Professional Documents
Culture Documents
DISCRETE
INEQUALITIES
VOLUME 3
SYMMETRIC
NONRATIONAL INEQUALITIES
A Glossary 211
i
ii Vasile Cîrtoaje
Chapter 1
1.1 Applications
1
2 Vasile Cîrtoaje
1 1 1 1 2
p +p +p ≥p +p .
a2 + 2bc b2 + 2ca c 2 + 2a b a2 + b2 + c 2 a b + bc + ca
p
1.10. Let a, b, c be nonnegative real numbers such that a + b + c = 3. If k = 3 − 1,
then XÆ p
a(a + k b)(a + kc) ≤ 3 3.
1.16. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
v v v
t a(b + c) t b(c + a) t c(a + b)
+ + ≥ 2.
a2 + bc b2 + ca c2 + a b
1.31. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
a b c
p +p +p ≤ 1.
4a2 + 5bc 4b2 + 5ca 4c 2 + 5a b
1.35. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
a2 + 2bc b2 + 2ca c 2 + 2a b p
p +p +p ≥3 a b + bc + ca.
b2 + bc + c 2 c 2 + ca + a2 a2 + a b + b2
1.36. Let a, b, c be nonnegative real numbers, no two of which are zero. If k ≥ 1, then
p
1.38. Let a, b, c be positive real numbers. If 0 ≤ k ≤ 1 + 2 2, then
a2 − bc b2 − ca c2 − a b
p +p +p ≥ 0.
ka2 + b2 + c 2 k b2 + c 2 + a2 kc 2 + a2 + b2
v v v
u
1 1 1 1 1 1
t t
(a + b + c) + + ≥1+ 1+ (a2 + b2 + c 2 ) + +
t
.
a b c a2 b2 c 2
1.45. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
v v v
t a2 + bc t b2 + ca t c2 + a b 1
+ + ≥2+ p .
b2 + c2 c2 + a2 a2 + b2 2
9
1.60. Let a, b, c be nonnegative real numbers such that a + b + c = 3. If k ≥ , then
8
p p p p
a2 + bc + k + b2 + ca + k + c 2 + a b + k ≥ 3 2 + k.
a k (b + c) + b k (c + a) + c k (a + b) ≥ 6.
a k (b + c) + b k (c + a) + c k (a + b) ≤ 2.
1.69. Let a, b, c be nonnegative real numbers, no two of which are zero. If m > n ≥ 0,
than
bm + c m c m + am am + bm
(b + c − 2a) + (c + a − 2b) + (a + b − 2c) ≥ 0.
bn + c n c n + an an + bn
b+c c + a a + b
a − 3 + b − 3 + c − 3 ≥ 2.
1.80. Let a, b, c be nonnegative real numbers, no two of which are zero, and let
2a 2b 2c
x= , y= , z= .
b+c c+a a+b
Prove that
p p p
x + y +z+ xy+ yz + z x ≥ 6.
12 Vasile Cîrtoaje
1.81. Let a, b, c be nonnegative real numbers, no two of which are zero, and let
2a 2b 2c
x= , y= , z= .
b+c c+a a+b
Prove that p p p
1 + 24x + 1 + 24 y + 1 + 24z ≥ 15.
1.84. If a, b, c are nonnegative real numbers, no two of which are zero, then
1 1 1 1 2
+ + ≥ +p .
a+b b+c c+a a+b+c a b + bc + ca
1.85. If a, b ≥ 1, then
1 1 1 1
p + ≥p +p .
3a b + 1 2 3a + 1 3b + 1
1
1.87. Let a, b, c be positive real numbers such that a + b + c = 3. If k ≥ p , then
2
(a bc)k (a2 + b2 + c 2 ) ≤ 3.
Symmetric Nonrational Inequalities 13
1.88. Let p and q be nonnegative real numbers such that p2 ≥ 3q, and let
v v
t 2p − 2w t 2p + w
g(p, q) = +2 ,
3 3
2p + 2w
s s
2p − w
+2 , p2 ≤ 4q
3 3
h(p, q) = p p ,
p + p + q, p2 ≥ 4q
p
where w =
p
p2 − 3q. If a, b, c are nonnegative real numbers such that
a + b + c = p, a b + bc + ca = q,
then
p p p
(a) a+b+ b+c+ c + a ≥ g(p, q),
p + 2w p−w
with equality for a = and b = c = (or any cyclic permutation);
3 3
p p p
(b) a + b + b + c + c + a ≤ h(p, q),
p − 2w p+w
with equality for a = and b = c = (or any cyclic permutation) - when
3 3
p2 ≤ 4q, and for a = 0, b+c = p and bc = q (or any cyclic permutation) - when p2 ≥ 4q.
aa + b b ≥ a b + ba .
14 Vasile Cîrtoaje
(a) a b + b a ≤ 1 + a b;
a2b + b2a ≤ 1 + a b.
1
2a a b b ≥ a2b + b2a + (a − b)2 .
2
2a a b b ≥ a2 + b2 .
Symmetric Nonrational Inequalities 15
2a a b b ≥ a2 + b2 .
a b b a + 2 ≥ 3a b.
a b−a + b a−b ≤ 2.
16 Vasile Cîrtoaje
1
1.108. Let a, b be positive real numbers such that a + b = 2. If k ≥ , then
2
kb ka
a a b b ≥ 1.
a−a + b−b ≤ 2.
a2b + b2a ≥ a b + b a ≥ a2 b2 + 1.
a3b + b3a ≤ 2.
a2b + b2a ≤ 1.
2a a b b ≥ a2b + b2a .
a−2a + b−2b ≤ 4.
where
1 1 1 1
A = (a + b + c + d) + + + − 16,
a b c d
1 1 1 1
2 2 2 2
B = (a + b + c + d ) 2 + 2 + 2 + 2 − 16.
a b c d
18 Vasile Cîrtoaje
n Ç
n
2
X X
(n − 1)2 ai4 + 2n − 1 ≥ ai ;
i=1 i=1
n
n
2
X q X
ai (n − 1)2 ai2 + 2n − 1 ≥ ai .
i=1 i=1
2
−1 − ≤ k < 1,
n−2
then X a1
≥ 1.
a1k + a2 + · · · + an
1.2 Solutions
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
Solution (by Nguyen Van Quy). Assume that c = min{a, b, c}. Since
b2 − bc + c 2 ≤ b2
and
c 2 − ca + a2 ≤ a2 ,
it suffices to show that
v
p t a2 + b2 + c 2
a2 − a b + b2 + a + b ≤ 3 .
2
Choosing k = 2, we get
v v
p t a2 + b2 t a2 + b2 + c 2
a2 − a b + b2 + a + b ≤ 3 ≤3 = 3.
2 2
X p p 2
6(a2 + b2 + c 2 − a b − bc − ca) ≥ 3a2 + 3b2 − 2a b − 3a2 + 3c 2 − 2ac ,
X X (b − c)2 (3b + 3c − 2a)2
3 (b − c)2 ≥ p p 2 ,
3a2 + 3b2 − 2a b + 3a2 + 3c 2 − 2ac
X (3b + 3c − 2a)2
(b − c)2 1 − p p 2 .
9a + 9b − 6a b + 9a + 9c − 6ac
2 2 2 2
Since p Æ
9a2 + 9b2 − 6a b = (3b − a)2 + 8a2 ≥ |3b − a|,
p Æ
9a2 + 9c 2 − 6ac = (3c − a)2 + 8a2 ≥ |3c − a|,
it suffices to show that
2
|3b + 3c − 2a|
X
2
(b − c) 1 − ≥ 0.
|3b − a| + |3c − a|
The equality holds for a = b = c, and also for b = c = 0 (or any cyclic permutation).
Symmetric Nonrational Inequalities 23
which is equivalent to
Æ Æ
(a + b + c)2 + 2(a − c)2 ≥ 3(a2 + b2 + c 2 ).
(a − b)(b − c) ≥ 0.
Æ
=2 3(a + b + c)2 + (a − c)2 .
Therefore, it suffices to show that
(a − b)(b − c) ≥ 0.
For k = −2/3 and k = 1, we get the inequalities in P 1.3 and P 1.4, respectively. For
k = −1 and k = 0, we get the inequalities
Xp p
a2 − a b + b2 ≥ 2 a2 + b2 + c 2 − 1,
Xp p
a2 + b2 ≥ 2 a2 + b2 + c 2 + 2.
First Solution (by Vo Quoc Ba Can). Using the Cauchy-Schwarz inequality, we have
X b2 + bc + c 2
X p 2 X
b + bc + c
2 2 ≤ (b + c)
b+c
b + bc + c 2
X 2 X a 2
= 2(a + b + c) =2 1+ (b + bc + c 2 )
b+c b+c
X 2a(b2 + bc + c 2 )
= 4(a2 + b2 + c 2 ) + 2(a b + bc + ca) +
b+c
2 2 2
X bc
= 4(a + b + c ) + 2(a b + bc + ca) + 2a b + c −
b+c
X 1
= 4(a2 + b2 + c 2 ) + 6(a b + bc + ca) − 2a bc .
b+c
Symmetric Nonrational Inequalities 25
where
(a + b + c)2 −a2 + a b + 2bc + ca
Sa = − 1 = ,
2a2 + b2 + c 2 + ca + a b 2a2 + b2 + c 2 + ca + a b
−b2 + bc + 2ca + a b
Sb = ≥ 0,
2b2 + c 2 + a2 + a b + bc
−c 2 + ca + 2a b + bc
Sc = ≥ 0.
2c 2 + a2 + b2 + bc + ca
According to
X a2
(b − c)2 Sa ≥ (b − c)2 Sa + (a − c)2 S b ≥ (b − c)2 Sa + 2 (b − c)2 S b
b
2 Sa Sb
2 a 2
≥ (b − c) Sa + (b − c) S b = a(b − c) + ,
b a b
it suffices to prove that
Sa S b
+ ≥ 0,
a b
which is equivalent to
First Solution (by Nguyen Van Quy). Assume that a = max{a, b, c}. Since
p p Æ
a2 + a b + b2 + c 2 + ca + a2 ≤ 2[(a2 + a b + b2 ) + (c 2 + ca + a2 )],
where
1
A = a2 + (b2 + c 2 + a b + ac), X = a2 + b2 + c 2 , Y = a b + bc + ca.
2
Write the desired inequality as follows
p p p p
2( A − X ) ≤ Y − b2 + bc + c 2 ,
2(A − X ) Y − (b2 + bc + c 2 )
p p ≤p p ,
A+ X Y + b2 + bc + c 2
b(a − b) + c(a − c) b(a − b) + c(a − c)
p p ≤p p .
A+ X Y + b2 + bc + c 2
Since b(a − b) + c(a − c) ≥ 0, we need to show that
p p p p
A + X ≥ Y + b2 + bc + c 2 .
Indeed,
The equality holds for a = b = c, and also for b = c = 0 (or any cyclic permutation).
Second Solution. In the first solution of P 1.5, we have shown that
X p 2 X 1
b2 + bc + c 2 ≤ 4(a2 + b2 + c 2 ) + 6(a b + bc + ca) − 2a bc .
b+c
Thus, it suffices to prove that
X 1 p p 2
4(a2 + b2 +c 2 )+6(a b+ bc +ca)−2a bc ≤ 2 a2 + b2 + c 2 + a b + bc + ca ,
b+c
which is equivalent to
X 1 Æ
2a bc + 4 (a2 + b2 + c 2 )(a b + bc + ca) ≥ 5(a b + bc + ca).
b+c
28 Vasile Cîrtoaje
Since X 1 9 9
≥P = ,
b+c (b + c) 2(a + b + c)
it is enough to prove that
9a bc Æ
+ 4 (a2 + b2 + c 2 )(a b + bc + ca) ≥ 5(a b + bc + ca),
a+b+c
which can be written as
9a bc Æ
+ 4 q(p2 − 2q) ≥ 5q,
p
where
p = a + b + c, q = a b + bc + ca.
For p2 ≥ 4q, this inequality is true because 4 q(p2 − 2q) ≥ 5q. Consider further
p
3q ≤ p2 ≤ 4q.
9a bc
≥ 4q − p2 .
p
which is Æ
4 q(p2 − 2q) ≥ p2 + q.
Indeed,
16q(p2 − 2q) − (p2 + q)2 = (p2 − 3q)(11q − p2 ) ≥ 0.
Third Solution. Let us denote
p p p
A = b2 + bc + c 2 , B = c 2 + ca + a2 , C = a2 + a b + b2 ,
p p
X = a2 + b2 + c 2 , Y = a b + bc + ca.
By squaring, the inequality becomes
X X
2 BC ≤ 2 a2 + 4X Y,
X
(B − C)2 ≥ 2(X − Y )2 ,
P 2
2
X (b − c)2 (b − c)2
2(a + b + c) ≥ .
(B + C)2 (X + Y )2
Symmetric Nonrational Inequalities 29
Since
B + C ≤ (c + a) + (a + b) = 2a + b + c,
it suffices to show that
P 2
2
X (b − c)2 (b − c)2
2(a + b + c) ≥ .
(2a + b + c)2 (X + Y )2
2(a + b + c)2 1
≥ ,
(b − c)2 (2a + b + c)2 (X + Y )2
P
which is X
2(a + b + c)2 (X + Y )2 ≥ (b − c)2 (2a + b + c)2 .
We see that
X X X X
(a + b + c)2 (X + Y )2 ≥ a2 + 2 ab a2 + ab
X 2 X X X 2
= a2 + 3 ab a2 + 2 ab
X X X
≥ a4 + 3 a b(a2 + b2 ) + 4 a2 b2
and X X
(b − c)2 (2a + b + c)2 = (b − c)2 [4a2 + 4a(b + c) + (b + c)2 ]
X X X
=4 a2 (b − c)2 + 4 a(b − c)(b2 − c 2 ) + (b2 − c 2 )2
X X X
≤8 a2 b2 + 4 a(b3 + c 3 ) + 2 a4 .
Thus, it suffices to show that
X X X X X X
a4 + 3 a b(a2 + b2 ) + 4 a2 b2 ≥ 4 a2 b2 + 2 a(b3 + c 3 ) + a4 ,
Consider the nontrivial case when no two of a, b, c are zero (Y 6= 0) and write the
inequality as X p
X − a2 + 2bc ≥ 2(X − Y ),
(b − c)2
P
X (b − c)2
p ≥ .
X + a2 + 2bc X +Y
By the Cauchy-Schwarz inequality, we have
P 2
X (b − c)2 (b − c)2
p ≥P p .
X + a2 + 2bc (b − c)2 X + a2 + 2bc
(b − c)2
P
1
p ≥ ,
(b − c) X + a + 2bc +
P
2 2 X Y
which is equivalent to X p
(b − c)2 Y − a2 + 2bc ≥ 0.
From p 2
Y− a2 + 2bc ≥ 0.
we get
p Y 2 − (a2 + 2bc) (a − b)(c − a)
Y− a2 + 2bc ≥ = .
2Y 2Y
Thus,
X p X (b − c)2 (a − b)(c − a)
(b − c)2 Y − a2 + 2bc ≥
2Y
(a − b)(b − c)(c − a) X
= (b − c) = 0.
2Y
The equality holds for a = b, or b = c, or c = a.
Symmetric Nonrational Inequalities 31
1 1 1 1 2
p +p +p ≥p +p .
a2 + 2bc b2 + 2ca c 2 + 2a b a2 + b2 + c 2 a b + bc + ca
(Vasile Cîrtoaje, 1989)
Solution . Let p p
X= a2 + b2 + c 2 , Y= a b + bc + ca.
Consider the nontrivial case when Y > 0 and write the inequality as
X 1 1 1 1
p − ≥2 − ,
a2 + 2bc X Y X
(b − c)2
P
X (b − c)2
p p ≥ .
Y (X + Y )
a2 + 2bc X + a2 + 2bc
By the Cauchy-Schwarz inequality, we have
P 2
X (b − c)2 (b − c)2
p p ≥P p p .
a2 + 2bc X + a2 + 2bc (b − c)2 a2 + 2bc X + a2 + 2bc
(b − c)2
P
1
p p ≥ ,
(b − c) a + 2bc X + a + 2bc (X + Y)
P
2 2 2 Y
which is equivalent to
X p
(b − c)2 [X Y − X a2 + 2bc + (a − b)(c − a)] ≥ 0.
Since X X
(b − c)2 (a − b)(c − a) = (a − b)(b − c)(c − a) (b − c) = 0,
the inequality becomes
X p
(b − c)2 X Y − a2 + 2bc ≥ 0,
X p
(b − c)2 Y − a2 + 2bc ≥ 0.
We have proved this inequality at the preceding problem P 1.7. The equality holds for
a = b, or b = c, or c = a.
32 Vasile Cîrtoaje
and
A = B = a2 + b2 + c 2 , C = a2 + b2 + c 2 .
We have
X + Y + Z = A+ B + C
and
A = B ≥ C.
Without loss of generality, assume that
a ≥ b ≥ c.
max{X , Y, Z} ≥ A, min{X , Y, Z} ≤ C.
Indeed, we have
min{X , Y, Z} − C ≤ Z − C = c(2c − a − b) ≤ 0.
Equality holds for a = b = c.
Lemma. If X , Y, Z and A, B, C are positive real numbers such that
X + Y + Z = A + B + C,
X ≥ A, Z ≤ C,
and hence
p p p p p p p p p p p p
X + Y + Z − A− B − C = ( X − A ) + ( Y − B ) + ( Z − C )
Symmetric Nonrational Inequalities 33
X −A Y −B Z −C X −A Y −B Z −C
≤ p + p + p ≤ p + p + p
2 A 2 B 2 C 2 B 2 B 2 C
C −Z Z −C 1 1
= p + p = (C − Z) p − p ≤ 0.
2 B 2 C 2 B 2 C
p
P 1.10. Let a, b, c be nonnegative real numbers such that a + b + c = 3. If k = 3 − 1,
then XÆ p
a(a + k b)(a + kc) ≤ 3 3.
which is an identity. The equality holds for a = b = c = 1, and also for a = 3 and
b = c = 0 (or any cyclic permutation).
(a − c)Ea ≥ (b − c)E b ,
which is equivalent to
Æ Æ Æ
(a − b) 3a b(a + b + c) + (a − c) a(2b + c)(2b + a) ≥ (b − c) b(2a + b)(2a + c).
This is true if
Æ Æ
(a − c) a(2b + c)(2b + a) ≥ (b − c) b(2a + b)(2a + c).
It is true if
a(2a b + 2bc + ca) ≥ b(2a b + bc + 2ca).
Indeed,
The equality holds for a = b = c = 1, and also for a = 0 and b = c = 3/2 (or any cyclic
permutation).
Solution. Denote
X c(a − b) + b(a − c)
p ≥ 0,
A+ b + c
X c(a − b) X c(b − a)
p + p ≥ 0,
A+ b + c B+c+a
X c(a − b)[a − b − (pA − pB)]
p p ≥ 0.
( A + b + c)( B + c + a)
It suffices to show that
p p
(a − b)[a − b + ( B − A)] ≥ 0.
Indeed,
p p a+b−c
2
(a − b)[a − b + ( B − A)] = (a − b) 1 + p p ≥ 0,
B+ A
36 Vasile Cîrtoaje
16
Notice that if k = , then the equality holds for a = b = c = 1, for a = 0 and b = c (or
9
any cyclic permutation), and for b = c = 0 (or any cyclic permutation).
( a)3
P
X Æ 2 27
a (a + b)(a + c) ≥ P a =P a .
(a + b)(a + c) (a + b)(a + c)
Symmetric Nonrational Inequalities 37
P 1.16. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
v v v
t a(b + c) t b(c + a) t c(a + b)
+ + ≥ 2.
a2 + bc b2 + ca c2 + a b
(Vasile Cîrtoaje, 2006)
4a bc ≥ 0.
The equality holds for a = 0 and b = c (or any cyclic permutation).
First Solution (by Nguyen Van Quy). Assume that c = min{a, b, c}. Since the equality
occurs for a = b and c = 0, we use the inequality
p p
c 2 + 9a b ≥ 3 a b.
Since
c(a + b) 2 c(a + b)(3a b − ac − bc)
a b(a b + bc + ca) − a b + = ≥ 0,
3 9
p
it suffices to show that f (c) ≥ 0 for 0 ≤ c ≤ a b, where
p
f (c) = (a + b)2 + 18c a b + [30a b + 10c(a + b)] − 34a b − 16c(a + b)
p
= (a + b)2 − 4a b + 6c(3 a b − a − b).
p
Since f (c) is a linear function, we only need to prove that f (0) ≥ 0 and f ( a b) ≥ 0.
We have
f (0) = (a − b)2 ≥ 0,
p p p
f ( a b) = (a + b)2 + 14a b − 6(a + b) a b ≥ (a + b)2 + 9a b − 6(a + b) a b
p 2
= a + b − 3 a b ≥ 0.
The equality holds for a = b and c = 0 (or any cyclic permutation).
Second Solution. Assume that c = min{a, b, c}. By squaring, the inequality becomes
X XÆ X
a2 + 2 (a2 + 9bc)(b2 + 9ca) ≥ 16 a b,
42 Vasile Cîrtoaje
X Æ p p p X
a2 + 2 (a2 + 9bc)(b2 + 9ca) + 2 c 2 + 9a b a2 + 9bc + b2 + 9ca ≥ 16 a b.
The Cauchy-Schwarz inequality gives
Æ p
(a2 + 9bc)(b2 + 9ca) ≥ a b + 9c a b.
hence p p p p
c 2 + 9a b a2 + 9bc + b2 + 9ca ≥ 3 a b (a + b + 4c).
