I know Gauss's divergence theorem for a vector field:


But how do you apply this to a scalar field? For example, if you wanted to find the surface integral of $z^2$ over a unit cube:


where $S$ is the surface of unit cube, how would you approach this using Gauss's divergence theorem?

  • $\begingroup$ Isn't that a dot product in the surface integral, not a cross product? $\endgroup$ – Muphrid Jun 15 '13 at 5:26
  • $\begingroup$ you must mean $S$ as the surface of the unit cube. $\endgroup$ – smiley06 Jun 15 '13 at 5:41

You can always write $ z^2 = z^2 \hat{n}.\hat{n} $ where $\hat{n} $ is the unit normal then using divergence theorem your expression becomes $$ \iiint_{|z|\leq 1} \nabla.(z^2\hat{n})dV $$

  • $\begingroup$ So does that mean the integral will become $\int_{-1}^1\int_{-1}^1\int_{-1}^1{2z}dz\space{dy}\space{dx}$? $\endgroup$ – Ataraxia Jun 15 '13 at 5:47
  • 2
    $\begingroup$ The normal vector is different for each face on the cube. $\endgroup$ – Shuhao Cao Jun 15 '13 at 5:50
  • $\begingroup$ Well in this case $ \hat{n} = \pm z^2 e_i $ for $ i \in \{1,2,3\} $, for the case of $ \pm z^2 e_3 $ you have $\pm 2z $ and otherwise $0$. $\endgroup$ – smiley06 Jun 15 '13 at 5:59
  • $\begingroup$ @smiley06 What? Then just try the double integral directly on the face $F$ that has $e_1$ as unit outward normal, see if you can get $$\iint_F z^2 dy\wedge dz$$ equal to 0. $\endgroup$ – Shuhao Cao Jun 15 '13 at 6:11
  • 1
    $\begingroup$ To be more explicit, you have to have a (nice) vector field defined everywhere on the volume. $\hat n$ is only defined (piecewise smoothly) on the surface, not inside. Indeed, there is no continuous unit vector field inside agreeing with $\hat n$ on the surface. $\endgroup$ – Ted Shifrin Jun 15 '13 at 11:27

Indeed we can use divergence theorem if we could find an $F$ such that: $$ \iint_S g\,dS = \iint_S F\cdot n\,dS.\tag{1} $$

Though the example you gave is not very illustrative it seems. Say the unit cube is $D = [0,1]\times[0,1]\times [0,1]$. The faces perpendicular to the $x$-axis have unit outward normals $(\pm 1,0,0)$, if we want $F\cdot (1,0,0) = z^2$ on the face $\{x = 1\}\cap D$, while $F\cdot (-1,0,0) = z^2$ on the face $\{x = 0\}\cap D$ as well?

Here we can't find a well-defined $F$ such that $F\cdot n = z^2$ on all six faces of the unit cube, and we have to evaluate directly: $$ \mathrm{Top} + \mathrm{Bottom} = \iint_{[0,1]\times[0,1]\times \{1\}} 1 \,dxdy + \iint_{[0,1]\times[0,1]\times \{-1\}} 0 \,dxdy, \\ \mathrm{Right} + \mathrm{Left} = \iint_{[0,1]\times\{1\}\times [0,1]} z^2 \,dzdx + \iint_{[0,1]\times \{-1\}\times[0,1]} z^2 \,dzdx, \\ \mathrm{Front} + \mathrm{Back} = \iint_{\{1\}\times[0,1]\times [0,1]} z^2 \,dydz + \iint_{\{-1\}\times[0,1]\times [0,1]} z^2 \,dydz. $$

Based on my teaching experience, the problem to illustrate (1) is actually taking $S$ to be the unit 2-sphere: $$ S = S^2 = \{(x,y,z): x^2 + y^2 + z^2 = 1\}, $$ so that the unit outward normal is $n=(x,y,z)$: $$ \iint_S z^2 \,dS = \iint_S (0,0,z)\cdot (x,y,z) \,dS = \iiint_{B(0,1)} \nabla\cdot (0,0,z)\,dV = |B(0,1)|. $$


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.