Chapter 11

Time-Dependent Rydberg–Fano Dynamics in CO 2

11.1Introduction

This chapter is where the preceding material comes together.

Chapter 10 gave the Fano profile: the asymmetric line produced when a resonance and a direct route lead to the same continuum. That result is stationary. It describes the spectrum obtained after waiting long enough for everything to settle, and says nothing about how the profile came to be.

However, the profile is not present at the instant of excitation. It builds up over the autoionization lifetime, while the Rydberg population that feeds it decays. With attosecond pulses that construction can be watched, and watching it requires abandoning the stationary picture and keeping the amplitudes and their phases — Chapter 6's coupled equations, with no perturbative expansion.

The chapter sets up the channel basis, derives the coupled equations by projection, states the approximations, and says what the calculation predicts. The equations of Sec. 11.4 are those of the research this course prepares for.(E. Saydanzad, V.-H. Hoang and U. Thumm, Time-resolved Rydberg dynamics in molecules, validated against the photoemission experiment of Biswas et al., Phys. Rev. A 110, 043106 (2024).) They look formidable; they are not. Every symbol has been defined in an earlier chapter, and the whole system is Theorem 6.2.1 with the terms written out.

We work in atomic units throughout, so ℏ = 1 .

Definition 11.1.1Optional research extension

This chapter is optional for entering Quantum Chemistry or Condensed Matter. Its first pass is the one-level decay and resonantly driven two-level models below. The CO 2 channel basis, continuum elimination, and pump–probe observables are the research extension reached by repeating the same amplitude bookkeeping with more channels.

11.1.1First solvable limits: decay, then drive

A single unstable level with energy E a and energy width Γ obeys

i c ˙ a = ( E a − i Γ / 2 ) c a , c a ( t ) = c a ( 0 ) e − i E a t e − Γ t / 2 . (11.1)

Thus P a ( t ) = | c a ( 0 ) | 2 e − Γ t . Restoring units changes the exponent to − Γ t / ℏ . If this effective non-Hermitian equation represents loss into an initially empty continuum, its accumulated probability is

P c o n t ( t ) = | c a ( 0 ) | 2 − P a ( t ) , P ˙ c o n t = Γ P a (11.2)

in atomic units. The missing norm has a physical destination; it has not ceased to exist.

Before adding decay or continua, drive a stable pair on resonance with the constant RWA Hamiltonian

H 2 = Ω 2 ( 0 1 1 0 ) . (11.3)

For ( c g ( 0 ) , c a ( 0 ) ) = ( 1 , 0 ) , c g = cos ⁡ ( Ω t / 2 ) and c a = − i sin ⁡ ( Ω t / 2 ) , so the two probabilities sum to one. The research equations below add three ingredients to this matrix: detunings, the loss term − i Γ n / 2 , and continuum amplitudes that receive the lost probability.

11.2The Physical Setup

The whole experiment in one diagram, and the whole of the channel basis that the coupled equations are written in. Everything the chapter derives is bookkeeping for these four arrows.
Figure 11.1. The whole experiment in one diagram, and the whole of the channel basis that the coupled equations are written in. Everything the chapter derives is bookkeeping for these four arrows.

An attosecond XUV pulse, centered near 17.6 eV with bandwidth about 0.4 eV, illuminates CO 2 in its ground state X 1 Σ g + . A near-infrared pulse follows after a controllable delay t d . Photoelectrons are collected and their energies recorded as a function of that delay.

Three things can happen, and the model is an accounting of them.

Pathway 1 — direct ionization. The XUV photon ionizes immediately, leaving CO 2 + in X 2 Π g (threshold 13.78 eV, Example 7.5.2) plus a photoelectron. Since 17.6 > 13.78 , this channel is always open.

Pathway 2 — resonant excitation and autoionization. The same photon instead excites a Rydberg state of the ( 3 σ u ) − 1 n l σ g series, converging to B 2 Σ u + at 18.08 eV (Chapter 9). These states sit near 17.5 eV, above the X threshold, so they autoionize into the same continuum that pathway 1 reaches.

Pathways 1 and 2 end in the same final state. They interfere. That is Chapter 10's Fano problem, now happening in real time.

