I suppose the quotient $GL(\mathbb{R}^n)/O(\mathbb{R}^n)$ has manifold structure. Is there a name for this manifold? Google isn't helping find it.

  • $\begingroup$ How about the case $n=1$ ? $\endgroup$
    – hardmath
    May 18 at 19:48
  • 1
    $\begingroup$ Polar decomposition does it. $\endgroup$
    – Ruy
    May 18 at 20:11
  • $\begingroup$ By the QR decomposition (or RQ decomposition if you want to consider the right action of $O(n)$ on $GL(\mathbb{R}^n)$) the quotient space is in bijection with the set of upper-triangular matrices with positive terms on diagonal. Set of such matrices is just $(\mathbb{R}_{> 0})^n \times (\mathbb{R})^{\frac{n(n-1)}{2}} \cong \mathbb{R}^{\frac{n(n+1)}{2}}$. $\endgroup$
    – pmp
    May 18 at 20:30

1 Answer 1


You are almost surely familiar with the case $n=2$, where the quotient space $GL(\mathbb R^2) / O(\mathbb R^2)$ is the hyperbolic plane.

In general this example fits into two large classes of examples, and you can apply the general name for those classes.

First, for any Lie group $G$ and any closed subgroup $H < G$, the quotient $G/H$ is a manifold, moreover the left action of $G$ on itself descends to a left action on $G/H$, and with respect to this action there is an invariant Riemannian metric (unique up to rescaling). This Riemannian manifold is called a homogeneous space for $G$. In the special case $GL(\mathbb R^2) / O(\mathbb R^2)$, this Riemannian manifold is isometric to the hyperbolic plane $\mathbb H^2$ (up to rescaling).

Second, $GL(\mathbb R^n)$ is a semisimple Lie group. For any semisimple Lie group $G$, up to conjugacy there is a unique maximal compact subgroup $H < G$. The quotient space $G/H$ is a special case of a homogeneous space, called the symmetric space of $G$. For your example, $O(\mathbb R^n)$ is indeed the maximal compact subgroup of $GL(\mathbb R^n)$, and so $GL(\mathbb R^n) / O(\mathbb R^n)$ is the symmetric space of $GL(\mathbb R^n)$.

As it turns out one can also kinda/sorta give a name to the particular symmetric space $GL(n,\mathbb R) / O(n,\mathbb R)$: it is the space of normalized positive definite quadratic forms on $\mathbb R^n$, or if you like the space of volume 1 ellipsoids in $\mathbb R^n$, centered on the origin. One way to see this is that $GL(n,\mathbb R)$ acts transitively on this set of ellipsoids, and the stabilizer of the roundest ellipsoid, that is the unit volume round ball centered on the origin, is $O(n,\mathbb R)$.

  • 2
    $\begingroup$ Quibbles: of course $GL_2(\mathbb R)$ is officially reductive, not semi-simple. Also, it's $SL_2(\mathbb R)/SO(2,\mathbb R)$ that is isomorphic to the upper half-plane. Yes, $GL_w$ contains the slightly larger $O(2,\mathbb R)$, but/and the quotient is the Cartesian product of the upper half-plane with a ray (isomorphic to the positive diagonal matrices...) $\endgroup$ May 18 at 21:16
  • $\begingroup$ Yes, I have to admit to lack of precise knowledge when it comes to things like this, being more of a geometric group theorist than a Lie groupie. $\endgroup$
    – Lee Mosher
    May 19 at 1:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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