Continuous symmetries of manifolds must form Lie groups

← All digests

Family 304 · Topology
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

Rotations, translations and matrix groups describe familiar continuous symmetries. They are Lie groups: their parameters form smooth manifolds, and their group operations are smooth. Could a much less regular locally compact group act faithfully by continuous transformations on an ordinary finite-dimensional manifold?

The Hilbert–Smith conjecture says no. The underlying manifold is only topological, and the transformations need only be continuous. Differentiability, a preserved metric, and a free action are not assumed.

The manuscript claims the conjecture in every finite dimension. Its central step excludes faithful actions of the additive group of $p$-adic integers. The final contradiction has an elementary numerical shape: a fixed nonzero invariant cannot be divided into arbitrarily many equal parts if those parts must lie in a lattice with a fixed denominator bound.

2. Background and essential definitions

The groups and spaces in the statement

A topological $n$-manifold is a space locally homeomorphic to $\mathbb R^n$. The theorem assumes it is connected, Hausdorff and second countable, and has no boundary. Hausdorff means that distinct points have disjoint neighborhoods; second countable means that the topology has a countable basis. These are standard conditions that exclude various pathological spaces.

A topological group is locally compact if every point has a compact neighborhood. A Lie group is a finite-dimensional smooth manifold whose multiplication and inversion are smooth. A countable discrete group can also be a Lie group of dimension zero; the conclusion does not require a positive-dimensional connected group.

An action is jointly continuous if the map

$$G\times M\longrightarrow M,\qquad(g,x)\longmapsto gx$$

is continuous in the pair of variables. It is faithful when only the identity fixes every point of $M$. A faithful action can have many fixed points for individual elements. This is weaker than a free action, which requires every point stabilizer to be trivial.

The p-adic integers

For a prime $p$, the additive group of $p$-adic integers is

$$\mathbb Z_p=\varprojlim_k\mathbb Z/p^k\mathbb Z.$$

An element is a compatible sequence of residues modulo $p,p^2,p^3,\ldots$. Two elements are close when they agree modulo a large power of $p$. This makes $\mathbb Z_p$ compact and totally disconnected, with arbitrarily small nontrivial open subgroups $p^k\mathbb Z_p$.

It is not a Lie group. Classical transformation-group theory reduces Hilbert–Smith to excluding its faithful actions. The reduction is about this specific compact group; it does not say that every non-Lie locally compact group is itself $\mathbb Z_p$.

Signature as a discrete quantity

For a nonsingular real symmetric matrix, the signature is the number of positive eigenvalues minus the number of negative eigenvalues. It is an integer and is additive under block sums. Witt groups package this kind of information by treating certain cancelling forms as zero.

The paper uses a sheaf-theoretic version. A sheaf records compatible local information over a space, and a direct image along a map can encode the topology of its fibers. Duality supplies analogues of symmetric forms. The essential feature for the argument is an additive invariant whose rational values still have uniformly bounded denominators.

3. Development history and previous partial results

Hilbert’s fifth problem asked how far differentiability assumptions could be removed from the theory of continuous groups. Gleason and Montgomery–Zippin resolved the locally Euclidean group formulation in 1952, and Yamabe extended the structure theory of locally compact groups.

Hilbert–Smith is the action problem: the group is not assumed locally Euclidean, but acts faithfully on a manifold that is. The distinction is substantial. Properties of the space on which a group acts do not immediately become local coordinates on the group itself.

The one- and two-dimensional cases were classical. Pardon proved the three-dimensional case in 2013 using local three-manifold topology, incompressible surfaces and mapping class groups. His result allowed arbitrary stabilizers rather than replacing the action by a free or smooth one. Pardon’s theorem

Another route added regularity to the transformations. Bochner–Montgomery established the differentiable precedent. Repovš–Ščepin proved the Lipschitz case on Riemannian manifolds; further results covered specified Hölder and quasiconformal actions. These strengthen continuity by controlling distortion of distances.

Dimension-raising results of Yang and Bredon–Raymond–Williams helped motivate these metric approaches. For example, averaging a metric under a sufficiently regular action can constrain the dimension of the orbit space, conflicting with the dimension increase forced by a hypothetical $p$-adic action. Such arguments need their metric and dimension hypotheses.

There are also more specialized advances, including symplectic results using Floer theory. The present manuscript claims to remove action-regularity restrictions and treat every finite dimension without an assumed dimension bound on the orbit space.

Its signature method has a separate history. Verdier duality, Balmer’s Witt groups, Woolf’s sheaf-theoretic constructions, and Ranicki–Weiss’s local chain methods provide precedents for extracting topological invariants from local duality. The new ingredient claimed here is a uniform integral comparison strong enough to keep denominators bounded for the continuous maps arising from the action.

4. What the new manuscripts claim

The September 23, 2026 manuscript states that a locally compact second-countable Hausdorff group acting faithfully and jointly continuously on a connected Hausdorff second-countable finite-dimensional topological manifold without boundary must be a Lie group.

The manifold may be noncompact, nonorientable, or nontriangulable. Fixed points and arbitrary stabilizers are allowed. For positive dimension, the main intermediate statement is that every jointly continuous action of $\mathbb Z_p$ on such a manifold has nonzero kernel and therefore factors through a finite quotient of $\mathbb Z_p$.

The last implication uses the closed-subgroup structure of $\mathbb Z_p$: a nonzero closed subgroup is open and has finite index. In dimension zero, a connected manifold is a point, and a faithful acting group is trivial.

