AMSP 2010 Algebra 3.

5: Day 10 Problems
Inequalities
Rearrangement and Chebyshev Inequalities
Theorem 1 (Rearrangement). Let x
1
, x
2
, . . . , x
n
and y
1
, y
2
, . . . , y
n
be real numbers (not
necessarily positive) with
x
1
x
2
x
n
, and y
1
y
1
y
n
,
and let be a permutation of 1, 2, . . . , n. Then the following inequality holds:
x
1
y
n
+ x
2
y
n1
+ + x
n
y
1
x
1
y
1
+ x
2
y
2
+ + x
n
y
n
x
1
y
1
+ x
2
y
2
+ + x
n
y
n
.
Proof. We prove the inequality on the right by induction on n. The statement is obvious
for n = 1. Suppose it true for n 1. Let m be an integer such that m = n. Since
x
n
x
m
and y
n
y
n
,
(x
n
x
m
)(y
n
y
n
) 0 =
x
m
y
n
+ x
n
y
n
x
m
y
n
+ x
n
y
n
.
Hence
x
1
y
1
+ + x
m
y
m
+ + x
n
y
n
x
1
y
1
+ + x
m
y
n
+ + x
n
y
n
.
By the induction hypothesis,
x
1
y
1
+ + x
m
y
n
+ + x
n1
y
(n1)
x
1
y
1
+ + x
m
y
m
+ + x
n1
y
n1
.
Combining these two we get the desired inequality.
To prove the LHS, apply the above with y
i
instead of y
i
.
Theorem 2 (Chebyshev). Let a
1
a
2
a
n
and b
1
b
2
b
n
be two similarly
sorted sequences. Then
a
1
b
n
+ a
2
b
n1
+ + a
n
b
1
n

a
1
+ a
2
+ + a
n
n

b
1
+ b
2
+ + b
n
n

a
1
b
1
+ + a
n
b
n
n
Proof. Add up the following inequalities (which hold by the Rearrangement Inequality):
a
1
b
1
+ a
2
b
2
+ + a
n
b
n
a
1
b
1
+ a
2
b
2
+ + a
n
b
n
a
1
b
2
+ a
2
b
3
+ + a
n
b
1
a
1
b
1
+ a
2
b
2
+ + a
n
b
n
.
.
.
a
1
b
n
+ a
2
b
1
+ + a
n
b
n1
a
1
b
1
+ a
2
b
2
+ + a
n
b
n
This gives the right-hand inequality.
By replacing y
i
with y
i
and using the above result we get the left-hand inequality.
Problems
1. Powers: For a, b, c > 0 prove that
1
AMSP 2010 Algebra 3.5: Day 10 Problems
(a) a
a
b
b
c
c
a
b
b
c
c
a
.
(b) a
a
b
b
c
c
(abc)
a+b+c
3
.
2. Prove the following for x, y, z > 0:
(a)
x
2
y
2
+
y
2
z
2
+
z
2
x
2

x
z
+
y
x
+
z
y
.
(b)
xy
z
2
+
yz
x
2
+
zx
y
2

x
y
+
y
z
+
z
x
.
3. (IMO 1978/2) Let a
1
, . . . , a
n
be pairwise distinct positive integers. Show that
a
1
1
2
+
a
2
2
2
+ +
a
n
n
2

1
1
+
1
2
+ +
1
n
.
4. (modied ISL 2006/A4) Prove that for all positive a, b, c,
ab
a + b
+
bc
b + c
+
ac
a + c

3(ab + bc + ca)
2(a + b + c)
.
5. (MOSP 2007) Let k be a positive integer, and let x
1
, x
2
, . . . , x
n
be positive real
numbers. Prove that
_
n

