03.02.08 · differential-geometry / manifolds

Myers-Steenrod theorem

shipped3 tiersLean: none

Anchor (Master): Myers-Steenrod 1939 Ann. Math. 40; Ebin 1970 Proc. Symp. Pure Math. 15; Palais 1957

Intuition [Beginner]

The Myers-Steenrod theorem answers a basic question about symmetry: on a Riemannian manifold, the collection of all distance-preserving maps (isometries) forms a smooth geometric object called a Lie group. This means the symmetry group is not just an abstract set but a space on which you can do calculus.

Consider the round sphere . Its isometries are rotations and reflections, forming the group . This is a 3-dimensional Lie group. The theorem guarantees that every Riemannian manifold, no matter how irregular, has an isometry group that is a Lie group. The dimension may be zero (no symmetries at all), but the structure is always there.

A striking consequence: on a smooth manifold, every distance-preserving map is automatically smooth. You cannot have a map that preserves Riemannian distances but fails to be differentiable. The metric constrains regularity.

Visual [Beginner]

A sphere with the group of rotations and reflections acting on it. Several positions of a triangle on the sphere are shown, connected by curved arrows indicating the isometry that moves one to the other. Below, a schematic of the group as a 3-dimensional manifold (itself a Lie group), with points representing individual isometries.

A sphere with a marked triangle in two positions connected by a rotation arrow, and below, a schematic 3-dimensional manifold representing the isometry group O(3).

Each point of the Lie group corresponds to one isometry of .

Worked example [Beginner]

Isometries of the round sphere . The round sphere has the standard metric inherited from . An isometry of is a map that preserves distances between all pairs of points.

The orthogonal group acts on by matrix multiplication: for , the map sends the sphere to itself and preserves all distances. So every orthogonal matrix gives an isometry.

The Myers-Steenrod theorem says these are all the isometries: . The identity component (rotations only) is , which is a 3-dimensional Lie group.

Step 1. has dimension , matching the maximum for Killing fields on .

Step 2. Each element of is either a rotation (determinant ) or a rotation composed with a reflection (determinant ).

Step 3. The identity component consists of rotations alone; it is connected and 3-dimensional.

What this tells us: the full isometry group has two components, and the Killing fields from 03.02.07 correspond to the tangent space at the identity of .

Check your understanding [Beginner]

Formal definition [Intermediate+]

Definition (Isometry group). Let be a Riemannian manifold. The isometry group is the group of all diffeomorphisms satisfying , equipped with the compact-open topology (the topology of uniform convergence on compact sets).

The compact-open topology on has as sub-basis the sets for compact and open . For a complete manifold, the Arzela-Ascoli theorem shows that closed subsets of the form are compact, making locally compact.

Counterexamples to common slips

  • A diffeomorphism need not be an isometry. The map on is a diffeomorphism but stretches the -direction, so it does not preserve the metric.
  • A homeomorphism preserving distances (in the metric-space sense) on a topological manifold need not be smooth. The Myers-Steenrod theorem requires a Riemannian metric; on a bare topological manifold with the induced distance function, the regularity result still holds because the distance function encodes the Riemannian structure.
  • The isometry group of a Lorentzian manifold (general relativity) need not be a Lie group. The Myers-Steenrod theorem applies specifically to positive-definite Riemannian metrics.

Key theorem with proof [Intermediate+]

Theorem (Myers-Steenrod). Let be a connected Riemannian manifold. Then:

(a) Every distance-preserving map is smooth and satisfies $f^ g = g$.*

(b) The isometry group , equipped with the compact-open topology, is a Lie group acting smoothly on .

Proof of (a). Let be a distance-preserving bijection. Fix a point and choose normal coordinates centred at , defined on a geodesic ball with less than the injectivity radius at . For each , the point is uniquely determined by and the angles between the geodesics from to and the coordinate axes. Since preserves all distances, maps geodesics to geodesics and preserves angles (the angle between geodesics from is determined by the law of cosines applied to the distances). In particular, maps into a geodesic ball about , and in normal coordinates is the restriction of a linear map composed with the exponential map. Since the exponential map is smooth and is its composition with a linear map and the smooth inverse exponential, is smooth on . As was arbitrary, is smooth on all of .

For the metric preservation: since is distance-preserving and smooth, and the Riemannian metric is determined by the distance function via , the pullback equals at every point.

Proof of (b). The identity component is generated by one-parameter subgroups, each corresponding to a Killing field by 03.02.07. The Lie algebra of is , the Killing algebra, which is finite-dimensional with . The exponential map of this Lie algebra gives a neighbourhood of the identity in diffeomorphic to an open set in (where ), providing the smooth structure. The group operations (composition, inversion) are smooth because they are smooth on each coordinate chart inherited from the Lie algebra.

Bridge. This theorem builds toward the full structure theory of Riemannian symmetry, where the Killing fields 03.02.07 appear again as the Lie algebra of the Lie group . The foundational reason the theorem works is that distance preservation forces the map to send geodesics to geodesics, and this is exactly the regularity mechanism that makes isometries smooth. The bridge is between the metric-space viewpoint (distances) and the differential-geometric viewpoint (smooth maps preserving the tensor ), and the smooth atlas 03.02.02 on provides the coordinate framework in which the smoothness of is verified.

