---
title: Adiabatic Ramsey Interferometry
url: https://www.emergentmind.com/topics/adiabatic-ramsey-interferometry
type: topic
---

# Adiabatic Ramsey Interferometry

Searching arXiv for recent and foundational papers on adiabatic Ramsey interferometry and closely related implementations.
Adiabatic Ramsey interferometry is a family of interferometric protocols in which the state preparation and readout operations of Ramsey spectroscopy are implemented through adiabatic control rather than solely through short non-adiabatic resonant pulses. Across current realizations, the adiabatic element may be RF dressing of trapped atoms, amplitude-shaped resonant spin rotations for pulsed neutron beams, half-STIRAP operations in a $\Lambda$ system, geometric transport within a degenerate dark-state manifold, or adiabatic ramps of a spin-boson Hamiltonian. The common structure is the preservation of coherent phase encoding between two effective interferometric operations, with the adiabatic control chosen to reduce sensitivity to pulse-area errors, detuning drift, phase-space broadening, or differential trapping shifts, and in some cases to convert weak symmetry-breaking perturbations directly into a spin-population bias [1907.01775] [2603.24049] [2407.16057] [2309.10192] [2603.29574].

## 1. Conceptual relation to conventional Ramsey interferometry

In conventional Ramsey interferometry, two coherent rotations, nominally $\pi/2$ pulses, are separated by a free-evolution interval. For ideal instantaneous pulses in a spin-$1/2$ implementation, the outgoing polarisation can be written as
$$
P(\Delta)=-\cos(\Delta T),
$$
while realistic experiments commonly parameterize the fringe as
$$
P(\Delta)=C\cos(\Delta T+\phi_0)+P_{\text{off}},
$$
with contrast $C$, residual phase $\phi_0$, and offset $P_{\text{off}}$ [2603.24049]. In two-level notation, the same structure is often written as $S(T)\propto \cos(\Delta T+\phi_0)$ [2407.16057].

Adiabatic Ramsey interferometry modifies the means by which the effective beam splitters and recombiners are realized. In RF-dressed shell interferometry, the Ramsey sequence is implemented by microwave coupling between two adiabatically dressed trapped states whose shell potentials are independently tuned to match geometry and curvature [1907.01775]. In pulsed neutron interferometry, the nominal $\pi/2$ pulses are replaced by resonant, circularly polarized RF fields with time-dependent amplitude envelopes chosen to equalize the pulse area seen by different velocities, thereby compensating the source phase-space distribution without monochromatization [2603.24049]. In NV-diamond nuclear-spin sensing, half-STIRAP pulses act as adiabatic $\pi/2$ equivalents that prepare and project a coherent superposition while suppressing intermediate-state occupation [2407.16057]. In the tripod implementation with ultracold strontium, the pulse operations are geometric SU(2) rotations in a two-dimensional dark-state manifold, and the free-evolution phase is generated by an effective geometric scalar term [2309.10192]. In the trapped-ion proposal, the protocol is generalized further: adiabatic ramps of a quantum Rabi Hamiltonian play the role of Ramsey operations, and the measured signal is a final spin-population bias,
$$
\langle \sigma_z(t_f)\rangle=\tanh(\pi\alpha\tau),
$$
rather than a sinusoidal fringe [2603.29574].

These variants show that “adiabatic Ramsey interferometry” is not a single pulse prescription but a broader interferometric design principle. What is preserved is coherent phase encoding between well-defined initial and final manipulations; what changes is the control manifold used to implement them.

## 2. Adiabatic control mechanisms and effective Hamiltonians

A defining feature of the subject is that adiabaticity is achieved in several technically distinct ways.

For RF-dressed shell interferometry in $^{87}\mathrm{Rb}$, adiabatic potentials arise from coupling Zeeman sublevels in a static magnetic quadrupole field with strong RF fields. The standard dressed potential is
$$
U_{m_F}(\mathbf r)=m_F\hbar\sqrt{\Omega_{\rm RF}^2(\mathbf r)+\delta^2(\mathbf r)}+Mgz,
$$
with detuning
$$
\delta(\mathbf r)=\omega_{\rm RF}-\omega_L(\mathbf r),\qquad
\omega_L(\mathbf r)=\frac{|g_F|\mu_B B(\mathbf r)}{\hbar}.
$$
With two RF tones of opposite circular polarization, the $|1,-1\rangle$ and $|2,+1\rangle$ states are dressed independently, permitting shell matching in both location and local curvature [1907.01775].