i=1
1
1 + x
i
__
n

i=1
x
i
_

_
n

i=1
x
k+1
i
1 + x
i
__
n

i=1
1
x
k
i
_
.
6. Prove that for any positive real numbers a, b, c the following inequality holds:
a
2
+ bc
b + c
+
b
2
+ ac
c + a
+
c
2
+ ab
a + b
a + b + c.
Convexity, Jensens and Karamatas Inequalities
Denition 1. A function f : I R (where I R is an interval) is said to be convex if
for any t [0, 1] and x, y I, the following inequality holds:
f(tx + (1 t)y) tf(x) + (1 t)f(y).
We say that f(x) is strictly convex if equality holds only when t = 0, 1 or x = y. If the
inequalities are reversed, we say that f(x) is concave or strictly concave.
You may think of convexity as meaning that the line segment joining two points of
the graph of f is always greater than the graph itself. Note that if you know something
is convex, you do not know that it is necessarily increasing or decreasing! However, the
absolute maximum of a convex function (if it exists) never occurs on the interior of the
interval of denition.
We have a number of examples of convex functions:
f(x) = x
r
for r 1 and x 0.
f(x) = log x.
2
AMSP 2010 Algebra 3.5: Day 10 Problems
f(x) = 1/x for x > 0.
f(x) = 1/(x
2
+ 1) for x 1/

2.
We have other characterizations of convex functions:
Theorem 3. A function f(x) is convex if and only if for any x
1
y
1
< x
2
y
2
(where
[x
1
, y
2
] is in the domain of f) we have
f(x
2
) f(x
1
)
x
2
x
1

f(y
2
) f(y
1
)
y
2
y
1
,
and strictly convex if and only if this inequality is always strict. A function f(x) which
is dierentiable everywhere is convex if and only if f

(x) is an nondecreasing function

of x, and strictly convex if and only if f

(x) is increasing. A function f(x) which is

twice dierentiable everywhere is convex if and only if f

(x) 0, and strictly convex if

f

(x) > 0 (though not conversely in general).

It is easy to nd the maximum of a convex function:
Theorem 4. If f is convex, then the maximum value of f(x) on the interval [a, b] is
attained when x = a or when x = b. If f is concave, then the minimum value of f(x) on
the interval [a, b] is attained when x = a or when x = b.
Jensens inequality essentially extends the elementary notion of convexity to any num-
ber of variables:
Theorem 5. Let f(x) be convex on an interval I, let x
1
, x
2
, . . . , x
n
I, and let
1
,
2
, . . . ,
n
be nonnegative real numbers (weights) with
1
+
2
+ +
n
= 1. Then

1
f(x
1
) +
2
f(x
2
) + +
n
f(x
n
) f(
1
x
1
+
2
x
2
+ +
n
x
n
).
If f(x) is strictly convex, then equality holds if and only if
i
= 1 for an i or all the x
i
are equal.
Proof. Induct on n.
Karamatas inequality is a generalization of Jensens.
Denition 2. The sequence x
1
x
2
x
n
majorizes the sequence y
1
y
2

y
n
, denoted
(x
1
, x
2
, . . . , x
n
) ~ (y
1
, y
2
, . . . , y
n
)
if
x
1
y
1
x
1
+ x
2
y
1
+ y
2
.
.
.
.
.
.
x
1
+ x
2
+ + x
n1
y
1
+ y
2
+ + y
n1
x
1
+ x
2
+ + x
n1
+ x
n
= y
1
+ y
2
+ + y
n1
+ y
n
3
AMSP 2010 Algebra 3.5: Day 10 Problems
Theorem 6 (Karamata Majorization). Let f : I R be convex on I and suppose that
(x
1
, . . . , x
n
) ~ (y
1
, . . . , y
n
), where x
i
, y
i
I. Then
f(x
1
) + + f(x
n
) f(y
1
) + + f(y
n
).
The reverse inequality holds if f is concave.
Proof. Let
c
i
=
f(x
i
) f(y
i
)
x
i
y
i
, S
k
=
k

i=1
x
i
, T
i
=
k

i=1
y
i
.
Then
n

i=1
f(x
i
) f(y
i
) =
n

i=1
c
i
(x
i
y
i
)
=
n

i=1
c
i
(S
i
S
i1
T
i
+ T
i1
)
= c
n
(S
n
T
n
)
. .
0
+
n1

i=0
(c
i
c
i+1
)(S
i
T
i
)
The last expression is positive since convexity implies c
i
c
i+1
and the majorization
condition implies S
i
T
i
.
Jensens and Karamatas inequalities are related to the idea of smoothing, which allows
you to make moves like Jensen without necessarily the assumption of convexity. More
precisely, given a desired inequality f(x
1
, x
2
, . . . , x
n
) 0 say, if x
i
and x
j
may both
be replaced by values x

i
and x

j
(often the same) such that f takes on a lesser value
than before, the problem may be reduced to that case (which may for example allow an
induction).
Problems
1. Trig: Let A, B, C be the angles of a triangle. Prove that
(a) sin A + sin B + sin C
3

