ‘These func- tions must be such that auiy _ , ¥ (24) =9, aui\! (50.12) 2 ( aU") which are the conditions that gu = ga = 0. Since the first of equations (50.12) expresses that the sum of the squares of three quantities is zero, and since the three quantities jaa =u), 51 + u*), and u! possess this property, it follows that the most general solution of the first of equations (60.12) consists of the above expressions multiplied by an arbitrary function of u', say, F(u'). Since the same argument applies to the second of equations (50.12), we have for equations (50.11) satisfying the conditions (50.12) ga sfa =u )FQu!) dul + af (1 =u )HOe) au’, (50.13) 2? = 3 J G+ 2 yr) du — 3 | (+e )H(e’) de’, 2 = ferceyae' + f emceyae. From these equations we have that the linear element of the surface is (60.14) ds’ = (it ulwyFw)H(u’) du du’. From this it is seen that for the surface to be real u' and w* must: be conjugate imaginary and H must be the function conjugate to F. It was in order to effect this result that the negative sign was used before the second integral in the second of equations (50.13). From (50.13) we have for the direction cosines of the normal to the surface _ (el +2), 10d = w+), a? 7 [fue from whieh one finds that the lincar element of the spherical representa- tion is (50.15) -X*, X*, x? 2 Adal du (60,16) ds’ = Apu 292 SURFACES IN SPACE (Cx. 1¥ Also by means of (38.13) one finds (50.17) dy = —F(u'), de =0, dn = —H(w’). From (50.14) and (50.17) one finds that the differential equations of the lines of curvature and of the asymptotic lines are respectively (60.18) Pw)du® — H(u)\du" = 0, (50.19) FQ) + HOE) = 0. From the form of these equations we have [50.2] When the minimal lines of a minimal surface are coordinate, the equations of its lines of curvature and of its asymptotic lines can be found by quadratures. Equations (50.13) are known as the equations of Enneper.* From these equations it is seen that the problem of minimal surfaces is that of functions of a single complex variable (see Ex. ). Weierstrass,t who also derived equations (50.13), remarked that these equations can be put in a form free of all quadratures. This is done by replacing F(u') and H(u’) by f”"(u') and ’”(u’) respectively, where the primes indicate differentiation, and then integrating by parts. ‘This gives the equations Ba pat + atro) —s0t) + LE way + EH Q2) ~ het), (6020) yeah uw! pat) = iulp) + dy — Le + Lt may +i ve) — in(u’), B= ubpt(ul) — $e) +R) — WO). It is clear that the surface so defined is real when f and h are conjugate imaginary functions of the conjugate imaginary coordinates u’ and wu’, In this case (50.20) may be written a= wi — uP yal) + uf’) — 2fee')], (50.21) f= MA tu yew) — auf’) + fu’), at = Huf"(w') — 2'u')], * 1864, 2, p. 107. + 1866, 1, p. 619, §50] MINIMAL SURFACES 293 where in these equations 9 denotes the real part of the expression following 9. Equations (50.20) are of particular value in the study of algebraic surfaces, for it is evident that when f and h are algebraic on the elimination of u' and «° from these equations the resulting equation will be algebraic in 2’, 2’, and 2°, The converse of this is also true.* EXERCISES 1. A minimal surface is a surface of translation (sce §42, Ex, 1); when its equa- tions are of the form (50.18) a necessary and sufficient condition that the gener- ating curves be congruent is that 1 1 Fa) ~an(- :) w w 2. When equations (50.11) are equations of a equations al snvfaee, sa alee are the al = cui + e*Ul G=1,2,3), where (= \/=Tand ais aconstant, ‘These minimal surfaces as @ takes different values ure called associate minimal surfaces; the normals to these surfaces at points with the same coordinates u® are parallel 3. When ain Ex. 2is equal to +/2, the surface is called the adjoint of the surface with the equation (30.11); the