Tutorials › Calculus III › Normal Vectors From Cross Products

Parametrized Surfaces · Tutorial 828 of 999

Normal Vectors From Cross Products

Use the cross product of a surface’s tangent vectors to compute and check a normal vector on common parametrized patches.

Intermediate 9 min read

What You'll Learn

  • Define a normal vector using the tangent vectors to a parametrized surface.
  • Compute a cross product from the component formula and verify its direction.
  • Explain why the cross product is perpendicular to both tangent vectors.
  • Find normal vectors for plane, sphere, and cylinder patches.
  • Recognize when a zero cross product means the parametrization does not provide a normal this way.

From Tangent Directions to a Normal

A parametrized surface gives a point \(\mathbf r(u,v)\) for each pair of parameters in a domain. Changing \(u\) while holding \(v\) fixed, or changing \(v\) while holding \(u\) fixed, traces curves on the surface. Their tangent vectors are the partial derivatives \(\mathbf r_u\) and \(\mathbf r_v\).

At a regular point, these two tangent vectors are linearly independent. Their cross product is perpendicular to both, so it points in a direction normal to the surface. This construction is useful because it turns the surface’s parametrization directly into a normal vector, without first finding an equation for the surface.

Definition: Let \(\mathbf r(u,v)\) be a differentiable parametrization of a surface. At a parameter pair \((u,v)\), the vector \(\mathbf r_u\times\mathbf r_v\) is a normal vector when it is nonzero. The parametrization is regular at that pair precisely when \(\mathbf r_u\) and \(\mathbf r_v\) are linearly independent, equivalently when \(\mathbf r_u\times\mathbf r_v\ne\mathbf 0\).

A normal vector need not have length one. Every nonzero scalar multiple of a normal is also perpendicular to the same tangent directions. If a unit normal is needed, divide a nonzero normal vector by its length. The cross product itself is often preferable in calculations because it retains useful information about the scale of the tangent vectors.

Computing the Cross Product

Write the parametrization in components as \(\mathbf r(u,v)=(x(u,v),y(u,v),z(u,v))\). First calculate both partial derivatives, then apply the component formula for the cross product:

$$ \mathbf r_u\times\mathbf r_v = \begin{vmatrix} \mathbf i & \mathbf j & \mathbf k\\ x_u & y_u & z_u\\ x_v & y_v & z_v \end{vmatrix} = \bigl(y_u z_v-z_u y_v,\ z_u x_v-x_u z_v,\ x_u y_v-y_u x_v\bigr). $$

The determinant notation is a compact way to remember the component formula. In practical calculations, expanding each component explicitly helps prevent sign errors. The result is a vector depending on \(u\) and \(v\); evaluating it at a particular parameter pair gives a normal vector at the corresponding surface point, provided the result is nonzero.

Worked Example: A Normal to a Plane Patch

Consider the plane patch parametrized by

$$ \mathbf r(u,v)=(1,2,-1)+u(2,-1,1)+v(1,1,2). $$

Differentiate with respect to each parameter. The constant vector disappears, and the coefficient of each parameter remains:

$$ \mathbf r_u=(2,-1,1), \qquad \mathbf r_v=(1,1,2). $$

Now compute each component of the cross product:

$$ \begin{aligned} \mathbf r_u\times\mathbf r_v &=\bigl((-1)(2)-(1)(1),\ (1)(1)-(2)(2),\ (2)(1)-(-1)(1)\bigr)\\ &=(-2-1,\ 1-4,\ 2+1)\\ &=(-3,-3,3). \end{aligned} $$

Check perpendicularity by taking dot products with both tangent vectors:

$$ (-3,-3,3)\cdot(2,-1,1)=-6+3+3=0, $$
$$ (-3,-3,3)\cdot(1,1,2)=-3-3+6=0. $$

The cross product is nonzero and perpendicular to both tangent directions, so it is a normal vector to the plane at every point of this patch. The check also confirms the signs in the component calculation.