For neutron beams, the control remains resonant and fixed in frequency, but the RF envelope is shaped in time. The spin dynamics are propagated by the Bloch equation
$$
\frac{\mathrm d\mathbf S(t)}{\mathrm dt}=\gamma\,\mathbf S(t)\times \mathbf B(\mathbf r(t),t),
$$
and the adiabaticity diagnostic used in RamseyProp is
$$
\eta(t)=\frac{\big|\mathrm d\hat{\mathbf B}/\mathrm dt\big|}{|\gamma|\,|\mathbf B|},\qquad
\hat{\mathbf B}=\frac{\mathbf B}{|\mathbf B|}.
$$
The condition $\eta(t)\ll 1$ characterizes adiabatic following of the local field direction. Importantly, this implementation explicitly distinguishes itself from adiabatic rapid passage via frequency chirps; its control variable is the amplitude envelope $f(t)$ of a resonant circularly polarized RF field [2603.24049].

In the NV-center implementation, the relevant structure is a $\Lambda$ system formed by the $^{14}\mathrm N$ nuclear-spin sublevels $|1\rangle\equiv |m_I=+1\rangle$, $|2\rangle\equiv |m_I=0\rangle$, and $|3\rangle\equiv |m_I=-1\rangle$. The effective interaction-picture Hamiltonian is
$$
H(t)=\frac{\hbar}{2}
\begin{pmatrix}
0 & \Omega_p(t) & 0\\
\Omega_p(t) & -2\Delta & \Omega_s(t)\\
0 & \Omega_s(t) & 0
\end{pmatrix}.
$$
Its dark state,
$$
|\mathcal D(t)\rangle=\cos\theta\,|1\rangle-\sin\theta\,|3\rangle,\qquad
\tan\theta(t)=\frac{\Omega_p(t)}{\Omega_s(t)},
$$
has no $|2\rangle$ component. Half-STIRAP stops the dark-state mixing at $\theta=\pi/4$, thereby creating an equal superposition that functions as an adiabatic analogue of a $\pi/2$ pulse [2407.16057].

In the geometric tripod scheme, three ground states are resonantly coupled to a common excited state, yielding a two-dimensional degenerate dark manifold. Adiabatic evolution inside that manifold is governed by the Wilczek–Zee connection
$$
A_\mu^{ij}(\boldsymbol\lambda)=i\langle D_i(\boldsymbol\lambda)|\partial_\mu D_j(\boldsymbol\lambda)\rangle,
$$
and the projected effective Hamiltonian takes the form
$$
\hat H_{\rm eff}=\frac{\hat{\mathbf p}^2}{2m}\otimes\mathds 1-\frac{\hat{\mathbf A}\cdot \hat{\mathbf p}}{m}+\hat Q+\hat w.
$$
Here $\hat w$ generates geometric beam splitting and recombination, while $\hat Q$ is the geometric scalar potential responsible for the interferometric phase during free evolution [2309.10192].

In the trapped-ion proposal, the probe is a quantum Rabi model with squeezing,
$$
\hat H(t)=\hbar\omega\,\hat a^\dagger\hat a+\frac{\hbar\Omega(t)}{2}\sigma_x+\hbar g\,\sigma_z(\hat a+\hat a^\dagger)+\hbar\xi\left(\hat a^{\dagger 2}e^{i\theta}+\hat a^2e^{-i\theta}\right),
$$
with parity symmetry
$$
\Pi=\sigma_z e^{i\pi\hat a^\dagger\hat a},\qquad [\hat H(t),\Pi]=0.
$$
An adiabatic ramp $\Omega(t)=\Omega_0 e^{-t/\tau}$ maps a unique paramagnetic ground state into a symmetry-protected cat-like manifold that becomes sensitive to weak parity-breaking perturbations [2603.29574].

Taken together, these formulations show that adiabatic Ramsey interferometry is better understood as an interferometric use of adiabatic following than as a single Hamiltonian model.

