TUTORIAL TEXT No. 1 ELEMENTARY INEQUALITIES RY D. $. MITRINOVIC Belgrade, Yugoslavia TN COOPERATION WiTHE E. 8. BARNES. Adelaide, Australia D.C. B, MARSH, Golden, Colorado, U.S.A J.R.M. RADOK “Adelaide, Australia 1964 P. NOORDHOFF LTD - GRONINGEN ‘THE NETHERLANDSGconrihe 1964 ty P. Neovo 48 Groningo, The Nerande This book o7 parts thereof may wot be reproduced in any form without writen permission of the publishers, Printed in The NetherlandsFOREWORD ‘The problems contained in this series have been collected over many years with the aim of providing students and teachers with material, the search for which would otherwise occupy much valuable time, Hitherto this concentrated material has only been accessible to the very restricted public able to read Serbian*. I greatly welcome the appearance of the Tutorial Texts based on my problem collection. Cooperation with colleagues in Australia and theU.S.A. and theirinitiative have madeit possible. For the preparation of this Text, I wish to thank E. S, Barnes, D.C. B. Marsh and J. R. M. Radok, D. S. Mirainovié, Belgrade, 23.111.1964, + Zbornik matematickih problema, Belgrade, Vol. I (1962), Vol. (1960), Vol. 111 (1960), published in cooperation with D. Adamovié, V. Devidé, D, Djokovié, D. Mibailovié, 2, Pop-Stojanovié, S. Prefié, J. Ultar,oa og 0.8 od 05 06 o7 08 0.9 0.10 oun 10 u TABLE OF CONTENTS dateoeiction ea auineiinte Inequalities involving mean values . Bemoulli’s inequality . Chebychev’s inequality Abel’s inequality. ‘The Cauchy Schwarz-Buniakowski inequality Young's inequality... Jensen's inequality... The Fejér-Jackson inequality Jordan's inequality Some integral inequalities : Some inequalities for symmetric functions Elementary inequalities ©... 2... Inequalities obtai Inequalities involving powers and factorials . . Inequalities involving finite sums and products . . Inequalities involving trigonometric functions Geometric inequalities... Inequalities involving mean values and symmetric functions The inequalities of Cauchy-Schwarz-Buniakowski, Holder, Minkowski and Chebychev Inequalities involving integrals Inequalities in the complex domain. Miscellancous inequalities... able from functional considerations 15 16 18 19 20 23 30 30 a1 32 36 58 a 91 101 ui 120 129 131 138 14aINTRODUCTION This tutorial text and problem collection is designed to intro- duce the student, at undergraduate or senior high school level, to the elementary properties of inequalities. A knowledge of algebra, geometry and trigonometry, together with a first course in calculus, provides a sufficient background for almost all the material. It is hoped that, as a collection of problems, the book will be a useful adjunct to regular course work, while the provision of an introductory survey and numerous complete solutions will give students an opportunity for independent study. For ease of reference, the material has been roughly classified according to subject matter or method of proof, although a precise classification of inequalities appears to be virtually impossible. For a careful development of the theory of inequali- ties, the reader is referred to the first two monographs listed below and to the comprehensive bibliographies given there. Many of the problems given in this text are to be found in the problem pages of periodicals such as the American Mathematical ‘Monthly which provides an abundant source of both elementary and advanced problems. GH. Harpy, J. B. Lirtuewoon, G. Porva: Inequalities, Cambridge University Press, London, 1934, E.P., Brexexnaca ~ R. Brueataw: Inequalities (Ergebnisse der Mathe- ‘matik), Julius Springer Verlag, Berlin, 1961 EB, Becemwmack — R, BELLMAN: An Introduction to Inequalities, ‘Random House New Mathematical Library, 1961. N. D. Kazaninore: Analytic Inequalities, Holt-Rinehart and Winston, New York, 1961 N.D. Kazaraworr: Geometric Inequalities, Wesleyan University Press ‘and Random House, 1961. D. A. Kryzuaxovsin: Elements of the theory of Inequalities, Moscow- Leningrad, 1938. GL. Neviazuses: Inequalities, Moscow, 1947. P. P, Konovxin: Inoqualities, Moscow-Leningrad, 1951.MEAN VALUES 9 § 0.1 Inequalities involving Mean Values O.1.1 For every set of positive numbers A= (ay, a, min (dy, da, - -» @y) S aa T Stotete 4 a is the harmonic mean of the numbers A their geometric mean, ir rit mean an (EEA a atm nr 1° Consider first the inequality a10 ELEMENTARY INEQUALITIES, Pxoor 1, For m= 2, this inequality is seen to be true and can ALE Ge By induction, Mart h—-Vdp is & 4 2 (AB yt ag Thus, it follows that Ania = HAgtA) & (AeA) B (GG) ie, = Gan, An 2 GAR, whence Aga 2 Gus This completes the inductive proof of the inequality (1). (On the basis of this proof, it is easy to show that equality holds in (1) if and only if a, =a, =... = ay ‘This result is obvious if m= 2. Assume that it holds for some n= k= 2. From the above proof, we see that if Ansa = Ge, then A,= 4, A= Gy, 4 =. Since A, =G,, we have a4=as.. and since A = G, we have fe = Ata = whence Finally, (1) certainly holds with equality if all a, are equal. Thus our assertion is established by induction. PxooF 2. The validity of a statement P(w) can be established by the method of regressive induction in the following manner:MEAN VALUES ry P(s) holds for infinitely many values of »; P(n) implies P(n—1) for all » > 1. Combine the methods of progressive and regressive induction ¥, It has just been seen that (1) holds for m= 2, Assume now that for all positive numbers ay, dy, ..., @, some of which may ‘be equal to each other, (1) holds for m = 2* (where & is some natural number). \ Form the sum at aet 2n), Opts cb hy, Ona tOnse to ” ne +m) From the inequality 8S Vit (Ody > 0), e we obtain 1 et Guat Oneat Mae a aes =) Hayat et St Using the assumption that the inequality (1) is valid for n = 2, the last inequality gives i (yay. ig). ‘Thus, the relation (1) holds for every me(2, 2, 28, ...}. Assume now that the inequality (1) has been proved for », and replace init a, by (Gat... +a, 4)/(1—1). Then 1 Alternatively, ascending and descending induction,rt) ELEMENTARY INEQUALITIES ‘When the left-hand side is simplified this relation assumes the form shay ie and, finally, srbaat ‘Thus it follows from the assumption of the truth of the relation (1) for 1 that it also holds for n—1. This completes the second proof of (1). 2° On the basis of the inequality (1) of 1°, we have for the numbers 1ay, 1/ay, .-., Ht, ee Lye ata a) a ay ‘This inequality becomes an equality if ee e) Tt follows from (1) that (3)MEAN VALUES 13. Consider now the relations (tant - +. +a) Sat -- FUG Meta ayt - tdyaty), (4) 2a,a, ZS a2-+aZ, since (a,—a,)* = 0. (5) Replacing 24,4, by a?