Solution Manual for A Journey into Mathematics Joseph J. Retman written by Joseph J. RotmanHis answer trickled through my head, like water through 3 sieve Lewis Carroll ieee 1.1. Find a formula for 1 + Zj2, jij and use mathematical induction to prove that your formula is correct oe « Adwseet The sums for n= 1,2, 3, 4,5 w 2, 6, 24, 120, 720 These are factorials; better, they are 2!, 3!, 4!, 5!, 6! We have been led to the guess S(n): 44 5j21 J = (n+ Dh a We now use induction to prove that the guess is always true The base step S(1) has already been checked; H-ts-en-the-tst For the inductive step, we must prove Sn + 1:4 + Eth jy = (n+ 20h Rewrite the left side as [1 + Zyb4 dij] + (n + 1)M(n + 1). By the inductive hypothesis, the bracketed term omthe-left-cide is (n + 1)!, and so the left side equals (n+ 1)!+(n+ 1)l(n + 1) = (n+ AU + (n+ A)] = (n+ 1) + 2) = (n + 2)! By induction, Stn) is true for all n21 1.2.1f r= 1, prove, for all nz 1, that tere rea rd a tem tc (r= DMr- 1) = The statement S$(1) is true because the left side is 1 (this is what the formula 1 +r +r2 +15 +~ + rt means when n = 2), while (r= 1)Ar - 4) #1 as well [r - 1 # 0 because ra 4), =~: S(4)=2.Inductive Step: We state S(n + 1) [nee rae Se eB a r= (ett - A)/Ur - 1) The inductive hypothesis allows us to rewrite the left side as (r= Mr - tye rhe OB at (r = 1)r™)/(r - 1) = (rth - 4)/(r - 1). One can also prove the formula by showing that (re tae re ree rd ee emt) e rhe The proof here does not use induction explicitly (but it uses other results needing induction in their proofs; eg. the generalized distributive law) 3 (r- Date rte? +rmd)s erd¢re reeds aretye (pert ree rd ee emt) ertree rene = net) era 4.3. Show, for all m2 1, that 10" Jeaves remainder 1 after dividing by 9 « We prove, by induction on n> 1, that there is an integer Gn with 107 = 9qq +1 Base Step: n = 1. Since 10 = 9x1 +1, we may set qi = 1. Inductive Step: We are to prove S(n + 1) 40Pt1 = 9q,.1 +1 for some integer Gn+iThe inductive hypothesis gives an integer dn with 10 = 9qn +1 Hence 10n+t = 10x10" = 10(9qn + 2) = 9(10gn) + 10 = 9(10an) + 9 +2 = 9(10an + 1) +4 Define dnet = 10dn * 1: it is an integer because dn is 14.1f a < re positive numbers, prove that a” sb" for all integers n ba 20 = Base step. ao = 1 = b? Inductive step. a®*? = aa" < ab", using the inductive hypothesis and Theorem 1.4(i), and ab" ¢ bb” = bO*!, again using Theorem 1.4(i). 1.5. (i) Prove that 12+ 22+ ~ +n? = n(n + 1)(2n + 1/6 « Base step: 1(1 + 1(2 + 1)/6 = 1 Inductive step [124 224 a n2]s (ne 2)? = Enln + 1)(2n + 19/6) + (n+ 1)? (n+ 1)(n + 2)(2n + 3/6. (ii) Prove that 13+ 22+ ~ 4n3= (14245 4 n)? « Base step: When n = 1, both sides equal 1Inductive step: 1343404 ne (ned = (1424-4 n+ ins 08 = [n(n + 1)/2]? + (n + 18 = (n+ 1)%(n + 2)°/4 e(142+s4¢n+(n4+1)), using Theorem 1.6 (iii) Prove that 144 244 ent = 8/5 + 4/2 43/3 - 0/30. « Base step: 41/5 + 1/2 + 1/3 - 1/30 = (6 +15+10- 1)/30= 1 Inductive step’ (144+ 244 + nf] + (n+1)4 5/5 + n4/2 +n5/3 - n/30 + (n+ 1)4 (n + 1)8/5 + (n + 1)4/2 + (n+ 1)9/3 - (n + 1)/30. 4.6. (i) Find a formula for ay = 13435458 +... + (an - 10%, and then prove that your guess is correct using induction « Comparing small values of a, with corresponding values of n4 and n2 ultimately leads to the guess: ay = 2n4 - n?. The proof by induction follows. Base step: If n= 1, then ay = 1, while 2x147-12 +1Inductive step: We must prove that ape = 13433453 +. + (2n- 154 (2n + 195 = 2(n + 4)4- (n+ 1)? Now 43433453 4..4(2n- 18+ (2n + 1)9 \edtats (13 + 33453 +. + (an - 158] + (2n + 18 = ant -n? + (2n + 1)? It is now a matter of high school algebra to expand this and compare it to the expanded form of 2(n + 1)*- (n+ 1)? (ii) Give a second proof of part (i) based on Exercise 1.5(ii) and the following observation: If ban = 1° +25 +3 3 then bon = an + 122+ 43 +... + (2n)9) = an + 913 + 23+ +n] = an + 8bp * By Exercise 1.5(ii), by = 4n2%(n +1)? and bon = %4(2n)2(2n + 1) By the hint, an = ben = (2n)@(2n + 1)? - B[4n2in + 1)7) Yalan2(4n?2 + an + 1) - 8n2(n? + 2n + 1)) n2(4n? + 4n + 1) - 2n2%(n? + 2n + 1)= 4nt+4n34n2-(2nt+4 = 2nt-n 1.7. i) If n = ab, where n, a, and b are positive integers, then either as /n or bs Jn. If, on the contrary, a> /W and b> Jn, then n=ab> J/n/n=n, a contradiction. (ii) If n 2 2 is a composite integer, prove that it has a prime factor p with p< Jn. « Since n is composite, n = ab, where 2 0. « We prove the inequality by induction on n 2 1. The base step n = 1 is obvious. For the inductive step, (2+ xpd s (1+ x2 + x)" > (1 + x)(1 + nx) [because 1+ x> 0] e1+(n+i)x4nx? 21+ (n+ t)x (The binomial theorem should not be used here.) 1.9. (i) Prove that 2" > n§ for all n 2 10 x Base step: 29 = 1024 > 103 = 1000. (Note that 29 = §12< 93 = 723.) Inductive step (n » 10 implies n 2 4): The inductive hypothesis is 2" > n°; multiplying both sides by 2 gives antl 2 2x27 > 2n3 = m3 + n3 2 nd + an? nd +3n24n2>n343n2+4n nd 43n24+3n41 = (n+ 1) wv (i) Prove that 2"> n4 for all n2 17. « Base step: 217 = 131072 > 174 = 83521 (Note: 164 = (24)4 = 216) Inductive step (n = 17 implies n 2 7). The inductive hypothesis is 2" > n4, multiplying both sides by 2 gives antl. 202%) and a nt 4nt2 nt + Sn né44n3+n32n4+ 4n3+7n?nf + 4n3 + 6n? +22 n#+ 4n9+ 6n? + Sn nf+4n3+6n2+4n+1=(n+1)4 1.10, Let g1(x), ... Gn(x) be differentiable functions, and let {(x) = g1(x)en(x). Prove, for all n > 2, that the derivative £60) = gf (xdga(x) nla) + Zi22 9768) @1-1 ve; (20814100 "e900 « Base step n = 2: the usual product rule for derivatives. Inductive step: Define h(x) = e104)" @n(x) = f00)/ens1(0). Rewrite the conclusion: f(x) = E924 ej (x)f/g j(0; now f(x) hOOgn+ 100)" W'Gdgn srl) + HOxe'y aa) [af 20h (x) /e (Oden 100 + 100/8n 41 O0]B gag 0 4.41, Prove that every positive integer a has a factorization a = 3m, where k20and m is not a multiple of 3 « Second form of induction on a 21 Base step: Take k = 0 and m= 1 Inductive step: f az 1, then a is either not a multiple of 3 or it is. If a is not a multiple of 3, then take k = Q and m =a. If a is a multiple of 3, then a = 3b. By the inductive hypothesis, b= 3'n, where @2 0 and n is not a multiple of 3. Hence, the desired factorization of a is a= 3'*tn1.12. Prove that 2% 4 Base step. If n = 4, then 2" = 24 = 16, while n! = 4! = 24, and 16 < 24 Inductive step. We must prove the statement for n+ 1, namely, 2"*1 < (n + 4)!, Multiplying both sides of the nth statement 2 > n! by 2, we have antl s 2x29 ¢ 2xn! But since 4 > sin Scot Bsec © Now sin @cot 9sec @ = sin 6(cos @/sin 8)(1/cos 8) = 1 Therefore, {4(sin 8 + cot © + sec © 1, Y%lsin 8 + cot @ + sec 8) > 1 and sin @ + cot @ + sec 6 2 3 1.45. Prove that if a1, a2, ~, a, are positive numbers, then (ay tag + + ay)(L/ay + 1/ag + ~ + L/an) = n2 « By Theorem 1.11, [(ay + az +~ + an)/n)® 2 ajagay and [1/ay + 1/ag + ~~ + 1/ay)/n]” 2 1/ayag~ an. Therefore, [(ay + ag + © + aq)/n)P[1/ay + 1/ag + + 1/ay)/n]? 2 ayag~a,/ayag~ay = 1 Taking nth roots, we have [lar + ag + ~ + ap)/n][1/ay + t/ag + ~ + 1/an)/nj 2 4, 2 and so (ay + az 4“ + anM(L/ay + 1/ag +» + Lan) 2 4.16. For every n= 2, prove that there are n consecutive composite numbers. «If 2¢asn+4,then a isa divisor of (n + 1); say, (n + 1)! = da for some integer d. It follows that (n + 1)! +a = (d+ Da, and so (n + 1)! +a is composite for all a between 2 and n+ 1.1 1.17. Show, for all r with O as well as the sum of the 8 boxes 2.14. Give another proof of the Pythagorean theorem (attributed to U.S. President James A. Garfield). Take a vertical line CC’ of length a + b, and construct two replicas of AABC as in Figure 2.28Figure 2.28 « Construct Figure 2.28 by drawing two copies of the original right triangle: AABC and AAB'C’, Notice that 4BAB = 90°, for the sum of the three angles at A is 180°, and the other two angles at A are the acute angles of AABC. Thus, CBBC’ is a trapezoid whose area T can be computed in two ways. On the one hand, Exercise 2.2 gives T = ‘(a + b)x(a + b); on the other hand, T is the sum of the areas of the three triangles: T = %ab + %ab + 4c? Therefore, Yala +b)? = ab + %c2, and high school algebra gives a? + b? = c? 2.15 (i) Ina triangle with sides of lengths 13, 14, and 15, what is the length of the altitude drawn to the side of length 147 « A “slick” way is to see that one can construct a triangle with sides of lengths 13, 14, and 15 by pasting together a right triangle with sides 5, 12, 13. and a right triangle of sides 9, 12, 15, as in the figure below9 12 13 5 It follows that the length of the altitude drawn to the side of length 14 is 12. Here is a less tricky solution. The altitude divides the side of length 14 into two pieces of lengths a and b, so that a+b = 14; it also divides the triangle into two right triangles. The Pythagorean theorem gives the equations h2 = 152 - h? = 132 - Hence, 15? (132 = b?) = 0, so a2 - pb? = 152 - 13% = that is, (a - bla +b) = But a+b = 14, so that a-b system: at+b=14 a-b= 4 ine: v2, that 226 - 169 = 56 We solve the a= 9 and b= 5. Finally, h? = 152 - a? = 225 - 81 = 144, and so h = 12. (ii) Find the area of this triangle « The area is %x12x14 = 84 2.16. Given a right triangle with perpendicular sides a and b, find the side s of the inscribed square.a Figure 2.29a « The area of the given triangle is A = “ab. A diagonal of the square divides the given triangle into two triangles, as in Figure 2.29. Thus A= Yass %bs = (a+ bs From “ab = %la +b)s, we obtain s = ab/(a + b) 2.17. Given a right triangle with legs a and b, find the radius r of the inscribed circle Figure 2.30a * Draw line segments from the vertices of the triangle to the center of the circle. The triangle of area %ab is thus divided into three triangles. It follows that Yab = “ar + “br + Yer, where c is the length of the hypotenuse, Thus a+b+te)r, and r= ab/(at+b+c),2.18. The length of the perimeter of a right triangle is 60 units, and the length of the altitude perpendicular to the hypotenuse is 12 units. How long are the sides? «If a and b are the lengths of the perpendicular sides and ¢ is the length of the hypotenuse, then at+b+c=60 and a?