Scribd Logo
Upload

Welcome to Scribd!

  • Geometric Inequalities With Polynomial 2

    Uploaded by

    Fustei Bogdan
    57 views14 pages

    Original Title

    Geometric Inequalities With Polynomial 2 (1)

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    57 views14 pages

    Geometric Inequalities With Polynomial 2

    Uploaded by

    Fustei Bogdan

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 14
    OCTOGON MATHEMATICAL MAGAZINE Vol. 22, No.2, October 2014, pp 728-741 Print: ISSN 1222-5657, Online: ISSN 2248-1893 http://www-uni-miskole.hu/~matsefi/Octogon/ 728 Geometric Inequalities with polynomial axy + 2yz + 2zx — x?—y?—z? Arkady Alt 17 ABSTRACT. This paper’s aim is to explore the usage of symmetric polynomial A (2,y, 2) = Qay + 2yz + Qer — 2? —y? — in various geometric inequalities related to triangle.In particular we will show how A (a,b,c) ,where a,b,c define a triangle, can be used along side of R,r,s to give a new interpretation ((A,r,s)- form) of Hadwiger-Finsler, Blundon’s and many others well known and new inequalities. Also we obtain the best quadratic (R, r)-minorant for A(a,b,c) and linear (s,r)-majorant for sum of medians. 1 INTRODUCTION: NOTATIONS AND BASIC CORRELATIONS Symmetric polynomial A (a, y, 2) := Qzy + Qyz + 22a — 2? —y? — 2? is not a positive definite quadratic form , really: trq-r\_, A(r+q tts )-@-2-8 2 And even requiring x,y, z > 0 doesn’t guarantee positivity of A (x,y, z). But A(x,y,2) acquires a special meaning for positive x,y, z since in this case inequality A(e,y, 2) > 0 is equivalent to triangle inequalities for numbers 2, J, V2, that is A(e,y,z) >0 => Vt+ Ju> Vz, fut Vz> Va,VetVE> Vu Received: 30.08.2014 2010 Mathematics Subject Classification. 26D15, 51M16. Key words and phrases. Geometric Inequality, Medians, Blundon Inequality, Hadwiger-Finsler Inequality. Geometric Inequalities with polynomial 2cy + 2yz + 22a — 22-y2 It is interesting to note that for positive numbers a, b,c, inequality A (a?, b,c?) > 0 characterizes a, b,c as side-lengths of a triangle with area ) al and inequality A (a4,b4,c') > 0 characterizes an acute triangle with side-lengths a,b,c. If A (a",b",c") > 0 for all natural n then a,b,c represent side-lengths of an isosceles triangle with the lateral side not less than the base [7]. Let a,b,c be side-lengths of a triangle AABC and let F,s, R, and r be, respectively, area, semiperimeter, circumradius, and inradius of AABC. Also, let rz be exradius corresponding to a side x € {a,b,c}. We shall add A =A (a,b,c) = 2ab + 2ac + 2be — a? to this list of triangle characteristics. Using these notations we can write down more representations for A = A (a,b, i A=E (#- 0-07) = 25 0( -0) = 45 (s-a)(0-9 = 4s° — 2 (a? +b? +c?) = 4 (ab+ be + ca) — 48? 5 A ii Since ab+be+ca=s?+r(4R +r), ora =4R +7, 2% =tanZ, we have cye 8 2 A=4r(4R+r) =4r Dor =4F tan 4. eve Ed Should also be noted that A (a?, b?,c?) = 16F? = 16r2s?. There are many ways to define a triangle. In particular, the most common way is to define a triangle as a triplet (a,b,c) of positive real numbers that satisfy the Triangle Inequalities: (TI) a+b>e,b+ce>a,c+a>b(or a,be< s) Triangle,defined in such a way will be denoted by T (a,b,c). Let T=s—ay=s—bz=s—cthena=y+2z,b=z+2,c=r+y, where ,y,2 > 0.Thus, any three positive numbers a, y, z determine a triangle T(y +2,z+2,2+y) and we will call such a representation of a triangle T (a,b,c) a Sree parametrization, because the numbers x,y,z do not depend on each other. In that case A = A (a,b,c) =A(yt+z,z+2,2+y) = 4 (cy +yz+z2). Let F(R,r,s) = 4R(R — 2r)’ — (s? ~ 2R? — 10Rr +r)”. Note note that three positive numbers R,r,s define a triangle with circumradius R, inradius r and semiperimeter s if and only if the Fundamental Geometric Inequality (FGI) F(R,r,s) > 0 holds ([1,p.4,inequality(12) ] or (2,p.54,Theorem 2 ]}). This inequality is most commonly used in the form . 