Hines of revature on cither surface correspond to the asymptotic lines on the other, and the tangents to the curves with the same equation g(u!, ut) = 0 on the two snefaces are orthogonal, 4. If xf denote the coordinates of the adjoint surface of the surface with the equation (50.10), thé equations of Ex, 2 may be weitten rhs cosa +sinaz ; the plane determined by the origin of coordinates, a point ? on a mininul surface, and the corresponding point on its adjoint contains the corresponding point on every associate surface and the locus of these points is an ellipse with center at the or . From §44 Ex, 11 it follows that for a mi ordinate curves form an isometri of the harmonic equation imal surface Ay 2! = 0; when the eo- rthogonal net, we have that 2° are solutions ae | oe au oui from whieh it follows that for any real minimal surface Basie tity) fir i), where f and ff are conjugate imaginary functions, which is in accord with equa- tions (50,13). * 1909, 1, p. 261. 294 SURFACES IN SPACE [cn Iv 6. When in equations (50.13) ia bew beé Pw =F Hoe) = 5S where a and b are constants, and one puts u! = e#*¥ and ut = ef, one ob- tains equations of the following form x! = b(cos a cosh at cos a? — sin w sinh w sin a) z* = b(cos a cosh a sin a? + sin a sinh w cos a), x= b(- @ cosa + asin a), which are equations of minimal helicoids (see §24, Ex. 6) 7. When a = Oin Ex. 6, the surface is the eatenoid (see §24, Ex. 4); when « = /2, the surface is the skew helicoid (sce §24, Ex. 7) 8. If two minimal surfaces correspond in such manner that at corresponding points the tangent planes are parallel, the minimal curves on the two surfaces, correspond; for two such surfaces the locus of the point whieh divides in constant ratio a line-segment joining corresponding points is 2 minimal surface, 9. The spherical representation of the lines of curvature of a minimal surface is an isometric orthogonal net (see §43, Hx. 11 and (46.5) 10. If one family of the lines of curvature on a ininimal surface are pline curves, those of the other system are plane curves also (see §46, Ex.4; $47, Ex, 15) HW. Show that the surface 2 = au! + sin wt cosh w, 2 = 1? +0 cos w sinh uw, B= VT— at cos wi cost is minimal and that its lines of c 12. The'surfaecs of center of a minimal surface are applicable to one another and to the surface of revolution of the evolnte of the catenary. 18, ‘The surface for which F = H = const., aay 3, is called the minimal surface of Enneper; it possesses the following propertie (a) it is an algebraic surface of the ninth degree whose equation is unaltered when z!, 2%, zt are replaced by z*, x!, —2* respectively; (b) it moots the planc 2° = Qin two orthogonal str: (©) if we put wl = a! — 7H, the equations of the surf ruture are plane curves. he lines; eare ga 3at+ sata — am, t= Bat saat = Ba? - and the curves i = const., a = const. ate the lines of eurvature; (d) the lines of curvature are rectifiable unicursal curves of the third ord they are plane curves, the equations of the planes being rand D+ das — 38a — 2a 0, at — ate — 3a? — 242° (@) the lines of curvature are represented on the unit sphere by a double family of circles whose planes form two pencils with perpendicular axes which arc tangent, to the sphcre at the same point; $50) MINIMAL SURFACES 295 (4) the asymptotic lines are twisted cubies; (g) the sections of the surface by the planes 2! = 0 and 2 = Oarc cubies, which are double curves on the surface and the locus of the double points of the lines of curvature; (h) the associate minimal surfaces are positions of the original surface rotated through the angle —a/2 about the 2°-axis,where a has the meaning of Ex. 2. ‘the surface is the envelope of the plane normal, at the midpoint, to the jo of any tivo points, one on each of the focal parabolas 0, stadt at a, t= 0, Qa 