- ME 251 Lecture 1
- Classical Physics '10
- CHECAL_CI1MultipleChoice
- Smithsonian Physical Tables 9th Revised Edition
- Syllabus PCS
- einstein.pdf
- Dimensional Analysis
- Chap 01 SM
- question on dimension.docx
- The Fairy Doll La Poupee Ued20.PDF
- 01_B Units
- Arihant Aieee Physics [Nitin m Sir]
- Consequence Analysis by Means of Characteristic Curves to Determine Damage to Buildings From Explosion
- Dam Busters 2 Solution
- Arihant AIEEE Physics
- Differential Equations Handouts
- Just the Maths - A.J.hobson (Complex Numbers)
- Assignment MA02 2017
- Full Syllabus
- 2016_Lecture-Physical Modeling (1)
- Lecture_Ch01 Introduction, Measurement, Estimating
- Ratna Output Spss
- Bt Btech Syllabus
- Introduction
- 3. Integration
- Current_Syllabus_for_B.Arch_and_BArch(ID).doc
- Wallis Rule
- Ae Notification-tsspdcl - Final
- ac1
- A Matlab Environment for Analysis of Fluid Flow and Transport Around a Translating Sphere
- Filosofía de Vida
- verdades no demostrables.pdf
- When is 0.999... Less Than 1.pdf
- Signo de una Permutación.pdf
- Cycloid
- La Anarquía Explicada a los Niños
- Capital Cities of the World
- Las matemáticas no me han servido para nada

Mathematics

for Chemists

P. G. Francis

Department of Chemistry,

University of Hull

LONDON

NEW YORK

Chapman and Hall

**First publis~d 1984
**

by Clulpmml and Hall LId

I I New Feller !..mIe. London EC4P 4EE

in Ihe USA by

Chapman and Hall

733 Third Avenue. New York NY 100/7

Publis~d

**10 1984 P. G. Francis
**

$Qftcover reprint of the hardcover 1st edition 1984

ISBN-13: 978-94-01()..8950-O

e-ISBN-13: 978-94-1109-5552-3

001: 10.10071978-94-009-5552-3

**This litle is available in bolh hardbound and paperback
**

editions. The paperback edition is sold subjtcllO t~ condition

tlull it slullillDI, by way of trade or ot~rwise, be lenl, re-sold.

hired aul. or olherwise circulated without the publisher's prior

consent in any form of binding or caverOlher titan tha.t in which

il is published and withoul a similar condition including this

condition being imposed on the subseqlUmt purcluJser.

All rights reserved. No part of this book may be reprinted, or

reproduced or utilized in any form or by any electronic ,

mechanicol or other meOllS. now known or hereafler inVf'nted.

including pholocopying and recording. or in any information

storage and retrieool system. wilhoul permission in writing

from Ihe Publisher.

Brhiala Library Cataloauill, ill PliblicatiOil Data

Francis, P.O.

Mathematics for chemists.

I. Chemistry- Mathematics

I. Title

510'.2454

QD42

**Francis. P. G. (Patrick Geoffrey), 1929Mathematics for chemists.
**

Bibliography: p.

Includes index

1. Mathcmatics-I961

2. ChemistryMathematics. I. Title.

QA37.2.F73 1984 510'.2454 84- 1802

Contents

Preface

ix

**1 Algebraic and geometrical methods
**

1.1 Natural numbers

1.2 Units and dimensional analysis

1.3 Functional notation

1.4 Quadratic and higher-order equations

1.5 Dependent and independent variables

1.6 Graphical methods

1.7 Some geometrical methods

1.7.1 Similar triangles

1.7.2 Triangular graph paper

1.7.3 Three-dimensional geometry

1.7.4 Circle, ellipse, parabola and hyperbola

1.7.5 Plane polar coordinates

1.8 Factorials and gamma functions

1.9 Probability

1.10 Complex numbers

1

1

4

8

8

10

12

14

14

14

15

17

21

22

23

24

2 Differential calculus

2.1 Significance and notation

2.2 The calculus limit

2.3 Differentiation of simple functions

2.4 The use of differentials; implicit differentiation

2.5 Logarithms and exponentials

2.6 The chain rule and differentiation by substitution

2.7 Turning points: maxima, minima and points of

inflection

2.8 Maxima and minima subject to constraint; Lagrange's

method of undetermined multipliers

29

29

30

34

36

38

39

41

45

3 Integration by parts 4.2 Power series and Taylor's theorem 2.9.4 Inversion of a series 2.11 The principles of Newtonian mechanics 3 4 Differential calculus in three or more dimensions.9.2 Reciprocal functions 4.9.5 Exact differentials 3.10 The evaluation of limits by L'Hopital's rule 2.1 Simple functions 4.9.2.2.7 Extensive and intensive variables.9.9.3 Standard forms of integral and numerical methods 4.2. Euler's theorem 3.2.4 Integration by substitution 4.5 Differentiation of integrals.1 Geometric series 2.2 Scalar and vector fields Integration 4.1 Significance and notation 3.2 Standard methods of integration 4.5 Expansion of algebraic functions 4.1 Scalar and vector products 3.8 Taylor's theorem in partial derivatives 3.2.5 Empirical curve fitting by power series 2.6 Relations between partial derivatives 3.vi Contents 2.6 Integration by partial fractions 4.3 The total differential 3.1 Significance and notation 4.2.9.2 An alternative approach to calculus 3.6 The Euler-Maclaurin Theorem 50 51 51 54 58 60 61 63 73 73 77 78 80 82 84 86 88 89 92 95 98 98 101 101 101 102 104 104 105 106 107 108 109 .9 Vectors 3. partial differentiation 3.4 Multiple integration 4. Leibnitz's theorem 4.4 General expression for a total differential 3.9 Series 2.3 Maclaurin's theorem 2.

Contents vii 5 Applications of integration 5. first degree 6.4 The statistical analysis of experimental data 7.1 Plane area 5.4.3.2.1 Separable variables 6.2 Root-mean-square error 7.4.3 Exact equations 6.6 Small-sample errors 165 115 117 120 122 123 124 124 127 129 132 138 140 140 142 144 145 147 148 156 159 160 160 163 165 166 168 169 170 171 .6. polar coordinates in three dimensions 5.3.4 Linear equations of first order 6.6 Applications of multiple integration 5.2.3 Linear equations of second order 6.1 Significance 7.8 Generalized dynamics 115 6 Differential equations 6.2 Interactions in a real fluid 5.2 Plane elements of area 5.2.2.2 First-order homogeneous equations 6.7 The calculus of variations 5.1 Laplace transforms 6.3 Linear differential equations 6.1 Significance and notation 6.2 Equations of first order.5 Curve length by integration 5.5 Propagation of error 7.3.4 Line integrals 5.2 General linear equation with constant coefficients 6.2 Fourier transforms 138 7 Experimental error and the method of least squares 7.3 Distribution of error 7.4 Integral transforms 6.1 Homogeneous linear equations with constant coefficients 6.6.1 The pressure of a perfect gas 5.3 Elements of volume.

Contents viii 7.8.8.8.6 Multivariable and non-linear least-squares analysis 174 176 176 179 180 183 183 184 Appendix SI units. physical constants and conversion factors. the Greek alphabet and a summary of useful relations Index 185 190 .7 7.2 Covariance and correlation coefficient 7.8.5 Weighting of observations 7.8.8 The normal distribution of error The method of least squares 7.4 Least-squares straight line with both variables subject to error 7.1 Linear relation between two variables 7.8.3 Uncertainty in the slope and intercept of the least-squares straight line 7.

often corresponding to a simplified physical model.Preface This text is concerned with those aspects of mathematics that are necessary for first-degree students of chemistry. When these are expressed in mathematical terms we can add mathematical rigour to produce theoretical structures in which only the basic premises. leave room for uncertainty. and that the connection between physical principles and their mathematical formulation requires at least as much study as the mathematical principles themselves. It is written from the point of view that an element of mathematical rigour is essential for a proper appreciation of the scope and limitations of mathematical methods. deduced from seemingly disconnected experimental observations. or axioms. who continue to give support and advice both in the application of established techniques and in the development of new approaches. These . As an experimental science. which is based upon a small number of premises. It is also a subject that draws particular benefit from having available two rigorous disciplines. The benefit of rigour is that it provides a degree of certainty which is valuable in a subject of such complexity as is provided by the behaviour of real chemical systems. we attempt in chemistry to understand and to predict behaviour by combining precise experimental measurement with such rigorous theory as may be at the time available. Chemists in particular need insight into threedimensional geometry and an appreciation of problems involving many variables. those of mathematics and of thermodynamics. Experimental scientists have added to this a rigour of a different kind. These provide the experimental laws of thermodynamics and quantum theory whose justification lies in the absence of disagreement with experiment. these seldom provide a complete picture but do enable areas of uncertainty to be identified. It is written with chemistry students particularly in mind because that subject provides a point of view that differs in some respects from that of students of other scientific disciplines. Mathematical rigour has been provided for us by generations of professional mathematicians.

It is a subject on which many admirable texts have been written. A proper acknowledgement of the many sources from which the author has drawn cannot be justly be made for a text in which the material was largely developed during the last two. The examples are. by the inclusion of a number of mathematical proofs and by the inclusion of a discussion of the treatment of experimental data.or three-hundred years. for which purpose worked examples are more valuable than exercises. given with their solutions.x Preface considerations are reflected in this text by frequent reference to physical principles. The later sections contain brief outlines of the principles and scope of more advanced techniques and are intended to point the way to more specialized texts. this is in order that the text may be used as a reference source. so that an initial reading of only the earlier sections in each chapter is possible. . the only justification for adding to their number being changes in emphasis. The subject matter is developed during each chapter. Readers are strongly recommended to attempt their own solutions to these examples before consulting the one that is given. in the main. It is assumed that the reader has previous mathematical knowledge extending to the elementary applications of calculus.

for example. The definition of angle is angle = (length of arc)/radius. and it is also used as the natural measure of angle. this unit being called the radian. that arise naturally. The circumference of a complete circle of radius r being 27tr. These can be regarded as the mathematical equivalent of natural units and occur so often in scientific formulae and equations that we consider how they originate. so that the numerical values obtained depend on the size of the units and are not significant in themselves. This ratio is a pure number that is independent of the size of the circle. 7t. We also have the conventional definition of such a full rotation . units (Section 1.CHAPTER 1 Algebraic and geometrical methods 1. There are some pure numbers. but arbitrary. Our first natural number.1 Natural numbers Quantitative measurements in science are usually made in terms of appropriate. this has a numerical value that is independent of any conventional scale of measurement. could be defined as the ratio of circumference to diameter for a circle.1a) and for a given angle. the angle corresponding to a full rotation is 2nr/r = 2n natural units of angle. where the number 3 corresponds to the three degrees of freedom of the particles. such as the gas constant R. (1. the heat capacity at constant volume of an ideal monatomic gas at a high enough temperature is 3R/2. do numerical values assume absolute significance.2). Only when there is a natural unit. the ones most commonly met in science being 7t and e. however.

e. is the one preferred by mathematicians and is often the one needed to solve the more difficult problems.1 b) = 41tr2/r2 = 41t. The wise rule is to use geometrical methods with care and to defer to analytical methods when in doubt. This division into geometrical methods and analytical methods is arbitrary since they are parts of the same whole and a combination of the two is frequently used in practice. . The above definition of 1t is geometrical and belongs in the field of diagrams and models. cos (J. called axioms. It is useful to divide the mathematical methods used in science into two categories. as such. an example being provided by the trigonometric functions sin (J.•. so that 1t radian = 180°. (1. on the other hand. The geometrical definition of sin (J is . This is generally the more powerful of the two mathematical methods and. Sometimes the connection between the two methods is not obvious. arises in a different way.3) . It is a useful geometrical concept that (area)/(radius)2 is constant for a given solid angle. The second category is algebraic and numerical. Our second natural number. The definition of angle in two dimensions as (arc length)/radius can be extended into three dimensions. tan (J. geometrical methods and analytical methods..2 Algebraic and geometrical methods as 360 degrees (360°).. For a complete sphere of surface area 41tr 2. SIn (J opposite side hypotenuse = -"'-=----- in a right-angled triangle. Geometrical methods. often seem simpler but are sometimes deceptively so.. Analytical methods are developed by strictly logical deductions from first principles. and does not depend on our ability to construct models and diagrams. and this is used in Section 5. The unit of solid angle is called a steradian. if we take a conical segment from a sphere we can define a three-dimensional solid angle as solid angle = (area of segment of spherical surface)/(radius)2.2) whereas the analytical definition is sin (J = (J3 (J - (J5 (J7 3! + 5! -71 + .. and since this expression does not contain r it is also a pure number that is independent of the size of the sphere. complete solid angle (1. (1.7.

(1. (1. belongs in analytical methods..Natural numbers 3 We can see that these two definitions are the same in a particular case by taking the angle eto be.6! + . As mentioned above. The derivative is used to define the slope of the tangent to a curve. 90° = nl2 radian. It is defined by the series x2 x3 e = 1 + x + 2! + 3! + (1. .5) gives eO = 1. Putting x = 0 in (1. .. and the angle between two straight lines can be obtained as the inverse of the cosine series. Then sin e = 1 by the geometrical definition.2! + 4! (J6 . and if we put e = nl2 = 1.. .5) so that when x = 1. = 2.3) are equivalent to each other. and only eX.8). This corresponds to the quite common situation where the rate of change of a quantity is proportional to the quantity itself. gives exactly the same quantity when it is differentiated (Chapter 2).. One way is to use the definition of cos e as the series cos (J = e2 e4 1 .000. . This is a particular case of the general rule that any quantity raised to the power of zero is unity. which follows from the laws of indices: and so that and when x = y. and that the sum tends towards a limit of 1..71828 . In this way. More fundamentally. geometry can be developed by the analytical approach leading eventually to showing that (1. e. our second natural number. .5708 into the series (1.4) This is then the derivative (Chapter 2) of sin (J. as in first-order reaction kinetics.3) it is easily shown that successive terms become rapidly smaller and smaller. The fact that the two definitions of sin e are always the same can be proved only by means of extensive analytical argument. 1 e = 1 + 1 + 2! 1 + 3! + . say.6) This arises naturally in science because e. exponential relations arise as a result of the simple probability that determines the Boltzmann distribution which underlies many physical phenomena (Section 2.2) and (1.

Notice that the symbols for units are written only in the singular. such as p for pressure. Dimensional analysis is a method of checking and predicting relations between physical quantities based simply on the principle that the dimensions must balance in any equation.99 g but 1. thus velocity has dimensions LT.x --:--::-- . the value of any quantity must first be divided by its units.325 kPa) = In(p/kPa) -In(101. This need not conflict with colloquial usage.1 and force (mass x acceleration) has dimensions MLT . so that force has units of kg m S-2. The dimensions of a physical quantity are defined as the appropriate combination of powers of the fundamentals mass (M). for example.2 Units and dimensional analysis The units used in the SI system are given in the appendix together with numerical values of the fundamental physical constants and conversion factors. The viscosity '1 of a fluid is defined by force per unit area '1 = velocity gradient force area distance velocity = .4. so that if p = 2 atm we have p/atm = 2. The dimensions are precisely analogous to the SI units of kilogram. not the units.2. In (p/atm) = In(p/101. length (L) and time (T).32S) = In(p/kPa) ..618. and this product follows the normal rules of algebra. Algebraic symbols are used to denote physical quantities. or they cannot be cleanly cancelled and we would also have absurdities such as 0.2. we may speak of a temperature difference of 10 kelvins but this is written as 10 K. metre and second. such as In(p/atm). Thus 2 atm is the product of the pure number 2 and the unit atm.4 Algebraic and geometrical methods 1. An important particular case is when taking logarithms. Since force = mass x acceleration it has dimensions MLT . The value obtained will then depend on the units being used since. A standard example of the technique is to predict the form for Stokes's law for the drag on a sphere moving in a viscous fluid. the symbol denotes the quantity. or a quantity divided by its units is a pure number.01 gms. A particular value of the quantity is then the product of a pure number and its units. since the definition of a logarithm only applies to pure numbers.

-p-y = -2. This has two valuable and practical uses: unless the units cancel. Finally.days" = 6 weeks. Land T to obtain for M. arguments are needed to show that the proportionality constant k is 61t. and when we wish to change to other units. so that oc = p = y = 1 and the required equation is force = ka"v. the new units over the old using appropriate conversion factors.x 7. we equate powers of each of the primary dimensions M. The same principle can be applied when using the practical SI units of kilogram (kg). for T. and we can construct 'unity brackets' by writing. The relation that is assumed is written as and dimensional analysis will enable us to find the numbers oe. . To convert.Units and dimensional analysis with dimensions 5 1 . M~ We then assume that the drag (resisting force) on the sphere will depend on the size of the sphere (radius a). so that in any equation connecting physical variables the units must cancel on the two sides of the equation. we can do so by a simple algebraic method called quantity calculus. rather than dimensional. The assumed relation is rewritten in terms of the dimensions (or units) of the quantities. y = 1.. x j22 x~ x L = ML -IT-I. its velocity (v) and on the viscosity of the fluid.eays. p and y. for L. if we write (1 week/7 days) this has the value unity. Physical. The latter is based on the principle that multiplication by unity leaves any quantity unchanged. the equation being used is wrong. as a fraction. 42 days into weeks we multiply by the appropriate unity bracket and cancel the units: 1 week 42 days = 42. As a simple example.. say. metre (m) and second (s). oe + p-y = 1.2 = L"(LT-I)P(ML -IT-I)y. thus MLT. Dependence on temperature will be taken care of through change in the viscosity of the fluid.t T'l.

x ~# x N.08205 litre atm mol. = INm.624 X 10. 1 N = 1 kgms. J sm x m = )g'~ .1 K . The unity brackets are based on the conversion factors 11= INm.Pt! X J molK x ~ x . If we wish to use SI units instead. n in moles (mol) R = 0.3~ 101.6 Algebraic and geometrical methods Example 1. Thus if the pressure p is in atmospheres (atm).3 m 3 .325 X 10 3.. .2 Show that Planck's constant has the dimensions of momentum x length. using four significant figures.atnr 10.mti.m Example 1.34 J s.. we make use of unity brackets based on the conversion factors 1 atm IJ Idm 3 IPa = 101.082057 ~. volume V in dm 3 (also called litres). From the appendix. The units of momentum x length are those of mass x velocity x length. =INm.1 The gas constant R in the equation p V = nRT has the units needed for the equation R=pV nT to have the same units on both sides. and temperature T in kelvins (K).2 .1. h = 6. = 1O. so that. and amount of substance. so that the units of momentum x length are kg x )4's'i.325 kPa.2 .x ~}If x ~}Ii = Js . R= 0.

8'. Thus the water exerts a pressure of about 2 x 101 kPa (2 atm).35 .m x 10.03 gcm -3) is under a pressure of about 3 atm. The pressure exerted by a column ofliquid of density p and height h is given by J</ ~ x .en1 )rf x 104 ~ ~ Pa~ x~ = 2.81pf .626 x 10 . We have that either or h I where h == 6. but can remember only that the spacing is 2B where B is either h 2 /8n 2 I or h/8n 2 I c. Example 1.35' X 108 m s .626 x 10.1.998 1= cm = 8n 2 c x 10..3 Suppose that we are given that the spacing between the lines in the microwave spectrum of HeI is 20.Qftl$ x--x--x--8 x 2.Units and dimensional analysis 7 Example 1.70 X 10.34 J sand c = 2..4 Show that a diver at a depth of66 ft of sea water (density 1.x 66j(x--II 103$ ~W it x 2.998 x 10 .J N~ 102. This difficulty can be resolved by adopting the strongly recommended rule that we substitute into formulae not just numerical values but the units as well..033 x lOs Pa = 203.3 kPa.34. so that either }(m kgm .47 kgm 2 • The second expression gives the correct units for the moment of inertia. x9.m 6..54.x12jn p=pgh= 1 .1 and that we need to calculate the moment of inertia I..7 cm.ern 8n 2 = 2. .J'.. . and so the total pressure on the diver is about 3 atm.

. and this would not be the same function as. Such equations will be satisfied by particular values of x called the roots of the equation. Variables are usually denoted by letters from the end of the Greek or Roman alphabets.. g(x. The simplest solution will be when the factors can be seen by .= 1 must first be multiplied through by x to give x 2 -x-2 = 0 (1. 1 atm). is used to show only that a dependence exists.8 Algebraic and geometrical methods 1. An equation is solved by finding the roots by any convenient method... if x = a is a root of the equation f(x) = 0. ..a)g (x) = O.y. . whereas the second labels a particular function as f(x. .. y.7) z =f(x. say x.. p). x. say..4 Quadratic and higher-order equations An algebraic equation in a single variable. . and constants are usually denoted by letters from the beginning of the alphabets. which depends on the temperature Tand the pressure p.. 1. is said to be of degree n when the highest power of x in the equation is x" after negative or fractional powers have been removed. )..3 Functional notation When the value of a quantity z depends on the values of one or more quantities x. which is the value off(x. equation (1. ). and so on. which means that the equation can be written as (x .8) z or The first form. An equation that contains x 3 is third degree (or cubic). Thus the enthalpy H of a substance. y.a) is a factor.9) and so it is a second-degree (or quadratic) equation.. b). by writing either = z(x. y. and an equation of degree n has n roots. . . Thus the equation 2 x x-. we call z. . ). . then (x . y. )... and at the particular values of the variables T = 298 K and p = 1 atm we write H (298 K. but this is a useful guideline rather than a universal rule. y) when x = a and y = b. We denote the value of a function at particular values of the variables by f(a.. y. variables and show z to be a function of x. (1. In general.7). y. is shown as H (1'. (1.

2) = 0 or (x + 1) = 0.4ac = -4-a-O. as x 2 -4 = (x+ 2)(x -2).(b2 -4ac).-b ± 1.1) gives the other.9) gives (x -2)(x + 1) = O.2. The factors of a difference between two squares are particularly useful.-- We now take square roots of both sides. A quadratic equation that cannot easily be factorized is solved by using the formula for solution of a quadratic.1) is a factor. The roots of a cubic equation such as x 3 -2x 2 -x+2 = 0 (1. factor.11) Corresponding formulae can be written for the solution to a cubic equation. giving x = 2 and x = -1 as the roots of the equation.Quadratic and higher-order equations 9 inspection.9). this means that (x . 2a 2a (1. x = 2 and x = -1.10) may be found by trial and error. which is obtained by 'completing the square'. By assuming small whole-number values of x we find that x = 1 satisfies the equation. so that the roots of (1. This will be satisfied if either (x . recognizing that this introduces ambiguity of sign. quadratic. Thus equation (1. The terms containing the variable x can be obtained by writing so that the original equation can be written as (x b + 2a )2 b2 = 4a2 - C ~ b2 . Given the equation ax 2 +bx+c = 0 we first divide through by a to obtain b c x 2 +-x+a a = o.J . Long division of the expression by (x . so that X = . but these are too cumbersome to be useful. it is better to find a . which in this case is the left-hand side of (1.10) are x = 1.

Thus the dissociation constant K of a solution of a monobasic weak acid is related to its degree of dissociation tX at concentration c by so that ctX 2 + KtX . temperature and pressure) of which two are independent (e. the length of a given bar of metal as a function of temperature.2) can be used in the initial numerical approximation. We then have three variables altogether (volume. Real solutions of a quadratic equation will not exist when b2 < 4ac. the length is fixed by the physical properties of the metal. which would imply that volume and temperature are independent variables and pressure the dependent one. We can then use complex numbers to write solutions containing real and imaginary parts as in Section 1. This is an example of one dependent and one independent variable.5 Dependent and independent variables When we say that a variable is independent we mean that it can assume any value. If we were considering the volume of a sample of gas.10 Algebraic and geometrical methods numerical factor by trial and error and to use the formula (1.K tX= = 0. Newton's method (Section 2. Example 1.9. temperature and pressure) and one dependent (volume).10.-K ± -1 V/ (K 2 + 4Kc).25). We can fix the temperature at any value we please. 1. so that length is not independent once the temperature is fixed. .5 In physical applications we may expect physical conditions to restrict the choice of roots. A simple case is if we were concerned with. Once having fixed the temperature. Thus we cannot specify which particular variables are necessarily dependent or independent. say. c and tX are all positive.11) for the remaining quadratic factor. what we can specify clearly is . this depends not only on temperature but also on pressure. so that only the positive sign before the root has physical meaning (see also Example 2. 2c 2c But K. Notice that we could equally well say that the pressure depends on the volume and the temperature. and this is an independent variable.g.

The number of independent variables is then the total number minus the number of equations (3 -1 = 2 for our sample of gas). we may choose to use the C values of the chemical potential of each component. A chemical example of the limitations on the number of independent variables set by the existence of relating equations is the phase rule. Thus in the simplest case of a single phase of a pure substance. giving C(P -1) equations.T) as the independent variables and obtain correct results in either case. C + 2 variables.l) or (Vm. instead of composition as a variable. Vm and T. we might choose to use either (p. particularly with problems involving chemical equilibria.P independent variables. giving P(C -1)+2. p V m = RT for a perfect gas). again. resulting in a freedom of choice which can cause confusion. The equations relating the variables are equality of chemical potential for each component in each pair of phases. C + 2 . In that case we have a Gibbs-Duhem equation relating the chemical potentials in each phase. To determine how many variables in a given problem are independent we need to know how many relevant variables there are (three for our sample of gas) and how many equations exist connecting these variables (one. so that P phases require P(C -1) composition variables. and is the number of variables that must be specified to define the state of a system completely. The temperature and pressure are two more variables. Each phase of C components requires C -1 composition variables (the 1 is a relating equation that the sum of the mole fractions is unity). giving. It is a characteristic of chemical problems that we have more than this minimum number of possible variables.Dependent and independent variables 11 that two are independent and the third (whichever one of the three we choose) is dependent. if we have as many simultaneous equations as variables we can solve the equations to obtain numerical values of the variables. the two independent variables may be chosen from p. The number of independent variables is also called the number of degrees of freedom of the system. with temperature and pressure. giving P equations and so. The number of independent variables is then P(C-1)+2-C(P-1) = C+2-P. The mathematical significance of this is that each equation (simultaneous equation if more than one) can be used to eliminate one variable from the problem. This need to have as many simultaneous equations as variables in order to obtain a unique solution is usefully borne in mind. It often . meaning that the number of independent variables is then zero. Alternatively.

and because they facilitate interpolation (reading intermediate values between the known points) and extrapolation (the projection of curves or lines beyond the range of the points). on the other hand. so the best choice of independent variables is then (p. The following terms and conventions are used in graph plotting. T) are chosen as the variables.12 Algebraic and geometrical methods happens that the relevant formulae are simpler in terms of one choice of independent variables than another. such as to absolute zero of temperature.1) are then particularly useful. The scales should be chosen so that the given points fill the graph. Graphs are always plotted with the dependent variable as ordinate and the independent variable as abscissa. the labels on the axes of the graph are length/m and TfK because length divided by metre and temperature divided by kelvin are both pure numbers. say. Experimental measurements on chemical systems are often conducted at a constant pressure of one atmosphere. for example if the volume is subsequently to be held constant this should be chosen as one of the variables in setting up the necessary equations because the simplest form of the result will then be obtained. but this should be calculated. are often simplest in terms of a fixed volume for the system and the independent variables used are (Vm. A long extrapolation is sometimes required. the total energy U and Helmholtz free energy A give the simplest formulae in terms of (Vm' T). not drawn. apart from short extrapolations. which implies that any straight line will have an angle of slope near to 45°.7.6 Graphical methods Graphical methods of presenting and analysing data are used because they reveal experimental scatter of points. The scales on a graph are pure numbers so that only pure number quantities should be plotted. because curvature can be detected or the choice of a best straight line can be made. graphical methods should be regarded only as a prelude to the analytical treatment described in Chapter 7. In precise work. Thermodynamic properties are expressed in terms of the enthalpy H and Gibbs function G when (p. 1. Theoretical treatments.T). length in metres as . there is no mathematical distinction between two variables as . As pointed out in the previous section. similar triangles (Section 1.T).a function of absolute temperature. The significance of this is that if the graph were to show. The vertical or y-axis is the ordinate and the horizontal or x-axis is the abscissa.

8. (1.k and intercept In Xo on the ordinate (y-axis).kt + In Xo and a straight line should be obtained by plotting In x against t (the independent variable. The best graph to plot is a straight line because visual estimation is then most precise and because simple analytical treatment is then available. a straight-line graph is much better than a curve.12) can be manipulated into that form before the scales of the graph are chosen. all graphs of the same quantities will then have the same appearance. which is the intercept on the y-axis. When the proper choice is not obvious. so that physical criteria are required. y = c. When a logarithmic scale is used the numerical values of any physical quantity must be divided by . or when numerical values of constant parameters are required. The best line passes through the 'centre of gravity' of the points and the sum of the squares of the deviations of the points from the line should be a minimum. and when y = 0. or gradient.Graphical methods 13 to which is the dependent and which the independent variable. this being the scaled tangent of the angle between the line and the x-axis and the value of the derivative dyldx (Chapter 2). temperature and composition are common independent variables and so should be on the x-axis. Such a visual choice of line will be subjective. Thus the first-order decay of quantity x with time t is given by This is converted into linear form by taking natural logarithms of both sides In x = . x = . The best straight line through a set of points can be fairly precisely estimated by eye using a tight cotton thread so as not to obscure any points. the intercept on the x-axis. Time. Many equations that are not originally in the linear form of equation (1. The proper choice is often obvious in that the independent variable is deliberately fixed and the dependent variable is then the result of an experiment. When a graph is used to test the agreement between experimental data and a theoretical form of equation. so that when the quality of the data merits it the analytical method of least squares should be used as described in Section 7. of the graph is m. with slope . The equation of a straight line is conventionally written as y = mx+c.12) When x = 0. so abscissa or x-axis). The slope. so that large deviations are much more important than small ones.clm. precedent should be followed.

hence b = -elm as before.1 Some geometrical methods Similar triangles If two triangles have corresponding angles equal.2 Triangular graph paper In an equilateral triangle. the sum of the perpendicular distances from any point onto the sides is equal to the height of the triangle. 1.7 1. This is . y I I I I I ~'-----.14 Algebraic and geometrical methods their units as described in Section 1.1 1. changing these units will change the intercept of a graph.2. then they are said to be similar and the lengths of corresponding sides are all in the same ratio to each other. In Fig. This is a particularly useful geometrical theorem.7. _ _ _ _ _ _ _ ---II x Fig.but will not affect the slope. 1.1 a straight line passing through points A and B fits experimental data plotted on paper shown in broken outline.7..N : I I IL. 1. The intercepts e and b would be most easily found by fitting the two points to y = mx +e by m = BN/AN and then e = YA -mx A • Then by similar triangles e/( -b) = BN/AN = m.

BOC and COA. involve more than two variables. that ofB is RO and ofC is PO. If a scale is chosen so that the height of the triangle is either unity (in mole fraction) or 100 (in percentage composition). 1.Some geometrical methods 15 used to represent the composition of three component mixtures on two-dimensional paper. 1. 1.7. In Fig.2. These correspond to diagrams in three or more . the amount of each of three components can be plotted from zero on a side to 1 (or 1(0) at the opposite apex. The sum of the perpendiculars OP + OQ + OR can be seen to be equal to h by noting that the total area of the triangle is the sum of the areas of AOB. The line BX in the figure is the locus of points that would be obtained by adding pure B to a mixture of composition X of A and C. each area being (a/2) times the corresponding height. In Fig.3 Three-dimensional geometry Many problems in science.2 1. the point 0 is where the percentage of A is QO. and particularly in chemistry. A B C ~------------------~ a Fig. Triangular graph paper is available on which lines parallel to the sides are drawn so that such scales can be used. the scale is such that h is 100.2 point 0 is anywhere within the equilateral triangle ABC of side a and height h.

When lines or planes are perpendicular to each other they are called orthogonal. In Fig. 1. An alternative approach to direction in space is to use the angles IX. m and n.b. Then by Pythagoras's equation. y. b = r cos p and c = r cos y. yand z-axes. y) plane. The coordinates of Pare (a. the factors cos IX. y) plane then gives projections OA = a and OB = b on the x.and y-axes. these being the projections of the line OP on to the x-. or normal. The projection ON in the (x. the angles being taken in the planes containing the line and the axis as also shown in Fig. y.and z-axes. 1.3. An equation containing two variables represents a line or curve in . p and y that the line OP from the origin makes with the X-. They are often treated by taking two-dimensonal sections in planes perpendicular to each other.and z-axes are then a = r cos IX. The projections of the line oflength r onto the X-.3 Pythagoras's equation in three dimensions then follows because r2 = ON 2 +PN 2 = a2 +b 2 +c 2. 1.16 Algebraic and geometrical methods dimensions. This is conveniently shown by drawing perpendicular. z c B A-~~~~------~---y A~ ____________ ~ x Fig.3 the point P is distance r from origin O.c). [2 + m2 + n 2 = 1. lines onto the z-axis and into the (x. cos p and cos yare called the direction cosines of OP and are denoted by I.

parabola and hyperbola An equation in two variables. At least one of the second-degree terms x 2.3 is denoted by (l/a. k. Band C in Fig. Equation (1. xy or y2 must be present. but any other coefficients may be zero. 1. when it is denoted by the Miller indices (h. (1.14) the conditions for the various kinds of curve are (a) Circle (b) Ellipse (c) Parabola (d) Hyperbola a=b and ab > h 2. These are the smallest whole-number ratios of the reciprocals of the intercepts on the axes. The general equation of a plane is ~+~+~ = 1. ab = h 2. Writing the general form of second-degree equation as ax 2 + 2hxy+by 2 +2gx+2fy+c = 0.13) abc as may be most easily checked by noting that the partial derivatives are then constants (Chapter 3). ° 1. (1. 1. say x and y. parabola or hyperbola. and if it contains only first powers of the variables and no cross-terms it can be written in the form of equation (1. represents a circle. The particular type of curve is readily identified from the relative values of the numerical coefficients. I). Similarly an equation in three variables represents a surface in three-dimensional space. ellipse. Thus the plane through A. The direction of a plane in space is often required in crystallography.3. which contains terms of the second degree (x 2. The use of reciprocals here avoids infinite values because a plane parallel to an axis is defined to have a corresponding Miller index of zero.Some geometrical methods 17 two dimensions.4 Circle.12) and represents a straight line. h = 0. and so on. l/c). Band C in Fig. ellipse.7. as can be seen by putting y = z = to show that the intercept on the x-axis is x = a.13) is that of a plane through the points A. xy or y2) but not higher. and if it contains only the first powers of the variables and no cross-terms it represents a plane. ab < h 2. l/b. multiplied by a common factor if necessary so as to give the smallest whole numbers. .

