oe DIFFERENTIAL ca & INTEGRAL CALCULUS VOLUME II DIFFERENTIAL & INTEGRAL CALCULUS F R. COURANT BLACKIEDIFFERENTIAL AND INTEGRAL CALCULUS by R. COURANT Vowume I In this volume Professor Courant dis- ‘cusses the more advanced parts of the Calculus, such as the theory of functions of several variables and multiple integrals, in a masterly manner, attractive and at the same time reasonably rigorous. Chap- fers on Calculus of Variations and Com- plex Variable have been specially written for the English edition. Print in Gre Brain From BLACKIE’S List ‘The Mathematical Discoveries of Newton. By H. W. Turnbull, MA, RS. 6s, 6d, net. The Absolute Differential Calculus. (Cal- clus of ‘Tensors.) By ‘Tullio Levi-Civit Professor of Rational Mechanics in the Uni versity of Rome, Fellow of R. Accademia Nazionale dei Lincei. Edited by Dr. Enrico Persico, Authorized translation by Miss M. Long, late scholar of Girton College, Cam- bridge. 730s. net. ‘Theory and Application of Infinite Series. By Dr. Konrad Knopp, Professor of Mathe- matics in the University of Tubingen, ‘Second Edition, 50s. net. ‘The Calculus of Observations. A Treatise on Numerical. Mathematics, By E. T. Whittaker, LL.D. ScD, PRS; and George Robinson, M.A., BSc. Fourth Edition. 355. net. A Simplified Presentation of Einstein’s Unified Field Equations. By Tullio Levi-Civita. Authorized translation by Joha Dougall, M.A., D.Sc, F.RSE, 3s. met. Mathematical Notes. By C. F. G. ‘MacDermott, M.A. Intended to help in th preparation and revision for such examina tions as ‘Mathematics for Science” at “A” level, the Qualifying Examination for the Mechanical Sciences ‘Tripos, “ Higher “Mathematics” for the Services Examination, etc. 38. 6d. Examples in Applied Mathematics. By F, W. Kellaway. For students at the G.CE, Advanced Level and O.N.C., &c. ‘Second Bdition 7s. 6d. net.DIFFERENTIAL AND INTEGRAL CALCULUSBLACKIE & SON LIMITED 16/8 Witt TV Street, Canine Crm Lavon, WiC "7 Sanbope Sree GUAssow BLACKIE & SON (INDIA) LIMITED ‘ols Pore Suet, Bown BLACKIE & SON (CANADA) LIMITEDDIFFERENTIAL AND INTEGRAL CALCULUS BY R. COURANT ‘rofestor of Mathexutica in New York Universicy ‘TRANSLATED BY E. J. McSHANE Professor of Mathematica in the University of Virginia VOLUME II BLACKIE & SON LIMITED LONDON AND GLASGOWion pba 2996 ‘Reprinted S997, 1940 94%, 1948 med), 2948.14 3045. 967 28a 900, 90.1981 8a 1055, 1080) roadie, fy. BBE ee soakenen, 95 ces 2998 ‘Printed in Ort Britain by Blache Son i, GlasgoePREFACE ‘Tho present volume contains tho more advanced parts of the differential and integral calculus, dealing mainly with functions of several variables, As in Volume I, I have sought to make definitions and methods follow naturally from intuitive ideas and to emphasize their physical interpretations—aims which are not at all incompatible with rigour. I would impress on readers new to the subject, oven more than I did in the preface to Volume I, that they are not expected to read book like this conseoutively. ‘Those who wish to get a rapid grip of the most essential matters should begin with Chapter II, and noxt pass on to Chapter IV; only then should they fill in the gaps by reading Chapter IIT and the appendices to the various chapters. It is by no means necessary that they should study Chapter I systematically in advance. ‘The English edition differs from the German in many details, and contains a good desl of additional matter. In particular, the chapter on differential equations has been greatly extended. Chapters on the calculus of variations and on functions of a ‘complex variable have been added, as well as.» supplement on real numbers. Thave again to express my very cordial thanks to my German publisher, Julius Springer, for his generous attitude in con- senting to the publication of the English edition. I have also to thank Blackie & Son, Ltd., and their staff, especially Miss ‘W. M. Deans, for co-operating with me and my assistants and relieving me of a considerable amount of proof reading. Finally,vi PREFACE I must express my gratitude to the friends and colleagues who have assisted me in preparing the manuscript for the press, Treading the proofs, and collecting the examples; in the first place to Dr. Fritz John, now of the University of Kentucky, and to ‘Miss Margaret Kennedy, Newnham College, Cambridge, and also to Dr, Schénberg, Swarthmore College, Swarthmore, Pa, R. COURANT. New Rocurutz, New Your. ‘March, 1936.CONTENTS Cuarre 1 PRELIMINARY REMARKS ON ANALYTICAL GEOMETRY AND VECTOR ANALYSIS 1, Rectangular Co-ordinates and Veetors- 4 2 The Ame fw rang, he Vote of Teen, he Yetor ‘Multiplication of Veotors = - 2 8. Simple Toor on Dtominanta of the Sond and Thin Order =~ » 4, Afine Transformations and the Multiplication of Determinanta - 27 (Onarren I FUNCTIONS OF SEVERAL VARIABLES AND ‘THEIR, DERIVATIVES 1. The Concept of Funotion in the Caso of Several Variables =~ 2 Continuity = = = = = 2 ee 3, The Derivatives of « Function - = 44. The Total Diflerntal of » Punotion nd ita Geometrical Meaning 5 Fustions of Wunctions (Compound Function) an the Into- duction of New Independent Variables - 6. Tho Mean Valoe Thor sn Tayl's There for Fuston of Several Variables - = =~ ao 11. The Application of Veotor Methods - = + + + + Ss sere ee APPENDIX 1. Tho Prinsiple of the Point of Accumlation in Sorel Dinen sions and ita Applications ~~ : 2. The Concoptof Limit for Functions of Several Variables = R Homogeneous Funetiont’= = = = = == “ 95, 101 108CONTENTS Cuarram IIT DEVELOPMENTS AND APPLICATIONS OF THE ‘DIFFERENTIAL CALCULUS 1. Implicit Bunotions == foe 2, Curve and Surin in Implicit Form > : 2: Spies of Festus, rvnoratons, a Mapas 4 Applications ~~ 5. Belin of Carve, Heil of Sra, and thes Hartge 6. Maxima nd Minima = APPENDIX 1, Suffolent Conditions for Extreme Values - 9 - = ++ 2, Singular Pointe of Plane Curves = - = = = = = 8, Singular Points of Surfaces - 4 Guerin Intern Bun? aod Tagrngy Reprostation of ‘the Motion of a Biuid = si . 6. Tangential Repreceatation of a Cheol Curve 7} + Carre IV MULTIPLE INTEGRALS 1. Ordinary Integrala as Funotions of a Parameter = 2, he tga ef Cantino Pension ovr» Repo th Pane or of Space 3. Reduction of the Mltiplo Integral to Repeated Single Integrals 4 Trunaformation of Mall Intagral 5, Improper Integrals = = 6. Geometrical Applications = = + 17. Physical Applications == = = SESESB APPENDIX 1, The Rxistenos of the Multiple Integral - 2 General Formula for the Area (or Volume) of « Region bounded by Sagmena of Stright Tine or Plane Aree (Gali Formula), The Polar Planimeter =~ 5 {. Volumes and Aress in Space of any Namber of Dimensions 4 Improper Intgrls ws Functions of « Parameter = + 5, The Fourier Integral = 7 ore 6. The Bnlerian Integrals (Gamma Function) aoa Page Mm 12 133 159 169 183 aeeee~ CONTENTS: 11. Differentiation and Integration to Fractional Order. Abel Integral Equation - = a 8, Note on the Dlnition ofthe Aree of a Curved Surfs > Curren V INTEGRATION OVER REGIONS IN SEVERAL ‘DIMENSIONS: 1, Line Integrala = E Comecion betwee Lin Tntorls and Double Integra in tho Plane. (The Integral Theorems of Gauss, Stokes, and Green) 3 Tatarpretation and Applications of the Integral Theorems for ‘the Plane zi 4, Surface Integrals - © Gann’ Theorem and Green's Theres in Space 6. Stokes's Theorem in Space = 4th Coneron been Dentin ad Totton for Soveral Variables a . APPENDIX 1. Remarks on Gauss's Theorem and Stokee's Theorem = + = 2. Repreeentation of « Source-free Vector Field as Curl + = Cuarrm VI DIFFERENTIAL EQUATIONS 1. The Diwontial Bquatins of the Motion of « Parte in Thre Dimensions : 2, Ramps on tho Mechanios of Parislo = 8 Further Example of Diferetial Equations 4. Linear Diflerential Equations = - 5. General Remarks on Differential Equations 6. The Potential of Attracting Charges = : 4, Farther Example of Partial Diflerental Equations + (Caarrm VII CALCULUS OF VARIATIONS 1. Introduction = Peeenar 2 Es Dental Brain he Sit Case 560 6 £8 ix 2 ge23 397 88 ‘ 20 438 450 403 gesCONTENTS Cuarren VII FUNCTIONS OF A COMPLEX VARIABLE 1. Introduction 2 Franti of tae Theory of Puntos of Cm Vs” 3. The Intogration of Analytic Functions 4. Couchy’s Formula and its Applications - 5. Applications to Complex Integration (Contour Integration) 6. Many-valued Functions and Analytic Extension ‘Miscellanoous Examples SUPPLEMENT Real Nambers and the Conoopt of Limit - Summary of Important Theorems and Formule, ‘Answers and Hints - Index BEERS RY 569 887 8sCHAPTER I Preliminary Remarks on Analytical Geometry and Vector Analysis N [in the intorpretation and application of the mathematical facts which {form the main subjeot of this second volume itis often convenient to wae ‘the simple fundamental conospta of analytical geometry and vector tanalyuis, Hence, oven though many reeders will already have a certain knowledge of these subject t sems advisable to summarize their elements ine brief introductory chapter. ‘Thio chapter, howover, need not be studied ‘before the rest of the book is read; the reader is advised: to refer to the facte ‘collected here only when be finds the ‘heed of them in stadying the Iter parte ‘of the book. 1, Reoranautan Co-onpntaras axp Vectors 1, Co-ordinate: Axes. ‘To fix « point in» plane oF in space, ‘sia well known, wo generally make uae of e rectangular co-ordinate ystan. In ‘tho plane we take two. perpendicular lines the warta and the y-axis in space veo take thro mutually perpendicular lines, the z-axis, the y-axis, and the z-axis, Taking the same unit of length on cash azia, we amign to each point of the plane an s-co-ordinate fmol e yooerdinate in the uraal way, or to cach point in space an ‘eco-ordinate, s y-co-ordinate, and » 2-co-ordinate (fig. 1). Conversely, ‘to every set of values (=, y) or (2: y, #) there corresponds just one point cof the plane, or of space, as tho case may be; © point i completely determined by ita co-ordinates. ‘Using the theorem of Pythagoras we find that the distance between two points (i) a0d (ay my) given by ra VRPT2 ANALYTICAL GEOMETRY AND VECTORS [Cuar. hilo the distance between the points with co-ordinates (2 yy #4) and Yo 4) is reVE— PG We ‘In setting up a system of rectangular axce we must pay attention to the orientation of the co-ordinate system. ‘in Vol. I, Chap. V, §2 (p. 268) we distinguished betwoun positive and y iz Ol ¥ le, Right-handed ayitem of eres Fig. 2—Leftsanded wate of ex negative sense of rotation in the plane. The rotation through 90° which ‘brings the postive z-axis of a plane co-ordinate eystem into the position of ‘the positive y-axis in the shortest way defines a sense of rotation. According ‘an this sense of rotation ie positive or negative, we aay that the eystam of ‘axes is right-handed or left-handed (ct. gs. 2 and 8). Tt confined to the plane. A similar distinction occurs with co-ordinate aystems 2 2 t Tr ‘Rg. 4—Righe-baded ser Fie, 4 Leland sxe in gpace. For if ono imagines oneself standing on the zy-plane with one’s hhend in the direction of the positive z-axis, it is possible to distinguish ‘two types of co-ordinate eystem by means of the apparent orientation of the oo-crdinato system in the ay-plane. If this aystem is right-handed the feystem in space ia slo anid to be right-handed, otherwise left-handed (C4-figs. 4 and 6). right-handed gystem corresponds to an ordinary right- handed screw; for if we make the 2y-plane rotate about the z-axis (in the senso presorbed by ite orientation) and simultaneously give it» motion ‘of translation along the postive 2-exs, the combined motion is obviously4 CO-ORDINATES AND VECTORS 3 that of a right-handed sorew, Similarly, a left-handed system corresponds toa left-handed sorew. No rigid motion in three dimensions can transform ‘left-handed system into a right-handed system, in what follows we shall always use right-handed systems of axes, ‘We may also assign an orientation to a system of threo arbitrary axes passing through one point, provided these axes do not all lie in one plane, just as we have done here for a system of rectangular axes. 2, Directions and Vectors. Formule for Transforming Axes. ‘An oriented lino 7 in space or in a plane, that is, @ line traversed in a definite sense, represents a direction; every oriented line that can be made to coincide with the line 1 in position and senso by displacement paral to yy itself represents the samo direction. Tt t is customary to specify a direction rela- tive to a co-ordinate system by drawing ‘an oriented halfline in the given dirco- tion, starting from the origin of the coordinate system, and on this half- = line taking the point with co-ordinates Fig, 6—The angie which «straight (6 Bs y) which is at unit distance from ‘ne makes with the anes the origin, ‘The numbers « 8, y are called the direction cosines of the direction. ‘They are the cosines of the three angles 8, 8, 8; which the oriented line 7 makes with the positive axis, yraxis, and eaxia* (¢f. ig. 6); by the distance formula, they satisfy the relation ne Uf we restrict ourselves to the zy-plane, a direetion can be speciiod by the angles 8,, 3, which the oriented line 1 having this direction and passing throogh the origin forms with the positive z-exis and y-axis; or by the direotion cosines «= co88,, 8 = co28, which satisfy the equation of Bho A lineaegment of given length and given direction we shall call a vedior; more epecifcally, a bound tector if the initial point is fixed in space, tand a free vector if the position of the initial point is immaterial. Tn the following pages, and indeed throughout most of tho book, we shall omit the adjectives free and bound, and if nothing is said to the contrary we shall always take the vectors to be frve veotors. We denote vectors by heavy type, eg. a, 8, c, x, A. Two free vectors are said to be equal if cone of them can be made to coincide with the other by displacement parallel to iteelf, We sometimes call the length of a vector ita absolute ‘value and denote it by | a |. *'The angle which ono oriented line forme with another may always be ‘taken aa being between 0 and 1 for io what follows only the cosines of such ‘angles will be considered.4 ANALYTICAL GEOMETRY AND VECTORS [Cuar. 1 from the initial and final pointe of «vector @ we drop perpen ical on an orinted line L wo obtain an oriented segment on T core- sponding to the veclr. If the orientation of this segment isthe samme as that of hwo eal ie length the component of © in the dretion ofl ifthe crientaions are opposite, wo oall tho negative Gf th length ofthe eogment the component of © fn te direction fl. Tho oomponent of on the direction off we denote by mf 8 is tho angle botmeen the direction of v and that off (et fig. 7) mo alaye a —wjasd—tb % Fie. 7-—Projection of « vector A veotor v of length 1 is called & unit vector. Its component in a diostion fia equal to the cosine ofthe angle between nd &. The components of vector v in the directions of the threo axes of 1 co-ordinate aystam are denoted. by tu ty ow If we transfer the initial pont of 0 tothe origin, we see that VirF I a By azo the direction cosines ofthe diction of, then salele w= lele a= | ele A free vector is completely determined by ita components vy ty ty ‘An equation | 088, between two vectors ia therefore equivalent to the three ordinary equations yw 4 ey ‘There are two diferent reasons why the use of vectors ie natural and ar(b+e)=(arb)re Fig. —Commatative hw of vector Fig, o—Associtive law of rector sation ‘aaiion advantageous, Firstly, many geometrical concepts, and a still greater ‘number of physical concepts, such as force, velocity, acceleration, &c. immediately reveal themselves as veotora indopendent of the particular co-ordinate system. Secondly, we oan set up simple rules for calculatingq CO-ORDINATES AND VECTORS 5 swith veotors analogous to the rules for calculating with ordinary numbers; bby means of these many arguments can be developed in « simple way, independently of the particular co-ordinate system chosen. We begin by defining the sum of the two vectors a and 5. For this ‘purpooe wo displace the vootor & parallel to itself until its initial point coincides with the final point of a. Then the initial point of @ and the final point of & determine a now vector c (see fig. 8) whose initial point is the initial point of @ and whose final point is the final point of 6, We call o the sum of a and & and write atone For this additive process the commuative law a+b and the aswociatice law at(b+e)—(a+ bomatb+e obviously bold aaa glance at fig. 8 and 9 shows ‘From the definition of vector addition wo at once obtain the“ projec tion theorem": the component ofthe eum of 20 or more vectors ina direction 1 ia equal tothe eum of the components of the individual eetore in that direc. tion, that i, (4 dye at by {in partionlar, the components of @ + Bin th directions of th coordinate a ate yy tat by tat De ‘To forma the sum of tro vectors we accordingly havo the following simple rule. The components ofthe um are gual to the ouma ofthe core ‘ponding components ofthe nammands. ‘Brory point P with co-ordinate (25 yy 2) may be determined by the patton veto fom th origin to P, whose componente in the directions of fhe axes are just the co-ordinates of the point P. We take threo unit ‘ecto in the directions of the thee axe, , in the s-diretion, e i the Jaivetion, oy in the =lvotion, If tho veotor 9 haa the componente fy ty ty then v= met Het ee Wo call 2, = t18y) P= they By = vyey the vector components of B. ‘Using the projection theorem stated above, we easily obtain the frans- formation formule which determine (2, , ’), the co-ordinates of a given ‘point P with respoct to the axes Ox’, Or/, O2' in terms of (2, y, 2), ita co- frdinates with respect to another set * of axes Oz, Oy, Or which bas the ‘same origin as the first ect and may bo obtained from it by rotation. Tho ‘threo new axes form angles with the three old axes whose cosines may be ‘1 js to be noted that in accordance with the convention adopted on 1p. 8 both systema of axes aro to be right-handed,6 ANALYTICAL GEOMETRY AND VECTORS [CuaP. expressed by tho following scheme, where for example 7, isthe cosine of the angle between the a-axis and the s-axis: lelyle lal ele ¥ || Bel Ye ¥ [eal Bale rom P we drop peepeniicuars to the axes Ox, Oy, O, thei feet being Py, Py Ps (ch fig. 1, ps1), Tho weotor from 0 to P ia then equal to {i sum of the vectors from 0 to Py from 0 to Pp and from O to Py. The direotion cosines of tho 2”-axia relative to the axes Or, Oy, O: are ay Pr thove of the y-axis ay By te and those of the z-axis ay By tx BY the projection theorem we know thet 2’, which is the component of the vector GP in the direction of the 2/-axis, must bo equal to the sum of the com: ponents of OP,, OP, OP in the direction of the axis, 60 that em ae But 1, {or aye is the component of « in the direction of the 2-axis, and soon. Carrying out similar arguments for yf and 2, we obtain the transformation formals wm me + By + yt ona + ba + ne vm ag + by tre eae tag tat v= Be’ + Ba + Be Bane tie +e ‘Since the components of # bound vector 0 in the directions of the axes are expreased by the formule aaa ey =m Ms =a ty in which (21, Yu %) are the co-ordinates of the initial point and (2, Ya 4) the co-ordinates of the final point of , it follows that the same trans- formation formule hold for the components of the vector as for the co-ordinates: and conversely ef = ayn + Bie ty tf = ath + Batt Yah ty gt + Bata + Yo 8. Scalar Multiplication of Vectors. Following conventions like those for the addition of vootors, we now define the product of a vector v by a number c: if v has the componentaq CO-ORDINATES AND VECTORS 7 ty ty em then co ia the vector with components cry ery ¢vy. ‘This de- finition agresa with that of vector addition, for v + 0= 20, 0+ 0+ 0 ‘= 30, and so on. If¢ > 0, cv has tho somo direction as v, and is of length @| |; if < 0, the direotion of co is opposite to the direotion of w, and ita length ia (—¢)| |. I e= 0, wo s00 that co is the zero vector with the components 0, 0, 0. ‘We can also define the product of two vectors + and 0, where this “multi- plication ” of weotors satis rales of ealeulation which are in part similar ‘to those of ordinary multiplication. ‘There are two diffrent kinds of vector multiplication. We begin with scalar multiplication, which ia the simpler and the more important for our purposes. ‘By the scalar product wv of the vecors 1 and 0 we mean the product of their absolute values and the covine of the angle 8 between their directions: 14|] 2] 0088, ‘The scalar product, therefore, is simply the component of one of the ‘veotors in the direction of the other multiplied by the length of the seoond ‘vootor. ‘From the projection theorem the distributive Jaw for multiplication, (6 ow w+ ow, {follows at once, while the commutative av, is an immediate consequence of the definition ‘On the other han, there isan easential diference between the scalar product of two vectors and the ordinary product of two number, for the Droducl can vanish although neither factor tanishes. 