8

I will here restate a question I asked earlier. It did not have much succes (probably by an incomplete introduction of the problem on my part).

I am reading a paper by Ricca ( http://www.maths.ed.ac.uk/~aar/papers/ricca.pdf p.1338 on the bottom) on the linking number. I would like to see that its definition as the degree of the map $\Psi$ is equal to that by intersection number.

A link $L$ has two component knots $\gamma_1$ and $\gamma_2$, these denote embeddings of $S^1$ in $\mathbb{R}^3$. Consider $$\Psi: \mathbb{T} \,\,(= C_1\times C_2)\to S^2: (s,t)\mapsto \textbf{n}(s,t) = \frac{\gamma_1(s) - \gamma_2(t)}{\bigl|\gamma_1(s) - \gamma_2(t)\bigr|} $$ Then $\text{deg} \Psi$ is one definition of the linking number. Taking a Seifert Surface $M$ of $\gamma_1$ (i.e. an orientable surface with boundary $\gamma_1$) then its intersection number with $\gamma_2$ will also give the linking number.

Now I would like to prove that these are equivalent. Therefore consider the orientable manifold
$$N = (M\times\gamma_2) \,\, \bigm\backslash \,\bigcup_{m\in M\cap \gamma_2} \mathcal{B}(m,\epsilon) $$ with boundary $$\partial N = (\gamma_1\times\gamma_2) \cup \bigl(\bigcup_{m\in M\cap\gamma_2}\mathcal{S}(m,\epsilon)\bigr) $$ Also consider $\overline{\Psi}:N\to S^2: (x,y)\mapsto \frac{y-x}{|y-x|}$. Its restriction to $\gamma_1\times\gamma_2$ is now $\Psi$.

The degree of $\overline{\Psi}$ restricted to $\partial N$ is now zero. So we get $$\text{deg}\biggl(\overline{\Psi}\biggm| {\gamma_1\times \gamma_2}\biggr) + \text{deg}\biggl(\overline{\Psi}\biggm|{\bigcup_{m\in M\cap \gamma_2} \mathcal{S}(m,\epsilon)} \biggr) =0 \text{.} $$ The first term gives the linking number defined by degree. How do I see that the second term gives the intersection number?

I know this has to do with the orientations, but I don't have enough feeling for this yet, to write this down rigorously. I really hope that someone could explain how this works or at least give me hint. Thanks in advance.

Lee Mosher
  • 135,265
  • 2
    I've done some edits to make this more readable. But having done so I can now see that your formulas for $N$ and $\partial N$ are flawed. For instance, the set you are removing from $M \times \gamma_2$ is not a subset of $M \times \gamma_2$, instead it is a subset of $M$. – Lee Mosher Jun 08 '15 at 14:14
  • 1
    @LeeMosher: Agreed. Confusingly, this is also the definition given in the paper referenced. Here's what I suspect is going on: View $N$ as $M\times S^1$. Given $m\in M \cap \gamma_2$, let $t_m$ be the point satisfying $\gamma_2(t_m)=m$. The points of the form $(m,t_m) \in N$ are the points where $\overline{\Psi}$ is undefined, so we remove a 3-ball around each such point. A well-chosen orientation convention for each $S^2$ in the boundary will make it so that the degree of $\overline{\Psi}$ is the negative of the sign of the corresponding intersection between $\gamma_2$ and the surface $M$. – Kyle Jun 08 '15 at 14:58
  • 1
    I should add that it is relatively straightforward, with the right picture, to see that the degree of the map on each $S^2$ is $\pm1$. If I can see why the, say, outward orientation makes the whole computation work, I'll post a full answer. – Kyle Jun 08 '15 at 15:28
  • @LeeMosher: Indeed this might be a bit confusing, however squirrel added what also I think is meant by the author. I suppose we can see $M\times \gamma_2$ as some kind of torus that is filled up. As such we can then take away these three-dimensional balls around the problematic points. Btw: thanks for the edits – Tom Ultramelonman Jun 08 '15 at 16:44
  • @squirrel: thank you for your efforts. I am looking forward to that full answer. – Tom Ultramelonman Jun 08 '15 at 16:45

1 Answers1

3

Here's the idea in a more general setting. If you have two oriented submanifolds $X$ and $Y$ intersecting transversally at a point in $\mathbb{R}^n$ (or in an arbitrary oriented manifold but viewed locally as living in $\mathbb{R}^n$ so that you can use vector operations), then the map $\psi: (X \times Y )\setminus (X \cap Y) \to S^{n-1} \subset \mathbb{R}^n$ given by $$(x,y) \mapsto \frac{y-x}{\|y- x\|}$$ encodes the sign of intersections between $X$ and $Y$. It suffices to consider a local model with oriented copies of $\mathbb{R}^k$ and $\mathbb{R}^\ell$ inside $\mathbb{R}^{k+\ell}$. The above map is defined on $(\mathbb{R}^k \times \mathbb{R}^\ell)\setminus \{(0,0)\}$. For convenience, let $\bar S^{n-1}$ and $S^{n-1}$ denote the unit spheres in $\mathbb{R}^k \times \mathbb{R}^\ell$ and $\mathbb{R}^{k+\ell}$, respectively, each oriented as the boundary of the unit ball. Note that the obvious map $\mathbb{R}^k \times \mathbb{R}^\ell \to \mathbb{R}^{k+\ell}$ is orientation-preserving if and only if the intersection $\mathbb{R}^k\cap \mathbb{R}^\ell$ is positive. It follows that the restriction $\bar S^{n-1} \to S^{n-1}$ of this map has degree positive (resp. negative) one when $\mathbb{R}^k$ and $\mathbb{R}^\ell$ intersect positively (resp. negatively). The restriction of $\psi$ to $\bar S^{n-1}$ is simply $(x,y)\mapsto y-x$, and we note that it factors as a composition $\bar S^{n-1} \to \bar S^{n-1} \to S^{n-1}$, where the first map is $(x,y)\mapsto (-x,y)$ and the second map is $(x,y)\mapsto x+y$. The first map has degree $(-1)^k$ and the second map has degree positive (resp. negative) one when the intersection is positive (resp. negative).

Now in our setting we have nearly the same setup, but we'll have to make one minor change that ultimately flips the degree. For each intersection point $(m_\alpha,t_\alpha) \in M \cap \gamma_2$ we remove a neighboring 3-ball from $M \times S^1$, obtaining a new manifold $N$. Note that each spherical boundary component $S^2_\alpha$ is oriented as the boundary of $N$, not the boundary of the (removed) 3-ball, so the orientation of $S^2_\alpha$ is the reverse of the sphere $\bar S^{n-1}$ in the above discussion. Thus, at a positive intersection point, the degree of $\overline{\Psi}$ is $-(1)(-1)^{k=2}=-1$. Simiarly, its degree is $-(-1)(-1)^{k=2}=1$ at negative intersection points.

Kyle
  • 6,263
  • 1
  • 23
  • 44
  • There was a pair of errors in my original answer. Each reversed the degree of (the same) map, so the answer appeared correct at the end. This approach may appear less geometric, so you should check out the previous (flawed) version to get the visual intuition. – Kyle Jun 09 '15 at 17:23
  • Thank you, this helps a lot. I think I am beginning to get some intuition with these orientations. Am I right to say that the first version would be correct if: 1) We notice that the orientation of the unit ball and and the chosen orientation do not coincide and 2) we take another orientation of these vectors (in the drawing). – Tom Ultramelonman Jun 09 '15 at 22:34
  • @TomBecker: No problem. And you're exactly right, those are the two errors I needed to remedy. – Kyle Jun 10 '15 at 00:25