3
2
.
(b) cos A + cos B + cos C
3
2
.
(c) cot A + cot B + cot C

3.
(d) tan A + tan B + tan C

3.
(e) sin Asin B sin C
3

3
8
.
2. Let 0 < a, b, c < 1. Prove that
a
1 a
+
b
1 b
+
c
1 c

3
3

abc
1
3

abc
.
4
AMSP 2010 Algebra 3.5: Day 10 Problems
3. Let n 2 and x
1
, x
2
, . . . , x
n
positive numbers whose sum is 1. Prove that
x
1

1 x
1
+
x
2

1 x
2
+ +
x
n

1 x
n

_
n
n 1
.
4. (USAMO 1977/5) If a, b, c, d, e are positive reals bounded by p and q with 0 < p q,
prove that
(a + b + c + d + e)
_
1
a
+
1
b
+
1
c
+
1
d
+
1
e
_
25 + 6
__
p
q

_
q
p
_
2
and determine when equality holds.
5. Let f(x) be a convex function dened on an interval I, and let x
1
, x
2
, x
3
I. Prove
that
f(x
1
) + f(x
2
) + f(x
3
) + 3f
_
x
1
+ x
2
+ x
3
3
_
2
_
f
_
x
1
+ x
2
2
_
+ f
_
x
2
+ x
3
2
_
+ f
_
x
3
+ x
1
2
__
.
Conclude that for a, b, c 0,
1
a
+
1
b
+
1
c
+
3
a + b + c

2
a + b
+
2
b + c
+
2
a + c
.
6. (Poland) Given a
1
, . . . , a
n
[0, 1], prove that
a
2
1
+ + a
2
n
a
1
+ + a
n
| +a
1
+ + a
n

2
.
7. (based o Vietnam 1998) Let x
1
, x
2
, . . . , x
n
, n 2 be positive real numbers with
1
x
1
+ 2010
+
1
x
2
+ 2010
+ +
1
x
n
+ 2010
=
1
2010
Prove that
n

x
1
x
2
x
n
n 1
2010.
8. (USAMO 1998/3) Let a
0
, a
1
, . . . , a
n
(0, /2) be numbers such that
tan(a
0
/4) + tan(a
1
/4) + + tan(a
n
/4) n 1.
Prove that tan a
0
tan a
1
tan a
n
n
n+1
.
9. (Romania 1999) Show that for all positive reals x
1
, . . . , x
n
with x
1
x
2
x
n
= 1, we
have
1
n 1 + x
1
+ +
1
n 1 + x
n
1.
Cauchy-Schwarz, H olders, and More
5
AMSP 2010 Algebra 3.5: Day 10 Problems
Theorem 7 (Weighted Power Mean). Let a
1
, . . . , a
n
and
1
, . . . ,
n
be positive numbers.
Dene the weighted power mean to be
P(a
1
, a
2
, . . . , a
n
; r) =
_

_
_

1
a
r
1
+
2
a
r
2
++
n
a
r
n

1
++
n
_
1/r
if r R0

1
++
n
_
a

1
1
a

2
2
a

n
n
if r = 0
max(a
1
, a
2
, . . . , a
n
) if r = +
min(a
1
, a
2
, . . . , a
n
) if r = .
Let a
1
, a
2
, . . . , a
n
be positive real numbers and r, s R with r > s. Then
P(a
1
, a
2
, . . . , a
n
; r) P(a
1
, a
2
, . . . , a
n
; s)
with equality if and only if a
1
= a
2
= a
n
.
In particular, QM-AM-GM-HM says
_
a
2
1
+ a
2
2
+ + a
2
n
n