1; oat aut; the planes normal to the two parabolas at the extremities of the join are the planes of the lines of curvature through the paint of contact of the given plane. BIBLIOGRAPHY ‘This bibliography contains only the books and memoirs which are referred to in the text. 1776 (1) Muusyier, J. B. Mémoire sur la courbure des surfaces. Published in 1827 1848 1853 1856 1859 1860 1804 1865 1366 1869 1870 1887 1888 1889 ay a ay a a) a) a @) a a aw (2) 3) a) qa) ay qa) 1785, in Acad. des sci., Paris, Mém. de math. et de phys. présentés par divers sgavans, vol. 10, pp. 477-510. Gauss, K. F. Disquisitiones generates circa superficies curvas. Werke, Géttingen, 1880, vol. 4, pp. 217-258. Bonnet, O. Mémoire sur la théorie générale des surfaces. Paris, Ecole palytechnique, Jour., vol. 32, pp. 1-146. Bowner, 0. Sur une propriété de maximum relative @ la sphére. Nou- velles annales, vol. 12, pp. 433-437. Murwarpr,G. Sulla teoria generale delle superficie. Istituto Lombardo, Giornale, n. s., vol. 9, pp. 385-398. Lam, C. Lecons sur les coordonnées curvilignes et leurs diverses applica- tions. Paris, Mallet-Bachelier. 368 pp. Bonnet, O, Mémoire sur Vemploi d’un nouveau systéme de variables dans Vétude des propriétés des surfaces courbes. Jour. de math. pures, ser. 2, vol. 5, pp. 153-266. Baurnau, E. Ricerche di analisi applicata alla geometria. Giornale di matem., vol. 2, pp. 267-282, 207-306, 331-330, 355-375. (This and 1865 (1) are also in his Opere, vol. 1.) Ennepen, A. Analytisch-geometrische Untersuchungen. Zeits. far Math. und Physik, vol. 9, pp. 96-125. Barras, E. Ricerche di analisi applicata alla geometria. Giornale di matem., vol. 3, pp. 15-22, 33-41, 82-91, 228-240, 311-314. Wererstrass, K. Untersuchungen uber Flachen deren mittlere Krim- mung wberall gleich Null ist. Preussische Akad., Monatsberichte, pp. 612-625, (Also in his Mathematische Werke, vol. 3.) Cunisrorrer, E.B, Uber die Transformation der homogenen Differential- ausdricke zweiten Grades. Jour. fr reine und angew. Math., vol. 70, pp. 46-70. BELTRAMI, Teoria fondamentale degli spazii di curvatura costante. Annali di matem,, ser. 2, vol. 2, pp. 232-255. Copazz1, D. Sulle coordinate curvilinee @'una superficie e dello spazio. Annali di matem., ser. 2, vol. 2, pp. 269-287. Dis, U. Sopra un problema che ei presenta nella teoria generale delle rappresentazione geograficke di una superficie su di un’ altra. Annali di matem., ser. 2, vol. 8, pp. 260-203. Darpoux, G. Legons sur la théorie générale des surfaces. Paris, Gauthier-Villars, vol. 1 Leumuven, M.M.C. Sur les lignes asymplotiques et leur représentation sphérique. Bulletin des sciences math., vol. 12, pp. 126-128. Darpoux, G. Lecons sur la théorie générale des surfaces. Paris. Gauthier-Villars, vol. 2. 296 BIBLIOGRAPHY 297 1901 (1) Ricci-Cursasrno, G. and Levi-Civrra T. Méthodes de caleul différ- ential absolu et leurs applications. Math. annalen, vol. 54, pp. 125 201, 608. 1902 (1) Brancnr, L. Lezioni di geometria differenziale, 2 ed. Pisa, Spoorri. vol. 1. 1904 (1) Fixe, H. B. A college algebra. Boston, Ginn. 595 pp. (2) Ricet-Cunsastro, G. Direzioni ¢ invarianti principali in una varielé qualunque. R. Istituto Veneto. Atti, vol. 63, pp. 1233-1239. 1907 (1) Bécuer, M. Introduction to higher algebra. New York. Macmillan. 321 pp. 1909 (1) Exsennanr, L. P, Treatise on the differential geometry of curves and surfaces. Boston, Ginn. 474 pp. 1910 (1) Vestn, O., and Youna, J. W. Projective geometry. Boston Ginn, vol. 1. (2) Darnoux, G. Legons sur les syst2mes orthogonaus, . . . Paris, Gauthier- Villars. 567 pp. 1917 (1) Levi-Crvira, T. Nozione di parallelismo in wna varield qualunque ¢ conseguente specificazione gevmetrica della curvatura riemanniana, Cireolo matem. di Palermo. Rendie., vol. 42, pp. 173-205. 