+a,! on the right-hand side of the iden- tity (4), we obtain the inequality (tayt i tay)? S ma tagt pat) ) which holds for all real a, (@) that Ifall the a, are positive, it follows from $a, S iolattaty a2), whence : er eapteee eta ttaMne aaa eure =(# ath ye Oy Tt will now be shown that min (a), ..., 44) S 7 ” (s) aap Without loss of generality, suppose that. 0 —1 and m is a natural number, then Bernoulli's in- equality states that (bie BL a) Poor. Apply the method of mathematical induction. If = 1, the relation (1) holds as an equality. Suppose that the relation (1) holds for » = & (= 1), ie. that (EA) & Leh. @) Multiplying this inequality by 144 (> 0), we obtain (LEA) B (LEA) (LEAR) = 1 (iE A) a, whence (L+ Ay SIT Rt AL (3) Since (2) implies (8), the proof is essentially complete. Next the generalized Bernoulli inequalities, required in differ- ential calculus, will be established: (l42)'> lta (-1<240,a>1ora<0), (4) (l44)* < Lar (-1 <2 40,0 0, this gives sgn {(L+2)*—1—av} = sgn {a(a—1)} (x #0), whence the inequalities (4) and (6) follow.16 ELEMENTARY INEQUALITIES § 0.3 Chebychev’s Inequality qj peers ‘ Peete: : Gilde) sae ( ProoF 1. Let Se =3a, Yoedn, Sab Sats then TE lesb 4,b,) = Erg e- a,b) = nYab~ ays, LTD, —4,b,) =F (na,b,—a, Z 0) = nYab— Lado, Hence nZad— LaBomAL Elegy ae tarb,—ab) = >> (a,—4y) baby) By (1), (244) (by —br) ZO (u, 7 = 1,2. m), whence follows Chebychev’s inequality, nZab~Ya Sp =O. If and only if a= a, =... =a, oF by the equality sign holds in (2) Pxoor 2. Let = Pa Gay = Pay s+ 2 By Bat . = day baby = Gay ee vy yOu = darCHEBYCHEY'S INEQUALITY i with p, = 0,¢, 20 (7 =2, 3,...,), whence (2) may be written, as a (3a)(Za) = (3 m+1-08)( 3 m1) ot nov b1—max 6 pg, & 304491 Pate Since (ot 1-4) (nt 1-7) (n-+1—max (f, j)} [n+ 1—min(, j)}, the last inequality becomes Bee tmx 0 fing) 2g, 2 0 i) Since n+1—max (i,j) > 0, min (i, j)—1 > 0, p,q, 29 for i, j = 2, 3,...,m, the inequality (8) is obvious. This second proof is due to D. Djokovié. GENERALIZATION. If Oeasas tea 0st aoe eb 054545... 5ty then da dd son Exampce, Ifa, 8, ¢, are positive numbers and ifm is a natural number, then (@tbte)* saat eter), Remarks. 2° From (3), it follows that equality holds in (2) if and only if18 ELEMENTARY INEQUALITIES 2 In the generalized Chebychev inequality, the conditions 0 Sa,0 In many books (CE, for instance, J. W. Archbold: Algebra, London 1958, pp. 5152), the conclusions corresponding to 1° and 2° above are incorrect. See also: C. V. Durell-A. Robson: Advanced Algebva, vol, 111, London 1948, pp. 870871. tabacrtaabaey 312 § 0.4 Abel's Inequality TE {y, yy. +, @q} and {b,, Og,..., Da} (Or SO, =~ by 2 0} are two sets of real numbers and if M and m, respectively, are the maximum and minimum of the numbers cease) then miby S Dy bagdyt + ayby SMOy, Proor, The sum abr tagdet Buby (=3.) may be written in the form Sab, = bts bet oo Hy —Sea)Pe ssl b) Erba bbe Pa) Sabo Since M = max(sy,8q,.. 454), ie. 4 SM, SM,...,5,5M, and, by hypothesis, bby = 0, bby = by, Z 0, by ZO, () we have the sequence of relations Si(0r—be) SM (by—~ba), S2(bo—bs) SM (bbs), «+ Sq1byn1—bp) SM (Oy1—bp)s SyDq SMB,ABEL'S INEQUALITY 19 Adding these expressions we obtain $1(01 ba) +S2(by—Bs)F «+ + +5 pa Carb Consequently, one arrives at the inequality @) Since m= min(sy,sq,...,5,), it follows that mSs,..., m sq, and so from (1), 1m (b,~ba) S 54(b4—Ba)s «<5 2(0na—Dn) S Sa-1On 1—ba)s oe Hence, summing once again, we obtain mb; S340, 6) This result completes the proof of the double inequality mb, < 3a,b, < Mby, known as Abel's inequality. § 0.5 The Cauchy-Schwarz-Buniakowski Inequality Consider the two sets of real numbers A= (ty dy, +, yh B= Gay Day ss bad and form the polynomial in 2: (2+ 9+ (Aa +0g)*4 «+ (aye by) 0) which is equal to (Ppa bate + 2(ay dy bagDyt «.. bayda)ie HEH OEE... +5.) e Since the quadratic trinomial (2) is the sum of squares of real numbers, its value is non-negative for all values of the variable 2;20 ELEMENTARY INEQUALITIES hence (ata -.. +a,2) (Bye + Batt -.. + byt) (Gaby }aabyt «+. + abyby) 6) ie, (Za5)'-(Ze)(Z4)s0 which is known as the Cauchy-Schwarz-Buniakowski inequality. ‘Equality holds in (4) if and only if a = by, dy = rby,... >, = 1,, where r is a constant of proportionality. § 0.6 Young's Inequality ‘Let #(2) be a continuous function on the interval [0, ¢] (¢ > 0) which is strictly increasing over this interval. Further assume that #0) = 0, @€ (0, ¢) and be (0, fle)} Fig. 1. ‘The area of the curvilinear triangle OA P (Fig. 1) isgiven by the integral f3/(z)dz, while the area of ORB is given by f/-'(2)dz, where /-!(x) is the inverse of f(z). On the basis of Figs. 1 and 2, we deduce the inequality fi teoaesf rear za, ayYOUNG'S INEQUALITY 2 known as Young's inequality. Tf and only if 5 = /(a), equality will hold. Fig. 2 Examptes, 1° The function f(«) = 2°1(p > 1) for all x (>0) satisfies the conditions under which Young's inequality is valid. In this case, the inequality (1) becomes fiene +f, led = ab, 0 5 whence ieaeped PV ppje-n > ad, a ? ‘This inequality is usually written in the form 1 tena Serceee (as20p>u2+2=1). @ p q ’ & g ® 2 The function log(1+z) also satisfies the conditions of ‘Young's inequality. In this case, the inequality (1) becomes Jfvecseyae+ fear = a, whence (1-+4) log (1+-a)—(1+-4) +(e) 2 ab (a,b 20). (3)22 BLEMENTARY INEQUALITIES 3° Starting from the inequality (2), we may derive Hilder’s inequality (Er)"(S6)"= Sab, “) where Oy, bp 20, p> 1, P+ = In (2), replace @ and 6 by to obtain Lbs a,by Ee ET ® an ? Sar ‘Summing now over y from 1 to m we obtain 3 abe whence Holder's inequality follows directly, because one may set Vp+1a Equality holds in (4) if b= aP7 (= 1,2,...,0) 4 Holder's inequality (4) may be used to derive Minkowski’s inequality 3 toot)" (3: ag) + Su)” > tim BO) Since . (GABA)? = ag (Oy by)? 2+ O4(4,4+5y)JENSEN'S INEQUALITY 23, we have (writing 5 for S24) ZX et by)? = Zee tbe + Fe (tebe)? (6) We now apply Hilder’s inequality to each of the sums on the right-hand side obtaining Talat S (Lary (Zlatbeyemrys, halo tO! S (ZH (S (tb) ee—H)oe Inserting these inequalities in (6), and noting that 9(f—1) = p, we obtain ZX lathe)? S (Len? v? + (L477) (E (Met be} Division of this inequality by ((e-+5,)*)* now gives (5). Nore. If p = 1, (6) is an equality. If, however, 0

0, while the function in Fig. 4 is concave and P < 0. Soy LO yg mM, Pai26 ELEMENTARY INEQUALITIES. ‘The condition P > 0 corresponds to Ja fe) % fe) 1) >, |e He) 1 (a2) f (1) — (@— a) f(a) + (ea) f(s) > 0, ie, Bye aon ei fley) + A 1), since, by assumption, 2—z, > 0. Hf 2, =4, 2 = 5, % = H(a-+) (a, be [2, fi), the preceding inequality becomes Ke) <2 1() 1, and show, first of all, that it is convex for x = 0. For k = 2, fay pity) ett arg? + 2a 1 (ES) EE ley eee [0 +). Since 27,2, < 22+2,8, the preceding inequality implies that ie) <4 ie, a) ‘This shows that the function ¢* is convex for « = 0 and k = 2. Suppose now that the function 2*[c = 0) is convex for k—=r (r any positive integer), ic., arte)" atay eye ‘Multiplication by the positive quantity #(¢, +2.) now brings the inequality (2) into the form ee" ete tay eytaay 7 7 (3) It follows from the identity Bey teyee + (ee) (Gag) = ay bag that Rey ban! Sal taj™, because —a, and z,'—x,' have the same sign. Introducing this result into (3), we obtain ey" ehh 2 228 ELEMENTARY INEQUALITIES, ‘This completes the proof that the function 2 (ka natural number > I, 2 = 0) is convex. ‘On the basis of our theorem on convex functions, we may write the inequality eae ey ctttett. tet = 208 he = (#20, 12.008), where equality holds for 2y = 2, 2° We will show that on the interval [ ), x) the function He) = sin x 1s concave For Kehler) Sinz tsints 9 0, log z is concave on (0, c0). It follows therefore that, if a, > 0 (k= 1,...,m), tog (J. 3.45) = 2 3 toga, = 10g (ay 0)30 BLEMENTARY INEQUALITIES Since log 2 is increasing, this gives artdat +4 Ay ” (ayy. ay)!" = Gy. ‘This is an alternative proof of the inequality of the arithmetic and geometric means (§ 0.1). § 0.8 The Fejér-Jackson Inequality L, Vietoris # has proved the inequalities: Sasinke >0, Sajcoske>0 (00, @ 2k—-1 eS te (ISk SH) ) Tf ag=1and a, = /k (b= 1, 2...) the conditions (2) and (8) are fulfilled and the inequalities (1) become a1 Spsinke>o W 0 (0<2n 542 1) 8) 4=({fuera)”, 2 (fier and replace a, } in (3) by |f(2)|[4, |g(#)|[B, respectively. We obtain Set gpl@etel sa%, ype + sleet, whence by integration 1 i dz= 1 ier 1 aoe lee ade say" A oe Thus ; |[[reretas| = and this is Hélder’s Inequality (2). “ Weyetelide SAB, § 0.11 Some Inequalities for symmetric functions * 1, Consider the function H(e) = (e—a)"(e—b)"(e—c)", where a, 8, ¢, 6 (2 1), q(2 1), r(Z 1) are real constants and aAb¢c#a, and its first derivative 1D. S. Mitrinovie: O nokim nejednakostima, Publikacife Elekivo- tehnidkog fakulieta u Beogradu, ser. Matematika 1 fizika, No. 29 (1959), 14,SYMMETRIC FUNCTIONS 33 f@) (2a)? (28) e—ey (pt gtr)at —[blb+e)-+4(e-+a) +7 (a+d)]x+ (pbc-+gea-+rab)}. Without loss of generality, it may be assumed that a 0. We also obtain the inequality ale 2(p+9+7) min(a, b,c) 0, and [pled +db-+ be) +9(da+ ac-+ed)-+r(ab-+bd+da)+s(be+ca-+ab)P —27(pigtris) (pbed + godatrdab-+sabe)* <0. If the zeros of the polynomial P(x) are denoted by 2, 2, ts (<2 <2) and we assume that a 0, if. x > 1. Por which values of x is f(r) <0? Hunr, Express /(2) as a polynomial in (z—1), 1.3 Prove the inequality (2k)! < 2™(b1)* (ea natural number). a Metuop 1, For k= 1, this inequality is valid. Consider any ‘k for which (1) holds. If the inequality (2b+1)Rk+2) < 241)? (E21) @) is true, it follows from (1) and (2) that (Ch+2)1 < (FI) E @) The validity of (2) may be established without difficulty, for it is equivalent to the inequalityELEMENTARY INEQUALITIES 37 <2 which holds for &>—1. Thus an inductive proof is easily constructed. MetHop 2. Since (PR)! = (2A)RR—1)11, (RN) = (2A) INP, the given inequality may be written in the form “OHM ew gyn” Wea? * ‘This last inequality is obvious, because 14 Prove the inequalit lator S 2 (laP+|o}") (6 21). PROOF. Without loss of generality, suppose that |al < [6]. ‘Then [a+b] S 26), whence Jat dl? S 27]? = 2°(\al-+ [0)"). 1.5 Ifa > b > 0, prove the inequality Yo-Yo < Yao SoLwtion 1. Setting a— = rewritten (nanatural number > 1). (1) >0, inequality (1) may be Vere < Verve. @) Assume that the inequality (2) does not hold for some 4, ¢ 06> 0), ie, that Vote 2 Vo+Ve. (3) Then pe pai Wore) = Wo+Ve)s, whence n(n—2) (4). , where the lst factor is 1 oF 2.38 ELEMENTARY INEQUALITIES seem aves S (2) Gi We and 23 ({) Wamwer ‘Thus, assuming (8) to be true, we are led to the false relation (4). Accordingly, the inequality (1) holds subject to the stated assumptions regarding the parameters n, a, b. (This sotution is due to D. Djokovié.) Sotvtion 2. Consider the function fee) = 29-1)" (n> 2 21), Then af (e) = ae <0 > 1). Therefore j(z) decreases for > 1. Hence, since j(1) = 1, we have f(z) <1 for x >1, ie, BL < (e-1)M" (e> 1), (5) ‘The inequality (6) is satisfied for x—a/d> 1, because a> 5 > 0, Therefore, setting x = a/b, we have G9 att—Bun (a8), ie, as was to be proved. 1.6 Ifa>6>0, then VeLB-VELR S a—b. a) Prove this inequality and find a bound for ate —YeFe (na natural number, # > 0,a > 5 > 0). Moreover, how must the inequality (1) be modified when a and 5 ate arbitrary real numbers?ELEMENTARY INEQUALITIES 39 Soxvtion. Start with the identity (ey) tatyt . 2. tay tty) and set ae 2 VaR y = OR Since x =a and y= b (k= 0) and a> 6 > 0, we obtain from (2) (ar ER YB E I) (a pa25-4 Lab B01) < aD", If the left- and right-hand members of the last inequ divided by the positive number apathy... abet oet, we obtain Vath Jere < ab. 8) For » = 2, the inequality (3) becomes VaR VLE ‘The inequality holds for a = } = 0, while k may be an arbitrary real number (positive, negative or zero). Ifa, , Rare any arbitrary real numbers, then \Varte—Ve+R| S |Ia)—[0| (This solution is due to D. Djokovié.) aa. 1.7. Ifa, 6, 6, d are real numbers and if ad—be = 1, w prove that E = at4bt+et+d?pactbd > 1. Soxvrton. The expression E may be written in the form ES 1B (ate)+ (a—a44 6-+a)*+ (b+e)4). ‘The expression in brackets vanishes if and only if a+e=0,a-d=0,b+d=0,b4e=0, 2) ie, ifa = = 0, But if this holds, (1) is not satisfied,40 ELEMENTARY INEQUALITIES Therefore E > 1 for all real a, 5, ¢, d satisfying (1). (See 1.23). 1,8. Solve the pair of inequalities (¥ #0). Soxvtion. Setting z|y = =, the given inequalities become 2 B-1<0, S-1<0. It follows from the first of these inequalities that haces Q) and from the second that s2 2) It follows from (1) and (2) that the given inequalities are both satisfied only if : <0, ic, if and y have opposite signs. 1.9: Solve graphically the inequality |izi-1| Sa @ = 0). 1.10 Show that for all a ia) SOLUTION. j{a) = 4 cos? a48 cos? a4-5 cos a1 = (2.c0s a1)? (cos a+-1) = (Wa)j(a) = 0. cos Ba}-4 cos 2a4-8 cos a+5 = 0. 1.11 Prove the inequality ytbaty O;m, n> 0) Hut. Consider the expression amnmamyt ayn tym, ie, (ey) ley"). 1.12 Prove the inequality alte 0;n a natural number), and hence the inequalityELEMENTARY INEQUALITIES 41 1 1 et Di 0; wa natural number). Huwr. Consider the product (2-1—1)(e—1), — See the preceding problem. 1.13 Prove that if a-4-08 = o?-4d* = 1, then lac-+bd| <1. Hint. Write aa cose, basing, c= cosy, d= sing. 1.14 Given the function He) = a-+og 2, a) find the values of « and hs for which the inequality Fle h) = fe +h) Heh) 242) <0 e) holds, Soto. The function f(z) is defined for x > 0, and for 2> 02th > 0,0—b>0, Fee, h) log(x-+h)-+log(x—h)—2 log x. Since oe F(a, h) = log = the inequality (2) holds for 0 < [Aj fle) > Ha) > KB) @>9), 1Qx)—f(—22) > 2ff(e)—1(—2)} @> 0), where f(x) = (a*—I)/x (a > 1) What happens if x < 0? 1.46 Prove the inequality VEPLVeb-B >a ( Va+e (= 0). @) Taking the square of (2) leads to Qab—(a+b)Vab > 0 which implies (since, from (2), a # 0 and b +0) abt > (ab)Pab < 4ab > (a-45)2, ie, 0> (a—0)8 8) This relation does not hold for any values of a, b, Hence it follows that (1) is always true. 1.18 Prove the inequality 14 cot a = cot $ (0 0) Does it then follow that lzy—ab) < (lal +[b|--0)4? 1.20 For what values of 2 is larybj0)?ELEMENTARY INEQUALITIES 43, ANSWER. Sto aro, b aCe (a <0), @ 1.21 For what values of x is ait), ANSWER. <2<0ore> (a<-1), Tha ma @} x\ (-1 f(a, 8, c,d) > f(a, B, d,c). Hit. Begin by considering the differences Ha, 5, 6, d)—H(a, 8, ds 0), H(a,c,b, d)~fla, b 6d). Generalise to the case Hla, b, c,d, e) = (a—B)*+ (b—c}*+ (c—-d)*+ (d—e)*-+-(e—a), where a 0), we have : 1 Ry < Bog (14+ 55) < gyELEMENTARY INEQUALITIES 45 Hence (1) holds if N satisfies the inequality 1 2 Thus, N = 6 suifices. 