a
1
+ a
2
+ + a
n
n

n

a
1
a
2
a
n

n
1
a
1
+
1
a
2
+ +
1
a
n
Theorem 8 (Cauchy-Schwarz). Let (a
1
, a
2
, . . . , a
n
) and (b
1
, b
2
, . . . , b
n
) be two n-tuples of
real numbers. Then
(a
1
b
1
+ a
2
b
2
+ + a
n
b
n
)
2
(a
2
1
+ a
2
2
+ + a
2
n
)(b
2
1
+ b
2
2
+ + b
2
n
)
with equality if and only if the two n-tuples are proportional, i.e. either every a
i
= 0 or
there is a real number with b
i
= a
i
for each i.
Theorem 9 (Titus Lemma). Let a
1
, a
2
, . . . , a
n
be real numbers and b
1
, b
2
, . . . , b
n
positive
real numbers. Then
a
2
1
b
1
+
a
2
2
b
2
+ +
a
2
n
b
n

(a
1
+ a
2
+ + a
n
)
2
b
1
+ b
2
+ + b
n
.
H olders inequality is a generalization of Cauchy-Schwarz; it allows an arbitrary num-
ber of sequences of variables, as well as dierent weights. First we need the following:
Theorem 10 (Young). For a, b > 0 and p, q > 0 such that
1
p
+
1
q
= 1,
ab
a
p
p
+
b
q
q
.
This is a special case of the weighted AM-GM inequality.
Theorem 11 (H older). Let a
1
, . . . , a
n
; b
1
, . . . , b
n
; . . . ; z
1
, . . . , z
n
be sequences of nonnegative
real numbers, and let
a
, . . . ,
z
be positive reals summing to 1. Then
(a
1
+ + a
n
)

a
(b
1
+ + b
n
)

b
(z
1
+ + z
n
)

z
a

a
1
b

b
1
z

z
1
+ a

a
n
b

b
n
z

z
n
.
Proof. First we prove that
(a
p
1
+ + a
p
n
)
1
p
(b
q
1
+ + a
q
n
)
1
q
a
1
b
1
+ + a
n
b
n
, (1)
6
AMSP 2010 Algebra 3.5: Day 10 Problems
when p, q > 0 and
1
p
+
1
q
, which is equivalent to the theorem statement for 2 variables.
Let A = (a
p
1
+ + a
p
n
)
1
p
(also denoted [[a[[
p
) and B = (b
q
1
+ + b
q
n
)
1
q
. Let a

i
=
a
i
A
and b

i
=
b
i
B
. Now that weve normalized (a
1
, . . . , a
n
) and (b
1
, . . . , b
n
) so that (a
p
1
+ +
a
p
n
)
1
p
= 1 and (b
q
1
+ + b
q
n
)
1
q
= 1, we can apply Minkowskis inequality.
a

1
b

1
+ + a

n
b

n

1
p
(a
p
1
+ + a
p
n
)
1
p
+
1
q
(b
q
1
+ + b
q
n
)
1
q
=
1
p
+
1
q
= 1
Multiplying by AB gives (1).
The general inequality follows by induction on the number of sequences. For example,
passing from 2 to 3 sequences, apply H older to with weights

a

a
+
b
,

b

a
+
b
, then with
weights
a
+
b
,
c
.
Theorem 12 (Minkowski). Let p > 1 and let a
1
, . . . , a
n
, b
1
, . . . , b
n
, . . . , z
1
, . . . , z
n
be posi-
tive numbers. Then
([a
1
[
p
+ +[a
n
[
p
)
1
p
+ ([b
1
[
p
+ +[b
n
[
p
)
1
p
+ + ([z
1
[
p
+ +[z
n
[
p
)
1
p
[([a
1
+ + z
1
[)
p
+ ([a
2
+ + z
2
[)
p
+ + ([a
n
+ + z
n
[)
p
]
1
p
Proof. We rst prove Minkowski for 2 sequences. We have
n