Therefore, it suffices to show that f (c) ≥ 0, where
p p X
f (c) = a2 + b2 + 2(a b + 9c a b) + 6 a b (a + b + 4c) − 16 ab
p p
= a2 + b2 − 14a b + 6(a + b) a b + c[42 a b − 16(a + b)].
p
Since fp(c) is a linear function and 0 ≤ c ≤ a b, it is sufficient to show that f (0) ≥ 0
and f ( a b) ≥ 0. We have
p p p 2
f (0) = (a − b)2 + 6 ab a− b ≥0
and
p p p
f ( a b) = a2 + b2 + 28a b − 10(a + b) a b ≥ (a + b)2 + 25a b − 10(a + b) a b
p 2
= a + b − 5 a b ≥ 0.
X
(b − c)2 Sa ≥ 0,
where
a(b + c − a) Æ
Sa = , A = (b2 + 4ca)(c 2 + 4a b) + bc + 2a(b + c),
A
b(c + a − b) Æ
Sb = , B= (c 2 + 4a b)(a2 + 4bc) + ca + 2b(c + a),
B
c(a + b − c) Æ
Sc = , C = (a2 + 4bc)(b2 + 4ac) + a b + 2c(a + b).
C
Since S b ≥ 0 and Sc ≥ 0, we have
X a2
(b − c)2 Sa ≥ (b − c)2 Sa + (a − c)2 S b ≥ (b − c)2 Sa + 2 (b − c)2 S b
b
bSa aS b
a
= (b − c)2 + .
b a b
Thus, it suffices to prove that
bSa aS b
+ ≥ 0,
a b
which is equivalent to
b(b + c − a) a(c + a − b)
+ ≥ 0.
A B
Since
b(b + c − a) a(c + a − b) b(b − a) a(a − b) (a − b)(aA − bB)
+ ≥ + = ,
A B A B AB
it is enough to show that
aA − bB ≥ 0.
Indeed,
p p p
aA − bB = c 2 + 4a b a b2 + 4ca − b a2 + 4bc + 2(a − b)(a b + bc + ca)
p
4c(a3 − b3 ) c 2 + 4a b
= p p + 2(a − b)(a b + bc + ca) ≥ 0.
a b2 + 4ca + b a2 + 4bc
The equality holds for a = b = c, and also for a = b and c = 0 (or any cyclic permuta-
tion).
Second Solution (by Nguyen Van Quy). Write the inequality as
p p p 2
a2 + 4bc + b2 + 4ca + c 2 + 4a b ≥ a2 + b2 + c 2 + 14(a b + bc + ca),
44 Vasile Cîrtoaje
p p p Æ
a2 + 4bc + b2 + 4ca + c 2 + 4a b ≥ a2 + b2 + c 2 + 14(a b + bc + ca).
Assume that c = min{a, b, c}. For t = 2c, the inequality (b) in Lemma below becomes
p p Æ
a2 + 4bc + b2 + 4ca ≥ (a + b)2 + 8(a + b)c.
t ≤ 2(a + b).
Then,
p
(a) (a2 + 2bt)(b2 + 2at) ≥ a b + (a + b)t;
p p p
(b) a2 + 2bt + b2 + 2at ≥ (a + b)2 + 4(a + b)t.
(a − b)2 t[2(a + b) − t] ≥ 0,
Solution (by Nguyen Van Quy). We see that the equality holds for a = b and c = 0.
Without loss of generality, assume that c = min{a, b, c}. For t = 4c, the inequality (a)
in Lemma from the preceding P 1.20 becomes
Æ
(a2 + 8bc)(b2 + 8ca) ≥ a b + 4(a + b)c.
Thus, we have Æ
(a2 + 9bc)(b2 + 9ca) ≥ a b + 4(a + b)c,
and also,
p p p p Æ
4
c 2 + 9a b a2 + 9bc + b2 + 9ca ≥ 3 a b · 2 (a2 + 9bc)(b2 + 9ca)
p Æ Æ
≥ 6 a b · a b + 4(a + b)c = 3 4a2 b2 + 16a bc(a + b)
Æ
≥ 3 4a2 b2 + 4a bc(a + b) + c 2 (a + b)2 = 3(2a b + bc + ca).
Therefore,
XÆ
(a2 + 9bc)(b2 + 9ca) ≥ (a b + 4bc + 4ca) + 3(2a b + bc + ca)
= 7(a b + bc + ca).
The equality holds for a = b and c = 0 (or any cyclic permutation).
b 2 3b2 c 2 3c 2
2 2 2 2
(a + a b + b )(a + ac + c ) = a + + a+ +
2 4 2 4
c 3bc a(b + c)
b
≥ a+ a+ + = a2 + + bc.
2 2 4 2
Then,
X a(b + c)
XÆ
2
(a + a b + b )(a + ac + c ) ≥
2 2 2 2 a + + bc = (a + b + c)2 .
2
The equality holds for a = b = c, and also for b = c = 0 (or any cyclic permutation).
First Solution. Without loss of generality, assume that c = min{a, b, c}. We see that the
equality holds for a = b and c = 0. Since
Æ
(a2 + 7ac + c 2 )(b2 + 7bc + c 2 ) ≥ (a + 2c)(b + 2c) ≥ a b + 2c(a + b),
1
Due to homogeneity, we may assume that a + b = 1. Let us denote d = a b, d ≤ .
4
Since
2a b
c≤ = 2d,
a+b
1
we need to show that f (c) ≥ 0 for 0 ≤ c ≤ 2d ≤ , where
2
f (c) = (1 + 5d)(1 + 7c + 28cd) − (6d + 5c)2 .
Since f (c) is concave, it suffices to show that f (0) ≥ 0 and f (2d) ≥ 0. Indeed,
and
Solution (by Nguyen Van Quy). Without loss of generality, assume that c = min{a, b, c}.
It is easy to see that the equality holds for c = 0 and 11(a2 + b2 ) = 38a b. By the AM-GM
inequality, the following inequality holds for any k > 0:
v v v
t 1 t 1 t 1
60 a2 + a b + b2 a2 + ac + c 2 + b2 + bc + c 2 ≤
3 3 3
v v 2
36 2 1 1 1
t t
2
≤ a + a b + b + 25k a + ac + c + b + bc + c
2 2 2 2 .
k 3 3 3
We can use this inequality to prove the original inequality only if the equality
v v 2
36 2 1 1 1
t t
2
a + a b + b = 25k a2 + ac + c 2 + b2 + bc + c 2
k 3 3 3
holds for c = 0 and 11(a2 + b2 ) = 38a b. This necessary condition if satisfied for k = 1.
Therefore, it suffices to show that
v
1 1 1
t
2 2
60 a + ab + b
2 2 b + bc + c + 36 a + a b + b
2 2
3 3 3
v v 2
t 1 t 1
+25 a2 + ac + c 2 + b2 + bc + c 2 ≤ 61(a + b + c)2 ,
3 3
50 Vasile Cîrtoaje
which is equivalent to
v
1 1 31c(a + b)
t
10 a + ac + c
2 2 b + bc + c ≤ 10a b + c 2 +
2 2 .
3 3 3
Since
v v
1 1 4 4
t t
2 a2 + ac + c 2 b2 + bc + c 2 ≤ 2 a2 + ac b2 + bc
3 3 3 3
v
4 4
t
=2 a b+ c ·b a+ c
3 3
4 4
≤a b+ c +b a+ c
3 3
4c(a + b)
= 2a b + ,
3
we only need to prove that
2c(a + b) 31c(a + b)
10 a b + ≤ 10a b + c 2 + .
3 3
This reduces to the obvious inequality
3c 2 + 11c(a + b) ≥ 0.
Thus, the proof is completed. The equality holds for 11(a2 + b2 ) = 38a b and c = 0 (or
any cyclic permutation).
a b c a+b+c
p +p +p ≥p .
b2 + bc + c2 c2 + ca + a2 a2 + ab + b2 a b + bc + ca
from which the desired inequality follows. The equality holds for a = b = c, and also
for a = 0 and b = c (or any cyclic permutation).
a b c a+b+c
p +p +p ≤p .
a2 + 2bc b2 + 2ca c2 + 2a b a b + bc + ca
(Ho Phu Thai, 2007)
which is equivalent to
p p p
a2 + 2bc + b2 + 2ca ≥ 2 a b + bc + ca.
The equality holds for a = b = c, and also for a = b and c = 0 (or any cyclic permuta-
tion).
Second Solution. By the Cauchy-Schwarz inequality, we have
X 2
a X X a
p ≤ a .
a2 + 2bc a2 + 2bc
Thus, it suffices to prove that
X a a+b+c
≤ .
a2 + 2bc a b + bc + ca
This is equivalent to
X 1 1
a − 2 ≥ 0,
a b + bc + ca a + 2bc
X a(a − b)(a − c)
≥ 0.
a2 + 2bc
Assuming that a ≥ b ≥ c, we get
X a(a − b)(a − c) a(a − b)(a − c) b(b − c)(b − a)
≥ +
a2 + 2bc a2 + 2bc b2 + 2ca
c(a − b)2 [2a(a − c) + 2b(b − c) + 3a b]
= ≥ 0.
(a2 + 2bc)(b2 + 2ca)
Clearly, this is just Schur’s inequality of degree three applied to x 2 , y 2 , z 2 . So, the proof
is completed. The equality holds for a = b = c, and also for a = 0 or b = 0 or c = 0.
54 Vasile Cîrtoaje
P 1.31. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
a b c
p +p +p ≤ 1.
4a2 + 5bc 4b2 + 5ca 4c 2 + 5a b
(Vasile Cîrtoaje, 2004)
First Solution (by Vo Quoc Ba Can). If one of a, b, c is zero, then the inequality becomes
an equality. Consider next that a, b, c > 0 and denote
1
a b c
x=p , y=p , z=p , x, y, z ∈ 0, .
4a2 + 5bc 4b2 + 5ca 4c 2 + 5a b 2
We have
bc 1 − 4x 2 ca 1 − 4 y2 ab 1 − 4z 2
= , = , = ,
a2 5x 2 b2 5 y2 c2 5z 2
and
(1 − 4x 2 )(1 − 4 y 2 )(1 − 4z 2 ) = 125x 2 y 2 z 2 .
For the sake of contradiction, assume that x + y + z > 1. Using the AM-GM inequality
and the Cauchy-Schwarz inequality, we have
1 Y 1 Y
x 2 y 2z2 = (1 − 4x 2 ) < [(x + y + z)2 − 4x 2 ]
125 125
1 Y Y
= (3x + y + z) · ( y + z − x)
125
x + y + z 3 Y
≤ ( y + z − x)
3
1 Y
≤ (x 2 + y 2 + z 2 )(x + y + z) ( y + z − x)
9
1
= (x 2 + y 2 + z 2 )[2(x 2 y 2 + y 2 z 2 + z 2 x 2 ) − x 4 − y 4 − z 4 ],
9
and hence
9x 2 y 2 z 2 < (x 2 + y 2 + z 2 )[2(x 2 y 2 + y 2 z 2 + z 2 x 2 ) − x 4 − y 4 − z 4 ],
X
x 6 + y 6 + z 6 + 3x 2 y 2 z 2 < x 2 y 2 (x 2 + y 2 ).
The last inequality contradicts Schur’s inequality
X
x 6 + y 6 + z 6 + 3x 2 y 2 z 2 ≥ x 2 y 2 (x 2 + y 2 ).
Thus, the proof is completed. The equality holds for a = b = c, and also for a = 0 and
b = c (or any cyclic permutation).
Symmetric Nonrational Inequalities 55
bc ca
Second Solution. In the nontrivial case when a, b, c > 0, setting x = 2
, y = 2 and
a b
ab
z= (x yz = 1), the desired inequality becomes E(x, y, z) ≤ 1, where
c2
1 1 1
E(x, y, z) = p +p +p .
4 + 5x 4 + 5y 4 + 5z
(12 − 5p)q − 25p2 + 16 > (12 − 5p)(4 + 5p) − 25p2 + 16 = 2(8 − 5p)(4 + 5p) > 0.
56 Vasile Cîrtoaje
1 25
x= = ,
yz (9t − 4)2
2
Indeed,
p 3(t − 1)2 (9t 2 + 14t + 7)
36t 4 − 32t 2 + 21 − 3t 2 − 2t = p ≥ 0.
36t 4 − 32t 2 + 21 + 3t 2 + 2t
in the form
9a bc
≥ 2(a b + bc + ca) − a2 − b2 − c 2 ,
a+b+c
Symmetric Nonrational Inequalities 57
30[ (b + c)]2
P
X 30(b + c)
≥P
5p2 + 2q + 3a2 + 6bc (b + c)(5p2 + 2q + 3a2 + 6bc)
120p2 120p2
= = .
10p3 + 4pq + 9 bc(b + c) 10p3 + 13pq − 27a bc
P
120p2 9
≥ ,
10p + 13pq − 27a bc
3 p
which is equivalent to
10p3 + 81a bc ≥ 39pq.
From Schur’s inequality p3 + 9a bc ≥ 4pq and the known inequality pq ≥ 9a bc, we have
This completes the proof. The equality holds for a = b = c, and also for a = 0 and b = c
(or any cyclic permutation).
58 Vasile Cîrtoaje
Remark. Using the same way as in the second solution, we can prove the following
inequalities for a, b, c > 0 satisfying a bc = 1:
p p p
a 4a2 + 5 + b 4b2 + 5 + c 4c 2 + 5 ≥ (a + b + c)2 ;
p p p
4a4 + 5 + 4b4 + 5 + 4c 4 + 5 ≥ (a + b + c)2 .
The first inequality follows from the the Cauchy-Schwarz inequality
X p X a
a 4a + 5
2 p ≥ (a + b + c)2
4a2 + 5
and the inequality X a
p ≤ 1, a bc = 1,
4a2 + 5
which follows from the inequality in P 1.31 by replacing bc/a2 , ca/b2 , a b/c 2 with 1/x 2 ,
1/ y 2 , 1/z 2 , respectively.
The second inequality follows from the the Cauchy-Schwarz inequality
2
X p X a
4a4 + 5 p ≥ (a + b + c)2
4a + 5
4
First Solution (by Vo Quoc Ba Can). Using the AM-GM inequality yields
X p X a(b + c)(a2 + 3bc)
a a2 + 3bc = p
(b + c)2 (a2 + 3bc)
X 2a(b + c)(a2 + 3bc)
≥ .
(b + c)2 + (a2 + 3bc)
9
=
24 + a + 3 a b
P P
2
27
= P
8( a) + 3 a2 + 9 a b
P P
2
P
9 a 3
= ≥ P .
11( a) + 3 a b
P P
2 4 a
Then, it remains to show that
9a bc
≥ 2(a b + bc + ca) − a2 − b2 − c 2 ,
a+b+c
which is equivalent to Schur’s inequality of degree three
X X
a3 + 3a bc ≥ a b(a + b).
Since Æ
2 (a + b + c)2 (a2 + 8bc) ≤ (a + b + c)2 + (a2 + 8bc),
it suffices to show that
X
a[(a + b + c)2 + (a2 + 8bc)] ≤ 2(a + b + c)3 ,
a3 + b3 + c 3 + 24a bc ≤ (a + b + c)3 .
The equality holds for a = b = c, and also for b = c = 0 (or any cyclic permutation).
Symmetric Nonrational Inequalities 61
P 1.35. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
a2 + 2bc b2 + 2ca c 2 + 2a b p
p +p +p ≥3 a b + bc + ca.
b2 + bc + c 2 c 2 + ca + a2 a2 + a b + b2
(Michael Rozenberg and Marius Stanean, 2011)
a2 + 2bc b2 + 2ca c 2 + 2a b 3
+ + ≥ (a + b + c).
b+c c+a a+b 2
This inequality is equivalent to
1X
a4 + b4 + c 4 + a bc(a + b + c) ≥ a b(a + b)2 .
2
We can prove this inequality by summing Schur’s inequality of fourth degree
X
a4 + b4 + c 4 + a bc(a + b + c) ≥ a b(a2 + b2 )
P 1.36. Let a, b, c be nonnegative real numbers, no two of which are zero. If k ≥ 1, then
X ak a k+1
− 2 ≥ 0,
a + b + c 2a + bc
X a k (a − b)(a − c)
≥ 0.
2a2 + bc
Assume that a ≥ b ≥ c. Since (c − a)(c − b) ≥ 0, it suffices to show that
a k (a − b)(a − c) b k (b − a)(b − c)
+ ≥ 0.
2a2 + bc 2b2 + ca
This is true if
a k (a − c) b k (b − c)
− ≥ 0,
2a2 + bc 2b2 + ca
which is equivalent to
The equality holds for a = b = c, and also for a = b and c = 0 (or any cyclic permuta-
tion).
a2 − bc b2 − ca c2 − a b
(a) p +p +p ≥ 0;
3a2 + bc 3b2 + ca 3c 2 + a b
a2 − bc b2 − ca c2 − a b
(b) p +p +p ≥ 0.
8a2 + (b + c)2 8b2 + (c + a)2 8c 2 + (a + b)2
(Vasile Cîrtoaje, 2006)
Symmetric Nonrational Inequalities 63
We have
X a2 − bc X (a − b)(a + c) + (a − c)(a + b)
2 =
A A
X (a − b)(a + c) X (b − a)(b + c)
= +
A B
a+c b+c
X
= (a − b) −
A B
X a − b (a + c)2 B 2 − (b + c)2 A2
= ·
AB (a + c)B + (b + c)A
X c(a − b)2 2(a − b)2 + c(a + b + 2c)
= · ≥ 0.
AB (a + c)B + (b + c)A
The equality holds for a = b = c.
(b) Let
Æ Æ Æ
A= 8a2 + (b + c)2 , B= 8b2 + (c + a)2 , C= 8c 2 + (a + b)2 b.
p
P 1.38. Let a, b, c be positive real numbers. If 0 ≤ k ≤ 1 + 2 2, then
a2 − bc b2 − ca c2 − a b
p +p +p ≥ 0.
ka2 + b2 + c 2 k b2 + c 2 + a2 kc 2 + a2 + b2
64 Vasile Cîrtoaje
Solution. Let
p p p
A= ka2 + b2 + c 2 , B= k b2 + c 2 + a2 , C= kc 2 + a2 + b2 .
therefore
X a2 − bc X (a − b)2 C1
2 = · ≥ 0,
A AB (a + c)B + (b + c)A
where
The equality holds for a = b = c, and also for b = c = 0 (or any cyclic permutation).
Second Solution. Write the inequality as
where p p p
A= b + c, B= c + a, C= a + b.
We have
X X
2 A(a2 − bc) = A[(a − b)(a + c) + (a − c)(a + b)]
X X
= A(a − b)(a + c) + B(b − a)(b + c)
X
= (a − b)[A(a + c) − B(b + c)]
X A2 (a + c)2 − B 2 (b + c)2
= (a − b) ·
A(a + c) + B(b + c)
X (a − b)2 (a + c)(b + c)
= ≥ 0.
A(a + c) + B(b + c)
Solution. If two of a, b, c are zero, then the inequality is clearly true. Otherwise, write
the inequality as
AX + BY + C Z ≥ 0,
where p p p
a2 + 4bc b2 + 4ca c 2 + 4a b
A= , B= , C= ,
b+c c+a a+b
X = (a2 − bc)(b + c), Y = (b2 − bc)(b + c), X = (c 2 − a b)(a + b).
Without loss of generality, assume that a ≥ b ≥ c. We have X ≥ 0, Z ≤ 0 and
X + Y + Z = 0.
In addition,
X − Y = a b(a − b) + 2(a2 − b2 )c + (a − b)c 2 ≥ 0
66 Vasile Cîrtoaje
and
It is true since
The equality holds for a = b = c, and also for a = b and c = 0 (or any cyclic permuta-
tion).
hence
v v
X X 1 tX X 1 tX X 1
a −2 a +1=1+ a2 ,
a a a2
v 2 v
tX X 1 tX X 1
a −1 ≥1+ a2 ,
a a2
v v v
tX X 1 u tX X 1
−1≥ 1+
t
a a2 .
a a2
The equality holds if and only if
X X 1 X 1 X
a2 = bc ,
bc a2
which is equivalent to
(a2 − bc)(b2 − ca)(c 2 − a b) = 0.
Consequently, the equality occurs for a2 = bc, or b2 = ca, or c 2 = a b.
Therefore, v
1 1 1
t
2(a2+ b2 + c2) + + −2 ≥ x + y −2
a2 b2 c 2
1 1 1
= (a + b + c) + + − 5.
a b c
The equality occurs for a = b, or b = c, or c = a.
p = a + b + c, q = a b + bc + ca, r = a bc,
which is equivalent to
(p − q − r + 1)2 ≥ 0.
The equality holds for p + 1 = q + r and q ≥ 1. The last condition follows from p + q −
r − 1 ≥ 0.
Second Solution. Since
and
(1 + b2 )(1 + c 2 ) = (b + c)2 + (bc − 1)2 ,
by the Cauchy-Schwarz inequality, we get
Æ
2(1 + a2 )(1 + b2 )(1 + c 2 ) ≥ (1 + a)(b + c) + (1 − a)(bc − 1)
P 1.45. Let a, b, c be nonnegative real numbers, no two of which are zero. Prove that
v v v
t a2 + bc t b2 + ca t c 2 + a b 1
+ + ≥2+ p .
b +c
2 2 c +a
2 2 a +b
2 2
2
(Vo Quoc Ba Can, 2006)
Solution. We may assume that a ≥ b ≥ c. Then, it suffices to show that
v v v
t a2 + c 2 t b2 + c 2 t a b 1
+ + ≥2+ p .
b +c
2 2 c +a
2 2 a +b
2 2
2
Let us denote v
t a2 + c 2 s
a
x= , y= .
b2 + c2 b
Since
(a − b)(a b − c 2 )
x2 − y2 = ≥ 0,
b(b2 + c 2 )
it follows that
x ≥ y ≥ 1.
