eee DESCRIPTION OF SHAPE AND FORMEllis Horwood Series MATHEMATICS AND ITS APPLICATIONS Series Editor: Professor G. M. BELL, Statistics and Operational Research Editor: Professor B. W. CONOLLY, Chelsea College, University of London THE MATHEMATICAL DESCRIPTION OF SHAPE AND FORM ERIC A. LORD, Ph.D., formerly Research Fellow, and C. B. WILSON, Ph.D., Professor of Architectural Science, Department of Architecture, University of Edinburgh Neither Pythagoras nor Euclid could have foreseen today’s importance of geometry in modern technological systems of computer-aided design and manufacture. This work offers the first survey of the mathematical principles of shape and form and embraces a very wide spectrum of their applications. It will stimulate the current and growing revival of interest in the applications of geometry and will encourage creative thinking about problems of form, as for instance, in the behaviour of environmental fields in their interaction with buildings. It will widen awareness of such versatility of application in other fields whose workers can now gain a clearer perception of how these rich mathe- matical concepts can be utilised. It is a wide-ranging survey of the applications of mathe- matics to problems of form description, presenting clearly and concisely a large number of applied geometrical methods. A unified viewpoint, based on the concept of mappings between geometrical spaces, is developed so that the individual methods are seen in the context of a frame- work of underlying principles. The topics treated fall broadly into two major categories: those based on concepts of continuity (differential and metrical properties of continuous curves and surfaces, topology, structure of singularities, etc.) and those based on concepts of discrete point-sets (crystallographic groups, graph theory, surface and curve fitting and interpolation, lattice descriptions in pattern recognition and image pro- cessing, etc.). A degree of mathematical expertise on the part of the reader is assumed, in particular a good knowledge of func- tions, vectors, matrices and differential equations. However, the esoteric notation and terminology prevalent in texts meant for specialist mathematicians is avoided. The book is the first of its kind; no other known work has attempted this kind of unified presentation of geometrical methods taken from a wide range of disciplines. As well as stimulating further the revival of interest in geometry, it is intended as a first step towards the future development of an adequate science of morphology. It contains over 200 diagrams and an extensive bibliography. Readership: Mathematicians, physical scientists and engineers, particularly those in engineering design, computer science, computer graphics, pattern recognition and image processing, biological sciences, geography, urban planning and design, architectural design, art and graphics.Eric A. Lord has been Research Fellow in the Department of Architecture, University of Edinburgh since 1977. He obtained a first in mathematics from the University of Liverpool in 1964 and was awarded the Hudson prize. He studied general relativity theory at Kings College London, leading to a Ph.D. He has held research fellowships in theoretical physics in various institutes, including the Insti- tute of Theoretical Physics, Cologne, and the Indian Insti- tute of Science, Bangalore. C. B. Wilson was appointed to the Chair of Architectural Science in the University of Edinburgh in 1968, prior to which he was a Director of Building Physics. Between 1961 and 1966 he was Lecturer in Environmental Physics at Liverpool University. Since 1974 he has been Editor of the international journal Building and Environment. He is a graduate of Liverpool with a first degree in Mathematical Physics and a Ph.D. in Theoretical Physics.THE MATHEMATICAL DESCRIPTION OF SHAPE AND FORM E.A. LORD, Ph.D. formerly Research Fellow, Department of Architecture University of Edinburgh and C.B. WILSON, Ph.D. Professor of Architectural Science Department of Architecture University of Edinburgh ELLIS HORWOOD LIMITED Publishers - Chichester Halsted Press: a division of JOHN WILEY & SONS New York - Chichester - Brisbane - TorontoFirst published in 1984 and Reprinted in 1986 by ELLIS HORWOOD LIMITED Market Cross House, Cooper Street, Chichester, West Sussex, PO19 1EB, England The publisher’s colophon is reproduced from James Gillison’s drawing of the ancient Market Cross, Chichester. Distributors: Australia, New Zealand, South-east Asia: Jacaranda-Wiley Ltd., Jacaranda Press, JOHN WILEY & SONS INC., G.P.O. Box 859, Brisbane, Queensland 4001, Australia Canada: JOHN WILEY & SONS CANADA LIMITED 22 Worcester Road, Rexdale, Ontario, Canada. Europe, Africa: JOHN WILEY & SONS LIMITED Baffins Lane, Chichester, West Sussex, England. North and South America and the rest of the world: Halsted Press: a division of JOHN WILEY & SONS 605 Third Avenue, New York, N.Y. 10158, U.S.A. © 1986 E.A. Lord and C.B. Wilson/Ellis Horwood Limited British Library Cataloguing in Publication Data Lord, E.A. The mathematical description of shape and form. — (Ellis Horwood series in mathematics and its applications) 1. Geometry I. Title II. Wilson, C.B. 516 QA445 Library of Congress Card No. 83-26685 ISBN 0-85312-722-0 (Ellis Horwood Limited) ISBN 0-470-20043-X (Halsted Press) Printed in Great Britain by Unwin Brothers of Woking. COPYRIGHT NOTICE — All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the permission of Ellis Horwood Limited, Market Cross House, Cooper Street, Chichester, West Sussex, England.Contents PREFACE 7 1. INTRODUCTION 10 2. GEOMETRY: AN OUTLINE OF ITS HISTORICAL DEVELOPMENT 13 Congruence and the Euclidean group. Cartesian coordinates. Plane curves. Three dimensions. Space curves. Gaussian surface theory. Gaussian and mean curvature. Curvilinear coordinates and non-Euclidean geometry. Projective geometry. Klein’s Erlangen program. Transformations. 3. TOPOLOGICAL SPACES 38 Homeomorphisms. Polygonal symbols. Classification of surfaces. 4. MAPPINGS 47 Submersion and immersion. Description of mappings in terms of coordinates. The design of boat hulls. Map projections. [lumination engineering. Deformations in Euclidean space. Conformal mappings in Euclidean space. Infinitesimal deformations. Time-dependent deformations. 5. SINGULARITIES OF ANALYTIC MAPPINGS 74 Curves in two dimensions. Space curves. Scalar field in two dimensions. Mappings from two dimensions to two dimensions. Surfaces in three- space. Critical points of algebraic curves and surfaces. Envelopes. Removal of critical points and multiple points. Blum’s medial axis description. Thom’s catastrophe theory. Vector field in two dimensions. Vector field in three dimensions. Boundaries in fluid flow.6. MAPPINGS THAT ARE NOT ANALYTIC Isolated exceptional points. Branch points. Lines of exceptional points. Edges and corners. Polyhedra. 7. MINIMISING PRINCIPLES Deformation of a curve. Fermat’s principle. The catenary. Flexible bars and river meanders. Minimal surfaces. 8. GENERATION OF SHAPES Shape grammars. Ulam’s modular patterns. Trees and river systems. Symmetry. Regular patterns in a plane. Crystallography. Tessellations and space filling. Spiral phyllotaxis. Spiral forms in three dimensions. Generation of surfaces. 9. DISCRETE SPACES Finite spaces. Graph theory. Planar graphs. Euler’s formula. 10. FITTING OF CURVES AND SURFACES Interpolation. Bivariate interpolation. Coons’ surface patches. Approximate methods of curve fitting. Smoothing. Surface modelling. Linear operators. 11. THE DISCRETISED EUCLIDEAN PLANE Scalar functions on a square lattice. Finite difference methods. Binary patterns. Lattice description of curves. 12. FOURIER METHODS The Fourier series. Fourier series in two dimensions. Walsh functions. The Fourier integral. Bandlimiting. The discretised Fourier transform. BIBLIOGRAPHY AND REFERENCES 118 129 136 179 186 215 230 249Preface A method of form description is a procedure for selecting and presenting information about the characteristic way in which an object occupies space. Methods of form description are, in practice, always chosen for a particular purpose, in a particular problem context. The fundamental issue involved in the choice of a method of form description is the identification of which information should be included as relevant to the problem, and what manner of presentation of that information best serves the process of solution of the problem. We first encountered the problem of form description in the particular context of problems of prediction of the behaviour of environmental fields in interaction with buildings. A survey of mathematical methods of form descrip- tion, covering a much wider problem area, was undertaken in order to gain some insight into the criteria that operate in the choice of a method of form description. The literature contains a widely scattered conglomeration of dis- parate and, at first sight, unrelated methods. In most situations, there are no well-established criteria for arriving at an appropriate mode of description; approaches to form description are usually highly intuitive and pragmatic ad hoc responses to practical needs. There is a distinct lack of, and an urgent need for, a rigorously developed science of morphology which would deal with the underlying principles of form description. The mathematics of form description, as a possible field of study in its own right, has not hitherto been rigorously developed, or even adequately recognised. Thus, in this survey, we are not presenting a compendium of unrelated mathematical techniques. Instead, we have attempted to present a unified view of the mathematics of form description, emphasising underlying mathematical principles. The large number of applications reviewed or cited are