MATHEMATICAL METHODS IN THE PHYSICAL SCIENCESme Fisica TR 269 2005 MATHEMATICAL METHODS IN THE PHYSICAL SCIENCES Second Edition MARY L. BOAS DePaul University JOHN WILEY & SONSCopyright © 1966 and 1983, by Joho Wiley & Sons, ne All sights reserved, Published simultaneously in Canad Reproduction or translation of any part of this work beyond that permite by Sections 107 and 108 ofthe 1976 United Sexes Copycight ‘Act without the permission ofthe copyright ‘owner is unlawful. Requests for permission ‘or further information shouldbe addressed 10 the Permissions Depsrment, Jon Wiley & Sons. Library of Congress Cataloging in Publication Data: Boas, Mary 1. Mathemticl methods inthe physical sciences, ibiograph:p. 743, Includes inden. 1 Mathematics 1961 1. Tite QAs72 B59 I9Ks S10 aB-1226 ISBN 0471-086400-1 Printed in the United States of America SAB/2 CON Local. To Ralph“in, Onyy, wo que PREFACE ‘This book is particularly intended for the student with one year of calculus who wants to develop, in a short time, a basic competence in each of the many areas of mathematics needed in junier to senior-graduate courses in physics, chemistry, and engineering, Thus it isintended to beaccessible to sophomores (or freshmen with AP calculus from high school) tmay also be used effectively by a more advanced student to review half-forgotten topics and learn new ones, either in independent study or ina class. Although the book was written specially for students ofthe physical sciences, students in any field (say, mathematics or ‘mathematics for teaching) may find it useful to survey many topics or to obtain some knowledge of areas they do not have time to study in depth. Since theorems are stated carefully, such students should not have to unlearn anything in their later work ‘The question of proper mathematical training for students in the physical sciences is of concern to both mathematicians and those who use mathematics in applications. Mathematicians are apt to claim that if students are going to study mathematics at all, they should study itincarefuland thorough detail. For the undergraduate physics, chemistry, or ‘engineering student, this means either (1) learning more mathematics than a mathematics major or (2) learning a few areas of mathematics thoroughly and the others only from snatches in science courses. The second alternative is often advocated; let me say why I think itis unsatisfactory. It is certainly true that motivation is increased by the immediate application of a mathematical technique, but there are a number of disadvantages.PREFACE ‘The discussion of the mathematics is apt to be sketchy since that is not the primary Students are faced simultaneously with learning a new mathematical method and applying it to an area of science that is also new to them. Frequently the difficulty in comprehending the new scientific area lies more in the distraction caused by poorly ‘understood mathematics than it does in the new scientific ideas. 3. Students may meet what is actually the same mathematical principle in two different science courses without recognizing the connection, or even learn apparently contradictory theorems in the two courses! For example, in thermodynamics students Jearm that the integral of an exact differential around a closed path is always zero, In electricity or hydrodynamics, they run into J3° dB, which is certainly the integral ofan ‘exact differential around a closed path but isnot zero! Now it would be fine if every science student could take the separate mathematics courses in differential equations (ordinary and partial, advanced ealculus, linear algebra, vector and tensor analysis, complex variable, Fourier series, probability, calculus of variations, special functions, and so on. However, most science students have neither the time nor the inclination to study that much ‘mathematics, yet they are constantly hampered in their science courses for lack ofthe basic techniques of these subjects, It is the intent of this book to give these students enough background in each of the needed areas so that they can cope successfully with junior, senior, and beginning graduate courses in the Physical sciences. [ hope, also, that some students will be sufficiently intrigued by one or ‘more of the fields of mathematics to pursue it further. Itisclear that something must be omitted ifso many topics are to be compressed into one course. I believe that two things can be left out without serious harm at this stage of a student’s work: generality and detailed proofs, Stating and proving a theorem in its most general form is important to the mathematician and to the advanced student, bu it is often ‘unnecessary and may be confusing to the more clementary student. This snot in the leas 10 say that che science student has no use for careful mathematics. Scientists, even more than mathematicians, need careful statements of the limits of applicability of mathematical Processesso that they can use them with confidence without having to supply proof of their validity. Consequently I have endeavored to give accurate statements of the needed theorems, although often for special cases or without proof. Interested students can easily find more detail in textbooks in the special fields; see References (pp. 741ff) Other texts on this subject almost invariably assume a degree of mathematical Sophistication and knowledge about advanced physics not yet attained by students at the sophomore level. Yetsuch students, ifgiven simple and clear explanations, can master these techniques in rapid succession. (They not only can, but will have ro one way or another, if they are going to pass their junior and senior physics courses!) These students are not resdy for detailed applications —these they will get in their science courses but they do need and ‘rant to be given some idea ofthe use ofthe methods and some simple applications, This T have tried to do for each new topic. Many other texts are slinted toward some particular field and contain a good many chapters on advanced topics in that special eld. tis the purpose of this text to cover in simple fashion just the basic methode PREFACE ie ing the 17 years since the publication of theft edition af his Book, have Become move than ever comvinced ofthe vale ofeesing tis materia tthe sophomore evel to Students inthe physical scence, Without i, they must struggle to learn about partial diferent equation, speil functions, Fourie seresand transform, probability, aleulus of vations, diagonalization of matrices, etc, they Tet these topics in ther junior senior ourss in quantum mechanics, electrodynamics, classi mechanics, opts, ee ‘Thus the basic character ofthis text fe unchanged. The principal changes inthe second edition are as follows. chapter 3 has been completely rewritten to include the row reduction method of solving + Guofsimutaneoseqationsandinvertig mares and ver algebra fas been ted tothis chapter. However, I disagree strongly with those who would neglect determinants, Cramer's rule, and the formula for the inverse of a matrix, in favor of row reduction, In spite of the importance of the computer, mathematical physics is much more than arithmetic, and formulasare indispensable in theoretical physics. Thus I have presented both computational and formula methods and have indicated their importance for different purposes. 2. The number of problems has been approximately doubled. Students need to solve many robles in order to master this materi and learn good problem-solving techniques, opriate problems are now included in the individual sections; also the last section of cach chapters se of misclancous problems: Nat the problem numbering stem Problem 7.2 means Problem? of Section 7; however, within Section 7, itis referred to as just Problem 2. Also note that equation numbers are in parentheses, but problem numbers are not; thus (7.2) means equation (7.2) 3. Important formulas, definitions, theorems, ete., have been boxed to make them easier to find 4. The old Chapter 4 (Partial differentiation and multiple integrals) has been split into two chapters (4 and 5). In the new Chapter 5 (Multiple integrals), I have added intro- 0.) Then from (1.2), the total distance traveled by the ball is 14+ 2-2=5. This is an answer to 4 mathematical problem. A physicist might well object that a bounce the size of an atom is nonsense! However, the infinite number of small terms of the series after 4 number of bounces, contribute very litle to the final answer (see Problem 1). Thus it makes litte difference (in our answer for the total distance) whether we insist that the bal rolls after a certain number of bounces or whether we include the entre seties, and itis easier to find the sum of the series than to find the sum of, ay, twenty terms. Series such as (1.3) whose terms form a geometric progression are called geumeri series. We can write a geomettic series in the form (1.6) atartar totaly. ‘The sum of the geometric series (if it has one) is by definition an S=lims,, where 5, is the sum of m terms of the series. By following the method of the example above, you can show (Problem 2) that a geometric series has a sum if and only if |r <1, and in this ease the sum is Sec.2 DEFINITIONS AND NOTATION 3 (8) ‘The series is then called convergent PROBLEMS, SECTION 1 1, In the example above, find the height ofthe tenth rebound, and the distance traveled by the bull after it touches the ground the tenth time. Compare this distance with the total distance traveled 2. Derive the formula (1.4) for the sum 5, ofthe geometric progression S, = a + ar + ar? + “+ ae”, Hint: Start by multiplying S, by rand subtracting the result fromS,. If necessary, you can look up the derivation in an algebra book. Show that the geometric series (1.6) ‘converges if and only if] < 1 also show chat if] < 1, che sum i given by equation (1.8) We can write $= 0.3333 °++ = dt rhy + rah ++. Any fraction whose decimal equivalent does not terminate can be written both as a repeating decimal and as an infinite geometric series. ‘Use equation (1.8) to find the fractions that are equivalent to the following repeating decimals 3. 0555555 4. 088181 5. 0583333 6 oot 2 077777 8. 0.26666, 9. 0.185185 10, 694444 He ose 12, Ina water purification process, one-nth of the impurity is removed in the first stage. In cach succeeding stage, the amount of impurity removed is one-ath of that removed in the preceding stage. Show that ifn = 2, the water can be made as pure a6 you like, but that if n= 3, at least one-half ofthe imparity will remain no matter how many stages are used. 13. Ifyou invest dollar at 6% interest compounded monthly,” it amounts to (1.005) dollars after n months. If you invest $ID atthe beginning of each month for 10 years (120 months), how much will you have atthe end of the 10 years? 14, A dollar due to be paid to you at the end of m months, with the same interest rate as in Problem 13, is worth only (1,00§)"* dollars now (because that is what wil amount ro SI after » months). How much must you deposit now in order be able to withdraw S10 a ‘month forever (starting one month from now)? (Assume the interest rate remains the sume forever.) 15. Some copying machines will make reduced copies. Suppose you copy a page 8 inches wide land it comes out } as wide. Then you eopy this, and so on indefinitely. How far will the original and all the copies reach if you ly them out side by side ona fong table? 2. DEFINITIONS AND NOTATION ‘There are many other infinite series besides geometric series. Here are some examples: @) PHP SPER ty 1 en ) 5 weINFINITE: st FS, POWER SERIES cht In general, an infinite ‘means an expression of the form 22) ay ag tay to tay te where the a,’s (one for each p¢ : i 8 stv integer n are numbers or fanctos given by me formula rT he ie in ech ease mean ht he sre never ne The es continue aeoding tothe lw of formation, which suposed to be eden on : be evident to you 0 he tme you reach thence das, I thre i apt to he dea abo how the oe formed, a general or th term is mien le this @) PHPae ger an?® (2.3) a an rate 2 + (n= 1)! i (The quantity ne factorial means, for integra the produc of all integers fom ns for example, 515" 3-2" 1 = 120, The quanity is deined tobe) fe Oso) itis easy to ee without the general term that each term is just th z rm is jst the square of the number of the term, that is n?. However, in (2.3b), if che formula forthe general term were missing, you could probably make several reasonable guesses for the next term. To be sure of the law of formation, we must either know a good many more terms or have the formula forthe general term. You should verify thatthe fourth term in (2.36) s —x"/6 We can also write series in a shorter abbreviated f i reviated form using a summation sign flowed by the formula forthe wth term, For example, (2) would be wraten (read “the sum of n? from m = 1 t0 2”). The series (2 3b) would be written ror hity. yr it / oI For printing convenience, sums like (2.4) are often written S24 n? 276 PROBLEMS, SECTION 2 is useful to write series both in the form a +43 + 45 + and in the form Dy ay. Write out several terms of the following series (chat oleh write them in the first form), whe 2Eo Se $ 2n(Qn + 1 Ean La Write the Following series in the abbreviated ¥ form. 2 bedede beige 1 yt “3G sates? 9 bade W d4b+ ata Me dbs igeeeseid 2 4 Seed CONVERGENT AND DIVERGENT SERIES 5 3. APPLICATIONS OF SERIES In the example of the bouncing ball in Section 1, we saw that it is possible for the sum of an infinite series to be nearly the same as the sum of a fairly small number of terms it the beginning of the series (also see Problem 1.1). This is the basis for the use of infinite series in applications. Many problems cannot be solved in terms of the so-called ‘lementary functions (powers and roots, trigonometric and inverse trigonometric Func- tions, exponentials and logarithms, and combinations of these). Or, even if they ean, the fesult may be too complicated to work with easily. In such eases, we find an answer in terms of an infinite series, and then use only as many terms as necessary to obtain the accuracy we desire. We shall see many examples of this later in this chapter and also in fater chapters. Differential equations are frequently solved by using series. We shall learn how to find series that represent functions; often a complicated function can be approximated by a few terms of its series (see Section 15). Then a definite integral, say fe dx, for which you cannot find the indefinite integral in terms of elementary fanctions, can be evaluated by expanding the integrand in a series and integrating term by term, 4, CONVERGENT AND DIVERGENT SERIES We have been talking about series which havea finite sum. We have also seen that there are series which do not have finite sums, for example (2.12) Ifa series has a finite sum, it is called convergent. Otherwise its called divergent. Is important to know whether a series is convergent or divergent. Some weird things can happen if you try to apply ordinary algebra toa divergent series, Suppose we try it with the following series ay SHIFDHA+BHIOF ‘Then, 2s S- $4484 OH SH -1 This is obvious nonsense, and you may laugh at the idea of trying to operate with such 2 violently divergent series as (41). But the same sort of thing can happen in more concealed fashion, and has happened and given wrong answers to people who were not careful enough about the way they used infinite series. At this point you probably ‘would not recognize that the series 2) legedadeds is divergent, but iti and the series 43) I-$4b-$44- is convergent as it stands, but can be made to have any sum you lke by rearranging the order ofthe terms! (See Section 8.) You can see from these examples how essential it s to Know whether a series converges, and also to know how to apply algebra to scries correctly. There are even cases in which some divergent series can be used (see Chapter 11), but in this chapter we shall be concerned with convergent series.6_ INFINITE SERIES, POWER SERIES om FOR CONVERGENCE; THE PRELIMINARY TEST_7 Before we consider some tests for convergence, let us repeat the definition of eonver= ence more carefully. Let us call the terms of the series a, so that the series is ay ay tay tay tag to tay t Remember that the three dots mean that there is never a last term; the series goes on ‘without end. Now consider the sums 5, that we obtain by adding more and more terms of the series. We define re 3) Ss ay tay tay to tay Fach S, is called a partial sum; it is the sum of the first m terms of the series. We had an example of this for a geometric progression in (1.4). ‘The letter » can be any integer: for each nm, S, stops with the mth term. (Since S, is not an infinite series, there ie ne question of convergence for it.) As m increases, the partial sums may increase without any limit as in the series (21a). They may oscillate as in the series 1—243—4+5—-+- (which has partial sums 1, —1, 2, ~2, 3,-+-) or they may have some more complicaced behavior. One possibility is that the S,’s may, afier 4 while, not change very much any more; the a,’ may hecome very small, and the S's come closer and closer to some value S. We are particularly interested in this case in which the S,’s approach a limiting value, say [= — (i is understood that 5 is a finite number.) If this happens, we make the following definitions, 4 If the partial sums 5, of an infinite series tend to a limit 5, the series is called convergent. Otherwise itis called divergent. b. The limiting value S is called the sum of the series. © The difference R, = S—S, is called the remainder (or the remainder after n terms). From (4.6), we see that (47) PROBLEMS, SECTION 4 A carl mathcmatcl definition of 3 convergent infinite series with sum Sis this: Given any seull postive number (sully elled¢ by matherstians), tt ple o ed nee An, tha IS ~Sy| «efor every >. In words ths met that movant how ga ee Of ll he pata sme yo terms fer le than hour om the hal eae Se tems more. Sle some e's and in! the conerpoding Ne be eal ane wigs secs TESTING sen Pa zig ee oe rs) when n> 3 (ce Problem 6.1); Hence the reminder $— 5, is male than the remainder of a geometric series. = 4 tine S la ey i om: "Te gener erm is es than Lae L e 1 1 1 gard <1 ase Problem 3 LoL Hi Doe Lh aaa Ga Saee zo 1 8 Hot: Compare Ss = SA 5, TESTING SERIES FOR CONVERGENCE; THE PRELIMINARY TEST We shal consider ere a few simple tests for covergence. These cts wil uaa some of the ides invlved in testing series for convergence and wil work for good many, but not all ese. There ate more completed tests which you can find in oer books. In some cases it may be quite a difficult mathematical problem to investigate the convergence of a complicated series. However; for our purposes the simple tests we gi here will be sufficient. ses saa First we discuss what we shall call the preliminary test. In most cas Sh apply this to a series before you use other tess hs do not tend to zero (that is, if Preliminary test. If the terms of an infinite ser 2 Tim,io 4 # 0), the series diverges. If lim,.,. 4, = 0, we must test further. b+aedege jes diverges and no the terms are tending to 1, so by the preliminary test, this series diverge further testing is needed.8_INFINITE SERIES, POWER SERIES PROBLEMS, SECTION 5 se the pretiminaey tt 10 dvi wheter the ft testing. Card: Do av say that seis i convergent he preliminary tex cannot dee hi De B+ 2 ee oS + On 4 . eat 2a iz i'n = ‘ 11. Using (46), give a proof of the preliminary test for convergence of an infinite series. Hint 5. 6. TESTS FOR CONVERGENCE OF SERIES OF POSITIVE TERMS; ABSOLUTE CONVERGENCE We are now going to consider four useful tests for series whose terms are all positive. If some of the terms ofa series are negative, we may stil want to consider the related series which we get by making all the terms positive; that is, we may consider the series whose terms are the absolute values of the terms of our original series. If this new series con- verges, we call the original series absolutely convergent. It can be proved that ifa series converges absolutely, then it converges (Problem 7.9). This means that if the series of absolute values converges, the series is sll convergent when you put back the original ‘minus signs. (The sum i different, of course.) The following foar tests may be used, then, «ther for testing series of postive terms, or for testing any series for absolute convergence.” A. The Comparison Test ‘This test has two parts, (a) and (b). (@) Let my omy tims + mg + be a series of positive terms which you know converges. Then the scries you are testing, namely a; +a tay tay + is absolutely convergent if |ay| dy forall » from some point on. Warning: Note carefully that neither |a| 2m, nor |a,| Sd, tells us anything. ‘That is, if a series has terms larger than those of a convergent series, it may still ‘converge or it may diverge—we must test it further. Similarly, if a series has terms smaller than those of a divergent series, it may still diverge, or it may converge. rest a1 4 b+ b+ de ++ for convergence, Example. Asa comparison series, we choose the geometric series 21 Epadetete des Notice that we do not care about the first few terms (or, in fact, any finite number of terms) in a series, because they can affect the sum of the series but or whether it converges. When we ask whether a series converges or not, we are asking what happens as we add more and more terms for larger and larger m. Docs the sum increase indefinitely, or does it approach a limit? What the first five o hundred or million terms are has no effect on whether the sum eventually increases indefinitely or ap- proaches a limit. Consequently we frequently ignore some of the early terms in testing series for convergence Tn our example, the terms of $., I/n! are smaller than the corresponding terms of 1/2" for all n > 3 (Problem 1). We know that the geometric series converges ‘because its ratio is $. Therefore Ss. I/n! converges also PROBLEMS, SECTION 6 1. Show that 1 > 2° for all > 3. Hints Write out a few terms; then consider what you multiply by to go from, say, 5! t0 6! and from 2* to 2° 2. Prove that the harmonic series Sz. ln is divergent by comparing it with the series LH d+ GED 4G +E +E +H + (8 terms each equal 10) + which is 1+d+4444+44 3. Prove the convergence of U7, 1/x® by grouping terms somewhat as in Problem 2 4. Use the comparison test to prove the convergence of the series in Problems 4.5 and 4.6 5. Test the following series for convergence using the comparison test. - 2 Hint: Which is larger, n or /? ola 6. ‘There are 9 one-digit numbers (Ito 9}, 90 twordigit numbers (10 to 99), How many threecdigit, fouredigit, etc, numbers are there? ‘The first 9 terms of the harmonic series al144494 00-44 are all greater than 5; similarly consider the next 90 terms, and so on. Thus prove the divergence of the harmonic series by comparison with the sence ol] + [90 terms each = ho] + ‘The comparison test is really the basic test from which other tests are derived. It is Probably the most useful rest of all for the experienced mathematician but it is often hard to think of a satisfactory m series until you have had a good deal of experienes with series. Consequently, you will probably not use it as ofien as the next three tests B. Integral Test We can use this test when the terms of the series are positive and not increasing, that is, when ag, 1 Say. (Again remember that we can ignore any finite number of terms of the series; thus the test can still be used even if the condition a,., N, then J* ay converges if J° ag dh is finite and diverges ifthe integral is infinite, (Phe integral isto be evaluated only at the upper limit no lower limit is needed.) To understand this test, imagine a graph sketched of a, asa function of n. For example, in testing the harmonic series J. 1/n, we consider the graph of the function 2= Un (Gimilar to Figs, 6.1 and 6,2) letting m have all values, not just integral ones, ‘Then the values of yom the graph at = 1, 2, 3, ---, are the terms of the series. In Figures 6.1 and 6.2, the areas of the rectangles arc just the terms of the series, Notice that in Figure 6.1 the top edge of each rectangle is above the curve, so that the area of the rectangles is greater than the corresponding area under the curve. On the other hand, in Figure 6.2 the rectangles lie below the curve, so their arca is less than the corresponding area under the curve. Now the areas of the rectangles are just the terms Of the series, and the area under the curve is an integral of y dn or a, de. The upper limit on the integrals is ao and the lower limit could be made to correspond to any FIGURE 6.1 OF POSITIVE TERMS 'S FOR CONVERGENCE OF SER © we 6.1), J ay dn is jesm of the series we wanted to start with, For example (see Figure 6.1), [Yay tex han the sum of the sre frm on, but se Figure 62) geste than the tam of the series from a, on. IF the integral is finite, then the sum of the series from ay on is finite, that is, che series converges, Note again that the terms at the beginning of a series have nothing 0 do with convergence. On the other hand, if the integral is infinite then the sum of the series from ay on is infinite and the series diverges. Since the beginning terms are of no interest, there is no need to use a lower limit on the integral, and you should simply evaluate { a, dn. (Also see Problem 16.) Example. Test for convergence the harmonic series 1) 1+$44 Using the integral test, we evaluate J Linainn (We use the symbol In to mean a natural logarithm, that is, 2 logarithm to che base «.) Since the integral is infinite, the series diverges. PROBLEMS, SECTION 6 series converge or diverge. Use the integral test to find whether the following re Hint and warning: Do aot use lower limits on your integrals (see Problem 16) D 8 10. 0 2. 13 ¥. 15. Use the integral test to prove the following so-called p-series test. ‘The series 1 convergent if p > 1, Ee * foe Caution: Do p = 1 separately 16 In tenga? for converges, det evn f6 9°? de © and concludes (erroneously) that the series diverges. What is vron area under the curve in a diagram such as Figure 6.1 oF 6.2. This example shows the danger of using a lower limit inthe integral test. Poet it: Although you cannot evaluate 17. Use the integral test to show that Ef. ¢°*" converges. Hit A vr fe impr you an show tht fate (which ltt in eee) by comping with §7 eo" de12_INFINITE SERIES, POWER SERIES cht ©. Ratio Test ‘The integral test depends on your being 1g able to integrate 2, di; this is not always easy! We consider another test which will handle many cases in which we cannot evaluate the integral. Recall that in the geometric series each term could be obtained by multiplying the one before it by the ratio r, that is, a, 1 =r, OF dy y/dy = r. For other series the Fatio dee /dy is mot constant bur depends on w; let us call the absolute value of this ratio p,. Let us also find the limit (if there is one) ofp, as n—» oo and eall this limit p Thus we define p, and p by the equations | 2) p= limp, to If you recall that a geometric series converges if |r] < 1, it may seem plausible that a series with p< I should converge and this is rue. This statement can be proved (Problem 30) by ‘comparing the series to be tested with a geometric series. Like a gcomettic series with seat ‘TESTS FOR CONVERGENCE OF SERIES OF POSITIVE TERMS 13 We find 1 1+ re the tex els ws noting and we mus te sme diferent tt. A word af warning A this example: Nove tat x= net 1 away les than 1. He cate ott ae this ro with p and Condude incorely that this sree converges. (tf Sry divergent ar we proved by the incl test) Remember that pismo! the sme as the ratio p, = | 4,4 4/a,|, but is the limit of this ratio as n—+ co. PROBLEMS, SECTION 6 Use the ratio test to find whether the following series converge or diverge: 2. Ir1> loa series ith p> 1 alo diverges Meblem 0) However en Ire eons " wis does not tell us anything; some series with p = I converge and some diverge, so ne mune nam : = 10 " find anather test (say one of the two preceding tests) To summatze the rato ee ft aise aia P <1, the series converges; ae ni x Tai 2, Gal D>, the series diverges. & mont » 5a ” * Go! gon Example 1. Test for convergence the series M0, Prove the ratio test. Hints IF Jays /anl—+p <1, take @ 50 that p- I diverges. Hint: Take p > a > and se the ps at (n= Wen 2 1 n+ DIG Doe DT we b. A Special Comparison Test p= limp, = tim 5 =0. “This test has two parts: (a) a convergence test, and (b) a divergence test. (See Problem Since p <1, the series converges Example 2. ‘Test for convergence the harmonic series en 37) (2) If 52.4 4, is a convergent series of positive terms and a, > 0 and a/b, tends to-a (nite) limit, then F32., a, converges. (b) If St. dy is a divergent seris of positive terms and «, 2 0 and a/d, tends to a limit greater than 0 (or tends to +0), then FP. a, diverges.cut There are really two steps in using either of these tests, namely, to decide on a comparison series, and then to compute the required limit. The first part is the most important; given a good comparison series it is a routine process to find the needed limit. ‘Phe method of finding the comparison series is best shown by examples Example 1. ‘Test for convergence < Jari Swot Remember that whether a series converges or diverges depends on what the terms are 88 m becomes lager and larger. We ae inerestd in the mh tera ae ovo: Think of m= 10'° or 10'°°, say; a little calculation should convince you that as m increases, 2n? — Sn + 1 is 2n? to quite high accuracy ‘Similarly, the denominator in our example is nearly 4? for lage m. ly Section 9, fact ly we se thatthe fact 3 In m, and 3"> n°. (You might like to compute 100° = 10°, and 31° > 5 x 10*7) ‘The denominator of the given series is approx. imately n° as in Example 1, Thus the comparison series is 7. 3%n3. prove this divergent by the ratio test. Now by test (b) Ir is easy t0 Sec.7 ALTERNATING SERIES _15 which is greater than zero, so the given series diverges. PROBLEMS, SECTION 6 ‘Use the special comparison test to find whether the following series converge or diverge @ Qn + 1K3n—5) % __ mln + 0) a ay R Lae Lea = % _(n—Inmt * laweees | Lacarei 7 37. Prove the special comparison test. Hint (part a): If a/b,» L and M > L, then a, < Mb, for large n. Compare Ss, a, with YS, MB, 7. ALTERNATING SERIES So far we have been talking about series of positive terms (including series of absolute values). Now we want t0 consider one important case of a series whose terms have mixed signs. An alternating series isa series whose terms are alternately plus and minus; for example, yr a 7 1 a an alternating serics. We ask two questions about an alternating series. Does it converge? Does it converge absolutely (that is, when we make all signs positive)? Let us consider the second question first. In this example the series of absolute values Lededed is the harmonic series (6.1), which diverges. We say that the series (7.1) is not abso- lutely convergent. Next we must ask whether (7.1) converges as it stands, If it had tumed out to be absolutely convergent, we would not have fo ask this question since an absolutely convergent series is also convergent (Problem 9). However, a series which is not absolutely convergent may converge or it may diverge; we must test it further, For alternating series the test is very simple: ‘Test for alternating series. An alternating series converges if the absolute value ‘of the terms decreases steadily to zero, that is, if [y+] < [ay] and lity dy I te — our example —— 1 and lim > = 0, so (7.1) converges.6 INFINI at +E SERIES, POWER St PROBLEMS, SECTION 7 Test the following series for convergence. 9. Prove that an absolutely convergent series Sx a, is convergent. Hint ‘Then the b, are nonnegative; we have Ib,| © 2la and ay = 6, — la 10. The following alternating series are divergent (but you are not asked to prove this). Show that 4,>0. Why doesn’t the alternating series test prove (incorrectly) that these series ‘converge? @) 2-$s4-bee- bed CONDITIONALLY CONVERGENT SERIES [A sseties like (7.1) which converges, but does not converge absolutely, is called con ditionally convergent. You have to use special care in handling conditionally convergent series because the positive terms alone form a divergent series and so do the negative terms alone. If you rearrange the terms, you will probably change the sum of the series, and you may even make it diverge! It is possible to rearrange the terms to make the sum any number you wish. Let us do this with the alternating harmonic series 1 — } +4—4+---. Suppose we want to make the sum equal to 1.5. First we take enough positive terms to add to just over 1.5. The first three positive terms do this: l4d4b= I> 15. ‘Then we take enough negative terms to bring the partial sum back under 1.5; the one term —} does this. Again we add positive terms until we have a little more than 1.5, and so on. Since the terms of the series are decreasing in absolute value, we are able (as we continue this process) to get partial sums just a little more or a little less than 1.5 bbut always nearer and nearer to 1.5. But this is what convergence of the series to the sum 1.5 means: that the partial sums should approach 1.5. You should see that we could pick in advance any sum that we want, and rearrange the terms of this series to get it. Thus, we must not rearrange the terms of a conditionally convergent series since its convergence and its sum depend on the fact that the terms are added in a particular order. Here isa physical example of such a series which emphasizes the care needed in applying ‘mathematical approximations in physical problems. Coulomb's law in electricity says that the force between two charges is equal to the product of the charges divided by the square of the distance between them (in electrostatic units; to use other units, say mks, we need only multiply by a numerical constant). Suppose there are unit positive charges at x See.9 USEFUL FACTS ABOUT SERIES 17 V2 VA, i, 8, >, and unit negative charges at x = 1, /3, 3, I, We vant to know the total force ating on the unit positive charge at x = 0 due to all the other charges. The negative charges attract the charge at x = 0 and try to pull it to the right; we call the forces exerted by them positive, since they are in the direction of the postive x axis, The forces due to the postive charges are in the negative + direction, and we call them negative. For example, the force duc to the postive charge at x= 9/3, is —(1 + 11/2)? = 4. The total force on the charge at x = 0 is, then, ey Fel-d+}-$44-24 Now we know that this series converges as it stands (Section 7). But we have also seen that its sum (even the fact that it converges) can be changed by rearranging the terms Physically this means that the force on the charge at the origin depends not only on the size and position of the charges, but also on the order in which we place them in thei positions! This may very well go strongly against your physical intuition, You feel that a physical problem like this should have a definite answer. ‘Think of it this way: Suppose there are two crews of workers, one crew placing the positive charges and fone placing the negative. If one crew works faster than the other, itis clear that the force at any stage may be far from the F of equation (8.1) because there are many extra charges of one sign, The crews can never place all the charges because there are an infinite number of them. At any stage the forces which would arise from the positive charges that are not yet in place, form a divergent series; similarly, the forces due to the unplaced negative charges form a divergent series of the opposite sign. We cannot then stop at some point and say that the rest of the series is negligible as we could in the bouncing ball problem in Section 1. But if we specify the order in which the charges are to be placed, then the sum 5 of the series is determined (Sis probably different from P in (8.1) unless the charges are placed alternately). Physically. this ‘means that the value of the force as the crews proceed comes closer and closer to S and we can use the sum of the (properly arranged) infinite series as a good approx imation to the force 9. USEFUL FACTS ABOUT SERIES ‘We state the following facts for reference 1. The convergence or divergence of a series is not affected by multiplying every term ofthe series by the same constant. Neither is it affected by changing a finite number of terms (for example, omitting the first few terms) 2, Two convergent series Si ay and Ey y ty may be added (or subtracted) term by term, (Adding “term by term” means that the nth term of the sum is a, + b,.) The resulting series is convergent, and its sum is obtained by adding (subtracting) the sums of the two given series 3. The terms of an absolutly convergent series may be rearranged in any order without alfecting cither the convergence or the sum. This is not true of conditionally conver- gent series as we have seen in Section 818_INFINITE SERIES. POWER SERU ut PROBLEMS, SECTION 9 Test the following series for convergence or divergence. Decide for yourself which test is easiest to use, but don't forges the preliminary test. Use the facts stated above when they apply 1 ye oF Dm (a? 4 6 Ee so Gay 7 9% 10. i 2. L. e i FD! Oe 16. 18. 9. 20, 2 Yay if eras 10. POWER SERIES; INTERVAL OF CONVERGENCE We have been discussing series whose terms were constants, Even more important and useful are series whose terms are functions of x. ‘There are many such series, but in this chapter we shall consider series in which the mth term is a constant times x" or a constant times (x ~ a)" where a is a constant. These are called power series, because the terms are multiples of powers of x or of (x — a). In later chapters we shall consider Fourier series whose terms involve sines and cosines, and other series (Legendre, Bessel, etc.) in which the terms may be polynomials or other functions By definition, a power series is of the form Sapte ag t aye tags? tase te or (0.1) ~ Sante — ay ao + aye — a) + agte — a)? + asl — a) + ___ POWER SERIES PERV) where the coefficients a, are constants, Here are some examples: (10.2) V2 WA Vint Whether a power series converges oF nor depends on the value of x we are consider- ing. We often use the ratio test to find the values of x for which a series converges. We illustrate this by testing each of the four series (10.2). Recall that in the ratio test we divide term n+ 1 by term m and take the absolute value of this ratio to get ,, and then take the limit of p, a8 n—» 00 to get p. Example 1. For (10.2a), we have Cay! ayr pert Pn 2p 2 ce ‘The series converges for p <1, that is, for |x/2| <1 or |x| <2, and it diverges for |x| > 2 (see Problem 6.30). Graphically we consider the interval on the + axis between = 2 and x= 2; for any x in this interval the serics (10.2a) converges. The end- points of the interval, x = 2 and x = —2, must be considered separately. When x = 2, (10.24) is Padtdaddes, which is divergent; when x= —2, (10.28) is 1+ 141+ 14+, which is divergent ‘Then the interval of convergence of (10.2a) is stated as —2 , mln — Dayle = a)? 2, r= Slay #43 Dayle —a) + Fan Wn — Daj ~ ay Ho, LOM) = nn — 1Y(n = 2) --- 1+ a, + terms containing powers of (* — a). [The symbol /'%(x) means the nth derivative of f(x).] We now put x= in each equation of (12.6) and obtain Fla =a, Fl@=ay, Sf" 27 ad £0) = Blas, oy fa) ma, [Remember that /"(a) means to differentiate /(x) and then put x = a; f"(a) means to find f(x) and then put x =a, and so on.) We can then write the Taylor series for f(x) about x = a 1 (28) J) =/0) +6 Of) + EOP 4 FSO OM) + ‘The Maclaurin series for f(x) is che Taylor series about the origin, Putting a = 0 in (12.8), we obtain the Maclaurin series for f(x) e s ey (29 fY=(O4 PO +FLO+T LO + +5 LO + ‘We have written this in general because itis sometimes convenient to have the formulas for the coefficients. However, finding the higher order derivatives in (12.9) for any but the simplest functions is unnecessarily complicated (try it for, say, &**). In Section 13,S. POWER cht we shall discuss much easier ways of getting Maclaurin and Taylor series by combining a few basic series. Meanwhile, you should verify (Problem 1, below) the basic series (13.1) to (13.8) and memorize them, PROBLEMS, SECTION 12 1. By the method used 10 obtain (12.5) [which is the series (13.1) below], verify each of the ‘other seties (13.2) t0 (13.5) below. 13, TECHNIQUES FOR OBTAINING POWER SERIES EXPANSIONS There are often simpler ways for finding the power series of a function than the successive differentiation process in Section 12. Theorem 4 in Section 11 tells us that for a given function there is just one power series, that is, series of the form Ts'o ay” ‘Therefore we can obtain it by any correct method, and be sure that it is the same Maclaurin series we would get by using the method of Section 12, We shall illustrate a variety of methods for obtaining power series. First ofall, itis a great timesaver for you to verify (Problem 12.1) and then memorize the following basic series | convergent for ay : alls; (132) : alls; (33) Merete heats alls; (13.4) Ind tayex-F+ -l0 Is, one 16, sin Dn (1-4 39) mn [ow ea uw [ere 0. 20, are sin x 21, cosh y= EE seo. QUESTIONS OF CONVERGENCE AND ACCURACY IN COMPUTATION 29 2 witt a WIRROFS ” 35. ome) 2% n(d +30) n. iB sinwat 2B Ine- ae wo, [eet * sin de — a a. | eS 3 n+ Ji4e)= | p V1 b Jive 3B, In cos x Hints: Method 1 Write cos x= 1-4 (cos x= 1) = 1-445 use the series you Know for In (I +4); replace u by the Maclaurin series for (cos x — 1) Method 2: In cos x= —[% tan x da. Use the series of Example 3 in method B HOE Hint 68 me oo Using method F abuve, find the frst few terms of the Taylor series for the following functions about the given points 1 a7. fle 35. fle sing, a= a2 fl)= con ann. fimatn ana sym Jan ant 38/0) 14, QUESTIONS OF CONVERGENCE AND ACCURACY IN COMPUTATION ‘The thoughtful student might well be disturbed about the mathematical manipulations we have been doing. How do we know whether these processes we have shown really give us series that approximate the functions being expanded? Certainly some functions cannot be expanded in a power series; since a power series becomes just ay when =0, it cannot be equal to any funetion (like 1/x or In.x) which is infinite at the origin. So we might ask whether there are other functions (besides those that become infinite at the origin) which cannot be expanded in a power series. Alll we have done so far is to show methods of finding the power series for a function if it has ane. Now is there a chance that there might be some functions which do not have series expansions, but for which our formal methods would give us a spurious series? Unfortunately, the answer is Yes"; fortunately, this is not a very common difficulty in practice. How- ever, you should know of the possibility and what to do about it, You may first think of the fact that, say, the equation eloxeeie T4e * is not valid for |x| 21. This is a fairly easy restriction to determine; from the beginning we recognized that we could use our series expansions only when they converged. But there is another difficulty which can arise. It is possible for a series found by the above methods to converge and still not represent the function being30_INFINITE SERIES, POWER SERIES cut expanded! A simple example of this ise" for which the formal series is 040+ 0-4 -- because ¢ "1 and all its derivatives are zero at the origin (Problem 15.26). It is clear that eis not zero for 7 > 0, so the series is certainly not correct. You can startle your friends with the following physical interpretation of this. Suppose that at ¢ = 0) car is at rest (zero velocity), and has zero acceleration, zero rate of change of acceleration, etc. (all derivatives of distance with respect to time are zero at £= 0). Then according to Newton's second law (force equals mass times acceleration), the instantaneous force acting on the car is also zero (and, in fact, so arc all the dcrivatives of the force). Now we ask “Is it possible for the car to be moving immedi- ately after + = 0?” The answer is “Yes”! For example, let its distance from the origin as a function of time be e~!