Pathway 3 — the probe. The delayed infrared pulse ionizes whatever Rydberg population survives, into the B 2 Σ u + continuum. No XUV route reaches this channel, so it is a clean measurement of the Rydberg population — and, because several Rydberg states were excited coherently by the broadband pump, of the phases between them.

That is why the experiment uses two colors. The coupled channel ( 𝒞 , here X ) shows interference between direct and resonant ionization; the uncoupled channel ( 𝒩 , here B ) shows quantum beats between coherently excited Rydberg states.

11.3The Channel Basis

Expand the molecular state in the available states, exactly as Theorem 10.2.1 prescribes — a sum over discrete states and an integral over each continuum — with the free-evolution phases extracted as in Chapter 6:

Following the notation of the manuscript Supplementary, 𝒞 labels continua coupled to the Rydberg manifold by configuration interaction, and 𝒩 labels continua that are not. For CO 2 in this chapter, 𝒞 ≡ X 2 Π g and 𝒩 ≡ B 2 Σ u + .

| Ψ ( t ) ⟩ = c g ( t ) e − i E g t | g ⟩ + ∑ n c ℛ n ( t ) e − i E ℛ n t | ℛ n ⟩ + ∫ d ε α c 𝒞 ( t , ε α ) e − i ( E 𝒞 + ε α ) t | 𝒞 , ε α ⟩ + ∫ d ε β c 𝒩 ( t , ε β ) e − i ( E 𝒩 + ε β ) t | 𝒩 , ε β ⟩ (11.4)
AmplitudeStateIntroduced in
c g neutral ground state X 1 Σ g + Ch. 7
c ℛ n Rydberg state ( 3 σ u ) − 1 n l σ g Ch. 9
c 𝒞 ( t , ε α ) coupled continuum (here X 2 Π g ) + electron energy ε α Ch. 9, 10
c 𝒩 ( t , ε β ) uncoupled continuum (here B 2 Σ u + ) + electron energy ε β Ch. 9, 10

The two continua differ in one respect: 𝒞 is coupled to the Rydberg series by configuration interaction — the states autoionize into it — and 𝒩 is not. That is not an assumption but a consequence of energy: the Rydberg states lie below the B threshold, so they cannot decay into it, only be driven into it by absorbing an infrared photon.

Because the expansion is over a complete orthonormal set, probability is conserved exactly:

| c g | 2 + ∑ n | c ℛ n | 2 + ∫ d ε α | c 𝒞 | 2 + ∫ d ε β | c 𝒩 | 2 = 1 . (11.5)

This is the check to apply to any numerical solution (Remark 6.2.3).

11.3.1The Hamiltonian

H ˆ ( t ) = K ˆ N + V ˆ N N + H ˆ e + D ˆ ( t ) , (11.6)

with K ˆ N and V ˆ N N from Eq. (7.2). The essential step is to split the electronic Hamiltonian according to whether it acts within a channel or between channels. Introducing projectors P ˆ a onto the individual channels,

H ˆ e = H ˆ ch + V ˆ CI , H ˆ ch = ∑ a P ˆ a H ˆ e P ˆ a , V ˆ CI = ∑ a ≠ b P ˆ a H ˆ e P ˆ b . (11.7)

H ˆ ch gives each channel its energies. V ˆ CI — the configuration interaction — collects the off-diagonal blocks and couples the Rydberg configurations to the open 𝒞 continuum. It is the autoionization coupling of Chapter 10, Eq. (10.4), under its research name.

The laser term is Chapter 6's dipole interaction with two pulses:

D ˆ ( t ) = D ˆ XUV h XUV ( t ) e i ( ω XUV t + ϕ XUV ) + D ˆ IR h IR ( t − t d ) e i ( ω IR t + ϕ IR ) + h . c . , (11.8)

with envelopes h , carrier frequencies ω , carrier phases ϕ , and the pump–probe delay t d appearing in exactly one place — the argument of the infrared envelope. This is the quantity the experiment varies.

11.4Deriving the Coupled Equations

The derivation is Theorem 6.2.1 applied to Eq. (11.4), and it is worth following once because the result looks more complicated than the procedure.