chosen as illustrations of those principles. Our aim is to encourage more flexible and8 creative mathematical thinking about problems of form, to enable a reader whose field of study involves some aspect of form description to gain a clearer perception of how his or her particular speciality fits into a broader mathe- matical background, and to provide a first step towards the future development of an adequate science of morphology. Many problems of form have physical and dynamical aspects, as well as geometrical aspects. The material properties of building components belong to the ‘form’ of a building in a broad sense, and have to be taken into account along with the geometry in the determination of, for example, heat flow. The generation of the form of a living organism is brought about by a complex inter- play of physical forces within the organism and between the organism and its environment. In order to restrict the scope of our subject matter to manageable proportions, we have chosen to concern ourselves in this work (except in a few instances) with the purely geometrical aspects of form. In common speech, the words form and shape are not clearly distinguished. We have found it expedient in this work to increase the precision of these terms in the following way. We have chosen the word shape to indicate those aspects of geometrical form which have to do with the external aspect that an object presents to the world. The word form has been reserved to indicate that some aspect of internal structure is also under consideration. For example, we shall call the morphology of a physical field the form of the field, whereas the geometrical properties of the external surface of an object constitute its shape. We assume that the reader posesses a certain degree of mathematical exper- tise. The readership we have in mind consists of engineers, scientists, computer programmers, etc. We presuppose a good grasp of, for example, functions, vectors, matrices, and differential equations. However, the esoteric notation and terminology employed by professional mathematicians (in topology and differ- ential geometry, for example) has been avoided. Because of the scope of the subject matter, the treatment of each topic is necessarily extremely concise; our intention was to achieve this without sacrific- ing intelligibility. What we have sacrificed, in the interests of brevity, is detailed explanation of formulae (which we take to be self-explanatory for a mathematic- ally well-educated reader who is prepared to think hard and go slowly), and discussions of mathematical subtleties that would be demanded in a more extended and rigorous approach. A word about the bibliography: clearly, a comprehensive listing of all the literature that could be construed as having something to do with ‘mathematical form description’ is out of the question. The bibliography is highly idiosyncratic, and is intended to be. The works that we have chosen to include seem to belong to the following three categories: those which have substantially assisted in the development of our own knowledge about the subject, those which treat the various branches of mathematics on which the principles of form description are9 based and which in our view are clearly written, and those dealing with appli- cations which serve as representative examples of the use of those principles. The authors wish to thank the Science and Engineering Research Council for financial support. Acknowledgements We are grateful to the following for permission to reproduce copyright material: E. L. Brill for fig. 228 Camper and Nicholson (Yachts) Ltd. for fig. 26 Cambridge University Press for figs. 38, 85 and 101 W. H. Freeman and Co. for fig. 160 P. S. Stevens for figs. 132 and 133 J. P. Duncan, G. W. Vickers and the publishers of Computer-Aided Design for figs. 192 and 193 which first appeared in ‘Simplified method for interactive adjustment of surfaces’, J. P. Duncan and G. W. Vickers, Computer-Aided Design, 12,6 (Nov. 1980). In a few other cases it has proved impossible to contact copyright holders. We regret this and hope that no offence has been given.Chapter 1 Introduction Mathematics is a language. The use of this language for the purpose of form description tends to highlight certain aspects of form, while suppressing others; mathematics, in common with other languages, is able to express some things more easily and naturally than others. The study of the singularities of curves and surfaces (Chapter 5) is a prime example of the way in which the use of the mathematical language predisposes the user to investigate certain particular aspects of form (or shape). What is expressed by a mathematical description is an idealised form — an abstract concept. The extent to which this abstract con- cept can be usefully regarded as a description of the form of an object existing in the real world (or potentially