Inequality 3

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Inequalities (Part III) (Note: When a = b = c = 1/3, there is equality.

A simple sketch of f(x) on [0,1]


shows the curve is below the tangent line at x = 1/3, which has the equation
1. Tangent Line Method 12 x 4 a 2 2a 1 12a 4
y . ) So we claim that for 0 < a < 1.
3 3a 2 2a 1 3
Below we will give some examples, where finding the equation of a
tangent line is the critical step to solving the problems. Multiplying out, we see this is equivalent to 36a3 15a2 2a+1 0 for 0 < a
< 1. (Note: Since the curve and the line intersect at a = 1/3, we expect 3a1 is a
Example 1. Let a,b,c,d >0 and a+b+c+d = 1. Prove that factor.) Indeed, 36a3 15a2 2a+1 = (3a1)2(4a+1) 0 for 0 < a < 1. Finally
adding the similar inequalities for b and c, we get the desired inequality.
6(a3+b3+c3+d3) (a2+b2+c2+d2) + 1/8.
Example 3. (1997 Japanese Math Olympiad) Let a,b,c > 0. Prove that
Solution. We have 0 < a,b,c,d < 1. Let f(x) = 6x3x2. Then we have to prove
f(a)+f(b)+f(c)+f(d) 1/8 (Thoughts: Since there is equality when a = b = c = d = (b c a) 2 (c a b ) 2 (a b c) 2 3
.
1/4, we consider the graph of f(x) and its tangent line at x = 1/4. By a simple (b c) 2 a 2 (c a) 2 b 2 (a b) 2 c 2 5
sketch, it seems the tangent line is below the graph of f(x) on the interval (0,1).
By calculus, the equation of the tangent line at x = 1/4 is y = (5x1)/8.) Solution. (Due to Titu Andreescu and Gabriel Dospinescu) As in the last
example, we may assume 0 < a,b,c < 1 and a+b+c = 1. Then the first term on the
We claim that for 0 < x < 1, f(x) = 6x3x2 (5x1)/8. This is equivalent to (1 2a) 2 2
48x 8x25x+1 0. (Thought: Since the graphs intersect at x = 1/4, we expect
3
left become 2 .
(1 a) a
2 2
1 (1 2a) 2
4x1 is a factor.) Indeed, 48x38x25x+1 = (4x1)2(3x+1) 0 for 0 < x < 1. So
the claim is true. Let x1 =12a, x2 =12b, x3 =12c, then x1 +x2 +x3 =1, but 1 < x1, x2, x3 < 1.
In terms of x1, x2, x3, the desired inequality is
Therefore,
1 1 1 27
f(a)+f(b)+f(c)+f(d) 5(a+b+c+d)/84/8 = 1/8. .
1 x1 1 x2 1 x3 10
2 2 2

Example 2. (2003 USA Math Olympiad) Let a,b,c > 0. Prove that (Note: As in the last example, we consider the equation of the tangent line to f(x)
= 1/(1+x2) at x = 1/3, which is y = 27(x+2)/50.)
(2a b c)2 (2b c a)2 (2c a b)2
8. So we claim that f(x) 27(x+2)/50 for 1 < x < 1. This is equivalent to
2a2 (b c)2 2b2 (c a)2 2c2 (a b)2
(3x1)2(43x) 0. Hence the claim is true for 1 < x < 1. Then f(x1)+f(x2)+f(x3)
27/10 and the desired inequality follows.
Solution. Setting a' = a/(a+b+c), b' = b/(a+b+c), c' = c/(a+b+c) if necessary, we
may assume 0 < a,b,c < 1 and a+b+c = 1. Then the first term on the left side of Exercise. (2008 CWMO) Let a,b,c be real numbers with a+b+c = 3. Prove that
the inequality is equal to
( a 1) 2 a 2 2a 1 1 1 1 1
2 2 .
f (a) 2 . 5a 4a 11 5b 4b 11 5c 4c 11 4
2
2a (1 a ) 2 3a 2 2a 1
By expanding both sides and rearranging terms, each of the following
2. Schurs Inequality inequalities is equivalent to the r = 1 case of Schurs inequality. These are
common disguises.
Sometimes in proving an inequality, we do not see any easy way. It will be
good to know some brute force methods in such situation. Below we will a) x3+y3+z3+3xyz xy(x+y)+yz(y+z)+zx(z+x),
introduce a simple inequality that turns out to be useful in proving inequalities by b) xyz (x+yz)(y+zx)(z+xy),
brute force.
c) 4(x+y+z)(xy+yz+zx) (x+y+z)3+9xyz.
Schurs Inequality. For any x, y, z 0 and real r,
Example 1. (2000 IMO) Let a, b, c > 0 and abc = 1. Prove that
xr(xy)(xz) + yr(yx)(yz) + zr(zx)(zy) 0. 1 1 1
( a 1 )(b 1 )( c 1 ) 1.
Equality holds if and only if x = y = z or two of x, y, z are equal and the third is 0. b c a
Solution. Let x = a, y = 1, z = 1/b = ac. Then a = x/y, b = y/z and c = z/x.
Proof. Observe that the inequality is symmetric in x, y, z. So without loss of Substituting these into the desired inequality, we get
generality, we may assume x y z. Then a = xy 0 and b = yz 0. Now the
left hand side of the inequality can be written as ( x y z ) ( y z x) ( z x y )
1,
y z x
xra(a+b) yrab + zrb(a+b) = xra2 + (xryr+zr)ab + zrb2.
If r 0, then xr yr. If r < 0, then zr yr. So xryr+zr 0 and Schurs inequality which is disguise b) of the r = 1 case of Schurs inequality.
follows. Now a > 0 implies x > y 0 and xra2 > 0. So, equality holds in Schurs
inequality if and only if a = 0 and zrb2 = 0, that is x = y and either y = z or z = 0. Example 2. (1984 IMO) Prove that
0 yz+zx+xy2xyz 7/27,
In using the Schurs inequality, we often expand out expressions. So to
where x, y, z are nonnegative real numbers such that x+y+z = 1.
simplify writing, we introduce the symmetric sum notation f(x,y,z) to denote
sym

