8

Consider a two-level atom, with ground state $|g\rangle$, excited state $|e\rangle$, having an energy gap $\hbar\omega_0=E_e-E_g$. Assume that the energy levels are sharp (i.e., no level widths). Let a sinusoidal electromagnetic radiation of frequency $\omega$, with its electric field $\vec{E}=\vec{E}_0\cos(\omega t)$, be incident on the atom, initially in state $|g\rangle$. Now, it is a well-known result from first-order time-dependent perturbation theory (and even from the exact solution?), that the probability of transition from $|g\rangle\to|e\rangle$ is nonzero even when $\omega\neq\omega_0$.

My doubts are the following.

Isn't it true that an atom can absorb or emit photons if and only if the photon energy exactly equals the gap (i.e., $\omega=\omega_0$). It seems to violate energy conservation. I am not able to make sense of this.

cconsta1
  • 1,257

4 Answers4

6

If the incident radiation is switched on at time $t=0$ then at any given time $t$ the atom has experienced radiation whose field amplitude, as a function of time, was, up till then, $$ H(t) \cos(\omega_i t) $$ where $H$ is a step function: $H = 0$ for $t<0$ and $H=1$ for $t \ge 0$ and I am using $\omega_i$ for the frequency of the incident radiation. So at time $t$, the atom has experienced not a wave for infinite time, but rather a pulse of duration $t$. To get some insight into the atom's response we can Fourier analyse this pulse. To construct the Fourier integral correctly, we have to include that, at the given time $t$, the atom has not experienced the radiation for times later than $t$. So the Fourier integral is $$ \tilde{f}(\omega) = \int_{-\infty}^\infty f(t') e^{-i \omega t'} dt' = \int_0^t e^{i \omega_i t'} e^{-i \omega t'} dt' $$ where for convenience I used a complex exponential in place of the cosine function. The integration is straightforward; one gets a phase factor multiplying the function $$ \frac{\sin((\omega_i - \omega)t)}{\omega_i - \omega} = t\; {\rm sinc}\left( (\omega_i - \omega)t\right) $$ This function has a width, in frequency, of order $$ \Delta \omega = 1/t . $$ What this is telling us is that, as far as the atom "knows", the radiation is not very monochromatic until sufficient time has passed, and it is not exactly monochromatic until infinite time has passed. Therefore, in a photon picture, the electromagnetic field incident on the atom is composed of photons in a range of frequencies. If $\omega_i$ is not too far from $\omega_0$, i.e. $$| \omega_i - \omega_0| < 1/t $$ then the range of modes with non-negligible excitation in the electromagnetic field includes the mode at $\omega_0$. So the atom can absorb from that mode.

Andrew Steane
  • 65,285
3

Isn't is true that an atom can absorb or emit photons if and only if the photon energy exactly equals the gap

This is true within the simplistic naive (often oversimplified and misleading) picture "the atom either absorbs photon or doesn't".

When you study mathematical theories of light-matter interaction based on Maxwell's equations and Schroedinger's equation, or Heisenberg's equations of motion for matrices, there is no such discrete event, not even in quantum theory, unless we are forced to put it in by hand due to "measurement" taking place. Instead, autonomously, light and matter evolve continuously in time.

Light of any frequency can excite the atom; it's just that intensity of this excitation (or probability of detection of big excitation) is very small, unless the frequency is within the small interval around the resonance frequencies (which are differences of the Hamiltonian eigenvalues divided by the Planck constant).

It seems to violate energy conservation.

Only if you assume EM field loses energy $\hbar \omega$ and the atom gains different energy $\hbar \omega_0$. But the process involves charged particles, and they can lose energy, or gain energy, via EM field. If the photon has more energy than necessary, the remaining energy can get scattered out, or absorbed by the atom in different way (kinetic energy); if the photon has too low energy, then more photons can provide enough energy, or some other source of energy can participate (energy from the environment). Anyway, we don't control the quantum state of both the EM field and the atom in such a way that we could demonstrate violation of energy conservation. There are quantum states that do not even have defined energy (superpositions of Hamiltonian eigenstates).

3

According to the energy-time uncertainty principle:

$$ \Delta E \, \Delta t \approx \hbar $$

Since $\Delta E = \hbar \, \Delta \omega$, it follows that:

$$ \Delta t \approx \frac{1}{\Delta \omega} $$

This means that if the incident photons have frequency $\omega \ne \omega_0$ (the transition frequency), the atom can still undergo a transition with small probability, provided the interaction time $\Delta t$ is short enough. In this context, $\Delta \omega = \omega - \omega_0$ effectively represents the frequency width of the transition.

As a back-of-the-envelope estimate, consider a gamma-ray photon with wavelength mismatch $\Delta \lambda \sim 10^{-13} \, \text{m}$. Then the frequency mismatch is:

$$ \Delta f = \frac{c}{\Delta\lambda} = \frac{3 \times 10^8}{10^{-13}} = 10^{21} \, \text{Hz} $$

So the angular frequency mismatch is:

$$ \Delta \omega \sim 10^{21} \, \text{rad/s} $$

and then:

$$ \Delta t \sim 10^{-21} \text{ to } 10^{-23} \, \text{s} $$

These are typical timescales for strong interactions. So yes, off-resonant transitions are possible, but highly suppressed and require extremely short interaction times. In practice, photon absorption is efficient only when $\omega \approx \omega_0$, i.e., on resonance.