1 the lengths of ve and 0 are nol zero the product 10 vanishes if, and only if, the to vera 1 and v are perpendicular to one another. “a order to exprem the scalar product in terms of the componenta of the two veotors, wo take both the vestors 1 and 0 with initial pointe at the origin, Wo denote their vector componenta by ty ata, and 4, Dy, Og respectively, no that m= 14, + my -+ te and Vem Oy + Oy} By In the equation wo = (14+ m+ 14)(0, + t+ 0) wo can expand the product on the right in socordanoo with the rules of calculation which ‘wo have just established; if we notice that the product 405 84,0 #0, teaey ts0s, and m0, Vans ocaute the factors are pecpendicular to one Aotier, we obtain 0 — 14,24-+ 4404+ sty. Now the factors on the Fight have the sume direction, oo that by deinition +40, — ty, Boy There ty ty ty And ty Hy vy are the componente of and v soxpoctively, Bence tee + Mh ‘This equation could have been takon as the definition of the scalar product, ‘and is an important role for calculating the scalar product of two veotors * Often called tho inner produc.8 ANALYTICAL GEOMETRY AND VECTORS [Cuar. ssiven in terms of their components. In particular, if we take + and v a8 unit vectors with direction cosines ay, ay cs and By fy By Fespectively, ‘the scalar product i equal to the cosine of the angle between x and 0, ‘which in accordingly given by the formula Br + OaBe + oBe The physical meaning of the scalar product is exemplified by the faat proved in elementary physics, that a force f which moves a parts of unit ‘nas through the divoted distance doos work amounting to fe. con8 4. The Equations of the Straight Line and of the Plane. Let a straight line in the zy-plane or & plane in ays-space be given. Jn onder to find their equations we erect & perpendicular to the line (oF x fp Pig. 19.—The equation ot « traght He the plane) and specify a definite “ positive direotion along the normal”, pperpendioular to the line (or it docs not matter whioh of the two [possible directions is taken as positive (cf. ig. 10). ‘The veotor with unit Tength and the direction of the positive normal we denote by st, ‘The points of the line (or plane) are characterized by the property that the position ‘vector 2 from the origin to them has constant projection p on the direc tion of the normal; in other words, the scalar product of this position ‘vector and the normal voctor # ia constant. If a (oF a By 7) are the = 2. For if we caleulte a, say, from 2, = 29+ ary, and mbt stitote thi value, «= (e, — %)/tvin the equation 2 = 29+ a, we obtain the expression given above. Let a straight line be given by Bm mya Y= b aaa tye We now es to find the equation ofthe plane which passe through the point (és Yo %) and in perpendicular to ths ine. Since the direction cosines of ‘he normal to this plane are a, B, , the canonical form of the required equation ia et yt PaO, and since the point (z» Yq %) lies on the plane Bm amy + By + Yee ‘The equation of the plane through (zy yy. %) perpendicular to the line with dirootion cosines 2, B, y is therefore ae — 2) + BY — Ye) + Ye — Be) = In the same way, the equation of a straight line in the 2y-plane which passes through the point (zp y) and is perpendicular to the line with Airection cosines a, fi ale — =) + Bly ~ w4) = 0. “Later we shall need a formula for 3, the angle Between two plance given by the equations ae + fy be —p =O. wet By t ye— w=.12 ANALYTICAL GEOMETRY AND VECTORS [Cuar. ‘Since the angle between the planet ia equal to the anglo between their ‘normal veotors, tho soalar product of these vectors is coa8, wo that, 008 8 = aa’ + BBE vy’. In the same way, for the anglo 8 between the two straight lines a+ py— p= Onda’ + By—p'=0 fn the zy-plane we have 08 3 = aa’ BB Exarenns 1. Provo that the quantities a, a, of axos, satisfy the relations eat hPa t re 0 att Bt tn erst Bib tree = 0, Gat + Bat + Ys yay + Bab t+ Yets= 0 eat + Bat + Ye 2, Ifaand b are two vectors with initial point O and final points A end B, then the vector with O as intial point and the point dividing 4 in the ratio 0:1—0 as final point is given by (= Ga + 06. 13. The centre of mass of the vertioes of a tetrahedron PQRS may be defined as the point dividing MS in the ratio 1:3, where Mf is the centro cof maas of the triangle PQR. Show thet this definition is independent of ‘the order in which the vertices are taken and that it agrees with the general definition of the centre of mass (Vol. I, . 288). 4, If in the tetrahedron PQRS the centres of the edges PQ, RS, PR, QS, PS, QR are denoted by A, 4’, B, B', O, O” respectively, then the lines Ad’, BB’, OO all pass through the centre of masa and bisect one another there, +ty (P: 6) defining « rotation 5. Let Py «-y Py bo m arbitrary partslos in space, with maases ‘my, My +8 my Teapestvely. Lat G be their centre of muss and let Bu «os fe denote the veotors with initial point G and final pointe Pe eees Py Prove that mips t Maps toes + aba 0 2. Tam Anna ov 4 Tuancre, Tae Vouume or A TETRAHEDRON, ‘ram Vucror Muyrucation or Vecrors 1, The Area of a Triangle. Im onder to caloulate the arce of a triangle in the 2y-plane we imogine it moved parallel to iteelf until one of ite vertices ia at the origin; let the other two vertices be Py(2y ys) and Py(zp ya) (cf. fg. 12). We write down,q VECTOR MULTIPLICATION 3 the oquation of the line joining P, to the origin in ite oanonical form — = at y= Vatu Var hhence for the distance J of the point P, from this line we have (except perhaps for sign) the expression 7 wae Vives Va Since the length of the segment OP, ia Va: y a ol z Fig. 12—To lntrate the method for ‘Fig. 12-—Determination of th sign ofthe {ding the area ofa angle ‘er of tangle ‘area of the triangle, which i the product of the “hase” OP, and the ‘altitude A, is given (except perhaps for sign) by the expression 24 = 2 Hate ‘Thia expression can be cither positive or negative; it changes sign if wwe interchange P, and Py. We now make the following assertion. ‘The expression A has a positive or negative ealue according as the sense in which the vertices OP,P, are traversed ia the same as the sense of the rotation associated with the co-ordinate azes, or not. Tnstead of proving the fact by ‘more detailed investigation of the argument given above, which is quite {eaaible, we prefer to prove it by the following method. We rotate the tri- ‘angle OP,P, about the origin O until P, lies on the positive z-axis, (The ‘ease in which 0, Py, P, li on line, 90 that A = 4(2yy4 — 24%) = 0, can be ‘omitted.) This rotation leaves the value of A unaltered. After the rotation 'P, has the co-ordinates 2’ > 0, y;’= 0, and the co-ordinates of the new Pyare 2; and yy The area of the triangle is now lowe Anjan, ‘and therefore hae the same sign as yy ‘The sign of ys, however, is the same as the sign of the sense in which the vertices OP,P, are traversed (cf. ig. 18). Our statement is thus proved.14 ANALYTICAL GEOMETRY AND VECTORS [Cxar, For the expression 247, — ayy. which gives twice tho area with ita proper sign, it is customary to introduce the symbolic notation am ann ana |%% which we call a tuo-rowad determinant, or determinant of the second order. If no vertex of the triangle is at the origin of the co-ordinate system, .. if th three vertioes are (zy Yah (Hi Ya) (Zw Ys) by moving the axes parallel to themselves we obtain the formula at Ie —% mm 2lu— me wv for tho area of the triangle, 2. Veetor Multiplication of two Vectors. in addition to the scalar product of two veotors we have the important ‘concept of the weclor product® The ‘eotor product {48} of the vootory a ‘od i defined ae follows (ct. fg. 14: We measure off a and 6 from a {ab} point O. Then @ and 6 are two sides Of parallelogram in space. ‘The veotor product [a8] = in a vector whose length io numerically equal to the ares of the parallelogram and whose dizoo- tion is perpendioula to the plane of the parallelogram, the sso of direction boeing auch that the rotation from a to and c= [a8) isright-handed. (That % fs, if we look at the plane from the taveaor prota oftmo final Point of the vector , wo see the Fi 14 mated» "shortest rotation from the direction of @ to that of & aa a positive rotation.) If and & lie in the same straight line, we must have {a] = 0, sinoo tho area of the parallelogram is zero, Rules of Calculation for the Vector Product. (1) 1a % O and 6 + 0, then [a6] = Oiif, and only if, aand & have the same direction or opposite directions. "For then, and only then, the area of the parallelogram with a and 2 1s desis equal to zaro, (2) The equation [a6] = —[6a] holds. f Ofen called the outer prafuc oter notations in un fort are aX by aneq VECTOR MULTIPLICATION 15 ‘This follows at once from the definition of (a6). (8) If and b are real numbers, then [aa0} ~ ab[ad}. For the parallelogram with sides aa and 66 has an ares ab times as great ae that of the parallelogram with sides @ and 6 and lies in the same plane as the latter. (4) The distributive law holds: [a{b + )] = {a8} + [ae}, (0 + )a] = [0a] + (ea). We shall prove the first of those formulie; the seoond follows from it ‘when rule (2) is applied. We shall now give a geometrical construction for the vector pro- uct [25] which will demonstrate the truth of the distributive law directly. ‘Lot £ be tho plane perpendicular to a through the point O. Wo project B B 5 @ oO A ig. 15-~To show that fa} = (08 {5 orthogonally on F, thus obtaining a veotor 6 (of. fig. 15). ‘Then [a6'] = {a8}, for in the first place the parallelogram with sides @ and & has the ‘samo bose and the same altitude as the parallelogram with sides and 2; and in the second plage the directions of (46) and [a6] are the same, since @, 3, 8’ ie in one plane and tho sense of rotation from a to 2 is the ‘samo as that from @ to 6. Since the veotors @ and & are sides of a rect- ‘angle, the length of (a8’] = [a8] is the product |@|| 8°]. If, therefore, wwe increase the length of 2” in the ratio |a| 1, we obtain a vector 6” which has the same length as [a6']. But [a6] =[a8'] ia perpen. Gicular to both @ and &, so that we obtain (a6} = [ad’] from 8” by & rotation through 90° about the line a. ‘The sense of this rotation must be positive when looked at from the fnal point of a. Such s rotation we shall call a positive rolation about the vector @ aa azis. ‘We ean therefore form [a8] in the following way: project ® orthogon- ally on the plane B, lengthen it in the ratio| |: 1, and rotate it positively through 90° about the vootor a. ‘To prove that (a(6 + c)] = [28] + [ac] we proceed as follows: & and ¢ are the sides OB, 00 of a parallelogram OBDO, whose dingonal OD in the sum 5+. Wo now perform the three operations of projection, Tongthening, and rotation on the whole parallelogram OBDC instead of on the individual vestore 6, 0, 8-4 ¢ we thus obtain a parallelogram OB,D,0, whoee siden OB;, 00, aro the vectors [28] and [ac] and whove16 ANALYTICAL GEOMETRY AND VECTORS [Cuar. Aiagonsl is tho product (a(4 + c)]. From this the equation [a6] + {ac} = [ad + o)] clearly follows (cf. fg. 16) ig. 16 —Dietibtive lw for the vector product of two vets and & (5) Let a and & be given by their components along the axes, ay dy dy and by, by, by respectively. What is the expression for the veotor product [26] in terms of the vector componentat ‘We express a by the gum of its vector components in the directions of the axes. If 4, ey 64 aro tho unit vootors in the directions of the axes, then = He, + ey + O5ey and similarly bm beer + bres + bye. By the distributive law we obtain [28] = (ayes) (yesN] + [aren (bees) + e303) yea) + Lletaea) (bve3)] + £0404) ea)] + (axes) (bt (ese) (Ove) + [ases) ees)] + [lases) bread), which by rales (1) and (3) may also be written [ab] = above] + above) + ahleved + abileyes)+ adhlese,] + aphferes) Now from the definition of vootor product it follows that [ese] = —Lesedh Hence ese] = —[eves, ¢s= (e011 = —[eres} [ab] = (adds — agbdey + (0abs — axbsdon + (abs — ado ‘The components of the veotor product (a8] = ¢ are therefore4q VECTOR MULTIPLICATION 9 ay [a olga ele Tin physica we use the veotor product of two vectors to represent moment. A foroe facting at the final point of the position vector 2 has the ‘moment [/2] about the origin. 3, The Volume of a ‘Tetrahedron. Wo consider a tetrahedron (ef. Sg. 17) whose vertices are the origin and threo other points Py, Py Py with eo-ondinates (2 Ys #1 (tp Yo #) & > Fig. 17-—Determinton ofthe volume of «tetrahedron (ee vo 24) reepectively. ‘To express the volume of this tetrahedron in terms of the co-ordinates of ita vertices we proceed as follows. The vectors x, = OP, and x,— OP, aro sides of triangle whose area is half the length of the veotor product [%,:}- ‘This vestor product has the direotion of the perpendicular from P, to the plane of tho triangle OPyPy; A, the length of thia perpendicular (the altitude of the tetrahedron), is therefore given by the soslar product of tho vostor x,= OP, and the unit vector in the direotion of (2,4 for bin equal to tho component of OP, in the direction of [-z,%. Since the absolute value of (2% is twico the area A of the triangle OP,P,, and since the volume V of the tetrahedron is equal to 44h, we have V=t((mda Or, sine the componente of (a5) ar given br rah feak fe ah [va eal velwwe can write (8 ANALYTICAL GEOMETRY AND VECTORS [Cuar. a aly glam tuft Blea a al ‘This also hold fo the case in which 0, Py, Ps ie-on straight line; in ‘his cat, iia true, tho dizeotion of [=] is indotorminate, x0 that ean no longer be regarded es the component of OP, in the direction of (2) Dut nevertheless A = 0, 80 that Y= 0, and this follows aso from the above ‘expression for V, sino in this esse all the components of [23] vanish, ‘Here again the volume of the tetrahedron is given with a dente sgn, 8 the ares of the triangle was on p. 18; end we can show that the sign is positive ifthe three axes OP,, OP, ,OP, taken in that order form a system af the same type (right-handed or left-handed, as the ease may be) as the fo-orlinato axes, and negative if the two syitoms are of opposite typo. For in the fit cago the angle 8 between [=] and 2 lice fa the interval 08 <5, and in the second case in the interval F <8 << x, a8 follows immediately from the definition of (2, and V is equal to [Lea] |x| 0088. ‘The expresion al +m nal” # afen[s | aalt la cccuring in our formulm may be express more brely by the symbol fan a te te lew which wo call a three-rowed determinant, or determinant of the third onder. ‘Writing out the two-rowed determinants in full, wo see that hom hte za Ua te alata — Bayes + BN — Evan + Bete — Ee ‘Tost as in the case of the triangle, we find that the volume of the tetra- Ihedron with vertices (ey toy Hs (Ein Vis 4 Cn Yo al (Ha Yor #4) 8 Bom WO A Bo oh As Hm Whe Hm Bxaseuns * 1, What is the distance of the point P( jy Yo %) from the straight line I given by aaattb yodtd ematst ‘Tho more dificult examplos are indicated by an asterisk.q VECTOR MULTIPLICATION 9 2*, Find the shortest distance between two straight lines J end in space, given by the equations zmatto yadtd ond reetts 8. Show that the plane through the threo points (zi &s (8: Yo Ha (ey Yo #4) a given by noe mov Ane moe moy Bo e/=O Boe woe A 4. In 8 uniform rotation let (a, 6, 7) be the direction cosines of the exis of rotation, which passes through the origin, and « the engular velocity. Find the velocity of the point (zy. 2). 5. Prove Lagrange’s identity [91 leeiy = (a9, 6. The area ofa convex polygon with the vertices Psy th) Pty sas ‘Palsu Ua) i given by half th abwolute value of ssl |e slag fet th tel” dye wl Yat Yn + 3, Some THeoReMs oN DeveRMINANTS OF THE SECOND np Taiap ORDER 1, Laws of Formation and Principal Properties. ‘The determinants of the second and third order oceurring in the cal culation of the ares of « triangle and the volumo of a tetrahedron, together with their genoralization, the determinant of the nth order, or nerseed deter. ‘inant, aro very important in thst they enable formal calculations in all branches of mathematica to be expressed in compact form. Here wo shall develop the propertics of determinants of the second and third order, those of higher order we shall neod but seldom. It may, however, be pointed out that all the principal theorems may be generslized at once for determinants with any number of rows. For the theory of theeo we ‘must refer the reader to books on algebra and determinanta.