0. Thus, by the definition of change of sign given, in Chapter 1, f changes sign from negative to positive. Conversely, we can reverse this argument to show that if f changes sign from negative to posi- tive, then x9 isa transition point from P to N. The proof of (b) is similar. Let us retum to the race between the tortoise and the hare. Denote by 1(2) the tortoise’s position at time t and hy e(0) the hare’s pasitinn at time ¢ The transition “the tortoise overtakes the hare at time t,” means that there is an open interval / about fo such that:CHANGE OF SIGN ANDOVERTAKING 21 1. If tL and # fo, then fle) >e(0). If we graph f and g, this means that the graph of flies below that of g for 1 just to the left of fo and above that of g for ¢ just to the right of fo. (See Fig. 2.3.) Notice the role of the interval /. (We could have taken a slightly larger one.) It appears in the deflniuon because die hare may overtake dhe (orwulse at a Laver time f. Point at which the tt (hare) tortoise overtakes the hare a 23 The tortoise overtakes the hare at t> T and the hare overtakes the w tortoise at fy. We now state the definition of overtaking for general functions. Definition Let f and g be two functions, A the set of x (in the domain of f and of g) sich that f(x) < g(x), and B the set of x (in the domain of f and g) such that f(x) > g(x). If x9 is a transition point from A to B, we say that f overtakes g at x9. In other words, f overtakes g at Xo if there is an interval / containing x» such that (f and g are defined on J, except possibly at x9) and 1, Forx in J and x x9, f(s) > a(x). We call an open interval / about x9 for which conditions 1 and 2 are true an interval which works for f and g at xy; i0., 1 ic emall enough co that in to the left of xo, fis below g, while in J to the right of xo, f is above g. (See Fig. 2-4) Clearly, if a certain interval J works for f and g at xo, so does any open inwerval J contained in £, as long as it stil contains Xo. If the tortoise and hare both fall asleep and start running the wrong way when they wake up, we still say that “the tortoise overtakes the hare” if the hare passes the tortoise when going in the wrong direction. (See Fig. 2-5.) In the general situation, when f overtakes g at xo, the graph of f may actually be going downward. It is only the change in f as compared with the change in g which is important.22 CHAPTER 2: TRANSITIONS AND DERIVATIVES ote) i fh i i 1 SS Fig, 2.4 If J isan inter- hte > val that works (for f over- x % taking g at xq), s0 is any 7 smaller interval J. ' At) (tortose) ! Fig. 29 The worwhe 1 Neat tra 2 1 ae ‘overtakes the hare (f over- te takes g) at to. Worked Example 2 For the functions f and z in Fig. 2-6. tell whether f over- takes g, g overtakes f; or neither, at each of the points x1, x2, Xs, x4, and xs. When overtaking takes place, indicate an interval which works. Fig. 28 What overtal occur at the indicated points? Solution Neither function overtakes the other at x1. In fact, in the interval @,x3) about ¥;, (2) > f(x) for x both to the right and to the left of x1. At xs, f overtakes g; an interval which works is J = (xy, x3). (Note that the number 2 is what we called xo in the definition.) At x3, g overtakes f; an interval which works is (2,4). Alxg, f overtakes g again an interval which works is (3,5). Neither function overtakes the other at xs.CHANGE OF SIGN ANDOVERTAKING 23 The relations between the concepts of overtaking and change of sign are explored in Problems 8 and 9 at the end of the chapter. If, while the tortoise is overtaking the hare at fo, a snail overtakes the tor- toise at fo, then we may conclude that the snail overtakes the hare at fy. Let us state this as a theorem about functions. Theorem 2. Suppose f, g, and h are functions such that f overtakes g at xo and g overtakes h at xo. Then f overtakes h at Xo. Proof Let I, be an interval which works for f and g at xo, and let J be an interval which works for g and h at xo. That these intervals exist follows from the assumptions of the theorem. Choose / to be any open interval about x9 which is contained in both J, and /:. For instance, you could choose / to consist of all points which belong to both /, and J>. (Study Fig. 2-7, where a somewhat smaller interval is chosen. Although the three graphs intersect in a complicated way, notice that the picture looks quite simple in the shaded region above tho interval J. You should rotu to tho figure frequently as you read the rest of this proof.) 1.27 f overtakes a and g overtakes h at xo. Wo will chow that the intorval J works for f and A. We begin by assuming that XT and x xo then f(x) > h(x), so I works. The next result provides a link between the concepts of linear change and transition, We know that a faster object overtakes a slower one when they meet, Here is the formal version of that fact for uniform motion. Its proof is a good exercise in the algebra of inequalities. Theorem 3 Let f,(x) = m,x + by and fa(x) = myx + bz be linear func- tions whose graphs both pass through the point (xo, Yo). If ma > my, then fz overtakes f, at Xo. Proof Since yo = fu(xo) = mixo + by, we can solve for by to get by Yo — m,xp. Substituting this into the formula for f,(x), we have f(x) ‘mx + Yo — miXo, Which we can rewrite as f(x) = my(x — Xo) * Yo. Similarly, we have fa(x) = mi;(x — xo) + Yo. From Fig. 2-8, we guess that it is possible to take J = (—s°,%). To finish the proof, we must show that x x implies f,(x) > fi(x)- y Thiel = milk ~ x3) + Yo (ee ¥o) Fig. 28 f; and fy both pass through (xo, yo) and , has larger slope.CHANGE OF SIGNANDOVERTAKING 25 Assume that x 0,x—xo ‘and mz — my have opposite signs so that (rms = my)(oe = 0) <0 a(x — X0) — mx = ¥0) SO so a(x — X0) < mix — Xo) and myx — Xo) + Yo xp, we have x — xo > 0, 0 (mm — m,)(x —xq) > 0, A chain of ‘manipulations like the one above (which you should write out yourself) Jeads to the conclusion s(x) > 7,(x). That finishes the proof. Terminology It is useful to be able to speak of graphs overtaking one another. If G, and Gz are curves in the plane which are the graphs of functions /,(x) and f(x), we will sometimes say that G» overtakes Gy at a point when what is really meant is that f; overtakes f,. Thus, Theorem 3 may be rephrased as follows (see Fig. 2-9): If the line 1, with slope m; meets the line J; with slope m, at (0,0), and if my > my, then 1; overtakes /, at Xo, X13; overtakes C at x21C overtakes /> at x3. Za i Fig. 29 /, overtakes /; at26 CHAPTER 2: TRANSITIONS AND DERIVATIVES (This statement only applies to those lines which are graphs of functions— we may not speak of a vertical line overtaking or being overtaken by any- thing.) Solved Exercises A. Where dove 1/se change sign? 5. Let f(x) = 3x + 2, g(x) = x + 2. Show in two ways that f overtakes g at 0: (a) by the definition of overtaking: (b) by Theorem 3. 6. Let f(x) = ~3 and g(x) = —x?. At what point or points does f overtake g? Construct an interval which works. Sketch. 7. Let f(x) = 2x? and g(e) = Sx — 3. Show that g overtakes f at x = 1. Is (-, 3) an interval which works? Is it the largest one? mx for various values of m. Can you find a transition point on the “‘m axis” where a cer- tain change takes place? 6. Let fe) = x? — 2x ~ 3 and g(x) = 2x — 2. Where does g overtake f? Find an interval which works. Sketch. 7. Let f(x) = x? — x and g(x) = 2x. At what point or points does f overtake 87 Does (3,0) work for any of these points! It not, why not? Sketch. 8, Let f(x) take g? —x? + 4 and g(x) = 3x — 2. At what point or points does f over- ind the largest interval which works. Sketch. 9. Let f(x) = 1/(1 — x) and g(x) =-x +1. (a) Show that f overtakes ¢ at x = 0. Find an interval which works. Sketch. (b) Show that f overtakes g at x = 2. Is (1,3) an interval which works? What is the largest interval which worke? The Derivative ‘The derivative was defined in Chapter 1. To rephrase that definition using the Ianguage of transitions, we shall use the following terminology. Let fhe a func-THE DERIVATIVE = 27 tion whose domain contains an open interval about xp. Let 4 be the set of numbers m such that the linear function f(xo) + m(x — x9) (whose graph is the line through (x0.f(xa)) with slope m) is overtaken by f at xa. Let B be the set of numbers m such that the linear function f(x) + m(x — x9) overtakes fat xo. Theorem 4 A number mg is a transition point from A to B if and only if mg is the derivative of f at xo. Proof First, suppose that mg is a transition point from A to B. Thus there is an open interval [ about my such that (i) if m, J and m, mo, then m is inB but not in A. Let m d28 CHAPTER 2: TRANSITIONS AND DERIVATIVES etd ifx<-1 (b) fe) = XQ WT 2x+3 ifx>-! wxSo (© fa) = . x itx>0 12. Is the following an acceptable definition of the tangent line? “The tangent line through a point on a graph is the one line which neither overtakes nor is overtaken by the graph.” If so, discuss. If not, give an example. Problems fOr PW 1 Describe the following as transition points. (a) Turning on a light, () 100° Centigrade. (c) An aircraft landing. (a) Sioning a contract. ‘What transition points can you identify in the following phenomena? (@) Breathing. (b) A heart beating. (©) Blinking. (@) Walking. (e) A formal debate. (O) Firefighting. (g) Marriage. (h) Receiving exam results. (@_ Solving homework problems. . Let A be the set of areas of triangles inscribed in a circle of radius 1, B the set of areas of circumscribed triangles. Is there a transition point from 4 to B? (You may assume that the largest inscribed and smallest circumscribed triangles are equilateral.) What happens if you use quadrilaterals instead of triangles? Octagons? |. For each of the following pairs of sete, find the transition points from A to Band from B to A. (a) A =(0,1);B = (~2,5) and (1,3). (e) those x for which x? <3; B = those x for which x? > 3. (c) A= those x for which x* <4; 8 = those x for which x? > 4. (W) A— those @ for witch die equation x? +a = O nas two real Toor; those @ for which the equation x? +a = 0 has less than two real roots. those x for which V7 — 4 exists (as a real number) those x for which Vx" — 4 does not exist (as a real number). . For which values of n (positive and negative) does x" change sign at 0?PROBLEMSFORCHAPTER2 29 6. For each of the following pairs of functions, find: 1, The set A where f(x) g(x). 3. The transition points from 4 to B and from B to A. 4, Intervals which work for f overtaking g and g overtaking f. ‘Make a sketch in each case, (@) fe) =x +25 gle () fe) =3 2x42; gQ)= () f(x) = -2x? + 4x — 3; gC @ se) = 242041; )=22 1 (©) fe) =x + ae; g(x) =x? - 2x (0 1G = Ux? #0: 2G = x? +1 © fe) = 1x? (#0); (x)= [1 +2)] — 2 #-2) 7. For which values of m does f(x) = m(x ~ 1) + 1 overtake g(x) = x? at 17 8. Show that f changes sign at xo if and only if f overtakes or is overtaken by the zero function (g(x) = 0 for all x) at x9. (This problem and the next one show that the concepts of overtaking aud sign change van Le defined i terms of one another.) 9. Let f and g be functions, and define h by h(x) = f(x) — g(x), for all x such that f(x) and g(x) are both defined. Prove that f overtakes g at xo if and only if h changes sign from negative to positive at xo, In Problems 10 to 17, let fle) = 2x? = Sx +2, (x) = Gx? + 2x, n(x) = 3x? Fx +5, k(x) = 3x — 4, ana Mex) = 2 +3 10. Pind the desivative of /(x) + x(x) atx = ive of 3f(x) — 2h(x) at x 12, Find the equation of me tangent line to (@ f@)atx= (b) g(x) atx = -2 (eo) he) ate = 100 (a) EG) at » = —108 (cketch) 13. Where does (x) overtake k(x)? 11. Find the deri 14. Where dase I(x) avertake the tangent line te (x) at x = —17 15. For what real number c is the line y = ck(x) parallel to the tangent line to Fox) atx = 2? 16. Where does g(x) + i(x) overtake f(x) + k(x)? What are the derivatives of g +/ and f+ at these points?17, 18. 19, 20, 2. 22. 