Step 1: project. Substituting Eq. (11.4) into i ∂ t | Ψ ⟩ = H ˆ | Ψ ⟩ and projecting onto each basis state gives, exactly as in the proof of Theorem 6.2.1, one equation per amplitude. Each term carries three factors: the amplitude it comes from, the matrix element connecting the two states, and a phase e i ( E final − E initial ) t .

Step 2: identify the matrix elements. Only three kinds are nonzero, and each was established earlier:

Step 3: apply the rotating-wave approximation to the XUV pulse (Def. 6.4.1), discarding the counter-rotating terms.

Step 4: eliminate the coupled continuum. The 𝒞 continuum appears in the Rydberg equation through the autoionization coupling. Solving the c 𝒞 equation formally and substituting back — a Laplace transform, with the Markov approximation that the coupling is flat across the narrow resonance — converts that coupling into two terms: a decay − i Γ n n c ℛ n / 2 and a principal-value shift, which is absorbed into the measured resonance energy exactly as F ( E ) was in Sec. 10.4. This is the Markov–Fano reduction, and it is where Γ enters the time domain.

The result is four equations.

The ground state loses population to the Rydberg states and to the 𝒞 continuum:

i c ˙ g = ∑ n c ℛ n h XUV ( t ) e i ϕ XUV e − i ( E ℛ n − E g − ω XUV ) t ⟨ g | D ˆ XUV | ℛ n ⟩ + ∫ d ε α c 𝒞 ( t , ε α ) h XUV ( t ) e i ϕ XUV e − i ( E 𝒞 + ε α − E g − ω XUV ) t ⟨ g | D ˆ XUV | 𝒞 , ε α ⟩ . (11.9)

Each Rydberg amplitude is pumped from the ground state, decays by autoionization, and is depleted by the probe:

i c ˙ ℛ n + i Γ n n 2 c ℛ n = ρ n c g h XUV ( t ) e − i ϕ XUV e i ( E ℛ n − E g − ω XUV ) t + ∫ d ε β c 𝒩 ( t , ε β ) h IR ( t − t d ) e i ϕ IR e − i ( E 𝒩 + ε β − E ℛ n − ω IR ) t ⟨ ℛ n | D ˆ IR | 𝒩 , ε β ⟩ . (11.10)

The coupled continuum receives amplitude from both routes:

i c ˙ 𝒞 ( t , ε α ) = c g h XUV ( t ) e − i ϕ XUV e i ( E 𝒞 + ε α − E g − ω XUV ) t ⟨ g | D ˆ XUV | 𝒞 , ε α ⟩ + ∑ n c ℛ n e i ( E 𝒞 + ε α − E ℛ n ) t Γ n n 2 π . (11.11)

The uncoupled continuum is fed only by the probe:

i c ˙ 𝒩 ( t , ε β ) = ∑ n c ℛ n h IR ( t − t d ) e − i ϕ IR e − i ( E ℛ n − E 𝒩 − ε β + ω IR ) t ⟨ ℛ n | D ˆ IR | 𝒩 , ε β ⟩ . (11.12)

11.4.1Reading the Terms

Four observations account for everything above.

Remark 11.4.1Every coupling is envelope × phase × dipole

That is the structure of Eq. (6.3): h ( t ) says when the pulse is present, the exponential carries the energy mismatch, and the matrix element says how strongly the states are connected. By Remark 6.2.2 a term matters only when its exponent is small — when the photon energy nearly matches the transition.

Remark 11.4.2The decay term is not from the laser

The term i Γ n n c ℛ n / 2 on the left of Eq. (11.10) carries no envelope and no dipole. It is autoionization, present whether or not any light is on, and it makes the effective Hamiltonian non-Hermitian in the Rydberg block. The factor 1 2 is because amplitudes decay at half the rate of populations: | c | 2 ∼ e − Γ t requires c ∼ e − Γ t / 2 .

Remark 11.4.3The two ends of the same process

The second term of Eq. (11.11) is Γ n n / 2 π — Eq. (10.9) exactly. It and the decay term of Remark 11.4.2 are the same physics viewed from opposite ends: population leaving the Rydberg state arrives in the continuum. Together they conserve Eq. (11.5).

Remark 11.4.4Interference lives in Eq. (11.11)

Its two terms add amplitudes into the same final state | 𝒞 , ε α ⟩ . Squaring the sum produces a cross term, and that cross term is the Fano asymmetry — now time-dependent, because the two contributions arrive on different schedules. The direct term follows the pump envelope and stops when the pulse does; the resonant term keeps arriving for as long as the Rydberg states take to decay.

