My Inequality Project PDF
My Inequality Project PDF
My Inequality Project PDF
M.Ramchandran
Preface v
Chapter 1. AM-GM Inequality 1
1. Arithmetic Mean - Geometric Mean 1
2. Proof 2
3. Beginners Practice Problems 4
4. Geometric Interpretations 7
5. AM-GM Tautogrid Technique 7
6. Nessbits Inequality 9
7. The Reverse Technique 10
8. The Weighted AM-GM Inequality 15
9. Method Of Balancing Co-efficients by AM-GM 16
10. Quasiliearisation 18
11. Equivalent Summation Technique 19
12. The G function 21
13. Problem Set 22
iii
Preface
This is the preface and it is created using a TeX field in a paragraph by itself containing \chapter*{Preface}.
When the document is loaded, this appears if it were a normal chapter, but it is actually an unnumbered
chapter.
v
CHAPTER 1
AM-GM Inequality
where ofcourse,
A.M. stands for the arithmetic mean of the two concerned non-negative real numbers - a and b.
Extending this idea to n-variables - a1 , a2 , a3 , .., an we get that -
Pn
a1 + a2 + a3 + ... + an ai
A.M. = = i=1
n n
Definition 2 (Geometric Mean). Geometric Mean of two real numbers is the collection of positive
real numbers is the nth root of the product of the numbers. Note that if it is even, we take the positive nth
root and it is mathematically expressed as -
G.M. = ab
where,
G.M. stands for Geometric Mean of the two concerned non-negative real numbers - a and b
Extending this idea to n-variables - a1 , a2 , a3 , .., an we get that -
n
1 1
Y
G.M. = (a1 a2 a3 ...an ) n = ( ai ) n
i=1
Recall the fact that for any real number x , we know that x2 0.
therfore we know that for all non-negative real numbers - a and b -
( a b)2 0
a + b 2 ab 0
a+b
ab
2
ring any bells?
yes you have probably spotted that LHS of the inequality is the arithmetic mean we discussed and
ofcourse the RHS is the geometric mean. Now this encourages the following proposition -
Theorem 1. Arithmetic Mean of some n non-negative real numbers is always greatern than or equal
to the Geometric Mean of the same
1
2 1. AM-GM INEQUALITY
is that true? if so where is the validity? by seeing that it is true for certaina two positive reals doesnt
imply a wider truth for any number of variables.
2. Proof
Proof. Lets proceede by proving the above statement for smaller numbers and eventually the general
case-
Since we know that A.M. G.M. for two variables we have -
p+q
pq
2
r+s
rs
2
also we have -
pq + rs q 1
pqrs = (pqrs) 4
2
combining we get that-
p+q
p+q+r+s + r+s pq + rs 1
= 2 2
(pqrs) 4
4 2 2
or,..
p+q+r+s 1
(pqrs) 4
4
which is AM-GM inequality for 4 variables!!!
With a similar idea we can do the same as above for 8 - variables by splitting into 4-4 and using AM-GM
for 4 variables.
Now something should be irking in your mind.... Cant we prove in the same way for any n?? - the
answer is NO. Why not? - well this covers only 2-powers not even even integers.. so this proof is incomplete
considering the general case. But a proof with the same idea can be given by induction for any 2-powers.It
goes as follows -
Consider 2k+1 variables - a1 , a2 , a3 , ..., a2k+1
Assume the truth of the statement for 2k , we shall prove it for 2k+1
Since it is true for 2k , we have -
a1 + a2 + a3 + ... + a2k 1
k
(a1 a2 a3 ..a2k ) 2k (1)
2
and,
a2k +1 + a2k +2 + a2k +3 + ... + a2k+1 1
(a2k +1 a2k +2 a2k +3 ...a2k+1 ) 2k (2)
2k
again,
1 1
(1) + (2) a1 + a2 + ... + a2k+1 (a1 a2 a3 ..a2k ) 2k + (a2k +1 a2k +2 a2k +3 ...a2k+1 ) 2k 1
= k+1
(a1 a2 ...a2k+1 ) 2k+1
2 2 2
or,
a1 + a2 + a3 + ... + a2k+1 1
(a1 a2 a3 ..a2k+1 ) 2k+1
2k+1
thus by induction it is proved for all powers of 2!!! this is a bit closer to the full proof..
