Rabi Oscillations

Point a clean, coherent laser at a two-level atom and turn up the amplitude, and the atom does not simply absorb a photon and stay lit. It flickers. The probability of finding it in the excited state climbs to one, falls back to zero, and climbs again — a rhythmic sloshing of population between ground and excited states that repeats for as long as the coherence survives. This is a Rabi oscillation, and it is the most literal picture we have of what "light driving an atom" actually means. It takes the transition dipole moment and the electric-dipole coupling $-\boldsymbol\mu\cdot\mathbf E$ from the two articles before it and asks the sharpest possible question: not how often does the atom jump, but how does its quantum amplitude evolve, instant by instant, while the field is on. The answer overturns the rate-equation intuition inherited from Einstein's $A$ and $B$ coefficients. Rate equations describe a one-way drift toward equilibrium; Rabi oscillations are reversible — during one cycle the atom borrows a photon's energy from the field and hands it straight back by stimulated emission. Only when something breaks the phase memory (spontaneous emission, a collision, laser jitter) does the exchange become one-way and the familiar absorption line reappear. Rabi oscillation is the parent phenomenon; incoherent absorption is its damped, forgetful child. Governing everything is one frequency, the Rabi frequency $\Omega=\mu E_0/\hbar$ , set by the dipole matrix element and the field amplitude — proportional to $\sqrt{I}$ , not intensity, the fingerprint of a coherent amplitude effect. The full story with damping is the optical Bloch equations, and the resonant-absorption line shape, the natural linewidth, and every qubit gate downstream are its consequences.

Key takeaways

A two-level atom driven by a near-resonant coherent field does not absorb once and stay excited — it cycles population reversibly between ground and excited states at the Rabi frequency.
On resonance the excited probability is $P_e(t)=\sin^2(\Omega t/2)$ , swinging fully between 0 and 1, with $\Omega=\mu E_0/\hbar$ .
Off resonance the oscillation is faster but incomplete: it runs at $\Omega_R=\sqrt{\Omega^2+\delta^2}$ and peaks at only $\Omega^2/\Omega_R^2<1$ .
A $\pi$ pulse fully inverts the atom; a $\pi/2$ pulse builds an equal superposition — the primitives of qubit control and Ramsey spectroscopy.
Because $\Omega\propto\sqrt I$ the Rabi rate is amplitude-linear, distinct from an intensity-linear absorption rate; decoherence damps the oscillation into the rate-equation steady state.
Key idea: a near-resonant coherent field turns a two-level atom into a driven two-state oscillator whose population is exchanged reversibly at $\Omega_R=\sqrt{\Omega^2+\delta^2}$ , with the coupling $\Omega=\mu E_0/\hbar$ fixed by the transition dipole moment and the field amplitude.

The core insight

A near-resonant coherent field turns a two-level atom into a driven two-state oscillator: population is exchanged reversibly at the generalised Rabi frequency $\Omega_R=\sqrt{\Omega^2+\delta^2}$ , with the coupling strength $\Omega=\mu E_0/\hbar$ set by the transition dipole moment and the field amplitude.

Where this lives today

The Rabi oscillation became a control knob. In a trapped-ion, neutral-atom, or superconducting processor, coherent control is implemented with calibrated drive pulses: an ideal resonant $\pi$ pulse implements a population inversion (a NOT operation in the appropriate qubit basis), a $\pi/2$ pulse prepares an equal superposition, and the pulse phase selects the rotation axis. Gate fidelity depends in part on how firmly $\Omega\gg\Gamma$ — how many coherent cycles fit before relaxation interrupts them — but also on detuning, pulse-shape errors, dephasing, and other control imperfections. Engineers watch live "Rabi flops" on a scope while tuning the drive, reading $\Omega$ off the flopping period. The same physics runs the world's clocks. Ramsey's two- $\pi/2$ -pulse sequence, with a long free-evolution gap, interrogates the atomic phase and gives fringes far narrower than a single Rabi resonance — the heart of caesium fountains and optical-lattice clocks stable to $10^{-18}$ , with detuning entering through the generalised Rabi frequency of equation (4). Strong resonant driving splits levels by $\Omega$ into an Autler–Townes doublet; far-detuned driving shifts them by $\Omega^2/4\delta$ , the light shift that traps atoms in optical tweezers and lattices; and a coherent $2\pi$ pulse can slide through a resonant medium without net absorption (self-induced transparency). From atomic clocks to superconducting qubits, the same two-level coherent-control physics remains a central building block of modern quantum technology.