11.4.2The Effective Pump Amplitude

One symbol in Eq. (11.10) remains: ρ n , the effective strength with which the XUV pumps the Rydberg state. It is not simply the bound dipole, because the Markov elimination of the continuum folds the direct route into the resonant one. The result is

ρ n = ( q n − i ) π Γ n n 2 ⟨ g | D ˆ XUV | 𝒞 , ε α ⟩ | ε α = E ℛ n − E 𝒞 (11.13)

evaluated at the on-shell photoelectron energy — the energy the electron must carry for energy to be conserved when the resonance decays.

Chapter 10's asymmetry parameter appears here in the combination ( q n − i ) , and that complex factor is the whole time-dependent Fano effect in one bracket. The real part is resonant excitation; the imaginary part is interference with the direct route; and q n measures their ratio, exactly as in Eq. (10.18). Setting q n → ∞ leaves the resonance alone; setting q n = 0 leaves only the interference term, which is the window resonance of Chapter 10.

11.5The Approximations

A model is only as good as its stated assumptions. Six are made here, and each is stated explicitly so that it can be judged.

  1. Two channels. One coupled continuum ( 𝒞 , here X 2 Π g ) and one uncoupled ( 𝒩 , here B 2 Σ u + ); the A and C ionic states of Example 7.5.2 are dropped. Justified because the experiment resolves photoelectron energy and can separate the channels.

  2. One vibrational level and one partial wave per channel. Nuclear motion is frozen — the timescale argument of Sec. 7.10: the pump is a few femtoseconds, nuclear motion takes tens. This is the weakest assumption for the longest-lived resonances, which survive for picoseconds.

  3. Configuration interaction restricted to the resonant coupling. Only ⟨ ℛ n | V ˆ CI | 𝒞 , ε α ⟩ is kept, set to Γ n n / 2 π ; continuum–continuum interactions are dropped.

  4. Independent resonances. Cross-widths Γ n m = 0 for m ≠ n , so each state decays at its own rate without coherent mixing through the shared continuum. Exercise E3 tests this.

  5. No infrared coupling within the Rydberg manifold. ⟨ ℛ m | D ˆ IR | ℛ n ⟩ = 0 , the parity argument of Example 9.7.1: both series are built on the same 3 σ u hole with g Rydberg orbitals, so both have total parity u , and one photon cannot connect two states of equal parity. Not an approximation so much as a symmetry.

  6. Predissociation neglected. The sharp states can also fall apart into neutral fragments, competing with autoionization. Omitting it makes the calculated autoionization yields upper bounds.

  7. No AC Stark shift. The states are taken to sit at their field-free energies for the whole calculation. They do not: while either pulse is on, the atom is dressed (Sec. 6.5) and every level moves. Sec. 11.5.1 works out what that costs.

Assumption (v) illustrates what this course was for. It removes an entire block of the Hamiltonian and costs nothing — no fitting, no parameter, no error — because it follows from the inversion symmetry of a linear centrosymmetric molecule. A symmetry argument proved in Chapter 3, applied to labels earned in Chapter 7, simplifies a research calculation in Chapter 11.

11.5.1The Shift the Model Omits

Assumption (vii) is the one most easily forgotten, because nothing in the equations complains about it. It is worth seeing what its removal would do.

While a pulse is on, the atom is not in an eigenstate of its own Hamiltonian; it is dressed by the field, and the levels move. For the intensities used in these experiments the field is far too weak to split anything visibly — there is no Autler–Townes doublet here — but it is easily strong enough to shift the resonance by a small amount Δ ( t ) , following the pulse envelope and vanishing with it.

The shift enters exactly one place: the energy denominator that fixes where the resonance sits relative to the continuum,

1 E b − E X − ε ⟶ 1 E b − E X − ε − Δ ( t ) . (11.14)

Comparing the two, the field-free formula is recovered by the replacement E b → E b − Δ ( t ) . The asymmetry parameter is a function of that energy, so it inherits the shift; expanding to first order in Δ ,

q ( t ) = q 0 − Δ ( t ) d q 0 d E b + 𝒪 ( Δ 2 ) (11.15)

valid whenever Δ is small compared with the energy scale over which q 0 varies — which, for a Rydberg series converging on a threshold, is itself small, so this is not a trivial condition.