1922 (1) Brancut, L. Sul parallelismo vincolato di Levi-Civita nella metrica degli spazi curvi, Acad. delle sci. fis. ¢. matem., Naples, Rendic., ser. 3, vol. 28, pp. 150-171. 1925 (1) Goursat, E. Cours d’analyse mathématique, 5. éd. Paris, Gauthier- Villars, vol. 1. 1926 (1) Exsexnanr, L. P. Riemannian geometry. Princeton university press. 262 pp, (2) Syxoe, J. L. On the geometry of dynamics. Royal Soc. London, Phil. ‘Trans. A, vol. 226, pp. 31-106. 1927 (1) Frye, HB. Calculus. New York, Macmillan, 421 pp. (2) Levi-Crvira, T. The absolute differential calculus. Tondon, Blaekie. 450 pp. 1931 (1) McConwent, A. J. Applications of the absolute differential calculus London, Blackie. 318 pp. Index 43, a4, a, 70, 72. ai, 12, Angle, of curves in a surface, 130-132, 187, 145; of vectors, 133. Applicable surfaces, 215; see Isometric surfaces. Arc of a curve, 9 Area, element of, 136; minimum, 200. Associate, vector, 194, 250, 252; mini- mal surface, 293, Asymptotic directions, 237, 241. Asymptotic Knox, 287-289, 213; eo- ordinate, 237, 238, 242, 260, 266; orthogonal, 288; straight, 237, 267; geodesic, 248; spherical representa- tion of, 260, 266, 267; plane, 249; osculating plane of, 237; tangential coordinates, 264, 265. 84, 18. Beltrami, 158, 189, 208, 275: Bertrand curves, 30. Bianchi, 194. Bianchi identities, 121. Binormal of a curve, 19. Bonnet, 190, 273. Calculus of variations, 200. Catenoid, 128, 242, 204, Center of curvature, of curve, 18; ofa curve in a surface, 227; prineipal, of a surface, 226. Characteristies of a family of surfaces, BA. Christoffel symbols, for space, 98, 102, 121; as components of a tensor, 100, 102, 108; for a surface, 149, 153, 154; relations between for a surface and its spherical representation, 257, 260. Circle, of curvature, 18; osculating, 19; superosculating, 24 Circles, orthogonal sphere, 267. Codazzi equations, 219, 230, 236, 238. system, on the Combescure transformation of curves, 29. Complementary surfaces, 275. Conformal correspondence, of two surfaces, 201-205; of a surface and its spherical representation, 255; of a surface with itself, 203; of a plane with itself, 204; of a sphere with the plane, 204; of a sphere with itself, 204; of a surface of constant curva- ture with the plane, 288. Congruenee of curves, 78. Conjugate directions, 231, 241. Conjugate net, 232, 243; coordinate, 282, 235, 236; orthogonal, 232; of plane curves, 286; spherical represen tation, 260; tangential coordinates, 263; isometric-, 235, 286; mean-, 240, 243, 260. Conjugate systems in correspondence, 243, Cone, 53, 57. Conoid, right, 50, 53, 129, 149, 221, 266. Coordinates, in space, 63-68; cartesian, 63, 70, 88; cylindrical, 83; sce Polar coordinates, ‘Tangential coordinates; ina surface, 48; 2!, 2? as, 137, 221. Coordinate curves, in space, 69; in a surface, 46, 127. Coordinate, net, 49; surface, 68, 81 Contraction of indices, 95. Contravariant, index, 89: tensor, 89, 90, 91; vector, 7, 82, 128; compo- nents, 85, 88, 127, 134. Covariant differentiation, 107-112; of sum, difference, outer and inner product of tensors, 111; of ais, a¥4, 6, 110; of gaa, 9%, fas, €8) 151 Covariunt, index, 89; tensor, 89, 90, 91; vector, 84, 127; components, 85, 86, 88, 127, 134. Cubic, twisted, 5, 7, 8, 16, 44 Curvature of a curve, 16; radius of, 16; center of, 18; cirele of, 18; constant, 209 300 19, 23, 24, 28; first, 22; socond, 22; geodesic, see Geodesic curvature Curvature of a surface, total, 151; see Gaussian curvature, Mean curvature. Curvature, of a quadratic form, 151; of a triangle, 183. Curvature, normal, of a surface, 224; radius of, 224, 227; eonter of, 224; principal radii of, 225; principal centers of, 226. Curvature tensor, see Riemann tensor. Curve, definition, 3; plane, 4; twisted, or skew, 4; length of, 9; are of, 9; minimal, or of length zero, 10; form of, 26; of constant curvature, 19, 24, 28; of constant torsion, 24, 28, 20, 30, Curvilinear coordinates, 47, 69. Cylinder, 8, 53, 61, 149. dup, d28, 215, 216, 221, 234, 262. 