Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

$\newcommand{\eS}{\mathscr{S}}$ $\DeclareMathOperator{\SO}{SO}$ $\newcommand{\eP}{\mathscr{P}}$ $\newcommand{\bR}{\mathbb{R}}$ $\DeclareMathOperator{\tr}{tr}$ Let $m>1$ be an integer and denote by $\eS_m$ the vector space of symmetric $m\times m$ real matrices. The group $\SO(m)$ acts by conjugation on $\eS_m$. The space of $\SO(m)$-invariant quadratic polynomials on $\eS_m$ is spanned by the two polynomials

$$ A \mapsto \tr A^2,\;\;A\mapsto (\tr A)^2. $$

Let $\SO(m-1)$ be the subgroup of $\SO(m)$ consisting of orthogonal transformations of $\bR^m$ that fix a unit vector $\eta$.

What is the space of $\SO(m-1)$-invariant on $\eS_m$. More precisely, can one explicitly write a basis of this space?

Clearly $\tr A^2$ and $(\tr A)^2$ are such polynomials, and so are

$$ A\mapsto (A^2\eta,\eta),\;\;A\mapsto (A\eta,\eta)^2, $$

where $(-,-)$ denotes the natural inner product on $\bR^m$. Are there any more $\SO(m-1)$ invariant quadratic polynomials? (I am inclined to believe that the above is the complete list.) Update After Robert Bryant's answer I lost my initial inclination.

share|improve this question

1 Answer 1

up vote 9 down vote accepted

This is a trick question, right? It's not true when $m=2$ because, then $\mathrm{SO}(m{-}1)=\mathrm{SO}(1)$ is trivial, so that all polynomials on $2$-by-$2$ symmetric matrices are invariants, and the four quadratics you mention clearly don't span the the six-dimensional space of all quadratics.

Moreover, for $m>2$, you are missing $\tr(A)(A\eta,\eta)$. When $m>2$, these five do span the space of $\mathrm{SO}(m{-}1)$-invariant quadratic polynomials, as you can see by realizing that, when $m>2$, the space $\mathcal{S}_m$ of symmetric $m$-by-$m$ matrices splits under $\mathrm{SO}(m{-}1)$ into a direct sum $$ \mathcal{S}_m = \mathbb{R}\oplus\mathbb{R}\oplus\mathbb{R}^{m-1}\oplus \mathcal{S}^0_{m-1} $$ of $\mathrm{SO}(m{-}1)$-irreducible modules, where the only equivalences are between the first two (trivial) summands, and where $\mathcal{S}^0_{m-1}$ means the trace-zero symmetric $(m{-}1)$-by-$(m{-}1)$ matrices. (These two projections to $\mathbb{R}$ are given by the invariant linear forms $\tr(A)$ and $(A\eta,\eta)$.) [This splitting is valid for $m=2$, of course, but then the third summand is another copy of $\mathbb{R}$ and the fourth summand has dimension $0$.]

However, you switched from quadratic polynomials at the beginning to all polynomials. Was that intentional, or did you just want quadratics?

share|improve this answer
@ Robert It was not a trick question. I wrote it early in the morning when my brain hadn't yet absorbed all the caffeine. I meant quadratic polynomials, and I now see that missed one. Thank you for the answer. –  Liviu Nicolaescu Aug 15 '12 at 14:32

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.