## 3. Platform-specific realizations

The present literature spans trapped neutral atoms, neutron beams, solid-state nuclear spins, ultracold atoms in synthetic gauge fields, and trapped ions.

| Platform | Adiabatic element | Principal readout |
|---|---|---|
| $^{87}\mathrm{Rb}$ bi-chromatic shell trap | Independently controlled RF-dressed shell potentials for $|1,-1\rangle$ and $|2,+1\rangle$ | Microwave Ramsey phase and spectroscopy |
| Neutron beam at ESS design study | Time-dependent amplitude modulation of resonant circularly polarized RF fields | Polarisation fringe near steepest slope |
| $^{14}\mathrm N$ in NV diamond | Half-STIRAP in a $\Lambda$ system | Optical population readout with phase-cycled I/Q extraction |
| $^{87}\mathrm{Sr}$ tripod scheme | Geometric adiabatic evolution in a dark-state subspace | Output-port populations after time of flight |
| Trapped-ion quantum Rabi probe | Adiabatic ramp into parity-sensitive cat manifold | Spin-state probabilities only |

In the bi-chromatic shell implementation, two RF tones at $\omega_1$ and $\omega_2$ with opposite circular polarizations separately dress the two hyperfine manifolds, enabling matched shell geometry and matched axial curvature. The experiment loads up to $2\times 10^5$ condensed atoms into the shell from a crossed-beam optical dipole trap, with condensate lifetime $\approx 7\,\mathrm s$ and thermal-cloud lifetime up to $80\,\mathrm s$; measured trap frequencies are $\omega_z/2\pi=50\text{–}300\,\mathrm{Hz}$ and $\omega_\rho/2\pi=8\text{–}13\,\mathrm{Hz}$ [1907.01775].

The neutron-beam framework combines McStas for phase-space generation, COMSOL field maps, and the RamseyProp spin-dynamics code. In the ESS case study, the static field is $B_0=100\,\mu\mathrm T$, the RF amplitude is $B_1=34.4\,\mu\mathrm T$, and a two-coil configuration uses a $7.7\,\mathrm m$ free-precession region. The source time structure at $14\,\mathrm{Hz}$ is exploited through an inverse-time envelope $f(t)\propto 1/t$, periodically reset and shifted by $t_0$ to compensate the extended moderator pulse [2603.24049].

The NV-center realization uses a $^{12}\mathrm C$-enriched [110]-cut diamond with initial nitrogen $\approx 13\,\mathrm{ppm}$ and NV $\approx 4\,\mathrm{ppm}$, biased near the ESLAC at $B\simeq 480\,\mathrm G$. The single-quantum nuclear transitions occur at $f_1\approx 5.089\,\mathrm{MHz}$ and $f_2\approx 4.797\,\mathrm{MHz}$, with measured $\langle\Omega_e\rangle/2\pi\approx 36.1\,\mathrm{kHz}$ and H-STIRAP pulse duration $t_p\approx 500\,\mu\mathrm s$ [2407.16057].

The geometric tripod interferometer uses the $689\,\mathrm{nm}$ intercombination line of $^{87}\mathrm{Sr}$, a bias field of $67\,\mathrm G$, and three resonant Gaussian laser pulses with $|\Omega_{01}|=|\Omega_{02}|=2|\Omega_{03}|\approx 2\pi\times260\,\mathrm{kHz}$, giving total $\Omega\approx 2\pi\times410\,\mathrm{kHz}$. State-dependent momentum transfer provides distinct output ports after $9\,\mathrm{ms}$ time of flight [2309.10192].

The trapped-ion proposal is explicitly aimed at weak nonlinearities arising from trapping-potential anharmonicity or higher-order Coulomb corrections. Representative simulation parameters are $\omega/2\pi\approx 5\,\mathrm{kHz}$, $g/2\pi\approx 4\text{–}9\,\mathrm{kHz}$, $\Omega_0/2\pi\approx 100\text{–}150\,\mathrm{kHz}$, $\tau\approx 3.0\text{–}3.5\,\mathrm{ms}$, and $\xi/2\pi\approx 0\text{–}1\,\mathrm{kHz}$ [2603.29574].

This cross-platform spread indicates that the unifying principle is operational rather than hardware-specific: adiabatic state engineering is used to improve or qualitatively alter Ramsey interferometric encoding.

## 4. Phase accumulation, observables, and signal models

Despite the diversity of implementations, the interferometric observable is always an accumulated relative phase or an equivalent population bias generated by a weak perturbation.