k=1
[a
k
+ b
k
[
p
=
n

k=1
([a
k
[ +[b
k
[)[a
k
+ b
k
[
p1
=
n

k=1
[a
k
[[a
k
+ b
k
[
p1
+
n

k=1
[b
k
[[a
k
+ b
k
[
p1

_
_
_
n

k=1
[a
k
[
p
_1
p
+
_
n

k=1
[b
k
[
p
_1
p
_
_
_
n

k=1
[a
k
+ b
k
[
(p1)
(
p
p1
)
_
p1
p
_
n

k=1
[a
k
+ b
k
[
p
_1
p

_
n

k=1
[a
k
[
p
_1
p
+
_
n

k=1
[b
k
[
p
_1
p
The general case follows by induction.
As a corollary, [[x[[
p
=
p
_
[x
1
[
p
+ +[x
n
[
p
is a valid norm in n-dimensional space for
p 1.
Theorem 13 (Schur). Let a, b, c 0 and r > 0. Then
a
r
(a b)(a c) + b
r
(b c)(b a) + c
r
(c a)(c b) 0
with equality i a = b = c or some two are equal and the other is 0.
Proof. Suppose WLOG a b c. Rewrite the inequality as
(a b)(a
r
(a c) b
r
(b c)) + c
r
(a c)(b c).
Both terms are positive.
7
AMSP 2010 Algebra 3.5: Day 10 Problems
Theorem 14 (Muirhead). Suppose (a
1
, . . . , a
n
) ~ (b
1
, . . . , b
n
). Then for any positive reals
x
1
, . . . , x
n
,

sym
x
a
1
1
x
a
2
2
x
a
n
n

sym
x
b
1
1
x
b
2
2
x
b
n
n
where the sum is taken over all permutations of n variables.
Problems
1. (Nesbitt) Let a, b, c > 0. Prove that
a
b + c
+
b
c + a
+
c
a + b

3
2
2. Prove Aczels Inequality: If a
2
1
a
2
2
+ + a
2
n
then
(a
1
b
1
a
2
b
2
a
n
b
n
)
2
(a
2
1
a
2
2
a
2
n
)(b
2
1
b
2
2
b
2
n
).
Hint: Consider the determinant of the quadratic
(a
1
x b
1
)
2
(a
2
x b
2
)
2
(ax
n
b
n
)
2
.
3. Prove that for all reals a, b, c,
(a
2
b + b
2
c + c
2
a)(ab
2
+ bc
2
+ ca
2
) (a
2
+ b
2
+ c
2
)(a
2
b
2
+ b
2
c
2
+ c
2
a
2
).
4. The numbers 1 x
1
, x
2
, . . . , x
n
1 satisfy x
3
1
+ x
3
2
+ + x
3
n
= 0. Prove that
x
1
+ x
2
+ + x
n

n
3
.
5. Let x, y, z be nonnegative real numbers. Prove that
(x
2
+ 1)(y
2
+ 1)(z
2
+ 1) (x + y + z xyz)
2
.
6. Let a, b, c be positive reals such that a + b + c = 1. Prove that

ab + c +

bc + a +

ca + b 1 +

ab +

bc +

ca.
7. (TST 2000/1) Let a
n

n0
be a sequence of real numbers such that a
n+1
a
2
n
+
1
5
for all n 0. Prove that

a
n+5
a
2
n5
for all n 5.
8. (IMO 2004/4) Let n 3 be an integer. Let t
1
, . . . , t
n
be positive real numbers such
that
n
2
+ 1 > (t
1
+ t
2
+ + t
n
)
_
1
t
1
+
1
t
2
+ +
1
t
n
_
.
Prove that for all i, j, k, the numbers t
i
, t
j
, t
k
are sides of a triangle.
8
AMSP 2010 Algebra 3.5: Day 10 Problems
9. (USAMO 2009/4) For n 2 let a
1
, a
2
, . . . , a
n
be positive real numbers such that
(a
1
+ a
2
+ + a
n
)
_
1
a
1
+
1
a
2
+ +
1
a
n
_

_
n +
1
2
_
2
.
Prove that max(a
1
, a
2
, . . . , a
n
) 4 min(a
1
, a
2
, . . . , a
n
).
10. (TST 2010/2) Let a, b, c be positive reals such that abc = 1. Show that
1
a
5
(b + 2c)
2
+
1
b
5
(c + 2a)
2
+
1
c
5
(a + 2b)
2