Exercises [Intermediate+]

Advanced results [Master]

Theorem 1 (Regularity of isometries). If is a Riemannian manifold (), then every distance-preserving map is . The proof uses the exponential map in normal coordinates: a distance-preserving map sends geodesic balls to geodesic balls, and in coordinates it becomes the composition of smooth maps.

Theorem 2 (Ebin-Palais slice theorem). Let be the space of Riemannian metrics on a compact manifold , and let act on by pullback. For each metric , there exists a slice (a submanifold transverse to the orbit) such that the neighbourhood of in the moduli space is homeomorphic to . The slice is given by , the -orthogonal complement of the tangent space to the orbit.

Theorem 3 (Compactness of isometry groups). If is compact, then is a compact Lie group. This follows from the Myers-Steenrod theorem (Lie-group structure) plus the Arzela-Ascoli argument (proper action on compact forces to be compact).

Theorem 4 (Myers-Steenrod for non-complete manifolds). Even for non-complete Riemannian manifolds, is a Lie group. The dimension may be zero, and the action need not be proper, but the group structure is smooth.

Theorem 5 (Jordan decomposition). Every isometry of a complete Riemannian manifold can be decomposed into an elliptic part (generating a compact subgroup), a hyperbolic part (acting by translation along an axis), and a parabolic part (a unipotent action). This is the geometric analogue of the Jordan decomposition in linear algebra.

Synthesis. The Myers-Steenrod theorem is the foundational reason that the symmetry structure of a Riemannian manifold is governed by Lie-group theory; the central insight is that the metric determines both the regularity of isometries (smoothness) and the topology of the isometry group (Lie-group structure via Killing fields). This pattern generalises from single isometries to the entire moduli space of metrics, where the Ebin-Palais slice theorem identifies the local quotient with a finite-dimensional orbifold. Putting these together with the Killing-field dimension bound 03.02.07, the isometry group has dimension at most , and this is exactly the constraint that makes the moduli-space picture finite-dimensional. The bridge is that the smooth structure 03.02.02 makes the exponential map available, identifying the tangent space at each point with a neighbourhood via normal coordinates, and this identification is what converts distance preservation into smoothness.

Full proof set [Master]

Proposition (Compactness of for compact ). If is a compact Riemannian manifold, then is a compact Lie group.

Proof. By Myers-Steenrod, is a Lie group. For compactness, let be a sequence in . Fix a finite dense set . The sequences lie in the compact set , so by passing to subsequences, assume for each . The family is equicontinuous with Lipschitz constant 1 (each is an isometry). By Arzela-Ascoli, a subsequence converges uniformly on to a continuous map . The limit is distance-preserving since . By Myers-Steenrod part (a), is smooth, so . The same argument applied to shows is surjective. The subsequence converges in , proving sequential compactness.

Connections [Master]

  • Killing fields and infinitesimal isometries 03.02.07. The Lie algebra of is the Killing algebra , and each one-parameter subgroup of isometries arises as the flow of a Killing field. The dimension bound on Killing fields from 03.02.07 is what makes a finite-dimensional Lie group rather than an infinite-dimensional one.

  • Smooth maps between manifolds 03.02.03. The Myers-Steenrod theorem converts a metric-space condition (distance preservation) into a smooth-map condition (). The machinery of smooth maps 03.02.03 then applies: composition of isometries is smooth, inversion is smooth, and the group operations inherit smoothness from the smooth structure on .

  • Topological manifolds 03.02.01. The compactness of for compact relies on the topological properties established in 03.02.01: second-countability provides the finite dense set needed for the Arzela-Ascoli argument, and the Hausdorff condition ensures limits are unique.

Historical & philosophical context [Master]

Sumner Myers and Norman Steenrod proved in 1939 that the isometry group of a Riemannian manifold is a Lie group [Myers-Steenrod 1939], resolving a question that had been open since the development of Riemannian geometry by Riemann and the Lie-group theory by Lie. Their paper established both the smoothness of individual isometries and the Lie-group structure of the full group.

David Ebin extended the theory in 1970 with the slice theorem [Ebin 1970] for the action of the diffeomorphism group on the space of Riemannian metrics, providing the framework for the moduli space of Riemannian structures. Richard Palais had earlier proved a related result in 1957 on the Lie-group structure of transformation groups. The Ebin-Palais slice theorem remains the central tool in the study of moduli spaces of metrics and Einstein metrics.

Bibliography [Master]

@article{myers-steenrod1939,
  author = {Myers, Sumner B. and Steenrod, Norman E.},
  title = {The group of isometries of a {R}iemannian manifold},
  journal = {Ann. of Math.},
  volume = {40},
  number = {2},
  pages = {400--416},
  year = {1939}
}

@incollection{ebin1970,
  author = {Ebin, David G.},
  title = {The manifold of {R}iemannian metrics},
  booktitle = {Proc. Symp. Pure Math.},
  volume = {15},
  pages = {11--40},
  year = {1970},
  publisher = {AMS}
}

@book{kobayashi-nomizu-vol1,
  author = {Kobayashi, Shoshichi and Nomizu, Katsumi},
  title = {Foundations of Differential Geometry},
  volume = {1},
  publisher = {Wiley Interscience},
  year = {1963}
}

@book{lee-riemannian,
  author = {Lee, John M.},
  title = {Introduction to Riemannian Manifolds},
  edition = {2},
  publisher = {Springer},
  year = {2018}
}