Why Parameter Changes Affect Normal Direction
A surface patch can be described using more than one pair of parameters. Changing the parameters may leave the geometric points of the surface unchanged while changing the order in which the parameter directions are traversed. Since the normal vector is formed from an ordered cross product of tangent vectors, changing that order can reverse the normal direction.
The sign of a two-dimensional Jacobian determinant records precisely this change. A positive determinant means that the new parameters preserve the orientation of the original parameter directions. A negative determinant means that they reverse it. The magnitude of the determinant also affects the size of the cross product, but the sign alone determines whether its direction is preserved or reversed.
Parameter Changes and Their Jacobian
Let \(\mathbf{r}(u,v)\) be a smooth regular surface parametrization. Regularity means that \(\mathbf{r}_u\times\mathbf{r}_v\ne\mathbf{0}\) at each point under consideration, so this cross product gives a nonzero normal vector.
Suppose new parameters \((s,t)\) are related to the old parameters by a smooth map \(\phi(s,t)=(u(s,t),v(s,t))\). The reparametrized surface is the composition \(\mathbf{R}(s,t)=\mathbf{r}(u(s,t),v(s,t))\). The Jacobian matrix and determinant of the parameter change are
To use the determinant to compare normal directions, the old parametrization must be regular and the parameter change must have nonzero determinant at the point. In particular, if the change is intended as a smooth reparametrization of a whole patch, it should be one-to-one on the domain being used, with a smooth inverse there. A zero determinant means that the two new parameter directions do not provide a nonsingular change of coordinates at that point.
Why the Determinant Controls the Normal
The chain rule gives the new tangent vectors as combinations of the old ones. Their cross product can then be simplified using bilinearity and the fact that a vector crossed with itself is zero.
\(\mathbf{R}_s=\mathbf{r}_u u_s+\mathbf{r}_v v_s\), and \(\mathbf{R}_t=\mathbf{r}_u u_t+\mathbf{r}_v v_t\), where the old partial derivatives are evaluated at \((u(s,t),v(s,t))\).
Distribute across the two sums. The terms involving \(\mathbf{r}_u\times\mathbf{r}_u\) and \(\mathbf{r}_v\times\mathbf{r}_v\) vanish.
Use \(\mathbf{r}_v\times\mathbf{r}_u=-(\mathbf{r}_u\times\mathbf{r}_v)\). The coefficient left over is \(u_s v_t-u_t v_s\), the Jacobian determinant.
In full, the expansion is
This proves the orientation rule: when \(\det(D\phi)>0\), the new cross product points in the same direction as the old one; when \(\det(D\phi)<0\), it points in the opposite direction. Since the determinant is nonzero, multiplying by it cannot turn the normal into the zero vector. If the determinant varies continuously on a connected domain and never vanishes, its sign cannot change there: changing from positive to negative would require passing through zero.
Worked Example: Reflection of a Rectangular Parameter Domain
Worked Example: Reflection of a Rectangular Parameter Domain
Consider the surface parametrization \(\mathbf{r}(u,v)=(u,v,u+2v)\) on the rectangle \(0\le u\le2,\ 0\le v\le1\). Its tangent vectors and normal are
Reflect the first parameter by setting \(u=2-s,\ v=t\), with \(0\le s\le2,\ 0\le t\le1\). This maps the rectangular parameter domain onto itself. Its Jacobian determinant is
The determinant predicts a reversal. Direct calculation checks it. The new parametrization and tangent vectors are
For a point check, take \(s=\tfrac12,\ t=\tfrac14\). Then \(u=\tfrac32,\ v=\tfrac14\), and both parametrizations give the point \((\tfrac32,\tfrac14,2)\). The old normal there is \((-1,-2,1)\), while the new cross product is \((1,2,-1)\). These are negatives of one another, as the determinant predicts. This verifies the direction change at the chosen point.
