MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

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

I'm pretty sure that this is a minor error, but I could use some help here. So the book I'm referring to in the title is this book (MR1164870).

On pg. 16-17, he is proving that the space of almost complex structures on a compact smooth surface without boundary is a smooth submanifold of the (1,1)-tensors. He does by expressing the space of almost complex structures is the intersection of traceless tensors and the tensors of determinant 1. He then uses transversality to conclude the argument.

I don't understand Tromba's proof of the statement that the space of tensors of determinant 1 is a smooth submanifold of the (1,1)-tensors. He says he uses the Implicit Function Theorem, but what he does doesn't really make sense to me.

I will use the notations that he uses in the book and present his argument, or at least my poor understanding of his argument.

Let $\mathcal{H}^s(T^1_1 M)$ denote the space of (1,1) tensors on a smooth surface $M$ of some genus $g$, which are weakly differentiable to order $s$ and are square integrable. (Let $s>3$.)

Let $\mathcal{A}^s = \{J\in \mathcal{H}^s| J^2=-id, \text{for any } x\in $M$, v\in T_x M, (v,Jv)\text{form an ordered basis} \}$. This is the space of almost complex structures.

Let $\mathcal{N}:=tr^{-1}(0)\subset\mathcal{H}^s(T^1_1 M)$. This is a linear subspace, thus is automatically a smooth submanifold.

Let $\mathcal{M}:=det^{-1}(1) \subset \mathcal{H}^s(T^1_1 M)$. It's easy to check that $\mathcal{A}^s = \mathcal{N} \cap \mathcal{M}$.

Claim: $\mathcal{M}$ is a $C^\infty$ submanifold of $\mathcal{H}^s(T^1_1 M)$ with $T_J\mathcal{M} =\{ H | tr JH = 0\}$

To use the Implicit Function Theorem, we want to show that $Ddet(J):T_J \mathcal{H}^s \to \mathbb{R}$ is surjective for every $J \in \mathcal{M}$.

It's easy to see that $Ddet(J)H = (det J) tr(J^{-1}H) = tr(J^{-1}H)$.

But now, Tromba goes onto say that $J^{-1} = -J$, and concludes that $Ddet(J)H = -tr(JH)$. From this expression, the surjectivity is obvious. But this doesn't make any sense, since $J$ is just an element of $\mathcal{M}$, and there is nothing in the definition of $\mathcal{M}$ that forces $J^2=-id$!!!

My questions are,

0) Is $\mathcal{M}$ a $C^\infty$ submanifold of $\mathcal{H}^s$, regardless of what is written here?

1) Why does he get to choose a $J$ such that $J^2 = -id$? Why is that sufficient to prove that $\mathcal{M}$ is a submanifold?

2) This argument does show that $\mathcal{A}^s$ is a $C^\infty$ submanifold of $\mathcal{N}$. Does that imply that $\mathcal{A}^s$ is a $C^\infty$ submanifold of $\mathcal{H}^s$? (i.e. is being a "$C^\infty$ submanifold" transitive?)

Thank you for your time. And I apologize if this is not an appropriate question here.

EDIT: I think some people misunderstood my question 2). What I was asking was whether transversality is necessary at all.

So we know that $\mathcal{A}^s=\det^{-1}(1)\subset\mathcal{N}$. Using IFT, we can conclude that $\mathcal{A}^s$ is a $C^\infty$ submanifold of $\mathcal{N}$, the same way one could conclude that $\mathcal{M}$ is a $C^\infty$ submanifold of $\mathcal{H}^s$.

Does it follow from this that $\mathcal{A}^s$ is a smooth submanifold of $\mathcal{H}^s$? (That is, without saying anything about transversal intersections?)

share|cite|improve this question
I realized that the answer to question 0) is "yes," or at least I see that the map $H \mapsto tr(J^{-1}H)$ is surjective. – BrainDead Jan 17 '11 at 4:43
Maybe I'm dense but isn't 1) just the Cayley-Hamilton theorem: $J^{2} - \text{tr}(J) \cdot J + \text{det}(J)\cdot\text{id} = 0$, so $J^{2} = - \text{id}$? – Theo Buehler Jan 17 '11 at 4:48
Why is $tr(J) =0$? – BrainDead Jan 17 '11 at 4:53
Sorry I confused $\mathcal{A}^{s}$ and $\mathcal{M}$. But then Cayley-Hamilton tells you that $J^{-1} = (\text{tr}(J)\text{id} - J)$, so $\text{tr}(J^{-1}H) = \text{tr}(J)\text{tr}(H) - \text{tr}(JH)$. – Theo Buehler Jan 17 '11 at 5:07
up vote 3 down vote accepted

A small typo: your definition of $\mathcal{M}$ is incorrect; shouldn't it be $\det^{-1}(1)$?

Anyway, I think you are mixing up two parts of the proof. The first is that $\mathcal{M}$ is a smooth submanifold. Indeed, using your computation, it is easy to see that taking $H= J$ you have $D\det(J)H = \mathop{tr}(J^{-1}J) = 2$ and so surjectivity follows.

The part about setting $J^{-1} = -J$ is only used to prove the transversality of $\mathcal{M}$ and $\mathcal{N}$ at the intersection: that if $J\in\mathcal{M}\cap\mathcal{N}$, then indeed $D\det(J)H = -\mathop{tr}(JH)$. In particular for such $J$ you have $Id\in T_J\mathcal{M}$ but not in $T_J\mathcal{N}$.

share|cite|improve this answer
Thank you for the answer. Yes, that should've been $\det^{-1}(1)$. I made the correction just now. But there is definitely a typo on pg. 17, since he says: "[after using $J^2=-id$ to get $Ddet(J)H = -tr(JH)$]...It remains to show that $H\mapsto -\tr JH$ is surjective. Then the implicit function theorem shows that $\mc{M}$ is a submanifold with tangent sapce $\ker(D\det(J))=\\{H|\tr JH =0 \\}.$ Thanks for the clarification. I thought perhaps I was missing some kind of a known theorem or some feature of the Implicit Function Theorem that allows me to only look at those with $J^2=-id$. – BrainDead Jan 17 '11 at 16:10
What about my question 2)? $\mathcal{A}^s$ is a $C^\infty$ submanifold of $\mathcal{N}$, and thus is a $C^\infty$ submanifold of $\mathcal{H}^s$? – BrainDead Jan 17 '11 at 16:12
@BrainDead: I don't have my copy of Tromba handy, so I cannot check the exact wording he used. If you are sure about your reading, then maybe you should send him an e-mail about the misprint. On question (2) I don't remember off hand whether any additional conditions are needed for the infinite dimensional transverse-intersection theorem. But the finite dimensional analogue of your statement is true: transverse intersection of two smooth submanifolds form a smooth submanifold of codimension the sum of the original codimensions. – Willie Wong Jan 17 '11 at 17:15
In particular: the finite dimensional analogue gives if $M,N$ two smooth submanifolds of $X$ such that along $x\in M\cap N$ you have $T_xM$ and $T_xN$ span $T_xX$, then $M\cap N$ is a smooth submanifold of $X$ with codimension $\mathop{codim}(M) + \mathop{codim}(N)$. – Willie Wong Jan 17 '11 at 17:20
@Willie Wong - Thank you for your answer. But what I was asking was whether the transverse-intersection theorem was necessary at all. I made an edit to my post to clarify what exactly I meant by 2). Please take a look. – BrainDead Jan 17 '11 at 21:55

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.