= Solution
Fix the standard <Jacobian matrix> convention $(Da)_{ij}=\partial_{v_j}a_i$. With column <gradients>, the <commuting velocity derivative for kinetic transport> is
$$
\boxed{G=\nabla_v f+t(Da)^T\nabla_x f.}
$$
If the symbol $J_a$ is instead defined with derivative indices as rows, the printed expression is this same formula. Under the standard convention it needs a transpose.
To prove the claim, put $D=\partial_t+a(v)\cdot\nabla_x$. Differentiating $Df=0$ in $v_j$ gives
$$
D(\partial_{v_j}f)=-\sum_i(\partial_{v_j}a_i)\partial_{x_i}f,
\qquad
D\left(t\sum_i(\partial_{v_j}a_i)\partial_{x_i}f\right)
=\sum_i(\partial_{v_j}a_i)\partial_{x_i}f.
$$
The terms cancel, so each component of $G$ satisfies the <transport equation>. Multiplication by the smooth coefficient $Da$ and differentiation preserve the same uniform <compact support>, hence $G\in\mathcal C_T$ componentwise. At $t=0$, $G(0)=\nabla_v f_{\mathrm{in}}$. The <L2 norm> calculation in part (d), summed over components, gives
$$
\boxed{\int|T(Da)^T\nabla_xf(T)+\nabla_vf(T)|^2\,dx\,dv
=\int|\nabla_vf_{\mathrm{in}}|^2\,dx\,dv.}
$$
One can also differentiate the explicit solution: the $t(Da)^T\nabla_x f$ term cancels the derivative of $x-ta(v)$, leaving the initial velocity derivative transported along the <characteristic curves>.
The transpose matters. For the shear $a(v)=(v_1+v_2,v_2)$, $\det Da=1$ but $Da\ne(Da)^T$. The printed field satisfies $Dg=(Da-(Da)^T)\nabla_xf$, which is nonzero for a generic compactly supported datum. \b[The untransposed formula is false with the standard convention.]
Back to article page