Equation (11.15) says the line shape changes shape while the pulse is on. Not its height and not its width: its asymmetry. Look back at Fig. 10.2, which draws the profile for a range of q ; a Stark shift slides the resonance along that family and back again as the pulse passes.

This is worth separating carefully from the time dependence the chapter already has. Figure 11.2 shows the profile forming, because the two interfering amplitudes arrive on different schedules and the interference is not complete until the resonance has finished decaying. That is a property of the resonance and happens with no field present at all. The Stark contribution is the opposite kind of thing: it is present only while the field is, it follows the envelope instantaneously, and it leaves nothing behind. In a delay scan the two are distinguishable — one tracks the pulse, the other tracks Γ − 1 .

Remark 11.5.1Where a dark state would hide

Assumption (iv) set the cross-widths Γ n m to zero, so that each Rydberg state decays into the continuum on its own. That assumption is doing more work than it looks.

If two bound states | b 1 ⟩ and | b 2 ⟩ decay into the same continuum, then the superposition α | b 1 ⟩ + β | b 2 ⟩ has ionization width

Γ dark = | α V X , b 1 + β V X , b 2 | 2 ,

and there is always a choice of α , β making it vanish. That superposition cannot ionize at all, however long one waits and however allowed each state is separately: it is a dark state (Sec. 6.5.2), and it would show up as a dip in the photoionization cross-section rather than as a peak.

Setting Γ n m = 0 removes the coherence that would build such a superposition, so the model cannot produce this feature. Whether the real molecule does is a question the model is not equipped to answer, and Exercise E3 is where it starts to be asked.

11.6What the Model Predicts

The prediction that makes the model worth writing down. The Fano profile of Chapter 10 is a steady-state object; here it is watched forming. The asymmetry needs the full lifetime to develop, so a probe arriving sooner than finds a line that is still symmetric — and the delay at which it stops being symmetric measures directly.
Figure 11.2. The prediction that makes the model worth writing down. The Fano profile of Chapter 10 is a steady-state object; here it is watched forming. The asymmetry needs the full lifetime to develop, so a probe arriving sooner than τ finds a line that is still symmetric — and the delay at which it stops being symmetric measures τ directly.

Solving Eqs. (11.9)–(11.12) for each delay t d gives | c 𝒞 ( ∞ , ε α ) | 2 and | c 𝒩 ( ∞ , ε β ) | 2 — the photoelectron spectrum in each channel as a function of delay. The two observables answer different questions.

The coupled channel (C, here X): a Fano profile assembling itself

Both pathways reach this continuum, so its spectrum carries the interference. Early in the pulse only the direct route has contributed and the spectrum is smooth. As the Rydberg states decay — over hundreds of femtoseconds for the sharp series, tens for the diffuse — the resonant amplitude accumulates and the asymmetric Fano profile of Chapter 10 grows in where a featureless background used to be.

The Fano line shape is not a property a state has; it is something that happens, on the timescale τ = ℏ / Γ .

Chapter 10's stationary formula, Eq. (10.16), is the t → ∞ limit of this.

The uncoupled channel (N, here B): quantum beats

No XUV route reaches this continuum, so its population is proportional to the Rydberg population surviving when the probe arrives. The broadband pump excited several Rydberg states coherently (Example 6.6.2), and their amplitudes carry phases advancing at different rates, e − i E ℛ n t . Any pair therefore beats at

ω n m = E ℛ n − E ℛ m , (11.16)

and the 𝒩 -channel yield (here the B channel) oscillates with t d at that frequency, superposed on the overall exponential decay.

Fourier transforming the delay dependence therefore recovers the level spacings directly. Because the photoelectron energy is resolved as well, pairs whose beat frequencies nearly coincide but whose electron energies differ can be told apart — which is what makes the technique more than a lifetime measurement.

Example 11.6.1Which beats are observable

Two members of the sharp series lie at 17.5395 eV ( 5 d , τ = 372.6 fs) and 17.7018 eV ( 6 d , τ = 1457.5 fs). Find the beat period and decide whether it can be seen.

