Define $V(x) = e^{x}$, $x \in \mathbb{R}$ and consider the Hamiltonian $H = - \frac{d^2}{dx^2} + V(x)$. The eigenvalue problem is $$ -\psi''(x) + e^{x} \psi(x) = E \psi(x)\,, \quad x \in \mathbb{R}\,. $$ A change of variables reduces this to a modified Bessel equation, so we can express the energy eigenfunctions in terms of well-known functions, see WolframAlpha.

Thus $$ \psi(x) = c_1 I_{- \lambda} \left( 2 e^{ x/2} \right) + c_2 I_\lambda \left( 2 e^{x/2} \right) $$ with $\lambda = 2 \sqrt{E}i$ and where $I$ is the modified Bessel function of the first kind.

Using the asymptotic form for the modified Bessel function around the origin we see that in the limit $x \to - \infty$ $$ \psi(x) \sim C_1 e^{-i \sqrt{E}x} + C_2 e^{i \sqrt{E}x}\,, \; x \to - \infty\,. $$ i.e. a solution to the free Schrödinger equation.

Now if $\psi_E$ is the eigenfunction of $H$ with energy $E$, I want to derive the asymptotic $$ \psi_E(x) \sim e^{i \sqrt{E}x} + R(E) e^{- i \sqrt{E}x}\,, \; x \to - \infty\,, $$ where $R(E)$ is supposed to be the reflection coefficient, telling us how an incoming plane wave reflects from our potential. But to derive $R(E)$ I feel like I need some additional boundary condition? Maybe I need to define the behavior for $x \to \infty$? (Edit: indeed it is enough to assume that the solution vanishes at $+ \infty$).

  • 3
    $\begingroup$ See page 13 of Liouville Field Theory. $\endgroup$
    – AHusain
    Commented Jul 29, 2017 at 10:34
  • $\begingroup$ I see you answered your own question, 0 at +infinity. Did it work? $\endgroup$
    – user196418
    Commented Dec 19, 2018 at 12:34

1 Answer 1


Indeed, there is a missing boundary condition: since the potential grows for $x\rightarrow +\infty$, the solution should vanish in this limit. Thus, even before solving the equation, we know that the wave is fully reflected, and the reflection coefficient has amplitude $1$.

Substitution $z=2e^{x/2}$ reduces the equation to the equation for the modified Bessel functions: $$z^2u''(z)+zu'(z)-[\lambda^2+z^2]u(z)=0$$ with $\lambda = 2i\sqrt{E}$. As the two linearly independent solutions of thsi equation one can take either $I_{\pm\lambda}(z)$ or $I_\lambda(z),K_\lambda(z)$. $I_{\pm\lambda}(z)$ diverges for $x\rightarrow+\infty$, but $K_\lambda$ (which is just a linear combination of $I_{\pm\lambda}(z)$) does not. We thus write the solution as $$ \psi(x)=CK_\lambda(e^{x/2})=\tilde{C} \left[I_{-\lambda}(e^{x/2})- I_{\lambda}(e^{x/2})\right]$$ The asymptotic form of this solution for $x\rightarrow-\infty$ is: $$ \psi(x)\rightarrow \tilde{C}\left[\frac{1}{\Gamma(1-2i\sqrt{E})}e^{-i\sqrt{E}x}-\frac{1}{\Gamma(1+2i\sqrt{E})}e^{i\sqrt{E}x}\right],$$ from which the reflection ciefficient is $$ R=\frac{\Gamma(1+2i\sqrt{E})}{\Gamma(1-2i\sqrt{E})}. $$

Remark: The last sentence in the OP indicates that they have already found the solution, whereas the math was done in the deleted answer by @2.1. However, I think it is worthwhile having it stated clearly.


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.