The "curly d"

symbol is used in

Mathematics for 2 separate purposes.

- For partial derivatives. If you have a real function of more than one variable, say f:
**R**^{3}→**R**, you can look at derivatives of f(x,y,z) with respect to any of the 3 variables. ∂f/∂y, for instance, is defined as follows: for any real x,z, define f_{x,z}(y)=f(x,y,z). Then f_{x,z}:**R**→**R** is a plain ordinary real function, and we can look at its derivative df_{x,z}/dy at any point y; define ∂f/∂y=df_{x,z}/dy. Note that ∂f/∂x, ∂f/∂y, and ∂f/∂z may all be defined at a point (x,y,z) but f may still not be differentiable at (x,y,z). But if f is differentiable, the gradient will satisfy ∇f=(∂f/∂x,∂f/∂y,∂f/∂z).
Used in this way, ∂f (and ∂y) will never appear on their own -- there is no concept of "differential" appearing here.

You may say "no big deal" upon hearing this. You will be right, but hey! It's traditional!

- For the boundary. If A⊂X is a subset of a topological space X, ∂A is the
*boundary* of A: it is defined by ∂A=cl(A)\int(A), where cl(A) is the *closure* of X, and int(X) is the *interior* of X. (In a topological space where the topology is described by converging sequences,) any point b∈∂X has a sequence of points s_{n}∈X converging to it and a sequence of points t_{n}∉X converging to it. This is easy to see: ∂A is the intersection of the closure of A with the closure of its complement. (In general, you need to use nets instead of sequences in the above formulation.)
∂A exactly describes what we would intuitively consider the surface of A.

#### IMPORTANT NOTE

No, it's not a

delta. Small delta looks like this:

δ, while our

tail twists the other way:

∂. It really

*is* a "curly d".

This web page brought to you by "The Everything2 Half-Crazed Alliance to Force You to Use a Better Browser Which Displays All Those Funny Symbols".