When a Surface Normal Fails
A flux integral uses an oriented surface element. For a parametrized surface \(\mathbf{r}(u,v)\), the cross product \(\mathbf{r}_u\times\mathbf{r}_v\) gives that oriented element: its direction is normal to the surface, and its magnitude measures the area scale of the parametrization.
If this cross product is zero, the two tangent vectors are linearly dependent, so the parametrization does not provide a normal direction at that parameter value. That fact matters, but it does not settle whether the surface is singular. A map can fail to give good coordinates at a point where the geometric surface is perfectly smooth. Conversely, a surface can fail to be smooth even when it has a unique limiting tangent plane.
Regularity and the Normal-Area Vector
Let \(\mathbf{r}:D\to\mathbb{R}^3\) be a continuously differentiable parametrization, where \(D\) is a region in the \(uv\)-plane. Write \(\mathbf{r}_u=\partial\mathbf{r}/\partial u\) and \(\mathbf{r}_v=\partial\mathbf{r}/\partial v\). At an interior parameter value, the parametrization is regular if \(\mathbf{r}_u\times\mathbf{r}_v\ne\mathbf{0}\). Equivalently, its derivative has rank 2 there.
The cross product combines the unit normal and the area scale. Where it is nonzero, the corresponding unit normal is \((\mathbf{r}_u\times\mathbf{r}_v)/\lVert\mathbf{r}_u\times\mathbf{r}_v\rVert\). For a continuous vector field \(\mathbf{F}\), the oriented flux over a regular parametrized patch is therefore
The sign depends on the chosen orientation: reversing the cross product reverses the flux. The formula applies directly on a regular patch with a consistent orientation. If the cross product vanishes, the unit normal from this formula is undefined there, and regularity must be assessed before using the formula across that point.
Why the Cross Product Measures the Surface Element
For small changes \(\Delta u\) and \(\Delta v\), the corresponding displacement vectors on the surface are approximately \(\mathbf{r}_u\Delta u\) and \(\mathbf{r}_v\Delta v\). They span a parallelogram whose area is the magnitude of their cross product.
The cross product also gives the parallelogram's oriented normal. Thus it supplies both the area scale and direction required for flux. If it is zero, the two first-order displacement directions span no area: the given parametrization has no surface element at that parameter value.
Worked Example: A Polar Parametrization of a Smooth Disk
Worked Example: A Polar Parametrization of a Smooth Disk
Parametrize the unit disk in the plane \(z=0\) by \(\mathbf{r}(u,v)=(u\cos v,u\sin v,0)\), with \(0\le u\le1\) and \(0\le v\le2\pi\). Locate the degeneracy and compute the upward flux of \(\mathbf{F}=(0,0,1)\).
Differentiate with respect to both parameters:
The cross product is
It vanishes at \(u=0\), where every value of \(v\) maps to the disk's center. This is a defect of these polar parameters, not a singularity of the disk: Cartesian parameters \(\widetilde{\mathbf{r}}(x,y)=(x,y,0)\) cover the same disk and have \(\widetilde{\mathbf{r}}_x\times\widetilde{\mathbf{r}}_y=(0,0,1)\) everywhere.
For the polar parametrization, the flux integrand is \(\mathbf{F}\cdot(\mathbf{r}_u\times\mathbf{r}_v)=u\). Hence
This agrees with the upward flux through a unit disk: the field has upward normal component 1 and the disk has area \(\pi\). The single collapsed center has zero area; the alternate Cartesian chart confirms that the underlying surface is regular.
A Parametrization Defect Can Be Removed
A point where the cross product vanishes can be harmless geometrically if another valid chart describes the same surface smoothly there. Such a point is a defect of the chosen parametrization. A regular change of parameters cannot repair a rank defect: if \((u,v)\) is replaced by a smooth coordinate change with nonsingular Jacobian, the chain rule multiplies the derivative by an invertible matrix and preserves its rank. Repairing the defect therefore requires a different chart, not merely a regular coordinate change within the defective chart.
Worked Example: A Cubic Parameter on a Plane
Consider \(\mathbf{r}(u,v)=(u^3,v,0)\), for \(-1\le u\le1\) and \(0\le v\le1\). Determine where the parametrization fails to be regular and identify the image.
The tangent vectors and their cross product are
The cross product vanishes when \(u=0\). But \(x=u^3\) ranges over \([-1,1]\), so the image is the ordinary planar rectangle \(-1\le x\le1,\ 0\le y\le1,\ z=0\). Its Cartesian parametrization \(\widetilde{\mathbf{r}}(x,y)=(x,y,0)\) is regular everywhere. The zero at \(u=0\) belongs to the cubic parametrization, not to the plane.
For example, with \(\mathbf{F}=(0,0,1)\), the cubic parameters give flux \(\int_0^1\int_{-1}^1 3u^2\,du\,dv\). Evaluating, \(\int_{-1}^1 3u^2\,du=[u^3]_{-1}^1=2\), so the flux is 2. The Cartesian chart gives \(\int_0^1\int_{-1}^1 1\,dx\,dy=2\) as well. Both calculations give the upward flux through a rectangle of area 2.
