I am somewhat going out "on a limb" here, since I am much more grounded in the physics side of things than I am in mathematics. Nonetheless, I am wondering if someone is able to comment on the following : given a connection form defined on a differential manifold, one can derive a tensorial invariant of this, the curvature form. If there is a soldering form present, one can also define torsion as an invariant of the connection form.

My question now is - are there any invariants other than curvature and torsion that can be derived from a connection on a differentiable manifold? My interest here is again motivated by physics, specifically gravitation theory; here using the torsion-free Levi-Civita connection will give us GR, and allowing torsion into the picture will give us Einstein-Cartan gravity. To me, this raises the question if - mathematically speaking - there are other options based on other connection invariants that I am not myself aware of.

I apologise if the question turns out to be meaningless - I am not a mathematician, and was not immediately able to find conclusive answers to this through a Google search.

  • 1
    $\begingroup$ there's an extension to the fundamental theorem of Riemannian geometry that states there's a unique metric connection for any given torsion; this highly depends on the ambient structure of your manifold: eg there's no such result for symplectic connections (IIRC); also note that if you force curvature to be zero, you end up with teleparallel gravity, which has basically the same degrees of freedom as ordinary general relativity recast into a different geometric framework $\endgroup$
    – Christoph
    Jan 21, 2014 at 16:53
  • 1
    $\begingroup$ The word "invariants" should be precised. If it is "invariant" relatively to diffeomorphisms, one thinks about scalar quantities. For instance, from the curvature tensor $R_{\mu\nu\alpha\beta}$, you have scalar quantities $R, R^2, R_{\mu\nu}R^{\mu\nu}, R_{\mu\nu\alpha\beta}R^{\mu\nu\alpha\beta}$. Here $R_{\mu\nu}$ is the Ricci tensor, and $R$ is the scalar curvature, which are obtained from contractions from the curvature tensor. $\endgroup$
    – Trimok
    Jan 21, 2014 at 18:25

1 Answer 1


Are there any other invariants other than curvature or torsion that can be derived from a connection on a differentiable manifold?

The $n$th Chern number $c_n$ of a manifold is a topological invariant. For example,

$$c_1=\frac{i}{2\pi}\int_{M} \, \mathrm{Tr}[\mathcal{R}]$$

where $\mathcal{R}$ is the curvature $2$-form. Technically, it is another quantity, even though it is derived from the curvature. Although, the fact that $c_1$ is a topological and not just geometrical invariant is not immediately obvious from the expression. If I interpret your 'invariance' as invariance under $\mathrm{Diff}(M)$, then scalars constructed from tensors through contractions such as,$^{\dagger}$


are invariants. Another example, interpreting 'invariance' as topological invariance, is the Chern-Simons term (used in topological quantum field theory),

$$S=\frac{k}{4\pi}\int_M \, \mathrm{Tr}\left[ A \wedge dA + A\wedge A \wedge A\right]$$

which meets your requirements as $A$ is both a gauge field, and a connection on a bundle of a manifold. There are many other examples of different types of invariants; your question is quite broad and ill-defined.

$\dagger$ The reason we check curvature scalars to see whether our space truly possesses a singularity is precisely because they are invariant under coordinate transformations, and we know inadequate coordinates can lead to the metric being singular when the space is not.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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