1. When a.18 Algebraic and geometrical methods Example 1.x = ±r. Having identified the type of curve represented by any particular equation.16) (c) Parabola (d) Hyperbola y2 = 4a. 1.x».15) when x = O.so that the equation represents a circle of radius r centred at the origin (Curve A in Fig. Thus y2 = 4a. From equation (1.and y-axes as asymptotes. Substituting V for y gives a = kj2 and b = h( = 0).y = ± rand wheny = O. so that ab = h2 (= 0) and the curve is a parabola (curve F in Fig. these being identified by seeing that as x tends to zero y tends to infinity.4 (xy = constant) has the x.7 The potential energy V for simple harmonic motion (Section 6. 1. An ellipse has major and minor axes with two foci on the long axis. There are standard forms of equation for each kind of curve.8 Show that these standard forms satisfy the general conditions stated above.3) is given by V = tkx 2. so that h 2 > ab and the curve obtained is a hyperbola (curve E in Fig. as here.4). (a) Circle (1. 2J . 2 _ y2jp2 = 1.4). Curve E in Fig. (1. 1. 1.y would give curve F.18) Example 1. > p the x-axis is the longer one and so contains the two foci (curve B in Fig. its position and orientation relative to the coordinate axes may be found by looking for symmetry and for points of intersection. and as y tends to zero x tends to infinity. These are x 2 + y2 = r2.4 whereas x 2 = 4a. when y = 0 and y = ± Pwhen x = 0. Equation (1.15) x 2ja.16) gives x = ± a. When.and y-axes.17) passes through the origin (y = 0 at x = 0) and is symmetrical about the x-axis (y = ± (a. A hyperbola lies between two intersecting straight lines called asymptotes.4). 1.6 A sample of perfect gas at fixed temperature obeys the relation p V = a positive constant. Substituting p for y and V for x we have a = b = 0 and h is positive. x 2ja.x. Example 1. (b) Ellipse (1. The parabola represented by equation (1.4). so that the axes of the ellipse are the x.x gives curve C in Fig. 2 + y2jp2 = 1.17) (1.

Example 1. cuts the x-axis at ± a: and does not cut the y-axis. If the cross-product xy is present. these are perpendicular to each other if a: = Pand the hyperbola is then rectangular (see also Example 2. The general form. the hyperbola is said to be rectangular. 1.9 The equation x 2 + y2 -2x+4y = 4 represents a circle because h = 0 and a = b.18). axis rotation is needed to eliminate that term. The equation is reduced to standard form by moving the origin to eliminate . The effect of this may be calculated analytically but can be seen by inspection in the simple cases usually met in physical applications. Two operations may be necessary to convert any given equation into standard form: displacement of the origin and rotation of the axes.Some geometrical methods 19 ----+-~------r-~~~_+------+_~---x Fig.4).30). The presence of terms in x and in y means that the centre of the circle is not at the origin.4 the asymptotes are perpendicular to each other. 1. equation (1. It is symmetrical about the x-axis and has asymptotes given by yiP = x/a: and yiP = -x/a: (curve D in Fig.

The standard forms of equation are useful when experimental data are to be fitted to such curves.X + 1/4 we obtain (x-l/2f = 1/4. Differentiation (Section 2.1/2)2 = x 2 .1)2 = X 2 .6.2x + 1 (y+2)2 = and that in y from y2+4y+4. Example 1. A linear plot may be obtained either by plotting y against x(1 . . oc/4).10 Identify the curve y = ocx(l-x) and suggest two alternative linear forms. Departure from symmetry about x = 1/2 can be allowed for by using a series of terms in powers of (x . This equation has the form ocx 2 . and also y = 0 at both x = 0 and x = 1.20 Algebraic and geometrical methods these linear terms. which is obvious by inspection. a linear graph is preferred when choosing the best fit to scattered points.x). y = oc/4.2) and radius 3.y/oc = -(y-oc/4)/oc. and using (x .ocx + y = 0 so that ab = h 2 and the curve is a parabola. or from yagainst (x -1/2)2.1/2). This becomes standard form by interchanging the variables and moving the origin to (1/2. As remarked in Section 1. The equation is reduced to standard form by writing x 2 . We can therefore write the equation as (x-l)2+(y+2)2 =9 so that it is a circle with centre at (1.X + y/oc = 0.8) shows the curve to have a maximum at x = 1/2. such as the heat and volume changes on mixing liquids or on mixing gases. Such a relation is often a good approximation to the properties of mixtures. the term in x can be obtained by using (x . A similar approach .

the relations between the coordinates being x = rcosO. 1. . When the x.11 Show that.5 may be defined either by (x. The angle 0 is generated by rotating the radius vector r from the x-axis in an anticlockwise direction. The radius r is defined always to be positive so that x and y assume their proper signs in the various quadrants due to the signs of cos 0 and sin O.5 Plane polar coordinates Coordinates such as x and y. are called cartesian coordinates.7.Some geometrical methods 21 is used to allow for anharmonicity of molecular vibrations in spectroscopy. a projectile moves in a parabola under the action of gravity. which are called polar coordinates. they are called rectangular cartesian axes. (1.19) Problems that possess circular symmetry will be solved most conveniently by using polar coordinates. We therefore have ab = h2 (= 0) and the curve of y against x is a parabola. which are plotted as lengths on graphical scales. y) or by (r. The vertical height y at time t is given by y = ut sin (X _!gt 2 • Simultaneously. 1.and yaxes are perpendicular to each other. Point P in Fig.O). the horizontal velocity is u cos (X so that the distance travelled in the horizontal direction is x = ut cos (X. Example 1. positive rotation is therefore conventionally anticlockwise. in contrast to rand 0. Eliminating t between these two equations gives x t=---. the vertical component of the initial velocity is u sin (X. neglecting air resistance and the curvature of the earth. U cos (X gx 2 y=xtan(X-22 U cos 2' (X so that a = -g/(2u 2 cos 2 (X) and b = h = o. Given velocity of projection u at an angle (X to the horizontal. a length and an angle. y = r sin O.

1. We anticipate subsequent chapters on integration to mention that this is achieved by use of the gamma function r(x) defined in Section 6.20) Thus 6! _ 6 x 5 x 4 x 3 x 2 x 1 _ 2 3! 3x2x 1 .. Since a graph could be plotted of n! against n and a smooth curve drawn through the points for integral values of n. 1.'tX-le-tdt (x >0).1 O.4 as r(x)=S. It is conventional to define O! = 1 so that the general relation (n + 1)! = (n + l)(n!) is still true when n = O..8 Factorials and gamma functions The factorial of a positive integer n is here denoted by n! and is the product n! = n(n-1)(n-2)(n-3) .21) . we wish also to be able to find the factorial of non-integral numbers.22 Algebraic and geometrical methods y p x Fig. (1.5 1. (1.

The objects in a group of nl may be rearranged . A fundamental principle is that the probability of an event is proportional to the number of ways in which that event can occur. by scoring any of (4.. By repeating the argument we see that the required number is NL This is then extended to find the number of ways of placing N objects into boxes that contain nl.22) Tables of this function are available. n2. 2). and so in N . We begin by placing the objects in line in N! ways.. An important quantity is the number of ways of arranging N objects in line. .721) (1. thus 2.1.1 ways. in successive boxes. For this we may choose the first as anyone of the objects. an approximation that is often used is Stirling's theorem for In n! in the form In n! 1.274 from the tables. = When n is very large.Probability from which x! = r(x + 1) = xr(x).721) r(l.721) (2. . . 3)or (3. 1). n2. (1. Each of the dice may fall in anyone of six ways. (1. and for any particular score on one we have six possibilities for the other. . values being necessary only in the range 1 < x < 2 since when x > 2 we use equation (1. and then divide the line into groups of n 1 . Thus in throwing two dice simultaneously we may obtain a total of5 in four ways. 23 (1.4). This argument applies to independent events. being typically a number of molecules in a sample of substance. Since each choice of the first may be followed by anyone of the choices of the second we multiply N(N -1) to find the number of ways of choosing the first two in the line. where zero means that the event will not happen and unity means that the event is certain.721) = 4..72l! = r(3.23) Probability We express the probability of an event as a fraction between 0 and 1. The second is chosen from the remaining N .9 ~ n In n .n. for large n.22). the total number of possible results is therefore found by multiplying 6 by 6 to give 36. which are common in scientific applications.721) = 2. and so in N ways assuming them to be distinguishable.721 r(2. The probability of scoring 5 is therefore 4/36 = 1/9. (2.

12 The equilibrium constant for the reaction H2 +D2 = 2HD may be calculated from simple probability theory. or the other way round. giving for the equilibrium constant which is a good approximation to the actual value of K at a high enough temperature. We therefore form twice as many HD molecules as H2 or D 2 . We suppose that a catalyst surface. that is to say no real solution. Since these rearrangements are not to be distinguished.24 Algebraic and geometrical methods in n 1 ! ways and these have so far been included in the total of N!.10 Complex numbers When equations are solved by simple algebraic methods we meet cases where no solution exists. We may form H2 (or D 2) only by both atoms being the same. this is used in Example 2. We assume that molecules are formed by random pairing of atoms on desorption from the surface. If the objects are placed at random into the boxes. but HD can be formed in two ways. on which dissociation into atoms occurs. is exposed to a gas mixture containing equal numbers ofR and D atoms. which is the product of the real number ± 2 and the imaginary number i. the probability of obtaining particular numbers in the boxes is then proportional to this factor.. 1. we must remove them from the product by dividing by nl!. ). We can define i as the imaginary square root of -1 and write the solution as x = ± 2i..We then have equal probability of any particular atom being H or D. Example 1. This can be extended to include numbers having both real and imaginary parts. which is written as the complex numbers 2 ± 2i. . thus x 2 -4x+8 = 0. either one site on the catalyst may be occupied by H and the other by D.17. hence the required number is N!/(nl ! n2! . x=±j(-4) has no real roots. x = 2±tj( -16) = 2±2j(-1). Thus the equation x 2 +4 = 0.

have any physical meaning in the real world.25). moreover. and the addition of complex numbers A and B can be shown as if we add vectors 0 A and AC to give . equate the imaginary parts. however. This is used in wave mechanics. a is called the real part and ib the imaginary part. Thus (1. If.ib) we call z* the complex conjugate of z. Thus (1. This should be regarded as an ordered pair of real numbers which have no relation between them. We must consider real and imaginary parts separately. In the general form of complex number a + ib both a and b are real numbers. we are concerned with finding a boundary beyond which real solutions no longer exist. some of which may be imaginary. and zz* is the real number given by (1. say. 1. It is sometimes useful to give a geometrical interpretation to complex numbers. because we can then equate the real parts and. This is called the Argand diagram and is shown in Fig.1. where the wavefunction I/J may be complex but I/JI/J* is a real probability. separately. again only real numbers having physical meaning. When the sign is reversed in the complex number z = (a + ib) to give z* = (a . The complex number A can be regarded as the directed line. complex-number theory cannot provide the physical scientist with any new concepts.Complex numbers 25 The definition and use of complex numbers in this way gives an advantage in the sense that an equation of degree n then always has n roots.25) This also illustrates another important point. This is done by plotting a two-dimensional graph showing real parts as abscissa and imaginary parts as ordinate. OA. Complex numbers are valuable. or vector.(c +id) = (a -c)+i(b -d). Thus if a+ib = c+id then a = band c = d. When we have an equation relating complex numbers this really constitutes two simultaneous equations.24) (a +ib) . The complex conjugate A * is then the reflection of the point A in the real axis. complex numbers can enable us to approach that boundary from the imaginary side as well as from the real one. and this has applications in stability problems. as a general analytical tool. These imaginary roots do not. that we can obtain real numbers from combinations of imaginary ones. The imaginary number i follows the normal rules of algebra with the additional rule that j2 = .6 where point A represents the complex number a + ib. but it does permit neater solutions to some problems.

26 Algebraic and geometrical methods Imaginary axis C(o+c+i(b+d)) .29) .. (1.26) and this does not contain i and so is a real number.6 ~C. Thus multiplication of complex numbers gives (a+ib)(c+id) = (ac-bd)+i(bc+ad) (1. so that a + ib = r(cos 8 + i sin 8). (1.9) must not be pursued too far because other allowed operations with vectors do not apply to complex numbers. The length OA is called the modulus IA Iofthe complex number A. and denote the modulus by r. . in polar notation. When we choose to use polar coordinates we use the angle 8.28) or. given by (1. We then have an alternative notation for complex numbers since a = rcos8 and b = rsin8.Q Real axis I 'A*(o-ib) Fig. r1 (cos 81 + isin ( 1) x r2 (cos 8 2 + isin ( 2) = r1r2[cos(8 1 +( 2)+isin(8 1 +( 2)]. . called the argument of the complex number. 1.27) The analogy between complex numbers and vectors (Section 3.

By definition x2 x3 eX = 1 + x + 2! + 3! + .27) we obtain a + ib r = r(cos 0 + isin 0) = reiO. = r1r2ei(O.iy = eO = 1. . exponential.31) This gives us a third.). and replacing the exponent by the complex number x + iy gives e x + iy = eXeiy and .. (1. 0 = tan -1 (b/a). (1.32) This notation is valuable because exponential functions are particularly easy to manipulate. Hence and similarly eiy = cosy + isiny (1.iy = cosy-isiny. .Complex numbers 27 This means that on the Argand diagram we multiply by adding arguments (0 1 + O2) and multiplying the moduli (rl r2). There is a particularly useful relation between the exponential of a complex number and the trigonometric functions. trigonometric identities can then be readily established. Thus the multiplication expressions (1. = 1 + ly - y2 2! - iy 3 3! + . the real terms are those in the series for cos y and the real coefficients of the imaginary terms are those for sin y.28) and (1. elY . This series has alternating real and imaginary terms.31) with (1.30) and (1.29) become (a +ib)(c +id) = r l eiO .j(a2 + b 2 )... so that multiplication and division involve rotating the vectors and changing their length. Example 1. One of the principal advantages of complex numbers is that they enable trigonometric functions to be rewritten in terms of exponentials. By comparing (1..30) e. = . x r2eiO.+O. form of notation for complex numbers.13 cos 2 y+sin 2 y = (cosy+isiny)(cosy-isiny) = eiYe.

33) Example 1.14 Use equations (1.31) from (1. remembering to treat real and imaginary parts separately.30) gives the useful substitutions eix+e.ix ---:-2-:-i- (1.ix cos x = ----:2:--- and . .33) to show that cos 2 x .28 Algebraic and geometrical methods Addition or subtraction of (1.sin 2 x = cos 2x. The application to complex numbers of the calculus techniques described in the following chapters is straightforward. SIn x = eix -e.

and is the partial derivative sign. ax which which which which means a large. Calculus also provides a clear and precise notation by which we can resolve problems that involve changes in many variables into ones involving changes in only one variable at a time.1 Significance and notation Calculus refers repeatedly to behaviour in the limit of infinitesimal changes. An appreciation of these uses of calculus requires not just the usual skills at the operations of differentiation and integration but an insight into the reasons why particular techniques are effective. Calculus is valuable in science in general. finite or macroscopic change in x. discussed in the next section. The techniques of differentiation and integration give connections between the values of properties and their rates of change. denotes an infinitesimal change in x.CHAPTER 2 Differential calculus 2. for many reasons. which are invaluable in the interpretation of experimental measurements. In this way. problems can be separated into parts by examining separately the behaviour at a point from the behaviour over a wider range. a clear understanding of this concept. !5x. . is essential. Amongst these are that it enables us to study the effects of infinitesimal changes in conditions. effects which are often quite simple even though the effect oflarge finite changes may be quite complicated. and in chemistry in particular. is the differential of x. dx. There is need to distinguish between the notation: L\x.

30 Differential calculus We use the following functional notation: means that y is a function of x. the slope of the chord becomes that of the tangent in the limit as (jx --+ o. To find the slope of the tangent we need to take account of the shape of the curve on either side of the point. and is written as dy/dx. or dy/dt. v. v and ware independent. y'(x) both denote the total derivative of y with respect to x. y =f(x) g(x). y is uniquely determined by an equation that contains both x and y but no other variables. we obtain the slope of the tangent to the curve at point A.2 The calculus limit The idea of a limit is fundamental to calculus and will be expressed both geometrically and analytically. so that dy dx = lim (jy. the slope of the chord will change until. 2. In Fig. a way of interpreting that is to point out that when A and Bare coincident we cannot obtain the tangent to the curve because the slope ofa line through a single point can have any value. being then denoted by (jx.1) . it does not imply that u. 2. only constants. The limiting value of (jy/(jx as (jx --+ 0 is called the derivative of y with respect to x. another function of the same variables would be written as. When point B is moved down the curve to approach A. 6x-O (jx (2. v and w. We are quite deliberately considering what happens as (jx --+ 0. for example. usually denotes differentiation of y with respect to time. or that given x.1 a curve y = f(x) is drawn with two points A and B on the curve separated by differences Ax in x and Ay in y. y = f(u. any other terms being constants. The difference in x between points A and B is denoted by Ax and as B is moved to approach A this difference becomes infinitesimally small. The line AB is a chord having a slope defined as Ay / Ax. not at (jx = 0. w) dy dx and means that y is a function of three variables u. when B coincides with A. f'(x) is the total derivative of the functionf(x).

The calculus limit 31 y y = '(x) I I I 6x . Since y = f(x) we can write (y+by) =f(x+bx) and by = (y+ by) . In analytical terms the limit (2. 6:%-0 . and then allow h to tend to zero to give dy = lim f(x + h) .JI r------ I I I x Fig.2) must exist. a function is said to be continuous if (2.2) An alternative notation is to denote a difference in x by a new symbol such as h.. A less severe condition concerns continuity.__ .1 The alternative analytical approach is to define by as the change in y for change bx in x. 2. dy = limf(x+bx)-f(x). the function then being said to be differentiable. dx h-+O h The definition of the derivative given by (2. dx /)x-+O bx (2.2) is used in the next section to obtain the rules for differentiation.3) lim f(x ± bx) = f(x). There are conditions that must be satisfied for calculus methods to be applicable to particular functions.y = f(x so that + bx) -f(x).f(x).

32 Differential calculus Condition (2.2. The straight line y = f(x) = x has the property .3) means that we can approach any point on the curve from either side and obtain the same value.3 is drawn to illustrate other terms that are used to describe the behaviour offunctions. An example of a discontinuous function is y = l/x as x -+ 0.00. but a function representing such a shape would not be differentiable at the corners. 2. 2.2 Fig. as shown in Fig. A differentiable function is one for which the slope of the line is continuous. Condition (2. when we approach zero in x from the positive side. l/x -+ + 00 whereas from the negative side l/x -+ .3) is not then satisfied and we say that l/x shows an infinite discontinuity at x = o. for example a line drawn in the shape of a square is continuous. This example allows us to introduce another concept. that of a single-valued function. 2. y y=t x Fig. a circle or square could give two possible values of y for a given value of x and vice versa.

The second derivative is thus seen to measure curvature. the slope is increasing with increase in x and the curve lies above its tangent as in . This gives the second derivative of y with respect to x. so that such functions can be used within certain ranges. which is written as d 2 y/dx 2 . and we can go on to higher derivatives d 3 y/dx 3 . meaning that changing the sign of x changes the sign of y.••• A straight line will have the same slope at any point on the line so that the first derivative is a constant. differentiable and single-valued.3 thatf( -x) = -f(x). 2. When the second derivative d 2 y/dx 2 is positive.4) for whichf( -x) = f(x).7. is zero. this means that the second derivative. and the methods of calculus can then be used. A function such as y = l/x is 'well behaved' except as x -+ 0. y = x 2 isa parabola (Section 1. . assuming that also to be differentiable. and this is called an odd function of x.The calculus limit 33 Fig. Most of the simpler functions used in physical applications are continuous. which may have to be specified.changing the sign of x does not change the sign of y and this is called an even function ofx. Having performed the operation of differentiation once we can do so again on the result of the first differentiation. On the other hand. which measures the rate of change of slope with change in x.

= _(x 2) = 2x.1 Put y = x 2 . Successive differentiation is then denoted by (d/dx)2 times y = d 2y/dx 2. then (y + <5y) Hence dy dx = (x + <5X)2 = x 2 + 2x<5x + (<5X)2.-+0 d dx .. 2..Y <5x lim (nx"-l ". etc.).34 Differential calculus Fig.2 Put y = x"..~~o = (2.3 Differentiation of simple functions The definition of the derivative given in the previous section can be used to obtain the rules for differentiation. 2. Example 2.-+0 + terms in <5x. examples of which will now be given. then (y + <5y) Hence dy dx = (x + <5x)" = (x + <5xHx + <5x)(x + <5x) . such a curve is called concave upwards. The notation for higher derivatives can be explained by regarding dy/dx as the product of the differential operator d/dx and the variable y..-+0 = so that dy ". a curve with decreasing slope is called concave downwards. Conversely.• = x" + nx"-l<5x + terms in (<5X)2.~~o = (y + <5y) .1. dx Example 2. etc..Y <5x lim 2x<5x + (<5X)2 <5x lim (2x + <5x) = 2x ". r = .. r = .4) (y + <5y) . When we take the limit all terms except the first contain <5x and so tend .

x-+OV(V+l5v) I5x I5x so that du dv v--udy = ~(~) = dx dx dx dx V v2 (2. v + I5v v 6.7) . 6.x-+0 l. If y = uv where u and v are both functions of x y + l5y = (u + l5u)(v + I5v) where l5u and I5v are the increments in u and v produced by an increment I5x in x.x-+0 the first term of which is zero.Differentiation of simple functions 35 to zero.6) The quotient y = u/v gives dy dx = lim [(u + l5u .uv) dx b-+O I5x .x-+0 which may be written as dy = lim 1 (v(u+I5U)-U _u(v+I5V)-v) dx 6. (I5U I5v I5Ul5v) = lIm v-+u-+--.~)/ I5x].5) The formulae for differentiating a product or quotient may be derived from these first principles. The first two terms contain the definitions of du/dx and dv/dx and the third gives ( I·1m 6. giving dy dx _ d dx = _(x") = nx"-l. and that of a product is the product of the limits.:)( uU I·1m -I5v) I5x 6. (2. Hence dy d du dv dx = dx (uv) = v dx + udx· (2. Hence dy = lim (U + l5u)(v + I5v) .x-+0 I5x I5x I5x Now the limit of a sum is the sum of the limits.

36

Differential calculus

Example 2.3

Show that if

x

then

dy

1

- = 1 - -2

dx

x '

2.4

The use of ditTerentials; implicit ditTerentiation

**The derivative dy /dx can be interpreted not only as the product of the
**

differential operator d/dx arld the variable y, but also as the ratio of the

differentials dy and dx. For this interpretation to be justified we must

be able to apply the usual rules of algebra to differentials, and this we

can in fact do. Introducing the notationf'(x) for the result of applying

the rules of differentiation derived in Section 2.3 to the function f(x) we

can write

dy = f'(x)

dx

**and can then multiply both sides by dx to obtain
**

dy =f'(x)dx

**and we call this the total differential of y. Notice that this is not called
**

the differential with respect to any particular variable; the variable x

appears only when we go on to write down the right-hand side of the

equation, and indeed it can be changed if we so wish without affecting

the left-hand side.

The rules for differentiation may then be written in the form

if y

= x"

then

if y

= uv

then

if y

= u/v

dy

= nx"-l dx,

= udv + vdu,

then dy = (vdu - udv)/v 2 •

dy

(2.8)

(2.9)

(2.10)

**This notation is especially important when we cannot conveniently
**

write y as an explicit function of x. Thus

y2 -2xy

=4

**implies that y is a function of x, but an inconveniently complicated one.
**

The derivative may be obtained most easily by differentiating each term

The use of differentials; implict differentiation

37

to give

**2ydy -2xdy -2ydx = 0
**

(y -x)dy = ydx

dy

dx

y

y-x

**When this method of obtaining derivatives is used the result will
**

often contain both x and y, whereas when we can begin by obtaining y

as an expliCit function of x the result will be a function of x only. The

original equation can be used to show that the results are the same.

Example 2.4

Show that implicit differentiation of xy - x 2 = 1 gives the same result

as Example 2.3.

Differentiating the expression gives

**xdy+ ydx -2xdx = 0
**

xdy = (2x - y)dx

2x - Y

dy

dx

X

2X2 -xy

x2

x 2 + (x 2 -xy)

x2

**But x 2 -xy = -1 so that, as before,
**

dy = 1-~

dx

x2 •

Implicit differentiation may also be applied successively so as to

obtain higher-order derivatives, given that differentiation of dx gives

d 2x, of xdx as a product gives xd 2x + (dX)2 and, since dx/dx = 1,

d 2x/dx 2 is the derivative of 1 and so is zero. From Example 2.4 we have

dy/dx = 1 _1/X2 so that d 2y/dx 2 = 2/x 3 • Ifwe choose to use this as an

example of successive implicit differentiation we also have

**xdy+ ydx -2xdx = 0
**

and differentiating again gives

**xd 2y+dxdy+ dydx+ yd 2x -2xd 2x -2(dx)2 = 0,
**

and division by (dX)2 gives

X

d2y

dy

d 2x

dx2 + 2 dx + (y - 2x) dx2 - 2 =

o.

38

Differential calculus

But d 2x/dx 2 = 0 so that

d2y

2dy

2

dx 2 = - xdx +~.

But dy/dx = 1 -1/x 2 so that, as before,

d2y

2

dx 2 = x 3 '

**1.5 Logarithms and exponentials
**

The exponential function eX, or exp(x), is defined by the series

for all x.

(2.11)

**Differentiation of this series can be seen to give the same series as a
**

result, so that when we differentiate eX with respect to x we get eX. Thus

if y = eX

then

dy = eXdx,

dy

dx

= eX.

(2.12)

**The natural logarithm of y is defined by the relation
**

then

x = Iny,

(2.13)

**and since eX is positive (or zero) for aU values of x, In y is defined only for
**

positive values of y. The notation In for natural logarithm is used, with

loge as an alternative, to distinguish it from the Napierian or decadic

logarithm log (or 10glo) which is defined by

if y

= 10%

Since In 10 = 2.3026 (or 10

bases are related by

then

z = logy.

= e2.3026),

(2.14)

**the logarithms to the two
**

(2.15)

so that

In y = 2.3026z = 2.3026 log y.

(2.16)

**The derivative ofa logarithm follows from the definition (2.14) and
**

from the fact that the reciprocal of dy/dx is dx/dy. Thus

dy

if y = eX,

dx

dy

- = eX = y

dx

1

1

= dy/dx = y'

The chain rule and differentiation by substitution

39

**which may also be written as
**

if x

= lny,

dx

1

dy

y

= d(lny) = -dy, - = d(lny).

y

(2.17)

Example 2.5

The rate of change of vapour pressure p of a liquid with temperature T

is related to the heat of vaporization !l.H by

dp

!l.H

dT= RT2P.

This can be transformed by using both dp/p

d(I/T) = - (I/T2)dT into

d (In p)

d(l/T)

=

dpT2

= d(lnp) and

!l.H

-p dT = -R·

Example 2.6

Show that if the rate coefficient k of a chemical reaction depends on

temperature T according to

k=Ae- E / RT

where A and E are constants, then

dInk

E

d(I/T) = - R

Example 2.7

Show that

**2.6 The chain rule and differentiation by substitution
**

The chain rule is used when introducing a new dependent variable into

a function that contains only one independent variable. Thus if y is

known as a function of x, a two-dimensional graph can be drawn and

dy/dx is defined. We may choose to introduce a new variable, say u,

which is also a function of x. We then have three variables x, yand u but

only one is independent and it is still a two-dimensional problem. If the

new variable were not a function of only the existing ones, we would be

moving into three dimensions, having then two independent variables,

Substitute then u = ax. This can be expressed as a general formula. This is best illustrated by examples. where a is a constant. Example 2. The introduction of a new dependent variable is called substitution.8 Let y = (2X2 _1)3.18) This follows from the definition of a derivative as a limit and from the fact that the product of two limits is the limit of the . The right-hand side could be evaluated algebraically and then differentiated term by term.9 Let y = eax . we can define three derivatives such as dy/dx. These are related by the chain rule dy dydu dudx' dx (2. and its purpose in differentiation is to convert an otherwise complicated function into one to which the simple rules can be applied.40 Differential calculus and would have to resort to partial differentiation as discussed in the next chapter. du/dx =a (2. then du/dx = 4x and dy/du = 3u 2 = 3(2x-l)2. A neater solution is to make the substitution u = 2X2 -1.19) . y and u of which only one is independent. that if y = f(x)" then dy = nf(x)"-lf'(x)dx.product. Given the three variables x. Hence dy dydu dudx -=--= dx 2 12x(2x-l). Example 2. dy/du and du/dx (du/dy is simply the reciprocal of dy/du).

The second derivative d 2 y/dx 2 is the rate of change of slope.= . when dy/dx = 0 we have a horizontal tangent and y is said to be stationary. if d 2 y/dx 2 is positive we have a minimum. then e/(x) (2.. a minimum as at B or a horizontal point of inflection as at C or D. and point A is a maximum. Turning points: maxima. Substitute u = (e hv /kT -1) then du dT _~ehv/kT kT2 and 1 u Z=-. where it is positive on the left and negative on the right. When both dy/dx and d 2 y/dx 2 are zero we continue with successive .20) Example 2. 2. dy/du = 41 e".10 Find the derivative with respect to temperature T of the expression Z = 1/(e hv /kT -1). minima and points of inflection and y = e". Conversely.= ae" = ae DX • In general if y = dy = e/(X)f'(x)dx. dz du Hence dz dT 2. If this is negative the slope is decreasing as at A. Hence dy dx dydu dudx . These may be distinguished by successive differentiation.Turning points: maxima. as at B. We then have a maximum as at A.7 dzdu dudT 1 . minima and points of inflection Fig. Since the derivative dy/dx measures the slope of a curve.u2 1 (e hv /kT _1)2 e hv /kT hv kT2 (ehv/kT _1)2 .4 is drawn to illustrate the various points A to E on a curve which can be detected by differential calculus and one point F which cannot.

Since the definition ofthe derivative dy/dx requires that a limit is taken of the slope of a chord as the endpoints move closer together. In general. 2. as can be seen by noticing that from D onwards the slope becomes increasingly negative towards E and then increases (becomes less negative) beyond E. Point F is where the curve suddenly terminates. This corresponds to horizontal point of inflection C. The conditions so far described are familiar ones.4 differentiation. we continue with successive differentiation until a nonzero derivative is found. d 2 (pH)/dm 2 zero and d 3 (pH)/dm3 negative. the conditions for a maximum (negative derivative) or minimum (positive derivative) apply. The endpoint of an acid-base titration is at the corresponding point of inflection on a rising curve of pH against quantity. this is called an endpoint maximum. d 2 y/dx 2 zero and d 3 y/dx 3 positive. m say. At E the slope is not zero but is nevertheless a minimum. These conditions are therefore d(pH)/dm positive. When n is odd. of added base. Point of inflection D is the converse. Point E is therefore characterized by dy/dx negative. The conditions d 2 y/dx 2 = 0 and d 3 y/dx 3 positive are those for a minimum in the slope dy/dx. the corresponding conditions for a horizontal point of inflection are used. When this is the nth derivative and n is even. such a limit is possible when approaching F from the left but . so that dy/dx = d 2 y/dx 2 = 0 and d 3 y/dx 3 is negative. where the slope decreases to zero and then increases again as we pass through the point. where the slope has a maximum value of zero being negative on either side. but points E and F also occur in physical chemistry.42 Differential calculus y x Fig.

= 3x 2 dx. Since n is even and the derivative is positive the point x = 0. Hence the fourth derivative is the first one to be non-zero at the turning point x = O. at x = 0. at x = 0.12 Show that the curve In y = x 3 + 1 cuts the y-axis at a horizontal point of inflection. minima and points of inflection 43 not from the right. This occurs when the vapour pressure ofa liquid is plotted against temperature. y dy dx = 3x 2 y.11 Given y = x4+4. Example 2. the curve vanishes at the critical point. Such a one-sided derivative cannot be found analytically.Turning points: maxima. Example 2. Differentiation of the equation gives dy . dy dx = 4x3 = 0 at x d 2y = 12x2 = 0 dx 2 d3y -=24x=0 dx 3 = 0. The second derivative may be obtained either by differentiating the equation a second time to give . d4y dx 4 = 24. y = 4 is a minimum.