1
3
.
11. (Crux Mathematicorum) a, b, c, d, e are positive reals multiplying to 1. Prove that
a + abc
1 + ab + abcd
+
b + bcd
1 + bc + bcde
+
c + cde
1 + cd + cdea
+
d + dea
1 + de + deab
+
e + eab
1 + ea + eabc

10
3
12. (IMO 2000/2) Positive reals a, b, c have product 1. Prove that
_
a 1 +
1
b
__
b 1 +
1
c
__
c 1 +
1
a
_
1.
13. (IMO 2003/5) Let n be a positive integer and let x
1
x
n
be real numbers.
Prove that
_

1i,jn
[x
i
x
j
[
_
2

2(n
2
1)
3

1i,jn
(x
i
x
j
)
2
.
Show that equality holds if and only if x
1
, . . . , x
n
is an arithmetic sequence.
14. (IMO 2008/2) Prove that if x, y, z are three real numbers, all dierent from 1, such
that xyz = 1, then
x
2
(x 1)
2
+
y
2
(y 1)
2
+
z
2
(z 1)
2
1.
15. (USAMO 2004/5) Let a, b, c be positive reals. Prove that
(a
5
a
2
+ 3)(b
5
b
2
+ 3)(c
5
c
2
+ 3) (a + b + c)
3
.
16. (USAMO 1997/5) Prove that for all positive reals a, b, c,
1
a
3
+ b
3
+ abc
+
1
b
3
+ c
3
+ abc
+
1
c
3
+ a
3
+ abc

1
abc
.
17. (MOSP 2007) Let a, b, and c be a nonnegative real numbers with
1
a
2
+ 1
+
1
b
2
+ 1
+
1
c
2
+ 1
= 2.
Prove that
ab + bc + ba
3
2
.
9
AMSP 2010 Algebra 3.5: Day 10 Problems
18. (USAMO 2003/5) Let a, b, c be positive real numbers. Prove that
(2a + b + c)
2
2a
2
+ (b + c)
2
+
(2b + c + a)
2
2b
2
+ (c + a)
2
+
(2c + a + b)
2
2b
2
+ (c + a)
2
8.
19. (ISL 2004/A5) Let a, b, c > 0 and ab + bc + ca = 1. Prove the inequality
3
_
1
a
+ 6b +
3
_
1
b
+ 6c +
3
_
1
c
+ 6a
1
abc
.
20. (ISL 2006/A5) Let a, b, c be the sides of a triangle. Prove that

b + c a

b +

a
+

c + a b

c +

b
+

a + b c

a +

c
3.
21. (TST 2007/3) Show that for reals x, y, z which are not all positive,
16
9
(x
2
x + 1)(y
2
y + 1)(z
2
z + 1) (xyz)
3
xyz + 1.
22. (IMO 2005/3) Prove that for all positive a, b, c with product at least 1,
a
5
a
2
a
5
+ b
2
+ c
2
+
b
5
b
2
b
5
+ c
2
+ a
2
+
c
5
c
2
c
5
+ a
2
+ b
2
0.
23. (IMO 2006/3) Determine the least real number M such that for all reals a, b, c,
[a
3
b + b
3
c + c
3
a a
3
c b
3
a c
3
b[ M (a
2
+ b
2
+ c
2
)
2
.
24. Consider any sequence a
1
, a
2
, . . . of real numbers. Show that

n=1
a
n

2

n=1
_
r
n
n
_
1/2
,
where r
n
=

k=n
a
2
k
.
25. (ISL 2004/A7) Let a
1
, . . . , a
n
be positive real numbers, n > 1. Denote by g
n
their
geometric mean, and by A
1
, . . . A
n
the sequence of arithmetic means dened by
A
k
=
a
1
++a
k
k
, k = 1, 2, . . . , n. Let G
n
be the geometric mean of A
1
, . . . , A
n
. Prove
that
n
n
_
G
n
A
n
+
g
n
G
n
n + 1
and establish the cases of equality.
Reference: Olympiad Inequalities by Thomas Mildorf
10