Trova il tuo prossimo libro preferito

Abbonati oggi e leggi gratis per 30 giorni
Mathematical Methods for Physicists and Engineers: Second Corrected Edition

Mathematical Methods for Physicists and Engineers: Second Corrected Edition

Leggi anteprima

Mathematical Methods for Physicists and Engineers: Second Corrected Edition

746 pagine
10 ore
Jun 11, 2012


This practical, highly readable text provides physics and engineering students with the essential mathematical tools for thorough comprehension of their disciplines. Featuring all the necessary topics in applied mathematics in the form of programmed instruction, the text can be understood by advanced undergraduates and beginning graduate students without any assistance from the instructor.
Topics include elementary vector calculus, matrix algebra, and linear vector operations; the many and varied methods of solving linear boundary value problems, including the more common special functions of mathematical physics; the calculus of variations, and variational and perturbation approximations applicable to boundary value problems and nonlinear differential equations; curve fitting and numerical approximation methods; the basic elements of probability and their application to physical problems; and integral equations.
Rather than aiming at a complete mastery of these complicated subjects, the text focuses on the fundamental applied mathematics the student needs to deal with physics and engineering problems. Instructors in those subjects will particularly appreciate this volume's function as a self-contained study resource, allowing them to devote fewer classroom hours to formal lectures in mathematics.
Jun 11, 2012

Informazioni sull'autore

Correlato a Mathematical Methods for Physicists and Engineers

Titoli di questa serie (185)
Libri correlati
Articoli correlati

Anteprima del libro

Mathematical Methods for Physicists and Engineers - Royal Eugene Collins

Mathematical Methods for

Physicists and Engineers

Second Corrected Edition

Royal Eugene Collins


Mineola, New York


Copyright © 1968, 1999 by Royal Eugene Collins

All rights reserved under Pan American and International Copyright Conventions.

Bibliographical Note

This Dover edition, first published in 1999, is an unabridged and corrected republication of the first edition of Mathematical Methods for Physicists and Engineers published by Reinhold Book Corporation in 1968.

Library of Congress Cataloging-in-Publication Data

Collins, Royal Eugene.

Mathematical methods for physicists and engineers / Royal Eugene Collins. — 2nd corr. ed.

p.   cm.

Includes bibliographical references and index.

eISBN 978-0-486-15012-3 (pbk.)

1. Mathematics. 2. Mathematical physics. 3. Engineering mathematics. I. Title.

QA37.2.C656 1999




Manufactured in the United States of America

Dover Publications, Inc., 31 East 2nd Street, Mineola, N.Y. 11501


In preparing this text I have drawn heavily on a great number and variety of texts on applied mathematics. Many are cited as desirable references for the student. I am indebted to all of these authors as my teachers. I am also indebted to my students over the years who, by their responses, have taught me how to teach this material in a more effective way each year.

I express my appreciation to Mrs. Gloria Parr and Mrs. Shirley Double for their patience with my poor handwriting in typing this manuscript, and to my son, Mr. Roy Collins, for his skillful preparation of the illustrations.



The physical sciences and engineering have become increasingly mathematical during the 20th century, and indeed the same trend is also observed in the biological sciences.

Advanced undergraduates, or beginning graduate students, in these fields find themselves in need of a certain mathematical tools in order to comprehend the physical principles, and their applications, encountered in their major courses. Here we emphasize mathematical tools rather than mathematics because the student of physics or engineering must use these elements of mathematics as tools to achieve the desired understanding of the physics rather than simply comprehend the mathematics itself. This distinction is important.

The mathematical tools most urgently needed, beyond a one-semester course in ordinary differential equations, as offered in the mathematics departments of most colleges and universities, are as follows: a knowledge of vector calculus, matrix algebra, and linear vector operations; the many and varied methods of solving linear boundary value problems, including a good familiarity with the more common special functions of mathematical physics; some knowledge of the calculus of variations, and variational and perturbation approximations applicable to boundary value problems and nonlinear differential equations; the ability to evaluate complicated integrals, or sum series; a facility for curve fitting and numerical approximation methods; a familiarity with the basic elements or probability and how these apply to physical problems; and, finally, some knowledge of integral equations.

