The zoo of integrals you already know
By the time you reach this rung you have collected a whole zoo of integrals, and they look like different beasts. There is the plain definite integral from Volume I, integral from a to b of f(x) dx. There is the line integral of work, integral of F dot dr along a curve. There is the double or triple integral of a density over a region. And there is the flux integral, integral of F dot n dA across a surface. Each comes with its own notation, its own setup, its own bestiary of formulas. The first job of this rung is to notice that they are not different beasts at all.
Look at what each integral physically does. To compute a line integral you chop the curve into tiny directed segments, and to each little segment you assign a number (how much work that step costs). To compute a flux integral you chop the surface into tiny oriented patches, and to each patch you assign a number (how much fluid crosses it). To compute a triple integral you chop the solid into tiny oriented boxes, and to each box you assign a number (how much mass it holds). In every case the recipe is identical: hand the integrand a small oriented piece of space, get back a number, then add up the numbers. The integrand is a machine that eats oriented pieces and returns numbers.
The simplest forms: 0-forms and 1-forms
Start at the bottom. A [[differential-form|0-form]] is just an ordinary scalar function f(x, y, z) — nothing new at all. Why call a function a form? Because of how it gets integrated: a 0-form is 'integrated' over a 0-dimensional region, namely a single point, and integration over a point means simply evaluating the function there. A function eats a point and returns a number. That is the degenerate, simplest member of the family, and it sets the pattern: degree zero pairs with dimension zero.
Now the first genuinely new object: a [[one-form|1-form]]. At every point it is a tiny linear meter that eats one vector — one step direction and length — and returns a number. The cleanest picture is a topographic map of a hill of height f. The 1-form written df is the gadget that, when you hand it a step, tells you how much you climb. Where the contour lines crowd together the climb per step is large, so df is 'dense' there; where the land is flat df returns almost nothing. In coordinates a 1-form reads like P dx + Q dy + R dz: the dx, dy, dz are the elementary meters that report how far your step went east, north, and up, and P, Q, R are how much each direction counts.
A 1-form is exactly what you integrate along a curve. The integral of P dx + Q dy + R dz along a path is the Volume-I line integral you already know — and the old notation F dot dr was a 1-form in disguise the whole time, with (P, Q, R) the components of F. The thing dr you happily integrated against was never really a vector you dotted with; it was the trio of meters dx, dy, dz waiting for a step. Recognizing this turns a formula you memorized into an object you can picture: hand the curve its steps one by one, let the 1-form measure each, and add.
Climbing up: 2-forms and the antisymmetry that makes area
A 1-form measured a one-dimensional step. The natural next question is: what measures a two-dimensional patch — a tiny oriented parallelogram of area? That is a 2-form. At each point it is a gadget that eats two vectors (the two edges of a tiny oriented patch) and returns a number: the signed area that patch represents, weighted by the field. A 2-form is precisely what you integrate over a surface, so the flux integral F dot n dA you computed in Volume I was a 2-form all along, just dressed in vector-and-normal clothing.
To build 2-forms we need a way to multiply two 1-forms that produces an area-meter. That product, the wedge, is written with the symbol dx ^ dy, and it has one defining quirk: it is antisymmetric, meaning dx ^ dy = - dy ^ dx, and therefore dx ^ dx = 0. This is not an arbitrary rule. Swapping the two edges of a parallelogram flips its orientation (clockwise becomes counterclockwise), so the signed area must change sign — that is exactly the determinant rule for the area of a parallelogram, [a, b; c, d] giving ad - bc, which also flips sign when you swap rows. And a 'parallelogram' with two identical edges is squashed flat, area zero, which is why dx ^ dx = 0. Antisymmetry is just oriented area written in symbols.
The general object: a k-form
Now we can state the general definition cleanly. A [[k-form|k-form]] is a gadget that, at each point, eats k tangent vectors — the k edges of a tiny oriented k-dimensional box — and returns a number: the signed k-volume that box represents, weighted by the field at that point. It is linear in each input vector (double an edge, double the answer) and antisymmetric (swap any two edges, flip the sign). Those two properties are the entire content of 'differential form.' In coordinates a k-form is a sum of basic pieces like dx_i ^ dx_j ^ ... (k factors wedged together) each multiplied by an ordinary function.
Antisymmetry has a striking consequence: in n-dimensional space you cannot build a form of degree higher than n. A k-form in three dimensions exists only for k = 0, 1, 2, 3. The reason is that any wedge containing a repeated dx (like dx ^ dx) collapses to zero, and with only n distinct coordinate differentials there is no way to wedge together more than n of them without a repeat. So in R^3 the top form is the 3-form dx ^ dy ^ dz — the volume form, the thing you integrate to get ordinary volume, which is exactly the triple integral dV from Volume I wearing its true name. Above the top dimension, the world of forms is simply empty.
FORMS IN R^3 -- one entry per dimension, each is what you integrate over that dimension
degree 0 f over a POINT (evaluate f) <- a scalar function
degree 1 P dx + Q dy + R dz over a CURVE (line integral) <- 3 components
degree 2 A dy^dz + B dz^dx + C dx^dy over a SURFACE (flux integral) <- 3 components
degree 3 g dx^dy^dz over a SOLID (triple integral) <- 1 component, the volume form
Component counts 1, 3, 3, 1 are the binomial coefficients C(3,k):
the number of ways to choose k distinct differentials out of {dx, dy, dz}.Why this language is worth the trouble
You might reasonably ask: I could already compute all these integrals — why rename everything? The payoff is unification, and it is enormous. Once integrands are forms, there is a single operator d, the exterior derivative, that turns a k-form into a (k+1)-form. The next guides show that d applied to a 0-form gives the gradient, d applied to a 1-form gives the curl, and d applied to a 2-form gives the divergence — the three operators of vector calculus are one operator seen at three degrees. This is the grad-curl-div dictionary collapsing into a single rule, and it is the heart of why this rung exists.
And the rung's headline promise — that Green's, Stokes', and the divergence theorem are one theorem in disguise — follows from the same move. In the language of forms all three become the single statement that the integral of d(omega) over a region equals the integral of omega over the region's boundary. One operator, one theorem, every dimension at once. That is the destination; this guide has only set the first stone by saying clearly what the objects being integrated actually are.
Two honest cautions before you go on. First, the wedge dx ^ dy is not ordinary multiplication; it is the antisymmetric product, and forgetting dx ^ dy = - dy ^ dx is the single most common beginner's slip — the signs are the meaning, never noise to drop. Second, the picture 'a form eats a tiny box and returns its weighted volume' is the right intuition, but the precise definition lives at each point on the tangent vectors there; forms are pointwise linear-and-antisymmetric machines, and 'tiny box' is shorthand for 'the parallelepiped spanned by the input vectors.' Hold the picture, but know it is a picture.