* By their definitions (pp. 14, 18) the determinants ete je 8) ana fae a jg h Ct, og. H. W. Turnbull, The Phoory of Determinants, Matrices, and In- variants (Blackie & Son, Ltd, 1929)20 ANALYTICAL GEOMETRY AND VECTORS [Cuar. tro expreesions formed in a definite way from their elements a,b, o, d and 4, , 6 dy 6 §, 9, b, k respectively. ‘The horizontal Lines of elements (such a d, ¢, f in our example) aro called rows and the vertical lines (euch as 64, 8) azo called columns. ‘Wo need not epend any time in discussing the formation of the two- rowed determinant lt 8) ad — be. je al For the three-rowed determinant we give the “diagonal rule” which exhibita the symmetrical way in which tho determinant is formed: a Wo repeat the frst two columns after the third and then form the product of each triad of numbers in the diagonal lines, multiply the pro- Gucts associated with ines slanting downwards and to the right by +1, the others by —1, and add, In this way we obtain fe pect Met a ob] 7 fh ake We shall now prove several theorems on determinants: (2) If the rows and columns of a determinant are interchanged, the value of the determinant is unaltered. That is, la bi_|a ec le aj |b af ede} [a a of ae sl=|b eal owe le f kl ‘This follows immediately from the sbove expressions for the determinants, 12) If tco rowe (or theo columns) of a determinant are interchanged, the sign of the determinant iv alterel, that ia, the determinant is multiplied by iin virtue of (1) this need only be proved for the columns, and it can ‘be verified at once by the law of formation of the determinant given above,q DETERMINANTS an (8) In sostion 2 (p. 18) we introduced three-rowed determinants by the equations ah aay me 2 alle al ha a[% Blan Using (2), we write Uhis in the form mh ath |= im te ‘then in the determinanta on the right the elements aro in the same order ‘as on the left, If we interchange the last two rows and then write down ‘the same equation, using (2), we obtain: Ahm hm A th % ‘and similarly hm 5 4m alaale ts ‘We call theso three equations the expansion in terms of the elements of the third row, the eooond row, and the first row respectively. By interchanging ‘columns and rove, which according to (1) does not alter the value of the determinant, we obtain the expansion by columns, Ana [m | nA hm | +4 vs we oh a + 2 al ams lila im aw zal ofS tals tleele ft mh ‘An immediate consequence of this is the following theorem: (A) If all the elements of one row (or column) are multiplied by a number f the value ofthe determinant is multiplied by p- ‘From (2) and (4) we deduce the following: (8) If the elements of to rows (or to columna) are proportional, tha if every clement of one row (or column) isthe product of the corresponding22 ANALYTICAL GEOMETRY AND VECTORS [Cuar. clement in the other row (or column) and the same factor ¢, then the de- terminant is equal to zero ‘For according to (4) we can write the factor outside the determinant. If wo then interchange the equal rows, the value of the determinant is ‘unchanged, but by (2) it should change sign. Hence its value is zoro. Tn particular, s determinant in which one row or column consists entirely of zeros has the value zero, as also follows from the definition of a determinant. (0) The eum of to determinants, having the axme number of rowe, which differ only in the elements of one row (or column) is equal to the determinant which coincides with them in the rows (or columns) common to the two de- terminants and in the one remaining row (or column) has the sums of the corresponding elements of the two non-identical rows (or columns). For example: abel ja mel ja btme ade slela a sald eta ff a hktl |o pl |g hep & For if we expand in terms of the rows (or columns) in question, which in our example consist of the eloments 8, ¢, A and m, m, p respectively, and ‘add, we obtain the expression somlp deermly ltca-ole Gh ( ‘which clearly is just the expansion of the determinant a btm e detn fs o hte in terms of the column B4+-m, etm, k++ p. ‘This proves the stato. ment. ‘Similar statements hold for two-rowed determinants, (1) If t each element of a row (or column) of a determinant we add the ame mile of the corresponding element of another row (or column), the tale of the determinant is unchanged. By (6) the new determinant is the sum of the original determinant and f determinant which has two proportional rows (or columns); by (5) this second determinant is zero.* ‘tho rale for expansion in torms of rows or oolumns may be extended to define determinants of the fourth nnd higher order: Given a system of Nxtoon ‘uumbere aye ay ahh G32) ae de for example, we define a determinant of the fourth ondar by the expreasionq DETERMINANTS 23 ‘The following examples illustrate how the above theorems are applied ‘to the evaluation of determinants. We have ja 0 Oo! 0 © 0|— ack ook as we can prove by the dingonal rule. A determinant in which the elements in the so-called principal diagonal alone differ from zero is equal to the product of these elements, ‘Braluation of a determinant: aoa 1-1 a= 200 11d -bo1d (second row added to the first) en aoa 1-1. =a. 4. Another example is tee) pos #} poe # Ly vf=|0 ys woal= 0 ye yaa veal lice | lore aoa 1f we now expand this in terms ofthe fist colimn we obtain — Lyte wre CoS |-w-9 eal, Phe|-0-9@-296-y, 2. Application to Linear Equations. Determinants are of fundamental importance in the theory of linear equations, In order to solve the two equations ast by= A, et dy=B, for # and y, we multiply the frst equation by ¢ and the second by a. and eC ale a dl -blea a al eae & dl alah oh waal ‘laa & md a tnd similarly we can introduce detorminanta of the fith, sith, ..., nth onder {in svcceasion. Tt turas aut thet in all ewential properties theas agree with the ‘determinants of two or threo rows. Dsterminanta of moro than threo rows, however, cannot be expended by the “diagonal rulo”. "We shall not consider farther dotala here.24 ANALYTICAL GEOMETRY AND VECTORS [Cuar. ‘subtract the second from the first; then we multiply the first equation by and the second by b and subtract. We thus obtain (be — ay = Ae — Ba (ad — bo) = Ad — Bb, or eb o|4 8) gle bale 4 *le al"|B alt Yc allo B 1 we astome thatthe determinant a eal in dierent from zero, these equations at once give the ection 45) led Ba e 3| “fa op ia oy ea) le a wtih can be verdod by ebetituton T, however, tho determina ot eS) ait he anne ad Ir | would lead to a contradiction if ether of the determinanta 4 2 | were diferent from zero, If, however, Ba ‘our formule tell us nothing about the solution. We therefore obtain the fact, which is particularly important for our ‘Purposes, that a system of equations of the above form, whose determinant is ‘different from zero, aboays has a wnique solution. If our system of equations is homogeneous, that is, if A= B= 0, our calculations lead to the solution 2 = 0, y = 0, provided that ‘For throo equations with three unknowns, ar} ly to = A, debate = 2B, wthyt+h=, Sneddon da on cen. We muy the it ation [¢ LL ah wind by =| fe aid ty [ sxe by [CL te and by 8 heed | 6 a 124 ee obaingq DETERMINANTS 25 Chali seal a vt he sot eA she salsa seo But by our formule for the expansion of a determinant in terms of the clements of « column, this equation can be written in the form At bbe! ode! [4 be alae gleule © sales l= Boe fl. lo & | AbR) [kee lOvE By ra (4) tha cofcents of y ad vain, oo that adel [Abo elec gl-|e e rh bel fone In the same way we drive the equations a bel Ja de vac mle 2 sh lp bt] |p ok Ja be JabaA s]de sl=|a e Bh lo atl lea ol Uf the determinant be ae 7 ph is not oro, the last threo equations givo us the value of the unknowns. Provided thet this determinant is not zero, tho equations oan be solved ‘uniqualy for 2, y, # If the determinant is zoro, it follows that tho right- band sides of the above equations must also be zero, and the equations ‘therefore cannot be solved unless 4, B, C satisfy the special conditions which are expressed by the vanishing of every determinant on the A= B= C=O, and if its determinant is different from zero, it again follows that z= y=: = 0. Tn addition to tho oases above, in which tho number of equations is equal to the number of unknowns, we shall occasionally mest with26 ANALYTICAL GEOMETRY AND VECTORS [Cuar. systems of two (homogenous) equations with three unknowns, az + by + cz= 0, det eyt fem 0. 1f the thes determinants bel pale e alla ala tre not ll eof fr exompe, Dy 0, oar ouatons cn it be sled forzend yt gen b male] 2D, Dy Dy ye DY" Dy = Dy: Dy: Dy Geometrically this haa the following meaning: wo are given two veotore wand 9 with the components a, Z, ¢ and d, «, f respectively. We sook a ‘eotor 2 which is perpendicular to # and v, that is, which eaties the equations ux=0, ox=0, ‘Thus 2 isin the direction of [+0]. Exaweues 1, Show that the determinant ab a lo & f k can alwaya be reduced to the form «00 op 0 00y merely by repeated application of the following processes: (1) interchang- ing two rows or two columns, (2) adding # multiple of one row (or column) to another rom (or column). 2, If the throe determinants, by by a], aia! law lee by by do not all vanish, then the necessary and sufcient condition for the existence of aolition of the three equations oe oy bye + by eet oyq DETERMINANTS 27 is D=|5 t, elo aal 5. State the condition that the two straight lines saat tb, sooth a yaad, ond yet td Patty realty cither intersect or aro parallel 4*, Prove tho propertios (I) t0 (7), given on pp. 20-28, for deter inant of the fourth onder (deined on p. 22 (footnote). 5, Prove that the volume of a tetrahedron with vertices (ey dy %) (Yo tah (to Yo Fs (te 20) given by Ana ah Blan a mH 4, Avvive TRANSFORMATIONS AND THE MULTIPLICATION or DErerwinants ‘We shall conclude these preliminary remarks by discussing the simplest facts relating to the so-called affine transformations; at the samo time ‘we shall obtain an important theorem on determinants, 1, Affine Transformations of the Plane and of Space. By a mapping or transformation of portion of space (or of plane) wo mean a low by which each point has sasgned to it another point of {oce (or point of tho plane) as tage point the point isa we call the Gripinl point, or sometimes the mate! (in antithesis to the image). We brain a physical oxpresion of the conoapt of mapping by imagining that fhe porn of space (or of the plane) in question is occupied hy some Aleforauble substance end that our transformation represents a deformation in which every punt of the substance moves from its originl position to a cerain final position, ‘Gain a rectangular sytem of co-ordinates, we tak (z, 9,2) as the co- ordinates of the orginal point and (2, #) as thos ofthe corresponding image point. ‘The transformations which ar not only the simplest and most easily tnderstood, but are also of fundamental importance forthe general ease, ro tho fine anaformations. Aa fine transformation iv one in which he covdinaton (os ys #) (or in th plane (2, y)) ofthe image point are28 ANALYTICAL GEOMETRY AND VECTORS [Cuar. expressed linearly in terms of those of the original point, Such a trans formation is therefore given by the three equations vaathytotm Yodtaytfetn Yagthythtp or in the plane by the two equations oat hytm Yretdtn, ‘with constant cooficients a,b, ...‘Theso antiga an image point to every ‘Point of space (or of the plane). ‘The question at once arises whether we ‘an interchange the relation of image point and original point, that i, ‘whether every point of space (oF ofthe plane) bas an original point corre- sponding to it. The necessary and sufficient condition for this is that the equations etltaa’om detaythay—n o wethyth=?—p shall be capable of being solved for the unknowns 2, ys £ (oF % ¥), no matter what the values of , yf, 7 are. By sootion 3 (p. 24) an affine fransformation Aas an inverse, and in fact a unique inverse," provided ‘that ita determinant an + ly=e—m eet dym yn aS | w ont Sh Repco is different from zero. Wo shall confine our attention to affine trans- formations of this type, and shall not discuss what happens when A=0. ‘By introducing an intermediate point (2”, y', 2”) we can resolve the ‘general affine transformation into the transformations 2 mw a + by tee b= vaathy Yom drteytife ot oe Vogts "OTe asd eave vadtm Yoyte oe Tote gree SN vawte ‘Here (2, y, #) i mapped first on (2, y’, #”) and then (2”, 2”) is mapped ‘on (2, 7,2). Sinoo the seoond transformation is merely a parallel translation ‘of the space (or of the plane) 08 a whole and is therefore quite easly under- ‘That fs, overy image point has one ad only one original point,q AFFINE TRANSFORMATIONS 29 stood, we may restrict ourselves to the study of the frst. We shall there- fore only consider affine transformations of the form FaatWyte yey v deta the oe eet gy ve thy + ke with non-ranithing determinants. "The results ef sdton 8 (p. 26) for Linear equations enable sto express the inverse tranaformation by the formule eLt +! way Mee TTE © yee Lay, ge thy tke ° in which a, ¥, . . . are certain expressions formed from tho coefficients 4,2, . » » Boonuse of the uniquenee of the solution, the criginal equations ‘ho follow trom these latter. In particular, from 2 y= = 0 it follows that 2 = y/ = ¢ = 0, and conversely. ‘The characteristic geometrical properties of afine transformations are stated in the following theorems. (1) In apace the image of a plane is a plane; and in the plane the image of a atraigh line i a straight Tine. ‘For by section 1 (p. 9) we can write the equation of tho plane (or the lino) in the form Az+ By+ r+ D=0 (or Az+ By + D=0), ‘The numbers 4, B, C (or 4, B) are not all zero. ‘The co-ordinates of the ‘image points of the plane (or of the line) satisfy the equation Ale’ + Uy + ee) + Bee + ey +12) + Oz + Wy +k2)+D=0 (or Ae’ + By) + Bev + dy) + D=0). ‘Hence the image pointe themselves lie on a plane (or # line), for the co- ffcienta (« frestes CA + PB 4+ KO amie of tho co-ordinates 2, y, # (or 2, y') cannot all be zero; otherwise the ‘equations (ce eate8 vAtaR ¢A+fB+¥C srould hold, and these we may regard as equations in tho unknowns A, B,jo ANALYTICAL GEOMETRY AND VECTORS [Cxap. (or A, B). But we have shown above that from these equations it follows that 4 (or A= B= 0). (2) Phe image of a straight line in space isa straight Line, ‘Thia follows immediately from the fact that a straight line may be roganed as the intersection of two planes; by (1) ita image is also the inter seotion of two planes and is therefore a straight line. (8) The images of two parallel planes of space (or of two parallel lines of ‘the plane) are parallel. ‘For if the images had points of intersection the originals would have to interscot in the original points of those intarsoctions. (4) The images of two parallel lines in space are two parallel ines. For as the two lines lic in a plane and do not intersect one another, the same is true for their images, by (1) and (2). ‘The images are therefore parallel "The image of a vector 2 ia of course a vector »’ leading from the image of the initial point of @ to the image of the final point of v. Since the components of the veotor are the differences of the corresponding co- ordinates of the initial end final points, under the most general afine transformation they are transformed according to the equations to any + Bey + Oy ty = dey + ee + Soy = 904 + Bg dy 2. The Combination of Affine Transformations and the Resolution of the General Affine Transformation. [Hf we map a point (x,y, 2) on an image point (2, y’,#) by means of the transformation eaart byte Yadehaytfe va get hy + ke and then map (2, 7) on & point (2, y”, 2”) by means of s second affine transformation 2 mage + by +e vim ds ey + he 2 = gy + hy +e, then we readily eo that (xy, #) and (2”, 2”) arwals related by an afine (teansformation: and in fact imag t by beg Yom deeb ey the Fm geet hy + he where the coeficients are given by the equationsy AFFINE TRANSFORMATIONS 3t as gat bd+ og be abt bet ah q=ae+ bf+ ob qadetedthn aa abtoetih h= det oft hb = Het Ret bg ym abt het bh, by= ge + af + hak We say that this lat transformation is the combination or resultant of the first two. If the determinants of the fit two transformations are different ‘rom zero, their inverses can be formed; hence the compound transforma tion also has an inverse, ‘The coefficients of the compound transformation fare obtained from those of the original transformation by multiplying corresponding elements of column of the first transformation and of & row of the socond, adding the three products thus obtained, and using this “product” of column and row aa the coefficient which stands in the column with the same number as the column used and in the row with the same number as the row used. In the same way, combination of the transformations emartly yg Maar thy vaatdy Vo + by ives the new transformation, x" = (aya + bye) + (mb + bydly Y= (eat dye + (ob + Ady. By & primitive transformation wo mean one in which two (or one) of the three (or two) co-ordinate of the image are the tame as the corre: sponding co-ordinates of tho originel points. Physically we may think of & primitive transformation ae one in which th spaco (or plane) undergoes ‘stretching in one direction only (the strtching of courwe varying from place to place) to that all the pointa are simply moved slong a family of parallel lies, A primitive affine tranaformation in which the motion takes place parallel to the s-axis ia analytically represented by formula of the type mart yay ly y The general affine transformation in the plane, ea artly vat dy, wwitn a non-vanishing determinant, can be obiained by a combination of primitive tranaformations. ‘In the proof we may assume * that a + 0. We introduce an intermediate = 164 ~ 0, thou +b 0, and wo can return to the caso @ +h 0 hy intorohanging a andy. Soch an Interchange, represented by the trnnaformation =p, ¥ 275, H ital elected by the three tcseaive primitive transformations Gry Goh , Eoratecy BoE Ro baat PENIS32 ANALYTICAL GEOMETRY AND VECTORS [Cuar. Point (Gn) by the primitive transformation Enartoy aH whose determinant a is different from zero. From & 7 we obtain 2, 7 by «second primitive transformation ad — be ja a. ‘This gives the required resolution into primitive transformations, Tn similar way the affine transformation in space vmod byte ade tf Ym gethy +h web ym let 4 ‘with the determinant ai— bet 1with a non-vanishing determinant, can be resolved info primitive traneforma- tions, Of the throo doterminanta 24] |e ol [bo act last legs st least one must be different from zero; otherwise, aa the expansion in feria of the elesenta ofthe last row shows, we should have do aes oak ‘Asin the provioun case, wo ean then assume without loss of generality (a) that [6 bf aod 2) at ot 0. ‘The firmb intermediate point (6 1» Os given by means of the equations gaart byte a oy t= ® ‘The determinant of this primitive transformation isa, which la not zero. or the second transformation to ©,» & we wish to put £/= & Um & and alo to have 7 =’: Ono primitive transformation then remains, Tf {nthe equation 1/ = y= de + ey + fe we introduce the quantities 7. & instead of = y. 5 we obtain the second primitive transformation in the formq AFFINE TRANSFORMATIONS 3 alata vat ‘he determinant of this transformation is * formation must then be = , = 3, The Geometrical Meaning of the Determinant of Transforma- tion, and the Multiplication Theorem. From the considerations of the previous section we can find the geometrical meaning of the determinant of an affine transformation and ft the same time an algebraic theorem on the multiplication of determinants. ‘We consider a plane triangle with vertices (0, 0), ( area ia given (scotion 2, p. 14) by the formula Alay my jv val Wo shall investigate the relation between A and tho azea A’ of its Image obtained by means of « primitive affine transformation mart by y ‘The vortiows of the image triangle have the co-ordinates (0, 0), (0, + bay ibe Caen fe Alan + by, a+ bys aly win alm % ‘This determinant, however, can be transformed by the theorems of section 3 (p. 22) in the following way: Marth at ny u Was (le tle Whose 4 a= Jan an} Zim ml Zin ws that is, a4, 2 err)34 ANALYTICAL GEOMETRY AND VECTORS [Cuar. If wo had taken the primitive transformation ver cot dy, wwe should have found in the same way that Amd, Wo see, therefore, that s primitive affine transformation hae the effeot of multiplying the ares of a trianglo by a constant independent of the tri- ‘angle.