This is a formidable list, and if the student were to try to completely master these many areas form the mathematician’s point of view, it would be more than a lifetime task. However, the essential needs of the student can be met without such complete mastery.

Most universities attempt to meet this need with a one-year course in the mathematics department, usually called Mathematics for Engineers and Physicists and commonly taught in the student’s junior year. However, rarely is even a small part of the above list actually treated in the course. As a result, a great amount of time must usually be devoted to the teaching of mathematics in each and every course in physics or engineering. This situation exists partially as a result of the form of most texts available for such a course and partially because of the method of presentation in the mathematics class.

The text presented here is intended to help remedy this situation in two ways: first, by including an adequate treatment of all the needed topics in applied mathematics; and second by presenting the material in the form of programmed instruction. The text has been designed to be entirely readable so that the material can be comprehended by advanced undergraduate, or beginning graduate, students without any assistance from the instructor. Consequently, class time can be devoted almost entirely to reviewing the problems executed by the student, thus dispensing with formal lectures on the mathematical methods.

It is recommended that the student be assigned a segment to read along with a corresponding set of problems to solve. These can then be reviewed and discussed in the following class session. Occasionally, the instructor may find it necessary to work out some of the more difficult problems or to enlarge on a particular topic with additional examples. Further, some of the text segments and problems sets are rather long and may require two or three class sessions to cover adequately. Nevertheless, it should be possible to cover all of this text in two semesters if this method of teaching is used. The author has found it to be most effective over a period of several years using essentially the material of this text as prepared notes issued to students.

Some of my mathematician friends may question the lack of rigor in this text, but they must remember the objective of the text and forgive me. The intent is to eliminate most of this same material from present physics and engineering courses by putting all of it into one course. The mathematician may wish to call the course something other than Mathematics; I prefer Mathematical Tools. This compaction of applied topics will also leave physics and engineering students free to include in their curricula additional courses in pure mathematics, subjects such as point set theory, topology, and others that are strongly recommended, especially for physics students. In addition, this method will allow students to concentrate on physics, or engineering, in their major courses. The end result is that courses in physics and engineering will be stronger, and students better educated.

R. E. Collins



1    Elementary vector calculus; the vector field

2    Matrix algebra and transformations in linear vector spaces; dyadics

3    Introduction to boundary value problems and the special functions of mathematical physics

4    Useful properties of some special functions of mathematical physics

5    Solution of linear homogeneous boundary value problems; separation of variables method and eigenfunction concepts

6    Elementary applications of the Laplace transform

7    Two-dimensional potential problems and conformal mapping; functions of a complex variable

8    The calculus of residues

9    Integral transforms; the solution of inhomogeneous partial differential equations

10  Inhomogeneous boundary conditions; Green’s functions

11  Introduction to integral equations

12  Variation and perturbation methods; introduction to nonlinear differential equations

13  Elements of probability theory

14  Miscellaneous topics: evaluation of integrals, summation of series, curve fitting, transcendental equations




Elementary vector calculus; the vector field

Here we present a brief but essentially complete presentation of those elements of vector calculus essential for the physicist or engineer. A knowledge of the elementary algebra of vectors is assumed.


We consider a point in space defined with reference to a set of rectangular cartesian coordinates by the three numbers, x1, x2, x3, which are the coordinates of the point, or the orthogonal projections of the point, on the three† axes. The directed line from the origin to the point is the position vector of the point. We define the base set of unit vectors on these axes as 11, 12, 13 where‡

Then we write,

as the representation of the position vector on these axes.

Now the differential displacement vector is just given by,

since these unit vectors, 1i, i = 1, 2, 3, have the same magnitude and direction at all points of space, i.e., dr is constructed as,


, the unit vectors factor out and Eq. (1.3) results.

The magnitude of dr is the line element, ds. This can be constructed as


by virtue of Eq. (1.1).