In the bi-chromatic shell interferometer, the microwave-coupled dressed states obey
$$
H_\mu/\hbar=\frac{\Delta_{\mu,\rm eff}(t)}{2}\sigma_z+\frac{\Omega_\mu}{2}\sigma_x,
$$
with effective detuning
$$
\Delta_{\mu,\rm eff}(t)=\omega_{\rm MW}-\omega_{\rm hfs}-\frac{1}{\hbar}\left[U^{(2,+1)}(\mathbf r,t)-U^{(1,-1)}(\mathbf r,t)\right].
$$
A standard Ramsey sequence then yields
$$
\phi=\int_0^T\Delta_{\mu,\rm eff}(t)\,dt+\phi_{\rm pulses}.
$$
Because the low horizontal confinement lets atoms spread into a quasi-2D sheet, spatially varying differential energies map directly into an imageable phase field,
$$
\phi(\mathbf r)=\frac{1}{\hbar}\int_0^T\Delta E(\mathbf r,t)\,dt.
$$
This is the basis for proposed imaging of DC, AC, RF, or microwave magnetic and electric fields, and even local gravity variations [1907.01775].

In the neutron-beam framework, the free-precession phase is
$$
\phi=\gamma\int B_z(t)\,\mathrm dt=\gamma\int \frac{B_z(z)}{v(z)}\,\mathrm dz,
$$
and the working-point sensitivity is obtained from the local linearization
$$
P(\Delta)\approx a+b(\Delta-\Delta_0),\qquad b=\mp CT.
$$
The uncertainty per fitted phase measurement is
$$
\sigma_\Delta=\frac{\sigma_a}{|b|}.
$$
This formalism makes contrast, flight time, and detected counts the primary sensitivity levers [2603.24049].

In STIRAP-based Ramsey with NV centers, the free evolution of the adiabatically prepared superposition leads to
$$
S(\tau)=S_0+C\cos(\delta\tau+\phi_0).
$$
The implementation uses four-phase cycling of the final pulse set to construct in-phase and quadrature observables,
$$
I=R_1-R_2+R_3-R_4,\qquad Q=R_5-R_6+R_7-R_8,
$$
and then
$$
z=I+iQ=r\,e^{i\Phi},
$$
which directly yields accumulated phase $\Phi$ and amplitude $r$ while suppressing single-quantum artifacts [2407.16057].

In the geometric tripod scheme, free evolution occurs inside the dark manifold under the geometric scalar term. In the asymptotic regime, the phase is
$$
\phi_{\rm geom}(T)=\frac{|Q_{11}-Q_{22}|}{\hbar}T=\frac{2p_r^2}{m\hbar}T.
$$
For balanced splitting, the two observable output ports obey
$$
P_{g_3}(T)\approx \cos^2\!\left(\frac{\phi_{\rm geom}(T)+\Phi}{2}\right),\qquad
P_{g_2}(T)\approx \sin^2\!\left(\frac{\phi_{\rm geom}(T)+\Phi}{2}\right),
$$
where $\Phi$ is an externally imposed phase jump [2309.10192].

In the trapped-ion scheme, the detected quantity is not a sinusoid but the final spin bias of an effective two-level model in the cat manifold,
$$
\langle \sigma_z(t_f)\rangle=\tanh(\pi\alpha\tau),\qquad
P_\uparrow=\frac{1+\langle \sigma_z\rangle}{2}.
$$
The corresponding classical Fisher information is
$$
F_C(\lambda)=(\pi\tau\,\partial_\lambda\alpha)^2\,\operatorname{sech}^2(\pi\alpha\tau),
$$
so weak symmetry-breaking nonlinearities are estimated from spin readout alone [2603.29574].

A central consequence is that adiabatic Ramsey interferometry should not be identified exclusively with cosine fringes. Depending on the adiabatic encoding mechanism, the observable can be a spatial phase map, a geometric port population, a phase-cycled quadrature signal, or a monotonic spin imbalance.

## 5. Adiabaticity, coherence, and dominant error channels

The performance of adiabatic Ramsey interferometers is governed by how well the system follows the intended dressed or dark manifold and by how effectively technical inhomogeneities are converted from dominant to subleading errors.