The manuscript also records a dynamical consequence. Suppose a homeomorphism $f$ is almost periodic in the sense that its powers have compact closure in the homeomorphism group with the compact-open topology. Some positive iterate $f^m$ is then the time-one map of a continuous real flow:

$$\Phi:\mathbb R\longrightarrow\operatorname{Homeo}(M), \qquad \Phi(1)=f^m.$$

The compact closure of the powers becomes a compact abelian Lie group. Its identity component is a torus of finite index, and some iterate lies in that torus and hence in a one-parameter subgroup. The statement concerns an iterate, not necessarily $f$ itself.

5. Why the advance matters

Importance: 5/5, provisional. If correct, the theorem resolves a foundational transformation-group problem in full finite-dimensional scope, with continuity as the only regularity of the action.

It would establish that a manifold constrains its locally compact symmetry groups much more strongly than local compactness alone does. The group must possess finite-dimensional smooth group structure even when the action is not assumed smooth.

The theorem does not say that the full homeomorphism group of a manifold is a Lie group: that group generally does not satisfy the local compactness hypothesis. Nor does it produce a smooth structure on the manifold or smooth every transformation. The rigidity concerns the acting group under the stated assumptions.

The method would also provide a way around the irregularity of orbit spaces. It obtains a discrete obstruction using well-controlled source and test spaces, while allowing the intervening quotient to be topologically complicated.

6. The main idea

Localize the hypothetical action

Assume a faithful $\mathbb Z_p$-action exists. Using the classical local reduction based on Newman’s rigidity theorem for periodic transformations, the paper passes to an open subgroup acting faithfully and orientation-preservingly on a connected invariant open subset $O$ of one coordinate chart.

Choose an odd integer $d>n$, with $d\geq3$, and enlarge the local space to

$$W=O\times\mathbb R^{d-n},$$

with the action trivial on the added coordinates. This is a device for making odd-dimensional sphere tests available; it is not a regularity assumption on the original manifold.

Compactify $W$ by one point, take its orbit space, and test that quotient by maps into finite-polyhedron spaces. A collapse map from $S^d$ supplies a fixed compact source. The quotient itself is not assumed to be a manifold or to have a convenient dimension bound.

Build a lattice before dividing anything

The manuscript constructs an abelian group $E_d(S^d)$ of signature-type classes and an orientation class $u_d$. After rationalization,

$$E_d(S^d)\otimes\mathbb Q=\mathbb Q u_d.$$

More importantly, it claims a positive integer $L_d$ such that every original class has rational image in

$$\frac1{L_d}\mathbb Z u_d.$$

This is the fixed lattice. Knowing only that the rationalized group is one-dimensional would be insufficient: rational numbers have arbitrarily large denominators.

The categories are generated by proper images of finite polyhedra under arbitrary continuous maps. Such images need not be constructible along a finite stratification, and their sheaf stalks need not have finite dimension. The paper develops duality, localization and homotopy invariance within this generated setting.

A comparison with more controlled semialgebraic maps first supplies ordinary signature integrality. Cutoff arguments and a doubling construction then bound the discrepancy between the two theories by a fixed power of two. A finite cover of the sphere gives the uniform denominator bound. Here “integral class” means a class before tensoring the abelian group with $\mathbb Q$; the sheaf coefficients themselves are real.

Faithfulness produces arbitrarily many equal parts

Characters of $\mathbb Z_p$ take values in roots of unity of $p$-power order. For each $k$, the paper partitions the character information into $p^k$ pieces and constructs corresponding integral signature classes.

There is a subtle extension issue: a projector defined away from the collapsed point need not extend as a direct-summand decomposition over that point. The manuscript instead uses the associated involution $2e-1$ and a sum of forms on an existing object. Localization lifts the resulting class without assuming that the projector’s image extends.

Faithfulness supplies a region of the orbit space with $p^k$ finite-quotient sheets. An odd-sphere homotopy argument concentrates a test map in that region after composing with a positive-degree sphere map. A coefficient symmetry then cyclically permutes the character pieces, making their rational classes equal.

The auxiliary degree may depend on the construction. It is cancelled only after rationalization; the original integral classes remain in the same fixed lattice. Their rational images sum to $4u_d$, so each is

$$\frac4{p^k}u_d.$$

For $p^k>4L_d$, this is a positive quantity smaller than the smallest positive lattice step $u_d/L_d$. It cannot lie in the claimed lattice. The contradiction excludes the faithful $p$-adic action, and the classical reduction then gives Hilbert–Smith.

7. Scope, limitations, and open questions

All topological hypotheses in the theorem matter to its stated scope: local compactness and second countability of the group; connectedness, finite dimension and the standard separation/countability conditions for the boundaryless manifold; and joint continuity and faithfulness of the action.

The proof does not replace faithfulness by freeness. Nor does it assume that orbit spaces admit triangulations, have dimension at most that of the manifold, or carry the duality theory used on the actual manifold sources and test spaces.

The odd-dimensional stabilization is an auxiliary construction. It is not a restriction to odd-dimensional original manifolds, and the manuscript treats both odd congruence classes modulo four in its invariant.

The conclusion leaves separate questions about smoothing an action, choosing invariant geometric structures, and quantitative bounds on symmetry groups. Its flow consequence likewise gives existence of a suitable iterate and a continuous flow, not an effective formula for either.

8. Sources and further reading

Pardon’s three-manifold theorem provides the earlier unrestricted low-dimensional result and the local reduction used here. Repovš–Ščepin’s Lipschitz theorem illustrates the distinct regularity-based approach.

Woolf’s Witt groups of sheaves gives background for the signature machinery. The source manuscript includes the classical locally compact group reduction, the precise generated categories, and the integral comparison on which its new obstruction depends.

Source manuscripts

• M1. OpenAI, The Hilbert–Smith conjecture in every finite dimension (September 23, 2026).

Sources consulted on 22 October 2026 at repository revision 1581d4c. Download the manuscript citations.