f(x,y,z)+ f(x,z,y)+ f(y,z,x)+ f(y,x,z)+ f(z,x,y)+ f(z,y,x). Solution. In Schurs inequality, all terms are of the same degree. So we first
change the desired inequality to one where all terms are of the same degree.
For example, x3 = 2x3 +2y3+2z3, x2y= x2y+x2z+y2z+y2x+z2x+z2y and xyz =
sym sym sym Since x+y+z = 1, the desired inequality is the same as
6xyz. Similarly, for a function of n variables, the symmetric sum is the sum of all 7( x y z )3
0 ( x y z )( yz zx xy ) 2 xyz .
n! terms, where we take all possible permutations of the n variables. 27
Expanding the middle expression, we get xyz + x2y, which is nonnegative
After expanding the r = 1 case of Schurs inequality, we get sym

3 3 3 2 2 2 2 2
x +y +z (x y+x z+y x+y z+z x+z y)+3xyz 0.2 and the left inequality is proved. Expanding the rightmost expression and
subtracting the middle expression, we get
In symmetric sum notation, it is (x
sym
3
2 x 2 y xyz ) 0.
7 12 5

54 sym
( x 3 x 2 y xyz ).
7 7
(1) Example 4. (2000 USA Team Selection Test) Prove that for any positive real
By Schurs inequality, we have numbers a, b, c, the following inequality holds
abc 3
(x 3
2 x 2 y xyz) 0. (2) abc max{( a b ) 2 , ( b c ) 2 , ( c a ) 2 }.
sym 3
12 2 5 2 2
Note ( x x y xyz) ( x 2 x y xyz ) ( x y xyz ). By the AM-GM
3 3 2
Solution. From the last part of the solution of example 3, we get
7 7 7
3(xyz)2/3 2(xy+yz+zx) (x2+y2+z2)
inequality, we have , which is the same as
2 6 6 6 1/ 6
x y 6 ( x y z ) xyz
sym sym
for any x,y,z > 0. (Note: this used Schurs inequality.) Let x a, y b
(x 2
y xyz) 0. (3)
sym and z c . Arranging terms, we get
Multiplying (3) by 2/7 and adding it to (2), we see the symmetric sum in (1) is a b c 3 3 abc 2( a b c ab bc ca )
nonnegative. This yields the inequality on the right.
( a b )2 ( b c )2 ( c a )2
Example 3. (2004 APMO) Prove that 3 max{( a b ) 2 , ( b c ) 2 , ( c a ) 2 }.
(a 2 2)(b2 2)(c 2 2) 9(ab bc ca) Dividing by 3, we get the desired inequality.
for any positive real numbers a,b,c.
Example 5. (2003 USA Team Selection Test) Let a,b,c be real numbers in the
Solution. Expanding and expressing in symmetric sum notation, the desired interval (0, /2). Prove that
inequality is(abc)2+ (a2b2+2a2)+8 9 ab. sin a sin(a b) sin(a c) sin b sin(b c) sin(b a) sin c sin(c a) sin(c b)
2 0.
sin(b c) sin(c a)
sym sym
sin(a b)
Since a2+b22ab, we get a2 ab. Also, from a2b2 + 1 2ab, we get
Solution. Observe that sin(uv)sin(u+v) = (cos 2vcos 2u)/2 = sin2 u sin2v. Let
sym sym
2 2
a b + 6 2 ab. Using these, the problem is reduced to showing
sym sym x = sin2 a, y = sin2 b, z = sin2 c. In adding up the terms, the left side of the
(abc)2 + 2 (ab 1 a2). inequality becomes
sym 2
To prove this, we apply the AM-GM inequality twice and disguise c) of the r = 1 x ( x y)(x z) y ( y z)( y x) z ( z x)(z y)
.
case of Schurs inequality as follow: sin(b c) sin(c a) sin(a b)