* Sinoe the goneral affine transformation oan be formed by com. Dining primitive transformations, the stetement remaina true for any affine transformation. In the ease of an affine transformation the ratio of the area of an image triangle to the area of the original triangle is constant ‘and independent of the choice of triangle, depending only on the coecients of ‘the transformation. Tn order to find this constant ratio we consider in ppartioular the triangle with vertioes (0, 0), (1, 0) and (0, 1), whose area A {is . Since the image of this triangle according to the transformation sant by Yatay ‘nas the vertioes (0, 0, (ao), (bs d) ita area ia fand wo thus seo that the constant ratio of area A’/A for an affine trans. formation is the determinant of the transformation. For transformations in space we can proceed in exactly the same way. Af we consider the tetrahedron with the vertioes (0, 0, 0), (zs Ys 4 (€x Yo *s (ep tr %) and subject it to the primitive transformation v= 5 ‘the image tetrahedron has the vertices (0, 0, 0), (az, + Buy + efi Yi 4), (arg + Bs + Cy Yo 2), (Oey + Bis + Cy Yo Hy 80 that its volume Fis amt byte amt Wat Oey 0g + yy + ce w we % 5 % 4 alm # a g]m te ws). “la a al 1 If no vertex of the triangle lies a the origin, the same fact hold, in virtue of the general formula for the area given on p. 14.i AFFINE TRANSFORMATIONS 35 Hence Ve ay, where ¥ ia the volume of the orginal tetrahedron. or the volume of the {mage given by tho primitive transformation vane Yadetatte var ‘wo similarly find that Vemer, and for the primitive transformation sve find that vow From this it follows that an arbitrary affine transformation has the effect of multiplying the volume of tetrahedron by a constant.* Tn onder to find this constant for the transformation at byte Yadtath get hy + he we consider the tetrahedron with the vertices (0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1), whose image has the vertices (0, 0, 0), (a, d, g), (6.6, (@ J, B. Bor tho volumes V” and V of the image and the original ‘we’ therefore have | v= | v Wet lok sae hence the determinant is the constant sought, abe ae | oh ‘Tho sign of the determinant also has @ geometrical meaning. For from ‘what we have seen in section 2 (p. 18) on the connexion between the sense of rotation and the volume of the tetrahedron or area of the triangle, it follows at onco that a transformation with a positive determinant preserves ‘the sense of rotation, while a tranaformation with a negative determinant revere i ‘If no vertox of the tetrahedron coincides with the origin, thi theorem. follows from the general formula for the volume of « tetrahedron (p- 18).36 ANALYTICAL GEOMETRY AND VECTORS [Cuar. We now consider the combination of two transformations vaatta waar tb tat Yadtarfe — Ymad tay tht gethy the 2 = oe + hay + 2m (at bya + ole + (b+ bye + cab + (aye+ yf + eke = (da + ed + fige + (db + ee + Lidl + (he + af + hike 2 = 0+ hd + kygle + (3b + he + kh + (xe + Inf + bake ‘As we pass from 2, y,2to 2, y/, 2 the volume of a tetrahedron is multiplied by a be ae jo wo puss from 2, ys # to 2", y's 2” by eh aah hh and by direct change fom sy, # toy’, 2” ibs multiplied by net hdbeg abt bet oh ae+d,f-+ ak dot adt ig Gbteet hh detest hk lnot hid hg abt hye + hh go + haf + bi ‘This gives us the following relation, known as the theorem for the ‘multiplication of determinant Cr Cs a hl | eh oe lo i a0, + Bade + ci, aia + Bea + chy ee + By fat cake soe + cide + Site iba + eee ibe hee + eafet Sake ts + hd, + Fay et Iyea + habs 46a + Bafa + By ‘As before, we call the elements of the determinant on the right the pro- lab | a be | ducta” of the sows of |, 4% f,| and tho columns of |d 4 fe[s at oy oe he Bl the intersection of the ¢th row and the #-th column of the product of the determinants there stands the expression formed from the i-th row of ah by Gy @ 4 fh| and the Ath column of |d ey Jy]. Sinve rows and & hy! a hy by ‘columns ate interchangeable, the product of the determinants can alsoy AFFINE TRANSFORMATIONS 7 bo obtained by combining columns and rows, eslumns and columns, ot rows and row, "For two-rowed determinants the corresponding theorem of course olds, namely [oe S| foe l= facet ee oe Js aL deel oes + abe cee + dad (combining rows and rows, &e). Exawrnns 1. Evaluate the following determinants: is 4 5 raid aad lew @i so, wir 24, @i2 34, Mi ye bo 7 139 3-17 le 2, Find the relation which must exist between a, b,¢ in order that the system of equations B+ dy poem det byt b= ba + by +7 may have a solution. 8°. (a) Prove the inequality abe Dale ve | 0 there shall correspond two positive numbers 8, and 3, such that [fE+h 1th) —-16 D1S€ whenever | h | S 8, and | k | S 8y. ‘The two conditions are equivalent. For if the original con- dition is fulfilled, go is the second if we take 8, = 8, = 8/4/2; and conversely, if the second con- dition is fulfilled, so is the first if for 8 we take the smaller of the ‘two numbers 8, and 8,. ‘The following facts are almost obvious: The sum, difference, and pro duct of continuous functions are 0 also continuous. The quotient of Fig, ¢—Bondary point continuous functions is continuous except where the denominator vanishes. Continuous functions of continuous functions are themselves continuous (cf. section 5, No. 1, p. 70). Tn particular, all polynomials are continuous, and alt rational fractional functions are also continuous except where the denominator vanishes.+ Fig 6 tustrates the cave whore (fy 9 lies onthe boundary of B. {Another obvious fact, which, however, is worth stating, is as follows age fenion (rg) ie continous tn a region Be and Se diferent from tere oto6 FUNCTIONS OF SEVERAL VARIABLES [Cxar. A funotion of soveral variables may have discontinuities of a much mote complicated type than a funetion of a single variable. For example, discontinuities may occur along whole arcs of curves, as in the case of the function «= y/z, whioh is discontinuous along the whole line 2 = 0. Moreover, a function f(z, y) may be continuous in z for exch fixed value ‘of y and continuous in y for each fixed value of 2 and yet be « discontinuous fupotion of sand y, Thai examplid bythe fanotion fs, 9) = 2. Bry J(0, 0)=0. TE wo take any fixed non-zero value of ys thin function ‘a cbriously continuous 28's funotion of 2, as the denominator cannot vanish. Tf y= 0, wo havo {(2, 0)— 0, which i a continuous fono- tion of =. Similarly, (2 9) is continuous in y for cach fxed value of 2 But at every polit on the line y= x, excopt the point = y = 0, tro have fis; y) 1s and there are pints of ths lin arbitrarily close {othe crigin. Henoo the fonction is discontinuous at the’ point zayno. Other examples of discontinuous fanotions will be found in Vol. I (p. 464). 2, The Concept of Limit in the Caso of Sovoral Variables. The concept of the limit of a function of two variables is closely related to the concept of continuity. Let us suppose that the function f(, y) is defined in a region R, and that (E, 2) is a point either within 2 or on its boundary. ‘Then the statement that the limit of f(z, y) as 2 tends to ¢ and y tends to 7 is 2 is to be understood as having the following meaning: for every €>0 there is a3 > 0 such that If —l (Em) interior point P of the region, it in pourible to mark of about P a neihtourhod, tty'a chek bldg en to Ree heh ie 3) 0 nowhere fea fo woe For the value of te function at Ps ay we can tack of about Pe cela wo ‘amall that the value of the function ia te lrcle diffe fom a by less than ai? and therefore ia oartainly got zero.1 CONTINUITY a7 For emphasis this is sometimes read “the double limit as 2 tends to £ and y tends to 7 of f(z, y) is”. Using the language of limits, we can say that a function f(y) is continuous at a point (¢, 7) if, and only if, pa fey =f). ‘We can see the matter in a new light if we consider sequences of points. We shall say that a sequence of points (2, 9s), (te Ys). +++ > Gm Yn> «+ tends to a limit point (Z, 7) if the distance VA Gea— 8F-+ Ya—7)?} tends to 0 a8 n increases. We can then show at once (ef. Vol. I, p. 47) that if f(a, y) > Las (x, y) > (.n), then lim f(t, ya) = 1 for every sequence of points (2 Ya) in R which tends to (£, 9). The converse is also true; if lim f(a: Ys) ‘exists and is equal tol for every sequence (tq) Yn) of points in R tending to (£, 7), then the double limit of f(z, y) as 7-> ¢ and y > 7 exists and is equal to. We omit the proof of this. In our definition of limit we have allowed the point (z, y) to ‘vary in the region R. If we 90 desire, however, we can impose restrictions on the point (2, y). For example, we may require it +o lie in a sub-region R’ of R, or on a curve C, or in a set of points MinR. In this case we say that f(z, y) tends to Las (z, y) tends to (€, 1) in R (or on C, or in M). Tt is of course understood that R (or C, or M) must contain points arbitrarily close to (g, 7) in order that the definition may be applicable. Our definition of continuity then implies the two following requirements: (1) a8 (x, y) tends to (£, 1) in R the function f(x, y) must possess a limit 1; and (2) this limit 1 must coincide with the value of the function at the point (£, 7). Tt is obvious that we could define continuity of a function, not only in a region R but also, for example, along a curve C, in the same way. 8, The Order to which # Function Vanishes.” If the function f(z, y) is continuous at the point (£, 9), the difference f(z, y) — f(é, 7) tends to zero as z tends to £ and y tends to 7. By introducing the new variables h—= 2— € and ‘This subweotion may be omitted on a first reading.B FUNCTIONS OF SEVERAL VARIABLES [Cuar. = y—7 wo can express this as follows: the function (A, #) (E+ h, n+ ) — f(g, 2) of the variables 4 and & tends to 0 ash and k tend to 0. Wo shall frequently meet with functions such as ¢(h, X) which tend to zoro* as hand & do. As in the case of one in- dependent variable, for many purposes it is useful to describe the behaviour of 4h, #) as h-r0 and k-»0 more precisely by between different “orders of vanishing” or “orders of magnitude” of ¢(i, i). For this purpose we base our comparisons on the distance = VBE = Ve FP of the point with co-ordinates 2== €-+ A and y= +k from the point with co-ordinates and 7, and make the following statement: A function $(b, k) vanishes as p->0 to the same order as p= Vi-+ EA, or, more precisely, to at least the same order as p, provided that there is a constant O, independent of h and k, such that the inequality |2#2:8| <0 fz holds for alt suffciently small values of por, more precisely, when there is a 80 such that the inequality holds for all values of hand & such that 0< Vi#-+ <8. Further, we say that Mh, H) vanishes to a higher order + than p if the quotient Hb: ') P tends to 0 as p> 0. This is sometimes expressed by the sym- bolical notation t (A, &) = ofp). ‘Tn the older literature the expressions “ (hb) bocomes infinitely small tush and k do" or ih, 4 i infinitesimal ” are also found. ‘These statements ave perfectly definite meaning if wo regard them simply es another way of sexing | “iis #) tends to'0 as hand f do". Wo neverdhclees prefer to avoid ‘expremsion "infinitely small entirely. ee Brntardst to bvold confaalon, ve wosld exproosy point out that higher onder of vanishing for p> 0 lmplice emellor values in the neighbourhood of fo for cxampl, of vnihon to higher Sedat than p and phn amalar than p, when p is noaly sero The lttat oe of course chosen bocange it ia the first letter of tho word. order. If we wish to express the statement that g(h, E)vanishoe to at leect the ‘Sime ordar ae p, but not necessarily toa higher ordar, we use th letter O instead. ‘ofc, writing 4(B, 4) = O(p). in this book, however, we abell not we this symbol.1m CONTINUITY 9 ‘Lot us now consider a few examples. Sinoo the components hand i of the distance ¢ in the directions of the 2- and ‘y-axes vanish to atleast the same order as the distance itself, The same is truo fora linear homogencous funotion ah + BE with constants a and 0, ot for the fantion pin. ‘or fixed values of « greater than 1 the power fp of the distance vanishes tos higher onler than 9; symbolically, p= ole) for a> 1. Similarly, a homogeneous quadratic polynomial ah? 4. Dak 4 eb? in the variables h and i vanishes to & higher order than pas po: ait + bhE-+ okt — o0). More generally, the following definition is used. If the ‘comparison function a(h, %) is defined for all non-zero values of (h, #) in a sufficiently small circle about the origin, and is not equal to zero, then $(h, 2) vanishes to at least the same order as «(h, #) a8 p> 0 if for some suitably chosen constant O the relation H(A, k) eenlse holds; and similarly, $(h, k) vanishes to a higher order than w(h, k), $(h, k) or lh, B) = ofo(h, Bit EG 7) + 0 when p> 0. ‘For example, the homogeneous polynomial ak* + Bhk + ol¥ is at least of the same order as 6%, since Lads 4 oe + od] (lo) + 181 + Le) + Also » = o(, Since lim (¢log) = 0 (VoL Lp. 196% vee Exacrurs 1 Disouse the behaviour of the functions (a) at — Sey, @) A Gott tt {in the neighboushood of the origin, ‘2, How many constante docs the general form of polynomial P(x, 9) of degree n contain?so FUNCTIONS OF SEVERAL VARIABLES [Cuar. 3, Prove that the expression att bt tty + at vanishes at z= y= 0 to at leat the same onder aa p*— (28+ yl 4. Find the condition that the polynomial Past + nyt og in of exactly the sare order as 6 in the neighbourhood of «= 0, y = 0 (he, both 5; and are bounded 1, Are the following functions continuous at 2 oy tev Btiyis OSte | (ORbartE fe Eta viet) lel gf a|itt. (ay | yield zr Vern fl 6, Find a 3(c)(p. 44) for thooe functions of Bx. 5 which aro continuous, 3. Tar Derivatives oF 4 Foncrion 1, Definition, Geometrical Representation, Hf in a function of several variables we assign definite numeri- cal values to all but one of the variables and allow only that one Sections of w= f(s 9) variable, say 2, to vary, the function becomes a function of one variable. We consider a function u = f(2, y) of the two variables and y and assign to y a definite fixed value y= yp=c. The function u= f(z, yo) of the single variable 2 which is thus11) ‘THE DERIVATIVES OF A FUNCTION st formed may be represented geometrically by letting the plane Y= yo cut the surface u= f(x, y) (of. figs. 6 end 7). The ‘carve of intersection thus formed in the plane is represented by the equation u= f(z, yo). If wo differentiate this function in the usual way at the point «= zr (we assume that the derivative exists), we obtain the partial derivative of f(x, y) with respect to x at the point (zq, yo). According to the usual definition of the derivative, this is the limit * Tim Loh wd) hoo h Yo) Geometrically this partial desivative denotes the tangent of the angle between a pazallel to the z-axin and the tangeat line to the curve seo Jee yo Tes therefore the slope ofthe eurface w= I, 3) in the direc tion of the x-axis. To represent these partial derivatives several different nota- tions are used, of which we mention the following: lim flo be ve) Sf eo 7 If we wish to emphasize that the partial derivative is the limit of a difference quotient, we denote it by 3a zp = falter Yo) = Halts Yo)- x a Hore we use a special round letter 2, instead of the ordinary @ ‘used in the differentiation of functions of one variable, in order to show that we are dealing with » function of several variables and differentiating with respect to one of them. Tt is sometimes convenient to use Cauchy's symbol D, men- tioned on p. 90 of Vol. I, and write x an Dabs but we shall seldom use this symbol. In exactly the same way we define the partial derivative of #1 (ee) i» point on the boundary of the region of definition, we make the resition that in tho passage to the limit the point (2 +, yq) mast always remain in the region.52 FUNCTIONS OF SEVERAL VARIABLES [Cuan Se, y) with respect to y at the point (zo, yo) by the relation ow y+ h) Be ro» Yo) lin Le tot = Slt 40) _ oe, ys) = Dy flew te ao ee u=f(, y) with the plane = 2, perpendicular to the z-axis. Let us now think of the point (z, yo), hitherto considered fixed, as variable, and accordingly omit the suffixes 0. In other ‘words, wo think of the differentiation as carried out at any point (, y) of the region of definition of f(z, y). ‘Then the two deriva- tives are themselves functions of = and y, ls, Dah = FEY and ule, n= sles n= LEY. For example, the function w= 2*4+ y* has tho partial derivatives 14, = 2x (in differentiation with respect to = the term y* is regarded aa a constant and so haa the derivative 0) and u, = 2y. ‘The partial derivatives of um aly are uy = Baty and ty = 2. ‘We similarly make the following definition for any number n of independent variables, Plena 4 te) jg Stat hey - 2 Bn) ~ f(r By «++» a) iy = falta Sy «65 20) = Defy ap +++ 4 Pr it being assumed that the limit exists. Of course we can also form higher partial derivatives of f(x, y) by again differentiating the partial derivatives of the“ first order ”, f,(z, y) and f,(z, y), with respect to one of the variables, and repeating this process. We indicate the order of differentia- tion by the order of the suffixes or by the order of the symbols Gx and dy in the “denominator”, from right to left,* and use the following symbols for the second derivatives: EO-B-n- ns (Gh) = ssaymfor= Po 3 (Z) i writen ‘In Continental usage, on the other band, eeay THE DERIVATIVES OF A FUNCTION 3 a (Y)_# 5G) = apa fos Dink We likewise denote the third partial derivatives by a =f. Oa \aat] ~ 08 ™ a a) of ay at) = Wyo a ( & (ey and in general the n-th derivatives by a (a ot ie &) an So 2 (a ay wy (& ayaeen See bee In practice the performance of partial differentiations involves nothing that the student has not met with already. For according to the definition all the independent variables are to be kept constant except the one with respect to which we are differentiat- ing. We therefore have merely to regard the other variables as constants and carry out the differentiation according to the rules by which we differentiate functions of # single independent variable. The student may nevertheless find it helpful to study ‘the examples of partial differentiation given in Chapter X of Vol. I (p. 469 et seq.). Just as in the case of one independent variable, the possession of derivatives is a special property of a function, not enjoyed even by all continuous functions.