But then what about 6, 10.. and 3, 5, 7... how do we prove them..?
with a little bit of trickery we get away with 3 - since the inequality is true for 4 we have -
1
a + b + c + (abc) 3 4 1 1
(abc) 3 4 = (abc) 3
4
and this re-arranges to -
a+b+c 1
(abc) 3
3
2. PROOF 3
which is the AM-GM inequality for 3 reals!!!! this can also be done by the following way -
a + b + c + ( a+b+c
3 ) a+b+c 1
[abc ( ]4
4 3
that is..
a+b+c a+b+c 1
[abc ( ]4
3 3
which re-arranges to -
a+b+c 1
(abc) 3
3
The above two proofs crucially modifies our way of looking at AM- GM inequality. Let us re-define
ourselves the meaning of AM,GM into -
A.M. = sum equaliser
G.M. = product equaliser
what does that mean ? It means that say A is the A.M. of 3 variables x, y, z and G the G.M. then -
x + y + z = 3A
and,
x.y.z = G3
If we know that the inequality is true for a certain m , then for any n such that m > n we can prove
the validity of AM-GM inequality as follows-
a1 + a2 + ... + an + (m n)A 1 mn
(a1 a2 a3 ..an ) m A m
m
mA n mn
(G) m A m
m
Am Gn Amn
or,
An Gn
= A G
thus we have prove the AM-GM inequality for any n as we know it to be true for any 2k .. or -
A.M. G.M.
for any n This is trivial looking inequality is probably the most celebrated of all that we all shall discuss in
this book.It has far and wide applications..
Proof. The inequality is true for 2n if it is true for n or it is true for all powers of 2 (already proved)
Suppose that the inequality is true for n numbers.We then choose
s
an =
n1
where,
s = a1 + a2 + a3 + .... + an
According to the inductive hypothesis, we get
s a1 a2 a2 a3 ...an s 1
s+ n( )n
n1 n1
1
s (n 1)(a1 a2 a3 ..an ) n1
Therfore if the inequality is true for n numbers than it will be true for n 1 numbers and by induction
(Cauchy Induction),the inequality is true for every natural number n .Equality occurs if and only if a1 =
a2 = a3 = ... = an
4 1. AM-GM INEQUALITY
3. Beginners Practice Problems
1. For a, b, c R+
0 .Prove that
(a + b)(b + c)(c + a) 8abc
Solution
multiplying the above inequalities we get the desired with equality for a = b = c
2.For a, b, c R+
0 . Prove that
a2 + b2 + c2 ab + bc + ca
Solution
a2 + b2 2ab
b2 + c2 2bc
c2 + a2 2ac
2
which we just proved! with equality for a = b = c = 3
and,
a2 c + b2 a + c2 b 3abc
multiply the above two inequalities to get the desired with equality for a = b = c
5.For a, b, c R+
0 .Prove that -
a4 + b4 + c4 abc( ab + bc + ca)
(Source:Nagasaki University 1970)
Solution
This problem seems a step tougher to novice. Expanding the RHS will not lead in the correct direction. Let
us try to transform this inequality into an equivalent one that for convenient sake looks simpler. One way
of this being done is by taking abc to the LHS -
a4 + b4 + c4 abc( ab + bc + ca)
a3 b3 c3
+ + ab + bc + ca
bc ac ab
a3 a3 b3 c3
+ + + 4a
bc bc ca ab
b3 b3 c3 a3
+ + + 4b
ac ac ba cb
c3 c3 a3 b3
+ + + 4c
ab ba bc ca
the three inequalities are true by AM-GM, add these inequality to get -
a3 b3 c3
+ + a + b + c ab + bc + ca
bc ca ab
notice that the last inequality is true and is equivalent to Problem 2 thus we have proved the inequality with
equality for a = b = c
(by M.Ramchandran)
aliter: Alternatively another ingenious AM-GM solution will be notice by AM-GM that -
3a4 + 3b4 + 2c4
abc ab
8
3b4 + 3c4 + 2a4
abc bc
8
3c4 + 3a4 + 2b4
abc ca
8
adding these inequalities we get the desired.