Now we do not always use rectangular orthogonal axes. Thus we now examine what happens to dr and ds under some arbitrary change of variables,


and these can be inserted into Eq. (1.3) for the dx’s and the order of summation interchanged to give,

In these equations the dq’s may have any dimensions, or units, but the x’s must all have the same dimensions in order to maintain dimensional homogeniety of our equations. We also note that the quantities in parenthesis in Eq. (1.10) are vectors in the rectangular space, but having components, ∂xi/∂qj, which are also not dimensionally homogeneous. Thus now multiply and divide each term here by a factor, hj, so that we have,

and require that the hj be chosen so that the factors hj dqj all have the same dimensions as the xi.

Furthermore we now note that the vectors forming each parenthesis above, denoted now as,

are dimensionless and will be unit vectors, if we require,

for each j. This condition then determines the hj in terms of the coordinate transformation by the relations,

so the hj are fixed except for sign by these equations†.

We now have the expression for dr in the form,

and here let us again look at the square of the line element, ds². We have,

This is distinctly different from our Eq. (1.7) for the rectangular system but it reduces to a very similar form if we require,

This states that these new unit vectors are mutually orthogonal, or perpendicular. Hence for any orthogonal coordinate system (i.e., Eq. (1.17) satisfied) we have,

and it is just such systems of coordinates that we will nearly always employ.

Note that the criterion of orthogonality, Eq. (1.17), has the form,

if we make use of vectors.


The elements hj, j = 1, 2, 3, introduced above are called the scale factors of the curvilinear coordinate system, qj. Here we note how area elements, and volume elements, are expressed in terms of these hj and the dqj.

Consider Eq. (1.15) for dr and choose dq2 = dq3 = 0, then we have a vector displacement we will call dr1 given by

, for dq1 > 0, and is a line element of length h1 dq1, from the definition of dr. Similarly if at the same point in space we let dq1 = dq3 = 0 we form another line element,

, dq2 > 0, and of length, h2 dq2.

The vector product, or cross product, is,

which in magnitude is the area of the parallelogram indicated in Fig. 1-1, with

FIGURE 1.1. Element of Area on the q1q2 Plane.

Obviously if the coordinate system is orthogonal, sin θ = 1. Also we note that the direction of this vector , i.e., from our definitions above we have the cyclic rule

just as for the original unit vectors.

If we now form the scalar, or dot, product of a line element,

with dA3 we obtain,

as the volume of the elemental parallelopiped. Obviously, for orthogonal systems this is just

and we will restrict our consideration to such systems. Here h1h2h3 is the Jacobian of the transformation from the xi to the qj.


We call attention to the fact that we generally require that it be possible to carry out the inverse transformation,

so that,

Substitution of these expressions for the dqj in Eq. (1.9) shows that the necessary conditions for this inverse to exist are,

Also, in order that we should be able to solve Eq. (1.9) for the dqj as functions of the dxi, it is required that the determinant of the coefficients should not vanish, i.e.,

These ideas become more concrete when we consider an example.

Example (1)

We here examine the spherical coordinate system in terms of the general ideas outlined above. The equations of transformation from the rectangular system, now denoted as x, y, z, to the spherical system, r, θ, ϕ, are:

as the equivalent of Eq. (1.8).

The scale factors are


which reduce to,

if we take the positive square roots.

Next we verify that the r, θ, ϕ system is an orthogonal system, i.e.,

and similarly with (r, ϕ), (r, ϕ) replacing (r, ϕ) here. This does indeed prove to be true for all these combinations as required by Eq. (1.14).

Having constructed the scale factors and shown that the r, θ, ϕ system is an orthogonal system we know from our discussion above that we can form a unit vector 1r in the direction of increasing r at any point in space, and similar vectors, 1θ and 1ϕ, in the directions of increasing θ and ϕ respectively, at this same point. These, 1r, 1θ, 1ϕ form a local orthogonal rectangular set of axes, i.e., as depicted in Fig. 1-2.

FIGURE 1.2. Unit Vectors for the Spherical Coordinate System.

