Deriving the Graph Area Factor

Why a Graph Can Have More Area Than Its Projection

A surface given as a graph sits above a region in the \(xy\)-plane. The region is its horizontal projection, but the graph can be tilted or curved above it. A small patch of the surface may therefore have more area than the corresponding patch in the plane.

The amount of local area expansion depends on the slopes in the \(x\)- and \(y\)-directions. We will derive that dependence directly from the tangent vectors to the graph. The result is a dimensionless factor that converts a small projected area into the corresponding surface area to first order.

Parametrizing the Graph

Let \(D\) be a region in the \(xy\)-plane, and suppose \(f:D\to\mathbb R\) has continuous first partial derivatives. Its graph is the surface consisting of points \((x,y,f(x,y))\), for \((x,y)\in D\). A parametrization is

Changing \(x\) while holding \(y\) fixed traces a coordinate curve on the surface. Its tangent vector is the partial derivative \(\mathbf r_x\). Changing \(y\) while holding \(x\) fixed gives the other coordinate tangent vector, \(\mathbf r_y\).

Here \(f_x\) and \(f_y\) are evaluated at the point \((x,y)\). The first two components record movement in the projected plane; the third component records the corresponding change in height. In particular, \(f_x\) is the slope along the \(x\)-coordinate curve, and \(f_y\) is the slope along the \(y\)-coordinate curve.

Deriving the Factor from the Tangent Vectors

At a fixed point, the two tangent vectors span a parallelogram. A small rectangle in the parameter plane is mapped, to first order, to a parallelogram spanned by scaled copies of these vectors. The cross product gives the area of that tangent parallelogram, so its magnitude is the local area expansion.

Compute the cross product using the coordinate formula. The first component is \(0\cdot f_y-f_x\cdot1=-f_x\); the second is \(f_x\cdot0-1\cdot f_y=-f_y\); and the third is \(1\cdot1-0\cdot0=1\). Thus

The squared length of this vector is the sum of the squares of its components. Taking the nonnegative square root gives the graph area factor:

The factor describes local area, not necessarily the exact area of a finite patch obtained by multiplying one factor by the patch’s projected area. If the slopes vary across the patch, the local factor varies as well. The surface-area formula accounts for that variation by integrating the factor over the projected region; the derivation here explains why that is the relevant quantity.

Why the Two Slopes Do Not Simply Multiply

The tangent vector in the \(x\)-direction has length \(\sqrt{1+f_x^2}\), and the one in the \(y\)-direction has length \(\sqrt{1+f_y^2}\). Those lengths describe the stretching of the two coordinate edges. But the tangent vectors need not be perpendicular, so multiplying their lengths does not in general give the area of the parallelogram they span.

This can also be seen from the area identity for two vectors: the square of the parallelogram’s area is the product of their squared lengths minus the square of their dot product. Here, \(\mathbf r_x\cdot\mathbf r_y=f_xf_y\), so

The final subtraction removes the contribution from the fact that the tangent directions are not generally perpendicular. It is precisely why the two one-direction length factors should not be multiplied as though they formed a right angle.

Worked Example: A Tilted Plane

Worked Example: A Curved Graph at a Point

Worked Example: Comparing Directional Slopes

How Slope Changes Area

If \(f_x=0\) and \(f_y=0\), both coordinate tangent vectors are horizontal unit vectors and the area factor is \(1\). If only one partial derivative is nonzero, the graph tilts in one coordinate direction, and the factor reduces to \(\sqrt{1+f_x^2}\) or \(\sqrt{1+f_y^2}\), respectively. If both are nonzero, their squared slopes both contribute to the factor.

Reversing the sign of either slope does not change the local area factor, because the formula uses squares. A slope of \(3\) and a slope of \(-3\) produce the same amount of local stretching in that direction, even though the graph tilts in opposite senses. Larger absolute slopes increase the factor; they cannot decrease it below \(1\).

One common error is to multiply the coordinate-direction length factors. For the plane \(z=x+y\), this mistaken calculation would be

But the tangent vectors are not perpendicular, so \(2\) is not the parallelogram area. The correct calculation is their cross-product magnitude, \(\sqrt{3}\). In general, use \(\sqrt{1+f_x^2+f_y^2}\), not \(\sqrt{1+f_x^2}\sqrt{1+f_y^2}\).