(by Mathias Tejs)
6 1. AM-GM INEQUALITY
Note It is not expected of the reader to get the above two proofs (if he/she is a newbie). Such proofs
come due to some strong observation and several wrong tries(like mine.. :P). Both involved splitting the
terms into terms with suitable co-efficients which shall come as time goes. So dont get disheartened or
awed.
6. For a, b, c R+
0 such that a + b + c + d = 1.Prove that -
1
ab + bc + cd
4
Solution Write the Inequality using the given condition as -
(a + b + c + d)2
ab + bc + cd
4
By AM-GM,
(a + b + c + d)2 (a + c) + (b + d) 2
=( ) (a + c)(b + d) = ab + bc + cd + da ab + bc + cd
4 2
with the equality for a = b = c, d = 0 or d = c = b, a = 0
(by M.Ramchandran)
7. For a, b, c R+0 .Prove that -
a2 + b2 b2 + c2 c2 + a2
+ + a+b+c
a+b b+c c+a
Solution Note that by AM-GM,
2(x2 + y 2 ) (x + y)2
for all non-negative reals x, y thus,
a2 + b2 b2 + c2 c2 + a2 a+b b+c c+a
+ + + + =a+b+c
a+b b+c c+a 2 2 2
with equality for a = b = c
8. For a, b, c R+
0 such that a + b + c + d = 1.Prove that -
a2 b2 c2 d2 1
+ + +
a+b b+c c+d d+a 2
Solution We have -
(a b) + (b c) + (c d) + (d a) = 0
a 2 b2 b2 c 2 c2 d2 d2 a 2
+ + + =0
a+b b+c c+d d+a
a2 b2 c2 d2 b2 c2 d2 a2
+ + + = + + +
a+b b+c c+d d+a a+b b+c c+d d+a
Solution By AM-GM,
a3 b+c+d 1 a
+ +
b+c+d 18 12 2
b3 a+c+d 1 b
+ +
a+c+d 18 12 2
c3 b+a+d 1 c
+ +
b+a+d 18 12 2
d3 b+c+a 1 d
+ +
b+c+a 18 12 2
Add these inequalities to get -
a+b+c+d1
LHS
3
and also frolm AM-GM,
(a + b + c + d)2 4(a + c)(b + d) = 4
or
a+b+c+d2
and the conclusion follows. With equallity for a = b = c = d = 41 .
(by mathlinks user : quykhtn-qa1)
5. For a, b, c R+ such that a + b + c = 2.Prove that-
a b c
+ + >2
b(a + b) c(b + c) a(a + c)
(Source:Own Inequality)
Solution By AM-GM,
a
+ (a + b)a + ab 3a
b(a + b)
b
+ (c + b)b + cb 3b
c(c + b)
c
+ (a + c)c + ac 3c
a(a + c)
or , X a
3(a + b + c) (a + b + c)2 = 6 4 = 2
b(a + b)
cyclic
but the equality case cant occur so the inequality sign becomes strict.
