Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

How would I go about showing that the vector field $X = x\frac{\partial}{\partial y} - y\frac{\partial}{\partial x}$ is tangent to the unit sphere in $\mathbb{R}^3$? I can see it pictorially but I'm not sure what would be required to actually prove it.

(This isn't homework! I understand the definitions of vector field as either a derivation of $C^\infty(M)$ or as a section of the tangent bundle $TM$, but I don't know how to compute anything about it so I'm trying to work through some examples.)

share|cite|improve this question
To be absolutely precise, we need a definition of "field tangent to a submanifold". Without being absolutely precise, I'd be happy with showing that $Xf=0$ on $M$ where $f$ is a defining equation of $M=\{f=0\}$ – user53153 Jan 25 '13 at 19:07
Your comment pretty much solves the problem. For your function $f$ is the standard equation for the sphere, $f=x^2+y^2+z^2-R^2$. Just directly apply the derivatives and show that it is zero. – F. Solis Jan 25 '13 at 19:13
@5PM Ok, so $Xf = 2xy-2yx =0$ everywhere on the sphere. This might be obvious, but why does $Xf$ being zero suffice? – Anna Jan 25 '13 at 19:16
Tangent vectors are the infinitesimal versions of flows. A flow maps a space into itself with $t$ as a parameter, say, $x'=F(x,t)$. We can look at the change in a function as we move: $df/dt=df(x'(t))/dt=df(F(x,t))/dt=Xf$, with the last equality being pretty much the definition of the tangent vector. To say that $X$ is tangent means that the flow it generates stays within the submanifold. If your function is constant in the submanifold the flow induces a change of zero. Further, in this case our function is not zero outside it, so that if the vector was not tangent, we would get $Xf\neq=0$. – F. Solis Jan 25 '13 at 19:55
up vote 4 down vote accepted

Here's how to make the argument from the comments a little more rigorous from basic properties of tangent bundles. Let $i: S^2 \to \mathbb R^3$ be the inclusion. For every $x \in S^2$ this induces an injection $i_* T_xS^2 \to T_x \mathbb R^3$. Then asking if a vector field $X$ on $\mathbb R^3$ is tangent to $S^2$ is the same as asking if for all $x \in S^2, X_x \in i_* T_x S^2$. Now consider the map $f: \mathbb R^3 \to \mathbb R, (x,y,z) \mapsto x^2 + y^2 + z^2$. Now $f \circ i : S^2 \to \mathbb R$ is constant so that $f_* \circ i_* = 0$. In other words, $i_* TS^2 \subset \ker f_*$. Since $f_* \ne 0$ on $S^2$, by dimensionality reasons we must have that $i_* TS^2 = \ker f_*$. Finally, a vector field is in $\ker f_*$ is equivalent to $Xf = 0$ since $Xf = df(X) = f_* X$.

share|cite|improve this answer

The unit sphere in $\Bbb R^3$ is nice in the sense that its tangent spaces admit a simple interpretation: If $r \in S^2$, then $$T_r S^2 = \{v \in \Bbb R^3 : r \cdot v = 0 \}.$$ Now we can make the identification $$x \frac{\partial}{\partial y} - y \frac{\partial}{\partial x} \leftrightarrow (-y, x, 0) \in \Bbb R^3.$$ Now for any $r = (x,y,z) \in S^2$, we have $$(x, y,z) \cdot (-y, x,0) = -xy + xy + 0= 0 \implies (-y, x, 0) \in T_r S^2.$$ This holds for all points $r \in S^2$, so $x \frac{\partial}{\partial y} - y \frac{\partial}{\partial x}$ is a vector field on $S^2$.

share|cite|improve this answer

Since we're explicitly considering an embedding, you can use the concept of a vector manifold to attack this problem. There are many benefits to doing so: in particular, you can use geometric (clifford) algebra to analyze the problem.

What Henry has done is clever: he's using the normal to the surface to characterize it, which is something you can always do in $\mathbb R^3$, but not something you can do generally. For instance, if your manifold $M$ has $\dim M = 2$ in $\mathbb R^4$, you can't construct a vector normal to it.

Geometric algebra allows you to characterize an orientable manifold by its pseudoscalar. This is what I will do here.

The unit sphere can be characterized by two angles $\theta, \phi$. We assign each point a vector in $\mathbb R^3$, which is obvious by the embedding. Let these vectors be $r$ (so that $r^2$, the second component of this vector, is $y$ as usual). We can then take, at any $r$, partial derivatives with respect to our angular coordinates to characterize tangent vectors.

$$\begin{align*}e_\theta &= \frac{\partial x}{\partial \theta} = \cos \theta \cos \phi e_1+ \cos \theta \sin \phi e_2 - \sin \theta e_3 \\ e_\phi &= \frac{\partial x}{\partial \phi} = \sin \theta (-\sin \phi e_1 + \cos \phi e_2)\end{align*}$$

For this problem, it helps to rewrite the variables back in terms of $x, y, z$.

$$\begin{align*}e_\theta &= \frac{zx}{\sqrt{x^2 + y^2}} e_x+ \frac{zy}{\sqrt{x^2+y^2}}e_y - \sqrt{x^2+y^2} e_z \\ e_\phi &= - y e_x + x e_y\end{align*}$$

We can now characterize the surface by its pseudoscalar $i$, which is formed by a wedge product of the tangent vectors.

$$i = e_\theta \wedge e_\phi = \sqrt{x^2 + y^2} (z e_{xy} + y e_{zx} + x e_{xy})$$

The pseudoscalar represents, for a surface, an area element. It is a function of position, as you can see. Because it has orientation, the pseudoscalar itself can tell you about the orientation of a manifold (indeed, if you were to get that the pseuscalar is multivalued, you would rightly conclude that the manifold is non-orientable). The pseudoscalar is used to perform projections under the geometric product. If a vector field lies entirely in the tangent space, then its wedge product with the pseudoscalar should be zero. We can see that this is indeed the case.

$$i \wedge X \propto (z e_{xy} + y e_{zx} + x e_{yz}) \wedge (x e_y - y e_x) = yx e_{zxy} - xy e_{yzx} = (yx -xy) e_{xyz} = 0$$

This approach is a bit more complicated for $\mathbb R^3$ than the standard approach, and the clifford algebra aspects of it may take some getting used to, but it's a powerful formalism when you have access to an embedding, and we could've just as easily used it for a 2-dimensional surface in $\mathbb R^4$ with almost no change in approach.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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