The physical picture

The cleanest way to see a Rabi oscillation is to stop thinking about photons and picture a spinning arrow. Any two-level system — ground $|g\rangle$ , excited $|e\rangle$ — maps exactly onto a spin-1/2, and its state maps onto a unit vector on the Bloch sphere: south pole "definitely ground", north pole "definitely excited", every point between a coherent superposition with a definite phase. The near-resonant field, in the frame that rotates with it, acts like a static effective field — or equivalently an effective rotation vector — and the Bloch state vector precesses about it. That precession is the Rabi oscillation, and the whole phenomenon reduces to one question: where does the torque axis point? On resonance, the effective rotation axis $\boldsymbol{\Omega}_{\rm eff}$ lies in the equatorial plane. A vector starting at the south pole precesses in a great circle straight over the top — ground to excited and back, a complete sweep with nothing out of reach. That is why resonant driving gives complete inversion. Detune the field and the axis tilts up out of the plane; the effective field grows longer, $\Omega_R=\sqrt{\Omega^2+\delta^2}$ , so precession speeds up, but the state now circles a tilted axis on a cone that never reaches the north pole. Faster but incomplete — the two signatures of off-resonant driving in one picture.

Figure

The Bloch-sphere torque picture. On resonance δ = 0 the drive torque Ω lies in the equatorial plane and sweeps the state vector from the ground pole g up to the excited pole e — a complete Rabi cycle. Detuned by δ = Ω the effective field tilts 45° and the vector precesses on a cone that reaches only the equator, so P_e peaks at just 0.5

Two facts fall straight out. First, in the ideal closed two-level model, energy is conserved: as the Bloch vector climbs, the atom gains energy from the driving field, and as it falls it returns that energy to the field by stimulated emission — the energy exchange is reversible over a coherent Rabi cycle. Second, the coupling scales with field amplitude, $\Omega=\mu E_0/\hbar\propto\sqrt I$ : for a fixed mode and polarization, doubling the intensity makes $\Omega$ rise by only $\sqrt2\approx1.41$ (41%), rather than by a factor of two as an intensity-linear rate would — a sharp distinction between coherent Rabi dynamics and incoherent rate-equation dynamics. The second figure plots the excited probability against the dimensionless clock $\Omega t$ . Read off the two lessons: the resonant curve reaches the ceiling $P_e=1$ (a $\pi$ pulse), while the detuned curve is faster and clipped at $0.5$ .

Figure

Excited-state probability P_e versus the dimensionless time Ωt. On resonance green it is sin²(Ωt/2), swinging fully between 0 and 1 with a π pulse at Ωt = π. Detuned by δ = Ω amber it runs faster at Ω_R = √2 Ω yet reaches only P_e,max = Ω²/(Ω²+δ²) = 0.5

The core equations

Everything starts from a two-level atom coupled to a classical field through the electric-dipole interaction. The bare atom carries a splitting $\hbar\omega_0$ between $|g\rangle$ and $|e\rangle$ ; the field adds the dipole coupling $-\boldsymbol\mu\cdot\mathbf E(t)$ recalled from the electric-dipole-approximation article. This Hamiltonian is the entire input — no relaxation yet, just a coherent drive.

\hat H = \hbar\omega_0\,|e\rangle\langle e| - \boldsymbol{\mu}\cdot\mathbf{E}(t) \qquad (1)

where: $\omega_0$ — atomic transition angular frequency; $|e\rangle,|g\rangle$ — excited and ground states; $\boldsymbol\mu$ — atomic electric-dipole operator; $\mathbf E(t)=E_0\cos\omega t\,\hat{\mathbf e}$ — classical driving field of amplitude $E_0$ and polarization $\hat{\mathbf e}$ . The reading: a two-state energy ladder shaken by an oscillating dipole force. The strength of that shaking is one number, the Rabi frequency: the dipole matrix element along the field, times the amplitude, over $\hbar$ . Being linear in $E_0$ it is linear in $\sqrt I$ , and it has units of angular frequency because it is the rate at which the driven atom cycles.