™") ‘This strange behavior is really the fault of the function itself and not of our method of finding series. The most satisfactory way of avoiding the difficulty is to recognize, by complex variable theory, when functions can or cannot have power series. We shall consider this in Chapter 14, Section 2. Meanwhile, let us consider two important questions; (I) Does the Taylor or Maclaurin series in (12.8) or (12.9) actually converge to the function being expanded? (2) In a computation problem, if we know that a series converges to a given function, how rapidly does it converge? That is, how many terms -must we use to get the accuracy we require? We take up these questions in order. The remainder R(x) ina Taylor series is the difference between the value of the function and the sum of m+ 1 terms of the series: (4. 1 : RAs) =f) [re FOF) 4 l= aa) +E Oe ore)| Saying that the series converges to the function means that lim, |R,(x)| = 0. There are many different formulas for R,(x) which are useful for special purposes; you can find these in calculus books, One suich formula is %) eo" (14.2) Ble) = where ¢ is some point between @ and x. You can use this formula in some simple cases to prove that the Taylor or Maclaurin series for a function does converge to the function (Problems 11 to 13) Serles Computation Now suppose that we know in advance that the power series of a function does converge to the function (within the interval of convergence), and we ‘want to use the series to compute values of the function. We would like, if possible, to éstimate the ‘error caused by using only a few terms of the series, [Although in Principle we could use (14.2), there are simpler methods which we shall discuss.] ‘There is an casy way to estimate this error when the series is alternating and meets the alternating series test for convergence (Section 7). In this case the error is (in absolute Value) less than the absolute value of the first neglected term (see Problem 1), See. QUESTIONS OF CONVERGENCE AND ACCURACY IN COMPUTATION _31 If.$ = Ya, is an alternating series with lags! , ‘The sum of the terms through — 35 is 0.656+, which differs from S'by about 0.01. ‘This is less than the next term = di = 0.015-+ “This estimate for the error may be completely misleading for convergent series that are not alternating. For instance, oa a = L6H94 Pre but it takes about 200 terms to get the second decimal place correctly, even though the 200th term is only ‘000025. (Also see Problems 9 and 10.) However, note that in this last example we are finding the sum of the power series Sarah, which is the largest x for which the series converges. If, instead, we ask for the sum of this £ (osm am 2, The next rerm is ({)%/6? = 0.0004, which is less than the error but still of the same order of magnitude. We can state the following theorem (Problem 2) which covers ‘many practical problems. IFS = Ya," converges for |x| <1, and if (44) lagi < lag] for > N, then S— Saye] See figure. Show that #/R = see 01 find wo terms of the series for sec # = Ics @, and uses = R6. Thus show thatthe distance in miles is appronimatly 3472 with hin fet i MISCELLANEOUS PROBLEMS rds of the ‘Show that it is possible to stack « pile of identical books so that the top book is as far as you like to the right of the bortom book. You may use as many books as you need, but for any desired distance it will be some finite number. Hint» Start at che top of the pile to see how to stack them. Each time place the pile already completed on top of another book so that the pile is just at the point of tipping. (In practice, of course, you can’t let them —— ——— a overhang quite this much without having the stack topple.) Find the distance from the right-hand end of each hook to the right-hand end of the one beneath it. (To find a general Formula for this distance, consider the three forces acting on book m, and write the equation for the torque about its right-hand end.) Show that the sum of these distances is a divergent series. (It is said that some students studying in the library tried this experiment one ‘evening and left the results, to the consternation of the librarian the next morning. Ut is suggested that you try it instead with one or more decks of cards, Also see “ Leaning Tower of The Physical Reviews,” American Journal of Phyricr, vol. 27, no. 2, p. 121.) The picture is a mobile constructed of dowels (or soda straws) connected by thin threads, Each thread goes from the left~ hhand end of a rod to a point on the rod below. Number the rods from the bottom and find, for rod m, the distance from its left end to the thread so tht all rads of the mobile will be horizontal. Hint: Can you see the relation between this problem and Problem 1? Show that Us. ln? is convergent. What is wrong with the following " proof ” that it diverges? oa Te eS Yat’ Jin = Ml+dede de Since the harmonic series diverges, the original series diverges. Hint: Compare 3m and / jebebeaea_anennrre SERIES, POWER SERIE cut Test for convergence: 43 s ‘ min) Find the interval of convergence, including end-point tests: By ean . Fe 10 Jin 2s Gat " a $e a fee FSD aa Find the Maclaurin series forthe following functions. 14 cos [in (+ 2)] 18 w() 6 * Vivane a nee {He ne Wo awans= [A Find the fist few terms ofthe Taylor series for the following functions about the given points. B. tingenr B. ynend a. ee Use series you know to show that ee eo (a dE | (ng)? 1 2. ns 4 4S 25, Evaluate the limit lim, +7/in os x both by series and by LHopital’s rule, Use Mactaurin series to evaluate % tn (S-ima) a in(* MW, 6 005 4COMPLEX NUMBERS 1, INTRODUCTION You will probably recall using imaginary and complex numbers in algebra, The general solution of the quadratic equation aay at tbe te=0 for the unknown 2, is given by the quadratic formula [Pee (1.2) If the discriminant d = (8? — 4ac) is negative, we must take the square root of a negative number in order to find 2. Since only positive numbers have real square roots, it is impossible to use (I.2) when d <0 unless we introduce a new kind of number, called an imaginary number. We use the symbol i = y/—1 with the understanding that 2 = —1, Then4 _COMPLEX. NUMBERS ch are real. In (1.2) we also need combinations of real and imaginary numbers. For example, the solution of We use the term complex mumber to mean any one of the whole set of numbers, real, imaginary, or combinations of the two like 1+ i. Thus, + 5, 171, 43+ i\/3 are all ‘examples of complex numbers Once the new kind of number is admitted into our number system, fascinating Possibilities open up. Can we attach any meaning to marks like sin i, e, In (1 +)? We shall see later that we can and that, in fact, such expressions may turn up in problems in physics, chemistry, and engineering, as well as mathematics, When people first considered taking square roots of negative numbers, they felt very uneasy about the problem. They thought that such numbers could not have any mean ing or any connection with reality (hence the term imaginary”). They certainly would not have believed that the new numbers could be of any practical use. Yet complex hhumbers are of great importance in a variety of applied fields; for example, the ele cal engineer would, to say the least, be severely handicapped without them. ‘The complex notation often simplifies setting up and solving vibration problems in cither dynathical or electrical systems, and is useful in solving many differential equations which arise from problems in various branches of physics, (See Chapters 7 and 8.) In addition, there is a highly developed field of mathematics dealing with functions of a complex variable (see Chapter 14) which yields many useful methods for solving prob- lems about fluid flow, clasticity, quantum mechanies, and other applied problems. Almost every field of either pure or applied mathematics makes some use of complex aumbers. 2. REAL AND IMAGINARY PARTS OF A COMPLEX NUMBER A complex number such as 5+ 3i is the sum of two terms. ‘The real term (not containing i) is called the real part of the complex number. ‘The coefficient of i in the ‘other term is called the imaginary part of the complex number. In 5 +34, $ is the real part and 3 is the imaginary part. Notice carefully that the imaginary part of a complex number is nor imaginary Either the real part or the imaginary part of a complex number may be zero. If the real part is zero, the complex number is called imaginary (or, for emphasis, pure imaginary). The zero real part is usually omitted; thus O + 5 is written just Si. If the imaginary part of the complex number is zero, the number is real, We write 7 + 0i as just 7. Complex numbers then include both real numbers and pure imaginary numbers as special cases, In algebra a complex number is ordinarily written (as we have been doing) as a sum like 5 + 31. There is another very useful way of thinking of a complex number. As we have said, every complex number has a real part and an imaginary part (either of which See.3 THE COMPLEX PLANE _45 may be zero). These are two real numbers, and we could, if we liked, agree to write 5 +31 as (5, 3). Any complex number could be written this way as a pair of real numbers, the real part frst and then the imaginary part (which, you must remember, is real). This would not be a very convenient form for computation, but it suggests a very useful geometrical representation of a complex number which we shall now consider. 3, THE COMPLEX PLANE In analytic geometry we plot the point (5, 3) as shown in Figure 3.1. As we have seen, the symbol (5, 3) could also mean the complex number 5 + 31. The point (5, 3) may then be labeled cither (5, 3) or 5+ 3i. Similarly, any complex number x + iy (x and y real) can be represented by a point (1, y) in the (x,y) plane, Also any point (x, ) in the (x,y) plane can be labeled x + iy a8 well as (x, »}. When the (x, 9) plane is used in this way to plot complex aumbers, itis called the complex plane, It is also sometimes called an Argand diagram. ‘The x axis is called the real axis and the y’ axis is called the imaginary axis (note, however, that you plot y and nr i). FIGURE 3.1 When a complex number is written in the form x +f, we say that i is in rectangue lar form because x and y are the rectangular coordinates of the point representing the ‘number in the complex plane. In analytic geometry, we can locate a point by giving its polar coordinates (r, 8) instead of its rectangular coordinates (x, »). There is a corre- sponding way to write any complex number. In Figure 3.2, cos 4, Gu) y=rsin 6, ‘Then we have rt iy =reos 6+ isin © Hoos +i sin 0), FIGURE 32 ‘This last expression is called the polar form of the complex number. As we shall see (Sections 9 t0 16), the expression (cos @ + 4 sin 6) can be written as”, so a convenient ‘way to write the polar form of a complex number is ~ ‘The polar form re” of a complex number is often simpler to use than the rectangular form,4 _ COMPLEX NUMBERS Example. In Figure 3.3. the point 4 could be labeled as (1, «/3) or as 1+ 4/3. Similarly, using polar coordinates, the point A could be labeled with its (r, 8) values as 2, n/3). Notice that r is always taken posi- tive, Using (3.3) we have iy) =, V3) ree aii alone Erin) = em This nes wo more ways bel pine in Figure 33 FiGure 3.2 Radians and Degrees In Figure 3.3, the angle x/3 is in radians. Ever since you studied calculus, you have been expected to measure angles in radians and not degrees, Do you know why? You have learned that (d/dx) sin x = cos x. This formula is nor correct —unless x is in radians. (Look up the derivation in your calculus book!) Many of the formulas you now know and use are correct only if you use radian measure: consequently that is what you are usually advised to do. However, it is sometimes simpler to do computations with complex numbers using degrees (especially when vou ‘must use a calculator, as often happens in practical work, for example in electric circuit problems). ‘Thus it is important to know when you can and when you cannot use degrees, You can use degrees to measure an angle and to add and subtract angles as long as the final step is to find the sine, cosine, or tangent of the resulting angle (with your calculator in degree mode) or to use tables with entries in degrees. For example, in Figure 3.3, we can, if we like, say that @ = 60° instead of 0 = 1/3. If we want to find sin (7/3 — m4) = sin (n/12) = 0.2588 (calculator in radian mode), we can instead find sin (60° ~ 45°) = sin 15° = 0.2588 (calculator in degree mode). Note carefully that aan angle is in radians unless the degree symbol is used; for example, in sin 2, the 2 is 2 radians or about 115 In forms, however, use radians. For example n using inte series, we sy that sin 0 3 0 for very small , Try tht on your calculator: you wl find thar wis roe ada mode bu notin degree mode. Asanoter example, consider ft dv(l 4 +2) cancun ln 1/4 = 0.785, Here are tan isnot an ange; itis the numeric value ofthe incor core answer 45 (obtained from a calculator in degree mode is wrong! Do not ase ence ae in reading an ar tan (rae sin oar cos) uf you are Sing an ong oe Sere Figure 32,0 = are un (9) 4. TERMINOLOGY AND NOTATION Both i and 7 are used to represent ./—1, j usually in any problem dealing with electricity since # is needed there for current. A physicist should be able to work with ‘ease using either symbol. We shall for consistency use é throughout this bool, sect TERMINOLOGY AND NOTATION _47 We often lbs» point with a single ter (Gor example, Pin Figure 32 and 4 in Figure 3.3) even though it requires two coordinates to locate the point. If you have staied vectors you will real that» vecor is represented by a sole lee, say w aithough it has in two dimensions) two components tis canomay to use 4 singe letter for a complex number even though we realize that it is actually a pair of real numbers. ‘Thus we write Zax ty = Hoos 0+ isin 0) sce pur ots ‘The quanti rh called the madefs or edule tate of and 0 called the angle of = (or the phase, or the argument, or the amplitude of =). In symbols: Je] = mod = VPey, (42) y (not iy), angle of = Phe values of 8 should be found from a diagram rather than a formula, although we do sometimes write @ = are tan (y/). An example shows this clearly -Lye-bre Vf Example. Weit ‘in polar form. Here we have (Figure 4.1). There are an infinite number of values of 8, (43) F 4 tnn, Where nm is any integer, positive or negative. The value 0-= 5n/4 is sometimes called the principal angle of the complex number 2 = ~~ i. Notice carefully, however, that this is not the same a the principal value 2/4 of arc tan 1 as defined in ealeu CURE 4 Jus. The angle of a complex number must be in the : - Sane quadrants the point representing the umber. or ur preent work any one of the values in (43) will do; here we would probably use either 52/4 or —3n/4, Then = V3( cos isin) = [We could also write 2 = ,/2 (cos 225° + # sin 225°).] ‘The complex number #— iy, obtained by changing the sign of jin called the compler conjugate or simply the conugate of =. We usually write the conjugate of 2x -+iy as 2=x~ iy. Sometimes we use 2* instead of 2 (in fields such as statistics oF quantum mechanies where the bar is used to mean an average value. sin is48 _COMPLEN NUMBERS cu? Notice carefully that the conjugate of 7) — 5 is —7i — 5; that is, ic is the j term whose sign is changed. Complex numbers come in conjugate pairs; for exame ple, the conjugate of 2 + 3/ is 2~ 37 and the conjugate of 2 — 3/is 2 + 31. Such a pair of points in the complex plane are mirror images of each other with the x axis as the mirror (Figure 4.2). Then in polar form, 2 and = have the same r value, but their 0 values are negatives of each other. If we write = = r(cos 0 + i sin 0), then “ay 2 = rfeos( ~6) + sin( —0) PROBLEMS, SECTION 4 Plot the following numbers in the complex plane. For each number, give the numerical value of its real part x, its imaginary party, its modulus or absolute value r, and one value of the ange 6 Label each plotted point in five ways as in Figure 33. Find and plot the complex conjugate of each number. oes hint a nif 4 fies 82 a not a3 9 ane Esme? = 0 feo + iain cos in nem E sin’) a eosin & Bee ee rr ee 16. Sos 0+ isin 0) Jaen 1 19 Sos 20° + isin 20°) 20, Hoos 110° — isin 110°) 5. COMPLEX ALGEBRA A. Simplifying to x + ly form Any complex number can be written in the rectangular form x + iy. To add, subtract, and multiply complex numbers, remember that they follow the ordinary rules of algebra and that? = =1 Example 1 (+P a4 ee P a1 4-122 To divide one complex number by another, first write the quotient as a fraction, Then reduce the fraction to rectangular form by multiplying numerator and denominator by the conjugate of the denominator; this makes the denominator real. Sees COMPLEX ALGEBRA 49 Example 2. 3th 6454? 34a 9 Iris sometimes easier to multiply or divide complex numbers in polar form. st sketch (or picture mentally) the sxample 3, To find (1 +) in polar form, we rome 4, so (1+ i) = V/2e4 we see that r= 9/2, and (4 t= (dete = 207 = FIGURE 5.1 FIGURE 52 from Figure §.2. The result i, of course, the same as in Example 1 Example 4. Write 1/[2(cos 20° +1 sin 20°)} in x41 form, By calculator we find 20° = 0.349 radians. Then 1 1 Yoos B+ i sin 2) ~ Hoos OSD +7 sin 0.39) ~ 2H" = geno. = 015fcos (0.349) + + sin (—0.349)) = 015(c08 0.389 ~ sin 0.49) = 047 - 0.17, by calculator in radian mode. We obtain the same result leaving the angle in degrees and using a calculator in degree mode: 0.5(cos 20° — isin 20°) = 0.47 ~ 0.17% PROBLEMS, SECTION 5 Follow the instructions for the problems in Section 4. Hint: First simplify each number to the Form x + iy or to the form 2". Use a calculator as needed to find 7,0, ftom x,y, ot t0 find x,y from r, 0; be careful to set it in radian mode or degree mode 8s appropriate 4 Peed 50_COMPLES NUMBERS. COMPLEX ALGEBRA st ch? sees ri ¢. Finding the Absolute Value of 2 : 8 162571 . Recall that the definition of || is 2 BE Ge TI a ats, 8, TP Gait N31 Ty (positive square root!) - 2 2 3+ 94, we can write Wo 12. Hoos 28° + fain 28) 3, = + san) . ea re Vee Vi ll M 2gemn Se oo negative since 22 = x? + y? and x and y are real. It (7 3.ip 18. (064 40.779 Note chat 2 is always real and not neg Find each of the following in rectangular (a + 4) form if 1 woe », 23 5. Complex Conjugate of a Complex Expression {It is easy t0 see that the conjugate of the sum of two complex numbers is the sum of the conjugates of the numbers. If Sst, and gen tiyy, then Athan itn —o, 182i +32). ‘The conjugate of 1+ 23) is Gr) FIO FID = ly +) — i, ty) 2-3) i+¢ then f+ ig, where and g are themselves complex, ther = i (not f— PROBLEMS, SECTION 5 28. Prove that the conjugate of the quotient of two complex numbers is the quotient of the onlugates, Also prove the corresponding statements for difference and product, Hint, It casier to prove the statements about produet and quotient using the polar coordinase to form; forthe difference, it is easier to use the rectangular form x + jy. is easy to see (5.1) using z = rel”. Then Remembering that the definition of [2] is |2 ecanulr form and computing i pch cet dh lung he fon ncaa fm and camping CESS Obie tthe thoi aw gen He ie my Ye quent othe ssl ales d's sar sateen fr prod PROBLEMS, SECTION 5 Use equation (5.1) to find the absolute value of 2. 31 Bw. (+29 2. 30 D. Complex Equations 1 always remember that Im working wth eats invling complex quant, we mut lass ent that asxmplet number seul» pi of ral number, To complex nbs ar if and only er weal ars are equal and thei imaginary prs are equal: For example, x + iy = 2+ 3i means += 2 and y = 3. In other words, any equaio ing complex numbers is rally Wo equations involving real numbersExample. Find x and y if (62) (et op? = 24 Since (x + Hy" = x* + 2isy — y*, (5.2) is equivalent to the tto real equations Day From the first equation y? second equation gives we find y= x oF WP=2 or 2? Since + is real, x* cannot be negative, Thus we find only that is, PROBLEMS, SECTION 5 Solve forall possible values ofthe real numbers x and y inthe fllowing equator Bortyaws 36 Die eBay ai 3, ". ° Booey 2-7 B rtyeyte Wo ety an MW Ory) 4 ares =0 ay +1) = R444 Ge—y— * ) + Gr —y— 1) 2x Hoortyed a 48 4. 4. (+09 8 : Bema St % H@se Vso) 50. Ito} ey = E. Graphs Eine the raphical representation ofthe complex number = as the point (x) in » WE can give geometrical meaning to equations and inequalities involving Example 1. Whats the curve made up of equation v mormer 7 Since 3) plane satisfying the the given equation is See.5 a 5 is the equation of a circle of radius 3 with center at the origin. Such an equation might describe, for example, the path of an clectron or of a satellite, (See Section F below.) 2, This is the circle (v — 1)? +)? Example 2. with > 05 this might be Example 3, (Angle of =) = 1/4. This is the half-line y the path of a light ray starting at the origin. Example 4. Rez >}. This is the half-plane x > 4 PROBLEMS, SECTION 5 Describe geometrically the set of points in the complex plane satisfying the following equations Sh [sl=2 8. Reno 8. [e-1<1 S [e-tl 1, the series diverges (Problem 6.14). For p = I (that is, ‘on the boundary of the circle of convergence) the series may cither converge or diverge. Tt may be difficult to find out which and we shall not in general need to consider the " ‘The four theorems about power series (Chapter 1, Section 11) are true also for comple series (eplace inerca! by cle of convergence) Also we ean now state for Prebrcm 2 what the cic of convergence is forthe quotient of tW0 series. Assume 10 has been canceled, Let r, and ry be the radii of start with that any common factor = convergence of the numerator and denominator series, Find the closest point to the srigin in the complex plane where the denominator series is zero; call the distance from the origin to this point s. Then the quotient series converges at least inside the smallest of the three circles of radii 7, r2, 5. (See Chapter 14, Section 2.) Example. Find the series for (sin )/[2(1 + s2)] and its circle of convergence. We shall soon sce that the series for sin = has the same form as the real series for sin x in Chapter 1. Using this fact we have Saas 73) 3 + Tas Bieeaast By ratio test, the series for (sin 2)/2 converges for all = (if you like, r, = 00), There is no r,, since we do not have an infinite series in the denominator. (The series we wrotes4_COMPLEX NUMMERS ee = cn? for + ti, *) is mot in the denominator.] The denominator 1+ 2? is zero when Points which are a distance 1 from the origin; thus s = I. By the theorem just stated, we then know that the series in (7.3) converges inside a circle of radius 1. We could also do this problem by noting that the series for 1/(1 +2") has a radius of convergence of 1. [See Chapter 1, (13.5), or use the ratio test.] Thus by Theorem 2 of Chapter 1, Section 11, the radius of convergence of the product series is 1. (Alo see Problem 17.) PROBLEMS, SECTION 7 Find the circle of convergence for each of the following complex power series, {equation (6.19) 9. wy (nytct Ber u uv. 13. oy Ont 3 4 ws F Enz4H we Se ninye n. (Check the statement that the series in equation (7.3) converges for ratio test. The best form of the series to une is lel <1 by using the 8. ELEMENTARY FUNCTIONS OF COMPLEX NUMBERS The so-called elementary functions are powers and roots, trigonometric and inverse ‘igonometric functions, logarithmic and exponential functions, and combinations of these. All these you can compute or find in tables, as long functions of real numbers. Now we want to find things like ‘These are not just curiosities for the amusement of the mather may turn up to be evaluated in applied problems. To be sure, the values of experimen. fal measurements are not imaginary, But the values of Re 2, Im s, |=], angle of = are fal, and these are the quantities which have experimental meaning. Meanwhile, mathe ematical solutions of problems may involve manipulations of complex numbers before ‘we arrive finally at a real answer to compare with experiment, Polynomials and rational functions ( ated. Here are some examples. if — 22+ 1, find 1 ~ 2) as you want them as #, sin (+9, oF Ini -matically inclined, but (quotients of polynomials) in = are easily evalu sec.8 ELEMENTARY FUN We substitute (1 — 1) for + t0 get fA -)= C= ALD H1 Iff(e) = (2? + Dz 3), find F4 — 2) Substitute = = (i — 2); (27 2 fe 9 = i-3 1-5" «OB Newt ne ant invests de posible memine of her anon of op nue We sl ew dee expen ori = ht he wl be he famines we Know forthe coping ral" exreone [or tame i 2x = 2 sin x cos x, or (d/dx)e* = e*]. We must, for consistency, define functions of aeraplex numbers so that any equations involving them reduce to correct real equations Sen eect bocomes 2 ay that iy when p= 0."These requirements wil be met if we define e bythe power series el) serie comers al ales ofthe complex mae = Probl 71) fo a othe sau af efor amy 2 Ife puts (ea, we gt the fair series for e% It is easy to show, by multiplying the series (Problem 1), that (8.2) tan dese enced eee eee ceteteete| & is defined by (8.1) by differentiating (8.1) term by term {Problem 2) Keon on PROBLEMS, SECTION 8 ‘Show from the power series (8.1) that Lodges 24 & 3. Find the power series for ¢ cos # and for e* sin x from the series for ¢ in the following ‘way: Write the series for &; put = = +. Show that e = e(cos y + sin y; take real and imaginary parts of the equation, and put y = 560_COMPLEX. NUMBERS 2 9. EULER'S FORMULA For real 8, we know from Chapter 1 the power s o sinO=0 ory # 05 = 1 nt From our definian (8.1), we cn write the sts for We write the series for e”, where @ is real ee GO GO G0)* 1+ 194 GE, CO" | Go” uo oUmarsimeeal Coe # (02) (he rearangement of terms i justified case th Now compare (9.1) and (9.2); the ast line in (022) just cox 8-4 1m B-We hen hone the very useful result we introduced in Section 3, known as Euler's formula: me eomm Thus we have justified writing any complex number as we did in (4.1), namely —] sec. EULER'S FORMULA 61 = ~x/2, From Figure 9.3, Se"? is re” with sry 0 —3i=-# 0,9 = —3, 90 3 2 is re with r= Vand 6 = Inn = n(2n); that is, 0 is an integral multiple of 2x. From Figure 9.4, x = so = 140151 FIGURE 9. It is often convenient to use Euler's formula when we want to multiply or divide ‘complex numbers. From (8.2) we obtain two familiar looking Faws of exponents which are now valid for imaginary exponents: 35) dee Remembering that any complex number can be written in the form re by (9.4), we get [eH Here are some examples of the use o ind (9.4). These problems can be done very les of the use of (9.3) and (9.4), These problems e ichly graphical rare : Examples, Find the values of 2e'"!*, o*, 3¢ (2, g2mei 2e*" is re” with r= 2, = 2/6, From Figure 9.1, ¥ = Viyohatin Jiti wo tte fey is re with r= 1, 0 = x. From Figure 9.2, x= -1, + i= 14 soe = 1 Note that r= Tan O= b. a. +i 15. dear ea dot 1 = J HV vn a, (LEY ase Ti, », 2. ( 4) a. ()" or 3 v es (4a a Grae (/3 - 98 Fa 2. (4,)" i 27. Show that for any realy, |e] = 1 Hence show that [| = efor every compler = 28. Show thatthe absolute value of soe vale of «produc of vo complex umber is ei to the product OF the absolute vas. Alo show thatthe able tale af the quae of on numbers isthe quotient ofthe absolute vals, Him Wate tenuis nthe nec Use Problems 27 and 28 east: Be 30. Je8H Bh. ise") a sey BL 2 He te) Be tem) 5 po +i a = ae, | rs iF x | SAND ROOTS OF COMPLEX NUMBERS 63 10. POWERS AND ROOTS OF COMPLEX NUMBERS Using the rules (9.6) for multiplication and division of complex numbers, we get De Moivre’s theorem (10.1) gee for any integral n, In words, to obtain the mth power of a complex number, we take the nth power of the modulus and multiply the angle by n. The case r =| is of particular interest. Then (10.1) becomes (10.2) ey" in n8. “| in 6" You can use this equation to find the formulas for sin 20, cos 20, sin 30, etc. (Problems 27 and 28) ‘The nth root of 2, 2!, means a complex number whose nth power is =. From (10.1) you can see that this is | a ‘This formula must be used with care (see Examples 2 to 4 below). Some examples will show how useful these formulas are Example 1. Find (1 +9)*. If +i, then r=/2, 0 Lai= 20 from Figure 5.1. ‘Then we have (14% 16 1 = 16, (See Figure 9.4.) 4, so in polar form Jaane m 6c Example 2. Find the cube rots of 8. We know that 2 is «cube rot of 8, but there are alo two complex cube rots of 8; let us see why. Plt the complex umber 8 (that is, += 8, 7 = 0) inthe complex plan; the polar coordinates of the point ae r= 8 and O'= 0, or 460", 720, 1OBD, exe. (We can use either degrees or raians here; vend the ‘end of Section 3.) Now by equation (10.3), 2! = re"; that is, to find the polar coordinates ofthe cube rot of a number re, we find the cube rot of and divide the angle by 3, Then the polar coordinates of @/& are (104) r= 2, = 0, 360°/3, 720/3, 1080/3 : a wy = 0%, 120%, 240%, 360° We plot these points in Figure 10.1. Observe that the point (2, 0°) and the point (2, 360°) are the same. The points in (10.4) are all’ on a circle of jew radius 2 and are equally spaced 360°/3 = 120° apart Starting with @ = 0, if we add 120° repeatedly, we just repeat the three angles shown. Thus, there are 12 240) exactly three cube roots for any number 2, always on a circle of radius 20. Show that (cosh 2 + sinh 2)" = cosh nz + sinh ne. Use this and a similar equation for to find formulas for cosh 2s and sinh 2s in terms of sinh z and cosh = 21. Using the defnions (122) and a table of exponents or ele, sketch graphs of sinh x and cosh x. = 22. Using (12.2) and (8.1, find the power series for sinh = and cosh = Find the real part, the imaginary part, and the absolute value of 23. cosh (i) 24. cos is) 28. sin (x —y) 26. cosh (2 — 31) 2. sin (4 439) 28. anh (149) Seo 13 _LOGARITHMS 71 Find each of the following i the «+ 4 form. sri 29. cosh 2 20, anh 2 5 Buns 4 in 35. cosh +2) 36 sinh (1+ 5) 13, LOGARITHMS In elementary mathematics you learned to find logarithms of positive numbers only; in fact, you may have been told that there were no logarithms of negative numbers. This is crue if you use only real numbers, but it is not true when we allow complex numbers as answers. We shall now see how to find the logarithm of any complex number = 9 0 (including negative real numbers as a special case). If (13.1) a then by definition 32) pains. (We use In for natural logarithms to avoid the cumbersome log, and to avoid confusion with logarithms to the base 10.) ‘We can write the law of exponents (8.2), using the letters of (13.1), as (13.3) ‘Taking logarithms of this equation, that is, using (13.1) and (13.2), we get (3.4) In 223 =m, +m, =Inz, +In 2 ‘This is the familiar law for the logarithm of a product, justified now for complex numbers. We can then find the real and imaginary parts of the logarithm of a complex ‘number from the equation (13.5) where Ln r means the ordinary real logarithm to the base ¢ of the real positive ‘number 7, | win z=In (re) = Lar + Ine” = Lar + 10, Since @ has an infinite number of values (all differing by multiples of 2n), a complex wumber has infinitely many logarithms, differing from each other by multiples of 27 The principal value of In = (often written as Ln 2) is the one using the principal value of 6, that is 0 < @ < 2x, (Some books use —x <0 < x.)cna Example 1. Find In (—1). From Figure 9.2, we see that the polar coordinates of the point 2 = —larer= land @=n, —z, 3m, In (=1) = Ln (1) + (x nm) Example 2. ind In (1-40). From Figure 5.1, for ¢=144 we find v= (/7, and O w/h & Dan. Then In(lag= a J+ (Fb aon) = 07 + (Gs a) Even a positive real number now has infinitely many logarithms, since its angle can be taken as 0, 2x, —2n, ete. Only one of these logarithms is real, namely the principal value Ln r using the angle # = 0. 14, COMPLEX ROOTS AND POWERS For real positive numbers, the equation In a? = # In a is equivalent to a =e, We define complex powers by the same formula with complex a and b. By definition, for complex a and & (a # ¢), aw) om ] [The case a =e is excluded because we have already defined powers of ¢ by (8.1).] Since In a is multiple valued (because of the infinite number of values of 0), powers a? are usually multiple valued, and unless you want just the principal value of In 2 or of a? You must use all values of @. In the following examples we find all valucs of each complex power and write the answers in the x + ty form, Example 1. Find all values of 1, From Figure 5.2, and equation (13.5) we find Ind Ln 1 + (x/2 £ Inn) = i(n/2 + 2nn) since Ln 1 = 0. Then, by equation (141), where ¢* = 23.14 (by calculator). Note the infinite set of values of #4 all real! Also read the end of Section 3, and note that here the final step is not to find sine or cosine of x + 4mm; thus, in finding In # = 48, we must not write 0 in degrees Example 2. Find all values of i! Using Ini from Example 1 we have Hat a gisitrn — ent, Now om +1 when m is even (Fig. 9.4), and | when m is odd (Fig. 9.2). Thus, Lti COMPLEX ROOTS AND POWERS 73 Sew COMPLEX ROOTS AND POWERS _73 igure §.1. Notice that although In has an infinite set of values, we find just 40 weer ae we should for naguvre ror (Campus she method of Secon 1 ‘which is easier for this problem.) Example 3, Find all values of (+a!) Using In (1 +1) from Exampl Section 13, we have (1 ika v2 i422) fpr ag Hn gilt tami dg 20 ata = dein ta tente2te (since e281 = 1) Siete 0 (th — Ln 9) + i sin (n/t — Ln J] by calculator. laws of exponents for complex Sometimes we have to be careful about using the laws of exponents for com numbers. For example, are a and the same? This means, are "4 ¢'°* and 2°91 the same? We do not have to take the same value of Ina in defining a and Tt is also not always true that (ab)' and a'#* 4°, so aa can have more values than a" have the same values or that (a"V' and a have the same values (Problem 24), PROBLEMS, SECTION 14 Find one value of exch of the following in x + ay form, nes) 9% (1) 1 (i ‘) 8. 19 cos (x +4 In 2) 20. 21 [in (-)1] an have more values than a. As examples compare ne pes 24. Show that s @ CaP and (= Uy andINVERSE TRIGONOMETRIC AND HYPERBK hod, you can find all the inverse trigonometric and hyperbolic logarithms. Let us look at one more example. gral tables you may find for the indefinite integral sine t+ VP $0 sinh" *G)2) anh“) are cos (5/4)SETS OF LINEAR EQUATIONS, ROW REDUCTION _ 8% S4_LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS ns (© Substitute ==1 into the second (€) Write the equations corresponding t0 cauation to get.y = —1; then substitute the matrix (d) and solve just a» in the {he values of = and (in general) y into left column; this is called. fact substine the first equation to get 2r=3, r= 3. tion ‘The solution is (x,y, ~10. As you can see, there is less writing in the matrix solution, but it exactly parallels the solution of the equations. Let us list the allowed operations on the matriv. (called clementary row operations) (2.3) i. Interchange two rows [see step (b)]i ii, Multiply (or divide) a tow by a (nonzero) constant [see step (@)}; iii, Add a multiple of one row to another; this inchudes subtracting, that is, iz using a negative multiple [sce steps (a) and (c)] | [The matrix process we have just carried out is known as Gausian elimination or row {eduction and the final matrix with zeros at the lower left (below the starsteps shown in (@)] is said to be in echelon (stairstep) form. The numbers on the starsteps [2,1 in (d)] are called pivots; took back at the example and you will see that we wsed ‘each pivot to obtain zeros below it Using the elementary row operations (2.3), the systematic procedure for solving a set ‘of equations is (note the example above). 1 Be sure the equations are in standard form (x terms lined up, tc., constants on right-hand side) and write the corresponding matrix, % [I necessary, interchange or combine rows to obtain a suitable pivot in the upper lefchand comer of the matrix. To avoid factions, try to obtain a pivot which divides evenly into the numbers below it. 3 Use the pivor and elementary row operations (2.3 il) to make all numbers below the pivot zeros. ‘+ Now ignore the frst row and first column of the matrix and repeat steps 2 and 3 on the remaining submatrix. If there are more than threc rows, next ignore the first two rows and first two columns, and so on until the matrix is in echelon form Finally, write the equations corresponding to the n ing at the bottom, back substitute to find the solution, ix in echelon form, and, start- Example. Solve the set of equations See.2 eduction press withowt stoping 19 is tne we shall jot cary trough the rw reduction pi Cin ish wp She edn be el ily Hh) are Ne rac ame ase feceaicy Pega ame fic T assis cave econ) Vor eetaer owen, © uar awe and tn Int io 4 condensed horsion beside each arrow what wreaton doing (this makes i easier to check back for errors). For example, to camaro ting two times row 1 from it, we write iene tat we change row 2 by sahtncang “ie ote Bini oie ht vnc wah fw we RoR a Haicte that we divide ow # BY (2) emit RA-+ (2) Then the row rlucion irecess 2 fllows 0 w -1 ow 1 1 224 08 22 4 fae 4s 100 32a 3 4 324 Loo rd Loo id sem, (0-2 4 23 sons, [2-2-4 0-3) arom z 0 10 1 10] Re 3Rr Oo 0 1 1 10 se 3 4 02 et ar) boos 242s meso, (0-2-4 O22 : aI vonas mee ty 00212 1 0 0 oo 1 a wom, (0 2 4 rts aaa Olea 0 0 1 0 Dot Ob oof aimee By back substitution we find =h ly=S—4e— Wests ral—pe ~3. -3,4 -1, 4). ‘Then the solution is (x, 94 5 2)LINEAR EQU: IONS; VECTORS, MATRICES, AND DETERMINANTS. cn PROBLEMS, SECTION 2 Solve the following sets of equations by reducing the matrix to row echelon form. ; fee ya o 4 1 Sr- 24% Bt smutty % (a= wHB0=3 Weak Bpte- Ha arty + wp [re5tac2 Pta~ rund to Problems 11 and 12 use row reduction to find complex numbers 1, 2, m satisfying the given uations, (Equations like these occur in the analysis of alternating current. cinits + ere, ve) dst dtae+ tie uy, 7 G+ 30+ +29 = 3444 (ow ~ 0 wal-i (tu i+ 20 -pe =a 49) iw + ML +p = 385 (at +9046 + 2pm 1421 13, Im order to expand in partial fractions Porte (rake ema 0 GaP we need to solve the following equations for 4, B, C: [4+ 2 mp 2A + (a +B © (44 ab aGay ‘Use row reduction to find 4, B, C in terms of a, b p, 4, and r. Kirchhoff’ laws for electrical networks such as those shown in the diagrams below are: (I) At «any junction the sum of all the currents flowing toward the junction is equal to the suim ofall the currents flowing away. (2) The algebraic sum of all the changes in potential around any closed loop is zero. (To apply this, go around each loop listing as increases in potential. volages when You go through a battery from — to +, and JR values when you go through a resistance ‘against the current 1; decreases in potential correspond to the opposite directions) Comaenn DI seed E eee MINANTS; CRAMER'S RULE #7 ‘You should not waste any time trying to decide which direction currents “really” flow, Just mak an arrow for each J, (as shown, say), and write and solve the Kirchhoffs law equations ssuming that the currents “really” flow as you have marked them. ‘Then any currents which ‘Speally” flow opposite tothe direction you have marked will come out negative ie we Ta} 3°98 oom van HIE a “Nh ih 3. DETERMINANTS; CRAMER'S RULE We have said that a matrix is simply a display of a set of numbers, it does nor have a numerical value. For a square matrix, however, there is a useful number called the determinant of the matrix. We shall outline some of the facts about determinants without proofs (for more details, see linear algebra texts), Evaluating Determinants To indicate that we mean the determinant of a square matrix (written det 4), we replace the large parentheses (or brackets or double bas) inclosing A by single bars. The value of det 4 it isa 1 by 1 matrix is just the value of the single element. For a2 by 2 matrix, a ao Gay 4 ( ) det A id — be ed 4 Equation (3.1) gives the value of a second order determinant. We shall describe how t0 evaluate determinants of higher order First we need some notation and definitions. It is convenient to write an mth order determinant like this: 2) an{4__LINFAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS chs Novice shat 235 isthe clement in the second row and the cird column; that ithe frst subscript is the number of the row and the second subscript fs the number of the column in which the element is. Thus the element a, i in tow # and column js an abbreviation for the determinant in (3.2), we sometimes write simply lay. that is, the determinant whose elements area. In this form i looks exactly ke the absolute value of the clement a; and you have to tell ftom the context which of these meanings i intended If we remove one row and one column fram a determinant of order m we have a decerminanc of order n — 1. Let us remove the row and column eantaning the clement 4, and call the remaining determinant M,,. The determinant is alle the miner of ay. For example, in the determinant G3) 73 ha naa|t “f obtained by crossing off the row and column containing 4, as shown. ‘The signed minor (—1'~4M,; is called the cofactor of a,j. In (3.3), the element + is in the second row ((= 2) and third column (j = 3), s0 #4 = 5, and the cofactor of 4 is (—1)°Mf,5 = M.It is very convenient to get the proper sign (plus or minus) for the factor (—1)!*? by thinking of a checkerboard of plus and minus signs like this G4) ete. bent e1tt oe - + ‘Then the sign (—1)'*/ to be attached to M,, is just the checkerboard sign in the same Position as «,;. For the element «25, you can see that the checkerboard sign is minus. Now we can easily say how to find the value of a determinant: Multiply each element of one row (or one column) by its cofactor and add the results, It can be shown that we get the same answer whichever row or column we use. sec} DETERMINANTS; CRAMER'S RULE _ 59 Example 1. Let us evaluate the determinant in (3.3) using elements of the third column. We get Ta 4-114 5-38 148, [As a check, using elements of the first row, we get 74 fit E 1s 73 = 48. A | N+ 18 42=4 ‘The method of evaluating a determinant which we have described here is one form of Laplace's development of a determinant. If the determinant is of fourth order (or higher), using the Laplace development ance gives us a set of determinants of order one less than we started with; then we use the Laplace development all over again 10 evaluate each of these, and so on until we get determinants of second order which we know how to evaluate. This is obviously a lot of work! We will see below how to simplify the calculation. A word of warning to anyone who has learned a special method of evaluating a third-order determinant by recopying columns to the right and multi~ plying along diagonals: this method does not work for fourth order (and higher). Usetul Facts About Determinants We state these facts without proof. (See algebra books for proofs.) 