# Do Lie algebroids pull back (along submersions)?

There are more general definitions, but for my purposes a Lie algebroid on a smooth manifold $X$ is a vector bundle $A \to X$, a map $\rho: A \to {\rm T}X$ of vector bundles over $X$, and a bracket $[,]$ making $\Gamma(A)$ into an $\mathbb R$-Lie algebra, such that $\rho$ induces is a map $\Gamma(A) \to \Gamma({\rm T}X)$ of Lie algebras and such that the Leibniz rule $[a,fb] = f[a,b] + (\rho(a)f)b$ is satisfied for $f\in \mathcal C^\infty$ and $a,b\in \Gamma(A)$.

So, suppose I happen to have a Lie algebroid $A\to X$ lying around, and also a smooth map $\phi: Y\to X$. (In my case, $Y \to X$ happens to be a vector bundle, so you can assume some fairly strong properties of the map.) Then I can certainly pull back the vector bundle $A\to X$ to $\phi^*A \to Y$. Is there a natural Lie algebroid structure I can put on this pullback? The answer is probably "how natural do you want it?": if $Y = \{{\rm pt}\}$ and $\phi({\rm pt}) = y\in X$, then the anchor map for $\phi^*A \to \{{\rm pt}\}$ must be trivial, but the only part of the fiber $A_y$ with a Lie algebra structure is $\ker \{A_y \overset\rho\to {\rm T}_y X\}$.

So the real question is:

Along what type of maps do Lie algebroids pull back? In particular, can I always pull back a Lie algebroid along a submersion?

-

As your example suggests, the vector bundle pullback is perhaps not the right thing to consider when wanting defining the notion of a pullback Lie algebroid. You have to go one step further and take the pullback of $\phi^*\rho: \phi^* A \to \phi^*TX$ along $d\phi: TY \to \phi^*TX$. The resulting bundle $\phi^{**}A$ (which exists when $d\phi - \phi^*\rho: TY \oplus \phi^*A \to \phi^*TX$ has constant rank, so in particular when $\phi$ is a submersion) then has a Lie algebroid structure with gives the right universal property you would want, if that's what you want. In the case where $Y$ is a point, this gives exactly your example.
This is all detailed in the paper "Algebraic Constructions in the Category of Lie Algebroids" by Higgins and Mackenzie. This may even have something to say about just trying to do something with $\phi^*A$ itself, but I haven't checked.
Awesome. In fact, awesomer than you know. At least, I'm trying to build a Lie algebroid that in the trivialized case comes in pieces, one of which is a Lie algebroid on $\phi^*A \to Y$. (Which is to say: I know what I want when $A\to X$ and $Y\to X$ are both trivialized bundles, but it's not obviously invariant, so I can't immediately glue.) But a dimension count suggests that this double pullback $\phi^{**}A$ has exactly the correct dimension to be the Lie algebroid I'm trying to construct. So now I just have to check if it really is. – Theo Johnson-Freyd May 31 '10 at 1:06
In particular, since in Theo's case $Y \to X$ was a vector bundle, then $phi^{**} A \to Y$ is the total space'' of a VB-algebroid $(\phi^{**}A, Y, A, X)$. See section 6.1 on arXiv:0810.0066. Not sure whether this is relevant, but since you mention "in trivialized case comes in pieces", it may. – Alfonso Gracia-Saz Jun 8 '10 at 2:09