8s, 8°, 84, 8, 25, 32 %, 125. D, D’, DY, 216, dre, dilen 01), Ofer, 2), 155, 158, 159, 160. Axe, 158, 159. Darboux, 115, 173, 233. Developable surface, 54, 57-61, 152; tangent planes, 147, 148; isotropic, 61; polar, 61, 149; rectifying, 61, 248; edge of regression, 61; tangent sur- face of a curve, 57; isometric with a plane, 147; Gaussian curvature, 150; geodesic in a, 178; of normals to a surface, 229; lines of curvature, 229, Differential parameters, of the first order, 155, 156, 159, 160; of the seeond order, 158, Dini, 210. Dini, surface of, 287. Direction cosines, of tangent, 12; gf principal normal, 18; of binormal, 20, 23. Divergence of a vector, 113, 155. Dupin indicatrix of a surface, 241. E, F, G, 126. esse, ef, 6, 93, 94. eign, oF, 94, 97. INDEX apy 628, 134, 138. cap, €°9, 135, 137, 138, 151. Edge of rogression, 56, 61 Flement, of area, see Area; Ii Linear Element. Ellipsoid, 50; see Quadries, central. Elliptie point of a surface, 242. Enneper, 248, 292. Enneper, minimal surface of, 204 Envelope, of a one-parameter family of surfaces, 54-58; edge of regression, 56, 62; of characteristi¢s, 56; of a one~ parameter family of planes, 57, see Developable surfaces; of a family of spheres, 62. Equations, parametric, of a line, 1; of acurve, 3; of a surface, 46. ‘valent representation of surfaces, E Ruler, equation of, 240; equations of, 177, 289. Evolute of a curve, 36, 38. Family, one-parameter, of surfaces, 54; characteristic, 54; of planes, 57; of spheres, 62; of curves in a surface, 138-141; of gepdesies, 174. rinula, of Green, 190, 192: of Liou- ville, 193. Frenet formulas, 25, 27; in general coordinates, 106; in a surface, 199. Fundamental quadr: form, of space, 70; first, of a surface, 124; second, of a surface, 215, Fundamental tensor, of space, 91; first, of a surface, 125, 262; sceond, of a surface, 215, 262. 9, 101, 125. os, 0°, 101. Goss 9°, 123, 125, 137, 151, 262. +r, 0. Gauss, 45, 151, 174, 184, 216, 219, 225, 252, 24. Gauss, equations of, 216; equation of, 219. Gauss-Bonnet theorem, 191. Gaussian curvature of a surface, 151. 154, 193, 225, 255 INDEX 301 Generator, of a developable surface, 59; of a surface of translation, 236, Geodesies, 170-179, 187, 246; equations of, 171, 179, 197; coordinate, 173; plane, 248; in a surface of Liouville, 176 Geodesic correspondence of two sur- faces, 205-211; of constant curvature, 209. Geodesic curvature, 186-193, 199, 201, 243-243; conter of, 246, 275; radius of, 276; curves of constant, 192, 267. Geodesic, circles, 180, 186; ellipses and hyperbolas, 185, 186, Geodesic parallels, 174, 200. polar coordinates, 180-182, torsion, 247-249, Geodesic triangle, 183, 184, 200. Gradient, 84. Group property, 68. hr, 258. hag, hf, 258, 256. Helicoid, 58, 129, 218; parameter of a, 129; isometric with a surface of revo- lution, 160; surfaces of center, 276; spherical, 287; pseudospherical, 287; minimal, 294. Helicoid, skew, 129, 145, 165, 242, 294 Helix, cireular, 14, 16, 19, 23, 30, 34, 375 conical, 15; cylindrical, 15, 19, 24, 28, 37, 38, 149, 178. Hyperbolie point of a surface, 242. Hyperboloid, 50, 242; rulings, 242; see Quadries, central. Index, dummy, 2; free, 2; contra- variant, 89; covariant, 89; lowering, 95; raising, 95. Indieatrix, of Dupin, see Dupin; spherical, of tangent, 17; of binormal, 2, Inner product of tensors, 95. Intrinsic, equations of a curve, 31, 147; geometry of a surface, 146; deriva- tive, 195. Inversion, 233, 236; preserves lines of curvature, 233, Involute of a curve, 35, 37, 20. Tsometrie surfaces, 147, 150, 166-169, 171, 197, Isometric, orthogonal net, 161-165; coordinates, 161; -conjugate net, 235, 236. K, 151, 154, 193, 225. Km, 225. «, 16. Ky) 187, 245. Kronecker