+ The area of the triangle can be computed in two ways: it is Yab as well as %4x12c. Therefore, ab = 12c It follows that (60 - c)? = (a +b)? = a? + b* +2ab = c2 + 2ab = c? + 24c, or 3600 - 120c = 24c. Thus, 144¢ = 3600, and hence c = 25. It follows that 35 and ab = 300 a+b Setting b = 300/a gives a+ 300/a = 35; we rewrite this as a® - 35a + 300 = 0. The quadratic form! a now gives a=15 and b= 20 2.19. A cylindrical column has height 30 feet and circumference 8 feet. A garland is wound evenly, in spiral fashion, 5 times around the column, reaching from top to bottom. Show that the garland must be at least 50 feet long. xThink of the cylinder as a 30 by 8 rectangle which has been rolled up. If we unroll the cylinder, we see five rectangles, each 6 by 8, as in the figure.The portion of the garland in any of these small rectangles must go, in some way, from the upper right corner to the lower left corner. The shortest such path is the straight line, that is, the diagonal, which has length 10, by the Pythagorean theorem, The garland is thus of length at least 5*10 = 50 feet 2.20. (i) Show that an equilateral triangle A with side lengths a has altitudes of height a/3/2 «The Pythagorean theorem gives a2 = (a/2)2 + h?, and this gives h = a/3/2. (ii) Show that the area of Ais a7 /3/4 a area + Yexaxal3/2 = a°/3/4 2.21. Find the area and circumference of the circumscribed circle of a regular hexagon whose sides have length s. « The radius of the circumscribed circle Is_s, and so Ag = 182 and L, = 2s2.22. Find the area and circumference of the circumscribed circle of an equilateral triangle whose side has length «Let r_ be the radius of the circumscribed circle and let h be the height of the equilateral triangle. By Exercise 2 20, h = 3/3/2 Moreover, Yh = 3f3/3 Thus A, = ares %ms* and Ly = 2nr =2ns/3/3 2.25. Is there a Pythagorean triple (a, b, ¢) with a = 1 or 2? 2 p2 = a2, so that (c + bc - b) 4; since there is only one x No. In a Pythagorean triple, Baiee eet et athenttag factorization of 1 into positive integers, s0 that ¢ + B = 1s pee rete ee acme deat from (c + be - b) «4. Either c+ b= 2=¢- B oF c+b=4 and ¢- b= 1; in the first case, b = 0, and in the second case, © = 5/2 (y that (781, 2460, 2581) is a Pythagorean triple x One can verify that 781? + 24602 = 25812 609961 + 6051600 = 6661561 or one can check that 761 = 4° ~ 2 +p, where q= 41 and p= 30 , 2460 = 2ap, and 2581 = Gi) Verity that (3 993, 7 972 024, 7 972 025) is Pythagorean triple «A hand calculator with an 8 or 9 digit display is inadequate for this problem30 (3993) = 15944049 (7972024)? = 63553166656576 (7972024)? = 63553182600625. Here is an alternative way. Note that c= b+1.If thisisa Pythagorean triple, then +b2=(b+1)* = b? + 2be4, and b= %(a® - 1). And this does, in fact, hold here, for (39932 - 1) = 7.972 024 (ii) Is (169568, 1139826, 1152570) a Pythagorean triple? « Yes., for one can find q = 8i3and g = 701 2.25. Show that no two of the Pythagorean triples in Eq. (2) are similar. x In each case, show that c/c’ = a/a’ and c/c' # a/b’. For example, (5, 12, 13) and (8, 15, 17), for 13/17 = 5/8 and 13/17 = 5/15 (as one checks with cross multiplication) 2.26. Prove that if b = 4(™31) for some m 2 2, then there is a Pythagorean triple (a, b,c) with c= b + 4. Conversely, prove that if there is a Pythagorean triple (a, b, c) with c= b+ 1, then there is an integer _m 2 2 with b= 4(™31) *If b= 4(Mg1) = 2m(m + 1) = 2m? + 2m, it is only a question of finding a with a2 + b2 = (b+ 1)? = b? + 2be1 Thus, a? = 2b+1 + 2m2+4m+i = (2m+1)*, and a= 2m+t Hence, (2m +1, b, b + 1) is a Pythagorean triple.If a2+b2 = (b+ 1)? = b? + 2b +4, then a? = 2b + 1. There are two consequences: a is odd, say,a = 2m+1;b= %(a*- 1) = sla + 1)la - 1). Hence, b= %”latiia- 1) lam + 232m = 4(™33) 2.27. Let & be a right triangle with legs a and b and hypotenuse ¢. In Exercise 2.17, one found that the radius r of the inscribed circle is given by r= ab/a+b+c) [r is called the inradius of A] Ii (a, b, ¢) is a Pythagorean triple, prove that the inradius of the corresponding right triangle is an integer. “ r= ab/la+b+c) = 2aplq? - p2)/l2qp + q? - p? + a? + p? = 2qp(q + pq - p)/2alp + a) = plq - p). 2.28, Let A be a right triangle with side lengths a, b, and c all integers. Prove that the height h of the altitude to the hypotenuse is a rational number. (Hint. What is area(A)?) « Area(A) = %ab = %he, and so h = ab/c is a rational number. 2.29. Find the values of g and h for t= %, %, 3/5, and 4/5. . t= % % 3/5 4/5 g- 5/13 7/25 8/17 9/44 mehr QR) drwebres Saws S3e h = 12/13 24/25 15/17 40/41 312.30. Find q and p in Theorem 2.12 for each of the following Pythagorean triples. (i) (7, 24, 25). (ii) (129396, 261547, 291805) «qe 526, p = 123, 2.31. Show that the same number can occur as a leg in two non-similar Pythagorean triples. « (6, 8, 10) and (8, 15, 17) are Pythagorean triples that are not similar, 2.32. Show that there are distinct Pythagorean triples (a, b, ¢) and (cx, p, c) having the same ¢ « if (a, b, ©) and (x, y, 2) are Pythagorean triples with c¢ = 2, then (az, bz, cz) and (cx, cy, cz) are Pythagorean triples with the same hypotenuse. Thus, (3, 4, 5) and (5, 12, 13) give distinct Pythagorean triples (39, 52, 65) and (25, 60, 65) 2.33. Show that every integer n = 3 occurs as a leg of some Pythagorean triple. «If n is even, say, n= 2k (k 2 2), then (n, k? - 1, k2 4 Lisa Pythagorean triple because (2k)? + (k? - 1)? = 42 + kt - 2k? + 4 ki‘ + 22 4 1 = (k? +1)? a similar calculations shows that if n is odd, say,n = 2k+4 (k 2 1), then (n, 2k(k + 1), 2k? + 2k + 1) is a Pythagorean triple.2.34, Use Heron's forrnula to solve Exercise 2.15(i) « The sides have lengths 13, 14, and 15, so that the semiperimeter s = %(13 + 14+ 15) = 21. 1f A is the area, then A2 = 21x8*7x6 = 7056, and so A = 84. 2.35, Let P = (1, 1),@= (2,3), and R = (-2, 2) be three points in the plane (i) Show that P,Q, and R are not collinear « It suffices to show that the slopes of the lines PQ and QR are different. The first slope is %, and the second slope is 4 (ii) Find area(APQR). = \ appro) « The distance formula 7 AS IPQl = (52.24; |0Pl = rohan IPR». /10 @ 3.16 Thus, the semiperimeter is s © 4.76, and so Heron's formula gives: (area)? = 2.24 « (4.76 - 2.24) « (4.76 - 4.12) x (4.76 - 3.16) = 4.76 x 252 + 0.64 « 1.6( 12.28, eae and area(APQR) 8 3.50. 2.36. To find the distance between two points A and C on opposite sides of a river, a distance of 100 feet 1s paced off on one side of the river from A toa point B with AB perpendicular to BC. The angle 4CAB is measured as 60° Find the distance IACI.wy Bo 400 7 Figure 2.40 * IAC] = |ABIcos60° = 100x% = 50. 2.37. A pole 30 feet tall casts a shadow 50 feet long. Find the approximate angie of elevation of the sun » If o is the angle of elevation, then cos a = 30/50 = 6, and 53.1" 2.58. New Orleans, Louisiana is due south of Madison, Wisconsin. If the latitude of New Orleans is 30°N, the latitude of Madison is 43°N, and the radius of the earth is 4000 miles, what is the distance between the two cities? = Since the radius of the earth is 4000, the circumference of the earth is 8000m. Now 13° (43 - 30 = 13) corresponds to 13/360 of the circumference, and the distance between the cities (on the surface of the earth) is 13/360*8000n (5) 907.6 miles. pie? 2.39. Assuming the fact that the arc and the chord subtended by a small central angle in a circle are approximately the same length, compute the diameter of the sun using the facts that the distance from the earth to the sun is about 93,000,000 miles and that the sun as seen from the earth subtends an angle of 0.53 degrees. 