The Jacobi field is defined as $$J^{''}(t)+R(\gamma^{'}(t),J(t))\gamma^{'}(t)=0$$ since it is a system of $2n$ order, spanned by $\{J,J^{'}\}$, so it is dimension of $2n$.
But I don't get if we add a condition of $J(0)=0$, then the dimension is $n$. ( I can't see why the system is immeddiately reduced to $n$ linear independent equations which is expained by our teaching assistant ).
Similarly, if we set the Jacobi field normal, that is $\langle J,\gamma ^{'}\rangle=0$, then the dimension of normal Jacobi field is $2(n-1)$.
It seems to be a trivial consequence, but I really don' see it. Can anyone else help explain it ?
A Jacobi field $J(t)$ determined by initial values of $J$ and $J'$ at a point. This proved by simply rewrite the Jacobi equation in terms of a parallel orthonormal frame along $\gamma$ and observe that the resulting equation is a system of linear second-order ODEs. For definiteness, here is a reference for this claim :
So if we denote $\mathfrak{X}(\gamma)$ as the space of smooth vector field along $\gamma$ and $\mathscr{J}(\gamma) \subseteq \mathfrak{X}(\gamma)$ be the set of Jacobi fields along $\gamma$. Then above proposition says that the map $ P : \mathscr{J}(\gamma) \to T_pM \oplus T_pM $ defined as
$$ J \mapsto (J(a), J'(a)). $$ is an isomorphism. Therefore it follows that $\text{dim }\mathscr{J}(\gamma) = 2n$.
Therefore the subspace of $\mathscr{J}(\gamma)$ that satisfy $J(a)=0$ (we can take $a=0$ from the beginning if we want) is just $P^{-1}(\{0\} \oplus T_pM)$ which is clearly have dimension $n$ since $P$ is an isomorphism.
Similarly for the subspace of normal Jacobi field, denoted by $\mathscr{J}(\gamma)^{\perp} \subseteq \mathscr{J}(\gamma)$ :
Sorry for the late answer by the way.