23. CHAPTER 2: TRANSITIONS AND DERIVATIVES Let f(t) = 21? — St + 2 be the position of object 4, and let h(t) = 307 +1+ 3 be the position of object B. (a) When is 4 moving faster than 5? (b) How fast is B going when 4 stops? (©) When does B tum around? A set of numbers 5 is called convex if, whenever x, and xz lie in S and x; oxo), then F(x) overtakes g(x) at x9. [#4im: Use a line with slope in the interval (¢'(%0), (%o))]- (b) Prove that, if f'(xo) <2’(xo), then f(x) is overtaken by g(x) at xo. (c) Give examples to show that, if ’(xo) = g'(xo), then it is possible that f(x) overtakes g(x) at xo, or f(x) is overtaken by g(x) at xo, or neither. (a) Solve Problem 6 by using (a) and (b). Explain how Fig, 2-10 illustrates the definition of the derivative of y J Fig. 2-10 This illustrates the derivative?3 Algebraic Rules of Differentiation We shall now prove the sum, constant multiple, product, and quotient rules af differential calculus. (Practice with these rules must be obtained from a standard calculus text.) Our proofs use the concept of “rapidly vanishing functions” which we will develop frst. (The reader will be assumed to be faniliar with absolute values.) Rapidly Vanishing Functions We say that a function f vanishes at xo, or that x9 is a root of f, if fle) = 0. IF we compare various functions which vanish near a point, we find that some vanish “more rapidly” than others. For example, let f(x) = x — 2, and let g(x) = 20(x — 2). They both vanish at 2, but we notice that the derivative of g vanishes at 2 as well, while that of f does not. Computing numerically, we find FO) = 1 20 f() = -1 20 fQ1) = 0.1 0.2 #09) = -0.1 0. (2.01) = 0.01 0.002 £(1.99) = —0.01 0.002 £2.01) = 0.001 0.00002, ‘f(1.999) = -0.001 0.00002 £(2.0001) = 0.0001 —_¢(2.0001) — 0.000002 (2.9999) = ~0.0001 _g(1.9999) = 0.000002 ‘As x approaches 2, g(x) appears to be dwindling away more rapidly than (x). Guided by this example, we make the following definition. Definition We say that a function r(x) vanishes rapidly at x9 if r(%0) and r'@%0) = 0. The following theorem shows that rapidly vanishing functions, as we have defined them, really do vanish quickly. This theorem will be useful in proving 3132 CHAPTERS: ALGEBRAIC RULES OF DIFFERENTIATION properties of rapidly vanishing functions, as well as in establishing the connec- tion between transitions and limits (see Chapter 13). Theorem 1 Let r be a function such that r(xq) = 0. Then r(x) vanishes ropily at ¥ = x if and only if, for every positive mumber c, there is an open interval I about x9 such that, for all x #xq in I, WW(x)i e(2 — xo), since e(+ — x9) overtakes r(x). We may rewrite these two inequalities as —c(xo — x) < r(x) < exo — x), ic. mex = x91 < r(x) < ete — x9, since be — x9l = xo — x whenx xp in J. Our overtakings imply that r(x) > “exe = x9) and r(x) < ex — x9), $0 el — x9) 0, there exists an interval / such that, for all x in J, x #Xo, In(x)I < etx — x9l.HAPIDLY VANISHING FUNIIONS 33 Reversing the steps in the preceding argument shows that, for all €> 0, the line y = —e(x — x9) is overtaken by the graph of r(x) at xo, while the line y = e(x — xo) overtakes the graph. Thus the set 4 in the definition of the derivative contains all negative numbers, and R contains all positive num. bers; therefore, 0 is a point of transition from A to B. Thus r'(xo) = 0, so 7 vanishes rapidly at x9. The next theorem shows the importance of rapidly vanishing functions in the study of differentiation. Theorem 2 A function f is differentiable at x9 and my = f'(eo) if and only if the function r(x), defined by 1x) = FX) ~ [70%0) + mo(x - X0)] vanishes rapidly at x». The function r(x) represents the error, or the remainder involved in approximating f by its tangent line at x9. Another way of stating Theorem 2 is that f is differentiable at xo, with derivative f"(co) = mo, if and only if f can be written as a sum f(x) = flo) + mo(x — x9) + r(x), where r(x) vanishes rapidly at x9, Once we know how to recognize rapidly vanishing functions, Theorem 2 will provide a useful test for differentiability and a tool for computing derivauves, Notice that Theorem 2, like Theorem 1, is of the “if and only if” type. Thus, it has two independent parts. We must prove that if f(x) is differentiable at xq and mo = f"(xo), then r(x) vanishes rapidly at xo, and we must prove as well that if r(x) vanishes rapidly at xo, then f(x) is differentiable at xo and img = f'(x0). (Each of the two parts of the theorem is called the converse of the other.) In the proof which follows, we prove the second part first, since it is convenient to do so. Proof of Theorem 2 Suppose that r(x) = f(x) — fo) — mo(x — xo) vanishes rapidly at x9. We will show that j"(xo) = Mm. We first show that, for m > mo, the line through (xo, f(xo)) overtakes the graph of f at xo. (Refer to Fig. 3-2 as you read the following argument.) To begin, note that if m > mo, then m — mg > 0. Since r'(xo) = 0, the line through (o,7(%o)) with slope m — mo overtakes the graph of r at x9. Since r(&o) = 0, the equation of this line is y = (m — mo)(x — xo). By the definition of overtaking, there is an open interval J around xo such that (x me y= 169) + max = x0) {tangent tne} hx) = Fle) ~ [fbre) + mb ~ el] ™ Line with siope m ~ me >0 3-2 Illustrating the proof of Theorem 2. (om — moNx — x0) x9 in J implies (mm — moN(x — x0) > f(&) — Fl%0) — mo(x — x0) (Ad) Adding f(x0) + mo(x ~ x9) to both sides of (A,) and (A2) gives Fo) + m(x —X0) < f(x), forx f(x), forx>xo int In other words, the line y = /(xq) + m(x — xo) overtakes the graph of fat Xo: Recall that m was any number greater than mo. Similarly, one proves that, for m < mo, the line y = f(xo) + ‘m(x — x9) is overtaken by the graph of fat xo, so f'is differentiable at xo, and mg must be the derivative of f at xo by the definition on p. 6 and Theorem 4, p. 27. This completes the proof of half of Thearem 2 Next, we assume that f is differentiable at xo and that 29 = f"(xo). Reversing the steps above shows that the line y = c(x — xo) overtakes the graph of r for ¢ > O and is overtaken by it for c <0. Clearly, r(x9) = 0, s0 vanishes rapidly at x9.‘THE SUM AND CONSTANT MULTIPLE RULES 35. Solved Exercises* I. Let gi(x) = x, go) = x7, g3(x) = x°. Compute their values at x = 0. 0.01, 0.002, and 0.0004. Discuss. 2. Prove: (@) If fand g vanish at xo, s0 does f+ g. (b) If f vanishes at xo, and g is any function which is defined at xo, then fg vauislies at.29. 3. Prove that (x — x9)? vanishes rapidly at x9. Exercises 1. Fill in the details of the last two paragraphs of the proof of Theorem 2. 2. Do Solved Exercise 3 by using Theorem 1. 3. Prove that f(x) =x? is rapidly vanishing at x 4. Let g(a) be a quadratic polynomial such that 6(0) — (2) Can g(x) vanish rapidly at some integer? (W) Can g@)v 5. The polynomials x? — 5x? + 8x — 4, x? — 4x? + Sx — 2, and x? — 3x7 + 3x — 1 all vanish at x9 = 1. By evaluating these polynomials for values of x Very near 1, on a calculator, try to guess which of the polynomials vanish rapidly at 1. (Factoring the polynomials may help you to understand what is happening.) lsh kapidly at more than one point? The Sum and Constant Multiple Rules The sum rule states that (f + g)\(x) = f'(x) +g'(x). To prove this, we must show that the remainder for f+, namely {lee) + 8D} — [{fl0) + e(%0)} — {F'G0) + 2’ Geo) HO — x0)] vanishes rapidly at xo, for then Theorem 2 would imply that f(x) + g(x) is dif- “Solutions appear in the Appendix.36 CHAPTER 3: ALGEBRAIC RULES OF DIFFERENTIATION ferentiable at xo with derivative my = f"(xo) + g'(%o). We may rewrite the remainder as LAG) — Flo) — f'CeoXx — x0)] + [a(x) — 80) — (ox — x0) ] - By Theorem 2, each of the expressions in square brackets represents a function which vanishes rapidly at x9, so we need to show that the sum of two rapidly vanishing functions is rapidly vanishing. ‘The constant muluple rule states that (af)'(x) = af (x). [he proot of this rests upon the fact that a constant multiple of a rapidly vanishing function is again rapidly vanishing. Theorem 3 proves these two basic properties of rapidly vanishing functions. Theorem 3 1. If riGe) and r2(x) vanish rapidly at xo, then so does ry(x) + ra(x)- 2. If r4(x) vanishes rapidly at xo, and a is any real number, then ar(x) vanishes rapidly at x9. Proof 1. Lat r(x) = r(x) + r2(3), where r(x) and r(x) vanish rapidly at Xo. By Solved Exercise 2, (xo) = 0; we will use Theorem 1 to show that r(x) vanishes rapidly at xo. Given > 0, we must find an interval / such that I(x)! < eux — xql for all x # xo in J. Since r4(x) and ra(x) vanish rapidly at xo, there are intervals J, and J, about xo such that x % x9 in J, implies that Irs(x)| <(¢/2) be —xo|, while x + xo in J> implies Ira(x)I< (€/2) be — xol. (We can apply Theorem 1 with any positive number, in- cluding ¢/2, in place of e.) Let J be an interval containing xo and contained in both 7, and J. For x # xo in J, we have both inequalities: Iry(x)| < (€/2) be = xpi and Ira(x)I<(€/2) lx —xo|. Adding these inequalities gives € Fix — x9! = etx —x9l Irs@x)I+ ira):

0, we apply Theorem 1 to 7; vanishing rapidly at x9 to obtain an interval 7 about xo such that In(x)! < (€/lal) le — xo for x # xo in I. (If a= 0, ar(x) = 0 is obvionsly rapidly vanishing, sa we need only deal with the case a #0.) Now we have, for x #9 in J,‘THE SUM AND CONSTANT MULTIPLE RULES 37 WGI = laryGo)l = tal Ir) < lal é be —x0l = elx —xol or n(x)! 0, while fis overtaken by (1/a) g at xo if a 0 and an interval J; about xo such that, for x in J, we have “B 0, since r(x) vanishes rapidly at xo, we can find an inter- val [about xo such that, for x # xp in, ir(2)! <(€/B) be — X91, where B is the bound from step 1. Now we have, for x # x9 in J, If(x)r()I = UfCailr(@) < B (eB) lx —xol = elx — xol. Thus, by Theorem 1, f(x)r(x) vanishes rapidly at xo. ‘We can now deduce the product rule from Theorem 5 by a computation.40 CHAPTER 3: ALGEBRAIC RULES OF DIFFERENTIATION Theorem 6 Product Rule. If the functions f(x) and g(x) are differentiable at Xo, then $0 ts the function f(x)e(x), and irs dertvative at Xo is F'%o)8%o) + 10)" %o) Proof By Theorem 2, f(x) = flcco) + F"Go)(e — x0) + rs(x) and g(x) = (0) + g(xo)(x — x0) + ra(x) where ra(x) and ra(x) vanish rapidly at xo. Multiplying the two expressions gives Fee) = Fore) Melxe) + 2'Cea Me — 0) + ra] + F'Co)x ~ xo) ato) + 8’ eax — x0) + ro(2)] + ri(x)[e(%o) + #'o)(x — x0) + ra()] = F0)8(xo) + ["(%o)s 0) + Flex0)e"(xo)] ( — Xo) + Fexo)ra(x) + ru()a C0) + F'Go)s' Go) ~ x0)” + FC oN& —x0)rale) + rue’ %o)(& — x0) + @)raX) Now each of the last two lines in the preceding sum vanishes rapidly at xo: F(&o)ra(x) + r1(x)g(Xo) is a sum of constant multiples of rapidly vanishing functions (apply 1heorem 3); 7"(xo}e"(XoXX — Xo)® 18 a constant multiple of (x — x0)?, which is rapidly vanishing by Solved Exercise 3; each of the next two terms is the product of a linear function and a rapidly vanishing function (apply Theorem 4), and the last term is the product of two rapidly vanishing functions (apply Theorem 4 again). Applying Theorem 3 ‘to the sum of the last two lines, we conclude that FO) (H) = Fxo)a (Xo) + [F'Go)e(%0) + Fo)’ %o)] (& — x0) + r(x), where r(x) vanishes rapidly at x9. Theorem 2 therefore shows that ‘(x)g(x) is differentiable at xo with derivative f'(xo)e xo) + F(xo)e’ (xo). Note that the correct formula for the derivative of a product appeared as the coefficient of (x — xe) in our computation; there was no need to know it in advance. Now that we have proven the product rule, we may use one of ite impor- tant consequences: the derivative of x” is nx"~1 (see Solved Exercise 7).THEQUOTIENTRULE 41 Solved Exercises 6. In step 1 of the proof above, estimate how large B must be so that -B-< FG) 0. (The case g(x) <0 is discussed at the end of this proof.) Then a bow tie argument like that on p.39 chowe that a(x) > £ g(xo) for x in some interval / about x9. (See Fig. 3-4 and Solved Exercise 9) Since 0 <4¢(xo) g' x.) Ialine with slope <9") Tangent ine Interval which worss for both overtakings Fig. 34 g(x) Is bounded below near xo by $9(x0). a "= Fae) ator vanishes rapidly at xp. Collecting terms over a common denominator, we eet r(x) Ae(%o)? — #(x)e(xo) + 2'(xo)e xXx ~ Xo) } a © 8G)8%o)* Since 1/le(x)e(xo)"1 lies between the values 0 and [2/efsxa)] [1/etxa)*]. it suffices to prove that the expression in braces, AG) — so)? Bs Gru) + e'Gru)sOG >u) vanishes rapidly at zero, Setting x = xo in r,(x), we get r,(xo) = 0. We aust show that 7}() = 0. However, now that we have proven the sum, constant multiple, and product rules, we may use them: FA) = 8" )a(0) + 8 eo)a" Ne — x0) + 8'o)e(e) wherever g is differentiahle Setting ¥ = v9, we get ri(o) = 0, <0 r4() vanishes rapidly at xo, and so does r(x). That finishes the proof for the case g(X0) > 0. If 30) < 0, we can apply the previous argument wo —g. Namely, writePROBLEMS FOR CHAPTERS = 43. ll ax) 8) Since g(x) is differentiahle at 9, 60 i¢ ) [existence of multiplicative inverses] . 