* All the same, this property is possessed by all functions of practical importance, except pethape at isolated exceptional pointa + For an explanation of the term “differentiable, which implies more than thatthe partial derivatives with respect to 2 aad y exit, we p. 60 en4 FUNCTIONS OF SEVERAL VARIABLES [Cnar. 2. Continuity and the Existence of Partial Derivatives with respect to x and 9. In the case of fumotions of a single variable, we know that the existence of the derivative of a function at a point implies the continuity of the function at that. point (cf. Vol. I, p. 97) In contrast with this, the possession of partial derivatives does not imply the continuity of a function of two variables: e.g. the re awit 4 (0,0) =0, has partial derivatives v everywhere, and yet we have already seen (p. 46) that it is discon- tinuous at the origin. Geometrically speaking, the existence of partial derivatives restricts the behaviour of the funetion in the directions of the 2- and y-axes only, and not in other directions. Nevertheless the possession of bounded partial derivatives does imply continuity, as is stated by the following theorem: Tf a function f(x, y) has partial derivatives f, and f, everywhere in a region R, and these derivatives everywhere satisfy the in- equalities function u (2, 9) [he 01 0 ‘and for all woficiontly amall values of had F Seal V) ~ €0, we arrive at the equivalent definition: ‘F(2) is said to be differentiable at the point x if fle-+ h) — f(z) = ha(e, h), where the quantity a(a, A) is continuous, as a function of h, at h=0. These ideas can be extended in a perfectly natural way to functions of two and more variables. We say that the function u= f(z, y) is differentiable at the point (2, y) if it can be approximated to in the neighbourhood of this point by a linear function, that is, if it ean be represented Set hy y+ B= fa, y) + Ah+ Bk+ eh+ exh, where A and B are independent of the variables hand k and whore « and ¢ tend to 0 as hand & do. In other words, themy) THE TOTAL DIFFERENTIAL 6 difference between the function f(z+h, y-+) at the point (+h, y+) and the function fla, y)+ Ah+ Bk which is linear in h and & must be of the order of magnitude* o(p), that is, ea eet le edad eee ren = y(i8 + I8) of the point (z + fh, y+ B) from the point (2, y). If such an approximate representation is possible, it follows ‘at once that the fanotion f(z, y) can be partially differentiated with respect to and to y at the point (z, y) and thet fom Aand f= For if wo put k= 0 and divide by h wo obtain the relation S@+h y—-fle 9) h Ate Sinoe ¢, tends to zero with h, as we pass to the limit +0 the left-hand side has a limit, and that limit is A. Similarly, we obtain the equation f,(z, y) = B. Conversely, we shall prove that a function u— f(z, y) is differentiable in the sense just defined, that is, it can be approxi- mated to by linear function, if it possesses continuous deriva- ‘tives of the first order at the point in question. In fact, we can ‘write the inoremont Au=fa@+hy+h—f@y) of the function in the form du={fe+hyt+h—feyt}+(/@9+H)—S@ vy} ‘As before (p. 54), the two brackets can be expressed in the form Au = hfsle + Oh, y+ b) + Hla, y + Ouk), using the ordinary mean value theorem of the differential calculus. Since by hypothesis the partial derivatives f, and f, aro continuous at the point (2, y), we can write Slat Oh y+ = fle +4 Tague ages ters om the’ inonelty Te heel ley) eae als, whee om Tel selon tae fo Gaudi srieenett # diet oltows from the Ors, since [eve , si iif i A tele (cL ineae on of the form Ol b+ [whee EOE ‘whence follows thatthe requitementa of the fst defition are also filed.62 FUNCTIONS OF SEVERAL VARIABLES {Cuar, and Sule, y+ Ob) = fults y) + ey whore the numbers ¢ and ¢, tend to zero as hand kdo. We thus obtain Au = hfe, 9) + Wile, 9) + eh + eck Hide, W) + Kids y) + G/T 8), and this equation is the expression of the above statement.” We shall occasionally refer to a function with continuous first partial derivatives as a continuously differentiable function. If in addition all the second-order partial derivatives are continuous, ‘we say that the function is twice continuously differentiable, and s0 on. As in the caso of functions of one variable, the definition of difierontiability can be replaced by the following equivalent definition: the function f(c, y) is said to be differentiable at tho point (2, y) if Se hy +k) —S(e, y) = ah+ Bk, where a and i depend on h and kas well as on x and y, and are continuous a8 functions of h and & for k= 0, k= 0. No further discussion is required to show how these considera- tions can be extended to functions of three and more variables. 2, Differentiation in a Given Direction. An important property of differentiable functions is that they not only possess partial derivatives with respect to z and y, or, as we also say, in the x and y-directions, but they also have partial derivatives in any other direction. By the derivative in the direction a. we mean the following: ‘We let the point (e+ A, y+ ) approach the point (2, y) in such a way that it is always on the straight line through (x, y) which makes the constant angle a with the positive z-axis. In other words, 4 and & do not tend to 0 independently of one another, but satisfy the relations A= peosa and k= where p is the distance /(i# + I) of the point (2+ h, y +B) “If we assamo the existance only, an not the continuity, of the derivatives fa and fy, the function is not neceasiily differentiable (cf. p65 ef oy.)1 ‘THE TOTAL DIFFERENTIAL 63 from the point (x, y) and tends to 0 as hand k do. If as usual we then form the difference f(z +h, y +k) —J(e, y) and divide by p, wo call the limit of the fraction tim 2+ pose, y + pain Foca) ? ‘(@, ») Do f@ N= the derivative of the function f(a, y) at the point (x, y) in the direction a, provided that the limit exists. In particular, when a=0 wo 0 and A= p, and we obtain the partial derivative with respect to 2; when a= 7/2 we have h=0 and = p, and wo obtain the partial derivative with respect to y. If the function f(r, y) is differentiable, we have fet hy +h) —S@, 9) = Met Wat ep plfecosa + fy sina + ¢). ‘As p tends to 0, so does «, and for the derivative in the direction a we obtain the expression Desf (% 9) = fa 0080 + fy sing; it is therefore a linear function of the derivatives f, and f, in the x- and y-directions, with the coefioients cosa and sina, ‘This result always holds good, provided that the derivatives f, and f, exist and are continuous at the point in question. “Thos for the radius vector r= V(@4-+ yf) from the origin tothe point (Gy) we have the partial dxivatives * you tee ptega tact md ye rea Veter Va pyr where 6 denotes the angle which the radius vector makes with the z-axin, Conseauectly, in the direction a the function r has the derivative Dayt = recota + r, sina = coaD cove + sin8 sine = co8(0 ~ a; fn particular, ia the direction of the radius veotor ito this derivative has tho value 1, while in the direction perpendicular to the radius vector it has the valve 0. ‘a the direction of the redias veotor the function = has the derivative Dyole) = 000 and the fonction y as the derivative Dio(y) = sin@; in the Neotion perpendicular to the radina veotor they have the deriva Denia? = —8ind and De q/ny = 080 respectively. The derivative of a function f(@, y) in the direction of the = sind,64 FUNCTIONS OF SEVERAL VARIABLES [Cuar, radius veotor is in general denoted by 2% ‘Thus we have the convenient relation G a_ a,. 8 an 0080 5. + sin, on here any dileentable function can be writen aftor th symbols aa a ie By It is also worth noting that we obtain the derivative of the function f(z, y) in the direction a if, instead of allowing the point Q with co-ordinates (+h, y + ) to approach the point P with co-ordinates (2, y) along a straight line with the direction a, we let Q approsch P along an arbitrary curve whose tangent at P has the direction a. For then if the line PQ has the direction B, we can write h= pcos 8, k= p sin, and in the formule used in the above proof we have to replaco a by B. But since by hypothesis f tends to a as p->0, wo obtain the samo expression for Deaf o, 9)- the same way, a differentiable function f(r, y, 2) of three independent variabies can be differentiated in a given direction. We suppose that the diroction is specified by the cosines of the three angles which it forms with the co-ordinate axes. If we call these three angles a, f, 7, and if we consider two points ( y, 2) and (2+ hy +k 240), where h z 1 cosa, 08 B, Pp c08y, then just as above we obtain the expression f.008a +f, cos 8 + f, cosy for the derivative in the direction given by the angles (a, B, 7). 3. Geometrical Interpretation. ‘The Tangent Plane. For a function u= f(x, y) all these matters can easily be illustrated geometrically. We recall that the partial derivative ‘with respect to x is the slope of the tangent to the curve in which the surface is intersected by a plane perpendicular to the zy-plane and parallel to the zu-plane. In the same way, the derivative inny THE TOTAL DIFFERENTIAL 65 the direction a gives the slope of the tangent to the curve in which the surface is intersected by a plane perpendicular to the zy-plane and making the angle a with the c-axis. The formula D., f(z, y) /,cosa + fysina now enables us to calculate the slopes of the tangents to all such curves, that is, of all tangents to the surface st a given point, from the ‘slopes of two such tangents, ‘We approximated to the differentiable function {= f(E, n) in the neighbourhood of the point (e, y) by the linear function SE D=fe + (E-Dfet+ - Who where ¢ and 7 are the current co-ordinates. Geometrically this Jinear function represents a plane, which by analogy with the tangent line to a curve we shall call the tangent plane to tho sur- face. ‘The difference between this linear function and the function F(E, 9) tends to zero as €—z—=h and y—y=k do, and in fact vanishes to a higher order than +/(H# + i). By the def- nition of the tangent to a plane curve, however, this states that the intersection of the tangent plane with any plane per- pendicular to the zy-plane is the tangent to the corresponding curve of intersection. We thus seo that all these tangent lines to the surface at the point (x, y, 0) lie in one plane, the tangent plane. This property is the geometrical expression of the differen- tiability of the function at the point (2, y, u=/(2, y)). I (& © ate current co-ordinates, the equation of the tangent plane at the point (2, y, u— f(x, y)) is f—u= (aft (o—wfe As has already been shown on p. 61, the function is differen- tiable at a given point provided that the partial derivatives are continuous there. In contrast with the case where there is only ‘one independont variable, the mere existence of the partial do- rivatives f_ and f, is not sufficient to ensure the differentiability of the function. “If the derivatives are not continuous at: the point in question, the tangent plane to the surface at this point may fail to exist, or, analytically speaking, the difference between S(e+ hy +k) and the function f(a, y)+ Afate, y) + Hla, y) which is linear in & and & may fail to vanish to higher order han (08 +). (aos)66 FUNCTIONS OF SEVERAL VARIABLES [Cxar. ‘This is cleaely shown by a simple example, We write fe y=0 it 20 of y= 0, J W=\2| i z—y=0 or tym. ‘Between thete lines we define the function in such a way that i¢ fs ropre ‘sented geometrically by planes. ‘The surface u = f(z, y) therefore consists of eight triangular pieces of planes, meeting in rooflike edges above the lines 2= 0, y= 0, y= and y= —z. This surface obviously has no tangent plane at the origin, although tho derivatives f,(0, 0) and f,(0, 0) tooth exikt and have the value 0, ‘The derivatives aro not continuous at the origin, however; in fact, as wo readily sce, they do not even exist on the ediges.* ‘The Total Difterential of a Function. ‘As in the case of functions of one variable, it is often con- venient to have a special name and symbol for the linear part of the increment of a differentiable function w= fla, y). We call this linear part the differential of the function, and write if(z, 9) du= ne Tea ae F dot fam act Fay. ‘The differential, sometimes called the total differential, is a function of four independent variables, namely the co-ordinates 2 and y of the point under consideration and the increments A * Another example of a similar typo is grvea by the Fonction ea fona pate Hatt yo «0 f en Gyn0 1 we introduce polar co-ordinates this becomes wm fein 2a, ‘The Sit derivatives with respect to and to y exe everywhere in the neighbour. hood ofthe origin and have the value a tho origin asl "han derivatives, However re not continuous at th org, for 2 “(vee vy Vee 7m) "ee the origi slong the z-xin tnd to 0, while it wo approtch srasin oy tts fo 1 is fanetlt w not dereniate at toe oy 1 thee peint to “Yanaoot plane to the murface w= fs 4) ox Por toe ‘uation {40, 0) = 0, 0} 0 chow that the tangeat plane would have to ‘Snide wilh the play's = 0. But at the pointe of tho line we Bave fin 2 Tan «2B “haw to dtanon of thn ga of the rice lane doesnot, as munt be tho cae with , anit to 8 higher order Chan ars é 3 F i1 THE TOTAL DIFFERENTIAL o and k, which are the differentials of the independent variables cor independent differentials. We need scarcely emphasize once more that this has nothing to do with the vague concept of “infinitely small quantities”. Tt simply means that du approxi- mates to Au— f(z + A, y + k) — f(a, y), the increment of the function, with’ an error which is an arbitrarily small frao- tion of (i+ ) (itself arbitrarily small), provided that A and & are sufficiently small quantities. Incidentally, we thus collect the expressions for the different partial derivatives in one formula, For example, from the total differential we obtain the partial derivative Loy putting dy = 0 and dz = 1. We again emphasize that to speak of the total differential of 4 function f(z, y) has no meaning unless the function is differen- tiable in the senso defined above (for which the continuity, but not the mere existence, of the two partial derivatives suffices). If the function f(z, y) also possesses continuous partial do- rivatives of higher order, we can form the differential of the differential df(c, y), that is, we can multiply its partial deriva- tives with respect to x and y by A= dr and k= dy respectively and then add these products. In this differentiation we must regard h and & as constants, corresponding to the fact that the differential df = hf, -+ Kf, is a function of the four independent variables 2, y, h, and &. We thus obtain the second differential * of the function, br Eiued Coe ksp ay = af) = Bee Tiere ye at . oy mB 2 dody + ae Similarly, we can form the higher differentials ef dep\=' hats dete 32h ead At Hae, * Wo shall lator so (p. 80 ef say.) that tho differentials of higher order intro- duced formally here cormspond exsoly to the terms of the corresponding order {a the Increment of the function.68 FUNCTIONS OF SEVERAL VARIABLES [Cxar. = aot + 4 OE “y= $4 E aatiy + Oar oy of + Aap ede + at and, as we can easily show by noon, in general mae aor (1) iy NY + 1) OF grays 2 +n) sagen detent + SE ay, ‘The last expression can be expressed symbolically by the equation ‘0 oy (Lact Lan)” = fate fay where the expression on the right is first to be expanded formally by the binomial theorem, and then the expressions OF an, OF dap er, seek deta, GO are to be substituted for the products and powers of the quan- tities fade and f, dy. For calculations with differentials the rule (fg) = fag + gf holds good; this follows immediately from the rule for the differentiation of a product. In conclusion, we remark that the discussion in this sub- section can immediately be extended to functions of more than two independent variables. aatay? a ay ‘5. Application to the Calculus of Error ‘Tho practical advantage of having the differential df = Rf + Kf, as convenient approximation to the increment of the function f(r, y), Au — S2+h, y+ B)—flz y) ns wo pass from (2, y) to (a+ h, y+), is exhibited particularly well in the ¢o-alled “ ealoulus of errors” (ef. Vol. I, p. 349). ‘Suppose for example, that we wish to find the possible error in the deter- mination of the density of a solid body by the method of displacement. If m ia the weight of the body in air and m ite weight in water, by Archi- modes’ principle the loas of weight (m ~ ia) is the weight of the water digpleood. If we are using the os. system of unite, the weight of thea) ‘THE TOTAL DIFFERENTIAL 69 ‘water displaced is numerically equal to ite volume, and hence to the ‘volume of the aolid. ‘The density « is thus given in terms of the indo- pendent variables m and 7 by the formula ¢ = m/(m — im). ‘The ecror in the measurement of the density « caused by an error dm in the mossure- ment of m and an error dm in the measurement of mis given approximately by the total differential dam am + 2 a, am + om By the quotient rao the parti derivatives are as as idm + man (mm ‘Thus the error in a is greatest if, say, dm is negative and dia is posti that is, if instead of m we measure too smsll an amount m ++ dm and instead of m too large an amount m + dm. For example, if « piece of brass vwoighs about 100 gm. in air, with a poasible error of 6 mg, and in water woighs about 88 gm. with « porsible error of 8 mg, the demity is given by our formula to within an error of about 88 ‘ about one per oent 5. Fonorrons or Functions (Comrounp Fuxcrions) AND THE Lwreopvorion ov New INDEPENDENT VARIABLES 1, General Remarks. ‘The Chain Rule Tt often happens that the function u of the independent variables z, y is stated in the form of a compound function uaflE m+.) where the arguments £, 9, ... of the function f are themselves functions of « and y: £= $29), 1= He, 9), We then say that FEn = fGe, 9), We, 9) is given as a compound function of z and y. Fe, 9)ye: FUNCTIONS OF SEVERAL VARIABLES [Cuar. ‘For example, the function ev sin(e + 9) Fey) ‘may be written as « compound function by means of the relations uae sing= (Gah Fem nasty Similarly, the function t= log (24 + x). arosin VI ‘can be expressed in the form =F= Fey) aro sin = f(E, ms E=VIRBHB, ne bogiet +). In order to make this concept moro precise, we adopt the following assumption to begin with: the functions ¢— ¢(z, y), 1= We, y),... are defined in a certain region R of the inde- pendent variables 2, y, As the argument point (2, y) varies within this region, the point with the co-ordinates (&, 7, - ..) always lies in a certain rogion $ of 9... -space, in which the fanction u=f(E, 7,...) is defined. The compound function 4 = FUG 9, He, y+ - = Fle, y) is then defined in the region R. In many cases detailed examination of the rogions R and 8 will be quite unnecessary, eg. in the first example given above, in which the argument point (=, y) oan traverse the whole of the zy-plano and the function w= ef sin’ ia defined throughout the Er-plane. On the other hhand, the second example shows the need for considering the regions ‘and § in the definition of compound functions. For the functions BaVI=B—F and y= logit +) tare defined only in the region 2 consisting of the pointe 0<2*-+ y* S1, that is, the region consisting of the circle with unit radius and cantre the origin, the centre being removed. Within this region | & | <1, while 7 ‘can have all negative values and the value 0. For the region 9 of pointa (€, 2) defined by these relations the function 7 are sin is defined, A continuous funetion of continuous functions is itself con- tinuous. More precisely: If the function u= {(€, 7, . ..) 