2 ones 100 = 1.25 Prove or disprove the following inequalities: 1°: [a)-+{5] < [a+b] < (a) +141, 2: [a)(b] S [ab] = (a][0)+ [a] +), 3: [Va] = [Vial #: (VaP Sn S [VnP+2[Va], Be: [Yn ons [Vaya np yan). (In 1° and 2°, a, 6 are real numbers, in 8°, a 20, in 4°, 2 = 0,1,2,..., in 5°, m is integral) 1.26 If a,b,t are positive numbers, show that at-+-bjt=2Vab. 1.27 By graphical methods, estimate the solutions of the inequality gee Va4+Va-1 > Vath. 1.28 Show that the number atkb Ma Teg (Ob reals A> 0) lies between a and 8, 1.29 Determine the region of the zy-plane in which the point (@, y) must lie in order that its coordinates satisfy the inequality (22 —4xry) | (22+ Sry+2y4) < 0. 1.30 Find the region of the plane in a Cartesian coordinate system whose points (x, y) satisfy the condition |e+al—[y—al| 0).46 ELEMENTARY INEQUALITIES Hint, Possible cases are Ve: efa>0,y—a> 0, 2: eta>0,y—a <0, 8: rpa <0, ya > 0, 0 #: eha<0,y-a <0, corresponding to the set of inequalities 1: —a 1 Resutt. xe (3/4, 2).48 ELEMENTARY INEQUALITIES 1.37 For what values of 2 is it true that (Vitex) <1 1.38. For what values of 2 is it true that 2 < (Bxt—162+10)/(2*—4e-+3) <9? 1.39 Ina Cartesian coordinate system, find the region of the plane for which 1 ayy) >0, GAT EPy") <0. 1.40 In a Cartesian coordinate system, find the region of the plane {@y)iy —he-s)} i.e, find the region whose points (2, y) satisfy the conditions y <2, y> Hes). Similarly, find the following regions {eye Ty {@ ure 0. 1.42. Solve the inequality (sin 8x)/(sin 2)* <0. 1.43. Solve the inequality (tan 8x)/(tan 2) > 0. 1.44 Solve the inequality cos ptsing > 1 (eg., by setting 2 = cos}, y = sing). 1.45 For which values of x is the inequality sin x > 2 cost 2—1 valid?ELEMENTARY INEQUALITIES 49 ‘Hint. Establish this first for @ e (0, T], where T is the funda- mental period of the function sin 2—2 cost 2+} 1. 1.46 Solve the inequalities: etal <1, 2: eke > 2, [Pate > 1, 4: sine-+|sine| > 1. 1.47 Solve the simultaneous inequalities: Yydr—4 > 0, Bxt—22—3y—6 > 0. 148 If atb+e= Oand a =>—$,b2—}, c= —}, prove that (4a-+1)!+ (46-41) (de+1)4 Ss 3. Hunt, (4a-41)! < 2a+1 (@ = —}). Generalize this result, 1.49 Find the region of the zy-plane for which the following inequalities are simultaneously satisfied: ES Ty yt > Be, 2 < Be, at > Dy, 1.50 Find the region of the ay-plane for which the following inequalities are simultaneously satisfied: sy Sb, xy Za y S me, y Zhe (0a Poor. Since at Var + 2a ave 3 2b 2b then at least one of these factors must be less than 1. 1.53 Prove the implication |at— 2043 | oan 48 | 1.54 Prove the inequality beim at 2met pt pom © a4 me pt Perens bom fee Proor. First of all, we have ata2me te Bee") Up ama Pe Fine pe @) 1 tap From (2), we see that y assumes smaller values for # > 0 than for x < 0, Thus it suffices, when bounding y from below, to examine only the case « > 0. For # > 0, we may write in succession Lome pe PP BLS eS (3)ELEMENTARY INEQUALITIES al For x < 0, we obtain, similarly, ae a ae ae OAD the aS — ape ap 2me 2 ep ue oe “att pe 2p _ bim sh bm a ys-l+ im () ‘Thus, (8) and (4) imply (I) 1.55 If 0 0, we fe— 1 <2. SoLuTioN. We write the inequalities in the form +E > |x? 22], 2—y > | 2-1. a Since the moduli jz*—2e| and [z—I| are non-negative, it follows from (1) that y > —} and y < 2; these inequalities yield the integral values y = 0 or y = 1. If we substitute inequalities 0 in (1), we have the simultaneous jet—20| <4, Jel] <2.52 ELEMENTARY INEQUALITIES Integral solutions of the first inequality are «= 0, 2 and of the second, z= 0, 1,2. Consequently, the common solutions are ©=0, 2 Thus, pairs of integral solutions corresponding to y =0 are: %=OH=0 m=%ZW=O. Substituting y = 1 in (1), we obtain Jette] <$, fel <1. ‘The integral solutions of the first inequality are = 0, 1, 2and of the second just # = 1. Hence the third integral solution is gol wel 159 atyb pct B |bepeatab}. 1.60 a(a—0)(a—c) +5(2—c) (@—a) +e(c—a) e—-8) = 0 (a, 8, ¢ = 0) (Schur). a | cd _ (ate) + 161 45 apb ced ath pot (@,4,¢,d>0). 1 162 ltatatt...ta"<4 (0 are 2 (o) (a,b > 0). 1.66 fa—(n4-1)(@~B}ar < O(a 1670 B>n (WEI). 168 2 > n+l (n=). 1.69 2 > nt (n=). 170 © 2" >n* (w= 10) 171 > nt (WB 8), 172 [eta] Sarle—al (lel S75 lal Sr) b> 0).FUNCTIONAL, CONSIDERATIONS 53 ay 1.73 (: >In (>) 174 1.75 1.76 4177 i ta>—t ats @> onan § 2. Inequalities Obtainable from Functional Considera- tions 21 Prove the inequality log(-+2) <= Souurion, Let I(e) = log(1+2) ~ Since /’(2) = ~a¥/2(1+2)* <0, f(x) 0), which is precisely the inequality (I) 2.2 Prove the inequality 2 (+2) @ 20) ‘Does this inequality hold for negative values of 2? Lott tee tient > 0). 23 Let fe) get g tsa54 ELEMENTARY INEQUALITIES Prove that, if 2 >0,2>0,e+2 <1, Het2) 0,2>0, #+2<1 that g(e+2) 1? 2.5 Given the funetion (fleyiee) (0 < lal < 2a), “© Vim eyele) fe = 9) whereFUNCTIONAL CONSIDERATIONS 55 J(e) = 2-2 cos x—w sine and g(a) prove the inequality h(e)> 0 (24 0 (0 <2 < 2a). gee) = Consider /(x) written in the form iz) = A sin® ($2) —4(G2) sin (x) cos (He) in® aaah = 4sin'@Qa) (0 onl 4a). Ii O 0. If x <@ < 2a, tan(}x) < 0 and, consequently, >0 (0<# 0 (t 0. When x = x, one has Het) = [2-2 00s #—a sin t]pay = 4 > 0. This proves the inequality (1). 2.6 Prove the inequality ylog’ +(-y) og 4 2-2" (O p—g (ew > 1), ay where p and g are real numbers (0

1, ie. the function is increasing on the interval (1, +00). The function j(z) therefore attains a ‘minimum at x = 1 which implies that fai = 0. Thus, f(x) > 0 for x > 1, which was to be proved. 2.8 Prove the inequality (y—a)a" log a < a¥—a" < (y—n)a" loga (x1). (1) Soturton. The function j(!) = a'(a > 1) satisfies the condi- tions for the Mean Value Theorem on the interval [z, y]. Hence wna Sa'loga (e 1) is increasing, it follows from (2) thatFUNCTIONAL CONSIDERATIONS a7 @loga<* yon I), which is equivalent to (1). This completes the proof. 2.9 Prove the inequality 1-a Ze*** ((x| sufficiently small). SoLUTION. We set /(z) = e-*-** and determine the equation of the tangent to the curve y = j(r) at the point (0,1) Since f'(z) = —(14-2x)e*~*, the equation of this tangent is, ya lan, Since (@) = (4a $421" <0 for we (—3(14-V2), 3(V2—1)), we conclude that y= f(z) is convex upwards in the neigh- bourhood of (0,1) Consequently, 1-# = e+" for j2| sufficiently small. 2.10 Prove the inequality ete ose oee)). o) Using this result, prove that the integral (ube ef 82 ae nn) is not greater than 1/(2m). Sovutioy. Consider the function i (x) = 2log x —58. ELEMENTARY INEQUALITIES Since 1 Qlge—-2— +0 W<2<)) Hence it follows that ‘The inequality is obvious, because for # > 0, sgn (x log x) = sgn (#*—1). It follows from (1) that £108 ae < Afra = 12m), (Revised from a solution by Z. Pop-Stojanovié.) 241 For which values of 2 is loge S Va? 2.12 Sketch the graph of the function Va-—Ve—a (a>0) and find the values of « for which Vena > 2FUNCTIONAL CONSIDERATIONS 59 2.13 Find the values of x for which Yea >b (nanatural number > 1; 4,6 > 0). 