Rabi Spectroscopy in Quantum Systems

• Rabi spectroscopy is the coherent interrogation of two-level quantum systems using time-limited electromagnetic pulses, enabling precise measurements of resonance frequencies and transition probabilities.
• It exploits square pulse-driven Rabi oscillations to determine critical parameters such as the on-resonance Rabi frequency, coherence time, and Fourier-limited spectral widths.
• The technique is widely applied in optical lattice clocks, superconducting circuits, and Floquet-engineered systems to probe noise, manage collisional shifts, and enhance metrological precision.

Searching arXiv for the cited papers and recent relevant work on Rabi spectroscopy. Searching for exact arXiv IDs and closely related Rabi spectroscopy papers. Rabi spectroscopy is the coherent interrogation of a two-level or effectively two-level quantum system by a time-limited, near-resonant electromagnetic field. In its standard form, a square pulse of duration $T$ drives population oscillations between the ground and excited states, and the resulting transition probability as a function of detuning or pulse duration yields the Rabi lineshape, the on-resonance Rabi frequency, coherence time, and systematic shifts. Across atomic clocks, superconducting circuits, semiconductor quantum dots, electron paramagnetic resonance, cavity quantum electrodynamics, trapped gases, and spin systems, the method serves both as a precision metrological tool and as a probe of driven many-body, Floquet, and noise-dressed dynamics (Liu et al., 2022, Yin et al., 2020, Matityahu et al., 2017).

1. Fundamental framework

For a two-level system driven by a square pulse of duration $T$, detuning $\Delta$, and on-resonance Rabi frequency $\Omega$, the standard transition probability is

$$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$

On resonance, this reduces to

$P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$

The pulse-area conventions are correspondingly defined by $\Omega T=\pi$ for a $\pi$ pulse and $\Omega T=\pi/2$ for a $\pi/2$ pulse (Liu et al., 2022).

The frequency-domain response of a square pulse has a sinc-like envelope with a Fourier-limited width that scales as $T$0, so longer interrogation narrows the main spectral lobe (Liu et al., 2022). This basic structure reappears across implementations. In optical lattice clocks, it defines the carrier and sideband line shapes of the clock transition (Yin et al., 2020, 0906.1419). In superconducting two-level defects and flux qubits, the same formalism is recast in rotating-frame language to connect Rabi decay to environmental noise at the Rabi frequency (Matityahu et al., 2017, Yoshihara et al., 2014). In cavity QED, the time-domain oscillation frequency is

$T$1

so that on resonance excitation swaps sinusoidally at frequency $T$2, while the stationary spectrum exhibits vacuum Rabi splitting (Yu et al., 21 May 2026).

In interacting or structured systems, the canonical two-level description is generalized rather than abandoned. In optical lattice clocks with motional structure, the observed signal is a thermal average over motional-state-dependent Rabi couplings (0906.1419, Yin et al., 2020). In periodically driven systems, the effective coupling to a given Floquet sideband is Bessel-renormalized, and the resonance condition is shifted by integer multiples of the modulation frequency (Liu et al., 2022, Yin et al., 2020). In many-body spin models, collective interactions transform the Rabi spectrum from a symmetric population transfer curve into an antisymmetric or Heisenberg-narrowed discriminator (Huang et al., 2022, Shaniv et al., 2017).

2. Optical lattice clocks and precision interrogation

Optical lattice clocks provide one of the most developed implementations of Rabi spectroscopy. In the fermionic $T$3 optical lattice clock, the clock transition is the ultra-narrow doubly forbidden $T$4 line at $T$5, with atoms confined in a one-dimensional optical lattice at the magic wavelength $T$6 (Liu et al., 2022). In the Floquet-engineered realization, about $T$7 atoms are cooled to $T$8 and $T$9 and trapped in a deep 1D lattice with $\Delta$0, $\Delta$1, $\Delta$2, and beam waist $\Delta$3 (Liu et al., 2022). Narrow-line spectra are acquired at low probe power of about $\Delta$4 with interrogation times of $\Delta$5 for scans and up to $\Delta$6 for time-domain oscillations (Liu et al., 2022).