34« The arclength in miles is approximately of[o* ii 53/360 «93,000,000 (= 430,000 Using the assumed fact, this is the sun's diameter 2.40. If a regular decagon whose sides are 7 inches long is inscribed in a circle, what is the radius of the circle? ™ Connecting each vertex of the decagon to the center of the circle dissects the decagon into isosceles triangles, with equal sides r (the radius of the circle), the other side of length 7, and with the angle at the center 36, The perpendicular bisector thus splits this triangle into two right triangles, each with hypotenuse r, angle 18°, and opposite leg 3.5. Hence, cos 18° =3.5/r, and so r = 3.5/cos18° = 3.68 2.41. Let a be an angle and iet A = (cos a, sin «). Show that if m_ is the slope of the line OA (where O denotes the origin), then m = tan « » The slope of the line joining points (a, b) and (c, d) is m = (b - d)/fa - c). Here, m = (cos a - 0)/(sin a - 0) = tan o 2.42. Let (g, h) be a point in the first quadrant lying on the arc of the unit circle, and let the line @ joining (g, h) and (-1, 0) have slope t. Prove that if the line L joining (h, g) and (-1, 0) has slope T, then T= (1- t)/(1 +t). « By Eqs. (2), and (3), we have g = 2t/(1 +t?) and h = (2 - 12)/(4 + t?); by Eq. (4), we have T = g/(h + 1). Substituting,36 T= 2 + 2a - 4 + #2) + After simplifying, the term on the right is (1 - t/(1 + t) 2.43, Prove that 1/(ese @ - cot®) - 1/(csc @ + cote) = 2cot & « Rewrite in terms of sin @ and cos 9; then use Theorem 2.14; after simplifying, each side is (1 - t®)/t 2.44, Prove that cote + cot? = cscte - csc®e. * Rewrite in terms of sin @ and cos @; then use Theorem 2.14; after simplifying, each side is (4 -2t4 + 18)/1604 2.45. Of all the triangles with a given perimeter p and with one side of a given length a, prove that the triangle having the maximal area is isosceles. «If p= 2s, then Heron's formula give (s ~ b)ls - c) = A2/s(s - a), where A is the area of the triangle, and so the arithmetic mean-geometric mean inequality (Theorem 1.11) gives (4ls - b +s - cl) > A%/s(s - a), with equality if and only if s- b= s~ ¢; ie. if and only if b = c. Thus, the triangle with largest area is isosceles. 2.46. Prove, for every triangle A, that (p/4)? 2 areal), where p is the perimeter of A. Can there be equality?«Let p2a+b+c= 2s. By Heron's formula, s(s - a)(s - bls - 6) = A’, where A = area(A), and the arithmetic mean-geornetric mean inequality gives (4ls+s-ats-b+s-cD42 A? But sts-ats-b+s-c=4s-(at+b+c) = 4s- 25 = 2s= p, 50 that (p/4)4 2 A2. Taking square roots, (p/4)? 2 A The inequality is always strict, for it can occur only when s=s-a s-b-=s-c.But s=s-a gives a= 0, which is not a length of a side of a triangle 2.47. Prove the Law of Sines: In a triangle AABC with angles ct, 8, and ¥ opposite sides of lengths a, b, and ¢, respectively, sin o/a = sin B/b = sin ¥/c « Since area(AABC) = %bh = %besin «, we have sin a/a = 2area(AABC)/abe Since area(AABC) = %ak = %acsin p, we have sin p/b = 2area(AABC)/abe. Therefore, sin «/a = sin p/b. The hint is essentially the whole proof! a72.48, For any two angles a and p, prove cosl%ele + pdlcos[%(a - pI) = (cos a + cos B) « cosl%ela + piicost(a - 8)] (cos%acos%s - sinYeasin%sp)(cos% acosp + sin%asin¥es) = cos2%acos*%p - sin?%asin24p cos2% acos?%p ~ (1 - cos@%a)(1 - cos*%p) cos2% acos?%p ~ (1 - cos*%a - cos?%p + cos?% cos” ap) = -1 + cos*%a + cos*p = -1 + Ylcos x +1) + 4(cos B+ 1) = Yacos x + 40s 6 2.49. Prove that cos 20°cos 40"cos 80° = 1/8. = Using Exercise 2.48, cos 40°cos 80° = %(cos 120° + cos 40°) [lo +p) = 80 and %(a - p) = 40 give o +p = 160,a- p= 80, so that oa = 120 and p = 40). Hence, cos 20°cos 40°cos 80" = cos 20°4[(cos 120° + cos 40°)] = -%cos 20° + %(cos 20°cos 40°). Using Exercise 2.48 again,39 cos 20°cos 40° = %4(cos 60° + cos 20°) [Ala +p) = 40, 4la - p) = 20 give a +p + 80,0 -p = 40, so that a = 60 and Bp = 20) Hence, cos 20°cos 40°cos 80° -%cos 20° + %(%4(cos 60° + cos 20°) = -%cos 20° + %(%(cos 60" + %cos 20°) = Ycos 60" 1/8 2,50. (i) Prove that cos x cos 2x cos 4x~-cos 2x © sin 20*1x/2"*1sin x. * The proof is by induction on n 2 0, and the idea is to multiply the left side by sin x Base step n = 0: The double angle formula sin 2x = 2sin xcos x gives cos x = sin 2x/2sin x. Inductive step: cos x cos 2x Cos 4x~cos 2°xcos 2"*1x = sin 29*!xcos 29*4x/29*1sin x = sin 29*2x/20*4sin x = sin 22*2y/2"*2sin x. (ii) Use part (i) to give a second proof of Exercise 2.49 = Exercise 2.49 follows from the case n = 2 cos x cos 2x cos 4x = sin 25x/25sin x = sin 8x/8sin x;40 set_x = 20° and obtain c0820"cos40"cos80" = sin 160°/8sin 20° The result follows, for sin 20° = sin(180° - 20°) 2.51, Prove that 2cos(n + 1)x = (2cos x)(2cos nx) - 2cos(n - 1)x. « We prove 2cos(n + 1)x + 2eos(n - i)x = (20s x)(2cos nx) by using Exercise 2.50 2eos(n + 1)x + 2cos(n - 1)x = 2[cos(n + 1)x + cos(n - 1)x] = 22eos([n + 1 +n - 1)/2)cos([n + 1 - n + 1)/2) = 4cos nx cos x = (2cos x)(2 cos nx). 2.52. Prove the half angle formula for tangent tan(8/2) = sin @/(1 + cos 8). « In Figure 2.41, we see that t = tan(@/2) is the slope of the line joining (-1, 0) to (cos 9, sin @). But Eq. (1) gives tan(@/2) = = h/g + 1) = sin @/(cos 8 + 1) 2.53. Find the indefinite integral Jsin 9d9/(2 + cos 8). » After substituting using Theorem 2.14, the integral is Jatai/(t + 123 + 12) The partial fraction decomposition is 4t/1 + 123 + 12) = aes + 12) - 243 + #7), so that the indefinite integral is41 Inia + t2| = InlS + #7) = Inia + 12)/(3 + 291 By Eq. (4), t = sin @/(1 + cos 9). Anything more is cosmetic 2.54. Find the indefinite integral [[sin 9 - cos elde/[sin @ + cos @] « After substituting using Theorem 2.14, the integral is fa(-a + at + t2)at/(a + 2t - 21 + t7) The partial fraction decomposition is Q(-1 + Qt + t7)/(2 + at - #204 + 42) = (Qt 2/1 + at - 02) + at/t + 2), so that the indefinite integral is -inli + 22) + ina + t= Inia + t2)/(4 + 2t - 2D) By Eq, (4) and some manipulation, the indefinite integral is -inlcos 8 + sin él 2.55. Find the indefinite integral [(/x - L)dx/(/x + 1). * The substitution /x = 1/t rewrites the integral as f-2 - ddtv(a + pe The partial fraction decomposition is 21 = yd + tt = ae + ast? - ast + a +t) Therefore, rata = tidt/(a + t)t3 = Ainia + 1/4 -4/t + 1/2,Recalling that 1/t = Jz, the indefinite integral is 4in(fx + 1) - 4/x + x, 2.56. Reduce the indefinite integral [x"/x7 + 1 dx, for any integer n = 0, to an indefinite integral of a rational function. = The appropriate substitution is: x= (14 t2)/0 - 13), so that y = 2t/(1 - t2), and dx = 4tdt/(1 - t2)? lor we are in the case of a hyperbola). The integral is rewritten as fat2(a + t2)Rat sca - +2943 The drudgery of partial fractions is not recommended. 2.57. (i) Show that the ellipse with equation x?/a? + y?/b? = 1 can be parametrized by x = acos @ and y = bsin ©. 2 x (acos @)?/a? + (bsin 9)?/b? = cos?@ + sin?9 = 1 (i) Show that if a> b, then the arclength of this curve is given by the integral 25")? Va? = (a? = b%)c0570 ao. « Ifa curve is parametrized by x = x(9) and y = y(@), then its arclength is given by J /(dx/de)? + (dy/de/ de. Here, the expression under the radical is a2sin29 + b@cos%e = a2 - (a? - b)cos?o (iii) Show that the tan 0/2 substitution rewrites this as an elliptic integral of the form f R(t, /f(t))dt, where R(u, v) isa rational function of two variables and f(t) is a quartic polynomial.« The substitution gives cos @ = (1 - t2)/(1 +t?) and de 2dt/(1 + t?). The integral becomes far + (at - b2)coste do = [VR art + 12)", where f(t) = b2t? + 2(2a2 - b2)t? + b?44 Chapter 3 3.1. li) For n> 1, define ky = 1 - (1/10)"; show that k. 7 1. « Clearly, ky ¢ kp «kg ¢ ~~ ¢ 1; moreover, = ky = 4-1 - (1/10)7] = (1/10)" < %P for all Gi) For nz 1, define é, = 1 - 2(1/10)"; show that & 71 ™ Clearly, @, ¢ &9 < #3 < ~ ¢ 1; moreover, 1 - 0, 2 1 - [1- 2(1/10)") = 2(1/10)" < 4? for all n. 3.2. Show that 1.1, 1.01, 1.004, ... approximates 1 from above. * Define Ky = 1+ (1/10) for all n> 1. Clearly, Ky > K2 > Kz > > 1 for all n; moreover w Ky - t= (1/10)9 « 4" for all n. 3.3, Use Theorem 3.1 to prove (again) that the sequence .9, .99, 999, ... approximates 1 from below. « The sequence in question is ky = 1 - (1/10)", which clearly is increasing. Now 1-kpet- 92 16% 5 We must also check, for all n> 1, that 1 - (2 = (1/20)44) « a4 - [1 = (2/10); that is, (1/10)P*2 « 44/10)" This last inequality follows from Exercise 1.4, for 1/10 < %\tnate 3.4 rafaty the proof of Theorem 3.3 to show that % is equal to (3333) (an unending string of 3's) «Let B+ 333K: )lan unending string of 3's). It is plain that B s % (because 3B <¢ 1), so that either B< % or B= %, Our strategy is to eliminate the first possibility. The sequence here is ky = %(1 - (1/10)"); that is, ky =.3, kg 33, kz = 333,¢). We show that this increasing sequence approximates % from below. For all nz 1, % - Ky = % ~ Volt - (1/1097) = 4(1/10)" < (1/20)" « 4 If B< %, then the Getting Close Principle says that there is some kn with B< kp; that is, there is some n 2 1 so that a string of only n 3's is larger than B = .33333~, a never- ending string of 3's. This contradicts the criterion for determining inequality between two numbers which are given as decimal expansions. We have eliminated the possibility B < ¥%, and the only remaining option is B= %. 