Why Parameter Order Matters
A parametrization does more than describe the points on a surface. Once an order for its parameters is specified, it also determines an orientation: the cross product of the parameter derivatives points to one of the two sides of the surface.
That choice matters whenever a problem prescribes a direction, such as upward on a graph or outward from a closed surface. A parametrization may describe the correct surface but induce the wrong orientation. The remedy is to compare its induced normal with a reference direction and use the sign of a dot product to decide whether the parameter order is suitable.
Parameter Order and the Induced Normal
Let \(\mathbf r(u,v)\) be a differentiable parametrization of a surface patch. Its parameter derivatives are tangent vectors to the surface. If they are linearly independent, their cross product is nonzero and normal to the surface.
The order is part of the orientation choice. The same point set can be parametrized with either order, but the resulting cross products point in opposite directions. In practice, calculate the cross product in the order stated, then compare it with the required direction. Do not infer the orientation from a sketch or from the order in which the coordinates appear in the parametrization.
The Dot-Product Test
Suppose \(\mathbf n_{\mathrm{ref}}\) is a nonzero reference normal at a point of the surface. For the order \((u,v)\), calculate
When the reference vector is a normal to the same surface, the sign gives a direct comparison of directions. A positive value means the two normals point to the same side; a negative value means they point to opposite sides. A zero value means that the test has not selected either side. In particular, if \(\mathbf n_{\mathrm{ref}}\) is a normal, a zero dot product would mean the induced normal is not parallel to it, so neither parameter order gives that prescribed normal direction.
The reference vector need not have length one. Multiplying it by a positive scalar changes the magnitude of the dot product but not its sign. It must, however, represent the direction being prescribed. For upward orientation, the reference vector is \(\mathbf k=(0,0,1)\); for downward orientation, it is \(-\mathbf k\). These vertical reference vectors need not themselves be normal to the surface. In that setting the test checks the sign of the vertical component: positive means upward, negative means downward, and zero means neither.
Why Swapping the Order Reverses the Induced Normal
The sign test works because the cross product records the orientation of the ordered pair of tangent vectors. Here is the complete algebraic justification.
For the order \((u,v)\), it is \(\mathbf N=\mathbf r_u\times\mathbf r_v\). Regularity ensures \(\mathbf N\ne\mathbf 0\), and the cross product is perpendicular to both tangent vectors.
The cross product is antisymmetric, so \(\mathbf r_v\times\mathbf r_u=-(\mathbf r_u\times\mathbf r_v)=-\mathbf N\). The opposite order therefore gives the opposite normal.
If \(\mathbf N\cdot\mathbf n_{\mathrm{ref}}>0\), the first order points to the reference side. After interchanging the derivatives, the dot product is \((-\mathbf N)\cdot\mathbf n_{\mathrm{ref}}=-(\mathbf N\cdot\mathbf n_{\mathrm{ref}})<0\). Thus exactly one of the two orders matches a nonzero reference normal.
This argument assumes the parametrization is regular at the point and that the reference vector genuinely represents the side to be selected. For a whole patch, a single parameter order must work across the patch: the relevant dot product must be nonzero and keep the required sign throughout. If it changes sign or vanishes, a single order does not give the prescribed orientation everywhere.
Worked Example: Choosing the Upward Order on a Graph
Worked Example: Upward and Downward Orientations
Consider the graph \(z=u^2+v^2\), parametrized by \(\mathbf r(u,v)=(u,v,u^2+v^2)\). Determine which order gives an upward normal, and which gives a downward normal.
First differentiate with respect to each parameter:
For the order \((u,v)\), calculate the cross product:
The reference direction for upward orientation is \(\mathbf k=(0,0,1)\). The dot product is
Thus the order \((u,v)\) induces an upward normal at every point. It is regular everywhere because the third component of its cross product is \(1\), so the cross product never vanishes.
For downward orientation, use the opposite order. Its induced normal is
Its dot product with the upward reference \(\mathbf k\) is \(-1\), so it points downward. Equivalently, its dot product with the downward reference \(-\mathbf k\) is \(1>0\). Both tests confirm the order choice.