\Omega \equiv \frac{\mu\, E_0}{\hbar}, \qquad \mu = \langle e|\,\hat{\mathbf e}\cdot\boldsymbol{\mu}\,|g\rangle \qquad (2)

where: $\Omega$ — on-resonance Rabi angular frequency; $\mu$ — transition dipole matrix element along the field; $E_0$ — peak field amplitude; $\hbar$ — reduced Planck constant. The reading: a bigger dipole or stronger field cycles the atom faster, and the dependence on $E_0$ (not $E_0^2$ ) is what makes Rabi coherent, not rate-like. Substituting $|\psi\rangle=c_g|g\rangle+c_e e^{-i\omega_0 t}|e\rangle$ into the Schrödinger equation and dropping the fast counter-rotating term (the rotating-wave approximation, derived below) collapses the dynamics to two coupled amplitude equations — the engine of the whole phenomenon.

i\dot c_e = -\tfrac{\Omega}{2}\,e^{-i\delta t}\,c_g,\qquad i\dot c_g = -\tfrac{\Omega}{2}\,e^{+i\delta t}\,c_e \qquad (3)

where: $c_e,c_g$ — excited and ground amplitudes; $\delta=\omega-\omega_0$ — detuning from resonance; $\Omega$ — Rabi frequency. The reading: each amplitude drives the other at rate $\Omega/2$ , and detuning appears only as a rotating phase — the seed of both the off-resonant speed-up and its incompleteness. Solving the coupled pair (Derivation C) gives an oscillation whose frequency is not $\Omega$ but the length of the effective-field vector $(\Omega,0,-\delta)$ — the generalised Rabi frequency, the key off-resonant quantity.

\Omega_R = \sqrt{\Omega^2 + \delta^2} \qquad (4)

where: $\Omega_R$ — generalised Rabi frequency; $\Omega$ — on-resonance value; $\delta$ — detuning. The reading: detuning adds in quadrature, so the atom always cycles at least as fast as $\Omega$ , faster the further you tune away. Frequency and amplitude together give the Rabi formula — the exact excited-state probability for a ground-state atom at any detuning, the master result of the article.

P_e(t) = \frac{\Omega^2}{\Omega^2+\delta^2}\,\sin^2\!\left(\frac{\Omega_R\,t}{2}\right) \qquad (5)

where: $P_e(t)$ — excited occupation; prefactor $\Omega^2/(\Omega^2+\delta^2)$ — the maximum transfer, unity on resonance; $\Omega_R$ — sets the period. The reading: a fixed-shape $\sin^2$ oscillation whose height is throttled by detuning and whose speed is boosted by it. For an ideal resonant pulse with a fixed phase, what matters is not the peak Rabi frequency but its time integral, the pulse area — the one scalar that determines the population transfer, so different pulse shapes with the same area give the same rotation angle.

\theta = \int_0^{\tau}\Omega(t)\,\mathrm{d}t \qquad (6)

where: $\theta$ — pulse area; $\tau$ — duration; $\Omega(t)$ — instantaneous Rabi frequency. The reading: $\theta=\pi$ fully inverts, $\theta=\pi/2$ makes an equal superposition, $\theta=2\pi$ returns to ground — the alphabet of coherent control. Real atoms decohere, and adding relaxation at rate $\Gamma$ damps the oscillation to a steady state. Balancing drive against decay gives the saturated population — the bridge from Rabi dynamics to the ordinary absorption line.

P_e^{ss} = \frac{1}{2}\,\frac{\Omega^2/2}{\delta^2 + \Gamma^2/4 + \Omega^2/2} = \frac{1}{2}\,\frac{s}{1+s},\quad s=\frac{\Omega^2/2}{\delta^2+\Gamma^2/4} \qquad (7)

where: $\Gamma$ — total relaxation rate; $s$ — saturation parameter; $P_e^{ss}$ — steady excited population. The reading: as $s\to\infty$ it saturates at exactly $1/2$ — stimulated emission balances absorption, and the atom can never be more than half-excited on average once coherence is gone.

Derivation AFrom the Schrödinger equation to the rotating-wave amplitude equations

The goal is to turn the dipole Hamiltonian of equation (1) into the two coupled amplitude equations (3) that generate Rabi oscillations, and to expose the one approximation — the rotating-wave approximation — on which the whole clean picture rests. We move into the frame co-rotating with the bare atomic phases and identify which of the two driving terms can be discarded and why, leaving a pair simple enough to solve in closed form while keeping all the physics that matters at optical frequencies.