3.5 (i). Show that if M_ is a positive number and ky / A, then Mk. 7 MA « By hypothesis, ky A and Ky- A‘ %4"A forall n21.It follows from Theorem 1.4 that MK, >» MKz > MK3 > ~ > MA and M(K, - A) = MK, - MA< %"MA for all n 21 * (iii) Use part(i) of this exercise, together with the fact (proved in Theorern’3.4) that the increasing sequence .9, .99, 999, . approximates 1 from below, to give a second solution to Exercise 3.4 dat,\edete 46 x In (i), take ky = 1 - (1/10)9, A= 1,and M= % 3.6. Prove that if a. 7 A andb, 7B, then a; + by, a2 + bg, a3 + bg, .. approximates A+B from below x It is easy to see, using Theorem 14, that a; + bi ¢ az+ be « as + by Knem < 2™*™A < 4A 3.8. Show that ky ¢ kz (kz <-~ ¢ 4 does not approximate 1 from below, where ky = 1- %" forall n21 ® If n= 1, then 1~- ky = % > %. Since one needs the inequality - kn < %" for all nz 4, this suffices. However, one sees that = ky = %P> %P forall n213.9. A pizzeria charges $2.50 for a 10” pizza and $5.00 for a 15" pizza (a 10” pizza is circular with diameter 10 inches). Should four hungry students order four 10" pizzas or two 15" pizzas? « The area of a 10° pizza is 257, so that four 10” pizzas give 100m square inches of pizza. The area of a 15" pizza is (7.5)°n = 56.25n square inches. Thus, two 15° pizzas having 112.50 square inches are more filling. 5.10. Complete the proof of Theorem 3.8 by showing that A'/A > r'2/r2 jeads to a contradiction: * lf A'V/A> v'2/r2, there is some number M' with M'/A = r'2/r2, so that A'/A > M'/A. Multiplying both sides by A gives Alo MM’. We have seen that area(P,) area(D'), where Py, is an inscribed regular 2"**-gon. Since M' < A’, the Getting Close Principle says that there is some inscribed polygon P; with M' < area(P;). Let Py be the corresponding polygon in D. By Lemma 3.7 area(P;)/area(P,) = ri2/r2 But M‘/A = r‘2/r2, so that area(P;,)/area(P,) = M'/A. Hence, area(P))/M' = arealP,)/A. This is a contradiction, for the right side is smaller than 1 (because P, is inside the disk D, hence has smaller area) while the left side is greater than 1 {for P, was chosen so that M'< area(P;)] 3.11. The inscribed polygon Py consists of 2°*1 congruent isosceles triangles, each with height hy. Prove that ha / r.Figure 3.10 « Clearly, hy < hg <3 < h. 1- cos 9, because 1 + cos @ > 1. 4849 3.12. For all n2z 1, let Py be the regular 2"-gon inscribed in a disk D of radius 1 (as constructed in the text). Prove, for all n> 1, that there are inequalities area(P,) < ¢ area(Py) + (4)"n x Theorem 3.6 gives area(P.) / area(D) = 1. Therefore, for all nei, 0 ¢ w- area(Py) < (4)"n Adding area(P,,) to all sides gives area(P,) <1 ¢ area(P,) + (4)"n 3.13, Let Py be as in Exercise 3.12, and let sy be the length of a side of Pp. (i) Prove that 5 = = In Figure 3.10, AOAB is an isosceles right triangle with equal sides of length 1. The Pythagorean theorem gives $1 = LB. (ii) Prove, for all n= 1 that —— sue 42-44 4 « In Figure 3.10, s, = !ABI and sys = IAC]. We begin by computing the altitude [OE] = hy; the Pythagorean theorem applied to AOEB gives 2 = JOE = 1 - 45% Similarly, 24. = lOFl= 1-454 nel nel Now IECl = 1-(OEl=a-h,=1- 1-454 The Pythagorean theorem applied to AACE gives50 ICEI? + |AEI? Sn+i = (1 - hy)? + (%s,)? 4- Qh, + he + 45% Mss, a-2/t~ Ms ei As + 48h Tis aeeiges The result follows from this if we rewrite the last term 3.14. (i) Prove that area(P) = 2. « By Exercise 3.12(i), each side of Py has length sj = /2; it follows that area(P,) = 2 (ii) Prove, for all n 24 that area(Ppy1) = 2"sp: « In Figure 3.10, we compute area(AGAC) by letting OC be the base and AE be the altitude. Notice that IOC] = 1, for OC is a radius, and that {ACI = %sp. Therefore, area(AOAC) = YIOCIIAE] = %(%5_) = %Sn: since Pye1 is dissected into 2"*? triangles congruent to AOAC, we have area(Py) = 29 t?%5_ = 2"5y 3.15. Use the previous three exercises to estimate (i) Show that 2¢w¢4 « Adisk D of radius 1 (and diameter 2) can be inscribed in a square Q with side lengths 2, and so51 m = area(D) < area(Q) = 2? = 4 By Exercise 312, a lower bound is area(P;) = 2 (ii) Show that 2,828427125 « m < 3.828427125. (Hint. Use Exercise 3.12 with the estimate m < 4 in the upper bound.) « When n 2, the inequality in Exercise 3.12 becomes area(Po) ¢ t ¢ area(Pa) + (%)?n Now area(Pa) = 251 = 2/2 = 2.828427125, by Exercise 3.130), so that 2.828427125 By Exercise 3.14(ii), area(P3) = 4s2, and by Exercise 3.13(ii), soe 42- V4- 54. Neh rx Since 51 = /2, we have so(=) 0.765366864 and“ area(Ps) = 4 x 0,765366864 ©)3.06147456 Hence, 3.06147456 < < 3.06147456 + (4)°GB Sn < (4)84 = % = 5, we have Since (% 3.06147456 < mw < 3.56147456. oe (iv) Show that sz = 0.390180643, arealPa)(=)3 121445144, and 3.121445144 < m < 3.371445144 * When n = 4, the inequality in Exercise 3.12 becomes area(P4) < 1 ¢ area(P4) + (%)4r By Exercise 3.14(ii), area(P4) = 853, and by Exercise 3.13(ii), \ area(Pq) = 8 x 0.390180643 Ys 121445144 Hence, 3121445144 < m < 3.121445144 + (%)4mr, Since (%)#m < (4)44 = % = 25, we have 3121445144 < mw < 3.371445144. (v) Repeat this procedure until you can estimate m well enough to see that its first digits are 3.14 « We keep iterating this process until we reach an inequality of the form 3.140 < m < 3.149. We recall that area(P,) = 2" 1s,-1, and we define the nth upper bound UB, = area(P,) + 4/2"n Sn-4 2n-l area(Py) area(P,) + 4/2" = peck = 2h Asnot 2828427125 3.828427125 © 2 1414213562 3 0.765366864 4 — 3.061467456 3561467456 es 390180643 8 3121445144 9 3.371445144 5 196034280 16 3.136548480 3.261548480 6 098135347 32 3.140331104 3.202831104 7 049082455 64 3.141277120 © 3.172527120 8 024543073 128 3.141513344 3.157138344 9 012271763 256 3.141571328 3.149383828 3.16. Let Py be as in Exercise 3.12. Show that area(Py) = 2"sin(360/2") « In Figure 3.10, area(P,) = 2"*tarea(AOAB). If an = 4OAB, then Op = 360/2"*!, Since the disk has radius the altitude hy = cos 4, and the base sp = IABI = 2|AEI = 2sin 4p) Therefore area(AQAB) = %hpSp = 408 Gy 28in &y = “SIN 2a. Since 2ap = 2x360/2"*! = 360/2", we conclude that area(Py) = 29*1x%sin 2oy = 2%sin 2a, = 2"sin(360/2"). 3.17. Let AABC be a triangle in the plane, as in Figure 3.17, and let 8 be a concave polygonal path from A to B lying wholly inside it. 1 @ has n> 2 edges, prove that IAC| + |CBI > length @.« The proof is by induction on n 2 2, the base step n = 2 being done in the text. For the inductive step, let p be an (n + 1)- edged path from A to 8; its first edge is AD and the remaining n edges forma path p' from D to B, as in the figure c 7 Extend AD, and let it meet CB in the point E. Since the path p is concave, it is entirely inside the triangle AAEB. Let us compute. A JACI + ICBI > JAE! + IEBI = \ADI + IDE) + [EBL > JADI + length(s') = length(e). The first inequality holds by the base step, for AE +EB is a concave 2-edged curve inside AABC; the second inequality holds by the inductive hypothesis applied to ADEB. 3.18. Show, in Figure 3.21, that IAC} + ICD] + IDBI > JAEl + JEBI.Figure 3.21a « Draw an edge from A to D. As Figure 3.21 is drawn, the path from A to E to B lies inside the region bounded by AB and the path from A to D to B. By Exercise 3.21, |ADI + IDBI > |AE| + [EBL But IAC] + CDI > JADI, and this gives the result ‘A more general interpretation involves any concave 2- edged path inside a region bounded by AB and a concave 3- edged path from A to B. One can show that the inside path lies either in AADB or AACB, but this involves taking the definition of concavity more seriously than is being done in the text 3.19. True or false: the arclength of the curve y = x? between 0 and 1 is greater than the arclength of the curve y = x° between 0 and 1. No fair integrating « False. Both are concave curves from A = (0,0) to B= (1, 1), and since x? s x2 <1 for all x in [0, 1], the quadratic curve is inside the region bounded by the line AB and the cubic curve. The principle of concavity now applies. 3.20. The front wheel of a tricycle is 3 feet in diameter, while its two rear wheels are 2 feet in diameter. If, on a straight road, the front wheel makes 64 revolutions, how many revolutions do the rear wheels make? « The tricycle travels 64*31 feet, so that the number of revolutions made by the rear wheels is 64x3n/2m = 96.ew on 3.24, Let aj, a2, a3, () be a constant sequence, that is, there is some number c with a, +c for all n. Prove that ay > © « Let €> 0. If we choose @ = 1, then lan -cl=le-cls O 4. 3.22. Prove that |x - Bl < cif and only if B-e € 5.23. (i) If by @ B= 0 and all by, = 0, prove that there is an integer @ and some number N with 1/lb_l 0 because B» 0.Since by > B, there is an integer 2 with Ib, ~ BI < IBI/2 for all _n 2 é Since Ib, - Bl is the distance between b, and B, we are saying that for all n 2 @, by, lies in the open interval (B - !BI/2, B + |BI/2) o “BoB B %IBlG If B > 0, every point in this interval lies to the right of %B, and so its distance to 0 is larger than %8; that is, Ibpl = lbp - Ol > %B. If B <0, every point by, is to the left of -%IBl, and so its distance to 0 is also greater than %IBI. By Theorem 1.4liii), we have 1/lbyl < 2/IBI for all nz & Define N = 2/IB|57 (i) Assume that a, > A and by — B. Prove that if B= 0 and all b, #0, then ap/by > A/B. « Given € > 0, we are going to tinker a bit. First, let's see what we need. lan/by ~ A/BI (Bay ~ Abp)/bpBl \(Bay - AB + AB ~ Abp)}/bpBI \(Bay, - AB)/b,BI + (AB - Abn)/bpBl lan - Al/lbp! + Iby - BIIAI/Ib,Bl Choose @ and N asin (i), so t/Ibpl A), and there is ¢" with IB - byl < elBI/2IAIN for all nz e" (since by — B). If we let ® denote the largest of ¢ @, and ot then all inequalities hold simultaneously when n 2 2: lan/by - A/BI ¢ lay - Al/Ibyl + Ibn - BPlAI/bpBI “lag - AIN + IAIIB - byIN/IBi f(b) and gap) > g(b), so part (ii) gives f(an)/glan) > f(b)/g(b). 