existing in the real world, such as the design of a man-made object intended for production) depends on the extent to which characteristics of the object correspond to characteristics of the idealised form. Thus, the application of a mathematical mode of form description singles out certain characteristics of an object and ignores others. The suitability of the description in a particular problem context depends on whether the character- istics singled out are relevant to the problem which gave rise to the need for description. All description is purpose-oriented, and the choice of descriptive method must, in practical situations, be guided by the problem context. We have attempted to emphasise this aspect of form description in our choice of illustrative examples of practical applications. Historically, methods of form description have arisen in two ways. In the first place, there is the pragmatic way, whereby an ad hoc descriptive method is devised as a response to a particular problem need. This process is well illu- strated, for example, by the methods of curve and surface fitting (Chapter 10) and by the methods of feature selection involved in devising character-recognition programs (Chapter 11). The mathematical methods involved here would prob- ably not have arisen out of the evolution of mathematical science in the absence of interaction with needs arising in the ‘real world’. The above process can be contrasted with the development of mathematics for its own sake, that takes place independently of practical problems. Euclid,Introduction 11 for example, regarded his development of geometry as a self-sufficient exercise in pure thought; applications to practical problems were incidental and of lesser interest. Many branches of mathematics are fully justified by their elegance and intrinsic interest; the question ‘what use is it?’ is irrelevant in such cases. Without this kind of mathematics, the development of mathematical science would be severely handicapped. From a utilitarian viewpoint, the exploration of mathe- matics for its own sake can be justified by observing the large number of develop- ments in mathematics which find an application only after their discovery — sometimes a long time after. The Fourier methods (Chapter 12) are a striking example. In view of these arguments, we have felt justified in including some mathematical methods of form description which at present have found no practical applications (some of the topology, for example). Our principal aim has been to reveal the possibilities of mathematics as a tool for exploring the nature of shape and form and for developing imaginative ways of thinking about shape and form, rather than simply to present a review of current applied methods. Natural forms, in particular biological forms (and, in a manner less amenable to mathematical analysis, geological forms) are to a greater or lesser extent determined by the forces that have produced them. The role of physical forces in the generation of biological form has been explored by D’Arcy Thompson [T3] in his fascinating and classic work On Growth and Form. The same prin- ciple operates in certain man-made forms (e.g. large suspension structures such as tents), and a related principle is at work when a man-made form is designed with reference to the forces that act on it (suspension bridges, dams, aeroplanes, etc.). A thorough treatment of this aspect of form determination would venture into physics, and so lies outside the scope of our subject matter, which we have restricted to the purely geometrical aspects of shape and form. However, in Chapter 7, which deals with minimising principles, we have touched upon this aspect of form description. There is a natural subdivision of our subject matter, around which the book has been structured. The idealised form associated with a mathematical method of description may be either continuous, as for example when a circle is described by an equation, or discrete, as for example when the idealised form is a finite or infinite set of separate points. We do not wish to apply these terms in a rigorous way; they nevertheless serve the useful purpose of broadly classifying descriptive methods and thus clarifying the structure of our subject matter. In crystallography (Chapter 8) discrete description is applied to objects whose form possesses characteristics which make discrete description appropriate and natural; the forms described have properties of discreteness, the discreteness is not an artifact imposed by the mode of description. In other circumstances, the discreteness may belong only to the mode of description and not to the form described, as for instance when the shape of a continuous curve is indicated by12 Introduction locating a finite number of points on it, or when an image is presented to a machine, for recognition, as an array of ‘grey values’ (Chapter 11). In these