= -2kda-x)+k 2(a-2x) =0 which is negative if k2 > k 1 • So that kl a-2x k2 = 2(a-x) for maximum velocity. and the relation between kl and k2 such that this occurs at x = a14. as before. x is variable. Put x = a14.13 The velocity v of a chemical reaction is given by v = kl(a _X)2 + k2x(a -x). d 2vldx 2 negative.4. This is the equation for a reaction which is second order in the reactant together with catalysis by-product.3 for the maximum to be at x = i. when x is zero both dyldx and d 2Yldx 2 are zero. then kl _ al2 _ 1 k2 .44 Differential calculus or by differentiation of the expression for dyldx to give. 3) The point at which the curve cuts the y-axis is when x = 0. .3 and so positive. and similarly :~ = :x (:~ ) :~ + y(6 +36x = (6x + 9x 4 ) = 54x 3 y+ 27x 6 y+ 6y. so that In y = 1. The required condition for a maximum is dvldx = 0. Find the condition for a maximum in velocity. d2 y d (dY) dy dx2 = dx dx = 6xy + 3x 2 dx = 6xy + 9x 4 y.718. 2. and so we have a horizontal point of inflection of type C in Fig. y = e = 2. From the above.3al2 . but d 3 Yldx 3 = 6y = 16. where dv dx . where kl' k2 and a are positive constants. Example 2.

and use the condition dA/dr = o.Maxima and minima subject to constraint 45 Example 2. We could insert Vas a constant and eliminate h by writing to give A = 2V/r+27tr2 as an equation in the variables A and r. which. The unknowns are the two coordinates of the required point so that the problem has a unique solution. . Thus the most economical shape to make a cylindrical tin is with the height equal to the diameter. that V is constant and A is a minimum.8 Maxima and minima subject to constraint. A simple example is to find the point on a known curve that is nearest to the origin. gives two simultaneous equations. We have two simultaneous equations to satisfy. h = 2r as the required condition. A neater solution is obtained by differentiating the expressions both for V and for A and putting d V = 0 for constant volume and dA = 0 for minimum surface area. with the equation of the curve as a constraint. This may be solved in various ways. Thus dV = 27trh dr + 7tr2 dh if dr = =0 and dA dh r 2h = 27th dr + 27tr dh + 47tr dr =0 if :~ = .14 Find the ratio of height to diameter for a cylinder so that the surface area is a minimum for a given volume. the distance r will be a minimum when dr = 0. 2. Put V = 7tr 2 h and A = 27trh + 27tr2. Lagrange's method of undetermined multipliers Lagrange's method is a general one for solving problems in maxima and minima in the presence of constraints that can be expressed as simultaneous equations.h : 2r and these simultaneous equations give 2h = h+2r.

We express that also in differential form by writing the equation of the curve with zero on the right-hand side and differentiating. a solution then being obtained in terms of multipliers whose values remain undetermined as in Example 2.7. The procedure in Lagrange's method is to write each of the simultaneous equations in differential form. The second equation is the constraint that the point (x. Example 2. however. when numerical values can be obtained for the multipliers as in the following Examples 2. y) as r2 = x2 + y2 and minimizing by differentiating and putting dr so that = 0. This gives xdx+ydy=O and so that 2lXxdx+lXdy = 0 x(l + 21X)dx + (y + lX)dy = O.22) dy+2xdx = O.23) . and to eliminate variables in the usual way by multiplying equations by factors and adding them together. when we have more variables than equations. the factors being the Lagrange multipliers.16. This is called the calculus of variations and is discussed in Section 5. The method is particularly useful. (2.22) by a constant factor IX (the Lagrange multiplier) and adding to equation (2.15 Find the minimum distance from the origin to the curve y = 2 . We solve these two simultaneous equations in the usual way by multiplying equation (2.15 and 2. but the function itself such that a stationary condition applies. Thus y_2+X2 = 0. y) must lie on the curve. (2. to give 2rdr = 2xdx+2ydy = 0 (2.17.x 2 • The first equation is obtained by expressing the distance r from the origin to any point (x.46 Differential calculus A related problem is to find not the maximum or minimum value of a given function.21) xdx+ydy=O. This provides an alternative method of solving problems in which we have as many simultaneous equations as variables.21).

16 Find the shape of a right circular cone having minimum surface area for a given volume. an algebraic solution of the simultaneous equations would give r as a function of x. y = t.Maxima and minima subject to constraint 47 We now choose ex so as to eliminate one of the differential variables. (2. we have a 2 = r2 + h 2 and the volume V and surface area A are given by V = nr 2hl3 A = nr2 + nra. d V = 0 giving t nrh dr + !nr2 dh = 0. The equation of the curve then (y+ex)dy gives x2 = 2.24) For a given volume. Example 2. dA = 0 giving 2nr dr + nr da + na dr = 0.25) . (1 +2ex) = 0. r = fil2 as before. x 2 = 3/2. 2hdr+rdh = O. vertical height h and slant height a. For minimum A.23) must then also be zero. since dy may take any value. so that = (y-t)dy = 0 and. (2. If the cone has base radius r. (2r+a)dr+rda = O. The other term in (2. so that r2 = x 2 + y2 = x 2 + (2 _ X 2)2 r2 = X4 - 3x 2 + 4. Differentiation then gives 2rdr = (4x 3 -6x)dx so that when dr/dx = 0.y =! and the required distance is obtained from Since in this case we have as many equations as unknowns. such as by putting ex = -1/2. 4x 3 = 6x.

= 0. we suppose a particular number nj of particles to be in each quantum state i. and also uses Stirling's theorem. so that 3r 2 + 2ra . An algebraic solution of the equations is possible but Lagrange's method gives the neater solution. n2. each state having a characteristic energy. In simplest form.23). it is this number of ways. then we .9 with maximization subject to constraints. that we seek to maximize subject to constraints. so that 2r+a-2h2Ia+r2Ia = O. It combines the probability theory of Section 1. This is an example of the use of Lagrange's method when the number of variables (the numbers n 1 .48 Differential calculus and from r2 + h 2 .a 2 = 0 rdr+hdh-ada = O. a = 3r. p = ria ex = . and this is called a distribution. Probability enters into the argument by giving the number of ways of obtaining a particular distribution (particular numbers of particles in each state). each of which may be in anyone of a large number of possible quantum states.a) (r + a) = 0. we put r-pa and exr+ph = 0.24) + ex (2.pa)da + (exr + fJh)dh = O. coupled with having an equation in which the sum of a series of independent terms is zero..25) + P(2. This is concerned with the most probable way in which a given amount of energy will be distributed amongst molecules. equation (1.a 2 = (3r .26) to give (2r + a + 2exh + pr)dr + (r . At any instant.26) We now take (2. if say x + y = 0 and if x and yare independent. Example 2.r2. (2.phlr = -hla. . and since only positive values have physical significance. But h2 = a 2 . Choosing ex and p to eliminate the terms in da and dh. ) is very large. The physical model is to suppose that we have a large number of identical particles.17 The Boltzmann distribution To determine the most probable distribution of molecules amongst quantum states: the Boltzmann distribution. or probability.. The argument turns on how many of these variables are independent.

. We apply Lagrange's method by adding (2. = constant (2. nj. The numbers nl. · The most probable arrangement is the one that can occur in the maximum number of ways.29) and p times the differential form of (2. but independence forbids that because the value of y is not determined by the value of x. Since we have three simultaneous equations.30).·.30) The variables in this problem are the numbers nl. when so that d In P = dP / P = 0 d InP = d InN! -d Innl! -d Inn2! ..27) nl·n2.. . and so are treated as continuous variables rather than integers.28) by use of Stirling's .Maxima and minima subject to constraint 49 must have x = 0 and y = 0. given by P = '" N! (2. more usefully...31) (2. E = nl81 + n282 + ... .· ni ......29) and since the ni particles in the state with energy 8i contribute ni8i to the total energy. = o.8. (2...... we modify (2..28) to at times the differential form of (2. These constraints give N = nl + n2 + . •••• We may regard the particles as objects and the quantum states as boxes.. from Section 1.. = o. Suppose that we have a fixed total number N of particles.28) If it were simply a matter of placing objects at random into boxes the most probable arrangement would be an equal number in each box. otherwise any non-zero value of x would require an equal and opposite value of y.32) Before doing the addition. . this is when dP = 0 or. n. n2' . that the total number of particles is fixed and that the total energy is also fixed. + 8i dni + .. The problem is complicated by constraints. n2.9 the number of ways of arranging N objects in boxes containing nl' n2... = constant. . . in each box is the probability P of that arrangement. all but two of these variables will be independent... (2... and each one has the same accessible set of quantum states. . .. which have energies 81> 82> ••• . . Hence dN = dnl +dn2 + and dE . that we seek are average values. (2. + ni + . +dni+ .. .. + ni8i + .. = 0 = 81 dnl + 82 dn2 + .

2.9 Series The use of series provides a valuable extension to algebraic methods.. The term e -cz may be written as a constant A. dN = 0. We may then use the series either to calculate its sum or as a source of successive numerical approximations.32) to obtain (In nl + (X + peddn 1 + (In n2 + (X + pe2)dn2 + + (In ni + (X + pe. and also problems that cannot be solved by simple algebra often yield to a solution in terms of series. Such series are useful in practice only when successive terms become smaller and smaller so that the series converges. . each term must be zero.. Also. e or sin 0. (2.) dn i + . The solution contains the undetermined multipliers (X and p. which is discussed in the next section. so that the term in N disappears from (2. equation (1. This leaves (2.23) In N! = N In N .33) We now add (2. . so that a formula exists for calculating any particular term.N giving for large N.31) + P(2. We can use a convergent series of terms to represent quantities that cannot be expressed in a simpler closed form. The subject of series arises at this stage in our discussion because the series with most physical significance can be based upon Taylor's theorem. such as 1t. giving the Boltzmann distribution law in the form ni = Ae-e. A mathematical series contains terms that are related to each other.33) + (X(2. . = O.34) We now choose the multipliers (X and p so as to eliminate n 1 and n2' Equation (2. each term containing only one of the independent variables ni so that in order for the sum to be zero. .34) is then a sum of terms. since N is constant.28).50 Differential calculus approximation.!kT. and physical reasoning identifies p with lila'. d InN! = In NdN. Hence (In ni + (X + pei)dni = 0.

. which may be derived as follows: ... The sum s of an infinite number of terms may be obtained by the device of multiplying each term in the series by x and subtracting the result from the original series.2 Power series and Taylor's theorem A power series is a general case of the geometric series discussed in the previous section.hv / kT +e. s = 1/(1-x). and is often a useful and significant way of expressing the values of physical properties.35) Successive terms will become smaller and smaller. Example 2. and the convergence of the series depends on the ratio of successive coefficients being less than the reciprocal of the variable. (2.hv / kT.. it is then called a polynomial of degree n.3hv / kT + ...2hv/ kT +e.Series 2.18 The theory of the heat capacity of a crystal gives rise to the series 1 +e. This can be seen to be a geometric series by making the substitution x = e.36) Subtracting (2. 2.37) A convergent power series is one of the most useful ways of expressing many mathematical functions. In general terms a power series has the form (2.. If the series is stopped at x". if -1 < x < 1. This is Taylor's theorem.1 51 Geomecric series This is a series of increasing powers of a variable in the form s = 1 + x + x 2 + x 3 + . so that the sum of the series is s = 1/(1-e. (2. thus sx = x + x 2 + x 3 + X4 + .9. In a mathematical power series the coefficients themselves form a series.36) from (2..35) we obtain s(1-x) = 1.9.hv/ kT ). so that the series converges. each of the successively higher powers of the variable being multiplied by a coefficient.. The values of the coefficients in a power-series expansion of a function can be expressed in terms of successively higher derivatives of the function.

... This form of Taylor's theorem applies when we have only one independent variable x.... . we denote the successive derivatives with respect to x by f'(x)... . ) and when h = 0.f"(x). .. can be expressed as a power series in h in the form f(a+h) =f(a)+alh+a2h2+a3h3+. (2. a2' .8. Thus f'(a + h) = :h (f(a) + al h + a2h2 + a3 h3 + . = al + 2a2 h + 3a3 h 2 + . .f(a). .38) When x is expressed as a + h the derivatives become h) f '( ) =f'( x a+ = df(a + h) = df(a + h) dh dx dh dx' but x = a+h so that dh/dx = 1 and so f '( ) x = df(x) dh' The technique is then to differentiate (2. which isf(a + h) . With more than one independent variable the theorem is expressed in terms of partial derivatives...38) successively with respect to h.. . .. Next f"(a+h)=2a2+2x3a3h+ .52 Differential calculus Given a continuous functionf(x) whose valuef(a) is known at x = a and which is successively differentiable.39) When this series converges it provides successive approximations to the functionf(x) in terms of the values of the function and its derivatives at the point x = a. .and the values of these derivatives at x = a by f'(a). (2.f"(a). and then to put h = 0 to give expressions for the coefficients al . so that at h = 0 f"(a) = 2a2' Hence Taylor's theorem can be written in the form f(a + h) = f(a) + hf'(a) + ~~ f"(a) + ~~ f"'(a) + ... .. We then assume that when x is increased from a to a + h the increase inf(x).. and this is discussed in Section 3.

and so will be a poor approximation if curvature is present so that the slopef'(x) changes with x. 2. suppose (xo + h) is the true root so that f(xo +h) = O.39) then provides successive approximations to an extrapolation of the curve from a known point A in terms of the known slope. (2. The first approximation is f(a + h) = f(a) + hf'(a).Series 53 A geometrical interpretation of Taylor's theorem is shown in Fig. Higher derivatives allow for higher orders of curvature and give better approximations to the true shape of the curve.2. This neglects any curvature. We can find an approximate value of h by using the first approximation from Taylor's theorem that f(xo + h) = f(xo) + hf'(xo) .40) y y= f(x) f(o+h) I I I hf'(o) f(o) _________________ JI A h I I I I o o+h x Fig. This change in slope is given by the second derivative and appears as the next term in the series. For increment h in x we obtain point B on the curve. Point A is on the curve y = f(x) at x = a.5. Given an approximate solution x = xo.S which is point C on the tangent drawn from A. Equation (2. The series (2. all at point A. curvature and successively higher derivatives. at x = a + h.40) is the basis of Newton's method of successive numerical approximation to a solution of the equationf(x) = O.

5645.2.0.5) = 0.54 Differential calculus so that h = _f(xo) !'(x o ) and our improved estimate of the root is f(xo) Xl = xo+h = x o . This gives Maclaurin's theorem. which is f(x) x2 = f(O) + x!.5666) = . Hence x = 1. correct to four decimal places.41) Xl. so that x = h..5) = .1555.6 X 10.6 .5666 .5666) = . The following examples illustrate the expansion of particular functions.2.2.3 Maclaurin's theorem Maclaurin's theorem may be regarded as a special case of Taylor's theorem.3333 Xl = 1. 2.. 1'(1.2x.( .2.4949. X2 = 1. .(054)/( .2. so that rapid convergence has been obtained.5 .(O) + 2! f"(0) + .5666 .5645. f(1.4949) = 1..0. and since the point at which the value of a function is often simplest is when x is zero.. f(1.0022 = 1.1555/( .f'(xo)" This can be repeated to improve upon the value (2.5.9. 1'(1.3333) = 1.5 is an approximate root of the equation In X = x 2 .5645) = .0054.5666 f(1.42) This may be used to develop series expansions of functions that can be successively differentiated.0. Put f(x) = lnx-x 2 +2 f'(x) = l/x .0.19 Given that x = 1. it is particularly useful to write the series for a = 0. (2. use Newton's method to find a solution correct to four decimal places. Example 2. Since Taylor's theorem gives the value of a function in terms of its value at a point and the derivatives at that point.

.3! + 5! .20 Let f(x) = eX.. the This is an example of the use of the theorem rather than a derivation of the series because eX is defined as the series.X)/2 x 2 X4 = 1 + 2! + 4! + .44) Similarly we can obtain x 2 X4 cos X = 1 .2! + 4! .Series Example 2. 1'(0) = eO f"(0) = 1. so that the expansion becomes (2. f" (0) = 0.21 Let f(x) = sin x f (0) = sin 0 = 0.43) This gives the useful approximation that.X)/2 = x3 x5 x+ 3! + 5! + . The hyperbolic functions sinh x and cosh x are defined by sinh x = (eX . f'(x) = eX.. !"'(O) = -1.. coshx = (eX +e. f"(x) = eX. = x for small x. f'''(x) = -cos x..sin x. (2..45) . so that x3 x5 sinx = x .. as a result of which the derivative is eX. eX series being rapidly convergent for small values of x. then f" (x) = . f'(x) = cos x.. ~ 1 + x. (2. f(O) then = eO = 1. 55 = 1... 1'(0) = cos 0 = 1. Example 2.e .. for small x.

56

Differential calculus

**and are useful because the terms do nl)t alternate in sign as do those of
**

sinx and cosx.

These examples show that the expansions of the functions are

consistent with the expressions for the derivatives. Maclaurin's

theorem gives a new series, however, in the following example.

Example 2.22

Let

= In(1 + x),

f'(x) = 1/(1 + x),

f"(x) = -1/(1 + X)2,

f(x)

f(O)

= In 1 = 0,

= 1,

1"(0) = -1,

1'(0)

so that

**This has been written as the variable y because the definition of the
**

logarithm gives

y = In(1 +x)

if

(1 +x) = e Y

so that we may also write

y2

e Y = (1 + x) = 1 + y + 2!

so that

y2

X

y3

+ 3! + ...

y3

= Y + 2! + 3! + ....

(2.47)

**Now equations (2.46) and (2.47) describe the same relation between x
**

and y; they differ in that the first gives y as an explicit function of x, and

the second gives x as an explicit function of y. This is an example of the

inversion of a series, which is discussed further in the next section.

Example 2.23 The binomial series

This is a series expansion of (1 + x)":

f(x) = (1 + xr,

f'(x) = n(1 + xr- 1 ,

f"(x) = n(n-l)(1 +xr- 2 ,

f(O) = 1,

1'(0) = n,

1"(0) = n(n - 1).

Hence

"_

n(n -1) 2 n(n - l)(n - 2) 3

(1 + x) - 1 + nx + 2!

x +

3!

x +

....

(2.48)

Series

57

**The successive coefficients of the powers of x are called binomial
**

coefficients and have characteristic values, for example

(1 +X)4

=

1 +4x+6x 2 +4x 3 +X4.

**The binomial series converges only when - 1 < x < 1. When x is
**

sufficiently small for x 2 not to be significant, we can neglect the thirdand higher-order terms so that

(1

+ x)" =

for small x.

1 +nx

(2.49)

Example 2.24

**The binomial expansion gives a useful approximation when an
**

expression contains an algebraic denominator, such as

1

(a + X)3 - a 3(1

1

+ x/a)3

)-3

= :3 [1 -3: + 6

(~r -10 (~y + ...J

1 (

= a3

1 +~

**which will be convergent so long as x < a, and terms beyond the second
**

can be neglected if x/a is small enough compared with unity.

Example 2.25

**Given the relation between the equilibrium constant K and the degree
**

of dissociation 0: for a solution of a weak acid at concentration c

0:2C

K=-1-'

-0:

**if we assume (1 - 0:) ~ 1 we obtain 0: = j(K/c). Use the binomial
**

expansion, assuming K ~ c, to show that successive approximations

are

0:

=

j(K/c) - (K/c)j2 + (K/C)3/2/8 - ....

**Equation (2.49) is a quadratic equation for
**

co: 2 +Ko: -K

0:

in terms of K and c;

= 0,

**so that the exact solution is given by formula (1.11) for solving a
**

quadratic equation as

0:

K

= - 2c

1 j

± 2c

2

(K +4Kc).

58

Differential calculus

**It is not possible by simple algebraic manipulation to separate out from
**

this the term J(K/c). We therefore resort to a series expansion by

writing

~ = -~±J(~)(1+ ~yI2,

so that the expression in the right-hand bracket is nearly unity, which

means that the binomial expansion will converge rapidly. Hence

~= -~ + J(~)(1+~ -~ 1~2+ .. .)

as required.

2.9.4 Inversion of a series

Given a convergent power series

y = x+a2x2+a3x3+ ...

(2.50)

**we can calculate y for a given value of x. Suppose, however, that we wish
**

to find x for a given value of y. We could do so in particular cases by

numerical trial and error, or we can invert (or revert) the series so as to

obtain an expression for x as a function of y.

Assume that a power series exists for x in the form

x = y+b 2y 2 +b 3y 3 +

...

and substitute into equation (2.50)

y = (y+b 2y 2+b 3y 3+ .. .)+a2(y+b 2y 2+b3y3+ ...)2

(2.51)

**For this to be true, the coefficients of the second and higher powers of y,
**

which appear on expanding the right-hand side, must all be zero. Hence

b2 +a2 = 0,

b3 + 2a2b2 + a3 = 0,

b2 = -a2'

b3 = 2a~-a3'

**and so on. The series produced is
**

x

= y - a2y2 + (2a~ - a3)y3 + (-

5a~ + 5a2a3 - a4) y4

+ (14a~ - 21a~a3 + 6a2a4 + 3a~ - as) yS + . . . .

(2.52)

Series

59

**The rate of convergence of the series (2.52) will not be the same as
**

that of the original series (2.50) so that different numbers of terms will

be necessary to obtain the same precision.

Example 2.26

**Given the series (2.46), obtain the series (2.47).
**

By comparing (2.50) with (2.46) we obtain a2 = -1/2, a 3 = 1/3,

a4 = -1/4, ... , so that (2.52) becomes

x

= y+ y2/2+ y3/6+ y4/24 + y5/120+ ....

which is (2.47).

An alternative method of inverting a series is by successive approximations. We write equation (2.50) in the form

x = Y -a2x2 -a 3x 3

and substitute successive approximations for x into the right-hand side.

The first approximation is x = y, which we substitute, retaining powers

of y up to y2, so that

This is then the second approximation. We next substitute the second

approximation and retain powers of y up to y3

X =

Y -a2 (y - a2y 2)2 - a3y 3

= Y - a2y2 + (2a~ -a3)y3,

and this is the third approximation. We continue by substituting the

third approximation for x 2 , the second approximation for x 3 and the

first for x\ retaining powers up to y4, and so on. This method can be

used for a convergent series of any kind.

Example 2.27

The equation of state of a real gas is a relation between pressure p,

molar volume v and temperature T. The virial equation of state is a

power series in the density l/v

pv = RT[1

+ B(l/v) + C(1/v)2 + D(I/v)3 + .. .].

**An alternative is to use a power series in pressure in the form
**

pv = RT+B'p+Cp2+D'p3+ ... ,

**and series inversion is used to obtain relations between the coefficients
**

B', C, D', ... and the virial coefficients B, C, D, ....

p = RT(l/v). Third approximation. and the successive approximations are used to substitute for p on the right-hand side. or so as to interpolate and extrapolate in a way that is not subjective. B'2 + RTC'..2 + RTC')RT/V3 + [B'3 + 3B'C'RT+D'(RT)2]RT/v4 • The required relations between the coefficients are B = D = B'3+3B'C'RT+D'(RT)2. the normal practice is first to choose variables. and so scales for plotting the data.5 Empirical curve fitting by power series We may wish to fit a curve to experimental points so as to produce an economical expression for the data. 2. In the absence of any theoretical guidance as to the form of equation to be used. p = RT(l/v) + B'RT(1/v)2. Substitute. the numerical values of the coefficients of which are found by curve fitting. retaining powers up to (1/v)2.. etc. First approximation. p = RT/v+ B'(l/v)[RT/v + B'RT/V2 + (B'2 +RTC')RT/v 3] + C'(l/v)(RT/v + B'RT/V 2)2 + D'(1/v)(RT/v)3 = RT/v+ RTB'/v2 + (B. so as to obtain as near to a straight line as possible.9. p = RT(l/v) + B'(l/v)(RT/v + B' RT/V2) + C'(1/v)(RT/v)2 = RT/v + B'RT/V2 + (B'2 + RTC')RT/V3 Fourth approximation.60 Differential calculus Write the series in the form p = RT(ljv) + B'p(l/v) + C'p2(1/v) + D'p3(1jv) + . This may be done in . Second approximation. A common practice is then to use a power series. e= B'.

and if more terms are used than are necessary the series can become oscillatory. all of which are indeterminate. the method depending on the estimation of limiting tangents.. so that the extrapolation to x = 0 gives a l ' and so on. discussed in Section 7. . We can then calculate and plot (y .. Such difficulties may be resolved by the use ofL'Hopital's rule.a)f'(a) + (x . we can plot y against x and extrapolate to x = 0... or to the ratios % or 00/00. The principal difficulty is that uncertain extrapolation is involved. In that case it is important to use the smallest possible number of terms.8. a set of duplicate measurements producing a different curve. Given an equation y =f(x) = ao+a l x+a2 x2 + ... The corresponding analytical procedure is to calculate values of the coefficients by the method of least squares. which is that if two functions f(x) and g(x) are both zero at x = a then lim f(x) = lim f'(x) g(x) x . the intercept being ao. g'(x)' (2.53) x . when the equation becomes (y -ao)/x = al + a2x + a3 x2 + . but difficulties arise when this leads to a product of zero and infinity.. This geometrical method has advantages in that the effect of successive terms is readily apparent and the increasing scatter of points which is inevitably produced shows when higher terms become irrelevant. This is because a power series does not necessarily converge.. .a)2J"(a) + g(x) = g(a) + (x -a)g'(a) + (x -a)2g"(a) + .. .. This can be shown by using Taylor's theorem to expand the two functions f(x) f(a) + (x .. This can usually be done by simple algebra.The evaluation of limits by L'Hopita/'s rule 61 two ways. a curve of complex shape that passes through every point merely reflects the accidental scatter of the data.. 2. ...10 The evaluation of limits by L'Hopital's rule We often need to find the limit of a function when the variables are taken either to zero or to infinity.' so that when f(a) = g(a) = 0 we may divide numerator and ..ao)/x against x.

.a)g" (a) + .a) to give f(x) g(x) f'(a) + (x -a)f"(a) + g' (a) + (x . It is most often used either when a = 0 or when a = 00.30 Find the condition for the hyperbola x2 y2 ---=1 a2 b2 to be rectangular (asymptotes perpendicular to each other). and it applies equally well when the functions f(x) and g(x) both tend to infinity. and as x -+ a the rule follows.. x-+O Example 2..00 at x = O. Example 2.29 Evaluate lim (x In x).1 11m -1-. We have b2 x 2 _ y 2 a 2 = a 2 b2 and the asymptotes are straight lines of slope equal to the value of dy/dx as x -+ 00 and y -+ 00. sin x 11 m-X x-+O would give % at x = 0. . This gives a product of zero and .-+0 - 11//X x 2 = lim "''''0 (-x) = O.28 . By implicit differentiation 2b 2 xdx -2a 2 ydy = 0. Example 2.62 Differential calculus denominator by (x . but differentiation of both the numerator and the denominator gives · cosx .. dy b 2 x dx = a 2 y' . . We therefore write the expression as a quotient and apply L'Hopital's rule: lim (x In x) x-+O = lim 1ln/ x = lim x-·O X ".

the choice of which distinguishes one approach from another.11 00 and e h. The high-temperature limit. The principles of Newtonian mechanics Mechanics is concerned with force.7./kT _1)2 gives 0/0.9.31 An alternative way of evaluating limits is by the use of series expansions. momentum and energy. and the conditions for equilibrium under the action of forces is one application of the vector analysis discussed in Section 3. shows that for very large T. so that the condition for a rectangular hyperbola is b = a. (dy/dx)2 = b 2/a 2.The principles of Newtonian mechanics 63 which gives 00/00 as x and y tend to infinity.. (e hY/ kT -1) -+ hv/kT so that the limit as T -+ 2. as in the treatment of Hamilton and Lagrange.= 1m 2 X"" 00 dx X"" 00 a dy/dx so that in the limit.10 gives an equation that occurs in the theory of the heat capacity of a monatomic solid. Expanding the exponentials in the form e h . In the Newtonian approach we regard force as fundamental. as T -+ 00. Force is a vector quantity. Example 2. The subject is deduced from a minimum number of first principles.. Differentiation of numerator and denominator with respect to x gives dy l' b2 . Example 2. dy/dx = ± b/a. and this is usually the one adopted in introductory treatments because it can be applied directly to particular cases. m 11 . mass. becomes k/hv. An alternative is to regard energy as fundamental./kT -+ 1. of the expression hv e h./kT kT2 (e h./kT = 1 + hv/kT + (hv/kT)2 + . . which is discussed in Section 4. The asymptotes will be perpendicular if they have slopes of ± 1.

32 The conventional order of definition of mechanical properties is. If the force is constant the impulse becomes F(t2 -td. force as mass x acceleration. When two or more bodies interact with each other the forces. this gives f t.55) Example 2. This is the basis of inertial navigation systems. In a time interval dt. so that impulse = force x time = change in momentum. if the velocity of a body in the x direction is v = dx/dt. Fdt = flnV. the product F dt will be d2x dv Fdt = m dt 2 dt = m dt dt = mdv = d(mv). knowing acceleration as a function of time we can calculate velocity. This order can be reversed with advantage in physical applications. (2. and that force = mass x acceleration. first.56). that a body changes its state of rest or uniform velocity only if acted upon by a force.54) This will apply in any direction in space. and knowing velocity as a function of time we can calculate the distance travelled. finally.54) as a consequence of (2.64 Differential calculus Newtonian mechanics can be based on three principles: that when bodies interact with each other the forces of action and of reaction are equal in magnitude and opposite in direction. and so . Knowing force and mass we can calculate the acceleration. then acceleration as rate of change of velocity and. d(mv) = mV2 -mvl' tl mV l The product mv is called momentum and the integral on the left-hand side is called impulse. (2. a force F in the same direction will produce acceleration dv/dt = d 2 x/dt 2 given by F d2x = m dt 2 ' (2. we can use a transducer to measure the force that is exerted on a sufficiently massive body as a result of acceleration of a containing vehicle. Anticipating our discussion of integration.56) We could equally well regard momentum as fundamental and obtain (2. velocity as rate of change of distance.

Energy may be defined as the result of doing work. In the earth's gravitational field the force exerted on a body of mass m is called the weight wand the field strength is denoted by g. (2. so that force dV F= . which is the principle of conservation of momentum. Energy that changes with position in a field is called potential energy and is denoted by V.potential gradient. (2.dx' = . so that the same applies to the changes in momentum. The field strength is defined as the force on unit mass or charge.58) . The corresponding force F exerted by an electric field of strength E on a charge e is F= Ee. dv dx Fdx = m-dx = m-dv = mvdv. If the force is constant the work done becomes F(X2 -Xl) so that work = force x distance = change in kinetic energy. The total momentum will therefore be constant.54) we see that g can be identified as the acceleration of free fall under gravity.57) The forces with which we are concerned are often due to interaction with a field. Xl VI The expression -!mv 2 is called kinetic energy and the integral on the left is called work. work F dx is done on the body and its energy increases. dt de Integration then gives I x> F dx = IV> mvdv = -!mv~ --!mvI. as with mass in a gravitational field or electric charge in an electric field. Since force is in the direction of decreasing potential energy we have dV= -Fdx. acting will be equal in magnitude and opposite in direction. When force F produces displacement dx in mass m.The principles of Newtonian mechanics 65 the impulses. By comparison with (2. When a body is given a displacement dx in the direction opposite to the force F exerted on it by a field. so that w=mg.