3.24, Prove that the sequence (an) given by ap = (-1)" does not converge.« If, on the contrary, a, ~ L, then for every € > 0, there is @ with IL - apl < € for all n 2 @ In particular, this would be true for ¢ = %. Thus, for all n2 @ a, lies in (L- %,L+ %), the open interval with center L and radius %. Since this interval has length 1, it cannot contain both numbers -1 and 1 (the distance between them is 2). This is a contradiction, for an and ay.4 have different signs, and so both cannot be in (L- %,L+%). 3.25. Prove that a convergent sequence has only one limit! if an? A and a, > L, then A= L «lf A#L,sete = “IL - Al> 0 (e is half the distance from A to L). Notice that the two open intervals of length 2e with midpoints L and A, respectively, do not overlap. Since an > A, there is @ with lan - Al L says there is | with ap lying in the open interval with midpoint L for all nz if n= larger of 2 and @', then a, lies in both nonoverlapping intervals, and this is a contradiction. Therefore, A = L. Leet s 5.26. (i) Let ay, ap, a3. (be a ee Gn ay 7 L, and 1 consider the new sequence az, a3, aq, (obtained by eliminating the first term. Show that the new sequence also converges to L. * Define a sequence by by bp = ane1i thus, by = az, bz = az, etc. Given € > 0, there exists @ with lay - Li< ¢ for all nz é Therefore, Ibn - LI = fane1 - LI< € for all n= @ (ii) Show that if (an) converges to L, then every subsequence of (aq) also converges to L. \taed x If by, bg, bg, is a subsequence of (a,), then each by = am with m2n.Givene>0, there is @ with lam - LI< € for all m 2alf n2é then by * am for m2n2 é@,so that lbp - LI = lam - Lie59 (iii) Use part (ii) to give another proof of Exercise 3.28 * The subsequence agn-1 > -1, while azn > 1. 3.27. (Sandwich theorem). Suppose that (a,), (by), and {cy} are sequences with ay 7 L, by > L,and ay < Cys bp for all nz 1. Prove that cy, > L. «Given €>0.Since a, 7 L and by > L, we have an~ bp > L-L-= 0. There is thus @ with lbp - apl < €/3 for all n2 @ (more tinkering afoot). Now by - an = (by - Cp) + (en = an) > ba - en so that lbp - cpl < €/3 for all n= &. Now IL - cpl = IL~ cy + bn - Bn * Qn ~ pl IL - anl+ lbp - enl* lan - bal Since a, 7 L, there is @ with IL - ap! < €/3 for all nz e Hence, if m 2 larger of @ and ¢', then for all n 2m, WL- cyl 4 = Since 0< % < 1, Theorem 3.19 shows that %" > 0. It follows thati- %"™ > 1-0-4. (ii) Show that (n- 1)/n > 1 «(n- 1)/n = 14 - 1/n. Now 1/n 0, as in the text, and so the result follows from Theorem 3.23. 5.29. Prove that a, > A ifand only if A- a, > 0. «If a, A, then for each ¢> 0, there is @ with IA - ayl 0, then for each € > 0, there is @ with (A = ay) - 01 A This is so easy that it looks hard. 3.30. (i) Prove that if Ir] <1, then r™ > 0. x In light of Theorem 3.19, we may assume that r = -s, where 0 0, there is @ with |s" - O| = [s"I< € for all nz @ But Ir? - Of = Ir®} = [(-1)"s"l = Is"| for all n, and so Ir? - O] ce forall n>e (ii) Prove that the geometric series [p29 r™ converges, with sum 1/(1 - r), whenever Ir] <4 « If r= 0, then the partial sum sp = 5j29 r= 0° = 1 (we made this agreement in the first chapter). Thus, (sp) is the constant sequence of all 1's, and hence sy > 1. If r= 0, then 1/(1 - r) = 1, as desired. We may now assume that -1 0, we have B = 3/10-10/9 zh, 3.32. Assume that there are two sequences converging to the same limit: aj > A and by 7 A. Prove that a \e4d at, by, a2, bz, a3, bs, ag, ba > A «Let €>0.Since ay > A, there is é’ with lan - Al¢€ for all n 2 U; since by > A, there is é" with [by - Al Al = Ibn - Al Al = lan - Ale, because m= 2n- 12 2@ implies 2n 2 2+ 1, and hence n2eet+nor’d 3.33. Define a sequence lan) by agn-1 = Sand azn = I/n; thus, the sequence begins 5, 1, 5, %, 5, 4,5, 4,9, ". Prove that (an) does not converge « The interval (4, 6), of radius ¢ = 1 and center 5, contains no term agp, but it does contain all odd terms agn-1. In words, there isno @ with [5 - a,l <1 for all n= @ Therefore, a, # 5. Now let L be any number other than 5. There is € > 0 so that (L - €, L + €) does not contain 5 (for example, take ¢ = lL - 5); that is, this interval contains no term agp. In words, there isno @ with IL - ay! <1 for all n2 & Therefore, ay # L. Alternatively, one can use Exercise 3.30(ii) 613.34, Define a sequence (ay) by agn-1* 1/n and agn = %™ The sequence begins: 1, %, %, 4, %, 1/8, %, 1/16, (i) Prove that ay 7 0, where a, = 1/n =« Given e> 0, there is @ with 1/n ¢¢ for all n 2 @, and there is ¢' with 4%" e for all n> é. Thus, if é" = @+ 2’, then ay e" (ii) Prove that the terms ay, do not get “closer and closer” to the limit 0; that is, if n> @, then it does not follow that lay - Ol «lag - OL « Since 2" > n for all n21, we have %"< 1/n, for all n. That is, agn-1 = 1/n> azn = %". But it is easy to prove, by induction, that 2" > n +1 for all n 2 2,50 that %™« 1/(n + 1); that is, azn ¢ @2n+1 = @2(net)-1- Thus, the sequence keeps oscillating up and down. 3.35, (i) Let (Py) be the sequence of inscribed 2"*!-gons in a disk of radius r, and let (Qn) be the sequence of circumscribed an*t-gons (notation as in Chapter 2). If py is the perimeter of Pp, and if qn is the perimeter of Qn, prove that pp» 7 2ur and qn > 2m x In Theorem 3.6, we proved that area(P.) 7 area(D); moreover, in Theorem 3.13, we saw that area(Py) = %2hpPp. AS area(D) = ar?, Theorem 3.21 gives %hnp, tr’, hence, hnPn —» 2nr2.Now hy > r, by Exercise 3.11 (and Theorem 3.21), so that Exercise 3.27(ii) gives Pn * ApPn/hn > 20r2/r = Quer In Theorem 3.11, we proved that area(Q*) \, area(D}; moreover, in Theorem 3.13, we saw that area(Qn) = %rdn. AS area(D) = mr2, Theorem 3.21 gives “rq, > mr, and so Theorem 3.18 gives qn > 2nr.(ii) Prove that sin(n/2™)/(n/2") > 4 « Position Py, as in Figure 3.23, so that the coordinates of X are (cos 8p/2, sin @,/2), where sin®,/2 = %by and by is the length of a side of Py. Note that @, = 2n/2"*1 = 1/2" Figure 3.23 Now Pp = 2°*4b, = 2°*2(%b,) 2P+2sin (64/2) = 2P*2sin(n/27*4), so that 2%72sin(n/2"*1) > 2m, by part (i), and hence 2°44sin(n/29*3) + Finally, Theorem 3.18 gives sin(n/2°*})(n/20*t) > 1 3.36. (i) Prove that if ay lies in (b - 1/n, b + 1/n) for all n24,then ay > b. «Given > 0, thereis @ with 1/€< & (because 1/n — 0). If n> @, then 1/n < 1/8, so that the interval (b - 1/n, b + 1/n) is inside of the interval (b - 1/é, b + 1/8). Therefore, if nz € then ay in (b- 1/n,b+1/n) implies an in (b- 1/¢, b+ 1/0); hence, lay - bl < 1/2 ¢ €. Therefore, an > b.64 Alternatively, one can use the Sandwich theorem (Exercise 3.31). For all n 21, we have L-tm¢ay¢L+i/n Since L- 1/n > L and L+i/n > L, we have ay 7 L (i), Assume that f(x) is continuous at a point b. If b = 0, prove that there is some interval (b - €,b + €) so that f(a) = 0 for all a in (b - €, b + €). Hint. Use part (i) « If the result is false, then for each € » 0, there is some ag in (b-€, b+) with f(a) = 0. In particular, for each n > 1, there is a, in (b- 1/n,b+4/n) with flaq) = 0. By part (i), we have ay — b. Hence, {{(a,)} is the constant sequence of all 0's, and so flan) > 0. But since f(x) is continuous at b, we have flay) — f(b) = 0, a contradiction: 3.37.If (ay) is a decreasing sequence if ay — t, prove that a, >t forall n2i « If {aq} is eventually constant, ie. there is some integer N so that an = aN«i = aN.2 +”, then it is easy to see that t = an, andso a,2t forall n We may assume, therefore, that for every m, there exists k>m with am > a, (strict inequality). Suppose that the result is false; that is, suppose that t > ag for some ¢. Choose & = t- ay. Since an — t, there is N with lay - th< e for n2N There is k2.N with ap> ax, so that traps t-aptay- a, = et (ar- ax), Since ay > a,, the term ag - a, is positive, and so t- ay? €. As kN, this contradicts la, - tl¢ea Chapter 4 2 4.1.lf ry and rzare the roots of {(x) = b and c in terms of ry and rz + bx + c, compute » By the hint, f(x) = x? + bx +c = (x - ry)(x - 19) = x2 = (ry + rp) + rire Therefore, b = - (ry #1rz) and c= ryr2 4.2. Wf a,b, and c are odd integers, prove that ax? +bx +c has no rational roots «If x = p/q isa rational root of ax? + bx +c = 0, then (@) ap? + bpq + cq? = 0. As in the preamble before the proof of Theorem 2.5, we can assume that p and q are not both even. There are now three cases (we shall use Exercise 2.4 throughout) Case 1.p iseven and q is odd In this case, ap? + bpq = p(ap + bq) is even, so that cq? must also be even. But since both c and q are odd, Exercise 2.4(ii) gives oddness of cq?, a contradiction Case 2,p isodd and q is even This is similar to the argument in Case |, Case 3. Both p and q are odd. In this case, each of the terms in Eq. (@) is odd, so that we have a sum of three odd integers being zero. Transposing, wehave the sum of two odd integers being odd, contradicting Exercise 2.4(ii) An