2.14 Factor the polynomial 2e*—32%4-1 and determine the values of for which it is positive, Hence prove the inequality He) = 2eosa-sectz—3>0 (0 <2 <}n) By considering the function frou o 3¢ (0-<2 < 4a), Sorution, First of all, we have 2x! 3at1 = (@—1)2(2e-+1). Hence 23-1 >0 (e> —f and x £1), a) Warp L<0 (@<—h. @) Since 0 < cos @ <1 for 0 <6 <4n, we find from (1) 2eos O—3co% 0-1 >0 (0< 0 < 4a), or Zeos B+sectO—3>0 (0< O < a). (3) Integrating the inequality (3) between the limits (0, 2), we obtain 2sine+tanz—3¢>0 (0 e. It follows from this result that (3) (4) since V7 << V8. For » = 7, direct calculations show that the inequality (4) holds. From (1), (2), (8), (4), we have the inequalities (ValY" < (VFTY® (n= 1,2,8,4, 5,8), (Vay > (vnFIY® (WB).FUNCTIONAL CONSIDERATIONS 61 2.17 Prove that x—tanh x > 0 (® > 0), and hence that &( mhz) >0 @>0). SoLuTioN. Consider the function Ne) = 2—tanh 2 a Since f(a) = tanbte>0 (@ 40), @) the function f(2) increases for # > 0, It follows from /(0) = 0 and (2) that j{e) =2-tanhe > 0 (@> 0) @) By (3), it follows from the relation @ /sinh:2) _ cosh al) and the inequality cosh 2 > 0 that aii (=) >0 (>). = (e—tanh 2) a 2.18 Prove the inequality Wr swith @ 20) (1) Sotution, Consider the function Ie) = 24-22 (w= 0) and its derivatives: f@) = 3e—2, f"(2) = — Fer (@> 0). For x = jg, the derivative /'(z) vanishes, while /"(z) <0. ‘Therefore f(z) attains its maximum value of $ at x Consequently, Vr—2r <2 (> 0), which was to be proved. For # = 0, the inequality obviously holds.02 ELEMENTARY INEQUALITIES 2.19 Prove the inequality H24c0st) > 3sint (> 0). w SoLUTION. Since 24 cos#> 0 for all 4, we may write (1) in the form Bsint no 2 eos Since , L008 t)# 10 = (Fe) 2° the function /() increases as ¢ increases. Since /(0) = 0, this establishes the inequality (1) 2.20 For (#1) an arbitrary positive number, prove that 1: P1> pel) (p> lor p <0), 2: P-1 1, then =0 @=1), <0 (0 (@>1). Thus, the function /(z) attains a minimum at 2 = 1. This is the proof of the inequality 1° for p > 1 It p <0, then PIs. OI). Hence =o @=)), <0 00 @>n. Consequently, 1° also holds for p <0.FUNCTIONAL CONSIDERATIONS 63 e: Wo). Hence SO O<#<1), f@))=9 @=)), <0 @>)). The function /(e) attains a maximum at 2 = 1. This proves the inequality 2°. 221 If a and b (\e} <1, || <1) are real numbers, then Vin#LVInF < VIB eT @ SoLUTIoN. Assume that (1) does not hold, but that Vic Vini > 2 VIR apo Then, squaring both sides, we find Vi-@ VI—# > 1—ab, ie, squaring again, (La!) (108) > (1—abjt = 0 > (a—B)%, On the basis of this contradiction, we conclude that the in- equality (1) is true. 2.22 If /'(#) is an increasing function, then F@+Y > fe+1)—f(@) > f(a). Q) Use this result to prove the inequality a (Vv@rip—} VED EV (WH =12,3, Sonution. By Lagrange’s theorem, He+)-f@) =/@+0) 0<6<1). From the hypothesis on f'(e) and the preceding equation, it follows thatot ELEMENTARY INEQUALITIES f+) > fe+)—f(e) > Fe), as was to be proved. ‘The inequality (1) may be written in the form Het) —fle) > £(@) > fe)—He-1). (3) Let /(z) = $28 and write (8) with x successively replaced by 1, 2,..., 9, Sum to obtain the inequality (2). (Revised from a solution by Z. Pop-Stojanovié.) Generalize the inequality (2) 2.23. Prove the inequality 1-2" >n(l—aje9 (0 <2 <2), 2.24 Prove the inequality peop

0), p and gis a? logta < at-na? 2.26 For what values of x is zoe 2 ae ayesi42? a Deets a Souution, If = —1, then, by Bernoulli's inequality (§ 0.2), (42) S145. @ For values of 2 for which sa 14$-F<0 @2B-1) Co) the inequality 1+2/2—29)8 < (1+-2)"* holds, ‘The polynomial 1-+2/2—a4/8 assumes negative values for = exterior to the interval (—2(V3—1), 2(V38+0) wyFUNCTIONAL CONSIDERATIONS 65, Since, in addition, the condition 2 = —1 is satisfied, we con- dude that 145- : < (+2) for 22 2(V841). (4) If x & —1 and if xis in the interval (A), ie, if -1S2 523+), (B) then 2+1 20 and 142/2—24/8 20, and after squaring the inequality 1+-2/2—24/8 S (142)! we obtain a(e—8) <0 which holds for OSrs8. (Cc) Conditions (B) and (C) jointly require that 0 < x = 2(V3+1). Collecting these results, we conclude that the inequality 14 Z-E sasaye holds for x = 0. Since, in addition, the inequality (2) holds for «= —1, we conclude that (1) holds for = 0. 2.27 Is the inequality 2* = 14-2 log 2 true for x > 0? 2.28 Show that, if xis real, then 2 ath - rs 3 Spe4i = Give a geometric interpretation of this result. 2.29 For what values of 2 is (a—ae)t§—Ba(a—2)8+ $a2(a—a)— fat (a—a)? < 0? 2.30 Prove the inequality (ay > Ger #0) ay66 ELEMENTARY INEQUALITIES ProoF. Consider the function ie) =(=)" eo, @) and its derivative re) ~ (8 + wos) 1. @) ‘The sign of the derivative is the same as the sign of the function ate boa (x) = pained 4) ae) = That 8s i Since oo ~ Wea Ota) * the function g(z) is decreasing and, consequently, ga gle) > gl+o0) = 0 ® On the basis of (5) and (3), we conclude that the function f(z) is increasing, whence fapa\e* fa\” ree °, 6) (GE)">G) @rereexn, © ‘The inequality (1) is the special case of (6) obtained when. (Proof by D. Djokovié.) 231 Prove the inequalities (l-detjsin ae < @—fatjoose (0<2 (@—Jx?+yFyrt)oosa (0< a gle) O), f'@)>e"@) =2 (<2 0, f (/(@)—8'(2)] dx = [f@)—-g@)]—[4(0) —8(0)] = flz)—g(@) > 0, which was to be proved. (Solution by D. Djokovié.) Remark by D. Adamovié, If we consider, instead of the given in- cquality, the equivalent inequality 2 < sin z//cosa (0 0). a) 2°; Starting with this inequality, derive the inequality af log p, 2zh 108 go, (2) for p, > 0,q,>0(¢=1,%...)) and68. ELEMENTARY INEQUALITIES 2m Ze (3) Soxvtion of 2°. Since p,{g, > 0, we deduce from (1) that Pegg bw oe Since q, > 0, the preceding inequality becomes, after multiplica- tion by qo be log SS bie p 8G, Pe ‘Summing both sides of this inequality over i, one finds Q Palo he 2 = (a). With (3), this inaquality gives > tog 20> 3 pulogePlogg) 20 = x Plog b, = 2 Plog de. as was to be proved. Equality holds in (2) if and only if 6, = 9, (= 1,2... 7). ‘The inequality (2) occurs in Information Theory. Seo, for example, L. Bril- louin: Science and Information Theory, New York 1958, pp. 13—14. 2.34 Prove the inequality 0< Vipe—1-tet het < fat @> 0). Proor. Consider He) = Y¥iFe—1—4et hat (@ > 0) and its derivatives £@) = 4042) 4492, f'@) >0. = al aa salFUNCTIONAL CONSIDERATIONS 69 Then Fe) =P) fr e@)de> 10) = 0, 112) = 10)+ f reeide > 10) = This proves the left-hand side of the inequality, Next, let 82) = lhe dette PTFE (o> 0), whence Bet aha! —A(l+2)28, BB+ 5 (la), 1 ~ BOM Gael > and so ere) —e"O+ [ecards > £”0) Be) =e 0+ fe 8 (z)de > g'(0) = 0, ate) = 0+ [e'(ehde > &(0) = 0. This proves the right-hand side of the proposed inequality. (Proof by D. Djokovié.) 235 Ifa>0,b>0,a+6=1, then aan: {, Reece: 4 1 5, (2) ea) BS SOLUTION. It follows from 2/5 < a+b = 1 that A(ab) 24. ) Furthermore70 ELEMENTARY nmQUALITIES and hence nit —ts-2(++ 3) (+5) From this result and (1), we obtain = (+42, ie, WL Ses Lek Nore, The proof may be obtained in a different manner, using the convex function f(x) = (z-+1/z)* for 2 > 0. In that ‘case we can prove the more general inequality (+2) + (+3) 5 (a> 0;4,5> 0,045 =1) The function Ie) = (+ yy (a>) for 0 0 (0<2<1a>0). Consequently, for a, 8 > and a+b = 1, HON) 3 (2) = 1a,FUNCTIONAL CONSIDERATIONS m1 ae b+ 5) ae aye & (0+) + (+5) aeaer = (This solution is due to D, Adamovic.) 