Consider the sphere $x^2+y^2+z^2=1$. Let $\mathbf x(u,v)$ be a parameterization for the sphere.

Say I was trying to find specifically the normal vector given by $$ \frac{\partial \bf x}{\partial u} \times \frac{\partial \bf x}{\partial v} $$ (to for example, evaluate a surface integral)

Question. Why is the value of this normal vector different when I evaluate it in cartesian coordinates vs spherical coordinates?

1. Cartesian coordinates $$\mathbf{x}(u,v)= \langle u, v, \pm\sqrt{1-u^2-v^2} \rangle$$ First consider the positive case.

Then, the partial derivatives $$ \mathbf{x}_u(u,v)= (1,0,\frac {-x}{\sqrt{1-x^2-y^2}}\mathstrut ) $$

$$\mathbf x_v(u,v)=(0,1,\frac {-y}{\sqrt{1-x^2-y^2}}\mathstrut ) $$

$$\mathbf{x}_u\times\mathbf{x}_v=\frac{1}{z}\langle x,y,z \rangle$$

2. Spherical coordinates

The sphere is given by $\rho = 1$ where $\phi \in [0,\pi]$ and $\theta \in [0,2\pi]$.

Computing the cross product in spherical coordinates gives $$\mathbf x_u \times \mathbf x_v=\Biggl\vert \begin{array} 1\mathbf e_\rho & \mathbf e_\phi & \mathbf e_\theta \\ 0 & h_\phi & 0 \\ 0 & 0 & h_\theta \end{array}\Biggr\vert = \rho^2\sin(\theta) \mathbf{e}_\rho = z\langle x,y,z \rangle $$

Where $\mathbf{e}_\rho$ etc are the unit basis vectors for spherical coordinates, and $h_\rho$ etc are their respective scale factors.


The unit normal vector is the same. Different parametrizations of the same surface give different non-unit vectors.

  • $\begingroup$ Thanks for the reply! But if I was to calculate a surface integral, would both vectors give the same answer? Also, if the two vectors are different, then where did the vector $\mathbf{x}_u \times \mathbf{x}_v $ (in cartesian coordinates) get mapped to in sphericals? I guess I assumed that the mapping would preserve the same vector. $\endgroup$ – user523384 Apr 20 at 10:03
  • $\begingroup$ @user523384, yes. Different parametrizations are related by a change of variable. $\endgroup$ – Martín-Blas Pérez Pinilla Apr 20 at 10:07
  • $\begingroup$ But the vector obtained from a cross product in one coordinate system need not be mapped to the cross product of the same vectors in the new coordinate system? (Just want to make sure I haven't misunderstood something) i.e. If $a\times b = c$ then $T(a\times b)=?= T(c)$? $\endgroup$ – user523384 Apr 20 at 10:14
  • $\begingroup$ @user523384, the direction is the same. The jacobian of the cov changes the lenght. $\endgroup$ – Martín-Blas Pérez Pinilla Apr 20 at 10:17
  • $\begingroup$ Hmm but the jacobian over here would be $\rho^2 \sin(\phi)$ I'm not sure why that explains a difference (since in cartesian it's not just (x,y,z) but 1/z (x,y,z)) $\endgroup$ – user523384 Apr 20 at 10:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.