Let $M^{2n}$ be a smooth manifold admitting two almost complex structures $J_0$ and $J_1$.

Suppose that $J_0$ and $J_1$ are both regular in the sense that the moduli space $$ \mathcal{M}_i:=\mathcal{M}(A; J_i)/PSL(2,\mathbb{C}) $$ of $J_i$-holomorphic spheres representing the homology class $A\in H_2(M;\mathbb{Z})$ is a smooth, compact manifold for both $i=0,1$. Suppose further that there exists a path $(J_t)_{t\in [0,1]}$ of regular almost complex structures on $M$ connecting $J_0$ and $J_1$.

In McDuff and Salamon's book they show that the manifolds $$ \mathcal{M}_t:=\mathcal{M}(A;J_t)/PSL(2,\mathbb{C}) $$ are oriented cobordant for all $t\in [0,1]$.

QUESTION: Are the moduli spaces $\mathcal{M}_t$ diffeomorphic for all $t\in [0,1]$? If not, how does the diffeomorphism fail?


The issue here is precisely what you mean when you say "regular path $J_t$ of almost complex structures".

I am not particularly expert in the theory of holomorphic curves, so I won't comment on that; I expect the situation is precisely analagous to the finite-dimensional setting I will describe below (I'm sure there are more technical issues involving the corner points).

Let $M$ be a smooth manifold, and $S \subset N$ a closed submanifold. Given a map $f: M \to N$ transverse to $S$, we have that $\mathcal M_f = f^{-1}(S)$ is a smooth manifold (with corners etc if $S$ and $N$ have them).

The condition "$f \pitchfork S$" is the precise analogy to "$J_t$ is regular" in this finite-dimensional setting.

Suppose you have a path of functions $f$, written as a map $f_t: I \times M \to N$. Then I would say that this is a regular path of functions if the map $I \times M \to N$ is transverse to $S$, and the same for $\{0,1\} \times M \to N$. Then the "parameterized moduli space" $\mathcal M_I$ is a smooth manifold with boundary equal to $\mathcal M_0 \sqcup \mathcal M_1$. This space is equipped with a smooth map $\mathcal M_I \to I$ which is a submersion near the boundary.

The point is that each individual $f_t$ is not assumed to be regular; spelling it out, it is entirely possible that for $(t,x)$, we have $f_t(x) \in S$, and $\text{Im}(df_t) \subset N_{f_t(x)} S$ is not the entirety of the normal space, but rather of codimension 1; and so that together with the additional vector $\frac{d}{dt} f_t(x) \in N_{f_t(x)} S$, we span the whole of the normal space.

The second point is that an individual $\mathcal M_t$ is regular if and only if $t \in I$ is a regular value of the projection $\mathcal M_I \to I$.

So if, in addition, $f_t$ is regular for each $t$, then by definition the projection $\mathcal M_I \to I$ is a submersion; as you know well, a proper smooth submersion is a fiber bundle, and this provides a global diffeomorphism $\mathcal M_I \cong I \times \mathcal M_0$.

In almost every interesting application, given any two regular functions / almost complex structures / whatever, it will be essentially impossible to find a path through regular (blahs) between them; but it will be possible to find a "regular path" of (blahs) between them in the sense given above. A simple case to think about the discussion above with is $S$ a hypersurface and $M$ a point! If $S$ disconnects the codomain, then to get from one side to another you will need to cross $S$. In this situation, $\mathcal M_I$ will be a finite set of points which project to the interior of $I$.


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.