59) For this to be true.6 represents a diatomic molecule consisting of masses m1 and m2 rotating about an axis perpendicular to the paper through point O. A molecule may be regarded as a set of discrete masses at the positions of the nuclei. or sliding on a rough surface introduces a resisting force which reduces the acceleration of the body. linear motion is converted into random thermal motion of the body and its surroundings. clockwise and anticlockwise moments being given opposite signs. In particular.7. For an extended body to be in equilibrium under the action of forces. some of the work done is converted into heat. An applied force not passing through the centre of mass will tend to produce rotation of the body. such as a fluid. Rotational properties are expressed in terms of polar coordinates (Section 1. We therefore introduce the principle of moments. as if the masses concerned were point particles. 2. or a charge is moved by an electric field. (2. the decrease in potential energy is equal to the work done. all of the assumptions that have been made above must be true. The motion of a structured body in three-dimensional space. The arguments used apply to translation of the centre of mass of an extended body when it is treated as if all the mass were concentrated at that centre of mass. so far. neglected structure. Denoting kinetic energy by T and potential energy by V we have T+V = constant.5). the only effect of the action offorce must be to increase the velocity of a body. but also the sum of the moments of the forces about any point must be zero. which by (2. This will be so only in the absence of friction. Mass m1 moves with tangential velocity V 1 in a circle of radius r 1 about O. which is an irreversible process in the thermodynamic sense. although the conservation of momentum always applies to real processes. conservation of energy in the simple mechanical sense does not unless some formulation which includes heat is adopted. not only must the sum of the components of the forces in any direction in space be zero.57) will be the increase in kinetic energy. Our discussion of mechanics has. such as that of a molecule in a fluid. Fig. Movement through a resisting medium. and all the forces were moved parallel to themselves so as to pass through that centre of mass. Thus.66 Differential calculus If a body falls freely under gravity. is resolved into linear motion (translation) of the centre of mass and rotation about the centre of mass. At the molecular level. the moment of a force about a point is defined as the product of the force and the perpendicular distance from the point to the line of action of the force. In time interval dt the radius vector r1 .

must exert equal and opposite forces on the two masses we have and the position of the axis of rotation will be where (2. this must exert a force on m j . this corresponds to an acceleration towards the centre given by · . (2 . so that a molecule that is freely rotating in space does so about its centre of mass.6 moves through angle dO and dS j = r 1 dO. the bond length would increase with increase in w. this being called centrifugal stretch. A general motion of a rigid body in space. measured in radian/second. or bond. which includes both translation and rotation. or for a chemical bond. From the triangle of velocities in Fig. or the mass exerts a force on the string given by (2. 2. Then mj moves through an arc of length (2.62) If the string were elastic. the linear velocity Vj will be constant in magnitude but changing in direction. Since the string. dO/dt.The principles of Newtonian mechanics 67 Fig. can be expressed as the sum of translation of the centre of mass and rotation about the centre of mass.63) This is the condition for finding centre of mass.60) where OJ is the angular velocity.6. At constant angular velocity w. I acce I centnpeta eratton = Vj dO dt = VjW = r l w 2 . 2.61) If we suppose the masses to be connected by a weightless string. This may be seen by considering some particular line fixed in the body and . in time dt the mass mj gains velocity Vj dO in the direction towards 0.

.68 Differential calculus perpendicular to an arbitrary axis of rotation. The quantity in brackets is called the moment of inertia I of the molecule about an axis through the centre of mass and perpendicular to the axis of the molecule. (2.67) In the general case of a molecule composed of masses m 1 .11).. rotation through the same angle about any parallel axis will then produce the same angle of rotation of the line. Then kinetic energy of rotation If the internuclear distance is . m2. . The moment of inertia about a particular axis is then 1= 'Lmi'f. m m1 m2 (2.64) = '1 +'2 then (2. so that T = tm 1 vI +tm2v~ = !(m1ri +m2d)co 2.63) gives and (2. the centre of mass of the hammer travels in a parabola (Example 1. together with translations of the line parallel and perpendicular to itself. Thus when an athlete throws the hammer.. m1 +m2 We define the reduced mass m* by 1 so that 1 1 . = t I (1)2. held by chemical bonds at various positions in space.2.66) (2.=-+-. We may therefore choose any parallel axis to describe the rotational part of the motion. the hammer meanwhile rotating about its centre of mass. for which we choose the one through the centre of mass. and the remainder of the motion will be translation.65) so that 1= m1 m 2 . the centre of mass is a point such that for any axis through that point where 'i is the perpendicular distance from mi onto the axis. The kinetic energy T of our rotating diatomic molecule is the sum of the kinetic energies of the two masses.

x sin ()2 to the moment of inertia about OX'. 2. Mass ml at point P(x. The total moment of inertia about the new axis OX' is then I' = L mi (Yi cos () . 2. the point P will assume new coordinates (x'. Of through the centre of mass as in Fig. y)will contribute m1 (PR)2 = m1 y2 to the moment of inertia about the x-axis OX.y.Xi sin ()2. .. ' .. = ON = OQ + QN = OQ + RM = x cos () + y sin ()..7 The mass ml then contributes m 1 (y cos () .and .7.y) \ \ \ \ \ \ \ \ \ \ \ \ \ M \ . Of' by rotating the axes through an angle (). The expression LmixiYi is called the product of inertia about the x.The principles of Newtonian mechanics 69 When the molecule is planar we can choose rectangular coordinate axes OX. OY' given by = PN = PM -MN = PM -QR = ycos()-xsin(). Ifwe then choose arbitrary axes OX'.. and will have a stationary value if dJ' Id() = O... This is when :~ = - 2L mi (Yi cos () - = -2Lmix. y' x' y yl\ \ \ \ \ P(X.. y') relative to OX'.0' \ \ x \ Fig. = Xi sin () (Yi sin () + Xi cos () o. The moment of inertia of our arbitrary planar molecule will therefore vary with the angle ().

8 -6 r2 I~ Ee .6. the moment of the force about the axis is called the torque. In three dimensions we find corresponding conditions.70 Differential calculus y-axes. then in field of strength E each mass experiences a force Ee.. The axes for which this applies are called the principal axes of the molecule. as in a molecular collision. we obtain the simplest expressions. 2. the torque about 0 is Fr 1 • The effect of such a single applied force.e respectively. in a direction perpendicular to the axis of the molecule. and the moments of inertia about the principal axes are called the principal moments of inertia of the molecule. and we see that if this is zero the moment of inertia will be a maximum or a minimum. We therefore consider a simpler case. When a force acts to cause rotation of an extended body about an axis. Thus if force F is applied to mass m1 in Fig. If the charges on m1 and m2 are + e and . and so will be the same about any axis Ee E m1 +6 0I r1 ~ . is complicated by the fact that it produces both rotation about the centre of mass and translation of the centre of mass. This moment depends only on the perpendicular distance r = r 1 + r2 between the parallel forces. and this is called a couple. When we choose these principal axes as the ones about which we express the rotational properties of the molecule..8. which is when the molecule is polar and the force is produced by interaction with an electric field in a direction perpendicular to the axis of the molecule. 2. as in Fig. that when the axes are chosen so that the products of inertia about each pair of axes are zero. The total moment of the two forces about 0 is then Eer 1 + Eer2 = Ee(r1 + r2). Fig. These forces will be equal in magnitude and opposite in direction. the moments of inertia will have maximum or minimum values. 2.

56) applies. From this definition and (2.9.70) the total angular momentum of the two masses about 0 being angular momentum = m 1v 1 r 1 +m2v2r2 = (m 1 d +m2rDw = Iw.71) .68) £e +e £ ma r----------~-i-n~9------~-e £e Fig.71 The principles of Newtonian mechanics perpendicular to the plane containing the forces. 2.lE sin e.69) In Fig. their resultant is zero. so that force = rate of change of linear momentum.9) passing through the mass points and perpendicular to the axis of the molecule. If the molecule were inclined at an angle e to the direction of the field. torque = Eer sin e = J.l being called the dipole moment of the molecule. when they are translated so as to pass through that centre. Such a pair of equal and opposite forces will have no effect on the position in space of the centre of mass of the molecule because. (2. 2. so that mass m1 has linear momentum m1 V1 and angular momentum m 1V1 r1 about an axis through O. in general. (2. For linear motion equation (2. The effect of the field is therefore to apply torque and produce pure rotation of the molecule.9 the product re = J.69) we obtain moment of force = torque = rate of change of angular momentum. (2. as in Fig. the perpendicular distance between the forces would be r sin e so that. 2. We define angular momentum as the moment of the momentum. (2.8 the forces and the momenta are both vectors (Section 3.

so that the total angular momentum about any axis remains unchanged. When rotation is possible.9.72) The relations used to describe angular motion are thus the same as for linear motion if we replace force by moment of force. When structured particles collide. and makes additive contributions to both potential energy and kinetic energy. (2. . so that kinetic energy = tml vf + tm2v~ = t(m1rf + mzd)w Z = tIw 2. we have a corresponding conservation of angular momentum.70) these produce equal and opposite rates of change of angular momentum. If we calculate the total angular momentum of the two bodies both about the same arbitrarily chosen axis. but for molecules we must include also the vibration of the chemical bonds. these impose additional constraints. the forces acting on impact contribute equal and opposite moments about that axis. This vibration is discussed in Example 6.72 Differential calculus The rotational kinetic energy of our molecule is simply the sum of the kinetic energies of the masses. Problems involving collisions between structureless particles are solved by using the principle of conservation of linear momentum. For a rigid body we need to consider only translation and rotation. mass by moment of inertia and velocity by angular velocity. but the principles outlined above still apply. In problems involving real molecules it may be necessary also to take account of the energies of allowed quantum states. By equation (2. The motion of a structured body in three-dimensional space is resolved into translation of the centre of mass and motion relative to the centre of mass. equal and opposite forces act on each one at the point of contact.

x) plane as in Fig.1. constant y means a plane parallel to the (z. The corresponding analytical problem is: Given a function z = z (x. 0 ux . we can apply the methods of two-dimensional calculus to this curve. Given a relation z = z (x. partial differentiation 3.y) (0 z/ uX. then !'l ) _ I· z(x+~x.. v. y).4). (3. how does z vary with x when y is kept constant. Ifthe surface cuts the plane in the curve shown. T) properties of a gas or the vapour pressure of a two-component mixture. which apply when the third variable is held constant. Thus we may consider p against vat constant Tfor a gas. and with y when x is kept constant? The general principle of taking two-dimensional sections of multidimensional space enables us to use the methods of differential calculus developed in the previous chapter.CHAPTER 3 Differential calculus in three or more dimensions. Here we have three variables and an equation connecting them. The analytical definition of this partial derivative is: If z is a continuous function of the independent variables x and y. y). A tangent at point P will have a slope which is 'dz/dx in the plane at constant y'.1 Significance and notation Typical three-dimensional problems are the (p. or p against mole fraction x at constant T for the mixture. 3. In such cases the geometrical method is to draw a three-dimensional diagram and then to consider twodimensional sections.y)-z(x.1) 1m i: 6" . so that two variables are independent (see Section 1.. This is written (oz/ox)" to be read as 'partial dz by dx at constant y'.. which defines a surface in three dimensions.

v. are called total derivatives.u.x plane at constant y Fig. at constant x and at constant z. and there are important differences between the properties of total and of partial derivatives. The rules for partial differentiation are the same as in the previous chapter. where only two variables were concerned. we define the partial derivative with respect to anyone of the independent variables by keeping the others constant.74 Three or more dimensions. Thus if z depends on x.u. y being kept constant. Analogously. denoted by (iJz/iJy)x and (iJy/iJx)z' The derivatives in the previous chapter.u. u and v we can write z = z(x..y. (iJz/iJy)x. say. When we have more than two independent variables.v' etc. y. we can define partial derivatives in the two other planes. 3. partial differentiation z y x z.v) and can define partial derivatives such as (iJz/iJx).1 where only the variable x is changed from x to x + {)x. One property that they have in common is that the reciprocal of. it being necessary only to remember that variables that are . (iJy/iJx)z is (iJx/iJy)z just as l/(dy/dx) = dx/dy.

6 does not apply to the partial derivatives when a different variable is held constant in each case. if we write oX / aTit may mean at constant v or at constant p. with partial derivatives a minus sign appears. The method of implicit differentiation described in Section 2.Significance and notation 75 being held constant are treated in the same way as any other constants. In the above example we can find (ox/oY)z by differentiating the equation at constant z. we normally have a wide choice of independent variables. For example. This may be written in the form 0 2 z/oxoy or more explicitly as o:. The higher-order differential coefficients such as (0 2 z/ ox2 )y are a straightforward extension of the rules for total differentiation.y == (:x (:.3xy + 7 then (OZ/ox)y = 4x . This relation is derived analytically in Section 3. such as z (x. Thus for some property X. as (oX/aT)v and (oX/aT)p.3y).3y and (OZ/oy)x = .3y)dx This example shows that the chain rule of Section 2.6).3x. however. )x)y == (:x)y (!. We therefore need either to use the full notation. these being different quantities (the relation between them is discussed in Section 3. which gives 0= 4xdx-3xdy-3ydx so that at constant Z = 3xdy at constant z (ox/oY)z = 3x/(4x . (4x .)x· These alternative forms all show that z is first to be differentiated with . thus (oz/ox)y(ox/oY)z = 3x = . if z = 2X2 . We can then use the abbreviated notation oz/ox to imply constant y without ambiguity. but a new possibility also arises in the form of a mixed second derivative.(oz/oY)x. it will be clear that partial differentiation with respect to one variable implies that the other is to be kept constant. or to begin with an explanatory statement such as 'using independent variables v. In physical applications.4 can also be used. y).T' so that oX/aT then means (oX/aT)v and not (oX/aT)p. When a function is defined in terms of particular independent variables.6.

If. From the above example we have oz oy = -3x. on the other hand.y) to a nearby point z (x + ~x. 02 Z oxoy = -3. here and in the next section.2 . Fig.Y) Z(X+h. the same result would be obtained. We can move from a point z (x. the z-axis being perpendicular to the plane of Y 8y O z(x. 3. that under proper conditions the order of successive differentiation may be reversed. partial differentiation respect to y at constant x. the result then being differentiated with respect to x at constant y. 3. The properties of partial derivatives are so important in chemistry that we give alternative proofs of that principle. A geometrical interpretation of a mixed second derivative is that it measures the rate of change of slope as we move in a perpendicular direction over a surface. corresponding to moving along two sides of a rectangle.Y+6Y ) fix x Fig. the order of differentiation were reversed by differentiating first with respect to x at constant yand then with respect to yat constant x. the rate of change of slope oz/ox with change in y is o oz oyox' The derivative in reverse order is the rate of change of oz/oy with change in x. Using x and y as independent variables.y + ~y) by changing one variable at a time.2 shows a projection of this into the (x. y) plane. This illustrates an important principle.76 Three or more dimensions.

y) plus the increase in slope on moving by either ~x or ~y. The expressions must be the same if a unique surface exists over which the alternative movements are made. is to consider the behaviour in terms of changes whose small magnitude is actually assessed at each stage of the argument. which has some advantages.y)-z(x. whereas change first in y and then in x gives oz (oz 0 oz ) ~z = oy ~y + ox + oy ox ~y ~x.y+h)] -[z(x+h. and the conditions for that are that the variables shall be independent and the function and its derivatives shall be finite.y)+hO(p). so p becomes larger.y) + O(P) ox y h where O(P) means a term which is zero to p decimal places. (3. The advantage of this interpretation of a derivative is that we can then multiply both sides by h to obtain h(!:)y = z(x+h. The two expressions will be the same so long as the order of successive partial differentiation can be reversed.3) .y)-z(x. An alternative approach.y+h)-z(x.2 An alternative approach to calculus The interpretation of calculus as the behaviour of functions in the limit of infinitesimal changes that was adopted in the previous chapter is the more conventional one. increase y by a small increment.2) ( oz) = z(x + h. We write (3. for which we again choose h. 3. As the increment h in x is made smaller. By changing x first and then y the change in z is oz (oz 0 oz ) ~z = ox ~x + oy + ox oy ~x ~y. and write the increase in the function again with an error O(P). We are now able to write the mixed second derivative by applying the same rule. continuous and single-valued. Hence 2 0 OZ 2 02 Z [ h oyox = h oyox = z(x+h.z(x.An alternative approach to calculus 77 the paper. In each case the expression in the brackets is the slope at z (x. This regards the derivative not just in the limit but as differing from a ratio of infinitesimal quantities by a term that is zero to a number of decimal places.y)]+h2 0(p).y) .

so that we can increase either variable by h without change in the other variable. we can in fact separate this as the ratio of dy to dx. y). presupposing that it is such a small increment that any conclusions we draw are actually valid only in the usual calculus limit. all derivatives are total derivatives. The gradients in these two planes are given by the partial derivatives at constant x and at constant y respectively. We then have the trivial . We have seen in the previous section that we can if necessary place this on a firmer footing by stating explicitly that we are in error by a quantity that is zero to a certain number of decimal places. Consider a particular path which moves first in the plane at constant x. 3. but with three or more variables we use partial derivatives. The geometrical interpretation of the total differential dz is that it is an increment in z.78 Three or more dimensions. because 82 z h2 8x8y = [z(x+h. (3. It could be regarded as the increase in altitude when we walk obliquely across a hillside.3 The total differential We have seen that if we have only two variables. the slope of a tangent is dy/dx.[z(x. partial differentiation When written in this form. and then turns through a right-angle to move in the plane at constant y so as to reach the desired finishing point. The important thing about a total differential is that it is a real quantity. the meaning of dz is the increase in z when we move over the surface from one point on the surface to another one which is very close to it.4) to be the same. For this to apply. This geometrical interpretation is here very useful because it is easily seen that we can move from one point to another on such a surface by a variety of routes. the same result is obtained by partial differentiation in either order.y)] .4) and simple algebraic manipulation shows the right-hand sides of (3.y+h)-z(x+h. y)]+h 2 0(P). and so twodimensional diagrams. the variables x and y must be independent. If we have an equation for a surface in three dimensions in the form z = z(x. y+h)-z(x.3) and (3. We have also seen that although the total derivative dy/dx denotes the operation of differentiation on the function y = f(x). which is the increment in y divided by the increment in x. We shall now see how we can use a total differential of this kind even with more than two variables. In two dimensions.

.v. so that if z = z(x. This equation means that the differential change in pressure produced by simultaneous differential changes dT in T and dv in v is given by the right-hand side of the equation. dx + (:. It is important to notice that this expression contains a mixture of total and partial differentials. dy + (3.6) Example 3.) x. Thus dz = (!:)y (:. v) so that dp = (Z)v (::)T dT+ dv. The total differentials dz and dx are potentially real increments.The total differential 79 relation meaning that the increment in y is the derivative times the increment in x. (3. These differential changes do not have any physical significance because they need to be integrated (Chapter 4) before they represent real changes.1 The pressure of a gas depends on the temperature Tand molar volume v.5) This is the definition of the total differential of z = z (x. so far. dz is.u.. the geometrical argument shows that we then add the increments in z produced by the component moves.v. .. Combining this expression with the principle that we can make a movement over a surface by movements in perpendicular planes. y). Thus p = p (T.. ) then dz = (!:) y.)x dx+ dy.y. . When the same relation is applied to a two-dimensional section of multidimensional space it becomes dz = (!:)y dx..u.u. .v. and integration will be possible only .. the increase in z when we move distance dx in the plane at constant y. When we have more variables a term is included for each.

w). but still only two of them are independent. x and y independent. (3. It was also stated that the definition applies even when the variables are not independent and this will now be shown. and applies whether or not the variables are independent. y). This will be shown in the next section since it provides useful practice in handling differentials.80 Three or more dimensions.4 General expression for a total differential The geometrical justification of the definition of a total differential given in the previous section was based upon the assumption that a surface could be drawn in space of dimensions equal to the number of variables.10) . y and w. This means that a relation must exist between w. 3. The definition of the total differential is the basis from which other useful relations are derived. which may happen in physical applications either by accident or by design. y. z. so that we can use these relations even when we do not use only independent variables. in which case R p v T and so that in this case dp = p P -dT--dv T v or dp p dT T dv v =--- The definition of a total differential is quite general. y). of course.w) (3. x and y. Suppose a quantity z depends upon the independent variables x and y. x. z = z(x. predict such properties by purely mathematical means. partial differentiation when we know the physical properties of the gas. We cannot. The existence of an equation for the surface means that all but one of the variables can be independent. (3. which we write as w = w(x.9) Since z may now be expressed as a function of any two of the variables we may also write z = z(x. (3.8) We now have four variables. If we make the physical assumption that the gas is perfect we then have pv = RT.7) Suppose we now choose to introduce another variable wand write z = z(x.

(3.17) Substitution of (3.) w dy + (!: ).10) and (3.8) in the first place and written the total differential to include the superfluous variable.18) which would be obtained if we had assumed (3. Each of these expressions contains the total differential of three of the four quantities z.X dy+ (!:) "X dw.11) dz = (:. Dividing (3. we can write the corresponding expressions for the total differential dz: from (3. from (3.15) dx+ (3. (3.14) dw.81 General expression for a total differential and z = z(y.15) will remain true if we choose to hold the fourth quantity constant.9) from (3. x.7).)x dy. dx+ (~. The number of independent variables that properly apply to a particular application is a physical problem. (:.11) all express w or z as a function of two variables.13) dz = (!:)w dX+(!:)x dw. (3. (3.9).12) and using (3. y and w. . (3.15) by dy at constant x gives (3. The significance of the above is to show that we do not make any mathematical errors as a result of a physical error in choosing that number.13) gives dz = (!:) w.14) and (3. (3. (3.11) Since equations (3.)x dw = (~:).16) and (3.12) dy.7) dz = (!:). (3. w). In particular.14) by dx at constant y we obtain and dividing (3. not a mathematical one.17) into (3.' dx+ (:z) y w.10) from (3.

2 Show that (2xydx + x 2 dy) is an exact differential. partial differentiation Exact differentials This term applies to an expression of the form X dx + Y dy. Example 3.5 Three or more dimensions.19) dz = Xdx+Ydy.20) = (!:)y dX+(!. This will be so if the relation (3. whereas o oy (2xy) = 2x o ox (x 2 ) = 2x and the test succeeds with the first expression whereas from the second expression o _(x 2 ) oy o =0 ox (2xy) = 2y and the test fails.)x dy and the test for exactness is to see if this is so. equation (3. .21) This operation is called cross-differentiation.82 3. This can be done by taking the mixed second derivative and checking that the order of differentiation can be reversed. z say. and means that a quantity. (3. Thus to confirm that and Y= (oz) oy x we calculate (~~)x = O~2:X = (::). so that (3. (x 2 dx + 2xy dy) is not. and is true for exact differentials. exists whose total differential is of that form. Applying the cross-differentiation test.21).19) is identical with dz (3.

22) dv = (!.22) gives work = -pdv = _p(ov) dT_p(ov) dp.) dT + (:. = = -p These would be identical only if (ov/OT)p were zero. If the volume is changed by dv at constant pressure. Ifwe move from one point (Xl' YI)on the surface to another (X2. Thus v = v(T. or op T p p We can show that work is not an exact differential by applying the cross-differentiation test. which is very significant.Exact differentials 83 Example 3.)T dp. Given the exact differential dG = vdp-SdT. Y2) by changes in X and y. This is not physically possible because the coefficient of expansion always finite and positive. Example 3. an inexact differential does not define a unique surface and the change in z will depend on the particular path followed from (Xl' YI) to (X2' Y2). the work done on the gas is . Substitution for dv from equation (3.4 Cross-differentiation of an exact differential is used to obtain the Maxwell relations in thermodynamics from the Gibbs equations. The above example shows that work is not a thermodynamic function of state. so that changes in the function are independent of path. . Conversely. equation (3. A thermodynamic function of state is one that can be expressed as an exact differential.21) )J)T -(!. (1/v) (ov/op)Tis The geometrical significance of this is that for an exact differential a unique surface exists when z is plotted against x and y. (:r[ -p(:. the change in z is uniquely defined by the initial and final values of X and y.3 The volume v of a sample of gas depends on the temperature T and the pressure p.)J)p O~2~.t-p O~2~. p) and (3.pdv. (:p[ -p(!.

at constant z. (&) du + av U. .24) can be used to obtain purely mathematical relations between experimental properties that are defined as partial derivatives.5 Equation (3. w. v. .. Dividing by dy at constant z (3. 3.. w. v.84 Three or more dimensions...w... u.1.23) At constant z. a straightforward extension of the above argument gives.v.6 Relations between partial derivatives We can use the definition of a total differential given in Section 3. . ) u. .. for any pair of terms. . (~:) = (~~) = v.)x dy... ' and so on. partial differentiation cross-differentiation gives When we have more than two independent variables. .. and the cross-differentiation test for an exact differential Adu+Bdv+Cdw+ .. .3 to obtain useful relations between partial derivatives... w..w. .)x dx+ dy. so that (!:)y dx = -(:. ) (&) dz = au V. . (&) dv + aw dw + U. y) then dz = (!:)y (:. w.. If z = z (x.. for z = z(u. dz = 0..24) which is the chain rule for partial derivatives mentioned in Section 3. becomes ( ( ~~ :~ ) u. (3. Example 3..

y). the new variable must be related to the existing ones to be useful. In the same way.23) by dy we obtain dz _ dy - (az) dx + (az) ax dy ay y (3.25) x· The variables most often used in physical applications are p. For example. the heat capacity at constant pressure C p = (aH/aT)p.Relations between partial derivatives 85 The coefficient of expansion IX = (l/v) (av/OT)p. so that . so that any experimental measurements of these three quantities which did not obey this relation would certainly be wrong. v and T.23).26) This equation relates partial derivatives when the variable being held constant is changed. isothermal compressibility (l/v) (av/ap)T and thermal pressure coefficient y = (op/OT)vare related by . If we choose to introduce a new variable.. only one variable is independent and we can write the total differential of z with respect to either one of the variables.= - ICy = _!(av) (a p ) = !(av) = V ap T or v v or p IX. into our relation z = z (x. The relation is useful also in that anyone quantity can be obtained by measuring the other two.6 The molar heat capacity at constant pressure Cp = (aH/OT)p. and that at constant volume Cv = (aU /OT)v are related because H = U + pv. Example 3. If a relation exists between the variables x and y in equation (3.23) by dy at constant u we obtain (3. two of which are independent due to the existence of the relating equation of state. Dividing equation (3. isenthalpic Joul~Thomson coefficient p = (aT/ap)H and isothermal Joul~Thomson coefficient ¢ = (aH/ap)T are related by¢= -pCp. dividing equation (3. u say.

)v (o.86 Three or more dimensions.)p +(~~)T(~)V = so that The term (OH/OP)Tcan be interpreted using the Gibbs equation dH = TdS+vdp and dividing by dp at constant T to obtain (~~)T =T(:!)T+ V. volume. temperature and refractive index. From equation (3. In mathematical terms this is a distinction between functions of first degree. A functionf(x) is of degree m in variable x if when x is multiplied by a constant. Hence (oH /oPh = 0 and perfect gas. and those of zero degree. pv = RTso that (ov/O/)p = R/p and (op/iJT)v = R/v. in the amount of substance.4. energy and entropy. partial differentiation To develop this into a relation between C p and C v we need to relate (oH/O/)vand (oH/O/)p.. say A. Euler's theorem The terms extensive and intensive are used in a physical context to distinguish between variables that are proportional to amount of substance (extensive) and those independent of amount of substance (intensive). f(AX) = Am f(x). Extensive variables include mass. ( OH) op T or P Cp = Cv+T(::)v (. This is as far as thermodynamic and mathematical relations can take us. Using the Maxwell relation of Example 3. and intensive variables include pressure.7 Extensive and intensive variables. .26) (o. we obtain so that = v _T(OV) . )p. 3. the value of the function is multiplied by Am. For a perfect gas.

If an extensive variable z(a. a functionf(xl.27) is homogeneous of degree m if (3. multiplying the amount of substance by A means multiplying the extensive variables by A but keeping the intensive variables constant.28) and Euler's theorem is Xl of uX I of uX2 of uX 3 ~+X2~+X3~+ . y) = Az(a. b. An intensive property. Ab.27). x. b. y) z(a. y) depends on extensive variables a. In physical problems we usually have more than one variable to consider. Then Z(Aa. y. = mf (3. b.7 For a mixture of substances 1 and 2 which is able to gain or lose numbers of moles dn l and dn2 of substance. b. x. This is true for any A. Partial differentiation with respect to Aat constant a. x. so put A = 1 to give OZ OZ a oa +b ob = z. Euler's theorem 87 A function of first degree is one that is proportional to the variable f(AX) = Af(x) for m = 1. Example 3. x 3 . some of which will be extensive and some intensive. x and y gives OZ d(Aa) OZ d(Ab) o(Aa) ~+ o(Ab) ~ oz OZ a o(Aa) + b o(Ab) = = z(a.. In general. x. is a zero-degree function f(AX) = for m f(x) = 0. x. the total energy U is given . intensive.27) When we have an equation in differential form for a function of state in terms of both extensive and intensive variables we may use Euler's theorem in the form (3.. extensive. x 2 . so that only the former appear in equation (3. band on intensive variables x.29) In the above example the function is first-degree in the extensive variables but zero-degree in the intensive ones. ••• ) (3. y). y). on the other hand.Extensive and intensive variables. b.

pdv + J.l2.-2... But from the given equation au as =T. a2f/axay by I". Thus if a function I(x. nl' n2) and Euler's theorem gives au au au au S as +vav+nlanl +n2an2 = U. v. .88 Three or more dimensions.S dT + J.ll +n2J. The treatment for more than one independent variable is analogous but with total derivatives being replaced by partial derivatives with respect to each independent variable.. II un2 U =TS-pv+nlJ.ll dnl + J. and so on. b) = alx+a2y+a3x2 + a4XY + asy2 + . We write U = U(S. On the other hand.(a+x.2. y) is known at point (a.8 Taylor's theorem in partial derivatives Taylor's theorem is derived for functions of a single independent variable in Section 2. b) we assume that for the point (a + x. integrating differential equations.. n l and n2 are extensive. b + y) the increment in the function may be written as a power series in the form I(a+x.. au and -. We denote the partial derivative with respect to x.. The variables S. 3. v. partial differentiation in differential form by dU = TdS . b+ y) -/(a.l2 dn2 where nl and n2 are extensive but p and Tare intensive. By using Euler's theorem in this way we are. Euler's theorem therefore gives terms in p and T not appearing. by I" and higher derivatives a2f/ax 2 by I"". Then I.ll dn l + J.9.--. b+ y) = al +2a3x+a4Y+ . in effect. so that au av= -p.. of/ox. for the Gibbs function G we have dG = v dp ..l2dn2.

and the (scalar) magnitude of Aisdenoted by A or IAI. b) = a2.(a. b).+ yfy+ 2!1 (x f""+2xyf.3..31) A+B= C. A+B-C=O. direction and sense. so that . b)/2 and a4 = f"y(a.. Band C we can write the vector equations (3. b+ y) = f(a. b) = at.5. ThenifOP. Taking second derivatives and then putting x = y = 0 gives a3 = f".(a. 3. .y+ y fyy) + 2 2 . velocity and force are vectors. Conversely.Vectors 89 and putting x = y = 0 gives f. .. A and B. 3. which have magnitude only. in the directions of A and B.3 We denore a vector quantity by a letter in bold-face type. In this way we obtain Taylor's theorem in the form f(a +x. they are then added together by joining the arrows head to tail as in Fig. Similarly fy(a. such as A. Mass and time are examples of scalar quantities. When C is obtained by adding together A and B. b). (3. 3. it is called the resultant of the two vectors. we may resolve vector C into its components. whilst displacement... A vector quantity may be represented by an arrow of length proportional to the magnitude of the quantity and in the appropriate direction and sense. This is used in Section 7. PQ and OQ represent the vectors A.C is a vector with the same magnitude and direction as C but opposite sense. where displacements OP and PQ are added to give OQ. b)+ xf. Q o~------~------~ A Fig.9 Vectors A vector quantity is one that has magnitude.30) where the derivatives are evaluated at the point (a. in contrast to scalar quantities..