1. If each element of one row (or one column) of a determinant is multiplied by a |. number £, the value of the determinant is multiplied by & 2. The value of a determinant is zero if 4 all elements of one row (or column) are zero; or if ’. two rows (or two columns) are identical; oF if . two rows (or two columns) ate proportional. 3. If two rows (or two columns) of a determinant are interchanged, the value of the determinant changes sign. 4. The value of a determinant is unchanged if 4. rows are written as columns and columns as rows; or if '. we add to cach clement of one row, & times the corresponding clement of another row, where & is any number (and a similar statement for columns). Let us look at a few examples of the use of these facts Example 2. Find the equation of @ plane through the three given points (0, 0, 0), (1, 2, 5), and (2, ~1, 0).90__LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS cus We shall verify that the answer in determinant form is t| H 0. By a Laplace development using elements of the first row, we would find that this is a linear equation in x, y, £5 thus it represents a plane. We need now to show that the three given points are in the plane. Suppose (x, y, =) = (0, 0, 0); then the first two rows of the determinant are identical and by Fact 2b the determinant is zero. Similarly if the point (x, y, 2) is either of the other given points, two rows of the determinant are identical and the determinant is zero, soll three points lie in the plane. Example 3. Evaluate the determinant 0 4 6 D=|-1 0 ¢ b-- 0 If we interchange rows and columns in D, then by Facts 4a and 1 we have 0 -« 6] 0 2 ~} aa ale meee bo 0 br 0 where in the last step we have factored —1 out of each column by Fact 1. Thus we have D = —D,s0 D=0. We can use Facts 1 to 4 to simplify finding the value of a determinant. The idea is to get as many zeros as possible in some row or column in order to have fewer terms in the Laplace development. This can be done efficiently by a method very similar to the process of reducing a matrix to row echelon form, We must be careful about two things, however. 1, We must keep track of row interchanges; each interchange multiplies the deter- minant by (~1) (see Fact 3). 2. We cannot just cancel a factor from a row; we must keep it as a factor in our answer. Finally, we have one freedom here which we did not have in the reduction to row echelon form. By Fact 4, we may do column operations as well as row operations, subject, of course, to the same two cautions just outlined above, We illustrate these ideas by evaluating a determinant. Following the method of row reduction makes the work routine; however, judicious use of column operations may keep the numbers smaller and so reduce the amount of work. We list the operations as we did for row reduction, seo.3 - DETERMINAN: 'S: CRAME! 'S RULE 91 Example 4. Evaluate the determinant D {$3 9 3] fantercange column 1 and column 4) o- | fore mins sign) _ [3 1 Fal tesctor2 out oFcotimn 3 ' SS —————— 3070109 Row reduction: R2— 3R1 =, 09 1 4 R3-RI Oar a 3 1 o 4 a2] 21-3) rape deeopment) 0-1 23 2 io (Factor 1 out of cohumn 1) 7 po + Hlmerchange columns 1 and 2 (Ce another = 1) ) yo 2 =3}_———_____ 1301 {Row reduction and Laplace development} 13 | =-2)0 103 1 3)_ eno = 0 39] 25 ge t= Cramer's Rule There is a formula in terms of determinants for the solution of linear equations in m unknowns, known as Cramer's rule. For a system of linear equations with numerical coefficients, row reduction (Sections 2 and 8) involves less arithmetic than evaluation of the determinants needed in Cramer's rule. However, if the coefficients are not numerical (see Example 6 and Problems 16 to 19 below and also Chapter 4, Section 7), and for theoretical purposes where a formula rather than a numerical answer is wanted, row reduction is not practical, and Cramer's rule is very Useful. Since examples of actual problems of this kind are usually rather complicated, it is best to learn the method of Cramer's rule by practicing first on equations with ‘numerical coefficients Let us first show the use of Cramer's rule to solve two equations in two unknowns. ‘Then we will generalize it to m equations in n unknowns. Consider the set of equations G3) ae thy sey, ane thy =e,92_LINEAR EQUATIONS; VECTORS, MATRICES, 1 we mulkiply the frst equation by bs, the second by by, and then subtract the results and solve for ry we get itaybs ag, 0) Goa) yaahooh aby = a2by Solving for y in a similar way, we get (3.60) Using the definition (3.1) of a second order determinant, we can write the solutions (3.6) of (3.5) in the form a hf | ler bol al a Ya: al ah b It is helpful in remembering (3.7) to say in words how we find the correct determin- ants. First, the equations must be written in standard form as for row reduction (Section 2). ‘Then if we simply write the array of coefficients on the left-hand side of 3.5), these form the denominator determinant in (3.7). This determinant (which we shall denote by D) is called the determinant of the coefficients, To find the numerator determinant for x, start with D, erase the x coefficients a, and az, and replace them by the constants ¢, and c from the right-hand sides of the equations. Similarly, we replace the y coefficients in D by the constant terms to find the numerator determinant in y. Example 5. Use (3.7) to solve the set of equations De + By ayy We find ‘This method of solution of a set of linear equations is called Cramer's rule. It may be used to solve equations in # unknowns if D +0; the solution then consists of fone value for each unknown. The denominator determinant D is the m by n deter- ‘minant of the coefficients when the equations are arranged in standard form. The numerator determinant for each unknown is the determinant obtained by replacing the column of coefficients of that unknown in D by the constant terms from the right-hand sides of the equations. Then to find the unknowns, we must evaluate [exch of the determinants and divide, seed DETERMINANTS; CRAMER'S RULE_ 93 the unknowns are letters, Cramer's rule is particularly useful when the coefficients of th . and especially when, as often happens in practice, we need a formula for just one of the unknowns. Example 6. Use Cramer's rule to find x in the following equations. (abe abe ay + bi (a+ Bs — ala — bly We write D and expand it by a Laplace development on the third column: a-b 02 ab a+b ~aa-b) 0 = A —a?Ha — 8) + a%(a + by) — Hala — 4)7) = a(2a? + 7b +b. ‘Then we write the x determinant, add the second column to the first column to introduce another zero in the third row, and expand using the third row jeae 2 02 1 oO ‘ -—° 6 lae-8) a6 0 aa) 0 _ aa — 8) “OD 48 2) ala — He? + 64) +209) | ® al is e = \ . ® A Q >| " a a ° ex © FIGURE 47 A eae in Figure 4.7. If we add the vector 4B to the vector A (head to tail, avin Figure 47), we have exo fom pont othe mide’ of he oppose side cee eel have the vector (A + 4B) = 4A +4B extending from 0 to on tio the 3 point for cach. We prove this by showing that P is the “3 point” on the re me aa ie vector oes Ro 8 (Figure 4.7c) is $A + B; this is the 8-48 Ths P and P axe the sme pinta ee dias have te 4 points” there, Note that we have made no reference to "yn ore components in this proof. ” # coordinate este oF © PROBLEMS, SECTION 4 1. Draw diagrams and prove (4.1) 2. Given the vectors making the given angles with the positive x axis A of magnitude 5, 0 = 45 Seed vecrors _99 B of magnitude 3, 0 = 30", Gof magnitude 7, 8 = 120° (a) Draw diagrams representing 2A, A — 2B, C ~ B, 48 —4C. (b) Draw diagrams to show that A+B=B+A, A-(B-O)=(A-B)+C, (ASB)+C=(A+OQ+B, (APB A B-O, =B.~ ‘Use vectors to prove the fllowing theorems from geometry 4. The diagonals of a parallelogram bisect each other. ‘4. ‘The line segment joining the midpoints of two sides of any erangle is parallel to the third side and half iss length. 5. Ina parallelogram, the two lines from one corner to the midpoints of the two opposite sides trisect the diagonal they cross, 6 In any quadrilateral (four-sided figure with sides of various lengths and—in general—four diferent angles), the lines joining the midpoints of opposite sides bisect each other. Hin Label three sides A, B, Cj what isthe vector along the fourth side? A tine through the midpoint of one side of a triangle and parallel toa second side biseets the third side. Hint: Call parallel vectors A and cA, 8. The median ofa trapezoid means the line joining the midpoints of the two nonparalel ses Prove that the median bisects both diagonals; that the median is parallel to the two parallel thas and equal to half the sum of their lengths. We have discussed in some detail the geometric method of adding vectors (parallelogram law or head to til addition) and its importance in stating and proving geometric and physical facts without the intrusion of a speciah coordinate system. There fre, however, many cases in which algebraic methods (using components relative 10 a particular coordinate system) are better, We shall discuss this next Vectors in Terms of Components We consider a set of rectangular axes as in Figure 4.8, Let the vector i be a unit vector in the positive x direction (out of the parr fonard you), and let j and k be unit vectors in the positive y and 2 directions, 1f 4, nd cf, are the scalar components ofa vector in the (x) plane, then id and 34, are ite vector components, and their sum is the vector A (Figure 49). Aid, +i, 14, FIGURE 48 FIGURE 49100 LINEAR EQUATIONS; VECTORS, MATRICES, AND_ DETERMINANTS cus Similarly, in three dimensions A A, +j4, + kA, It is easy to add (or subtract) vectors in this form: Eeecie Tf A and B are vectors in two A+B=(4, +54) + 68, +58) fos just the familiar result of adding components; the unit vectors i and j serve to Keep tack of the separate components and allow us t0 write A as a single algcheaty expression, Ay By) + i(A, +B) Multiplication of Vectors ‘There kinds of are two kinds of product of two vectors, One, caled the scalar product (or dot product or inter produc), gives a result which 4 Salar: the other, called the vector product (or cross product or outer product), given 5 Scalar Product By dfnton, the saa pod of and B (ten AB) scr equ (0 the magnitude of times the magnitude of Bence te ice gs angle @ between A and B: 7 * — = 7 You should observe from (4.2) that the comm cation: tative law (4.3) holds for scalar multipli- (43) A-B=B-A A wseful interpretation of the dot product is shown in Figure 4.10. Since BI cos 0 is [BI =8, 1A) =6 Projection of Bon A AB=6-4224 Or, projection of A on B B-A=3-8=24 FIGURE 4.10 the projection of B on A, we can write (44) ACB or alternatively AA| times (projection of B on A) A+ B= |B| times (projection of A on B), Also we find from (4.2) that (45) A+ A= IAP cos 0° = AP? = 42, VECTORS 101 sect — 2 c atthe square fometimes A? is written instead of |A|? or A®; you should understand that the sq Srt vector always means the square of ts magnitude or its dot product with el From Figure +11 we can sce that the projection of B+ C on A is equal to the projetion of B on A plus the projection of C an A. Then by (44) of (46) -A-(B+C)= [Al times (projection of (B + C) on A) = [A| times (projection of B on A + projection of € on A) HA SBHASC FIGURE 4:11 ‘This is the distributive law for scalar multiplication. By (4.8) we get also BHA-C 7) (B+Q-A=B-A4COA I is useful’ to know A - B in component form as well as our definition form (+2) We write an A+ B= (id, +4, +kA,) « (GB, + 5B, + KBD). By the distributive law we can multiply this out getting nine terms such as 4, Bi * iy A,Byi*j, and so on, Using the definition of the scalar product, we get «il cos = 11-1 = Ay and similarly, j j= 1 kk 1s 9) 552 fil jf eas 90° = 1-1 0-0, and similarly, i -k=j “k= 0 Using (4.9) in (4.8), we get 4.19) A-Ba A,B, + A,B, + A,B, ‘| Equation (4.10) is an important formula which should be memorized. ‘There are several immediate uses of this formula and of the dot product, we can find the angle between Angle Between Two Vectors Given the vectors, oe them by using both (4.2) and (4.10) and solving for cos 8. For example, let us find # angle between the vectors A = 3i + 6j + 9k and B= —2i + 3) +k. Using (4.2) and (4.10) we get As B= [AI |B) cos i) 1A 34/4 cos 0 = 21, Free102_LINEAR EQUATIONS: VEGTORS, MATRICES, AND DETERMINANTS cus Perpendicular and Parallel Vectors cos = 0; thus [wm A+ A,B, If two vectors are perpendicular, then if A and B are perpendicular vectors. IF two vectors are parallel, their components are proportional; thus (when no compox nents are zero) (13) VECTORS _103 geet ‘Then we have the useful results Serie ie oie ees predueat ae Gl (Of course, if B, = 0, then A, = 0, ete.) Vector Product The vector or cross product of A and B is written Ax B. By definition, A x B is a vector whose magnitude and direction are given as follows | ‘The magnitude of A x B is (uy) IA.) = 1A11B| sin 0, wore is the postive angle (< 180") berween A and B ‘The direction of A x B is perpendicular to the plane of A B and B and inthe sense C of advance of 4 ight-handed o Screw tated from Avo B asin Figure 4.2 ) “ raure 412 It is convenient to find the direction of C= A x B by the following righthand rule. ‘Think of grasping the line C (or a screwdriver driving a rightchanded screw in the direction C) with the right hand. The fingers then curl in the direction of rotation of A into B (arrow in Figure 4.12) and the thumb points along C = Ax B. Perhaps the most startling result of the vector product definition is that A x B and B % A are not equal; in fact, A x B= ~B x A. In mathematical language, vector ‘multiplication is not commutative We find from (4.14) that the cross produet of any two parallel (or antiparallel) vectors has magnitude |A x B] = 4B sin 0° = 0 (or AB sin 180° = 0). Thus Ax B= Oif A and B are parallel or antiparallel, I" 15) “ Ax A =0 for any A. kxiaj A ud nyt renter hes 0 wise ly rnd ink inal tga 1, ening and she de nce peste # dt, we eerste products (or example, ix} =H), reading the ober egate products (or amplei k= ~}) mY FIGURE 4.14 FIGURE 4.13 7 way we have labeled Ie is well to ote here that the results (4.17) depend upon the way we have Sheed the ane in Figure 413, We have arranged the (s 2) axes 60 that turin ofS axis ino they axis (Qhrough 90" coresponds to the rotation ofa righihanded sere advancing inthe postive < direction, Such a coordinate stem is cae ghost ‘pstem. If we used a left-handed system (say interchanging + and y), ‘This would be confusing; conse quaions in (4.17) would have thet signs changed. This re quently, we practically always use right-handed aan AXP O) AM BEA XC. Ie for difficult but very tedious to prove104_LINEAR EQUATIONS; VECTORS, MATRICES, AND_DETERMINANTS as (ee, for example, Thomas and Finney, . 509) so we shall assume and use it without proof. Then we have (418) A x B= G4, + 54, + kA) x GB, + 5B, + kB) L The second line in (4.18) is obtained by multiplying out the first line (getting. nine Products) and using (4.16) and (4.17). The determinant in (4.18) is the most convenient way to remember the component form of the vector product. You should verify (by ‘multiplying out the determinant using the elements of the first row) that the determi. nant is really identical with what is in the line above it. Since Ax B is a vector perpendicular to A and to B, we can use (4.18) to find a vector perpendicular to two given vectors. = iA, B, ~ A,B,) + (A,B, — A,B.) + K(A,B, ~ A,B.) k ij A, A, B. B, a Example. Find a vector perpendicular to both A. 21+ j—kand B =i +3) — 2k ij ok AxBa]20 01-1) =i(-243)— (44) +h6—1) la see] i+ 3} + 5k PROBLEMS, SECTION 4 9. Let A= 2) + 3jand B= 3B, A~B, B+ 2A, 4A +B), 1. 1 A+Be4j—~iand AB how to find A and BB geometrically 5i. Show graphically, and find algebraically, the vectors —, 5+ 3), find A and B algebraically. Show by a diagram M. Let 31 —j + 4h, 7)~ 2k, i~ 3) + 1 be three vectors with tails at the origin. ‘Then their heads determine three points 4, B, C in space which form a triangle. Find vectors represen ting the sides AB, BC, C4 in that order and direction (for example, 4 to B, not B to 1) and show that the sum of these vectors is zero, 12. Find the angle between the vectors A= =: 13, If A= 4i— 3k and B= —2i + 2)~k, find the scalar projection of A on B, the scalar Projection of B on A, and the cosine of the angle between A and B. + j—2k and B= 243}, 14, Find the angles between (a) the space diagonals of a cube; (b) a space diagonal and an ‘ge; (©) a space diagonal and a diagonal ofa face, 15, Let A 4. 2k (a) Find a amit vector in the same direction a8 A. Hint: Divide A by IAL. (b) Find 2 vector in the same direction as A but of magnitude 12. (c) Find « veetee Perpendicular co A. Hint: There are many such vectors; you ate 0 find one of them, {d) Find a unit vector perpendicular to A. See hint in (a). ss 5 LINES AND PLY j= 2) + 4K, and another unit 16 er inthe same dcon Borat ak. Show tht the ver sm of thse ‘Steet the angle between A and B. Hin: Shteh the hombus hing the two unit Yectors a ace ies tn Find tree sectors (none oF them pall 1 4 coordinate avs) which have lengths and iesins such tha they could he ade ico ih tangle — 2k are orthogonal (perpendicular). Find a third vector s 4 th and 51+ 18. Show thar 21—j +4 perpendicular to both 1 to both i — 55 + 2h and Si — j — 4K 19. Find a vector perpendicular to both i — 31 20. Find a vector perpendicular co both i + j and i — 2k 21. Show that BJA| + A[B| and AJB| ~ BJA| are orthogonal en 22, Square {A +B); interpret your result geometrically, Mint: Your answer is 2 Taw which you Tearned in trigonometry 2B. FA =21— 3) +k and A B= 0, dos it follow thar B = 0? (Either prove that it does oF give a specific cumple 10 show that it doesn't) Answer the same question if A x B = 0. ‘And again answer the same question if A B= O und A x B= 0 24. What is the value of (A x B)? + (A BY? Use vectors as in Problems 3 to 8, and also the dot and crass product, (0 prove the following theorems from geometry rice the sum of the 25, The sum of the squares of the diagonals of a parallelogram is equal to twice the squates of two adjacent sides ofthe paralletogram. 26, The median to the base of an isosceles triangle is perpendicular tothe base 27. Ina kite (foursided figure made up of two pairs of equal adjacent sides) the diagonals are perpendicular. 28, The diagonals of a rhombus are perpendicular and bisect each other, 5. LINES AND PLANES A from (1, 2, 3) to (x, 9 2) is Aar-C= (uy Asixtiy the G4 2438 =U, 2a hy Be—H oF (ie — DF = 2) + RE =D.106 _LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS cus (2.9.8) FIGURE 5:1 Thus we have two ways of writing vector equations; we m: Note the possible advantage of writing (1, 0, —2) for i — bss written, there is less chance of accidentally confusing i — 2k with i — A a on the a hand, 5j is simpler than (0, 5, 0). ae ae be eqation ofa stright ae through (ro,95) with 6» to, Seppe, inn of he slp, we ate gen ves in he dtm ofthe tine, sy ia + jb (Figure 5.2). Then the line through (¥q, yo) and in the direetion A ix FIGURE 5.2 determined and we should be able to write its equation. from (x9, yo) to any fast ¥— x9 and y — yo 62) ‘The directed ne segment point (x9) onthe line the vector f~ re with componcare i — x0) + 10 — 90) This vector is parallel to A = ia + jb, Now if t parallel, their componen Wo vectors are parallel, their compon are proportional. ‘Thus we can write (for a, b #0) 63) or Loot seo.5 LINES AND PLANES _107 ‘This is the equation of the given straight line. As a check we see that the slope of the line is m = Bja, so (5.3) is the same as (5.1) “Another way to write this equation is to say that if'r — ty and A\ are parallel vectors, ‘one is some scalar multiple of the other, that is, 6a) Fore AQ or rato tA ‘phere tis the scalar multiple. We can think of fas a parameter; the component form of (Gul) is a set of parametric equations of the Hine, namely Yo + bt. 65) inning 1 eso orginal equation ofthe ie lint i ens can be ase, We want the equations ofa straight ting through a given point (2p, Yoo 20) and parallel to a given vector A= ai +6) + ck Hine hough a on Dd 7A cre iJon) a8) Ceo) ee ern Stopnents # age y Yer #20 ae propor 1 te vane bro Aand we hve (commetri equations ofa straight line, ae £9) 6.6) If, for instance, happens to be zero, we would have to write (5.6) in the form xa% (symmetric equations of a straight Gr Fine when As in the two-dimensional case, equations (5.6) ot (5.7) could be written remy tah a (parametric equations ee Sot AK oi y=Io+ be ora straight line) raiy tet, ‘The parametric equations (5.8) have a particu- larly useful interpretation when the parameter t ‘means time. Consider a particle m (electron or bil~ liard ball) moving along the straight line Lin Figure 5.3. Position yourself at the origin and watch m move from Po to P along /.. Your line of sight is the vector 1; it swings from ro at =O t0 7 ig + At at time’. Note that the velocity of m is | Ais a vector along the line of motion. FIGURE 5.3 dejaSAND DETERMINANT: ns Going ack 10 90 through the point (x9, yo) and perper SNE ena sees throu the poi Cx.) and perpendicular to's given settn Nad hee ge Ft = (8 soi #6 — yj tes along the line ‘This time we want this vector perpendicular sw Ny recall that two vectors are perpendicular if their dot product is zero, Setting sro ar prendclar P ting the dot product of N and 69 aly ~ 0) + AG yo roa 6 ‘This the desired equation ofthe stright ine Z from Figure 5.4 that the slope of the line L is pe " ae tan 0= ~cor d= —a/b, FIGURE 5.4 FIGURE ss In three dimensions, we use this method to write the is @ given point in the plane (Figure 3.5) equation ofa pane. IF tap yore and (5.3, =) is any ether point in the plane. the Sere Fa Fy = — soli +O — roi + (2 ~ 2a)k isin the plane. If N = ai + 4j + ck is normal (perpendicular) to the plane, then N and F To are perpendicular, so the equation of the plane is N = (r —r,) 0, or -— js als — 0) + BO — 90) + (2 — 29) or art by tes (equation ofa plane) | where d = ag + by9 + 0 (i ato + bye + | If we are given e are given equations like the ones above, we can read backwards an on ; ones above, we can read backwards to find A or N. pia hat the equations (5.6), (5.7), and (5.8) are the equations of a straight which is parallel to the vector A = ai + bj + ck, and either equation in (3.10) is the equation of a plane perpendicular to the vector N = ui + 6) 4 ok Example 1. Fg the equation of he plane BO, 3, 0), CO, 1, —2). “ ' fehb. seo$ Te {A vector joining any pair of the given points lies ia the plane, Two such vectors are 3.0) (1 1, I) = (3, 2, 1) and AC = (1, 0, —3). The cross product of AB = these ewo vectors is perpendicular ro the plane. This is _ , fi ik N= (4B) x (AC)=|3 2-1) = ~61 +8) — 2k. ee Now we write the equation of the plane with normal direction N through one of the given points, say B, by using (5.10) =x = 2) + 8G = 3) — 0 or jr~dy tote (Note that we could have divided N by —2 to save arithmetic.) Example 2. Find the equations of a line through (1, 0, ~2) and perpendicular to the plane of Example 1 ‘The veetor 31 — 4] + k is perpendicular to the plane of Example 1 and so parallel 1o the desired line, Thus by (5.6) the symmetric equations of the line are +2 ao (=) ~ 2k + (31 — 4) + ky or, if you by (5.8) the parametric equations of the line are r like, # =, 0, —2) +, 4 I Vectors give us a very convenient way of finding distances between points and lines ‘or planes. Suppose we want to find the distance from a point P to the plane (5.10) FIGURE 5.6 (This means the perpendicular distance.) (See Figure 5.6.) We pick any point Q we like in the plane (just by looking at the equation of the plane and thinking of some simple numbers x,y, + that satisfy it). The distance PR is what we want. Since PR and RQ are perpendicular (because PR is perpendicular to the plane), we have from Figure 5.6 PQ cos 8. Gan) PR= From the equation of the plane, we can find a vector N normal to the plane. If we divide N by its magnitude, we have @ unit vector normal to the plane; we denote this unit vector by n. Then | PL (PQ) c0s 8, which is what we need in (5.11) to findH0_LINEAR EQUATIONS: VECTOR: STERMINANTS cn PR. (We have put in absolute val igns because PO cht be ue signs becane Fim might be negative, wheres (PQ) cos 0, with @ acute as in Figure 5.6, is positive. _ " . Frample 3. Find the stance fom the point PCI, —2,3) whe 3x ~ 2y + 1=0. " fared) = ‘One point in the plane is (1,2) al this point 0. Then the vector fom P 10-0 is FG =(1, 2,0) - 0, -2,3)= (0,4, -3) = 4j — ak From the equation of the plane we get the normal vector N=3i-2)+k We get n by dividing N by IN| = 1H. Then we have IPR| = IPQ - nl = 14) — 31) - Gi — 3} + EY TAI = |(-8 — 34 = We can find the distance fom a point Pwo alin ina sine way, In Figure $7 we want the perpendicular distance PA We select any point an the ne [tate any (3.2) sang the equations ofthe Hine} eal ths ; point. Then (see Figure 57) PR = PQ sin. et A he K 4 vector along the line and u a unit vector along the line, i obtained hy diving A by its magnitude Then [PG x ul = |POI sin 6, we pick so we get PR| = |PO x ul FIGURE 57 Example 4. Find the distance from P(1, 2, —1) to the line joining Praca 1) joining P,(0, 0, 0) and Tet A= P\PL= — Tine is (Piss 4 2k isis vecoralogtheline “Then a unit yector along. the VM i + 2k), Let us take to be P,(0, 0,0). Then PO — —i — ‘so we get for the distance | PR) » : ae 3 +k) x (i + ky FIGURE 5.8 HS AND PLANES IIL Ic is also straightforward to find the distance between two skew fines (and if you really want to appreciate vectors, just look up this caleulation in an analytic geometry book thar doesn't use vectors!) Pick two points P and Q, one on each Tine (Figure 5.) ‘Then [PO + n|, where m is a unit vector perpendicular to both lines, isthe distance we seant, Now if A and B are vectors along the two lines, then A x B is perpendicular 10 both, and a is just A x B divided by [A x Bi Example §. Find the distance between the lines r=i—2j+(/— ky and rejj—k +i If we write the first line as F = ro ++ AF, then (the head of} r9 is a simple choice for P, so we have (4, =2,0) and A Similarly, from the second line we Bnd 2=0.2%-) md B $j + band n= (1//3Ki +5 +h). Also. o 4 -D=-i+4j—k. ‘Then Ax B ro Thus we get for the distance between the lines 17d al i+4i-k i+i+h) Example 6, Find the dircetion of the line of interscetion of the planes x—2y +32—4and 2x ty — ‘The desired direction will be parallel to both planes, hence perpendicular to the two normal vectors to the planes, namely i — 2) + 3k and 2i + j ~ k that of the cross product of these normal vectors, which is easily found to be , or using matrix notation corresponding to (6.8) ‘ (6.10) r= Br, where (S05 Sy) 2) sect MATRIX OPERATIONS _119 ‘The net result of these two rotations is a rotation through the angle 0 +, and we would expect 2", y” and -, 9, to satisfy ehe equations fi x 008 (0 + 6) +y sin (0+ $), oat y= x sin (0+ 6) +9 cos + 9). [A direct substitution of (6.3) into (6.9) justifies our expectations. See Problem 8.] ‘Again using matrix notation, we write (6.11) as Cr, (6.12) r (teen) Ese) =(’) —sin (8+) cos (8 + 0), Ny If we substitute (6.8) into (6.10), we get where (6.13) P= BAP) = (BAy. (Whe last step is correct because matrix multiplication is associative.) Comparing (6.12) and (6.13), we see that we might expect the matrix C to be equal to the matrix product BA; let us check this, We write the matrix equation BA = C with the elements in C ‘expanded by the trigonometric addition formulas: 6.14) (af (Gy = sin 6 cos ¢)\—sin 8 cos 8, See Se) sin @ cos cos ¢ sin 8 —sin g sin @ + cos 6 cos 8, Observe thar the elements in C are exactly what they should be according to our “row times column” rule, Thus the way in which we have defined matrix multiplication makes it possible for us to combine the matrix equations (6.8) and (6.10) to get (6.13), treating the letters as if they represented ordinary numbers (except that we must not change the order of the factors). The product matrix BA then correctly represents the rotation matrix for the resultant rotation through an angle (0 + 9). We have discussed this example in detail because it was a simple problem for which we knew the answer (that is, the © matrix) so that we could check our results However, in a more general problem we might not be dealing with a rotation of axes bbut simply with a linear change of variabies, say x sarthy, ext dy, (6.15) where a, b, cd ate given constants. Similarly, x* and y” might be another pair of variables given as linear functions of »’ and y’, and perhaps we might have even another set of variables x”, »” given in terms of s”, »". Then instead of making the algebraic substitutions in the successive sets of equations, we could find the final set of variables in terms of x,» by writing the matrix equations [like (6.8), (6.10), (6.12), (6.13)}, and multiplying the matrices as in (6.14). (See Problems 10 to 13.) The same method can be120_LINEAR EQUATION: tused for any number of variables, say sy 3, 2 in three dimensions; this method is, for ‘example, very useful in working with three-dimensional rotations As another example of the use of matrix multiplication, let us write the following set of equations in matrix notation, eo 6.16) -irty 1 { x-yt2e= Using letters to represent the matrices, Loo : 5 im a= (2 1 of, (}. e-( i}, 1-13 d = we can write (6.16) as (18 Ask Now if (6.18) were an orinary algebraic equation, te would sole i fr t get (619 reat Since 4 is a matrix, we can do this only if we can give a meaning to A”! such that (6.19) gives the solution of (6.16). Let us discuss how to do this, of a Matrix A unit matrix is a square matrix which has every element of the ‘main diagonal (upper left to lower right) equal to 1 and all other elements equal to ‘tro. For ample 100 on ‘) oot is a unit matrix of order three (that is, three rows and three columns). A unit matrix is called 1 or U or J or £ in various books. You should satisfy yourself that in multipli- cation, a unit matrix acts like the number 1, that is, if A is any matrix and U is the unit matrix conformable with A in the order in which we multiply, then UA = A and AU = A (Problem 16) ‘The reciprocal or inverse of a number A is the number 1/A so that the product AWA = 1, We define the inverse of a matrix (if it has one) as the matrix A~" so that AA'=U and A~'4= U, Note that only square matrices can have inverses (otherwise we could not multiply both 44-' and 1A). Actually some square matrices do not have inverses either; this will become clear later. We are going to consider two methods of finding the inverse of a matrix. Recall that we discussed two methods of solving simultaneous equations, one useful for numerical computation (row reduction) and the other a formula (Cramer's rule). Similarly, we will consider a method of finding the inverse of a numerical matrix, and also a formula for the inverse which is very important in theoretical work os _ MATRIX OPERATIONS 121 Let us try to find the inverse of the matrix A in (6.17). If we write ea wi=(i 2 2) then we want to satisfy the matrix equation 447! = U: 10-1) [x my ms) [1 0 0 20 [2 1 ‘| (> Ja Js (: 1 ‘| r-1 Ws 2 =) loon ‘This is a set of nine equations in che nine unknowns in (6.20). [Multiply each row of 4 times each column of A~' in (6.21) and set the result equal to the corresponding ‘element of U] Verify that the x,y, 21 equations are: Satisfy yourself that the second column of A4~' = U gives a set of equations in x3, Yay % With exactly the same leftchand sides as the x), ys, £1 equations, but with constant terms (0, 1, 0) on the right-hand sides. Similarly’ the vy, y3, = equations are the same except that the rightchand sides are (0, 0, 1). Now consider the matrices we ‘want to row reduce to solve these sets of equations: For a1, ¥15 21 For #2592542 For 34.93, 83 Loo iy (to -1 oy fa to 1a a) (2 tot) (2 toe r-ro2 of (ta 2 of Via tt ‘The first three columns are the same in all three matrices. We can, then, combine the three problems by row reducing the matrix 1 0-1 1 0 0 -2 1 0 0 1 0 1-1 2 0 01 ‘Then the first four columns will give us x,y» £15 the first three columns and column. 5 will give us x2, y2, £2, and so on, First we row reduce the above matrix: Pisb tye is In this lat step we went 2 little farther than echelon form and used the pivor 1's to obtain zeros above as well as below the pivots; this final form is called the reducedLINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS cus echelon form. In solving sets of simultancous equations this step was not necessary because ack substitution is easy. Here, however, it is worth while because {-" is noe just the last three columns! Satisfy yourself of this. For example, the first four column give x22, and these are the elements of the fist column of ~!. ‘Thus zi (629 aa(i H ad We can now very easily state the method of finding 4” given a square matrix 4 of forder n, if A has an inverse [(©23) Write 4 and (eat to on the righ) 4 wat paw of Re Game oc / Considering this a8 one large matrix of rows and 2x columns; spoly elementary row operation to brig it to reduced echelon form. Ifthe fae > columns now form a unit matrix, then the last columns form A | (IF the first m columns do nor form 4 unit matrix, then A~! does not exist; we will {earn more about this in Section 8.) Using A and A~* from (6.17) and (6.22), you can ‘etify that their product in both orders is the unit matrix (Problem 17). Let us mow use 4” from (6.22) in (6.19) to solve equations (6.16). We find : pine sy Jeon (E23) (pe | ‘} ee hen Any et of reqs cn be writen inthe mati fom Ar =a in (18, and, if 4~* exists, then the solution is given by r= 47" as in (6.19). The amount of eral ton vale nshng naevus se hr metho. Honenr, hs cam iss neh nee ep is acter mah of sling squats Nese het oe eee fet tet maces mate pa te complcaed ene te ee esti may unknowns ange way eon fe a ak “am he snp nm how see pode eae i Inthe wor ne els ul fr ene Real ha ma sie we cn the fet The aor oe cee ea 4 mists cacy the ne things he oar tt eater he ee GAN]. Namely, the cofactor C., of the element 4;; in row i and column j is « symber equal to (~1)'*/ times the value of the determinant remaining when we crosy off row ¢ smd colunj Then fd 4" td he coeee Coat at ese fet ie rea 85 so the solution is: x = 1, y MATRIX OPERATIONS _123 Sec.6 at (6.24, atagG CT where Cy = cofactor of ay — 4 hat det 4 1 be zero (see Section 8). Although (6.24) is particularly useful Note that det A must not be zero (see be Cramer’ Ue) on ork, you should pacce using it (8 me sai f aol umet : Z the formula means. to learn what simple numerical problems in order to I 2 0 - a[o to bee a? +B? The cofactors of the elements are Example, Find 4”! given | on 10 _|o 0 =o, | =-b, « o| | —b ae “ =a+h, - ae. semdron: =| “tl =nm [ft a. 0-6 @ a | | 4 dee: [0 a aC Qe cejoat-cbheraaty (0 eer 0 PROBLEMS, SECTION 6 — 1. A-college keeps a record in the form of a matrix (that is, 4 table) of the numberof receiving various grades in various courses each year wat we couse sa 2, onge Si Pe va 3 tata) [sa22 ¢ waar) [vase > 32s} lavas i tate Wiad What does the sum of these matrices represent? What does their difference represent? sa—si ve that AB 4 BA. Show In Pobems 2 and 3, fad AB, BA, 4B, 54, Ms SA — 3B. Oar that 48-4 Bt Show that de (48) = des (BA) = (det det Bb that det (A+ 8) # dt A # et Bde 5A Sidet 4, and det 38 + 3 det B. (In Problem 2, show thar det 38 = 9 det By a et 3B = 27 der B.)m 2 MATRIX OPERATIONS _125 LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS cus set (fo eC} 11, The equations (See inaructios for aa 102 110) Problems 2 ana'3 vin ost ie gake-ywita, HovyBto ren av : (2h). eft i), : 3-1 ) roducts of two of these matrices (4B, BA, 42, etc); P, ec). Compute the product of each of the matrices in Problem 4 with orders, that is, AT and ATA, etc) is transpose (in both ‘The Pauli spin matrices in quantum mechanics are “(0 #( Pir of these matrices anticommute that is, AB = —B4, etc, Show that the commutator of A and By that is, AB ~ BA, is 2iC, and similasy for other paits in eyclc order, srenttily 4 complex number 2; = rye! by 2, we get dey = re!“ that is, a complex number with the same r, but with its ange increased by 0, We can say that ig Lota. ram the origin tothe point zy has been rotted by the angle Jy. Ta Figure 1, Senin ines through angle @ which i equivalent to rotating the vector ¢ threugh the angle leaving the axes unchanged. Thus if z= r+ijy we should’ fd Fils (+i). Take real and imaginary parts of this equation to obtain equations (6.3). Nerf that (61D isthe resul of substituting equation (6.3 into (6.9) and using erigonomer- tic addition formulas, Use Cramer’ rule to solve the rotation equations (6.3) for x and y in terms of x’ and y' ‘Show that your results correspond to a rotation through the angle — write each set a5 a matrix equation and solve for x,y" in terms of s, » by multiplying matrices These equations represent rotations of axes in two dimensions. By comparing ‘them with (6.3) find the rotation angles and check your results Given the equations ¥+yV3, —¥eV5 +9), { ee to), How +y, 13, ‘ent rotations of axes in three dimensions. Find the matrix of the resultant rotation and escribe geometrically the net result of the two rotations “The rectangle in the diagram represents some electrical network (we are not concerned with ir le tes at te da ini ony an tte pcre and slg a an a the output curent and volage, In certain casei may be true that f, and Combinins of Ty and yy the network then ald «ner fuera network Spo these equations hod B= ABs + Bh, I= CE, + Dla, where 4, B,C, D are constants. Write these equations in matrix form. The square matrix sin 7. Comidr two In pour eqution cle, n eect cc theory, the taamieion matic 7. idenal networks ineascade a shown, and find Ty in ore of Tyy Ey. Show that the fh hk A {—— P tes i : ” ove ranamion marie 7 Under crn ccumsaners the const 4B, Dan ere erases aae omens D=cote Bab sih 1 col anh i Using these values of the constants and the formulas for hyperbolic functions from Chapter 2, show that cosh 2b sinh 2 n 4 sinh 2e cosh 2a + (See Pipes and Harvill, pp. 208) boar (a) Find the matrix produce @ (7, _T)Ms. 16. 1", In problems 18 to 2 formula (6.24). 2. 2. 2 LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS chs () Show, by multiplying out the matrices, that the following equation represents an ellipse “Wom ‘The following matrix product is used in discussing a thick lens in ar: 4 (' (RY 1 ot — yr, ot Jae iXo 1} where dis the thickness of the lens, mis its index of refraction, and Ry and Ry are the radi of curvature of the lens surfaces. It can be shown that element 4,2 of A is — If whete fis the focal length of the lens. Evaluate A and det A (which should equal 1) and find I/F. [See the American Journal of Physics, 48, 396 (1980),] ‘The following matrix product is used in discussing two thin lenses in air: 6M GY where f, and fa are the focal lengths of the lenses and d isthe distance between them. As in Prolem 1, ment Sek iif whe fe fr length othe combiaonFin det M, and Wf. " " Find a Prove that the unit matrix U has the propery which we associate with the number 1, that is, Ua = A for any matrix A which is conformable with U, and AU = A for any matrix for which the matrices are conformable in this order. For the matrices in (6.17) and (6.22), verify that 44~! = U and A~1A = U, Gnd te iene other mais 6) by rw sedan, () by he (3) » 6) Given mas Loo 4 co eer ee 420 a (@) Find 4~', B-*, B*4B, and B>1A~*B, (©) Show that the last two matrices are inverse that is, that their product is the" unie Problem 22(b) is a speciat case of the general theorem that the inverse of a product of matrices is the product of the inverses in reverse order, that is, (4B)"' = B-'A" 1. Prove this theorem. Hint- This is easy! Seo LINEAR COMBINATIONS, LINEAR FUNCTIONS, LINEAR OPERATORS _127 In Problems 24 to 27, solve exch set of equations by the method of Finding the inverse of the cnelicient natriv. = eee : {aa yais teracs a hoes meee Paes (att (av42y4 28, Prove formula (6.24). Hint Consider the product of the mattices AC. Use Problem 3 29. Use (6.24) 10 find A", given cos eas Osin sind sing woe —oudong ence | 0. in om 0 30, Use the method of solving simultancous equations by finding the inverse of the matrix of coefficients, together with the formula (6.24) forthe inverse of a matrix, to obrain Cramer's rule 7. LINEAR COMBINATIONS, LINEAR FUNCTIONS, LINEAR OPERATORS Given two vectors A and B, the vector 3A — 2B is called a “Linear combination” of A and B. In general, a linear combination of A and B means aA + SB where a and & are scalars, Geometrically, if A and B have the same tail and do not lie along a line, then they determine a plane, You should satisfy yourself that all linear combinations of A and B then lie in the plane. It is also rue that every vector in the plane can be written as a linear combination of A and B; we shall see how to do this in the next section. "The vector r = ix + iy + ke with cil at the origin (which we used in writing equations of lines and planes) is a linear combination of the basis vectors i,j, k ‘A funetion of a vector, say /(), is called linear if | aay Fley + £2) = fey) +/(62), and Slur) = af), where a is a scalar, For example, if A= 2i+3)—k is a given vector, then S(t) = A+ r= Be + 3y — = is a linear function because Fle, #4) = Ao (ry 4g) =Aoey Arey = fle) Sle) and flat) = As ar =aA +r =af(e) On the other hand, f(r) = |r| is not a linear function, because the length of the sum of ‘two vectors is not in general the sum of their lengths. That is, ry tral # deal + Ira] =f) +L (ead fle, try)12m _LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS. ons a8 you can see from Figure 7.1, Also note that although we call y= me +b a linear equation (jt is the equation of a straight line), the function (x) = me +h is not linear (unless # = 0) because igure 7.1 Fey #45) = mley ag) + HS Comey +) ones +B) = flor) + Flay We can also consider vector functions of a vector +. The magnetic field at each point (x, 7, 2) that is, at the head of the vector f, is a vector B = iB, +48, +kB,. The components B., B,, B. may vary from point to point, that is, they are functions of (x, 9, 2) or F. Then’ [rset see con [ Fir, +12) =Fle,) + Fl) and Fler) where a is a scalar. For example, F(r) = fr (where b is a scalar) is a linear vector function of You know from calculus that 7 d 4d d 73) =f) +) == fs) + = eta) an a) ge L100 + ated) = FSG) + Fa) and ad qo 4 bese, where & is a constant, We say that d/dv is a“ linear operator” [compare (7.3) with (7.1) and (7.2)]. An “operator” or “operation” simply means a rule or some kind of instruction telling us what to do with whatever follows it In other words, a linear ‘operator is a linear function, ‘Then is a linear operator if on O(A + B)=O(4) +.0(B) and Of) = KOLA, | where & is a number, and A and B are numbers, functions, vectors, and so on, Many of the errors people make happen because they assume that operators are linear when they are not (see problems) Matrix Operators Consider a set of equations like (6.3), (6.7) oF (6.8), say = r\_(l =r - a5 o (")= ce Gos GGG) For every point (x9) these equations give us a point (x) If we think of eich point of the (x, y) plane being moved to some other point (with possibly some points like the origin not being moved), we can call this process a mapping or transformation of the fen.7 LINEAR COMBINATIONS, LINEAR FUNCTIONS, LINEAR OPERATORS 129 plane into itself. (See Chapter 10.) All the information about this transformation is contained in the matrix 1-1 roy We say that this matrix is an operator which maps the plane into itself. Any matrix (square or not) ean be thought of as an operator on column matrices r. Since Moy tny= Ar + Ar, and Afr) = KAN), the matrix A is a linear operator. Example 1,_ Is square root 4 linear operator? We are asking, is /A+FB the same we JA + JB? The answer is no; aking the square root is not near operation Example 2. Is taking the complex conjugate a linear operation? We want to know whether (4+ B) = 4+ B and (FA) = 4A. The first equation is true; the second equation is true if we restrict & to real numbers. PROBLEMS, SECTION 7 [Are the following linear functions? Prove your conclusions by showing that /(r) satisfies both of ‘equations (7.1) of that it does not saisy atleast one of them. 1. fle) = A 2+ 3, where A is a given vector. 