Worked Example: Standard Polar Parameters
Worked Example: Standard Polar Parameters
Let \(\mathbf{r}(u,v)=(u,v,u^2+v^2)\), a parametrization of a paraboloid. Use polar parameters \(u=\rho\cos\theta,\ v=\rho\sin\theta\), where \(\rho>0\). The determinant of this change is
Thus standard polar coordinates preserve the orientation wherever \(\rho>0\). To check the formula directly, the reparametrized surface and its tangent vectors are
At \(\rho=2,\ \theta=\tfrac{\pi}{6}\), the tangent vectors are \((\tfrac{\sqrt3}{2},\tfrac12,4)\) and \((-1,\sqrt3,0)\). Their cross product is
For comparison, the original normal is \(\mathbf{r}_u\times\mathbf{r}_v=(-2u,-2v,1)\). At the corresponding point, \(u=\sqrt3,\ v=1\), so this normal is \((-2\sqrt3,-2,1)\). Multiplication by the Jacobian determinant \(\rho=2\) gives \((-4\sqrt3,-4,2)\), exactly the new cross product. The positive factor confirms that the direction is preserved.
Worked Example: Reversing the Polar Angle
Worked Example: Reversing the Polar Angle
Use the same paraboloid and the standard polar parametrization \(\mathbf{P}(\rho,\alpha)=(\rho\cos\alpha,\rho\sin\alpha,\rho^2)\) on \(1<\rho<3,\ -\tfrac{\pi}{3}<\alpha<\tfrac{\pi}{3}\). Change parameters by setting \(\alpha=-\theta\) on the same symmetric parameter domain. The map reverses the angular coordinate, and its determinant is \(-1\). In the original \((u,v)\) parameters, the composite change is \(u=\rho\cos\theta,\ v=-\rho\sin\theta\), with determinant \(-\rho\).
The reparametrized surface and its tangent vectors are
At \(\rho=2,\ \theta=\tfrac{\pi}{6}\), these tangent vectors are \((\tfrac{\sqrt3}{2},-\tfrac12,4)\) and \((-1,-\sqrt3,0)\). Therefore
At the corresponding old parameters \(u=\sqrt3,\ v=-1\), the original normal is \((-2\sqrt3,2,1)\). The composite determinant is \(-\rho=-2\), and multiplying gives \((4\sqrt3,-4,-2)\), exactly the computed new cross product. The negative sign verifies that the new parametrization reverses the normal direction, even though it covers the same surface patch.
A Common Mistake: Replacing the Determinant by Its Magnitude
The absolute value of the determinant measures a scale factor, but it discards the information needed to determine orientation. For the rectangular reflection above, the determinant is \(-1\). A mistaken calculation might replace it with its absolute value and claim
This conclusion is wrong: direct calculation gave \((1,2,-1)\), the opposite vector. The correction is to retain the signed determinant:
Use the determinant itself when deciding orientation. Its absolute value is not a substitute for its sign.
What Happens When the Determinant Is Zero?
If \(\det(D\phi)=0\), the parameter change collapses the two parameter directions at that point. The cross-product formula then gives \(\mathbf{R}_s\times\mathbf{R}_t=\mathbf{0}\), even if the original normal was nonzero. There is no new normal direction to classify there using this parametrization. The sign test applies only where the original patch is regular and the parameter change has nonzero determinant.
In standard polar coordinates, for example, the determinant is \(\rho\), which vanishes at \(\rho=0\). The polar change is therefore not nonsingular at the origin. This does not contradict the positive orientation result for \(\rho>0\); it identifies the point where that coordinate change cannot supply a regular pair of parameter directions.
Check Your Understanding
Use the signed determinant and the cross-product formula to answer each question.
- If a parameter change has determinant \(5\), what happens to the direction of the normal? By what factor is the cross product multiplied?
- For the change \(u=s,\ v=-t\), calculate the determinant and state whether orientation is preserved or reversed.
- Why does a zero determinant prevent the orientation rule from determining a new normal direction?
- For standard polar parameters with \(\rho>0\), what is the Jacobian determinant, and what does its sign imply?
- Explain why replacing a negative determinant by its absolute value gives the wrong orientation conclusion.