From
1 1 (x − y)(x y − 1)
x+ − y+ = ≥ 0,
x y xy
we have v v v
t a2 + c 2 t b2 + c 2 s
a tb
+ ≥ + .
b2 + c 2 c 2 + a2 b a
Therefore, it is enough to show that
s v v
a t b t ab 1
+ + ≥2+ p .
b a a +b
2 2
2
s
a
Putting t = , the inequality becomes
b
s
1 1 t
t + −2≥ p − .
t 2 t +1
2
We have
(t − 1)2 (t − 1)2
s
1 t
p − = ≤ 2
t +1
2 t +1
s
1
2 t
2(t 2 + 1) p + 2+1
2 t
(t − 1) 2
1
≤ = t + − 2.
t t
The equality holds for a = b and c = 0 (or any cyclic permutation).
Symmetric Nonrational Inequalities 71
X p X p
≥ (2bc + bc + a bc) = 3(a b + bc + ca) + a bc.
and Xp
a b (a + b) ≥ 2(a b + bc + ca).
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
Solution. Assume that a ≥ b ≥ c, consider the non-trivial case b > 0, and write the
inequality in the following equivalent homogeneous forms:
X Æ
a (4a + 5b)(4a + 5c) ≥ 3(a + b + c)2 ,
X X X p p 2
2( a2 − a b) = a 4a + 5b − 4a + 5c ,
X X 25a(b − c)2
(b − c)2 ≥ p p ,
( 4a + 5b + 4a + 5c)2
X
(b − c)2 Sa ≥ 0,
where
25a
Sa = 1 − p p .
( 4a + 5b + 4a + 5c)2
Since
25b 25b
Sb = 1 − p p ≥1− p p =0
( 4b + 5c + 4b + 5a)2 ( 4b + 9b)2
and
25c 25c 25
Sc = 1 − p p ≥1− p p =1− > 0,
( 4c + 5a + 4c + 5b)2 ( 9c + 9c)2 36
we have
X a2
(b − c)2 Sa ≥ (b − c)2 Sa + (a − c)2 S b ≥ (b − c)2 Sa + 2 (b − c)2 S b
b
a b a
= (b − c)2 Sa + S b .
b a b
Thus, it suffices to prove that
b a
Sa + S b ≥ 0.
a b
We have p p
25a a( 4a + 5b − 4a)2
Sa ≥ 1 − p p =1− ,
( 4a + 5b + 4a)2 b2
p p
25b b( 4b + 5a − 4b)2
Sb ≥ 1 − p p =1− ,
( 4b + 4b + 5a)2 a2
and hence
p p p p
b a b ( 4a + 5b − 4a)2 a ( 4b + 5a − 4b)2
Sa + S b ≥ − + −
a b a b b a
v v
t 4a2 5a t 4b2 5b
a b
=4 + + + − 7 + − 10
b2 b a2 a b a
q p
= 4 4x 2 + 5x − 8 + 2 20x + 41 − 7x − 10,
Symmetric Nonrational Inequalities 73
where
a b
x= + ≥ 2.
b a
To end the proof, we only need to show that x ≥ 2 yields
q p
4 4x 2 + 5x − 8 + 2 20x + 41 ≥ 7x + 10.
Indeed,
p p
15x 2 − 60x − 228 + 32 20x + 41 ≥ 15x 2 − 60x − 228 + 32 81 = 15(x − 2)2 ≥ 0.
3
The equality holds for a = b = c = 1, and also for c = 0 and a = b = (or any cyclic
2
permutation).
and
9c 9c 9
Sc = 1 − p p ≥1− p p =1− > 0,
( c + 3a + c + 3b)2 ( 4c + 4c)2 16
we have
X a2
(b − c)2 Sa ≥ (b − c)2 Sa + (a − c)2 S b ≥ (b − c)2 Sa + 2 (b − c)2 S b
b
a b a
= (b − c)2 Sa + S b .
b a b
Thus, it suffices to prove that
b a
Sa + S b ≥ 0.
a b
We have p p
9a a( a + 3b − a)2
Sa ≥ 1 − p p =1− ,
( a + 3b + a)2 b2
p p
9b b( b + 3a − b)2
Sb ≥ 1 − p p =1− ,
( b + b + 3a)2 a2
and hence
p p p p
b a b ( a + 3b − a)2 a ( b + 3a − b)2
Sa + S b ≥ − + −
a b a b b a
v v
t a2 3a t b2 3b a b
=2 + + + − + −6
b2 b a2 a b a
Æ p
= 2 x 2 + 3x − 2 + 2 3x + 10 − x − 6,
where
a b
x= + ≥ 2.
b a
To end the proof, it remains to show that
Æ p
2 x 2 + 35x − 2 + 2 3x + 10 ≥ x + 6
Indeed, p
3x 2 − 44 + 8 3x + 10 ≥ 12 − 44 + 32 = 0.
p
The equality holds for a = b = c = 1, and also for c = 0 and a = b = 3 (or any cyclic
permutation).
Symmetric Nonrational Inequalities 75
This is true if p p
(a − c) a(b + c) (b − c) b(c + a)
p p ≥p p .
a2 + bc + a(b + c) b2 + ca + b(c + a)
Since Æ Æ
a(b + c) ≥ b(c + a),
it suffices to show that
a−c b−c
p p ≥p p .
a2 + bc + a(b + c) b2 + ca + b(c + a)
Moreover, since
p Æ p Æ
a2 + bc ≥ a(b + c), b2 + ca ≤ b(c + a),
Indeed, we have
because
Since
(b2 + ca)(c 2 + a b) − bc(a + b)(a + c) = a(b + c)(b − c)2 ≥ 0,
it suffices to show that
X X
a(b + c)(a2 + bc) + 2 bc(a + b)(a + c) ≥ 4(a b + bc + ca)2 ,
which is equivalent to X
bc(b − c)2 ≥ 0.
(b) Write the inequality as
X p p
a(b + c) a2 + 2bc ≥ 2(a b + bc + ca) a b + bc + ca,
X p p
a(b + c) a2 + 2bc − a b + bc + ca ≥ 0,
X a(b + c)(a − b)(a − c)
p p ≥ 0.
a2 + 2bc + a b + bc + ca
Since (c − a)(c − b) ≥ 0, it suffices to show that
This is true if
a(b + c)(a − c) b(c + a)(b − c)
p p ≥p p .
a2 + 2bc + a b + bc + ca b2 + 2ca + a b + bc + ca
Since
(b + c)(a − c) ≥ (c + a)(b − c),
78 Vasile Cîrtoaje
a b
p ≥p .
(a + 2b)(a + 2c) (b + 2c)(b + 2a)
This is true if p p
a b
p ≥p .
(a + 2b)(a + 2c) (b + 2c)(b + 2a)
Indeed, we have
which is equivalent to
X Xp X X
a b(a2 + b2 ) + 2a bc bc ≥ 2 a2 b2 + 2a bc a,
X X Xp
a b(a − b)2 ≥ 2a bc( a− bc),
X (a − b)2 Xp p
≥ ( a − b)2 .
c
p p p
Using the substitution x = a, y = b, z = c, the last inequality becomes
X (x 2 − y 2 )2 X
≥ (x − y)2 ,
z2
y +z− x z+ x − y y +z− x z+ x − y 2z
A+ B = + ≥ + = 2 > 0.
x2 y2 x2 x2 x
p
The equality holds for a = b = c = 1, and for a = 0 and b = c = 3 (or any cyclic
permutation).
Second Solution. Assume that a ≥ b ≥ c, and write the inequality as
X p
a a b + 2bc + ca ≥ 2(a b + bc + ca),
X p
a a b + 2bc + ca − b − c ≥ 0,
X a(a b + ac − b2 − c 2 )
p ≥ 0,
a b + 2bc + ca + b + c
x y z
+ + ≥ 0, (*)
b+c+A c+a+B a+ b+C
Symmetric Nonrational Inequalities 81
where
p p p
A= a b + 2bc + ca, B= bc + 2ca + a b, C= ca + 2a b + bc,
B 2 − A2 = c(a − b) ≥ 0, C 2 − B 2 = a(b − c) ≥ 0.
Solution. Without loss of generality, assume that c = min{a, b, c}. Using Minkowski’s
inequality gives
p p p
p p Ç p p p p
a + 4bc + b + 4ca ≥ ( a + b)2 + 4c( a + b)2 = ( a + b) 1 + 4c.
Solution (by Nguyen Van Quy). Assume that c = min{a, b, c}. Using Minkowski’s in-
equality yields
p
p p Ç p
b + c + 7bc + a + c + 7ca ≥ (a + b)2 + 4c 2 + 7c( a + b)2 .
2 2 2 2
Since
−7c 5(5c − 1)2
f 00 (x) = p − p <0
4 x 3 4 [5x 2 + (5c + 1)x + c]3
1
f (c) is concave. Thus, it suffices to show that f (c 2 ) ≥ 0 and f ≥ 0.
4
Write the inequality f (c 2 ) ≥ 0 as
Æ
5 (1 + 5c 2 )(c + c 2 ) ≥ 6c 2 + 9c.
p
Making the substitution t = 4c + 1, t ≥ 1, the inequality becomes
Thus, the proof is completed. The equality holds for a = b and c = 0 (or any cyclic
permutation).
Solution. Without loss of generality, assume that c = min{a, b, c}. Using Minkowski’s
inequality, we have
p
Æ Æ Ç p
(a + c)2 + 3ac + (b + c)2 + 3bc ≥ (a + b + 2c)2 + 3c( a + b)2 .
p
Ç p Æ p
(a + b + 2c)2 + 3c( a + b)2 ≥ 21(a b + bc + ca) − a2 + b2 + 5a b.
Since p
00 −3c 21(3c − 1)2
f (x) = p − p <0
4 x 3 4 [3x 2 + (3c + 1)x + c]3
1
f (c) is concave. Thus, it suffices to show that f (c 2 ) ≥ 0 and f ≥ 0.
4
86 Vasile Cîrtoaje
1
Using the substitution 3t 2 = 4c + 1, t ≥ p , the inequality becomes
3
(t − 1)2 (3t 2 + 6t − 4) ≥ 0.
a b + bc + ca = 1.
where
9
Sa = (b + c)2 − a2 − bc(b + c)2 .
10
In addition, since
bc(b + c)2
Sa ≥ (b + c)2 − a2 − bc(b + c)2 = (b + c)2 − a2 − ,
a b + bc + ca
a(b + c)3 − a2 (a b + bc + ca)
= ,
a b + bc + ca
it is enough to show that X
(b − c)2 Ea ≥ 0,
88 Vasile Cîrtoaje
where
Ea = a(b + c)3 − a2 (a b + bc + ca).
and
Ea E b (b + c)3 (c + a)3
+ = + − 2(a b + bc + ca)
a2 b2 a b
b3 + 2b2 c a3 + 2a2 c
≥ + − 2(a b + bc + ca)
a b
(a2 − b2 )2 + 2c(a + b)(a − b)2
= ≥ 0,
ab
we get
2 Ea Eb
X
2 2 2 2
(b − c) Ea ≥ (b − c) Ea + (a − c) E b ≥ a (b − c) + ≥ 0.
a2 b2
p
The equality holds for a = b = c = 1, and also for a = b = 3 and c = 0 (or any cyclic
permutation).
Lemma. If x ≥ 0, y ≥ 0 and d > 0, then
Æ 1
2 (x + d)( y + d) ≥ x + y + 2d − (x − y)2 .
4d
Proof. We have
Æ 2(d x + d y + x y) 2(d x + d y + x y)
2 (x + d)( y + d) − 2d = p ≥ (x+d)+( y+d)
(x + d)( y + d) + d 2 +d
Applying Lemma from the preceding P 1.56 for x = 3a b, y = 3ac2 and d = 2, we have
Æ 9
2 (2 + 3a b)(2 + 3ac) ≥ 3a(b + c) + 4 − a2 (b − c)2 ,
8
X Æ X X 9 X
2 bc (2 + 3a b)(2 + 3ac) ≥ 3a bc (b + c) + 4 bc − a bc a(b − c)2
8
X 9 X
= 6a bc a + 4q − a bc a(b − c)2 .
8
Therefore, it suffices to show that
X X 9 X
1 + 3a bc a + 4q + 6a bc a − a bc a(b − c)2 ≥ 4q + 4q2 ,
8
which is equivalent to
X 9 X
1 + 9a bc a − 4q2 ≥ a bc a(b − c)2 .
8
Since X
a4 + b4 + c 4 = 1 − 2(a2 b2 + b2 c 2 + c 2 a2 ) = 1 − 2q2 + 4a bc a,
from Schur’s inequality of fourth degree
X X X
a4 + b4 + c 4 + 2a bc a≥ a2 ab ,
we get X
1 ≥ 2q2 + q − 6a bc a.
Thus, it is enough to prove that
X X 9 X
(2q2 + q − 6a bc a) + 9a bc a − 4q2 ≥ a bc a(b − c)2 ;
8
90 Vasile Cîrtoaje
that is, X X
8(q − 2q2 + 3a bc a) ≥ 9a bc a(b − c)2 .
Since X X X X 2 X
q − 2q2 + 3a bc a= a2 ab − 2 a b + 3a bc a
X X X
= bc(b2 + c 2 ) − 2 b2 c 2 = bc(b − c)2 ,
Also, since
8 − 9a2 = 8(b2 + c 2 ) − a2 ≥ b2 + c 2 − a2 ,
This is true if
a(c 2 + a2 − b2 )(a − c)2 ≥ b(a2 − b2 − c 2 )(b − c)2 .
a ≥ b, c 2 + a2 − b2 ≥ a2 − b2 − c 2 , (a − c)2 ≥ (b − c)2 .
1 1
The equality holds for a = b = c = p , and for a = 0 and b = c = p (or any cyclic
3 2
permutation).
Solution. (a) If two of a, b, c are zero, then the inequality is trivial. Otherwise, by
Hölder’s inequality, we have
X t 2a + bc 2
v
( a)3
P
9
a ≥P =P a .
3 3a
2a + bc 2a + bc
( bc)2 ( bc)2
P P
X bc
≥P = = 1.
2a + bc bc(2a + bc) 2a bc a + b2 c 2
P P
3
The equality holds for a = b = c = 1, and for a = 0 and b = c = (or any cyclic
2
permutation).
(b) Write the inequality in the homogeneous form
X Æ
a a(a + 4b + 4c) ≥ (a + b + c)2 .
where X X
Sa = a 2 a2 a b − 3a2 b2 c 2 .
Assume that a ≥ b ≥ c. Since Sa ≥ S b ≥ 0 and
X X
S b + Sc = (b2 + c 2 ) a2 a b − 6a2 b2 c 2
X X
≥ 2bc a2 a b − 6a2 b2 c 2
X
≥ 2bca2 a b − 6a2 b2 c 2 = 2a2 bc(a b + ac − 2bc) ≥ 0,
we get X
(b − c)2 Sa ≥ (c − a)2 S b + (a − b)2 Sc ≥ (a − b)2 (S b + Sc ) ≥ 0.
3
The equality holds for a = b = c = 1, and for a = 0 and b = c = (or any cyclic
2
permutation).
Since
b2 (a − c)2 ≥ a2 (b − c)2 ,
p p
a( B + 3b + c + a) ≥ b( A + 3a + b + c),
4(c + a − b) 4(a − b − c)
p p +1≥ p p − 1.
A+ C B+ C
Write the second inequality as
a2 B − b2 A
p p + (a − b)(a + b + c) ≥ 0.
a B+b A
94 Vasile Cîrtoaje
Since
p p p p p
a B + b A − 4a b > 2(a + b) a b − 4a b = 2 a b(a + b − 2 a b) ≥ 0.
and
The equality holds for a = b = c = 1, and also for a = 3 and b = c = 0 (or any cyclic
permutation).
9
P 1.60. Let a, b, c be nonnegative real numbers such that a + b + c = 3. If k ≥ , then
8
p p p p
a2 + bc + k + b2 + ca + k + c 2 + a b + k ≥ 3 2 + k.
and follows immediately from Jensen’s inequality applied to the convex function f :
[0, ∞) → R defined by Æ
f (x) = (2x + 3)2 + 8k − 9.
Symmetric Nonrational Inequalities 95
2bc − ca − a b 2ca − a b − bc 2a b − bc − ca
+ p + p ≥ 0,
A1 + A2 B1 + B2 C1 + C2
p p p p
X b(c − a) c(b − a)
p +p ≥ 0,
A1 + A2 A1 + A2
p p
X c(a − b) c(b − a)
p +p p ≥ 0,
B1 + B2 A1 + A2
p
X
c(a − b)( C1 + C2 )[( A1 − B1 ) + ( A2 − B2 )] ≥ 0,
p p p p p p
2(a + b − c) 2a + 2b + c
X
2
c(a − b) ( C1 + C2 ) p p +p
p p
≥ 0.
A1 + B1 A2 + B2
p
Without loss of generality, assume that a ≥ b ≥ c. Clearly, the desired inequality is true
for b + c ≥ a. Consider further the case b + c < a. Since a + b − c > 0, it suffices to show
that
2(b + c − a) 2b + 2c + a
2
a(b − c) ( A1 + A2 ) p p +p +
p p
B1 + C1 B2 + C2
p
2(c + a − b) 2c + 2a + b
2
+b(a − c) ( B1 + B2 ) p p +p
p p
≥ 0.
C1 + A1 C2 + AC2
p
Since
b2 (a − c)2 ≥ a2 (b − c)2 ,
it suffices to show that
2(b + c − a) 2b + 2c + a
b( A1 + A2 ) p p +p +
p p
B1 + C1 B2 + C2
p
2(c + a − b) 2c + 2a + b
+a( B1 + B2 ) p p +p
p p
≥ 0.
C1 + A1 C2 + A2
p
From
a2 B1 − b2 A1 = 8c(a3 − b3 ) + 8k(a2 − b2 ) ≥ 0
96 Vasile Cîrtoaje
and
a( B1 + B2 ) ≥ b( A1 + A2 ).
p p p p
2(b + c − a) 2b + 2c + a 2(c + a − b) 2c + 2a + b
p +p p +p p +p p ≥ 0.
B1 + C1 B2 + C2 C1 + A1 C2 + A2
p
This is true if
2b −2b
p +p p ≥0
B1 + C1 C1 + A1
p
and
−2a 2a 2a
p +p p +p p ≥ 0.
B1 + C1 C1 + A1 C2 + A2
p
1 1 1
+p ≥p .
C1 + C2 + B1 +
p p p p
A1 A2 C1
Since
1 1 4
+p ≥p ,
C1 + C2 + C1 + A1 + C2 +
p p p p p p
A1 A2 A2
it suffices to show that
B1 + 3 C1 ≥ A1 + A2 + C2 .
p p p p p
4
B1 + 3 B1 + 2
p p p p p p p p p
4 C1 − A1 − A2 − C2 ≥ 4 C1 − A1 − A2
B1 + 2 B1 −
p p p p
≥4 A2 − A2
= 2(3 B1 − A2 ).
p p
Symmetric Nonrational Inequalities 97
In addition,
Thus, the proof is completed. The equality holds for a = b = c = 1. If k = 9/8, then the
equality holds also for a = 3 and b = c = 0 (or any cyclic permutation).
Solution. Since
(a3 + 2bc)(a + 2bc) ≥ (a2 + 2bc)2 ,
it suffices to prove that
X a2 + 2bc p
p ≥ 3 3.
a + 2bc
By Hölder’s inequality, we have
2 X
a + 2bc
X 2
X 3
p (a2 + 2bc)(a + 2bc) ≥ (a2 + 2bc) = (a + b + c)6 .
a + 2bc
which is equivalent to
X
(a + b + c)4 ≥ (a2 + 2bc)(a2 + 6bc + ca + a b).
Indeed,
X X
(a + b + c)4 − (a2 + 2bc)(a2 + 6bc + ca + a b) = 3 a b(a − b)2 ≥ 0.
The equality holds for a = b = c = 1, and also for a = 3 and b = c = 0 (or any cyclic
permutation).
98 Vasile Cîrtoaje
= b3 c 3 + a(b5 + c 5 ) + a2 b2 c 2 .
Therefore, it suffices to show that
X X
2 b3 c 3 + 2 a(b5 + c 5 ) + 6a2 b2 c 2 ≥ 3(a2 + b2 )(b2 + c 2 )(c 2 + a2 ).
which yields
p 2a(b2 + c 2 ) + 4bc(b + c) + 6a bc
BC ≥ ,
b+c
p
4 a(b2 + c 2 ) + 8 bc(b + c) + 36a bc
P P
X BC
2 ≥ ,
(c + a)(a + b) (a + b)(b + c)c + a)
p
12 bc(b + c) + 36a bc
P
X BC
2 ≥ .
(c + a)(a + b) (a + b)(b + c)c + a)
On the other hand, according to the known inequality Iran-1996,
X a b + bc + ca 9
≥ ,
(b + c) 2 4
(see Remark from the proof of P 1.70), we have
X A X a b + bc + ca X a 9 X a
= + 3 ≥ + 3 .
(b + c)2 (b + c)2 b+c 4 b+c
Thus, it suffices to show that
12 bc(b + c) + 36a bc
P
X a
3 + ≥ 18.
b+c (a + b)(b + c)c + a)
This is equivalent to Schur’s inequality of degree three
X X
a3 + 3a bc ≥ bc(b + c).