alternative, though more complicated proof, can be based on Exercise 2.8(ii): /b* - 4ac is rational if and only if b? - 4ac is a perfect square If b2 - 4ac = m2, then m? must be odd (being of the form odd - even), and hence m must be odd; say, m = 2k - 4. In this case, we have b2 - m? = 4ac, so that (b + m)(b - m) = dac. If we write b= 22-1, then 4ac = (b+ m)(b - m) = [2k + 22 - 212k - 22] = A(k + @- 1k - @) alk? -k- 0 +2) It follows that k2- k - (é2 - @) = ac is odd. But it is easy to see that k2-k is always even, for it factors as k(k - 1); that is, it is a product of two consecutive integers, one of which must be even. Hence, [k? - k] - (2 - @] is even, a contradiction, 4,3, Find the points where the line with equation y = 2x + 2 intersects the circle of radius 1 and center (0, 0) « Ifa point (x, y) lies on the graphs of x*+y*=1 andof y = 2x +2, then x? + (2x +2)? = 1; hence, 5x? + 8x + 4 = 1. The quadratic formula gives x = (-B + J64 - 60)/10 = (-8 + 2)/10 = -1 or -6 If follows that there are two points of intersection: (-1, 0) and (-0.6, 0.8) 4.4, Suppose that a rectangle having sides of lengths x and y has area A and perimeter p. If p* - 16A 2 0, show that the roots of the quadratic in Eq. (1) are x and y 66« The quadratic is 222 - pz + 2A can be rewritten in terms of x and y 222 - pz + 2A = 222 - (2x + 2y)z + 2xy. Setting either z= x or z= y gives 0 4.5. Find and estimate the roots of 10°50x? + 2x - 9 =» Now /4 + 36/102 is very close to /4 = 2. Therefore, the root x = [+2 - JE + 367109%1/(2/1089) = -4/(2/205°) = -2%10°°. To calculate the second root, use the variant expression’ xe [2 + JF + 56710 /(2/10°) = 182+ /4 + 56/10) = 18/4 = 45. 4.6, Suppose a cannon on the ground makes an angle of 45°. (i) If the initial velocity of a shell is vg = 80 feet per second, what is the horizontal distance R the shell travels when it hits the ground? « The quadratic equation is -16t? + 80(1//2)t = 0, so that the time 1 it takes to hit the ground is t = 5/J2 seconds Substituting this value of t into Eq. (3) of Chapter 4, one finds R = 80(1//2)5(1//2) = 200 feet. (ii) What should the initial velocity vo of a shell be in order that it hit a target 400 feet away? « In this problem, Eqs. (2) and (3) are: 400 = votcos 45° = vot //2 0 = -162 + votsin 45° = -16t? + vot//Z,68 where t is the time needed to hit the ground. Hence, 16t? = 400 and t = 5. Therefore, 400 = vo5//2, and vo = 80/2 \ Laat 4.7. Prove, for any numbers aj, a2, “, ay, that [lay + apt + ap)/ni? < (a4 + ab + ~ + ah)/n ft Ee *Set by =b = by = 1 in Cauchy's inequality (Theorem 4.4) (ayb1 + agba + ~ + anbal® ¢ (aj + 03+ ~ + adb} + b3 + ~ + bad, to obtain lay tage ~ tans nlaf tage tan Now multiply both sides by 1/n? 4.8. Rewrite each of the following complex numbers in the form a + bi (i) (3 + 4102 - i) «10+51 (i) (2 + 0? «Qi (iii) [1/2 + A/F wi (iv) (3 + 4i)(3 - 41) ~ 2569 (v) (3 + 4/4 + iD. « 4(7 +i) 4.9. Prove the cancellation law for cormplex numbers: if u, V, and z are complex numbers with zu = 2v, and if z = 0, then usy « Since zu = 2v, we have {(1/z)zlu = [(1/z)z]v, and so u = v 4.10. Show that if u and v are nonzero complex numbers, then their product uv is also nonzero. * If uv = 0, then 0 = (1/uuv = [(1/u)ulv = v. This contradicts the assumption that v * 0. 4.11. Prove that if z is a complex number on the unit circle, then 1/z = z. on the unit circle, z = a + bi, where a% +b? = 1 a2 +b? = 1,andso 2 = 1/z » For z But 22 4.42, Let z= a+bi Prove that if z is areal number, then z Z; conversely, if z= Z, then 2 is a real number «Let z©s+tilf 2 isa real number, then t = 0, that is, z = s +0 hence, Z=s- Oi = s+ Oi = z Conversely, if z= 2, then 3 + ti = s - ti, One equation of complex numbers gives two equations of real numbers: ¢ = s, of course, and t = -t. Hence t=0 and z=s+ti=s+0i= 5; therefore, z is real. 4.13. If z= a + bi, show that (x - 2)(x - 2) is a quadratic polynomial having real coefficients. = (x - 2x = = (x +a - ib)(x - a + ib) = (x - a)? - (ib)? = (x - al? +b? = x - 2ax + a? +b?70 4.14. lf 2 and w are complex numbers, prove that Z+w-+z+w and zws2+w s+ti and w=p+qi.Now z+w = (stp) +(t+ qb, ztwels+p)-(t+qh; Also, zw = (sp - tq) + (sq +tp)i, so that Zw = (sp - tq) - (sq + tpi, and this is equal to zw = (s ~ tip - qi) 4.15. Let (x) = ax? + bx + c, where the coefficients a, b, and ¢ are complex. Prove that if the roots of f(x) = 0 are not real numbers, then they are complex conjugates. « Quadratic formula: If D = b? - 4ac, then f(x) has a complex root if and only if D< 0; write D = -E in this case. The roots are -b/2a +i JE/2a) and -b/2a - i JE/2a, and these are conjugates. Alternatively, suppose that f(z) = 0. Thus, ~ eee az? + bz+ c+ azt+be+t =a = a22+bz+¢, because a,b,c are real = (2) Therefore, Z is a root of f(x). 4.16. lf u is a root of a cubic polynomial f(x) having real coefficients, then its conjugate O is also a root of f(x).71 elf fe) = x3 + bx? + cx +d, then O=ud+bu2+cutd Repeated application of Exercise 4.14 gives o= 0- (a+ but tcurd) u> + bue +cu+ed +but+ca+d (b, ¢, dare real) Therefore, is a root of f(x). (This result and proof can be generalized to polynomials of any degree having real coefficients.) 4.17. \f § 1s a cube root of unity, show that &=¢ « ¢ = -% + i/3/2, and computing ¢? gives -% - iV3/2. 4.18, Find the square roots of 3 ~ 4i « (2 - i). 4.19. Find the roots of x? + (2+ ix + 2i = 0. « Using the quadratic formula and the previous exercise, x= A(-2-i4 J3= 4) = B-2-14 (2-1) = i or -2. 4.20. Prove the binomial theorem for complex numbers: If z and w are complex numbers, then for all n 20, = The proof is identical to that of Lemma 1.18 (if one replaces x by z throughout) and Corollary 1.21nD 4,21. Prove that if u, v, and w= are roots of a cubic polynomial x3 + bx? + cx +d, then b= -(u+v+w)and d= -uvw. «x8 + bx? + cx td = (x - ul(x - v(x ~ w) = x3 - (ut v + wx? + (uy + uw + vw)x > uv]. 4.22. (i) Find /8 + 151 « In polar form, 2 = 8 + 15i = 17(cos 62° + isin 62°); 62° is the angle @ in the first quadrant with cos 6 = 8/17 and sin @ = 15/17 a JB+ 151 = /i7cos 31° + isin 31° 533 + 2.1231. (Of course, the other square root is the negative of this one.) One can also use the method of Theorem 4.10 (i) Find V8 + 151 epee x VERT = YVT7(cos 15.5° + isin 15.5°)G 1.964 + 542i The other fourth roots are obtained by multiplying this one by i, -1, and -i. (The problem has been ambiguously worded, however, so that a student with only one fourth root has solved it.) 4.23. Find an 8th root of 977i. Nene? . 9-71 G /T30(cos 322° + isin 322°); 322° is the angle 9 in the fourth quadrant with cos 6 = 9//130 and sin 9 = -7//130 8S Fi = 1.356(cos 40° + isin 40°) ©) 1.403 + 875i. \odgynty.73 (The other eighth roots are obtained from this one by multiplying by the powers of e2™/8) 4.24, Show that cos 1.25664 + isin 1.25664 is a fifth root of unity (1.25664 = 2n/5 = 72°). spe ran x By De Moivre's theorem, (e!72) = e560 = 4.25. Prove that if ¢ is a complex cube root of unity, then g(t - 624 - o)* = 318 * c(t - e204 - 9)? = sla + sya - 098 (because 1 (c+ c1 - 6) [because ¢% + ¢ +1 = 0] = 1430-3524 53 = 3c - 34? [because ¢3 = 1] = 3g - 8) = 33 {because ¢ = -% 4 i/3/2]) 4.26. For every positive integer n, show that there is a polynomial f,(x) of degree n and with integer coefficients so that cos n@ = f,(cos 9)74 « By De Moivre'’s theorem, cos nx + isin nx = (cos x + isin x)" The binomial theorem gives (cos x + isin x)" = 5 420 (5) cos? Sx i sind: and so cos nx is the real part of this. The real part is the sum of all the terns with j even cos nx = Ej even (') cos? Sx (-1)7? (sin2x)/? But sin@x = 1 - cos®x, and so this substitution gives a formula for cos nx as a polynomial in cos x 4.27. (i) Prove that 2cos @ = el +e “ 28+ en !® = cos 9 + isin @ + cos(-8) + isin(-9) = 2cos © (ii) Use De Moivre's theorem to give a new proof of Exercise 2.51 2cos(n + 1)8 = (2cos 8)(2cos mB) - 2cos(n - 1)8 * By (i), (2cos @)(2cos ne) = (el? + e7!)(eiM + e71P8) a clinstey gilt-msy giln-2)e 4 g-iln+i)e elintiey erinst)8 , giln-tle 4 gill-n)e = 2cos(n + 1)@ + 2cos{n - 1)8. 4.28, (i) It can be proved that if z+ a+ ib, then e@elb = e®%(cos b + isin b)In contrast to real exponentiation, show that e? can be a negative real number. «If zim, then e'™= - (ii) If 2 and w are complex numbers, prove that 2, eM = git «Let 2+ a+ib and w=c+id. Then QW 2 cagibetgld - eid eretelPe % 8 ” s = ettteibtid . gatcgilbed) . gztw (iii) If w = e%, where z is a complex number, then define log(w) = z. In contrast to real logarithms, show that -1 has a logarithm; indeed, show that it has infinitely many logarithms. « We know that e!” = -1, so that log(-1) = im. On the other hand, log is not single valued, for (e'")2"-1 = -1 for all integers n, and so log(-1) = (2n - 4)in for every integer ni 4.29. For any complex number 2, define cos z = “lel + e7/?) In contrast to real cosines, show that cos z > 1 is possible. «If b is real, then cos(ib) = %(eMiP) + eG) = Yle® + eb) = cosh(b) But cosh(b) > 1 (indeed, we saw in Chapter 1 that there is equality if and only if b = 0) 4.30. (i) Write 1/(e!