6. For x, y, z R+ such that xyz = 1 .Prove that -
x3 y3 z3 3
+ +
(1 + y)(1 + z) (1 + x)(1 + z) (1 + x)(1 + y) 4
(Source: IMO Shortlist 1998)
Solution By AM-GM
x3 1+y 1+z 3x
+ +
(1 + y)(1 + z) 8 8 4
6. NESSBITS INEQUALITY 9
3
X x 1X 3
= (2x 1)
(1 + y)(1 + z) 4 4
cyclic cyclic
Equality for x = y = z = 1
7. For a, b R+
0 such that a + b = 1 Prove that -
a2 b2
+ 13
1+b 1+a
Solution By AM-GM,
a2 b+1 2a
+
b+1 9 3
b2 a+1 2b
+
a+1 9 3
1
add these to get the result with equality for a = b =
2
6. Nessbits Inequality
This a very famous,well-known and well discussed inequality. Most problems are probably stronger than
this (that is the job of the proposers - if they keep the questions down to elementary inequality then what
is the fun?) but nevertheless it is a must to know this beautiful inequality-
Theorem 2. For a, b, c R+
0 the following inequality holds -
a b c 3
+ +
b+c c+a a+b 2
Proof
Note that for any positive real x, y, z we have by AM-GM -
x y z
+ + 3
y z x
Consider the following expressions -
a b c
S= + +
b+c c+a a+b
b c a
M= + +
b+c c+a a+b
c a b
N= + +
b+c c+a a+b
we have ofcourse: M + N = 3. According to the Lemma,
a+b b+c c+a
M +S = + + 3
b+c c+a a+b
a+c b+a c+b
N +S = + + 3
b+c c+a a+b
Therefore,
M + N + 2S 3
or
2S 3
or,
3
S
2
Practice
10 1. AM-GM INEQUALITY
1. Extend the same idea to prove Nessbits inequality for 4 non-negative reals -
a b c d
+ + + 4
b+c c+d d+a a+b
2. For a, b, c R+
0 , Prove that -
a2 + 1 b2 + 1 c2 + 1
+ + 3
b+c c+a a+b
(Source : Indian RMO 2006)
3. Prove Nessbits inequality using the Tautogrid Technique
(Left to readers)
a3 X ab2 ab2
X X
X b a+b+c
2 2
= a 2 2
a = a =
a +b a +b 2ab 2 2
cyclic cyclic cyclic cyclic
and Voila! we have a solution and it is correct and it uses AM-GM inequality.
Here the main idea is -
l t1
at a k ab
=
kat1 + lbt1 k kat1 + lbt1
This is only a random form. To justify my statement that we can do several types of problems and to get
you used to this technique,
The strength and importance of this technique cant be more revealed than the following problems.
Problems
1. Let a, b, c R+ , then prove that we have
a3 b3 c3 a+b+c
+ +
a2 + ab + b2 b2 + bc + c2 c2 + ca + a2 3
7. THE REVERSE TECHNIQUE 11
Solution
a3 b3 c3
LHS = + +
a2 + ab + b2 b2 + bc + c2 c2 + ca + a2
X ab(a + b) X ab(a + b) a+b+c
= (a 2 2
) (a )=
a + ab + b 3ab 3
cyclic cyclic
The application of this technique doesnt come by just seeing the solutions given by the Author.. for
really *learning* it, the reader is advised to try the following problems before succumbing to seeing the
solutions.
2. Let a, b, c R+ such that , a + b + c = 3 . Prove that
a b c 3
2
+ 2
+ 2
1+b 1+c 1+a 2
Solution
We have;
a ab2
2
=a
1+b 1 + b2
. Summing up cyclically we get that
X a X ab2 X ab2
= a + b + c a + b + c
cyc
1 + b2 cyc
1 + b2 cyc
2b
1 3 3
= 3 (ab + bc + ca) 3 =
2 2 2
since from trivial inequality we have that
1
(ab + bc + ca) (a + b + c)2 = 3
3
. Therefore we are done.
Can we extend this problem to four variables? The answer is yes
cyc
a + 2b2 3
Hence it suffices to show that
2 2 2 2 2 2
a3 b3 + b3 c3 + c3 a3 3
But,
2 2 2 2 2 12
a3 b3 + b3 c3 + c3 a3 [ab + ab + 1 + bc + bc + 1 + ca + ca + 1]
3
2 2
= 1 + (ab + bc + ca) 1 + 3 = 1 + 2 = 3
3 3
because
(a + b + c)2 3(ab + bc + ca) ab + bc + ca 3
. Hence we finish our proof here.
Note that this is also true for ab + bc + ca = 3. However, here I pose a challenge for the readers.
5. For a, b, c R+ such that a + b + c = 3,Prove that -
a2 b2 c2
+ + 1
a + 2b3 b + 2c3 c + 2a3
Solution Obviously
a2 2ab3 ab3
3
=a 3
a2
a + 2b a + 2b 3 a b2
3
2
3
=a b a2
3
Hence we have
X a 2
2 3
2 + c 3 b2 + a 3 c2
a + b + c b a
cyc
a + 2b3 3
2
a+b+c [b(a + a + 1) + c(b + b + 1) + a(c + c + 1)]
9
2 2
= 3 {2(ab + bc + ca) + 3} 3 {6 + 3} = 1
9 9
The last inequality is true by AM-GM, and since we have
1
ab + bc + ca (a + b + c)2 = 3.