For RF-dressed shell traps, adiabatic following requires
$$
|\dot\delta|\ll \Omega_{\rm RF}^2,
$$
with Landau–Zener loss
$$
P_{\rm LZ}\simeq \exp\!\left(-\frac{\pi\Omega_{\rm RF}^2}{2|\dot\delta|}\right).
$$
The experimental loading sequence uses long ramps, approximately $400\,\mathrm{ms}$ for bias overlap and approximately $200\,\mathrm{ms}$ for final transfer, with $\Omega_{\rm RF}/2\pi$ typically $100\text{–}370\,\mathrm{kHz}$, keeping non-adiabatic loss negligible. Spectroscopically, the mono-chromatic shell gave broad lines of $\Delta\nu\approx1000\,\mathrm{Hz}$ and no Rabi oscillations, whereas the bi-chromatic shell reduced the linewidth to $\Delta\nu\approx100\,\mathrm{Hz}$ for $10\,\mathrm{ms}$ microwave pulses and produced clear Rabi oscillations with $\Omega_{\rm MW}/2\pi=146\pm1\,\mathrm{Hz}$ and decay time $\tau=18\pm2\,\mathrm{ms}$. The text further states that radial mismatch can be made $\lesssim10\,\mathrm{mHz}$, implying dephasing times on the order of $10\,\mathrm s$ when only the lowest radial mode is occupied [1907.01775].

In the neutron case, the adiabaticity metric $\eta(t)$ is used as a trajectory-resolved diagnostic rather than as a global theorem. The practical target is narrowing the flip-angle distribution $\theta$ around $\pi/2$. For the full HIBEAM E5 spectrum into a $0.3\,\mathrm m$ coil at $16\,\mathrm m$, the baseline flip-angle spread is $\sigma_\theta=0.67\,\mathrm{rad}$; time-dependent amplitude modulation reduces this to $0.17\,\mathrm{rad}$, and a $0.3\,\mathrm{ms}$ timing window centered at $7.08\,\mathrm{ms}$ reduces it further to $0.02\,\mathrm{rad}$. In the two-coil Ramsey study, the corresponding detuning uncertainty improves from $\sigma_\Delta\approx0.6\,\mathrm{Hz}$ to $\approx0.14\,\mathrm{Hz}$ with modulation and to $\approx0.013\,\mathrm{Hz}$ with timing restrictions [2603.24049].

For half-STIRAP in NV centers, the adiabatic criteria are
$$
|\dot\theta(t)|\ll \Omega_e(t),\qquad
|\dot\xi(t)|\ll \sqrt{\Omega_e^2(t)+\Delta^2}.
$$
Experimentally, STIRAP 4-Ramsey shows wide stable regions of accumulated phase over detuning $\Delta$ and effective Rabi frequency $\Omega_e$, in contrast to the narrow operating stripe of conventional DQ 4-Ramsey. Under typical conditions, the detuning sensitivity of the conventional scheme is $\approx 2.1\,\mathrm{deg/kHz}$, while STIRAP 4-Ramsey reduces this sensitivity by about a factor of $5$ and can be tuned to null the first-order detuning dependence. Individual STIRAP traces also show approximately $30\%$ larger amplitude and approximately $20\times$ smaller ripple than DQ 4-Ramsey [2407.16057].

In the geometric tripod realization, adiabatic following requires the nonadiabatic generator to remain small compared with the bright-state gap, summarized in the text as $\|\hat w\|\ll \hbar\Omega(t)$. A pulse-width window of $3\,\mu\mathrm s\lesssim \sigma_t\lesssim 15\,\mu\mathrm s$ gave time-independent geometric splitting; below that range adiabaticity breaks, while above it the neglected kinetic and spin-orbit terms become important. At the chosen $\sigma_t=2.5\,\mu\mathrm s$, the fraction of atoms remaining in the dark subspace averages $\approx97(2)\%$. The measured interferometric frequency is $2\pi\times19.8(16)\,\mathrm{kHz}$, and thermal dephasing at $T_0=50(3)\,\mathrm{nK}$ yields a damping time $\tau=23(4)\,\mu\mathrm s$ [2309.10192].