contexts, problems of obtaining an approximate continuous description from a given discrete description arise (interpolation). Continuous descriptions, expressed in terms of known continuous func- tions, arise only in very special cases. They are usually concerned with approxi- mations in which the broad ‘smoothed out’ features of a shape or form are singled out because they have significance in a particular problem context. An example is the use of trend-surface analysis in cartography (Chapter 10). Numerical descriptions are by their nature discrete. Nevertheless, the principles of continuous description are important in discussing numerical methods, as a conceptual basis. A generic example is the numerical description of the topo- graphy of a piece of land by locating a finite number of points on its surface. The finite discrete description is conceived of as a description of a continuous surface (which is itself an idealised conception of the actual topography). The ideal of continuity is thus inherent in the topographical problem, even though the actual data are necessarily discrete. Chapters 1-7 are concerned with continuous methods of description. Chapters 8-11 are concerned primarily with discrete methods of description. The Fourier methods (Chapter 12) represent a kind of hybrid.Chapter 2 Geometry: an outline of its historical development Geometry is the branch of mathematics which abstracts from our perceptual experience of the shapes of objects in space and the positional relationships of objects to each other. The usefulness of mathematics as a tool for understanding the world results from the process of abstraction involved in mathematical description, which focuses on the general rather than the particular. For example, the formula for the volume of a cylinder is applicable to any cylinder. The earliest geometrical methods were concerned with the search for formulae of this kind, in response to practical problems of mensuration and surveying (the word ‘geometry’ comes from the Greek ynowerpta, ‘earth-measuring’). Formulae were often empirically derived, with little understanding of general principles or the notion of mathematical proof. Euclid recognised the underlying logical principles inherent in this body of knowledge. His systematic development of those prin- ciples is undoubtedly one of the greatest achievements of the human mind. A fascinating survey of the development of geometrical techniques, from the earliest beginnings to the early twentieth century, is Coolidge’s A History of Geometrical Methods [C8] . In this chapter we present a (necessarily highly condensed) survey of the principal developments that have taken place in geometry since Euclid. Our aim is to convey the ‘flavour’ of each branch of geometry and to emphasise the unity of the subject matter by indicating the way in which the various branches are interrelated. 2.1 Congruence and the Euclidean Group A concept of fundamental importance in Euclidean geometry is that of the congruence of two figures. Suppose we are given two figures in the Euclidean plane (or in a three-dimensional Euclidean space). And suppose that, for every point in one of the figures, we can identify a corresponding point in the other, in such a way that the distance between any two points in one figure is equal14 Geometry: An Outline of its Historical Development (Ch. 2 to the distance between the corresponding two points in the other. We then say that the two figures are congruent to each other. The correspondence between two congruent figures is a particular example of a mapping between two sets of points. Let A and B be two sets of points, and, for each point P in A, choose a point P’ in B. We then say that we have established a mapping from A to B, which associates, with every point P in A, a unique image P’ in B. If every image is the image of only one point, the mapping is called one-to-one, and if every point of B is an image, the mapping is called a mapping of A onto B. In an obvious sense, every one-to-one mapping of A onto B has an inverse, which is a one-to-one mapping of B onto A. In order to emphasise this two- way property of a one-to-one onto mapping, one speaks of a one-to-one corre- spondence between A and B. Under a one-to-one correspondence, every point of either A or B belongs to a unique pair of points, of which one is in A and the other is in B. Thus: Two figures in a Euclidean plane (or a Euclidean 3-space) are congruent if there is a one-to-one correspondence between A and B that preserves distances. A one-to-one mapping of a Euclidean plane onto itself that preserves dis- tances is called an isometry. Two figures in the plane are congruent if, and only if, there is an isometry for which one is the image of the other. Examples of isometries in a plane are: rotation about a fixed point, through a given angle, translation in a fixed direction, through a given distance, and reflection in a given line. The combination of two isometries is itself an isometry. There is an identity isometry, namely the mapping which associates every point of the plane with itself. Every isometry has an inverse. Hence, all the isometries of the Euclidean plane constitute a group, called the group of motions of the Euclidean plane, or the two-dimensional Euclidean group. Similarly, congruence in three-dimensional Euclidean space establishes the concept of the three-dimensional Euclidean group. 