3
6. For a, b, c R+ such that a + b + c = 3, Prove that -
a+1 b+1 c+a
+ + 3
b2 + 1 c2 + 1 a2 + 1
7. THE REVERSE TECHNIQUE 13
Solution Since
a+1 b2 (a + 1) b2 (a + 1) ab b
2
= a + 1 2
a + 1 =a+1
b +1 b +1 2b 2 2
. Therefore we have
X a+1 a+b+c 1
2+1
3+ (ab + bc + ca)
cyc
b 2 2
3 3
3+ =3
2 2
Equality for a = b = c = 1
Hence proved.
7. For a, b, c R+ such that a2 + b2 + c2 = 3, prove that
1 1 1
+ + 1
a3 + 2 b3 + 2 c3 + 2
(Pham Kim Hung)
Solution Note that
X1 a3
X 1
= 1 3
cyc
a3 + 2 cyc
2 a +2
3 1 X a3 3 1
= =1
2 2 cyc 3a 2 2
8. For a, b, c > 0. Prove that -
a3 b3 c3 1 2
a + b2 + c2
+ +
a+b b+c c+a 2
Solution
X a3 X a2 b
= a2
a+b a+b
X a2 b X 1 a2 b
= a2 + b2 + c2 a2 + b2 + c2
a+b 2 ab
X1 Xa
= a2 + b2 + c2 (a ab) a2 + b2 + c2 (a + b)
2 4
1 1
= a2 + b2 + c2 (a2 + b2 + c2 + ab + bc + ca) a2 + b2 + c2 (a2 + b2 + c2 )
4 2
1 2 2 2
= (a + b + c )
2
Therefore we are done. Equality occurs if and only if a = b = c.
2
=3 3(a + b + c) = 3 2 = 1
9
Therefore we are done. Equality occurs if and only if a = b = c = 1.
Practice Problems
10. For a, b, c, d R+ such that a + b + c + d = 1.Prove that -
a2 b2 c2 d2 1
+ + +
a+b b+c c+d d+a 2
11. Given a, b, c, d R+ , show that we have
a4 b4 c4 d4 a+b+c+d
+ 3 + 3 + 3
a3 + 2b 3 b + 2c 3 c + 2d 3 d + 2a3 3
(by Gibbenergy)
13. For given four positives a, b, c, d with sum 4; show that
a b c d
+ + + 1
1 + b c 1 + c a 1 + d a 1 + a2 b
2 2 2
=1
Therefore ab+c bc+a ca+b 1
(by mathlinks user: rem)
Solution 3 Note that -
1
ab+c bc+a ca+b = 1
aa bb cc
the last inequality is true because for a, b, c individually greater or lesser than 1 we have -
aa a, bb b, cc c
Solution
a a a 3
3a + ( + b) + ( + 2b) + ( + b + 4c) 3a + ab + 2 ab + 3 abc
4 2 4
(by Aravind Srinivas)
k(x2 + y 2 ) + z 2
Solution Lets experiment with some values of k shall we? Let k = 10, so we are now required to find the
minimum of this non-symmetric expression-
10(x2 + y 2 ) + z 2
How do we apply a classical basic inequality like AM-GM for this? Well it does seem horrendously difficult,
so lets take a sneak-peek at the magical solution? By AM-GM we have the following inequalities,
2x2 + 2y 2 4xy
1
8x2 + z 2 4yz
2
1
8y 2 + z 2 4zx
2
and summing up these we have,
Hurray ! We did it!!(dont get envious i like you was flabbergasted at this cause it aint mine) how can
a guy arrive at this?? why not pair up 1 and 9 , 5 and 5 or something? All answers shall be revealed in the
following lines. back to our general problem, Lets choose some l k , then apply AM-GM this way,
lx2 + ly 2 2lxy
1 p
(k l)y 2 + z 2 yx 2(k l)
2
1 p
(k l)x + z 2 xz 2(k l)
2
2
summing up we get this - p
k(x2 + y 2 ) + z 2 2lxy + (yz + zx) 2(k)
now we have a condition given - xy + yz + zx = 1 so for obtaining a numerical value we have to have the
co-efficients
p in the RHS the same and without a variable hopefully. so intuitively lets just equate them-
2l = 2(k l) and solving this we obtain that
1 + 1 + 8k
l=
4
and ofcourse substituting this in the equation we get the minimal value we are looking for to be -
1 + 1 + 8k
2
so we observe that there is a unique pair of integers - l, k l that show us the way by AM-GM. Now that is
the reason we were seeing the use of 8,2 and not 3,7 etc.. sure enough you can check the credibility of the
pairings now!
