0% found this document useful (0 votes)
580 views4 pages

Andre Rzym - Muirhead Inequality

1) Muirhead's inequality states that if one sequence of real numbers majorizes another, then the symmetric mean of the first sequence will be greater than or equal to the symmetric mean of the second sequence. 2) Majorization means that the terms of one sequence sum to the same total and are at least as large when arranged in descending order. 3) The inequality provides a tool for proving relationships between polynomial expressions and means of real variables. Several examples are given to illustrate its application.

Uploaded by

Dijkschneier
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
580 views4 pages

Andre Rzym - Muirhead Inequality

1) Muirhead's inequality states that if one sequence of real numbers majorizes another, then the symmetric mean of the first sequence will be greater than or equal to the symmetric mean of the second sequence. 2) Majorization means that the terms of one sequence sum to the same total and are at least as large when arranged in descending order. 3) The inequality provides a tool for proving relationships between polynomial expressions and means of real variables. Several examples are given to illustrate its application.

Uploaded by

Dijkschneier
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Muirhead’s Inequality

Andre Rzym
November 2005

Introduction

Muirhead’s inequality requires a little work in order to understand what


it actually is, but it is well worth the trouble.

As well as being a useful tool, quoting it is a great way to intimidate


people who are not familiar with it. To quote the diary notes (from the
UKMT Yearbook) of the UK team leader for the ’05 IMO, when a ‘jury’
decides which questions will be used:

‘. . . many leaders don’t know the technique known as Muirhead’s


inequality, and some of those who do seem to think that it is an
advanced exotic technique . . . ’

‘. . . it turns out that Muirhead’s inequality might be better known


among the students than parts of the jury . . . ’

A good deal of the material for this article is taken from Inequalities by
Hardy, Littlewood, Pólya, with examples culled from various sources.

Majorisation
0 0
Suppose we have two sequences of reals, (α1 , . . . , αn ) and (α1 , . . . , αn ),
0
which we refer to as (α) and (α ).
0 0
We say that (α) majorises (α ), or equivalently that (α ) is majorised by
(α), if the following conditions are met:
0
• αi ≥ 0 ∀i, αj ≥ 0 ∀j
0 0
• α1 + . . . + αn = α1 + . . . + αn
0
• If the elements of (α) and (α ) are arranged in descending order, then
0 0
α1 + . . . + αi ≥ α1 + . . . + αi 1≤i≤n

0
Note that there is no requirement that the αi , αi be integers.

1
0
If these conditions are satisfied then (α) majorises (α ), which we write
as
0
(α)  (α )

For example:
• (0, 3, 0)  (1, 0, 2)

• (4, 0, 0, 0)  (2, 0, 2) since the number of elements is different

• (5, 0, −1)  (2, 2, 0) since terms cannot be negative

• (2, 1, 1, 1)  (1, 1, 1, 1) since 2 + 1 + 1 + 1 6= 1 + 1 + 1 + 1

• (4, 1, 1, 1)  (3, 3, 1, 0) since 4 + 1  3 + 3

Symmetric Means

Suppose we have n positive reals, xi and a sequence (α). Then we can


construct a function

F (x1 , x2 , . . . , xn ) = xα1 1 xα2 2 . . . xαnn

We can, of course, permute the xi to give, for example,

F (x2 , x1 , . . . , xn ) = xα2 1 xα1 2 . . . xαnn

We denote the sum over all n! permutations of the xi by


X
!F (x1 , x2 , . . . , xn )

Finally, we define the symmetric mean as


1 X
[α] = !F (x1 , x2 , . . . , xn )
n!
Obviously, [α] is a function of x1 , x2 , . . . , xn . If we swap any of the xi , [α]
is unchanged. That is why it is called a symmetric mean!

For example:

• If (α) = (0, 3, 1) then F (x1 , x2 , x3 ) = x32 x3

• If (α) = (2, 1) then !F (x, y) = x2 y + y 2 x


P

2
• If (α) = (1, 3, 2) then the symmetric mean
1
[α] = (x1 x32 x23 + x1 x33 x22 + x2 x31 x23 + x2 x33 x21 + x3 x31 x22 + x3 x32 x21 )
3!

Muirhead’s Inequality

Now that we have defined majorisation and symmetric means, the state-
ment of Muirhead’s inequality is easy:
0 0
Muirhead’s Theorem: If (α)  (α ) then [α] ≥ [α ]. There is equality if
0
and only if (α) and (α ) are identical or all the xi are equal.

For example:
• For positive x, y, prove that 2(x5 + y 5 ) ≥ (x2 + y 2 )(x3 + y 3 ).
Solution: We have:

(5, 0)  (3, 2)

therefore by Muirhead’s inequality:


1 5 1 3 2
(x + y 5 ) ≥ (x y + x2 y 3 )
2! 2!
so

x5 + y 5 ≥ x3 y 2 + x2 y 3

hence

2(x5 + y 5 ) ≥ x5 + x3 y 2 + x2 y 3 + y 5 = (x2 + y 2 )(x3 + y 3 )

• BMO1 2002 Round 1 Q3: For positive real x, y, z such that x2 + y 2 +


z 2 = 1 prove that
1
x2 yz + xy 2 z + xyz 2 ≤
3
Solution: We have

(4, 0, 0)  (2, 1, 1)

and

(2, 2, 0)  (2, 1, 1)

3
therefore
2 4 2 2
(x + y 4 + z 4 ) ≥ (x yz + xy 2 z + xyz 2 )
3! 3!
and
2 2 2 2 2
(x y + y 2 z 2 + z 2 x2 ) ≥ (x yz + xy 2 z + xyz 2 )
3! 3!
i.e.
x4 + y 4 + z 4 ≥ x2 yz + xy 2 z + xyz 2
and
x2 y 2 + y 2 z 2 + z 2 x2 ≥ x2 yz + xy 2 z + xyz 2
Adding the first to twice the second gives:
(x4 + y 4 + z 4 ) + 2(x2 y 2 + y 2 z 2 + z 2 x2 ) ≥ 3(x2 yz + xy 2 z + xyz 2 )
Rearranging:
(x2 + y 2 + z 2 )2 ≥ 3(x2 yz + xy 2 z + xyz 2 )
But x2 + y 2 + z 2 = 1, so
1 ≥ 3(x2 yz + xy 2 z + xyz 2 )
1
x2 yz + xy 2 z + xyz 2 ≤
3
Exercises

• BMO1 1996 Round 1 Q5: Let a, b and c be positive real numbers.


Prove that
4(a3 + b3 ) ≥ (a + b)3
and
9(a3 + b3 + c3 ) ≥ (a + b + c)3

• For non-negative real x, y, z, prove that the RMS Mean is greater than
or equal to the Arithmetic Mean, which in turn is greater than or equal
to the Geometric Mean, which in turn is greater than or equal to the
Harmonic Mean. I.e.:
r
a2 + b 2 + c 2 a+b+c 3
≥ ≥ (xyz)1/3 ≥
3 3 1/x + 1/y + 1/z

You might also like