2.36 Prove that i a b—a tana <——" (0 Sa 0). Proor. Let us assume that for some 2 (> 0), the inequality WF 0) (2) does not hold, but rather that yReitis @>0). Then, raising both members to the n™™ power, we obtain + (hit Qa t +0) Lie (14+ 44) Since2 ELEMENTARY INEQUALITIES (2) ae+- + (i) pero (e > 0; m an integer = 2), the relation (8) is false and, consequently, (2) is valid. Next, consider the function Ne) = YF ta at and its successive derivatives, re) = [rayon 1 * re) =" te), (—ny(—20) * Since 7’"(z) > 0, 7°40) = 0, f'(0) =0 and f(0) = 0 for z >0 and n 2, it follows successively that /"(z), /’(z) and j(x) are increasing functions, whence re) (Haynes, fz) >0@> On = 2), This proves the inequality 1 i 2c Fs integral & 2). Leben hat < FE (> 0; w integral 2). (4) The inequalities (2) and (4) together yield (1) Question. What happens if x = 0 or m= 1? 2.38 Prove that slog 2+et—ay PRooF. Consider the function 0 @>9 Hy) =e loge-et—ay (2 > 0) and its derivatives fy) =et—2, fy)FUNCTIONAL CONSIDERATIONS 3 Since /”(y) > 0 for all y, the function /{y) attains a minimum of zero for y = logz-+1 (@ > 0). Consequently Hy) 20 (@>0), as was to be proved. 2.39 Prove that 1” Gril +5) @>9. @) 0,1 > oH aa ae (+0)! > GH)! Now (1) follows from (2) and (8). ma, (@) 2.40 Prove that a function f(x) having a second derivative has no zero in the interval (a, 4) if there exists ce (a,3) such that for every 2 € (a, 6) £(c)—2f(0)f"(@) < 0. () Sonvtion 1. Let us assume that the converse statement is true. If ye (@,6) isa zero of f(a) and ce (a, b) satisfies the given condition of our problem, then we have the following: (yne} =e) 1M) = e+ fOr f(a) = 0, wela, 5),4 ELEMENTARY INEQUALITIES He) + aerate = 0(a = ye), which implies: £*(€)—2Ale)f" (2) 2 0. From this contradiction it follows that f(z) has no zero in the interval (a, 6). Sorution 2. From (1) it follows that j(c) #0. For 2 #¢ we obtain 24(c)F (@)—2F(€}F (e) = 2F(C)F"(E) (ee) Z (ee) *(e) according as x 2 ¢. In both cases Hoe) = 2PO)+2N0) fF He)de ae rot f [27(6)f (e+ @—e)F(e) de = H12/(e)+ (eo) (cP. Henee, f(z) cannot vanish, (Problem and solution 1 by S, Presié, Solution 2 by D, Djokovie.) 2AL lasinz-+b cos 2| < VaHT 2.42 cost x-+sint x =}. 243 SEE gene zy, 244 WE> ss @>). 245 05 (e+a)i|(et+e+1) 5 4(e—a+1). 246 at—atl < (x,—a)"— a)" (0 Sa Sm, Sm). 247 Di Seb le S1 S251 p>. 248 y(a¥—1) <2(at"—1) (a> 03a 41;0<2 n0 0 and 2 #1).2.51 2.52 2.53 2.54 2.55 2.56 2.87 2.58 2.59 2.60 261 2.62 2.63 2.64 2.65 2.66 2.67 2.68 FUNCTIONAL CONSIDERATIONS. ey 142 nat ree Ssbi sen 142%... pete beget. Sate a x Zz as Ge I)oos 575 > Lh cos F (= 2) (ea) > (ea) O<2 0). 2x/(2+2) 0). a> (La)logt(+2) (@> —0). Hlog(1+2)—2| Sa? (x) <4). Lx loge V1+a8) = V1i+z, log(1+-2) > (arc tan x)/(I+2) (> 0). 2a are tan. = log(1-+23). (e+ Dlog(e+1)—x loge > 0 (> 0) 248242460 log x > 6a (2 > 1). atllog 2| <1(ae) (0< 2 <1; a>0). flog(1+2)| = eure @>-1). eet Nery t 1 geo), wet ) 0 and x 41). (0<2<)).76 2.69 2.70 271 2.72 2.73 2.74 2.75 2.76 297 2.78 2.79 2.80 2.81 2.82 2.83 2.84 2.85 ELEMENTARY INEQUALITIES 1 yt satsenpe5>0). 1 o> gt ©>ok= Slee & b>0;0>0). eV oaias expe > (142) >exp 24 ey >o pe>(1+2)>enZt wy>o, log (4: VIM) <1 ttoge (@ > 0). a-d 6 = < log a—log b < e<(1+2)4* (e> 0) > Y(l-z) 0<2 < 1/2). athe — lee (e>0). ePoosz0)POWERS AND FACTORIALS 7 1 30 2.86 2.87 2t}-82-+1224 log 2 > Sxt+-1 (@ > 1) 2.88 @<2<)). 2.89 log (I-+cos.2) < log 2—Jx* (0< x < 2). 2.90 log (1+2%) 0). 2.91 y>0;r> I), 2.92 oe y>00 e422 (EET) > (Bn) > 1) SoLutiow 1, 1°: For » , the inequality n> 2n—a)!t w is valid. Assuming that (1) holds for » = &, ie., a > kW)! (2) then the inequality obtained by multiplying (2) by (4+1)*+/A* is also valid; thus, (ETE > (BRA) (RE LEHR, 3) If (QR-A)N (AE A)EYRE > (2A-E A), (4) then (YEAS RED 6)8 ELEMENTARY INEQUALITIES ‘The inequality (4) is valid. Indeed, instead of (4), we may consider (+1 (R+1)AE > 0. 6) After application of the binomial expansion, the left-hand side of (6) becomes BAY yg (REY ge he AED (ayes (aire OY (el) which is positive for all natural numbers &, From hypothesis (2), ie., /(R) > g(#), we have shown that HG41) > ge+1). Since, in addition, (1) holds for » = 2, it follows that it holds for every me {2,3,4,...}. The second inequality may be proved in a similar manner. SotuTion 2. 1°: From (a—b}* & 0, it follows that a 0(2a—8), « with inequality if @ 43. If we let a= m and b= 2h—-1 (b=1,2,..., successively we have wt 1+ @n—1), 3+ @n—3), nt > (2n—1) +1, Multiplication of these inequalities yields n> (QI) n> (2n—I)IE 2: Thin (1) weset a = n41,0 = 2k (k= 1,2,...,), we obtain (+t = 2+ 2m, (tut S 4 (2n—2), (+t > 2m 2. After multiplication, we obtainPOWERS AND FACTORIAES 0 (+L > (mn) > (tI) > mit Golution by B. Mesihovié.) 3.2 Prove the inequality (tly 4+2)" > a(n)? (n> 1). Hint, Use the inequality (-+1)" > (28)!! (n> 1) of the preceding exercise and the inequality 2" > 3", 3.3 Prove the inequality me< Ce (a natural number > 1). a Proor. We start with the inequality gy" = (uF mm” >2(k=1,2,3,. Setting = 1, 2, 8,..., #1, we obtain the inequalities: G)>a Gas Gay =a eEy > After multiplication, we obtain (nets >a, Read whence the inequality (1) follows immediately, 3.4. Prove the inequality # a)! al (nt? (032,84...) a SOLUTION, For = 2, the relation (1) certainly holds, because 16/3 < 4/@N)t = 6 Suppose that (1) is valid for some natural number # = & (= 2), that is to say80 ELEMENTARY INEQUALITIES URL) < (2A) /(R1) or, alternatively, (2A) (RY? > 4°/(R+D). (2) After multiplying both sides of (2) by (2441) (2-+2)/ (+), we have (R42) YET)? > R41) Bh+2)4/ (R41), which may be written in the form (2R+2)! _ 4h (4-2) (2R+1) (py eae Since (b+2)(2h+1) serie 7+ spre? 1 A>) wwe have (@hta a eer {(k+1) IP R+2 ‘The proof of the inequality (1) now follows by induction. 3.5 Prove the inequality Qn Galt (nt From the inequalities it follows thatPOWERS AND FACTORIALS SL 2-4-6... @n) Ne gtgrg gery 7 Hence Ni< ine N <.yVi. 3.6 Prove the inequality BEAL... (@n)1> {(-+1)!)" (a natural number = 2). (1) Soxution, For # = 2, the inequality (1) is valid. Let us suppose now that it holds for 1 = k—1, ie, 2141... (2R—2)t > (ANE. ‘Then, 2LAL... (2)! (2B)! > (BR) ICRI 7 eer = (2h) (R21)... TYE > EDMENY = (LI), because each of the factors 2k, 2k—1,..., R-+2 is greater than +1. Consequently, the inequality (1) is proved by induction. 3.7 Prove that n! > n"* (w a natural number > 2). PROOF. 1! = 1-2...m, (Wl}?= 18-28... which may be written in the form (nt)? = (1 -m} 2Q1—1)} B(W—2)}... (HH). fe, () where 7(1 SS) is a natural number. For all such 7, r(n—r +1) Ba, since (71) (r—n) SO. 2) If we set r= 1, 2,..., m successively, we obtain from (2) Lim 15; 20-1) > msrp) > mpm Ln82 ELEMENTARY INEQUALITIES Hence it follows from (1) that (nlyE> wm (n= 3,4,...), nl > at 1 = 8,4, 0.) 3.8 Prove the inequality guru > ml (wa natural number > 2) Hint, Write 2-0 in the form Bhs = att amt, and use the inequality 2" > n+1 (n = 2) which may be proved by the method of complete induction, 3.9 Show that ifm, a, ..., 2 ate natural numbers satisfying the equation ztayt... +2, = np (p a natural number), then altel... +2_! 