= Liouville transformation of a second-order equation
{c}
{title2=$U=\dot x^{-1/2}u(x(t))$}
A <Liouville transformation of a second-order equation> combines a change of independent variable with an amplitude factor that removes the induced first derivative. For $u_{xx}-f(x)u=0$, let $x=x(t)$ be smooth with $p=dx/dt>0$, put $U(t)=u(x(t))$ and $w(t)=U(t)/\sqrt{p(t)}$. The <chain rule> gives
$$
U_{tt}-\frac{p_t}{p}U_t-p^2f(x(t))U=0,
$$
and substitution of $U=\sqrt p\,w$ gives
$$
w_{tt}=\left[p^2f(x(t))-\frac12\frac{p_{tt}}p+\frac34\left(\frac{p_t}p\right)^2\right]w.
$$
This specifies which quantity receives the inverse-square-root rescaling.
Where $f>0$, choose $p=f(x(t))^{-1/2}$, so the principal term $p^2f$ is one. The coordinate is $t=\int\sqrt{f(x)}\,dx$ and the original amplitude factor is $u=f^{-1/4}w$. Neglecting the remaining coefficient-variation terms produces the exponential <WKB approximation> $u\simeq f^{-1/4}e^{\pm\int\sqrt f\,dx}$ only when those terms are suitably small in the asymptotic regime. Turning points $f=0$ require a different local scaling; negative $f$ gives the corresponding oscillatory formulation.
Back to article page