deltas, 25, 63. Lagrange, 290, Lamé, 158. Lelicuvre, formulas of, 266. Levi-Civita, 196, 199, 251 Line of curvature, 228, 220; equation, 228, 277; coordinate, 280; conjugate, 232, 233; normal curvature of, 229; geodesic torsion of, 248; geodesic, 2S; two surfaces intersecting in a, 248; spherical representation of, 2553 osculating plane of, 259; plane, 230, 249, 260, 266, 268, 276; spherical, 24 under an inversion, 233. Linear clement, of a curve, 9; of space, 70; of a surface, 124; of the spherical representation, 252. Linearly independent (constant coefi- cients), 262, Lines of length zero, see Minimal lines. Lines of shortest length, 175. Liouville, formula of, 193; surface of, 176. Loxodromic curve, 145. Mainardi, 219, ‘MeConnell, 194. Mean-conjugate net, 240, 243, 260. Mean curvature of a surface, 225. Meridian of a surface of involution, 49. Metric tensor, of space, 91; of a surface, 124, 125, Meusnier, theorem of, 224. Minimal curve, 10, 16; ona surface, 124; ‘ona sphere, 154. Minimal surface, 238, 249, 288-295; asymptotic lines, 238, 260, 292; minimal lines, 238, 291; spherical 302 Tepresentation, 255, 260, 204; heli- Coidal, 242, 294; of revolution, 242; Tines ‘of curvature, 243, 292, 204; Algebraic, 208; surfaces of center, 204; adjoint, 203; associate, 293; of En- Reper, 294. Monge, 46, 290. Normal plane to a curve, 12. Normal section of asurface, 228; prinei- Pal, 225, 229; radius of curvature of, 228, 294; center of curvature of, 227 Normal to a curve, 18; sce Principal normal, Binormal. Normal to a surface, 58, 213, 229. Order of contact, 13, 16. Orthogonal net, 131, 141, 144, 136, 205; Coordinate, 131, 193; of geode 178; sce Isometric orthogonal net. rthogonal coordinate surfaces, 81. Orthogonal trajectories, of a family of curves, 140; of a family of geodesics, 174, 175, 180. 'sculating circle of a curve, 19. Oseulating plane, 13, 16; equation of, 14; stationary, 23; through a fixed Point, 28; tangent plano of tangent Surface, 51; edge of regression, 7; ©f asymptotic line, 237; orthogonal trajectories, 27, 148. Osentating aphere, 13, 44. Outer product of tensors, 04, 97. Paraboloid, 145, 165, 242; fundamental quantities, 166, 222; rulings, 242: lines of curvature, 231, 236; of nor- mals to a ruled surface, 221. ‘arullel curves on a surface, 157, 158, 174, Parallels, geodesic, 174. . Parallel surfaces, 272, 273, 276; funda- Mental quantities, 272; lines of curvature, 272, 276; of surface of Constant curvature, 273; of surface Of revolution, 276; asymptotic lines, 276, ‘Arallel vectors, in space, 107, 113; in A surface, 196-200; 249-252. INDEX Parametric equations of a line, 1; of a curve, 3; of a surface, 46. Partial differential equations of the first order, 14-HS, 122; mixed sys- tem, 118. Plane, 1, 217; surface isome 150. Plane curve, 4, 6; curvature, 17; in- trinsie equations, 147, Plateau problem, 290. Point of a surface, ordinary, 51; singu- lar, 51; clliptic, 242; hyperbolic, 249; umbilie, 225, 227. Polar coordinates, 64, 82, 88, 102, 107, 113. Poler linc of a curve, 37. Polar, surface of a curve, 41, 52; de- velopable, 61, 149. Principal center of normal curvature, with a, al direetions, for a tensor tas, 144; of normal curvature, 224. Principal normal of a curve, 18, 24, 30, 31 Principal radii of normal curvature, 223. Pscudosphere, 284 Peondosphorical surface, 277, 281-284, 286; of revolution, 281-284; lines of curvature, 285, 280; asymptotic lines, 242, 286; geodesics, 184, 287, 288; of Dini, 287; surfneos of center, 287 Quadratic form, definite, 151, 152: indefinite, 151, 152, 155; see Funda- mental quadratic form. Quadries, central, equations, 49; funda- mental quantities, 165, 221; lines of curvature, 231, 236; asymptotic lines. 