In the earlier Floquet-engineered Sr clock, the same spectroscopic logic was used to resolve multiple Floquet bands while preserving clock sensitivity. The clock laser at $\Delta$7, with linewidth $\Delta$8, interrogated about $\Delta$9 atoms at $\Omega$0 in a quasi-1D lattice at $\Omega$1, with trap frequencies $\Omega$2 and $\Omega$3 (Yin et al., 2020). The frequency was scanned in $\Omega$4 steps with probe times $\Omega$5–$\Omega$6, yielding linewidths of a few hertz (Yin et al., 2020).

The observable in these clock settings is not purely internal. In a 1D optical lattice clock, atomic motion modifies carrier Rabi flopping and sideband spectra through motional-state-dependent matrix elements. The carrier coupling for a trapped atom in state $\Omega$7 is

$\Omega$8

with Lamb–Dicke parameters set by the trap frequencies and probe geometry (0906.1419). This leads to inhomogeneous excitation across a thermal ensemble and dephasing of the ensemble-averaged Rabi oscillations. In the same system, longitudinal sideband spectra provide direct access to trap frequencies and temperatures, while transverse motion and small probe–lattice misalignment generate carrier inhomogeneity that becomes central for collisional shifts (0906.1419).

Rabi spectroscopy also enables high-resolution sideband thermometry and excitation counting in interacting trapped gases. In a finite-temperature trapped Bose gas of $\Omega$9, a controlled displacement $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$0 between spin-dependent traps turns the internal-state Rabi drive into a probe of motional sidebands (Allard et al., 2016). With $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$1 and $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$2 for a $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$3 pulse, the technique resolves carrier, red, and blue sidebands across the BEC transition, and in a nearly pure condensate with $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$4 yields an upper temperature bound of $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$5 from the absence of the red sideband (Allard et al., 2016).

3. Floquet engineering and sideband structure

A major extension of Rabi spectroscopy is its use in periodically driven systems. In Floquet-engineered optical lattice clocks, periodic modulation of the lattice laser frequency generates an effective time-dependent detuning in the clock transition. In the $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$6 clock, a sinusoidal voltage applied to a piezoelectric transducer modulates the lattice laser frequency as

$$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$7

with $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$8 in the degenerate Rabi experiment (Liu et al., 2022). In the co-moving lattice frame, this yields a Landau–Zener–Stückelberg–Majorana-type Hamiltonian with a periodically modulated $$P_e(T,\Delta)=\frac{\Omega^2}{\Omega'^2}\sin^2\!\left(\frac{\Omega' T}{2}\right),\qquad \Omega'=\sqrt{\Omega^2+\Delta^2}.$$9 term (Liu et al., 2022).

The corresponding Floquet resonances occur at $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$0, and the effective coupling on the $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$1th sideband is governed by Bessel functions. In the resolved Floquet sideband approximation,

$P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$2

In the earlier Sr implementation, the effective Hamiltonian for the $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$3th Floquet band is

$P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$4

so the effective Rabi frequency is directly renormalized by $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$5 (Yin et al., 2020).

Experimentally, this produces multiple sharp peaks separated by $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$6, with linewidths remaining at a few hertz for $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$7 (Liu et al., 2022, Yin et al., 2020). The ability to suppress individual sidebands at zeros of $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$8 is central. In the degenerate Sr clock, at $P_e(T)=\sin^2\!\left(\frac{\Omega T}{2}\right).$9 the carrier sideband was suppressed below background, and at $\Omega T=\pi$0 the first sideband was suppressed, with time-domain excitation remaining near zero for the suppressed bands up to $\Omega T=\pi$1 (Liu et al., 2022). In the earlier Floquet sensitivity study, the zeroth band vanished at $\Omega T=\pi$2, and the first band was suppressed near $\Omega T=\pi$3 (Yin et al., 2020).