Note
2. Let x, y, z, t be real numbers satisfying xy + yz + zt + tx = 1. Find the minimum of the expression -
5x2 + 4y 2 + 5z 2 + t2
Solution Here k = 5, so we chooose l 5,
lx2 + 2y 2 2 2lxy
2y 2 + lz 2 2 2lyz
1 p
(5 l)z 2 + t2 2(5 l)tx
2
Summing up these results, We conclude that
p
5x2 + 4y 2 + 5z 2 + t2 2 2l(xy + tz) + 2(5 l)(zt + tx)
The condition xy + yz + zt + tx = 1 suggests us to choose a number l(0 l 5) such that -
p
2 2l = 2(5 l)
. A simple calculation yields l = 1, thus the minimum of 5x2 + 4y 2 + 5z 2 is 2 2
18 1. AM-GM INEQUALITY
Note Using the same method, solve the following problem :
Let x, y, z, t be arbirary real numbers. Prove that -
1
2 2 2 2 2kl 2
x + ky + z + lt (xy + yz + zt + tx)
k+l
3. Let x, y, z be positive real numbers with sum 3. Find the minimum of the expression
x2 + y 2 + z 3
(Pham Kim Hung)
Solution Let a and b be teo positive real numbers.. Then,by AM-GM inequality we have -
x2 + a2 2ax
y 2 + a2 2ay
z 3 + b2 + b2 3zb2
Combining these we have x2 + y 2 + z 3 + 2(a2 + b3 ) 2a(x + y) + 3b2 z with equality for x = y = a and
z = b.In this case, we could have 2a + b = x + y + z = 3(?). Moreover, in order for 2a(x + y) + 3b2 z to be
represented as x + y + z, we must have 2a = 3b2 (??).Accordin to (?) and (??) we can find out that,
1 + 37 3b 19 37
b= ,a = =
6 2 12
Therfore the minimum of x2 + y 2 + z 3 is 6a 2(a2 + b3 ) where a, b are as determined.The proof is completed.
4. For x, y, z R+ such that xy + yz + zx = 1 , Prove that - 15x2 + 7y 2 + 3z 2 6
(Own Inequality)
Solution Rewrite the Inequality as -
7
5x2 + y 2 + z 2 2
3
By AM-GM we have ,
x2 + y 2 2xy
1
4x2 + z 2 2xz
4
4 2 3 2
y + z 2xz
3 4
now adding these we get the desired.
10. Quasiliearisation
This is a very intrigueing idea due to Russian problem proposer - Fedor Petrov
We know by AM-GM that ,
2ab a2 + b2
Introduce a parameter and get the following:
b2
2ab ta2 +
t
Then read the last inequality from the other point: for any positive a, b there exist positive t such that
b2 b
2ab = ta2 + (t = )
t a
Or, we may write:
b2
2ab = min(ta2 + )
t
11. EQUIVALENT SUMMATION TECHNIQUE 19
How may this observation help? Put
X X
a= (x2i ), b = (yi2 )
Then for appropriate t we have:
b X 2 yi2 X
2 a b = ta + = (txi + ) (2xi yi )
t t
So, we get the Cauchy-Schwarz Inequality. A lot of other inequalities also may be proved by this idea
Problem Prove that for any four nonnegative reals a, b, c, d the following inequality holds-
1 1 1
(ab) 3 + (cd) 3 ((a + c + b)(a + c + d)) 3
(Source:Proposed at 239 Lyceum Traditional Olympiad)
(Author : Fedor Petrov)
We have
1 1
3(AB) 3 Ax + By +
xy
And for any positive A and B there exist appropriate x and y ,for which equality holds
1 1
(AB) 3 (AB) 3
x= ,y =
A B
Let
A = (a + c + b), B = (a + c + d)
in terms of the problem. For some positive x, y we have -
1 1 1
3(AB) 3 = Ax + By + = (a + c + b)x + (a + c + d)y + =
xy xy
1 1 1 1 1 1
(a+c+b)x+(a+c+d)y+ + = a(x+y)+bx+ + c(x+y)+dy+ 3(ab) 3 +3(cd) 3
x(x + y) y(x + y) x(x + y) y(x + y)
By AM-GM and we are done!