2s) = As (eke) 3. fer. Are the following linear vector functions? Prove your conclusions using (7.2) 4 Fl) sr mies iy the 5. F(r) = Ax, where A is a given vector 6 Fle) =e +A, where A isa given vector Are the following operators linear? 7. Definite integral with respect to x from 0 to 1; the objects being operated on are functions of, 8. Find the logarithm; operate on postive real numbers. 9. Find the square; operate on numbers or on functions 10. Find the reciprocal; operate on numbers or on functions 11, Find the absolute value; operate on complex numbers. 4 e 2 > D stand for 4, D? for £5, D =F, and so on, Are D, D?, D? linear? Operate on 12, Let D stand for , D? for 5, DY = 75 functions of + which can be differentiated as many times as needed,130_LINEAR. EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS as 13. Asin Problem 12, is D? 420 + 1 linear? 14. Find the maximum; operate on functions of 15. Find the transpose; operate on matrices. 16, Find the inverse; operate on square matrices. 17. Find the determinant; operate on square matrices, 8. GENERAL THEORY OF SETS OF LINEAR EQUATIONS Linear Dependence; Rank of a Matrix We say that the three vectors A= i+ j, Ba ith and C= 2+) +k are lincarly dependent because A+B—C=0. The ‘two vectors i and j are linearly independent because there are no numbers 4 and b (not both zero) so that the linear combination ai + bj is zero. In general, a set of vectors ie linearly dependent if some linear combination of them is zero (with not wil the cock, ficients equal to zero). In the simple examples above, it was easy to see by inspection whether the vectors were linearly independent or not. In more complicate need a method of determining linear dependence. Consider the set of vectors 1) (1, 4, —3), 3, 2, D, 2 —1, 3), and (3, —6, 1); ‘we want to know whether they are linearly dependent, smaller linearly independent set. Let us row reduce th sven vectors and if so, we want to find a 1 matrix whose rows are the ji 4 (82) 52 2-1 3-6 9 07 0-9 13 “lo 0 0 0 0 In row reduction, we are forming linear combinations of the rows by elementary row operations (Section 2), All these operations are reversible, so we could, if we liked, reverse our calculations and combine the two vectors (9, 0, 7) and (0, —9, 13) to obtain cach of the four original vectors. (Problem 26). Thus vectors in (8.1). (83) The number of nonzero rows remai eee i, there are only two independent sec. GENERAL THEORY OF SETS OF LINEAR EQUATIONS _131 By 4 denon sina 1 that fr vetrs we sy atthe Fonction fie files fle) linearly dependent if some linear combination of them is identically zero, that foe here are coments fy Bayes es nota sera such that (84) By fla) + ka fala) +0 + be fl) = 0. For example, sin?x and 1 — cos? are linearly dependent since sin?y = (1 = cos*x) = 0. Bu sn and co sre Hcy indepen dace there reno sumbers by a fy sh tee eee is zero for all x (Problem 34). \ h _ nc ‘We shall be particularly interested in knowing that a given set of functions is linearly independent. For this purpose the following theorem is useful (Problems 33 to 37, and Chapter 8, Section 5) TE AG), L402), fe) have derivatives of order n — 1, and if the determinant fil) fl) fle) fie) fx) Fx) Fite) File) Fale) Ee SEM) FEM) FP) “then the functions are linearly independent. (See Problem 33.) ‘The determinant WV | called the Wronstian ofthe functions. #0, me be ‘There is no solution, since 2 # 5; we say that the equations are inconsistent. It will not always be so obvious when a larger set of equations is inconsistent. However, in the132_LINEAR EQUATIONS: VECTORS, MATRICES, AND DETERMINANTS ons process of reducing the matin to echelon frm, we wil id 4 sow which except the last entry. For the equations above, we have. Bich is al zeros eee eo ) Gi )—683 ‘he tow pode he “eqton” O80) 3, Tha # rw which a zeros excep for the lst ety tls us thatthe equations ae incosise Consider the equations “ : (8.2) ‘We see that these are really just one equation. The solution (or solution sct) consists of all points on the line x + y = 2. Let us reduce the matrix to echelon form a (: 1 2) (lie ee 0 0 of ‘The row of zeros here, like the rows of zeros in (8.2), indicates linear dependence. ‘The rank of the reduced matrix in (8.8) is 1; th i 7 5 there is one independent equation in (8.7), Next we want to consider sets of homogeneous equations (zero right-hand side), « o (0728 wm [atone =0 © pray 1 1 0) 1 1 ft) 1 w ( = ae rar dom G 3 yo te . = OG 2-0 bo ‘The equations corresponding to the reduced matrices are r+ y= 0, x+y =0, of yao, yao OP eae Solution: x = y =0, Solution: AU poins on » + y Cenocr ee rare Eeeel eee erect ele eeties se sls ay = 0fae nt er us |ncese a] tal eee noe sora’ (aS sas eee ga ST eer namely 2. This reflects what we could see in (8.9), that we really have just ong San Wlees aan eevee aon ne ‘Let us summarize the facts for homogeneous equations: ae see.8 - GENERAL ‘THEORY OF SES OF LINEAR EQUATIONS 130, {@.10) Homogeneous equations are never inconsistent; they always have the solu- : tion “all unknowns = 0” (often called the “trivial solution”). If the number of independent equations (that is, the rank of the matrix) is the | sume as the number of unknowns, this is the only solution. If the rank of the matrix is less than the number of unknowns, there are infinitely many solutions. ‘A very important special case is a set of homogencous equations in unknowns. By (6.10), these equations have only the trivial solution unless the rank of the matrin is less than coefficients is a zero row. But then the determinant D of the coefficients is zero. ‘Thus ‘we have an important result (see Problems 30 to 37; also see Chapter 10, Section 4): ‘This means that at least one row of the row reduced by » matrix of the FG&Ati) A system of m homogeneous equations in n unknowns has solutions other ethan the trivial solution if and only if the determinant of the coefficients is Now let us return to the general problem of solving m (inhomogeneous) equations in n unknowns. We write these in matrix form as Ar=2, where 1 has m rows (corresponding to m equations) and columns (corresponding to unknowns), r is the column matrix of unknowns, and & is the column matrix of constants [sce (6.16) to (6.18), for example]. Here we must distinguish carefully between the matrix 1 (called the matrix of the coefficients) and the matrix containing also a last column of constants (called the augmented matris). Look at (8.6); the reduced A matrix (first two columns of the augmented matrix shown) has a row of zeros and so has rank 1; the augmented ‘matrix has rank 2, Note that the equations are inconsistent. In (8.8), the rank of and the rank of the augmented matrix are the same (namely R= 1), but the number of unknowns is n = 2 > R = 1; we could give y any value and find «in terms of y. In the examples in Section 2, we had rank 4 = rank of augmented matrix =», the number of tunknowns, and there was a unique solution. The following summary covers all these cases for @ set of m linear equations in » unknowns, {frank A-< ran ofthe augmented matin the equation are inconsistent | and there is no solution, b. IF R= rank rank of the augmented matrix, and R= (number of unknowns) there is one solution this ean be found using Cramer's rule if m= nor if a set of» independent equations is selected co IF Rm then the R unknowns corresponding to the (nonzero) pivos in the row-reduced matrix can be found in terms of the remaining wR unknowns. Example, Solve the equations ( ( (8.13)14_LINEAR EQUATIONS; VECTORS, MATRICES, AND D1 es ‘We reduce the augmented matrix to row echelon form 11-2 7 roi. 7 2-1-4 2) fo -3 0 —2 Seaesae ogee 0 9 0 3 ore lee cues 0-4 0-1 jo. -2 7 “Re + 10 4 BF fo 1 0 4 caamen Ogee leer mans] 11-2 7 (14) —jo too 4 o 0 0 0 lo 0 0 ol q (8.15) By back substitution, we find emket oy ‘The solution set consists of all points on the line which is the intersection of these two planes. An interesting way to write the solution is the vector form (6.16) r Qs +3, 4,2) 84,0) + (2,0, 1s 12 = hiss the parame form of he equation of ssigh ner = J eqatons ofa sigh line += ry + At Note that this Slaton isan example of (B12) wth m = (number of eqns "=3 (numberof unknowns) R = 2 (rank ofboth mates); then we sole sniooes (ad) omen he phot Ts in) te thew Another point of ines i o consider the homagencous equations cosespending 10 (8.13) (that is, consider dr = & with & = 0) “ : 7 oo -2 wan ee siete 5 4 wf) 2 0 -1 -6} ** 3-1 ‘The solutions of (8.17) are (8.15) with zero right-hand sides, that is - (8.18) -AR EQUATIONS 135, see8 Comparing (8.16) and (8.18), we see that the solution of the homogeneous equations (is a straight line through the origin; the solution of the equations Ar = & is a parallel straight line through the point (3, 4,0). We could say that the solution of js the solution (8.18) of the homogencous equations plus the particular solution r= G40). ‘The outline (8.12) answers the question of when a square matrix has an inverse, We saw that solving (6.16) or (6.18) (dr, A square), and finding A~', were equivalent problems. By (8.12), there is a unique solution for m equations in ® unknowns if the rank of 4 ism If A has rank 1, its reduced echelon form is a unit matrix and det AZ 0. If a matrix A is square and det 4#0, then A has an inverse A”. If det A =0 or A is not square, A is called singular. A singular matrix. does | Coe Linear Vector Spaces We have used extensively the vector r= ix + jy + ke to sean a vector from the origin to a point (x, y, 2). There is a one-to-one correspondence between the vectors r and the points (x,y, 2), and the collection of all such points or all such vectors makes up the three-dimensional space often called Ry (R for real) or Vs (V for vector) of Ey (E for Euclidean). Similarly, we can consider a two- dimensional space V’, of vectors r= ix + jy or points (x, y), making up the (x, plane; 1’, might also mean any plane through the origin, And 1’, means all the vectors fom the origin to points on some line through the origin We also use s, y, 2 to mean the variables or unknowns in a problem. Now applied problems often involve more than three variables. By extension of the ideas of Vy, itis convenient to call an ordered set of m numbers a point or a “vector” in the n= dimensional space V,. For example, the four-vectors of special relativity are ordered sets of four numbers; we say that space-time is four-dimensional. (See Problern 12.14 of Chapter 10.) A point of the “phase space” used in classical and quantum mechanics is an ordered set of six numbers, the three components of the position ofa particle and the three components of its momentum; thus the phase space of a particle is the six-dimensional space Vg If we start with several vectors, and find linear combinations of them in the usual ‘way (by components), then we say that the original set of vectors and all their linear combinations form a linear vector space (oF just vector space or linear space). Note that if + is one of our original vectors, then r —r = 0 is one of the linear combinations; thus the zero vector (that is, the origin) must be a point in every vector space. A line or plane not passing through the origin is nat a vector space. Suppose we start with the four vectors in (8.1). We showed in (8.2) that they are all linear combinations of the two vectors (9, 0, 7) and (0, —9 13). Now two vectors (not Parallel) determine a plane; all Tinear combinations of the vectors lie in the plane. [The plane we are talking about is the plane through the three points (9, 0, 7), (0, —9, 13) and the origin.] We call this plane a subspace V, of Vy and we say that either the original four vectors or the two independent ones spun the subspace. A set of linearly independent vectors which spans a vector space is called a basis. Here the vectors (819)VECTORS, MATRICES, AND DETERMINANTS as (, 0, 7) and (0, —9, 13) are one possible choice as a basis for the subspace V’p; another choice would be any two of the original four vectors since in (8.2) no two of the vectors. are dependent, Note carefully that the dimension of the subspace is equal to the number of basis vectors, We can use the vector space language to discuss matrices and sets of simultaneous equations, Consider (8.6) and (8.8). We can think of the rows of these matrices as vectors; in (8.6) the row vectors in the reduced matrix are (1, 1, 2) and (0, 0, 3). ‘These two vectors determine a plane through the origin called the rom space of the matrix. It is a vector space of dimension 2; the row vectors span the space, and since they are independent they also form a basis. A simpler basis is (1, 1, 0) and (0, 0, 1). Note that the dimension of the row space is the rank of the matrix, In (8.8), the row vectors span only a one-dimensional space V,, the straight line through the origin in the direction (1, 1, 2). The rank of the matrix is 1 Consider the homogeneous equations (8.17). Recall that row times column multipli= cation corresponds to finding the dot product of a row vector of 4 and the column vector r; (8.17) says that all these dot products are zero, that is, the corresponding vectors are perpendicular. This means that the solution set r = (x, y, 2) is the vector space which is perpendicular to the row space of , From the reduced matrix (8.14) we can see that the row space of 4 in (8.17) is a plane; the vectors (1, 1, —2) and (0, 1, 0) form a basis for the row space [a simpler basis is (1, 0, —2) and (0, 1, 0)]. The subspace of Vs perpendicular to a plane is a line; the solution (8.18) of (8.17) is a line Perpendicular to”the row space plane (Problem 38). This line [the solution of the homogeneous equations (8.17)] is a vector space V, since it goes through the origin, ‘The parallel line (8.16) which is the solution of the inhomogeneous equations is not a vector space since it does not go through the origin. PROBLEMS, SECTION 8 Solve the following sets of simultaneous equations by reducing the matrix to row echelon form. ety— 222 tater rby 223 e+ dy — fae 4 ay r+ 2y mrt y- 224 4 ra bres d- td =6 (x-24 6 ey (2427-3 Bet yt sey — athe ae + Let aye 10. 2r + 3y—2e= 1 Bray (ats: xa 12 2 las yes ( r= y4tens thy ed (anaes (2e+ y—Se=7 16 4 ry 1 [3r-Sy— = {2 435 18. dr 2y +52 [r- +2 (2+ qt art st en? 20, 4p +2949 4 36 — De + Be [op + 39-49 + 304 54 Be - PBR ii - 4a oH t4 :I os wedaga iit id ee ws wesc amt ste nrc se if oi 99, 0, ); how do you get the right y componer In Problems 27 to 29, find out whether the given vectors are dependent or independent; if they are dependent, find a linearly independent subset. In other words, find the dimension of the subspace spanned by the vectors, and a basis for it. Write each of the given vectors as a linear combination of the basis vectors. My -2, 94H 1D (=2, 1, 4,0, 9) 2B ,1, 0, (=1, 8,31, 0, 2), @ —15, 23,5, 1,0, 4, 9, (1,0, 5), (6, 14, 5)138_LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS cn 30. For what values of 2 does the following set of equations have nontrivial solutions for x, 9, =? For each value of 4 find the corresponding solutions for x, y, 2. (Comment: This is an example of what is called an eigenvalue or characteritic value problem in mathematical Dhysies; the values of 2 are the eigenvalues. See Chapters 10 and 12) [-U4 arty + +02 Br + (2-9) 31. Find a condition for four points in space to lie in a plane. Your answer should be in the form of a determinant which must be equal to zero, Hint: The equation of a plane is oF the form ax + by + cz = d, where a, b, ¢, d are constants. ‘The four points (¥y, Yu, =), Gaon xe all to satisfy this equation, When can you find 4, not al zero? 32, Find a condition for theee lines in a plane to intersect in one point, Hint: See Problem 31 33. Prove that if the Wronshian (8.5) is not zero, the functions fx, fz, s++y fq are linearly independent. (Note: This is equivalent to proving that if the functions are linearly depen- dent, then HY = 0. However, if 17'= 0, the functions may or may not be linearly dependent. see, for example, Wilcox and Curtis, p.119.] Hints for proof: Suppose (84) were true; you want to find the #s. Differentiate (8:4) repeatedly until you have a set of w equations for the unknown &s, Then use (8.11) In Problems 34 to 37, use (8.5) to show that the given functions are linearly independent. 3. sin a, c0s x 35. sins 3. eet 37. sin, cos 4 sin x, x 608 + 38. Verify thatthe line (8.18) is perpendicular to the row space of 4 in (8.17), 39. Show that any vector V in a plane can be written as a linear combination of two non- parallel vectors A and B in the plane; that is, find « and 6 so that V= aA + 6B. Hint Find the cross products A x V and B x V; what are Ax A and B x B? Take components perpendicular to the plane to show that (BxV-n (BxAyn where m is normal to the plane, and a similar formula for 6 40. Use Problem 39 to write V= i+ 5) as a linear combination of A=2i4j and B = 3i—2), Show that the formulas in Problem 39, written as 2 quotient of 2 by 2 determinants, are just the Cramer's rule solution of simultaneous equations for a and &. = 41. As in Problem 40, write V = 4i — 5) in terms of the basis vectors i 4j and Si + 2) 42, In Problems 6, 11, and 18, find a basis for the row space of the matrix 1 of coefficients Show thatthe Solution set isa vector space which is orthogonal tothe row space of A. sect SPECIAL MATRICES 139 9. SPECIAL MATRICES In this section we want to summarize the definitions of various special terms which are used in work with matrices. 'A unit matrix is a square matrix (of whatever order is needed in the problem we are doing) which has every element of the main diagonal equal to 1 and all other elements ‘equal to 0. For example, ( 00 0. ot ‘| oy oot isa unt matrix of order three, A anit matrix is called 1 or U' or Jor fin various books. (See Section 6.) ‘The Kronecker 5 is defined by if i=y, 02) it ity For example, 54, = 1, 5:2 =0, 622 = 1, 53, =0, and so on. In this notation a unit matrix is one whose elements are 5,, and we can write 03) u 54) ‘The Kronecker 3 notation is useful for other purposes. For example, since (for positive integers m and 1) nif m=m 04) f. cos mx cos my di = f an We can write 5) | cos me cos ma dit = 7° Bam This is the same as (9:4) because Syq = Oif m sm, and Sng = 1 if m= A zero or null matrix means one with all its elements equal to zero; it is often abbreviated by 0, but we must be careful about this. (See Problems 1 and 2) A square matrix whose determinant is zero is called singular; all non-square matrices are also called singular. ‘There are several special matrices which are related to a given matrix 4, We outline in (9.6) what these matrices are called, what notations are used for them, and how we get them from A. The notation |4| or det A means the determinant of the matrix 4; whenever it is used, A is understood to be square, The cofactor of an element in @ square matrix means exactly the same thing as the cofactor of that element in the determinant of A.Mo_LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERMINANTS chy 0.6) Name of matrix Notations for it Hor to get it from A anspase of A, oF AT or For A’ or A transpose a Interchange rows and columns in 4, Complex conjugate Aor At Take the complex conjugate of A. of each element ‘Transpose conjugate, Hermitian conju- sate, Hermitian adjoint, of (in quantum mechanics) adjoint, but note that adjoint has another meaning (in algebra)—see below At (read A dagger) Ts ce the complex conjugate of each element and transpose (that is, inter- change rows and columns), Adjoint or adjugate adj 4 or 4 Replace each element by its of 4 (A must be square), ‘cofactor and then transpose. Reciprocal or inverse at Formula (6.24): Divide each of A (A must not ‘element of adj A by det A. be singular) Computation: See Section 6, ‘There is another set of names for special types of matrices. We list these and their definitions for reference. on A matriy is called if it satisfies the equation ee symmetric A=aT (matrix skew-symmetric or A real pure imaginary orthogonal : Hermitian (matrix = its transpose conjugate) unitary AA = U, that is, At = A (inverse = transpose conjugate) (Sce the problems for examples of the use of these terms.) se SPECIAL MATRICES 41 PROBLEMS, SECTION 9 10. 2. Find AB and BA given sh a(S) Observe that 1B isthe aull matrix; if we call tO, then AB =O, but neither 4 nor 2 is 0 sae ens) a AD, bur C # Dand A #0. Given and 4 as in Problem 1, show that AC Given the matrix find the transpose, the adjoint, the inverse, the complex conjugate, and the transpose conjugate of 4. Verify that 4 = AW! w the unie matrix. Repeat Problem 3 given (4) Show thac (48)" = BAT, thai the transpose of a product of two matics is equal 10 the product ofthe transpose in reverse order (b) Use (a) 10 show thatthe matrix A" is symmetric. (See (9.7) Give mumerical examples of: a symmetric matrix; 2 skew-symmetric matrix; areal matrix; 2 pre imaginary matrix Show thar the matrix of a vocation (6.4) is orthogonal; show that a unit matrix is orthogonal. Show that the matrix in Problem 6.29 is orthogonal ‘Show that the determinant of an orthogonal matrix is +1 or 1. Hints What is the determinant of AA when A is orthogonal and how do det A and det A" compare? Show that if @ matrix is orthogonal and its determinant is +1, then each element of the matt is equal to its own cofactor. Hint: Use (6.24) and the definition of an orthogonal Show that a real Hermitian matrix is symmetric. Show that a real unitary matrix is orthogonal Show that the definition of a Hermitian matrix (4 = 4") cam be written a= dy (that is, the diagonal clements are real and the other elements have the propery that 2,2 =) ete). Construct an example of a Hermitian matrix.| I } M2_LINEAR EQUATIONS, VECTORS, MATRICES, AND DETERMINANTS as sec MISCELLANEOUS PROBLEMS 18) 13. Show that the following matrix is unitary matt, 12, What is weong with the folowing argument? “If we add the first row of 2 determinant ro i - the second row and the second rom to the frst row, then the first two rows of the H+ V3) determinant are identical, and the value of the determinant is zero. Therefore all determin- fants have the value zero." 13. (@) Find the equations ofthe ine through the points (4, ~1 2) and (3, 1,4. (©) Find the equation of the plane through the points (0, 0, (1,2, 3) and (2 1). 14. Show that the two: meaning {6) Find the distance from the pont (1,1, 1) £0 the plane 3x~ 2) + 62 = 12 ‘Show that the emo meanings of ain in (26) ae the same for & unitary mats wich {@) Find the distance frm the point (i, 0,2) tothe line r= 24 + j—k+ (2) 2k {e) Find the angle between the plane in (6) and the line in () 15. Show thatthe Pali spin matrices (Problem 6.6 are Hermitian Hho Given the fine ew 3k —j + (2-4 j= 2p 16. Let Gy = (—1'*/My be the cofactor of element 4, inthe determinant (2) Find the equation ofthe plane containing the line and the point (2, 1,0) the equations combined in (6) Find the perpendicular distance between the Tne and the axis ‘4 Find the equation of the plane through the point (2, 1, 0) and perpendicular tothe line. Lay, , {@) Find the equations of the ine of intercon ofthe plane in (8) andthe plane y = 2 G 15,_ a) Write the equations ofa straight line through the points (2, 7, ~1) and (8, 7,3. Taye, on (6) Find the equation of the plane determined by the two lines ¢ = (@~ 2)-+ 1) and eee (i — 3) + (©) Find the angle which the line in (a) makes withthe plane in (). (@) Find the distance from (1, 1, 1) tthe plane in (b) {6) Find the distance fom (1,6, —3) tothe line in (a). 10. MISCELLANEOUS PROBLEMS she Dern the rma kn Problems 1 to 8, sole the genset of equations by reducing the mati to echelon fr, Say Ho + Bot cod scometrially what the solution i (one point ll points om a line ur ons ran ‘he solution sa ine, writeitsvectorequation " eee forthe distance from (rp, Yo, £9) 19 ax + by + cz = : [3+ 2y— [r+ 5) 422 an Roem ye ees 17. Given the following set of matrices, find or mark as meaningless these matrices: 4°, A~', (srry ers AB, A, A'BY, BAT, BA", ABC, ABYC, BAG, A, A’, BYC, B-'C, C-"4, CBE (ety —_ Loar (ae! or A : [e-bt ee a-() Ih a-G 5a) af 18 Given, 102 F a-(: -3 ‘}. find AA, AN A, A 104 19, By multiplying matrices, find , y" in terms of x» given 9. Use Cramer's rue to sole Problem 3 Pe ee 1H. Show that if each element of one ow (or cohamn) of a determinant isthe sum of two terms scare eae grea the determinant can be written ab sum of mot we terms, i % WFtwo determinants; for example 29, Asin Problem 19, nd 2’, ai ie tb ay neti tis] day bia aa au fiatbs an} = fase ai as| + fer biz ap | a rth an] fas a Fes, bs2 asa Use this result to verify Fact 4b of Section 3 "im terms of x,y, 2, and describe geometrically the rotation of axes, given Hryi+ ed Hav -yy2+ fe V2 [Yad ty yn tevi-2/d [e=4eort 2-9ir a 23 INANTS. LINEAR EQUATIONS; VECTORS, MATRICES, AND DETERM ‘There is a one-to-one correspondence between two-dimensional vectors and complex nuns bers. Show that the real and imaginary parts of the product 2,2 (the star denotes comples conjugate) are respectively the scalar product and + the magnitude of the vector product of the vectors corresponding to 2, and 2, ‘The vectors A = ai + bj and B = ci + dj form wo sides of a parallelogram, Show that th area of the parallel am is given by the absolute value ofthe determinant a) al (Also see Chapter 6, Seetion 3.) The plane 2x + Sy + 62 = 6 intersects the coordinate axes at points P,Q, R forming 4 triangle. Find the vectors PQ and PR. Write a vector formula for the area of the triangle PQR, and find the area,PARTIAL DIFFERENTIATION 1, INTRODUCTION AND NOTATION Ify = f(a), then dy/de can be thought of either as the slope of the curve y = /(x) or as the rate of change of y with respect to x. Rates occur frequently in physies; time rates such as velocity, acceleration, and rate of cooling of a hot body are obvious examples. ‘There are also other rates: rate of change of volume of a gas with applied pressure, rate of decrease of the fuel in your automobile tank with distance traveled, and soon Equations involving rates (differential equations) often need to be solved in applied problems. Derivatives are also used in finding maximum and minimum points of a curve and in finding the power series of a function. All these applications, and more, occur also when we consider a function of several variables. Let 2 be a function of two variables x and y; we write = = f(x,y). Just as we think of y= f(x) as a curve in two dimensions, so it is useful to interpret 2 = f(x,y) weometrically. If x, y, © are rectangular coordinates, then for each x, y the equation fives us a value of =, and so determines a point (x,y, 2) in three dimensions, All the Points satisfying the equation ordinarily form a surface in three-dimensional space (see Figure 1.1), (It might happen that an equation would not be satisfied by any real Points, for example x? + y? + 2? = —1, but we shall be interested in equations whose Braphs are real surfaces.) Now suppose + is constant; think of a plane x= const. inersecting the surface (see Figure 1.1). The points satisfying 2 =/(x,y) and x =onst. then lie on a curve (the curve of intersection of the surface and the = const, plane; this is AB in Figure 1.1). We might want the slope, maximum andcus Mee dey fr the slope Hwee oe i awaly& fncton of wo arabe x and y with one of them (x) temporarily a A ° with respect 00) Sina, we can old constant and Bnd Gale the pana erative of 2 with e- , spect 1o 1. TE these pata derivatives are dierent further, We write partial derivative of ee Ceey” te dy Re dx ey Other notations are often useful If < = f(s, 9), we may use sy oF f, oF fy for ef Fs and corresponding notations for the higher derivatives Example, Given "7 then ty — ye, ye — aye We can also consider functions of more variables than two, although in this case itis ‘ot so easy to give a geometrical interpretation. For example, the temperature. T of the ai i a room might depend on the point (x,y, 2) at which we measured it and on the ime 1; we would write T= T(x, y, z, 0). We could then find, say, 27/2), meaning the rate at which T is changing with y for fixed x and = at one instant of time 1 ie oss which is frequently used in applications (particularly thermodynamics) is =/¢4),. meaning Csié when = is expressed as a function of x and». (Note two Aiflerent uses of the subscript y; in the example above, f, meant df/2y. A subscript on 4 partial derivative, however, does not mean another derivative, but just indicates the variable being held constant in the indicated partial differentiation.) For umple, sel arr opecrioy ann Notation i 4. ‘Then using polar coordinates * and 1, (recall that x =r eos say 7 ), We ean write = in several other ways. For each new expres- sin 0,8 +9? y=rsin 0 ton let us Find 02:7 cos! O—? sin? ( cos? 0 — sin? 0), fe) (@) = -% ) ( ‘These three expressions for G=/¢r have different values and are derivatives of three different functions, so we distinguish them as indicated by writing the second indepen- dent variable as a subscript. Note that we do nor write 2(x, y) oF 2(r, Ws = is ume variable, but it is equal to several different functions, Pure mathematics books usually Firs O) = tr, x) = Mrs) tes then avoid the subscript notation by writing, say (@zjér)p can be written as just €//¢r, and similarly | ete.) would be inconvenient and However, this multiplicity of noration (2 confusing in applications where the letters have physical meanings. For example, in thermodynamics, we might need (2 ) ep) a well as many other similar partial derivatives, Now T means temperature (and the fother letters similarly have physical meanings which must be recognized). If we wrote T = Alp, 2) = Ble, #) = Clp, w) = DX, p) and similar formulas for the eight commonly ‘used quantities in thermodynamics, each as functions of pairs from the other seven, we would not only have an unwieldy system, but the physical meaning of equations would be lost until we translated them back to standard letters. Thus the subscript notation is essential, ‘The symbol (22/0r), is usually read “the partial of = with respect to r, with x held constant.” However, the important point to understand is that the notation means that = has been written as a function of the variables + and x only, and then differentiated with respect tor. A litle experimenting with various fanetions f(s, y) will probably convince you that (81 xMafieyy = (CleyMef en); this i usually (but nox always) true in applied problems. Te ean be proved that if 72//ex7y and @//eyCx ave both continuous, then they te equal. (See Bartle, p- 241)(M8_PARTIAL DIFFERENTIATION chs PROBLEMS, SECTION 1 1. Wen = x7(s* + y*), find u/ar, Ou/dy. 2 its en dou 3. Mem tn JP EFT, find 20/04, 25/00, dal. 4 For warty? 20746, find Dwi fay? ome find Oays? and OPw/2y* atthe poms where 5 Bet ty — Dey, fin and Sey? ett deity? atthe pins where ‘ For u = * 00s y, (a) verify that 224/00) = Ou/2yde (b) verity that G4/ds? + awidy! 0s 0,9 =r sin 0, find the following partial derivatives. 7 to 2, Repeat Problems 7 to 24 if = 2, POWER SERIES IN TWO VARIABLES Just as in the one-variable case discussed in Cha point) for a function of two variables is uni of finding it (see Chapter 1 for methods) ipter I, the power series (about a given ique, and we may use any convenient method Example 1. Expand /(s, 9) = sin x c write and multiply the series for sin x and sy in a two-variable Maclaurin series. We cos y. This gives : ry Sy sin x cosy =(x— 4 a 5-3 (Se Y0-Reo)or fae Example 2. Find the two-variable Maclaurin ser in equation (13.4) of Chapter 1 by x — y to get In(l+x~5) ies for In (1 + x — 5), We replace x == (eZ + PB + yo Rt yy 2 +88 Py ty? 84 ‘The methods of Chapter 1, used as we have just shown, obtaining power series for many simple functions f(x, »). However, it is also con venient, for theoretical purpose, to have formulas for ike cosh " series or the Maclaurin series for f(x, thee a2 provide an easy way of lor 25 see, for example, Problem 8.2. Following a POWER SERIES IN TWO VARIABLES 149 seo.2 ss similar to that sed in Chapter 1, Section 12, we can find the coefficients of reer series fora function of two varabies (9) (assuming that i can be expanded in Pomerer series), To ind the series expansion of (3,7) about the point (a, #) we write Feo) aa sete of powers of (=a) and (7B) nd then diferente this equation repeatedly as follows a FG, 9) = doo F 410% z 4 doghy — BY? + agole = a)? + angle — a" ~ 4) 4) + aga — Bt aol — 0)? Hayle = ay =) tayo — ally — BF + dos — 8) 4 fo= Ayo + Magolt — a) Farid = 8) 40s fy = oy + ayy — 9) + Moaly — 8 + Sr Jog = yy + terms containing (x ~ a) andor (y ~ 6 2ayg + terms containing (x ~ a) and/or (y seve arte any» fo dervaies show the ie Yu sald be Oe a ee oe ey Brae 71] Now puting += 57 = in we gs Fla, 6) = ao0s fa b= ay0> Sits = dors © Seslas 6) = Baroy faglty BY = aryy te with Remember that f,(a, #) means that we are to find the partial derivative of / wit Cre eee a pnd simi te ter death] Sub tuting the values for the coefficients into (2.1), we find (23) fle, 9) = F(a, B+ fila, bee = a) + fla, By = 4) 1 A Efe, BXs ~ a? + Bos Bs — ak = + fle BK 99 Phen the and y— ‘This can be written in a simpler form if we put second-order terms (for example) become os) Fi Uae, OH + Pas BE + fle, DEI We can te isin he frm 1 é ey 2s E(s2e2)a bbe squared and then aterm Tike ‘an be shown (Problem 7) that the if we understand that the parenthesis is. to Hales vHC/dy) fla, B) is to mean Hf ,{a,B). It 6 third-order terms can be written in this notation as 1,24 22) pla, =F, Whales) + Hoole DT 26) Ldeeg)reo FPLa7 150_PARTIAI, DIFFERENTIATION ‘TOTAL DIFFERENTIALS 151 _ ous and so on for terms ofan n of any order. Thus we can write the series (23 in the form en Ke 2402); | (9) f+02)re fe S$) ram. | | ‘The numbers appearing in the nth order terms a the expat ve ire the familiar binomial coef the expansion of (p + 4)"] divided by (1!) familiar binomial coefficients [in PROBLEMS, SECTION 2 Find the two-variable Maclaurin series for the following functions 1. cos x sinh y 2 costs ty) a matey te 8. Tam eer 7. Verify the coefficients of the third-order terms for Fy) by [2.6) or n= 3 in (2.3) of the power series finding the third-order partial derivatives in (2.1) and substituting + = 8. Find the ewo-varil th ble Maclaurin series for e* cos y and e* sin y by finding the series for = 27 and taking real an king real and imaginary parts. (See Chapter 2) 3. TOTAL DIFFERENTIALS The graph igure 3.1) of the equation y = f(x) is a curve in the (x, y) plane and Bay wee ay ae! «) is the slope of the tangent to the curve at the definition : 7 ‘he come (32) Point (x, y). In calculus, we use Ax to responding change in y (see Figure 3.1), By FIGURE 3.1 Wecshall now define the differential de of the independent variable as 63) dy = Ax, 1m Figure 3.1 and equation (3.1), we can see ydx is the change in y along the tan- (or linear approximation) to Ay. However, dy is not the same as Ay. Frot that Ay isthe change in y along the curve, but dy fom line, We say that dy isthe tangent approximation If y = (U0 represents the distance a particle has gone a8 a function of #, tar id is the velocity. The actual distance the particle has gone between time ¢ and se + dt is Ay. The tangent approximation dy = (dy/d*dt is the distance it would fave gone if it had continued with the same velocity dy/dt which it had at time "You can see from the graph (Figure 3.1) that dy is a good approximation to Ay if dx js small, We can say this more exactly using (3.2). Saying thar dy/dx is the limit of diye as Ax means that the difference Ay/Ax ~ dy/dt +0 as Ax+0. Let us call this diference €; then we can say Example. Ga) where e +0 as Ar+0, or since dx = Ax 65) Ay ‘The differential dy = y'dx is called the principal part ds, you can see from (3.5) that dy is then # good approximation to Ay. ‘in our example, suppose y = 17, 1 = 1, dt = 0.1. Then Ay = (Lb? = P = 021, where e+ 0 as Ar—+0. a +6) dx, 1 of Ay; since € is small for small D gy =2-1-@1)=02, ar Ay =)" +0 dr = 2+ 0.0.1) = dy be dt = 0.2 + 001 ‘Thus dy is a good approximation to Ay. is for a function of two variables, = = /(% 9) ts a surface and that the derivatives (f/x, ‘Je, at a point, ate the slopes of the two tangent lines to the surface in the x andy The symbols Ax = dr and Ay = dy represent changes in the independent variables x and y. ‘The quantity Sz means the corresponding change in = along the surface, We define d= by the equation oe , G6) den at 5 We want to do something similar to thi We have said that this equation represent directions at that point.182 PARTIAL DIFFEREN TH = cus Enza FIGURE 3.2 ‘The differential ds is called the total differential of = ‘meaning of ds. Recall (Figure 3.1) that for y = f(x), tangent Tine; here we shall see that dz Figure 3.2, PORS is a surtace, PABC PDEF is « horizontal plane through P. P), PS is the curve of intersection of this Let us consider the geometries ‘dy was the change in y along th is the change in 2 along the tangent plane. [+ is the plane tangent to the surfice at P, an: Thus PSCF is the plane y = const. (throug!, plane with the surface, and PC is the tangent line to this curve and so has the slope df/2r; then (just as in Figure 3.1), it PF wan weentts CFE (eles) ds, Similarly, PQAD is a plane = const, intersecting the Fyre Gace ees Cr Py een yearn DA = (fs) dy. From the igure, GE = CF, and BG = 4D, so ects d1aL ae ‘Thus, as we said, dz is the change in and y by dy. In the figure, ER = 2, the change in 2 along the surface. From the geometry, we ean reasonably expect de to be a goad approximation to As i 4k and dy are small. However, we should like to say this more accurately in on ‘equation Seerespondting to (3.5). We can do this if f/x and dfiéy are continuous functions, Br Uefinition a7) along the tangent plane when x changes by di L004 Av, + AY) Sly y), By adding and subtracting a term, we get BB) Asm fle + bey) — flay) + fle + Any» + By) — fle + Ae, 9, Recall from calculus that the me differentiable function f(x), 69) 'an value theorem (law of the metn) says that for a S00 + AX) ~f0) = (ANS), is between x and x + Ax. Geometrically this says (Figure 3.3) that there is a tangent ling somewhere between and x + Av which has the same slope as the line TOTM, DIFFERENTIALS 153 sos ae — FIGURE 3.3, e a like (3.9) with y as the variable; y, will mean a value of y between ya Pen a, Met ava gy G10 approach zoro, Let us call these quantities €, and €,. ‘Then we can Lon)ara(Ses} a ls te, Av te Ay Gap a=(4 ¥ fe, and €: Was Avand Ay +4) fr 2 and (1 we nt at 3, Eaton (1 Te 9 t= ced el gah weap rm he ge ide woh o/ly ae comtinans) ds 4 good appronnaton to Bs for sm Thee sae he pn Pat functions af any number of variables. Iu then by def re comtinuous and and du is a good approximation to Aw if the partial derivatives of fa ds, dy, de, ete, are smal. PROBLEM, SECTION 3 = 1 Gamer oct 3) ham ern rar prin 2.3 rae aie Set iene changes from a tou + Av Becomes fu arb Aye hange Aen Ja) whom cinges andy changes from bro b+ is then A Wat Ont A = fla, Bd are and that they Use the series (27) to obtain (3.11) and to see explicitly what €, and , are and approach zero as Av and Ay +0)ls 4. APPROXIMATE CALCULATIONS USING DIFFERENTIALS T-et us consider some examples of the use of differentials in approximate calculations. Example 1. Find approximate the value of 1 (1p TE (x) = 1/x?, the desired difference is af imately df= d(1/x") with 10'S, de when n= 109 (10 +1) ~ f10"%). But AFis approx Ts -2x 10°48, “108 Note that we can't do this problem on a calculator (unless it carries atleast 16 figures) since we are subtracting two numbers which are nearly equal to each other Example 2. Show that, for large » and small a, ee Grae (means “approximately equal 10"). This is essentially the same problem as Example 1. We find d(1/x?) with x =n, de = a; this gives Sa (This result is of use in quantum mechanics; see, for example, French, Modern Physics, pp. 113-114) Example 3. The reduced mass wt of a system of two wt sme! mit. If my is increased by 1% Unchanged? Taking differentials of the equatio masses m, and ms is defined by what fractional change in mz leaves j m and substituting dm, = 0.01m,, we find O= — my? deny ~ m3? dims, dmz dm 0.01m, ds a at or 2 = 001m im. Hor example, if'm = ma, mz should be decreased by 1%: if my = 3myy my should be decreased by 3% ; and so on, mm Example 4. The resistance R of a wire is proportional to its length and inversely Broportional to the square of its rads, that is, R= #1. If the relative ctor in lenetn measurement is 5% and the relative error in radius measurement is 10%. fale relative error in & in the worst possible ease ‘The felatve eror in / means the actual error in measuring / divided by the length measured. Since we might measre (either too large oF too small the rete ere ot set IMATE 3 a ight be either +0.05 of —0.05 inthe worst cases. Silay dh] mi O10. We mane the largest vale which [4R/AT could haves we can find dR/R by Siterentating In R. From R = tr? we find In R= Ink +In f= 2 In. ‘Then ie dla R71 or tn the worst case (that i, rgest value of | dR/R), ll! and dir might have opposite ‘Poa the two terms would add. Then we woul have an) _ at) , || argest | +2 Largest | 05 + 2(0.10) 7 Example 5. Estimate the change in sls) [ota when x changes from 1/2 to 101n/2, Recall from calculus that dfdx = (sin x)/x. ‘Then we want df = (df/ds) dx with x = x/2 and dx = 0.01n/2. Thus a= 222 orni2 = 001 #2 PROBLEMS, SECTION 4 1. Use differentials to show that, for very large m, L oy 2 Use differentials to show that, for large w and small 4, Var Find the approximate value of 10 3. The thin lens formula is where J is the focal length of the Tens and ¢ and # are the distances from the lens to the object and image, If '= 15 when o = 10, use differentials to find i when o = 10.1 4. Do Problem 3 if) = 12 when o = 18, to find if » = 175. 5. Let R be the resistance of A, = 25 ohms and Ry = 15 ohms in parallel. (See Chapter 2, Problem 16.6.) If R, is changed to 25.1 ohms, find Ry s0 that R is not changed. 6 The acleraton of greity canbe found from the length and period of «pends the formula is ¢ = 4n%/7, Find the relative error in g in the worse case ifthe relative error in is 5%, and the relative errr in T is 2%.a 7 Fauomb’s lw for the force berween owo charges gy and gz at distance ix F = bgypai® Tans the tcurve err in qs in the worst ease ifthe relative error ing, 8 Bos iG and in F, 2% 8 About how much (in percent) does an error of 19% in a and b affect a2"? 