This Floquet formulation generalizes beyond clocks. In continuous-wave EPR, magnetic-field modulation at frequency $\Omega T=\pi$4 generates multiphoton sidebands weighted by Bessel functions, and Rabi resonance occurs when the modulation frequency matches the dressed-state splitting (Saiko et al., 2016). In nano-device chains, joint ac and dc driving produces Rabi–Bloch oscillations whose current and dipole spectra contain narrow lines at $\Omega T=\pi$5 in the ultrastrong regime (Levie et al., 2017). In proton-spin dressed-state spectroscopy, a strong off-resonant field produces multiple dressed-state transitions involving several dressing quanta, again requiring a beyond-RWA treatment (Schulthess et al., 16 Mar 2026).

A related development is the use of Rabi spectroscopy to detect super-Bloch oscillations in optical lattice clocks. In that setting, a static force $\Omega T=\pi$6 and periodic lattice drive are tuned near the condition

$\Omega T=\pi$7

leaving a residual effective force proportional to $\Omega T=\pi$8 and yielding a super-Bloch period

$\Omega T=\pi$9

Two-pulse Rabi spectroscopy then maps the quasimomentum evolution into a time-dependent return probability, allowing force metrology through the SBO period (Xiao et al., 2023).

4. Noise spectroscopy, decoherence, and driven-frame diagnostics

Rabi spectroscopy is also a frequency-selective probe of environmental noise. In superconducting two-level tunneling defects coupled to a phase qubit, the decay of Rabi oscillations was used to extract noise spectral density in the MHz range by varying the Rabi frequency $\pi$0 (Matityahu et al., 2017). The driven TLS Hamiltonian in the eigenbasis contains a resonant drive term and couplings to low- and high-frequency noise channels, and at resonance the total Rabi decay rate is

$\pi$1

Here $\pi$2 is energy relaxation from noise near the TLS splitting, $\pi$3 is rotating-frame relaxation from noise near $\pi$4, and $\pi$5 is second-order pure dephasing from slow thermal TLSs (Matityahu et al., 2017).

The key quantitative finding in that system was that the Rabi dephasing rate vanishes at the symmetry point $\pi$6 and scales as

$\pi$7

with measured coefficients $\pi$8 ranging from $\pi$9 to $\Omega T=\pi/2$0 across the four defects studied (Matityahu et al., 2017). This directly identifies quasi-static interacting defects as the dominant source of driven-frame dephasing.

An analogous strategy was used for a superconducting flux qubit under strong driving. There, Rabi frequencies from $\Omega T=\pi/2$1 up to $\Omega T=\pi/2$2 were achieved, approaching a qubit splitting of $\Omega T=\pi/2$3 where the rotating-wave approximation breaks down (Yoshihara et al., 2014). The measured Rabi-envelope decay was decomposed into a quasi-static contribution and an exponential term,

$\Omega T=\pi/2$4

with

$\Omega T=\pi/2$5

By subtracting the relaxation contribution, the flux-noise PSD $\Omega T=\pi/2$6 was extracted over the MHz–hundreds-of-MHz range, reaching about $\Omega T=\pi/2$7 near $\Omega T=\pi/2$8 (Yoshihara et al., 2014).

In semiconductor quantum dots, photon-echo implementations of Rabi spectroscopy isolate decoherence during the pulse itself rather than only between pulses. For trion ensembles in InGaAs quantum dots, the echo amplitude under two-pulse excitation obeys $\Omega T=\pi/2$9 for delta-like pulses, and spatial beam shaping was used to remove ensemble-averaging over local pulse areas (Grisard et al., 2022). With a flattop second pulse, Rabi rotations persisted up to $\pi/2$0, and the residual damping was quantitatively fit by a dressed-state phonon model with spectral density

$\pi/2$1

yielding $\pi/2$2 and $\pi/2$3 (Grisard et al., 2022). This identified acoustic phonons as the dominant intrinsic loss channel during strong picosecond excitation.

These implementations clarify a common point: Rabi spectroscopy is not only a way to determine a resonance frequency. It is also a driven-frame spectrometer of noise, damping, and microscopic couplings, with the effective probe frequency set by $\pi/2$4 rather than by the bare transition alone (Matityahu et al., 2017, Yoshihara et al., 2014).