2R? + 10Rr —r? — 2(R—2r) /R(R—2r <8? < S 2K? + 10Rr — r? + 2(R- 2r) /R(R—2r). 730 Arkady Alt A 4r? Since A= 4r(4R+r) =} R= then three positive numbers A,r and s.determine some triangle T (a,b,c) with inradius r, semiperimeter s and A= A (a,b,c) if and only if “e (FA. :) 20s 72r2s2A + 2A? — AS > 16r2s? (27r? + 452). ; In general, any inequality in the form G(R,r, s) > 0 is equivalent to its : A-4 (A,r, s)-form, obtained by replacing R. with ere namely to inequality _ 2 eee, ns) >0. o( 16r ~ For example, (A,r, s)-form of inequalities 8r(4R+7r) 4V3F 4V3rs< A <> sV3<4R4r. Since 16Rr = A — 4r? then inequality 16Rr — 5r? < s? (Gerretsen) can be rewritten in (A,r, s)-form A A ayz(e+y+z) <(cytyz+22)? <> Le w-z? 20; exe 9% (DG) <=> 4 (ay + yz +22) <(e@ ty +2)? + ——— = Alayy,z) < rtytz (t1,2) s Sayz —A(a+y +z) (ay t+yzt 2a) +(a+y+2)° 20 Ya(x—y) (x - z) > O(Schure Inequality). ye Geometric Inequalities with polynomial 2zy + 2yz + 22a —2?—y?—z? 731 7 AND 7r~!— TRANSFORMATIONS Here we will consider two triangle transformations where A plays an important role and which will allow us to obtain new geometric inequalities and establish equivalence of several well-known geometric inequalities. 2.1. 7 transformation. Let a, = a(s—a),b, = 6(s ~b),c, =¢(s—c). Numbers a,,b,,c, are positive and satisfy the triangle inequalities, and therefore determine a triangle T (a7,br,¢r) Indeed, b; +¢r ay = b(s — 6) -+e(s —c)—a(s—a) = 8(b+c—a)—b2— b+)? —a? +20? -2(P +2 2 — (6-0)? (b+e) = Gre) P= O-O _9(5_8)(s—0) and cyclically we have c, +a, ~b, = 2(s~c)(s—a) and a, +b, —c; =2(s—a)(s—6). Let s,,F,,R-,ry be semiperimeter, area, circumradius and inradius of the triangle sa b(s— T (ar, Be, ¢r). Then s, = 22— 9) +5(s= 8) +e(s— 2) C+P= =] and since 2 by — a, = tor 8 _ (5 _ 4) (5—c) and the cyclic s, ~b; =(s —c)(s— a), 5c, = (s —a)(s— 5) we obtain s 3A? +36Ar? < 4A? <> 36r2 4F? < 6 +9r?-—- <> a or2\ A? A 2(,-—-)<— os P< > __. 2 (0 T) So YS Gao) Remark 2. Using free parametrization (a,b,c) = (y+ z,z+2,2+y) we can 3 rewrite inequality F? < in the form }> y?z? (x — y) (x — z) > 0.The eye * G4(A — 972) 734 Arkady Alt latter inequality can be obtained from Schure Inequality 2 x(a — y) (a — z) > 0 by Food replacing (z,y,z) with (yz, z«,21y). 3 (R,r)- majorants, minorants. 3.1. The family of (R,r)-linear majorant for semiperimeter s First we will prove Lemma 1. Let 4 and v be non-negative real numbers. Inequality 2R? + 10Rr —r? +2(R—2r)/R(R—2r) < (uR+ur)? with equality condition of R = 2r holds if and only if 2 2 and t (ut + vy" 5 viel 20? + 10t-1+2( (#2) ve t ac oe woe equality occurs if t= 2. Then 1 < 2 a 1< ii (ut +v) ero =Ites =O ACES too 242 + L0t—1 42 (t 2) VEE—2 For t = 2 we have 2-2? + 10-2-1=(p- Dan)? <> Wa (rv)? 3 Qty =3V3 <> v =3V3 —2y and, therefore, 3/3 —2n >0 => p< Ye ues Sufficiency. Let y € 2, wf) Since uR-+ (3V3 — 2) r = (BR ~2r) + 3V3r isn’t decreasing in ys and 2R? + 10Rr —r? + 2(R — 2r) \/R(R — 2r) <4R? + 4Rr + 3r? then suffice to prove 4R? + 4Rr + 3r? < (2Ru + (3V3—2n) nr)” for p= We have (2R + (3V3 — 4) r)” — (48? + 4Rr + 3r?) = 4r (3V3 — 5) (R-2r) > 0. From s? < 2R? + 10Rr — 1? + 2(R—2r) /R(R — 2r) and the lemma above it follows: Corollary 1. For any 2< p< eve holds inequality s < pR + (3V3 — 2u) r. Corollary 2. For any 2