9. Show that the approximate relative error (4/)/f of a product = gh is the sum of the Approximate relative errors of the factors A force of $00 nt is measured with a Possible eror of 1 nt. Its component in a direction 60) {uy from its line of ation is required, where the angle is subject to an error of 0$" Whey '8 (approximately) the largest possible error in the component? Assuming that your calculator isn’t handy, show how 1 ve f make a quick estimate (to 10 decimal plas) of M498)" — GOVE, Hint: Colder uy) 2 gi 12. Asin Problem 11, estimate ¢/ BOO? TIE 13. Without using calculator, estimate the change in length of a space diagonal of a box whose dimensions are changed from 200 x 200 x 100 to 201 x 202 x 99, 14. Estimate the change in i'r changes from 0.7 t0 0.71 1S. Foran ideal ga of N’ molecules, the numberof molecules with speeds as de> 0 IMPLICIT DIFFERENTIATION Given x +e = 4, find de/di and d?x/de?. Pec aus Ste coespniy and i n e sph wh at {x/dt, In other words, x 1s a function of t even though we “Ee he eqn einer of deena Fins To di, Saute That ura funeion of rand as diferente each term of the equation wi ‘pect Chis is eld implicit diferentaon). We get fy ade oo Hy ott a fe aie is here, and write first dr +e dx = dt; dividing Alternatively, we could use differentials here, and ican sno Bnd higher derivatives by impli difeematon (but do mtu i Ist ti snc we have not ven an} meaning tothe derive or dierent of itera, Let us eerenate each term af (61) mith fspet 0 Ne EE @s dx | (ds)? (6.2) Bite gt ( a Solving for d?x/dr® and substituting the value already found for ds/dt, we get a) (63) +e sey ‘This problem is even easier if we want only the numerical values of the derivatives at 4 point. For x =0 and ¢ = I, (6.1) gives fey Hy og Be ata ai 160_PARTIAL DIFFERENTIATION c and (6.2) gives Implicit differentiation is the best method to use in finding slopes of curves with ‘complicated equations. Example 2. Find the equation of the tangent line to x? — 3y3 +4 re iy + ay +2150 at the ‘We differentiate the given equation implicitly with respect to + to get 4 292% sey Be Ba, Substitute x = 1, y= 2 oy 3-352 49 dete * ‘Then the equation of the tangent Tine is PROBLEMS, SECTION 6 1. Ir pet = © (where a and C are constants), find deidp and dsp? 2. Wye!” = sin x find dyjde and d?y/ds? at (0, 0) 3. fs = 9%, find dyide at 2, 4). 4. Wae? = ye find dy/ds and dy/ds? for y #1 5 If 2y? — yx* = 6 is the equation of a curve, find the slope and the € quation of 1 line at the point (1, 2). ™ : latina 6, In Problem 5 find d2y/ds? at (1, 2. 7. My" ~ ey = 8 the equation of uve, fd the pe and th line at the point (3, ~ 1). _ " ‘we anee 8. In Problem 7 find d2y/ds? at (3, —1), 9. Bor the curve 29 4 92 (8, 0, and at (0, 8), find the equations ofthe tangent fns at A/F, —2/3), at 10, For the curve se” + ye 0, find the equation of the tangent line a the origin Cation. Substiture x = y a eon 45 Soon 28 you have differentiated. M1. In Problem 10, find y” atthe origin, MORE CHAIN RU seat ee 7. MORE CHAIN RULE ‘P(x, 9), where x and y are functions of 1, Now suppose Above we have considered ‘( ») as before, but x and y are each functions of two variables s and 1. Thei a function of ¢ and 1 and we want to find 62/05 and dz/dt. We show by some examples how to do problems like this. Example 1. Find 0:/2s and @:/ér given in (6+, We take differentials of each of the three equations to get ols +0) (ds dt, dy dvtsdy dx oe Substituting dx and dy into ds, we get a) dz = y cosls +1) (ds + di) + x(ds — di) = Ly cos (6 +1) +] ds + Ly 008 (5 + 1) = 3) |, 2 is a function of one variable ¢, and we can divide (7.1) fe ~ dt on the left we write ¢=/¢1 Now if s is constant, ds by de [see (5.1) and the discussion following it). For because that is the notation which properly describes what we are finding, namely, the rate of change of with s when s is constant. Thus we have y cos (5+ 1) =x ‘and similarly yeas (s+) +2, Note that in (7.1) the coefficient of ds is 0=/@s and the coefficient of dé is d=/ét [also ‘compare (5.3)]. If you realize this, you can simply read off O:/@s and Gz/ét from (7.1), We can do problems with more variables in the same way Example 2. Find du/és, du/ét, given yos-Penk We find iu = 2x de + 2a dy + By dr 4 (2x + 2y)ds + 2r dt) + (2x — In zYds — 2 dt) (dt) 2 e+ By —In 2) det (0+ Bene Then Hosea ying Haar srine 2162_PARTIAL DIFFERENTIATION chs IF we want jst one deat, wy du ‘ Fc. sy hl we can sve some work by ein d= 0 wo start with, To make it clear that we have done this, we write - aoa tu, (Qe + WQr de) + Qe — Ing - (o +2Ins 2) ra = dt) ~ he subscript s indicates that s is being held constant, Then dividing by di, we h: P 1 viding by dr, we have Cwict as before. We could also use derivatives instead of differentials. By an equation like (5.1), we have _ (7.2) where weave item all the 1 both s and ¢. Using (7.2), we get . ‘een 2a= y+ eins f= e+ 29928 + Qe = In g -1 +( 1c is sometimes asf to write chain rule formu fi : rule formulas in matric form (ee Chapter 3, Seaton 6) Given, as above, w= /( 3 2) le Dy Ms Dy me can ee (7.2) in the following matrix form: eae ‘avations The which is reminiscent of a but be careful of this for two reasons: (a) It may be helpful in remembering. the formula but to use it you must understand that it means the matrix product (7.3). (b) The symbol O(u, 2)/¢(x,y) usually means a determinant rather than a matrin of partial derivatives—see Chapter 5, Section 4] Again in these problems, you may say, why. not just substitute? ain you may say, why. not just substitute? Look at this Seo 7 MORE CHAIN RULE_163 Example 3. Find dz/dt given 2 =y and Payee, wsin gaye’ From the = equation, we have dz = dx —dy We need ds and dy; here we cannot solve for x and y in terms of f. But we can find de and dy in terms of di from the other two equations and this is all we need. Take differentials of both equations to get 2 dx + By dy = 2 dt, sin de +x 0s 1 dt = (ye +) dy, Rearrange terms: sdetydy=eds, sin t dx — (7+ Det dy = =x cos ¢ de Solve for dx and dy (in terms of di) by determinants: td y pew leeetd =O 4 De] <1 + We + ay cost Ts y | as DO —ysin sint -Q + Ne} and similarly for dy. Substituting dx and dy into the formula for dz and dividing by dt, we get da/dt We can also do problems like this when + and y are given implicitly as functions of ‘wo variables + and 1. Example 4. Find dz/és and dejér given s=Pto, Sa yaws, Poyaeer We have dz = 2x dx + x dy + y dx. To find dx and dy from the other two equations, we take differentials of each equation Dede + By? 3x? de — 2y dy adits, 2s ds + Dede oy164_PARTIAL DIFFERENTIATION cus We ean solve these two equations for dx and dy in terms of ds and dé to get sdetids 3% te = [Bde 2a dy = By] (29s — by?) dt + (—29t — 692) ds a tay = HF x? —2y| and 4 similar expression for dy. We substitute these values of dr and dy into dz and find ds in terms of ds and di just as in Example 1; we can then write 02/61 and ds'cr just as we did there (Problem 11). Notice that if we want only one derivative, say @= cr we could save some algebra by putting ds = 0 in (7.4). Also note that we can save some algebra if we want the derivatives only at one point, Suppose we were asked for @s, cs and dz/¢r at x= 3,y = 1, s= 1, 1=5. We substitute these values into (7.4) to get 6dr 43dy Wde—2dy= dt +5 ds, 10 dr +2 ds, We solve these equations for dx and dy and substitute into ds just as before, but the algebra is easier with the numerical coefficients (Problem 1). So far, we have been assuming that the independent variables were “natural” pairs like x and y, or s and ¢. For example, we wrote dx/2s above, taking it for granted that the variable held constant was 1. In some applications (particularly thermodynamics), it is not at all clear what the other independent variable is and we have to be more explicit, We write (Cx/09); this means that s and ¢ are the two independent variables, that x is thought of as a function of them, and then x is differentiated partially. with Fespect (0 5, Suppose we try to find from the three equations of Example 4 a rather Peculiar looking derivative. Example 5. Given the equations of Example 4, find (0s/02), First, let us see that the question makes sense. ‘There are five variables in the three equations. If we give values to two of them, we can solve for the other thrce; that is, there are 10 independent variables, and the other three are functions of these two. If = and are the independent ones, then s, f, and y are functions of = and x; we should be able to find their partial derivatives, for example (5/02), which we wanted. To carry ‘out the necessary work, we first rearrange equations (7.4) and the dz equation to get Qe +y) dr de ds, 2 ds 2s de+ De dt + 2y dy From these three equations we could solve for ds, dt, and dy in terms of dx and ds (by Aeterminants or by elimination—the same methods you use to solve any set of linear MORE CHAIN RULE 165 Set ivative of (8, 2) es, 2), oF p(s, 2) with ations), Then we could find any partial deri 7 ) Sapect o'r or = For example, 0 find (s/s), we tom the ist equation Bete ds, that ne would not ned to dféxentte al three equations f we wanted only sis Na foc, youtould sways Wok ahead to so how much diftrentiaon fe neces seis ibn, ne mun wit the thee equtos fo din rns of de aad deve sa deel ome wor [fee man ol (olde). by puting dx 0 c San ith. Te make it clear that we have done this we write ds, and dz,. Then we get -dz, 0 -*| 0 + ~3"| om | = (sy + 65") d= = a eh as QF? = 2A) ros -¥F | my _ 9 +357 (2). Example 6, Let x, y be rectangular coordinates and r, @ be polar coordinates in a plane, Then the equations relating them are + c08 8, a5 y=rsin 6, or re JP ay, Ce Osan t% Sonpse we watt 2: Remembering that i y=), de and deft a reciprocals, you might be tempeed to find Cor by tating the recipes of 28 is eu oF en 2s can! (y/x) we get hich is easier to find than 20/0, This is wrong. From @ 0 yt oy a a Tro) A From x = 1 cos 8 we gst 78) Be rsin d= —y166_PARTIAL DIFFERENTIATION. cus ‘These are not reciprocals. You should think carefully about the reason for this; 20/2 ‘means (80/0x),, whereas éx/20 means (@x/60),. In one case y is held constant and in the other case r is held constant; this is why the two derivatives are not reciprocals, I is true that (20/¢x), and (2x/0), are reciprocals. But to find (2x/C), directly, we have to express x as a function of and y. We find x — y cot 0, so we get , 2) get acd ne 9) (5), W498! = FF Sh which is the reciprocal of 80/@x in (7.7). ‘his is a general rule: 4/00 and do/Ou ae wo wuly cereals; they are rei | ci th ter independent ables (sds wo) these in bth cnn | You can sce this clearly from the equations involving differentials. From a= wetan (3/0) we en Bd (: +4) x8 can) dy = cos 0 dr ~r sin 9 d0 =" dr — y a, dy yd (7.10) do From x = cos 0, we get From (7.10), ify is constant, dy = 0, and we can write 7.12) 40, y= Fads where the y subscript indicates that y is constant, From (7.12), we then find either (5), and these are reciprocals. From (7.11), however, we can find (2/26), or (28/0), these are again reciprocals of each other, but are different from the derivatives found from (7.12), Its interesting to write equations like (7.11) in matrix notation: dx, . ‘dx\ [er @0\/dr\ (cos —r sin @\/dr dr cae (3) ey x () = ¢ or c0s (i) 2 ih) er 36 where 4 stands for the square matrix in (7.13). Similarly, we can write , aus («)- ar ar) aa)-G) % MORE. CHAIN RULE _167 se so wwe have writen the square matrix as 47 since by (7-13), #(G)-a)= (a) “Then, finding A”! (Problem 9) and using (7.14), we have a sy 28 ‘ino bcos We can simply read off the four partial derivatives of 7, 0, with respect to x,y, from (7.18). (Also see Problem 9.) Also using (7.5), and specifically noting that x and y are Independent vaibles, we have (%),=s0e (2) a doy > Mand fou Jy (uA) isa masimnum paint Pat G0 Bh fgg (Soke that this includes ox fo, <0, that by, and fy of opposite si.) Hint: St Jog = Ae fog ™ Bey = Ci th the second derivative terms in the Taslor series aye AH + BME + CEP, thin can be writen Uh + BELA)? = (C— BNE Find out wher this expression #8 positive Fora small & [that iy all (xs) near (a, 4} aso tnd out when i is negative forall small A, &, and when if has both positive and negative values foe Seal Use the fats stated in Problem 2 to find the maximum and minimum points of the funetiony in Problems 3 t0 6 Bo tty de aye tO 4 8 dtrty-vaw-dy 6 3. Given = = (y= 12)(y = 2s2), show that = has neither maximum nor x minima st Ut, although = has a minimum on exery straight line through (0,1 BA roof gutter is to be made from a long strip of sheet metal, 24 em wide, by bending up equal amounts at cach side through equal angles, Find the angle and the dinensi will make the carving capacity of the guter as large as possible Ns 9. An aquarium with rectangular sides and bottom (and no tap) is tw hold 5 gal. Pind its proportions so that it will use the least amount of material 10, Repeat Problem 9 ifthe bottom is to be three times as thick asthe sides. 11. Pind the most economical proportions fora tent as inthe figure, with no floor 12, Find the shortest distance from the origin tothe surfice = = ay +5, 13, Given particles of masses m, 2m, and 3m at the points (0,1), (1, 0), and (2.3) find the point P about which their total moment of inertia will be least. (Recall that co find the ‘moment of inertia of m about P, you multiply m by the square ofits distance from P.) 14, Repeat Problem 13 for masses my. may my at (ys Jah 3sa%h (ssa) Show thar the Point you find is the center of mass, 1S. Find the point on the line through (1,0, 0) and (0, 1,0) that is closest to the fine =.=. Also find the point on the line x = == that is closest to the fine throwgh (1,0, 0) and (0, 1,0). w+ b such that the sum of the 16. Curve fitting by lease squares: Find the straight tine y squares of the deviations at x= —1, 0, 1 fram the observed points (1, 2) (th (1, 9)* 172_PARTIAL DIFFERENTIATION, 64 minimum, that is,(—2 — y,)? + (0 —y)? 4 3 — yy}? in an (3) (99) are on the line cha, 17. Repeat Problem 16 0 find the best straight line fi xo the points (1, 0, (2 —1), @, 8) 9. MAXIMUM AND MINIMUM PROBLEMS WITH CONSTRAINTS; LAGRANGE MULTIPLIERS Example 1. A wire is bent to fit the curve y = stretched from the origin to a pe length of the string. int (x, 2) on the curve, Find (x, ») to minimis =? (Figure 9.1), A. string is the We want to minimize the distance d this is equivalent to minimizing they are related by the equation of the curve. is what we mean by a constraint, Problems i applications There are several ways to do methods: (a) elimination, (b) impli (@) Elimination The most obvious method is to climinat FIGURE 2.1 FFF fom the rig tothe pone (9) 37 +y?, But x and y are not independent: ‘This extra relation between the variables involving constraints occur frequently in a problem like this. We shall discuss che following icit differentiation, (c) Lagrange multipliers +1 rey. Then we want to +1 a ___ CONSTRAINTS; LAGRANGE MULTIPLIERS _173 ‘This is just an ordinary calculus problem: as a 20, x sim and whichis a not inmenely obvious which of these points is ms timum, so nti imple problem ics worth while to ind the scond deaive: “ 2s? ae 2 ara (rclative maximum), tr aty=+V$ (minimum) tvby=t ‘The minimum we wanted then occurs at nose i had not een possible cose for and 0) Impl Oiarentaton Supine tha ote p Matar: ne could il do the prom, From fs" ry me bd ¥ 1) Yu deds rt dy oe Fades rye From an equation like y= 1— 2? relating x and , we could find dyin terms of dx even ifthe equation were not solvable for y. Here we get dy = 2s ds. Eliminating dy from df, we have ‘ gatrtyte oe Larente inna treet =P sy no, Thi neously ith the equation of the cure This equation must now be solved simultaneously with Vm, We get r~ 4a(1 — a2) = Ox = 0 or t.V/ as before To test for maxima or minima we need d*f/dx?. Differentiating df {dx in (9.1) with — QB * ae Atx=0, we find y= 1, dyide = 0, dy/ds* = —2, 90 this is a maximum point. At x=f? 174_PARTIAL DIFFERENTIATION chs this point is the required mir Notice particular c inimum. Notice particularly here that you could do every step of (b) even if the equation of the curve could not be solved for y. We can do problems with several independ bles by independent variables by methods similar to those wwe have just used in Example 1. Consider this problem Example 2, Find the shortest distance from the orig We want to minimize the distance d= /s* + y? +2) from the origin to a point (,y, 2) on the plane. This is equivalent to minimizing f= d? =x? +)? 422 if =~ 2y~2e = 3, We can climinate one variable, say x, from f using the equation of t plane. Then we have Fusing the en f the to the plane x ~ fa G+ dy + 2p ty? ta? Here fis a function of the two independent variables y and 2, so to minimize / we set Oley = 0, Offoe = 0. = 2B +2y 422) 2427-0, Is 918 23 +.2y + 2e) 242 =O. Solving these equations for y and 2, we get’ plane we get x = §. Then Sain = 3) + BP + GP = 1, doin = 1 4, s0 from the equation of the Ivis clear from the geometry that there i a minimum distance from the origin to a plane; therefore this is it without a second-derivative test. (Also see Chapter 3, Section 5 for another way to do this problem.) Problems with any number of variables cam be done this way, or by method (b) if the equations are implicit. () Lagrange Multipliers However, methods (a) and (b) can involve an enormous amount of algebra. We can shortcut this algebra by a process known as the method of Lagrange multipliers ot undetermined multipliers. We want to consider a problem like the one in (a) or (b). In general, we want to find the maximum or minimum of a function f(x, y), where x and y are related by an equation g(x, y) = const. Then fis really a function of one variable (say x). To find the maximum or minimum points of f, we set dfidx = 0 or df= 0 as in (9.1). Since ¢ = const., we get dp = 0. (92) set CONSTRAINTS; LAGRANGE MULTIPLIERS _175 Jn method (b) we solved the dd equation for dy in terms of dx and substituted it into JF this often involves messy algbra. Instead we shall multiply the d@ equation by 2 this it he undetermined muliplicr—we shall fd its value later) and adit co the df Guaion; then we have 03) We now pick 2 so that oa) [hat is, we pick 2 = —(Ef/ey)MCd/ep), but it isn't necessary to write it im this complicated form! In fact, this is exactly the point of the Lagrange multiplier 43 by ting the abbreviation 4 for a complicated expression, we avoid some algeba.] Then from (9.3) and (9.4) we have 0.6) Flr, 9) =Sle.9) + Ab 9) of two independent variables x and y and we wanted to find its maximum and mini~ mum values. Actually, of course, « andy are not independent; they are related by the 4 equation. However, (9.6) gives us a simple way of stating and remembering how to get equations (9.4) and (9.5). Thus we can state the method of Lagrange multipliers in the following way <—————— ee | @.7) To find the maximum or minimum values of f(x, y) when x and y are related by the equation (sy) = const., form the function F(x, y) as in (9.6) and set the two partial derivatives of F equal to zero [equations (9.4) and (9.5)} ‘Then solve these two equations and the equation (x, ») = const. for the three unknowns x, y, and i, As a simple illustration of the method we shall do the problem of Example 1 by Lagrange multipliers. Here Seay ty Oe dayteAl 176_PARTIAL DIFFERENTIATION chy and we write the equations to minimize Faas ida +5? + iy +2), (0.8) We solve these simultaneously with the @ equation y+ 4? in (9.8), either x =O or A= =I, If 1, the second equation gives y and then the $ equation gives x* = 4. These are the same values we had before. The method offers nothing new in testing whether we have found a maxijnum or a minimum, so we shall not repeat that work; if it is Possible to see from the geometry or the physics what we have found, we don’t bother tw test Lagrange multipliers simplify the work enormously in more complicated problems, Consider this problem. 2 Example 3. Find the volume of the largest rectangular parallelepiped (that is, box), with edges parallel to the axes, inscribed in the ellipsoid Let the point (x, y, 2) be the corner in the fi it where the be I first octant where the box touches the ellipsoid. Then (x, », 2) satisfies the ellipsoid equation and the volume of the box is 8xyz (since there are 8 octants). Our problem is to maximize f(x, y, 2) = XJ) £ are related by the ellipsoid equation ie, where Pe ys By the method of Lagrange multipliers we write Aa naaseidntoes f and set the three partial derivatives of F equal to 0: or Beta = Bre $4 Bay +2 sed CONSTRAINTS: LAGRANGE MULTIPLIERS 17 We solve these three equations and the equation ¢ = 1 simultaneously for x, y, 5, and J. (Although we don’t faze to find 4, it may be simpler to find it first.) Multiply the first equation by 2, the second by y, and the third by =, and add to get ‘Using the equation of the ellipsoid, we can simplify this to = Raye, Ways +=0 or Substituting 4 into the @F/0x equation, we find that de yz — ys = 0, From the geometry it is clear that the corner of the box should not be where yor = is equal to zero, so we divide by yz and solve for x, getting “The other two equations could be solved in the same way. However, itis pretty clear from symmetry that the solutions will be y? = 40 and 2? = 4e?, Then the maximum volume is ab Baye = 3 ‘You might contrast this fairly simple algebra with what would be involved in method (@). There you would have to solve the ellipsoid equation for, say, 2, substitute this into the volume formula, and then differentiate the square root. Even by method (b) you ‘would have to find G2/2x or similar expressions from the ellipsoid equation, We should show that the Lagrange method is justified for problems involving several independent variables. We want to find maximum or minimum values of (95 2) if (x,y, 2) = const. (You might note at each step that the proof could easily be extended to more variables.) We take differentials of both the J and the ¢ equations. Since 4 = const., we have dq) = 0. We put df= 0) because we want maximum and minimum values of f° Thus we write x, 2 ar Sat 99) do We could find dz from the d@ equation and substitute it into the df equation; this ‘corresponds to method (b) and may involve complicated algebra. Instead, we form the sum F = f+ Ap and find, using (9.9), 0.10) dF=af+idg «(Lei B)ar (Leakall 178_PARTIAL DIFFERENTIATION cus ‘There are two independent variables in this problem (since x, y, and < are related by = const.). Suppose x and yy are the independent ones; then = is determined from the equation, Similarly, dy and dy may have any values we choose, and ds is determined. Let us select 2 so that an) £4,% Then from (9.10), for dy = 0, we get (9.12) and for de = 0 we get (0.13) ate We can state a rule similar to (9.7) for obtaining equations (9.11), (9.12), and (9.13). eee ee eee nee | 0.14) To find the maximum and minimum values of f(x,y, 2) if (8, 94 2) = const, we form the function F = f+ 29 and set the three | partial derivatives of F equal to zero, We solve these equations and the equation $ = const, for x,y, 2, and 4. (For a problem with still more variables there are more equations, but no change in method. It is interesting to consider the geometric meaning of equations (9.9) to (9.13). Recall that x, y, 2 are related by the equation (1, », think of solving the ¢ equation for const, We might, for example 24,9). Then x and y are independent vari= ables, and 2 is a function of them. Geometrically, 2 2(x, 9) is a surface as in Figure 3.2. If we start at the point P of this surface (see Figure 3.2) and inerease x by ds, y by dy, and = by dz as given in equation (3.6), we are at a point on the plane tangent to the surface at P. That is, the vector dr = ide +jdy+kdz (PB in Figure 3.2) lies in the tangent plane of the surface. Now the second equation in (9.9) is a dot product {see Chapter 3, equation (4.10)] of de with the vector (called the gradient of @ and written grad ¢; see Chapter 6, Section 6fF.). We could write the second equation in (9.9) as dq = (grad ) - dr = 0, Recall [Chapter 3, equa- tion (4.12)] that if the dot product of two vectors is zero, the vectors are perpendicular, ‘Thus since dr lies anywhere in the plane tangent to the surface $ = const., (9.9) says that grad @ is perpendicular to this plane, or perpendicular to the surface ¢ = const. at P. The first of equations (9.9) says that grad /'is also perpendicular to this plane. Thus grad ¢ and grad fare in the same direction, so their components are proportional; this is what equations (9.11), (9.12), and (9.13) say. We ean also say that the surfaces = const. and f= const. are tangent to each other at P; that is, they have the same tangent plane and their normals, grad ¢ and grad f, are in the same direction, sec. CONSTRAINTS; LAGRANGE MULTIPLIERS _179 liers if there are several ¢ equations. We can also use the method of Lagrange multipliers if th Suppose we want to find the maximum oF minimum of f(s, 95 = #) if y(t yy = ®) = const. and 2(, ¥, 5m) = const. There are two independent variables, say x and y We write 9 far Lad 8 yg Oba pg Ot gg Oty 0.15) dy = Go de + Sy + GE de +S dw = 0, obs be Oo de 4 SE dy + or a 2 1a + 82 dw =0 dbs tte Again we could use the dp, and dz equations to eliminate dz and dw from df (method b), but the algebra is forbidding! Instead, by the Lagrange method we form the function F= f+ 2b, + Az 2 and write, using (9.15), 916) dP = df + dy dds + da dbs oy 0.18) gta Gite and for dv = 0, we have (9.19) ‘As before, we can remember the method of finding (9.17), (9.18), and (9.19) by thinking: (9.20) To find the maximum or minimum of f subject to the conditions y A const. and @ = const, define F= {+ Aid: + 4ad2 and set each of the | partial derivatives of F equal to zero. Solve these equations and the ‘equations for the variables and the 2's.; 180_PARTIAL DIFFERENTIATION ch, Example 4. Find the minimum distance fom the origin to the intersection of xy with 7x + 24s = 0. : ve orome We are to minimize x? + Te +24 subject to the two conditions xy =6 and 0. By the Lagrange method, we find the three partial derivatives of F Pty bet + Ae $2 + Aaxy and set each of them equal to zero, We get Det Tay t Ary (9.21) yt Ays 2s + 24a, ‘These equations can be solved with ay = 6 and 7x + 24 £12 0 t0 get (Problem 10) F7/10. y= £52, ‘Then the required minimum distance is (Problem 10) da JP PROBLEMS, SECTION 9 1. What proportions will maximize the area shown in the figure (rectangle with isosceles triangles at its ends) ifthe perimeter is given? Wha repens wil mina te wl of rok in he rm of with one conical end and one flat end, if the surface area is given? “ 3. Find the largest reetangular parallelepiped (box) that ean be shipped by pare a anes ipped by parcel post (length Pind the largest box (with faces parallel tothe coordinate axes) that can be inscribed in 5. Find the point on 2x + 3y 4 0 for which 4x? + 9? + 24 is minimum, 6 A box has three of its faces in the coordinate planes and one vertex on the plane 2x + 3y +42 = Find the maximum volume for the box Repeat Problem 6 ifthe plane is ax + by + 8. A point moves in the (x, x) plane on the line 2x + 3y —4~0. Where will it be when the sun of the squares of its distances from (1, 0) and (—1, 0) is smallest? =a. see.10 ___ENDPOINT_OR BOUNDARY POINT PROBLEMS 181 49, Find the largest triangle that can be inscribed in the ellipse (x?/a?) + 07/6") = 1 (assume the triangle symmetric about one axis of the ellipse with one side perpendicular to this axis). 10. Complete Example 4 above Find the shortest distance ftom rhe origin to the line of intersection of the planes n dety—2= land sy + 12, Find the right triangular prism of given volume and least area if the base is required 0 be a right triangle, 10. ENDPOINT OR BOUNDARY POINT PROBLEMS So far we have been assuming that if there is a maximum or minimum point, calculus will find it. Some simple examples (see Figures 10.1 1o 10.4) show that this may not be true, Suppose, in « given problem, x can have values only between O and 1; this sort of restriction occurs frequently in applications. For example, if x = |eos 6), the graph of fla) =2— 2 exists for all real x, but it has no meaning if x = |eos 61, except for x between 0 and 1. As another example, suppose x is the length of a rectangle whose perimeter is 2; then x < 0 is meaningless in this problem since + isa length, and 1 > 1 is impossible because the perimeter is 2. Let us ask for the largest and smallest values of each of the functions in Figures 10.1 to 10.4 for 0 1. If y = «/x, "is discontinuous at the origin Example 1. A piece of wire 40 em long is to be used to form the perimeters of a square and a circle in such a way as to make the total area (of square and circle) « ‘maximum. Call the radius of the circle r; then the circumference ofthe citcle is 2x7, A length 40 — 2nr is left for the four sides of the square, so one side is 10 ~ dr ‘The total area is A P+ (10 ~ dre Then aA a nr + 2010 fxn) = + )- lor. We dAfdr = 0, we get A= 564 Now we might think that this is the maximum area, But let us apply the second derivative test 10 see whether we have a maximum. We find eA ® St a(t 4 2) > aenm(it foo wwe have found the minimum arca! The problem asks for a maximum, One way to find it would be to sketch 4 as a function of 7 and look at the graph to see where 4 hay its largest value. A simpler way is this. is a continuous function of r with a continuous derivative, If there were an interior maximum (that is, one between pO. and 2nr = 40), calculus would find it. Therefore the maximum must be at one end or the other, At At Qn 0 A= 100, 0, r=637, 4 (6.37)? = 1274. We see that A takes its largest value at r= 6.37; 4= 127+ is then the desired ‘maximum. It corresponds to using all the wire to make a circle; the side of the square AA similar difficulty can arise in problems with more variables. .DPOINT OR BOUNDARY so Example 2. The temperature in a rectangular plate bounded by x = 0, y = 0,» =3, and y = Sis T= ay? = xy + 100. ind the hottest and coldest points of the plate x ‘We first set the partial derivatives of T equal to zero to find any interior maxima and minima. We get “The only solution of these equations is x = y = 0, for which 7 ~ 100. ‘We must next ask whether there ae points around the boundary of the plate where TT has a value larger or smaller than 100, ‘To sce that this might happen, think of a raph of plotted as a function of x and y; this is a surface above the (x, 9) plane ‘The mathematical surface docs not have to stop at x= 3 and y = 5, but it has no meaning for our problem beyond these values. Just as for the curves in Figures 10.1 to 104, the graph of the temperature may be increasing or decreasing as we cross a boundary; calculus will not then give us a zero derivative even though the temperature at the boundary may be larger (or smaller) than at other points of the plate. Thus we must consider the complete boundary of the plate (nor just the corners!) ‘The lines | x= 3, and y= 5 are the boundaries; we consider each of them in turn On x = and y = 0 the temperature is 100. On the line + = 3, we have T= 3)? —9y + 100. We can use calculus to see whether T has maxima or minima as a function of y along this line. We have Simitarly, along the line y Finally, we must find T’at the corners. At 0,0), 0, 5), and (3,0), T= 100. At G3, T= 130. Putting all our results together, we see that the hottest point is ($5) with T= 1314, and the coldest point is (3, 3) with T= 934ae 184_PARTIAL DIFFERENTIATION cus Example 3, Find the point or points closest to the origin on the surfaces @ Pe -He~8, (b) We want to minimize f= 17 + y? +2? subject to a condition [(a) or (by]. If we climinate 17 in each case, we have a ) fae 14st tn both (a) and (b) the mathematical faction fy, 2) i define fr ally and problems, however, this is not true. In (a), since x* > 0, we have x? = 8 + 4 We are interested in minimum values of f(y. 2) in (a) only in the region [Compare Example 2 where T(x, ») was of interest only inside a rectangle.) ‘Thus we look for ‘interior minima in (a) satisfying yz > —2; then we substitute = = —2)) into (10.28) and find any minima on the boundary of the region of interest. In (hy ao.) B+ dys ty? + 2, (102) mete Peyad Por our 0 so since x? = 2? — 12 0, we must have 2° > 1. Again we try to find “interior” minima satisfying 2? 2 15 then we set 2? = 1 and look for boundary minima. We now carry out these steps, From (10,23), we find (10.3a) These values satisfy the condition yo > —2 and so give points ise the_region of interest. We find from (10.la), x7 = 8, x= +24/2; the points are (+2,/2, 0, 0) at distance 243 tom the gin, Net we camider he bounder e ts ee ap ho (10.2%, 4 sooty ss, . FV Remembering that x= 0, we have the points (0, /2, ~/2) and (0, —/2, /3) a distance 2 from the origin. Since 2< 2/2, these boundary points are closest to the origin, (0.44) Answer to @): (0, V2, ~/2), (0, —/2, /). Prom (10.2b) we find (10.36) ENDPOINT OR BOUNDARY POINT PROBLEMS _15 gee) 1, there is no minimum point inside the region of = 1. From (10.1b), x = 0, and from (10.2b) ferth Lax since «= 0 does not satisfy jaterest, So we look at the boundar , y= 0, ay “hus we find the points (0, 0, +1) at distance 1 from the origin, Since the geometry tells us that there must be 2 point or points closest to the origin, and calculus tells us that these are the only possible minimum points, these must be the desired points (lob) Answer to (b): (0,0, £1). In both these problems, we could have avoided having to consider the boundary of the region of interest by eliminating = to obtain fas a function of x and y. Since x and Jy ate allowed by (10.12) of (10.16) to take any values, there are no boundaries to the Tegion of interest, In (b) this is a satisfactory method; in (a) the algebra is complicated In both problems, Lagrange multipliers offer a more routine method. For example, in (@) we write Patty tt tie —4 Ss oy ae ty We obtain the same results as above, namely, the points (¢24/2,0,0), (0, £.V2, F./2), the points (0, /2, ~/2), (0, —/2, x/2) are closer to the origin by inspection. Part (6) eam be done similarly (Problem 15) We see that using Lagrange multipliers may simplify maximum and minimum prob- lems. However, the Lagrange multiplier method still relies on ealculus; consequently, it ean work only if dhe maximum and minimum can be found by calculus using some set of variables (x and y, not y and =, in Example 3). For example, a problem in which the maximum of minimum occurs at endpoints in all variables cannot be done by any method that depends on setting derivatives equal to zet0, Example 4. Find the maximum value of y’— x for nonnegative xand y if x* Here we must have both s and y between 0 and 1. Then the values y tive y — x its largest value; these are both endpoint values which cannot be found by aaleulus PROBLEMS, SECTION 10 1. Find the shortest distance from the origin to x? — 2. Find the largest and smallest distances from the origin to the conic whose equation is Sx? — 6xy 4+ 5)? — 32 = O and hence determine the lengths of the semiaxes ofthis con.aii tard 186 _PaRTL DIFFERENTIATION, 3. Repeat Problem 2 for the conic 61? + 4xy + 3)? = 28 ind the shortest distance from the origin to each of the following quadric surfaces. Hint: See Example 3 above. 6 Hy 4224 + Say = 8 Find the largest ¢ for which 2x + 4y = 5 and x? + © wy. Find the hottest point (or points) on the 12, and find the temperature there. 8, Ifthe temperature atthe point (x,y, 3) is T= ay surface of the sphere x? + y? + 22 9. The temperature T of the circular plate x? +5? < 1 is given by the hottest and coldest points ofthe plate Bet = 3)? — De Find 10. The temperature at a point (x, 3,2) in the sphere x24)? 22<1 is given by T= y? + xz. Find the largest and smallest values which T' takes (2) on the circle y = 0, x7 + 2 = (b) on the surface 2? 4 y? +24 = 1, (6) in the whole sphere. 11. The temperature of a rectangular plate bounded by the lines x = +1, y = 41, is given by T= 2x? ~ Sy — 2x + 10, Find the hottest and coldest points ofthe plate 12, Find the largest and smallest values of the sum of the acute angles that a line through the ‘origin makes with the three coordinate axes. 13, Find the largest and smallest values of the sum of the acute angles that a line through the origin makes with the theee coordinate planes 14. The diagram shows a parking lot, 20 by 60 yd. A contractor has to run a power line from .1 10 G; he can put it om poles around ABC at $60 yard or 0 underground part or all the way at $75 a yard. If he is © R_# honest and wants to minimize the cost, to what point D fo Gif any) should he run the underground part? If he is dishonest and wants the maximize the cost, but will run Bw 4 the Tine straight along AD (because an inspector is watching), where will D be? 15, Do Example 3b using Lagrange multipliers. ‘11. CHANGE OF VARIABLES ‘One important use of partial differentiation is in making changes of variables (for ‘example, from rectangular to polar coordinates). This may give a simpler expression oF 4 simpler differential equation or one more suited to the physical problem one is doing, For example, if you are working with the vibration of a circular membrane, or the flow of heat in a circular cylinder, polar coordinates are better; for a problem about sound waves in a room, rectangular coordinates are better. Consider the following problems, CHANGE OF VARIAULES 187 gee ANGE OF VARIABLES 187 Example 1. Make the change of variables y= + 38 = 4 ~ st in the mace equation FOF qty 2a and solve the equation, We use the equations aia) and equations like (7.2) to find OF _ OF & ox or or oo OF _ oF & ie hep to sayin war what we have writen in (13): To find the pata of ocon with reece ty we find its paral with rapt oF plus its paral with ae ST be vec wih toes to me Bed ce prt nis cop of TSErs ae antl wh pet tot and Ply by the eonean I ire co we tis in ooo notion ne Chapter, Secon 7 aay ans) that is, the r derivative of OF/és is zero, ‘Then OF/0s must be independent of r, 50 27/81 = some faction of + aloe, We integrate with rapes! to + to, find F=f() + “const”; the “constant” is a constant as far as 5 is concerned, but it may {)g(r) = 0. Thus we find that the solution of be any function of r, say g(r), since ( (11.6) is a7) F=f) + 640).a ti PARTIAL p ip Then, using (11.2), we find the solution of (11.1) ans) F (ee) + ee + 00), Where /and ¢ are arbitrary functions, Example 2. Write the Laplace equation or (119) in terms of polar coordinates r, 0, where x =r cos 0, (Lo) yer sin 0. Note that equations (11.10) give the old variables x and y’ in terms of the new oncs, 1 and (, whereas (11.2) gave the new variables r and s in terms of the old ones. In thig situation, there are several ways to get equations like (11.3). One way is to write oF any and then solve (11.11) for @F/éx and OF needed partial derivati using (7.16), y (Problem 5). Another way is to find the cs of r and @ with respect to x and y [for methods and results, Pxample 6, equation (7.16) and Problem 7.9] and then write as in (11.3), n12) In finding the second derivatives, it will be convenient to use the abbreviations G = OP Jéx and H = OF ey. Thus, cova 2t 2000? (11.13) Z re 2g gf, e007 aE ino Eg 8 o 8 Tw ven aus ANGE OF VARIABLES 189 geet ee Pr lar they are correct if fow equations (11.12) are correct for any function F; in particular they * te teptce by G ar by He Let apace F by @ nthe ft equton (113) and eplace F by H in the second equation. ‘Then we have a 2G _sin 0G Fem es 8 oH a ais) Substituting (11.15) into (11.14), we get ee OF ata ac oH =e 0S + sin T+ a6) We find the four partial derivatives of Gand H which we need in (11,16), by differen- tiating the right-hand sides of equations (11.13) P oo 0 HE a 2r Hot a7 tt agg BE cos 9 oF Oe ee er aG (11.18) Finally, substituting (11-18) into (11.16) gives or oF ati) We next discuss a simple kind of change of variables which is very useful in thermo- dynamics and mechanics. This process is sometimes known as a Legendre transform= tion. Suppose we are given a function /(x, 9); then we can write (11.20)rata 190 PARTIAL 1 Let us call fae and df/@y = q; then we have (en) Oa pdr tady. If we now subtract from df the quantity digy), we have Y~ day) = pdx +a dy—gdy—ydg or (11.22) AS) = 9 dey dg. If we define the function g by (11.23) e=f-o then by (11.22) (2s de yg Because dx and dy appear in (11.24), it is convenient to think of g as a function of » and q. The partial derivatives of ¢ are then of simple form, namely, (11.28) fan Ba- er a Similarly, we could replace the p dx term in df by —x dp by considering the function £— xp. This sort of change of independent variables is a Legendre transformation. See Problems 10 and 11 for applications PROBLEMS, SECTION 11 1. In the partial differential equation put s=y +21, ¢= y+ Sr and show that the equation becomes the method of solving (11.6), solve the equation 01 =0, Fellowing 2. Asin Problem 1, solve bby making the change of variable w 3. Suppose that w = /(x, 3) satisfies Pur xvaat equation v=u—s, and show that a sttsies Ce/du Cr 1. Hence solve the 4. Verify the chain rule formulas 10. 0 CHANGE OF VARIABLES 191 and similar formulas for Collect couficents of dx and dy; these are the values of OFYéx and OF (ey Solve equations (11.11) to get equations (11.12) (0) -42)- Reduce the equation and y by the change to 2 differential equation with constant coefficients in d?y/de*, dy/ of variable # = cos @ and show that the Legendre ‘Change the independent variable from x 0 0 by equation adv _5# - Bw 2eyno0 Oe ‘vecomes ay © 5 co 0% + 2y =0. ae 6 Change the independent variable from x to 4 = 24/¥ inthe Bese equation ady dy dea a-ay= and show that the equation becomes 4 weet —aye Biteh rw Ihr =e 0086, = # sin show that Given du = T ds — p de, find a Legendre transformation giving (@) a function f(T, 0); (8) a Function My) {e) 4 function x7 9). ‘Hint for (e)+ Perform a Legendre transformation on both terms in de2 FERENTIATION Me Given Lg i) such thar db = p dg + p dj find Hip, 4) so that dH = § dp — f de Comments: L and H are functions used in mechanics called the Lagrangian and the Hamilz tonian. The quantities j and f are actually cime derivatives of p and g, but you make no yx ‘ofthe fact inthis problem. Treat # and ¢ as if they were two more variables having nothing to do with » and q. Hin: Use'a Legendre transformation. On your first try you wit probably get —H1. Look at the text discussion of Legendre transformations. and satan, yourself thac ¢ = gy —f would have been just as satisfactory as g = f-— gy in (11.23), 12. DIFFERENTIATION OF INTEGRALS: LEIBNIZ’ RULE According to the definition of an integeal as an antiderivative, if aFx) a soe, then (122) [vo t= FO} = Fa) Fa), where @ is a constant. If we differentiate (12.2) wi 4 12.3) = (123) br by (12.1). Similarly, respect to x, we have as) is) 10) de= 4 Ce) — Re) [y () dt = Fla) — Fs), (24) F0 dr = Fo 765 Example Find [sin at By (12.3), we find immediately that the answer is sin x, We can check this by finding the integral and then differentiating. We get ———— cour + 4/3 and the derivative of this is sin x as before By replacing x in (12.3) by v, and replacing x in (12.4) by w, we can then write f “(ry aereo ) Ef rower set {PERENTIATION OF INTE ~ 4 (26) mal Suppose w and are functions of x and we want df/dx where r l IO de wd, the answer depends onthe Knits and, Findig en the itr seve, th sands Fig Ha eral dieenaton problem; 1 Te» fin af and which mnctions of x. We can write a dtl du | Ode aan de tude” dv de tds means to dierent I with pc ae ee nde pte, Sar, ern thas = cnt and we cine (128) to get af/éu = —/(u). Then we have [ pe ie te 128) é cC 0) d= fo) 2 fy beample 2. Find diet = [1d By (12.8) we get 1 ap Leryn ta Lea et BY eG We cold sso integrate fist and then diferente with respect 10 wpe sat ae as This last method seems so simple you may wonder why we need (12.8). Look at another example. Example 3. Find di/ds if Here the indefinite integral cannot be evaluated in terms of elementary functions; however, we can find dffdr by using (12.8). We get(OF INTEGRALS; LEIBNIZ’ RULE_195 194_PARTIAL DIFFERENTIATION Bear tnen chs se _DIFFERENTIATIC | : except that we need to find C. Finally, we may want to find difde when Im {¥ f(xy) dt, where a and b are ‘his is the value ofthe integral / which we wanted, eacee thar we need c9 nd constants: Under not too restrictive conditions, When «= 0, /= (ln 1) +C=C. Brom the orginal inepra, when «= 0 grand (# — Tn cis (? = 1m ¢ = 0, 9 FO, Thus C = and we hav 5 (129) Er odt= that is, we can differentiate under the integral sign. [A set of sufficient conditions for this to be correct would be that f8 f(s, 1) dt exists, Of/Or is continuous and I is convenient to collect formulas (12.8) and 12FG, N/2s| s aC), where J2 (0) de exists. For most practical purposes this means ther | if both integrals in (12.9) exist, then (12.9) is correct.] Equation (12.9) is often useful in evaluating definite integrals, Te=in(a +), (12.9) into one formula known as Leib! rule ae : 2.) z( fle dt pon tpn + [ a Example 4. Find (2! dr for odd n, #> 0 Fist we evaluate the integral gxample 6 Find diide if | refit camp Now we calculate successive derivatives of / with respect to &. HP Layee 1 cae Ga [teense or [Pew aed Repeating the differentiation with respect to &, we get By (12.11) we get { i Continuing in this way (Problem 17), we can find the integral of any odd power of 1 aed PROBLEMS, SECTION 12 ve as as 1d send to zero. - bn and Bu and ao ir its 28 ~— Example 5. Evaluate 2 we ty find = * fore sine de ° r=[‘S a azo ol me > lat oe First we differentiate J with respect to a, and evaluate the resulting integral. 