Why the Cross Product Is Perpendicular

The perpendicularity is not an additional assumption about surfaces. It is an algebraic property of the cross product. For vectors \(\mathbf a=(a_1,a_2,a_3)\) and \(\mathbf b=(b_1,b_2,b_3)\), its component formula is

$$ \mathbf a\times\mathbf b = (a_2b_3-a_3b_2,\ a_3b_1-a_1b_3,\ a_1b_2-a_2b_1). $$

Taking a dot product with \(\mathbf a\) and expanding gives pairs of equal terms with opposite signs:

$$ \begin{aligned} \mathbf a\cdot(\mathbf a\times\mathbf b) &=a_1(a_2b_3-a_3b_2)+a_2(a_3b_1-a_1b_3)+a_3(a_1b_2-a_2b_1)\\ &=a_1a_2b_3-a_1a_3b_2+a_2a_3b_1-a_1a_2b_3 +a_1a_3b_2-a_2a_3b_1\\ &=0. \end{aligned} $$

The same expansion with \(\mathbf b\) gives

$$ \begin{aligned} \mathbf b\cdot(\mathbf a\times\mathbf b) &=b_1(a_2b_3-a_3b_2)+b_2(a_3b_1-a_1b_3)+b_3(a_1b_2-a_2b_1)\\ &=a_2b_1b_3-a_3b_1b_2+a_3b_1b_2-a_1b_2b_3 +a_1b_2b_3-a_2b_1b_3\\ &=0. \end{aligned} $$

Thus \(\mathbf a\times\mathbf b\) is perpendicular to both vectors. Taking \(\mathbf a=\mathbf r_u\) and \(\mathbf b=\mathbf r_v\) proves that the cross product of the surface’s tangent vectors is perpendicular to both coordinate tangent directions. At a regular point those directions span the tangent plane, so the cross product is normal to that plane.

Why the nonzero condition matters: If \(\mathbf r_u\times\mathbf r_v=\mathbf 0\), the two tangent vectors are linearly dependent. The parametrization then fails the regularity condition at that parameter pair, and this cross product does not provide a nonzero normal there.

Normals to a Sphere Patch

For a sphere of radius \(R>0\) centered at the origin, one useful parametrization is

$$ \mathbf r(u,v)=(R\cos u\cos v,\ R\cos u\sin v,\ R\sin u). $$

The parameter \(u\) controls latitude and \(v\) controls position around the vertical axis. Differentiating component by component gives

$$ \mathbf r_u=(-R\sin u\cos v,\ -R\sin u\sin v,\ R\cos u), $$
$$ \mathbf r_v=(-R\cos u\sin v,\ R\cos u\cos v,\ 0). $$

Worked Example: A Normal to a Sphere

Take \(R=2\), \(u=\pi/6\), and \(v=\pi/3\). The surface point is

$$ \begin{aligned} \mathbf r\left(\frac{\pi}{6},\frac{\pi}{3}\right) &=\left(2\cdot\frac{\sqrt3}{2}\cdot\frac12,\ 2\cdot\frac{\sqrt3}{2}\cdot\frac{\sqrt3}{2},\ 2\cdot\frac12\right)\\ &=\left(\frac{\sqrt3}{2},\frac32,1\right). \end{aligned} $$

At these parameter values, the tangent vectors are

$$ \mathbf r_u=\left(-\frac12,-\frac{\sqrt3}{2},\sqrt3\right), \qquad \mathbf r_v=\left(-\frac32,\frac{\sqrt3}{2},0\right). $$

Using the component formula,

$$ \begin{aligned} \mathbf r_u\times\mathbf r_v &=\left( \left(-\frac{\sqrt3}{2}\right)0-(\sqrt3)\left(\frac{\sqrt3}{2}\right),\ (\sqrt3)\left(-\frac32\right)-\left(-\frac12\right)0,\ \left(-\frac12\right)\left(\frac{\sqrt3}{2}\right) -\left(-\frac{\sqrt3}{2}\right)\left(-\frac32\right) \right)\\ &=\left(-\frac32,-\frac{3\sqrt3}{2},-\sqrt3\right). \end{aligned} $$

Verify the result against each tangent vector:

$$ \begin{aligned} \left(-\frac32,-\frac{3\sqrt3}{2},-\sqrt3\right) \cdot \left(-\frac12,-\frac{\sqrt3}{2},\sqrt3\right) &=\frac34+\frac94-3=0,\\ \left(-\frac32,-\frac{3\sqrt3}{2},-\sqrt3\right) \cdot \left(-\frac32,\frac{\sqrt3}{2},0\right) &=\frac94-\frac94+0=0. \end{aligned} $$

The cross product is nonzero, so it is a valid normal at the chosen point. As a further check on its location, the point vector and this normal are scalar multiples: \(\left(-\frac32,-\frac{3\sqrt3}{2},-\sqrt3\right)=-\sqrt3\left(\frac{\sqrt3}{2},\frac32,1\right)\). This is consistent with the fact that a sphere’s radius direction is perpendicular to its tangent plane.