Worked Example: Outward and Inward Orientation on a Cylinder
Worked Example: Selecting the Cylinder’s Side
Parametrize the cylindrical wall of radius \(2\) by \(\mathbf r(\theta,z)=(2\cos\theta,2\sin\theta,z)\). Find the order that gives the outward normal, then identify the inward order.
The parameter derivatives are
For the order \((\theta,z)\), their cross product is
The outward radial reference direction is \(\mathbf n_{\mathrm{ref}}=(\cos\theta,\sin\theta,0)\). The dot product is
Therefore \((\theta,z)\) gives the outward orientation everywhere on the cylindrical wall. For the order \((z,\theta)\), the cross product is
and its dot product with the outward radial reference is \(-2\). This order points inward. The calculations are valid for every \(\theta\) and \(z\); the two parameter derivatives are independent everywhere on this wall.
Worked Example: Checking an Outward Order on a Sphere
Worked Example: Sphere Coordinates Away from the Poles
For a sphere of radius \(3\), use the parametrization \(\mathbf r(\phi,\theta)=(3\sin\phi\cos\theta,3\sin\phi\sin\theta,3\cos\phi)\), with \(0<\phi<\pi\). Check whether the order \((\phi,\theta)\) gives an outward normal.
The derivatives are
Taking the cross product in the stated order gives
For a sphere centered at the origin, the position vector \(\mathbf r\) points radially outward. Use it as the reference normal. Since \(\mathbf r=3(\sin\phi\cos\theta,\sin\phi\sin\theta,\cos\phi)\), the dot product is
Thus the order \((\phi,\theta)\) points outward on the stated parameter range. The strict inequality uses \(0<\phi<\pi\), where \(\sin\phi>0\). At the poles, \(\sin\phi=0\) and this spherical parametrization is not regular, so the same parameter coordinates do not provide a valid oriented normal there.
Common Mistake: Reading the Sign from the Wrong Order
A frequent error is to calculate the cross product in one order, but then interpret it as though the order had been reversed. For the graph in the first example, for instance, someone might write:
This is incorrect: \((-2u,-2v,1)\) was \(\mathbf r_u\times\mathbf r_v\). The reversed order must negate every component:
Its vertical component is \(-1\), not \(1\), so it points downward. A reliable check is to write the parameter derivatives in the same order as the intended cross product before expanding the determinant, then test the result against the reference direction.
Using the Test on a Whole Patch
Pointwise calculations are not always enough. A single ordered parametrization gives a continuous induced normal on a regular parameter domain, but the prescribed direction may vary from point to point. To use one order across the entire patch, verify that the dot product keeps the required sign everywhere. In particular, for a patch that should be upward, the third component of \(\mathbf r_u\times\mathbf r_v\) must be strictly positive everywhere; for downward, it must be strictly negative everywhere.
If the test is zero at a point, switching parameter order only changes zero to zero. If its sign changes across a connected regular patch, neither fixed order matches the reference direction throughout. The right response is not to ignore the zero or choose an order by visual guesswork: check whether the reference direction is appropriate, whether the parametrization is regular, and whether one orientation can be maintained over the patch.
Matching parameter order to a prescribed normal is a small calculation with an important role: it ensures that the geometric direction required by a problem is the one actually encoded by the parametrization. A clear reference vector and a checked dot-product sign make that choice explicit and dependable.
Check Your Understanding
For each question, state the relevant cross product or dot-product test and explain what its sign implies.
- For a regular parametrization \(\mathbf r(u,v)\), what normal vector is induced by the order \((u,v)\)?
- If \((\mathbf r_u\times\mathbf r_v)\cdot\mathbf n_{\mathrm{ref}}<0\), which parameter order matches a nonzero reference normal?
- What test determines whether an induced normal is upward or downward?
- Why does a zero dot product not become nonzero when the parameter order is swapped?
- For the cylindrical parametrization \(\mathbf r(\theta,z)=(a\cos\theta,a\sin\theta,z)\), with \(a>0\), which order gives the outward orientation?