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
100 Vasile Cîrtoaje
A ≥ B ≥ C,
Therefore,
AX + BY + C Z ≥ BX + BY + BZ = B(X + Y + Z) ≥ 0.
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
Symmetric Nonrational Inequalities 101
hence v 2
Xt bc X 1
≤ .
3a + 6
2 a2 +2
Therefore, to prove the original left inequality, it suffices to show that
X 1
≤ 1.
a2 +2
102 Vasile Cîrtoaje
Solution. Let
E = a k (b + c) + b k (c + a) + c k (a + b).
We consider two cases.
Case 1: k ≥ 2. Applying Jensen’s inequality to the convex function f (x) = x k−1 , x ≥ 0,
we get
E = (a b + ac)a k−1 + (bc + ba)b k−1 + (ca + c b)c k−1
(a b + ac)a + (bc + ba)b + (ca + c b)c k−1
≥ 2(a b + bc + ca)
2(a b + bc + ca)
k−1
a (b + c) + b2 (c + a) + c 2 (a + b)
2
=6 .
6
Symmetric Nonrational Inequalities 103
a2 (b + c) + b2 (c + a) + c 2 (a + b) ≥ 6.
(a b + bc + ca)(a + b + c) − 3a bc ≥ 6,
a + b + c ≥ 2 + a bc.
It is true since Æ
a+b+c ≥ 3(a b + bc + ca) = 3
and 3
a+b+c
a bc ≤ = 1.
3
= 3(a k−1 + b k−1 + c k−1 ) − a k−1 b k−1 c k−1 (a b)2−k + (bc)2−k + (ca)2−k .
Since
3 − x2 3 − x2
1≥ ≥ ,
2 2x
the conclusion follows. Thus, the proof is completed. The equality holds for a = b =
c = 1.
a k (b + c) + b k (c + a) + c k (a + b) ≤ 2.
Solution. Denote by Ek (a, b, c) the left hand side of the inequality, assume that a ≤ b ≤
c and show that
Ek (a, b, c) ≤ Ek (0, a + b, c) ≤ 2.
The left inequality is equivalent to
a b k−1
(a + b k−1 ) ≤ (a + b)k − a k − b k .
c
Clearly, it suffices to consider c = b, when the inequality becomes
2a k + b k−1 (a + b) ≤ (a + b)k .
which is equivalent to
cd(2 − cd)(k−1)/2 ≤ 1.
Since 2 − cd ≥ 1, we have
P 1.69. Let a, b, c be nonnegative real numbers, no two of which are zero. If m > n ≥ 0,
than
bm + c m c m + am am + bm
(b + c − 2a) + (c + a − 2b) + (a + b − 2c) ≥ 0.
bn + c n c n + an an + bn
(Vasile Cîrtoaje, 2006)
AX + BY + C Z ≥ 0,
106 Vasile Cîrtoaje
where
bm + c m c m + am am + bm
A= , B = , C = ,
bn + c n c n + an an + bn
X = b + c − 2a, Y = c + a − 2b, Z = a + b − 2c, X + Y + Z = 0.
Without loss of generality, assume that a ≤ b ≤ c, which involves X ≥ Y ≥ Z and X ≥ 0.
Since
Indeed, we have
First Solution. Among a − 1, b − 1 and c − 1 there are two with the same sign. Let
(b − 1)(c − 1) ≥ 0, that is,
1
t ≤ , t = b + c − 1.
a
By Minkowsky’s inequality, we have
v v
2
1 2 3
u u
1 3 t
p p
b − b+1+ c −c+1= + + +
t
2 2 b− c−
2 4 2 4
p
≥ t 2 + 3.
Thus, it suffices to show that
p p
a2 − a + 1 + t 2 + 3 ≥ a + b + c,
which is equivalent to p
a2 − a + 1 + f (t) ≥ a + 1,
where p
f (t) = t 2 + 3 − t.
Clearly, f (t) is decreasing for t ≤ 0. Since
3
f (t) = p ,
t2 + 3 + t
1
f (t) is also decreasing for t ≥ 0. Then, f (t) ≥ f , and it suffices to show that
a
1
p
a2 − a + 1 + f ≥ a + 1,
a
which is equivalent to
v
p t1 1
a2 − a + 1 + + 3 ≥ a + + 1.
a2 a
By squaring, this inequality becomes
v
1 2
t
2 (a − a + 1) 2 + 3 ≥ 3a + − 1.
2
a a
Indeed, by the Cauchy-Schwarz inequality, we have
v v
1 1
t t
2 (a − a + 1) 2 + 3 = [(2 − a)2 + 3a2 ] 2 + 3
2
a a
2−a 2
≥ + 3a = 3a + − 1.
a a
108 Vasile Cîrtoaje
1 3
p
x2 − x +1− x ≥ −1
2 x + x +1
2
1 1 1
+ + ≥ 1,
a2 + a + 1 b2 + b + 1 c 2 + c + 1
which is just the known inequality in P 1.43. Indeed, the above inequality is equivalent
to
1− x (1 − x)(2 + x)
p ≥ ,
x − x +1+ x
2 2(x 2 + x + 1)
p
(x − 1)[(x + 2) x 2 − x + 1 − x 2 − 2] ≥ 0,
3x 2 (x − 1)2
p ≥ 0.
(x + 2) x 2 − x + 1 + x 2 + 2
(MEMO, 2012)
which follows immediately from the inequality in P 1.43. Therefore, using the substitu-
tion x = a1/5 , x > 0, we need to show that
2x 5 + 3 3
≥ 8 ,
2x + 2x + 1
10 5 x + x4 + 1
which is equivalent to
2x 4 (x 5 − 3x 2 + x + 1) + x 4 − 4x + 3 ≥ 0.
and
4
p
x4 + 3 = x4 + 1 + 1 + 1 ≥ 4 x 4 · 1 · 1 · 1 = 4x.
The equality holds for a = b = c = 1.
Second Solution. Making the substitution
p p p
x = 16a2 + 9 − 4a, y = 16b2 + 9 − 4b, z = 16c 2 + 9 − 4c, x, y, z > 0,
which involves
9 − x2 9 − y2 9 − z2
a= , b= , c= ,
8x 8y 8z
we need to show that
yields
x + y + z ≥ 3.
Use the contradiction method. Assume that x + y + z < 3 and show that
hence
p
(3 + x)(3 + y)(3 + z) ≥ 64 4 x yz.
Therefore, it suffices to prove that
which follows immediately from P 1.31. Therefore, using the substitution x = a1/13 ,
x > 0, we only need to show that
p p
25x 26 + 144 + 5x 13 ≥ 6 5x 18 + 4.
This is true if
25x 26 + 144 ≥ (18x 5 − 5x 13 )2 ,
which is equivalent to
5x 18 + 4 ≥ 9x 10 .
By the AM-GM inequality, we have
5x 18 + 4 = x 18 + x 18 + x 18 + x 18 + x 18 + 1 + 1 + 1 + 1
9
p
≥ 9 x 18 · x 18 · x 18 · x 18 · x 18 · 1 · 1 · 1 · 1 = 9x 10 .
which involves
9 − 4x 2 9 − 4 y2 9 − 4z 2 3
a= , b= , c= , x, y, z ∈ 0, ,
5x 5y 5z 2
we need to show that
involves
x + y + z ≤ 3.
Use the contradiction method. Assume that x + y + z > 3 and show that
Since
12x 2 3( y + z − x)( y + z + 3x)
9 − 4x 2 < 3(x + y + z) − = ,
x + y +z x + y +z
it suffices to show that
27AB ≤ 125x yz(x + y + z)3 ,
112 Vasile Cîrtoaje
where
125
B≤ (x + y + z)3 .
27
Therefore, it suffices to show that
A ≤ x yz,
which is a well known inequality (equivalent to Schur’s inequality of degree three).
Solution. (a) First Solution (by Pham Thanh Hung). By squaring, the inequality be-
comes XÆ
(b2 + 3)(c 2 + 3) ≥ 3(1 + a + b + c).
Since
1
(b2 + 3)(c 2 + 3) = (b + c)(b + a)(c + a)(c + b) = (b + c)2 (a2 + 3) ≥ (b + c)2 (a + 3)2 ,
4
we have
XÆ 1X 1 X X
(b2 + 3)(c 2 + 3) ≥ (b + c)(a + 3) = 2 bc + 6 a = 3(1 + a + b + c).
2 2
The equality holds for a = b = c = 1.
Second Solution. Write the inequality as follows:
Æ Æ Æ
(a + b)(a + c) + (b + c)(b + a) + (c + a)(c + b) ≥ a + b + c + 3,
Æ X p p 2
2 a + b + c − 3(a b + bc + ca) ≥ a+b− a+c ,
1 X X (b − c)2
p (b − c)2 ≥ p p 2 ,
a+b+c+ 3(a b + bc + ca) a+b+ a+c
Symmetric Nonrational Inequalities 113
X Sa (b − c)2
p p 2 ≥ 0,
a+b+ a+c
where p p 2 Æ
Sa = a+b+ a+c −a−b−c− 3(a b + bc + ca).
The last inequality is true since
Æ Æ
Sa = 3(a + b + c) + 2 (a + b)(a + c) − 3(a b + bc + ca)
Æ Æ
>2 a2 + (a b + bc + ca) − 3(a b + bc + ca) > 0.
Third Solution. Use the substitution
p p p
x = a2 + 3 − a, y = b2 + 3 − b, z = c 2 + 3 − c, x, y, z > 0.
A = 5a2 + 3, B = 5b2 + 3, C = 5c 2 + 3,
which is equivalent to
Since
and
25a2 + 120 − 4(b − c)2 − 25bc = 25a2 + 120 − 4(b + c)2 − 9bc = 3(7a2 + 8a + 28 − 3bc),
Consider a fixed, a ≥ 1, and denote x = bc. So, we only need to prove that f (x) ≥ 0
for
a2 − 6a + 9
3a − 2a2 ≤ x ≤ ,
4
where
a − 6a + 9
2
Since f is concave, it suffices to show that f (3a − 2a ) ≥ 0 and f
2
≥ 0.
4
Indeed,
a2 − 6a + 9
375
f = (25a4 − 140a3 + 286a2 − 252a + 81)
4 16
375
= (a − 1)2 (5a − 9)2 ≥ 0.
16
Thus, the proof is completed. The equality holds for a = b = c = 1, and also for a = 9/5
and b = c = 3/5 (or any cyclic permutation).
k
Lemma. Let b, c ≥ 0 such that b + c ≤ 2. If k > 0 and 0 ≤ m ≤ , then
2k + 2
Æ
(k b2 + 1)(kc 2 + 1) ≥ m(b − c)2 + k bc + 1.
p p p p a2 + b2 + c 2 + 33
A ( B + C) + BC ≥ ,
6
p p a2 + b2 + c 2 + 33
q
A(B + C + 2 BC ) + BC ≥ ,
6
where
A = a2 + 1, B = b2 + 1, C = c 2 + 1.
Applying Lemma from the preceding problem P 1.74 for k = 1 and m = 1/4 gives
p 1
BC ≥ (b − c)2 + bc + 1.
4
which is equivalent to
Æ
6 2(a2 + 1)[3(b + c)2 + 8 − 4bc] ≥ 2a2 − (b + c)2 + 54 − 4bc,
Æ
6 2(a2 + 1)(3a2 − 18a + 35 − 4bc) ≥ a2 + 6a + 45 − 4bc.
Consider a fixed, a ≥ 1, and denote x = bc. So, we only need to prove that f (x) ≥ 0
for
a2 − 6a + 9
3a − 2a2 ≤ x ≤ ,
4
where
f (x) = 72(a2 + 1)(3a2 − 18a + 35 − 4x) − (a2 + 6a + 45 − 4x)2 .
Symmetric Nonrational Inequalities 117
a2 − 6a + 9
Since f is concave, it suffices to show that f (3a − 2a ) ≥ 0 and f
2
≥ 0.
4
Indeed,
a2 − 6a + 9
f = 144(a4 − 6a3 + 13a2 − 12a + 4)
4
= 144(a − 1)2 (a − 2)2 ≥ 0.
The equality holds for a = b = c = 1, and also for a = 2 and b = c = 1/2 (or any cyclic
permutation).
p p p p a2 + b2 + c 2 + 21
A ( B + C) + BC ≥ ,
2
p p a2 + b2 + c 2 + 21
q
A(B + C + 2 BC ) + BC ≥ ,
2
where
A = a2 + 3, B = b2 + 3, C = c 2 + 3.
Applying Lemma from problem P 1.74 for k = 1/3 and m = 1/9 gives
p 1
BC ≥ (b − c)2 + bc + 3.
3
Therefore, it suffices to show that
v
t 2 1 a2 + b2 + c 2 + 21
A[B + C + (b − c)2 + 2bc + 6] + (b − c)2 + bc + 3 ≥ ,
3 3 2
118 Vasile Cîrtoaje
which is equivalent to
Æ
2 3(a2 + 3)[5(b + c)2 + 36 − 8bc] ≥ 3a2 + (b + c)2 + 45 − 4bc,
Æ
3(a2 + 3)(5a2 − 30a + 81 − 8bc) ≥ 2a2 − 3a + 27 − 2bc.
Consider a fixed, a ≥ 1, and denote x = bc. So, we only need to prove that f (x) ≥ 0
for
a2 − 6a + 9
3a − 2a2 ≤ x ≤ ,
4
where
f (x) = 3(a2 + 3)(5a2 − 30a + 81 − 8x) − (2a2 − 3a + 27 − 2x)2 .
a − 6a + 9
2
Since f is concave, it suffices to show that f (3a − 2a ) ≥ 0 and f
2
≥ 0.
4
Indeed,
f (3a − 2a2 ) = 27a2 (a − 1)2 ≥ 0,
a2 − 6a + 9
27 4
f = (a − 8a3 + 22a2 − 24a + 9)
4 4
27
= (a − 1)2 (a − 3)2 ≥ 0.
4
The equality holds for a = b = c = 1, and also for a = 3 and b = c = 0 (or any cyclic
permutation).
(b) By squaring, the inequality becomes
p p p p 27 − a2 − b2 − c 2
A ( B + C) + BC ≥ ,
2
p p 27 − a2 − b2 − c 2
q
A(B + C + 2 BC ) + BC ≥ ,
2
where
A = 3a2 + 1, B = 3b2 + 1, C = 3c 2 + 1.
Consider a fixed, a ≥ 1, and denote x = bc. So, we only need to prove that f (x) ≥ 0
for
a2 − 6a + 9
3a − 2a2 ≤ x ≤ ,
4
where
f (x) = 3(3a2 + 1)(11a2 − 66a + 111 − 8x) − (15 + 15a − 4a2 − 2x)2 .
a − 6a + 9
2
Since f is concave, it suffices to show that f (3a − 2a ) ≥ 0 and f
2
≥ 0.
4
Indeed,
f (3a − 2a2 ) = 27(a − 1)2 (3a − 2)2 ≥ 0,
a2 − 6a + 9
27
f = (9a4 − 48a3 + 94a2 − 80a + 25)
4 4
27
= (a − 1)2 (3a − 5)2 ≥ 0.
4
The equality holds for a = b = c = 1, and also for a = 5/3 and b = c = 2/3 (or any
cyclic permutation).
Remark. Similarly, we can prove the following generalization.
• Let a, b, c be nonnegative real numbers such that a + b + c = 3. If k > 0, then
v
p p p t 8k(a2 + b2 + c 2 ) + 3(9k2 + 10k + 9)
ka2 + 1 + k b2 + 1 + kc 2 + 1 ≥ ,
3(k + 1)
3k + 1 3k − 1
with equality for a = b = c = 1, and also for a = and b = c = (or any
2k 4k
cyclic permutation).
120 Vasile Cîrtoaje
which is equivalent to
p
A(A + B + C + 210) ≤ (A + 105 − BC)2 .
which is equivalent to
Since
32(a2 + b2 + c 2 ) + 219 = 32a2 + 32(b + c)2 − 64bc + 219 = 64a2 − 192a + 507 − 64bc
and
Since
a − 6a + 9
2
f is decreasing, hence f (x) ≥ f . Therefore, it suffices to show that
4
a − 6a + 9
2
f ≥ 0. We have
4
a − 6a + 9
2
f =9[9a2 + 10a + 20 − (a2 − 6a + 9)]2
4
− (32a2 + 3)[64a2 − 192a + 507 − 16(a2 − 6a + 9)]
=9(8a2 + 16a + 11)2 − (32a2 + 3)(48a2 − 96a + 363)
=192a(a − 1)2 (18 − 5a) ≥ 0.
Thus, the proof is completed. The equality holds for a = b = c = 1, and also for a = 0
and b = c = 3/2 (or any cyclic permutation).
Lemma. If b, c ≥ 0 such that b + c ≥ 2, then
Æ
(32b2 + 3)(32c 2 + 3) ≤ 5(b2 + c 2 ) + 22bc + 3.
Case 2: a ≤ b + c. We have
b+c a + b c + a b+c c + a
− 3 + − 3 + − 3 ≥ − 3 + − 3
a c
b a b
b+c c + a 2b b + a (a − b)(2b − a)
= 3− + 3− ≥6− − =2+ ≥ 2.
a b a b ab
a
Thus, the proof is completed. The equality holds for = b = c (or any cyclic permuta-
2
tion).
b + c c + a a + b
+ +
≥ 2.
a b c
b + c |a − c| |a − b|
+ + ≥2
a b c
for all a, b, c > 0. Without loss of generality, assume that b ≥ c.
For b ≥ c ≥ a, we have
b + c |a − c| |a − b| b+c
+ + ≥ ≥ 2.
a b c a
For b ≥ a ≥ c, we have
x yz = x + y + z + 2 > 2
x − y − z + 2 = x yz
involves
x + y +z ≥2
x + y + z − 2 = x + y + z − (x yz − x + y + z) = x(2 − yz),
p
x(1 − yz) + 2(1 − yz ) ≥ 0,
p p
(1 − yz )[x(1 + yz ) + 2] ≥ 0,
hence
yz ≤ 1 < 2.
P 1.80. Let a, b, c be nonnegative real numbers, no two of which are zero, and let
2a 2b 2c
x= , y= , z= .
b+c c+a a+b
Prove that
p p p
x + y +z+ xy+ yz + z x ≥ 6.
Symmetric Nonrational Inequalities 125
we have
12a bc + 2 bc(b + c)
P
Xp
yz ≥
(a + b)(b + c)(c + a)
8a bc
= + 2.
(a + b)(b + c)(c + a)
a b c 4a bc
+ + + ≥ 2,
b + c c + a a + b (a + b)(b + c)(c + a)
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
Second Solution. Write the inequality as
X Xp p
4 (x − 1) ≥ ( y − z)2 .
Since
X X (a − b) + (a − c) X a − b X b − a X (a − b)2
(x − 1) = = + =
b+c b+c c+a (b + c)(c + a)
X (b − c)2
= ,
(a + b)(a + c)
and
p p 2(b − c)2 (a + b + c)2
( y − z)2 = p p 2 ,
(a + b)(a + c) b2 + a b + c 2 + ac
we can write the inequality as
X
(b − c)2 Ea ≥ 0,
126 Vasile Cîrtoaje
where
(a + b + c) 2
Ea = (b + c) 2 − p p 2 .
b + a b + c + ac
2 2
Therefore,
a+b+c
Ea ≥ (b + c) 2 − = b + c − a,
b+c
and hence X X
(b − c)2 Ea ≥ (b − c)2 (b + c − a) ≥ 0.
The right inequality is just Schur’s inequality of third degree.
Third Solution. Using the Cauchy-Schwarz inequality, we have
a b c (a + b + c)2 (a + b + c)2
+ + ≥ = .
b+c c+a a+b a(b + c) + b(c + a) + c(a + b) 2(a b + bc + ca)
Using Hölder’s inequality, we have
s v 2
(a + b + c)3
s
a t b c
+ + ≥ 2 .
b+c c+a a+b a (b + c) + b2 (c + a) + c 2 (a + b)
6a bc p ≥ (p2 − q)(4q − p2 ), p = a + b + c,
we obtain
6a bc ≥ (4q − 1)(1 − q).
Therefore,
1 2 1 2
+ − 12 ≥ + (4q−1)(1−q)
− 12
q q − 3a bc q q−
2
Symmetric Nonrational Inequalities 127
1 4 (1 − 3q)(1 − 4q)2
= + 2 − 12 = ≥ 0.
q 4q − 3q + 1 q(4q2 − 3q + 1)
P 1.81. Let a, b, c be nonnegative real numbers, no two of which are zero, and let
2a 2b 2c
x= , y= , z= .
b+c c+a a+b
Prove that p p p
1 + 24x + 1 + 24 y + 1 + 24z ≥ 15.
(Vasile Cîrtoaje, 2005)
Solution (by Vo Quoc Ba Can). Assume that c = min{a, b, c}, hence z = min{x, y, z},
z ≤ 1. By Hölder’s inequality
s v 2
a t b 2
a (b + c) + b2 (c + a) ≥ (a + b)3 ,
+
b+c c+a
we get
8(a + b) 8
= = .
a + b + 2c 1+z
Using this result and Minkowski’s inequality, we have
v
p p q p p t 48
1 + 24x + 1 + 24 y ≥ (1 + 1)2 + 24( x + y)2 ≥ 2 1 + .