Edit: My previous answer used the uncertainty principle and assumed that the levels have width, but the OP specifically mentioned sharp levels (i.e., no width), so here is the mathematical situation:

I'll follow Griffiths' Intro to Quantum Mechanics, 1st ed., Chapter 9, p. 298 onward.

For a two-level system with states $|g\rangle$ (ground) and $|e\rangle$ (excited), the wavefunction under a time-dependent perturbation $H'(t)$ evolves as:

$$ \Psi(t) = c_g(t) \psi_g e^{-i E_g t/\hbar} + c_e(t) \psi_e e^{-i E_e t/\hbar} $$

Plugging this into the Schrödinger equation with $H = H_0 + H'(t)$ gives:

$$ \dot c_g(t) = \frac{-i}{\hbar} \left[ H'_{gg} c_g(t) + H'_{ge} c_e(t) e^{-i \omega_0 t} \right] $$

$$ \dot c_e(t) = \frac{-i}{\hbar} \left[ H'_{eg} c_g(t) e^{i \omega_0 t} + H'_{ee} c_e(t) \right] $$

where $\omega_0 = (E_e - E_g)/\hbar$, and $H'_{eg}, H'_{ee}$, etc., are the matrix elements, e.g., $H'_{eg} = \langle e|H'|g \rangle$.

Now assume the interaction is turned on at $t = 0$ and lasts for time $t$, and that the system starts in state $|g\rangle$, so $c_g(0) = 1$, $c_e(0) = 0$.

Griffiths is vague on the next step, but for an interaction Hamiltonian of the form

$$ H'(t) \sim e \vec{r} \cos(\omega t) $$

and if the states $\psi_g$ and $\psi_e$ are parity eigenstates (e.g., hydrogen 1s and 2p), then $\langle g | \vec{r} | g \rangle = \langle e | \vec{r} | e \rangle = 0$ by parity. So $H'_{gg}(t) = H'_{ee}(t) = 0$, and only the off-diagonal terms contribute.

To first order in time-dependent perturbation theory (see p. 302):

$$ c_e(t) \approx \frac{-i}{\hbar} \int_0^t H'_{eg}(t') e^{i \omega_0 t'} dt' $$

Assume $H'_{eg}(t) = V_{eg} \cos(\omega t)$, where $V_{eg} = \langle e | V(\vec{r}) | g \rangle$. Then:

$$ c_e(t) = \frac{-i V_{eg}}{\hbar} \int_0^t \cos(\omega t') e^{i \omega_0 t'} dt' $$

Using $\cos(\omega t') = \frac{1}{2} \left( e^{i \omega t'} + e^{-i \omega t'} \right)$:

$$ c_e(t) = \frac{-i V_{eg}}{2\hbar} \int_0^t \left( e^{i(\omega + \omega_0)t'} + e^{i(\omega_0 - \omega)t'} \right) dt' $$

Integration yields:

$$ c_e(t) = \frac{-i V_{eg}}{2\hbar} \left[ \frac{e^{i(\omega + \omega_0)t} - 1}{i(\omega + \omega_0)} + \frac{e^{i(\omega_0 - \omega)t} - 1}{i(\omega_0 - \omega)} \right] $$

Simplifying:

$$ c_e(t) = \frac{V_{eg}}{2\hbar} \left[ \frac{1 - e^{i(\omega_0 - \omega)t}}{\omega_0 - \omega} + \frac{1 - e^{i(\omega + \omega_0)t}}{\omega + \omega_0} \right] $$

If $\omega_0 + \omega \gg |\omega_0 - \omega|$, i.e., the driving frequency is near resonance, we can neglect the second term:

$$ c_e(t) \approx -i \frac{V_{eg}}{\hbar} \cdot \frac{\sin\left(\frac{(\omega_0 - \omega)t}{2}\right)}{\omega_0 - \omega} e^{i(\omega_0 - \omega)t/2} $$

Then the transition probability becomes:

$$ P_{g \to e}(t) = |c_e(t)|^2 \approx \left( \frac{V_{eg}}{\hbar} \right)^2 \cdot \frac{\sin^2\left( \frac{(\omega_0 - \omega)t}{2} \right)}{(\omega_0 - \omega)^2} $$

This is the familiar sinc-squared shape, sharply peaked at $\omega = \omega_0$.

So even with sharp levels, off-resonant transitions are allowed mathematically. I made a quick plot of $P(\omega)$ vs $\omega$, and you can see the sharp peak in the figure below:

enter image description here

cconsta1
  • 1,257
0

Atoms can tunnel between states without absorption at all, and yes, this does violate classical energy conservation, but it is a known fact that, in general, energy conservation does not hold in quantum mechanics. It only holds for the energy operator expected value.

Secondly, there's a multi-photon (non-linear) absorption. Even when the energy level gap is $\hbar\omega_0$, the atom can still absorb $n$ photons per single event of energy ${\hbar\omega_0}/{n}$ to go to the same energy level. Surely process probability exponentially decreases as $n \to \infty$, and the photon flux must be high enough (laser and so on). But this mechanism does not violate energy conservation.

cconsta1
  • 1,257