However, as can be seen from the equations defining these unit vectors, 1r for example,

formed as prescribed by Eq. (1.12), is not a constant vector. That is, although its length is always unity, its direction varies from point to point in space. For example we can differentiate 1r with respect to θ or ϕ and the result is not zero, whereas all derivatives of 1x, 1y, and 1z are zero.

From our general formulation we see that, for example,

is an element of area oriented in space such that the vector r from the origin is orthogonal to this area, i.e., this is the magnitude of the element of area on a spherical surface of radius r at the point r, θ, ϕ, and directed outward from the surface.



We consider some continuous function, F(gj), of a set of orthogonal coordinates, qj, j = 1, 2, 3, having continuous derivatives; at least continuous first derivatives. The total differential of this function is given by,

and represents the difference in the values of F at the two points [(q1 + dq1), (q2 + dq2), (q3 + dq3)] and (q1, q2, q3).

Note that if we multiply and divide the first term by h1, the second by h2, and the third by h3 this appears as,

and this has the form of the scalar, or dot product, of the displacement vector, dr, having components hi dqi, with another vector,

which we call the gradient of F, or just "grad F". Thus we see

in any orthogonal coordinate system, with ∇F defined by Eq. (1.42).

Note that the magnitude of the gradient, |∇F|, is

Hence we can write Eq. (1.43) above as,

since ds = |dr| and we here define γ as the angle between the direction of ∇F and dr. This follows from the definition of the scalar product. Since |∇F|, being just a function of the qj, has a fixed value at any point in space, and we can choose some fixed magnitude for the displacement ds, we ask in what direction must we make the displacement in order to achieve maximum change in F, i.e., dF = maximum. Quite obviously this occurs for γ = 0, or cos γ = 1. Thus the gradient vector is always directed in the direction of the maximum space rate of change of F.

Also observe that dF = 0 corresponds to γ = π/2, but since dF = 0 defines the surface of constant F, we see that the vector ∇F is always at right angles to the surfaces of constant F. This is indicated in Fig. 1-3.

FIGURE 1.3. Surfaces of Constant F(x, y, z) and Their Relation to the Gradient Vector, ∇F.



Just above we saw that we could form a vector, ∇F, from some scalar function, F(qj) in a certain way. The components of this vector were each a function of the coordinates. We call such a vector a vector field. A vector field assigns, at each point of space, a vector, say V(r), whose direction and magnitude will in general be different at each point, r, of this space.

Such vector fields may be formed in many ways, the gradient field is just a very special case. Thus we write for the general vector field,

where the components, Vi(r), are distinct functions of the coordinates which we here indicate in the general way as the position vector, r.

Now we visualize some surface, S, in this space as indicated in Fig. 1-4. We choose a positive side for the surface, then subdivide it into infinitesimal elements of area, ΔSi. At the center of each such element we erect a unit vector, ni, normal to the area and, in this way, have the vector element of area

such as for element number 1 in our picture.

Also at the center of each element we evaluate the vector field, i.e., V(r1) for the first element in our picture, then form the sum, IN,

over all such area elements making up S, here assumed to be N in number.

FIGURE 1.4. Subdivision of a Surface for Construction of the Surface Integral of a Vector Function.

In the formal limiting process we begin with small elements ΔSi then show that as all ΔSi → 0 and N → ∞, in any way the above sum approaches a unique limit. In such fashion we define the surface integral of V over S as,

Example (2)

Compute the vertical force on a hemisphere, whose base is horizontal, when the external curved surface, above this base, is exposed to a uniform pressure P.

Since by definition of the pressure the force on an element of area dS on the surface, with outward normal, is – P dS, we have as the vertical component of this force

Then since in spherical coordinates

we have, using Eq. (1.38) for the representation of 1r,

Integrating and observing that both P and r² are constant and can be brought outside the integral, we obtain

so that

where the minus sign indicates the resultant is in the direction of the –z axis.