48 continuous in the region 8, and the functions £= $(x, ys n= WX, ¥), -- « are continuous in the region BR then the compound function u—= F(x, y) is con- tinuous in R.1 FUNCTIONS OF FUNCTIONS n The proof follows immediately from the definition of con- tinuity. Let (ya) be @ point of R, and let 5 7... be the corresponding values of £, 7,.... Then for any positive ¢ the difference SE —FEo M0 ++) is numerically less than «, provided only that the inequalities Lé-&1<8 12-0] <8. are all satisfied, where 3 is a sufficiently small positive number. But by the continuity of g(a, y), Y(a, y),. - « these last inequalities are all satistied if le—l<» ly—vol<» where y is a sufficiently small positive quantity. ‘This establishes the continuity of the compound function, Further, we shall prove that a differentiable function of differentiable fanctions is itself differentiable. ‘This statement is formulated more precisely in the following theorem, which st the same time gives the rule for the differentiation of compound functions, or so-called chain rule: Uf £= dx, y), 7= Vx, yh, --. are differentiable functions of x and y in the region R, and {(8, 9, ...) is a differentiable function of & 9, .. «in the region 8, then the compound function A ($e, y), He, y+) = Fle, y) is also a differentiable function of x and y, and its partial deriva- tives are given by the formulae Fe=febet fhe +--+ Fy=fibut fu + or, briefly, by tebe tte ees y= 4 + tyne +» ‘Thus in order to form the partial derivative with respect to 2 we must first differentiate the compound function with respect to all the functions £, 7, ... which depend on 2, multiply each of these derivatives by the derivative of the function with respect to z, and then add all the products thus formed. This is tho generalization of the chain rule fora FUNCTIONS OF SEVERAL VARIABLES [Cuar. functions of one variable discussed in Vol. 1, Chapter IIT (. 158), Our statement can be written in a particularly simple and suggestive form if we use the notation of differentials, namely due = thgdls + teydy = tied E + ud + ie(Eutlt + Exdy) + (ned + nydy) + (ugbe + Wine + ++ )én-b (ugk + tyne +» «Ay. ‘This equation means that the linesr part of the increment of the compound function u—f(é, 7 -- -)= F(a, y) can be found by first writing down this linear part as if g 7, .. . were the inde- pendent variables and subsequently replacing d&, dn, . .. by the linear parts of the increments of the functions = 4(2, y), 7 = Wz,y),.. ... This fact exhibits the convenience and flexibility of the differential notation. Tn order to prove our statement we have merely to make use of the assumption that the functions concerned are differentiable. From this it follows that if we denote the increments of the independent variables z and y by Azand Ay, the quantities £, 9,... change by the amounts AE= baba + dudy + ade + ndy e+ bAy + abet radu, where the numbers ¢, ¢y)- ++» 71» Ya +. tend to 0as Ax and Ay do, or as y(Ac! + Ay?) does. Moreover, if the quantities 7, .. tundergo changes Ag, Ay, ..., the function «= f(é, 1 --.) is subject to an increment of the form bu = fe-+ fbn. of 8AE+ BAN t+ where the quantities 8,, 8,, ... tend to 0 as Ag, Ay, . . . do, or as (AG + An?-+ ...) does (and may be taken as exactly zero when the corresponding increments AZ, An vanish). If in the last expression we take the increments Ag, An, . . - 108 those due to a change of Az in the value of 2 and a chango of Ay in the value of y, as given above, we obtain bu= (fobs t+ Site + +» de + Seba + file + Day + Aa + yAge1 FUNCTIONS OF FUNCTIONS ” ‘Hore the quantities ¢ and y have the values est Sita +o + bedi + Hada + 8 tbs tees v= Sets t Siva t +++ by81 + tyBa + 18 + ya8at oes On the right we have a sum of products, each of which contains at least one of the quantities ty... 5 Py Ym + «By By «++ From this we seo that © and 7 also tend to 0-as Az and Ay do. By the results of the preceding section, however, this expresses the statement asserted in our theorem. It is obvious that this result is quite independent of the number of independent variables x,y, ..., and remains valid o.g. if the quantities € 7, ... depend on only one independent variable 2, so that the quantity w is a compound function of the single independent variable z. If we wish to calculate the higher partial derivatives, we have only to differentiate the right-hand sides of our equations with Tespect to x and y, treating f,, f,, ... a8 compound functions. Confining ourselves for the sake of simplicity to the case of three functions £, 9, and {, we thus obtain Waa Saab + Ste? + Fiche + Yorks t+ Sifeet Sites + Sikes fcbob e+ Syinatet Saheb + Silbatet bone) + fetal te) t SiglGabe + bole) + Seber + Site + Seben Uy Seabe® + Sos? + feces + Beaks + Bytaby + Yecbaby tebe + Sites + Seow + fee at yen tay’ 2. Bxamples.* 1. Let us consider the function une + taanang We put Ea atainty, n= Meysinzsing, Cay pe ty sins ny + Say conesing, y= Os Bet siny ony, ny = 2esinz ciny + 2eysinz cong, Sy — 295 ugmetteed, + We would emphasize that the following dfferentiations ean also be carried cout directly, without using the chain ralo. ° cane)74 FUNCTIONS OF SEVERAL VARIABLES [Cur Hence tg Det katoeaan He sinty + y sina sing + xy 0082 sing) and a, = 2etity + todeason + (a8 giny cosy + zsinz siny + sysinz cosy +). 2. Tn the case of the function wwe put E= 2+ 97, and obtain con(at + yt uy = By costa + 9") at sin(2t + ot) + 2 con(at + 9), Ygy = —Aey sina + 94), wee Aan ae + #) + Beotet 3. In tho caso of the fonetion wm aro tan(at + ay + yh int + 94) ttematain ‘tne wo ete SF erate et ty erat 3. Change of the Independent Variables. ‘A particularly important application of the facts developed on pp. 69-74 occurs in the process of changing the independent variables. For example, let w= f(€, 7) be a function of the two independent variables £, 7, which we interpret as rectangular co-ordinates in the éy-plane. If we introduce new rectangular co-ordinates 2, y in that plane (cf. p. 6) by the transformation Ear t By, 2= oF + asm, n= att By y= BE+ Ban the function u = f(é, 7) is transformed into a new function of z andy, SE 1) = Fe, y), and this new function is formed from f(£, 7) by a process of com- pounding such as was described on p. 69. We then say that new independent variables 2 and y have been introduced into the relation u = f(£, 1) between the independent variables € and 7 and the dependent variable 1.Wy FUNCTIONS OF FUNCTIONS on ‘The rules of differentiation given on p. 71 at once yield Uy thay + Ua thy = HB + 4B» where the symbols u,, tty denote the partial derivatives of the function F(z, y), and the symbols u, tu, denote the partial derivatives of the function f(, 7). ‘Thus the partial derivatives of any function are transformed according to the same law as the independent variables when the co-ordinate axes are rotated. This is true for rotation of the axes in space also. Another important type of change of the independent variables is the change from rectangular co-ordinates (¢, y) to polar co-ordinates (r, 0) which are connected with the rectangular co-ordinates by the equations t= 10088, r= Vert 9), in@, 9 = are.cos = aresi y = Very Pcs +A) On introducing the polar co-ordinates we have w= fla, y) = f(r 0088, r sin®) = F(r, 8), and the quantity 1 appears as a compound function of the inde- pendent variables r and 8, Hence by the chain rule we obtain = 42 = wu, 0080 — uy 4, 2 = 4S = w, 008 — 4 OE, We = theta + 080 =u, sind + u, °°! sam tr sin + ty ty = ated la ot "These yield the equation watt ud ut t Lud, which is frequently of use. By the chain rule the higher derivatives are given by Me He SO Hy BEE — gE HDD 4 tPE ay S088 809,76 FUNCTIONS OF SEVERAL VARIABLES [Cuar. hy = then thr 0080 sn — thy SESE + 4, SOHO ee sin?@— cost? | sin9 cos +4 ee eee ; q @ 20 Map = the int Uy SE yy sind 1 cos sin — a4, 28a ‘This leads us to the following formula, giving the expression appearing in the well-known “Laplace's” or “ potential ” equa- tion Au = 0 in terms of polar co-ordinates: = uu, 1, 1ilf, 2 (, 2), ee Aum teat tay tert tant Me a{ 5( a) +S. Of the formule y= u, Y= u_cosd + wy sin8, uy Mgy + yt = —ugs sin @ + uyr cos, which express the rules for the differentiation of a function f(a, y) with respect to r and 0, the first is the expression for the derivative of f(c, y) in the direction of the radius vector r which ‘wo previously met with on p. 64. In general, whenever we are given a series of relations defining a compound function, u=f ms E=$@y), 1= Hey). ‘we may regard it as an introduction of new independent variables 2, y instead of & 7,.... Corresponding sets of values of the independent variables ascign the same value to ts, whether it is regarded as a function of f, 7, ... or of z, y- Inall cases involving the differentiation of compound functions w=fEn-.) the following point must carefully be noted. We must distin- guish clearly between the dependent variable w and the func- tion /(é, 7, «. .) which connects u with the independent variables1) FUNCTIONS OF FUNCTIONS n fn, ‘The symbols of differentiation 14, w,,.. . have no mean- ‘ing until the functional connexion between u and the indepen- dent variables is specified. When dealing with compound functions u=flE 7 --)= Fle, y), therefore, we really should not write Up thy OF ty ty, but should instead write fy, f, oF Fa Fy respec- tively. Yet for the sake of brovity the simpler symbols tp, thy up, 0 often used when there is no risk that confusion will arise. ‘Tho following example will serve to show that the result of differentiating 1 quantity depends on the nature of the fanctional connexion between it and tho independent variables, that is, it depends on which of tho independent variables are kept fixed during the differentiation. With the “identical” transformation = 2, q—y the function w= 26+ deoomes w= 2r+y, and we bave wy— 2 uy= 1. If however, wo introduce the new independent variables & = (as before) and E+ = 6, we find that w= +0, so:that uy— 1, y= 1. That is, differentiation with respect to the same independent variable 2 gives different results in the to different eases. ‘Exauruse 1. Prove that the tangent plane to the quadsio ast yt totam tthe point (ee, Yo) is 24 + byyy + eazy = Le 2, If w= uz, is the equation of a cone, then Marty — Mes? = 0. 3, Prove that if a function f(2) is continuous and has a continuous derivative, then the derivative of the function fie) 2 1 Ife) a 2 Iya) % 1 ‘vanishes for a certain value between =, and ay. 4. Let f(z, y, 2) be a function depending only on r= V (et + y* + 2), fie let f(z, 2) = g(r). (2) Calculate fan + Soy + Sor () Prove thot if fon + fey + Jur 0st follows that f= © +B (where 4 and 6 are constant) r BU Sey Hy oes By) = 00) = Ve + aE oe + mM caonlate Sane, + gay tot Daag (2) = (cl. Bx. 2 p. 68).78 FUNCTIONS OF SEVERAL VARIABLES [Cuar. 6%. Find she expression fr fea + fey + fe in Uhree-dimensional polar co-ordinate, i, transform to the variables 1,8, @ defined by = rein8 coup y 2 = F008, Compare with example 4(a). 17, Prove that the expression sin sing. San + for is unchanged by rotation of the co-ordinate system. ‘8 Provo that with the linear transformation em ab + Bn yo toy fast Wh Seylts Yh yt 9) 200 respectively transformed by the seine law 5 the coeficients a,b,c of the polynomial at + Bhay + of. 6, Tak Mean Vanur Tazonem anp Tavior’s TazoneM vor Fowcrions oF SEVERAL VARIABLES 1. Statement of the Problem. Preliminary Remarks. ‘We have already seen in Vol. I (Chapter VI, p. 320 et seq.) how a function of a single variable can be approximated to in the neighbourhood of a given point with an accuracy of order higher than the n-th, by means of a polynomial of dogree n, the Taylor series, provided that the function possesses derivatives up to the (n+1)-th order. ‘The approximation by means of the linear part of the function, as given by the differential, is only the first step towards this closer approximation. In the case of functions cof soveral variables, o.g. of two independent variables, we may also seek for an approximate representation in the neighbourhood of a given point by means of a polynomial of degree n. In other words, we wish to approximate to f(r+ h, y+) by means of a “Taylor expansion ” in terms of the differences / and &, By a very simple device this problem can be reduced to what we already know from the theory of functions of one variable. Instead of considering the function f(a + h, y + k), we introduce yet another variable ¢ and regard the expression, FO) =fle+ Me y+ ty1m ‘TAYLOR'S THEOREM 9 ‘88 a function of t, keeping 2, y, h, and k fixed for the moment. As ¢ varies betwoen 0 and 1, the point with co-ordinates (+, y+ I) traverses the line-segment joining (2, y) and WHA y +h. We begin by calculating the derivatives of F(t). If we assume that all the derivatives of the function f(z, y) which we are about ‘to write down are continuous in a region entirely containing the line-segment, the chain rule (section 5, p. 71) at once givea F) = Met In F'O = When + Uikfey + fon and, in general, we find by mathematical induction that the n-th derivative is given by the expression FO = Met () Wferay + () Pf ah oR. which, as on p. 68, can be written symbotielly in the form FW) = (Wat Ifa). In this last formula the bracket on the right is to be expanded by the binomial theorem and then the powers and products of the quantities & ana g are to bo replaced by the corresponding n-th y « In all theso derivatives the argu- ments «+ ht and y+ it are to be written in place of @ and y. 2. The Mean Value Theorem. In forming our polynomial of approximation we start from ‘8 mean value theorem analogous to that which we already know for functions of one variable. This theorem gives a relation between the difference f(z + &, y-+ R) — fla, y) and the partial derivatives f, and f,. We expressly assume that these derivatives are continuous. On applying the ordinary mean value theorem to the function F() we obtain Ft) — FO) PO FO) _ wre,80 FUNCTIONS OF SEVERAL VARIABLES [Cuar. ‘where 0 is a number between 0 and 1, and from this it follows that Let Wy SWI) apse Ot, y+ 68) + Kf + Oht, y + kt). If we put t = 1 in this, we obtain the required mean value theorem for functions of two variables in the form FO+K, y+>R—F(@, y= Mel + Oh, y+ Ok) + Kft Oh, y+ Ok) = Mle + HAE, 0)- That is, the difference between the values of the function at the points (x +h, y +k) and (x, y) is equal to the differential at an ‘intermediate point (£, n) on the line-segment joining the two points. It is worth noting that the same value of 8 occurs in both fo and fp ‘The following fact, the proof of which we leave to the reader, is a simple consequence of the mean value theorem. A function ‘(2 y) whose partial derivatives fy and f, exist and have the ‘value 0 at every point of a region is a constant, 3. Taylor’a Theorem for Several Independent Variables. If we apply Taylor's formula with Lagrange’s form of the remainder (ef. Vol. I, Chapter VI, p. 324) to the function F(0) and finally put ¢= 1, we obtain Taylor’s theorem for functions of ‘two independent variables, Se + by y+ =H W + (fale M+ Hiker WV} oe a a ce +4 intent (iMate abot Hole 0} + Ry where R,, symbolizes the remainder term 1 Ry= we eet Gh, y+ Ok) + kyla + Oh, y+ Ok), o<é@ 0 and k-+0 we have Ry = 0 VET BP}. In the case of Taylor’s theorem for functions of one variable the passage (n> ©) to infinite Taylor series played an im- portant part, leading us to the expansions of many functions in power series. With functions of several variables such process jis in general too complicated. Here to an even greater degree than in the case of functions of one variable we lay the stress rather on the fact that by means of Taylor’s theorem the incro- ment (e+ h, y-+#)—f{e, y) of @ function is split up into increments df, @f, . . . of different orders. Ry Exanets 1. Bind the polynomial of the second dogree which best’ approximates to the function sins siny in the neighbourhood of the origin.& FUNCTIONS OF SEVERAL VARIABLES [Cuar. 2. If fle, v) is @ continuous function with continuous first and seoond derivatives, then = tim (2) 3) + $0, 0) fd = 8, Prove that the function e-*"* ean be expanded in a series of the form yn which converges for all values of = and y and that () H,(2) is & polynomial of degree n (so-called Hermite polynomials). (0) H'f2) = lly -s(2). (0) Bas — 2Hly + 20g (@) H',— 22Hl'y + 2nHly = 0. 4. Find the Tylor series for the following functions and indicate their range of validity: 1 @) dyert, v 7, Tae Avpucation or Vector Merops ‘Many facts and relationships in the differential and integral calculus of several independent variables take a decidedly clearer and simpler form if we apply the ideas and notation of vector analysis. We shall accordingly conclude this chapter with some discussion of the matter. 1. Vector Fields and Families of Vectors. The step which connects vector analysis with the subjects just discussed is as follows. Instead of considering a single ‘vector or a finite number of vectors, as in Chapter I (p. 3), we investigate a vector manifold depending on one or more con- tinuously varying parameters. If, for example, we consider a solid body occupying ® portion of space and in a state of motion, then at a given instant each point of the solid will have a definite velocity, represented by a ‘veotor #, We say that theso vectors form a vector field in the region in question. The three components of the Held vector then appear as three functions Mh Fy Fa) Male Be ty Ugly, tp %)rai) VECTOR METHODS 83 of the three co-ordinates of position, which we here denote by (2y, Zt) instead of (2, ys 2). ‘A case of a velocity field is represented in fig. 8, which shows the velocity field of a solid body rotating about an axis with constant angular velocity. ‘The forees acting on the points of a moving solid body likewise form a vector field. As an example of a force field we consider the attractive force per unit mass exerted by a heavy particle, according to Newton’s law of gravitation. By Newton's law all the vectors of this field of force are directed towards the attracting particle, and their lengths are inversely proportional to the square of the distance from the particle. If we pass to a new rectangular co-ordinate system by rotation of axes, all the vectors of the field will have new components with respect to the new system of axes. If the two co-ordinate systems are connected by aquations of the form (Chapter I, section 1, p. 6) fis aint Bim t yt far oat + Bate + 72% sagt + Bata + aa = aby + aybs + anh = Bibi + Babe + Babs a= nb + abe + Yaka8 FUNCTIONS OF SEVERAL VARIABLES [Cuar. respectively, then the relations between the components 2, tp) ty with respect to the 2-system and the components «,(f,, &, £4), alr, fy Es), a(S, £4» &4) with respect to the new €-system are given by the equations of transformation oy at + Bata + ms = a + Bit + Yala y= hy + Bata + Yas and thy aye age + ay ty = Bret Bat + Baty y= nent Yen + Yay respectively. (Cf. Chap. I, p. 6.) The components 1, ay, 0» in the now system thus arise from the introduction of the new variables and the simultaneous transformation of the functions representing the components in the old system. ‘When in physical applications each point of a portion of space hhas assigned to it a definite value of a function ws = f(ty, 2%, %), such as the density at the point, and we wish to emphasize thst the property is not a component of a vector, but on the contrary is a property which retains the same value although the co- ordinate system is altered, we say that the function is a scalar function or scalar; of, if we wish to emphasize the association ‘between the values of the function and the points of the portion ‘of space, we speak of a scalar field. Thus