5. Interactions, collisions, and many-body generalizations

In dense atomic ensembles, interactions distort the ideal Rabi lineshape and can shift the inferred resonance. In 1D optical lattice clocks with ultracold fermions, state-dependent Rabi couplings make atoms partially distinguishable, enabling $\pi/2$5-wave collisions even though identical fermions in the same internal state would otherwise forbid them (0906.1419). The resulting dynamic mean-field shift depends on the two-body correlation

$\pi/2$6

and on the time-dependent population imbalance during the interrogation (0906.1419).

For 3D optical lattice clocks at unity filling, short-range collisional shifts are suppressed, and long-range electronic dipole–dipole interactions become the dominant interaction mechanism (Liu et al., 2019). Starting from a Lindblad master equation, the Rabi clock shift was derived to first order in the interaction strength and shown to factorize as

$\pi/2$7

where $\pi/2$8 depends on the pulse parameters and $\pi/2$9 depends on lattice geometry (Liu et al., 2019). For representative $T$00 parameters, the predicted shift is $T$01, corresponding to a fractional effect at the $T$02 level (Liu et al., 2019).

A particularly important result for metrology is that collisional shifts in Rabi-interrogated optical lattice clocks can be cancelled by operating slightly over $T$03 pulse area. In the Yb clock analysis, the total collisional shift was decomposed into homogeneous $T$04-wave, inhomogeneous $T$05-wave, and inhomogeneous $T$06-wave contributions,

$T$07

with the inhomogeneous terms scaling as $T$08, where $T$09 quantifies Rabi-frequency inhomogeneity (Lee et al., 2015). Because the homogeneous $T$10-wave contribution becomes negative for over-$T$11 pulses while the inhomogeneous terms remain positive, the total shift can be nulled for sufficiently small $T$12. The analysis concluded that an over-$T$13 pulse combined with inhomogeneity below $T$14 allows a fractional uncertainty on the level of $T$15 in both Sr and Yb clocks (Lee et al., 2015).

Beyond collisional cancellation, interactions can be used constructively to sharpen the spectroscopic signal. In antisymmetric Rabi spectroscopy, the many-body Hamiltonian

$T$16

is interrogated after preparing an equal superposition state with a fast $T$17 pulse (Huang et al., 2022). Measuring $T$18 yields an exactly antisymmetric signal,

$T$19

for all $T$20, $T$21, and $T$22 provided the initial state has even Dicke-basis parity (Huang et al., 2022). Because the zero crossing remains pinned to $T$23, the resonance is free of collision-shift bias. For small interaction strength, the slope at resonance is enhanced relative to conventional Rabi spectroscopy, and for stronger interactions the protocol can beat the standard quantum limit in specific regimes while remaining robust to detection noise (Huang et al., 2022).

6. Correlated, Heisenberg-limited, and non-Hermitian variants

Rabi spectroscopy has also been generalized to correlated-spin and non-Hermitian settings. In the two-ion Heisenberg-limited Rabi protocol, the effective Hamiltonian

$T$24

conserves parity, so the even and odd two-ion subspaces behave as effective two-level systems with doubled detuning (Shaniv et al., 2017). Starting in $T$25 or $T$26, the correlated transition probabilities become

$T$27

and similarly for the odd subspace with $T$28 (Shaniv et al., 2017). The measured narrowing factors were $T$29 and $T$30 in the even and odd subspaces, close to the ideal factor of $T$31 for two ions (Shaniv et al., 2017).

In cavity QED, multi-modal time-domain spectroscopy extends the Rabi concept to a three-cavity architecture with one emitter (Yu et al., 21 May 2026). The middle-cavity projection of the zero-energy supermode determines the effective emitter–mode coupling,

$T$32

so abruptly tuning $T$33 switches the Rabi oscillation off by driving $T$34 (Yu et al., 21 May 2026). A generalized sensor method then reconstructs the nonstationary transient spectrum with computational scaling reduced from $T$35 to $T$36 (Yu et al., 21 May 2026). This suggests that “Rabi spectroscopy” can refer not only to scanning a simple two-level resonance but also to time-resolved monitoring of driven dressed-state structure in multimode systems.

