On pg 73 of "Tensors, Relativity and Cosmology"

The generalized Stokes's theorem in arbitrary $N$-dimensional space is given by: $$\int_c A_mdx^m=\frac{1}{2}\int_S F_{mn}dS^{mn} \tag{1}$$

where $F_{mn}$ is the curl tensor of the vector $A_m$, $F_{mn}=A_{n,m}-A_{m,n}$ (, denotes covariant differentiation here) and $dS^{mn}$ is the contravariant tensor of an infinteseimal element of the surface $S$ $(dS^{mn}=dx^m \wedge dx^n$).

In the three-dimensional metric space the RHS of (1) is equivalent to the ordinary curl A definition

I tried to expand the RHS of (1) to obtain $$\frac{1}{2} \left(\frac{\partial A_n}{\partial x^m}-\frac{\partial A_m}{\partial x^n} \right)dx^m \wedge dx^n$$ since the Christoffel symbols vanish in the three-dimensional Euclidean metric space.

It appears that $$\frac{\partial A_n}{\partial x^m} - \frac{\partial A_m}{\partial x^n}$$ is the definition of curl A but how do I convert $\frac{1}{2}dx^m \wedge dx^n$ into dS to obtain Stokes's theorem in the ordinary vector notation?


1 Answer 1


Here, you still have the curl in antisymmetric tensor form, not a vector form. Once you are in three dimensions and there are no peculiarities with the metric, you have the correspondence:

$$(\operatorname{curl}{\bf A})_i=\frac12\epsilon_{ijk} \left(\frac{\partial A_{j}}{\partial x^k}-\frac{\partial A_{k}}{\partial x^j}\right) $$ where $\epsilon_{ijk}$ is the Levi-Civita tensor. The inverse of this expression is: $$\epsilon_{inm}(\operatorname{curl}{\bf A})_i= \left(\frac{\partial A_{n}}{\partial x^m}-\frac{\partial A_{m}}{\partial x^n}\right) $$

When you use this on your expression, you get $$\frac12\epsilon_{inm} dx^m \wedge dx^n$$ which is just the cross product you require to get to $dS_i$. Note that antisymmetric property of $\wedge$ goes together nicely with antisymmetric Levi-Civita and the one-half in front.

In terms of differential geometry, you convert a 2-form into 1-form with the Hodge star operator, which is a generalized and more formally correct version of what we did above.


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.