Step 1. With the fast atomic phase factored out, $|\psi\rangle=c_g|g\rangle+c_e e^{-i\omega_0 t}|e\rangle$ , substitute into $i\hbar\,\partial_t|\psi\rangle=\hat H|\psi\rangle$ with $\hat H_{int}=-\mu E_0\cos(\omega t)(|e\rangle\langle g|+|g\rangle\langle e|)$ . Projecting onto $\langle e|$ and writing $\cos\omega t=\tfrac12(e^{i\omega t}+e^{-i\omega t})$ gives $i\dot c_e=-\tfrac{\Omega}{2}(e^{i(\omega-\omega_0)t}+e^{-i(\omega+\omega_0)t})c_g$ . Why: factoring $e^{-i\omega_0 t}$ moves us to the interaction picture, where only the slow drive–atom mismatch survives explicitly.

Step 2. Split the two exponentials by frequency. The first, $e^{i\delta t}$ , oscillates slowly at the detuning; the second, $e^{-i(\omega+\omega_0)t}$ , oscillates at $\approx2\omega_0$ (hundreds of terahertz) and averages to nothing over the slow timescale $1/\Omega$ . Why: a term that flips sign a million times per Rabi cycle contributes zero net drive.

Step 3. Discard the fast term. This is the rotating-wave approximation, valid whenever $\Omega,|\delta|\ll\omega_0$ , and it leaves exactly the coupled pair of equation (3). The neglected term is not truly zero — it produces the small Bloch–Siegert shift of order $\Omega^2/4\omega_0$ , measurable in strongly driven microwave and circuit-QED systems. Why: the RWA is a controlled truncation, not a fudge, and it fails in the ultrastrong-coupling regime where $\Omega$ approaches $\omega_0$ .

Go deeper · Derivation B · 1 minThe on-resonance solution complete inversion and pulse areas

Derivation BThe on-resonance solution complete inversion and pulse areas

The goal is to solve the amplitude equations on exact resonance and read off the two results that make Rabi oscillations useful: complete, deterministic inversion, and the $\pi$ and $\pi/2$ pulses that are the primitive gates of coherent control. Setting $\delta=0$ removes the rotating phases and reduces the coupled system to a single harmonic-oscillator equation whose conclusions run all the way to quantum computing.

Step 1. On resonance $\delta=0$ , equation (3) becomes $i\dot c_e=-\tfrac{\Omega}{2}c_g$ and $i\dot c_g=-\tfrac{\Omega}{2}c_e$ . Differentiate the first and substitute the second to decouple them: $\ddot c_e=-(\Omega/2)^2 c_e$ . This is simple harmonic motion in the amplitude at frequency $\Omega/2$ . Why: eliminating $c_g$ turns two coupled equations into one textbook oscillator, and the appearance of $\Omega/2$ — half the Rabi frequency — in the amplitude is exactly why the probability, which squares it, oscillates at $\Omega$ .

Step 2. With $c_e(0)=0$ , $c_g(0)=1$ the solution is $c_e(t)=i\sin(\Omega t/2)$ , $c_g(t)=\cos(\Omega t/2)$ , so $P_e(t)=|c_e|^2=\sin^2(\Omega t/2)$ , and probability is conserved: $|c_e|^2+|c_g|^2=1$ . Why: unitary evolution pins total probability to one, so population genuinely sloshes between the states rather than leaking away — the signature of coherent, lossless dynamics.

Step 3. Evaluate at special times. At $\Omega t=\pi$ , $P_e=1$ : a $\pi$ pulse completely inverts the atom. At $\Omega t=\pi/2$ , $P_e=1/2$ : a $\pi/2$ pulse prepares the equal superposition $\frac{1}{\sqrt2}(|g\rangle+i|e\rangle)$ used in Ramsey interferometry. At $\Omega t=2\pi$ , the atom returns to the ground state with the state vector acquiring an overall minus sign, reflecting the spin- $1/2$ nature of the quantum state under a $2\pi$ rotation. Why: the outcome depends only on the accumulated area $\theta=\int\Omega\,dt$ , so calibrating the area gives robust gates independent of pulse shape.

Go deeper · Derivation C · 1 minThe generalised Rabi formula and the damped steady state

Derivation CThe generalised Rabi formula and the damped steady state

The goal is twofold: first to solve the amplitude equations at arbitrary detuning for the master Rabi formula (5), showing how detuning both speeds up and caps the oscillation; second to add relaxation and watch it decay into the incoherent rate-equation steady state — showing that coherent cycling and irreversible absorption are the same physics on two timescales.