5. Forall x,y, and z, x+(y+z)=x-y +x+z [distributivity) On the basis of multiplication axiom 4, we can define the operation of division by 4-26) (yo) ‘One often writes xy forx +).ADDITION AND MULTIPLICATION AXIOMS 47. Implicit in one niles for addition and multiplication is that the numbers x +y and xy are uniquely specified once x and y are given. Thus we have, for example, the usual rule of algebraic manipulation, “if x = z, then xy =zy.” In fact, x =z meaus Urat x and 2 ate dhe sate nunber, 99 iat utultiplying eacle of them by y must give the same result. In principle, one could prove all the usual rules of algebraic manipulation from the axioms above, but we will content ourselves with the few samples siven in the exercises below. ‘A final remark: we can define 2 = 1+ 1.3=2+1 obtain the fractions. As usual, we write x? for x +x, x° for x’ and via division etc. Solved Exercises* 1. Prove that a= 0 ~0 by mulliplying die equality 0+ 0 ~0 by a. 2. Prove that (x + y)? =x? + Ixy +y?, 3. Prove that 2+3=6. 4, Prove that (—x)-y =—(ay). a,¢_ adtbe 5. Prove that +77 bd ifb #0 andd #0. Exerci Prove the following identities. L (w= 2. (xy =x? - dy ty? 3. TE a —Y? 4 ®-@ (040,440) eo 5. Gre @#0040) i a,b 6. ro(G+p) $4242 (@#0,640) a (2) =-1 #0) *Solutions appear in the Appendix.48 CHAPTER 4: THE REAL NUMBERS Order Axioms From now on we will use, without further justification, the usual rules for alge- braic manipulations. The previous exercises were intended to convince the reader that these rules can all be derived from the addition and multiplication axioms. We turn now to the order axioms. MI, Order Axtoms There ts a relaion “$< such dua, for certain patty « and y of real numbers the statement “x i true [compara bility]. 5. Ifx x (y is greater than or equal to x) if x x (vis strictly greater than x) if x cy. 8. Let a and b be numbers such that, for any number ¢ with ¢ 0 for all x.‘THE COMPLETENESS AXIOM = 49 Exercises Prove the following statements. Thi 12, Wa> 1, then 1fa<1 13. Ifa? 0 and <0, then (a+)? <(@ by. 15. If@ and b have the same sign, then (a + b)? > (@—b)?. 16. Ifx 0; and (2) a is less than every positive real number. Prove that there are no infinitesimal real numbers.* The Completeness Axiom We need one more axiom to guarantee that irrational numbers exist. In fact, the rational numbers (i.¢., quotients of integers) satisfy all of the axioms in I, II, and m To motivate the last axiom, we consider the problem of defining V2. Con- sider the set $ consisting of all numbers such that 00, then x? >0. (b) x and y have the same sign if and only if xy > 0.52 uN 12. 13, CHAPTER 4: THE REAL NUMBERS (©) Ifa 2ifa>0. ‘Use me axioms for adaition and multiplication to prove me folowing: 4 (@) 3-$=4-3 () ¥=64+1 (ce) 46245-6431 @) B=aR (©) -—(-(@)) +2=0 (f) -@+b)=-a-b @_c_e-c ws 55 #0 @ Let 5 be the set of x such that x* < 10. Show that 5 is convex and describe 5 a9 an interval. Discuss how this can be used to prove the existeuve of YT0. ). Prove the following inequality using the order axioms: 20? favo 1: ¢ 2 Prove that V2 < V3. [Hint: Assume /2 > V/3 and derive a contradiction.) ) » wherea 0, prove that there is a positive integer n such that 0< I/n f(x») there is an open interval J about xp such that, for those x in J which also tie in D, fe) <2. If f is continuous at every point of its domain, we simply say that f is continuous or fis continuous on D. Waring It is tempting to define a continuous motion f() as one which never passes from (to) to f(t) without going through every point be- 54‘THE DEFINITION OF CONTINUITY 55 tween these two. This is a desirable property, but for technical reacone it is not suitable as a definition of continuity; (See Figure 5-4). The property by which continuity is defined might be called the “principle of persistence of inequalities”: fis continuous at xo when every strict inequality which is satisfied by f(a) continues to be satisfied by f(x) for x in some open interval about xo. The intervals J and J in the definition may depend upon the value of cy and ¢2. The definition of continuity may also be phrased in terms of transitione using the idea of Salued Fearciee 7, Chapter 13 Another way to paraphrase the definition of continuity is to say that f(x) is close to f(xo) when x is close to x9. The lines y = ¢, and y = cy in Fig. 5-1 provide a measuse of closeness. The following example lustrates this idea. Worked Example I The mass y (in grams) of a silver plate which is deposited fon a wire during @ plating process is given by a function f(x), where x is the length of time (in seconds) during which the plating apparatus is allowed to operate. Suppose that you wish to deposit 2 grams of silver on the wire and that 143) = 2. Being realistic, you know that you cannot control the time precisety, but you are willing to accept the result if the mass is less than 0.003 gram in error. Show that if f is continuous, there is a certain tolerance r such that, if the time is within r of 3 seconds, the resulting mass of silver plate will be acceptable. Solution We wish to restrict x so that /() will satisfy the inequalities 1.997 < (8) < 2.003. We apply the definition of continuity, with xo = 3, cy = 1.997, and cy = 2.003. From condition 1 of the definition, there is an open interval containing 3 such that 1.997 < f(x) for all x € I. From condition 2, there is. such that f(x) < 2,003 for all x €J. For r less than the distance from 3 to either endpoint of J or J, the interval (3 — r,3 + 1] is contained in both and J; for x in this interval, we have, therefore, 1.997 0 Show that g is not continuous at x; ‘Solutions appear in the Apps!56 CHAPTERS: CONTINUITY 2. Let fix) be the absolute value function, f(x) = \x|. Show that f is contin- uous at x9 = 0. 3. Let f be continuous at x» and suppose that f(x) ¥ 0. Show that 1//(x) is defined on an open interval about xo. 4, Decide whether each of the functions whose graphs appear in Fig. 5-2 is continuous. Explain your answers. rc) tl Ae aate 7 Fig. 5-2 Which functions (o) (a) are continuous? Exercises 1. Let f(x) be the step function defined hy -1 ifx0 Show that f1s discontinuous at 0. 2, Show that, for any constants @ and 6, the linear function f(x) = ax + b is Gontinuuus at.ag 2. 3. Let f(x) be defined by x41 ifx<] F@)=4? — ifl O such that x° + 8x? +x < 1/1000 if 0-< x<6. 8. Let f be continuous at x9 and A a constant. Prove that f(x) + 4 is contin- uous at xo. Exercises 7. Why can’t we ask whether the function (x? atl? 1)/(? — 1) is continuous 8. Let flx) Can you define f(0) so that the resulting function is continuous at all x? 9, Find a function which is continuous on the whole real line, and which is differentiable for all x except 1,2, and 3.(A sketch will do.)60 CHAPTERS: CONTINUITY 10. In Solved Exercise 7, show that 5 = 1/2000 works; i., x? + 8x? +x < 1/1000 if 0 d to the reader. We look at the set 5 consisting of all those x in [a,b] for which f d. By the continuity of fat, we must have f(x) >d for all x in some open interval about c; but this contradicts the fact that f c, this contradicts the statement that S = [a,c]. Thus, case 2 cannot occur. The only possibility which remains, and which therefore must be true, is that $= [a,b] . Thus, f(@) d, or f(@) > d and f(b) 0. Prove that f must have a critical point (2 point where f’ vanishes) somewhere on the interval (—1,1). 15, Let f(x) = x* — x? + 35x — 7, Prove that f has at least two real roots. 16. (a) Give a direct proof of the Intermediate Value Theorem. (b) Use the intermediate value theorem to prove the alternative version. Increasing and Decreasing at a Point Intuitively, if we say f(x) is increasing with x at xo, we mean that when x is in- creased a little, fc) increases and when x is decreased a little, fie) decreases. Study Fig. 5-5 to see why we say “a little.”INCREASING AND DECREASING ATAPOINT 63 Fig, 55 x; is a little greater than xo and f(xy) > Flegli az 10 @ lot greater than xo and now fxs) < F(x). Definition Let f be a function whose domain contains an open interval about x9 1. fis said to be increasing at x9 if the graph of f overtakes the hori- zontal line through (xo, f(%o)) at xo. 2. fis said to be decreasing at xo if the graph is overtaken by the line at x9. (See Fig. 5-6.) 1 tweeter increasing nor frsincreasing decreasing tx a Crea Fis decreasing at x: pyri TK tis increasing 1 at t . Fig. 56 Where is f in- ry me creasing and decreasing? Thus, f is increasing at xo if and only if f(x) — (x0) changes sign from negative to positive at xo. Similarly f is decreasing if and only if f(x) — f(o) changes sign from positive to negative at x9. If we substitute the definition of “overtake” in the definition of in- creasing, we obtain the following equivalent reformulation, Definition’ Let f be a function whose domain contains an open interval about xo. 1, f is said to be increasing at xo if there is an open interval J about Xo such that:64 CHAPTERS: CONTINUITY @ FG) Feo) forx>xo int 2. fis said to be decreasing at xo if there is an open interval J about Xo such that: @ fee)> Fox) forx xq ns Pictorially speaking, a function fis increasing al x9 when moving x a liwle to the left of xo lowers f(x) while moving x a little to the right of x9 raises f(x). (Che opposite happens if the function is decreasing at xo.) Worked Example 3 Show that f(x) = x* is increasing at xo Solution Let I’be (1,3). If x xo isin J, then 2 4 = x3, We have verified @) and (i) of part 1 of Definition’, eo f ie inoreasing at 2. The transition definition of the derivative of f at xo tells us which lines overtake and are overtaken by the graph of f at Xo. This leads to the next theorem. Theorem 3 Let f be differentiable at xo, 1. IF"(&o) > 0, then f18 mereasing at Xo. 2. If fo) <0, then f is decreasing at Xo. 3. If f'(&o) = 0, then f may be increasing at xo, decreasing at xo, or neither, Proof We shall prove parts 1 and 3; the proof of part 2 is similar to 1. The definition of the derivative, as formulated in Theorem 4, Chap- ter 2, imcludes the statewent drat any line Uurough (x9, /(G%0)) Whose slope is less than "(xo) is overtaken by the graph of f at xo. If f"(xo)> 0, then the horizontal line through (xo, f(0)), whose slope is 0, must be overtaken by the graph of fat xo; thus, 'is increasing at x9, by definition. The functions x°, —x?, and x?, all of which have derivative O at x» = 0, establish part 3; see Solved Exercise 15. The following is another proof of Theorem 3 directly using the original definition of the derivative in Chapter 1.INCREASING AND DECREASING ATA POINT 65 Alternative Proof From condition 1 of the definition of the derivative (p. 6), we know that if m < f"(xo), then the function f(x) — [f(%0) + m(e — x)] changes cign from negative to positive at xy. Thus if 0 < F'(&o), then we may choose m = 0 and conclude that f(x) ~f(zo) changes sign from negative to positive at xo; that is, fis increasing at x9. This estab- Ushes part 1 of the theorem. Part 2 is similar, and the functions x, —x°, and x" establish part 3 as in the first proof. Worked Example 4 Is x* — x° — 2x? increasing or decreasing at 2? Solution Letting f(x) = x* — x? — 2x®, we have f'(e) = Sx* — 3x? — 4x and f'(-2) = 5(-2)* — 3(-2)? — 4(—2) = 80 — 12 + 8 = 76, which is positive. By ‘Theorem 3 part 3, x* — x? — 2x? is increasing at —7. Theorem 3 can be interpreted geometrically: if the linear approximation t0 f at xy (that is, the tangent linc) is an increasing or decicasing funutivi, test F itself is increasing or decreasing at x9. If the tangent line is horizontal, the behavior of f at xo is not determined by the tangent line. (See Fig. 5-7.) ua '(x,) >0; fis increasing at x; #'lx2) <0; fis decreasing at x2 #'Uxs) = f'xa) =0; fis neither increasing nor decreasing at x3 and xq F'lxe) =0: fis decreasing at ve f'lxe) =0; fis increasing at x, We can also interpret Theorem 3 in terms of velocities. If f(¢) is the posi- tion of a particle on the real-number line at time f, and (fo) > 0, then the particle is moving to the right at time fo; if f"(fo) <0, the particle is moving to the left. Combined with the techniques for differentiation in Chapter 3, Theorem 3 provides an effective means for deciding where a function is increasing or decreasing. Solved Exercises 13. The temperature at time x is given by f(x) = (x + 1)/(x — 1) forx # L.Isit getting warmer or colder at x = 0?66 CHAPTERS: CONTINUITY 14, Using Theorem 3, find the points at which f(x) = 2x? — 9x? + 12x +5 is increasing or decreasing. 