® - 1) in the form a + ib. “ t/(el® - 1) = 1/(cos @ - 1 + isin 9) (cos 8 - 1 - isin $)/(2 - 2cos 9)76 Hence, a = (cos @ - 1)/(2 - 2cos @) and b = sin 9/(2 - 2cos 9). (ii) Prove, for all n = i and for all 6, that 1+ els i284 4 giN® = feilm*t)8 - a)/fel* - 1) « The proof is identical to the proof of Exercise 1.2. (iii) Prove the identity: (2 - 2cos@)(sin @ + sin2@ + sin3e) = -(cos4@ - 1)sine + sin4a(cose - 1). * By part (ii), 1 + el + el28 + elS® = (e!48 - 1)/(el® - 1). Rewrite the right hand side, using part (i) and De Moivre’s theorern, and multiply both sides by 2 - 2cos @. The imaginary parts of the two sides are equal, and this is the desired identity 4.31. (i) If cos 3a is positive, show that there is an acute angle p with 3a = 3p or 3a = 3(p + 90°), moreover, show that the collection of numbers cos p, cos(p + 120°), cos(p + 240°) coincides with the collection of numbers cos(p + 90°), cos(p + 210°), cos(p + 330°) « Each collection of cosine values consists of cos B, -Y%ecos p + /3/2sin 6, and -%cos p - /3/2sin B. (ii) If cos 3a 15 negative, show that there is an acute angle B with 3a = 3(p + 30°) or 3x = 3(p + 60"); moreover, show that the collection of numbers cos(p + 30°), cos(p + 150°), cos(p + 270°) coincides with the collection of numbers cos(s + 60°), cos(p + 180"), cos(a + 270°) « Each collection of cosine values consists of -sin p, /3/2cosp - “sing, -/3/2cosp - “sing. 4.32, Consider the polynomial {(X) = x2 + x? - 36 that arose in the castle problem in Chapter 2.17 (i) Show that 3 is a root of {(X) and find the other two roots as roots of the quadratic {(X)/(X - 3). « x5 4X2 - 36 = (X - 3X2 + 4X + 12) The other roots are thus -2 + iJ8. (ii) Use the cubic formula to find the root 3 of f(X). « The substitution X = x - % gives the reduced cubic f(x) = x? - %x - 970/27 Thus, q = -%, r = -970/27, R = 940896/27, and 3 ¥%(970/27 + /940896/27) = (1/54)(970 + (940896) © 35.926; hence, a © 3.2996 ) afeprnt Ax It follows that h = -3q/g = 1/9g @ .0336, and so g+h = 333329 3% A root of f(x) is thus 3% - % = 3. (iii) Show that the discriminant of {(X) is negative, and find its rea] root (which is 10/3) using cosh x The discriminant formula gives a negative value, so that f(x) has complex roots. In the reduced cubic fix) = x3 - %x - 970/27, we have q=-% and r = -970/27, so that -4q/3 = 4/9>0 It follows that we are in the hyperbolic cosine case, and t= %we 78 c Now cosh p = ~4r/t? = 485, so that /6(5 6.8773 and p/3 = 2.29243. But cosh(2.29243)(A4.999 © 5, so that tcosh(B/3) = % x5 = 10/3. aca 4.33, (i) Show, for all a and b and for all j2 1, that ad b) = (a b)lav"} + al"2b + ad 3p? + + abl? + bind) * After multiplying out, all terms cancel except ad - bi) Gi) If f(x) = cyx? + cp-pxPo2 + ~ + cyx + cg be a polynomial of degree n. For any number u, show that there is some polynomial q(x) of degree n- 1 with f(x) = (x = w(x) + flu) * f(x) - f(a) = ZyBo cyxd - Zj8o cyud = Tj2o ej(x) - uw); now use part (i) to factor x! - ul = (x - u)hy(x) for all j 2 1, where h(x) is some suitable polynomial. Hence, 0 _ 40 f(x) = fu) = (x - walx) + cg(x® - u9) = (x - walx), where q(x) = jt, hy(x). (iii). Show that u is a root of f(x) if and only if x-u isa factor of {(x). «If u isa root of f(x), then f(u) = 0 and f(x) = (x - uda(x), by (ii) Conversely, if f(x) = (x - uMg(x), then flu) = (u - u)(q(u) = 0, and so u isa root of f(x) 4.34. Show that every cubic polynomial f(x) = x9 + bx? + cx +d having real coefficients b,c, and d has at least one real root. % Suppose that f(x) has a nonreal root u; by Exercise 4.16, U is also a root of f(x). By Exercise 4.33(iii), we have79 fix) = (x - ullx - DOr - vd, where v_ is another root of f(x). Now the constant term is d = uuv, so that v = d/uii is real (for ut = 0 and is real) 4.35. Show that if cos 3a = r, then the roots of f(x) = 4x3 - 3x -r are cos a, cos(a + 120°), and cos(a + 240°). « Since r = cos 3a = cos3(a + 120°) = cos 3(a + 240°), the 3 triple angle formula (Corollary 4.16(ii)) cos 3a = dcos*a ~ Scoso shows that all three cosines are roots of (x). 4.36 Find the roots of f(x) = x5 - 3x41 * Since f(x) is already reduced, we can use the cubic formula at once. Hence, q = -3, r= 1, and R = -3. It follows that g° -% + iJ3/2 = t; since ¢ is a cube root of unity, g must be a 9th root of unity. By De Moivre’s theorem, g = cos 40° + ising0" Now h = -q/3g = 1/g = @ = cos 40" - isin40’, and so a root is 4.37, Find the roots of f(x) = x° - 9x + 28 = -1, h = -3, and the roots -4,2+i1f8,2-i18 The complex roots emerge most quickly if one writes tx) = (0 + A(x? - 4x + 7) and then uses the quadratic formula on the quadratic factor80 4.38. Find the roots of f(x) = x°- 24x? - 24x - 25 « The reduced polynomial is f(x) = x - 216x -1241. Here, a = -216, r= -1241, R = 47089, JR = 217, 5 = 729,g = 9, and h = 8. Therefore, a root of f(x) is g +h = 17, and soa root of f(x) is 17 +8 = 25. One can factor f(x) = (x - 25)(x? + x + 1), and so the other two roots of f(x) are -% + i/3/2 4.39. (i) Find the roots of x°- 15x - 4 using the cubic formula « Wehave q=-1 gthe°vo+ Jin +°42- J=i2t = -4, R = 484. The roots are: mgs tthe ¢ V24 Jani + 62% 3V2- Jia etches 52/26 Jota +e V2- Stal It is not at all clear how to estimate the roots in the form above, From Theorem 4.24, however, we do know that all the roots are real numbers because the discriminant -27r? - 4q5 = -27x(-4)2 - 4(-15)3 = 13068 is positive Gi) Find the roots using the trigonometric formula rot beet we tele? «t= S906) 447 and cos 3a (e179. Hence 3a @ 80° and « & 27°. Hence, cos 27° & 891, cos 147° @ ~.839, cos 267° © -.052 Hence, the roots are: S2cos 27° 447x891 G 3.989 4 J®cos 147° (S) 4.47 x(-.839)G -3.750; S2cos 267°\F 4.47 x(-.052) © -.232. Once one sees that 4 is a root, it is simplest to use long division te find the other two roots. Since81 x3 ~ 15x2 - 4 = (x - 4)(x? + 4x + 1), the other two roots are -2 + /3. 4.40. Find the roots of x>- 6x +4 * Roots are 2,-1+/3,-1- 3 7 x? - 20x 4.41. Find the roots of x “This is a realistic problem; the implementation of the formula is rather long (but see the remark at the end of this solution) Set L = -15, M = -20, and N = -6. Then Eq. (16) is: jo + ab j4 + (L2 - anny? - M? = j® - 3054+ 249)? - 400 We now make this cubic (in j*) reduced with the substitution i? = y + 10, to obtain owe? Fly) = y>- Sly + 90. Apply the cubic formula: q = -51, r = 90, R = -11552. Thus, /R = 107.5) s %(-90 + 107.5i) -45 + 53.81, and De Moivre’s theorem gives a cube root: g@ 3 + 2.8: Now h = -q/3g = 51/3[3 + 2.81] = 17/(3 + 2.81 1® 3-28i = g. Therefore, g+h= g + & = 6 is a root of Fly) (alternatively, one could have checked for integer roots of Fly) using Theorem B). Therefore, 16 (for j2 = y +10) and j = 4. Eqs. (15) are thus am = 16-15 - 20/4 = -4 2@ = 16-15 + 20/4 Thus, m = -2, @= 3, and we have the factorizationx4 = 15x2- 20x - 6 = (x2 + 4x + 3x? - 4x - 2) The quadratic formula applied to each of the two factors gives the desired roots: -1,2+/,2- Je In particular, all the roots are real. Remark. It is natural to use Theorem B to check first for rational, hence integral roots. One finds that -1 and -3 are roots (the candidates are #1, +2, #3, and +6). Therefore, long division gives x4 15x2 - 20x - 6 = (x + 1)(x + 3)[x? - 4x - 2) The other two roots are thus 2+ /6 and 2 - £. One now understands the unpopularity of the quartic formula. 4.42, (i) Show that cos 28 = (1 ~ tan@9)/(1 + tan@e) «(1+ tan2e)/(1 + tan2e) = (1 - sin@e/cos?9)/(1 + sin@e/cos*e) c0s2@ ~ sin20}/(cos?e + sin?e) 29 - sints cos’ = cos 28. (i). If ris a rational number, show that the only rational values of tan rt are Q and +1 * If tan rw is rational, then part (i) shows that cos 2rm is rational. By Theorem 4.25, cos 2rm = 0, +%, or +1. Hence, sin 2rm = 1, #%, or 0. If cos 2rm = 0, then tan 2rm = sin 2rm/cos 2r is not defined. In the other two cases, we have sin 2rn/cos 2rm = +1 or 0.83 4.43. (i) Give an example of two positive irrationals whose sum is rational « J2+ and -/2 +2 (ii) Show that /2 + JB is a root of x4 -10x? +1 = (3 + JB)? = 5 + 26, (2 + B)4 = 49 + 20/6, and so (/2 + S3)4 102 + J8)2 +1 = 49 + 20% +5+ 2 +1 = 0. (iii) Use Theorem B to show that /2 + J3 is irrational « Were /2 + 3 rational, then Theorern B would say it is either Lor -1 4.44, (i) Prove that log,6 is irrational « If loggé = r/s, where r and s are positive integers, then 5r/s = 6 and 5” = 6%, This cannot be, for 57 is odd and 6° is even Gi) Prove that logg15 is irrational « If logg15 = r/s, where r and 8 are positive integers, then 6t/® = 15 and 6" = 15%, This cannot be, for 6" is even and 15° is odd84 Glossary of Boring Terms G1. (i) For any formula 9, prove that gv (~@) is a tautology. «@ » | is true if at least one of p or W is true (ii) For any formula , prove that g + (~@) is always false. «@ -y is false if either g or is false G.2 Prove that ~(~y) is logically equivalent to 9 «If @ is true, then ~@ is false, and so ~(~g) is true; if g is false, then ~w is true, and so ~(~@) is false G.3. Prove that g @ p and (g = y) ~ (y = 9) are logically equivalent = The truth table for (p 4 p) « (W > g) is the same as that of gey G4. Show that @ = W is logically equivalent to (~g) ~ & « The