Sign up ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $\omega$ be an $n$-form on $S^n$, nowhere vanishing.

Is there a Riemannian metric $g$ on $S^n$, so that its volume form is $\omega$, and $(S^n,g)$ is homogeneous? Is it unique, and if not, what transformations of $S^n$ relates such metrics?

For example, for $n=1$, the unique metric with these properties is $\text{d} u^2$, where $u=\frac{1}{2\pi}\int\omega$.

share|cite|improve this question

1 Answer 1

up vote 18 down vote accepted

There is. By Moser's Theorem, there exists a diffeomorphism $f:S^n\to S^n$ such that $f^*\omega = c\omega_{std}$ where $c>0$ and $\omega_{std}$ is the standard volume form homogeneous under $\mathrm{SO}(n{+}1)$. Now use $f$ to transfer the right multiple of the standard metric (under which $S^n$ is homogeneous) back to one for which $\omega$ is the volume form.

About uniqueness: (I forgot to answer this question on the first pass; also I am fixing a few errors in the cases below in this edit.) The answer about uniqueness of homogeneous metrics with a given volume form depends on the dimension $n$. Obviously, when $n=1$, it is unique. It's also unique (up to volume-preserving diffeomorphism) when $n=2$ (and, in fact, for all even $n$). However, it is not unique (up to volume-preserving diffeomorphism) when $n$ is odd.

The reason has to do with this: Suppose $g$ is a metric on $S^n$ that is homogeneous, and let $G$ be the identity component of the group of $g$-isometries. Then $G$ is compact, and $S^n = G/H$ where $H\subset G$ is a closed, connected subgroup of $G$ acting effectively on $S^n$. It is known, from the work of Borel, which pairs such $(G,H)$ have the property that $G/H$ is diffeomorphic to $S^n$ for some $n$, so we know all of the possibilities and can work out the isometry groups in each case.

When $n=2k$, the only possibility is $G = \mathrm{SO}(2k{+}1)$ with $H = \mathrm{SO}(2k)$. The $G$-invariant metric $g$ is then uniquely determined by specifying its total volume. (While $S^6 = \mathrm{G}_2/\mathrm{SU}(3)$ would appear to be a counterexample, in this case $\mathrm{G}_2$ is not the full isometry group of the invariant metric, $\mathrm{SO}(7)$ is.)

When $n = 4k{+}1>1$, we can have $G = \mathrm{SU}(2k{+}1)$ and $H = \mathrm{SU}(2k)$, and there is a $1$-parameter family of $G$-invariant metrics with a given total volume. The full isometry group of such a metric is $\mathrm{U}(2k{+}1)$, except that one element of the family (the metric of constant sectional curvature) has (connected) isometry group $\mathrm{SO}(4k{+}2)$.

When $n=4k{+}3$, we can have $G = \mathrm{Sp}(k{+}1)$ and $H=\mathrm{Sp}(k)$, and there is a $5$- (when $k=0$) or $6$- (when $k>0$) parameter family of $G$-invariant metrics on $S^n$ with a given total volume. Some of these are isometric though; one can get rid of $3$ of the parameters that way. Some of these metrics have slightly larger isometry groups than $\mathrm{Sp}(k{+}1)$, such as $\mathrm{Sp}(k{+}1){\cdot}S^1$ or even $\mathrm{Sp}(k{+}1){\cdot}\mathrm{Sp}(1)$, and one, the metric of constant sectional curvature, has (connected) isometry group $\mathrm{SO}(4k{+}4)$.

Finally, there is the exceptional case of $S^{15} = \mathrm{Spin}(9)/\mathrm{Spin}(7)$, which has a $1$-parameter family of inequivalent $\mathrm{Spin}(9)$-invariant metrics with the same total volume. (Even though we also have $S^{15}=\mathrm{Sp}(4)/\mathrm{Sp}(3)$, these homogeneous metrics are not comparable, except at the constant curvature metric, which belongs to both families.)

share|cite|improve this answer
Thanks, that's exactly what I needed! – Cristi Stoica Jul 23 '12 at 16:27
Here's a link to Moser's paper… – Cristi Stoica Jul 23 '12 at 16:28

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.