Step 1. Off resonance, eliminate $c_g$ from equation (3) to get $\ddot c_e-i\delta\dot c_e+\tfrac{\Omega^2}{4}c_e=0$ , with roots $i\delta/2\pm i\Omega_R/2$ and $\Omega_R=\sqrt{\Omega^2+\delta^2}$ . With $c_e(0)=0$ the solution is $c_e\propto e^{i\delta t/2}\sin(\Omega_R t/2)$ , and $|c_e|^2$ gives $P_e=\dfrac{\Omega^2}{\Omega^2+\delta^2}\sin^2\!\big(\tfrac{\Omega_R t}{2}\big)$ . Why: the detuning enters as a phase $e^{i\delta t/2}$ that drops out of $|c_e|^2$ but survives inside $\Omega_R$ — hence faster oscillation, throttled amplitude.

Step 2. Read the prefactor. On resonance it is $1$ and $P_e$ reaches the ceiling; for $|\delta|\gg\Omega$ it collapses to $\Omega^2/\delta^2\ll1$ — tiny fast wobbles underlying the AC Stark (light) shift. The maximum transfer $P_{e,\max}=\Omega^2/(\Omega^2+\delta^2)$ is unity only on resonance. Why: this is the quantitative content of the Bloch-sphere cone — a tilted axis cannot reach the excited pole, and the shortfall is exactly $\delta^2/\Omega_R^2$ .

Step 3. Add relaxation. Recasting the amplitudes as the Bloch vector $(u,v,w)$ , $w=2P_e-1$ , with coherences decaying at $\Gamma/2$ and inversion at $\Gamma$ , gives $\dot u=\delta v-\tfrac{\Gamma}{2}u$ , $\dot v=-\delta u+\Omega w-\tfrac{\Gamma}{2}v$ , $\dot w=-\Omega v-\Gamma(w+1)$ . For weak damping $\Gamma\ll\Omega$ on resonance, $P_e(t)\approx\tfrac12[1-e^{-3\Gamma t/4}\cos\Omega t]$ : many cycles before settling. Why: the envelope is the death of coherence — count surviving cycles and you measure $\Omega/\Gamma$ , the qubit figure of merit.

Step 4. Set the derivatives to zero for the steady state: $P_e^{ss}=\tfrac12\,s/(1+s)$ , $s=(\Omega^2/2)/(\delta^2+\Gamma^2/4)$ , saturating at $1/2$ as $s\to\infty$ , with power-broadened width $\sqrt{\Gamma^2+2\Omega^2}$ . Why: in the overdamped limit $\Gamma\gg\Omega$ the coherences follow adiabatically and $P_e$ obeys a single rate equation — the Einstein picture with its Lorentzian line re-emerges as the memoryless limit of Rabi dynamics.

Limiting cases

The behaviour is set by three rates — coupling $\Omega$ , detuning $\delta$ , relaxation $\Gamma$ — and each corner of that space has a clean, nameable limit.

Case	Result	Meaning
Exact resonance, no damping	$P_e(t)=\sin^2\!\left(\tfrac{\Omega t}{2}\right)$	Complete, undamped inversion — the ideal two-state oscillator; a $\pi$ pulse reaches $P_e=1$ .
Large detuning $\lvert\delta\rvert\gg\Omega$	$P_{e,\max}=\dfrac{\Omega^2}{\Omega^2+\delta^2}\to\dfrac{\Omega^2}{\delta^2}\ll1$	Fast, tiny wobbles; the atom stays essentially in the ground state — the regime of the AC Stark / light shift and dipole traps.
Strong damping $\Gamma\gg\Omega$	$P_e(t)\to P_e^{ss}$ (overdamped)	Oscillations vanish; monotonic approach to the rate-equation steady state — the Einstein absorption regime.
Weak damping $\Gamma\ll\Omega$	$P_e(t)\approx\tfrac12\!\left[1-e^{-3\Gamma t/4}\cos\Omega t\right]$	Underdamped Rabi oscillations, many cycles visible, relaxing to $P_e^{ss}=1/2$ .
Strong saturation $s\to\infty$	$P_e^{ss}\to\tfrac12$	Population caps at one-half; stimulated emission balances absorption — an atom cannot be driven fully excited on average.