+kA z. and A. A. and may be operated upon like any other scalar quantity.32) A = iAx+jA. Thus we may multiply vector s by scalar a to obtain a vector of magnitude as and in the same direction and sense as s. 3.90 Three or more dimensions. Arguments of this kind may be used to show that momentum.. A useful analytical device is to define a set of vectors i. (3. j and k in the directions of rectangular coordinates in space and each of unit length.4 the line OP represents vector A and gives scalar projections Ax. We may then add the components to obtain (3. y.33) When a body is free to rotate in space it does so about its centre of mass . for example. differentiation of the scalar magnitude s of displacement s with respect to time t gives the velocity vector v in the same direction as s and of magnitude ds/dt. In Fig. and this is the component of A in the x direction. on the x. z y x Fig. We may also differentiate with respect to a scalar quantity. partial differentiation The magnitude s of vector s is a scalar.4 Addition or subtraction of vectors is then achieved by adding or subtracting the components giving. The product iAx is then a vector of magnitude Ax in the direction of the x-axis. acceleration and force are also vector quantities. 3.and z-axes.

o about a horizontal axis it has angular momentum 1c.3) of the axis of rotation.5. If it is supported at one end only. It may be shown that a rotation through angle w about that arbitrary axis is then the same as the sum of rotations wi. There is.. m and n are the direction cosines (Section 1. one difference between the vectors used to represent these angular properties and the vector quantities previously considered. then the component rotations about the axes obey Pythagoras's equation (w1)2 + (wm)2 + (wn)2 = w2 W+ m2 + n2) = ro 2 • and the arrows may be treated as any other vector quantity. ' . where I. so that any arbitrary axis of rotation makes angles (x. ~ __~x____+T~____~~/~ mg _ .and z-axes. Example 3.o.11). wm and wn about the three coordinate axes. are non-localized in the sense that they may be freely moved parallel to themselves.8 When a gyroscope spins with angular velocity c. it follows that angular velocity. on the other hand.Vectors 91 (Section 2. f3 and y with the X-.5 . angular momentum (the moment of linear momentum) and torque (the moment of a force) may be represented by vectors in the same way. Using the same arguments about scalar manipulation as above. 3. namely that the vector is localized in the axis of rotation. represented by a vector located in the axis of rotation as shown in Fig. It follows from this that if we represent angular rotation by an arrow drawn in the axis of rotation. mg I 'I . y. 3.\ .7. however.. Vectors based on displacement of an extended body. of length proportional to the angle of rotation and in the conventional sense that the rotation is in the direction of a right-hand screw. We then define three rectangular coordinates through that centre of mass as origin.-1. mgxdf ~ Fig.

the unit vectors i. This product is a scalar quantity and is the magnitude of one vector multiplied by the projection of the other upon it. B and as the vector (or cross) product A x B. couple = rate of change of angular momentum.70). j and k are defined to be orthogonal so that i.9. and when they are perpendicular to each other the scalar product is zero. From equation (2. rotates in the horizontal plane in an anticlockwise direction from that point of view. The sign will be positive or negative according to whether the angle () is acute or obtuse.92 Three or more dimensions. so that in time interval dt angular momentum mgxdt is generated.B = ABcos(}. which therefore precesses about the direction of the magnetic field. The two angular momentum vectors are drawn in the figure as when viewed vertically downwards. Nuclei that behave in this way are detected by the absorption of radiation of frequency equal to the precession frequency in a nuclear magnetic resonance spectrometer. A and B.i = j.1 Scalar and vector products The product of two vectors. This rotation of the gyroscope about the vertical axis is called precession.k = 1 and . these forces generate a couple mgx represented by a vector in the horizontal plane perpendicular to the axis of rotation. represented by a vector into the plane of the paper in Fig. partial differentiation the weight mg of the gyroscope is balanced by upthrust mg at the support. The scalar product is defined as A. the direction of precession being the opposite of what would occur if the gyroscope were a wheel running on a horizontal table. the precession frequency depending on the magnitude of the magnetic field strength at the particular nucleus. the resultant. the field imposes a couple on the spinning dipole. 3.j = k. and so the axis of rotation of the gyroscope.34) where () is the angle between the two vectors. as the scalar (or dot) product A. When the vectors are expressed in component form. (3. This also applies to the effect of a magnetic field on a spinning magnetic dipole. such perpendicular vectors are called orthogonal. When the two vectors are parallel it is simply the product of their magnitUdes. is defined in two ways.5 in the direction perpendicular to the axis of rotation. 3.

which is the area of the parallelogram based on the two vectors as in Fig.7. The vector product represents the magnitude of one vector multiplied by the component of the other perpendicular to the first.r = A cos O.37) The moment ofa vector about a point. When the order of writing the product is reversed. can be expressed as the vector product of F and the radius . (3.j = j.B = (iA" + jAy + kA. that plane surface area may be represented by a vector of length proportional to the area and in the direction normal to the surface. AxB Fig. 3. 93 (3. When plane surface area A is viewed obliquely from the direction making an angle 0 to the normal to the surface. 3.i = 0 and the scalar product becomes A. the area seen is A cos O.B.k = k. (iB" + jBy + kB.) = A"B" + AyBy + A. expressed in the form of a scalar product: effective area = A.(A x B).6.Vectors i..6 This illustrates another useful principle.36) in the direction perpendicular to the plane containing A and B and in the sense that a right-hand screw brings one from A to B through an angle of less than 1[. 3. such as that of force F about o in Fig. the direction of the vector reverses so that B x A = . If we define the line of sight by unit vector r this may be.35) The vector product or cross-product A x B is defined as a vector of magnitude IA x BI given by IA x BI = ABsinO (3.).

which is when 0 = O. The increase in potential energy when the dipole is rotated through an angle 0 as shown in Fig. Each end moves a distance x = (r .7 vector r of any point A on the line of F. V=2 Sa" Eedx = Eer Sa 8 sin 0 dO = . being defined as the force exerted on unit positive charge.E cos 8.11) and its potential energy are the vector and scalar products respectively of the dipole moment and the field strength.40) The condition that a scalar product of zero means that the vectors are . 3.8 is the work done against the forces ± Ee acting at the ends of the dipole. When a dipole is subjected to a field. since moment of force = F x r = Fr sin 0 (3.94 Three or more dimensions.39) The potential energy V will be a minimum when the dipole is aligned with the field.p. 3.38) in the direction perpendicular to both F and r and in the sense of the right-hand screw convention.r cos 0)/2 against the direction of the field so that. From equation (2. An electric dipole consisting of charges ± e separated by distance r has dipole moment p. the torque on the dipole (Section 2. which is a vector on the axis pointing from the negative to the positive end and of magnitude p. dx = !rsinOdO. anticipating our discussion of integration. so that V= -p. (3. is a vector directed from the positive to the negative of the two charged plates that generate the field.68) we have torque = JJE sin 0 = JJ x E.E (3. = reo Electric field strength E. \ \ rsin 9 r Fig. partial differentiation o.

A gravitational. at any point. on the other hand. such as density. to give f: f(x)g(x)dx =0 (3. .p * to be orthogonal.". This is used in wave mechanics when we replacef(x) by the wavefunction . temperature. Under these conditions we may replace the summation in (3. and g (x) by the complex conjugate . b). electric or magnetic field.Vectors 95 E Ee .42) as the condition for the functionsf(x) and g(x) to be orthogonal in the interval (a. or A may be replaced by a continuous functionf(x). If the quantity is scalar.8 orthogonal (perpendicular to each other) may be used to explain the general definition of orthogonal functions.p. We may extend equation (3. If we then allow the number of dimensions to tend to infinity we are supposing A to be made up of an infinite number of components.35) by writing (3.41) where Ai is the component of A in the ith coordinate direction. 3. as well as magnitude. is a vector field because it has direction and sense. pressure or concentration. x: +I----*+e -e Ee Fig.p and . 3. we have a scalar field and the normal methods of calculus apply.p * (Section 1.10) so as to obtain the condition for ..41) by integration over the continuous variable.9.2 Scalar and vector fields A region of three-dimensional space in which a quantity varies from point to point is called a field.

This notation becomes useful in more complicated cases. In a fluid (liquid or gas) a flow of material can occur either in parallel streamlines or in the form of eddies or swirls. so that this notation is useful in three-dimensional problems of these kinds. is constant. = I ox + J oy + Tz· (3..45) The significance of this is that by adding together the gradients in the three coordinate directions as vectors we obtain the magnitude and direction in which the rate of change in q.43) becomes V q. Gradients of temperature.oq. ioq. = gradq. in this case the only. is greatest. grad tP is a vector that is normal to the contour surfaces over which q. where". the work done in a displacement dr is the scalar product F. In three dimensions we may write gra d '" 'I' .43) This may also be written in operator form.44) V = i~+ j~+ k~. potential gradient is mg in the vertical direction. If the field exerts force F on a particle. we have oq. meaning that the maximum. (3. where i is a unit vectQr in that direction.oq./oy = 0 and oq./ox = oq.grad". and z is the vertical axis.dy+Fzdz. dr given by work = F. This applies to mass in a gravitational field if we neglect air resistance. the direction and strength of the electric field at a point due to a surrounding distribution of charges in space can be found as . When we are concerned with vector quantities in three-dimensional space. z) in the x direction. .46) . new possibilities arise. We can express the gradient of a scalar field as a vector by writing. concentration or electric potential will produce a flow of heat./ox for the gradient of q./oz = mg. Thus if tP is the (scalar) potential of mass m in the earth's gravitational field. for example.96 Three or more dimensions. k0q. diffusion of material or an electric current. (3. ox oy oz so that (3. If we consider the work done in moving from one point to another in space.dr = F"dx+F). partial differentiation When we differentiate a scalar quantity with respect to a coordinate direction in space we obtain the gradient in that direction. is the total potential due to the superimposed charges. (x. we define 'nabla' as the operator (3. so that grad cP = kmg. when this is independent of the path taken between the points we say that we have a conservative field. y.

Vectors 97 This work will be independent of path and the field will be conservative if (3. The nabla operator (3.49) and detects sources or sinks in the field. such as of. and in fact measures rotation of the field in that plane. The derivatives in the parallel directions. so that both scalar and vector products of V with F can be written.. or if and oFz of. of.. oy = oz· (3.. defined by d· F _ IV - of.+ ... It can be seen from the above definitions that V. These concepts provide concise notation in problems involving three-dimensional fields. ox + oy + OZ ' (3. _ I_ curI F - (3. This is called the curl of the vector field.) oy oz oz ox ox oy . and it depends on the changes in F in the (y.46) is an exact differential and the cross-differentiation identity applies (Section 3.47) We may therefore express departure from the condition of the work being independent of path as departure from (3..-OF.. z) plane. The significance of curl F is that it measures eddies or swirls in the field: if we move from one point to another along a path that is opposed to the eddies we do more work than if we move along a path that is in the same direction as the eddies.+J .F = divF and V x F = curiF.44) can be used for vector fields by noting that it can be regarded as a vector. this would apply if heat were being supplied at a point in a flowing fluid because thermal expansion would then create increase in velocity and so a positive divergence of the velocity field. (3./ox. of..48) is the x component of the vector curl F. Equation (3.5). give the components of the increase in magnitude of the vector.48) The first bracket in (3.48) contains the derivatives of the components of F in perpendicular directions.. such as ofz/ ox.) _(OF" OFz) k(OFy . defined by (OFz OF. This is a scalar quantity called divergence.50) ..47).

1) This is called an indefinite integral because the addition of any constant c to the integral f (x) will give the same differential f' (x) dx. we have a definite integral and the integration constant cancels out. then the integral off' (x) dx is f(x): If' (x) dx = f(x) + C.CHAPTER 4 Integration 4.2) When the integration is between particular values of x. (4.1 Significance and notation Integration can be defined as the inverse of differentiation or as the limit of a sum. A definite integral can also be defined as the limit of a sum. K. this means that constants can be taken outside the integral sign: I Kf'(x)dx = K If'(X)dX+C.f(a). if we multiply f(x) by a constant. If the differential of a functionf(x) isf' (x) dx. Also. say from x = a to x = b. thenf' (x) dx is multiplied by the same constant. r f' (x) dx = [f(X) I = f(b) . and these are discussed in the following sections.3) This definition of integration as the inverse of differentiation gives the rules for integrating simple functions. (4. which is . (4.

The area of the shaded rectangle is ybX and the sum of the areas of such rectangles becomes the area under the curve as bx -+ o. a succession of such steps along tangents will follow the curve.7) .2.f(a). starting at x = a and ending at x = b.6) In the limit as bx -+ 0. (4. 4.1 Another interpretation is illustrated in Fig. (4. and the sum of the steps between a and b becomes lim L x=b 6s-0 x=a dY d bx = X i b a dy -d dx = X 1/ (b) I (a) dy = f(b) . which is (a.5) b Fig.1. so that the increment 0 y in y along the tangent for increment bx in x is dy by = dx bx. (4. We can follow the curve y = f(x) starting from the point A. = area abBA f = y dx f x a f(x) dx.Significance and notation 99 the reason for the use of the integral sign: f f(x) dx = 6~~0 :t: f(x) bx.4) where 1: denotes the sum produced by adding successive increments bx to x. by making small steps bx in x. (4. 4. 4. f(a». The proof of the equivalence ofthese two definitions requires careful mathematical analysis. but a geometrical interpretation is shown in Fig. A tangent to the curve has slope dy/dx.

4. (4. The change in X from temperature Tl to temperature T2 at constant pressure is X(T2 .2 Thus f(b) = f(a) + or f(b) = f(a) + i dx (4.100 Integration y y = f(x) feb) ------------------ f(o) b a x Fig.1 If a quantity X depends on temperature T and pressure p.P)-H(Tl'P) = r' CpdT.p) = f:'(~)' dT. p). .8) (x) dx. integration will produce a volume. If X is the enthalpy H then (oH/OT)p is the heat capacity Cp and H(T2.p)-X(T1 .9) abdY dx s: f' Example 4. if we know a curve (y = f(x». integration produces a curve. JT 1 These interpretations of integration may be summarized in the form that: if we begin by knowing the properties at a point (the values of Xo and dy/dx). and if we know an area. integration produces an area. we write X = X (T.

+l X" dx = n + 1 + C. f [j(x) + 9 (x)] dx 4.2 Standard methods of integration 4. f eX dx = eX + C. these are discussed in Section 5. but functions that are not single-valued are often met.2 = ff(X) dx + f 9 (x) dx.1.2. dx dy then = eY = x.1 in the first example above. produces the original function. Taking reciprocals dy dx Hence 1 x x = eY . the rule does not apply. when differentiated. n + -1. fdX = x+C. Reciprocal functions When n = .1 Simple functions The integral of a function is that quantity which. Continuity usually applies in physical applications. feaxdx = ~eax+c. .Standard methods of integration 101 It is assumed here that the function to be integrated is well-behaved in the sense that it is continuous and single-valued. This particular case is evaluated by the following device: put y = In x.. 4. Thus f X.2.

102 Integration 4.3 Integration by parts This applies to the integral of a product of two functions. so that f u dv = uv . that the factors of the integrand must be chosen so that differentiation of one part and integration of the other shall yield a simple integral. minus the integral of (derivative of first function times integral of second)'. Example 4. we might wrongly regard the above example as a product of terms taken in the opposite order. as .2 This can be used to illustrate an important principle. as 'the integral of a product of two functions is the first function times the integral of the second. This may not be the case if the parts are incorrectly chosen.f v duo If we denote a functionf(x) by u and a derivative 9 (x) dx by dv this gives ff(X)g(X) dx = f(x) f g(x) dx .2. It is useful also to express this in words.ff' (x) ( f 9 (x) dx )dX. From the differential of a product we have d (uv) = u dv + v duo We can integrate term by term to give f d(uv) = f u dv+ fv duo But f d(uv) = uv.

~ sin 0 cos 0 + c 2 2 sin 2 0 dO 7t = -.Standard methods of integration 103 thus producing only a more complicated integral. Denoting the integral by the symbol I.sin 0 cos 0 + f (1 . If this happens. the rule is to try again with the factors the other way round.sin 0) dO - sin 0 cos 0 + .sin 0 cos 0 + Hence 2 2 f dO -I. we could have 1= f sin 2 0 dO = f cos 0 dO . Example 4.3 Integration by parts can sometimes yield the same integral again. 4 When a definite integral is integrated by parts the application of the limits is straightforward so long as we remember that the limits mean taking the value of the integral at the upper limit minus that at the lower limit.4 The integral of In x is obtained as the product of In x and unity: fIn xdx = f(ln x). Thus r u dv = [ uv I -r v duo Example 4. which can then be evaluated by the solution of an algebraic equation.l dx = x In x - f ~ x dx =xlnx.fdX = x In x -x +c. for example "/2 o ~ .sin 0 cos 0 + c. 1= i so that. . 21 = 0 .

the limits for u become the values of u = x 2 + 4 at x = 0 and x = 2.. du = dx/x.5 Expansion of algebraic functions When 1= f f(X) g(x) dx.2. Example 4. This can. where d In x therefore substitute U = In x. the factor of 2 being immaterial. This is the product of l/x and In x. the expression can be expanded by division of the denominator into the numerator. we substitute U = x2+4. one part being the differential of the other part.4 Integration by substitution This applies when the integral can be regarded as the product or ratio of two parts. ="21 fdU -. be avoided by returning to the original variable before taking the limits. Example 4.= 1ln U + C = 11n(x2 +4)+C. Then I du = 2x dx. We f Udu = 1u 2= 1(In X)2 + C.104 Integration 4. Thus if Example 4. I = = (l/x) dx. the change in variable means that the limits of integration must also be changed to the values of the new variable at those limits. 4.5 I = fX~!X4 Since d (x 2 + 4) = 2x dx. of course. .2.6 1= fnxx dx. When substitution is used to evaluate a definite integral. or u = 4 and u = 8. wheref(x) is of higher degree (contains higher powers of x) than g(x).5 were between limits of 0 and 2 for x.

. = . (a) for factor (x + a) use A x+a (b) for factor (x + a)2 use A B . 3B = 2.ln (x 2 +2)+C. The rules for obtaining partial fractions depend on the factors in the denominator of the expression.6 Integration by partial fractions Algebraic functions with denominators that can be factorized can be expanded by resolution into partial fractions. Thus if x if x = 1.2.2... A = 1..7 I f 105 f f f = x:: 2 dx = (X .-. This is the inverse of the simple addition of fractions.x+2 (x+2)(x-l) x-I we can equally well resolve the right-hand side of this equality into the left... since 1 1 2x+ 1 .+ -:-------:-..3... Since this must be true for any value of x.Standard methods of integration Example 4.X22: 2 ) dx = x dx - X22: 2 dx x2 = 2. We define parameters A and B by writing 2x+1 (x+2)(x-l) A B = x+2 + x-1 = A(x+1)+B(x+2) (x+2)(x-l) ...3A and so and so = .= -:------:. 4. B = 1. or by equating constant terms and the coefficients of x. x+a (x + a)2 . which will be true if A(x+ 1)+ B(x+2) = 2x+ 1. .-----..+ . we can find A and B either by suitable choice of x to eliminate one of the parameters...

Although it is generally possible to differentiate a continuous function. published at intervals by the Chemical Rubber Publishing Company.3 Standard forms of integral and numerical methods. =In ( ~ 4. dX f --+ dx f -dxf -+ x x+2 x-2 1= -2 X2 -4) +c.J 1t e -". For integrals which cannot be evaluated in closed analytical form.2. Hence B and so = 8. geometrical and numerical methods are available at least to obtain an .2 A (x + 2) (x . and A = -2. = C = 1.4x x (x + 2) (x Then ABC =-+--+-2) x x+2 x.'x'd x·=-21X and this is discussed in Section 5. a particularly comprehensive set is included in the Handbook of Chemistry and Physics.8 I=f x 8dx -4x 3 Put 8 8 --:=---= x 3 .106 Integration (c) for factor (ax 2 + bx + c) Ax+B ax 2 +bx+c use Example 4. Extensive tables are available giving the integral of many standard functions. USA. One important standard form is 1 00 o . of x and constant terms gives B=C 0. integration only gives a closed expression in particular cases.2) + Cx (x + 2) Equating coefficients of x A+B+C = 2.2) + Bx (x .

or from right to . both of which mean integrate with respect to x and then integrate the result again with respect to x. 4. The Euler-Maclaurin formula is particularly useful and this is discussed in Section 4. If we write ff(X. This applies also when we want to integrate a quantity which appears as a curve on a recorder trace. An alternative.Multiple integration 107 approximate solution. This is a double integral and the solution will contain two integration constants. it cannot be evaluated without a known relation between x and y so as to be able to substitute for y to obtain an integrand containing x only. keeping y constant. Multiple integration is most useful when we have more than one independent variable. An integral can be interpreted as the area under a curve.4 Multiple integration Just as we may repeat the operation of differentiation to obtain higherorder derivatives. or by cutting out the shape and weighing it. The area can be found by counting squares or vertical strips.Y)dX with two variables x and y in the integrand. analogous to partial differentiation. both of which mean integrate first with respect to x. and details are given in texts on numerical analysis. and then with respect to y. The notation used is to show the required number of integral signs together with the variables of integration. The limits c and d apply to x and the limits a and b apply to y. y) dx dy or rf dy f(x. We do this when defining the multiple integral rf f(x. y) dx. Numerical methods are based on the same principle. keeping x constant. Thus f ff(X) dx dx or f dx f f(x) dx. so that we can draw the curve and measure the area between particular limits to evaluate the definite integral. is to integrate with respect to one variable whilst regarding other variables as constants. so we may repeat integration to evaluate multiple integrals.6. Multiple integrals are evaluated from the inside to the outside in the first form shown.

Leibnitz's theorem Differentiation is the inverse of integration so that the operations of differentiation and integration.9 Jl1 Jo12 x 3 2 ydxdy= 0dy= Jl1 "3 ydy f3 [X3 YJ2 1 3 38 =~ [~J: = 332 . both with respect to the same variable. Notice that the limits of integration must be applied successively. will cancel. This order of integration does not. If. we have a function of more than one independent variable we may differentiate with respect to one variable and integrate with . in fact.5 2 ydydx = y2 Jl I' x 2 JoI' [X2 -2. however. If the integral were written in reverse order we would have JoI' Jo11 x 4. = [i:I 15' Since the variable y occurs in the limits of the integration with respect to x it is essential that the inner integral be evaluated first.0dx = Jo 2 dx Differentiation of integrals. matter so long as the limits of integration for one variable do not contain the other variable.108 Integration left in the second form. Example 4. not simultaneously. d dx f f(x) dx = fdf(X) f ~ dx = df(x) = f(x).

y) dx (4.2a 2 • 4. This will apply so long as the function and the series used both converge sufficiently rapidly in the interval considered. a oy f = fof(X. We wish to obtain a series expansion of the definite integral f~' f(x)dx so that the integral may then be evaluated. Thus if we differentiate the integral 1 2 1 xe. so that the order of differentiation and integration may be reversed.y)dx= a(y) aY a(y) ob oa dx+f(b'Y)a-f(a'Y)a· Y Y (4. these operations are independent of each other.11) Example 4. which is a ib(Y) a y ib(Y) of(x. We first expandf(x) as a . f(x. y) f(x. Since the variables are independent. The Euler-Maclaurin theorem can then be used. we may be able to expand the function as a series and then integrate term by term. and is derived here by the use of Taylor's theorem.10) This will have meaning when the integral is between limits.11 Leibnitz's theorem can be used to obtain an alternative form for an integral by differentiation with respect to a parameter other than the variable of integration.ax dx = o 2a 00 with respect to a we obtain 1 00 - 0 2 x 3 e -ax dx = 1 d 1 da 2a = . and complications arise if the limits of integration contain the variable of differentiation in which case Leibnitz's theorem is used.6 The Euler-Maclaurin theorem When we are concerned with a function for which no simple closed form of integral exists.y) oy dx.The Euler-Maclaurin theorem 109 respect to another.

we put h = (Xl . in terms of the values at the end points of the interval. .12) and then integrate to obtain r l Jxo f(x)dx = (Xl -xo)f(a)+ + ~-~-~-~ 1 2! ° I' (a) (Xl .a) I' (a) + (x 2! a)2 I" (a) + ..in the form f' (x) = I' (a) + (x ..t)2 I" (a) ... we adopt the device of taking the sum of the values of the function at the end points. (4.. the terms containing odd powers of (x .! .h/2..f 4 (a).a) = h/2 and (xo . so that (Xl . By taking the differences between the values of these expressions at the limits. 3 4 (4.a)3 ...!3 (x).13) We now use Taylor series again to expressf(a). .a) cancel since (Xl .(xo .110 Integration Taylor series about the value f(a) at x = a in the form f(x) = f(a) + (x . Xl) the terms containing derivatives of odd order will cancel. we have f(a) = f(x) .f" (a).. . Again using (4. We also write Taylor expansions off' (X). .(xo . and this is used to eliminate f(a) from (4..a) .(x .. and we obtain .! (x a)2 f3 (a) + a)2 fS (a) + .x o)..a)f" (a) + f3 (x) = f3 (a) + (x - (x a)f4 (a) + .. . .a)3 f" ( ) a + . When we substitute x = Xo and x = Xl and add the resulting expreS"sions.. giving h hS to[Xl f(x) dx = hf(a) + 223! I" (a) + 2 5! f4 (a) + . .. ..a).a) = .a)I' (a) (x ...a) = .12). the constant terms and those containing even powers of (x .13). 3! Ifwe now choose a to be the mid point of the interval (xo. but the differences of the values of the derivatives. . .

.... 3) 2 f (rx)=}.15) (4.. the intermediate terms cancel to leave only the values at the ends of the whole range. all at the end points of the integration range x = Xo and x = Xl' The advantage of this form of equation is that we can divide the range of integration into small parts. .. (4. in particular. JX__ 1 and the right hand side of (4.15) and (4.(1 1)) h +214 (1 3! (1 2! -3! 4! -5! -terms in 4(f 3(xd-f 3) (xo) (P(Xd-f 5 (xo»).) ).etc.13) gives f~If(X)dX = ~h (f(Xd+f(Xo») 1 (2!1. Substitution of (4.17) then simplifies. (4.1 ( f (xd-f (Xo) -2h2 3!f 6 (rx).. + rxo .2~:!f4(rx) - 4 3 ..!" (x).18) .17) This formula can be used to evaluate the integral in terms of the values of the functionf(x) and of the derivatives of odd order I' (x).The Euler-Maclaurin theorem 111 will cancel to leave !"(rx) = ~ (I' (xd -1'(Xo») .) +f(xi+ I») = Jo f(x..22 3! )h2 (f ' (xd -I'(xo) ) 1) .) -~ (f(Xo) +f(x. .. which we show symbolically as rxof(x) dx = Jxo r XI + r X2 Jxo JX 1 + . .14). Hence we obtain the Euler-Maclaurin formula in the form (4.16) into (4. (4. The first term can be expressed in terms of the sum of the values of f(x) at the n points.16) and so on. if h = 1 this gives ~ ito (f(X.1 . and in the terms containing the derivatives.

.... and compare the result with example 5. and direct summation of the series gives L e-«2n2= 1+e-O.2S+e-4+ 00 "=0 The sum of this series to eight terms is 2. The two most common applications of this formula are to the numerical evaluation of convergent. so that we have agreement with the Euler-Maclaurin formula.772454. if this range is from zero to infinity and if the values of the derivatives approach zero as x tends to infinity. o "=0 2 r: 1 The integral is shown in example 5... the formula simplifies to roo Jo f(x)dx 1 00 1 1 = i~of(xJ-2f(0)+ 12f'(0)-720f3(0) 1 + 30240 f 5 (4. but intractable.j 'It/2a. 1/1209600.. This expression arises in calculating the contribution from translation in a particular direction to the partition function in statistical .112 Integration and the numerical coefficients of subsequent terms can be shown to be 1/30240... Hence by (4.5 we obtain fooo e-«2x2dx = ~: = 1. from which we obtainf(O) = l.1.S+e-l+e-2. 00 ~ e _«'n' --.22).5. in statistical mechanics).272454.21) contains the derivatives at both ends of the range of integration. Example 4.J(oo) = O. Putting IX = 0..g. In order to obtain rapid convergence we must have a small enough value of IX.and all of the derivatives of odd order are zero at both limits. and to the replacement of the sum of a convergent but intractable series by an integral (e. We putf(x) = e-«2 x2.22) (0) . .1 to have the value .12 Use the Euler-Maclaurin theorem to evaluate e-«' x' dx when a = 0. 1 e _«'X2 d x = £. The formula as written in (4. integrals which do not possess a solution in a closed form (numerical quadrature).

f5(0) = 120 «()/T)2 . the correction of .13 Following the previous example. We then require to evaluate the sum L e -n'h'j8ma 2kT = L e -n2h2j8ma 2kT -1. 00 00 n=1 n=O 1 The Euler-Maclaurin formula gives ~ e -n2h2 j8ma 2kT -_ . n=O 00 0 e -n2h2j8ma2kT +-1 2 _ (2nmkT)1/2 ~ h2 a+ 2' so that ~ ... f3(0) = -12()/T+12«()/T)2.The Euler-Maclaurin theorem 113 mechanics. This is also an intractable series. The integral may be evaluated by making the substitution u = 1(J + 1W/T to give 1 00 o (21 + l)e. 1'(0) = 2 -()/T..J (J+l)8/TdJ = T 0' .. and all relevant terms become zero as 1 -+ 00. whereas the corresponding integral can be evaluated in closed form. n=1 e _n 2h2j8ma 2kT = (2nmkT)1/2 _~ h2 a 2' Since the value of the integral is usually a very large number. .J (J+l)8/T the terms up to «()/T)2 are 1(0) = 1. When 1(1) = (21 + 1)e. .t... Example 4. the relation which corresponds to the rotational contribution to the partition function is L 00 (21 + 1)e -J(J+l)8/~ J=O (where () = h 2 /8n 2 Ik).t..1/2 given by the Euler-Maclaurin formula may usually be neglected and we say that the sum may be replaced by the integral for the purposes of evaluation....

..22) then gives. This applies except for small molecules at low temperatures. 18 1(8)2 4(())3 ) Y +315 Y + . 1+ 3Y +15 In this case. to terms in (8jT)2.114 Integration The Euler-Maclaurin formula (4. f: (2J + l)e- J=O J (J+1)6/T =!.. ..~(2 -~) 8 2 12 T +_1_ (-12 ~+ 12 (~)2) 720 T T - 30~40 ( 120 (~ T( =fi y) + . the sum may be replaced by the integral when T.}> ()..+! . for which the series expansion can be used..

the x-axis and the ordinates x = a and x = b. the integral ff(X)dX gives the area between the curve y = f(x). If. f(a) wheref(a) andf(b) are the values of y =f(x) at x = a and at x = b respectively.b f if a = = a f(x)dx+ (b) xdy ib Xdxdy dx f(a) a [j(x) + x!'(x)] dx .1 Plane area As shown in Section 4. 4. as can be seen by reference to Fig. instead.CHAPTER 5 Applications of integration 5. we required the area between the curve and the y-axis. we sum horizontal strips of length x and thickness dy to obtain i f (b) xdy.1.2. The sum of these two areas must be the difference between the areas of the rectangles a xf(a) and b xf(b). This can be proved analytically as follows: i b f(x)dx+ f.

116 _Applications of integration = r:: d(xf(x» X-b [ = xf(x) ] x=a = bf(b)-af(a).1. this is because area has the signs shown in the quadrants of Fig. y x Fig..)(r2 -x 2)dx the result would be zero. 5. This condition will not be met if the equation contains quadratic or higher-degree terms in y.1 If we simply integrated f: ydx = f: ± . value. The area between two curves can be found as the difference between the area under one curve and the area under the other. as the sum of equal positive and negative areas. unique. centred at the origin. as shown in Fig. 5. These relations give the area between a curve and one of the axes. the function y = f(x) shall have a single. this means that for any given value of x. 5. Thus the equation x 2 + y2 = r2 represents a circle ofradius r.1. The various manipulations used here should be carefully noted. The conditions necessary for this interpretation of integration to be valid are that the function should be continuous and single-vaiued. .

cos 2 0 = sin 2 0: r area of quadrant = J"/2 = rsinO(-rsinOdO) _r2 r J"/2 sin 2 0dO = 1[r 2 /4 by using the result of Example 4. The area of a circle drawn entirely within the positive quadrant is then 1[r2. Alternatively.r sin 0 dO.1. being that . dx = .3) is that we first obtain the ordinate y by the inner integral and then sum the areas of vertical strips by the outer integral. (5. so as to take advantage of the relation 1 .2) If we now combine equations (5.2 Plane elements of area In Section 4. In the same section. so that y = s: dy. This integral may be evaluated by using the substitution x = r cos 0.3.3) The geometrical interpretation of equation (5. the use of integration to obtain the area between a curve and the x-axis as area = [YdX (5. the area in the first positive quadrant of our circle being obtained by choosing the positive root: area of quadrant = IJ (r2 - x 2) dx.1) is justified as the sum of areas of vertical strips of height Y and width dx. integration is interpreted as giving an increment in Y by moving along the curve y = f(x). we must restrict the range of integration. 5. rf: (5.1) and (5.Plane elements of area 117 When a function is not single-valued.2) we obtain the area as a double integral area = dy dx. we can interpret the double integral by regarding dy dx as a differential element of area.

which follows from the definition of angle in polar coordinates as angle = (arc length}jradius. This requires the element of arc length ds = r de. shown in Fig. 5. Then dA = dydx and area = fdA = ff dYdX .118 Applications of integration of a rectangle of sides dy and dx. Given that the circumference of a circle is 2nr the area of an annular strip becomes. dA=271"rdr Fig. dA = 2nrdr.3. In the same way. 5. we can obtain the area of a circle. using plane polar coordinates. is obtained by increasing r by dr and () by d(}. as dr --+ 0. by summing the areas of annular strips of radius rand thickness dr as shown in Fig.2 The area of a circle of radius a then becomes fdA = f: 2nrdr = na 2• The differential element of area in plane polar coordinates.2. Then dA = rd(}dr and area of a circle = Jor Jor a 21t r d() dr = 2n f: r dr = na 2 • . 5.

As a definite integral (i. This is because the polar equation of a circle centred at the origin is simply r = a.3 By using polar coordinates we have avoided the difficulty met in Section 5. (5. which is a single-valued function. Example 5.Plane elements of area 119 dA=rdOdr Fig. (5.5) We now multiply the two integrals (5. so we may rewrite the integral in terms of variable y as 1 = 1 000 e -IX 2 Y 2 dy. the element of area dx dy becoming r dO dr. We now transform to polar coordinates. y) plane.5) together. this may be written as This double integral is over the positive quadrant of the (x.4) This may be evaluated by the following device. between limits).e. 5. The limits for the same positive quadrant .1 where the area of only a quadrant of the circle could be found.1 An integral that arises frequently in physical applications has the form 1= 1 000 e -IX 2 x 2 dx.4) and (5. using x 2 + y2 = r2. Since they have common limits. the variable x is immaterial.

but when x = L'Hopital's rule (Section 2.120 Applications of integration become 0 to n/2 in () and 0 to [2 = foOO f"/2 e 00 _". the differential element of volume is that of . The first term in [1 is zero when x = 0. 2 This has the effect of introducing a factor r into the integral.2 in r.6) Example 5. such as Integrating by parts.2 r dr.2. gives where [=1 .3 Elements of volume. so that r d() dr 0 = foo -n e 0 _".fOO 2 20: 2 0 e-'" 2x 2dx . polar coordinates in three dimensions In rectangular coordinates.10) 00 we need to use I·1m 1 2 2 =0' · x IIm---=== 00 e'" x x _ 00 20: 2 xe'" x X _ so that 5.In dx = --.2.2 The previous example can be used in the integration of similar forms. 20: (5. choosing x as the first part. which can now be evaluated by the substitution Hence so that [= f OO o e -'" 22 x .