    0. Geometric Inequalities with polynomial 2ey + 2yz + 2er — 2?—y? 735 So, Blundon’s Inequality s < 2R + (33 — 4) r (that corresponds to pz = 2) gives the best (R,r) —linear majorant for s. 4 For w= 4. we obtain inequality s< “LR + (ov -2 V3 sqlan +r) <= > V3s<4R+r = 43F lors < pA+ (48V3 - 36) r?2 <=> 12 (43 — 3) Hu s (HF) 5, 22v3 9 16r (s — 3V3i Since s > 3V3r then ars — Br? = 6s (seaisv.30)) ++ 36r? is decreasing in In particular, if pp = 2, rs+4r2 p= a = 16(2V3-3), = 1 (ue-22) <25-2aseovi- 3) r = 25—3 (2V3 — 3) rand, . For we have 8 therefore, Ma + my + me < 2s -3 (2v3-3) r. 3 Thus, only p= ee provide the linear (s,r) —majorant for sum of medians. Inequality (M), when it already established, can be proven by a shorter way: Since ab + be + ca = 8? +1 (4R +r) and a? + 6? +c? = 2(s?—r(4R +1) then T(a? +B +c?) +2 (ab + be + ca) (tte +m +m)? pg UEP re) + lob bet on) 14(s? —r(4R +r) + 2(s? $r(4R +r) ee —r = 5? suffices to prove 4s? — 3r? — 12Rr < (2s ~ 3 (2V3—3)r)”. Since s << Wp we have(2s — 3 (2V3 — 3) r)” — (4s? — 3? — 19Rr) = 12r (R - s (2V3 - 3) —r (9V3 — 16)) > or(n- rc v3 -3)—r (oV5~16)) = 6(0V8—19)r(R-20) 20 r? — 12Rr and, therefore, it (Inequality (M) as a conjecture was proposed by Konstantin Knop in private communication). 738 Arkady Alt 4 More inequalities with A In conclusion, we will consider a few inequalities with A. First we will present a chain of inequalities with A (some of them are already well known). Inequality 1. Let a,b,c be lengths of sides of a triangle. Then 9a?b?c? Ee 3abe (a +b +c) éAc< 8abe (ab + be + ca) ab? + bc? + 2a? ~ a +P +2 ~~ (a rb)(b+a (cra) ~ oat c myin { Babee + +0) a ap). at+tb+e ab +be+ca Proof. Since a,b,c are positive, we have we <3VEOE > 3Yabe 3(ab+be+ ca) < (a+b+c)? anc Babe (ab + be-+ ea) Sabe (a+b)(6+e)(c+a) satbte 9(a +8) (b+) (c+) (as a side note: all of these inequalities hold for any positive a, 6,¢). So it remains to prove: <= 8(a+b+c)(ab+be+ca) < Babe(a +b +c) 8abe (ab + be + ca) a+ P+ aes s°S GH O+o(c+a) 907620? < Sabe(a+b+e) OP +POF CR ~ 8 +P EA Babe (a? +b? +0?) < (a+b +c) (070? + bc? + 2a?) . Proof. (Inequalities (1) and (2). Using free parametrization of the triangle namely,(a, b,c) = (y+z,z+2,¢+y), denoting p := ry + yz + 22,q:= yz, and due to homogeneity of the inequalities, assuming x + y + 2 = 1,we obtain s = 1,A = 4p,abe = p—q,a? +b? +c? = 2(1 — p) 0262 +.2e2 +202 = (1 =p)” +49, (a +6) (b +c) (¢+a) = 2+ p-+q and inequalities (1) and (2) becomes, respectively, 3(p-4) < 4p < 8@-9CFP) Lg l-p “2+p+q 3(p—4)(1—p) S$ (1—p)? +4q => (7-3p)q- (4p —1)(1-p) D0. First, lets prove inequality 3(p-4) . oe ee i-p < i S<4p <=> 3q¢+p—4p? > 0. Since p = ry+yz+ Geometric Inequalities with polynomial 2zy + 2yz + 2za —2?-y?—z? 739 4p — (et+yt+2)? 3 then 1 = gand > x(x-y)(t-z) >0 = q> 3 cyclic } sor O p(1— 4p) > 0 and fort