for every vector field 1 the quantity | ¢|*= 2 + ug! + ug? is a scalar; for it represents the square of tho length of the vector and therefore retains the same value independently of the co-ordinate system to which the components of the vector are referred. In the examples above the vector field ze is given us to begin with, and its components with respect to any system of rect- angular co-ordinates are therefore determined. If, conversely, in a definite co-ordinate system, say an 2-system, there are given ‘three Functions t(2%, ty 2); (24 2%), ty 74), these three functions define a vector field with respect to that system, tho components of the field being given by the three functions. To obtain the expressions for the components a, a a, in any other system we have only to apply the equations of transfor- mation deduced above.1) VECTOR METHODS 85 In addition to vector fields, we also consider manifolds of vectors called families of vectors, which do not correspond to each point of a region in space, but are functions of a parameter. We express this by writing 1 = 1(). If we think of 1 as a position ‘vector measured from the origin of co-ordinates in wtt'sD80°, then as t varies the final point of this vector desoribes # curve in space given by three parametric equations, = 80 = HO, t= xO Vectors which depend on a parameter ¢ in this way can be differentiated with respect to ¢. By the derivative of a vector ‘u(t) we mean the vector #(0) which is obtained by the passage to the limit im e+ = HO) oF and which accordingly has the components CT. ee wa Gh wah wa ‘We soe at once that the fundamental rules of differentiation old for vectors. Firstly, it is obvious that if whe waw to. w Further, the product rule applied to the scalar product of two vectors a and 9, #0 — ttt tty t MY (of Pe 7), gives (42) _ eo! + w'v. In the same way we obtain the rule a {sel = [uo'] + [#'0] for the veotor product.86 FUNCTIONS OF SEVERAL VARIABLES [Cnar, 2. Application to the Theory of Curves in Space. Resolution of a Motion into Tangential and Normal Components. Wo shall now make some simple applications of theso ideas, If x(t) is @ position vector in zzzzyspace which depends on a parameter t, and therefore defines curve in space, the vector (0) will be in the direction of the tangent to the curve at the point corresponding to t. For the vector x(t-+ #) — x(t) isin the direction of the line-sogment joining the points (?) and (¢-++ A) (cf. fig. 9); therefore so ia the vector 2¢+ = 2), which differs from it only in the factor 1/k, As k—>0 the direction of this chord approaches the dinec- tion of the tangent. If instead of £ wo introduce as parameter the length of ‘the arc of the curve meas- ‘ured from a definite starting-point, and denote differentiation with respect to # by means of a dot, we can prove that batt at this may also be written in the form Fie. —Diflerestinion ofthe poston a= w= 1, ‘The proof follows exactly the same lines as the corresponding proof for plane curves (cf. Vol. I, Chap. V, p. 280). ‘The vector 2 is therefore of unit length. If we again differentiato both sides of the equation 2a: — 1 with respect to s, we obtain aE = 0. This equation states that the vector # with components #(s), (6), #4(s) is perpendicular to the tangent. ‘This vector we call the cureature vector oF principal normal vector, and its absolute value, that is, its length ka = VP + E+ eM), Wwe call the curvature of the curve at the corresponding point.1 VECTOR METHODS 87 ‘The reciprocal p= 1/k of the curvature we call the radius of curvature, as before. The point obtained by measuring from the point on the curve a length p in the direction of the principal normal vector is called the centre of curvature. ‘We shall show that this definition of the curvature agrees with that given in Vol. I, Chap. V (pp. 280-8). For 2% is a vector of unit length. If we think of the vectors (6 + h) and 2(8) as measured from a fixed origin, then the difference (s + h) — (6) will be represented, as in fig. 9, by the vector joining the final points of the vectors £(s) and 2(s-+ h). If ais the angle between tho vectors -£(8) and 22(s ++), the length of the vector joining ‘their final points is 2 sin a/2, since 2(s) and 22(8 + h) are both of ‘unit length. Hence if we divide the length of this vector by and let b> 0, the quotient tends to the limit 1. Consequently fm = FV (ale +0) — 2008+ le + A) — ao) + (éals +B) — 28))7}. Here tho limit on the right is exactly V/(h*+ 22+ &%}) But a/h is the ratio of the angle between the directions of the tangents at two points of the curve and the length of are between those points, and the limit of that ratio is what we have previously defined as the curvature of the curve. "The curvature vector plays an important part in mechanics. We suppose that a particle of unit mass moves along a curve x(t), where t is the time. The velocity of the motion is then given both in magnitude and in direction by the vector 2'(), where the dash denotes differentiation with respect to ¢. Similarly, the acceleration is given by the vector x(). By the chain rale wwe have ds wane +e (where the dot denotes differentiation with respect to 8), and also ®s, 2 (a) Maen (3) . In view of what we already know about the lengths of 2 and 2, this equation expresses the following facts: ‘The “ acceleration vector ” of the motion is the sum of two88 FUNCTIONS OF SEVERAL VARIABLES [Cuar, ‘vectors. One of these is directed along the tangent to the curve, and its length is equal to * that is, to tho acceleration of the point in ita path (the sangential acceleration). ‘The other is directed towards the centre of curvature, and its length is equal to the square of the velocity multiplied by the curvature (the normal acceleration). 3. The Gradient of a Scalar, ‘We now return to the consideration of vector fields and shall give a brief discussion of certain concepts which frequently arise in connexion with them. Let u= fic, % %) be any function defined in a region of Zz_r-space; that is, according to the terminology previously adopted, u is a scalar quantity. We may now regard the three partial derivatives ate he omfy in the z-eystem as forming the throo components of a veotor 4. If we now pass to a now system of rectangular co-ordinates, the Eaystem, by rotation of axes, the new componente of the vector #4 aro given according to the formule of p. 6 by the equations ath, + Bitty + r1M4s agtty + Bgtly + yyy oat + Pata + ete On the other hand, if wo introduce the rectangular co-ordinates £y &y &y a8 new independent variables in the funotion / (2,2, %), the chain rule gives Se = ats + foPa + far Se Sata faBa t+ Loo Seg= fata + SusBa + fase fe a=Sy =o and we thus see that in the now co-ordinate system also the components of the vector # are given by the partial derivatives of the function f with respect to the three co-ordinates. Thus to every function f in three-dimensional space there corresponds Hence——_ ~ 1 VECTOR METHODS 89 1 definite vector, whose components in any rectangular co- ordinate system are given by the threo partial derivatives with respect to the co-ordinates. We call this vector the gradient of the function, and denote it by se = grad f. For a function of three variables tho gradient is an analogue of the derivative for functions of one variable. Tn order to form a graphical idea of the meaning of the gradient, we shall form the derivative of the function in the direction (a, a», a4), Whore ayy dg, dy are the three angles which this direction makes with the axes, so that costa, + costa, + costa, = 1. For this derivative we have already obtained the expression Df = fo, 0080 + fe, If we think of a vector e of unit length in the direotion (ay, a3, 4), this vector will have components ¢, = 008 a3, = C08 dy, €4= C08. ‘Thus for the derivative of the function in the direction (a, ap, a) we obtain the expression Def =e grads, the scalar product of the gradient and the unit vector in the direction (ay, a, a), Le. the projection of the gradient on that vector (of. Chap. I, p. 7). Iti this fact that accounts for the importance of the concept of gradient. If, for example, we wish to find the direction in which the value of the function increases or decreases most rapidly, we must choose the direction in which the above expres- sion has the greatest or least value. This clearly occurs when the direction of e is the same as that of the gradient or is exactly opposite to it. ‘Thus the direction of the gradient is the direction in which the function increases most rapitlly, while the direction opposite to that (of the gradient is that in which the function deoreases most rapidly; the magnitude of the gradient gives the rate of inorease or deorease ‘We shall return to the geometrical interpretation of the gradient in Chapter III (p. 124). We can, however, immedi- ately give an intuitive idea of the direction of the gradient. If in the first instance we confine ourselves to vectors in two dimen- 08.4 + fa, 008.5.\LP 9° FUNCTIONS OF SEVERAL VARIABLES [Cuar, sions, we have to consider the gradient of a function f(z, y). We shall suppose that this function is represented by its contour ines (or level lines) fan= in the zy-plane. ‘Then the derivative of the funetion f(z, y) in the direction (cf. p. 62) of these level lines is obviously zero. For if P and @ are two points on the same level line, the equation SUP) — f(Q) = 0 holds (the meaning of the symbols is obvious), ‘and the equation will still hold if we divide both sides by A, the distance between P and Q, and then let A tend to 0. The projec- tion of the gradient in the direction of the tangent to the level line is therefore zero, and hence at every point the gradient is perpendicular to the level line through that point. An exactly analogous statement holds for the gradient in three dimensions. I we represent the function f(xy, 2, %) by its level surfaces Sy % 2) the gradient has the component zero in every direction tangent to a level surface, and is therefore perpendicular to the level surface. In applications we frequently meet with vector fields which represent the gradient of a scalar function. The gravitational field of force may be taken as an example. IE wo denote the co-ordinates of the attracting particle by (Ex Ew &) those of the attracted particle by (zy #p ), and their masses by m and -H, the components of the force of attraction are given by the expressions boa a VGH aT Gat GaP an VG = PH P+ PP a VG BF GaP ‘Here C is a constant with the value yf, where y is the gravitational constant”. (The factors vee tare the cosines of the angles which the line through the two points makesny VECTOR METHODS o with the axes.) By differentiation wo see at once that these components fare the derivatives of the function o VG at GPF Ga with respect to the co-ordinates 2, Z 2 respectively. ‘The foree veotor ‘pert from « constant factor is therefore the gradient of the function 1 1 7 VG as Gat Ifa field of force is obtained from a scalar function by forming the gradient, this scaler function is often called the potential function of the field. Wo shall consider this concept from a ‘more general point of view in the study of work and energy (Chapter V, p. 350, and Chapter VI, pp. 415, 468-81). 4, The Divergence and Curl of a Veotor Field. By differentiation we have assigned to every function or scalar a veotor field, the gradient. Similarly, by differentiation wwe can assign to every vector field a certain scalar, known as the divergence of the vector field. Given a specific co-ordinate system, the zsystem, we define the divergence of the vector # as the function div 24 = Ot ie. 2, Or," Or, © Oey) ie, the sum of the partial derivatives of the three components ‘with respect to the corresponding co-ordinates. Suppose now ‘that we change the co-ordinate system to the é-system. If the divergence is really to be a scalar function associated with the vector field and independent of the particular co-ordinate system, we must have tiv ee Os Oy Be diem ae + ag, * ah where 0, ow are the components of 2 in the é-system. In fact, the truth of the equation uy, i Oe Bi, By Bo ae + ba, t ae, 96, * BB tA can be verified immediately by applying the chain rule and the transformation formule of p. 84.2 FUNCTIONS OF SEVERAL VARIABLES (Cua. Here we content ourselves with the formal definition of the divergence; its physico-geometrical interpretation will be dis- cussed later (Chapter V, section 5, p. 388). We shall adopt the same procedure for the so-called curl * of a vector field. ‘Tho curl is itself a vector r= cule whose components r, fp fp are defined by the equations yn O_O, G2 Bey on, Bey In order to show that our definition actually gives a vector independent of the particular co-ordinate system, we could verify by direct differentiation that the quantities a Bits tee, 5 Bey _ Gry Sy Gey mag, a’ OE a” 08 8," which define the curl in terms of the new co-ordinates, are con- nected with the quantities r,, 7, 75 by the equations of transfor- mation for vector components. Here, however, we shall omit these computations, since in Chapter VI, section 6 (p. 396) we shall give a physical interpretation of the curl which clearly brings out its vectorial character. The three concepts of gradient, divergence, and curl can all be related to one another if we use a symbolic vector with the @@ a " components —-, —, =. This bolic vector is often called ponent Ba Bay Gay an nablatt and is denoted by the symbol V. The gradient of a scalar field f(a, 2, 2%), grad f, is the product Vf of the scalar quantity f and the symbolic vector V, that is, it isa vector with the com- P xe, ¥ wee ‘The ourl of a vector field u(x, 2p, 9), curl 2¢, is the vector product [Vee] of the vector 2 and the symbolic vector V; finally, tho divergence is the scalar product div ae Y= * Often eal rotation (with the abboe Alter « Hebrew stringed instramentuN) VECTOR METHODS 93 In conclusion we mention a few relations which constantly recur. The curl of a gradient is zero; in symbols carl grad f= 0. As we easily see, this relation follows from the reversibility of the order of differentiation. The divergence of a curl is zero; in symbols div curl # = 0. ‘This also follows directly from the reversibility of the order of differentiation. ‘The divergence of a gradient is an extremely important expres- sion frequently occurring in analysis, notably in the well-known “ Laplace's ” or “ potential equation”. It is the sum of the three “ principal ” second-order partial derivatives of a function; in eymbole a, ast where Af is written as an abbreviation for the expression on the right.* ‘The symbol a, a, e 8= ana aa at div gn f= af— FE +f is called the Laplacian operator. Finally, we may mention that the terminology of vector analysis is often used in connexion with more than three inde- pendent variables; thus a system of n functions of m independent variables is sometimes called a vector fiold in n-dimensional space. The concepte of soalar multiplication and of the gradient then retain their meanings, but in other respects the state of afiairs is more complicated than in the caso of three dimensions. Baers 1. Find the equation of the so-called oseulating plane of @ curve 2= 0. y= (0 = Ma) at the point fy ic. the limit of the planes passing, through three points of the curve as those points approach the point with parameter 2, Show that the curvature vector and tho tangent veotor both Lie in the cscalating plane. * Tho notation Vif is alo used.oa FUNCTIONS OF SEVERAL VARIABLES [Cuar. 3, Let a= (8) be an arbitrary curve in space, such that the vector (6) is threo times continuously differentiable (s is tho length of are). Find the centre of the sphere of closest contact with tho curve at the point «. 4. If C ia a continuously differentiable closed curve and A a point not on @, there is a point B on O which has a shorter distenco from than any other point on . Prove that the line AB is normal to tho 6. If x = (8) is curve on a ephere of unit radius, the equation EA — 34) = (28) wats 8. hx = a9 en presse ection of ere the the vor 2E ith lt inh eating ea 17. ‘The limit of the ratio of the angle between the osculating plance ft two neighbouring points of a curve and the length of aro between, these two points, Le. the derivative of the unit normal veotor with respect to the are (s), is called the forsion of tho ourve. Let Es), Eis) denote the unit vectors along the tangent and the curvature vector of the curve (8); by E,(s) we mean the unit vector orthogonal to E, and Ey (the so-called binormal vector), which is given by [6,8]. Prove Frenet’s formulae b= te bile + Bale u-- shore Ijp = bis the earvature and /s the tonion of 40). 8, Using the vectors E,, ‘expressions for (a) the vector X, (b) the vector from the point = to the centro ofthe sphere of cnet cotatt at 0. Show that cure of sao tron «plane curve, 10%, Prove that if = wa, 9) ropreents the surtace formed by the tangenia fan aitray curve then (a) eeryonclating plan ofthe carve ie tangent plane to testa () ws 9) ats the oation Weatiyy — tay! = 0. AL, Prove that ‘curl curl 4¢= grad div 1 — Ave.1 ‘THE POINT OF ACCUMULATION 95 Appendix to Chapter II 1, Tae Priore or Tue Pousr or Accumutation IN SEVERAL ‘Diwexstons AND 11s AvPLicaTioNs If we wish to refine the concepts of the theory of functions of several variables and to establish it on a firm basis, without reference to intuition, we proceed in exactly the same way as in tho oase of functions of one variable. It is sufficient to discuss ‘these matters in the case of two variables only, since the methods are essentially the same for functions of more than two inde- pondent variables. 