Solution.

The spacing is

Δ E = 17.7018 − 17.5395 = 0.1623   eV ,

and with 2 π ℏ = 4.136 eV fs the beat period is

T = 2 π ℏ Δ E = 4.136 0.1623 = 25.5   fs .

Now the question that decides whether the experiment works. The beating survives only as long as both states do, so the shorter lifetime governs:

372.6 25.5 ≈ 15   beat periods .

Fifteen oscillations before the envelope decays is ample — the Fourier transform will show a clean, narrow peak.

Contrast the diffuse series, where the 5 s state lives 15.9 fs. Its beats against a neighbor would be damped within a single period, so no oscillation is observable at all: that channel reports a decay and nothing more.

The same experiment therefore measures a frequency for one series and only a lifetime for the other, and the difference traces back through Γ τ = ℏ to states differing only in l — that is, to the centrifugal barrier of Chapter 3.

First-pass self-checks

The factor of two that decides whether the bookkeeping closes. Time is measured in lifetimes =/. The amplitude falls as e^- t/2 and the survival probability as its square, e^- t/; the continuum population is the difference from one, and probability is conserved at every instant because the two lower curves sum to one by construction. Using the amplitude's exponent for a population — the error Self-check 1 is built to catch — overestimates the survival at one lifetime by 0.61 against 0.37, and the accounting no longer closes.
Figure 11.3. The factor of two that decides whether the bookkeeping closes. Time is measured in lifetimes τ = ℏ / Γ . The amplitude falls as e − Γ t / 2 ℏ and the survival probability as its square, e − Γ t / ℏ ; the continuum population is the difference from one, and probability is conserved at every instant because the two lower curves sum to one by construction. Using the amplitude's exponent for a population — the error Self-check 1 is built to catch — overestimates the survival at one lifetime by 0.61 against 0.37 , and the accounting no longer closes.
Example 11.6.2Self-check 1: one level decays and one channel fills

A teaching level has Γ = 20 meV and begins with unit population. Find the level and continuum probabilities after one lifetime.

Solution.

The lifetime is τ = ℏ / Γ = 32.9 fs. At t = τ , P a = e − 1 = 0.367879 and P c o n t = 1 − e − 1 = 0.632121 . Their sum is one. Applying the amplitude decay e − Γ t / ( 2 ℏ ) directly to probability would miss the square and fail this accounting check.

Example 11.6.3Self-check 2: solve the driven pair

In atomic units let Ω = 0.2 in Eq. (11.3). When is the initially empty level first fully populated?

Solution.

P a = sin 2 ⁡ ( Ω t / 2 ) first reaches one when Ω t / 2 = π / 2 , hence t = π / Ω = 15.708 atomic units of time. At that instant P g = 0 and P a = 1 . This is the lossless benchmark to recover when all widths and continuum couplings are set to zero.

Example 11.6.4Self-check 3: account for every channel

At one time a teaching calculation gives P g = 0.05 , ∑ n P ℛ n = 0.40 , and integrated continuum probabilities P 𝒞 = 0.35 , P 𝒩 = 0.20 . Does it pass the norm check?

Solution.

Equation (11.5) gives 0.05 + 0.40 + 0.35 + 0.20 = 1.00 , so it passes. Omitting either continuum would make an apparent loss. Each continuum contribution is an energy integral of a probability density, not the density at one photoelectron energy.

11.7Summary

11.8Exercises

  1. Account for every term. Take Eq. (11.10). For each of its three terms state (a) which physical process it represents, (b) which pulse if any drives it, and (c) which earlier chapter defined the quantity in it. Then explain why the second term contains c 𝒩 rather than c 𝒞 .

  2. Beats and lifetimes. Repeat Example 11.6.1 for 7 d ( 17.8000 eV, τ = 3378.5 fs) and 8 d ( 17.8650 eV, τ = 1485.2 fs). Give the beat period and the number of observable periods. Then find the pair from Chapter 10's table whose beat would be hardest to observe, and say why.

  3. Testing assumption (iv). Independent resonances requires that neighboring resonances not overlap: widths small compared with spacing.