4 is therefore r dO dr dz. 5.5. Fig.4) or two angles and one length (spherical polar coordinates. . (5. Fig. The volume element in Fig.5 Spherical polar coordinates x = r sin 0 cos cfJ y = r sin 0 sin cfJ z = r cos 0 d V = r2 sin 0 dO dcfJ dr The corresponding elements of volume are constructed by increasing r by dr. y) plane and negative below.z) z y x Fig. z z P(r.O. 5. The volume element in Fig.5 is therefore r sin edcfJ r dO dr = r2 sin 0 de dcfJ dr. 5. 5. 5. the radius to be considered is r sin 0.Elements of volume. so that volume = f dV = fffdXdYdZ. 5.7) Polar coordinates in three dimensions may be defined in terms of either one angle and two lengths (cylindrical polar coordinates. 5. producing an arc of length r sin 0 dcfJ.5). dy and dz. polar coordinates in three dimensions 121 a cube of sides dx. When 0 is increased by dO the point Pat radius vector r moves in an arc of length rdO. In each case volume will be positive above the (x. 0 by dO and cfJ by dcfJ. When cfJ is increased by dcfJ in Fig.4 Cylindrical polar coordinates x = r cosO y = r sin 0 Z=Z dV = rdOdrdz Fig.

Change in z then depends only on the initial and final states.7).3 The volume of a cylinder of radius a and height h with base centred at the origin is Example 5.5 that.8) . A physical application of this is when z is a thermodynamic function of the state defined by x and y.4 The volume V of a sphere of radius a centred at the origin must be found from that of the positive hemisphere. We may write the change in z as the integral of dz between the endpoints. There is a general analytical method for obtaining the relations for changing the variables in integration when changing the coordinate system. 5. being independent of path. The geometrical interpretation of this is that we can then plot a three-dimensional graph of z against x and yand obtain a unique surface. In three dimensions we obtain f dZ = f(X dx + Ydy). if z is a unique function of x and y.4 Line integrals It is shown in Section 3. In two dimensions.122 Applications of integration Example 5. (5. with one independent variable. the expression dz = X dx + Y dy satisfies the cross-differentiation test and is an exact differential. this gives equation (4. V = 4nr 3 /3. so that the limits of integration are 0 to n/2 for e and 0 to 2n for tjJ: 1V = = f: f: xf:1 f: 2nr 2 dr = sin e de dtjJ dr = f: f:x r2 dtjJ dr j nr3. This makes use of lacobians and is described in texts on calculus.

we denote distance travelled by s. which clearly depends on the path taken between the endpoints. The general characteristic of a line integral is that it contains two variables and so. which means along a particular curve. S.8) contains both the variables x and y.8). Then for increments dx in x and dy in y we can apply Pythagoras's theorem to obtain ds 2 = dx 2 + dy2. y) plane and so define a curve.) (dx 2 + dy2) = . The line integral is then evaluated along that curve in the (x. . The length of curve between points (Xl' s where = i2 yd and (X2' Y2) is then ds ds = . if X dx + Y dy is an exact differential it can be written as single derivative dz and so integrated. To evaluate an integral containing two variables x and y we usually need to know an equation connecting them so as to be able to express the integral in terms of a single variable. y) plane. This corresponds to choosing a particular path. and this is discussed in the next section. However. and it need not contain both terms on the righthand side of (5.Curve length by integration 123 The right-hand side of (5. This also means that the line integral of an exact differential over a closed path is zero because the endpoints then coincide. if not an exact differential. can be evaluated only for a known equation connecting the variables. An example is when the variable of integration is the distance along a curve. Ifwe suppose that we move from one value of z to another across the surface in three dimensions along a particular path we can project this path into the (x..)[1 + (dy /dx)2] dx. so can be regarded as an integral in the (x. y) plane. A line integral can be defined and used even when it does not contain an exact differential. Thus we may use the condition that the line integral is independent of path as an alternative definition or test for a thermodynamic function of state.S Curve length by integration When a point moves along a curve. Between given endpoints the left-hand side gives the change in z and the right-hand side is called a line integral.. The line integral is then independent of path. the change in z depending only on the endpoints.

6 Applications of multiple integration Multiple integration is discussed from the analytical point of view in Section 4. Hence the circumference of a complete circle is s = 2nr. This is illustrated in the following.4 and in geometrical terms in Sections 5.()d()=r t I2d ()=i r . The equation of the circle is x 2 + y2 2xdx+2ydy dy x y dx = r2 = 0 x The length of arc is therefore Jo S=f"'(1+ 2 r X2 -x2 )dX=f"'J(: Jo r-x 2)dx. and finite solutions are then obtained by multiple integration. x.5 The circumference of a circle of radius r centred at the origin is found from that of the positive quadrant.dx= -rsin()d().1 The pressure of a perfect gas This is given by the kinetic theory of gases in the form of the well known equation pV=tNmu 2 • . J[1+(dy/dx)2]dx.6. 5. in terms of physical variables of various kinds.2 and 5. Example 5. 5. s= £J[1~(x/r)2]dX= [12 -~~i. for which we need to know the equation connecting x and y. Putx=rcos().3. 1 This integral cannot be evaluated until we know dy/dx in terms of x only.124 Applications of integration so that s= 1 2 ds= iX. It is generally useful in physical applications. This is because differentials can be combined to great advantage in many applications.

the fraction of molecules with speed between u and u + du depends upon u and is written as v( u) duo We multiply together the above three factors to obtain the number of molecules in a volume element d V which are moving towards an element of area du on a sphere of radius r with a speed between u and u+du as N du (5. In a perfect gas. or (N / V)d V. and the total solid angle for a sphere is 47t.6 we consider molecules moving from an element ofvolume at P in the gas to collide with an element of area dA on the surface of a wall.1). and force is rate of change of momentum. with angle cjJ in the plane perpendicular to the paper (compare Fig. the element of area dA can be resolved into an element perpendicular to the radius vector r as du = dA sin () sin cjJ. Any particular molecule will be in a random position and moving in a random direction with a particular velocity. (5.5). 5.9) dN = V dV 47tr2 v(u)du. the density is N / V and the number of molecules in an element of volume d V is the density times that volume. this being defined as area/(radius)2 (Section 1. In Fig. Using spherical polar coordinates. Random position and random direction lead to simple calculation. molecules of mass m are assumed to move freely and independently of each other. The speed of a particular molecule is a more complicated problem because some molecules move quickly. For N molecules in volume V.10) . The fraction of molecules that will be moving towards an element of surface du on a sphere of radius r is du/47tr2. Random direction in space can be dealt with using solid angle. An alternative derivation is presented here in a form which is extended in the next section to include the properties of real fluids. and there is a most probable speed at a given temperature. we can calculate the pressure exerted by collision of gas molecules on unit area of wall from the rate of change of the component of momentum perpendicular to the wall.Applications of multiple integration 125 It is derived in standard texts in ways that avoid mathematical complexities such as multiple integration. but this is at the cost of physical clarity. Since pressure is force per unit area. some slowly. 5. We have a continuous distribution of velocity amongst the molecules and this is expressed as a velocity distribution v(u).

and the corresponding change in momentum is 2mu sin 0 sin l/>. When a molecule from P collides elastically with the wall.9). The component of velocity u perpendicular to the wall is u sin 0 sin l/>.6 The volume element at P is d V = r2 sin 0 dO dl/> dr.¢» Fig. The total pressure is now found by integration.lJ. 5. The factor that remains to be considered is the number of collisions in unit time.11) The pressure of the gas is the force exerted per unit area on the wall by elastic collisions of the molecules on the wall. is obtained by mUltiplying the number of collisions.126 Applications of integration P(r.10) and (5.11). by the change in momentum per collision: dp N dA sin 0 sin l/> r2 sin 0 dO dl/> dr 4 2 v(u)du 2mu sin 0 sin l/> V nr =- = N 2m sin30 sin 2 l/> uv(u) dO dl/> dr dA duo V n The differential variables show this to be a five-fold integral. from equations (5. it is the rate of change of momentum perpendicular to the wall on unit area. and we . its component of momentum perpendicular to the wall becomes reversed in sign. a molecule from P will reach the wall in unit time if the distance r is less than or equal to u. dp. (5. The differential expression for the pressure. (5.

6. sin 2 ¢ d¢ = n/4. except at high temperature. which is the ratio of the number of pairs at separation r to the random number.. The number of pairs of molecules at separation r depends upon this potential and we denote by g(r) the pair distribution function. Using the notation of Section 4A we write p = 4 N 2m V n (00 uv(u) du (./2 Jo dr = u. The remaining integral is the average value of the square of the velocity. = 1N = --m 3 V 1 00 0 u2 v(u)du. Jo Jo Jo Jo Jo It is important here to evaluate the integrals from right to left because the variable u appears in the limits of the integration with respect to r. the attractive forces are dominant. so that p 1. We denote by w(r) the potential energy of interaction of a pair of molecules separated by a distance r in the bulk fluid. the attractive force is then -dw(r)/dr. ("/2 Jo sin 3 () d() = 2/3. so that the pressure exerted on the walls is less than it would be for a perfect gas. which we write as u 2 . f: f: We have dA (. The velocity u is integrated from 0 to 00. We obtain unit area by integrating A from 0 to 1. and unit time by integrating r from 0 to u.2 Interactions in a rea/fluid In a real fluid the molecules exert attractive and repulsive forces on each other and.Applications of multiple integration 127 now consider the ranges of integration. . so that pV=jNmu 2 • 5. Integration of e and ¢ from 0 to n/2 ensures that only single values of sin () and of sin ¢ are used and we introduce a factor of 4 so that point P moves through a complete hemisphere based on the wall./2 sin3 ed() ("/2 sin2 ¢ d¢ (u dr (1 dA..

5. This gives an area visible at radius r of r 2dA sin e sin 4>1(oc 2r2). e. perpendicular to the plane.7 We use point P and unit area at 0 to define an element of volume on the right-hand side within which molecules would be visible from P. The volume element d V R in the range r to r + dr on the right-hand side is then dVR = dA sine sin 4> dr/oc 2 • (5. Then d V L= (ocr)2 sin ede d4> d(ocr) = oc 2r3 sin ede d4> doc. An element of area dA at 0 gives a component dA sin e sin 4> perpendicular to the radius vector r. Since each molecule in d V Linteracts with all .12) Point P is defined by spherical polar coordinates (-ocr. the random number of molecules in d V R is (N 1V) d V R and in d V Lis (N 1V) d V L. exerted by the molecules on one side on the molecules on the other. z p Fig. 5. We define fraction oc of the radius vector as being on the left-hand side of the plane. so that the solid angle subtended at P becomes dA sin () sin 4>1(ocr)2. Molecules on the left-hand side interact with those on the right-hand side. by restricting the interaction to a window of unit area we can calculate the total force per unit area.128 Applications of integration Fig.7 shows a plane Oz drawn within bulk fluid. 4» and we construct a volume element d VL at P by using oc as the variable at constant r. (5.13) The number of pairs of molecules at this separation r is g(r) times the random number.

7. (N)2. 3 () . and this is extended to simultaneous equations in Section 2.(dw(r)/dr) sin () sin ¢. which may also be written as Pint = dw(r) -61 (N)2 V Jo('" g(r)r~4nr2 dr. dPinl' is then dPint = dw(r). The differential contribution to the pressure from the intermolecular forces. with a compensating factor of 4 to cover a complete hemisphere on each side of the plane. dr . 0 to 1 in (X to include all pairs for a given direction of the radius vector r. 2 3 dw(r) = .dw(r)/dr perpendicular to the plane is . we can show that the shortest distance between two points in a plane is a straight line. This is an important equation in the theory of liquids anO gases. 0 to n/2 in () and ¢. and that between two points on a sphere is by the great circle route. 5.7 The calculus of variations The determination of stationary (maximum and minimum) values by differentiation ofa known function is described in Section 2. .The calculus of variations 129 molecules in d V R' the number of pairs is (~y dVLdVRg(r).8.~ sm () sm ¢ V (X2 r3 sm () d() d¢ d(X dA sm () sm ¢ (X2 g(r) . Another class of problems involving stationary conditions is when the function itself is unknown. . and 0 to co in r: Pint = N)2 -4 ( V = l<Xl 0 dw(r) r3g(r)~dr N)2 -1n ( V Jo("/2 sin 3 ()d() Jo("/2 sin2¢d¢ fl 0 d(X JoedA.( N)2 V sm sm ¢r g(r)~drd()d¢d(XdA. We again have a five-fold integral and the limits are 0 to 1 in A for unit area. . dw(r) Jo(<Xl r3g(r)~dr. by finding the function which . For example. The component of the attractive force .

We have seen in Section 5.15) the device of regardingf as dependent not just on the two variables x and y but also on the third superfluous variable y' = dy/dx.15) 1= ff(X. For this we need to differentiate I with respect to some parameter that produces variation in the function y and so in the integral. JX1 (5. which we shall later do.14) Here.8. 5.4. We have introduced in (5. and can be written in general terms as (5. as in other problems of this kind. (5. we may still use the expression for the total differential in the presence of such a superfluous variable. y') dx 08 = O. y')dx where.fis known but y is not. as in Fig. When 8 = 0 we obtain the simple form y = 4>(x). Change in 8 produces a change in y at constant x for given 4>(x) and '1(x). . we have a known integral which depends upon an unknown function y = 4> (x). again.17) Now change in 8 at constant x. The integrand in (5. 4>(x) and '1(x) produces changes in y and . As shown in Section 3. y.6 that curve length in two dimensions is given by the integral s= rx. y.16) where '1(x) is yet another function defined as arbitrary but for the condition that '1(x) = 0 at the endpoints x = a and x = b.15) is then a function of 8 and we can use the stationary condition 01 08 = o. and 8 is an adjustable parameter. This is achieved by the device of writing the unknown function as the sum of two terms y = 4>(x) + 8'1(X). (5.130 Applications of integration describes the path of minimum length.J [1 + (dy/dx)2] dx. If we now vary 8 for given 4>(x) and '1(x) we produce a variation in the curve of y against x between fixed endpoints. The problem now takes the form of finding the function y = 4>(x) which makes the integral I stationary. The integrand is a known function of the unknown function y. As shown in the previous section we may interchange the order of differentiation and integration so that 01 08 = i b a of (x.

-"1= cp(x}+ £l'}(x) =cp(x) .19) By definition.. oy'OJ) dx Ja = O. so that (5. 5. x Fig.= .. y aY Ya aXy (5.18) by parts gives J. .20) contains '1(x) which may be any function.20) Now equation (5.. . y . so that the only way in which the integral must be zero is when the bracketed . .. -'" "'.18) becomes (5. (5.b 81'1 oj .b d oj J.The calculus of variations 131 y '.oj a'dx = dx = 81'1 da''1 dx ..= ..16) dy de dy' d dy d dy d'1 .8 in y' so that oj oe ojdy oy de oj dy' oy' de -=--+-and from (5. '1(x) = 0 at x = a and at x = b so that the integrated part is zero and (5..18) Integration of the second term in (5.•b'1 .17) becomes (b( '1 oyoj + '1..= '1 de dedx dxde dx ' - = '1(x). = . [OJ Jb J.

one set for y and one set for z.21) in each of the variables x and y.21) which is called Euler's equation.11 was based upon Newton's laws of motion. y. on which the theory can be built. y') is defined by (5. however. This is not the only basis. rather than . Whenf(x. y. y'. This is the equation of a curve of constant slope.132 Applications of integration part is zero. Hence af _~ af = 0 ay dxay' .6 Show that a straight line is the shortest distance between two points in a plane. which is a straight line. y. In place of (5. a useful alternative formulation being due to Hamilton and Lagrange.20) to be zero is then that we have an Euler equation of the form of (5.21) gives d y' dx y'(1 + y'2) = 0. The application of this technique to three (or more) variables requires only a straightforward extension of the argument. The condition for the integral in (5. Example 5. This regards kinetic and potential energy.8 Generalized dynamics The discussion of dynamics in Section 2. (5. ay . z.20) likewise.17) becomes the sum of two sets of terms and (5. or y' is independent of x. Equation (5. y') = y' (1 + y'2) and since we are regarding y and y' as separate variables af = o. z'). 5. and substitution into (5.15) an additional variable z will introduce also z' as f = f(x.14) we have f (x. We define two arbitrary functions in place of the one function '1 (x) and two parameters in place of the single parameter e.

which is the change in the product of momentum and length. For motion of a particle of mass m in the x direction. not direction as well. An early formulation of quantum theory was in this form. which means that we can easily change coordinate systems. the Wilson and Sommerfeld rule being that p dx = nh over a closed cycle of the motion. so that 1= it2 Ldt = it2 (T.22) t. This difference is called the Lagrangian function L and the postulate is called Hamilton's principle.Generalized dynamics 133 force. or twice the work done by the applied force.V) is twice the change in potential energy. Denoting momentum by p we have that work = fFdX and so that f L dt = 2 f f F dx dt = 2 J p = f Fdt . and we can apply the methods of the previous section. This in turn. An abstract formulation of this kind is of course more difficult to visualize than Newton's laws because it is not so clearly related to everyday experience. We can use the calculus of variations from the previous section to show that the two formulations of the principles of mechanics are the same. ff F dt dx = 2 f p dx. and p dx is action. It was shown in ExaPlple 1. is to be stationary. which arises because it is expressed in terms of the kinetic and potential energies. The difference in (T .2 that such a relation is dimensionally correct. (5. means that we are not confined to any particular system of coordinates because we need only to express position in space. The Lagrangian function L then depends on x J . so that we write that potential energy as V(x). Its great advantage is generality.V)dt t. such as gravity. produces a force F on a particle in the x direction. it moves so that V decreases and T increases by the same amount. both of which are scalar quantities. Hamilton's principle is related to the little used concept of action. If a field. integrated with respect to time. is stationary. as fundamental and postulates that a system will move in such a way that the difference between the kinetic energy T and the potential energy V. the kinetic energy T = ! mi 2 • When force F acts on the particle in the same x direction the potential energy will change.

11. By using the more powerful technique of equation (5. where it was assumed that masses m1 and m 2 were rigidly held at a separation r from each other. so that d aL .67) becomes in terms of the angular velocity {j = OJ rotational KE = !m*r 2 82 • (5. We use the polar coordinates shown in Fig.22) is the same as Newton's law. --=mx dt ax ' and we obtain F = mx. which is Newton's law.25) Since the distance r between the particles is now a variable.24) we can consider not only that case but the more general one.26) . by each other with a force that is a function of the separation r. We define the reduced mass m* of the two particles when they are at separation r as in (2. mx. and aL ax = . this will apply either to a chemically bonded molecule or to the general case of the interaction between two particles in three-dimensional space. The rotation of a diatomic molecule was considered on the basis of Newton's laws in Section 2. ax (5. We suppose the two particles to be moving in space. x. or repelled.23) we have aL _~ aL = 0 ax dt ax ' aL ax -= av. x) = !mx 2 . where the masses are attracted. Thus Hamilton's principle (5. we also have radial kinetic energy given by (5.134 Applications of integration (through V) and on x (through n and we have L(t.23) Euler's equation (5.24) --=F ' which is the force acting on the particle. so that both distance r and angle () change with time.66) by and the rotational kinetic energy from (2. 5.9 to define the positions of masses m1 and m2 at distances r 1 and r2 from the centre of mass 0 of the pair of particles. The potential energy may be written in the general form V(r).64) and (2. (5.21) then becomes and from (5.V(x). and so relative to each other.

24) both in terms of the angular coordinate oL _~~ = oe dt oe -~(m*r20) = o. When the attractive and .= m*re ----(m*r) = o.9 But.26) becomes Hence radial KE = -!m*r2 • (5. this is shown to be constant. In equation (5..27) (5. and we can write Euler's equation (5.2 oV d or dt . rand e. 8. and m*f is the radial momentum.Generalized dynamics 135 Fig. from (2..28) We now have L in terms of the coordinates r. m*r8 2 is the centrifugal force in the direction of r opposite to the attractive force between the particles -oV/or.30).29).65) and and (5..30) In equation (5.29) dt and in terms of the radial coordinate oL or d oL dt or . m*r 20 = 10 is the angular momentum. (5. 5. (5.

ellipse or circle for decreasing values of the relative velocity of approach. or that the work done in any displacement is independent of the velocity (no 'friction').29) and (5.V) aqi aT aqi if V is independent of qi' (5.9. The formal definition is (5.32) Again for the simple case of motion in the x direction. it may be shown that when particles having the same charge approach each other the path of one. this gives pi=mi. " and if qi were an angular coordinate 0 we would have T = tIlP.7. In these generalized terms we can define also the Hamiltonian function H which is often used in quantum mechanics. .9. This approach to dynamics is called 'generalized' because the equations are the same in any coordinate system. as illustrated in Fig. relative to the other as focus. or 'conjugate' 'generalized momentum' by Pi = aL aqi = a(T. corresponding to gravitational attraction of planetary bodies or the electrostatic interaction between charged particles.31) The significance of this may be seen from the simple case when qi is an x cartesian coordinate. In the simplest case of circular motion the radial momentum is zero and r is constant.136 Applications of integration centrifugal forces balance. When we denote a generalized position coordinate by qi we can also define a corresponding. the equations represent conic sections (Section 1. and again aT d 1 '2 ' aqi = (2 10 ) = 10 = Pi' de The condition in (5. By considering the polar equations of conics.31) that V is independent of qi means that we have a conservative system (Section 3. The momentum is then Pi = mx and the kinetic energy T is -! mi 2.2). and for unlike charges we may obtain also a parabola. In the important case of an inverse square law attraction. or otherwise.4). so that aT = aqi ~(-21mi2) = di mi = p. is a hyperbola. d(m*r)jdl = 0 and the radial momentum is constant.30) depends on the form of V(r). The form of the general motion represented by (5. 5.

31) so that the first and last terms cancel to give dH ~L dqi' = qidpi - uqi Hence oH uPi . In turn L depends on qi (for T) and on qi (for V). . .33) gives the rate of change of generalized momentum and so may be regarded as a generalized force.33) ~=qi' and oH Oqi oL Oqi .33) are simpler than the equation in terms of the Lagrangian function L in that they have a degree of symmetry.34) Equations (5. (5. Thus so that for the ith coordinate " dPi-:.Generalized dynamics 137 and so that H = mx 2 -(!mx2 = = V) !mx 2 + V T + V = total energy. however. in most problems.32) that H depends on Pi> qi' and on L. the Lagrangian function is found first and then the Hamiltonian. .= -po (5..oLd qi-~ aLd' d H = Pi d qi+qi qi' uqi uqi But the generalized momentum Pi is defined by equation (5. This means that the Hamiltonian function has the simple meaning of being the total energy so long as the system is conservative.31) gives We can then write the equation in terms of H by examining the parameters on which H depends. The expression (5..32) and (5. We can express Lagrange's equation in terms of coordinate qi as which with (5. we have in equation (5.= .

CHAPTER 6 Differential equations 6. as dy = 2xdx. heat capacity). so the solution of a differential equation requires the separation of the variables into terms that can be individually integrated. A trivial example of a differential equation is dy dx = 2x. momentum (force. distance or temperature of such quantities as energy (thermal conductivity. These theories give rise to equations containing derivatives. the purpose of which is to eliminate the differentials. viscosity) or amount of substance (fluid flow. reaction kinetics. whereas if more than two variables are relevant. and comparison with experiment requires the solution of such equations by integration. This may be written with the variables on separate sides of the equation. . partial derivatives and partial differential equations will be obtained. The methods of integration discussed in the previous chapter can be used only when the integrand can be expressed in terms of a single variable. In physical theory we may have simple relations for the rates of change with time. heat of reaction). Problems involving only two variables will give rise to total derivatives and ordinary differential equations.1 Significance and notation Differential equations arise frequently in mathematical and physical theory because the calculus limit so often gives simple relations.

. Then by putting x = Xo and t = 0. The value of the integration constant can be determined by the use of boundary conditions. if an equilibrium system is disturbed by a small amount Ax it will return to equilibrium at a rate which to a first approximation may be assumed to be proportional to the displacement. An alternative interpretation of this form of relation is in terms of a time constant. A simple physical example is when the rate of change of a quantity x with respect to time t is proportional to the quantity itself. In and In Xo = (:J C.t /<. or relaxation time. The integration constant C is important in differential equations. Not all differential equations possess an analytical solution. linearity and homogeneity. so that after time T the displacement will have decreased to lie = 0.f kdt . Then giving Ax = Axoe.Significance and notation 139 We may then integrate each side of the equation. This classification is based upon the concepts of order.37 of its initial value. This is called the time constant. but standard methods are available for particular classes of equation. In x = -kt+C. to obtain fdY = f 2XdX Y = X 2 +c. such as knowing the initial value Xo of x at t = o. degree. of the system and is seen to be the reciprocal of the first-order rate constant k in reaction kinetics. = -kt. so that dx -= dt giving -kx ' fd: = .

6.140 Differential equations The order of a differential equation is that of the highest derivative occurring in it. first degree 6. giving dy+4xdx = dx. which may be written as dy = (1 . then higher-order linear equations. homogeneous.2). A first-order equation is homogeneous if each term contains the variables to the same power (Section 6. that of a second-order equation contains two constants. and their range of application. Example 6. and so on. is very wide and provides scope for considerable ingenuity in the methods of solution.1) is a second-order linear equation.1 dy dx +4x = 1. In the following sections we consider first-order first-degree equations. with no cross-terms. on the other hand.2. the solution is merely integration.4x)dx. The term 'homogeneous' is used in various ways.2 Equations of first order. is homogeneous if all terms contain the dependent variable. The subject of differential equations. and so on. A linear equation. This chapter does no more than introduce the techniques and some typical applications. exact or linear. dy/ dx being first order. so that f(x) = 0 in (6. classed as separable variables. .1 Separable variables If a differential equation can be written with all terms containing one variable on the left-hand side and all terms containing the other variable on the right-hand side. d 2 Y/ dx 2 being second order.1). Multiply both sides by dx. The solution of a first-order differential equation contains one arbitrary constant. so that dy d 2y ao(x) dx 2 +adx) dx +a2(x)y =f(x) (6. A linear equation is one containing only the first powers of a variable and its derivatives. The degree is the power of this derivative.2.

each of which corresponds to a particular value of the integration constant C. The variables are not separable as such due to the terms xdy and ydx.2). Example 6. Thus if dy dx 2x-y x+1 the above procedure gives (x + I)dy = (2x . but the combination xdy + ydx is actually the differential of the product xy. giving 2 = -I +C. C=3 so that the solution to the differential equation becomes y = x(I-2x)+3.Equations of first order.ydx. then we substitute x = 1 and y = 2 into the equation so as to eliminate the integration constant. so that x 2 +C I+x y=--.y)dx. first degree 141 Integrate both sides fdY = f(1-4X)dX Y = X_2X2+C Y = x(1-2x)+C. so that dy + d(xy) = 2xdx which may be integrated term by term to give f dY + fd(XY) = f 2XdX . y+xy = x 2 +C.2 An equation may sometimes be separated into integrable terms even when the variables are not completely separated. The geometrical interpretation of this solution is a family of curves. . Suppose we require the solution passing through the point (1. xdy + dy = 2xdx .

kt ). x and xdy/dx being both first degree. -In(a -x) = kt + C. together with the concentrations of the two species at time t.2.tion is described as 'first order' when the rate of production of product is proportional to the first power of the reactant concentration.3 A chemical rea{. Example 6. For the reaction A ~ B (a-x) x the rate coefficient k is shown. This is a separable variables equation. If the variables are x and y we can eliminate y by the substitution y = zx. so that dx -(a-x . and the integration constant C is found from the boundary condition that x = 0 at t = 0. and so on. the ratio dy/dx being zero degree. 6.4 Multiply by dx to give xdy+2ydx We now substitute y = ZX.142 Differential equations Example 6. x 2 and xy both second degree. a-x x = a(l _e. An equation will be homogeneous if all terms are of the same degree. This is a different use of the term 'order' from the mathematical definition. giving C = -In a and In(_a) = kt.2 First-order homogeneous equations A first-order equation contains only the first differential such as dy/dx. The standard method of solution is to replace one variable by introducing a new one that is the ratio of the variables.) = kdt. so that dy = zdx + xdz. . = O. Then dx dt = k(a -x).

**Equations of first order, first degree
**

and

**x(zdx + xdz) + 2zxdx
**

zdx + xdz + 2zdx

3zdx + xdz

143

= 0,

= 0,

= 0.

**The equation will now be of the separable variables kind, to be solved as
**

before. Thus we collect all terms in dx on one side of the equation and

terms in dz on the other, to give

3zdx = -xdz,

dx

dz

x

In x = -! Inz+C,

In x 3 z = C',

x 3 z = C",

where C, C' and C" are constants, which are actually related to each

other by C' = 3C = In C", but the only point of interest is that a

constant has been retained in the correct place. We finally return to the

original variables by the substitution z = y/x, to give

x 2 y = C.

The correctness of the solution may be checked by differentiation and

substitution into the original equation.

Example 6.5

An equation that is homogeneous in the terms containing the variables

but also contains an additive constant can be reduced to a simple

homogeneous equation by using a substitution to eliminate the

constant. Thus if

dy

x-+y=3

ax

we substitute z

=

= dy, to give

xdy+ (y -3)dx = 0,

xdz+zdx = 0,

y - 3, dz

dx

dz

x

z

Inx = -lnz+C

Inzx = C,

zx = C',

144

Differential equations

so that

6.2.3

x(y -3) = C.

Exact equations

**An exact differential is defined in Section 3.5 as being of the form
**

X dx + Ydy where a function z(x,y) exists with this as its differential:

dz

=

Xdx+Ydy

(!:)y

=

dX+(:;)x dy,

so that

and

Y- ( -az)

- ay x·

**The test for such an exact differential is the cross-differentiation
**

identity

If a differential equation either has this form or can be converted into it,

direct integration is then possible.

Example 6.6

dz

=

2(x + y)dx + (2x + l)dy.

**We first test whether or not this is exact, as in Section 3.5, by seeing if
**

cross-differentiation applies, which means looking for equality of

a

ay (2 (x + y»

and

a

ax (2x+ 1).

**In this case both are equal to 2 and so it is an exact equation. It may be
**

possible to see by inspection what the function of x and y is that gives

the differential equation, but if not we may proceed as follows.

The solution may be obtained from the equations

2(x+ y) =

(::)y

and

**From the first equation
**

z = f2(X

=

+ y)dx

x 2 + 2xy + C(y),

Equations offirst order, first degree

145

**this integration being the inverse of partial differentiation with respect
**

to x so that C(y) is a constant only so far as x is concerned; it may be a

function of y. Then

(aoz) .= a0 (x +2xy+C(y»

y

Y

x

2

dC(y)

= 2x+-- = 2x+ 1

dy

so that

dC(y) = 1,

C(y) =y+C

dy

where C is now a numerical constant. The required solution is therefore

z = x 2 + 2xy + y + c.

6.2.4 Linear equations offirst order

These have the form

dy

dx + P(x)y

=

Q(x)

(6.2)

**where P(x) and Q(x) do not contain y but may be functions of x or
**

constants. Writing the equation in the form

dy + P(x)ydx

= Q(x)dx

(6.3)

**the solution is obtained by multiplying both sides by the integrating
**

factor

efP(x)dx.

This has the effect of converting the left-hand side of the equation into

an exact differential by making cross-differentiation apply (Section 3.5).

We show this by multiplying the left-hand side of (6.3) by a functionf(x)

to give

f(x)dy + P(x) yf(x)dx

and cross-differentiation gives

o~~) = OP(~:f(X) = P(x)f(x),

so that

f

df(X)

f(x) =

lnf(x) =

f(x)

f

f

P(x)dx

P(x)dx,

= e fP(x)cIx,

146

Differential equations

**omitting any integration constant so as to obtain the simplest solution.
**

Example 6.7

dy

dx +2xy = 4x.

Multiply by ef2xdx = e x2 gives

eX2 dy+ 2xyeX2 dx = 4xex2dx

d(ye X2 ) = 4xe x2 dx.

Integrating both sides

ye X2 = 2e x2 + C,

y = 2+Ce _X2.

Example 6.8

Successive kinetically 'first-order' chemical reactions give linear differential equations. An extension of Example 6.3 is

A~ B ~C

(a-x)

(x-y)

**where the rate coefficients kl and k2 are shown, together with the
**

concentrations of the various species at time t. The rate of production

of C is given by

**This may be seen to be a first-order linear differential equation by
**

writing it in the form

dy

dt +k 2 y = k 2 x.

The integrating factor is e$k 2dt = e k2t, giving

e k2tdy + k2 ye k2 tdt

= k2 xe k2tdt,

d(ye k2t ) = k 2 xe k2t dt.

The right-hand side contains the variable x, which may be eliminated

by applying the result of Example 6.3 to the first stage of the reaction.

Thus

Substitution and integration then gives

y = a-

k2 a e -k,t+ Ce -k2t.

k2 -kl

Linear differential equations

147

**The integration constant C may be found from the initial conditions
**

that Y = 0 at t = 0, giving

Y = a[1-(k 2 e -k,t-k 1 e -k,t)/(k 2 -kdJ.

**Fig. 6.1 shows typical curves of (a - x), (x - y) and y, each plotted
**

against time t.

A-B-C

f

Fig. 6.1

**6.3 Linear differential equations
**

A linear equation is one containing only the first powers of a variable

and its derivatives, the general form of linear equation of order n being

dny

dn - 1 Y

ao(x) dxn+adx) dxn- 1 + " .+an(x)y =f(x).