15. Decide whether each of the functions x*, creasing, or neither at x = 0. and x? is increasing, de- Exercises 17. If f(0) = 5 — 4 + 227 isthe position of a particle on the real-number line at time ris it moving to the left or right at ¢ = 1? 18, Find the points at which f(x) = x? — 1 is increasing or decreasing. 10. Find the pointe at which v3 — 2y? 4 2y is incrnasing ar decraasing, 20. Is f(x) = 1/(x? + 1) increasing or decreasing at x = 1, -3, 3, 25, -36? 21. A ball is down upwad wid The ball’s height above the ground at time t is h(t) the bail rising? When is it falling? 22. (a) Prove that, if f(xo) = 0, then f is increasing [decreasing] at x9 if and only if f(a) changes sign from negative to positive [positive to nega- tive] at xo (b) Prove that f is increasing [decreasing] at xo if and only if f(x) — fo) ‘changes sign fram negative ta positive [positive to negative] at xo. initial velucity uf 30 ureters per second. 30t — 4.912. When is Increasing or Decreasing on an Interval Suppose that fis increasing at every point of an interval [2,5] . We would expect, ‘F(b) to be larger than f(a). In fact, we have the following useful result. Theorem 4 Let f be continuous on (a, b], where a f(a) (7) fla) is decreasing at each point of (b,c); Fle) <#(b) is neither increasing nor decreasing at a, b, ¢ 5-9 f is increasing at : * all points of (0,1), but #11) ie not greater than ene =i #(0). Proof of Theorem 4 We proceed in several steps. Step 1 Wa FG); td derive a contradiction. In fact, if f(a) > F(y) for some y &(@, 0), the continuity of f at a implies that f > f(y) on some interval [a,7). Choosing x in (a,r) such that x <,y, we have /(x) > f(»), contradicting step 1. Using the continuity of f at 5, we can prove in a similar manner that () flee.) for all X1 0, decreasing if f' <0). Proof By the intermediate valua thearem, f” is averywhere positive or everywhere negative on (a,b). Suppose it to be everywhere positive. Let X1 0 at all points of its domain, but f(1) = ~2 is not greater than f(~1) = 2. What is wrong here? 17. Let a g'(x) for all x in (0,8). Tf fla) = o(0), peowe that f(5) > eb) (Hint: You may assume that f— g is continuous at a and 5.) (b) Give a physical interpretation of the result of part (a), letting x be time 26. Let f be a polynomial such that f(0) = f(1) and '(0) > 0. Prove that f'(&) = 0 for some x in (0, 1). The Extreme Value Theorem The last theorem in this chapter asserts that a continuous function on a closed interval has maximum and minimum values. Again, the proof uses the complete- ness axiom. We begin with a lemma which gets us part way to the theorem. Boundedness Lemma Let f be continuous on [a,b]. Then there is a num- ber B such that f(x) b, and fis ‘continuous on [a,b], then there is some x in [2, 5] such that f(x) = x. 8, Give an example of (discontinuous) functions f and g on [0,1] such that f and g both have maximum values on [0, 1], but f+g does not. 9. Let f(x) = x17 + 3x4 — 2 and g(x) = Sx® — 10x + 3. Prove that there is a number xo such that f(zq) = g (za). 10. (2) ‘Prove that, if fis increasing at each point of an open interval J, then the set S consisting of all those x in / for which f(x) € (0, 1) is convex. (b) Show that S might not be convex if fis not increasing on J. 11, Show by example that the sum of two functions with the intermediate value property need not have this property. 12, Prove that the intermediate value theorem implies the completeness axiom. [Hint: If there were a convex set which was not an interval, show that you could construct @ continuous function which wakes on only the values 0 and 1,]6 Graphing In this chapter we continue our development of differential calculus by ana. lyzing the shape of graphs. For extensive examples, consult your regular calculus text. We will give an example not usually tound in calculus texts: how to graph a general cubic or quartic. Turning Points Points where the derivative f" changes sign arc called turning points of f; they separate the intervals on which f is increasing and decreasing. Definition Mf f is differentiable at xo and f’(x0) point of f. If f is defined and differentiable throughout an open interval con- taining xo, we call xo a turning point of f if f' changes sign at xo. 0, we call xq a critical If f' is continuous at xo, then a turning point is also a critical point (why?). A critical point need not he a turning point, however, as the function y =x3 shows. To investigate the behavior of the graph of fat a turning point, itis useful to treat separately the two possible kinds of sign change. Suppose first that ” changes sign from negative to positive at x9. Then there is-an open interval (2, 5) about xo such that ’(x) < 0 on (@,x9) and f"(x) > 0 on (xo, b). Applying the remark following the corollary to Theorem 5 of Chapter 5, we find that f is decreasing on (a, xo] and increasing on [xo, b). It follows that f(x) > f(x) for x in (@, x0) and f(x) > f(%o) for x in (2, b). (See Fig. 6-1.) y on (axe) on (x9,} y= fix) 6-1 f has a local minimum point at xo. 74TURNINGPOINTS 75 Notice that for all x # xo in (a,b), (x0) < s(x); that is, the smallest value taken on by f in (4,) is achieved at xo. For this reason, we call xo a local mini- ‘mum point for f. In case f’ changes from positive to negative at xo, a similar argument shows that f behaves as shown in Fig. 6-2 with f(xo) > f(x) for all x # x9 in an open interval (@, b) around x. In this case, Xp is called a local maximum point tors. fincressing ‘fF cresing YR onlaxel on(xe.b) to Lay = fix) y= fix) Fig. 6-2 f has a local maximum point at xa. Worked Example 1 Find and classify the turning points of the function f(x} 23 + 3x? — 6. Solution The derivative f"(x) = 3x? + 6x — 6 has roots at 1 +-V3; it is posi- tive_on (—ce,—1 — V3) and (—1 + -V3.00) and is negative on (—1 — V3,-1 + V3). Changes of sign occur at —1 — V3 (positive to negative) and —1 + V3 (negative to positive), so —1 — V3 is a local maximum point and —1 +-V3 isa local minimum point. We can summarize the results we have obtained as follows. Theorem I Suppose that f' is continuous in an open interval containing xo. 1. If the sign change of f" is from negative to positive, then x9 is a local minimum point. 2. If the sign change of f' is from positive to negative, then Xp is a local maximum point. 3. If f' is continuous, then every turning point is a critical point, but a criticial point is not necessarily a turning point. Solved Exercises 1. Find the turing points of the function f(x) = 3x* — 8x? + 6x? — 1. Are they local maximum or minimum points?76 CHAPTER6: GRAPHING 2. Find the turning points of x" for n = 0, 1,2, 3. Does g(x) = 1/f(x) have the same turning points as f(x)? Exercises 1. For each of the functions in Fig. 6-3, tell whether xo is a turning point, a local minimum point, or a local maximum point. ++— (2) (©) (a Fig. 62 Identity the turning points. 2. Find the turning points of each of the following functions and tell whether they are local maximum or minimum points: @ fa)=4e Ext 44x41 OH) F) (© fle)=axt+bx +0 (@#0) (@) f 3. Find a function f with turning points at x = —1 and x = 4, Can a function be increasing at a turning point? Explain. If g isa function such that g(x9) ~ O and g'(xo)> 0, Uhemg is imuicasing al 19, 90. 4 changes sign from negative to positive at x9. Similarly, if g(vo) = O and #'(X0) <0, then g changes sign from positive to negative at xo. Applying these assertions to g = f”, we obtain the following useful result. Theorem 2 Second Derivative Test Let f be a differentiable function, let Xo be a critical point of f (that is, f'(xo) = 0), and assume that f""(xo) exists 1. If f'"(%0)> 0, then xq is a local minimum point for f.ATESTFOR TURNING POINTS 77 2. IF Fle) <0, thon xy io 2 local maximum point for f. 3. If F"(%o) = 0, then xo may be a local maximum or minimum point, or it mav not be a turing point at all. Proof For parts 1 and 2, we use the definition of turning point as a point where the derivative changes sign. For part 3, we obscrve that the func- tions x*, —x*, and x? all have f"(0) = f"(0) = 0, but zero is a local mini- mum point for x*, a local maximum point for —x*, and not a turning point for x? Worked Example 2 Use the second derivative test to analyze the critical points of f(x) =x? — 6x? +10, Solution Since f'(x) = 3x? — 12x = 3x(x — 4), the critical points are at 0 and 4, Since f"(x) = 6x — 12, we find that f”"(0) = —12 <0 and f"(4) = 12>0. By Theorem 2, er0 is a local maximum point and 4 is local minimum point. When f(a) = 0 the second derivative test is inconclusive. We may still use the first derivative test to analyze the critical points, however. Worked Example 3 Analyze the critical point xq = ~1 of f(x) = 2x* + 8x? + 12x? + 8x +7, Solution The derivative is f"(x) = 8x? + 24x? + 24x + 8, and f"(-1) = -8 + 24- 24+ 0, so —1 is a critical point. Now f"(x) = 24x? + 48x + 24, so f'(-1) = 24 — 48+ 24 = 0, and the second derivative test is inconclusive. If we factor f’, we find f"() = (x? + 3x? + 3x + 1) — 8 (x + 1)®, Thus —1 is the only root of f’, f'(—2) = -8, and f’(0) = 8, so f” changes sign from negative to posi- tive at —1; hence —1 is 2 local minimum point for f. Solved Exercises 4. Analyze the critical points of f(x) =? —x. 5. Show that if (xq) — (#9) — 0 and f"(x0) > 0, then x9 is not a turning point of f. Exercises 5. Analyze the critical points of the following functions: @ fe) (©) g(5) = s/(1 +8)78 — CHAPTER6: GRAPHING © h)=p + (lp) 6. Show that if f"(vo) = f"(xa) = 0 and f”"(¥o) <0, then x9 isnot a turning point for f. Coneavity The sign of £”"(¢o) has a useful interpretation, even if /"(x9) is not zero, in terms of the way in which the graph of 1s "benaing” at x9. We first make a prelimi- nary definition. Definition If f and g are functions defined on'a set D containing xo, we say that the graph of f lies above the graph of g near x if there is an open interval f about x9 such that, for all x # xo which lie in and D, f(x) > a(x). The concepts of continuity, local maximum, and local minimum can be expressed in terms of the notion just defined, where one of the graphe ic a horizontal line, as follows: 1, fis continuous at xo if (@ For every c > f(xo), the graph of f lies below the line y = ¢ near Xo, and (b) For every c r(9)- But r(0) = f(0) — M&x0) =80 CHAPTER 6: GRAPHING H(0) — fl0co) = 0, s0 r(x) > O for all x # x9 in Isie., fe) — Ix) > 0 for all x #.x9 in J; ie., f(x) > U(x) for all x # xo in J; ie., the graph of flies above the tangent line at Xo, so fis concave upward at x9. The case where f' is decreasing at xo is treated identically; just reverse all the inequality signs above Since f" is increasing at xo when f"(xo) > 0, and f" is decreasing at xo wien (x9) <0, we obtain the following vest. (See Fig. 6-6.) Fx) <0 p90 \ Fig. 6-6 f Is wuncave up- i ward (holds water) when feoncave upned | fooncavedowmwar —f"" > 0 and is concave malixe downwerd. {sheds woter) when f” <0. Corollary Let f be a differentiable function and suppose that f"(xo) exists. 