1+z
The equality holds for a = b = c, and also for a = 0 and b = c (or any cyclic permuta-
tion).
we have
(x 2 − 7)a + 3x 2 b + 3x 2 c = 0
3 y 2 a + ( y 2 − 7)b + 3 y 2 c = 0 ,
3z a + 3z b + (z − 7)c = 0
2 2 2
which involves
x2 − 7 3x 2 3x 2
3 y2 y − 7 3 y2
2 =0 ;
3z 2 3z 2 z2 − 7
that is,
F (x, y, z) = 0,
where X X
F (x, y, z) = 4x 2 y 2 z 2 + 8 x2 y2 + 7 x 2 − 49.
We need to show that F (x, y, z) = 0 involves x + y + z ≤ 3, where x, y, z > 0. To
do this, we use the contradiction method. Assume that x + y + z > 3 and show that
F (x, y, z) > 0. Since F (x, y, z) is strictly increasing in each of its arguments, it is enough
Symmetric Nonrational Inequalities 129
then
x + y + z ≤ 3.
For the sake of contradiction, assume that x + y + z > 3 and show that F (x, y, z) > 0,
where
X x2 1
F (x, y, z) = − .
2x 2 + 7 3
130 Vasile Cîrtoaje
Since F (x, y, z) is strictly increasing in each of its arguments, it is enough to prove that
x + y + z = 3 involves F (x, y, z) ≥ 0. This is just the inequality in P 1.29. We give here
another proof. By the Cauchy-Schwarz inequality, we have
( x 3/2 )2 ( x 3/2 )2
P P
X x2
≥P = P .
2x 2 + 7 x(2x 2 + 7) 2 x 3 + 21
Therefore, it suffices to show that
X X
3( x 3/2 )2 ≥ 2 x 3 + 21,
G(x, y, z) − G(x, t, t) = y 6 + z 6 + 6 y 3 z 3 − 8t 4 + 6x 3 ( y 3 + z 3 − 2t 3 ).
Since y 3 + z 3 − 2t 3 ≥ 0, we get
6x 3 ( y 3 + z 3 − 2t 3 ) ≥ 4x 3 ( y 3 + z 3 − 2t 3 ) ≥ ( y 3 + z 3 + 2t 3 )( y 3 + z 3 − 2t 3 ).
Thus,
G(x, y, z) − G(x, t, t) ≥ y 6 + z 6 + 6 y 3 z 3 − 8t 4 + [( y 3 + z 3 )2 − 4t 6 ]
= 2( y 6 + z 6 + 4 y 3 z 3 − 6t 6 ) = 2[( y 2 + z 2 )3 − 3 y 2 z 2 ( y 2 + z 2 ) + 4 y 3 z 3 − 6t 6 ]
= 4(t 6 − 3t 2 y 2 z 2 + 2 y 3 z 3 ) = 4(t 2 − yz)2 (t 2 + 2 yz) ≥ 0.
Also,
7
G(x, t, t) = x 6 + 2t 6 + 6(t 6 + 2x 3 t 3 ) − (x 2 + 2t 2 )3
9
2
= (x − t)2 (x − 2t)2 (x 2 + 6x t + 2t 2 ) ≥ 0.
9
Symmetric Nonrational Inequalities 131
X Æ
3
3 X 2 X b 2 + c 2
a2 (b2 + c 2 ) ≤ a(b + c) · .
(b + c)2
X b2 + c 2 p4
− 1 ≤ − 3,
(b + c)2 6q2
X 2bc p4
+ ≥ 3,
(b + c)2 6q2
where
p = a + b + c, q = a b + bc + ca.
9 p2 p4
− + 2 ≥ 3,
2 q 6q
which is equivalent to
2
p2
−3 ≥ 0.
q
The equality holds for a = b = c = 1.
132 Vasile Cîrtoaje
P 1.84. If a, b, c are nonnegative real numbers, no two of which are zero, then
1 1 1 1 2
+ + ≥ +p .
a+b b+c c+a a+b+c a b + bc + ca
(Vasile Cîrtoaje, 2005)
p = a + b + c, q = a b + bc + ca, r = a bc,
p2 + q 1 2
≥ +p .
pq − r p q
1 1 2
+ ≥p ,
b c bc
which is obvious. For b = c = 1, the inequality becomes as follows:
1 2 1 2
+ ≥ +p ,
2 a+1 a+2 2a + 1
1 1 2 2
− ≥p − ,
2 a+2 2a + 1 a + 1
p
a 2(a + 1 − 2a + 1)
≥ p ,
2(a + 2) (a + 1) 2a + 1
a 2a2
≥ p p .
2(a + 2) (a + 1) 2a + 1(a + 1 + 2a + 1)
So, we need to show that
1 2a
≥ p p .
2(a + 2) (a + 1) 2a + 1(a + 1 + 2a + 1)
1 2a
≥ ,
2(a + 2) (a + 1)(3a + 2a + 1)
which is equivalent to (a−1)2 ≥ 0. The proof is completed. The equality holds for a = 0
and b = c.
P 1.85. If a, b ≥ 1, then
1 1 1 1
p + ≥p +p .
3a b + 1 2 3a + 1 3b + 1
2 2
x=p , y=p , x, y ∈ (0, 1],
3a + 1 3b + 1
the desired inequality can be written as
v
t 3
xy ≥ x + y − 1.
x2 y2 − x2 − y2 + 4
t = x + y − 1, p = x y, 1 ≥ p ≥ t ≥ 0.
Since
x 2 + y 2 = (x + y)2 − 2x y = (t + 1)2 − 2p,
134 Vasile Cîrtoaje
1 1 1
p +p ≥ .
3b + 1 3b1 + 1 2
1 1 1
p +p ≥ ,
3c + 1 3c1 + 1 2
1 1 1 1
p + ≥p +p ,
3a + 1 2 3b1 + 1 3c1 + 1
Symmetric Nonrational Inequalities 135
which is equivalent to
1 1 1 1
+ ≥p +p .
3b1 c1 + 1 3b1 + 1 3c1 + 1
p
2
From the inequality in the preceding P 1.85, the conclusion follows. The equality holds
for a = b = c = 1.
1
P 1.87. Let a, b, c be positive real numbers such that a + b + c = 3. If k ≥ p , then
2
(a bc)k (a2 + b2 + c 2 ) ≤ 3.
Solution. Since 3
a+b+c
abc ≤ = 1,
3
p
it suffices to prove the desired inequality for k = 1/ 2. Write the inequality in the
homogeneous form
3k+2
a+b+c
(a bc)k (a2 + b2 + c 2 ) ≤ 3 .
3
According to P 2.57-(a) in Volume 1, for fixed a+b+c and a b+bc+ca, the product a bc is
maximal when two of a, b, c are equal. Therefore, it suffices to prove the homogeneous
inequality for b = c = 1; that is, f (a) ≥ 0, where
From
3k + 2 k 2a
f 0 (a) = − − 2
a+2 a a +2
p p
2(a − 1)(ka2 − 2a + 2k) 2(a − 1)(a − 2)2
= = ,
a(a + 2)(a2 + 2) a(a + 2)(a2 + 2)
it follows that f is decreasing on (0, 1] and increasing on [1, ∞); therefore, f (a) ≥
f (1) = 0. This completes the proof. The equality holds for a = b = c.
136 Vasile Cîrtoaje
P 1.88. Let p and q be nonnegative real numbers such that p2 ≥ 3q, and let
v v
t 2p − 2w t 2p + w
g(p, q) = +2 ,
3 3
2p + 2w
s s
2p − w
+2 , p2 ≤ 4q
3 3
h(p, q) = p p ,
p + p + q, p2 ≥ 4q
p
where w =
p
p2 − 3q. If a, b, c are nonnegative real numbers such that
a + b + c = p, a b + bc + ca = q,
then
p p p
(a) a+b+ b+c+ c + a ≥ g(p, q),
p + 2w p−w
with equality for a = and b = c = (or any cyclic permutation);
3 3
p p p
(b) a + b + b + c + c + a ≤ h(p, q),
p − 2w p+w
with equality for a = and b = c = (or any cyclic permutation) - when
3 3
p ≤ 4q, and for a = 0, b + c = p and bc = q (or any cyclic permutation) - when p2 ≥ 4q.
2
b + c = p − a,
(a + b)(a + c) = a2 + q
and
p
p Ç Æ
a+b+ a+c = a+p+2 a2 + q,
we get p p p
a+b+ b+c+ c + a = f (a),
where Ç Æ
f (a) = p−a+ a+p+2 a2 + q.
p
From
a2 + q + 2a
p
−1
f 0 (a) = p + p ,
2 p − a 2 a 2 + q · a + p + 2p a 2 + q
q
Symmetric Nonrational Inequalities 137
Therefore, f (a) is increasing on [a1 , a2 ] and decreasing on [0, a1 ] and [a2 , p], where
p + p2 − 3q
p p
p − p2 − 3q
a1 = , a2 = .
3 3
(a) From
we get
f (a1 ) = f (a4 ) = g(p, q).
p+2
p p
p−2 p2 − 3q p2 − 3q
a3 = ≥ 0, a4 = .
3 3
Since a3 ≤ a1 ≤ a2 ≤ a4 , it follows that f (a) is increasing on [a1 , a2 ] and decreasing on
[a3 , a1 ] ∪ [a2 , a4 ]. Thus,
f (a) ≤ max{ f (a2 ), f (a3 )}.
138 Vasile Cîrtoaje
we get
f (a2 ) = f (a3 ) = h(p, q).
Case 2: p2 ≥ 4q. Assume that a = min{a, b, c}, a ≤ p/3. From
0 ≤ bc = q − a(b + c) = q − a(p − a) = a2 − pa + q,
p+
p p
p− p2 − 4q p2 − 4q
a5 = , a6 = .
2 2
Since a6 > p/3, it follows that a ∈ [0, a5 ]. Since a1 ≤ a5 ≤ a2 , f (a) is decreasing on
[0, a1 ] and increasing on [a1 , a5 ]; thus,
It remains to show that max{ f (0), f (a5 )} = h(p, q). Indeed, from a52 + q = pa5 and
r q Æ p
a5 + p − a5 = p+2 pa5 − a52 = p + 2 q,
p p
we get
f (0) = f (a5 ) = h(p, q).
a + b + c = a b + bc + ca = 4,
then p
2(1 + 10) p p p p
p ≤ a + b + b + c + c + a ≤ 2(1 + 2),
3
with left equality for a = 8/3 and b = c = 2/3 (or any cyclic permutation), and right
equality for a = 0 and b = c = 2 (or any cyclic permutation).
Symmetric Nonrational Inequalities 139
a + b + c = 4, a b + bc + ca = 5,
then p p
p p p p p 10 + 2 7
2+2 3≤ a+ b+ b+c+ c+a ≤ p ,
3
with left equality for a = 2 and b = c = 1 (or any cyclic permutation), and right equality
for a = 2/3 and b = c = 5/3 (or any cyclic permutation).
a + b + c = 11, a b + bc + ca = 7,
then p p p p p Æ p
3 6≤ a+b+ b+c+ c+a≤ 11 + 11 + 7,
with left equality for a = 31/3 and b = c = 1/3 (or any cyclic permutation), and right
equality for a = 0, b + c = 11 and bc = 7 (or any cyclic permutation).
ln x ≤ x − 1, x > 0.
f (x) = x − 1 − ln x, x > 0.
From
x −1
f 0 (x) =
,
x
it follows that f (x) is decreasing on (0, 1] and increasing on [1, ∞). Therefore,
f (x) ≥ f (1) = 0.
a b(a − b)(a3 − b3 ) ≥ 0.
1
Taking now x = in the inequality ln x ≤ x − 1 yields
a
a ln a ≥ a − 1.
Similarly,
b ln b ≥ b − 1,
and hence
3 3
ln a a b b = a3 ln a + b3 ln b ≤ a2 (a − 1) + b2 (b − 1) = a3 + b3 − (a2 + b2 ).
3 3
Thus, to prove the right inequality a a b b ≥ 1, it suffices to show that a3 + b3 ≥ a2 + b2 ,
which is equivalent to the homogeneous inequality
For b ≤ 1, we have
a a+b (a a−b − 1) ≥ 0 ≥ b2b (b a−b − 1).
For b ≥ 1, the inequality is also true since
Case 2: a ≤ 1. Since
a2a + b2b ≥ 2a a b b ,
it suffices to show that
2a a b b ≥ a a+b + b a+b ,
which can be written as a b a
b
+ ≤ 2.
b a
By Bernoulli’s inequality, we get
a
a b a−b b b−a a b(a − b) a(b − a)
b
+ = 1+ + 1+ ≤1+ +1+ = 2.
b a b a b a
The equality holds for a = b.
Conjecture 1. If a, b are positive real numbers, then
aa + b b ≥ a b + ba .
Solution. Assume that a ≥ b. We will show that if a ≥ 1, then the inequality is true.
From
a a−b ≥ b a−b ,
we get
a b ba
bb ≥ .
aa
Therefore,
a b ba (a a − a b )(a a − b a )
aa + b b − a b − ba ≥ aa + − a b
− b a
= ≥ 0.
aa aa
Consider further the case 0 < b ≤ a < 1.
First Solution. Denoting
a
c = ab, d = bb, k= ,
b
where c ≥ d and k ≥ 1, the inequality becomes
c k − d k ≥ c − d.
Since the function f (x) = x k is convex for x ≥ 0, from the well-known inequality
we get
c k − d k ≥ kd k−1 (c − d).
Thus, it suffices to show that
kd k−1 ≥ 1,
which is equivalent to
b1−a+b ≤ a.
Indeed, since 0 < 1 − a + b ≤ 1, by Bernoulli’s inequality, we get
ba ab a
c= , d= , k= ,
a b + ba a b + ba b
where c + d = 1 and k ≥ 1, the inequality becomes
ck a + d k−b ≥ 1.
Symmetric Nonrational Inequalities 143
a1−b ≥ b1−a ,
ln x
f (x) = .
1− x
It is enough to prove that f (x) is an increasing function. Since
g(x) 1
f 0 (x) = , g(x) = − 1 + ln x.
(1 − x)2 x
c 2k − d 2k ≥ c 2 − d 2 .
The left inequality follows from Lemma below for x = (c/d)2 . The right inequality is
equivalent to
k(cd)k−1 > 1,
144 Vasile Cîrtoaje
b
(a b)a−b > ,
a
1+a− b
ln a > ln b.
1−a+ b
For fixed a, let us define
1+a− b
f (b) = ln a − ln b, 0 < b < a.
1−a+ b
If f 0 (b) < 0, then f (b) is strictly decreasing, and hence f (b) > f (a) = 0. Since
−2 1
f 0 (b) = ln a − ,
(1 − a + b)2 b
we need to show that g(a) > 0, where
(1 − a + b)2
g(a) = 2 ln a + .
b
From
2 2(1 − a + b) 2(a − b)(a − 1)
g 0 (a) = − = < 0,
a b ab
it follows that g(a) is strictly decreasing on [b, 1), therefore g(a) > g(1) = b > 0. This
completes the proof. The equality holds for a = b.
Lemma. Let k and x be positive real numbers. If either k > 1 and x ≥ 1, or 0 < k < 1
and 0 < x ≤ 1, then
k−1
x k − 1 ≥ k x 2 (x − 1).
Proof. We need to show that f (x) ≥ 0, where
k−1
f (x) = x k − 1 − k x 2 (x − 1).
We have
1 k−3 k+1
f 0 (x) = k x 2 g(x), g(x) = 2x 2 − (k + 1)x + k − 1.
2
Since k−1
g 0 (x) = (k + 1) x 2 − 1 ≥ 0,
g(x) is increasing. If x ≥ 1, then g(x) ≥ g(1) = 0, f (x) is increasing, hence f (x) ≥
f (1) = 0. If 0 < x ≤ 1, then g(x) ≤ g(1) = 0, f (x) is decreasing, hence f (x) ≥ f (1) =
0. The equality holds for x = 1.
Remark. The following more general inequality holds for 0 < k ≤ e and any positive
numbers a and b (Vasile Cîrtoaje, 2006):
a ka + b kb ≥ a kb + b ka .
Symmetric Nonrational Inequalities 145
2a a b b ≥ a2b + b2a .
Note that Conjecture 1 has been proved for a, b ∈ (0, 1] by A. Coronel and F. Huancas
(2014), and also by L. Matejicka (2014).
(a) a b + b a ≤ 1 + a b;
a b ≤ 1 + b(a − 1) = 1 + b − b2
and
b a = b · b a−1 ≤ b[1 + (a − 1)(b − 1)] = b2 (2 − b).
(a) We have
a2b + b2a ≤ 1 + a b.
b2a ≤ b2 .
(1 + a b − b)2 + b2 ≤ 1 + a b,
b(2 + 2a b − a − 2b − a2 b) ≥ 0.
2 + 2a b − a − 2b − a2 b ≥ 0,
which is equivalent to
2(1 − a)(1 − b) + a(1 − a b) ≥ 0,
4(1 − a)(1 − b) + a(a − b)2 ≥ 0.
Since a ≥ 1, it suffices to show that
Indeed,
which is equivalent to
(a + b)(a3 + b3 ) ≥ (a2 + b2 )2 ,
a b(a − b)2 ≥ 0.
The equality holds for a = b.
1 − a b + a2 b2 − a a b b ≥ 1 − a b + a2 b2 − (a2 − a + 1)(b2 − b + 1)
= (a + b)(1 − a)(1 − b) ≥ 0.
To prove the inequality x x ≤ x 2 − x + 1, we show that f (x) ≤ 0 for x ∈ (0, 1], where
We have
2x − 1
f 0 (x) = ln x + 1 − ,
x2 − x +1
148 Vasile Cîrtoaje
(1 − x)(1 − 2x − x 2 − x 4 )
f 00 (x) = .
x(x 2 − x + 1)2
Let x 1 ∈ (0, 1) be the positive root of the equation x 4 + x 2 + 2x = 1. Then, f 00 (x) > 0
for x ∈ (0, x 1 ) and f 00 (x) < 0 for x ∈ (x 1 , 1), hence f 0 is strictly increasing on (0, x 1 ]
and strictly decreasing on [x 1 , 1]. Since lim x→0 f 0 (x) = −∞ and f 0 (1) = 0, there is
x 2 ∈ (0, x 1 ) such that f 0 (x 2 ) = 0, f 0 (x) < 0 for x ∈ (0, x 2 ) and f 0 (x) > 0 for x ∈ (x 2 , 1).
Therefore, f is decreasing on (0, x 2 ] and increasing on [x 2 , 1]. Since lim x→0 f (x) = 0
and f (1) = 0, it follows that f (x) ≤ 0 for x ∈ (0, 1]. The proof is completed. The
equality holds for a = b = 1.
2c c d d ≥ c 2 + d 2 .
(1 + x)1+x (1 − x)1−x ≥ 1 + x 2 ,
We have
2x
f 0 (x) = ln(1 + x) − ln(1 − x) − ,
1 + x2
1 1 2(1 − x 2 ) 8x 2
f 00 (x) = + − = .
1 + x 1 − x (1 + x 2 )2 (1 − x 2 )(1 + x 2 )2
Since f 00 (x) ≥ 0 for x ∈ [0, 1), it follows that f 0 is increasing, f 0 (x) ≥ f 0 (0) = 0, f (x)
is increasing, hence f (x) ≥ f (0) = 0. The proof is completed. The equality holds for
a = b.
a2b + b2a ≤ 1 + a b
and
2a a b b ≥ a2 + b2 .
Therefore, it suffices to show that
1
a2 + b2 ≥ 1 + a b + (a − b)2 .
2
which is an identity. The equality holds for a = b = 1.
2a a b b ≥ a2 + b2 .
2x x ≥ x 2 + 1, x > 0.
We have
2x
f 0 (x) = ln x + 1 − ,
x2 +1
x 2 (x + 1)2 + (x − 1)2
f 00 (x) = .
x(x 2 + 1)2
Since f 00 (x) > 0 for x > 0, f 0 is strictly increasing. Since f 0 (1) = 0, it follows that
f 0 (x) < 0 for x ∈ (0, 1) and f 0 (x) > 0 for x ∈ (1, ∞). Therefore, f is decreasing on
(0, 1] and increasing on [1, ∞), hence f (x) ≥ f (1) = 0 for x > 0. This completes the
proof. The equality holds for a = b = 1.
2a a b b ≥ a2 + b2 .
where
Also, we have
B = [a b(a + b) − 3a b + 1]2 + C,
Symmetric Nonrational Inequalities 151
where
and
C = (a b − 1)2 + 2D,
where
We have
2(2x 3 − 3x 2 + 4x − 1)
f 0 (x) = 1 + ln x − ,
g(x)
x 8 + 6x 4 − 32x 3 + 48x 2 − 32x + 9 (x − 1)2 h(x)
f 00 (x) = = ,
g 2 (x) g 2 (x)
where
h(x) = x 6 + 2x 5 + 3x 4 + 4x 3 + 11x 2 − 14x + 9.
Since
h(x) > 7x 2 − 14x + 7 = 7(x − 1)2 ≥ 0,
we have f 00 (x) ≥ 0, hence f 0 is strictly increasing on (0, ∞). Since f 0 (1) = 0, it follows
that f 0 (x) < 0 for x ∈ (0, 1) and f 0 (x) > 0 for x ∈ (1, ∞). Therefore, f is decreasing
on (0, 1] and increasing on [1, ∞), hence f (x) ≥ f (1) = 0 for x > 0.
152 Vasile Cîrtoaje
First Solution. Using the substitution a = b x, where x > 0, the inequality becomes as
follows:
bx+b
b x + b2 2
2 2
bx b
(b x) b ≥ ,
2
x+1
b x + b2 2
2 2
x
(b x) b ≥ ,
2
x+1
+
2 2
x 1
b x+1 x x ≥ b x+1 ,
2
x+1
x +1 2
2
x
x ≥ .