truth table for (~g) v y is the same as that of 9 > y. G.5. Show that ~(p > y) is logically equivalent to p ~ (~W). « Both have the same truth table’ 9 Teenie y erence See eee reece ae G.6. Show that (y v y) «1 lp» (~~ fh o (-@)l (g » ¥)) is logically equivalent to85 x Both have the same truth table: eae Teer eee TSeee Fetes G.7. Show that g @ | and ~[(y ~ (~)) ~ (ps (-@))] are logically equivalent « Both have the same truth table: 9 Tae eee eee ¥ Pe eeu ee Reee G.8. Given statements and y, define | by the truth table 9 Teed a = y Pre eee oly oF FF T (i) Show that (p | y) @ (~g) » (~) = Both have the same truth table. (ii) Show that ~g @ glo « Both have the same truth table (ii) Show that @ > y @ (-@) | (-y) * Both have the same truth table(iv) Show that yy vw @ ~(y | y) * Both have the same truth table 6.9. (Associativity). For any formulas 9, y, and 8, prove that gv (pv 6) @ (gv yy @ and gr(prelelgr pas are tautologies * Both have the same truth table 9 sTeecre gee ates Eee Perechsceh: ¥ Bree eee eee eee eee ° Hee ee eae eee Te eee Tee eae Tene eee errno G.10. (Distributivity) Let @, , and 9 be formulas. (i) Prove that g v (y ~ 8) @ (y » Wy) ~ (yg ~ 8) is a tautology. * Both have the same truth table @ Seeger ceeeee Tce Tota tect Page cee y ee eee ee peter eee eee ieee 8 Eee eee ib ce Meee etek Treen eetEEeetee Perea Peete eee eee (ii) Prove that g ~ (wv 8) @ (@ ~ ) » (gy « 9) is a tautology. * Both have the same truth table. 86T T T F F F F F G.14. (i) Define a sequence 11,12)", Tn“) Where rp is the world record for the fastest mile ever run in or before year n. Show that this sequence converges. « This sequence is decreasing, i@., mn 2 nei for all n= 1, and it is bounded below by 0 (ii) Conclude that there is a time t so that runners can run a mile in a time arbitrarily close to t, yet no runner will ever run a mile even a millisecond faster than t Exercise 3.30 (in the section on sequences) states the theorem that every decreasing sequence that is bounded below does converge to some some number 1, and it proves that ry 2t for all n> 1. G.12, (i) Prove that (a, b) = (a’, b') if and only if a = a’ and b=b' «If asa’ and b=b’, the on is obvious. For the converse, assume that (a, (a, b)) - {a', {a’, b'}}. There are two cases Ia =a’ and (a, b) = fa’, b'): Il. a = (a', b') and (a, b) = a’ In case I, a = a’, so that (a, b) = (a’, b'} = {a, b'}. Therefore, (a, b) - (a) = (a, b’) - fal. If a = b, the left side is ernpty, hence the right side is also empty, and so a = b’; therefore, b = b’ If a # b, then the leftside is (bh; the right side is also nonempty, so that it equals (b’) Therefore, b = b’, as desired In case Il, a = {a’, b') = {(a, b), b’}. Hence, a ¢ fa, b) € {la, b), b} =a, contradicting the axiom that a € x € a is always false Therefore, case I] cannot occur (ii) Give a forma! proof that (a, b) = (a', b’) if and only if a=a and b= b' x As it says in the text, “Once in your life, go into a secluded room and actually write a complete formal proof.” I have already done this; it is now the reader's turn Gii) Prove that la, b] = [a’, b'] if and only if a = a’ and b = b’ «If a=a' and b= b’, the result is obvious. For the converse, assume that {(a], (a, b)} = {{a'l, (a’, b')}. There are two cases a=b and a#b. If a= b, then (a, b) = (a) and so {(al, {a, b}} = (Ca), fa)) = ({a)}, the set whose only element is (a). It follows that (la}} = {{a'}, fa’, b']) is a 1-element set, so that (a) = [a’, b’); this, in turn, gives a’ = b' (for (a’, b’) is a 1-element set). Therefore, b=a-=a'= bias desired If a « b, then (a) = (a, b), and so {fa}, (a, b)} is a 2- element set; therefore, {{a'), {a’, b'}} is also a 2-element set, so that (a’) « (a', b'), and so a’ * b’. The given equation {la}, (a, b)} = (fa), fa’, b')} gives (a) = {a') and (a, b) = {a', b') (the other possibility {a} = {a', b') and (a, b} = (a'] cannot occur because (a} has exactly one element and {a’, b'] has exactly 2 elements). It follows that and so {b) = (a, b) - {a} = {a, b’) - (a) = (b’). Therefore, b = b', as desired. G13. For any subset X of a set U, prove that X U (~X)=U and XN (-X)= 8 88«If u € U, then either ué X or u ¢ X; that is, X U (~X) = U. There isnmo u€U with uc X and u¢X G14. Define X- Y= (ue Xu @ YI g = glu) and = lu) are formulas, show that Vig > y) = ~[V(g) - V(yI] = We use the observation in the text that if a = a(x) and p= a(x) are true for every x, then V(a) = Vip). By Exercise G4, @ > w and (~g) v W are logically equivalent; hence, vig = y) = Vil-p) ~ wy) = V(~g) U Vy) = ~VOp) U Voy). By a De Morgan law, ~V(g) U Vig) = ~[V(g) 9 ~VOp)] = -[V(p) - Vp]. G.15. Show that V(y @ ) = ~[V(g) & V(yII. « We use the observation in the text that if a = a(x) and B= p(x) are true for every x, then V(c) = V(p). By Exercise G.7, @ @y and ~i(p » (~w)) v (y > (~@))] are logically equivalent It follows that Vig & w) = ~Vilg « (~g)] » [y » (~g))) = ~({Vig) A V(= pil U [¥(y) 9 V(~9)]) = ~[Vig) 4 (Vp) G.16, Show that Vio | w) = ~V(y) 1 ~V(y), where g = g(x) and y = w(x) * Since @| p is defined as (~g) « (~y), it follows that Vo |p) = Vil=@) * (~y)) = Vig) 9 Vie) = =V0p) 9 ~ VOW) G.17. (Associativity). For any subsets X,Y, and Z of aset U, prove that X U(YUZ)=(XUY)UZ and XN(YNZ)= (KAY AZ « The proof involves showing that each side is contained in the other. It is not very enlightening, for it hinges on the definition of the connective “or” given in the text. For example, if u € X U(Y UZ), then u€X or uc YUZ;thatis,ucX or ueY or u€ Z.Hence,u€X€Y or u€ Zandso uc (KUY)UZ The reverse inclusion is proved similarly The proof for M is proved in the same way, hinging on the definition of the connective “and” 89Remark. If one knew that every subset of U is of the form V(q) for some formula @ = g(n), then one could use Exercise 6.9 together with the observation in the text that if a = a(n) and p = pn) are true for every n, then V(«) = Vip). This is, in fact, true, but readers may not be able to prove it (it involves describing subsets of U by 2-valued sequences) 6.18. (Distributivity) For any subsets X,Y, and Z of a set U, prove that X U(Y 0 Z) = (X UY) (XU Z) and X N(Y UZ) = (XN Y) U(X N 2). « This proof is similar to that of Exercise G17, showing that each side is contained in the other; again, it is Just a question of looking at the definitions of the connectives Remark. If one knew that every subset of U_ is of the form V(g) for some formula @ = y(n), then one could use Exercise G.10 together with the observation in the text that if « = a(n) and p = a(n) are true for every n, then V(«) = V(p). This is, in fact, true, but readers may not be able to prove it (it involves describing subsets of U be 2-valued sequences) G.19. Prove that the sequence ((-1)") does not converge to 1 « As in the text, one must show (Be > OM YN)(3n = NIM-1)" - 112 el Choose € = 2 (the distance between -1 and 1; of course, any smaller positive choice of © works as well). For any choice of integer N, choose an odd number n 2 N. With these choices, M-aP - t= -1 - 1] = 22 €, as desired 6.20. (i) lf {n(x) = x", prove that fy(x) converges pointwise to L(x) on (0, 1], where Lix) = 0if Os x< 1 and LO) = 1 « By Theorem 3.20, x" + 0 whenever 0 s x ¢ 1, while the sequence x" is just the constant sequence a, = 1 when x = 1, and hence it converges to 1. Therefore, fy(x) converges pointwise to L(x) on (0, 1} 9091 (ii) Write the definition of pointwise convergence with quantifiers, and then write its negation « With quantifiers, the definition of pointwise convergence is (ve > 0)(We € fa, BIEN) ¥n = N)llfple) - L(c)i < e] The negation of pointwise convergence is: (Be > 0)(de € fa, DVCVND(3n = NDllip(c) - L(t =] G.21. (i) Write the definition of uniform convergence with quantifiers « (ve > OAN)(¥e € fa, bIK¥n = NDIlf,(c) - Llc)! « el. Comparing the definitions of pointwise and uniform convergence, one sees that the middle two quantifiers are reversed (the text remarks earlier that the order of the quantifiers is important) (ii) Show that if {,(x) converges uniformly to L(x) on la, bl], then {,(x) converges pointwise to L(x) on [a, b). « The second basic principle about quantifiers given in the text is (SN)(¥e € la, b]) implies (Vc € [a, b](3_N), and this applies here ii) Write the negation of uniform convergence. “ (de > OMYNI(3e € La, DIAN 2 NVlify(e) - L(cil 2 €l (iii) If f,lx) = x", prove that f,(x) does not converge uniformly to L(x) on [0, 1], where L(x) = 0 if 0 s x < 1 and Li) = 4 « Choose © = %; given N, choose c+ (%)!/N and n = N. ThenIfp(c) - L(odh = Ie? - Ol = cP = % > % = €, as desired, Therefore, this sequence does not converge uniformly to L(x) G.22. Prove that the empty set @ is a subset of every set X. «The negation of (vx € Bike B= x € X) is (ax € ox € B) » (x ¢ XI. But (3x € g(x) is false for any formula g(x), because there isno x € @.Hence, the negation of (Vx € Bke @ exe Xl] is false, and the original formula is true (ii) Prove that there is only one empty set « Suppose @ is a second empty set; that is, 6 also has no elements. As in part (i), @ ¢ X for every set X; in particular, © ¢ @; similarly, @ C O,andso G= o