    1. Using Γ n ∝ ( n ∗ ) − 3 (Thm. 10.6.1) and the Rydberg level spacing Δ E ∝ ( n ∗ ) − 3 (Eq. (4.62)), show that the ratio Γ / Δ E is independent of n ∗ .

    2. Evaluate that ratio for the 5 d state using its width and its spacing to 6 d .

    3. Comment: does the assumption get better or worse as n increases, and why is that answer surprising?

  4. The pump amplitude. From Eq. (11.13), compute | ρ n | in terms of q n and Γ n n , and find the phase of ρ n . What happens to both as q n → ∞ , and what physical situation does that limit describe?

  5. Norm conservation. Show analytically that with the probe off and the pump over,

    d d t [ ∑ n | c ℛ n | 2 + ∫ d ε α | c 𝒞 | 2 ] = 0 .

    using Eqs. (11.10) and (11.11). Identify which two terms cancel, and relate the result to Remark 11.4.3.

11.9Project: Time-Resolved Fano Buildup

The problem. Solve a reduced version of Eqs. (11.9)–(11.12) numerically and watch both signatures appear: the Fano profile assembling itself in the coupled channel ( 𝒞 , here X ), and quantum beats in the uncoupled channel ( 𝒩 , here B ). This is the capstone of the course, and a small version of the research code.

Reduce as follows: keep the ground state, two Rydberg states (so there is something to beat), and both continua on energy grids. Take the pump weak enough that c g ≈ 1 throughout, which removes Eq. (11.9).

Use real parameters: E 𝒩 = E B = 18.08 eV, and for the two Rydberg states the 5 d and 6 d values — energies 17.5395 and 17.7018 eV, lifetimes 372.6 and 1457.5 fs, infrared dipoles 1.000 and 0.765 , q 5 d = 2.00 and q 6 d = 3.69 . Take a Gaussian XUV pulse of 0.4 eV bandwidth centered at 17.6 eV and a longer infrared pulse. Work in atomic units.

  1. On paper. Write the reduced equations explicitly for two Rydberg states with c g = 1 . Confirm by inspection that with the probe off each c ℛ n decays as e − Γ n n t / 2 once the pump has passed, so | c ℛ n | 2 decays at Γ n n .

  2. On paper. Convert every parameter to atomic units and state the time step you will need. It must resolve the fastest phase in the problem — identify which exponent in Eqs. (11.10)–(11.12) that is.

  3. On the computer. With the probe off, integrate the c ℛ n and c 𝒞 equations and plot the 𝒞 -channel spectrum | c 𝒞 ( t , ε α ) | 2 at several times: during the pulse, at τ / 4 , τ and 4 τ of the shorter-lived state. The asymmetric profile should grow out of a smooth background. Compare the final spectrum with Eq. (10.16) using the same q and Γ .

  4. On the computer. Verify that ∑ n | c ℛ n | 2 + ∫ d ε α | c 𝒞 | 2 is conserved once the pump is over — the analytic result of Exercise E5. Report the drift; this tests whether your continuum grid is dense and wide enough, and you should state what you needed.

  5. On the computer. Switch the probe on and scan t d over several hundred femtoseconds. Plot the total 𝒩 -channel yield ∫ d ε β | c 𝒩 ( ∞ , ε β ) | 2 against t d : exponential decay with oscillations superposed.

  6. On the computer. Fourier transform that delay trace and identify the beat frequency. Compare with E 6 d − E 5 d from Eq. (11.16) and with the 25.5 fs period of Example 11.6.1.

  7. What has to move. Two animations. First, the 𝒞 -channel spectrum (here X ) against time with the stationary Fano profile drawn behind as a dashed target — the moving curve should converge onto it. Second, the two Rydberg populations and the 𝒩 -channel yield (here B ) as the delay is scanned, so the beating is visible as it is measured.

The check. Part (c) is the one to get right and part (d) is how you know you have. If the built-up profile does not converge to Eq. (10.16), the usual causes are a continuum grid too narrow to hold the resonance wings, or a grid spacing coarser than Γ . Both show up first as drift in (d).

Be ready to answer. Your part (c) spectra converge to the stationary Fano profile, which Chapter 10 derived without any reference to time. Explain what the time-dependent calculation tells you that the stationary one cannot — and identify one experimental observable that would be identical in both descriptions, and one that would not.