1. If f"(xo)> 0, then f is concave upward at Xo. 2. If f"(%o) <0, then f is concave downward at Xo. 3. If F'"(%o) = 0, then f at xo may be concave upward, concave downward, or neither. The functions x*, —x*, x3, and —x® provide examples of all the possibil- ities in case 3. Solved Exercises 6. Show: (@) x* is concave upward at 2er0; (©) x4 is concave downward at zeros (c) x? is neither concave upward nor concave downward at zero. 7. Find the points at which fle) ~ 3x3 8x + 12 ia concave upward and those at which it is concave downward. Make a rough sketch of the graph.INFLECTION POINTS 81 8. (a) Relate the sign of the error made in the linear approximation to f with the second derivative of f. (0) Apply your conclusion to the linear approximation of 1/x at xo = 1. Exercises 7. Find the intervals on which each of the following functions is concave up- ward or downward. Sketch their graphs. @ Ur (®) 1/1 +2) (9 P+ 4x? Be +1 8. (a) Use the second derivative to compare x* with 9 + 6(x — 3) for x near 3. (b) Show by algebra that x > 9 + 6(x — 3) for allx. Inflection Points ‘We just saw that a function f 1s concave upward where (x) > 0 and concave downward where f(x) < 0. Points which separate intervals where f has the two types of concavity are of special interest and are called inflection points. An example is shown in Fig. 6-7. The following is the most useful technical defini- tion. Definition The point xo is called an inflection point for the function f if f' is defined and differentiable near 9 and f”” changer eign at xy. ” ve fia Coneave | Concave | Concave upward | downward | upward Fig. 6-7 x, and xq are x * x inflection points. Comparing the preceding definition with the one on p. 74, we find that inflection points for f are just tuning points for f'; therefore we can use all our ‘techniques for finding turning points in order to locate inflection points. We summarize these tests in the following theorem.82 CHAPTER6: GRAPHING Theorem 4 4. {J X98 an injlection pomt Jor J, then ¢"(%o) =U. 2. If f'"(%o) = O and f'"(xo) # 0, then xq is an inflection point for f. If you draw in the tangent line to the graph of f(x) at x = a in Fig. 6.6, ‘you will find that the line overtalces the graph of f at 2. Tho opposite can ooour, ‘to0; in fact, we have the following theorem, which is useful for graphing. Theorem 5 Suppose that xa is an inflection point for f. 1. If f"" changes from negative to positive at xo, in particular, if fo) = 0 and f""(x9) > 0, then the graph of f overtakes its tan- ‘gent line at Xo. 2. If f" changes from positive to negative at xo, in particular, if £"(xo) = 0 and f'"(9) <0, then the graph of f is overtaken by its tangent line at xo. We invite you to try writing out the proof of Theorem 5 (see Solved Exer- cise 10). The two cases are illustrated in Fig. 6-8. Recall that in defining the derivative we stipulated that all lines other than the tangent line must cross the graph of the function. We can now give a fairly = fd : A, = Nx) | y= fle) =itx) A op int a y= te) ote | 4 + 2 Fig. 6-8 The tangent line yenn-nm 08686. the rach when (0) £3) <0 F(x) — Nx) changes sign.THEGENERALCUBIC 83 complete answer to the question of how the tangent line itself behaves with re- spect to the graph, Near a point x9 where /""(ro) #0, the tangent line lies above or below the graph and does not cross it. If F"(xo) = 0 but f”"(xa) #0, the tan- gent line does cross the graph. If f"(xy) and f(y) are both zero, one must examine f” on both sides of. x9 (or use f’”"(xo)) to analyze the situation. Solved Exercises 9. Discuss the behavior of f(x) at x = 0 for fix) = 10. Prove Theorem 5. [Hfint: See Solved Exercise 5.] 11. Find the inflection points of the function f(x) = 24x* — 32x3 + 9x? +1. Exercises 9. Pind the inflection points for 2", 1 @ pusitive integer. How dues the auswet depend upon n? 10. Find the inflection points for the following functio the tangent line behaves with respect to the graph. (@) x8 -x (b) x#—x7 +1 © @-1 (@) 1/0. +x) 11. Suppose that f(a) = £%xo) = F"Géa) = 0. but F"Geo) # 0. Ts x0 a turning point or an inflection point? Give examples to show that anything can happen if f"(o) = 0. In each case, tell how The General Cubic ‘Analytic geometry teaches us how to plot the general linear function, f(x) = ax +b, and the general quadratic function f(e) = ax? + bx + ¢ (@ #0). The methods of this section yield the same results: the graph of ax +h isa straight line, while the graph of f(x) = ax? + bx + c is a parabola, concave upward if a> 0 and concave downward if a < 0. Moreover, the turning point of the para- Wola oceuis whien f"(a) = 2ax 1 b= 0; that is,» = —(6/2a). ‘A more ambitious task is to determine the shape of the graph of the gen- eral cubic f(x) = ax? + bx? + ox +d. (We assume that a #0; otherwise, we are dealing with a quadratic or linear function.) Of course, any specific cubic can be plotted by techniques already developed, but we wish to get an idea of what all possible cubics look like and how their shapes depend on a, b, c. and d.84 CHAPTER: GRAPHING ‘We begin our analysis with some simplifying transformations. First of all, we can factor out @ and obtain a new polynomial f;(x) as follows: ‘The graphs of f and f, have the same basic shape; if a > 0, the y axis is just re- scaled by multiplying all y values by a: if @ <0, the y axis is rescaled and the graph is flipped about the x axis. It follows that we do not lose any generality by assuming that the coefficient of x? is 1. Consider therefore the simpler form AQ) =P + byx? Hoye td where by = b/a, ¢, = c/a, and dy = dia. In trying to solve cubic equations, mathematicians of the early Renaissance noticed a useful trick: if we replace x by x — (61/3), then the quadratic term drops out; that is, ti a -4) =x tox tds where ¢, and dz are new constants, depending on by, c1, and dy. (We leave to the reader the task of verifying this last statement and expressing cy and da in terms of by, c;,and dy.) The graph of f,(x — b,/3) = f(x) is the same as that of f,(x) except that it is shifted by 51/3 units along the x axis. This means that we lose no generality hy accuming that the enafficient of x? is 2era—that is, we only need to graph Sax) = 2x34 c9x + dz. Finally, replacing fa(x) by fa(2) = fa(z) — dz just corre- sponds to shifting the graph d> units parallel to the y axis. We have now seduced the graphing of the general cubic to the case of graphing f,(x) = x° + cx. For simplicity let us write f(x) for f(x) and ¢ for ¢2. To plot f(x) = x* + cx, we make the following remarks: 1. fis odd. 2. fis defned everywhere. Since f(x) = x7(1 — ¢/x?), /(x) 1s large postuive [negative] when x is large positive [negative] . 3. FG) — Sx? 1c. If > 0, f(x) > 0 for all x, and f is increasing cvery- where. If c = 0, f'(x)> 0 except at x = 0, so f is increasing everywhere. If c <0, f(x) has roots at +V—c/3; f is increasing on (—», -/—c/3] and [V—c/3,o0) and decreasing on [~V/—c/3,V/—c/3] . Thus —/— 0. Zero is an inflection point.THEGENERAL CUBIC 85 (ife<0) We fE) IE é/¥)- 0 (ife0 (a) Type! (b) Type Fig. 6-9 The three types (e) Type il of cubics. ‘Thus there are three types of cubles (with w > 0). Type: fis increasing at all points; f” > 0 everywhere. Type Il: fis increasing at all points; Oat one point. Type III: f has two turning points. ‘Type II is the transition type between types I and III. You may imagine the graph changing as c begins with a negative value and then moves toward zero. ‘Asc gets smaller and smaller, the turning points move in toward the origin, the86 CHAPTER6: GRAPHING bumps in the graph becoming smaller and smaller until, when ¢ = 0, the bumps merge at the point where /"(x) = 0. As c passes zero to become positive, the ‘bumps disappear completely. Golved Exercises 12. Convert 2x + 3x? +x + 1 to the form x? + cx and determine whether the cubic is of type I, II, or III. Exercises 12. Convert x3 + x2 4 3v 4+ 1 to the form x? + ex and determine whether the cubic is of type I, I, or Ill. 13. Convert x? — 3x? + 3x + 1 to the form.x* + cx and determine whether the cubic is of type I, I, or IIL 14, (a) Find an explicit formula for the coefficient c, in f,(x — (b;/3)) in terms of by, ¢y, and dy, and thereby give a simple rule for determining whether the cubic x° + bx? + cx +d is of type I, Il, or I. (©) Give a rule, in terms of 2, b, ¢, d, for determining the type of the gen eral cubic ax? + bx? + ox +d. (© Use the quadratic formula on the derivative of ax? + bx? + cx +d to determine, in terms of a, b, c, and d, how many turning points there are. Compare with the result in (b). The General Quartic Having given a complete description of all cubics, we turn now to quartics: fex)=ax* bx tex? +dete (@#0) Exactly a2 we did with cubiee, we can ascume that @ — 1, b — 0 (by the eubeti- tution of x — (6/4) for x), and can assume that ¢ = 0 (by shifting the y-axis). ‘Thus our quartic becomes (with new coefficients!) f(x) =x* + cx? + de Note that SQ) = 409 + 2x tdTHEGENERALQUARTIC 87 and I") = 12x? + 2e From our study of cubics, we know that f’(x) has one of three shapes, according to whether ¢ > 0, ¢ = 0, or ¢ <0 (types I, Il, and I}. It turns out that there aro 9 types of quartice, six of which may be divided into paire which, are mirror images of one another. Type I c > 0. In this case, f"(x) is everywhere positive, so f is concave upward and "is increasing on (—s,<:). Thus, J" has one root, which must be a local minimum point for f. The graph looks like type I in Fig. 6-11. Type Il c = 0. Here the cubic f'(x) is of type Il and is still increasing. ‘J'@) = 12x? > 0, so there are no inflection points, and fis everywhere concave upward. We can subdivide type I acording to the sign of d. ‘ype lly: d > U. Looks like type 1. (See Solved Exercise 13.) Type Ia: = 0. Looks like type 1, except that the local minimum point is Matior because 7," aind "all vanish hese. ‘Type Ils: d <0, Looks like type I. (See Solved Exercise 13.) ‘The graphs are sketched in Fig. 6-11. Type IT c < 0. This is the most interesting case. Here, the cubic f"(x) has two turning points and has 1, 2, or 3 roots depending upon its position with respect to the x axis. (Fig. 6-10.) To determine in terms of c and d which of the five possibilities in Fig. 6-10 holde for 'Ge) — 4x3 + dex +d, we notice from the figuro that it ouffices to determine the sign of f"(x) at the turing point of f” (Which are inflection points of f). These points are obtained by solving 0 = f"(x) = 12x? + 2c. They occur when x = =V/—c/o. Now J'(—V—c[6) = 5 c/—<7b + dis the value at the local minimum and f'(/— 320.7 The Mean Value Theorem ‘The mean value theorem is, like the intermediate value and extreme value theorems, an existence theorem. It asserts the existence of a point in an interval Where u fuuctivn has # paitivular Uchavior, but it does not tell you how to find the point. Theorem | Mean Value Theorem. Suppose that the function f is contin- uous on the closed interval [a,b] and differentiable on the open interval (@,b). Then there is a point xo in the open interval (a,b) at which F'%o) = 16) -F@|/@ 2). In physical terms, the mean value theorem says that the average velocity of a moving object during an interval of time is equal to the instantaneous veloc- ity at some moment in the interval. Geometrically, the theorem says that a secant line drawn through two points on 2 smooth graph is parallel to the tan- gent line at some intormodiate point on the curve. There may be more than one such point, as in Fig. 7-1. Consideration of these physical and geometric inter- pretations will make the theorem believable. 1g 1h \ Fig. 7-1 Tangent lines (aoe % parallel to the secant line. We will prove the mean value theorem at the end of this section. For now, wwe will concentrate on some applications. Our first corollary tells us that if we know something about f"(x) for all x in [a,b] , then we can conclude something about the relation between values of f(x) at different points in [a,b]. Corollary 1 Let f be differentiable on (a,b) {and continuous on (a,b]]- ‘Suppose that, for all x in the open interval (a,b), the derivative f'(x) be- longs 00 a certain sct 5 of real numbers. Then, for any two distinet points x, and xz in (a,b) [in [a,b], the difference quotient 9394 CHAPTER 7: THE MEAN VALUE THEOREM £2) — fe) =x belongs to S as well. Proof The difference quotient stays the same if we exchange x, and x2, so we may assume that x, 98 CHAPTER 7: THE MEAN VALUE THEOREM Solved Exercises 5. Let fix) = x* — 9x + 26x? — 24x. Note that f(0) = 0 and f(2) = 0. Show without calculating that 4x? — 27x? + 52x — 24 has a root somewhere strictly between 0 and 2. 6. Suppose that f is a differentiable function such that f(0) = 0 and f(1) = Show that f"(xo) = 2xo for some x9 in (0, 1). Exercises 5. Suppose that the horses in the horserace theorem cross the finish line with ‘equal yelocitics. Must they have had the same accclesat during the race? mm at some Une 6. Let fx) = lel — 1. Then f(-1) = f(1) = 0, but f"(x) ie never equal to zero on [—1,1] . Does this contradict Rolle’s theorem? Explain. Problems fOr ChOP SS 1. Suppose that (4?