In the trapped-ion proposal, the adiabatic condition is expressed through the gap to the excited manifold, with $\tau\gg1/\tilde\omega$ and, in the effective two-level picture, $|\dot\Omega(t)|\ll \Delta_0(t)^2$. Weak pure dephasing,
$$
\partial_t\hat\rho=-i[\hat H(t),\hat\rho]+\frac{\Gamma}{2}\left(\sigma_z\hat\rho\sigma_z-\hat\rho\right),
$$
attenuates the spin signal approximately as $e^{-\Gamma T}\tanh(\pi\alpha\tau)$, but the work states that super-Heisenberg scaling with mean phonon number persists for small $\Gamma$ [2603.29574].

A common thread is that adiabatic protocols do not eliminate technical requirements; they redistribute them. Pulse-area sensitivity can be reduced, but stability of RF amplitudes, polarization purity, dark-state isolation, or ramp duration becomes correspondingly decisive.

## 6. Metrological roles, comparative advantages, and limitations

The documented applications span imaging sensors, beam interferometers, solid-state frequency references, synthetic-gauge interferometers, and nonlinear quantum metrology.

The bi-chromatic shell interferometer is positioned as a fully trapped matter-wave interferometer with low horizontal confinement, 2D sheet formation, insensitivity to homogeneous DC magnetic fields, and direct spatial imaging capability for magnetic, electric, RF, or microwave near fields, as well as local gravity variations. Its measured sensitivity to homogeneous $B_z$ is near zero, with slope $(2\pm10)\,\mathrm{Hz/G}$ up to $3.5\,\mathrm G$ [1907.01775].

The neutron-beam framework is explicitly directed toward experiment design at the European Spallation Source, including searches for axion-like particles. In the ESS geometry studied, time-dependent amplitude modulation improves phase sensitivity by a factor of approximately $4$ for a $10\,\mathrm m$ setup starting $15\,\mathrm m$ from the moderator, and timing restrictions improve it by approximately $44\times$ relative to baseline, at reduced throughput [2603.24049].

In NV diamond, the principal metrological motivation is long-term stability in gyroscopes and frequency standards. The reported reduction in phase sensitivity to detuning drift, together with suppression of intermediate-state occupation and cleaner fringes, directly addresses the bias-stability limitations of ensemble nuclear-spin sensing under RF gradients and temperature drift [2407.16057].

In the tripod scheme, the significance lies in replacing dynamical pulse-area control by geometric SU(2) operations in a non-Abelian dark manifold. The work points to multiport interferometry, atomtronic switching, recoil measurements, and inertial sensing as natural extensions, while showing that the free-evolution phase can arise from a geometric scalar term that remains operative when the lasers are nominally off [2309.10192].

In the trapped-ion proposal, the application is the estimation of weak nonlinear couplings such as cubic anharmonicity or higher-order Coulomb processes. For odd-$k$ nonlinearities, the scaling
$$
\alpha\approx 2 f_k \bar n^{k/2}\quad \Rightarrow\quad \Delta f_k\propto \bar n^{-k/2}
$$
yields super-Heisenberg precision in mean phonon number. For $k=3$, the paper gives $\Delta f_3\propto \bar n^{-3/2}$, and it emphasizes that this scaling is accessible with spin readout only and even for an initial thermal motional state [2603.29574].

Several clarifications follow from this literature. First, adiabatic Ramsey interferometry is not synonymous with adiabatic rapid passage; one implementation explicitly uses fixed-frequency resonant excitation with amplitude shaping instead of chirps [2603.24049]. Second, the method is not tied to two bare states. It may operate in dressed internal states, in a $\Lambda$ system, in a degenerate dark-state manifold, or in a parity-protected spin-boson ground-state multiplet. Third, robustness is conditional rather than absolute: each platform trades conventional $\pi/2$-pulse sensitivity for a different stability budget, such as RF amplitude matching at the $\lesssim10^{-5}$ level for long shell-trap coherence, sufficient pulse overlap and area for STIRAP adiabaticity, or precise control of ramp times and gap structure in trapped ions [1907.01775] [2407.16057] [2603.29574].

Taken as a whole, the field defines adiabatic Ramsey interferometry as a technically heterogeneous but conceptually coherent generalization of Ramsey spectroscopy. Its central innovation is to embed interferometric preparation, accumulation, and readout within adiabatically controlled dressed-state dynamics, thereby extending Ramsey methods to regimes where conventional square-pulse control is limited by inhomogeneity, phase-space dispersion, differential trapping, geometric gauge structure, or nonlinear symmetry-breaking physics.

Source: https://www.emergentmind.com/topics/adiabatic-ramsey-interferometry