Suppose we have given a vector field, V(r), and select a volume τ bounded by a surface S. Then we can form the integral of V over the closed surface S and divide the result by the in closed volume τ. If V(r) is a continuous vector function in the infinitesimal neighborhood of a point r then the limit

will exist.† Here S encloses the point r in the limit, and we call the quantity defined here the divergence of V at r.

Before proceeding with further analytical enlargement on this definition it will be helpful to have some physical concept of the meaning of the divergence of a vector. This is best conveyed by thinking here of V as the velocity of a fluid flowing in space. Then V dS represents the volume of fluid per unit time flowing out from the volume τ through the area dS. Thus the numerator on the right in Eq. (1.55) represents the total volume of fluid per unit time leaving τ. As we let τ → 0 we see Div[V(r)] as the volume of fluid per unit time, per volume of space, leaving the neighborhood of the point r, i.e., as pictured in Fig. 1-5.

FIGURE 1.5. Flow through the Surface of a Volume Element Inclosing a Point.

Note that our definition says nothing about the shape of the volume at any stage of the limiting process. We will not go through the process of a mathematical proof here, but we state that in the limit the shape does not matter. Physically one can see that in the case of an incompressible fluid, the amount per unit time passing through one surface will also pass through another surface inclosing the first. Parallel ideas are used in the mathematical proof.

Let us then choose our volume element in the form of a rectangular parallelopiped with sides h1 Δq1, h2 Δq2, h3 Δq3, surrounding the point r = (q1, q2, q3), as indicated in Fig. 1-6. Observe that the contribution to the surface integral of V · dS over S on the end shaded and

FIGURE 1.6. Volume Element in the Curvilinear Coordinatee q1, q2, q3.

indicated by (a) is approximately,

, indicate values of these coordinates at some mean point on this end. (This is just an application of the mean value theorem of calculus.)

We can write similar expressions for the other surfaces, for example on the end opposite to (a) we have,

where the minus sign results because the area element vector is outward from the surface while V is inward so that the cosine of the angle between V and dS is (–1). Collecting all such integrals and adding to form the integrand of Eq. (1.55), then dividing by the volume, Δα = h1h2hqqq3, we have:

Hence taking limits as all Δqj → 0 we obtain,

in any orthogonal curvilinear coordinate system.

Example (3)

The divergence of V in spherical coordinates can readily be written down since we have already shown that for q1, q2, q3 → r, θ, ϕ we have the scale factors as given in Eq. (1.36). Thus

which simplifies to,

for spherical coordinates. Here we have used the obvious notation,

for the vector V.


In rectangular coordinates hx = hy = hz = 1 and the divergence of V appears as

where Vx, Vv, Vz are the rectangular components of V. Also note that the gradient of a scalar F appears in this coordinate system as,

Thus if we define the vector operator del, or nabla as it has sometimes been called, as,

then we can look upon Eq. (1.63) as the dot, or scalar, product of this operator and the vector V, thus,

Here we must execute the dot operations, then the derivative operations.

Note that this operator structure is not valid in all systems of coordinates, only in rectangular coordinates. However it is often convenient to carry out various manipulations with as an operator until the last step; say for example the end result is symbolically ∇ · V. We interpret this as the divergence and to write it out, say in spherical coordinates, we must go back to Eq. (1.59) for the proper form.


Frequently in physical problems we encounter the divergence of a vector V which itself is the gradient of a scalar function. Symbolically, in operator form, we write

where the scalar operator, ∇², represented here in rectangular coordinates as the dot product of the operator of Eq. (1.65) with itself,

is called the Laplacian operator.

If we use our previous definition of Div and grad, Eqs. (1.42) and (1.59) we see the general form for ∇²F in any orthogonal system as,

for any function F. Deleting F we then have the general representation of the Laplacian operator in any orthogonal system.



Here we develop one of the most valuable theorems of vector calculus. It is based on the formal definition of the divergence of a vector field as given in Eq. (1.55).

Consider the surface integral of a vector function V(r) over a closed surface S, bounding a volume τ. It is not necessary that S be simply connected, but we will here put a restriction on the function V(r). This vector function must be continuous throughout τ and have first derivatives everywhere in τ. We depict the domain schematically in Fig. 1-7, where we have indicated the case of the surface, S, not being simply connected. Here S is made up of two nonconnected surfaces S′ and S″, so we are considering the integral,

FIGURE 1.7. Subdivision of a Volume into Elements.