A different generalization appears in PT-symmetric Rabi problems, where a periodic non-Hermitian perturbation replaces the usual Hermitian drive (Joglekar et al., 2014). In the two-level case,

$T$37

Near the resonance $T$38, the effective oscillation frequency becomes

$T$39

so oscillatory dynamics exists only for $T$40, while at exact resonance any nonzero $T$41 breaks PT symmetry and leads to exponential growth or decay (Joglekar et al., 2014). This replaces the familiar Hermitian Rabi resonance by a Floquet exceptional-point cone with threshold $T$42 (Joglekar et al., 2014).

7. Applications, precision strategies, and methodological contrasts

A unifying feature of Rabi spectroscopy is that its informational content depends strongly on how the time-domain and frequency-domain data are used. In muonium hyperfine spectroscopy, fitting the full time evolution of the Rabi oscillation rather than a frequency-swept resonance curve was shown to improve precision by a factor of about two at an optimized detuning near $T$43 (Nishimura et al., 2020). Applying the method to zero-field ground-state muonium yielded

$T$44

reported as the world’s highest precision under zero-field conditions (Nishimura et al., 2020). The improvement arises because the time trace separates detuning from drive amplitude while using the full decaying Rabi signal of a short-lived system (Nishimura et al., 2020).

In optical clocks, a comparable principle underlies the use of Fisher information to compare Floquet-engineered Rabi bands. For binary readout of the excited fraction, the classical Fisher information for detuning estimation is

$T$45

In the Floquet-engineered Sr clock, the maximum Fisher information of lower-order Floquet bands remained comparable to the undriven case, with the finite-temperature undriven benchmark near $T$46 (Yin et al., 2020). This supports the interpretation that periodic lattice modulation redistributes spectral weight without depleting metrological sensitivity (Yin et al., 2020).

Several recurrent trade-offs emerge across platforms. Longer interrogation sharpens the line through Fourier scaling but increases sensitivity to technical decoherence (Liu et al., 2022, Yin et al., 2020). Stronger drive boosts signal but can produce power broadening, excitation-induced dephasing, or beyond-RWA shifts (Grisard et al., 2022, Yoshihara et al., 2014). Inhomogeneity, whether from thermal motion, beam profile, or local field variation, damps Rabi oscillations and biases line-shape-based inference unless explicitly modeled (0906.1419, Grisard et al., 2022, Nishimura et al., 2020). Periodic driving adds sidebands and Bessel control but demands resolved-band conditions and careful handling of residual magnetic, light, or motional shifts (Liu et al., 2022, Xiao et al., 2023).

A common misconception is that Rabi spectroscopy is intrinsically lower resolution than Ramsey spectroscopy and therefore mainly a calibration tool. The literature suggests a more qualified view. Ramsey interrogation generally yields narrower fringes for long coherence times, but Rabi spectroscopy offers direct control over pulse area, natural access to sideband engineering, simpler error signals in some settings, and a particularly transparent mapping between drive amplitude, dressed-state structure, and noise sensitivity (Liu et al., 2022, Yin et al., 2020, Matityahu et al., 2017). In interaction-engineered or correlated settings, Rabi-type protocols can also access regimes not naturally available to standard Ramsey methods, including collision-shift-free antisymmetric locking (Huang et al., 2022) and Heisenberg-narrowed correlated rotations (Shaniv et al., 2017).

Taken together, these developments establish Rabi spectroscopy as a broad family of driven-state interrogation methods rather than a single two-level textbook protocol. Its modern forms include motional sideband spectroscopy, Floquet-band spectroscopy, rotating-frame noise spectroscopy, collision-aware clock interrogation, multimode transient spectral reconstruction, and correlated-spin metrology (Liu et al., 2022, Matityahu et al., 2017, Yu et al., 21 May 2026, Shaniv et al., 2017). This suggests that the central concept of Rabi spectroscopy is not merely coherent population transfer, but the use of controlled coherent driving to convert otherwise inaccessible Hamiltonian, dissipative, or interaction parameters into spectroscopically resolvable structure.