From a moving point to a moving sheet
In the curves rung you learned to see a curve not as a frozen picture but as a moving point gamma(t), traced over one clock value t. A surface is the natural next step: instead of one clock you turn two knobs. A surface parametrization is a vector-valued function of two parameters, written x(u, v) = (x(u,v), y(u,v), z(u,v)), and as the pair (u, v) ranges over a flat patch of the plane, the point x(u, v) sweeps out a sheet sitting in space. Freeze v and let u run: you get a curve drawn on the surface, a u-curve. Freeze u and let v run instead: a v-curve. The whole surface is woven from these two families of coordinate curves, like the warp and weft of a fabric.
Just as a curve needed regularity — a velocity that never stalls — a surface needs its own honesty condition. At a point we differentiate x in each direction separately to get two velocity vectors: x_u = (partial x / partial u) and x_v = (partial x / partial v). Each is tangent to its own coordinate curve. We call the patch a regular surface when x_u and x_v are never parallel — equivalently, their cross product x_u x x_v is never the zero vector. That single demand guarantees the two coordinate directions genuinely span a flat two-dimensional plane at the point, with no collapse to a line or a spike. It is the surface's promise that, up close, it really does look two-dimensional.
The tangent plane: the surface's best flat guess
Pick a point P on the surface. The two velocity arrows x_u and x_v live at P, and because they are not parallel they span a flat plane through P — the tangent plane. This is the exact surface analogue of the tangent line to a curve: the best flat approximation that kisses the surface at P. Any direction you might set off walking along the surface from P is, to first order, a combination of x_u and x_v. A tangent vector is therefore w = a x_u + b x_v for some numbers a and b — the pair (a, b) is just how much of each coordinate direction your step contains.
Here is the crucial subtlety, and the reason a surface is harder than a curve. The two basis arrows x_u and x_v are generally not unit length, and generally not perpendicular to each other. A step that reads as '(a, b) = (1, 0)' in coordinates — one unit along u — might be a long stride or a baby step in actual space, depending on how stretched the u-curves are there; and the u and v directions might meet at a slanted angle rather than a clean right angle. So the coordinate label (a, b) does not tell you the real length of your step or the real angle between two steps. Something has to translate coordinate bookkeeping into honest geometry. That translator is the first fundamental form.
Three numbers E, F, G that hold all the geometry
The fix is wonderfully economical. To measure lengths and angles among tangent vectors, all you need is the dot products of the two basis arrows with each other — and there are only three of those. We name them E, F, G: E = x_u . x_u (the squared length of the u-step), G = x_v . x_v (the squared length of the v-step), and F = x_u . x_v (which records the angle between them, since u . v = |u||v| cos theta). These three functions of (u, v) are the entire data of the first fundamental form. E tells you how stretched the u-direction is, G how stretched the v-direction is, and F whether the coordinate grid is slanted — F = 0 exactly when the u-curves and v-curves cross at right angles.
Now watch the payoff. Take any tangent step w = a x_u + b x_v. Its squared length is just the dot product w . w, and expanding it with E, F, G gives |w|^2 = E a^2 + 2F ab + G b^2 — a tidy quadratic in your coordinate components (a, b). If you let the step become infinitesimal, with components (du, dv), this becomes the line element ds^2 = E du^2 + 2F du dv + G dv^2: the squared length of a tiny displacement on the surface. This formula is the first fundamental form. It is the surface telling you, at every point and in every direction, exactly how a coordinate change of (du, dv) translates into real distance.
the first fundamental form E = x_u . x_u F = x_u . x_v G = x_v . x_v length of a tangent step w = a x_u + b x_v : |w|^2 = E a^2 + 2F ab + G b^2 the line element (tiny displacement (du, dv)) : ds^2 = E du^2 + 2F du dv + G dv^2
A worked example: the unit sphere
Let us make this concrete on a surface you can hold in your hand. Parametrize the unit sphere by latitude and longitude: x(u, v) = (sin u cos v, sin u sin v, cos u), where u is the angle down from the north pole (the polar angle) and v is the angle around (longitude). The u-curves are the meridians, the v-curves are the circles of latitude. Differentiating gives x_u = (cos u cos v, cos u sin v, -sin u) and x_v = (-sin u sin v, sin u cos v, 0). A short computation of the three dot products yields E = 1, F = 0, and G = sin^2 u.
