Morse distance functions characterization in terms of normal exponential map

84 Views Asked by At

I'm reading the book "Classical and Modern Morse Theory with Applications" by Mercuri, Piccione and Tausk.

We want to characterize the points $q\in \mathbb{R}^{n+p}$ such that the distance function $L_q(x)=\langle q-f(x), q-f(x) \rangle$ is Morse. Here, $M$ is an $n$-dimensional Riemannian manifold, $L_q : M \to \mathbb{R}$ and $f:M \to \mathbb{R}^{n+p}$ is an isometric immersion.

It was also obtained before that if $\xi (x)= q-f(x)$, then $\overline{\nabla}_X \xi =-X$, so $$dL_q (x) X = -2\langle X, \xi \rangle$$ and $$d^2 L_q (x) (X,Y)=-2Y \langle X, \xi \rangle =2 \langle(Id-A_{\xi})X,Y \rangle,$$ where $A_{\xi} X = - \overline{\nabla} _X \xi$ denotes the shape operator at $x$ on the $\xi$ direction.

Having this in mind, the book continues saying that this will be done in terms of the $\textit{ endpoint map }$ or normal exponential map: $$E:\nu M \to \mathbb{R}^{n+p}, \; \; E(x, \eta)=f(x) + \eta,$$ where $\nu M$ denotes the normal space corresponding to $f$.

And it goes:

We compute the differential of $E$. Let $(x, \eta) \in \nu M $ and $\gamma (t) = (x(t),\eta(t))$ be a curve in $\nu M$ such that $x(0)=x, \; \eta (0)= \eta$. Then: $$dE(x, \eta)(\dot{\gamma}(0))= (x(t)+\eta(t))'(0)=\dot{x}(0)+(\dot{\eta}(0))^T + (\dot{\eta}(0))^{\perp},$$ where, for $z\in T_x \mathbb{R}^{n+p}$, $z^T$ y $z^{\perp}$ denote the projections of $z$ in $T_x M$ and in $\nu _x M$, respectively. In particular, taking $x(t)=x, \; \eta(t)=x+t\eta,$ we get that the differential of $E$ along the fibres is the identity (which was geometrically obvious). In particular, $dE (x, \eta)$ and $(Id-A_{\eta})$ haver kernels of the same dimension.

The part that I still don't understand is: "In particular, taking $x(t)=x, \; \eta(t)=x+t\eta,$ we get that the differential of $E$ along the fibres is the identity (which was geometrically obvious)".

I am confused because I think that if we take the dot product $\langle x(t), \eta (t) \rangle = \langle x , x +t\eta \rangle = \langle x , x \rangle \neq 0$ and this should equal $0$ so that the curve is in $\nu M$, so I don't understand why is it taking that curve.

And even if we take that curve, in what sense is the differential of $E$ "along the fibres" the identity and why is that geometrically obvious?. If you could explain some intuition here it would be great, but I am also confused with how to obtain that the differential of $E$ is the identity, since it is a transformation between the tangent space of the normal bundle at $x$ and $\mathbb{R}^{n+p}$, am I wrong?

The last part "$dE (x, \eta)$ and $(Id-A_{\eta})$ have kernels of the same dimension" sounds like the most important to make the characterization, but I have no clue about how to show that.

I would be really thankful if you could help me with any of these parts!