Common trapRabi oscillation is just absorption happening fast

It is not — and conflating the two is the most common error here. Ordinary absorption is a rate process: a one-way, irreversible jump whose rate scales linearly with intensity $I$ (the Einstein $B$ coefficient). A Rabi oscillation is coherent and reversible: population flows both ways, and its rate $\Omega=\mu E_0/\hbar$ scales with the amplitude $E_0\propto\sqrt I$ . Double the intensity and an absorption rate doubles while $\Omega$ grows by only $\sqrt2$ . The rate-equation picture is not wrong — it is what Rabi dynamics becomes once decoherence erases the phase faster than one Rabi period, so the atom can no longer return the energy it borrowed. Treat absorption as the damped, forgetful limit of Rabi cycling, never a competing description. A second trap: the clean picture rests on the rotating-wave approximation, which quietly fails at ultrastrong driving.

Worked examples

1 · Rabi frequency and π-pulse time for a real atom (confidence builder)

A near-resonant laser drives an alkali D-line transition with dipole matrix element $\mu=2.5\times10^{-29}\ \mathrm{C\,m}$ (about $3\,ea_0$ , a realistic value) at intensity $I=100\ \mathrm{W/m^2}$ . Find the field amplitude, the Rabi frequency in rad/s and in ordinary frequency, and the duration of a resonant $\pi$ pulse.

Step 1. Intensity to field: for a plane wave $I=\tfrac12 c\varepsilon_0 E_0^2$ , so $E_0=\sqrt{2I/(c\varepsilon_0)}=\sqrt{2(100)/(2.998\times10^8\cdot8.854\times10^{-12})}=\sqrt{7.54\times10^{4}}=\mathbf{274\ V/m}$ — the peak field the dipole feels.

Step 2. Rabi frequency from equation (2): $\Omega=\mu E_0/\hbar=(2.5\times10^{-29})(274.7)/(1.055\times10^{-34})=6.51\times10^7\ {\rm rad/s}$ , i.e. $\Omega/2\pi=10.36\ {\rm MHz}$ . This sits far below the transition frequency ( $\sim5\times10^{14}$ Hz), so the RWA is superb.

Step 3. A $\pi$ pulse needs $\Omega\tau=\pi$ , so $\tau_\pi=\pi/\Omega=\mathbf{48\ ns}$ ; a $\pi/2$ pulse takes $\mathbf{24\ ns}$ . In that window the atom is carried deterministically from ground to fully excited.

Step 4. Scaling check: doubling the intensity to $200\ \mathrm{W/m^2}$ raises $E_0$ and $\Omega$ by $\sqrt2$ — from 10.4 to $\mathbf{14.6\ MHz}$ , a 41 % increase, not the 100 % an intensity-linear rate would show. Result: $E_0=274$ V/m, $\Omega=6.5\times10^7$ rad/s (10.4 MHz), $\tau_\pi=48$ ns, $\Omega\propto\sqrt I$ .

2 · How detuning throttles the transfer

The same atom is now driven off resonance. Compute the generalised Rabi frequency and the maximum excited population for detunings $\delta=\Omega$ and $\delta=2\Omega$ , and say what happens to the oscillation period.

Step 1. For $\delta=\Omega$ , equation (4) gives $\Omega_R=\sqrt2\,\Omega=\mathbf{1.41\,\Omega}$ (41 % faster), and equation (5) gives $P_{e,\max}=\Omega^2/(\Omega^2+\delta^2)=\mathbf{0.50}$ — the oscillation tops out at the equator, the clipped amber curve above.

Step 2. For $\delta=2\Omega$ , $\Omega_R=\sqrt5\,\Omega=\mathbf{2.24\,\Omega}$ and $P_{e,\max}=1/5=\mathbf{0.20}$ : twice the detuning more than doubles the speed while cutting the reach to a fifth.

Step 3. The trend: detuning always shortens the period ( $\Omega_R>\Omega$ ) and lowers the ceiling (prefactor $<1$ ), the two locked together through the vector $(\Omega,0,-\delta)$ . Result: $\delta=\Omega\Rightarrow\Omega_R=1.41\,\Omega,\ P_{e,\max}=0.50$ ; $\delta=2\Omega\Rightarrow\Omega_R=2.24\,\Omega,\ P_{e,\max}=0.20$ .