a3 b3 X a3 X b3 1 X a 3 + b3 a+b+c
2 2
= a b = 2 2
= 2 2
= 2 2
a + ab + b cyc
a + ab + b cyc
a + ab + b 2 cyc a + ab + b 3
But there exists another nice solution using the reverse technique. The solution runs as follows:
a3 X
X ab(a + b) X ab(a + b)
2 2
= a 2 2
a+b+c
cyc
a + ab + b cyc
a + ab + b cyc
3ab
(a b)2
a b c a b b c b (b c)(a c)
G(a, b, c) = + + 3= + 2 + + 1 = +
b c a b a c a a ab ca
this factorization plays an important role in many proofs
Note
The inequality - G(a, b, c) 0 is a cyclic inequality and thus no pair-wise order can be assumed.
I shall present some properties of this important function.
Properties
1. For a, b, c, k R+ we have -
G(a, b, c) G(a + k, b + k, c + k)
Proof WLOG : c = min{a, b, c} We have ,
(a b)2 (b c)(a c)
G(a, b, c) = +
ab ca
So it is enough to prove that -
(a b)2 (b c)(a c) (a b)2 (b c)(a c)
+ +
ab ca (a + k)(b + k) (c + k)(a + k)
This is true as k > 0 and by assumption - (a c)(b c) 0. Hence proved with equality for a = b = c
2. For a, b, c, k R+ we have -
G(a, b, c) G(a + b, b + c, c + a)
Proof We only have to show that -
(a b)2 (b c)(a c) (a b)2 (b c)(a c)
+ +
ab ca (a + c)(b + c) (a + b)(a + c)
It is trivial to see that the inequality is true.
Note The following problem (equivalent to the property discusses above) was asked in the Mathlinks
contest 2003 -
a b c a+b b+c c+a
+ + + +
b c a b+c c+a a+b
The following properties(same conditions as above)can be solved using the same method -
3. For a, b, c, k R+ we have -
G (a b)2 , (b c)2 , (c a)2 2
(by M.Ramchandran)
5. Let a, b, c, k R+ , then If and only if the sum of any two of {a, b, c} is less than 2
(by M.Ramchandran)
6. For a, b, c, k R+ and let a b c then,
a2 b2 c2
G( , , ) G(a, b, c)
bc ca ab
(by M.Ramchandran)
7. For a, b, c, k R+ and let k max{a2 , b2 , c2 } then,
G(a, b, c) G(a2 + k, b2 + k, c2 + k)
(Pham Kim Hung)
8. For a, b, c, k R+ we have -
G(a3 , b3 , c3 ) 3G(a2 , b2 , c2 )
(by M.Ramchandran)
9. For a, b, c, k R+ we have -
G(a2 , b2 , c2 ) G(a, c, b)
(by M.Ramchandran)
If more properties are invited to be shared with the author by e-mail.
(1 c)(1 d) 1
13. PROBLEM SET 25
1
(a 1)(b 1) (a + b 2)2 1
4
Therefore -
(a 1)(b 1)(1 c)(1 d) 1
and the conclusion follows. Equality for a = b = c = d = 1, a = b = 2, c = d = 0 and cylic permutations.
9. Let a, b, c R+
0 and a + b + c = 3 . Prove that -
p p p
a 1 + b3 + b 1 + c3 + c 1 + a3 5
(Pham Kim Hung)
Solution We know by AM-GM that,
X p X p X 1
a 1 + b3 = a (1 + b)(1 b + b2 ) a(2 + b2 )
2
cyclic cyclic cyclic
If first one occurs we are done and if the second one occurs then replace a by a1 ; b by 1
b ; c by 1
c and we
get the same expression as the first one and thus we have proved !!