4. Use LHipitals rue weve in =P | a (teint 1 5 1 - a rolte ssn a Now we integrate di/da with respect to @ to get J back again (plus an integration . constant) “s eee Lae me di ‘Hint: Use differentials, | in a 2 i h196 PARTIAL DIFFERENTIATION MISCELLANEOUS PROBLEMS 197 tu ferace at we mt (8), (2 eee [evtnn oot change of variable + (6) Show thaty = | W709 da satises 9° =F) son sin the integral) 9. fF y= [" sin ae dt, find bby evaluating the integral and then diffe i aa lifferetiaing, 1 (©) Show that y 7 eae de atten” =F, (b) by diferentiating frst and then evaluating the integral 10. Find dys explicit if om i 1. Fiea 4 (* 43, MISCELLANEOUS PROBLEMS ind & | (=) de both by evaluating the integral ist | HJ, 64 by cvatatig he ine fi and by diterening fe ILA funtion fs. 92) x led bamoqeneaus af degre i 91) = Ft 2. or xample, 2 In (a) is homogencous af degree 2 since i a Fina 4 [te ie, 2 Tn (x) is homogen : |, intra i | “ee (a) n= 4 (sin at b 13, Find 4 ("S22 al. v4 le'ssheoesn on homogeneous functions says tha if fis hamogencous of degree then = p mevins Sy ah 5 x o.% a SA. Given that bier Hy y Ls Lag 7 | 2 dilerentate with respect toy and so evaluate ae) iy fv ae Prove this theorem, Hints: Differentiate len 0) = flay 2) with respect 0 f and Lo» then ke r= 1 Tr convenient 10 eal 22s) =; (ha is the partial derivative oF with 15. Given th ae fespect to its st variable), fs ~ of/6iy), and so on. Or, you can at first eal =a en that Drees fe~ (Both the definition and the theorem can be extended (0 any number of rs variables ) [- differentiate with respect to a to show that {find the distance (a) Given the point (2,1) in the (x, ») plane and the line 3x + 2y from the point to the line by using the method of Chapter 3, (8) Solve part (a) by writing a formula forthe distance from (2,1) t0 (45) and minimizing sin br d the distance (use Lagrange multipliers) (©) Derive the formula and differentiate with respect to & to show that for the distance from (tos yo) #0 ax + Ay = © by the methods suggested in parts (a) and (6). AG 16. In Kinetic theory we have to evaluate in werrls of the form I= "Fe" 2, Given tht In Problems 3 to 6, assume that x,’ and r, 0 are rectangular and polar coordinates. [lo dea Voie, eta Foe 2,456.05 Bm _ 3 4 Find ie 17. Complete Example 4 to obtain (12.10) fo ds arina (5 & . 7. About how much (in percent) does an error of 1% in x and y affect 18. Show that w(x, y) satises uy, + 4, attevs. 2, 2. 23 ys, 25, PARTIAL DIFFERENTIATION _ ons Assume that the Earth is perfect sphere. Suppose that rope lies slong the equator with is guts Gstened so that it fis exactly, Now let the rope te made 2 fe longer, and let ir ke eid the same distance above the surface of the Farth at all points of the equator, Aton how high up is it? (For example, could you crawl under? Could fly?) [28 +2? = 38, (a + 3)? = 6, If w= (7 cos 8%", find Ow /c0, Ie ay and find dz. Y 4 y sAfind 2 and £2 by impicit dtferentiation de 8 ga By implicit di Pe brant tet (se Tea flrs, 2x +3, ¢= dv, find (Gm), in terms of (and its derivatives. Mm = fla, x* + y%,21y), find (Ont), (compare Problem 14), 12), rove the» Find the shortest distance from the origin to the surface x + 10, Find the shores distance from the origin to the line of intersection of the planes det 2a, Br~ y-2e= Hl, (2) using vector methods (see Chapter 3, Section 5); (b) using Lagrange multipliers. Find by the Lagrange multiplier method the largest value of the product of three postive ‘numbers if their sum is 1 Find the largest and smallest values of te) 4 902 Find the hottest and coldest points on a bar of length $ if T= 4x ~ 22, where ¥ is the distance measured from the lft end 12 +3 ifs = 00s 0 Find the hoetese and coldest points ofthe region y? ) plane [4 is shown also in Figure 2.4 (a and b)]. We are going to consider to ways of evaluating this double integral. We think ofthe teiangle f cut up int litle rectangles A = Av Ay (Figure 24) and the whole solid cut into vertical columas of height = and base AA (Figure 2.3), We want the (limit of the) sum of the volumes of these columns. First add wp the columns (Figure 24a) fora fied value of + prodacing the volume ofa slab (Figure 23) of thickness Ax. This corresponds to integrating with respect to y (holding x constant, Figure 2a) ftom y =0 to y on the line that i, 9 ea) Sib of volume Fhe FIGURE 2.3, o FIGURE 24 DOUBLE AND TRIPLE INTEGRALS 201 sree have found isthe are ofthe Sabin Figure 2.3; ts volume isthe ar times Fo lap the somes the sab hs conesponds integrating (2) Saree ea [\ aneer eevee We could summarize (2.1) and (2.2) by writing Capon Wie cal (23) an ered (repeat nega. Mail nega te usually evated by wey gel Nowe thatthe Inge parenthesi23) ae noc ally nes an ef wate the care ging the its on a nepal at always write f1_9, not just [& Sim we cl a op he umes 44) y is incartingwth os : s Pty ening valine oa ab (gud y Figure 240) Rom c= tol 9? sing tor are tou) sin Figure 23 ad then ad up the slumes of he lbs by semen apc toy fomy =O toy = 2 (Pigue 240), We wre an LL ([ieena)o= feo =| et 92 = (4 yt- VD dy =% As the geometry would indicate, the results in (2.2) and (2.4) are the same; we have ‘two methods of evaluating the double integral by using iterated integrals. Often one of these two methods is more convenient than the other: we choose whichever method is easier. To see how to decide, study the following sketches of areas A over which we want to find ff, f(s, 9) dy dy. In each ease we think of combining litle rectangles dx dy’ to form strips (as shown) and then combining the strips to cover the whole area Arcs shown in Figure 23: Integrate with sespect oy frst, Note that the top and bottom of area A are curves whose equations we know; the boundaries at x =a and = b are either vertial straight lines or else points zi 1 ate : e FIGURE 25201_MULTIPLE INTEGRALS cy, We find 5) | [rtssa9 de dy = ii (re fess7) 6) de Areas shown in Figure 2.6: Integrate with respect to x fest. Note that the sides of area are curves whose oqutions we Tnow: the Bounds either horizontal straight lines or else points. : eee as 5 & rioune 26 We find 26) | Floy) de dy = | (f° (9) «) 4y. Areas shown in Figure 27:1 Figure 27: Integrate n ithe order, Noe that thes res the requirements for both (2.5) and (2.6). " meats yt) a. a < SP eS roune 27 We find en | Flasy) de dy = [ re Sa, 9) dy de = f re Sls, y) dr dy An important special case is a double integral 7 integral over a rectangle (both x and y limits are constant) when fl 3) is 4 product /(e,9) =e@)io) Then nn (2.8) f Sx, 9) dx dy = { f B(x)h(y) dy de -([o0a(0r. Wh en ates are more complicated than those shown, we may break them more simpler areas (Problems 9 and 10). break them int ewo or so} DOUBLE AND TRIPLE INTEGRALS 205 We have considered a double integral as a volume just as we often think of a single einite integral as an area. However, we also use single integrals for other purposes Shan finding areas. Similarly, now that we know how to evaluate a double integral, we van use it t0 find other quantities besides volume. Example 2, Find the mass of a rectangular plate bounded by #=0, x= 2 y= 0, TE its density (mass per unit area) is f(x, y) = xy. The mass of a tiny rectangle A= Ax Ay is approximately f(x, 9) Ax Ay, where f(s, 9) is evaluated at some point BaA, We want to add up the masses of all the A's; this is what we find by ‘alvating the double integral of dM = xy dx dy. We call dMf an element of mass and think of adding up all the dA’ to get M. Fo) ue [foao-f [oae (flan A tripe integral of f(x, yy 2) over a valume ¥ written fffy f(x, yy 2) dx dy dey is also defined as the limit of a sum and is evaluated by an iterated integral. If the integral is over a box, that is, all limits are constants, then we can do the x, 95 = integrations in any order. If the volume is complicated, then we have to consider the geometry as we did for double integrals to decide on the best order and find the limits. ‘This process can best be learned from examples (below and Section 3) and practice (see problems) Example 3, Find the volume of the solid in Figure 2.3 by using a triple integral Here we imagine the whole solid cut-into tiny boxes of volume Av Ay Az; an element of volume is de dy dz. We first add up the volumes of the tiny boxes to get the volume of a column; this means integrating with respect to = from 0 to 1+y with x and y constant. Then we add up the columtis to get a slab and the slabs to get the whole volume just as we did in Example 1. Thus ey v= [ffecare LLC = LL Lem fe [ “as gana as in (2.1) and (2.2), Or, we could have used (2.4).26_MULTH INTEGRALS Example 4. Find the mass of the solid in Figure 2.3 if the density (mass per unit slum) i's An kent af ass ey of mass just as we add up elements of volume; y ae in Eee that is, the limits are the same as in Qn) M | rf acenenet =] o|, GO hase ae on fico tyre soem -f fit 2s)? — 11+ 3 — 24)? fae a =} floss 20) a2 PROBLEMS, SECTION 2 In Problems 1 to 6 valute the iter [ [oe 2 fl [mae 2 (ee [fee 8 fl flee 6 f Poee In Probl 7 018 eval he db the double ih oer he aes desi To Sd the nin sketch the area and compare Figures 2.5 to 2.7. aaa 1. Sax — 390 de dy, where A isthe triangle with vertices (0, 0, 2,1, 0) 8. {fc 69? ew x de dy, whore isthe area inclosed by the curves fo? on ca in x the = ais and 9. Jf sin x de dy where A is the area shown in Figure 28. | a 10, Jf de dy where isthe aren in Figure 28 2 riaune 28 Th. fa x de dy, where Ais the area between the parabola y= and the straight line 12, fy de dy over the triangle with vertices (1, 0), (0,2), and 2 0). sek DOUBLE AND TRIP! 13, ff 207 de dy over the eangle with vertices (0, 0. (2, 1, 4. ff 2°" de dy over the area bounded by y= ',y= 27%, and x= In 4 15, {de dy over the area bounded by y = In y= ¢+ 1— and the axis 1s, {f 0 + 29° de dy over the quadsiateral with vertices (1, 3), 3. (2 6) (6, 8 An {f Ga de dy over the tingle with vertices (0, 0) (1, Dy (hs 2) 1B. {fy Me dy ove he area bounded by y= 33, 4 +y = 2,and the y axis. In Problems 19 to 24, use double integrals to find the indicated volumes 19, Above the square with vertices at (0, 0), (2, 0}, (0, 2), and (2, 2), and under the plane pe8arty. 20, Above the rectangle with vertices (0, 0), (0, 1), 2, 0), and (2, 0, and below the surface = 36rd 2) 21. Above the triangle with vertices (0, 0), (2, 0), and (2, 1), and below the paraboloid wae ny 22, Above the triangle with vertices (0, 2) (1,1), and (2, 2), and under the surface = = sy 2B. Under the surface = = y(+ + 2), and over the atea bounded by x+y =0,y=1y= Vir 24. Under the surface = ~ 1/( + 2), and over the area bounded by y=: andy? + In Problems 25 to 28, sketch the area of integration, observe that itis Tike the areas in Figure 23, and so write an equivalent integral with the integration in the opposite order. Check your ‘work by evaluating the double integral both ways. « [ [ee fl [Leone m (vee a [fom In Problems 29 t0 32, observe that the inside integral eannot be expressed in terms of elementary functions. As in Problems 25 to 28, change the order of integration and so evaluate the double integral », j i 8 sea Fy PP dy de aa ees rae a. oe x Say i f wo lay ° le ae 33. A lamina covering the quarter circle x* +3? <4, x>0, y > 0, has (area) density x + Find the mass ofthe lamina 3H. A dielectric lamina with charge density proportional ro y covers the area between the parabola y = 16 — x? and the x axis, Find the total charge 38, A triangular lamina is bounded by the coordinate axes and the line x + y = 6. Find its mass ifits density at each point Pis proportional tothe square of the distance fom the origin 10 P.205_MULTIPLE INTEGRALS cus 36. A partially silvered mirror covers the square area with vertices at (41, +1). The fraction of incident Tighe hich it reflects at (x, 9) is (r—y)?/4. Assuming a uniform intensity of incident light, find the fraction reflected In Problems 37 to 40, evaluate the tiple integrals a. [eae PL Roeewe 0 [LP fenaa 41. Find the volume between the planes 2 = 2x +3y-+6 and z= 2s +7y +8, and over she ‘tiangle with vertices (0, 0), 3,0), and (2, 1). 42, Find the volume between the planes z= 2e + 3y +6 and 22s +7) +8 and over the square in the (x,y) plane with vertices (0, 0), (1, 0), (0, 1), (1 43. Find the volume between the surfaces 2 = 2s? + y? 4 12 and c= x? 432 +8 and over the triangle with vertices (0, 0), (1,0), and (1, 2) 44. Find the mass of the solid in Problem 42 ifthe density at (x, y, 2) is proportional ty. 45. Find the mass of the solid in Prablem 43 ifthe density is proportional to x 46. Find the mass ofa cube of side 2 if the dens from the center of the cube. 's proportional to the square of the distance 47. Find the volume in the fist octant bounded by the coordinate planes and the planc rt dytend, 48. Find the volume in the frst ectant bounded by the cone , where M is the mass of the body, J, is the moment of inertia of the body about an axis through the center of mass and parallel to the given axis, and d isthe distance beeween 2 For a thin rod of length / and uniform density p find @ M, (b) Jp about an axis perpendicular to the rod, (©) I'sbout an axis perpendicular 0 the rod passing through one end (see Problem 1), 3. A thin rod 10 fe long has a density which varies uniformly from 4 10 24 Ib/f. Pind (6) Jy about an axis perpendicular to the rod, (@) F sbout an axis perpendicular tothe rod passing through the heavy end. 4. Repeat Problem 3 for «rod of length ! with density varying uniformly from 210 1 5. For a square lamina of uniform density find / about (a) a side, (b) a diagonal, (6) an axis through a comer and perpendicular to its plane. Hint: See the perpendicular axis theorem, Example If, + 6. A triangular lamina has vertices (0, 0), (0, 6) and (6,0), and uniform density. Find @ 85, 1, (©) Zp about an axis parallel to the x axis. Hint: Use Problem 1 carflly 7A rectangular lamina has vertices (0,0), (0,2), (3, 0), (3, 2) and density ay. Find @) M, 25, (© Tals (4) Ly about an axis parallel 10 the = axis. Hint: Use the parallel and perpendicular axis theorems 8, For a uniform cube, find / about ane edge.CHANGE OF VARIABLES IN INTEGRALS; JACOBIANS 217 26 MULTIPLE INTEGRAL cus 9. For the pyramid inclosed by the coordinate planes and the plane x + y +2 {@) Find te volume. (b) Find the coordinates ofits centroid (o) Ue the density is 2, find Mand © 10. A uniform chain hangs in the shape of the ctenary y = cosh x between x= —1 and x= 1 Find (a) its length, (6) 5 11. A chain in the shape » @ Mm, b) m5 7 between x= —1 and x has density [+], Find Prove the following two theorems of Pappus 12, The area 4 inside a closed curve in the (x, ») plane, » 2 0, is revolved about the + avis The volume of the solid generated is equal co. times the circumference of the cirle taved by the centroid of A. Hint: Write the integrals forthe volume and for the centroid 13. An are in the (x,y) plane, y 2 0, is revolved about the x ats, The surface area generated is {equal to the length of the ar times the circumference of the circle traced by the centeuid the ae. 14, Use Problems 12 and 13 to find the volume and surface area of a torus (doughnut) 15, Use Problems 12 and 13 10 find the centroids of a semicircular area and of a semicircular are. Hint: Assume the formulas = 4nr?, V = $n? for a sphere 16 Let a curve y =/(a) be revolved about the x axis, thus forming a surface of revolution Show that the eross sections of this surface in any plane x = const [that is, parallel to the 0. 2) plane] are citeles of radius /(s), Thus write the general equation of a surface of revolution and verify the special ease /(x) = x? in (3.9) In Problems 17 to 30, for the curve » = v/%, between x and x= 2, find 17, The area under the curve 18. The ae length. 19. The volume ofthe solid generated when the area is revolved about the ¥ ais 20, The curved ares ofthis solid 21, 22,23. The centroids of the arc, the volume, and the surface ate, 24, 28, 26, 27. The moments of inertia about the « axis of a lamina in the shape of the plane area under the curves of a wire bent along the are of the curve; of the solid of revolutions and of a thin shell whose shape isthe curved surface of the solid (assuming constant density forall these problems) 28. The mass of a wire bent in the shape of the arc if its density (mass per unit length) is \/- 29, ‘The mass ofthe solid of revolution if the density (mass per unit volume) is | 292 30. ‘The moment of inertia about the y ans ofthe solid of revolution if the density is [2y2 31. (a) Revolve the curve y = 5°', from x 1 to x= ony about the x ais to create 4 sunice and a volume, Write integrals forthe serfice acs and the volume Find the volume, and show thatthe surface aenis infinite. Hin: The surface aren integral ot ens to erie, ‘hut you cam easily show that itis greater than 1! which you can evaluate set 8) The flowing question i challenge 16 your bili to fe oxsther your mbes Gltoons and psa! Tico Tn ayo ound ate lume snd on nine es eh re tacos lenving what si tothe sure, Apps ou have painted a ine se With spay sty voeraenienry ie Fete eyo 4. CHANGE OF VARIABLES IN INTEGRALS; JACOBIANS In many applied problems, it is more convenient to use other coordinate systems instead of the rectangular coordinates we have been using. For example, in the plane we foften use polar coordinates, and in three dimensions we often use cylindrical coordi- nates or spherical coordinates. It is important to know how to set up multiple integrals directly in these coordinate systems which occur so frequently in practice. That is, we ‘need to know what the area, volume, and arc length elements are, what the variables r, @, etc., mean geometrically, and how they are related to the rectangular coordinates. We are going to discuss finding elements of area, etc., geometrically for several important coordinate systems. However, if we are given equations like x =r cos &, y= 7 sin 8 relating new variables to the rectangular ones, it is useful to know how to find the elements of area, etc., algebraically, without having to rely on the geometry. We are going to discuss this and illustrate it by verifying the results which we can get geo- ‘metrically for the familiar coordinate systems. In the plane, the polar coordinates 7, @ are related to the rectangular coordinates x, y by the equations x= 10s 6, (4.1) yar sin 8. FIGURE 4.1 Recall that the area element dy dx was obtained by drawing a grid of lines x = const., J = const., which cut the plane into little rectangles dx by dy; the arca of one rectangle was then dy dx. We can make a similar construction for polar coordinates by drawing lines @ = const, and circles r= const.; we then obtain the grid shown in Figure 4.1 Observe that the sides of the area element are not dr and 40, but dr and r d8, and its area is then (42) dA = dr 4 dO =F dr db.2X MULTIPLE INTEGRALS ons FIGURE 42 rly, we can see from Figure +2 that the are length element ds is given by d= dP +7 dB, an (Fern foe (43) Example 1. Given a semicircular sheet of material of radius a and constant density find {a) the centroid of the semicircular area; (b) the moment of inertia of the sheet of material about the diameter forming tt straight side of the semicircle (a) In Figure 4.3, we see by symmetry that p= 0. We want ty, find ¥. By (3.4), we have fora [oars coanginghe 19 pr onan pug ace tn weg | N eft rant pomrana We caleulate the integrals and find FIGURE 43 (8) We want the moment of inertia about the y axis in Figure 4.3; by definition this is [x2 dil In polar coordinates, dM = p dt = pr dr d0, We are given that the density pis constant. Then we hase tap [eraa=o | fe P eosin dr dO = p ™ GHANGE. OF VARIABLES IX INTEGRALS: JACOMLANS. Since the mass of the semicircular object is Map [rican we have (44) Cylindrical coordinates: y= reos = rsin 0 WV mdr dO de dee PaO? + da Spherical coordinates: v= rin O c08 rsin@sing@ y =r eos O 2 sin dr dO dd dP 4 Pd +? sin? Oag?220_ MULTIPLE INTEGRALS us Cylindrical coordinates are just polar coordinates in the (x, ») plane with = for the third vatiable. Note thar the spherical coordinates r and @ in Figure 4.5 are different from the cylindrical or polar coordinates r and 8 in Figures 4.4 and 4.1. Since we seldom use both systems in the same problem, this should cause no confusion, (If necessary, use p or R for one of the r's, and use @ instead of @ in cylindrical coordinates.) Watch out, however, for the discrepancy in notation for spherical coord fates in various texts. Most ealculus books interchange @ and @, ‘This can be confusing later since the notation of Figure 4.5 is almost universal in advanced mathematics (rartial differential equations, special functions), in applications ro the physical sciences, and in reference books of formulas and tables. You will need to learn a number of Useful formulas involving spherical coordinates [for example, (4.7), (4.19), and (4.20) below; also see Chapter 10, Section 9 and Chapter 13, Section 7]. It is best to learn these formulas in the notation that is used in advanced books, We will need the volume clements in these two systems [and also the are length elements —see equations (4.18) and (4.19)}. To find the polar coordinate area clement in Figure 4.1, we drew a grid of curves + = const., @ = const. In three dimensions we want to draw a grid of surfaces. In cylindrical coordinates these surfaces are the cylinders r= const., the half-planes = const. (through the 2 axis), and the planes = const, [parallel to the (x, y) plane]. One of the elements formed by this grid of surfaces is sketched in Figure 44. From the geometry we see that three edges of the clement are dr, r d8, and dz, giving the volume element arr [a 4V =r dr d0 dz (cylindrical coordinates). ] Similarly for the spherical coordinate case, we draw the spheres r = const. the cones 4 =const., and the half-planes = const. The volume elements formed by this grid (Pigure 4.3) have edges dr, r dB, and r sin Odd; thus we get an aV =? sin 0 dr dO dp Jacobians, For polar, cylindrical, and spherical coordinates, we have seen how to find area and volume elements from the geometry. However, it is convenient to know an algebraic way of finding them which we can use for unfamiliar coordinate systems (Problems 16 and 17) or for any change of variables in a multiple integral (Problems 19 and 20). We state without proof (see advanced caleulus books) some theorems which tell us how to do this. First, in two dimensions, suppose x and y are given as functions of two new variables s and 1. ‘The Jacobian of x, y, with respect to 5, t, is the determinant in (4.8) below ; we also show abbreviations used for it (spherical coordinates). ax, y) _ [8 “ Ms, ‘ CHANGE OF VARIABLES IN INTEGRALS; JACOBIANS 221 set sd in the s £ system by the area clement -qhen the area clement dy dv is repla =. 7 the Jacobian in (4.8) 1 | sth absolute val of the Jacobian in notin “NEE he Jcbian af with respec the pr cites 72 nd th et that (it) snd our geomeirc method give the sume result (42) for the polar ordinate area clement. We have [lds dt & 2) ooo —rsina Cy a, A) ~ | ay sin ® cos 8 (49) ee arn elements dr 40 ain 4.2 TH at Jeans extends to more arabes, Als it no essary sr wth rectangular coordinates; let us state a more general theorem. ieee a [[frm ne) aasce ‘in some set of variables 1 0, m. Let r, 5, # be another set of variables, related to u, 2, ow by given equations weed P= =m 5 1. en if eu au oe far se atu, 2) a np MD a eo & os Fis the Jacobian of w, %, m with respect to r, 5, f, the triple integral in the new variables is 1) [fp una ea, where, of course, f and J must both be expressed in terms of, , f, and the limits ‘must be properly adjusted to correspond to the new variables. for cylindrical coordi We can use (4.12) to verify the volume element (4.6) f : (Problem 15) and the volume element (4.7) for spherical coordinates. Let us do thea 222_ MULTIPLE INTEGRALS calculation for spherical coordinates. From (4.5), we have ay Jae oe ae | | ae, ee» | Hew d_ | BA) _ |sinesing reosOsing —rsinO cos 0,8)” | a 20 a6 a é 4| =P sin O[ sin? (sin? 0 — cos? 8) — cos? sin? 0 — cos? 0] =P sind. ‘Thus the spherical coordinate volume element is dV’ = +? sin 0 dr dO dé as in (47). Example 2. Find the = coordinate of the centroid of a uniform solid cone (part of fone nappe) of height # equal to the radius of the base r. Also find the moment of inertia of the solid about its axis 4 If we take the cone as shown in Figure 4.6, its set. CHANGE. OF VARIABLES IN INTEGRALS; JACOBIANS _223 Using the value of Mf from (4.15), we write [in the usual form as a multiple of Af F soidsphereofratiusabouta diameter ample 3 Find the moment oferta of slid W ® If we use spherical coordinates, the equation of the sphere is r = a. Then the mass is ce farno [L[ psmearane 4.10) 2+ 2n = na'p. Go no one's sure! The moment of ner bau the «axis is Jorsoanen [ [ [tes on sn oar anes fr, using the value of Mf, we get equation in cylindrical coordinates is r =<, since at any height 2, the cross section is a circle of radius equal to the height. To find the mass we must integrate dM = pr dr dB dz, where p is the constant density. The limits of integration are 6: 012, Ow 2 Oh ‘Then We have + rane 48 anfoamol’ [ [wrdaacmp me ['S foo-[ Deane (4.15) For the moment of inertia about the 2 axis we have LL Coeaer [ee 7 Example 4. xis of the solid ellipsoid inside ‘We want to evaluate p[ffeoe mt refffernene where the triple integrals are over the volume of the ellipsoid. Make the change of variables x = ar’, y = by’, 2 = ca); then s'? +/+ 2? = 1, so in the primed variables we integrate over the volume of a sphere of radius 1. ‘Then M = pate [ff ay de = pd solume of phere frais Using (4.16), we have M = pabe- 4x 1? = $rpabe Similarly, we find wt [Joo ewer24_MULTWPLE INTEGRALS cus where the triple integral is over the volume of a sphere of radius 1. Now, by symmetry [Jeo=(ffea ffowafifen where? = 4 49"? 4 2 ind we are integrating over the volume inside the sphere [leer Ef [peso ara te ntlet [fJear ov [fora] nie oo, or, in terms of MI, T= 4Ma? +P), In order to find are lengths using spherical or cylindrical coordinates, we need th are length clement ds, Recall that we found the polar coordinate are length clement > (Figure 4.2) as the hypotenuse of the right triangle with sides dr and + d0. From Figure 41 you can see that ds can also be thought of as a diagonal of the arca clemen Similarly, in eylindrical and spherieal coordinates (see Figures 44 and 43), the an length element ds is a space diagonal of the volume element. In cylindrical coordinates (4.4), the sides of the volume element are dr, rd8, given by ¢ (Cylindrical coordinate [a : canner In spherical coordinates (4.5), the sides ofthe volume element are dvd, ‘so the arc length element is given by neiaimnoae so the are length element 1s 18) ds = dP + Pd ede G19) dt = dr? +7 dO? +7 sin? 84g? (spherical coordinates), ____| Itiis also convenient to be able to find arc lengths algel le to find arc lengths algebraically. Let us do this for polar coordinates; the same method applies in three dimensions. From (4.1) we have dy = cos @ dr —r sin 8 dO, dy im O dr +r cos 8 dO. ‘Squaring and adding these two equations, we get dS weds? + dy? (cos? @ + sin? 6) dr? + 0- dr dO + (sin? 8 + cos? 6) dB? sd a ae se GIANG oF ‘VAR as in (43), Using the same method for eylindrical and spherical coordinates (Problem 31) you can verify equations (4.18) and (4.19). Example 5. Express the velocity of a moving particle in spherical coordinates. If s represents the distance the particle has moved along some path, then dsidr is the velocity of the particle. Dividing (4.19) by d#?, we find for the square of the velocity We have just sn how 10 find the are length element ds in polar coordinates (or other systems) by calculating /as® + dy?, You might be tempted to try to find the ee gement by computing dx Uy, but you would discover that this does not work (we sreet use the Jacobian to get volume or area elements). You can see why by looking at Fygure 4 The element of area 7dr dO at the point (xy) i not the same as the einent of area de dy at that point. Then consider Figure +2; the element of are ds is the hypotenuse ofthe triangle with legs dr and r dO, and itis also the hypotenuse of the trangle wth legs dt and dy. Thus ds is the same rlement for both x. and 8 and this is why we can compute ds in polar coordinates by calculating v/dx? + dy?. These Comments hold for ather coordinate systems, too. We can always find ds by computing Yast + dy oc Jae + dy? + dey, but we cannot compute area or volume elements Srecy from the rectangular ones—we Must ove the Jacobian or ele. geomerical smethods nates) PROBLEMS, SECTION 4 1. Given a circle of radius find by integration using polar coordinates (a) its area: (6) the centroid of one quadrant; (©) the moment of inertia of «circular lamina about 2 diameters (@) its circumference. (@) the centroid of + quarter cite ae. @ “ind the moment of inertia of a circular disk (uniform density) about an axis through its center and perpendicular to the plane ofthe disk () Find the moment of inertia of a sold right circular eylinder (uniform density) about its (6) Do (a) using Problem le and the perpendicular axis theorem (Si ction 3, Example 19,226_MULTIPLE INTEGRALS _ cas 4. Fora sphere of radius 4 find by integration (2) its volume; (©) its surface areas (6) the centeoid of solid hemispheres (d) the centroid of the curved surface area of the hemispheres (e) the moment of inertia of the whole spherical shell (chat is, surface) about a. diameter (assuming constant density) (@) Write a triple integral in spherical coordinates for the volume inside the cone 7 + y/ and between the planes = I and = = 2, Evaluate the integral (6) Do (a) in cylindrical coordinates, 6. Find the mass of the solid in Problem 5 if the density is (x? +)? + work by doing the problem in both spherical and cylindrical coordinates +. Cheek your 7. (a) Using spherical coordinates, find the volume cut from the sphere + 0 and y > 1 or 2 lamina of density about the = axis in the shape of the area of Problem 6, find its moment of inertia For the spherical cap eut from the sphere x? + 92 4 (0,0, find (3) the volume (ee Problem 6 and Problem 3.12); (b) the = coordinate ofthe centroid (ase cylindrieal coordinates. bby a horizontal plane through Find the centroid of the area above y = x° and below y = ¢(¢ > 0), (2) Find the centroid of the area between the s axis and one arch of y = sin « (b) Find the volume formed if the area in (a) is rotated about the x aus. {) Find J, of a mass of constant density occupying the volume in (b). ‘Show that the = coordinate of the centroid of the volume inside the elliptic cone y BoRth O 90, this will come out negative, so in general we should say that the volume is JA (B x ©)]. Any side may be used as base, so, for example, B = (C x A) must also bee either plus or minus the volume. There are six such triple scalar products, all equal except for sign [or twelve if you count both the type A- (Bx C) and the type x © -A)] ‘To write the triple scalar product in component form we first write Bx C in determinant form [Chapter 3, equation (4.18)] ijk Gay Bxc=|a, B, 8, ac ¢ Now A: (Bx C) = AJB x ©), + 4,(B x ©), + 4B x C),, and this is exactly what we get by expanding, by clements of the first row, the determinant in (3,2); this determinant is then equal to A» (B x C) eee 8.2) pee Recalling that an interchange of rows changes the sign of a determinant, we can now easily write out the six (or twelve) products mentioned above with their proper signs You should convince yourself of, and then remember, the following facts: The order of the factors is all that counts; the dot and cross may be interchanged. If the order of factors is eyclic (one way around the circle in Figure 3.3), all such triple sealar products are equal, If you go the other way, you get another set all equal to each other and the negatives of the first set, For example, fh (B x ©) or (A xB) C. triple Vector Product This is written A x (B x ©). Before we try to evaluate it evcan make the following observations. B x C is perpendicular to the plane of Band TA x (Bx C) is perpendicular to the plane of A and (B x C); we are particularly fmveested in the fact that A x (B x C) is perpendicular to (B x C). Now (see Figure 44) any vector perpendicular to Bx G lies in the plane perpendicular to B x C, that is, the plane of B and C. Thus A x (B x C) is some vector in the me plane of Band C, and can be written as some combination BB+ OC, where « and 6 are scalars which we want to find. (See Chapter 3, Section 8, Problem 39.) One way to find a and 6 is to ‘write out A x (B x C) in component form. We can simplify this ‘work by choosing our coordinate system intelligently; recall that a yector equation is true independently of the coordinate system. Given the vectors A, B, C, we take the x axis along B, and the y B axis in the plane of B and C; then Bx C is in the = direction. FIGURE 3.4 “The vectors in component form relative to these axes are B= 3,5, Ga) CaCitGi, AS Ait A+ Ack Using (3.4) we find Bx C= Bei x (Ci + Ci) = BeCyi i) = BoC, 6s Ax Bx ©) = A,B,G,4 xk) + 4, BoC XW) = ABC —j) + 4, BoC). We should like to write A x (B x C) in (3.5) as a combination of B and C; we can do this by adding and subtracting 4, B, Cyi 66) Ax (BK C)= ~ ABC + Gi) + (AG + ACB ai Each of these expressions is something simple in terms of vectors ABy=A°B, 4,C,+4,C,= 4°, Cait Gi=G Bia B Using (3.7) in (8.6), we get 68) Ax (BX ©) =(A- OB—(A- BI G7)240_ VECTOR ANALYSIS ‘This important formula should be learned, but not memorized in terms of letters, because that is confusing when you want some other combination of the same letter. Learn instead the following three facts: (3.9) The value of a triple vector product is a linear combination of the two vectors in the parenthesis [B and C jn (3.8)]; the coefficient of each vector is the dot product of the other two; the middle vector in the triple product (B in (3.8)] always has the positive sign, ‘This method also covers triple vector products with the parenthesis first; by (3.9), the value of (B x C) x A is (A= BJC ~(A- C)B. This is correct since it is just the negative of what we had above for A x (B x C), ‘An Application of the Triple Scalar Product We have shown that the torque of a force F about an axis may be written as r x F in one special case, namely when rand F are in a plane perpendicular to the axis. Now let us consider the general case of finding the torque produced by a force F about any given line (axis) J. in Figure 3.5 Let & be a vector from some (that is, any) point on L to the line of action of F; let O be the tail of r. Then we define the torque about the point O to be x x F. Note that this cannot contradict our previous discussion of torque because we were talking about torque about a line before, and this definition is of torque about a point. However, we shall show how the two notions are connec~ ted, Also notice that r x F is not changed if the head of r is moved along F; for this just adds a multiple of F to r, and F x F = 0(see Problem 10) We shall now show that the torque of F about the line Z through O is a - (r x F), where n is a unit vector along L. To simplify our calculation, choose the positive = axis in the direction n; then n = k. Think of a door hinged to rotate about the = axis as in Figure 3.6, Let a force F be applied to it at the head of the vector r. We first find the torque of F about the = axis by elementary methods and definition, Break F into its components; the 2 component is parallel to the rotation axis and produces no torque FIGURE 3.5 Figure 2.7 FIGURE 2.6 TRIPLE PRODUCTS 241 J it (pulling stsight up or down on a door handle doesnot tend 10 open or close edu’). The and 7 componente cane soe beter if we drew them i the (, 7) ‘Figure 17; note thatthe wand ayes te rtated 90° clockwise fron their a amin order to compte this igure more easily with Figure 6) The trgue about vai produce by F and P, stP, YP, by the elementary defnion of torque ihe ant show tha hii he same av n= (@ F) or hee kG B). Using (3.2) we find oot > [rb ke@x B= =aF,~3Fy ‘To summarize: a In Figure 3.5, the torque of F about point 0 is r x F. The torque of F ‘about the line L through O is n-(r x F) where n is a unit vector along L. ‘This proof can casily be given without reference to a coordinate system. Let the symbols || and 1 stand for parallel and perpendicular to the given rotation axis n. Then any vector (F oF r, say) can be written as the sum of a vector parallel to the axis and a veetor perpendicular to the axis (that is, somewhere in the plane perpendicular to m): rar tey FL+F, Then the torque about O produced by F is FX P= (ry 46) KF +Fy) ar x Fler xP) try xP te x Fy ‘The last term is zero (cross product of parallel vectors). Also ry and Fy are parallel to 1; therefore their cross products with anything are in the plane perpendicular to n, and the dot produet of n with these is zero. Hence we have ae Gx P= ve. x FD, Now r, and F, are in a plane perpendicular to n; thus the torque about n produced by F, is (by Section 2) r, x F, . But since only the component of F perpendicular to 1 produces a torque about a, r, x F, is the total torque about n produced by F. The Vector torque r, x F, is in the -in direction since r, and Fare perpendicular to n; the dot product ofthis vector torque with the unit vector n gives a scalar torque of the same magnitude; the + sign indicates whether the torque is in the :i:n direction Example. If F =i + 3) ~ k acts at the point (1, 1, 1), find the torque of F about the line r= 34 + 2k + ( kyr. We first find the veetor torque about a point on the line, say the point (3, 0, 2). By 3.10) and Figure 3.5, this is r > F where r is the vector from the point about which weothe “ee 22_VECTOR ANALYSIS; cas a TRIPLE PRODUCTS 243 ‘want the torque, to the point at which F acts, that is, from (3,0, 2) to (I, 1, 1)5 then 9. The force F = 24 — | — Sk acts a the point (~5, 2,1). Find the torque due to F about she P=, 1 DG, 0,2) = (2, 1, 1), The vector torque is congin and aboot the line 2s = —4y = 2 i 5, ow that x Fee x F i k In Figure 3.5, let ¥ be another vector from 0 10 the fine of F. Sh 13a, aed unis of distance ditded. by force, ut his Tati televant for the vector proof) setlce thar moving the tail of + along m docs not change n= x F, Hint: ‘The triple ‘The torque about the line isn « (x x F) where n is unit vector along the fine, namely Stir product is not changed by interchanging the Joc and the cro ie a ia tee eee ee IIL Waite out dhe twelve tiple sae prodcts involving A, By and C and verify the fits stated n(x F) = Hi — 2) +8) - i 3) 7k) = 1 just above 3.3). Simplify (A ~B)? — [(A x B) xB] -A by wsing 39) Applications of the Triple Vector Product In Figure 3.8 eae ) z that the wiple salar product of (AX B), (B x C), and (Cx A), is equal to the (compare Figure 2.6, suppose the particle m is at rest on a rotating 13, Prove that the i =D, and oT Mure of the triple sealar product of A, By and C. Hint: First lt (Bx C)=D, a rigid body (for example, the earth). ‘Then the angular momentum of cua of the rl ‘m about point O is, by definition, L =r x (mv) = mr xv. In the anaes discussion of Figure 26, we showed that v=o xr. Thus, Prove the Jacobi identity: A x (BX C) +B x (Cx A) + Cc (AB) = 0. Hints Espand L = mr x (@ x 1). See Problem 16 and also Chapter 10, Section cach triple product asin equations (38) and (39), De eee 15. In the figure wy is @ unit vector in the direction of an ‘As another example, it is shown in mechanics that the centripetal incident ay of light, and uy and u, are unk vectors in acceleration of m in Figure 3.8 is a = @ x (@ x 1). See Problem 17. the directions of the reflected and refracted rays. fw is 2 anes unit vector normal to the surface 4B, the lows of optics Say that 0; 05 and m, sin 0, =m, sin Op, where m, and th are constants (indices of fefraction), Write these Taws PROBLEMS, SECTION 3 in vector form (using dot or cross products). 