(6.4)

**There is a general method of solution for such an eqtUltion only when
**

the coefficients ao, a1> ... are constants, and this is discussed in Section

6.3.2. An equation in which all terms are functions of the dependent

variable Y is called homogeneous, which is whenf(x) = 0 in (6.4), and

this is discussed in Section 6.3.1.

An important consequence of an equation being linear is that linear

combinations of solutions may also be used. For example, if y = oc(x)

148

and y

Differential equations

**= fJ(x) are both solutions of the homogeneous equation
**

dy

d 2y

ao dx2 +al dx +a2Y = 0,

**then y = ex(x) + fJ(x) is also a solution. This follows because the
**

equation may then be separated into parts after making the substitution, one part containing ex and the other part containing fJ, and both

parts are zero since ex and fJ are both solutions of the equation.

A physical system is called linear if its variables are linearly related,

meaning that multiplication of one variable by a factor increases

another by the same factor. Thus a linear amplifier is one for which if we

double the input we double the output. For a 'first-order' chemical

reaction the rate is proportional to concentration and a series·of such

reactions gives rise to a linear differential equation, as in Example 6.8.

The differential equations describing linear physical systems will be

linear not only in the dependent variable y and its derivatives but also in

the independent variable x. We may use linear differential equations

not only to describe physical systems that are truly linear, but also as a

first approximation to non-linear systems. This is because the effect of

small enough disturbances will be given by the first, linear, term in a

Taylor expansion.

6.3.1

Homogeneous linear equations with constant coefficients

**These have the form
**

d"y

d,,-l y

(6.5)

aO-+al--1 + ... +a"y = o.

dx"

dx"The substitution y = em,x reduces the solution of the equation to that of

the auxiliary. algebraic equation in m.

Example 6.9

Put

dy

m,x

dx=me,

then

em" (m 2 + m - 2)

= O.

This equation will be satisfied by solving the auxiliary equation

m2+m-2 = 0,

Linear differential equations

149

**which factorizes to give either m = 1 or m = - 2, so that the differential
**

equation is satisfied by either y = eX or y = e - zx. It is also satisfied

by y = C1e x and by y = Cze-zX, where C 1 and C z are arbitrary

constants. Furthermore, since the equation is linear, we may also

write

y = Clex+CZe-zx.

This last solution contains the two arbitrary constants expected in the

solution of a second-order equation and is the most general form, the

earlier solutions corresponding to values of zero or unity for the

arbitrary constants.

This method gives solutions whenever the auxiliary equation has real

roots. If the auxiliary equation were to give equal roots, however, we

would obtain only a solution containing one arbitrary constant. In that

case, multiplication of that root by Czx will be found to produce the

required second solution.

A solution to the differential equation may exist even when the roots

of the auxiliary equation are imaginary since these appear as imaginary

exponents in the solution. From Section 1.10

e iy = cos Y + i sin y

and

e -iy = cos Y - i sin y.

**Now, the arbitrary constants used in the solution of the differential
**

equation may also be complex numbers and these will then produce

real solutions, as in the following example.

Example 6.10

Put Y = em.>:, so that

emX(mZ + 1) =

°

**and the auxiliary equation becomes
**

mZ+ 1 = 0,

m

=

±i,

**giving the general solution
**

y

= Ae ix + Be -ix.

Taking A and B to be complex numbers, we put A = a + ib and

150

Differential equations

B = a - ib, giving

y = (a + ib)(cos x + i sin x) + (a - ib)(cos x - i sin x)

= 2acosx -2bsinx,

or

y = Ccosx+Dsinx,

**which is wholly real.
**

Example 6.11

An equation of the form of Example 6.10 is obtained for simple

harmonic motion, when a particle of mass m experiences a restoring

force proportional to its displacement x. Then

force

d 2x

dt

= m= -kx,

2

d2 x k

= O.

d t 2 +-x

m

Put

then

a = ±i.j(klm)

giving

x

=

Ae it.J (kim) + Be - it.J (kim),

**and when A and B are allowed to be complex,
**

x

=

C cos [t

.J (kim)] + D sin [t .J (kim)].

By using the trigonometric identity

**sin(A+B) = sinAcosB+cosAsinB
**

we can write the solution in the form

x

= A

sin(wt + 4»,

**where the constants A and 4> are simply related to the constants C and
**

D, and

w = .j(klm).

This equation suggests the interpretation in terms of polar coordinates shown in Fig. 6.2; point P moves in a circle of radius A with

.

or if x is the fluctuating dipole moment accompanying vibration of a molecule. so that the kinetic energy T becomes T= !mv 2 = !mw2A 2cos 2 (wt+fjJ) = !mw2A2[1-sin2(wt+~)] = !mw 2(A2 _x 2). force = -potential gradient = -dV/dx. when x = A it is wholly potential. and x is the projection ON of OP onto the vertical axis. = 21tJ(m/k). independent of x. and the angle fjJ gives an arbitrary zero from which the motion begins at time t = O. A graph of x against t has the form of a sine wave as shown in the figure.7). so that T= !kA 2 -!kx 2 giving total energy T + V = ! kA 2 . and the curve of V against x is a parabola (Example 1.152 Differential equations angular velocity w. From equation (2. Point N moves up and down on the vertical axis with simple harmonic motion. and we have an interchange between kinetic and potential energies during the motion. so that 21t T= w v = (1/21t)J(k/m). If OP is a vector (Section 3.58). so that the potential energy for simple harmonic motion is given by V= - f Fdx = !kx 2 . so that interaction with the radiation can occur and an absorption spectrum will be observable.9) representing the dipole moment of a rotating polar molecule. Thus when x = 0 the energy is wholly kinetic. But w 2 = kim. . The amplitude of the motion is A. The time period T is that taken for one complete cycle of the motion and the frequency v is l/T. an oscillating electric field is produced which will be the same as that of electromagnetic radiation of the same frequency. The velocity at displacement x is given by dx v = dt = wA cos (wt + fjJ).

66) and so that d2r F= -m*- dt 2 and the motion will be as described above with the usual substitution of reduced mass m* for m. 6. This applies both to the vibration of an object immersed in a fluid such as air or a liquid. assume that the restoring force is proportional to the displacement. Given that we resolve the general motion of a molecule into motion of the centre of mass and motion relative to the centre of mass.3.65) and (2.12 Simple harmonic motion will occur only in the absence of frictional losses. to a first approximation.3 But from equations (2. and to electrical oscillations in the . Example 6. the motion will be as shown in Fig. zero displacement corresponding to the equilibrium separation r 0 so that x becomes (r . is proportional to the velocity.r 0). so that r1 r2 14----.Linear differential equations 153 The vibration of a diatomic molecule occurs by extension and compression of a chemical bond and we can. 6. in vibrationally excited states the restoring force will no longer be proportional to the displacement and the curve of potential energy against r will depart from the parabola corresponding to simple harmonic motion. each atom vibrating relative to the centre of mass O.-------*------1 I • 1 I ~ I 0 • • m2 I 1 ~-------------------~ ro Fig. The restoring forces on the two masses are equal and opposite. often to a good first approximation. The motion of real macroscopic systems will be subject to resistance which. This simple model of molecular vibration will apply only to small displacements.

The process then becomes irreversible and potential and kinetic energy are gradually con .154 Differential equations presence of electrical resistance.= m dt dt 2 ' where the frictional. This is called the critical damping condition. When o < {32 < 4mk. and so simple harmonic motion. the solution then has the form of Example 6. which means on the extent of the damping. so that the vibration decays at a rate depending on the resistance. giving the solution This also gives exponential decay. . or damping. When {32 = 4mk.. When heavily damped. force is {3 times the velocity and acts in the same direction as the restoring force when dx/dt is positive and so the displacement increasing. the auxiliary equation has equal real roots. If {3 were zero we would have purely imaginary roots to the auxiliary equation as in Example 6.11. The behaviour then depends on the relative magnitudes of {32 and 4mk.4). This is called damping.9.4. with the solution a= - 2~ [{3 ± ({32 - 4mk)J. erted into heat. {32 > 4mk and the auxiliary equation has real roots. but allowing the decay to occur most quickly (curve B in Fig.say -al and -a2. This is a homogeneous linear equation with constant coefficients and the usual substitution x = eat gives the auxiliary equation a 2 {3a k m m +-+-=0. of the form a = d±if. The restoring force is then given by dx d2x force = -kx -{3. which corresponds to exponential decay of any imposed displacement. 6. less than the critical amount of damping is present and the roots of the auxiliary equation will be complex. the damping force being just sufficient to prevent oscillation. This is illustrated as curve A in Fig. 6.

the real part is exponential decay and the imaginary part gives periodic oscillation of frequency f/(2n). 6. 6.4. and in more complicated control systems by adjusting the control signal input which depends upon the derivative dx/dt.p/(2m) and f = [J (4mk - = J(~.:~2). The critical damping condition is very important in practice because that gives the fastest response without oscillation.Linear differential equations 155 x Fig. This is achieved by adjusting the value of p: in a simple moving-coil galvanometer system by adjusting the circuit resistance to equal that for critical damping. We notice that the frequency of the damped oscillation is lower than the value (1/2n)J(k/m) which would be obtained in the absence of damping. . the result being a damped oscillation of form C in Fig.4 where d (for decay) andf (for frequency) are real numbers given by d = . p2)J/2m The solution is then the product of terms arising from the real and the imaginary parts.

however. A particular integral is therefore y = x.1 y ao dxn + a1 dx n.156 Differential equations 6. the same function with arbitrary coefficients should be used as a trial solution. however. Example 6. which is a first-degree polynomial. To distinguish these two kinds of solution. as in Example 6. the rest can then be solutions of the corresponding homogeneous equation. We will see that this separation of solutions has not only mathematical significance but physical significance as well. The complementary function for this equation was found in Example 6. if successful.6) We again use the principle that the n adjustable constants expected in the solution of an nth-order linear equation are obtained by adding together the n separate solutions.2 General linear equation with constant coefficients This has the form dRy dn . it is often possible to find a solution by trial.1 + . the values of the coefficients will be found. In simple cases. in particular. Substituting this into the equation gives a-2(ax+b) = 1-2x.13 d 2y dx 2 dy + dx -2y = 1-2x. whenf(x) is a polynomial. or an exponential or trigonometric function. we try multiplying the trial solution by x. The general solution of the equation is then the sum of these two. which is satisfied by a = 1 and b = O. (6. so that we try the general first-degree polynomial y = ax + b as a solution.9.. This is substituted into the equation and.3. we also take advantage of the' fact that only one of these solutions need give the right-hand side of the equation. Heref(x) = 1-2x..9 to be and the general solution is the sum of the complementary function and . details of which are given in texts on differential equations. if not. the ones with f(x) = 0 are called the complementary function and the solution with the given value of f(x) is called the particular integral. There are systematic methods for finding a particular integral for equations of various types. + anY = f(x). In this case. which is withf(x) = O.

x • The right-hand side of the original equation is again a polynomial of first degree. so long as the coefficients are constants. or y = CleX+C2e-2x+x. and the equation becomes dy eX dx + ye X so that = 3xe x.Linear differential equations 157 a particular integral. m= -1 to yield the complementary function y = Ce. emX(m + 1) = 0. Example 6. and put y = emx. so that the general solution becomes y = Ce. so we try y = ax + b to give dy dx + y = a+ax+b = 3x if a = 3.2.2.3 as a particular integral.3. b = . We are here comparing the method that applies to a linear equation of any degree. dy dx + y = 0.4 by solving the equation dy dx +y = 3x. ye X = 3e X(x-1)+C y = 3x-3+Ce. with the method that applies to any linear equation of the second degree.x + 3x -3.4 is to multiply the equation by eJPdx = eJdx = eX.14 Show that the method of this section gives the same solution as the method of Section 6.x . giving y = 3x . We write the linear homogeneous form of the equation. The solution corresponding to Section 6. d(yeX) = 3xe xdx.

In order to sustain oscillation the forcing function must be 'in tune' with the oscillation.12 and have interpreted the damping as a dissipation of kinetic and potential energy as heat.15 We have discussed damped harmonic motion in Example 6. PAp = -F. which is satisfied if k = mp2. A particular integral is sought by using a trial function of the same form as the righthand side of the equation. and the equation becomes -mp2 x . when we apply a sinusoidal forcing function .12.Ap sin pt + Bp cos pt. so we put x dx dt = A cos pt + B sin pt. A = -F/(Pp) = -(F/p).j(m/k). Example 6. in the notation of Example . The same equation is solved by a third method in Example 6.158 Differential equations as before. or as incident radiation of a particular frequency on a molecular system. = -Fsmpt. The complementary function will be as in Example 6. or as an applied alternating voltage to an electrical circuit.16. = y(k/m).F sin pt the equation becomes force = dx d2 x -kx-Pdt-m dt 2 . B=O. A common physical situation is when such oscillation is sustained by a forcing function in the form of a periodic input of energy which may be applied mechanically. p The particular integral is then The damped oscillation conditions are. .pApsin pt + PBpcos pt + kx = F sin pt.

12.Pt /(2m)(C 1 cosft + C 2 sinft) - ~ J(. remembering that when Y = Yl the expression .3 Linear equations of second order The general second-order linear equation can be reduced to a firstorder linear equation if a solution can be found by inspection. together with the final term in the equation.Linear differential equations 159 6.J (kim). x = e. 0< p2 < 4mk and a = d ± if. which is the value of w for simple harmonic motion (Example 6. so that dy du dYl dx = Yl dx + u dx' d2y dx 2 d2u dYl du d 2Yl dYl du = Y 1 dx2 + dx dx + U dx 2 + ~ dx' Substitution into (6. then put Y = Y1U(X). If the system were displaced and then the forcing function applied. 6.11).3. in spite of the fact that the real damped oscillation frequency is lower. Given the equation (6.7) if Y = Yl is a solution to the corresponding homogeneous equation withf(x) = 0. This corresponds to applying a forcing function at the frequency w/2n which the system would display if it were undamped.7).) cos [t J(~) J The physical significance of this is that the motion is the damped harmonic motion that would follow a displacement in the absence of the forcing function. The condition that was chosen to obtain the particular integral was when P = .Pt /(2m)(C 1 cosft + C 2 sinft). it would display damped motion for a time until the contribution from the first term (the complementary function) in the general solution had decayed. so that the complementary function is (compare Example 6.and the general solution is x = e. The condition when the forcing frequency is equal to the natural undamped frequency is called resonance. followed by sustained oscillation given by the second term only. this is the completely contrary condition to critical damping.10).

gives d2u + 2 dYl dx 2 du =f(x). or adding or subtracting terms in f(x). ••• are also simple algebraic functions of the new variable p. We consider two forms of such integral transformation: Laplace transforms. dx dx Yl This is a first-order linear equation. which is why this is called a transformation. is given in tables. Only a simple example will be given to illustrate the method. from which u(x) may be found by integration. which may be solved for du/dx.8) The limits of integration with respect to x mean that the variable x disappears after integration. which allows it to be useful in solving differential equations.7) is zero. The tables of transforms are then consulted again in ·the inverse sense of: knowing the transform L(y). The second reason is that it is a linear operator. which means that multiplying by a constant.PXf(x)dx. which are useful in solving differential equations. is that the Laplace transform of derivatives dy/dx. (6. 6. which can be solved to give the Laplace transform of y in terms of p. simply means multiplying. we obtain instead an expression in terms of the new variable p. . This may be achieved not only by algebraic substitution but also by the use of substitutions in the form of integrals. The terms in a linear differential equation can in this way be converted into an algebraic equation in p. This is useful in solving differential equations for three reasons: the first is that the integral can be evaluated for simple functions and. the reader being referred to a text containing a full table of transforms for more details. The third reason.160 Differential equations on the left-hand side of (6. which are used in instrumentation.4. This technique is particularly useful for the equations describing electrical circuits.1 Laplace transforms The Laplace transform L(f(x)) of the functionf(x) is defined by L(f(x)) = I'O e . 6. adding or subtracting terms in L(f).4 Integral transforms Substitution methods are often used to transform from one independent variable to another. and Fourier transforms. indeed. find yand so the solution of the equation. d 2 y/dx 2 .

9) so that r(x) is the transform of t x .1 • This may be evaluated by integration by parts.8) as pft foro xfte-PXpdx = pft+l L(xft) = n! (6. defined by (6. (6.11) we obtain the form (6. (6.PX ].y(O) + pL(y).Integral transforms 161 A particular example of a Laplace transform is the gamma function r(x).10) Repeated integration by parts then gives r(x + 1) where = x!r(l) so that If we substitute n for x and px for t in (6. We write the function for (x + 1) as r(x+ 1) = to tXe-tdt = -[tXe-t].14) .PX dx dx = [ye. where y(O) is the value of y when x = O.12) In particular. when n = 1 and when n = 0 L(x) = 1/p2 and L(l) = lip.13) The Laplace transform of dyldx is given by L(dyldx) roo dy = Jo e. (6. + p LOO ye-PXdx = .+x LOO tX-le-tdt = xr(x).

This equation may conveniently be solved by using Laplace transforms. From a table of transforms. which are defined by x = 0 and x = a.X ) = 1/(P+1) which. We now take the inverse transform of each term. together with (6. and the wavefunction t/! must be zero at the walls.yeO) + pL(y) + L(y) = 3/p2.162 Differential equations Example 6. .14 dy dx + y = 3x. Solving for L(y) we obtain L(y) 1 = 3 p+1 y(O) + p2(p+1) 1 = p + 1 yeO) - 3 3 3 p+ p2 + P + 1 where we have used partial fractions for the last term.13).pt/!(O) -1/1'(0).17 The Schrodinger equation in one dimension has the form a 2t/! ax 2 = - 8n 2m ~ (E . Example 6. or by integration similar to (6.14).16 As a simple example we use Laplace transforms to solve the same equation as Example 6. Taking Laplace transforms of each term gives .x = (Yo + 3)e. the potential energy V does not change and so may be taken as zero. For free translation of a particle of mass m between walls separated by distance a.V)t/!. we find L(e.14 when C = Yo + 3. gives y = y oe.x -3+3x+3e. X which is the same solution as obtained by the complementary function and particular integral approach in Example 6. or by evaluating the integral. From tables.-3 + 3x. L(d 2t/! /dx 2 ) = p2 L(t/!) .

the resolution of a given function into a sum of sine and cosine terms.4. The inverse of this. leaving the required signal less obscured by noise.1/1' (0) = - 163 = 0 and 1/1(0) = 0. 81t 2 m ~ EL(I/I).a = n1t. we then obtain the Fourier transforms of!(w) as F(t) = -1 f+oo !(w)e-iwtdw 21t . Again from tables L(sin AX) = A/(P2 + A2) giving the solution to the equation 1/1 and at x = Hence 6.sm AX. so that a graph of sin wt against t has the form of a wave of frequency w/21t.00 .2 Q.Integral transforms Taking transforms of the equation. Putting 81t 2mE/h 2 = A2 we obtain L(I/I) = I/I'(0)/(p2 +A2). 1/1' (0) . with V p2 L(I/I) . can be adjusted to fit any particular curve as closely as we may wish. A. If we superimpose sine and cosine functions of different frequencies we produce a complicated pattern which. In the limit when the number of terms becomes infinite we may replace the sum by an integral. Sine and cosine functions are periodic. Fourier transforms Fourier transforms are of practical importance because their use enables instruments to be designed and used in such a way that results can be obtained very quickly. for which there are standard procedures. This is an advantage because measurements of weak signals always have superimposed upon them the effects of random disturbances called noise. = -A. 1/1 = 0 so that sin AQ = 0. is Fourier analysis. Since noise produces signals of random sign these average to zero for a large enough number of repeated measurements. given enough terms.

that is observed can be transformed into a superposition of exponential decays in time of a set of frequencies. secondly. that the resulting decay. Their application in spectroscopy is to transform measurements of the decay with time of the response induced by an irradiating pulse into the required spectrum as a function of frequency. the set being the required spectrum. first. or relaxation in time.10. The complex exponential function is expressed in sine and cosine terms in Section 1. . so that a spectrum of absorption frequencies can all be excited simultaneously and.164 Differential equations and /(w) = f:: F(t)eiwtdt which relate functions of frequency (w) and functions of time (t). that a pulse is equivalent to a superposition of different frequencies with the major components near the frequency of the pulsed radiation. An appropriate text should be consulted for the derivation and use of Fourier analysis and transforms. The mathematical and physical principles may be brought together by observing.

Experimental errors are of two kinds. however.CHAPTER 7 Experimental error and the method of least squares 7. A typical systematic error would arise if length were measured using a ruler calibrated at one temperature but used at a higher one. it is as absurd to measure the length of a piece of string to 0. This implies that the results have been analysed with great care because two figures are quoted in the uncertainty.01 mm as it would be to weigh a block of gold to the nearest kilogram. The least that should be done is to quote only figures that are believed to be reliable. systematic and random. All measurements would then be short by an amount proportional to the .12335 and 2. that such measurements will not give results outside the bounds set on the existing measurement. Some properties can and should be measured more precisely than others. such as 2.123 ± 0.024. It is preferable to give an explicit statement of precision.12345. which is the most that can ever be achieved. This chapter is concerned with the methods of analysing experimental data which are necessary to achieve such a high level of confidence in the statement of results.15. it does imply.1234 implies that it is between 2. so that a result given as 2.05 and 2. the result and the uncertainty both being quoted to the same number of decimal places. and 2. It does not imply that it is not possible to obtain greater precision in the measurement because improved equipment and better instruments may well produce a more precise result.1 implies that it is between 2. for example.1 Significance Experimental measurements must be made and used with their probable accuracy in mind.

mechanical vibration or thermal or electrical noise. Such errors can be analysed statistically in terms of the distribution of error about a mean value. 7. the mean will not be the true value and the curve may not be symmetrica1. Fig. due to poor control.7. we sum over such a large number that the differences between adjacent results becomes so small that we can replace sums by integrals. this distribution is a curve as in Fig. discussed in Section 7. the results will scatter about the arithmetic mean x given by x = LXi. The term 'random' is used in error analysis in a particular sense. For this reason sums and integrals occur in this chapter. depending on whether or not we are assuming an infinite number of measurements. If we then add the results together. when errors are of random sign. Such a curve could be drawn if a large enough number of repeated measurements were made. which means the number of independent measurements made. Since the methods used to assess the size of experimental error are statistical. This is the normal. 7. n (7.1 shows the probability that a particular result will be obtained in the measurement of a quantity x. which is the arithmetic mean. Instability.1.1) . produces a scatter in the experimental results. The curve is then symmetrical about a maximum. or from inadequacy in the design of apparatus or in the technique used. The simplest case is the one already mentioned. the sample size. is very important. Systematic errors arise from inadequacy in the calibration of instruments. The uncertainty in the final result needs to be known and the conventional way of expressing that uncertainty is discussed in the next section. We do not therefore expect repeated measurements to give the same result. The statistical theory is developed by assuming at first an infinitely large number of measurements. and the quality of a set of measurements is assessed by the spread of results obtained. If significant systematic error is present in addition to such random error. which shows the number of times a particular error occurs as a function of the size of that error. and an error of given size will usually be equally often positive as negative. 7.2 Root-meaD-square error If the measurement of a quantity x is repeated n times. or Gaussian distribution of error. Small errors occur more often than large ones.166 Experimental error and method of least squares length being measured.

~ I >c' ~ .: ~ s ....r..

4) The mean value x of x is obtained by summing (integrating) the product of x and its probability p(x) over all possible values of x.2) The relation between this and the standard deviation q(x) is discussed in Section 7.3) 7.1. We can then define only the probability of obtaining a result in an interval of X between X and X + dx as p(x)dx.X)2.168 Experimental error and method of least squares and this is the best result obtainable from the measurements. (7. s(x).x). however. (7. The final result of the experiment is to be expressed as x ± Ax. This is the smooth curve shown in Fig. where p(x) is a continuous function of x. such as the number of matches in a box. (7. rather than discrete.5) . (7.x) because that is zero by definition ofx.6. 7. We therefore use instead the square of the deviation. We ensure that p(x) is a fraction between 0 and 1 by normalizing the function to unity. we define the probability p(x) of obtaining a particular result X as the fraction of a very large number of samples which contain that number. In order that Ax shall be a useful measure of uncertainty it should reflect the spread in the values of Xi obtained. and define the root-mean-square error. which means making the total area under the curve unity by f +OO _ 00 p(x)dx = 1. (Xi .X)2 = (l/n) L (xf .2XiX + x 2) = Lxt/n-2xLxdn+x2 = Lxt/n-x 2. x f +OO = _ 00 xp(x)dx. where Ax is a conventional assessment of the uncertainty of the measurement x of x. variables. In physical measurements. We cannot use for this the average value ofthe error (Xi . by s(x) -J II (Xi - = X)2 /n]. we are usually concerned with continuous.3 Distribution of error For discrete variables. (Xi . The root-mean-square error can be calculated most easily from a set of values of X by use of the relation S(X)2 = (l/n) L (Xi . The experimental error in a particular result Xi is then the difference between that value and the arithmetic mean.

So long as dl independent variables are identified and their values controlled in the experiment. We are more often concerned with measurements of more than one variable. From the statistical point of view there is no difference between these two applications because the uncertainty in the coefficients of a fitting equation can be regarded as a measure of the degree to which the assumed relation fits the data.4 The statistical analysis of experimental data The methods used in the statistical analysis of data can be classified either from the mathematical or from the physical points of view. Statistical texts discuss the concepts of covariance and correlation coefficient (both considered here) and of tests of significance and confidence limits (not considered) as measures of the extent to which variables are related rather than as measures of the uncertainty in the values of the fitting parameters. . and this form of correlation analysis is not considered here. to determine how closely a set of results follows a particular distribution. This has value as a way of testing for systematic error of a kind that distorts the distribution.The statistical analysis of experimental data 169 The root-mean-square error O'(x) in x is then given by 0'(X)2 = L+oooo (x -x)2 p (x)dx. and results will be subject only to experimental error. Statistical methods can then be used either to test whether or not the variables depend on each other.6) 7. both of which will now be considered. This is a physical concept. and for well defined chemical systems is easily determined. if a sufficiently large number ofreadings is taken. On the other hand. the distribution curve can be drawn. or to find the most probable values of numerical coefficients in an equation that is assumed to fit the data. The results will scatter and. properties and materials may not be rigorously defined and the independent variables may not be identifiable and controllable. A true value will not then exist and relations are being sought between quantities which both show legitimate variations. Another way of comparing the various aspects of statistical analysis is in terms of the relevant number of independent variables. a true value of any experimental quantity will exist. such as the 'chi-squared' test not considered here. The simplest case is when measurements of a single quantity are repeatedly made. The latter application is the more important one in physical science. (7. Only a small number of results will be available in real cases and there are standard tests.

y)p(x.10. (X2' Y2). Suppose a quantity z(x. which is normalized to unity by the two-dimensional analogue of (7. for which we use the formula for propagation of error. y)dxdy. and that measurements are repeated n times to obtain pairs of results (Xl' yd. (7.170 Experimental error and method of least squares 7. y)dxdy. Yn). Knowing the uncertainty in the results we then need to calculate the uncertainty in the related quantity. Assuming an infinite number of measurements. y)dxdy 00 _ 00 (7.· .~) (y - y)2 ] (7. y) depends on experimental variables x and y. X= f _+OO f+oo xp(x. We define the probability of obtaining a result in the range (x to x + dx.4) The root-mean-square errors cr(x). The arithmetic mean value z of z is then z= f:: L+: z(x. L+: L+: (z .11) 00 -00 _ 00 -00 and we require an expression for cr(z) in terms of cr(x) and cr(y). y to Y + dy) as p(x.7) We can also write the mean value of x in this form. y)dxdy.Z)2 p(x. (7. (7.12) . y)dxdy.y)2p(X.8) to express a departure of z from the arithmetic mean in the form (z -Z) = (oz) oX y (x -x)+ (oz) (y _ oy x + 2 (o~:y) (x -x)(y + higher-order terms. We use Taylor's theorem (Section 3.(xn . cr(y) and cr(z) in x. y and z respectively are defined as cr(X)2 +OO f+oo = f_ (x - cr(y)2 cr(Z)2 = X)2 p(x. the mean values will be free from experimental error so that zis obtained from x and y. (7.9) +OO f+oo =f (y .5 Propagation of error Experimental results are often substituted into equations so as to obtain values of related quantities. y) y)+~2 [(02~)(X _X)2 ox + (:. . y)dxdy..8) because the integration with respect to y is simply the integral of the probability of obtaining any value of y for a particular value of x.

:)(!.15) The mean squares of the errors U(X)2 and U(y)2 are called the variances in x and y.Small-sample errors 171 We assume the errors to be small so that only the first-order terms in the Taylor expansion are significant..y)p(x.Y (y - yf + 2 (.. the integral in (7. This curve is characterized by the form of the distribution. The covariance is therefore zero if the variables are independent. if z is a function of independent variables x. (7. The significance of covariance is that it determines whether or not the variables are independent of each other. before which we consider relations that apply to any form of distribution. A particular form of distribution is considered in the next section..16) Small-sample errors We aSSllme that a particular distribution curve such as Fig.13) into (7. The arithmetic mean has been defined for a series of n results by equation (7..5) and (7. y) = L+oooo Y + (.1) and the corresponding root-mean-square error as s(x) by equation (7.) (x -x)(y .14) (x -x)(y .. These have also been defined in terms of integrals by (7. y. 7.. y) is called the covariance of x and y.6 = (!: y U(X)2 + (. y U(y)2 + (. If they are independent. the formula for propagation of error is U(Z)2 7.. In general.2). and both of these integrals are zero by definition of the arithmetic mean.15) separates into the product of two integrals each with respect to one variable only. the arithmetic mean and the rootmean-square error. y)dxdy. and the new term u(x. (7. y) (7.11).1 exists for the results of any experimental measurement.~ Y U(U)2 + . .y). (7.. Then (z _Z)2 = (!:Y (x _X)2 + (.6). assuming higher derivatives to be negligible so that the first derivatives are constants to be taken outside the integrals. These become the same onlv as the number of results . we obtain U(Z)2 = (!:Y U(X)2 where u(x. L+: U(y)2 +2 (!:) (!. u.13) Substituting (7.) u(x.

The mean xn of a sample of n results is only an estimate of the mean Xoo of the distribution. It is seldom worth while to repeat a reading more than about 10 times. automatic repetition is especially useful because the effect of even high-frequency noise can then be averaged out.17) This relation shows that when n duplicate measurements are made of the same quantity. so that a different sample of points will give a different value of xn. and the mean is calculated so as to obtain a more reliable estimate. Here we are seeking not the error in . in principle. Thus the mean of nine duplicate readings will reduce the uncertainty by a factor of one-third. except when automatic recording is available. The effect of small-sample error on the estimate s(x) of the standard deviation can also be calculated. In the same way. the uncertainty is thereby reduced by the factor 1/ In. but to improve the precision by one decimal place we would need to take 100 readings. an infinite number of points to establish the curve of distribution of error with complete certainty. is only an estimate of that of the distribution u(x).172 Experimental error and method of least squares tends to infinity because we require. n (7. denoted by s(x). the root-mean-square error of a sample of n results. By definition and since _ Xl X2 n n Xn=-+-+ the integral separates into the sum of n integrals of the form _ 2 U(Xn) = so that n 1 f+oo L 2 i=ln _ 2 (Xi -Xoo) p(x)dx = n U(X)2 L i=l -00 _ U(Xn) U(X) = Tn· -2- n U(X)2 = --. each of which is subject to error with standard deviation u(x). we obtain a set of values of xn which will themselves follow a distribution about the true mean Xco" We can then calculate the standard deviation of the mean u(xn ). only as n tends to infinity will the means of different samples become the same and equal to the mean of the distribution. When it is possible to take readings very quickly. which is a measure of the uncertainty in xn due to the fact that we have too few results. which is called the standard deviation. and then repeat the set of n measurements a large enough number of times. If we obtain n results and calculate the mean X n .

19) the middle term having disappeared because L (Xi .20) and combining (7. if we fit n sets of values of several experimental variables to an equation containing k numerical coefficients. by definition (7.) = 0 for the arithmetic mean..... but the error in our estimate of that error.k) degrees of freedom and the standard deviation of the fit is (7. (7.18) We introduce X" into this expression by writing the algebraic identity (Xi-X<x» = (xi-x....)-(x.-X<x»2. If the true mean of the distribution.-X<x»+ (X. This can be interpreted by noting that the calculation of the mean involves the use of an equation containing the n variables.22) .)2 -2(Xi -X. x<x>' were known we could obtain a good estimate of the standard deviation from only n results by using the relation U(X) 2 L (Xi -X<x»2 = =------'-- n (7.)2+n(X..18).x.)2+ n(X. which is of second order in small quantities.. or we have (n -1) degrees of freedom.21) which is the equation for the standard deviation of a set of measurements ofa quantity x.)2 ' n- (7.-X<x»2 = ns(x)2 + nu(x. The estimate s(x) of the standard deviation u(x) is said to be biased by the smallness of the sample. this bias being corrected by the factor nj(n -1)..Small-sample errors 173 the measurement..l X.)(X. so that (n -1) variables remain independent..)2 = ns(x)2 + U(X)2 so that n ()2 ( )2 =--SX UX n-1 = L (Xi . By analogy..20) gives nu(x)2 = L(Xi-x. We can then write the required relation between s(x) and u(x). (7.-x<x» so that (Xi -X<x»2 = (Xi -X.-X<x»2. and L(Xi-X<x»2 = L(x i -x. we then have (n .19) and (7.