2.2 Cartesian Coordinates The first major advance in the development of Euclidean geometry, since the time of Euclid, was Descartes’ introduction of the coordinate system. A coordin- ate system for the Euclidean plane assigns to every point in the plane a pair of numbers (x, y). Geometrical problems concerning the properties of figures in the plane then become algebraic problems. In terms of a Cartesian coordinate system, the metrical properties of figures are dependent on the prescription s? = (x, —x2) +(¥1 --y2)? (2.1) for the distance between two points (x,, ¥,) and (x2, y2). For ‘infinitesimally close’ points (x, y) and (x + dx, y + dy), we can write ds* = dx? + dy? (2.2)Sec. 2.3] Plane Curves 15 This latter form is more convenient for studying the differential properties of curves, and the generalisation to curvilinear coordinate systems. We now have a structure on R,, the set of ordered pairs of real numbers, which is equivalent to the structure of plane Euclidean geometry. In terms of Cartesian coordinates, an isometry of the Euclidean plane can be described as follows: Let (x’, y’) be the coordinates of the image under an isometry, of the general point whose coordinates are (x, y). Since an isometry must map any straight line to a straight line, the isometry will be described by a linear relationship, of the form x'=ax+Byta ; (2.3) yryxtdy tb The condition that distances shall be preserved (ds’ = ds for any two ‘infinite- simally close’ points and their images) then restricts the matrix a8 a=(¢ ) (2.4) ATA=1 (2.5) to be orthogonal: (where the superscript T denotes matrix transpose). It will frequently be found convenient to employ an abridged notation. We denote the Cartesian components of a general point of the plane by (x, , x2) instead of (x, y), and introduce a summation convention: if an index appears twice in an expression, a summation over all the values of that index is implied. Thus, in the present context, (2.2) can be written in the form ds? = dxy dx (2.6) and the linear expressions (2.3) are expressed simply by Xq =Agg Xg + ag (2.7) The orthogonality of the matrix is expressed by _,. _ jl whena=8 7 {0 otherwise Cd) (ie: 5g denotes the set of components of the unit matrix). 2.3 Plane Curves The intuitive picture of a curve as the locus of a moving point leads to the parametric description of a plane curve, which expresses the Cartesian coordin- ates of points of the curve as continuous functions of a parameter t, that takes16 Geometry: An Outline of its Historical Development (Ch. 2 all values in some interval and which increases as we move along the curve ina chosen direction: a. (29) y=n(t) For any portion of a curve that is nowhere parallel to the y-axis, the x- coordinate can be chosen as the parameter, and we then have the description y =f) (2.10) An alternative description is by means of a function of two variables, o(x,y)=0 (2.11) which can be regarded, in principle, as being obtained by eliminating the para- meter from the two equations (2.9). A special form of the parametric description is obtained by choosing the parameter to be length s along the curve, measured from some arbitrarily chosen reference point on the curve. We then have a special form for the parametric description, et (2.12) y=y(s) If the two functions here are differentiable, the unit tangent vector to the curve at every point on it can be defined. Its components are (x, ¥), where the dot denotes differentiation with respect to s. Abrupt ‘corners’ on a curve are places where the functions are not differentiable. In this chapter, we shall always assume that any functions that we write down are continuous functions, and that any derivative appearing in formulae actually exists. An important descriptor of the form of a plane curve is the curvature kK at each point on the curve, defined as the reciprocal of the radius of curvature p. This is the radius of the circle touching the curve at the point under consider- ation, about whose centre the curve is turning as it passes through the point. From Fig. 1 it is readily apparent, from considering the limit as the point P’ is moved to P, that K=y (2.13) where the dot denotes differentiation with respect to arc length and p is the angle between the x-axis and the tangent to the curve. Also, x=cosy, ypHsiny (2.14) The intrinsic equation of a curve expresses the curvature asa function of