Differential forms and wedge products

At a point of a smooth manifold, a k-form is a multilinear alternating map of k tangent vectors to a real number. Alternating means that exchanging two input vectors changes the sign. If two inputs coincide, the value must therefore equal its own negative and is zero over the real numbers. A differential k-form assigns such a map smoothly at every point. A zero-form is simply a smooth function, connecting this language to ordinary calculus.

In two coordinates, the two-form dx∧dy evaluates the ordered pair (u,v) as $u^xv^y-u^yv^x$. This is the oriented coordinate area of their parallelogram. Reversing the order reverses the sign; parallel inputs give zero. Calling this coordinate area does not yet give physical area in an arbitrary Riemannian metric. On an oriented Riemannian manifold, the metric supplies its volume form. Without a chosen orientation it supplies a volume density instead.

The wedge product combines a p-form with a q-form into a (p+q)-form. For one-forms, moving one factor past another introduces a minus sign. Thus dy∧dx=−dx∧dy and dx∧dx=0. Bilinearity then reduces coordinate computations to ordinary expansion followed by these sign rules. The pattern is graded commutativity: moving all p slots past all q slots produces pq exchanges. Two two-forms therefore commute, unlike two one-forms.

An odd-degree real form wedges with itself to zero by graded commutativity, but an even-degree form need not. For instance, on four-dimensional Euclidean space the two-form dx∧dy+dz∧dw has square $2\,dx\wedge dy\wedge dz\wedge dw$. This example prevents the mistaken rule that every differential form squares to zero. On an n-dimensional manifold, forms of degree greater than n vanish because any n+1 tangent vectors are linearly dependent and alternation forces the result to zero.

Given a smooth map F from M to N, its derivative sends tangent vectors forward. A form on N can therefore be pulled back to M by applying it to the pushed-forward vectors. This construction works even when F has no inverse and even when the source and target dimensions differ. For a function f on N, the pullback is composition f∘F. For differentials, it obeys $F^*(df)=d(f\circ F)$, recovering the chain rule.

For a surface parametrization, pulling back a two-form produces an expression in the parameter differentials that can be integrated with an orientation. A singular parametrization can collapse a nonzero area measurement to zero, reflecting a rank-deficient derivative. This is not a failure of the form. It records how the map acts on tangent directions. In applications, first identify the degree, the manifold, and the ordered vectors being measured; only then attach interpretations such as work, flux, or signed area.