(abc)2 +2 3(abc)2/3 9abc/(a+b+c) This is nonnegative by the r = 1/2 case of Schurs inequality.
4(ab+bc+ca) (a+b+c)2
= 2(ab+bc+ca)(a2+b2+c2) = (ab 1 a2).
sym 2
3. Muirheads Inequality For the next example, we would like to point out a useful trick. When the
product of x1, x2, , xn is 1, we have
Muirheads inequality is an important generalization of the AM-GM
[(p1, p2, , pn)] = [(p1r, p2r,, pnr)]
inequality. It is a powerful tool for solving inequality problem. First we give a
definition which is a generalization of arithmetic and geometric means. for any real number r.

Example 2. (1995 IMO) For any a, b, c > 0 with abc = 1, prove that
Definition. Let x1, x2, , xn be positive real numbers and p = (p1, p2, , pn) n.
1 1 1 3
The p-mean of x1, x2, , xn is defined by 3 3 .
a (b c) b (c a) c (a b) 2
3

1
[ p]
n! S n
xp1(1) xp(2 2 ) xp(n n ) , Solution. Multiplying by the common denominator and expanding both sides,
the desired inequality is
where Sn is the set of all permutations of {1,2,, n}. (The summation sign
means to sum n! terms, one term for each permutation in Sn.) 2(a 4b4 b4c4 c 4a 4 ) 2(a 4b3c a 4c3b b4c3a b4 a3c c 4a3b c 4b3a)
2(a3b3c 2 b3c3a 2 c3a3b2 )
n
1 1
For example, [(1,0, ,0)] n! x (1) n xi is the arithmetic mean of x1, x2, , 3(a 5b 4 c 3 a 5 c 4 b 3 b 5 c 4 a 3 b 5 a 4 c 3 c 5 a 4b3 c 5b 4 a 3 ) 6a 4b 4 c 4 .
Sn i 1

1 1 1 This is equivalent to [(4,4,0)]+2[(4,3,1)] + [(3,3,2)] 3[(5,4,3)] + [(4,4,4)]. Note


xn and , , , x1 x 2 x n is their geometric mean.
1/ n 1/ n 1/ n
4+4+0 = 4+3+1 = 3+3+2 = 8, but 5+4+3 = 4+4+4 = 12. So we can set r = 4/3
n n n and use the trick above to get [(5,4,3)] = [(11/3,8/3,5/3)] and also [(4,4,4)] =
[(8/3,8/3,8/3)].
Theorem (Muirheads Inequality). Let x1, x2, , xn be positive real numbers and
p, q n. If p q, then [p] [q]. Furthermore, for p q, equality holds if and Observe that (4,4,0) (11/3,8/3,5/3), (4,3,1) (11/3,8/3,5/3) and (3,3,2)
(8/3,8/3,8/3). So applying Muirheads inequality to these three majorizations and
only if x1= x2 = = xn. adding the inequalities, we get the desired inequality.
Since (1,0,,0)(1/n,1/n,,1/n), the AM-GM inequality is a corollary of Example 3. (1990 IMO Shortlisted Problem) For any x, y, z > 0 with xyz = 1,
Muirheads inequality. prove that
x3 y3 z3 3
Example 1. For any a, b, c > 0, prove that .
(1 y)(1 z) (1 z)(1 x) (1 x)(1 y) 4
(a+b)(b+c)(c+a) 8abc.
Solution. Multiplying by the common denominator and expanding both sides,
Solution. Expanding both sides, the desired inequality is the desired inequality is
a2b+a2c+b2c+b2a+c2a+c2b 6abc 4(x4+y4+z4+x3+y3+z3) 3(1+x+y+z+xy+yz+zx+xyz)
or [(2,1,0)] [(1,1,1)]. This is Muirheads inequality for (2,1,0)(1,1,1). or 4[(4,0,0)] + 4[(3,0,0)] [(0,0,0)] + 3[(1,0,0)] + 3[(1,1,0)] + [(1,1,1)].
For this, we apply Muirheads inequality and the trick as follow: where (1) and (3) are by Muirheads inequality and the remark, (2) is by
Muirheads inequality, (4) is by the fact, Muirheads inequality and the remark
[(4,0,0)] [(4/3,4/3,4/3)] = [(0,0,0)], and (5) is by the remark.
3[(4,0,0)] 3[(2,1,1)] = 3[(1,0,0)],
3[(3,0,0)] 3[(4/3,4/3,1/3)] = 3[(1,1,0)] Considering the sum of the leftmost parts of these inequalities is greater
and [(3,0,0)] [(1,1,1)] . than or equal to the sum of the rightmost parts of these inequalities, we get the
Adding these, we get the desired inequality. desired inequalities.