distance along the curve:Sec. 2.3] Plane Curves 17 Figure 1 K=K(s) (2.15) Whereas the descriptions (2.9) and (2.11) contain information about the position and orientation of the curve relative to the coordinate axes, as well as infor- mation about the shape of the curve, the intrinsic equation describes only the shape of the curve, without reference to any coordinate system. Integrating (2.13) and (2.14) gives y =| cds (2.16) x = [cos wds, y = {sin wds (2.17) These equations provide a prescription for obtaining a parametric description of a curve (with arc length as the parameter) from an intrinsic description. The three constants of integration of course are the missing information required to specify a particular position and orientation.18 Geometry: An Outline of its Historical Development [Ch. 2 2.4 Three Dimensions If (x, y, z) are the Cartesian coordinates of a general point in a three-dimensional Euclidean space, the distance ds between two ‘infinitesimally close’ points is given by ds* = dx? + dy? + dz? (2.18) The coordinates (x, y, z) of a point are of course the components of its position vector r, so (2.18) can be written more simply as ds? = dr-dr (2.19) A set of points in Euclidean 3-space whose position vectors are continuous single-valued functions of two parameters, r=r(u,v) (2.20) describes a surface. The two families of curves u = const. and v = const. lie on the surface, and are called coordinate curves. Every pair of values (u, v) deter- mines a unique point on the surface. The numbers (u, v) are the curvilinear coordinates of the point. Alternative descriptions of a surface are analogous to the descriptions (2.10) and (2.11) of a closed curve. The description z=f(x,y) (2.21) is Monge’s form for the description of a surface. It can be regarded as being obtained from the parametric form (2.20) by choosing x and y as parameters. A surface can also be described by a single relationship satisfied by the Cartesian coordinates of its points, $(x,¥,z)=0 (2.22) Monge’s form is, of course, a special case of such a description. The parametric description of surfaces encounters the difficulty that, for some surfaces, a single parametrisation for the whole surface does not exist. For example, if longitude and latitude are used as parameters for the surface of a sphere, the assignment of a pair of curvilinear coordinates to each point fails for the two poles. In fact, any curvilinear coordinate system on a sphere necessarily fails at at least one point. For this reason, it is necessary in general to regard a surface as covered by a set of overlapping ‘coordinate patches’, on each of which is imposed a different parametrisation. A different parametric description (2.20) then describes each patch of the surface, and the complete description of the whole surface consists of the collection of these separate descriptions. The abridged notation of §2.2 is readily extended to three dimensions. Writing x; (@=1,2,3) for the Cartesian coordinates of a general point in a Euclidean three-space, the summation convention enables us to write (2.19)Sec. 2.5] Space Curves 19 in the form ds? = dx; dx; (2.23) Denoting by x; the coordinates of the image, under an isometry, of a point whose coordinates are x;, the isometry is expressed by x; = ay Xj +a; (2.24) in which the matrix is orthogonal: Qik Qk a by (2.25) 2.5 Space Curves A parametric description of a curve in Euclidean three-space is related to the idea of a curve in 3-space as the locus of a moving point. Distance along the curve (from some arbitrarily chosen point on it) can be used as the parameter; we then have a description of the curve by an equation of the form r=r(s) (2.26) Denoting differentiation with respect to s by a dot, the vector t=r (2.27) is a tangent vector to the curve at every point. By dividing (2.19) by ds”, we see that it is a unit vector. Differentiating t-t = 1 shows that t is perpendicular to t, that is, it is normal to the curve. The plane containing the vectors t and t for a point on the curve is the osculating plane at that point. The unit vector normal to the curve, in this plane, is denoted by n, and called the principal normal. We can then write t=xn (2.28) The factor x is called the curvature of the curve. It is clearly the curvature of the plane curve obtained by projecting the space curve onto the osculating plane at the point under consideration. The binormal to the curve, at any point, is a unit vector perpendicular to the osculating plane, defined by b=txn (2.29) We now have a triad of orthogonal unit vectors that changes its orientation as we move along the curve. The plane containing the vectors n and b is the normal plane and the plane containing b and t is the rectifying plane. Since n is a unit vector, n is perpendicular to it and is therefore a linear combination of t and b. The coefficient of t isn-t = —n-t = —x. Hence we can write a= —Kt+7b (2.30)