1. ‘The Principle of the Point of Accumulation. ‘We again base our discussion on Bolzano and Weierstrass's principle of the point of accumulation. A pair of numbers (x, y) will be called a point P in space of two dimensions, and may be represented in the usual way by means of a point with the rect- angular co-ordinates 2 and y in an 2y-plane, We now consider a bounded infinite sot of such points P(z, y); that is, the set is to contain an infinite number of points, and all the points are to lie in a bounded part of the plane, so that |z| 0 there is a sufic N= N(e) such that the distance between the points P, and Pry Vj — Xa) Gu— Yad tt less than «whenever both n and tm are greater than N. 2. Some Concepts of the Theory of Sets of Points. The general concept of a limit point is fundamental in many of the more refined investigations of the foundations of analysis based on the theory of sets of points. Although these matters are not essential for most of the purposes of this book, we shall mention some of them here for the sake of completeness. A bounded set of points, consisting of an infinite number ofuy] THE POINT OF ACCUMULATION 7 points, is said to be closed if it contains all its limit points; that is, limit points of sequences of points of the set are again points of the set. For example, all the points lying on a closed curve ‘or surface form a closed set. For functions defined in closed sets ‘wo can state the two following fundamental theorems: A function which is continuous in a bounded closed set of points assumes a greatest and a least vale in that se. A function which is continuous in a bounded closed set is uniformly continuous in that se. ‘The proofs of these theorems are so like the corresponding proofs for functions of one variable that we shall omit them. ‘The least upper bound of the distance between the points P, and P, for all pairs of points P,, P,, where both points belong to a set, is called the diameter of that sot. If the set is closed, this ‘upper bound will actually be assumed for a pair of points of the set. The student will be able to prove this easily, remembering ‘that the distance between two points is a continuous function of the co-ordinates of the points. ‘By using the theorem that a continuous function on a bounded closed set does assume ite least value, we can readily establish the following fact: if a point P does not belong to a closed set M, 1 positive least distance from P to M exists; that is, a point Q of ‘M exists sich that no point of M has a smaller distance from P ‘than @ has. This enables us to show that the closed regions defined in section 1 (p. 41) are actually closed sets according to the definition here. For lot C be a closed curve, and let R be tthe closed region consisting of all points interior to C or on C; ‘we have to show that all the limit points of R belong to R. We assume the contrary, i.e. that there is a point P not belonging to B which is a limit point of R. Then, in particular, P does not lie on C; hence by the theorem above it has a positive least distance from © (0 being a closed set). We can therefore describe a circle about P as centre, s0 small that no point of C lies in the circle; we have only to make the radius of the circle less than the least distance from P to C. The point P is outside C, since otherwise it would belong to R; and since every point in the small circle can be joined to P by a line-segmont which does not cross the curve C, every point of the circle lies outside C, and s0 no point of the circle belongs to R. But we assumed that P is a limit point of R, which requires that the circle should ‘ easy98 FUNCTIONS OF SEVERAL VARIABLES [Cuar. contain an infinite number of points of R. Hence the assumption that there is a limit point of R which does not itself belong to R leads to a contradiction, and our assertion is proved. The extension to closed regions R bounded by several closed curves is obvious. ‘A useful property of closed sets is contained in the theorem on shrinking sequences of closed ses: If the sets M,, My, My, . . . are all closed, and each set is con- tained in the preceding one, then there is a point (£, 9) which belongs to all the sets. In each of the sets Mf, let us choose point P,. ‘The sequence P,, must either contain an infinite number of repetitions of some one point, or else an infinite number of distinct points. If one point P is repeated an infinite number of times, thon it belongs to all the sets; for if M, is any one of the seta, P belongs toa set M,,, where m >, and M,, is contained in My. If there aro an infinite number of distinct points P,, then by the principle of the point of accumulation they possess a point of accumulation (£, 7). This point belongs to each M,. For when- ever m>n the point P,, belongs to M,, since it is a point of Mz which is contained in M,. Hence (2, 7) is o limit point of points P,, of Mf,, and since M,, is closed, (E, 7) is a point of I. ‘Thus in either caso there exists a point common to all the sets ‘M,, and the theorem is proved.* “A set is said to be open if for every point of the set we can find 1 circle about the point as centre which belongs completely to the set. An open set is connected if every pair of points A and B ‘of the set can be joined by a broken (polygonal) line which lies entirely in the set. The word “domain” is often used with the restricted meaning of a connected open set. As examples we have tho interior of a closed curve, or the interior of circle with the points of a radius removed. ‘The points of accumulation of a domain which do not themselves belong to the domain are called the boundary points. The boundary B of a domain D is a closed set, Hore we shall sketch the proof of this statement, ‘+The assumption that the sets fy a closed is eaentil, as the following cxamps shore Tat Bf be the ot 00 <4, Bach ots conti in the preceding, but no point belongs to all the ses. For if» ~ 0 the point belongs to no st, whil f= > Ot Belongs to no st Ay for which 2 <Hi} ‘THE POINT OF ACCUMULATION 99 A point P which is a limit point of B does not belong to D, for every point of D lies in a circle composed only of points of D and heneo devoid of points of B. It is also a limit point of D, for arbitrarily close to P we can find a point Q of B, and arbi- trarily close to Q we can find points of D. Hence P belongs to B. Tf to a domain D we add its boundary points B, we obtain a closed set. For every limit point of the combined set is either a limit point of B and belongs to B, or is a limit point of D and belongs either to D or to B. Such sets aro called closed regions, and are particularly useful for our purposes. ‘Finally, we define a neighbourhood of a point P as any open set containing P. If we denote the co-ordinates of P by (é, 7), ‘the two simplest examples of neighbourhoods of P are the circular neighbourhood, consisting of all points (x, y) such that @— + y—n<® and the square neighbourhood, consisting of all points (2, y) such that —€|<8 and |y—a| <3 3, The Heine-Borel Covering Theorem. A further consequence of the principle of the point of secumu- lation, which is useful in many proofs and refined investigations, is the Heine-Borel covering theorem, which runs as follows: Tf corresponding to every point of a bounded closed set M a neighbourhood of the point, say a square or a circle, is assigned, it is possible to choose a finite number of these neighbourhoods in such @ way that they completely cover M. ‘The last statement of course means that every point of M belongs to at least one of the finite number of selected neighbourhoods. By an indirect method the proof can be derived almost im- ‘mediately from the theorem on shrinking closed sets. We suppose that the theorem is false. The set M, being bounded, lies in a square Q. This square we subdivide into four equal squares. For at least one of these four squares, tho part of M lying in or con the boundary of that equaro cannot be covered by a finite number of the neighbourhoods; for if each of the four parts of 1M could be covered in this way, M itself would be covered. ‘This part of M we call M,, and we see st once that Mf; is closed.too FUNCTIONS OF SEVERAL VARIABLES [Cxar. ‘We now subdivide the square containing M, into four equal squares. By the same argument, the part M, of If, lying in or on the boundary of one of these squares cannot be covered by a finite number of the neighbourhoods. Continuing the process, ‘we obtain a sequence of closed sets By, My, My, ... , each en- closed in the preceding; each of these is contained in a square ‘whose side tends to zero, and none of them can be covered by a finite number of the neighbourhoods. By the theorem on shrink- ing sequences of closed sets we know that there is @ point (€, 7) which belongs to all these sets, and hence a fortiori belongs to M. To the point (&, 7) there accordingly corresponds one of the neighbourhoods, containing a small square about (£, 1). But since each M, contains (£, ) and is itself contained in a square whose side tends to 0 as 1/n does, each M, after a certain nis completely contained in the small square about (é, 7), and is therefore covered by one neighbourhood of the set. The assump- tion that the theorem is false has therefore led to a contradiction, and the theorem is proved. Exaweuss 1. A convex region R may be defined as a bounded and closed region with the property that if 4, Bare any two points belonging to 2, all pointe of the segment 4B belong to R. Prove the following state- mente: (ait If A is © point not belonging to R, there is a straight line passing through which has no point in common with R. (8)* Through every point P on the boundary of F there is a straight line 7 (a eo-alled “ line of support") euch that all pointa of B lie on one tnd the same side of 1 oF on Titel. (6) If a point 4 lies on the same side of every line of support as the points of F, then 4 is also a point of R. (@) Tho centro of mass of B ia a point of R. (@) A closed curve forms the boundary of a convex region, provided ‘that it has not more than two points in common with any straight line. (/)* A closed curve forms the boundary of a convex region, provided that ita curvature is everywhere positive. (It in asrumed that if the ‘whole curre is traversed the tangent makes one complete revolution.) 2. (a) If 8 is an arbitrary closed and bounded set, thero is one “least convex eavelope” Hof 8, ie. 0 set which (1) contains all pointa of 8, (2) is contained in all convex sets containing 8, (@) is convex.1 THE CONCEPT OF LIMIT ror (8) E may also be described in the following way: A point P is in B if, and only if, for every straight line which leaves all points of on one and the same side, P is also on this side, (©) The centre of masa of $ is a point of B. 2. Tue Conczpt or Luar ror Foxcvions oF Suvmrar Variances ‘We shall find it usoful to refine our conceptions of the various limiting processes connected with several variables and to consider them from a single point of view. Here wo again restrict our- selves to the typical case of two variables. 1, Double Sequences and their Limits. In the case of one variable we began with the study of se- quences of numbers a,, where the suffix m could be any integer. Here double sequences have a corresponding importance. These are sets of numbers Gua With two suffixes, where the suffixes m and n run through the sequence of all the integers independently of one another, so that we have e.g. the numbers May he Gan M199 Say May ae Aas = + + + Examples of such sequences are the sets of numbers Se om me We now make the following statement: The double sequence tom, converges as n> co and m-> © to @ limit, or more precisely’ a “double limit”, 1 if the absolute Aifference | aoe, — | is less than an arbitrarily’ small pre-assigned positive number e whenever n and m are both sufficiently large, that ts, whenever they are both larger than a certain number N depend- ‘ng only on «. We then write Thus, for example,toz2 FUNCTIONS OF SEVERAL VARIABLES [Cuar. and tim "2" — tim (, 2) capoeira Following Cauchy, we can determine, without referring to the limit, whether the sequence oonverges or not, by using the following criterion: ‘The sequence to, converges if, and only if, for every «> 0 a number N= N(e) exists such that | aq — Sy | <€ whenever the four suffizes n, m, n', m! are all greater than N. “Many problems in analysis involving several variables depend on the resolution of these double limiting processes into two successive ordinary limiting processes. In other words, instead of allowing n and m to increase simultaneously beyond all bounds, wo first attempt to keep one of the suffixes, say m, fixed, and let ‘n alone tend to ©. The limit thus found (if it exists) will in general depend on m; let us say that it has the value J. We now let m tend to «©. The question now arises whether, and if so when, the limit of |,, is identical with the original double limit, and also the question whether we obtain the same result, no matter which variable we first allow to increase; that is, whether we could have first formed the limit lim a, A, and then the limit lim A, and still have obtained the same result We shall begin by pining © goal ee of the poston from & few examples. In the case of the double sequence dg, = —!_, when min xed wo obviously obtain the rat lim dq fy = 0, aad therefore iis = Oy the same sels obtained Ho peor the page othe EEE in tha reve ent Vor thn soqace On Em however, we obiain lim ym = nd consequently Ura fg = 3s ‘while on performing the passages to the limit in the reverse order we fret obtaina] ‘THE CONCEPT OF LIMIT 103 Tira agg, = by = 0 and then oe lim 24 = 0. Jn thin case, then, the remult of the aucosiive limiting proceses i not Independent of their order: iz (iy) in (i. gg) {In addition, if wo let » and m increat beyond all bounds simultaneously, we find that the double lini fails to exis* “Another example i given by the sequence om 2, iy th able init im yest and haste valu 0, sc hem ‘merator of the fraction can never exoced 1 in absolute val, while thedenomi- ‘ator increases beyond all bounds. We obtain the same limitif we frst let m tend to 2; wo find that lim dq Xy—0, 20 that lim cover, wo wish to perform te pasages tothe limit inthe reverso order, keeping ‘mifixed and letting n increas beyond all bounds, we encounter the dificalty that lim sinn does not exist. Hence the resolution of the double limiting process into two ordinary limiting procestes cannot be carried out in both ways. ‘The position can be summarized by means of two theorems, ‘The first of these is as follows: If the double Limit lim aq lim gm = ty exists for every value of m, then the limit lim Ip also exists, and lim 1, =. Again, if the double limit oxista and exists, and the simple limit has the value J, and the limit lim dam — 2 exists for every value of n, then lim d, also exists and has the value J. In symbols: T= lim dy = lim (lim yq) = lim (lim dyn); ‘Hor if such # limit xistod it wosld necemarily have the value 0, inca se can nak tater clove tO by choosing »ltge enough aid choos mo nt. On th other hd, y= | ihensver'n ~ my no unter how lane in, “Toon ewo facta contra the maunption that the out him wake Bat oven win lim (im gq) ~ lim (Udy) the double limi lis my {ail to exist, an is shown by the example Gym 4 oma104 FUNCTIONS OF SEVERAL VARIABLES [Cua the double limit can be resolved into simple limiting processes and this resolution is independent of the order of the simple limiting processes. The proof follows almost at once from the definition of the double limit. In virtuo of the existence of lim dam = J, for every positive e there isan N = N(e) such that the relation | aq —1|<« holds whenever n and m are both larger than N. If we now ‘keep m fixed and let n increase beyond all bounds, we find that | lim Goq—1|=|ln—1| Se. This inequality holds for any positive « provided only that m is larger than W(«); in other words, it is equivalent to the statement lim (lim dyq) = ‘The other part of the theorem can be proved in a similar way. ‘The second theorem is in some respects a converse of the first, Tt gives a sulicient condition for the equivalence of a repeated limiting process and a double limit. This theorem is based on the concept of uniform convergence, which we define 1 follows: The sequence tga, converges as 2 -> cto the limit hy uniformly sn mn, provided thai the limit Lira tm, = Ip exists for every m. and in addition for every positive e i is possible to find an N= Nie), depending on ¢ but not on m, such that |Iq — Mpa | <€ whenever a>N. oo 5 n@tm)" mem uniformly to the limit iq = 2, an wo se immediately from the estimate Yat For example, the #0queneo dm = 1 arm |e — wwe have only to pot 4 2. On the other hand, the condition for uniform convergence dow nt holdin the cao of tho suet yg = For fixed valuos of m the equation lim dam = %,— 0 is always true; but the convergence isnot writer. or it any patioalar vale, aay 1/100, ia tanigoel to e then no tater how lage avalos of h we choose thero are Sheaye valoss fru for whi | SgeIy|— m ezoeds 0. Wo have nly to take m= 2n to obtain dqq = fy wich ee valve dieing trom the kins 0 by more than 1/100.uy ‘THE CONCEPT OF LIMIT 105 ‘We now have the following theorem: If the limit Yi Oggy = by exists uniformly with respect tom, and if further the limit lira y= existe then the double Tiit litt ym exists and has the value iad Tim (im dq) = lit Ge We can then reverse the order of the passages to the limit, provided that Titn ayg = A, exists By making uso of the inequality J enee — E] SS | dam — Foe | | Doe = ‘the proof can be carried out just as for the previous theorem, and we accordingly leave it to the reader. 2, Double Limits in the Case of Continuous Variables. ‘In many cases limiting processes oocur in which certain suffixes, 8. m, are integers and increase beyond all bounds, while at the same time one or more continuous variables =, y, ..., tend to limiting values gm, .... Other processes involve continuous variables only and not suffixes. Our previous discussions apply to such cases without essential modification. We point out in the first instance that the concept of the limit of a sequence of func- tions fy(2) of falz, y) a8 > 00 can be classified as one of these limiting processes. We havo already seen (Vol. I, Chap. VIII, - 998—the definition and proofs can be applied unaltered to functions of several variables) that if the convergence of the sequence f,(z) is uniform the limit function f(z) is continuous, provided that the functions f,() are continuous. ‘This continuity gives the equations £(8)= Jim, fe)= a (lima fa(2)): Jim (im Sala), which express the reversibility of the order of the passages to the limit n> «© and 2—> & Further examples ofthe part played by the question ofthe reversibility of the order of passages to the limit have already occurred, e theorem on the order of partial diferentiat106 FUNCTIONS OF SEVERAL VARIABLES [Cuar. other examples later. Here we mention only the case of the function ¥ For fixed non-zero values of y wo obtain the limit lim f(z, y)—= —1, while |. ‘Thus for fixed non-zero values of we have lim f(s 9) Bm, lin J 9) Lin, (lin fe, De ‘and tho order of the passages to tho limit is not immaterial. ‘This ia of courte connected with the discontinuity of the function at the origin. In conclusion we remark that for continuous variables the resolution of a double limit into successive ordinary limiting pro cesses and the reversibility of the order of the passages to the limit are controlled by theorems which correspond exactly to those estab- lished on p. 103. for double sequences. 3. Dini’s Theorem on the Uniform Convergence of Monotonic Sequences of Functions. In many refined analytical investigations it is useful to be able to apply a certain general theorem on uniform convergence, which we shall state and prove here. We already know (Vol. T, p. 387 ef 2eg.) that a sequence of functions may converge to a continuous limit function, even though the convergence is not uniform. In an important special case, however, we can conclude from the continuity of the limit that the convergence is uniform. ‘This is the case in which the sequence of functions is monotonic, that is, when for all fixed values of z the value of the function ‘ola) either increases steadily or decreases steadily as n increases, ‘Without loss of generality we may assume that the values inorease, or do not decrease, monotonically; we can then state the follow: ing theorem: If in the closed region R the sequence of continuous functions 4,(2z, ¥) converges to the continuous limit function f(x, y), and if at each point (x, y) of the region the inequality Susalt, y) 2 fale, y) holds, then the convergence is uniform in R. ‘The proof is indirect, and is a typical example of the use of the principle of the point of accumulation. If the convergence