1 Wf As 2) jk, BH 23 4k CH 54k, find (ABYC, ABC), (Ax -€ ABO). (AXB) XC, Ax XC) For Problems 210 6, given A =i + j — 2k, B= 2i—j + 3k, C 2 Find the work done by the fore B ating on an object which undergoes the displacement C 3. Find the total work done by forces A and B if the object undergoes the displacement C 16 In the discussion of Figure 38, we found for the angular momentum, the formula Hint: Can you ad the to forces firs = nr x (0). Use (39) 10 expand shen product. Ii perpendicular too (ha . lane), show that you obtain the elementary formula 4. Let 0 be the til of B and let A be a force acting at the head of B. Find the torque of A eee eee about 0; about a tine through © perpendicular to the plane of A and B, aboot a line ‘angular momentum = mer. ‘through © parallel to C: 17, Expand the tiple product for a = «x (@ x £) given in che discussion of Figure 3.8.1 ris rpendicular to © (Problem 16), show that a = —0%r, and so find the ckmentary result 5. Let A and © be drawn from a common origin and let C rotate about A with an angular ae and of magnitude 0*/. ‘elocity of 2 radjse. Find the velocity of the head of © that the acceleration is toward the center ofthe circle and of mag Problem 5, draw B with its 18, Two moving charged particles exert forces on each other because eich one creates a 6 In Problem 5, draw B with its tila the head of A. If the Figure is rotating as in Problem magne field in which the other moves {eee Problem 4.6). ‘These two forces are pro= 5, find the velocity of the head of B. With the stme diagram, let B be 2 fore; find the Eris joining the part torque of B about the head of C, and about the line C. " 7 portional tov % fv % #] and v2 x [¥1 %(—0)} where is the vector joining the partie . ‘es. By using (33), show that these forces ae equal and opposite (Newton's third 1a") i 7. A force F = 21 ~ 3) + k acts at the point (1,5, 2). Bind the torque due to F and only it x (v; % ¥2) =O, Compare Problem 14 eeeraeeee: 19, The force F=i + 3) + 2k acts at the point (I 1,0 (8) about the y axis; (6) about the line 2/2 = y/t (2, A vector force with components (1, 2, 3) acts at the point (3, 2,1). Find the vector torque about the origin due to this force and find the torque about exch coordinate axis, (@) Find the torque of the force about the point ( from (2, —1, 8) (1, 1, D. (b) Find the torque ofthe force about the line x Note thatthe line goes through the point (2, =1,5). Careful! The vector + goes 5k 4 (ij + he24 VECTOR. AN, os 20. The tne B= 21 shai the po Fhe fre B= 21 Se ats a the pia (3, 10 Find the trate oF abot ech othe t= Gh) +0) ~ ay, rote a ae owe 4. DIFFERENTIATION OF VECTORS WASiAL 4, thls, where i i k functions of 1, then we define the deriv: dA/dt by the equation ee eee an Oh the ty | i eae rarneay | - — ee = Ths the derivative of vector A means a vector whose components are the derivatives of the components of A. Example 1. Let (sy, 2) be the coordinates of a moving particle at time 1; then x Ys £ are functions of ¢. The vector displacemer i wis ae hhe vector displacement of the particle from the origin at | | (42) ix + iy + ke, where is «vector fom the origin tothe parti heer iain to the parce atte 1. We say that isthe esion ves or vetr coordinate ofthe parle ‘The comonene of eve of the particle at time ¢ are dx/dt, dy/dt, dz/dt so the velocity vector is ° | 4 + agg | (4.3) vata a ‘The acceleration vector is Co) ie The product of a scalar and a vec differentiated by the ordinary calcula of caution: The order of the factors Drove these facts (4.5) by writing out tor and the dot and cross products of vectors are s rules for differentiating a product, with one word must be Kept in a cross product. You can easily components (Problem I) and using (4.1), de deny ian as) ! ‘TION OF VECTORS _148 et second term in (d/di)(A~B) can be written B-dA/dr if you like since A-B=B-A. But the corresponding term in. (d/di)(A x B) must nor be rurned ound unless you put a minus sign in front of it since Ax B= ~B x A. Example 2. Consider the motion of a particle in a circle at constant speed. We can then write +o) ro or co) A we a0 or viaad. (a Also differentiating r - v = 0, we get atvey reas ae? 48) r ‘The first of equations (4.7) says that r is perpendicular to v; the second says that a is perpendicular to v. Therefore a and rare either parallel or antiparalel (since the motion is in a plane) and the angle @ between a and r is either 0° or 180°. From (4-8) and the definition of scalar product, we have 49) reas [el|al cos 6 = —0% ‘Thus we see that cos 6 <0 so @ = 180°. ‘Then from (4.9) we get rilal(-1)= 2? or (4.10) We have just given a vector proof that for motion in a circle at constant speed the acceleration is toward the center of the circle and of magnitude «/r So far we have written vectors only in terms of their rectangular components using the unit vectors i, j, k. It is often convenient to use other coordinate systems, for cxample polar coordinates in two dimensions and spherical ot cylindrical coordinates in three dimensions (sce Chapter 5, Section 4, and Chapter 10, Sections 6 to 9). We shall consider using vectors in various coordinate systems in detait in Chapter 10, but it will be useful to discuss briefly here the use of plane polar coordinates. In Figure 41, think of starting at the point (x,») or (*, 0) and moving along the line @ = const. in the direction of increasing r. We call this the “r direction”; we draw a unit vector (that is, 4 vector of length 1) in this direction and label ite,. Similarly, think of moving along the circle r= const. in the direction of increasing 8. We call this the “@ direction”; we ddraw a unit vector tangent to the circle and label it e». These two vectors e, and ey are the polar coordinate unit basis vectors just as i and j are the rectangular unit basis vectors. (Other books may use a, and ay, u, and ug, mand I, etc; there is no polari oan | | FIGURE 4.1 FIGURE 42 coordinate unit basis vector notation which is as standard as i and j for rectangular coordinates.) We can now write any given vector in terms of its components in the directions e, and ey (by finding its projections in these directions). ‘Ther cation here, however. In rectangular coordinates, the unit vectors and j are eonsien. fy magnitude and direction. The polar coordinate it vect ea but thir directions change from point vo point (Figure 42) This edeaine derivative of a vector written in polar coordinates, we must differentiate the unit vectors 48 well as the components [compare (4.1) where we differentiate the components only] FIGURE 43 One straightforward way to do this is to ex and j. From Figure 4,3, we see that the x ‘Thus we have spress the unit vectors e, and ey in terms of i and y components of e, are cos @ and sin 0. = [oe cr cea TS) Similarly (Problem 7) we find &=—isinOtjos@ ~*«~d: ee aac Differentiating e, and ey with respect to 1, we get (4.12) de, (4.13) ds dey —___ ne — ‘We can now use (4.13) in calculating the derivative of any vector which is written in terms of its polar components. DIFFERENTIATION OF VECTORS 267 gxample 3. Given A= 4,6, + dgey, where 4, and dy are functions of 1, find anit. We get aA, ik ae at Using (4.13), we find aA. 4A, 10 dy do PR Beg eA, + ep 8 Ae aa a a We can find higher-order derivatives if we like by differentiating again using (4.13) cach time to evaluate the derivatives of e, and ey PROBLEMS, SECTION 4 1. Verify equations (45) by writing out the components 2. Let the position vector (with its tail at the origin) of a moving particle be eli) = Pi — 25 +(e + 2k, where # represents time. {a) Show that the particle goes through the point (4, —4, 8). At what time does it do this? (@) Find the velocity vector and the speed of the particle at time 1; at the time when it passes through the point (4, —4, 8) (6) Find the equations of the line tangent to the curve described by the particle and the plane normal to this curve, at the point (4, —4, 8). 3. As in Problem 2, if the position vector of a particle is r = (44 3h +) ~ Sik, at what time docs it pass through the point (1, ~1, 5)? Find its velocity at this time. Find the ‘equations of the line tangent to its path and the plane normal to the path at (I, ~1, 5). 4. Let es x(t) bea vector whose length is always I (it may vary in direction). Prove that either + is constant vector oF dr/d is perpendicular tor, Hint: Differentiate F + ¥. "The postion of a particle at time # is given by r= ios ¢+ j sin t+ kr, Show that both the speed and the magnitude ofthe acceleration are constant, Describe the motion. 6. ‘The farce acting on a moving charged particle in a magnetic field B is F = a(w x B) where is the electric charge of the particle, and visits velocity. Suppose that a particle moves in fhe (2,3) plane with a uniform B in the 2 direction, Assuming Newton's second law, im dvjdi = F, show that the force and velocity are perpendicular and that both have con- stant magnitude, Hint» Find (d/dt\v *v) 7. Sketch a figure and verify equation (4.12). 8. In polar coordinates, the position vector of a particle is Using (4.13), find the velocity and acceleration of the particle 9, The angular momentum of « particle m is defined by 3. Show that sme X (ded) (see end of Section foneees hala 248 VECTOR ANALYSIS 10, LFV(9 isa vector funetion oft find the indetnite integral f(x Ba 5. FIELDS Many physical quantities have diferent values at diferent points in space. For ble, the temperature in a room i diferent at diferent points high new a reget te near an open window, and ao on. The electric field around a point charge Siaee non the charge and decrees as we go ava frm the care. Sia the paiioa fre acting on atte depends on is distance rm the eh, The veo a of water in a stream is lage in rapids and in narrow channels and small ores Rigo and where the stream is wide. In all these examples there is a Particular region of sj = whichis of interest forthe problem at hand, at every point of thi tpion ae ohenca quantity ha a value. The term field is used to mean both the rexom and the coe at the physi quay inthe eon fr expe eric Rll, estan Feld) If he physical quantiy isa scalar (for example temperature), we spent of clay ele. Leah auantty is «vector (for example, electric field, fore, or velocity), we speckota tose {feld. Note agsin a point which we discussed in “endpoint problems” in Chapter 4 Section 10: Physical problems are often restricted to cerain regions af space sed ogy mathematics must take account of this. eer FIGURE 5.1 A simple example of a scalar field is the gravitational potential energy near the earth its value is = mes at every point of height 2 above some arbitrary reference level Cwhich we take as the (x, y) plane). Suppose that on a hill (Figure 5.1) we mark a series of curves each corresponding to some value of = (curves of constant elevation often called contour tines or level lines). Any curve or surface on which a potential is constant is called an equipotential. ‘Thus these level lines are cquipotentals of the gravitational field since along any one curve the value of the gravitational potential ‘energy mez is constant. The horizontal planes which intersect the hill in these curves ae equipotential surfaces (or level surfaces) of the gravitational field As another example, let us ask forthe equipotential surfaces in the eld of an electric point charge 4. The potential is V = 9 - 10°g/r (in mks units) at a point which is at distance r from the charge. The potential V is constant if r is constant; that is, the ‘quipotentils of this electric field are spheres with centers at the charge. Similarly we could imagine drawing a set of surfaces (probably very iregular) in a room so that at | pees DIRECTIONAL DERIVATIVE; GRADIENT 209 every point of single surface the temperature would be constant, These surfaces Mould be like equipotentials; they are called isothermals when the constant quantity is the temperature. 6, DIRECTIONAL DERIVATIVE; GRADIENT that we know the temperature T(x, y, 2) at every point of a room, say, or of a Supper Scaring aa given point we could ask for the rate of change of the Temperature with distance (in degrees per centimeter) as we move away from the werting point. ‘The chances are that the temperature increases in some directions and Seereases in other directions, and that it increases more rapidly in some directions than thers. Thus the rate of change of temperature with distance depends upon the direction fh which we move; consequently it is called a directional derivative. In symbols, we stant to find the limiting value of A7/As where As isan element of distance (are length) jp given direction, and AT is the corresponding change in temperature; we write the directional derivative as dds. We could also ask for the direction in which dT/ds has jis largest value; this is physically che direction from which heat flows (that is, heat flows from hot to cold, in the opposite direction from the maximum rate of temperature increase, ‘Before we discuss how to calculate directional derivatives, consider another example. Suppose we are standing at a point on the side ofthe hill of Figure 5.1 (not at the top), and ask the question “In what direction does the hill slope downward most steeply from this point?” This is the direction in which you would start co slide if you lost your footing; it isthe direction most people would probably call “straight” down. We ‘want to make this vague idea mote precise. Suppose we move a small distance As on the hill; the vertical distance Az which we have gone may be positive (uphill) ot negative (downhill) or zero (around the hill). Then Az/As and its limit ds/ds depend upon the direction in which we go; d2/ds is a directional derivative. ‘The direction of steepest slope is the direction in which d2/ds has its largest absolute value. Notice that since the gravitational potential energy of a mass m is V= mgs, maximizing dejds is the same as maximizing dVjds, where the equipotentils on the hill are Vis, 9) = meets, 9) = const [Let us now state and solve the general problem of finding a directional derivative. We are given a scalar field, that is, a function 44s, »5 2) Lor ots, ») in a two-variable problem; the following discussiore applies to two-variable problems if we simply drop terms and equations containing 2]. We want to find d¢/ds, the rate of change of > with distance, at a given point (xo, Yor o) and in a given direction. Let w= ia + jb + ke be 4 unit vector in the given direction, In Figure 6.1, we start at (ro, Jos 20) and go a distance 5 (520) in the direction u to the point (x, 5 2); the vector joining these points is us since w is a unit vector. Then, (2.95 2) — (#05 oy 20) = us = (ai + Bj + chs or = ty tas, Za 1) yao t by oko) ae FIGURE 6.1— 280_VECTOR ANALYSI sirection w [se Chapter equation G8)] mith the dance Gane oye (x,y, 2}, then } becomes a function of just the one variable That is, along ‘he measured from (x9, ¥o, 29). Since @ depends on s alone, we can find ddjds: me 4b Ob de Ody Ob de ds” ax ds Oy ds” On ds 4 84 ar ay * Ge © (6.2) This is the dot product of uw with th i Ob/éa) + OOK n the vector i(2p/ex) + j(Ob/ay) + k(OH/2=). ‘This ecto is ealled the gradient of and is writen grad @ or Vé (read “del 4"). By Pa Soe (63) Vp = grad 6 =i 4 5 om end oni Prien ‘Then we can write (6.2) as (64) SHVeru (directional derivative), i Example 1. Find the directional derivative of direction A = 2i ~ 2) +k. Here u is a unit vector obtained by dividing A by |A|. Then we have ay tae at (1,2, 1) in the u=4Qi- 3 +h) Using (6.3) we get 2 28 Fi Bans ait vo os ay the Vo at the point (1, 2, —1) ‘Then from (6.4) we find Baa, nev F.2D =V6-u= The gradient of a function has useful geometrical and physical meanings which we shall now investigate. From (6.4), using the definition of a dot product, and the fact DIRECTIONAL DERIVATIVE; GRADIENT 251 shat [| = 1, we have a, 6» ib where 0 is the angle between w and the vector Vd. ‘Thus dp/ds is the projection of Vd on the direction 2 (Figure 6.2). We find the largest value of déids eamely |V6)) if we go in the direction of Vo {that is, O= 0 in Figure 6.2). If we go in the site direction (chat is, @= 180° in Figure 6.2) tre find the largest rate of decrease of , namely doids = -1V0| Vo! cos 8 FIGURE 62 Example 2. Suppose that the temperature Tat the point (x, 2) is given by Tas! — 9 +92 +273; in which direction is the temperature increasing most rap- idly a (—1,2,3),and whatis the rate? Here V7'= (2x + y2)i + (—2y + x2)) + 49k = 7 Dk at (1, 2, 3), and the increase in temperature is fastest in the direction Gf this vector. The rate of increase is dTjds = |VT| = /16 +49 +4 = V/69. We can tlso say that the temperature is decreasing most rapidly in the direction — V7’; in this direction, d7/ds = —/69. Heat flows in the direction —WT7 (that is, from hot to cold). Next suppose u is tangent to the surface q = const. at the point P(t9,.¥o. 0) (Figure 6.3). We want to show that dd/ds in the direction uw is then equal to zero, Consider AG/A: for paths PA, PB, Pt approaching the tangent u. Since g= const, on Pos the surface, and P, A, B, C, etc. are all on the surface, AG =0, and Ad/As=0 for such paths, Bur déjds in the tangent direction is the limit of AG/As as As—» 0 (that is, as PA, PB, ete., approach, 1), 50 d@lds in the direction ui is zero also. Then for 1 along the tangent to d = const,, Vo « u= 0; this means that Vo is perpendicular to 1. Since this is true for any w tangent to the surface at the point (xy, Joy os then at that point FIGURE 63, ‘The vector Vo is perpendicular to the surface g = const. Since |Vq| is the value of the directional derivative in the direction normal (that is, perpendicular) to the surface, it is often called the normal derivative and written 1VO| = didn. ‘We now see that the direction of largest rate of change of a given function with distance is perpendicular to the equipotentials (or level lines) ¢ = const. In the tem= perature problem, the direction of maximum dT/ds is then perpendicular to the iso- thermals, At any point this is the direction of W7' and is called the direction of the temperature gradient, In the problem of the hill, the direction of steepest slope at any point is perpendicular to the leve! lines, that is, along Vz or VV’252_VECTOR ANALYSIS; Example 3. Given the surface xy and normal line at (1, —2, 3). ‘This is a level surface’ of the fun direction of the gradient Ver = 3°93 i+ 2 ys] + ey 7 = 364 +4 at (1, ~2,3), A simpler vector in the same direction is 91 +L Then (Ge Chapter the equation of the tangent plane is ‘ ale He = 1) = 3 +2) 42-3) and the equations of the normal line are 66) In (6.3) we have written Vo in terms of its rectangular components. It will be useful to write it in plane polar coordinates also (sce Chapter 10, Section 9 for other coordi- nate systems), What we want are the components of V¢ in the directions e, and e, According to (6.4), the component of Vb , ent of Vp in any direction w is the directional derivative dpjds in that direction. ‘The clement of are length ds in the r direction is dr, so the directional derivative in the r direction is dé/dr (8 constant) which we write as 0¢/¢r In the 0 direction, the element of arc length is r d0 (Chapter 5, (1/29/08). Thus we have (6.7) PROBLEMS, SECTION 6 1. Find the gradient of (1.2, —D, ‘Starting from the point (1, 1), in what direction does the function =x decrease most rapidly? 3. Find the derivative of ay? + ye at (1,1, 2) in the direction of the vector 24 —j + 2k 4. Find the derivative of ze" cos y at (1, 0, m/3) in the direction of the vector i + 2j 5. Find the gradient of @ = = sin y ~ rz at the point (2, x/2, —1). Starting at this point, in what direction is @ decreasing ‘most rapidly? Find the derivative of in the direct +3 ° ome an 6. Find a vector normal to the surfice 7 +y?—2=0 at the point (3, 4,25). Find the ‘equations ofthe tangent plane and normal line to the surface at that point Find the direction of the line normal to the surface xy +y?s +22 +1=0 ‘e+ str +1 = 0 at the point (Q, 2, =1), Write the equations ofthe tangent plane and normal line at this point, 2 cng 2, find the equations of the tangent plane jon w= s*y%s, so the normal dircetion is the Sector soe directional derivative in the @ direction is d/(r d®) (r constant) which as a 4 15, 16. n. DIRECTIONAL DERIVATIVE; GRADIENT_253 (a) Find the directional derivative of = 22 + sin y — xs in the direction i + 2) — 2k at the point (I, x/2, —3). (b) Find the equation of the tangent plane and the equations of the normal line to @ = § at the point (1, x/2, —3) (@) Given § =? — ys, find V9 at (1,1, D. (b) Find the directional derivative of @ at (1,1, 1) in the dtection i~ 2) + k (©) Find the equations ofthe normal line tothe surface 3? —y2z = O a (I, Is D. ‘Suppose that the temperature in the (x, ) plane is given by T= sy —s: Sketch 2 few jeothermal curves, corresponding, for instance, to T= 0, 1, 2, ~1, ~2. Find the direction jn which the temperature changes most rapidly with distance from the point (1, 1), and the tnaximumn rate of change, Find the directional derivative of Tat (1, 1) in the direction of the vector 314} Heat flows in the direction —VT" (perpendicular to the isothermals) ‘Sketch a few curves along which heat would flow. (@) Given g = 2? 9%, sketch on one graph the curves @=4, @= 1, O=0, >= 1, G4. If G is the electrostatic potential, the curves = const, are equiporentials, and the electric field is given by E= —V@. If ¢ is temperature, the curves @ = const fare isothermals and V@ is the temperature gradient; heat flows in the direction —V¢. (t) Find and draw on your sketch the vectors —Vo at the points (x3) =(E1, £0, (0. +2), (42,0). Then, remembering that V@ is perpendicular to @ = const. sketch, ‘without computation, several curves along which heat would flow (see (3) For Problem 11, (a) Find the magnitude and direction of the electric field at (2,1) (b) Find the direction in which the temperature is decreasing most rapidly at (—3, 2) (6) Find the rate of change of temperature with distance at (1, 2) in the direstion 31 —j Let @ =< cosy. Let g represent either temperature or electrostatic potential, Refer to Problem 11 for definitions and find (@) ‘The direction in which the temperature is increasing most rapidly at (1, —m/4) and the magnitude of the rate of increase, a (b) The rate of change of temperature with distance at (0, 1/3) inthe direetion i + iV/3. (6) The direction and magnitude ofthe electric field at (0m). (@) The magnitude of the electric field at x = —1 any y (a) Suppose that a hill (as in Fig. 5.1) has the equation z= 32— 2? —4y7, where ‘z= height measured from some reference level (in hundreds of feet). Sketch a contour imap (that is, draw on one graph a set of curves «= const); use the contours < = 32, 19, 12, 7,0 (b) If you stare at the point (352) and in the direction i+ j, are you going uphill oF downhill, and how fast? Repeat Problem 4b for the following points and directions. @) G2, 445 () (3,0), 4143) (2, ~H4i (@) (-4, =D, 41-3) Show by the Lagrange multiplier method that the maximum value of dds is |V¢p|. That is, maximize d@ds given by (6.3) subject to the condition a? + 6? +c = 1. You should get two values (+) for the Lagrange multiplier 2, nd two values (maximum and minimum) for dgjds. Which is the maximum and which is the minimum? Find Vr, where r= JF y%, using (6.7) and also using (6.3). Show that your results are the same by using (411) and (4.12)254 VECTOR ANALYSIS As in Problem 17, find equivalent 1 Ve Wy the following gradients in two ways and show that your answers are 2. Wi?) 7. SOME OTHER EXPRESSIONS INVOLVING V If we write Vo as [i(@/éx) + j(C/2y) + K(0/02)], we can then call the bracket V. By itself V has no meaning (just as d/dx alone has no meaning; we must put some function after it to be differentiated). However, it is useful to use V much as we use didx to indicate a certain operation. We call V a vector operator and write oa a 7.1) Vail+joeec. ay EtIBt*; Ic is more complicated than d/dr (which is a scalar operator) because V has vector properties too. So far we have considered V@ where ¢ is a scalar; we next want to consider whether V can operate on a vector. Suppose V(x,.y, 2) is a vector function, that is, the three components V,, V,, V- of V are functions of x, 9, 2 VO, 94 2) = C0 95 8) + IO 9, 2) + KV Ge, 9, 2) (The subscripts mean components, not partial derivatives.) Physically, V represents vector field (for example, the electric field about a point charge). At each point of space there is a vector V, but the magnitude and direction of V may vary from point to point We can form two useful combinations of V and V, We define the divergence of V abbreviated div V or V + V, by (7.2): = nin yn HH ] (7.2) V-V=divV Ge oy tae | We define the curl of V, written W x V, by (7.3) (3) VxVeaniv WW.) (av, _ wW, (22) (Ha _ Ve HZ (Se a) SOME. OTHER, ‘You should study these expressions to sce how we are using V a5 “almost” a vector. tions of divergence and curl are the partial derivative expressions, of course Te th sim of he frmus (2) and (2.2) 0 toe for A= Band 8 B Taps us to remember V- Vand Vx V, But you must remember to put the partial erative “components” of V defore the components of V in each term [for example, Grevaluating the determinant in (7.3)]. Note that VV is a scalar and Wx V is a in gor compare A B and A x B). We shall discuss later the meaning and some of the Ipplcations of the divergence and the curl of a vector function "The quantity Vo in (6,3) is a vector function; we can then let V = V6 in (7.2) and find ¥ “V6 = div grad 6. This is a very important expression called the Laplacian of itis usually written as V24. From (6,3) and (7.2), we have ‘The Laplacian is part of several important equations in mathematical physics Vg = is Laplace’s equation. is the wave equation. 1 VO ve is the diffusion equation or equation of heat conduction. ‘These equations arise in numerous problems in heat, hydrodynamics, electricity and magnetism, aerodynamics, elasticity, optics, ete.; we shall discuss solving such equations in Chapter 13. There are many other more complicated expressions involving V and one or more scalar or vector functions, which arise in various applications of vector analysis. For reference we list a table of such expressions at the end of the chapter. Notice that these are of two kinds: (1) expressions involving wo applications of V such as V-- Vp = V8; (2) combinations bf V with two functions (vectors or scalars) such as ¥ x (GV). We can verily the given values of any of these expressions simply by writing ‘out components. However, itis usually simpler to use the same formulas we would use if V were an ordinary vector, being careful to remember that V is also a differential operator. Let us illustrate this method by evaluating V x (V x V), We use the formula (3.8) for A x (B x C) being careful to write both V's before the vector function V which they must differentiate. Then we get vx(VxV) WV) (VV = Hv «Vy — VV286 VECTOR ANALYSIS: ‘This is a vector as it should be; the Laplacian of a vectors V2V, simply means a vector whose components are V2V,, V°V,, VV, As a second example let us find V - (GV), where @ is a scalar function and V is « vector function. Here we must differentiate a product, so our result will contain tws terms. We could write these as (75) V (OV) = Ve (OV) + Vy * (GV), where the subscripts on V indicate which function is to be differentiated. Since @ is a scalar, it can be moved past the dot. Then Vo (OV) = (Vo) V=V-(¥O), where we have removed the subscript in the last step since V no longer appears after V Actually you may see in books (V¢) - V meaning that only the ¢ is to be differentiated, but i is clearer to write it as V « (V). [Be careful with (Vd) x V, however: assuming that this means that only g is to be differentiated, the clear ‘way to write it» ~V x (V@); note the minus sign.] In the second term of (7.5), @ is a scalar and is nor To find the constants, recall that the magnitude of the force on m at the surface of the earth is me 9, GREEN'S THEOREM IN THE PLANE ‘The fundamental theorem of calculus says that the integral of the derivative of a function is the function, or more precisely FIGURE 9.1 Next evaluate § Q(x, 9) dy around C in the counterclockwise direction, that is, so that Lis always to our left as we go around C. (The symbol § means an integral around a Gesed curve back to the starting point.) Along the horizontal sides of A in Figure 9.1, Jp 20, so the integrals are zero. Along the right side, x =, and y goes from ¢ to d Along the left side, x = a, and y goes from d toc. Thus, @3) bana [aw way | Qa») dy [ LOU, ») ~ Qa, wy) dy. ‘Then by (9.2) and (9.3) we have 04) [ao fou. Similarly, you can show that (Problem 9.1) 0) ff Combi Pas ing (9.4) and (9.5), we have 06 rasan {[(2 2) act ne oy | Ree 10 Seo We have proved (9.6) for a rectangle. We next consider an area A in the (x, ») plane bounded by'a closed eurve; we asume that fan be regarded asthe limit of a st of small rectangles (See Figure 9.2) For each rectangle (2.0) is true, The sum ofall the double integrals in (96) overall the rectangles in Figure 9.2 is the double integral over the whole area. ‘The line integrals over interior common boundaries cancel each other; for example, in Figure 9.2, the integrals trvund rectangles I and 2 are in opposite directons along their common boundary DE. What i lef 5 just ae ee ' the line integral around (Thus (0.6) is valid for an area Z (02) i EI ay gy | [ I 1. ay — f 1Q(b, ») ~ Qa, 99] dy. like 4 in Figure 9.2. Let us use the notation 2A (read 2 he Os 7 boundary of A) to mean the curve C in Figure 9.2. We are going to consider some useful generalizations of this theorem to two and three dimensions, Let P(x, ») and Q(x, y) be functions with continuous first partial derivatives. We consider the double integral of (0/0x)Q(x, y) over the rectangle 4 in Figure 9.1 and show that the double integral is equal to a line integral around the boundary of the rectangle. First write the double integral as an iterated integral and perform the © integration using (9.1) to get Figure 92264_VECTOR ANALYSIS a, Then we have Greens theorem in the plane 8Q oP’ WG-5)« ‘The line integral is in the counterclockwise direction; P = Pls, y) and Q = Qtx, 5) have continuous first partial derivatives at every point of ] § Pdr+Qdy Using Green's theorem we can evaluate either a line integral around a closed path or a double integral over the area inclosed, whichever is easier to do. Example 1. In Example 1, Section 8, we found the line integral (8.2) along several paths (Figure 8.2). Sup- a0 pose we want the line integral in Figure 8.2 around the fa closed loop (Figure 9.3) from (0, 0) to (2, 1) and back 7 as shown. From Section 8, Example 1, this is the work 4, — done along path 2 minus the work done along path 3 “°°! (ince we are now going in the opposite direction); we eee find W, ~ W = 3-4 = —1. Let us evaluate this using Green's theorem, From (8.2) and (9.7) we have Wwe £ (ayy | dr dh $: 5 4 | ty as before Example 2. In Section 8, we discussed conservative forces for which work done is independent of the path. By Green’s theorem (9.7), the work done by a force F around a closed path in the (x, y) plane is {yraesrane f(D) w If (2F,/0x) — (@Fy/0y) = 0 (note that this is the = component of curl F = 0) then Hf around any closed path is zero, which means that the work from one point to another is independent ofthe path (also see Section 11) re GREEN'S THEOREM IN THE PLANE 269 ‘There are two useful ways to apply (9.7) to the integration of vector functions. First et 03) 2 ‘Then where V=i¥, +iV, ) by (7.2) with V, FIGURE 9.4 10) deside+jdy —(angent) is a tangent vector, and the vector 1) nds (outward normal), nis a unit vector and ds = fd? + dy, idy —jds is a normal vector (perpendicular to the tangent) pointing out of area A. Using (9.11) and (9.8), we can write (9.12), Px + Ody = —V, det Vy dy = (WV, + iV) (idy 5 ds) =Vinds ‘Then substitute (9.9) and (9.12) into (9.7) to get 0.3) [foe varar= [vena This is the divergence theorem in two dimensions. Tt can be extended to three dimen sions (also see Section 10). Let represent a volume; then ér (read boundary of 2) ‘means the closed surface area of 1. Let dr mean a volume clement and let d mean an clement of surface area, At exch point of the surface, let be a unit veetor perpendicu270_VECTOR ANALYSIS. lar to the surface and pointing outward. Then the divergence theorem in three dimen. sions says (also see Section 10) (4) fff div V dr ff Venda. Divergence theorem ‘To see another application of (9.7) to vector functions, we let (13) Q=0, P=V,, where Vail, +i¥, Then (0.19) QP hy Me out Vy k de Gy oe Gy Oi) es Beymer ba nV oe ow [fests ay G.v a ‘This is Stokes’ theorem in two dimensions. It can be extended to three dimensions (Section 11). Let @ be an open surface (for example, a hemisphere); then Go means the curve bounding the surface (Figure 9.5). Let n be a unit vector normal to the surface. Then Stokes’ theorem in three dimensions is (also see Section 11) FIGURE 0.5 oan ffieato-naee [vou Stokes’ theorem, | | ‘The direction of integration for the line integral is as shown in Figure 9.5 (see also Section 11), PROBLEMS, SECTION 9 1. Derive (9.5) by following a method similar to that used in obtaining (94). In Problems 2 so 5 use Green's theorem [formula (9.7)] to evaluate the given integrals, 2, § de dy — ay de around the square with vertices (0, 2) (2,0), (~2, 0), and (0, —2), THE DIVERGENCE AND THE DIVERGENCE THEOREM 771 3 foxy debe dy, where C is as sketched, 4. feet cosy ds — e* sin y dy, where C is the broken line from 4 = {in 3,0)t0.D ~ (0,1) and then from D to B = (—In2,0) Hint; Apply Green's theorem to the integral around the closed curve ADBA. 5. folyet= 1) dr +o dy, where © is the semicircle through (0, ~10), (10, 0) and (0, 10). (Compare Problem 4) 6 For a simple closed curve C in the plane show by Green's theorem that the ara inclosed is Levan 7. Use Problem 6 to show thatthe ares inside the ellipse + is A= nab, 4 208 8, y = sin 0,0<0.< 2n, 8. Use Problem 6 to find the ates inside the curve x22 + 8 = 4, 9 Apply Green's theorem with P= 0, O = 4x? to the triangle with vertices (0, 0), (0,3, {G,0). You will then have {fx de dy over the wrangle expresed as a very simple line integral. Use this to locate the centroid of the triangle. (Compare Chapter 5, Section 3.) Exaluate each of the following integrals inthe easiest way you can, 10. § (2y dr ~ 3x dy) around the square bounded by x = 3, x =5,y= Vandy = 3, UL fee sin s~y) ds + (2 — 9%) dy, where C i the wianle inthe (9) plane with vertices (0,.0),0, 1, and 2,0) 1. [G2 27) de + Qay +3) dy along the x axis from (0,0) to (/3,0) and then along a tireular ate fom (/5, 0) (1,2) 10. THE DIVERGENCE AND THE DIVERGENCE THEOREM ‘We have defined (in Section 7) the divergence of a vector function V(x, y, 2) as as 5 ye ox oy oz (io. divV= v We now want to investigate the meaning and use of the divergence in physical applic tions. Consider a region in which water is flowing. We can imagine drawing at every point 4 vector v equal tothe velocity of the water at that point. The vector function v then represents a vector field. ‘The curves tangent to v are called stream lines. We could inAECHOR ANALYSIS maw = Ch the same way discuss the flow of a as, of heat, of electricity, oF of particles (sy fron: radioactive source). We ate going to show that if v represents the velocity of flow any of these things, then div w iy related to the amount of the substance which flo, ‘out of a given volume. This could be different from zeto either because of a change density (more air flows out than in as a room is heated) oF because there is a souree sink in the volume (alpha particles flow out of but not into a box containing alpha-radioactive source). Exactly the same mathematics applies to the electric J magnetic fields where v is replaced by E or B and the quantity. corresponding ‘outflow of a material substance is called flu FIGURE 103 For our example of water flow, let V = vp, where g is the density of the war ‘Then the amount of water crossing in time ¢ an area 4° which is perpendicular to direction of flow, is (see Figure 10.1) the amount of water in a cylinder of cross sect 4 and length v1. This amount of water is (10.2) (ep), The same amount of water crosses area 4 (see Figure 10.1) whose normal is inclined angle to v. Since 4’ = A cos 0, (10.3) vA'p = vip cos 8. ‘Then if water is flowing in the direction v making an angle (with the normal nt surface, the amount of water crossing unit area of the surface in wis time is cio.) rp cos = cos 0= Ven if n is a unit vector. Now consider an clement of volume de dy dz in the region through which th ‘water is flowing (Figure 10.2). Water is flowing either in or out of the volume di dy « through each of the six surfaces of the volume element; we shall calculate the ne! outward flow. In Figure 10.2, the rate at which water flows into uy dy dz throug surface 1 is [by (10.4)} Vi per unit area, or (V +i) dy ds through the area dy dz » FIGURE 102 seo ____AME_ DIVERGENCE AND ‘THE DIVERGENCE THEOREM 278 surface 1. Since V+ i= 1, we find that the rate at which water flows across surface 1 is Ve dy de. A similar expression gives the rate at which water flows on! through surface 2, except that T’y must be the » component of V at surface 2 instead of at furface 1. We want the difference of the two 1’, values at two points, one on surface I and one on surface 2, directly opposite each other, that is, for the same’ and 2. These wo values of 1, differ by AV, which can be approximated (as in Chapter 4) by dI”, For constant yy and =, dV, =(€Vex) de. ‘Then the net outflow through these two surfaces is the outflow through surface 2 minus the inflow through surface 1, namely, (Ean We get similar expressions for the net outflow through the other two pairs of opposite (05) [(s a surface 2) — (Vat surface 1)] dy di surfaces through top and bottom, and (10.9) through the other two sides. ‘Then the total net rate of loss of water from dv dy dz is dx dy (0) (BBs div Wav dy ds or V> Pde did or If we divide (10.7) by dx dy dz, we have the rate of loss of water per unit volume. This is the physical meaning of a divergence: It is the net rate of outflow per unit volume evaluated at a point (let dv dy d2 shrink to a point). This is outflow of actual substance for liquids, gases, or particles; ic is called flux for electric and magnetic fields. You should note that this is somewhat like a density. Density is mass fer unit volume, but it is evaluated at a point and may vary from point to point. Similarly, the divergence is evaluated at each point and may vary from point to point [As we have said, div V may be different from zero either because of time variation of the density or because of sources and sinks. Let v source density minus sink density wet mass of fluid being created (or added via something like a ‘minute sprinkler system) per unit time per unit volume; 1p = density of the fluid = mass per unit volume; pict = time rate of increase of mass per unit volume: Then Rate of increase of mass in dx dy de = rate of ereation minus rate of ourward flow, oF in symbols op Se dy d: ane W de dy dz —V-V de dy“ia i 4 VECTOR ANALYSIS - os pee SF DIVrRGENCE AND THE DIVERGENCE THEOREM 2s Cancelling dv dy de, we have ‘The value of V = V on the let is of course, am average value of W = V in dy but if we -vey Give (10.13) by de and let dr shrink to a point, we have a definition of V=W at the co ein oa (ios) vevey (01) Ve V= sim J [fone If there are no sources or sinks, then y = 0; the resulting equation is often called 1c = equation uf continaaty. (See Problem 13.) 