16.Prove that for a, b, c R+ -
r r r
2a 2b 2c (a + b + c) (ab + bc + ca)
+ + p
a+b b+c c+a (a + b2 + c2 ) (a2 b2 + b2 c2 + c2 a2 )
2
Solution r
2b 4b
b+c 3b + c
r
2c 4c
c+a 3c + a
r
2a 4a
a+b 3a + b
So that,
r r r
2a 2b 2c a b c
+ + 4 + +
a+b b+c c+a 3a + b 3b + c 3c + a
From Titus Lemma and the following well-known inequality a2 + b2 + c2 ab + bc + ca
a b c a2 b2 c2
+ + = 2 + 2 + 2
3a + b 3b + c 3c + a 3a + ab 3b + bc 3c + ca
2 2
(a + b + c) (a + b + c)
2 2 2
3 (a + b + c ) + ab + bc + ca 4 (a2 + b2 + c2 )
Therefore,
2
r r r
2a 2b 2c (a + b + c)
+ + 2
a+b b+c c+a a + b2 + c2
We also have, (just prove analogously)
r r r r r r
2a 2b 2c 2ac 2ba 2cb
+ + = + +
a+b b+c c+a ac + bc ba + ca cb + ab
2
(ab + bc + ca)
a b2 + b2 c2 + c2 a2
2
So that,
!
2 2
r r r
2a 2b 2c 1 (a + b + c) (ab + bc + ca)
+ + 2 2 2
+ 2 2
a+b b+c c+a 2 a +b +c a b + b2 c2 + c2 a2
(a + b + c) (ab + bc + ca)
p
(a2 + b2 + c2 ) (a2 b2 + b2 c2 + c2 a2 )
(from AM-GM inequality)
The proof is completed. Equality holds if and only if a = b = c
(by mathlinks user : leviethai)
13. PROBLEM SET 29
17. For a, b, c R+
0 . Prove that -
1 1 1 9
(ab + bc + ca) 2
+ 2
+
(a + b) (b + c) (c + a)2 4
(Source : Iran 1996)
Solution This is one the most famous, discussed and celebrated inequality of all times. In the year asked
it was percieved as a very difficult inequality not solved by any elementary methods but this notion was
wronged by this solution given by the Vietnamese Inequality Solver well known for his beautiful solutions -
Va Quoc Ba Can (mathlinks uers id : canhang2007)
Without loss of generality, we may assume that a b c. Then, we will show that
1 1 1 1 2
2
+ 2
+ 2
+
(a + b) (a + c) (b + c) 4ab (a + c)(b + c)
Indeed, this inequality is equivalent to
1 1 2 1 1
2
+ 2
(a + c) (b + c) (a + c)(b + c) 4ab (a + b)2
or
(a b)2 (a b)2
2 2
(a + c) (b + c) 4ab(a + b)2
This is true because
4ab 4b2 (b + c)2
and
(a + b)2 (a + c)2
Now, using the above estimation, it is sufficient to prove that
1 2 9
(ab + bc + ca) +
4ab (a + c)(b + c) 4
Since
ab + bc + ca 1 c(a + b)
= +
4ab 4 4ab
and
2(ab + bc + ca) 2c2
=2
(a + c)(b + c) (a + c)(b + c)
it is equivalent to
c(a + b) 2c2
4ab (a + c)(b + c)
or
(a + b)(b + c)(c + a) 8abc
The last one is true according to the AM-GM Inequality, so our proof is completed
It stands out as one of the best solutions for the inequality.
18. Let a, b, c R+ such that abc = 1 . Prove that -
ak bk ck 3
+ +
a+b b+c c+a 2
for any positive integer k
(Source : China Northern Mathematical Olympiad 2007)
Solution Re-write the LHS as -
3 X ak1 b
ak1 + bk1 + ck1 +
2 a+b
cyclic
30 1. AM-GM INEQUALITY
by applying AM-GM to the denominators in the RHS we get -
3 1
X
RHS ak 2 b 2
cyclic