Normals to a Cylinder Patch

A cylinder of radius \(a>0\) around the \(z\)-axis can be parametrized by

$$ \mathbf r(u,v)=(a\cos u,\ a\sin u,\ v). $$

Its partial derivatives are

$$ \mathbf r_u=(-a\sin u,\ a\cos u,\ 0), \qquad \mathbf r_v=(0,0,1). $$

Worked Example: A Normal to a Cylinder

Let \(a=3\), \(u=\pi/6\), and \(v=2\). The point on the cylinder is

$$ \mathbf r\left(\frac{\pi}{6},2\right) =\left(\frac{3\sqrt3}{2},\frac32,2\right). $$

At this point the tangent vectors are

$$ \mathbf r_u=\left(-\frac32,\frac{3\sqrt3}{2},0\right), \qquad \mathbf r_v=(0,0,1). $$

Their cross product is

$$ \begin{aligned} \mathbf r_u\times\mathbf r_v &=\left( \left(\frac{3\sqrt3}{2}\right)(1)-0,\ 0- \left(-\frac32\right)(1),\ \left(-\frac32\right)0-\left(\frac{3\sqrt3}{2}\right)0 \right)\\ &=\left(\frac{3\sqrt3}{2},\frac32,0\right). \end{aligned} $$

The dot-product checks are

$$ \left(\frac{3\sqrt3}{2},\frac32,0\right) \cdot \left(-\frac32,\frac{3\sqrt3}{2},0\right) =-\frac{9\sqrt3}{4}+\frac{9\sqrt3}{4}=0, $$
$$ \left(\frac{3\sqrt3}{2},\frac32,0\right)\cdot(0,0,1)=0. $$

The normal is nonzero and perpendicular to both tangent vectors. It points radially away from the cylinder’s axis at this point; the calculation itself establishes the perpendicularity.

Common Mistake: A Sign Error in the Middle Component

A frequent error is to change the sign of the middle component when expanding the determinant. For example, in the plane calculation above, an incorrect expansion might give

$$ (2,-1,1)\times(1,1,2)=(-3,3,3). $$

This cannot be the correct cross product because its dot product with the first tangent vector is

$$ (-3,3,3)\cdot(2,-1,1)=-6-3+3=-6\ne0. $$

The middle component must be \(z_u x_v-x_u z_v\), not its negative. Substituting the plane’s tangent vectors gives \(1\cdot1-2\cdot2=-3\), so the corrected cross product is \((-3,-3,3)\). Its dot products with the two tangent vectors are both zero, as required.

When This Construction Applies

The procedure is the same for any differentiable parametrization: compute \(\mathbf r_u\), compute \(\mathbf r_v\), take their cross product, and check that it is nonzero. When it is nonzero, it supplies a normal to the tangent plane at the corresponding point. The plane, sphere, and cylinder examples illustrate different geometries, but none requires a different rule.

If the cross product vanishes, do not report the zero vector as a normal. A normal vector must be nonzero, and the vanishing cross product says that the two parameter tangent vectors do not span a tangent plane at that parameter pair. The parametrization may be singular there, or a different local parametrization may be needed to describe the surface regularly.

For a reliable calculation, keep the parameter dependence until after differentiating, write the cross product components in order, and use dot products as a final check. A correct nonzero result must have zero dot product with each of the two partial derivative vectors.

Check Your Understanding

Answer these questions using the cross-product definition and the checks developed above.

  1. What condition on \(\mathbf r_u\times\mathbf r_v\) makes it a normal vector obtained from the parametrization?
  2. Why does \(\mathbf r_u\times\mathbf r_v\) have zero dot product with each of \(\mathbf r_u\) and \(\mathbf r_v\)?
  3. For \(\mathbf r(u,v)=(u,v,u+v)\), compute the two tangent vectors and their cross product. Verify both dot products.
  4. What does \(\mathbf r_u\times\mathbf r_v=\mathbf 0\) tell you about the regularity condition at that parameter pair?
  5. Why is it not necessary for a normal vector obtained from a cross product to have length one?