009) v Equation of continuity If Cp/ct = 0, then = (10.10 Vevey, In the ease of the electric field, the “sources” and “sinks” are electric charges and {equation corresponding to (10.10) is div D = y, where is the charge density and 1) the electric displacement. For the magnetic field B you would expect the sources to ‘magnetic poles; however, there are no free magnetic poles, so div B We have shown that the mass of fluid crossing a plane area A per unit time is 4V-n, where n is a unit vector normal to A, v and p are the velocity and density of the —/ uid, and V = vp. Consider any closed surface, and let do represent an area element on the surface (Figure 10.3). For example: for a plane, da = dv dy; for a spherical surface, always. do =? sin 0 d0 dd. Figure 103 Let n be the unit vector normal to do and pointing ou of the surface (varies direction from point to point on the surface). Then the mass of uid flowing o through do is Vn dor by (10.4) and the toral outflow from the volume inclosed by th surface is ony [fy nde, where the double integral is evaluated over the closed surface. We showed previousi that for the volume element dt = dx dy dz, the outflow from de is (10.2) VeVue Ic is worth noticing here another way [besides (7.2)] of defining the divergence. We write (10.11) for the surface of a volume element dt, we have two expressions fir the total outflow from dr, and these must be equal. Thus (10.13) VeVi I Venue. Je we start with (10.14) as the definition of ¥ + V, then the discussion leading to (10.7) is a proof that V - V as defined in (10.14) is equal to V + V as defined in (7.2) ‘The Divergence Theorem (10.17) Consider a large volume + imagine it cut up into volume elements ds, (a cross section of this is shown in Figure 10.4). ‘The outflow from each dr, is VW de,; let us add together the outflow from all the de, to get, (0.15) DV Vade, FicuRe 104 We shall show that (10.15) isthe outflow from che large volume +. Consider the flow between the elements marked w and & in Figure 104 across their common face, An outflow from a to B is an inflow (negative outlow) from b to a, so that in the sum (10.13) such ourflows across interior faces cance. ‘The total sum in (10.13) then equals just the total outflow from the large volume, As the sizeof the volume elements ends to zero, this sum approaches a triple integral over the volume, oo [[Jeove We have shown that both (10.11) and (10.16) are equal to the total outflow from the large volume; hence they are equal to each other, and we have the divergence theorem as stated in (9.14) coi — ff vevse= ff vende. Divergence serem (a poimts out of the closed surface «.) Notice that the divergence theorem converts a volume integral into an integral over a closed surface or vice versa; we can then evaluate whichever onc is the easier to do Jn (10.17) we have carefully written the volume integral with three integral signs and the surface integral with two integral signs, However, itis rather common to write only one integral sign for either case when the volume or area clement is indicated by single differential (de or da). Thus we might write [ff de or f dt or [ff dr dy ‘meaning the same thing. When the single integral sign is used to indicate a surface or ‘volume integra, you must sce from the problem or the words under the integral what is really meant. To indicate a surface integral over a coved surface or a line integral216_VECTOR ANALYSIS around a closed curve, the symbol § is often used. ‘Thus we might write ff do or § dy or § do, all meaning a surface integral over a closed surface. A different notation for the integrand V-n do is often used. Instead of using a unit vector and the scalar magnitude da, we may write the vector do meaning a vector of magnitude da in the direction nj thus de means exactly the same thing as nde, and we may replace V - s do by V = de in (10.17) Example of the Divergence Theorem Let V = ir + jy + kz and evaluate § V'-n da over the closed surface of the cylinder shown in Figure 10.5. By the divergence theorem this is equal to {V+ V dr over the volume of the cylinder. (Note that we are using single inte- eral signs, but itis clear which integeal is a volume integral and which a surface integra.) We find from the definition of diver- sence FIGURE 10.5 Then by (10.17) ee I view paseo fas times volume of cylinder = 3na*h. Ie is harder to evaluate § Vn do directly, but we might do it to show an example of calculating a surface integral and to verify the divergence theorem in a special case We need the surface normal n, On the top surface (Figure 10.5) n= k, and there VensVek=s =k Then J ovenae forms mat e™—“‘SECCC surface is zero, On the curved surface we might see by inspection that the vector iv + jy is normal to the surface, so for the curved surface we have ixtiy _ivtiy Very a If the vector n is not obvious by inspection, we can easily find it; recall (Section 6) that if the equation of a surface is G(x, », 2) =const., then Vp is perpendicular to the surface. In this problem, the equation of the cylinder is x74)? =a; then seem “THE DIVERGENCE AND THE DIVERGENCE THEOREM _277 2si + yi, and we get the same unit vector nas above, Then for arty VO {Fred sure we Bd The value of §V-ndo over gath + 2na7h = 3na*h as before, the whole surface of Ye Law The divergence theorem is very important in aera ved, we net ain electri hnown ae Gaus’ lw. Let us drive this re ae ti ein to more ar Confomb' nw (11. Cala a (oi ve er ns unin) ves forthe ci elds et 2 pin carey athe ] origin Coulomb's law €) = 9 10? in a vacuum in mks units.) ‘The (€ is the dieleetric.constant, and 1/(4n€) = | clectric displacement D is defined by D = €E; then te, (10.19) D= ves sctin of FIGURE 108 Let @ be a closed surface surrounding the point charge at the origin; let do be an element of area of the surface atthe point r, and let n bea unit normal 1 de (Figures 10,3 and 10.6). Also (Figure 10.6) let 4 be the projection of da onto a sphere of radius rand center at O and let dQ be the solid angle subtended by do (and 4) at O. Then by definition of solid angle 1 (10.20) a= 74a” 28_VECTOR ANALYSIS. From Figure 10,6 and equations (10.19) and (10.20), we get 4 (v0.2) Ds nde =D cos Ode = Dd > 1 Garman We want to find the surface integral of D+ n da over the closed surface 0; by (10.21 this is wf Denwe=t [an (¢ inside 0), ‘This is a simple case of Gauss’s law when we have only one point charge 43 for most purposes we shall want Gauss’s law in the forms (10,23) or (10.24) below. Before we derive these, we should note carefully chat in (10.22) the charge 4 is inside the closed surface a, If we repeat the derivation of (10.22) for a point charge ¢ outside the surface (Problem 13), we find that in this case D-ndo=0 Next suppose there are several charges 4; inside the closed surface. For each y, and the D, corresponding to it, we could write an equation like (10.22). But the total electric displacement vector D at a point due to all the g, is the vector sum of the vectors D,. Thus we have therefore for any charge distribution inside a closed surface (1023) § Dn de = wal care inside the coed ice. Gates iw Uf, instead of isolated charges, we have a charge distribution with charge density (hich may vary from point to point), then the total charge is f p-d, so 02% § Dendo= foes Gauss tw Since (by Problem 13) charges outside the closed surface @ do not contribute to the integral, (10,23) and (10.24) are correct if D is the total electric displacement due to all _— ‘THE DIVERGENCE AND THE DIVERGENCE THEOREM 279 charges inside and outside the surface. The total charge on the right-hand side of these fquations is, however, just the charge inside the surface o. Fither (10.23) or (10.24) is talled Gauss’s law. . ‘We now want to see the use of the divergence theorem in connection with Gauss's law. By the divergence theorem, the surface integral on the left-hand side of (10.23) or (10.24) is equal to foc wounded bye ‘Then (10.24) can be written as [vipa [oa Since this is true for every volume, we must have VD =p; this is one of the ‘Maxwell equations in electricity. What we have done is to start by assuming Coulomb's law; we have derived Gauss’s law from it, and then by use of the divergence theorem, wwe have derived the Maxwell equation VD =p. From a more sophisticated view- point, we might take the Maxwell equation as one of our basic assumptions in electri- city, We could then use the divergence theorem to obtain Gauss’s law nas) § Dendem [ vepdew | pde~ sot ge inde by 0 From Gauss’s law we could then derive Coulomb's law (Problem 14): more generally wwe can often use Gauss’s law to obtain the electric field produced by a given charge distribution as in the following example Example. Find E just above a very large conducting plate carrying a surface charge of C coulombs per square meter on each surface. ‘The electric field inside a conductor is zero when we are considering an electrostatic Problem (otherwise current would flow). From the symmetry of the problem (all huri- zontal directions are equivalent), we can say that E (and D) must be vertical as shown in Figure 10.7. We now find § D «n do over the box whose cross section is shown by the dotted lines. The integral over'the bottom surface is zero since D =O inside the conductor. The integral over the vertical sides is zero because D is perpendicular to n there. On the top surface D+ n= |D| and [Dn de =|D|- (surface area). By Figure 10.711. Given that B 280_VECTOR ANALYSIS cng (10.25) this is equal to che charge inclosed by the box, which is C'- (surface area). ‘Thus we have | D| «(surface area) = C - (surface area), or |] = C and || = Cle PROBLEMS, SECTION 10 1. Evaluate both sides of (10.17) if V= vay? ixtiy+ke, and + is the volume << 1, and so verify the divergence theorem in this case 2. Given V = 41 + y?] + 27k, integrate V earth's radius R, but this is not a correct formula for r V290_VECTOR ANALYSIS Example 2. Given V =i? — V=VKA We find div Sut +S (2 str 2e 422 2e=0, ‘Thus V is solenoidal and we proceed to find A. We are looking for an A such that ie aaa oe & & A, A, Ay quay curl A= ~ ye + K(2? — 2229, ‘There are many A’s satisfying this equation; we shall show first how to find one of them and then a general formula for all. It is possible to find an A with one zero component; let us take 4, = 0. Then the y and 2 components of curl A each involve just one component of A. From (11.11), the y and 2 components of curl A are aA, o, an) Qe= Pte ” or" ole If we integrate (11.12) partially with respect to x (that is, with y and 2 constant), we find 4, and A, except for possible functions of y and 2 which could be added without changing (11 an) Ay= Bx 2x? + fy 2), A, = bayz + fils 2), Substituting (11.13) into the x component of (11.11), we get (ig) We nom ast ans (1.1). The mush ey hee a his an easily be done by inspection. We could take f; = 0, f; = $y2%, of fy = 0, fy = —4y": and so forth. Using the second choice, we have : a Ys anasy A= j(2x — 2x) + kQaye — by*2), You may wonder why this process works and what div V = 0) has to do with it, We an answer both these questions by following the above process with a general V rather than a special example. Given that div V = 0, we want an A such that V = curl A. We tty to find one with A, = 0. Then the y and 2 components of V = curl A are 24, «aL16) dye th(? —2ex), find A such thay “a1i7) _i [THE_CURL AND STOKES’ THEOREM 291 en we have V, de +f) A, -[ow +e ‘phe « component of V = curl A is 1,24 11s) Since div V a9) into (11.18), getting ‘This is correct with proper choice of A(s 2). ‘When we know one A, for which a given V is equal to curl A, all others are of the form (11.20) AG Vu, ‘where u is any scalar function. For (see Problem 7.17b), V x Vu = 0, so the addition of Yu to A does not affect V, Also we can show that all possible A's are of the form (11.20). For if V = curl Ay and V-=curl Ag, then curl (Ay — 3) = 0, s0 Ay ~ Ap is the gradient of some scalar function. 'A careful statement and proof that div V =0 is a necessary and sufficient condition for V = curl A requires that V have continuous partial derivatives at every point of a region which is simply connected in the sense that every closed surface (rather than closed curve) can be shrunk to a point in the region (for example, the rexion between two concentric spheres is not simply connected in this sense) PROBLEMS, SECTION 11 1. Do case (b) of Example I above. 2. Given the vector A= (04 = 9 + DH (@) Find Vx A (8) Evaluate ff (V % A) “de over a rectangle in the (x,)) plane bounded by the ines r= 08 yb (6) Evalue § A ae around the boundary of the rectangle and thus verify Stokes! theorem for this case Use either Stokes’ theorem or the divergence theorem to evaluate each of the following integrals in the easiest possible way. 3. ffearnces Curl (271 + 22) —y*k) +m do, where o is the part of the surface above the (s, y) plane292 VECTOR ANAL 6. [FV mde over the closed surface of the tin can bounded by x2 4)? = 4, 40 Sfeurl OF + 23) mado, where o is the surface in the frst extant ‘made up of part of the plane 2 + ay 4 42 2, aed triangles in the (4 2) and (1, £} planes, as indicated inthe figure 5 Mfr omdo oer the surface in Problem 4, where Foe iv + ip ke, Hint: See Problem 103. / V= 2nd yi 4 (24 ok {f (curl V) «mde over any surface whose bounding curve is in the (4,1) plane. where +0) Yi +r — ok 8. curt (yi 02k) «ml over the slsed surface ofthe ellipsoid Waruine: Stokes: theorem applies only co an open surface. Hints: Could you cut the o Surface into two halves? Also see (d) in the table of vector identities 9. F[V onde over the entire surfice of the volume in the first octane twonele: 2 3B. 14. " 16 and the coordinate planes, where Vewre Sf tourt Vim do over the V~ devi + (04 anja Y+ Gas — 2enj— of the surface 9 = = 99? above the (4.9) plan h {LV no over the entire surface of a cube of edge 2 in the first octane, where Vs6? y+ yi JV de around the cincle x? 4? + 2v = 0, where F = = 29. §V de around the boundary of the square with vertices (1, 0), (0,1), (1, 0) (0, —1), i Vai + Sy BW. Jo(s? —y) dr + (04) dh, where C is the parallelogram with vertices at (0,0, (2,0) Gye 3B. fa? =) de + Gey + 3) dy along the x axis from (0,0) to (4/ 0) 0 (1,2). Hint: Use Green's theorem 0) and then along « circular are from (4/ Table of Vector Identities involving V ——————— ey Notice carefully that ¢ and y are scalar functions; U and V are vector functions. (a) V+ Vo = div grad @ = V2 = Laplacian (b) Vx Vo = curl grad @ = 0 (©) VOW» V) = grad div V ey, @ Vw div cul V =0 (©) Vx (V x V) = curl curl V = V(V + V) — V?V = grad div V— Laplacian V (©) V-@V)= VW) +V-(¥4) (@) Vx (GV) = HV x V) —V x (9) (h) V-(UxXW=V-(V x L)-U (Vx) () VxUx Y= WU -(WU- WVU) + UW) G@) VU -VW=Ux WX V+ (UW EV x (Ox UE: DU () V-(¥G x Vy) =0 ee eee——— possesses “high fidelity.” ‘To find out which harmonic are the important ones in tees sgn, we egund int Powe ee, ‘The traf the sr ih rge coefficients then represent the important harmonics (Frequencies). ee was Since snes and cosines are themsles petit sems rater str to ue ser OF then ater than poner sri, repeat pre Rint, There is ante important reason. The coefficients of a power series are cosine yo wil rel (Chapter 1, Section 12), by finding successive derivatives of the ee a a panded cece, nl camino fenton with deratives af alors cane Send in pew sre. Many pre finns in pti ae ot ction o not dieretale Figure 32) Fount, Fourier sees (une power st) represent discontinuous functions of functions whose graphs have corners. On the other hand, Fourier series do not usually converge as apidly a8 power series and much more care is necded in manipulating them. For eximple, you cannot usually difeentate a Fourier series term by term. (For more deal, se references on Fourier series.)ais 304 FOURIER SERIES Our Problem then is 9 expand 2 given periodic function in a series of sines ang cosines. We shall rake this up in Section 5 after we have done some preliminary work PROBLEMS, SECTION 3 Draw a graph over a whole petiod of each of H the following combinations of 4 fundamen) ‘musical tone and some of its overtones Le sine 4 sin ar 2 206 + cos 2r 3. sin ne sin 2ne + § sin 3ar 4. cos Int + cos 4nt + $ cos 6xt 5 Using the definition end of Section 2) of 4 perindic function, show that a sum of tem responding toa fundamental musial tone and its overtones has the period ofthe fan mental 6 Draw a graph of sin 2r + sn 2 + 9/3). Hint: Use a trigonometry formula t write this as g single harmonic. What are the petiod and amplitude? 7. Asin Problem 6, draw a graph of cos nx — cos n(x — }) ‘The voltage in a telegraph line is, under certain circumstances, given by asin(s-2) 4 Asin (1-2 +2) where © is the velocity and ¢ isthe time for a wave to travel the length of the line. Tis represents the combination of a wave starting at one end and the reflection of the wace rea she other end. What are the wavelengths and frequencies of the waves? Stetch the gph of Lila) when A= 2, v= 3, c= 1, both as a function of ¢ for x= 2 (over the time inerval {or which is between 1 and 2) and asa function ofr (between O and 3) when t=} Fes 9. A periodic amplitude modulated (AM) radio signal has the form =A B in 20 in 261 — ) The factor sin 20/(¢~ x0) is called the carver wave; it has a very high fequency called radio frequency; fis of the order of 10° eycles per second). ‘The amplitude of the corer wave is (A +B sin 2af) This amplitude varies with time—hence the term “ample ‘odulaton””—with the much smaller frequency of the sound being transmitted (called adie frequency; Fis ofthe order of 10? cycles per second). In order to See the general appearance of such a wave, use the following simple but unrealistic data to sketch @ graph ory ay 2 function of for x= 0 over one period of the amphiude function: d= 3 B = 1, fl, J, = 20, Using trigonometric formulas, show that y can be writen as acum of thee waves of freauences ff. +f, and f,~ J; the fist of there isthe carrer wave and the other we are called sie bands. 4, AVERAGE VALUE OF A FUNCTION ‘The concept of the average value of a function is often useful. You know how to find he average of a set of numbers: you add them and divide by the number of numbers his process suggests that we ought to get an approximation to the average value of 4 AVERAGE VALUE OF A FUNCTION 30s FIGURE 4.1 function /(x) on the interval (a, 6) by averaging a number of values of f(s) Figure 4.1) «) Average of f(x) on (a, #) is approximately equal 10 L064) + fle) +--+ fly) This should become a better approximation as n increases. Let the points #4, #3, ~~ be Ax apart. Multiply the numerator and the denominator of the approximate average by ‘Ax. Then (4.1) becomes: (4.2) Average of f(x) on (a, 6) is approximately equal to er Leen) n Ay Now n Ax =~ a, the length of the interval over which we are averaging, no matter what and Av are, If we let m+ 20 and Av—+ 0, the numerator approaches fj f(x) dx, and we have Zl a) de ey Average of (2) on (a, 8) = BLO) 4 ge value of a given function is zero, In applications, it often happens that the average value of a gi ti For example, the average of sin x over any number of periods is zero, The average Yalu of the velocity of a simple harmonic osilator over any: number of vibrations is zero. In such cases the average of the square of the function may be of interest. For example, if the alternating electric current flowing through a wire is described by a sine function, the square root of the average of the sine squared is known as the root Square or effcctive value of the current, and is what you would measure with an a-c ammeter. In the example of the simple harmonic oscillator, the average kinetic energy (average of $yr0*) is 4m times the average of Now you can, of course, find the average value of sin? x over a period (say —x to 2) ring ic. There is an easier way by looking up the integral in (4.3) in tables and evaluating is : Which is well worth Knowing. By considering the graphs of cos? x and sin? x searate 06_FOURIER SERIES. lo E FIGURE 42 (Figure 4.2), you ean probably convince yourself thatthe area under them is the sy for any quarterperod ftom 0 to n/2,x/2 to mee (Albo see Prablems 2 and 13) Tae (4) f sin? x anf cos? x de. Similarly (for integral n 4 0), But since sin? nx + cos? mx = 1, 4.6) (sin? nx + cos? nx) de f dr = 2r, Using (4.5), we get a7 ‘Then using (4.3) we see that: ‘The average value (over a period) of sin? nx = the average value (over a period) of cos? mx We can say all this more simply in words, By (45), the 7 7 simply Is. By (4.5), the average value of sin? nat equals the average value of cos? nx. The average value of sin? nx + cos? nx = 1 ig 1, There- fore the average value of sin* mx or of cos? mx is 4. (In each case the average value is taken over one oF more periods.) 48) ? die [sont x sin? ne de zf cost ne ds = PROBLEMS, SECTION 4 1. Show that if /(a) has period p, the average value of fis the same 1 of k , lue of fs the same over any interval of length 2. Hints Write °F f(x) dx as the sum of two integrals (a to P, and p to a-+ p) and make ‘the change of variable x ~ 1 + pin the second integral 2. (a) Prove that fg? sin? x d in one of the integrals [5° cos? x de by making the change of variable x = $n — 1 s FOURIER COEFFICIENTS _307 (b) Use the same method to prove that the averages of sin# (wna) and cos? (nsf) are the same over a period. tn Problems 3 to 12, find the average value of the function on the given interval, Use equation (48) if applies. Ifa average value is zero, you may beable to determine this from a sketch fe Tae) 6 mr ao arenes oo (02) a sinas m (82) 9. sin? By on (0, 4 10. cosx on (0, 3n) Ih. sins +sin?.< on (0,20) 2. cost 7 on (0,4) 13. Using (4.3) and equations similar to (4.5) 10 (4.7), show that if Hb — 0) is a mutple of x Use the results of Problem 13 co evaluate the following integrals without calculation. wa [Pee o Poe()e » [ow Ge [leo ama 6) [sot ove 5. FOURIER COEFFICIENTS We want to expand a given periodic function in a series of sines and cosines. To simplify our formulas at first, we shall start with functions of period 2x and call ‘our basic functions sin nx and cos'nx instead of sin no and cos mot. Later we shall sce how we can change the formulas to fit a different period. The functions sin x and cos x have period 2n; so do sin mr and cos mx for any integral 1 since Sin n(x + 2n) = sin (ny + nm) = sin mx. (It is true that sin ry and cos ny also have shorter periods, namely 2n/n, but the fact that they repeat every 2x is whar we are interested in here, for this makes them reasonable functions to use in an expansion of a function of period 2x.) Then, given a function f(x) of period 2: 6.1) (3) = da + 0, 008 x + a3 cos 2x + ay cos 3x + 4b; sin x + by sin Dy + by sin 3x40,308 FOURIER SERIES and derive formulas for the coefficients 4, and ,. (The reason far writing 4a a8 the constant term will be clear later—it makes the formulas for the coefficients simpler te remember—but you must not forget the 4 in the series!) In finding formulas for 4, and 2, in (3.1) we need the following integrals le | sin mx cos mx dx = 0. 62) ‘The average value of sin ms cos nx (over a petiod) ‘The average value of sin mx sin nx (over a period) (o men, Lh gear ees la nore Te wer nie sae mean fossa [% men, i i Cat m=n#0, We have already shown that the average values of sin? nx and cos? nx are 4. The last integral in (5.2) is the average value of 1 which is 1. To show that the other average values in (5.2) are zero, we could use the trigonometry formulas for products like sin 0 cos @ and then integrate. An easier way is to use the formulas for the sines and cosines in terms of complex exponentials. [See (7.1).] We shall show this method for one integral [anne wa [fl Set ete 63) de. @ 2 We can see the result without actually multiplying these out. All terms in the product are of the form e**, where & is an integer % 0 (except for the cross-product terms when n= m, and these cancel), We can show that the integral of each such term is zero because * = 7" = cos fx (since sin kn = (0). The other integrals in (5.2) may be evaluated similarly (Problem 12). ‘We now show how to find a, and b, in (5.1). To find ag, we find the average value on (=n, 2) of each term of (5.1). 64) sin x ds FOURIER COEFFICIENTS _508 5 _ cepa on the sighchand side of (5.5) are sero excep the fs, oe ae cans toe ae ovof co mz cue at wih # =O and mF O By 62) ‘T pecause they are integral (that is, mm). Then we have itt tol [yao El JoeB ao 66 ay Given (2) to be expanded in a Fourier series, we can now evaluate ao by calculating integral in (5.6). ; aa esse ind as, multiply both sides of (5.1) by oos x and again find the average value of each term: e Bn Ef omnes aot ‘i 2 424" cog ede ta, t | cost dr tal Bl, tangy [cos tremrdes > gn x cos x de + a ‘This time, by (5.2) all terms on the right are zero except Le" cost x de = 4. 2 Je - Solving fora), we have € ce [00 08x de. ‘The method should be lear by now, se we shall next find 2 general formula for a, Multiply both sides of (5.1) by cos mx and find the average value of each te fod f cos ne de +a, [i coer rn 68) xls cos nr de = 80 tay [cos 2x con ms ds + aa) ale sin x cos mx dx + +h ‘except the one By (5.2), all terms on the right are z le a), cos? mx de Solving for a,, we haveFOURIER COEFFICIENTS _311 310_FOURIER_ SERIE: a s Notice that this includes the m= 0 formula, but only because we called the constant 1us ag =}, and all other a, = 0. To obtain a formula for 6,, we multiply both sides of (5:1) by sin nr and tte <1 jysinnae tea} [*o-sinne de [1 sin ne de] average values just as we did in deriving (5.9). We find (Problem 13) wae) fo Es 4 Dee ~ * 1 bok seo sin near a2 [sin ne ar =t ~Ane-r-n _! [0 for even eel (5.9) and (5.10) will be used repeatedly in problems and should be = ie for odd n. Expand in a Fourier series the function f(x) sketched in Figure 5.1. ‘This on might represent, for example, a periodic voltage pulse. The terms of our Fourier series would then correspond to the different a-c frequencies which are com. bined in this “square wave” voltage, and the magnitude of the Fourier coefficients ‘would indicate the relative importance of the various frequencies, fy FIGURE 5.1 Note that /(x) is a function of period 2x. Often in problems you will be given /(x) for only one period; you should always sketch several periods so that you see clearly the Periodic function you are expanding. For example, in this problem, instead of a sketch, you might have been given It is then understood that /(x) is to be continued periodically with period 2x outside the interval (—x, 2) ‘We use equations (5.9) and (5.10) to find a, and 2 [vo comes [1 comme de] 0, -a 0, and f(x) = —1 for every x for which sin (I/x) <0, this function would have an infinite number of discontinuities, You can see that most functions you are apt to meet in applied work will not behave like these, but will satisfy the Dirichlet conditions, Finally, if y = 1/x, we find Lb 0 the function I/x is ruled out by the Dirichlet conditions. On the other hand, if F(a) = Va/ Te, then so the periodic function which is 1/x/TaT between —z and 2 can be expanded in a Fourier series. In most problems it is not necessary to find the value of J. | f(a)| drsSM_FOURIER SERIES _ let us see why. If /(2) is bounded (that is, all its values lie between tM for some positive constant M), then [ire des [aden anon and so is finite. Thus you can simply verify that the function you are considering ig bounded (if it is) instead of evaluating the integral Figure 6.1 is an example of 5 function which satisfies the Dirichlet conditions on (—m, 2), MSN 9 HM A We see, then, that rather than testing Fourie series for convergence as we did power series, we instead check the function we want expanded; if it satisfies the Dirichlet conditions we are then sure that the Fourier series, when we get it, will converge to the function we expanded except at jumps where it converges to the midpoint of the jump We can now verify that the series (5.12) actually represents the function we started with (Figure 5.1) at all points in between the points wa. (We have already noted that at nz the series gives the value 4, which is half-way between 0 and 1 as Dirichlet’s theorem says) Between —x and m the given /(x) is single-valued (one value for cach 1), bounded (between +1 and 0), has 4 finite number of maximum and minimum values (one of each), and a finite number of discontinuities (at — x, 0, and x), and therefore satisis the Dirichlet conditions. Dirichlet’s theorem then assures us that the series (6.12) actually converges to the function f(x) in Figure 5.1 at all points except « = nt It is interesting to see a graph of the sum of a large number of terms of a Fourier series. Figure 6.2 shows several different partial sums of the series in (5.12) for the funetion in Figure 5.1. We can see that the sum of many terms of the series closely approximates the function away from the jumps and goes through the midpoint of the jump. ‘The “overshoot” on either side of a jump bears comment. It does not disappear 38 we add more and more terms of the series. It simply becomes a narrower and narrower spike of height equal to about 9% of the jump. This fact is called the Gibbs phenomenon We ought to say here that the converse of Dirichle’s theorem is not tuc—if a function fails to satisfy the Dirichlet conditions, i still may be expandable in a Fourier series. The periodic function which is sin (I/x) on (—m, x) is an example of such a function. However, such functions are rarely met with in practice. feud COMPLEX FORM OF FOURIER SERIES 315 Number of terme FIGURE 62 PROBLEMS, SECTION 6 1 t0 11. For each of the periodic functions in Problems §.1 to $.11, use Dirichle’s theorem to find the value to which the Fourier series converges at x =0, +n/2, tm, +2n the series in Problems 5.1, 58 and 12, Sketch a graph of the sum of three terms of each of the r 5.11, and compare this approximation with the graph of f(s). Mint: Sketch each term separately (on the sime axes) and add the terms graphically 7. COMPLEX FORM OF FOURIER SERIES Recall that real sines and cosines can be expressed in terms of complex exponentials by the formulas [Chapter 2, (11.3)] ay If we substitute equations (7.1) into a Fourier series like (5.12), we get a series of terms Of the forms e™* and ¢~'**. This is the complex form of a Fourier series. We can also find the complex form directly; this is often easier chan finding the sine-cosine form, We can then, if we like, work back the other way and [using Euler's formula, Chapter 2, (9.3)] get the sine-cosine form from the exponential form,ho 316_FOURIER SERIES _ We want to see how to find the coefficients in the complex form directly. We assume SO) = cote tee bee be ge "| and try to find the «,'s. From (5.4) we know that the average value of ¢ on (— x, 2) is zero when & is an integer not equal to zero. To find ¢g, we find the average valuey ot the terms in (7.2) if aft [average values of terms of th GC po hs ee cane a) te dg * Ylorm eM with ban integer #0 =m +0, 4) ory f Se) de. ‘To find cy, we multiply (7.2) by e~"* and again find the average value of each term. {Note the minus sign in the exponent. In finding a,, the coefficient of cos ny in equation (5.1), we multiplied by cos nx; but here in finding the coefficient c, of e*, we multiply by the complex conjugate e~'"™) a5) ("pene dene [ete LE a 7.3) Bl yom amon f atar| ele de ee earer 1a | metas ‘The terms on the right are the average values of exponentials e'*, where the & values are inegers. Therefore al thse terms are zero except the one where FOr this the term contuning ey. We then have Ne ae 1 zl fom 3 flees Rl soem ae Note that this formula contains the one for ¢o (no 4 to worry about here!). Also, since (7.6) is valid for negative as well as positive m, you have only one formula to memorize here! You can easily show that for real f(s), ¢, = Z, (Problem 12), ‘ OTHER INTERVALS _317 Example, Let us expand the same f(s) we did before, namely (5.11). We have from * 76) an ‘Then 8) It is interesting to verify that this is the same as the sine-cosine series we had before: We could use Euler's formula for each exponential, but it is easier to collect terms like 5 ee ) a9) 1.2 nyt iinet Hsin de 9 which is the same as (5.12) PROBLEMS, SECTION 7 1 to 11. Expand the same functions as in Problems $.1 ro 5.11 in Fourier series of complex ‘exponentials &* on the interval (—m, x) and verify in each case (by using Euler's formula) that the answer is equivalent to the one found in Section 5 12, Show that if areal f(a) is expanded in a complex exponential Fourier series T2.. ye, then ¢_, = i, where Z means the colnplex conjugate of cy 13. Ef) = dao + EF a, com wr + DF by sin mr = TE eye use Euler's formula to find a, and by in terms of and ¢.,, and to find ¢, and ¢, in terms of e, and by 8. OTHER INTERVALS ‘The functions sin nx and cos nx and e* have period 22. We have been considering (=x, x) as the basic interval of length 2n. Given f(x) on (=x, 7), we have first sketched it for this interval, and then repeated our sketch for the intervals (x, 3x), Gx, 52), (—3n, —), ete. There are (infinitely) many other intervals of length 2x, any314 FOURIER semis ‘one of which could serve as the basic interval. If we are given f(2) on any interval of length 22, we sketch /(3) for that given basic interval and then repeat it period, ically with period 2x. We then want to expand the periodie function s0 obtained, in , Fourier series, Recall that in evaluating the Fourier coefficients, we used average Valuc, over a period. The formulas for the coefficients are then unchanged (except for the Timits of integration) if we use other basic intervals of length 2x. In practice, the intervals (—a, x) and (0, 22) are the ones most frequently used. For f(x) defined ne (0, 2) and then repeated periodically, (5.9), (5.10), and (7.6) would read wel [revered ne! [dened (8.1) ” 4 =} Prom dey | and (5.1) and (7.2) are unchanged. FIGURE 1 Notice how important it is to sketch a graph to see clearly what function you are talking about. For example, given f(x) = 2? on (—n, n), the extended function of period 2x is shown in Figure 8.1. But given f(x) = x* on (0, 2n), the extended periodic function is different (see Figure 8.2). On the other hand, given f(x) as in our example G.ID, oF given f(x) =1 on (0,2), f(x) =0 on (x, 2n), you can easily verify by sketching that the graphs of the extended functions are identical. In this case you would fet the same answer from either formulas (5.9), (5.10), and (7.6) or formulas (81). LLL! FIGURE 8.2 Physics problems do not always come to us with intervals of length 2x, Fortunately, it is easy now to change to other intervals. Consider intervals of length 2/, say (1, 1) ‘oF (0, 21). The function sin (nz) has period 21, since sn sn (Ee) mn OTHER INTERVALS 9 5 similarly, cos (nns/l) and" have period 2. Equations (5.1) and (7.2) are now replaced by We have already found the average values ner « period of all the funeions we need to use to find a,, by, and ¢, here. The period is now of length 2/ (—f to 1), so in finding average Values of the terms we replace pa ml, Ya, er reriod is $ Bone nence of costes Then the mas 9), (610) a (78) tr the cui become 6.3) = For the basic interval (0, 2/) we need only change the integration limits to 0 to 21. ‘The Dirichlet theorem just needs z replaced by / in order to apply here 0, = ce ae solution containing one arbitrary constant. Although this solution may be referred to as the “general solution,” show that y = 1 is a solution of the differential equation not obsainabic from the * general solution” by any choice of the arbitrary constant. The solution y = 1 is called a singular solution; y = —1 is another singular solution. Sketch a number of graphs of the “ general solution "for different values of the arbitrary constant and observe that y jis tangent to all of them. This is characteristic of a singular solution—its graph is tangent at each point to one of the graphs of the “general solution.” Note that the given differential equation is not linear; for linear equations, all solutions are contained in the general solu- tion, but nonlinear equations may have singular solutions which cannot be obtained from the “general solution” by specializing the arbitrary constant (or constants). Thus @ nonlin- car frstorder equation in x and y may have two (or more) solutions passing through 1 tiven point in the (x, y) plane, whereas a linear first-order equation always has just one such Solution, Show that any continuous curve made up of pieces of y= 1,” = ~ 1, and the sine curves of the “ general solution,” gives a solution of the above differential equation. Stetch such solution curve on your graphs In Problems 13 to 15, find singular solutions of the differential equations in the indicated problems by inspection and show thatthe singular solution cannot be obtained from the * general Solution” by specializing the arbitrary constant. 13, Problem 2 14, Problem 3 15, 6. Problem 7. Find the “general solution” of the equton y= /F by separation of wah, Find panic solution ishing y= 0 when x= 0. Show the the singular solution y= Cannot be obsined frm the gene solution. Sketch pap of the "general lon” ot sever values ofthe abitary costam, and observe that eich of them i tangent tothe Singur solution. Thus thre are tw sltions ping through any pant onthe rst in Parc, there ae two slutons saistying x= 0, Probie 17 and 8 ae peel Probl ading to hin ierencal equation 17, The velocity of a particle on the x axis, x= 0, is alway th is, & 0, is always numerically equal to the square oot of its displacement x. If v= 0 when x= 0, find x as a function of ¢. Show that the given conditions ae satisfied if the particle remains at the origin for any arbitrary length of time fo, and then moves away; find x for ¢ > fy for this case. 18, Let the rate of growth ddr of a colony of bacteria be proportional to the square root of the number present 2t any time. If there are no bacteria present at ¢ =O, how many are there at a later time? Observe here that the “general solution gives an unreasonable 19. (a) Consider alight beam traveling downward into the ocean, As the beam progresses, it is partially absorbed and its intensity decreases. The rate at which the intensity is decreas- ing with depth at any point is proportional to the intensity at that depth. The pro- portionalty constant 4 is called the linear absorption coefcient. Show that if the a 2, %, (0) Note that the differential equation and its solution inthis problem are mathematically the same as those in Example 1, although the physical problem and the terminology are Uliferent. In discussing radioactive decay, we call 2 the decay comtant, and we define the halflife T of a radioactive substance a8 the time when N= £No (compare balf-value thickness). Find the relation between 2 and 7 Consider the following special cases of the simple series circuit (Figure Ll and equa tion (1.29), {@) RC circuit (hat is, L = 0) with = 0; find 4 as a Function of # if ag isthe charge on the capacitor at = 0. (6) RL circuit (that is, no capacitor; this means 1/6 T= Ig atr=0. (©) Again note that these are the same differential equations as in Problem 19 and Example 1. The terminology is again different; we define the time constant + for a circuit a5 the time required for the charge (or current) to fall zo Le times its initial value. Find the time constant for the circuits (a) and (b), If the same equation, say ym ooe#, represented either radioactive decay oF light absorption or an RC or RL reat, what would be the relations among the halflife, the half-value thickness, and the time constant? Suppose the rate at which bacteria in a culture grow is proportional to the number present at any time, Write and solve the differential equation for the number N of bacteria as a function of time 1 if there ace No bacteria when f = 0. Again note that (excepe for a change of sign) this is the same differential equation and solution as in the preceding problems. ©) with V 0; find_0) given Solve the equation for the rate of growth of bacteria ifthe rate of increase is proportional to the number present but the population is being reduced ata constant rate by the removal of bacteria for experimental purposes. ‘Heat is escaping at a constant rate [dQ/dt in (1-1) is constant] through the walls of 2 long cylindrical pipe. Find the temperature T at a distance r from the axis of the esinder if the and temperature T= 100 and the outside wall his r= 2 and ‘inside wall has radius 7 T=0 Do Problem 23 for a spherical cavity containing a constant source of heat, Use the same radii and temperatures asin Problem 23. ‘Show that the thickness of the ice on a lake increases with the square root of the time in cold weather, making the following simplifying assumptions. Let the water temperature be 2 Constant 10°C, the air temperature a constant —10°C, and assume that at any given time the ice forms 2 slab of uniform thickness x, The rate of formation of ice i proportional (0 the rate at which heat is transfered from the water to the ait. Let ¢ = 0 when x = 0 [An object of mass m fills from rest under gravity subject to an air resistance proportional 10 its velocity. Taking the y axis a8 positive down, show that the differential equation ofMé_ ORDINARY DIFFERENTIAL, EQUATIONS om ‘motion is m(du/de) = mp — ke, where & is 4 postive constant, Find v as a function oft, and find the limiting value of v as 1 tends to infinity; this limit is called the ferminal veloc an you find the terminal velocity directly from the differentia! equation without solving ‘Hint: What is dejdt after © has reached an essentially constant value? Consider the following specific examples ofthis problem. (@) A person drops from an airplane with a parachute. Find a reasonable value of k (6) In the Millian oil drop experiment to measure the charge of an electron, tiny electric cally charged drops of ol fall through air under gravity or rie under the combination oF sravity and an electric field. Measurements can be made only after they have reached terminal velocity. Find a formula for the time required for a dcop starting at test tp reach 99% ofits terminal velocity. 27. According to Newton's law of cooling, the rate at which the temperature of an object changes is proportional to the difference between its temperature and that of its surround. ings. A cup of coffee at 200° in a room of temperature 70” is stirred continually and reaches 100" after 10 min, At what time was it at 120°? 28. A lass of milk at 38° is removed from the refrigerator and left in a room at temperature 70°, Ifthe temperature of the milk is $4 after 10 min, what wil its temperature be in half an hour? (See Problem 27) 29. A solution containing 90% by volume of alcohol (in water) runs at 1 gal per min into a 100-gal tank of pure water where itis continually mixed. ‘The mixture is withdrawn at the rate of I gal per min, When will it start coming out $0% aleohol? 30. If P dollars are left in the bank at interest J percent per year compounded continuously, find the amount 1 at time ¢, Hint: Find dA, the interest on A dollars for time de FFind the orthogonal trajectories of each of the following families of curves. In each case sketed Several of the given curves and several of their orthogonal trajectories, Be careful to eliminate the constant from dy/dx for the original curves; this constant takes different values for different ‘curves of the original family and you want an expression for dy/ds which is valid forall curves of the family crossed by the orthogonal trajectory you are trying to find. 33. y=Ax" (Assume that wm is a given number; the different curves of the family have different values of &,) arth 3. LINEAR FIRST-ORDER EQUATIONS AA first-order equation contains y' but no higher derivatives. A linear firstcorder equa tion means one which can be written in the form 6) y+ Py=Q, LINEAR FIRST-ORDER EQUATION sed where P and Q are functions of x. To sce how to solve (3.1), let us first consider the simpler equation when Q = 0. The equation 4 62) ytPy=0 or Ba —Py is separable. As in Section 2, we obtain the solution as follows: dy #46 ; 6s bya afrtees es 1= fa Tn a 09 “ep and we can write (3.3) as y = Ae“! or 66) ytd ‘We can now see how to solve (3.1). If we differentiate (3.6) with respect to x and use G5), we get Yd aytPady’ on Lod) ave a yt Za ve + ye Ay + Py), which is the leftchand side of (3.1) multiplied by e!, ‘Thus, we can write (3.1) (times ¢') 68) ety = ely +B) = Oe Since Q and ¢! are functions of x only, we can now integrate both sides of (3.9) with respect to x to get ‘This is the general solution of (3.1). It is worth while to memorize (3.9) since first- order linear equations occur frequently.348_ ORDINARY DIFFERENTIAL EQUATIONS Example 1. Solve x*y’ — 2xy = I/x. In the form of (3.1), this is From (3.9), we get etn gate Example 2 Radium decnys to radon which decays lon which decays to pon. Ifa = 0, seme ‘is pure radium, how much radon does it contain at time 1? ™ Let = number of radium atoms at ¢ umber of radium atoms at time f, umber of radon atoms at time f, 4, and 23 = decay constants for Ra and Rn, As in Section 2, we have for radium EM a AN Ny = Noe * ‘The rate at which radon is being created is the rate at which radium is decaying, namely ZN) oF ZyNoe "But the radon is also decaying at the rate ZN. Hence, Dl a AN, Aga, oF 4 4,N, AN, = Noe a This equation is of the form (3.1), and we solve it as follows: t= fica t G.10) =ANo foc dee HNO seed LINEAR FIRST-ORDER EQUATIONS 349, if Ay # Aa. (Por the case Ay = Az, see Problem 19.) Since Np =0 at ¢=0 (we as- sumed pure Ra at ¢ = 0), we must have ‘Substituting this value of ¢ into (3.10) and solving for N3, we get 4sNo (ert, N= PROBLEMS, SECTION 3 Find the general solution of exch of the following differential equations. LL ytyee 2 ry thy=l 3 dy +Qxy— x07") de = 0 4 ey ty = 209? A pee 6 fF ay 2 (bey + dey a tee 8 (elnnytyelns 9. 12) dy — (ay +2) ts 10, yf ty tnh r= 2 My ty cose sin de v. cosy -rtany Ik det (re) dy 4. Hint for Problems 12 ta 14: Solve for x in terms of . 18, Water with a small salt content (5 Tb in 1000 gal) is flowing into a very salty lake at the rate ‘of 4 10° gal per hr. The salty water is flowing out at the rate of 10° gal per br. If at some time (say # = 0) the volume of the lake is 10° gal, and its salt content is 107 1b find the salt Content atime f. Assume that the satis mixed uniformly with the water in the lake at all times 16, Find the general solution of (1.2) for an RL circuit (I/C= 0) with V = Vo cos ox (@ = const). 17, Find the general solution of (1.3) for an RC circuit (L ), with V = Fo €0s or. 18. Do Problems 16 and 17 using V = Vee, and find the solutions for 16 and 17 by taking real parts ofthe complex solutions 19, If dy = 4g = Aim 10), then f e294 de =f dt Pind Ni For this case 29, Extend the radioactive decay problem (Example 2) one more stage, that is let Zs be the

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5835)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brene Brown

  Rating: 4 out of 5 stars

  4/5 (1093)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (852)
• • Magazines
  • Podcasts
  • Sheet music
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (612)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1720)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (590)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1194)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (903)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (541)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2107)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (350)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1029)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1872)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (824)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (122)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (443)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Tóibín

  Rating: 3.5 out of 5 stars

  3.5/5 (1948)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (405)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (809)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4772)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2259)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4214)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1930)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1905)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2526)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3973)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2410)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (74)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (880)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (105)