The probability P of obtaining a set of results Xl' X2' .x) = 0 dm dm dx' so that L [d In P(Xi .Zcalc) is the deviation of an experimental value of Z from the fitting equation.23) and din P = L din P(Xi .. =0. times equation (7.m) = L din p(xi . (7.24) (compare Section 2.x)] = o... Xl .. When error is defined as departure from a mean value m.m) of obtaining error (Xi .25) .(xl-x)+ d( ) +A.24) Solving equations (7.23) and (7. we have (7.24) simultaneously by adding A. is the product of the probabilities of obtaining each result. This means that errors of given magnitude occur equally often with positive as with negative sign. . 7. and the sum of the errors is zero.23) to equation (7. Xi' . or dlnP _ 1 dP _ 0 dm-Pdm-· Introducing the condition that m is the arithmetic mean x. We can find an equation for the probability P(Xi .7 The normal distribution of error We assume random sign error in a set of independent measurements of an experimental variable x. P = np(xi-m) which will be a maximum when dPjdm = 0. this must be the arithmetic mean because L(xi-m) so that m = LXi-nm = 0 = Lxdn = x.(x 2 -x)+ .x)jd(xi ..X x 2 -x (7.174 Experimental error and method of least squares where (Zi .m) by maximizing the probability of obtaining a set of independent errors subject to the condition that m is the arithmetic mean.8) we obtain dlnp(xi -x) _ dlnp(x2 -x) _ d( ) +A..

or A(Xi .X) and 0' = = ± 1/ J A = ± 0'.x) = 00 !. from which K f: oo exp[ -A(X i -x)2/2]d(xi -x) K = =K J(n/2A) =! J(A.x)2/2]. This gives an integral of the standard form discussed in Example 5.J(A/2n)A(x i -x)exp[ -A(X i -x)2/2] 2 d p(xi -x)/d(x i _X)2 = J(A/2n)A[A(xi _X)2 -1] exp[ -A(Xi .1 = 0 (Xi .x)d(xi .x) = K exp [ .x)d(xi . The curve will have the shape shown in Fig. dp(xi .1.x)/d(xi .x) and integration gives P(Xi . Writing the first and second derivatives of P(Xi .The normal distribution of error 175 Hence.x) - x) to = 1. so that P(Xi -x) is a maximum at the arithmetic mean. din P(Xi .x) is symmetrical about Xi = x. This can be identified with the .x) = . The integration constant K is determined by normalizing p(x i unity. and the halfwidth ofthe curve can be expressed as the error at the point of inflection.A(Xi . 0' J 1 (2n) exp[ -(Xi-X)2/20'2] (7. P(Xi .A(Xi .1.j2n) and P(Xi .x) = J (A/2n) exp [ . so that 00 5-+00 P(Xi - f: x)d(xi .x)2/2].x)2/2].x) Then P(Xi. At Xi = x the first derivative is zero and the second is negative.X)2 . 7.A(Xi .x) = . as required. by suitable choice of A. say. this is when the second derivative is zero.26) is called the standard deviation.x) with respect to (Xi -x). or since P(Xi .

. . In general. . both X and Y will be in error.2.-J (27t) )n exp ( ..1 Linear relation between two variables Given that experimental variables X and yare related by the equation y = mx+c. A theorem that is important in practice is the central limit theorem.2 i i -x)2/20. (Xn' Yn). We can expect experimental measurements to follow this distribution so long as only random sign errors are present.L (X i2- 20- X)2). (7.2 ]d(Xi. Such averaging can occur in the use of measuring instruments. the means of which .X ) by Example 5. (Xi' Yi)..28) we wish to obtain the most probable values of m and c to fit a set of independently measured points (Xl' yd. which states that if mean values are taken of samples having any distribution. Y2). (x 2 . the probability of obtaining a set of results Xl' X 2 . 7..B. so that we then obtain a normal distribution of error even when the quantity being measured does not itself follow that distribution. . . This is the principle ofleast squares.8 The method of least squares Given a normal distribution of error. (7.26) describes the normal or Gaussian distribution of error. those means follow a normal distribution. •. . Xi is p = n[p(X i -x)] =( 1 0.176 Experimental error and method of least squares root-mean-square error o-(x) by 0-(X)2 = L+". 7. We assume a normal distribution of error for each variable at each point.'" (x i -x)2 p (x i -x)d(x i -x) = 2 L'" (x -x)2{1/[0-v'(27t)]}exp[ -(x = 0.. that the sum of the squares of the deviations is a minimum.27) This will be a maximum when L (Xi - X)2 is a minimum. Equation (7.

If we assume the standard deviation a(x) to be the same at each of the n points. 7. The observed point is then found at X.y.Yi)' i -Xi)2 +L(Yi-Yi)2)] )n exp [_(L(X2a(x)2 2a(y)2 ' which will be a maximum when a(y)2 L (Xi - XY + a(x)2 L (Yi . y (ii' Jij ) 6-. and similarly for a(y).29) In Fig. 7.x.30) . curves illustrating the distributions parallel to the x.YY is a minimum. (7.177 The method of least squares distributions lie on the straight line = mxi+c.- I? X ~-x xi I I x Fig.) and in the planes perpendicular to the paper and parallel to the axes.2. these curves should be interpreted as being centred on (Xi' y. An error in x displaces the point from the centre of its distribution to. Yi (7.2 When we assume independent random errors in x and in y.and the y-axes have been drawn displaced for clarity..) so that p_( - 1 21ta(x)a(y) = P(Xi -Xi)P(Yi . and error in y to the position of the square.Yi). the position of the triangle.) is the product of the probabilities P(Xi . we obtain P(Xi -Xi' Yi . the probability of obtaining point (Xi' y. say.) and P(Yi .

Yi)2 is a minimum with We put di = Yi . y) is calculated. When we assume error to be vested only in Y we have Xi = Xi' and condition (7.32) . = L(Yi-mxi-c)2 is a minimum. (7. If a group of n points from a chosen range of x is taken and the mean (x.mxi.c and call this the residual. to divide the points into groups at the upper and lower regions of the variables. therefore.mxi.178 Experimental error and method of least squares Further assumptions are necessary for the best line to be found because so far we have more unknowns than data points. this mean point will then have a smaller uncertainty than the individual points by the factor 1/ J n (equation (7. this is the deviation of the point (Xi' Yi) from the line measured parallel to the y-axis. We now regard D as a function of the two unknowns m and c.31) :C L(Yi. (7. D = Ld. calculate the mean point of each group. A solution of this equation will be (:~) = (~~) = 0 giving and since we may now omit the subscript i without causing confusion. We then assume the better known variable to be free from error.c) = 0. The least-squares condition is then that the sum of the squares of the residuals shall be a minimum. A recommended procedure when the errors in both x and Y are significant is. so that dD = (:~)dm+(~~)dC = o. Ly-mLx-nc = O.17)). and use these two points to define the straight line. Lxy-mLx 2 -cLx = 0. this being used as the independent variable x.mX i . In real applications the errors in the two variables are often not the same.c)2 = -2L(Yi.30) becomes L (Yi .

the slope m of the least-squares line Y = mx + c can be expressed in terms of u(x. c= (7.35) S(X)2 = [L(x -x)2]/n. y). y) ... y)of u(x.2 Covariance and correlation coefficient The covariance u(x.38) The values of m and c for the least-squares line will be uncertain because only a limited number of points are available. Y)/S(X)2.y) of variables x and Y is defined in Section 7.. S(X)2 -+ U(X)2 and L (y .33) (7.37) so that the estimated value of m is m = s(x. n n Hence for an infinite number of points U(y)2 = m 2u(x)2 and (7. As the number of points tends to infinity the uncertainty in the line tends to zero. (7. (7.LX~> L(x -x)(Y.The method of least squares 179 Solving (7.32) may be written as L(Yi-mxi-c) =0 we see that minimizing the sum of the squares of the residuals incidentally makes the simple sum of the residuals also equal to zero. u(x.31) and (7.. On the other hand. (x.x).. y) = mu(x)2 = u(x)u(y). if x and Y are linearly dependent.y) = [L(x -x)(y-y)]/n.34) shows that the least-squares line passes through the 'centre of gravity' of the points.36) s(yf = [L(y .y) . s(x. S(X)2 of U(X)2 and S(y)2 of U(y)2 are s(x. so that S(y)2 = L(y_y)2 = m 2 I(x-x)2 = m2s(x)2. and since (7. 7.38) gives u(x. y).39) . y). (7.32) simultaneously for m and c gives the equivalent forms of equation m = nLxy .34) Equation (7. For a finite number of points. m is given by (7..y) L(x -x)y = = .5 and shown to be zero if x and yare independent.8.33) and the estimates s(x. (7... (7.y)2]/n..-------'-'=nLx 2 -(Lx)2 I(X-X)2 L(x-xf LYLx 2 -LxLxy _ _ nLx2 _ (LX)2 = y-mx. y).. m L (x .

41) which may be written in the alternative forms 2 [~::<x .9). The error bYi is the difference between Yi and the value mX i + c given by the line.y)2 nCLY+ nmLxy .3 Uncertainty in the slope and intercept of the least-squares straight line Equations (7. This may be seen by plotting the graph. y) of the correlation coefficient. and this provides a test for such a linear relation.(Ly)2 nLy2 _(Ly)2 (n LXY . and the variables should then be changed to convert a curve into a straight line before applying the least-squares analysis. or that the supposed linear relation does not apply.. (7.y)]2 r = ~"'--'---.. y) = s(x.35}-(7.40) show that p = 1.~--=--.37) are not the same as (7.15).34) for m and c were derived assuming the errors to be vested entirely in the variable y.43) . (7. equations (7. When x and yare linearly related. 7. When this is applied to a finite number of points we can calculate only an estimate r(x. y)/s(x)s(y). it is not possible to separate these two possibilities for a given set of results. A value of r that is less than unity is therefore evidence either of limitations in the number and precision of the points...x)(y .. This will be zero if x and yare independent since u(x. y) will then be zero.-":-".LX Ly)2 (7. y) -----:--.39) and (7.33) and (7. L(X -.10) and (7. given by r = r(x.--~ u(x)u(y) (7.8.42) This will not be unity even when a true straight line exists because small-sample error means that expressions (7. (7.180 Experimental error and method of least squares The ratio p= u(x.. A correlation coefficient that is much less than unity does not imply independence of the two variables since it may be that a curve rather than a straight line fits the data.:.x)2L(Y .40) is called the correlation coefficient.

By combining equations (7. are derived from the data leaving (n .44) (7.46) = 0.33) and by propagation of error (Section 7. «(jm)l = -)2 ~( 2 2 2 x.x) assumed to be precisely known. and assuming this to be the same for all y..47) with (1'.17) (by)2 = (1'. Again assuming a mean value (1'. In which. = -+-=. for all values of y.2) in (7. given by equation (7.44) and . the error in the mean value of y is given by equation (7..5). together with (7.45) arises because two numerical coefficients._X)2]2 I(X.I(X. (xo .2) degrees of freedom. Writing the mean value of byf as (1'.-x (1'y = (1'y = nay [I(X.-X)2 n nIx 2 -(Lx)2 The uncertainty in the intercept c is then found by putting Xo (7.-'--:---=---'-.44).45). with (x. .X)2 n(1'.. When the least-squares line is used to estimate the value Yo of y corresponding to the value Xo of x we have Yo = y+m(x o -x) so that (bYO)2 = «(jy)2 + (xo _x)2«(jm)2.r2)[ n ~>2 (Iy)2] n(n -2) (7. (7. gives S:)2 (oyo = (1' 2 y (1-+ n (xo .-X)2 nIx 2 -(Ix)2 L. m and c.The method of least squares 181 From equation (7.X)2) (1'. where - (1 .45) The divisor (n .

The correlation coefficient for the points shown is r = 0. The error ± be in the intercept on the y-axis is due mainly to the second term in the bracket in (7. The least-squares line.46). 7.47).47) we obtain Example 7. is drawn with alternative slopes (b and e) of ± 15m. 7. The curves d and e are drawn to show the errors ± 15Yo at particular values of Xo as given by equation (7. a. y) has error bars at ± a y/ -J n. it can be seen that as Iowa value as this indicates a considerable scatter in the points. corresponding to uncertainty in the slope.1 The result of applying this analysis to a set of points is shown in Fig.182 Experimental error and method of least squares (7.3. y x Fig.3 .775. The mean point (x. with an additional contribution from the first term which corresponds to the uncertainty in y.

and will coincide for an infinite number of points if the linear relation applies so that p = 1. whereas m' is dx/dy.50) As the number of points tends to infinity. having used the same uncertainty (1 y for all values of y..49) I S(y)2 (y .5 Weighting of observations We have assumed so far that all experimental points are equally reliable.=-. s(x. The ratio of the slopes of these lines is related to the estimate r of the correlation coefficient. we would then obtain a different straight line by interchanging x and y in the equations.= mm .x)(y .----'---'(7. The simplest way of achieving this is to repeat the . y) and bisects the angle between the alternative lines. be experimental reasons for regarding some points as more reliable than others. the estimated correlation coefficient r tends to the value p free from small-sample error. s(xf s(yf (7. we therefore compare m with 11m'.= =--.--~--..33) and (7. y) I (x . this can be included in the analysis by weighting the points.1 assumes that the error is vested in only one of the two variables and equations (7. otherwise we might equally well assume the error to be vested only in x.The method of least squares 7. y)2 2 = = r .8..38) s(x. The ratio of the slopes is then m 11m' -.4 183 Least-squares straight line with both variables subject to error The calculation of the least-squares line in Section 7. 7. Since the least-squares line passes through the 'centre of gravity' of the points. the two lines obtained by assuming the error to be vested either in y or in x will intersect at that point.y) m = .34) for m and c assume the error to be in y.y)2 Now m is the slope dyldx.48) We write the alternative line as x = m'y + c' and by interchanging x and y we obtain . It is conventional to assume that if the errors in the two variables are both significant the best fit to the data passes through (x. however. The alternative lines will then move closer together as the number of points is increased. s ( x .8.y) m = S(X)2 = I(x _X)2 (7. with slope equal to the mean of m and 11m'. y) I(x -x)(y .8. Experimental conditions may justify this assumption. From equation (7. There may.

The proper weight to be used in precise work can be seen from the expression for the normal distribution. some of which make use of expressions for the partial derivatives with respect to the coefficients to accelerate convergence.184 Experimental error and method of least squares entry of any particular point into the analysis w times. the analytical methods of this chapter are still applicable and remain useful in simple cases. 7.27). It is necessary to remember only that these weighting factors must be included in an effective total number of points. equation (7. . thus making that point w times as significant as a single point. Partial differentiation of L: (d f) with respect to each numerical coefficient in turn gives a set of normal equations. so that each point should be weighted by the reciprocal of its standard deviation. Standard programs are available for this purpose.6 Multivariable and non-linear least-squares analysis When fitting results to a more complicated equation than that of a straight line. Assuming the error to be vested in a particular variable.8. such as might be estimated by the formula for propagation of error. It is important when using numerical methods to allow only the smallest possible number of search parameters. When these equations are linear in each coefficient the solution is straightforward. which can then be solved. an excessive number will produce an undulating curve or surface which may fit the data precisely but be physically misleading. the difference between an experimental value of the variable and that predicted by an assumed equation is the residual d j • The principle of least squares is that L: (d f) is a minimum. for which a text on statistical analysis should be consulted. An alternative to the analytical approach is to use trial-and-error searching by computer to find values of the coefficients that produce the smallest value of L: (d j )2. A polynomial fit in which the coefficients alternate in sign may be evidence of this effect. otherwise linearization may be achieved by using trial values of the coefficients. the quantity to be minimized is the sum of the squares of the deviations divided by (12.

2. Energy is measured in joule (J). the Greek alphabet and a summary of useful relations SI units SI stands for the agreed international system of units. The kelvin is defined as 1/273. The mole is the amount of substance containing as many elementary units as there are atoms in 0.APPENDIX SI units. without changing the triple point. Pressure is then force per unit area in N m . This means any subsequent refinements in absolute temperature measurement will alter the size of the degree and so. and of the gas constant from measurements of the properties of gases. counting this number of atoms. The ice point is 0. . The unit offorce is the newton (N).012 kg of 12c. together with the kelvin (K) unit of temperature. say. which is based upon the kilogram (kg).16 of the thermodynamic temperature of the triple point of water. metre (m) and second (s). which is 1 N m.01 K below the triple point so that O°C is 273. and power in watt (w) which is 1 J s . the mole (mol) for amount of substance and the ampere (A). physical constants and conversion factors. defined as that which produces an acceleration of 1 m s . The candela (cd) unit ofluminous intensity is also a basic unit.15 K. in effect. this unit being called the pascal (Pa).1. Other units are derived from these basic units. This allows experimental measurement of the Avogadro constant by. the absolute temperature of the boiling point of water.2 in a mass of 1 kg.

1 (1979) are: Avogadro constant L Boltzmann constant k gas constant R Planck constant h Faraday constant F mass of proton charge on electron e mass of electron speed of light in vacuum permittivity of vacuum normal atmosphere zero of Celsius scale standard acceleration of free fall 6.23 J K.186 SI units.1 8.1 m. usually the thermochemical calorie of 4. The volt (V) is then the potential difference such that a current of 1 A dissipates 1 watt. with the recommendation that neither the word litre nor its symbol. conversion factors The ampere (A) is that current which produces a force of 2 x 10.109534(47) X 10. this definition being rescinded in 1964.6021892(46) X 10. A.380662(44) x 10. 1.85418782(5) x 10.022045(31) X 1023 mol. Other units that are still in use are the calorie in various forms.184 J. which is 10. . this is 101.2 and a standard atmosphere is the pressure produced by a mercury column 0. This is because the litre was previously defined as 1. and the Angstrom.2 exactly The notation used is that the figure(s) in brackets is (are) the standard deviation of the last figure(s) quoted.325 kPa. Physical constants Recommended values of the fundamental physical constants are published by the International Union of Pure and Applied Chemistry. Some of the values given in Pure and Applied Chemistry. The faraday constant (F) can then be measured as the number of coulombs of charge carried by one avogadro of electrons.80665 m s .1 6.6726485(86) x 10. defined as 100 kPa. and the coulomb (C) is 1 A s. physical constants.000028 dm 3 .34 J s 9.76 m high under standard gravity at ODC. The preferred unit is the bar.1 mol.648456(27) X 104 C mol. Standard gravity is 9.15 K exactly 9.19 C 9.01325 x 105 Pa exactly 273.2 1.99792458(1) X 108 m S-l 8.31 kg 2. should be used to express results of high precision.626176(36) x 10.27 kg 1.1 1.1 1.10 m. 51.31441(26) J K.80 665 m s . The word 'litre' is now regarded as a special name for the cubic decimetre.12 C 2 N.7 N m -1 between long straight parallel conductors 1 m apart in vacuum.

7 J 1 farad = 1 F = 1 A S V-I = 1 m .325 kPa 1 cal = 4. conversion factors 187 Conversion factors for units 1 amp = I A = I J s .Sf units.I 1 atm = 101.2 =lm.3 A-I I watt = 1W = 1J s .5 N 1 electron volt = 1 eV = 1.I kgs. physical constants.30 debye = 10.6021 x 10.184 J 1 coulomb = 1 C = 1 A s = 1 J V-I 0.6 m I newton = I N = 1 m kg s .1333 kPa 1 volt = 1 V = 1 J A -I S .2009 gallon (USA) 1 inch = 2.I V . (j E '1 0 K ).546 litre = 1.540 cm 1 joule = I J = I N m 1m 3 = 106 cm 3 1 micron = 1 p. p.2 1 pound (UK) = 1 Ib = 0.2 kg .4536 kg I torr = 1 mmHg = 0.30 C m 1 dyne = 10.I = 1 m 2 kgs. gamma delta epsilon zeta eta theta iota kappa lambda mu N 0 v ~ 0 n 1t P 1: (J p T T Y <l> u X 'I' n 4> X '" OJ nu xi omicron pi rho sigma tau upsilon phi chi psi omega .3 The Greek alphabet A B r i\ E Z H e I K i\ M IX alpha p beta y .I = I m 2 kg s .I S4 A 2 1 gallon (UK) = 4.m = 10.19 J 1 erg = 10. = I p.2 lpascaI=1 Pa=INm.

dy du du dx = (!:)Y dX+(.. (1 + xt +.. x = --±.. physical constants..2 x 2 + . conversion factors 188 Summary of useful relations cos 2 A + sin 2 A = 1 cos 2 A . XS 5.... · f(x) 11m .)x dy CX) .3026 log x = 2. y)..() x~a 9 X I' (x) ·ff() () 0 = 1·x~a 1m ~() 1 a = 9 a = or 9 X z = z(x.sin 2 A = cos 2A sin 2A = 2sinA cos A sin (A±B) = sin A cos B±cos A sin B cos (A ±B) = cos A cos B+ sin A sin B eXeY = eX+ Y a-X = 1jaX if y = eX then x = In y In x = loge x = 2.1) 2! x 2 + . = 1 + nx + n(n .. (a+xt=an+nan-1x+ f(x) x2 = f(O) + xl' (0) + 2! f" (0) + .30261og 10 X log 1o lO-x = -x b 1 ax 2 +bx+c = 0. dz dy dx n(n -1) 2! an...SI units.j(b2 -4ac) 2a 2a .. SIn x x3 = x -3-'.

:)y = -(. conversion/actors (..~)% f udv = uv - e iy f vdu = cos y + i sin y e.)x (.iy = cos y -i siny 189 .SI units. physical constants.

71. 73 reciprocal of. complex. 25 Argument of complex number. 122. 121 Damped harmonic motion. 173. 118 Area by integration. 26 Asymptotes. 136 Continuous function. 80. 74 second. 67. 182. 70 rectangular cartesian. 31. 39 Circle. 18. 60. 56 Boltzmann distribution. 33 Convergence of series. 75 partial. 25 Complex numbers. 70 Covariance. 30. 21 spherical polar. 21 Approximation. 74 . 67. 129 Central limit theorem. 180. 121 Correlation analysis. 169 Correlation coefficient. 153 Damping. 46. 171. 82. 75. 1. 119 differential element of. 117 Argand diagram. 10. 91 Anharmonicity. 33 total. 133 Angular momentum. 24. 116 Complex conjugate. 34 Conjugate. 154 exponential form. 121 plane polar. 21 cylindrical polar. 176 Centrifugal force. 53. 55. 183 Critical damping. 118. 150 rectangular. 67. 145 Curl of vector field. 171 Derivative mixed second. 50-61 Coordinates cartesian. 10. 17. 32. 149. 144. 27 Concave upwards. 12. 99. 97 Curvature. 154 Dependent and independent variables. 179. 40. 12 Action. critical. 43 Cross differentiation. 135 Angular velocity. 62 Axes principal. 84 total derivatives. 31. 48 Calculus of variations. 25. numerical. 21 Binomial series. 21. 183 Couple. 184 Cylindrical polar coordinates. 115. 96. 135 Chain rule partial derivatives.Index Abscissa. 48. 33 Curve fitting. 95 Conservative field. 154 Critical point. 91. 97. 57 Arc length.

140. 186 Gaussian distribution of error. even and odd. 171 systematic. 144. 132 Exact differential. 18. 152 Direction cosines. pair. 10. 122. 36. 147 linear. 70. 91 Discrete variables. 12. 83. 86. 9 quadratic. 132 Generalised momentum. 64 Independent variable. 10. 150 Helmholtz free energy. 63. 138 exact. 11. 16. 22. 24. 166. 82. 154 Implicit differentiation. 31 Differential exact. 48. 132 Harmonic motion damped. 1. 65. 86. 166 Generalised dynamics. 144 Exponential functions. 66. 12. 51 Gibbs equations.27. 65 scalar and vector. 82. 134 principal moments of. 171 Indices laws of. 88 Gibbs function. 8 Functions. 149. 168. 97 Dynamics generalised. 152 Enthalpy. 89.153 forced. 37 Impulse. 154 capacity. 132. 100 of vaporization. 127 perfect. 152 gravitational. 75 of integrals. 17 Equilibrium constant. 145 implicit. 158 simple. 107. 152 generalised. 3 Miller. 34 Dimensional analysis. 68. 83. 136 Geometric series. 6. 137 Fourier transform. 145. 71. 75 higher order. 140. 169 small-sample. 94.62 Hyperbolic functions. 7. 17 End point maximum. 161 Field electric. 165 Euler-Maclaurin theorem. 88 Gradient of scalar field.Index Differentiable function. 64. 127 Force. 63. 161 Gas real. 70 product of. 66. 95 Graphical methods. 132 Newtonian. 95 First order reaction. 33 Gamma function. 154 kinetic and potential. 8. 25 root. 39 Hamiltonian mechanics. 144 first order homogeneous. 12 Hyperbola. 38 Extensive and intensive variables. 18. 39 cross. 163 Functional notation. 59. 69 . 109 Euler's theorem. 127 Divergence of vector field. 176 root-mean-square. 36. 140 separable variables. 100 Equations cubic. 65. 8 second degree. 91 Heat. 148 order and degree of. 22. 80. 5 Dipole moment. 166. 10. 66. 172 normal distribution of. 124 Gas constant. 78 Differential element of area. 6. 117 of volume. 140 Differentiation by substitution. 94. 120 Differential equations. 42 Energy heat. 55 Imaginary number. 86 191 Factorial. 12. 80. 36. 17 Inertia moment of. 17. 85. 12 Gyroscopic motion. 57 Error in arithmetic mean. 97. 142 linear. 144 total. 174 random sign. 13 Fluids. 166. 86 Euler's equation. theory of. 108 rules for. 82. 140. homogeneous. 24. 168 Distribution function. 63 Ellipse.

120. 25. 38 Newtonian mechanics. 119 by partial fractions. 186 Polar coordinates cylindrical. 30 expansion in series. 119 curve length by. 61. 160 Least squares. 99. 120 Limit calculus. point of. 152 Kinetic theory of gases. 38 Logarithmic series. 179 weighted. 129 subject to constraint. 120 by using polar coordinates. 124 pH. 150 spherical. 85 Kinetic energy angular. 1. 11 Planck's constant. 105. 72. 86 Mean arithmetic. 76. 12 Orthogonal. 122 Logarithm. 65. 68. 73 chain rule for. 40 Line integral. 10. 123 L'Hopital's rule. 45. 122 Intercept. 56 Maclaurin's theorem. 94. 68. 120 by substitution. 102. 21 Propagation of error. 64. 83. 139. motion of. 119. 6. 34 nabla. 115. 56. 70 Momentum angular. 133. 124 Nabla operator. 125 Multiple integration. 174 Projectile. 124 replacing summation by.95 Parabola. 35. 183 straight line. 112 Parts. 170. 174 Lagrangian mechanics. 152 Power series. 58 Joule-Thomson coefficient.92. 83. 49. 95 vector. 124 Kinetics. 168. 7. 178. 174 Numerical quadrature. 68.153 Moment of a force. 132. 78.21 Partial derivatives. 96. 107. dependence on.164 Normal distribution of error. 16. 162 Partition function rotational. 80. 63 L'Hopital's rule for. 23. 91. 14 Inversion of a series. 13. 100. 66. 70 Probability. 70. 118 of curve. 122 Perfect gas. 98 area by. 65. 26 Molecular properties. 66. 21. method of. 166. equations for. 184 error in both variables. 6. 118.92. 120 Path. 71. 18. 41 Integration. 91 Principal axes of molecules. 51. 132 Laplace transform. reaction. 74 relations between. 75. 146 Lagrange's method of undetermined multipliers. 11.134. 71. 102. 17. 170 Partial fractions. 123 multiple. 108 Length of arc. 105 by parts. 135 linear. 13. 113 volume by.60 Precession of gyroscope. 112 Operator differential. 63 Newton's method. 166 error in. 95 Ordinate. 117. 42 Phase rule. 84 reciprocal of. 44. 120 of sum or product. 48. 84 Taylor's theorem in.192 Index Inflection. integration by. 67. 183 Leibnitz's Theorem. 72 linear. 17 Modulus of complex number. 113 translational. 41 calculus of variations. 54 Maxima and minima. 121 Potential energy. 95 Napierian logarithm. 107. 181 . 61. 117. 53 NMR. 88. 104. 13. 45 Maxwell's relations. 176. 70 Moment of inertia. 172 Miller indices. 86. 121 plane. 134 principal.

12. 60 sine. 170 Torque. calculus of. 39 integration by. 55 exponential. 41 Units. 166 Taylor's theorem. 176 Reaction kinetics. 2. 68. 178. 25. 21 Systematic error. 92 Series. 80. 136. 89 product. 72 sign of. 154 Rotation dynamics. 129 193 Stirling's theorem. 159 Resultant. 7. 125. 150 damped. 30. 90 localized. 49 Stokes's law. 51. 184 Resonance. 152. 121 Standard deviation. 4 Straight line equation of. 8 imaginary. 50-63 binomial. 83. 55 SI units. 139 Residual. 70. 58 logarithmic. 39 Total differential. 10. 122 Thermodynamic properties. 185 Similar triangles. 170. 96 . 66. 90 differentiation of. 23. 179 Substitution differentiation by. 14 Simple harmonic motion. 86 Variance. 89 Root-mean-square error. I. 171 Variations. 65. 44. 104. 78 Transformation. 165. SI. 153 forced. 2. 171 extensive and intensive. 72. 45. 13 least-squares. 53 substitution in series. 12. 3. 139 Total derivative. 91 operator. 92 Volume by integration. 38. 13. 50--61 cosine. 158 Simultaneous equations. 87 Time constant. 106. 18. 56. 11. 120 Successive approximation Newton's method. 109. 41. 3. 10. 4. 89 components of. 128 Spectroscopy.24. 166. 119 Stationary values. 168. 51 inversion of. 55 sinh and cosh. 169. 74 chain rule.48. 21 Scalar. 88. 172. 83. 176. 155 Rectangular coordinates. 166. 94 Thermodynamic function of state. 162 Radian. 149.Index Quantity calculus. 83 Thermodynamic relations. 173 Standard forms of integral. 85. 160 Triangular graph. 181 Root of an equation. 122 Weighting in least-squares analysis. 129 Vector. 59 Symmetry. 146 Real and imaginary parts. 46. 21. 27. 120. circular. 168. 45 Single-valued functions. 95. 67 Random sign error. 32 Solid angle. 51. 95 product. 36. 100. 185 Variables dependent and independent. 56 convergence of. 15 Trigonometric identities.61. 31. 56 power. integral. 139. 149. 83. 4. 121 Reduced mass. 164 Spherical polar coordinates. 188 Turning point. 183 Work. 134 Relaxation time. 148 in partial derivatives. 5 Quantum theory. 55 geometric.

- ME 251 Lecture 1Uploaded bySeth Ankamah
- Classical Physics '10Uploaded byAhmad Munjin
- CHECAL_CI1MultipleChoiceUploaded byJelain Humarang
- Smithsonian Physical Tables 9th Revised EditionUploaded byadelbadr
- Syllabus PCSUploaded byWaleed Mallick
- einstein.pdfUploaded byBauEngenharia
- Dimensional AnalysisUploaded byDhanyaUnni
- Chap 01 SMUploaded byuser0056
- question on dimension.docxUploaded byMantu Das
- The Fairy Doll La Poupee Ued20.PDFUploaded byElíGomaraGil
- 01_B UnitsUploaded bydeelol99
- Arihant Aieee Physics [Nitin m Sir]Uploaded byaashray18521
- Consequence Analysis by Means of Characteristic Curves to Determine Damage to Buildings From ExplosionUploaded byuetian7296931
- Dam Busters 2 SolutionUploaded byfeiphung
- Arihant AIEEE PhysicsUploaded bybansalcomics
- Differential Equations HandoutsUploaded byEllimacOdnumreb
- Just the Maths - A.J.hobson (Complex Numbers)Uploaded byKriss_Ps
- Assignment MA02 2017Uploaded bysandip kumar
- Full SyllabusUploaded bySanthana Bharathi
- 2016_Lecture-Physical Modeling (1)Uploaded byola
- Lecture_Ch01 Introduction, Measurement, EstimatingUploaded byfafa2019
- Ratna Output SpssUploaded byOkta Manus
- Bt Btech SyllabusUploaded byadddata
- IntroductionUploaded byFarah Aliah
- 3. IntegrationUploaded byBenjamin Hi
- Current_Syllabus_for_B.Arch_and_BArch(ID).docUploaded bySoumyaJain
- Wallis RuleUploaded byJunMonteron
- Ae Notification-tsspdcl - FinalUploaded bysandy
- ac1Uploaded bybudo ako
- A Matlab Environment for Analysis of Fluid Flow and Transport Around a Translating SphereUploaded byendrajuwana

- Filosofía de VidaUploaded byAngel Tec López
- verdades no demostrables.pdfUploaded byAngel Tec López
- When is 0.999... Less Than 1.pdfUploaded byAngel Tec López
- Signo de una Permutación.pdfUploaded byAngel Tec López
- CycloidUploaded byAngel Tec López
- La Anarquía Explicada a los NiñosUploaded byAngel Tec López
- Capital Cities of the WorldUploaded byAngel Tec López
- Las matemáticas no me han servido para nadaUploaded byAngel Tec López