2 nt) 3.10 If both sides of the inequality 2sina—sintz < 1 (2 (1—sin z)* 2 0) are multiplied by sin (0 <2 LITRE L) +! Then we find that 2) 1 (ety Bk+1 (2h+-2)* > aya’84 ELEMENTARY INEQUALITIES 12K4-28K2-4 198-44 > 128428824 208-44, 19% > 208, which cannot be valid when & > 0. Consequently, the relation (2) is valid. By induction, this proves (1). 3.44 Ifn— a, (n, natural numbers), then » ye n/t (a!) Sn" / TT (%,"*), a) PROOF. Since every term in the expansion of (7,-+1ty-+ .«.+)" is positive and less than the sum of all the terms, ih ans (1) n! sm (1) Equality holds in (2) if m = my This solution is by Chih-yi Wang (Mathematics Magazine, vol. 31, No. 2, 1957, p. 113) 315 Ifa, = x(e—1)(2—-2)... ( % = 1, prove the inequality 1) (ranatural number), (2m) -12n—3)q_g(28—B)aig +s > Maatnnatnns «=» where the products on both sides extend over all non-negative indices of the same parity as n—1, and # > 1 SoLuTioN. Since «= (Gr if @ is a positive integer, the products on the left and rightchand sides may be written in the forms 2) og 45 (20 2n—3 L (i) ony a) (v5)!POWERS AND FACTORIALS 85 R= (02) wor (9) oayi(," 5) (wo)... dean, og =" Gy) (n—2—1)1, From these expressions we have (2-1) R_ den (y—ay—1) _ dean atin! o> Eh ana DGB! Gayl (ra) OM al al on as tea (o0-ta)* From the characteristic symmetry of the binomial coefficients and the fact that the middle coefficient of an expansion is the greatest, we have (2) <(%) oer eden Because (4) = (2n)!/(m)*, we thus have R/L <1, ie, L > R, which proves the given inequality. (This solution is by S. Presi). 3.16 Use mathematical induction to prove the inequality n log n—n < log n! < (n+4)log n—n+1 a ( a natural number > 1) Souution. We set (0) = mlog n—n, o(n) = (n-+-$)log n—n +1. Let us first prove the inequality ‘u(n) 1) (3) It then follows that log Al+log(k+1) < (k+4)log k—k+-1-+log(k+1). If we show that (E+ pllogh—A-+1-+log (+1) < (B+ Slog A+1)—(R+1)-+1, (9) ‘then we shall have shown that the inequality (7) holds for 8 = k4+1, if it holds for # = A.POWERS AND FAGTORIALS 87 ‘The inequality (9) may be written in the form A 4) og (b+8) log pT +1 <0, or -+o7 <0 (10) Setting we find that ee+@e+iP For x > 0, it follows that /'(z) is positive, whence f(z) is an increasing function for positive x, As > +00, f(r) -* 0. Hence Hz) <0 for x>0. This completes the proof of (10), and the proof of (1) follows by induction, 3.17. Prove the inequality n't > (41) (w a natural number > 2) a) Proor, For m= 3, (1) is true. Let us suppose that (1) holds form = k, ie., that BEALS (R+I)F (h > 2). (2) Since a 9). Rthy, Ate (tie > Ata) = SoS, it follows that (3)88 ELEMENTARY INEQUALITIES From (2) and (8), after multiplying, we obtain (ert > (b+2)4, whence the truth of (1) follows by induction, 3.18 Prove J. van Hengel’s inequality (mtr) (+4) log mnt 9 B2L Ant) < (12839... mys) <4An41) (> 1) 3.22 2S 14nVE (1) 3.23 YT < "Yn (22), ame nis(f) =o. 3.25 (nl) 2) 3a7 <1 (f) ), aVk~ Nk ‘Sk+1 3:71... (4n—1) oe 398 Saris. Gang) 2) 4:7-10... (@n+4) 3-5-8... (Bn+2) atte Sadpbd* ped" (a, b, c,d > 0). ate(cdyei4 < cdt(atb)e4 (a, 0, 6,d > 0). ardh(a+b) S (ato (a,b > 0) Daye (HEE +a) a < (a+8)*(a—b)e < (+ ) (ah . 4, a+b > 1 (a,b > 0). a > @d* (a> 0 >0). priveaingaiioraenyr/inverr) > b+ gtr) (p,q, 7 natural numbers) (14a)!-"(1-ay4# (at—BHN(ar 45°) (a> b>0, p> 9) (pa+-gb)r** > (ptg)*- arb (a,b, p,9 > 0p #9). arB-*4 (I-a)*(1—B)-* 51 (0 0). 349 a> (“EY > 0,020) abpby cries 350 (EMEA ener (ae > 0 3.51 (bed-eda-+-dab+abe}** we @ieidnapeidinnctioisegaitiot (a tb pepdymaiett (@,4,0,.d>0) aa (1i2)"> (raat) oo nen, 28s (02) ast (ey oC) Oey 3.55 2ab(c+1) 0). ase wr (Es faye om. 3.57 ml-ful > (m—1)M4(n41)! (m—n > 1). 358 I1(;) sy 2%) 3.59 = pa—(P-1)b (a,b > 0,6 > 1)SUMS AND PRODUCTS o 3.62 3.63 aus (-2)">(-2)" ocececn 365 (I+ atj—(1ay Sa (aS 0,1 = 2) § 4. Inequalities Involving Finite Sums and Products 4.1 Prove the inequality wt 42> aVaFI-l) (WB). () Vai Baa aatiy Souvrion 1. For m= 1, the inequality (1) certainly holds. Let us suppose that (1) is valid for m = k, ic, that 2(Ve+I—2). ios 444 ts wari-14— Va VBS VEEL Vievl If we can show that 1-1)+(1/Vk+1) > 2(VkF2—-1) (R21), (2) then the proof of (1) will follow by induction, Now (2) is equivalent to 2VREL+ (1/VRFL) > VEE. If, on the contrary, 2VRFI+(1VEFI) S 2VRPR, (3)92 ELEMENTARY INEQUALITIES then thas aVEPN ETH 5958. Since hypothesis (8) leads to a contradiction, it follows that the inequality (2) holds for & 2 1. Sorutiow 2. Since the function f(x) = 1/Vz is monotone decreasing, 3m > rene oS 1 oati-0), which was to be proved. 4.2 Prove the inequality = 1 1 1 2 wotintinny > 2B” 2 tierra "> 4.3 Prove the inequality aoe >). Sowvtion. For 1 Sy Sa, wy < 2n, Thus, 1j(n+s) = 1/20 for cach of the m terms of j(n), whence f(t) > nj(2n) = 1/2. 44 Prove the inequality Lio mol Tapa Faept Soturion. Let us suppose that the given inequality is valid for some natural number n= k, that is to say fi) > a) les of this inequality are increased by fl Sl @=1,2,3,...) If both ee ee Teel tae aaa EaSUMS AND PRODUCTS 93. we obtain a 1 _ ee 2 ee eps * HS BRE ) If 20 1 geye* aera ea? (8) it will follow from the induction hypothesis that /(#-+1) > 1 Now (3) is equivalent to oA 2 Rea" Sha BRS" ie, 2(8k+3) 2 Geral wpa" eoeeeee Gera 1 Gray’ which holds for all & = 1. Since, in addition, f(1 for all natural numbers 1. 13/12 > 1, weconchude that j(1) > 1 4.5 Starting with the inequalities Te el) Sas en ea prove that m wee 3m+1 (eT mE SZ OWEN) Bh T) * dow} OMFD) for m a natural number.94 ELEMENTARY INEQUALITIES 4.6 Show that if a, =a, >a, >=... 24, =0 and then Sershe e) ‘Proor. After multiplication by a,—dy.4, (1) becomes rai) Sa Stes) FH HELL m), @) where dyi1 = 0. Summing both sides of the relation (8) from k= 1 to k=, we obtain Sars San a Win Sevan’ inqusty (Ze9J'=(3+) (3%). the retin () yids (Y's whence follows (2). olen Equality holds in (2) if and only if a, = 2, (y= 1, 2,..., m). 47 Mansa F< bin Sm € Yc 2 Ye (i- 1, then (6), after division by 1—a*, becomes (eb Natt Stat... pat. (8) Since a > 1, we have the inequalities aA Lat > at... aH > att, whence follows (8). ‘Thus (5) holds for all a = 0. This establishes the inequality (1). 4.9 Prove the inequality $2) @-mpeuart ct a (a natural number, 2 a real number). Poor. We start with the binomial expansion 3 (i) t= an ® After differentiating with respect to ¢ and multiplying by ¢, asa Sli) After another differentiation tion by #, this yields nt(L-+ey. (3) respect to and multiplica- 3 (7) B= nl(1 pt) n(n—1)e(L pe 4 @ = mill nt) Ifa # 1, the relations (2), (3), (4) with ¢ = a/(1—z) become, respectively, 6) @)SUMS AND PRODUCTS 97 i()eonm Direct verification shows that (5), (6), and (7) also hold for z = 1. If we multiply both sides of (5) by %%, (6) by (—2nz) and (7) by 1, and add, we obtain na(l—2-+na). m 5 (rat—anet.seie(2) (nyt = means) ata, ie, = (ven (7) (2) = ne(1—2). (8) Next, since a(l—z) = f= (4), we have a(l—2) = 1/4. (9) From (8) and (9), the inequality (1) follows. 