/dx?) [/() — 2g(x)] = 0. What can you say about the relationship between f and g? 2, Suppose that f and g are continuous on [a,b] and that f" and g’ are con- tinuous on (a, 5). Assume that F@)=g0@) ana fio) =g00) Prove that there is a number c in (g, 5) such that the line tangent to the graph of f at (¢, f(c)) is parallel to the line tangent to the graph of g at a). 3. Let f(x) = x7 ~ x8 — x* + 2x + 1, Prove that the graph of f has slope 2 somewhere between —1 and 1. 4. Find the antiderivatives of each of the following: (a Tay = fx — ax? aL (b) fx) = 6x8 = 1239 + 15x — 11 () fedaxte Wd txt eee (a) f(x) = (fx?) + 2x (e) fe) = 2x(x? + 7)10. . FR) . (a) Let f(x) = x5 + 8x4 — 5x? +15, Prove that somewhere between —1 and PROBLEMS FOR CHAPTER? © 99 . Find an antiderivative F(x) for the given function f(x) satisfying the given condition: @) fix) = 1x8; FU) =3 on (a,b), what can you say about /? O the tangent line to the graph of f has slope —2. (b) Let f(x) = Sx* + 9x9 - 11x? + 10, Prove that the graph of f has slope 9 somewhere between —1 and 1. . Let f be a polynomial. Suppose that f has a double root at ¢ and at b. Show that /"(x) has at least three roots in [a,b]. . Let f be twice differentiable on (a, b) and suppose f vanishes at three dis- tinct points in (@,b). Prove that there is a point xo in (¢,5) at which £40) — 0. Use the mean value theorem to prove Theorem 5, Chapter 5 without the vvothesis that f’ be continuous.8 Inverse Functions and the Chain Rule Formulas for the derivatives ot inverse and composite functions are two ot the most useful tools of differential calculus. As usual, standard calculus texts should be consulted for additional applications. Inverse Functions Definition Let the function f be defined on a set A. Let B be the range of values for f on A; that is, © B meane that y = f(«) for some x in A. We say that f is invertible on A if, for every y’ in B, there is a unique x in A such that y = f(x). If fis invertible on A, then there is a function g, whose domain is B, given by this rule: g(y) is that unique x in A for which f(x) = y. We call g the inverse function of f on A. The inverse function to fis denoted by f™ Worked Example 1 Let f(x) = x*. Find a suitable A such that f is invertible on A. Find the inverse function and sketch its graph. Solution ‘The graph of f on (~22, <0) is shown in Fig. 8-1. Fory in the range of ‘f-that is, y > O-there are two values of x such that f(x) = y: namely, -Vy and +Vy. To assure only one x we may restrict f to A = (0,0). Then B {0,~) and for any y in B these is exactly one x ind such that /(a) ~ y namely, (9). (Choosing A = (—=,0] and obtaining x = we is also accept- able.) We obtain the graph of the inverse by looking at the graph of f from the back of the page. If we are given a formula for y = f(x) in terms of x, we may be able to find a formula for its inverse x = g() by solving the equation y = f(x) for x in terms of y. Sometimes, for complicated functions f, one cannot solve y = f(x) to get an explicit formula for x in terms of y. In that case, one must resort to thearetical results which gnarantee the existence af an inversa function. ‘Thesa will be discussed shortly. 100INVERSE FUNCTIONS 101 y= fx) =e x — Restrict domain 1010.) Fig. R-1. A restriction af fle) = ¥? and its inverse Worked Example 2, Does f(x) = x° have an inverse on (—2, 2)? Sketch. Solution From Fig. 8-2 we see that the range of f is (—0,e:) and for each y © (see, 00) there is exactly one number x such that f(x) = y—namely. x = Yy = FQ) (negative if y <0, positive or zero if y > 0)~so the answer is yes. We can regain x as our independent variable by observing that if f(y) = Wy, then Fe) = YE Thus we can replace 3» by x co that the independent variable has a more familiar name. As shown in Fig. 8-2(6) and (c), this renaming does not affect the graph of f. a y= fix) = w= Ve I (a (w © Fig. 8-2 The inverse of the cubing function is the cube root, whatever you call the variables. Solved Exercises* 1, If'm #0, find the snverse function for s(x) = mx + 0 on (—2», 09). 2. Find the inverse function for f(x) = (ax + b)/(ex +d) on its domain (c #0). ‘Solutions appear in the Appendix102 CHAPTER 8: INVERSE FUNCTIONS AND THE CHAIN RULE 3, Find an inverse function g for f(x) = x? + 2x + 1 on some interval con- taining zero. What is g(9)? What is g(x)? 4, Sketch the graph of the inverse function tor each tunction in Fig. 8-3. es KA. (a) ) ) V Fiy. 09 Sketch jeise Turnetivns, 5. Determine whether or not each function in Fig, 8-4 is invertible on its domain. vt y y / QQ Kk, y ——— fa) (b) to Fig. 8-4 Which functions are invertible? Exercises 1, Find the inverse for each of the following functions on the given interval: (a) fx) = 2x+5.on [4,4] (b) fle) = —Fx +2 on (—o2, 0) (c) A(Q) =t— 10 0n [0,7) © as) = s+ YEs+ I) on HF © fe) =x* on (—e, 20) () F@) =x* on (0, 1] 2, Determine whether each function in Fig. 8-5 has an inverse. Sketch the in- verse if there is one. 3. Sketch a graph of f(x) invertible. x/(1 + x?) and find an interval on which f isATESTFORINVERTIBILITY 103 y= fix va (a) tb) (a Fig. 85 Which functions are invertible? 4, Enter the number 2.6 on a calculator, then push the x? key followed by the Vx key. Is there any round-off error? Try the Vx key, then the x? key. Also try a sequence such as x?, VX. x?, Vx, ... . Do the errors build up? Try pushing the x? key five times, then the Vx key five times. Do you get back the original number? Try these experiments with different starting numbers, 5, If we think of a French-English dictionary as defining a function from the cet of French wards to the set of English words (does it really?), how is the inverse function defined? Discuss. A Test for Invertibility ‘A function may be invertible even though we cannot find an explicit formula for the inverse function. This fact gives us a way of obtaining “new functions.” There is a useful calculus test for finding intervals on which a function is in- vertible. Theorem I Suppose that f is continuous on a,b] and that f is increasing at each point of (a,b). (For instance, this holds if f'(x) > 0 for each x in (@,B).) Thon f is invertible on [a,b], and the inverse f is defined on Y@,F@)1. Uf f is decreasing rather than increasing at each point of (a,b), then f 1s st invertible; in this case, the domain of f~* 8 (/(0),SCa)l - Proof If f is continuous on [a,b] and increasing at each point of (@,b), we know by the results of Chapter 5 on increasing functions that f is in- creasing on [a,b]; that is, if @ 0. Solution Since f is differentiable on (—e,<), it is continuous on (—e,<«) and hence on [0,8]. Dut f(x) — 2x > 0 for 0 0 or f'(x) <0 for ail x in an open inter- val I containing x», 60 that hy Thenrem I there is an inverse function g to f, defined on an open interval containing f\xe) = yo, with g(o) = x0. Then g is differentiable at yo and Pi i £00) 7) FEO) Proof We will consider the case f'(x) > 0. The case f'(x) < 0 is entirely similar (in fact, one may deduce the case f’ < 0 from the case f’ > 0 by considering —f). From Theorem 1, f is increasing on / and the domain of g is an open interval J. The graphs of typical f and g are drawn in Fig. 8-7, which you should consult while following the proof. Let x = Ay) be 2 line through (Yo, £(V0)) = ('o.X0) (Fig. 8-7) with slope m > 1/f"(xo) > 0. We will prove that / overtakes g at Yo. We may write I(y) = my + b, so iyo) = xo. Define I(x) = (x/m) — (b/m) which is the inverse function of I(y). The line » = ie) passes through (eo, Fle)) and has slope 1/m ix) ifxisinh,x>xo and f@) xo corresponds to y > Yo if ‘y = fe), and likewise x <2 corresponds to y Yo, and let y = f(x) so x > xo and x is inh. Hence Since y = f(x) and x = g(y), this becomes 2 y= ie., 1g) = my +b> 80) since m > 0. Similarly, one shows that if y 0.] Composition of Functions ‘The idea behind the definition of composition of functions is that one variable depends on another through an intermediate one. Definition Let f and g be functions with domains Dy and D,. Let D be the set consisting of those x in Dg for which g(x) belongs to Dy. For x in D, we van evaluate /lg(a)), and we call the result A(x). The resulting Func- tion A(z) = f(e()), with domain D, is called the composition of f and g. Itis often denoted by. Worked Example 5 If flu) =u? + 2 and g(x) = V3" #1, what ish =f -g? Solution We calculate h(x) = flg(x)) by writing u = g(x) and substituting in flu). We get u = Vx? +1 and A(x) = flu) =u? + 2 = (Vx? +1)? + 2= (7 + 17/742 Since each of f and g is defined for all numbers, the domain D of h is (co, «<). Solved Exer 14, Let g(x) = 4 1 and fu) =u?. Find f-g and g-f. 15, Let h(x) = x% + 3x"? + 1, Write A(x) as a composite function f(g(x)). 16. Let (x) =x — 1 and g(x) = Vx. (a) What are the domains Dy and Dy? (b) Fina yg and g +. Wnat are their domains? (©) Find (f+g)(2) and @ -/)(2). (a) Sketch graphs of f, g, f-g,and gs -f. 17. Let i be the identity function i(x) = x. Show that («f= f and f+i = f for any function f.110 CHAPTER 8: INVERSE FUNCTIONS AND THE CHAIN RULE Exercises 16. Find f -g and g -f in each of the following cases. @ 2) =x; fa) =V¥—7 (©) g@) =x"; f(x)=x* (7, s rational) © e@)=1/ -2); fa) = -VE (@) ste) = Gx — 24x +1); A) = (2x — NOx + 3) 17. Write the following ae compositions of simpler functions: @) AG) = 4x7/(@? - 1) (h) AG) =? + br 4 073/24 (s/o ora) (©) hw) =VO =a) tu) 18, Show that the inverse f of any function f satisfies f -f! = fand f7 +f: where i is the identity function of Solved Exercise 17. The Chain Rule Theorem 3 Chain Rule. Suppose that g is differentiable at x» and that fis differentiable at g(xo). Then f +g is differentiable at xo and its derivative there ts f'(g(Xo))*8'o)- As with the algebraic rules of differentiation, to prove the chain rule, we reduce the problem to one about rapidly vanishing functions (See Chapter 3). Write Yo = g(x) and f0) = 100) +F' Vo — yo) +r) where r(y) is rapidly vanishing at yo. Set y = g(x) in this formula to get FEE) = Fo) + Fe Eo) KE) — 860) + rg) Suppose we knew that r(g(x)) vanished rapidly at xo. Then we could differen- nate the righthand side with respect to x at x= xy to obtain what we want (using the sum rule, the constant multiple rule, and the fact that the derivative of a constant is zero). In other words, the proof of Theorem 3 is reduced to the following result.THECHAINRULE 111 Lemma Let r(y) be rapidly vanishing at yo and suppose that g(x) is dif- ferentiable at xo, where y 9 = g(xo). Then r(g(x)) is rapidly vanishing at xo. Proof We will use the characterization of rapidly vanishing functions given by Theorem 1 of Chapter 3. First of all, we have r(g(o)) =r(v0) = 0, since r vanishes at yo. Now, given any number ¢ > 0, we must find an open interval J about xo such that, for all x in J with x # x0, we have ir(e(2))I < ele — x0. Pick a number M such that M > |g'(xo)|. Since r vanishes rapidly at Yo and e/M is a positive number, there is an open interval J about Yo such ‘that, for all y in J with y *yo, we have ()1<(c/M) y —yol- Next, let J be an open interval about x» such that, for x # xo in J, we have lg(x) — g(0)I< Mx — xo. Such an interval exists because -M < ¥(&o) 0, so we are done. Worked Example 6 Verify the chain rule for f(u) = u? and g(x) =x? + I. Solution Let A(x) = f(g) = [e@)]? = @? + 1)? = x6 + 2° + 1. Thus A(x) = 6x5 + 6x*. On the other hand, since f'(u) = 2u and g'(x) = 3x,112 CHAPTER 8: INVERSE FUNCTIONS AND THE CHAIN RULE FG) 1G) = (2+? + 1) 3x7 = G5 + Gx? Hence the clutn rule fs verified in this case. Solved Exerci 18, Show that the rule for differentiating 1/g() follows from that for 1/x and the chain rule. 19. How should we choose J in the proof of the lemma so that g(x) isin J for allx in? Exercises 19. Suppose f is differentiable on (a, b) with f(x) > 0 for all x, and g isdifferen- tiable on (c,d) with g'(x) <0 forx in (c,d). Let (c,d) be the image set of f, so g +f is defined. Show that g +f has an inverse and calculate the derivative of the inverse. 20. Let f, g, hall be differentiable. State a theorem concerning the differentia- bility of K(x) = h(e(Ax))). Think of a spenific example where yon wonld apply this result. 21. Prove that 2 LAGI’ =r 7°"! 