Surface Integrals of Scalar Fields:
Area Distortion & Numerical Integration

In vector calculus and physical modeling, we frequently encounter the need to accumulate physical quantities across curved surfaces. Examples range from calculating the total electrostatic charge on a curved conductor to computing heat flux or estimating the mass of a shell with variable density. These problems are solved using surface integrals of scalar fields.

Unlike integration over flat Euclidean domains, integrating over a curved surface requires translating the coordinate space of a flat 2D parameter region onto a manifold embedded in 3D space. This process introduces local geometric distortion, requiring a correction factor known as the Jacobian scaling factor to reconcile the difference between the parameter space area and the actual physical area.

To perform integration over a curved 2D surface $S$ in $\mathbb{R}^3$, we define a vector-valued position function $\vec{r}(s, t)$ that maps a flat parameter coordinate space $T \subset \mathbb{R}^2$ to 3D space:

At any point on the surface, moving infinitesimally in the directions of the parameters $s$ or $t$ defines two tangent vectors along the surface grid curves. These are calculated using partial derivatives:

Taking the cross product of these tangent vectors yields the surface normal vector $\vec{n}(s, t)$, which is perpendicular to the tangent plane at that point:

When mapping a flat grid $(ds \times dt)$ onto a curved manifold, the area is stretched, compressed, or tilted. For example, mapping a flat grid to a sphere maps the top parameter boundary line to a single point (the pole), compressing the area.

The magnitude of the normal vector represents the local area expansion factor, which serves as the Jacobian scaling factor $J(s, t)$. Geometrically, the area of the differential parallelogram spanned by $\vec{t}_s \, ds$ and $\vec{t}_t \, dt$ is:

Using this relationship, we define the surface integral of a scalar field $f(x, y, z)$ over a parameterized surface $S$ as:

If the scalar field is uniform, $f(x, y, z) = 1$, the integral yields the exact geometric surface area of the manifold:

To compute surface integrals programmatically over arbitrary shapes, we subdivide the parameter space $T$ into a grid of cells of size $\Delta s \times \Delta t$. For each cell centered at $(s_i, t_j)$, the simulation performs the following steps:

1. Partial Derivatives: Evaluates $\vec{t}_s$ and $\vec{t}_t$ using numerical central differences: $$\frac{\partial \vec{r}}{\partial s} \approx \frac{\vec{r}(s + \epsilon, t) - \vec{r}(s - \epsilon, t)}{2\epsilon}, \quad \epsilon = 10^{-3}$$
2. Jacobian Calculation: Computes the local cross product magnitude $J_{i, j} = \|\vec{t}_s \times \vec{t}_t\|$.
3. Scalar Field Evaluation: Computes $f(x, y, z)$ at the spatial coordinate $\vec{r}(s_i, t_j)$.
4. Riemann Sum Accumulation: Accumulates the discrete sum over all grid subdivisions: $$\iint_S f(x, y, z) \, dS \approx \sum_{i} \sum_{j} f\big(\vec{r}(s_i, t_j)\big) \, J_{i, j} \, \Delta s \, \Delta t$$

The interactive simulation provides a diagnostic environment containing five different parametric manifolds:

• Paraboloid (Bowl): $\vec{r}(s, t) = [s\cos t, s\sin t, s^2]^T$ where stretching increases quadratically with radius $s$.
• Hyperbolic Paraboloid (Saddle): $\vec{r}(s, t) = [s, t, s^2 - t^2]^T$, illustrating saddle curves.
• Hemisphere (Dome): $\vec{r}(s, t) = [\cos s\cos t, \sin s\cos t, \sin t]^T$, highlighting spherical coordinate compression near the poles ($t \to \pi/2$).
• Wavy Ripple: $\vec{r}(s, t) = [s, t, A\sin(\omega(s^2+t^2))]^T$, which contains oscillatory derivatives.
• Torus (Donut): $\vec{r}(s, t) = [(R + r\cos t)\cos s, (R + r\cos t)\sin s, r\sin t]^T$, illustrating outer-edge stretching and inner-edge compression.