4.10 Prove that 5 il -3 ff] Zntn (na natural number), ‘where [a] represents the greatest integer not exceeding a. 4.11 From the graph of the function 2(p > 0) deduce the inequality vcs wide (0 1, i 1 at +—> 1 (wa natural number > 1) we find ra LL ntin att teat 4.14 Prove the inequality Tee i: aon va (Hz 2). va v3 vn PROOF. If k I/y/m. Setting k= 1 2,.....1, in succession, and adding the inequalities so obtained we find =vn which is the desired inequality. (Solution by B. Mesihovié.) ais (5 wy" = (34)" 416 TLQ+m) > 1436, (> 0 or —1< a <0;n>1) 417 TI0—a)>1-Sq 0D.4.19 4.20 422 4.23 4.24 4.25 4.26 427 4.28 4.29 4.30 sus AND PRopucrs 99 = Ie \(& 1 Tate) <1/ (34) (3a <1a>0) Ht (la) <1 / (t+ 34) (1> a, > 0). Tote) sts Be. +3 (n205=3) a def) z let $0 Sa51) Hate) n). s,102 ELEMENTARY INEQUALITIES Prove that (S24S2y"" S Y(sin de) (0 <2 <2) a) ANSWER, Sy = {cos H(m-tnt-L)e sin $(m—n)2}/(sin fe), Sy = {sin 4(m-bo-+1)a sin $(m—n)x}(sin 42). Note. How does the inequality (1) change if ¢ (—o0, + 00)? 5.3 Solve the inequality sin x > sin 3 graphically. 54 If 0 2 coste—1? Sorvtiow, The given inequality may be written in the form 2sintztsin2—1> 0 2(sin 2+1)(sin 24) > 0.TRIGONOMETRIC FUNCTIONS 103 Since sin x > —1 for all « # 87/2-+2h2, the above inequality is satisfied provided that sina—}> Ox (x/6-42hz, Sx/6-+2kx) (k= 041,42...) 5.7 Show that if is a natural number, then ne See |Sh (eh k= 0,41, 2, ) ay SoLvtioN 1, For m= 1, the relation is valid. Let us suppose that (1) holds for n = , ie., that sin ke Sinz @) Consider the quotient sin(k-fI)x _ sin he Ste 8 2 4008 be sing sing It follows from this result that in( + He | | Tsing he SG | eos 2l-+leos Az From (2) and the preceding relation, it follows that sin(&-+1)2! sine [san t This proves (1) by induction. SoLvrion 2. The inequality (1) may also be proved by starting with the identities “in Spe sine. = 2cos e+-cos 324+... +c08(2p—1)z} (p = 1) = 142fo0s Befoos 4+... +008 2px} (p21). From these it follows thatlos ELEMENTARY INEQUALITIES in 2p | i: sing Aja S21. sing 5.8 Show that, if & is a natural number and if A and B are real numbers, then cos RB—cos kA | ces Boe set : a cos Bees | * (008 A #008 BY a Proor. Starting with the identity cos p—cos q = —2sin H(p-+g)sin HP—9), we find that cos KB—cos kA _ sin $8(B+A) sin }4(B—A) “cos B—cos Asin $(B4-A) sin }(B—A) emsa—on | | sin $4(B+-A) | |sin $4(B—A) cos B—cos A |~ | sin }(B+-A) | | sin }(B—A) |" Since |sin nz/sin x] Sm (n a natural number), @) the preceding inequality gives precisely the inequality (1). ‘The inequality (2) was proved in 5.7. 8.9 If A and B (cos B 4 cos A) are real numbers and if &(> 1) is a natural number, then 0s RB cos A—cos kA cos B ee er 7 cos B—cos A = % So.vniox. From the inequality |sin re| Sr |sina| (r a natural number) it follows that sin rx sin sy sing sin y sin ra sin sy--sin se sin ry) sing siny Snes (7s natural numbers). If we now set r= 241, 5 2 = \(4+B),y = (4—B), we obtain the inequality (1).‘TRIGONOMETRIC FUNCTIONS 105 This solution was given by L. E. Bush (The American Mathe- ‘matical Monthly, vol. 64, 1957, p. 651). 5.10 Is there ana such that e * t>1 45g > cost > 1 “ © /2)2 pe ae) 5.1 What conditions must be satisfied by the positive numbers @ and b so that the following inequality is valid: ae (etesine a+b cose (@ 20)? a) SOLUTION. If we set 6a = & (> 0), the inequality (1) becomes 2) We cannot have & > 1, because when # tends to a root of the equality Iho # = 0, {and such roots will exist), then the value of the function (L+4) sing +R cos tends to infinity and (2) cannot hold. For & = 1, the inequality (2) becomes 2sine 7214 cose Iv =2tande (220), which is false. Consequently, we must take & <1, Then (2) may be written in the form: z—sin 2-+R(x cose—sin x) = 0, zoose—sine 1 ein ® E lim f(x). fe) = (z= 0). Here we take /(0) =106 ELEMENTARY INEQUALITIES If m denotes the minimum value of the function /(z), we seek the conditions under which m 2% —1k = —a/b, atmb = 0. 8) We shall show that m = /(0). First of all, we use L’Hépital's method to find reose—sing | —ssing (0) = lim-—__—. = lim #0) ao 6t—sing es01— cose sing—xcose |, wsing—2cosa = tim SPE EONS _ pig FINE Pos 2 aang ae eee ey We must prove that /(2) = —2, i., that woose—sing 2—2vt2sine (220), 22+ cost) ZSsinz (220), asin ele) arg =0 20) ® Since a 2Feosa)eosa-+sinte _ /1—cos2\* fo = Soa” Gena) = the function g(x) is increasing, so that lz) 2g) =0 (@ 20), which is identical with the inequality (4). Hence m = —2, and the condition (3) may be written asa 2 25, Thus, we have proved the inequality = (ot bsinz afb cose (Solution by D. Djokovié,) 5.12 Prove Everitt’s inequalities: 2 (@=%>0220).TRIGONOMETRIC FUNCTIONS 107 sin _ 2(1—cos 9) o e sin6 240030 Geos . (0<9s2) sin 4G _ 44.005 0-08 cornea @ sing a Nore. Concerning these inequalities and their generalizations see: The Mathematical Gazette, vol. 44, 1960, p. 52—54, 5.13 Prove the incquality VAtVB+VvC 43, where A= tan ftany-+5, B= tany tanats, C = tana tan p+5, a> 08> 0,7>0athty = ba Bld If 0 tanz-+tany. 5.15 Prove that sinO+}sin26>0 (0<6<2), a) sin 0-4-}sin 2044 sin30>0 (0< 9 O and L+cosd>0 (0<0< (1) follows immediately, In order to prove (2), we consider the identity sin 0-43 sin 20444 sin 30 = sin 0-4sin 0 cos 0-+4 sin 0 (4cost0—1)108 ELEMENTARY INEQUALITIES 4 ssin 0(2+8 cos 6+4 cos? 0) = }sin O{(1-+c0s 6)*-+(1+c05 6)-+3 cost 9}, from which (2) follows at once. ‘The inequalities (1) and (2) are particular cases of Jackson's inequality (see § 0.8). 5.16 Prove that sin 6+sin 26+... +sin n6+43 sin(n+1)6 = 0 (05952). (1) Poor. The left-hand side of (1) may be written in the form 4 (3 sina sin ws). e) Since S sin ko 1 = sin O cot 30. ‘This expression is non-negative for all values of G€ (0,2) Finally, (1) is trivially true if @ = 0 or =. 5.A7 Prove that arccos# = VI—a! (-1Se@<+41).* **) This assumes the principal-value branch of the inverse function,‘TRIGONOMETRIC FUNCTIONS 109 Proor, For # = 0, #2 sint = Vi—cost, Hae [—1, +1], we may set ¢= are cos x i.e, 2 = cost. Conse- quently, mea a Viel (21s cee): 8.18 Prove that arcsing <2+Vz (0<2<1). Proor. If we set « = sin ¢ (0 <¢ S a/2), the given inequality becomes sint < Vsing, a) By Taylor's formula, a — Zoot (0<9<1), tsint = “cos 64 < & sin ¢ = 08 OF < The inequality (1) follows from the inequality Pe qc Vsnt, B2e (0<2< 2/2) 5.23 1B cose = 1—$a, 5.24 tang >ztde (0<0< 2/2). 5.25 arctane 0) 5.27 —logcose1 (0<2< 2/2). 5.34 tang a+ +A, (2) The inequality (2) is equivalent to 1 We sine > — (a+), @) We i where a is the angle opposite the side @. By the law of cosines, 2be cos a = Beat, ) Squaring (3) and (4) and adding, we obtain the inequality BULL Cat > at tothe, which is equivalent to (2+ (2c) + tet)? < 0, which is false From this result it follows that the inequality (1) is valid. 64 — Let Pbe any point in the interior of the triangle A BC and let |PA| = a, |PB| = y, |PC| =z. Let pg, r be the distances of P from the sides BC, CA, AB, respectively. Prove the inequality aye (N+P) +0) 65 With the notation of 6.4, prove that wtyte = p-+g+r). Soxvtioy, If L, M, N are the feet of the perpendicular from P to BC, CA, AB, respectively, we have MN = (¢+r-+297 cos a)}, MN = sina,