3 · Saturating a damped transition

Drive the same atom on resonance ( $\delta=0$ ) with relaxation rate $\Gamma$ . Find the steady-state excited population at (a) the saturation point $\Omega^2/2=\Gamma^2/4$ and (b) hard drive $\Omega^2/2=100\,\Gamma^2/4$ .

Step 1. On resonance $s=(\Omega^2/2)/(\Gamma^2/4)$ , so case (a) $s=1$ (drive equal to decay) and case (b) $s=100$ . From equation (7), $P_e^{ss}=\tfrac12\,s/(1+s)$ : case (a) $\mathbf{0.25}$ ; case (b) $\tfrac12(100/101)=\mathbf{0.495}$ — within 1 % of the ceiling.

Step 2. However hard you drive, $P_e^{ss}\to1/2$ : once coherence is lost, stimulated emission empties $|e\rangle$ as fast as absorption fills it. This incoherent ceiling is what the coherent $\pi$ pulse of Example 1 briefly beats, reaching $P_e=1$ before decoherence catches up. Result: $s=1\Rightarrow0.25$ ; $s=100\Rightarrow0.495$ ; ceiling $1/2$ .

Physics of the portrait

Rabi oscillations carry their discoverer's name because he found the resonance formula not in light but in magnetism — and the two turned out to be one equation.

Origins · 1 minIsidor Isaac Rabi and how this was discovered

Isidor Isaac Rabi

1898–1988 · molecular-beam magnetic resonance and the two-level rotation formula

Isidor Isaac Rabi was born in Rymanów, in what is now Poland, and raised in New York; at Columbia he built the molecular-beam laboratory that defined mid-century atomic physics. In 1937 he analysed a spin in a static magnetic field driven by a small rotating field at the resonant frequency, and derived the formula — maximum flip probability $\Omega^2/(\Omega^2+\delta^2)$ , oscillation at $\sqrt{\Omega^2+\delta^2}$ — that now bears his name. The next year his group (Rabi, Millman, Kusch, Zacharias) turned it into the molecular-beam magnetic-resonance method, measuring nuclear magnetic moments by sweeping a radio-frequency field across a beam and watching the resonant dip; for this he received the 1944 Nobel Prize. The formula was written for spins, but the mathematics is identical for a two-level atom driven through its transition dipole — which is why the coherent light–matter cycling here carries his name. His student Norman Ramsey extended the method into the separated-oscillatory-fields technique behind atomic clocks, and the same resonance logic runs through NMR, MRI, and every qubit driven by a calibrated pulse today.

History

1937

— Rabi derives the resonant spin-flip formula for a field rotating at the transition frequency

1938

— Rabi, Millman, Kusch and Zacharias demonstrate molecular-beam magnetic resonance

1944

— Rabi is awarded the Nobel Prize in Physics for the resonance method

1950

— Ramsey introduces the separated-oscillatory-fields method, sharpening the Rabi resonance

1969

— McCall and Hahn report self-induced transparency, a macroscopic 2π-pulse effect in optics

Going further

Texts. Foot, Atomic Physics, Ch. 7 (the interaction of atoms with radiation) · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 2 · Allen & Eberly, Optical Resonance and Two-Level Atoms (the definitive coherent-transient reference) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., Ch. 13–14 · Demtröder, Laser Spectroscopy, Vol. 1 · Steck, Quantum and Atom Optics (open lecture notes) for the optical Bloch equations. The key results to keep in view are: the RWA amplitude equations and $P_e=\sin^2(\Omega t/2)$ at $\delta=0$ ; the general Rabi formula (5) with $\Omega_R=\sqrt{\Omega^2+\delta^2}$ and $P_{e,\max}=\Omega^2/\Omega_R^2$ ; the steady state $P_e^{ss}=\tfrac12\,s/(1+s)$ , $s=(\Omega^2/2)/(\delta^2+\Gamma^2/4)$ , with power-broadened width $\sqrt{\Gamma^2+2\Omega^2}$ ; the coupling $\Omega=\mu E_0/\hbar\propto\sqrt I$ ; and the anchor numbers $E_0=274$ V/m, $\Omega/2\pi=10.4$ MHz, $\tau_\pi=48$ ns for $\mu=2.5\times10^{-29}$ C·m at $I=100$ W/m².

Prerequisites. Electric Dipole Approximation · Transition Dipole Moment · Stimulated Emission. Leads to. Radiative Lifetime and Natural Linewidth · Lorentzian Line Shape · Resonant Absorption.

