Muirhead's inequality compares two symmetric sums of monomials by looking only at their exponents. It works for symmetric, homogeneous inequalities in non-negative variables. You write down two exponent vectors, run a short test, and the inequality is settled.
The statement
For non-negative reals and a vector , write
where the sum is over all permutations . With three variables, has six terms: . And is .
Say that majorizes when both are listed in decreasing order, they have the same sum, and every partial sum of is at least the matching partial sum of :
Muirhead's inequality says that if majorizes , then for all non-negative . For positive variables, equality needs all of them equal unless the two vectors are the same.
The test
Take two exponent vectors. Sort each from largest to smallest. Check the sums match. Then compare partial sums one by one. If every partial sum of the first is at least that of the second, you have the inequality.
For example, majorizes , which majorizes . That is a chain: . Both steps are Muirhead.
The proof idea
The whole thing rests on a two-variable step. If and , then
Both brackets have the same sign, so the product is non-negative. In words: pull two exponents apart while keeping their sum, and a symmetric sum goes up.
Any vector majorized by can be reached from by finitely many steps that bring two exponents closer together and keep their sum. Here is how one step acts inside the symmetric sum. Say the step changes the exponents into and leaves the others alone. Group the terms of by which variables carry the unchanged exponents. In one group, two variables carry and in some order, so the group contributes , where is the product of the other powers. The new sum has in the same group, and by the identity above the first is at least the second. Add over the groups, and one step can only lower the total. That is the proof. A related idea sits behind smoothing, where you move two variables toward each other and watch the expression change monotonically.
The signals
- The inequality is symmetric. Swapping any two variables gives the same expression.
- Every term has the same degree, or you can multiply by a constraint to make that true.
- The variables are non-negative.
- Both sides look like sums of monomials, perhaps after expanding.
- AM-GM nearly works but needs a different weight on each variable.
Example: a classic
Prove for non-negative . This is the AM-GM warm-up in the archive.
Expand: . Then majorizes , so . The total is at least .
Check with numbers: at the product is , the symmetric sum is , and . Also .
Example: after homogenising
Let with . Prove .
The two sides have different degrees, so replace the constraint by a multiplier. Since , we can write the right side as . Then the claim is . Expand the right side: . So the claim becomes , which is Muirhead with .
Homogenising is a habit worth building. Any constraint like or is a tool for matching degrees.
Example: a quartic
Prove . On the left, , because each fourth power comes from two permutations, the two that swap the other two variables. On the right, for the same reason. The vectors are and , and the first majorizes the second: , , . So , and dividing by gives the claim.
Where it goes wrong
- Comparing vectors that are not comparable. Take and . Their sums match, but while , so neither majorizes the other. At the first symmetric sum is and the second is . At the first is and the second is positive. No inequality holds in either direction.
- Using it on a cyclic sum. is not symmetric, so Muirhead says nothing about it. AM-GM often handles it.
- Using it where the inequality is sharper than Muirhead. Schur's inequality is stronger than the Muirhead version , because . At Schur is an equality, while the Muirhead version reads . See Schur's inequality: when to use it.
- Forgetting non-negativity. For negative variables the monomials change sign and the argument breaks.
- Skipping the proof when a grader expects it. Some graders want the step written out. When the case is small, such as against , write the AM-GM step in one line and nobody can object.
What to do with this
When you meet a symmetric inequality, expand it, write both sides as sums of monomials, and put the exponent vectors next to each other. If one majorizes the other you are done in a line. If neither does, the inequality is either false or needs a finer tool. The AM-GM and Cauchy-Schwarz page lists the ones to try next. The problem archive has more inequalities to practise on.
Related guides
- How to prove an olympiad inequality, step by stepMost failed inequality proofs pick a tool in the first minute. The order of attack that picks it last: equality case, degree, substitution, shape, audit.
- Nesbitt's inequality: three proofs and how to spot it in disguiseNesbitt's inequality takes two lines to prove once you know it. The skill is seeing it inside a harder problem, and choosing which of three proofs to use.
- Jensen's inequality: how to set one up on a competition problemJensen is one line of theory and three lines of setup, and the setup is where it goes wrong. The shape that calls for it, and the two checks that save you.
- AM–GM or Cauchy–Schwarz: how to know which one to useThe two inequalities solve different shapes, not different difficulties. A shape test that picks one in seconds, and the equality case that settles the ties.
- The rearrangement inequality in olympiad problems: when to use itThe rearrangement inequality says a sum of products is largest when both lists are sorted the same way. Proof by swapping, the signals, and Nesbitt by rearrangement.
Techniques in this guide
- Smoothing & NormalisationNudge the variables toward equality and watch the bound improve.
- Classical InequalitiesEquality tells you where to aim.
Competitions this guide helps with
- IMOsix proof problems, three on each of two days · 4.5 hours on each of two consecutive days
- APMOfive proof problems in one paper · 4 hours
- AIME15 questions in two parts, eight and then seven, each answer an integer from 0 to 999 · Two parts of 90 minutes each
- USAJMOsix proof problems, three on each of two days · 4.5 hours on each of two consecutive days
- USAMOsix proof problems, three on each of two days · 4.5 hours on each of two consecutive days
- BMO Round 2four problems needing full written solutions · 3.5 hours
- IOQM30 questions, each answered with a one- or two-digit number on an answer sheet (2025 paper) · 3 hours
- RMOsix problems, each needing a detailed proof · 3 hours
- INMOsix proof problems · 4.5 hours
- SMO OpenRound 1: short questions of equal weight. Round 2: full written solutions · 2.5 hours for Round 1, 4 hours for Round 2
Train it on Lemma
Every technique here is taught explicitly, in order. 128 lessons, 1237 curated problems and unlimited generated practice at six difficulties. Free to start, no card, and the paid plan opens with 3 free days.
Find your level