Let $M$ be a manifold and $g$ a metric on $M$. Let $TM$ denote the tangent bundle of $M$, and denote points in $TM$ by $(x,v)$ where $v \in T_xM$.

The Levi-Civita connection of $(M,g)$ induces a splitting of the double tangent bundle $TTM = V \oplus H$, where $V$ is the vertical distribution, defined by $V_{(x,v)} = T_{(x,v)}T_xM$ (i.e. the tangent space to the fibre), and $H_{(x,v)}$ is the horizontal distribution, which is determined by the connection.

Suppose $A:TM \rightarrow TM$ is a map such that $A(x,v)\in T_xM$ (so the map $A(x,\cdot)$ is a map from $T_xM$ to itself for all $x \in M$).

How does one use the splitting described above to define "partial derivatives" $\nabla_xA$ and $\nabla_vA$, which should be maps:

$(\nabla_xA)(x,v):T_xM \rightarrow T_xM$,

$(\nabla_vA)(x,v):T_xM \rightarrow T_xM$.

These should have the property that if $\gamma(t)$ is a curve on $M$ and $u(t)$ is a vector field along $\gamma$ (so $u(t) \in T_{\gamma(t)}M$ for all $t$), and $\nabla_t$ denotes the covariant derivative along $\gamma$, then

$\nabla_t(A(\gamma,u)) = (\nabla_xA)(\gamma,u) \cdot \dot{\gamma} + (\nabla_vA)(\gamma,u) \cdot \nabla_t u$

(here on the LHS, $A(\gamma,u)$ is itself a vector field along $\gamma$, so the notation $\nabla_t(A(\gamma,u))$ is meaningful).

The expression above "makes sense" intuitively, but I can't get the formalism to work properly.


2 Answers 2


Let us write the isomorphism

$T_{(x,v)}TM = H_{(x,v)}TM \oplus V_{(x,v)})TM \cong T_xM \oplus T_xM$


$\xi \simeq (\xi^h,\xi^v)$,

so that $\xi^h \in T_xM$ and $\xi^v \in T_xM$.

Here the identification $H_{(x,v)}TM \cong T_xM$ is given by the restriction of $d_{(x,v)}\pi$ to $H_{(x,v)}TM$ (where $\pi:TM \rightarrow M$ is the projection), and the isomorphism $V_{(x,v)}TM \cong T_xM$ is canonical.

The key point is that under these identifications, if $z$ is a curve on $TM$, say $z(t)=(\gamma(t),u(t))$ then

$\dot{z}(0) \simeq (\dot{\gamma}(0),(\nabla_tu)(0))$.

So suppose $x \in M$ and $v,w,y \in T_xM$. Let $\gamma$ be a curve in $M$ such that $\gamma(0)=x$ and $\dot{\gamma}(0)=w$, and let $u$ be a vector field along $\gamma$ such that $u(0)=v$ and $(\nabla_tu)(0)=y$. Let $z(t)=(\gamma(t),u(t))$.

Think of $A$ as a map $TM \rightarrow TM$, so that the differential $dA$ is a map

$d_{(x,v)}A:T_{(x,v)}TM \rightarrow T_{(x,v)}TM$.

Then given $w\in T_xM$, if $\xi_w$ is the unique vector whose horizontal component is $w$ and whose vertical component is zero (i.e. $\xi^h = w$ and $\xi^v = 0$), then we define


and similarly if $\zeta_w$ is the unique vector whose horizontal component is zero and whose vertical component is $w$ (i.e. $\zeta^h = 0$ and $\zeta^v = w$), then we define


Then it follows that

$d_{(x,v)}A(\xi) \simeq ((\nabla_xA)(x,v)(w),(\nabla_vA)(x,v)(y))$,

and these two maps have the properties you're looking for.


Denote $E=TM$. We have a vertical projection $P_{(x,v)}:T_{(x,v)}E\to E_x$ determined by the splitting. The covariant derivative of a vector field $(\gamma,u)$, regarded as a curve in $E$, is the vertical projection of its derivative. We also have linear maps $d\pi:T_{(x,v)}E\to T_xM$ (where $\pi:E\to M$ is the bundle projection), the horizontal lift $h:T_xM\to H_{(x,v)}$ such that $h\circ d\pi=id_H$, and the natural isomorphism $i:T_xM=E_x\to V_{(x,v)}$ (which is the only thing depending on the fact that $E$ is the tangent bundle and not an arbitrary vector bundle). In this notation, we have $\xi=h\circ d\pi(\xi)+i\circ P_{(x,v)}(\xi)$ for every $\xi\in T_{(x,v)}E$.

Define $\nabla_x A=P_{A(x,v)}\circ (dA)\circ h$ and $\nabla_v A=P_{A(x,v)}\circ (dA)\circ i$ where $dA:T_{(x,v)}E\to T_{A(x,v)}E$ is the ordinary derivative. Then, for a vector $\xi\in T_{(x,v)}E$ (the derivative of our vector field $(\gamma,u)$) we have $$ \begin{aligned} \nabla_t (A(\gamma,u)) &= P_{A(x,u)}\circ dA(\xi) = P_{A(x,u)}\circ dA \circ h\circ d\pi(\xi) +P_{A(x,u)}\circ dA \circ i\circ P_{(x,v)}(\xi) \cr &=(\nabla_xA)(d\pi(\xi)) + (\nabla_vA)(P_{(x,v)}(\xi)) =(\nabla_xA)\cdot\dot\gamma + (\nabla_vA)\cdot\nabla_t(\gamma,u) . \end{aligned} $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.