242. Pity Raise, 100-108, 150, 151, 207. o, 16. Tectifying developable, 61, 248. Representation, equivalent, 20%; see Spherical represent: Riccati equation, 33, 34 Ricei, tensor, 108, 122, 13% 112, identities, IN] Riemann, symbols, 101; tensor, 101, 103, 120-122, 150, 153. Scalar, 78, 91; relative, 92; density, 92. Singular, point on a surface, 51, 125 curve on asurface, 125. Skew curve, 4. Sphere, 49, 53, 128, 145, 154, 217, 258, 267. Spheres, family of, 62, 275. Spherical curve, 16, 43, 44. Spherical indicatrix, of the tangents to a curve, 17; of the binormals of a curve, 22. Spherical point of a surface, 225. Spherical representation of a surface. 252-260; fundamental quantities, 253, 260; lines of curvature, 255; asymp- totie lines, 280, 267; conjugate net, 260; area of closed portion, 254: isometric net, 267. Spherical surface, 277-281; parallels to, 273; of revolution, 278-281 ; geodesics, 287; lines of curvature, 285, 286: asymptotic lines, 286. Spiral surface, 146, 179, 231. Stereographic projection, 206. Sunmation convention, 1. Superosculating circle, 24, Surface, definition, 2; imbedded in space, 128. Surfaces of center, 268-272; fundamen- tal quantities, 271; Gaussian curva- ture, 272; asymptotic lines, 276; lines of curvature, 276. Surface of constant mean curvature, 278. Surface of constant Gaussian curva- ture, 170, 186, 287; see Spherical surface and Pseudospherica! surface. Surface of Liouville, 176, 180. Surface, of positive curvature, 226; of negative curvature, 226, Surface of revolution, 49, 149, 160; ineridian, 49, 165; fundamental quan- tities, 128, 218; Gaussian curvature, 153; loxodromic curve, 145, 192; geodesics, 178, 192; lines of curva- ture, 230, 236, 259; asymptotic lines, 303 242; of constant curvature, 169; surface isometric with a, 160, 165, 169, 179, 277; principal radii, surfaces of center, 275; p: surfaces, 276. Surface of translation, 236, 24 generating curves, 236. Surface of Weingarten, 276, 27. Symmetry with respect to, a point, 24, 218; a plane, 24. Synge, 194. 7, 22. ty 247. Tangent, to a curve, 11; positive seuse, 11; tan surface, 50, Tangent plane toa surface, 50; distance to, 53; osculating plane of asymptotic line, 237, Tangent snrface of a eurve, 38; edge of regression, 40; form of, 40; tangent plane, 51; gencrators, 38; lines of curvature, 230; linear element, 128. Tangential coordinates of a surface, ‘Tehebychef net, 160, 200, 236, 242, ‘Tensor, contravariant, 80, 90; co- variant, 80, 90; mived, 89, 90; orde 90, 91; zero, 91; symmetric, 92, 9 skew-symmetric, 92, 93; relative, 03, $4; addition, subtraction, multiplica- tion af fensars, 94; innor praduet, 95; outer product, 94, 97; quotient. law, 97. ‘Torus, 227. ‘Torsion, of a curve, 22, 24, 28, 29, 30; of ana 248; of a geodesic, 248. ‘Tractrix, 29; surface of revolution of a, 284; helicoid whose generating curve isa, 287. ‘Transform of a function, 78. ‘Transformation of coordinates, in space, 04-08; inverse, 65; linear, 63, 79; product, 67; in a surface, 124: positive, 133. Transformation, of Combescure, 29; by reciprocal radii, see Inversion. constant, 28, nptotic line, 304 ‘Triply orthogonal family of surfaces, 82, ‘Twisted curve, 4; cubie, 5, 7, 8, 16, 44; of nth order, 5. Umbilical point, of a surface, 225, 227; of a quadric, 227. Vector in space, contravariant, 7 covariant, 84; length, 80, 85; unit, 80; significance of components, 82, 86; contravariant and covariant com- ponents, 85, 88. Vectors in space, independent, 79, 88; angle of two, 81, 85; perpendicular, 81, 83. INDE! Vector in a surface contravariant, 126; covariant, 127; length, 128; unit, 128; contravariant and covariant com- Ponents, 127; signifieance af com- ponents, 134, Veetors in a surface, angle of, 133; perpendicular, 134, 135, 136; asso- ciate, 194, 250, 252; parallel, 196-200, 249-252. W, 260. Wooysr 211 Weierstrass, 202. Weingarten, surface of, 276; lines of curvature, 277.

You might also like