Alternate Solution. Since


Remark. For the following example, we will modify the trick above. In case xyz
1, we have [(p1, p2, p3)] [(p1r, p2r, p3r)] for every r 0. Also, we will use
x5 x 2 x5 x 2 ( x 3 1) 2 ( y 2 z 2 )
the following 3 2 0,
x y z
5 2 2
x ( x y z ) x( x y 2 z 2 )( x5 y 2 z 2 )
2 2 2

[ p] [q ] p q we have
Fact. For p, q n, we have . (This is due to the AM-GM
2 2 x5 x 2

y5 y 2

z5 z2 x5 x2 y5 y2 z5 z 2
2 3 2 3 2 2 2 3 2 2 2
x p1(1) x p (n n ) x q1(1) x q n( n ) x y z
5 2 2
y z x
5 2 2
z x y
5 2
x (x y z ) y ( y z x ) z (z x y )
2 2

inequality x ( p(11 ) q1 ) / 2 x ( p( nn ) q n ) / 2 . Summing over 1 1 1 1


2 (x2 y2 z 2 )
x2 y 2 z 2 x y z
Sn and dividing by n!, we get the inequality.)
1
( x 2 y 2 z 2 yz zx xy )
x2 y2 z2
Example 4. (2005 IMO) For any x, y, z > 0 with xyz 1, prove that ( x y ) 2 ( y z ) 2 ( z x) 2
0.
2( x 2 y 2 z 2 )
x5 x2 y5 y 2 z5 z 2
0.
x5 y2 z2 y5 z 2 x2 z5 x2 y2
Solution. Multiplying by the common denominator and expanding both sides, Proofs of Muirheads Inequality
the desired inequality is equivalent to [(9,0,0)]+4[(7,5,0)]+[(5,2,2)]+[(5,5,5)]
[(6,0,0)] + [(5,5,2)] + 2[(5,4,0)] +2[(4,2,0)] + [(2,2,2)]. Let p q and p q. From i = 1 to n, the first nonzero pi qi is positive. So
there will be a negative pi qi later. It follows that there are j < k such that pj > qj,
To prove this, we note that
pk < qk and pi = qi for any possible i between j, k.
(1) [(9,0,0)] [(7,1,1)] [(6,0,0)]
Let b = (pj+pk)/2, d = (pjpk)/2 so that [bd,b+d] = [pk, pj] [qk, qj]. Let c
(2) [(7,5,0)] [(5,5,2)] be the maximum of |qjb| and |qkb|, then 0 d < c. Let r = (r1,,rn) be defined
(3) 2[(7,5,0)] 2[(6,5,1)] 2[(5,4,0)] by ri = pi except rj = b + c and rk = b c. By the definition of c, either rj = qj or
(4) [(7,5,0)] + [(5,2,2)] 2[(6,7/2,1)] 2[(9/2,2,1/2)] 2[(4,2,0)] rk=qk. Also, by the definitions of b, c, d, we get p r, p r and r q. Now
(5) [(5,5,5)] [(2,2,2)],
x ( x x (u
pj r
n!([ p] [r ]) x pk
( j ) (k ) xj( j ) xrk( k ) ) b d
v b d u b c v b c ),
S n S n

where x is the product of x p ( i ) for i j, k and u = x(j) , v = x(k). For each


i

permutation , there is a permutation such that (i) = (i) for i j, k and (j) =
(k), (k) = (j). In the above sum, if we pair the terms for and , then x = x
and combining the parenthetical factors for the and terms, we have
(ub+d vbd ub+c vbc)+(vb+d ubd vb+c ubc)= ubd vbd (ud+c vd+c) (udc vdc) 0.
So the above sum is nonnegative. Then [p] [r].

Equality holds if and only if u = v for all pairs of and , which means x1=
x2 = = xn. Finally we recall r has at least one more coordinate in agreement with
q than p. So repeating this process finitely many times, we will eventually get the
case r = q. Then we are done.

You might also like