2
It is true if f (x) ≥ 0 for all x > 0, where
x 1 x2 + 1
f (x) = ln x − ln .
x +1 2 2
We have
1 1 x g(x)
f 0 (x) = ln x + − 2 = ,
(x + 1)2 x + 1 x + 1 (x + 1)2
where
x2 − 1
g(x) = ln x − .
x2 + 1
Since
(x 2 − 1)2
g 0 (x) = ≥ 0,
x(x 2 + 1)2
g is strictly increasing, therefore g(x) < 0 for x ∈ (0, 1), g(1) = 0, g(x) > 0 for x ∈
(1, ∞). Thus, f is decreasing on (0, 1] and increasing on [1, ∞), hence f (x) ≥ f (1) =
0. This completes the proof. The equality holds for a = b.
Second Solution. Write the inequality in the form
a+b a2 + b2
a ln a + b ln b ≥ log .
2 2
Without loss of generality, consider that a + b = 2k, k > 0, and denote
a = k + x, b = k − x, 0 ≤ x < k.
Symmetric Nonrational Inequalities 153
1 1 2k(x 2 − k2 )
f 00 (x) = + +
k+x k−x (x 2 + k2 )2
8k2 x 2
= 2 .
(k − x 2 )(x 2 + k2 )2
Since f 00 (x) ≥ 0 for x ≥ 0, f 0 is increasing, hence f 0 (x) ≥ f 0 (0) = 0. Therefore, f is
increasing on [0, k), hence f (x) ≥ f (0) = 0.
3 − a2 3 − b2
2 2
(a + b ) + ≤ 4,
1 + a2 1 + b2
which is equivalent to
a2 b2 (2 + a2 + b2 ) + 2 − (a2 + b2 ) − (a2 + b2 )2 ≥ 0,
(2 + a2 + b2 )(1 − a2 )(1 − b2 ) ≥ 0.
1 + x2
To prove the inequality x 2x ≥ for 0 < x ≤ 1 , we show that f (x) ≥ 0, where
3 − x2
1 1
f (x) = x ln x + ln(3 − x 2 ) − ln(1 + x 2 ).
2 2
We have
x x
f 0 (x) = 1 + ln x − − ,
3− x 2 1 + x2
1 3 + x2 1 − x2
f 00 (x) = − −
x (3 − x 2 )2 (1 + x 2 )2
(1 − x)(9 + 6x − x 3 ) 1 − x2
= − .
x(3 − x)2 (1 + x 2 )2
We will show that f 00 (x) > 0 for 0 < x < 1. This is true if
9 + 6x − x 3 1+ x
− > 0.
x(3 − x)2 (1 + x 2 )2
Indeed,
9 + 6x − x 3 1+ x 9 1+ x 1
− > − = > 0.
x(3 − x)2 (1 + x )
2 2 9x x(1 + x)2 1+ x
Since f 00 (x) > 0, f 0 is strictly increasing on (0, 1]. Since f 0 (1) = 0, it follows that
f 0 (x) < 0 for x ∈ (0, 1), f is strictly decreasing on (0, 1], hence f (x) ≥ f (1) = 0. This
completes the proof. The equality holds for a = b = 1.
Symmetric Nonrational Inequalities 155
a b b a + 2 ≥ 3a b.
We have
1− x 1+ x 6x
f 0 (x) = − ln(1 + x) + ln(1 − x) + − + ,
1 + x 1 − x 1 − 3x 2
1 00 −1 2(x 2 + 1) 3(3x 2 + 1)
f (x) = − + .
2 1 − x 2 (1 − x 2 )2 (1 − 3x 2 )2
1
Making the substitution t = x 2 , 0 ≤ t < , we get
3
1 00 3(3t + 1) t +3 4t(5 − 9t)
f (x) = − = > 0.
2 (3t − 1)2 (t − 1)2 (t − 1)2 (3t − 1)2
Therefore, f 0 (x) is strictly increasing, f 0 (x) ≥ f 0 (0) = 0, f (x) is strictly increasing, and
hence f (x) ≥ f (0) = 0. This completes the proof. The equality holds for a = b = 1.
Solution (by Vo Quoc Ba Can). Without loss of generality, assume that a ≥ b. Using the
substitution
c = a − b,
we need to show that
(b + c)−c + b c + c 2 ≤ 2
for
0 ≤ b ≤ 1 − c, 0 ≤ c ≤ 1.
156 Vasile Cîrtoaje
We claim that f 0 (x) > 0 for x > 0. Then, f (x) is strictly increasing on [0, 1 − c], hence
f (x) ≤ f (1 − c) = (1 − c)c − (1 − c 2 ).
f (x) ≤ 1 + c(−c) − (1 − c 2 ) = 0.
Since
c[(x + c)1+c − x 1−c ]
f 0 (x) = ,
(x + c)1+c x 1−c
we have f 0 (x) > 0 for x > 0 if and only if
1−c
x + c > x 1+c .
2c
Let d = 1−c . Using the weighted AM-GM inequality, we have
1 1+c
1 − c d 1+d 1 + c 1 − c 1−c 1−c
1−c
x +c =1· x +d · ≥ (1 + d) x = x 1+c .
2 2 2 2
c = a − b.
We have
x x(3x 2 + 6x − 4)
f 0 (x) = − ln 1 + + ,
2 (2 + x)(16 − 3x 2 )
g(x)
f 00 (x) = ,
(2 + x)2 (16 − 3x 2 )2
where
g(x) = −9x 5 − 18x 4 + 168x 3 + 552x 2 + 128x − 640.
Since g(x 1 ) = 0 for x 1 ≈ 0, 88067, g(x) < 0 for x ∈ [0, x 1 ) and g(x) > 0 for x ∈ (x 1 , 2],
f 0 is strictly decreasing on [0, x 1 ] and strictly increasing on [x 1 , 2]. Since f 0 (0) = 0
5
and f 0 (2) = − ln 2 + > 0, there is x 2 ∈ (x 1 , 2) such that f 0 (x 2 ) = 0, f 0 (x) < 0 for
2
x ∈ (0, x 2 ), and f 0 (x) > 0 for x ∈ (x 2 , 2]. Therefore, f is decreasing on [0, x 2 ] and
increasing on [x 2 , 2]. Since f (0) = f (2) = 0, it follows that f (x) ≤ 0 for x ∈ [0, 2].
Lemma 2. If 0 ≤ x ≤ 2, then
x x 1 2
1− + x ≤ 1,
2 4
with equality for x = 0 and x = 2.
Proof. We need to show that f (x) ≤ 0, where
1 2
x
f (x) = x ln 1 − − ln 1 − x .
2 4
We have
x x2
f 0 (x) = ln 1 − − ,
2 4 − x2
−1 8x
f 00 (x) = − .
2 − x (4 − x 2 )2
Since f 00 < 0 for 0 ≤ x < 2, f 0 is strictly decreasing, hence f 0 (x) ≤ f 0 (0) = 0, f is
strictly decreasing, therefore f (x) ≤ f (0) = 0 for x ∈ [0, 2].
1
Conjecture. If a, b are nonnegative real numbers such that a + b = , then
4
a2(b−a) + b2(a−b) ≤ 2.
a b−a + b a−b ≤ 2.
Since the case a − b = 0 is trivial, consider further that 0 < a − b < 2 and use the
substitution
c = a − b.
So, we need to show that
a−c + (a − c)c ≤ 2
for
c
0 < c < 2, c ≤ a ≤2+ .
2
Equivalently, we need to show that f (x) ≤ 0 for
h ci
x ∈ c, 2 + , 0<c<2,
2
where
f (x) = x −c + (x − c)c − 2.
The derivative
c[x 1+c − (x − c)1−c ]
f 0 (x) =
x 1+c (x − c)1−c
has the same sign as
We have
c(2x − 1 − c)
g 0 (x) = .
x(x − c)
5
Case 1: c = 1. We need to show that f (x) ≤ 0 for x ∈ 1, , where
2
x 2 − 3x + 1
f (x) = .
x
Indeed, we have
(x − 1)(2x − 5) + (x − 3)
f (x) = < 0.
2x
160 Vasile Cîrtoaje
1+c 1+c
c
it follows that g(x) is decreasing on c, , and increasing on , 2 + . Then,
2 2 2
1+c 1+c 1−c
g(x) ≥ g = (1 + c) ln − (1 − c) ln ,
2 2 2
and it suffices to show that
1+c 1−c
1+c 1−c
> .
2 2
This inequality follows from Bernoulli’s inequality, as follows
1+c
1+c 1 − c 1+c (1 + c)(1 − c) 1 + c 2
= 1− >1− =
2 2 2 2
Symmetric Nonrational Inequalities 161
and 1−c
1−c 1 + c 1−c (1 − c)(1 + c) 1 + c 2
= 1− <1− = .
2 2 2 2
h ci
Since g(x) > 0 involves f 0 (x) > 0, it follows that f (x) is strictly increasing on c, 2 + ,
2
and hence c
f (x) ≤ f 2 + .
2
c
So, we need to show that f 2 + ≤ 0 for 0 < c < 2, which follows immediately from
2
Lemma 1.
The proof is completed. The equality holds for a = b.
Lemma 1. If 0 ≤ c ≤ 2, then
c −c c c
2+ + 2− ≤ 2.
2 2
Proof. According to Lemma 2, the following inequalities hold for c ∈ [0, 2]:
c −c c3
2+ ≤ 1 − c ln 2 + ,
2 9
c c c3
2− ≤ 1 + c ln 2 − .
2 9
Summing these inequalities, the desired inequality follows.
Lemma 2. If
−2 ≤ x ≤ 2,
then
x x x3
2− ≤ 1 + x ln 2 − .
2 9
Proof. We have
ln 2 ≈ 0.693 < 7/9.
If x ∈ [0, 2], then
x3 x3 8
1 + x ln 2 − ≥1− ≥ 1 − > 0.
9 9 9
Also, if x ∈ [−2, 0], then
x3 7x x 3 8 + 7x − x 3
1 + x ln 2 − ≥1+ − >
9 9 9 9
2(x + 2)2 + (−x)(x + 1)2
= > 0.
9
162 Vasile Cîrtoaje
x3
x
f (x) = ln 1 + d x − − x ln 2 − , d = ln 2.
9 2
We have
9d − 3x 2 x x
f 0 (x) = + − ln 2 − .
9 + 9d x − x 3 4 − x 2
Since f (0) = 0, it suffices to show that f 0 (x) ≤ 0 for x ∈ [−2, 0], and f 0 (x) ≥ 0 for
x ∈ [0, 2]; that is, x f 0 (x) ≤ 0 for x ∈ [−2, 2]. According to Lemma 3, the inequality
x f 0 (x) ≤ 0 is true if x g(x) ≥ 0, where
9d − 3x 2 1 2
x x
g(x) = + − d− − x .
9 + 9d x − x 3 4 − x 4 32
We have
9d − 3x 2
1 2
x x
g(x) = −d + + + x
9 + 9d x − x 3 4 − x 4 32
2
d x − 3x − 9d 2 64 − 4x − x 2
=x + ,
9 + 9d x − x 3 32(4 − x)
hence
x 2 g1 (x)
x g(x) = ,
32(4 − x)(9 + 9d x − x 3 )
where
it follows that h(x) is decreasing. Since h(0) = 0, we have h(x) ≥ 0 for x ≤ 0, and
h(x) ≤ 0 for x ∈ [0, 4); that is, xh(x) ≤ 0 for x < 4.
1
P 1.108. Let a, b be positive real numbers such that a + b = 2. If k ≥ , then
2
kb ka
a a b b ≥ 1.
We have
2k(1 + x 2 ) 1 1
f 0 (x) = − k ln(1 − x 2 ) + +
1− x 2 (1 + x) ln(1 + x) (1 − x) ln(1 − x)
2k 1 1
> + +
1− x 2 (1 + x) ln(1 + x) (1 − x) ln(1 − x)
1 1 1
≥ + +
1− x 2 (1 + x) ln(1 + x) (1 − x) ln(1 − x)
g(x)
= ,
(1 − x ) ln(1 + x) ln(1 − x)
2
where
g(x) = ln(1 + x) ln(1 − x) + (1 + x) ln(1 + x) + (1 − x) ln(1 − x).
If g(x) ≤ 0, then f 0 (x) > 0, f (x) is strictly increasing, and hence
Since
1+ x
h0 (x) = ln > 0,
1− x
h(x) is strictly increasing, h(x) > h(0) = 0, g 0 (x) < 0, g(x) is strictly decreasing, and
hence g(x) < g(0) = 0. This completes the proof. The equality holds for a = b = 1.
Solution. For a = b = 1, the equality holds. In what follows, we will assume that
a > 1 > b. Taking logarithms of both sides, the inequality becomes in succession
p p
a ln a + b ln b ≥ 0,
p p
a ln a ≥ b(− ln b),
1 1
ln a + ln ln a ≥ ln b + ln(− ln b).
2 2
Setting a = 1 + x and b = 1 − x, we need to show that f (x) ≥ 0 for 0 < x < 1, where
1 1
f (x) = ln(1 + x) − ln(1 − x) + ln ln(1 + x) − ln(− ln(1 − x)).
2 2
We have
1 1 1
f 0 (x) = + + .
1− x 2 (1 + x) ln(1 + x) (1 − x) ln(1 − x)
As shown in the proof of the preceding P 1.108, we have f 0 (x) > 0. Therefore, f (x) is
strictly increasing and
f (x) > lim f (x) = 0.
x→0
1
1 − a a+1 b b+1 ≥ (1 − a b)2 .
3
(Vasile Cîrtoaje, 2010)
Symmetric Nonrational Inequalities 165
a−a + b−b ≤ 2.
0 < b ≤ 1 ≤ a < 2,
which is equivalent to
1−a 1−b
a 2 ≥b 2 ,
166 Vasile Cîrtoaje
1−b
(a b) 2 ≤ 1,
a b ≤ 1.
The last inequality follows immediately from the AM-GM inequality
1
ab ≤ (a + b)2 = 1.
4
The equality holds for a = b = 1.
a2b + b2a ≥ a b + b a ≥ a2 b2 + 1.
a b + b a − a2 b2 − 1 = (a b − 1) + (b a − 1) + 1 − a2 b2
b(a b + 1)(a − 1) a(a b + 1)(b − 1)
≥ + + 1 − a2 b2
2 2
= (ab + 1)(a b − 1) + 1 − a2 b2 = 0
and
a2b + b2a − a b − b a = a b (a b − 1) + b a (b a − 1)
a b b(a b + 1)(a − 1) b a a(a b + 1)(b − 1)
≥ +
2 2
a b(a b + 1)(a − b)(a b−1 − b a−1 )
= .
4
On the valid assumption a ≥ b, we only need to show that a b−1 ≥ b a−1 , which is
equivalent to
b−a a−b a−b
a 2 ≥b 2 , 1 ≥ (a b) 2 , 1 ≥ a b, (a − b)2 ≥ 0.
Symmetric Nonrational Inequalities 167
For both inequalities, the equality holds when a = b = 1, when a = 0 and b = 2, and
when a = 2 and b = 0.
Lemma. If x, y ∈ [0, 2] such that (1 − x)(1 − y) ≤ 0, then
y(x y + 3 − x − y)(x − 1)
xy −1≥ ,
2
with equality for x = 1, for y = 0, for y = 1, and for x = 0 and y = 2.
Proof. For y = 0, y = 1 and y = 2, the inequality is an identity. For fixed y ∈ (0, 1) ∪
(1, 2), let us define
y(x y + 3 − x − y)(x − 1)
f (x) = x y − 1 − .
2
We have
x y + 3 − x − y (x − 1)( y − 1)
0
f (x) = y x y−1
− − ,
2 2
f 00 (x) = y( y − 1)(x y−2 − 1).
Since f 00 (x) ≥ 0 for x ∈ (0, 2], f 0 is increasing. There are two cases to consider.
Case 1: x ≥ 1 > y. We have f 0 (x) ≥ f 0 (1) = 0, f (x) is increasing, hence f (x) ≥ f (1) =
0.
Case 2: y > 1 ≥ x. We have f 0 (x) ≤ f 0 (1) = 0, f (x) is decreasing, hence f (x) ≥
f (1) = 0.
a3b + b3a ≤ 2.
x2 x3 x2 x3
f (x) = e3(1−x)(x− 2 + 3 ) + e−3(1+x)(x+ 2 + 3 ) .
168 Vasile Cîrtoaje
Since f (0) = 2, it suffices to show that f 0 (x) ≤ 0 for x ∈ [0, 1). From
15 2 9x 2 5x 3 4
f 0 (x) =(3 − 9x + x − 4x 3 )e3x− 2 + 2 −x
2
15 2 9x 2 5x 3 4
−(3 + 9x + x + 4x 3 )e−3x− 2 − 2 −x ,
2
it follows that f 0 (x) ≤ 0 is equivalent to
3 6 − 18x + 15x 2 − 8x 3
e−6x−5x ≥ .
6 + 18x + 15x 2 + 8x 3
For the nontrivial case 6 − 18x + 15x 2 − 8x 3 > 0, we rewrite this inequality as g(x) ≥ 0,
where
Since g(0) = 0, it suffices to show that g 0 (x) ≥ 0 for x ∈ [0, 1). From
1 0 (6 + 8x 2 ) − 10x (6 + 8x 2 ) + 10x
g (x) = −2 − 5x 2 + + ,
3 6 + 15x 2 − (18x + 8x 3 ) 6 + 15x 2 + (18x + 8x 3 )
it follows that g 0 (x) ≥ 0 is equivalent to
Since
This reduces to 6x 2 +200x 4 ≥ 0, which is clearly true. The equality holds for a = b = 1.
(1 + x)3(1−x) + (1 − x)3(1+x) + x 4 ≤ 2.
1
(1 + x 1−x ≤ (1 + x)2 (2 − x 2 )(2 − 2x + x 2 ),
4
1
(1 − x 1+x ≤ (1 − x)2 (2 − x 2 )(2 + 2x + x 2 ).
4
Therefore, it suffices to show that
which is equivalent to
t 3 − 8t 2 + 31t − 34 < 0
and
17t 3 − 38t 2 + 16t + 8 > 0
for all t ∈ [0, 1]. Indeed, we have
Lemma. If −1 ≤ t ≤ 1, then
1
(1 + t)1−t ≤ (1 + t)2 (2 − t 2 )(2 − 2t + t 2 ),
4
with equality for t = −1, t = 0 and t = 1.
170 Vasile Cîrtoaje
We have
2t 2(t − 1)
f 0 (t) = 1 + ln(1 + t) − + ,
2 − t 2 2 − 2t + t 2
t 2 g(t)
f 00 (t) = ,
(1 + t)(2 − t 2 )2 (2 − 2t + t 2 )2
where
g(t) = t 6 − 8t 5 + 12t 4 + 8t 3 − 20t 2 − 16t + 16.
Case 1: 0 ≤ t ≤ 1. From
it follows that g is strictly decreasing on [0, 1]. Since g(0) = 16 and g(1) = −7, there
exists a number c ∈ (0, 1) such that g(c) = 0, g(t) > 0 for 0 ≤ t < c and g(t) < 0
for c < t ≤ 1. Therefore, f 0 is strictly increasing on [0, c] and strictly decreasing on
[c, 1]. From f 0 (0) = 0 and f 0 (1) = ln 2 − 1 < 0, it follows that there exists a number
d ∈ (0, 1) such that f 0 (d) = 0, f 0 (t) > 0 for 0 < t < d and f 0 (t) < 0 for c < t ≤ 1. As
a consequence, f is strictly increasing on [0, d] and strictly decreasing on [d, 1]. Since
f (0) = 0 and f (1) = 0, we have f (t) ≥ 0 for 0 ≤ t ≤ 1.
Case 2: −1 < t ≤ 0. From
it follows that f 0 is strictly increasing on (−1, 0]. Since f 0 (0) = 0, we have f 0 (t) < 0
for −1 < t < 0, hence f is strictly decreasing on (−1, 0]. From f (0) = 0, it follows that
f (t) ≥ 0 for −1 < t ≤ 0.
Conjecture. If a, b are nonnegative real numbers such that a + b = 2, then
a−b 2
a3b + b3a + ≥ 2.
2
0 < a ≤ 1 ≤ b < 2.
3 3
There are two cases to consider: 0 < a ≤ and ≤ a ≤ 1.
5 5
3 7
Case 1: 0 < a ≤ . From a + b = 2, we get ≤ b < 2. Let
5 5
3
f (x) = x x , 0 < x < 2.
Since
3
f 0 (x) = 3x x −2 (1 − ln x) > 0,
7 7
f (x) is increasing on , 2 , and hence f (b) ≥ f ; that is,
5 5
15/7
3 7
bb ≥ .
5
Using Bernoulli’s inequality gives
15/7
7 7 2 8/7 7 16 51
3
b ≥
b = 1+ > 1+ = > 2.
5 5 5 5 35 25
hence
3 3
a a + b b > 2.
3
Case 2: ≤ a ≤ 1. By Lemma below, we have
5
3
2a a ≥ 3 − 15a + 21a2 − 7a3
and
3
2b b ≥ 3 − 15b + 21b2 − 7b3 .
Summing these inequalities, we get
3 3
2 a a + b b ≥ 6 − 15(a + b) + 21(a2 + b2 ) − 7(a3 + b3 )
= 6 − 15(a + b) + 21(a + b)2 − 7(a + b)3 = 4.
From
h0 (x) = 3(−5 + 14x − 7x 2 ),
s s s
2 2 2
it follows that h(x) is increasing on 1 − ,1 + , and decreasing on 1 + ,∞ .
