What does an exterior (second) derivative such as in $d^2=0$ have to do with second derivatives as in single- or multi-variable calculus? Is this a correct start:

Calculus derivatives are good for Taylor expansions (and thus optimization), and curvature.

Exterior derivatives are needed for integration -- and are essential for generalizing the Fundamental Theorem of Calculus (i.e., Stokes theorem).


Good question! Here's a start. The ordinary derivative in one-variable calculus is a Lie derivative along a special vector field on $\mathbb{R}$; in particular, it is not a special case of the exterior derivative. The exterior derivative is instead some kind of "universal derivative": it records all of the information you would need to determine the derivative of a function along any vector field, for example. In particular, unlike the ordinary derivative, the exterior derivative of a function is a different kind of object, namely a $1$-form. Roughly speaking, a $1$-form is "the kind of thing that pairs with a vector field to return a number," so you can see the relationship there to what I said above.

  • $\begingroup$ Thank you for the starters -- but isn't it the opposite? The calculus derivative contains all information, while exterior derivatives are projections that extract interesting info? Think of the second derivative of a function f: R^2 -> R^3 (and the associated area form). I like to think of a derivative as a (multi-)linear functional valued map (with many symmetries). My best intuit is that the reduction (projection) onto the alternating part is exactly what makes a generalization of the fundamental theorem (to Stokes' theorem) natural. How far off is this? I am thinking of the next time I am t $\endgroup$ – user128171 Feb 12 '14 at 7:17
  • $\begingroup$ What is the relation between 2nd derivative of a 0-form in R and lie derivative of exterier derivative of the same 0-form? Are they somehow related? $\endgroup$ – tantuni May 25 '18 at 7:52

That $d^2 = 0$ is probably something you were already taught in vector calculus.

For instance, you probably remember that $\nabla \times \nabla \phi = 0$ for $\phi$ a scalar field, or that $\nabla \cdot (\nabla \times E) = 0$ for $E$ a vector field. It's a good exercise to show that both of these can be written as $d^2 f = 0$ and $d^2 E = 0$.

Of course, you know that $\nabla$ can be written in terms of partial derivatives:

$$\nabla = (dx) \frac{\partial}{\partial x} + (dy) \frac{\partial}{\partial y} + dz \frac{\partial}{\partial z}$$

You should also know that the above vector calculus identities rely upon the equality of mixed partial derivatives:

$$\nabla \times \nabla f = \frac{\partial^2 f}{\partial x \partial y} dx \times dy + \frac{\partial^2 f}{\partial y \partial x} dy \times dx + \ldots$$

But since $dx \times dy = -dy \times dx$, the equality of mixed partial derivatives reduces this to zero. The logic is similar for the vector field case.


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.