Ask the tutor: Want to see the Rabi flop trace turn into a smooth absorption line as you dial up the decoherence rate Γ? Ask me to walk through the optical Bloch equations step by step.

Why does the excited population oscillate at Ω but the amplitude at Ω/2?

Because probability is amplitude squared. On resonance the amplitudes evolve as $\cos(\Omega t/2)$ and $\sin(\Omega t/2)$ , each cycling at half the Rabi frequency. But the observable is $P_e=|c_e|^2=\sin^2(\Omega t/2)$ , and $\sin^2 x=\tfrac12(1-\cos2x)$ doubles the frequency, so $P_e$ oscillates at the full $\Omega$ . It is the same factor of two that makes a spin-1/2 need a $4\pi$ rotation to return exactly to itself while its population repeats every $2\pi$ of $\Omega t$ . Keep the two clocks separate: amplitudes tick at $\Omega/2$ , populations at $\Omega$ .

Is energy conserved during a Rabi oscillation?

Yes, completely. While $P_e$ climbs the atom absorbs energy; while it falls the atom returns that energy by stimulated emission, coherently and in phase with the drive, so the net exchange over a cycle is zero. This is exactly what makes Rabi cycling reversible and unlike absorption. Net one-way transfer appears only when decoherence (spontaneous emission, collisions) interrupts the descent so the atom cannot give back what it borrowed; then the exchange is irreversible and you measure genuine absorption. Reversible cycling and irreversible absorption are the same interaction with and without a working phase memory.

When is the rotating-wave approximation safe to use?

Whenever the Rabi frequency and detuning are both far below the transition frequency, $\Omega,|\delta|\ll\omega_0$ . For an optical transition $\omega_0\sim5\times10^{14}$ Hz while a strong $\Omega$ is $\sim10^7$ – $10^{10}$ Hz, so the RWA is excellent — the discarded term oscillates at $\approx2\omega_0$ and averages away. It matters only under extreme driving: the neglected term gives the Bloch–Siegert shift $\sim\Omega^2/4\omega_0$ , seen in strongly driven NMR and superconducting circuits, and in the ultrastrong-coupling regime where $\Omega\sim\omega_0$ the simple formula fails and the full quantum Rabi model is needed. For essentially all of atomic and optical physics it is nothing to worry about.

What is the difference between the Rabi frequency Ω and the generalised Rabi frequency Ω_R?

$\Omega=\mu E_0/\hbar$ is the on-resonance coupling — the cycling frequency when the laser sits exactly on the transition, $\delta=0$ . The generalised Rabi frequency $\Omega_R=\sqrt{\Omega^2+\delta^2}$ is the actual oscillation frequency once you detune; because detuning adds in quadrature, $\Omega_R\ge\Omega$ always, and the two coincide only on resonance. The distinction matters twice: $\Omega_R$ sets how fast the population oscillates, while the ratio $\Omega^2/\Omega_R^2$ sets how high it reaches. A detuned atom cycles faster yet never fully inverts — both facts encoded in the effective-field vector $(\Omega,0,-\delta)$ of length $\Omega_R$ .

Why can't the steady-state excited population exceed one-half?

Once coherence is destroyed the atom obeys rate equations in which the field drives absorption ( $g\to e$ ) and stimulated emission ( $e\to g$ ) at equal rates. At $P_e=1/2$ these balance — as many atoms driven down as up — so the population stops climbing however hard you drive, and spontaneous emission only pulls it back further. Hence $P_e^{ss}=\tfrac12\,s/(1+s)\to1/2$ . A coherent $\pi$ pulse can momentarily beat this ceiling and reach $P_e=1$ , but only within the decoherence time; once phase memory is lost the half-excitation cap reasserts itself. Full inversion is a coherent privilege, not a steady-state option.

How is a Rabi oscillation actually observed in the lab?

You prepare the atom (or ion, or qubit) in the ground state, drive it resonantly for a controlled time $t$ , switch off, and read out the excited population — often by fluorescence, since $|e\rangle$ scatters photons. Plot $P_e$ versus $t$ and you see the sinusoidal "Rabi flop" of period $2\pi/\Omega$ , giving $\Omega$ directly; scanning the amplitude confirms $\Omega\propto\sqrt I$ and scanning the detuning maps out $\Omega_R$ . In practice the flops decay: the envelope's rate measures decoherence, and the number of clean cycles before it dies is the coherence figure of merit qubit engineers work to maximise.