    a Now, lets prove 8(p—g) (1 2(p—a) (1 (p g) (+p) = px 2@-DAty) _, P 2+p+q 2+pt+aq (cyt yz+ 2a) _ p? =5 then inequality 4p < (3p +2) q > 0.Since q = zyz (x+y +2) < 2 3 9 1 P?— (8p+2)q>p? ~ (3p +2) = Sp? (1 3p) 20. Thus, it remains to prove inequality (7 — 3p) q — (4p — 1) (1 — p) > 0. - p) (4p — Note that q > max {o, Conte» Since See (cea) (ee 6 eyelic 0S q> GD =) poo cp (1 4p) (1 ~p) 2 0 and for 7

    (7 -3p) Qo p)ip—) ~(4p-1)(1—p) = (1 =p) (1 ~ 3p) (4p — 1) >0 Remark 4. In reality, inequality A < 3\Va?b2c? holds for any real a,b,c and it is A= form (up to replacement (a,b,c) with (x, 3, 29) ) of well known [5],[6] inequality (x3 + y + 23)? +3 (wy2)? > 4 (0%y9 + y923 + 2929), where 2,y,2 € R. Inequality 2.Let a,b,c be lengths of sides of a triangle. Then for any real positive z,y,2 holds inequality: ri xbe 4 yee. abe (a) yte zta arty 2 we () 22 ze ytz ty Remark 5. Inequality (a) is a geometric version of algebraic inequality, proved by M.S.Klamkin for any positive a,b,c,x,y,z and presented as Inequality 1 in [1], p.33 without proof and with reference to original article. Inequality in (b) is also a geometric version of algebraic 740 Arkady Alt inequality proved by D.S.Mitrinovic,J.E. Pecaric for any positive a,b,c and real x,y,z such that z+y,y+z,z+2>0 and presented as Inequalities 6 and 10 in [1], p.34 with easy proof. z So, we will prove only the inequality (a) in the form P ly quality (a) = ayaa Proof. Applying Cauchy Inequality to triples ( ——2__ y 2 and ~ \ Vary +2)’ Joe +a)' Vez(e+y) (var (+2), Jon(e+ a), Jez(et )) we obtain > 2 2 2 x (e@+y+z) azr(y+z)- ooo (tty tcl eS ——V 2 =: Rot) varyan = eter) — Laary Sages axe — (e+y+z)? 4(ab+be+ca) we 5 _ Ate te tea) | ; Since Gia 2 Gath Te-ray (inequality (D), [4]) and Ex 8 (ab-+ be + ca) abe 2 A > ain J —~_ > 2 2 A we obtain 5 5 2 ate (+0 6+0 (e+a) Inequality 3. Let a,b,and c be lengths of sides of a triangle ABC and P be any point in the triangle, Let da,dy,d- be distances from point P to sides a,b,c respectively. Then 4F? A (a?,b?,c?) djdg <= SS dady + dyde+deda < a ZA (a,b,c) Proof. Since ada + bd + cde = 2F then by replacing (x,y,z) and (a,8,~) in 2 oer (inequality (C), [4] ) with (ada, bd, cd.) and (a,>,¢), respectively, we obtain Remark 6. This inequality was proven in [4],p-460 as inequality (DP) and originally represented as maximization problem and since Mody + dyile + dedy = A (da + dy, dy + de,de + dq) it can be rewritten as A (da + dy, ly + do, de + da ) A (a,0,0)

    You might also like