7G) (cational power of a function rule). Problems fOr Chap te 5 SS 1, Prove that, if fis continuous and increasing (or decreasing) on an open inter- val J, Grout Use iuveise funttivu "is euuLinuuus as Well. 2. Let f be differentiable on an open interval J. Assume f" is continuous and f'Gca) # 0. Prove that. on an interval about xo. f has a differentiable in- verse g. 3, Suppose that f is differentiable on an open interval J and that f” is contin- uous. Assume /“(x) # 0 for all x in J. Prove f has an inverse which 18 dif- ferentiable. 4 tet f (x) f(0)= 0. 5. Show that the inverse function rule can be derived from the chain rule if you assume that the inverse function is differentiable (use the relation FAGG) =x). sf (om times). Express £0) in terms of f'(0)if10, 1. 12. PROBLEMS FOR CHAPTERS 113 Let f(x) =x — 3x +7. (@ Find an interval containing zero on which fis invertible. Denote the in- verse by g. (b) What is the domain of g? (c) Calculate g'(7). Let fx) = (ax + b)/lex + d), and Let g(x) = (rx + sy/tex tu). (a) Show that f+g and g -/ are both of the form (kx + D/(mx +n) for some k.l.m,andn. (b) Under what conditions on a, b, c,d, r,s, t, u does f+g=g +f? If f isa given differentiable function and g(x) = f(V), what is (x)? FAVS, what is g(x)? Find a formula for the second derivative of f+g in terms of the first and second derivatives of f and g. If fis differentiable and has an inverse, and g(x Find a formula for the second derivative of g(y) if g(y) is the inverse func- ton of 7x). Derive a formula for differentiating the composition fy * f functions. fn ofm9 The Trigonometric Functions The theory of the trigonometric functions depends upon the notion of arc length on a circle, in terms of which radian measure is defined. It is possible to develop this theory from scratch, using the integral (just as for the logarithm), but intuition is sacrificed in this approach. At the expense of some rigor, we shall take for granted the properties of arc lengths of circular arcs and the attend- ant properties of z and the trigonometric functions. The Derivative of sin 6 and cos 6 We recall that if an arc length @ is measured along the unit circle in the x, y plane counterclockwise from (1,0) to a point P, then the coordinates of P are (cos@, sin). (See Fig. 9-1.) The measure of the angle swept out is 6 radians. The graphs of sin@ and cos@ are shown in Fig. 9-2. Fig. 9-1 The defi cos@ and sin 8. The key to differentiating the trigonometric functions is the following Jemma. Lemma For —n/2 <0 1 — (67/2) and hence the lemma. Theorem 1 We have sin'(Q)=1 cos'(0)=0 Proof Form> 1 we have sind o0 and sin@>mé if8<0 i.e., for m > 1, the line y = mé (in the 6,y plane) overtakes y = sind at 6=0. Since 1 — (67/2) ls continuous and equals 1 at @ = 0, 1f m <1, there is an interval / about 0 such that 1 — (62/2) > m if @ € 1. Thus GE and 9>0 implies siné > mo and 6€/ and 6 0 on (—n/2,1/2), sinx is increasing on this interval, so sinx has an in- verse on the interval [—1/2,7/2] . The inverse is denoted sin” y.* We obtain the graph of sin" y by interchanging the x and y coordinates. (See Fig. 9-4.) ys y=sinx 94 The sine func- ‘Although the notation sin® y is commonly used to mean (sin. * y does not mean (sin) = 1/sin y. Sometimes the notation arcsin y is used for the inverse sine function to avoid confusion.120 CHAPTER 9: THE TRIGONOMETRIC FUNCTIONS The values of sin” y may be obtained from a table for sinx. (Many pocket calculators can evaluate the inverse trigonometric functions as well as the trigo- nometric functions.) Worked Example 3 Calculate sin” sin (0.342). sin* 0. sin*(—1), sin (3), and Solution Since cin z/2 —"1, cin 1 — n/2. Similarly, ein! 0 — 0, ein ( 1) — 1/2. Also, sin(—n/6) = —}, so sin (-}) = —n/6. Using a calculator, we find sin (0.342) = 20°. We could have used any other interval on which sinx has an inverse, such as [n/2,3n/2],, to define an inverse sine function; had we done so, the function obtained would have been different. The choice [—n/2,7/2] is standard and is the most convenient. Let us now calculate the derivative of sin"! y. By the formula for the deriv. ative of an inverse, Cle ay" ~ @jax)sinx ~ cosx where y = sinx. However, cos?x + sin? x = 1, s0 cosx = 1 —y?. (The negative root does not occur since cosx is positive on (—1/2,7/2).) Thus tnt A= (1 yy", cyt iy Fra vy, y Notice that the derivative of sin” y is not defined at y = +1 but is “infinite” there. This is consistent with the appearance of the graph in Fig. 9.4. Worked Example 4 Differentiate h(y) = sin (3y*). Solution From the chain rule, with u = 3y?, HO)=C -wyt 7G = 6y(1 -9y57"? Worked Example 5 Differentiate f(x) xsin™ (2x). Solution Here we arc using x for the variable name. Of course we can usc any letter we please. By the product and chain rules, '@) = @) sin? (2x) tx tint 2x)PROBLEMS FOR CHAPTERS = 121 sin t 2x 2x(1 4x2)? It is interesting to observe that, while sin” y is defined in terms of trigono- metric functions, its derivative is an algebraic function, even though the deriva- tives of the trigonometric functions themselves are still trigonometric. Solved Exercises 4, (a) Calculate sin” (4), sin (--V3/2), and sin (2). (b) Simplify tan (sin). 5. Calculate (d/dx) (sin 2x)*/?. 6. Differentiate sin? (/T —x?), 0. Ditterentiate the indicated functions: (@) @&* - 1) sin* @) (©) (sin x}? (c) sin® [7/(t+ 1] (domain = ? 6. What are the maxima, minima, and inflection points of f(x) = sin™ x? 7. Show that cosx on (0,7) has an inverse cos and 1 Leostx ~ - ax vi-x* 8. Show that tanx is defined for all x, takes values between —1/2 and 7/2, and. Problems for Chapter 9 1. Differentiate each of the following functions: (@) fe) — Vx + 200 3x (8) fo) —Veoex (©) fx) = (Gin 3x? +2) (@) FO) =? cos x + tanx)?? (©) £06) = cot (sind + Vcos* 38 + 0°) (© f= +V1—Pyersinr) 2. Differentiate sec [sin (y — 2)] by (a) simplifying first and (b) using the chain rule right away.122 CHAPTER 9: THE TRIGONOMETRIC FUNCTIONS 3. (a) What is the domain of cos" (x? — 3)? Differentiate. (b) Sketch the graph of cos (x? ~ 3), 4, Show that f(x) = secx satisfies the equation f” +7 — 2? = 0. 5. Is the following correct: (d/dx) cos x = (—1)(cos® x)[(d/dx) cosx]? 6. Find a function f(x) which is differentiable and increasing for all x, yet FG) /0Y8 =e, Worked Feample § If h>1 and p and q are rational numbers with p 0. We shall show that b? > 1, s0 69/6? > 1 and thus b? > bP. Suppose that z = m/n. Then B? = (6")"". However, b™ =b+b+...+B (im times) > 1 since b > 1, and (6”)'/" > 1 since 6” > 1. (The nth root cl!” of a number ¢ > 1 is also greater than 1, since, if c!/" <1, then (cl/")" =c<1 also.) Thus b* > 1 if z > 0, and the solution is complete. As a consequence, we can say that if b > 1 and p \x° +»), a, b, c,d rational num- bers): @ x-Vxy-» (0) x-y © Var tVyxttxty @) x-2vx-8 (©) x+2V3x+3 3. Solve for x: (a) 10* = 0.001 () S¥=1 () *=0 @ x-2Vx—3=0 (factor) 4, Do the following define the same function on (a) (—1, 1), (b) (0,3)? fiy=x!? Ae) — Tot fax) = (1%)? (which, if any, are the same?) 5. Based on the laws of exponents which we want to hold true, what would be your choice for the value of 0°? Discuss. 6. Using rational exponents and the laws of exponents, verify the following root formulas. @ ) SUF HWE 1 P 7. Find all real numbers x which satisfy the following inequalities. @ x? >x" @ 220 () xP >x/4, p,q positive odd integers,p > (a) x1/4 > x1, p,q positive odd integers, p > q 8. Suppose that b> O and that p = rational powers, that b”/" = pm’ m'/n'. Show, using the definition of Le. b? is unambiguously defined.THE FUNCTION fix)=6" 127 The Function f(x) = bX Having defined f(x) = d* if x is rational, we wish to extend the definition to allow x to range through all real numbers. If we take, for example, b = 2 and compute some values, we get: 3 2 These values may be plotted to get an impression of the graph (Fig. 10-1). It seems natural to conjecture that the graph can be filled in with a smooth curve, i.e.. that b* makes sense for all x. Fig. 10-1 The plot of some points (x, 2*) for rational x. To calculate @_number like 2V3, we should be able to take a decimal approximation to V3 ~ 1.732050808 ..., say, 1.7320, calculate the. rational power 21-7920 — 917320/10090, and hope to get an approximation to 2¥ >, Experi mentally, this leads to reasonable data. On a calculator, one finds the following: 1}2 1.7 | 3.24900958 1.73 | 331727818 1.732 | 3.32188010 1.73208 | 3.32199523 1.7320508 | 3.32199707 1,732050808 | 3.32199708128 CHAPTER 10: THE EXPONENTIAL AND LOGARITHM FUNCTIONS The values of 2* as x gets closer to V3 seem to be converging to some definite number. By doing more and more calculations, we could approximate this number to as high a degree of accuracy as we wished. We thus have amethod for generating the decimal expansion of s sunber which could be called 23 To define 2V* and other irrational powers, we shall use the transition idea. Let b be positive and let x be irrational. Let 4 be the set of all real num- bers a which are less than or equal to b?, where p is some rational number and larly, let R he the cat of numbers > AY where q ie some rational nee number and q > x (Fig. 10-2). Fig. 10-2 Powers of 2 with rational exponents less than ./3 go into set A {along with all numbers less than such powers) and all powers of 2 tional exponents larger than V3 go into set B along with all numbers larger than such powers). Theorem I There is exactly one transition point from A to Bif b> (and from B to A if 0 1 and p and sativual with p < q we had 6? 1. we have expa x 1, B® has a unique in- verse function with domain (0,0) and range (—2e,se). This function is denoted ogg. Thus x = logs y is the number such that 6% =y. Worked Example 7 Find log; 9, logo (10*), and logs 3. Solution Let x = logs9. Then 3* = 9. Since 3? = 9, x must be 2. Similarly, logio 10° is a and logy 3 = + since 91” The graph of logy x for b > 1 is sketched in Fig. 10-3 and is obtained by flipping over the graph of expp.x along the diagonal y = x. As usual with inverse functions, the label » in logy» is only temporary to stress the fact that logy y is the inverse of y = exp, x. From now on we shall usually use the variable name x and write logy x. ig. 103 The graphs of ¥ = expy x and y = logy x compared. Notion that for b> 1, logy x is increasing. If b <1, oxpp x is decreasing and so is logs x. However, while expp x is always positive, logp x can be either positive or negative. From the laws uf exponents we can read off corresponding laws for logy x:130 CHAPTER 10: THE EXPONENTIAL AND LOGARITHM FUNCTIONS ogy (xy) = logy x + logy y and logy 6) = logy x — logy y @ logy (”) = y logy x @ logy x = logs, (c) loge (x) Gi) For instance, to prove (i), we remember that logp x is the number such that exp (logy x) = x. So we must check that expp (logy x + logy y) = expp (logp ay). But due lefthund side is expp (logy x)expy (logy y) = ay as is the right-hand side. The other laws are proved in the same way. Solved Exercises 5. How is the graph of exp, px related to that of exp, x? 6. Simplify: (V3 + 2V3yqv3 _ av), 7. Match the graphs and functions in Fig. 10-4. a @ Cy) © @ 10-4 Match the graphs and functions: ( yaxe® (ey youve (d v=(/3* Ud) y= (1/V3)* 8. Find log: 4, logs 81 and logyo 0.01. 9. (a) Simplify log, (6?*/25) (b) Solve for x: logs x = logy 5 + 3log, 3.CONVEX FUNCTIONS 137 Exercises 9. simpnry: 9 =F HF We 10. Give the domaine and ranges af the following finctions and graph them @ y= 20) @) ya © yar 11, Graph y = 3°*? by “‘shufuing™ dee graph of y — 3® wu units tw the left. Graph y = 9(3%) by “stretching” the graph of y = 3* by a factor of 9 in the ‘y-axis direction. Compare the two results. In general, how does shifting the ‘graph y = 3¥ by k units to the left compare with stretching the graph by a factor of 3* in the y-axis direction? 12, Consider s(x) = (~3)*. For which fractions x ts f(x) defined? Not defined? How might this affect your ability to define f(n)? 13. Graph the following functions on one sot of axes. (@ fe =2* () s@)=x7+1 (©) he) =x #1 ‘Can you make an estimate of /'(1)? 14, Solve for x: (@ log.5 =0 (b) loga(x?) = 4 (0) 2logsx +1og94 = 2 15. Use the definition of log,x to prove: (a) logs (x”) =y logy x (b) logy x = logy (c) loge (x) Convex Functions* ‘We shall use the following notion of convexity to prove Theorem 1. Definition Let fx) be a function defined for every rational [real] x. We call f convex provided that for every pair of rational [real] numbers x1 and x2 with x1 1, fx) = 6* defined for x rational, is (strictly) convex, Proof First of all, we prove that for; Fee2) + {UA%s) — 12) Mees — Xa) HO — 22) Sketch. What if f is strictly convex? Make up a similar inequality of the form f(x3) > something.134 CHAPTER 10: THE EXPONENTIAL AND LOGARITHM FUNCTIONS Exercises 16. If f(x) is twice differentiable on (22, ) with f(x) continuous and fx) > 0, prove that f is convex. [Hint: Consider g(x) = 4f(e) + (1 — A)fCe1) — ‘fOx+(1 = 2)x;) and show that g is increasing, and g(x,) = 0.] 17. Show that f(x) belis convex. Proof of Theorem 1 Let us suppose that b> 1 and that x is a given irrational number (the case b <1 is dealt with similarly). Let A be the set of a such that a , and lot B be tho cot of > 84, whore q ia rational and q >». Lemma I A and D are convex and hence intervals, A — (yuo) oF (2, a9] and B = (Bo, °°) or [Bo,%*) for some ao < Bo. Proof Suppose y; and yz belong to A, andy, Bo. If ao > fo, then Bp belongs to A 20 lo < 2 for some p 64 for some q > x. But if p fo is impossible, so it must be that a x. Then b? belongs to A, s0 bP < ao and similarly b? > fo. Picking a smaller p and larger q will insure that BP fo, If we choose d such that d is rational and 0 < (6% — f)/(b% — bP) < A< (OF — a9)/(64 — BP) <1, then we will have