Read what those numbers are saying. E = 1 means a step in u — moving along a meridian — has true length equal to its coordinate change, which is right: u is literally an angle, and on a unit sphere arc length along a meridian equals the angle subtended. F = 0 means meridians and latitude circles cross at right angles, which you can see on any globe. And G = sin^2 u is the interesting one: the real length of a longitude step depends on where you are. Near the equator (u = pi/2) we have sin u = 1, so longitude steps are full size; near the north pole (u near 0) sin u shrinks toward 0, so a full 360-degree change in v traces an ever-tinier circle. That single factor sin^2 u is the whole reason the lines of longitude crowd together and meet at the poles.
What the first fundamental form lets you compute
Once you hold E, F, G, you can measure everything a surface-dweller could ever measure — without ever leaving the surface. Length of a curve on the surface: feed its (u(t), v(t)) into ds^2 and integrate the speed, exactly the arc-length integral from the curves rung, now reading the metric off E, F, G. Angle between two tangent directions: take their inner product using E, F, G in place of the ordinary dot product, then read off cos theta. And area: a tiny coordinate rectangle du by dv maps to a tiny parallelogram on the surface with area |x_u x x_v| du dv, and a clean identity turns that into square root of (EG - F^2) du dv. So surface area is the integral of square root of (EG - F^2) over the parameter region.
- Parametrize the patch as x(u, v) and compute the two tangent vectors x_u and x_v by partial differentiation; confirm it is a regular surface (x_u x x_v is never zero).
- Form the three dot products E = x_u . x_u, F = x_u . x_v, G = x_v . x_v — this is the first fundamental form.
- Get length of a surface curve by integrating ds^2 = E du^2 + 2F du dv + G dv^2 along it; get the angle between directions from the E, F, G inner product.
- Get area by integrating square root of (EG - F^2) over the parameter region — the local stretch factor of the parametrization.
Notice the profound word in all of this: intrinsic. Every one of these quantities — length, angle, area, and soon the straightest paths, the geodesics — is computed from E, F, G alone, never from how the surface happens to sit in the surrounding space. A two-dimensional creature confined to the surface, who could never peek into the third dimension, could still measure all of them. This is the extrinsic-versus-intrinsic distinction, first met for curves, now sharpened to the central theme of surface geometry: the first fundamental form is the surface's inner world, the geometry it owns regardless of how it is bent into space.
Where this opens out
The honest boundary to draw is this: the first fundamental form captures measurement on the surface, but not yet how the surface bends in space. A flat sheet of paper and that same paper rolled into a cylinder have the identical first fundamental form — you can roll paper without stretching, so all intrinsic distances and angles are preserved — even though one is flat and one is visibly curved. So E, F, G alone cannot see the difference between flat and rolled. To detect bending we will need a second machine, the second fundamental form, which watches how the surface's normal direction tilts as you move — and that is precisely the subject of the next guide, on principal curvatures.
And yet — here is the spark the rest of the rung lives on — the first fundamental form is more powerful than the paper example suggests. Gauss's astonishing Theorema Egregium will reveal that one particular measure of bending, the Gaussian curvature K, can be computed from E, F, G alone, even though it looks like it should need the way the surface sits in space. That is why a flatlander really can detect curvature from inside, why the sphere's map must distort, and why this modest table of three dot products is the seed of everything ahead, all the way to the Gauss-Bonnet theorem, where curvature and topology shake hands. The full proofs lie a few guides up the ladder; what you have built here is the language they will all be spoken in.