The positive direction of dS is everywhere outward from the volume τ, which is the shaded region, i.e., as indicated by the normal vector n in our picture.

Now we subdivide τ into volume elements, Δτi, i = 1, 2, . . . N, as indicated by the dotted lines in our picture, and consider the surface integrals of V(r) over the closed surfaces, ΔSi, bounding these volume elements.

For example now consider the sum of two such integrals, say for the elements 1 and 2 in our picture. We have,

Note that a part of ΔS1 is on S′ and that ∇S2 has parts on S′ and S″ also. But most significant is that along the interface between Volume Elements 1 and 2 we have dS1 directed from τ1 into τ2, while dS2, at the same point and of the same magnitude, is directed from τ2 into ∇τ1. Thus for this particular element we see, since V has but one value, i.e., is continuous across the interface,

on the common interface.

Thus we see that in the sum on the right in Eq. (1.70) the contributions on the common interface all add to zero. From this it is evident that,

since in the sum the only parts which do not cancel each other are those surface elements bounding the whole region.

In this sum we multiply each term by ∇τi/∇τi = 1 so that it appears as,

where ∇Si incloses the volume ∇τi. We then take the limit as all ∇τi → 0, and see that, from the definition of the divergence, each bracket in the sum approaches ∇ · V at a point in the infinitesimal volume, ∇τi . Also we see the limit of the sum as the volume integral over the whole region bounded by S, (S′ and S″). Thus we have,

as the divergence theorem.

Example (4)

We illustrate the power and utility of the divergence theorem by applying it in the derivation of the heat flow, or diffusion,† equation.

The law of heat conduction states that the flow of heat is parallel but oppositely directed to the gradient of the temperature, T, and the quantity per unit area, per unit time, flowing in that direction is given by,

where k is the thermal conductivity of the medium and is essentially constant. We call J the heat flux density vector.

We also define the quantity of heat per unit volume of the medium as, CρT, where ρ is the mass per unit volume and C is the specific heat per unit mass of the medium. Actually this is approximate, but is sufficiently accurate in most situations.

Now in some arbitrary fixed region in the medium, of volume τ, the quantity of heat is,

and this may be increasing at the rate

where the time derivative d/dt can be taken inside as a partial derivative since τ is a fixed region of the coordinate space. Note that we assume C and ρ constant in time.

Since heat is a form of energy, and is conserved, this rate of increase must arise as a result of flow of heat into τ if there are no sources in τ. From above the net rate of flow of heat into τ is given by,

where S is the surface bounding τ. Thus we have,

for any region τ free of heat sources.

Here we apply the divergence theorem to the surface integral to obtain,


where we assume k to have the same value everywhere. Since this integral must vanish for every volume element τ, free of heat sources, we conclude that,

in all regions free of heat sources. This is the heat-flow equation.



Let there exist a vector field, V(r), and a space curve C. Such a curve can be prescribed as the intersection of two surfaces in three dimensions, say,

where f(qj) and g(qj) are functions of the orthogonal coordinates, qj, j = 1, 2, 3. Such a curve is depicted in Fig. 1-8. We have at each point of the curve C, specified by the position vector r, a line element ds, and correspondingly a displacement dr directed along the curve. We take dr positive going from end point A of C to end point B.

FIGURE 1.8. A Space Curve as the Intersection of Two Surfaces.

At each such point on C the vector V has a value and we form the scalar, or dot product, V · dr. Then we sum, or integrate, to form,

as the line integral of V(r) on C from

Hai raggiunto la fine di questa anteprima. Registrati per continuare a leggere!
Pagina 1 di 1


Cosa pensano gli utenti di Mathematical Methods for Physicists and Engineers

0 valutazioni / 0 Recensioni
Cosa ne pensi?
Valutazione: 0 su 5 stelle

Recensioni dei lettori