7 7 7
3
Then, it suffices to show that f ≥ 0 and f (2) ≥ 0. Indeed
5
3 6
f = , f (2) = 1.
5 125
Write now the desired inequality as f (x) ≥ 0, where
3 3
f (x) = ln 2 + ln x − ln(3 − 15x + 21x 2 − 7x 3 ), ≤ x ≤ 2.
x 5
We have
x2 0 x 2 (7x 2 − 14x + 5)
f (x) = g(x), g(x) = 1 − ln x + ,
3 3 − 15x + 21x 2 − 7x 3
g1 (x)
g 0 (x) = ,
(3 − 15x + 21x 2 − 7x 3 )2
where
In addition,
g2 (x) = 11x 5 + 3g3 (x), g3 (x) = −20x 5 + 49x 4 + 21x 3 − 56x 2 + 24x − 3,
g3 (x) = (4x − 1)g4 (x), g4 (x) = −5x 4 + 11x 3 + 8x 2 − 12x + 3,
g4 (x) = x 5 + g5 (x), g5 (x) = −6x 4 + 11x 3 + 8x 2 − 12x + 3,
g5 (x) = (2x − 1)g6 (x), g6 (x) = −3x 3 + 4x 2 + 6x − 3,
g6 (x) = 1 + (2 − x)(3x 2 + 2x − 2).
Therefore, we get in succession g6 (x) > 0, g5 (x) > 0, g4 (x) > 0, g3 (x) > 0, g2 (x)
> 0,
3
0
g1 (x) ≥ 0, g (x) ≥ 0, g(x) is increasing. Since g(1) = 0, we have g(x) < 0 on ,1
5
3
and g(x) > 0 on (1, 2]. Then, f (x) is decreasing on , 1 and increasing on [1, 2],
5
hence f (x) ≥ f (1) = 0.
Symmetric Nonrational Inequalities 173
e−10u (u0 + v 0 ) ≥ u0 − v 0
For the nontrivial case u0 − v 0 > 0, we rewrite this inequality as g(x) ≥ 0, where
u00 + v 00 u00 − v 00
g 0 (x) = −10u0 + − 0 ,
u0 + v 0 u − v0
where
62 3 27 4
u00 = 14x + x , v 00 = 5 + 17x 2 + x .
3 2
Thus, the inequality g 0 (x) ≥ 0 is equivalent to
a1 t + a2 t 2 + a3 t 3 + a4 t 4 + a5 t 5 + a6 t 6 + a7 t 7 ≥ 0,
where t = x 2 , 0 ≤ t < 1, and
50t 2 (1 − t)(6 + 9t − t 2 − 5t 3 ) ≥ 0,
13
Case 2: 0 ≤ x ≤ . According to Lemma below, it suffices to show that f (x) ≤ 2,
20
where
11 3 1 4
2
+ 12 x −2 x 2
+ 11 3 1 4
12 x + 2 x )
f (x) = e2x−2x + e−(2x+2x .
If f 0 (x) ≤ 0, then f (x) is decreasing, hence f (x) ≤ f (0) = 2. Since
11 2 2 11 3 1 4
f 0 (x) = (2 − 4x + x − 2x 3 )e2x−2x + 12 x − 2 x
4
11 2 2 11 3 1 4
−(2 + 4x + x + 2x 3 )e−(2x+2x + 12 x + 2 x ) ,
4
0
the inequality f (x) ≤ 0 is equivalent to
11
x3 8 − 16x + 11x 2 − 8x 3
e−4x− 6 ≥ .
8 + 16x + 11x 2 + 8x 3
For the non-trivial case 8 − 16x + 11x 2 − 8x 3 > 0, rewrite this inequality as g(x) ≥ 0,
where
11 3
g(x) = −4x − x − ln(8 − 16x + 11x 2 − 8x 3 ) + ln(8 + 16x + 11x 2 + 8x 3 ).
6
176 Vasile Cîrtoaje
1
(16 + 24x 2 )(8 + 11x 2 ) − 22x(16x + 8x 3 ) ≥ (8 + 11x 2 )[(8 + 11x 2 )2 − (16x + 8x 3 )2 ].
4
Since
This reduces to 77x 4 ≥ 0. The proof is completed. The equality holds for a = b = 1.
13
Lemma. If −1 < t ≤ , then
20
p 11 3 1 4
1 − t ln(1 + t) ≤ t − t 2 + t − t .
24 4
Proof. We consider two cases.
13
Case 1: 0 ≤ t ≤ . We can prove the desired inequality by multiplying the following
20
inequalities
p 1 1 1 3
1 − t ≤ 1 − t − t2 − t ,
2 8 16
1 1 1 1
ln(1 + t) ≤ t − t 2 + t 3 − t 4 + t 5 ,
2 3 4 5
1 1 1 3 1 1 1 1 11 3 1 4
1 − t − t2 − t t − t2 + t3 − t4 + t5 ≤ t − t2 + t − t .
2 8 16 2 3 4 5 24 4
1 1 2 1 3 1
f (t) = ln 1 − t − t − t − ln(1 − t).
2 8 16 2
Since
1 8 + 4t + 3t 2 5t 3
f 0 (t) = − = ≥ 0,
2(1 − t) 16 − 8t − 2t 2 − t 3 2(1 − t)(16 − 8t − 2t 2 − t 3 )
Symmetric Nonrational Inequalities 177
1 1 1 2
f (t) = ln(1 − t) − ln 1 − t − t .
2 2 8
Since
−1 2(2 + t) −3t 2
f 0 (t) = + = ≤ 0,
2(1 − t) 8 − 4t − t 2 2(1 − t)(8 − 4t − t 2 )
f (t) is decreasing, and hence f (t) ≥ f (0) = 0.
The second inequality is equivalent to f (t) ≥ 0, where
1 2 1 3 1 4
f (t) = t − t + t − t − ln(1 + t).
2 3 4
178 Vasile Cîrtoaje
Since
1 −t 4
f 0 (t) = 1 − t + t 2 − t 3 − = ≤ 0,
1+ t 1+ t
f (t) is decreasing, and hence f (t) ≥ f (0) = 0.
The third inequality reduces to the obvious inequality
t 4 (10 − 8t − 3t 2 ) ≥ 0.
b(3 − a b) b(1 + b − a b)
a b+1 + b a+1 − 2 ≥ −b(a − 1)2 + (a − 1)2 − (a − 1)4 .
2 6
Since
b(3 − a b) b
−b(a − 1)2 + (a − 1)2 = (a − 1)4 ,
2 2
Symmetric Nonrational Inequalities 179
we have
b b(1 + b − a b)
a b+1 + b a+1 − 2 ≥ (a − 1)4 − (a − 1)4
2 6
a b(1 + b) ab a b(1 − a b)2
= (a − 1)4 ≥ (a − 1)4 = .
6 6 6
The equality holds for a = b = 1, for a = 2 and b = 0, and for a = 0 and b = 2.
(b) To prove the right inequality we apply Lemma 2 below. For x = a and k = b,
we have
(b + 1)b (b + 1)b(b − 1)
a b+1 ≤ 1 + (b + 1)(a − 1) + (a − 1)2 + (a − 1)3
2 6
(b + 1)b(b − 1)(b − 2)
+ (a − 1)4 ,
24
b(b + 1) b(b + 1) a b(b + 1)
a b+1 ≤ 1 + (b + 1)(a − 1) + (a − 1)2 − (a − 1)4 + (a − 1)5 .
2 6 24
Also, for x = b and k = a, we have
a2 + b2 + 2 a2 + b2 + 2 ab
a b+1 + b a+1 − 2 ≤ −2(a − 1)2 + (a − 1)2 − (a − 1)4 − (a − 1)6
2 6 12
a2 + b2 − 2 a2 + b2 + 2 a2 + b2 + 2
≤ (a − 1)2 − (a − 1)4 = (a − 1)4 − (a − 1)4
2 6 6
ab a b(1 − a b)2
= (a − 1)4 = .
3 3
The equality holds for a = b = 1, for a = 2 and b = 0, and for a = 0 and b = 2.
Lemma 1. If x ≥ 0 and 0 ≤ k ≤ 1, then
1 k(1 − k)
f 0 (x) = x k − 1 − k(x − 1) + (x − 1)2 ,
1+k 2
1
f 00 (x) = x k−1 − 1 + (1 − k)(x − 1),
k(1 + k)
1
f 000 (x) = −x k−2 + 1.
k(1 + k)(1 − k)
(k + 1)k (k + 1)k(k − 1)
x k+1 ≤ 1 + (k + 1)(x − 1) + (x − 1)2 + (x − 1)3
2 6
(k + 1)k(k − 1)(k − 2)
+ (x − 1)4 ,
24
with equality for x = 1, for k = 0, for k = 1, for k = 2.
Proof. For k = 0, k = 1 and k = 2, the inequality is an identity. For fixed k, k ∈
(0, 1) ∪ (1, 2), let us define
(k + 1)k (k + 1)k(k − 1)
f (x) = x k+1 − 1 − (k + 1)(x − 1) − (x − 1)2 − (x − 1)3
2 6
(k + 1)k(k − 1)(k − 2)
− (x − 1)4 .
24
We need to show that f (x) ≤ 0. We have
Case 1: x ≥ 1, 0 < k < 1. Since f (4) (x) ≤ 0, f 000 (x) is decreasing, f 000 (x) ≤ f 000 (1) =
0, f 00 is decreasing, f 00 (x) ≤ f 00 (1) = 0, f 0 is decreasing, f 0 (x) ≤ f 0 (1) = 0, f is
decreasing, f (x) ≤ f (1) = 0.
Case 2: 0 ≤ x ≤ 1, 1 < k < 2. Since f (4) ≤ 0, f 000 is decreasing, f 000 (x) ≥ f 000 (1) = 0, f 00
is increasing, f 00 (x) ≤ f 00 (1) = 0, f 0 is decreasing, f 0 (x) ≥ f 0 (1) = 0, f is increasing,
f (x) ≤ f (1) = 0.
a2b + b2a ≤ 1.
1
0≤ b≤ ≤ a ≤ 1.
2
Applying Lemma 1 below for c = 2b, 0 ≤ c ≤ 1, we get
which is equivalent to
a2b ≤ 1 − 4a b2 − 2a b(a − b) ln a.
Similarly, applying Lemma 2 below for d = 2a − 1, d ≥ 0, we get
which is equivalent to
b2a ≤ 4a b2 + 2a b(a − b) ln a.
Adding up these inequalities, the desired inequality follows. The equality holds for
a = b = 1/2, for a = 0 and b = 1, and for a = 1 and b = 0.
Lemma 1. If 0 < a ≤ 1 and c ≥ 0, then
(1 − c)e x + c ≥ e(1−c)x ,
(c − 1)e x + e(1−c)x ≥ c,
which yields
f 0 (x) = (c − 1)[(c − 1)e x + e(1−c)x − c] ≥ 0.
1 − d + d ln(b + d) ≥ 1 − d + d ln d > 0,
Since
d 2 (e x − 1 − x)
f 0 (x) = ≥ 0,
(1 − d − d x)(1 − d e x )
f is increasing, and hence
To complete the proof, we only need to show that −d 2 + d(1 + d) ln(1 + d) ≥ 0; that is,
(1 + d) ln(1 + d) ≥ d.
−d
This inequality follows from e x ≥ 1 + x, where x = .
1+d
Conjecture. If a, b are nonnegative real numbers such that 1 ≤ a + b ≤ 15, then
2a a b b ≥ a2b + b2a .
Solution. Taking into account the inequality a2b + b2a ≤ 1 in the preceding P 1.119, it
suffices to show that
2a a b b ≥ 1.
a−2a + b−2b ≤ 4.
184 Vasile Cîrtoaje
Remark. According to the inequalities in P 1.119 and P 1.121, the following inequality
holds for all positive numbers a, b such that a + b = 1:
1 1
a2b + b2a + ≤ 4.
a2a b2b
Actually, this inequality holds for all a, b ∈ (0, 1]. In this case, it is sharper than the
inequality in P 1.103.
Symmetric Nonrational Inequalities 185
First Solution. We can obtain the desired inequality by summing the inequalities
p p p p
1 − a + 1 − b ≥ c + d,
p p p p
1 − c + 1 − d ≥ a + b.
Since p p Æ
4
1 − a + 1 − b ≥ 2 (1 − a)(1 − b)
and v v
p p tc + d 4 c +d
t 2 2
c+ d ≤2 ≤2 ,
2 2
the former inequality above holds if
c2 + d 2
(1 − a)(1 − b) ≥ .
2
Indeed,
Second Solution. We can obtain the desired inequality by summing the inequalities
p p 1 p p 1
1 − a − a ≥ p (1 − 4a2 ), 1− b− b ≥ p (1 − 4b2 ),
2 2 2 2
p p 1 p p 1
1 − c − c ≥ p (1 − 4c 2 ), 1−d − d ≥ p (1 − 4d 2 ).
2 2 2 2
To prove the first inequality, we write it as
1 − 2a 1
p p ≥ p (1 − 2a)(1 + 2a).
1−a+ a 2 2
1
Case 1: 0 < a ≤ . We need to show that
2
p p p
2 2 ≥ (1 + 2a)( 1 − a + a).
186 Vasile Cîrtoaje
p p p p
Since 1 − a + a ≤ 2[(1 − a) + a] = 2, we have
p p p p
2 2 − (1 + 2a)( 1 − a + a) ≥ 2(1 − 2a) ≥ 0.
1
Case 2: ≤ a < 1. We need to show that
2
p p p
2 2 ≤ (1 + 2a)( 1 − a + a).
p
Since 1 + 2a ≥ 2 2a, it suffices to prove that
Æ
1 ≤ a(1 − a) + a.
Indeed,
p
Æ p p p 1 − a (1 − 2a)
1−a− a(1 − a) = 1 − a ( 1 − a − a) = p p ≤ 0.
1−a+ a
where
1 1 1 1
A = (a + b + c + d)+ + + − 16,
a b c d
1 1 1 1
2 2 2 2
B = (a + b + c + d ) 2 + 2 + 2 + 2 − 16.
a b c d
(Vasile Cîrtoaje, 2004)
Solution. By squaring, the inequality becomes
A2 + 4A ≥ B.
Let us denote
x y z x2 y 2 z2
f (x, y, z) = + + − 3, F (x, y, z) = + + 2 − 3,
y z x y 2 z2 x
where x, y, z > 0. By the AM-GM inequality, it follows that f (x, y, z) ≥ 0 and F (x, y, z) ≥
0. We can check that
and
B = F (a, b, c) + F (b, a, d) + F (c, d, a) + F (d, c, b).
Since
we get
B = f 2 (a, b, c) + f 2 (b, a, d) + f 2 (c, d, a) + f 2 (d, c, b) + 4A.
Therefore,
a1 − 1 a2 an
+p + ··· + p ≥ 0,
3a1 + 1 + 2 3a2 + 1 + 1 3an + 1 + 1
p
a2 an a2 + · · · + an
+ ··· + p ≥p ,
3a2 + 1 + 1 3an + 1 + 1 3a1 + 1 + 2
p
1 1 1 1
a2 p −p + · · · + an p −p ≥ 0.
3a2 + 1 + 1 3a1 + 1 + 2 3an + 1 + 1 3a1 + 1 + 2
The last inequality is clearly true. The equality holds for a1 = 1 and a2 = · · · = an = 0
(or any cyclic permutation).
Second Solution. We use the induction method. For n = 1, the inequality is an equality.
We claim that Æ
3a1 + 1 + 3an + 1 ≥ 3(a1 + an ) + 1 + 1.
p p
188 Vasile Cîrtoaje
a1 = · · · = ak = a, ak+1 = · · · = an = b,
Clearly, this inequality follows by the weighted AM-GM inequality. For n ≥ 3, the equal-
ity holds when a = 0, one of ai is also 0 and the other ai are equal to b.
Lemma. Let 0 ≤ a < b and a1 , a2 , . . . , an ∈ [a, b]. Then, the expression
a1 + a2 · · · + an − n n a1 a2 · · · an
p
For the sake of contradiction, assume that there exits a1 ∈ (a, b) such that
p f (a1 ) > f (a)
p p n
and f (a1 ) > f (b). Let us denote x i = n ai for all i, c = n a and d = b (c < x 1 < d).
From
we get
Analogously, from
we get
1 1 1
+p + ··· + p ≥ 1.
1 + (n2 1 + (n2 − 1)a2 1 + (n2 − 1)an
p
− 1)a1
1 1 1
+p + ··· + p < 1.
1 + (n2 1 + (n2 1 + (n2 − 1)an
p
− 1)a1 − 1)a2
1
xi = p , 0 < x i < 1, i = 1, 2, · · · , n.
1 + (n2 − 1)ai
1 − x i2
Since ai = for all i, we need to show that
(n2 − 1)x i2
x1 + x2 + · · · + x n < 1
implies
(1 − x 12 )(1 − x 22 ) · · · (1 − x n2 ) > (n2 − 1)n x 12 x 22 · · · x n2 .
190 Vasile Cîrtoaje
which follows immediately from the AM-GM inequality. The equality holds for a1 =
a2 = · · · = an = 1.
Symmetric Nonrational Inequalities 191
which yields
n − xi
ai = , 0 < x i < n, i = 1, 2, · · · , n,
(n − 1)x i2
we need to show that
x1 + x2 + · · · + x n < n
implies
(n − x 1 )(n − x 2 ) · · · (n − x n ) > (n − 1)n x 12 x 22 · · · x n2 .
By the AM-GM inequality, we have
x + x + · · · + x n
1 2 n
x1 x2 · · · x n ≤ <1
n
and
v
t x1 x2 · · · x n
n − x i > (x 1 + x 2 + · · · + x n ) − x i ≥ (n − 1) n−1
, i = 1, 2, · · · , n.
xi
Therefore, we get
which involves
(n − 1)2 ci
ai = , 0 < ci < n, i = 1, 2, · · · , n,
(n − ci )2
we need to show that
c1 + c2 + · · · + cn < n
implies
n − c 2 n − c 2 n − c 2
1 2 n
··· > c1 c2 · · · cn .
n−1 n−1 n−1
Clearly, it suffices to show that
c1 + c2 + · · · + cn = n
implies
n − c 2 n − c 2 n − c 2
1 2 n
···
≥ c1 c2 · · · cn .
n−1 n−1 n−1
Popoviciu’s inequality (see the preceding P 1.128) applied to the convex function f (x) =
− ln x, x > 0 gives
n − c n−1 n − c n−1 n − c n−1 c + c + · · · + c n(n−2)
1 2 n 1 2 n
··· ≥ c1 c2 · · · cn ,
n−1 n−1 n−1 n
n − c n−1 n − c n−1 n − c n−1
1 2 n
··· ≥ c1 c2 · · · cn ,
n−1 n−1 n−1
n − c 2 n − c 2 n − c 2 2
1 2 n
··· ≥ (c1 c2 · · · cn ) n−1 .
n−1 n−1 n−1
Thus, it suffices to show that
2
(c1 c2 · · · cn ) n−1 ≥ c1 c2 · · · cn .
194 Vasile Cîrtoaje
(n − 1)[n(a − 1)2 + b2 − 1] ≥ 0,
Solution. We use the induction method. For n = 2, the inequality is equivalent to the
obvious inequality
1
a1 + ≥ 2.
a1
Assume now that the inequality holds for n − 1 numbers, n ≥ 3, and prove that it holds
also for n numbers. Let a1 = min{a1 , a2 , . . . , an }, and denote
a2 + a3 + · · · + an
x= y=
p
, n−1
a2 a3 · · · an ,
n−1
q Æ
f (a1 , a2 , . . . , an ) = (n − 1)(a12 + a22 + · · · + an2 ) + n − n(n − 1) − (a1 + a2 + · · · + an ).
By the AM-GM inequality, we have x ≥ y. We will show that
To prove this inequality, we use the induction hypothesis, written in the homogeneous
form
q Æ
(n − 2)(a22 + a32 + · · · + an2 ) + n − 1 − (n − 1)(n − 2) y ≥ (n − 1)x,
which is equivalent to
a22 + · · · + an2 ≥ (n − 1)A2 ,
where v
tn − 1
A = k x − (k − 1) y, k= .
n−2
So, we need to prove that
q q p
a12 + (n − 1)A2 − a12 + (n − 1) y 2 ≥ n − 1 (x − y).
A2 − y 2 x−y
q q ≥p .
a1 + (n − 1)A2 + a1 + (n − 1) y 2
2 2 n−1
Since x ≥ y and
A2 − y 2 = k(x − y)[k x − (k − 2) y],
we need to show that
k[k x − (k − 2) y] 1
q q ≥p .
a12 + (n − 1)A2 + a1 + (n − 1) y 2
2 n−1
196 Vasile Cîrtoaje
k[k x − (k − 2) y] 1
p ≥p .
y 2 + (n − 1)A2 + n y
p
n−1
We claim that Æ p
y 2 + (n − 1)A2 ≤ k n − 1 [k x − (k − 1) y] .
If this inequality is true, then it is enough to prove that
k[k x − (k − 2) y] 1
p p ≥p .
k n − 1[k x − (k − 1) y] + ny n−1
k[k x − (k − 2) y]
s
n
≥ 1, m= .
k[k x − (k − 1) y] + m y n−1
k[k x − (k − 2) y] k[k x − (k − 2) y]
> = 1.
k[k x − (k − 1) y] + m y k[k x − (k − 1) y] + k y
Using the relation a1 y n−1 = 1, we can write the right inequality in (*) as
Æ
(n − 1)[(n − 1) y 2n + 1] ≥ (n − 1) y n − t y n−1 + 1,
where Æ
t = n− n(n − 1).