Checking Whether the Surface Itself Is Smooth
A surface singularity is a failure of the surface to be a smooth embedded surface near the point. Locally, a smooth embedded surface can be described, after a suitable choice of coordinates, as a smooth graph over a plane (with the appropriate boundary version if the surface has a boundary). Lack of a unique tangent plane is one way smoothness can fail. However, a unique tangent plane alone does not guarantee that the surface is smooth or has a regular parametrization there.
Worked Example: A Cusp with a Unique Limiting Tangent Plane
Consider the surface described by \(z^2=x^3\), \(x\ge0\), with \(y\) free. A parametrization is \(\mathbf{r}(u,v)=(u^2,v,u^3)\), for \(u,v\in\mathbb{R}\). Its image satisfies \(z^2=u^6=(u^2)^3=x^3\), and every point of the described surface is obtained by taking \(u\) with the sign of \(z\) and \(u^2=x\).
Its tangent vectors and their cross product are
At \(u=0\), the cross product vanishes. Unlike the disk and plane examples, this defect does not simply disappear by choosing another smooth chart for the same smooth surface. To see the issue, fix a value of \(y\). For each small \(x>0\), the surface contains two points, with \(z=x^{3/2}\) and \(z=-x^{3/2}\), both arbitrarily close to the origin. Thus, near the origin, the surface is not a single graph over the \(xy\)-plane. More generally, its cusp cross-section is not a smooth embedded curve at the origin. The surface fails to be a smooth embedded surface there.
Nevertheless, both branches approach the same tangent plane \(z=0\): along \(z=\pm x^{3/2}\), the ratio \(z/x=\pm\sqrt{x}\) tends to zero as \(x\) tends to zero. This example shows why a unique limiting tangent plane does not establish smoothness. The plane is a limiting tangent plane, but the cusp still fails to be a smooth embedded surface at the cusp line.
What a Degeneracy Means for Flux
At a zero of \(\mathbf{r}_u\times\mathbf{r}_v\), the normal obtained by normalizing that vector is undefined. The correct response depends on the geometry and on the extent of the degenerate set.
- If another regular chart covers the point, use that chart or divide the surface into regular patches. Check that the chosen orientations agree where patches meet.
- If the parametrization is regular except on a negligible set, a parameter-domain integral may sometimes be interpreted piecewise or as an improper integral. This requires appropriate integrability, a consistent orientation on the regular part, and a parametrization that covers the surface with the intended multiplicity.
- If the surface itself is not smooth, the usual formula for one regular oriented surface does not apply across the singularity as stated. One may need to split the surface into regular pieces, specify orientations on those pieces, and treat their common edges or singular set carefully.
An isolated parameter value is not automatically harmless, and a whole line of parameter values mapping to one point need not represent a line of surface area. Conversely, a degeneracy along a parameter set cannot be ignored merely because it looks small: determine what portion of the surface it represents and whether the rest of the flux setup is valid. In particular, a vanishing cross product at a parameter value does not justify assigning an arbitrary unit normal there.
A Reliable Workflow
Differentiate the parametrization and solve \(\mathbf{r}_u\times\mathbf{r}_v=\mathbf{0}\) on the parameter domain.
Determine which surface points those parameter values represent; do not infer the geometry from the parameter set alone.
Look for a regular alternative chart or a local smooth graph. A limiting tangent plane is useful evidence, but is not by itself a smoothness test.
Use regular patches where possible, verify their orientations, and account for any singular part or repeated coverage.
Common Mistake: Calling Every Zero a Surface Singularity
A common incorrect conclusion is: “The cross product is zero, so the surface has no tangent plane there and the surface is singular.” This confuses a property of a particular parametrization with a property of its image. In the polar disk example, \(\mathbf{r}_u\times\mathbf{r}_v=(0,0,u)\) is zero at \(u=0\), yet the image is a smooth disk with a regular Cartesian parametrization.
The reverse overclaim is also unsafe: “The limiting tangent plane is unique, so the surface is smooth.” The cusp example has the unique limiting tangent plane \(z=0\), but is not a smooth embedded surface at its cusp. The corrected approach is to test the cross product to locate loss of rank, then inspect the surface itself for a valid smooth local chart.
Check Your Understanding
For each question, distinguish carefully between the parametrization and the geometric surface.
- What does \(\mathbf{r}_u\times\mathbf{r}_v=\mathbf{0}\) establish, and what does it not establish by itself?
- For \(\mathbf{r}(u,v)=(u\cos v,u\sin v,0)\), where does the cross product vanish, and why is the disk still smooth there?
- Why can a regular change of parameters not restore rank at a point where the original parametrization has rank less than 2?
- Why does a unique limiting tangent plane fail to guarantee that a surface is smooth?
- List two conditions that should be checked before interpreting a flux integral through a parametrization that is degenerate at some parameters.