x. By Solved Exercise 11 with x, = p, xy = No +(1—2)q, andx) =q, we get Ap (1=2)q X@ =P) pr > prrc-na (-~-4@) ie, dw? + (1-2) aBPTI-AM > 5, which is impossible, since MBP + (1 — A)? < Ap. Similarly, if Ao + (1 =2)q 1). Proof If x; b**”, and let € = b*b” — b**”. Pick a rational number r > x + y such that 0” — 5**” x and p; >y. Since the laws of exponents are true for rationals, we get T =P > PP Hence BDU

f(xo). Refer to Fig. 10-7 and the defi- nitions on pp. 54 and 31. Pick x1 0, we need €2 < (€ — F(%o))/ [/le2) — fleo)]). Then if 0 d by using the inequality Flos) — feo) See) > feo) #2 (& - x0) and an argument like the one just given. The case x > xo is similar, Thus there is an open interval J about xo such that f(x) >d ifx is inJ. Thus, by the definition of continuous function, f is continuous atx. Solved Exercises 12. Suppose that f(x) is convex, a 0. Exercises 18. Give the details of the part of the proof of Theorem 3 dealing with the cased 0 for allx OI) = x5 Differentiation of the Exponential Function Now we turn our attention to the differentiability of b*. Again, convexity will be an important tool. Theorem 4 If b> 0, then f(x) = B* is differentiable and S@)=FOME), ie, LH =f'O-0 Proof From the equation f(x) = f(x —xo)f(xo) and the chain rule, all we need to do is show that f(x) is differentiable at x = 0. Refer t0 Fig. 1U%, Let x; > U and consider the line through (0,1) and (¥2, flea), ie, yer fda} - 1@) This line overtakes f(x) at x = 0. Indeed, if 0 In(x) by Solved Exercise 11, with x, 0, replacing x,,2, andx3. Fig. 10:8 The geomery needed for Theorem 4.DIFFERENTIATION OF THE EXPONENTIAL FUNCTION 139 In exactly the same way, we see that for x, <0, the line J, passing through (x1 ,f(21)) and (0, 1) is overtaken by fat x = 0. Going back to the definition of derivative in terms of transitions (Theorem 4, p. 27), we let A= the set of slopes of lines /; which are overtaken by fat x = 0 and B = the set of slopes of lines /, which overtake fat x = 0 Let A = (—e,a) or (—e,a] and B = (8,00) or [6,<2).We know that a 0. Then qx fled=1_ foe) 2 o- aH My aye bt Te ~ i ‘Now we may use the convexity inequality which tells us that f(¢) < L() if 0 ¢ But g(r) = b%(b" — 1) is continuous, and g(0) = 0. Thus if r is near enough to zero we would have b*(b' — 1) Hence expp (x) = expo (¥ logo) Differentiate by using the chain rule: exp) () = expto (& Logie b) * Logi b Setx =0: exp} (0) = expo (0) *logio b If we pick b so that expo (0) logiob = 1 aM ‘then we will have exp) (x) = expp (x), as desired. Solving (1) for b, we have 1 10810 = 50D) That is, We denote the number expro [1/expjo(0)] by the lette: e. Its numerical value is approximately 2.7182818285, and we haveDIFFERENTIATION OF THE EXPONENTIAL FUNCTION 141 expe (2) = expe (x) Although we started with the arbitrary choice of 10 as a base, itis easy to show (see Sulved Exercise 15) that any initial choice of base leads to the same value for e. Since the base e is so special, we write exp(x) for expe (x) =e*. Logarithms to the base ¢ are called natural logarithms. We denote loge x by Inx. (The notation logx is generally used in calculus books for the common logarithm logyox.) Since e! = e, we have Ine = 1. Worked Example 8 Simplify in(e*) + In(¢*). ‘Solution By the laws of logarithms, n (e*) + ine) = In(e* + €*) = Ine?) = 2. We can now complete our differentiation formula for the general expo- nential function expp x. Since & = e?, we have b* = e*M?. Using the chain rule, we find “Thus te mysterivus favtur eaph (0) tuins out to be just the natural Togerithon of b. Worked Example 9 Differentiate: (a) f(x) = e**; (b) g(@&@) =3*. Solution (@) Letu = 3x so e** =e" and use the chain rule: This expression cannot be simplified further; one can find the value In3 ~ 1.0986 in a table or with a calculator.142 CHAPTER 10: THE EXPONENTIAL AND LOGARITHM FUNCTIONS Solved Exercises 14. Differentiate the following functions. (@ e* () (© xe* (@) exp? + 2x) @x 15. Show that, for any base b, expp (1/exph (0) =e. 16. Differentiate: @ a () © rm @ Ginx)? 17. Prove that for t > 0 and b > 1, we have bf — 1 <(6* log, b)t. Exercises 21. Differentiate the following functions. @ et (b) sin(e*) (©) 3-21 (a) er* (e) tan (3?*) (D el 433 22. Differentiate (assume f and g differentiable where necessary): (a) 2 +2 — ye tH (b) e* —cos(@ +e2*) © (ey 8) @ ("1 +1ye! 1) © fox) e +266) () fer © fee @) AE +200) 23. Show that f(x) =e isan increasing function. 24. Find the critical points of fx) =x7e™. 25. Find the critical points of f(x) = sinxe*, —4n 0 (b) Find (d/dx)(x"). 20. Use logarithmic differentiation to calculate dy/dx, if y is given by y= (2x 3 7/Vxt 1. 21. Differentiate y =x"),PROBLEMS FOR CHAPTER 10 145 Exercises 27. Differentiate: (@) In(2x +1) (b) In(x? — 3x) (©) In(tanx) (a) (nx)? (e) (e -2x)InQQx+1) ( **B* 28. Problems for Chapter 10 rh 10. (@® [In(tan 3x)] /(1 + Inx?) ‘Use logarithmic differentiation to differentiate: @ y=x* () y=x* © y= Ginx) @ y= +1"? (©) y = - 2)? (4x + 3)°/7 ‘Simplify: (@) e*[in(e*!) Different @ ee (b) x © 14x? 8sinx @ = (©) In +x) (© (nxy?* (g) sin(x* + 1) + logy (14x ~ sinx) in(e!-*)] (b) en 3402) . Sketch the graph of y = xe™*; indicate on your graph the regions where y is increasing, decreasing, concave upward or downward. . Pind the minimum of y =a* for in (0,=). ‘Suppose that f is continuous and that f(x +») = /(x)f() for all x and y Show that /() = 6* for some b. [Hint: Try showing that f is actually dif- ferentiable at 0.) Let f be a twice differentiable convex function. Prove that /"(x) > 0. Let f be an increasing continuous convex function. Let f be the inverse function. Show that —/” is convex. Apply this result to log, x. . Suppose that f(x) is a function defined for all real x. If.xy 1+x*, forx 2 1. [Hint: Note that e > 2 and show that the difference between these two functions is increasing.) 12. We have seen that the exponential function exp (x) estiefies the following relations: exp (x) > 0, exp(0) = 1, and exp’ (x) = exp (x). Let f(x) be a function such that O 0 Prove that 0 < f(x) < f(O)exp(x) for all x > 0. (Hint: Consider g(x) = Fledlexp 13. Let f(x) be an increasing continuous function. Given x9, let A= the set of f(x) whiere x x0 Show that /(9) is the transition point from A to D. 14. Show that the sum f(x) + g(x) of two convex functions f(x) and g(x) is con- vex. Show that if f is strictly convex, then so is f+ g. Use this to show that ‘f(e) = ax? + bx +c is strictly convex if a > 0, where a, b, and c are con- stants. 15. Suppose /(x) is defined and differentiable for all real x (a) Does f(x) > O'for all x imply that f(x) is convex? (b) Does f(x) > 0 for all x imply that f(x) is canvex? Strictly convex?11. The Integral An this chapter we define the integral in terms of transitions; .., by the method of exhaustion. The reader is assumed to be familiar with the summation notation and its basic properties, as presented in most calculus texts. Piecewise Constant Functions In the theory of differentiation, the simplest functions were the linear functions F(x) = ax +b. We knew that the derivative of ax + b should be a, and we defined the derivative for more general functions by comparison with the linear functions, using the notion of overtaking to make the comparisons. For integration theory, the comparison functions are the piecewise constant functions, Roughly speaking, a function f on [«,6] is piccowiao constant if 2,5] can be broken into a finite number of subintervals such that f is constant on each subinterval. Definition A partition of the interval [a,b] is a sequence of numbers (Cost1,--+5tn) such that asta Sty So ae a where Af180 CHAPTER 11: THE INTEGRAL We shall verify shortly that the definition is independent of the choice of adapted partition If all the k's are nonnegative, the integral f2 f(r)dt is precisely the area “under” the graph of f—that is, the area of the set of points (x,y) such that 4 <4 ky At; = (2G) + OE) + GE) Sa etataee which ie the exme anewer we obtained from the first partition. Theorem 1 Let f be a piecewise constant function an [a,b]. Suppase that (tostisss-stn) and (S0,515---4Sm) are adapted partitions, with f(t) = kion (teat) and f(t) = jz on (s4-1,5). Then 2 kyan= > rds Proof The idea is to reduce the problem to the simplest case. where the second partition is obtained from the first by the addition of a single point. Let us begin by proving the proposition for this case. Assume, then, that m= m+ 1, and that (S0,515+++ 55m) = Costisevosttsbesfieto--eata)i ie., the s-partition is obtained from the t-partition by inserting an extra point 1 between 1) and 1141. The relation between s’s, t's, /’s, and k's is illustrated in Fig. 11-5. The sum obtained from the s-partition 1s 3 HAS; = jr (S1 ~50) + °° + fa Gin — 8) + i242 — 9141) F777 TInt Ont] — 9m) fh te ie oy Thee flows ee ce ee —— i i inet than the (partition.152 CHAPTER 11: THE INTEGRAL Substituting the appropriate &’s and /’s for the sand s’s converts the sum. to Kex(ts — to) #00 + ra t(ta— 1) + Reason — te) + 000 helt — tnt) We can combine the two middle terms: Fats <1) + krGait — te) = Reea(te Hann — any =D ke tj i which is the sum obtained from the ¢-partition. (For a numerical illustr tion, see Worked Example 2.) This completes the proof for the special case. To handle the general case, we observe first that, given two parti- tions (Fo,..-5tn) and (S0,--+45m), We can find a partition (to,...,tp) which contains hoth of them talzing all the points foy.--sfys8ose+es8rus elit inating duplications, and putting the points in the correct order. (See Solved Exercise 1, immediately following the end of the proof.) Adding points to an adapted partition produces another adapted partition, since if a function is constant on an interval, it is certainly constant on any subin- terval. It follows that the w-partition is adapted to f if the s- and t-parti tions are. Now we can get from the r-partition to the u-partition by adding points one at a time, By the special case above, we see that the sum is un- changed each time we add a point. so the sum obtained from the u-parti- tion equals the sum obtained from the r-partition. In a similar way, we can get from the s-partition to the w-partition by adding one point at a time, so the oum from tho u partition oquals the oum from the »partition. Since the sums from the f- and s-partitions are both equal to the sum from the u-par- rion, they are equal to each other, which is what we wanted to prove. Solved Exercises 1. Consider s- and r-partitions of [1.81 as follows. Let the s-nartition be (J, 2, 3,4,7, 8), and let the r-partition be (1,4,5, 6,8). Find the correspond- ing u-partition, and show that you can get from the s- and r-partitions to the u partition by adding ono point at a time.UPPER AND LOWER SUMS AND THE DEFINITION 183 2. Let (0) be defined by 2 if0