This inequality is true if
that is,
2(n − 1)t y 2n−1 − t 2 y 2n−2 − 2(n − 1) y n + 2t k y n−1 + n − 2 ≥ 0.
Since
t 2 = n(2t − 1),
we can write this inequality in the form
2t y n−1 B + C ≥ 0,
where
B = (n − 1) y n − n y n−1 + 1, C = n y 2n−2 − 2(n − 1) y n + n − 2.
Symmetric Nonrational Inequalities 197
and Æ
2n−2
n y 2n−2 + n − 2 ≥ (2n − 2) y n(2n−2) · 1n−2 = 2(n − 1) y n .
The proof is completed. The equality holds for a1 = a2 = · · · = an = 1.
Solution. We use the contradiction method. Assume that the reverse inequality holds,
namely
1 1 1 n
+p +p >p ,
1 + pa1 1 + pa2 1 + pan 1+p
p
1+p
p
p
1 + pai = , 0 < x i < p + 1, i = 1, 2, · · · , n,
p
xi
we need to show that x 1 + x 2 + · · · + x n > n yields
x1 + x2 + · · · + x n = n
yields
1+p 1+p 1+p
−1 − 1 ··· − 1 ≤ pn.
x 12 x 22 x n2
Denoting
p n
1 + p = q, 1<q≤ ,
n−1
198 Vasile Cîrtoaje
n Ç
n
2
X X
(n − 1)2 ai4 + 2n − 1 ≥ ai ;
i=1 i=1
n
n
2
X q X
ai (n − 1)2 ai2 + 2n − 1 ≥ ai .
i=1 i=1
Solution. According to the preceding P 1.132, the following inequality holds for any
real m:
n
X 1
≤ 1.
(n − 1)2 + (2n − 1)aim
Æ
i=1
hence 2
n n
X q X
ai2 (n − 1)2 + (2n − 1)aim ≥ ai .
i=1 i=1
Setting m = −4 and m = −2 gives the desired inequalities. The equality occurs for
a1 = a2 = · · · = an = 1.
X b1k
≥ 1,
b1k + (b2 + · · · + bn )/r k−1
200 Vasile Cîrtoaje
and we see that it suffices to prove it for r = 1; that is, for a1 a2 · · · an = 1. On this
hypothesis, we will show that there exists a positive number p such that
p
a1k a1
≥ p p p.
a1k + a2 + · · · + an a1 + a2 + · · · + an
If this is true, by adding this inequality and the analogous inequalities for a2 , . . . , an , we
get the desired inequality. Write the claimed inequality as
p
a2 + · · · + anp ≥ (a2 · · · an )k−p (a2 + · · · + an ).
For
(n − 1)k + 1
p= , p > 1,
n
this inequality turns into the homogeneous inequality
p−1
p
a2 + · · · + anp ≥ (a2 · · · an ) n−1 (a2 + · · · + an ).
k+1
Jensen’s inequality applied to the convex function f (x) = x 2 yields
P k+1
X k+1 n(n − 1) 2 1≤i< j≤n a1 a2 2
(a1 a2 ) 2 ≥
1≤i< j≤n
2 n(n − 1)
Symmetric Nonrational Inequalities 201
Therefore,
P k+1 P
2 1≤i< j≤n a1 a2 2 2 1≤i< j≤n a1 a2
≥ ,
n(n − 1) n(n − 1)
hence P
X k+1 n(n − 1) 2 1≤i< j≤n a1 a2 X
(a1 a2 ) 2 ≥ = a1 a2 .
1≤i< j≤n
2 n(n − 1) 1≤i< j≤n
and we see that it suffices to prove it for r = 1; that is, for a1 a2 · · · an = 1. On this
hypothesis, we will prove the desired inequality by summing the inequality
p
a1k a1
≤ p p p
a1k + a2 + · · · + an a1 + a2 + · · · + an
which is equivalent to
Adding this inequality and the analogous inequalities for a3 , · · · , an yields the inequality
(*). Thus, the proof is completed. The equality holds for a1 = a2 = · · · = an = 1.
which is equivalent to
X 1
≥ 1.
n − k + 1 + (k − 1)a1
This inequality follows immediately from the AM-HM inequality
X X 1
x1 ≥ n2 ,
x1
for x i = n − k + 1 + (k − 1)ai , i = 1, 2, · · · , n.
1 p
Case 2: k ≥ n − . Let us denote r = n a1 a2 · · · an and bi = ai /r for i = 1, 2, · · · , n.
n−1
Note that r ≥ 1 and b1 b2 · · · bn = 1. The desired inequality becomes
X b1
≤ 1,
b1k r k−1 + b2 + · · · + b n
and we see that it suffices to prove it for r = 1; that is, for a1 a2 · · · an = 1. On this
hypothesis, it suffices to show that
p
(n − 1)a1 a1
+ p p p ≤1
a1k + a2 + · · · + an a1 + a2 + · · · + an
for a suitable real p; then, adding this inequality and the analogous inequalities for
p
a2 , · · · an yields the desired inequality. Let us denote t = n−1 a2 · · · an . By the AM-GM
inequality, we have
p
a2 + · · · an ≥ (n − 1)t, a2 + · · · + anp ≥ (n − 1)t p .
where
q = (n − 1)(k − 1).
Choosing
(n − 1)(k − n − 1)
p= ,
n
the inequality becomes as follows
2 2
−2n −3n
(n − 1)t q (t n − 1) − (t n − 1)(t n + tn + · · · + 1) ≥ 0,
2 2
−2n −3n
(t n − 1)[(t q − t n ) + (t q − t n ) + · · · + (t q − 1)] ≥ 0.
1
The last inequality is clearly true for q ≥ n2 − 2n; that is, for k ≥ n − . Thus, the
n−1
proof is completed. The equality holds for a1 = a2 = · · · = an = 1.
2
−1 − ≤ k < 1,
n−2
then X a1
≥ 1.
a1k + a2 + · · · + an
(Vasile Cîrtoaje, 2006)
p
Solution. Let us denote r = n a1 a2 · · · an and bi = ai /r for i = 1, 2, · · · , n. Note that
r ≥ 1 and b1 b2 · · · bn = 1. The desired inequality becomes
X b1
≥ 1,
b1k /r 1−k + b2 + · · · + bn
and we see that it suffices to prove it for r = 1; that is, for a1 a2 · · · an = 1. On this
hypothesis, by the Cauchy-Schwarz inequality, we have
( a1 )2 ( a1 )2
P P
X a1
≥P = P P .
a1k + a2 + · · · + an a1 (a1k + a2 + · · · + an ) ( a1 )2 + a11+k − a12
P
Case 1: −1 ≤ k < 1. Using Chebyshev’s inequality and the AM-GM inequality yields
X 1 X 1−k X 1+k X X
a12 ≥ a1 a1 ≥ (a1 a2 · · · an )(1−k)/n a11+k = a11+k .
n
2
Case 2: −1 − ≤ k < −1. It is convenient to replace the numbers a1 , a2 , · · · , an
n−1
(n−1)/2 (n−1)/2
by a1 , a2 , · · · , an(n−1)/2 , respectively. So, we need to show that a1 a2 · · · an = 1
involves X X
q
a1n−1 ≥ a1 ,
Symmetric Nonrational Inequalities 205
where
(n − 1)(1 + k)
q= , −1 ≤ q < 0.
2
By the AM-GM inequality, we get
X 1 X n−1 X X 1
a1n−1 = (a2 + · · · + ann−1 ) ≥ a2 · · · an = .
n−1 a1
a11−k + n − 1 a11−k + n − 1
=P ≤P ,
a1 + 2 1≤i< j≤n ai a j a1 + n(n − 1)
P p
hence
a11−k + n(n − 1)
P
X 1
≤ .
a1 + n(n − 1)
P
a1k + a2 + · · · + an
Therefore, it suffices to show that
X X
a11−k ≤ a1 .
206 Vasile Cîrtoaje
X p
n−1 a1
+ p p p − 1 ≤ 0,
a1k + a2 + · · · + an a1 + a2 + · · · + an
where
(n − 1)(k − 1)
p= ≥ 0.
n
To complete the proof, it suffices to show that
p
n−1 a1
≤1− p p p.
a1k + a2 + · · · + an a1 + a2 + · · · + an
Let
x= x > 0.
p
n−1
a1 ,
By the AM-GM inequality, we have
n−1 n−1
a2 + · · · + an ≥ (n − 1) n−1 a2 · · · an = n−1
p =
p
a1 x
and
p
Æ n−1 n−1
a2 + · · · + anp ≥ (a2 · · · an ) p = q p =
n−1
.
n−1
a1 xp
n−1 x (n−1)p
n−1
≤1− n−1
,
x (n−1)k + x x (n−1)p + xp
which is equivalent to
x 1
≤ ,
x (n−1)k+1 +n−1 x np + n − 1
(n−1)k+1
x − x np+1 − (n − 1)(x − 1) ≥ 0,
x q (x n−1 − 1) − (n − 1)(x − 1) ≥ 0,
where
q = (n − 1)(k − 1) + 1 ≥ 1.
Symmetric Nonrational Inequalities 207
and
(x q − 1)(x n−1 − 1) ≥ 0,
(x n−1 − 1) − (n − 1)(x − 1) = (x − 1)[(x n−2 − 1) + (x n−3 − 1) + · · · + (x − 1)] ≥ 0.
This completes the proof. The equality holds for a1 = a2 = · · · = an = 1.
For k = n + 1, this inequality is an equality. Otherwise, for 1 < k < n + 1, we rewrite the
inequality as
n
X 1
≥ 1,
i=1
n − k + 1 + (k − 1)ai
1 1 1
+ + ··· + ≥ 1.
a1k + a2 + · · · + an a1 + a2k + · · · + an a1 + a2 + · · · + ank
X a1 (a k + a k + · · · + a k )
2 3 n
≤ a1k + a2k + · · · + ank ;
a2 + a3 + · · · + an
+ + +
k k k
X a 2 a3 · · · a n
a1 a1k−1 − ≥ 0;
a2 + a3 + · · · + an
X a1 a2 (a k−1 − a k−1 ) + a1 a3 (a k−1 − a k−1 ) + · · · + a1 an (a k−1 − a k−1 )
1 2 1 3 1 n
≥ 0;
a2 + a3 + · · · + an
!
X aik−1 − a k−1
j a k−1
j − aik−1
ai a j + ≥ 0;
1≤i< j≤n
1 − ai 1 − aj
X ai a j (aik−1 − a k−1
j )(ai − a j )
≥ 0.
1≤i< j≤n
(1 − ai )(1 − a j )
Since the last inequality is true for k > 1, the proof is finished. The equality holds for
a1 = a2 = · · · = an .
210 Vasile Cîrtoaje
Appendix A
Glossary
a1 + a2 + · · · + an ≥ n n a1 a2 · · · an ,
p
2. WEIGHTED AM-GM INEQUALITY
Let p1 , p2 , · · · , pn be positive real numbers satisfying
p1 + p2 + · · · + pn = 1.
3. AM-HM (ARITHMETIC MEAN-HARMONIC MEAN) INEQUALITY
If a1 , a2 , · · · , an are positive real numbers, then
1 1 1
(a1 + a2 + · · · + an ) + + ··· + ≥ n2 ,
a1 a2 an
211
212 Vasile Cîrtoaje
5. BERNOULLI’S INEQUALITY
For any real number x ≥ −1, we have
a) (1 + x) r ≥ 1 + r x for r ≥ 1 and r ≤ 0;
b) (1 + x) r ≤ 1 + r x for 0 ≤ r ≤ 1.
In addition, if a1 , a2 , · · · , an are real numbers such that either a1 , a2 , · · · , an ≥ 0 or −1 ≤
a1 , a2 , · · · , an ≤ 0, then
(1 + a1 )(1 + a2 ) · · · (1 + an ) ≥ 1 + a1 + a2 + · · · + an .
6. SCHUR’S INEQUALITY
For any nonnegative real numbers a, b, c and any positive number k, the inequality
holds
a k (a − b)(a − c) + b k (b − c)(b − a) + c k (c − a)(c − b) ≥ 0,
with equality for a = b = c, and for a = 0 and b = c (or any cyclic permutation).
For k = 1, we get the third degree Schur’s inequality, which can be rewritten as follows
For k = 2, we get the fourth degree Schur’s inequality, which holds for any real numbers
a, b, c, and can be rewritten as follows
where the equality holds for a = b = c, and for a/m = b = c (or any cyclic permutation).
This inequality is equivalent to
X X X X
a4 + m(m + 2) a2 b2 + (1 − m2 )a bc a ≥ (m + 1) a b(a2 + b2 ),
X
(b − c)2 (b + c − a − ma)2 ≥ 0.
Another generalization of the fourth degree Schur’s inequality (Vasile Cirtoaje, 2004):
Let α, β, γ be real numbers such that
1 + α + β = 2γ.
The inequality
X X X X
a4 + α a2 b2 + β a bc a≥γ a b(a2 + b2 )
1 + α ≥ γ2 .
7. CAUCHY-SCHWARZ INEQUALITY
For any real numbers a1 , a2 , · · · , an and b1 , b2 , · · · , bn we have
214 Vasile Cîrtoaje
8. HÖLDER’S INEQUALITY
If x i j (i = 1, 2, · · · , m; j = 1, 2, · · · n) are nonnegative real numbers, then
! v !m
m
Y n
X n uY
X m
m
xi j ≥ .
t
xi j
i=1 j=1 j=1 i=1
9. CHEBYSHEV’S INEQUALITY
Let a1 ≥ a2 ≥ · · · ≥ an be real numbers.
a) If b1 ≥ b2 ≥ · · · bn , then
n n n
X X X
n ai bi ≥ ai bi ;
i=1 i=1 i=1
b) If b1 ≤ b2 ≤ · · · ≤ bn , then
n n n
X X X
n ai bi ≤ ai bi .
i=1 i=1 i=1
10. MINKOWSKI’S INEQUALITY
For any real number k ≥ 1 and any positive real numbers a1 , a2 , · · · , an and b1 , b2 , · · · , bn ,
the inequalities hold
1
k k k
n n n
X 1 X X
aik + bi k k
≥ ai + bi ;
i=1 i=1 i=1
1
k k k k
n n n n
X 1 X X X
aik + bik + ci
k k
≥ ai + bi + ci .
i=1 i=1 i=1 i=1
11. REARRANGEMENT INEQUALITY
(1) If a1 , a2 , · · · , an and b1 , b2 , · · · , bn are two increasing (or decreasing) real se-
quences, and (i1 , i2 , · · · , in ) is an arbitrary permutation of (1, 2, · · · , n), then
and
n(a1 b1 + a2 b2 + · · · + an bn ) ≥ (a1 + a2 + · · · + an )(b1 + b2 + · · · + bn ).
(2) If a1 , a2 , · · · , an is decreasing and b1 , b2 , · · · , bn is increasing, then
and
n(a1 b1 + a2 b2 + · · · + an bn ) ≤ (a1 + a2 + · · · + an )(b1 + b2 + · · · + bn ).
(3) Let b1 , b2 , · · · , bn and c1 , c2 , · · · , cn be two real sequences such that
b1 + · · · + bk ≥ c1 + · · · + ck , k = 1, 2, · · · , n.
If a1 ≥ a2 ≥ · · · ≥ an ≥ 0, then
a1 b1 + a2 b2 + · · · + an bn ≥ a1 c1 + a2 c2 + · · · + an cn .
Notice that all these inequalities follow immediately from the identity
!
X n n
X i
X i
X
ai (bi − ci ) = (ai − ai+1 ) bj − cj ,
i=1 i=1 j=1 j=1
where an+1 = 0.
12. MACLAURIN’S INEQUALITY and NEWTON’S INEQUALITY
If a1 , a2 , . . . , an are nonnegative real numbers, then
and
Sk2 ≥ Sk−1 Sk+1 , (N ewt on)
where
v X
u
u
u a i1 a i2 · · · a i k
k 1≤i1 <···<ik ≤n
Sk = u
u
.
t n
k
216 Vasile Cîrtoaje
f (αx + β y) ≤ α f (x) + β f ( y)
f (x) + (n − 1) f ( y) ≥ n f (s)
The inequality
a + a + ··· + a
1 2 n
f (a1 ) + f (a2 ) + · · · + f (an ) ≥ n f
n
holds for all a1 , a2 , · · · , an ∈ I satisfying a1 + a2 + · · · + an = ns if and only if
f (x) + (n − 1) f ( y) ≥ n f (s)
The inequality
a + a + ··· + a
1 2 n
f (a1 ) + f (a2 ) + · · · + f (an ) ≥ n f
n
holds for all x 1 , x 2 , · · · , x n ∈ I satisfying a1 + a2 + · · · + an = ns if and only if
f (x) + (n − 1) f ( y) ≥ n f (s)
f (x) + (n − 1) f ( y) ≥ n f (s),
If a1 , a2 , . . . , an ∈ I such that
a1 + a2 + · · · + an = S = const ant,
218 Vasile Cîrtoaje
then
(a) E is minimum for a1 = a2 = · · · = an−1 ≤ an ;
(b) E is maximum for either a1 = a or a < a1 ≤ a2 = · · · = an .
On the other hand, it is known the following result concerning the best upper bound
of Jensen’s difference.
Best Upper Bound of Jensen’s Difference-Theorem (Vasile Cirtoaje, 1989). Let p1 , p2 , · · · , pn
be fixed positive real numbers, and let f be a convex function on a closed interval I = [a, b].
If a1 , a2 , · · · , an ∈ I, then Jensen’s difference
is maximum when some of ai are equal to a, and the others ai are equal to b; that is, when
all ai ∈ {a, b}.
if
a1 ≥ b1 ,
a1 + a2 ≥ b1 + b2 ,
·····················
a1 + a2 + · · · + an−1 ≥ b1 + b2 + · · · + bn−1 ,
a1 + a2 + · · · + an = b1 + b2 + · · · + bn .
Let f be a convex function on a real interval I. If a decreasingly ordered vector
−
→
A = (a1 , a2 , . . . , an ), ai ∈ I,
then
f (a1 ) + f (a2 ) + · · · + f (an ) ≥ f (b1 ) + f (b2 ) + · · · + f (bn ).
Glossary 219
p = a + b + c, q = a b + bc + ca, r = a bc,
Æ p
s = p2 − 3q = a2 + b2 + c 2 − a b − bc − ca.
From the identity
it follows that
1 + A + B = C + D,
3(1 + A) ≥ C 2 + C D + D2 .
p = a + b + c, q = a b + bc + ca, r = a bc.
has the highest coefficient A1 = 0, we can prove the inequality g6 (a, b, c) ≥ 0 as in the
preceding case 1.
Notice that sometimes it is useful to break the problem into two parts, p2 ≤ ξq and
p2 > ξq, where ξ is a suitable real number.
x 1 + x 2 + · · · + x n = a1 + a2 + · · · + an ,
Sn = f (x 1 ) + f (x 2 ) + · · · + f (x n ).
0 < x 1 = x 2 = · · · = x n−1 ≤ x n ,
x 1 + x 2 + · · · + x n = a1 + a2 + · · · + an ,
0 < x 1 = x 2 = · · · = x n−1 ≤ x n ,
0 < x 1 = x 2 = · · · = x n−1 ≤ x n ,
0 ≤ x 1 = x 2 = · · · = x n−1 ≤ x n ,
0 ≤ x 1 = x 2 = · · · = x n−1 ≤ x n ,
0 ≤ x 1 = x 2 = · · · = x n−1 ≤ x n ,
0 ≤ x 1 = x 2 = · · · = x n−1 ≤ x n ,
The Equal Variable Theorem (EV-Theorem) for real variables has the following
statement (Vasile Cirtoaje, 2012).
EV-Theorem (for real variables). Let a1 , a2 , · · · , an (n ≥ 3) be fixed real numbers, let
x 1 ≤ x 2 ≤ · · · ≤ x n be real variables such that
x 1 + x 2 + · · · + x n = a1 + a2 + · · · + an ,
Sn = f (x 1 ) + f (x 2 ) + · · · + f (x n )
S = {(x 1 , x 2 , · · · , x n ) : x 1 + x 2 + · · · + x n = s, x i ≥ 0, i = 1, 2, · · · , n}.
224 Vasile Cîrtoaje
If
F (x 1 , x 2 , x 3 , · · · , x n ) ≥
n x + x x + x o
1 2 1 2
≥ min F , , x 3 , · · · , x n , F (0, x 1 + x 2 , x 3 , · · · , x n )
2 2
for all (x 1 , x 2 , · · · , x n ) ∈ S, then F (x 1 , x 2 , x 3 , · · · , x n ) is minimal when
s
x1 = x2 = · · · = x k = , x k+1 = · · · = x n = 0;
k
that is, s s
F (x 1 , x 2 , x 3 , · · · , x n ) ≥ min F , · · · , , 0, · · · , 0
1≤k≤n k k
for all (x 1 , x 2 , · · · , x n ) ∈ S.
Notice that if
x + x x + x
1 2 1 2
F (x 1 , x 2 , x 3 , · · · , x n ) < F , , x3, · · · , x n
2 2
involves
F (x 1 , x 2 , x 3 , · · · , x n ) ≥ F (0, x 1 + x 2 , x 3 , · · · , x n ),
then the hypothesis
F (x 1 , x 2 , x 3 , · · · , x n ) ≥
n x + x x + x o
1 2 1 2
≥ min F , , x 3 , · · · , x n , F (0, x 1 + x 2 , x 3 , · · · , x n )
2 2
is satisfied.