WorksPhysics and Astronomy
Aharonov-Bohm blockade of complete reversal in coupled Rabi cavities
Abstract
When a common bias reverses two bistable fields, the system can pass through either of two single-reversal configurations on its way to the favored state. We show that interference between these two orders of reversal can block the final step. In a dimer of coupled quantum Rabi cavities, the four field configurations form a loop, and a complex photon-hopping amplitude threads it with an Aharonov-Bohm flux. Near half a flux quantum, the two paths to complete reversal interfere destructively: in microscopic simulations, a slow uniform sweep reverses one field with probability 0.998, although the final bias favors reversing both. A four-state model calibrated at zero bias predicts how the window of this Aharonov-Bohm blockade narrows with sweep time. The blocked state remains coherent. A differential pulse followed by a return sweep converts the relative phase of the two single-reversal configurations into a fringe in the joint field signs, with contrast 0.95. In a trimer, a phase gradient imprinted between sweeps determines whether the next sweep adds or removes a reversed field: reversing the second of two gradients changes the most probable outcome from three reversed fields to one.
Introduction
A slow bias sweep reverses a bistable system by carrying its ground state from one minimum to the other. The scenario is ubiquitous, from magnets in a field ramp to qubits swept through an avoided crossing [1, 2, 3, 4, 5], and sweeps of many coupled bistable units underlie quantum annealing [6, 7], including proposals based on networks of nonlinear oscillators [8, 9]. Even two identical units driven by a common bias pose a question that a single unit cannot. The final bias favors reversing both, but either unit can reverse first, and the two intermediate configurations have the same bias energy. The system thus has two routes to the favored configuration. Can interference between these two orders of reversal prevent the system from reaching it?
Quantum interference can suppress transitions that energetics alone would allow. A geometric phase can quench tunneling between opposite spin orientations [10, 11, 12, 13], path interference shapes transition probabilities in multistate Landau–Zener models [14, 15, 16], and in a -magnet a suitable sweep rate can select one spin to flip while suppressing flips of the other [17]. In real space, the same principle confines a particle to an Aharonov–Bohm cage when a lattice plaquette encloses half a flux quantum [18, 19], and caging also occurs in the Fock space of two photons [20]. In the problem posed above, the interfering paths are of a different kind. They are the two orders in which distinct cavity fields reverse, and their relative phase is set by the hopping amplitude on a single physical bond.
Coupled quantum Rabi cavities supply both ingredients that this question requires: bistable fields and a controllable hopping phase. The quantum Rabi model [21, 22, 23] undergoes a superradiant transition in the limit of an infinite qubit-to-cavity frequency ratio [24], the single-mode counterpart of the Dicke transition [25, 26]. At large but finite ratios, its lowest doublet defines two states localized at opposite field displacements, which play the role of the two minima of a bistable field. Trapped ions have been used to simulate the Rabi model [27] and, following a theoretical proposal [28], to probe its superradiant transition [29], and coupled Rabi sites support collective and chiral phases [30, 31, 32, 33, 34, 35]. Complex hopping amplitudes, for their part, can be engineered through artificial gauge fields for neutral atoms [36] and photons [37, 38, 39], and trapped-ion chains support phonon hopping [40] and spin couplings with tunable phases [41, 42]. We ask what such a phase does to the reversal of two bistable fields under a uniform bias sweep.
We show that the four field configurations of a Rabi dimer form a loop in a synthetic dimension [43], threaded by an Aharonov–Bohm flux [44] that a single complex hopping amplitude controls. Near half a flux quantum, destructive interference between the two orders of reversal blocks complete reversal, even when a slow sweep reverses one field with near-unit probability. We refer to this effect as Aharonov–Bohm blockade. A four-state model, calibrated from zero-bias energies and matrix elements of the bias operator, predicts how the window of the blockade narrows with sweep time. The blockade does not erase phase information: a local pulse followed by a return sweep reveals the coherence between the two single-reversal configurations. In a trimer, the same kind of phase becomes a means of control, and the direction of a local phase gradient determines whether the next sweep adds or removes a reversed field.
Section II introduces the model and the sweep protocol and derives the configuration ring from a symmetry of the zero-bias Hamiltonian. Section III establishes the blockade in the dimer, tests its coherence, and turns to stepwise reversal in a trimer. Section IV summarizes the results, and Appendixes A–C collect the microscopic calibration, the finite-sweep theory, and the pulse protocols.
Model and sweep protocol
We consider two quantum Rabi cavities coupled by photon hopping with a complex amplitude,
where annihilates a photon in cavity , , are Pauli matrices of the two-level system in cavity , is the ratio of the qubit and cavity frequencies, and is the dimensionless light–matter coupling. We set and the cavity frequency to unity, and use , , and unless stated otherwise. Only the uniform bias varies during a sweep. The bond phase is held fixed and, as shown below, controls the flux through the configuration loop. In the trimer, the hopping is oriented cyclically around the triangle.
An isolated cavity becomes superradiant for as [24], and its field is then bistable. Eliminating the two-level system adiabatically yields a double-well potential for the field quadrature , with a barrier of height 1.41 for our parameters; tunneling between the two wells is suppressed exponentially in (Appendix § 1.2.1). We denote the dressed states localized in the two wells of a cavity by and the four configurations of the dimer by , , , and , the first sign referring to cavity 1 (Figure 1(a)). The bias shifts each well in proportion to its mean field, so the configuration with signs acquires the bias energy . The projected bias coupling is approximately the magnitude of the field displacement in a localized state; at the first zero-bias ground-state crossing, , we find . Reversing one field therefore releases the energy , and a positive bias favors . The single-reversal configurations and , by contrast, have the same bias energy throughout the sweep, which keeps the two orders of reversal on an equal footing.
Figure 1. Interference blocks complete reversal near half a flux quantum. (a) Two Rabi cavities with bistable fields, coupled by photon hopping , and the four configurations in the plane of the two fields. Arrows mark single flips with amplitude , and dotted diagonals mark direct double flips with amplitude . (b) Magnetization during the sweep at and for (solid, microscopic model; dashed, two independent cavities at and the block at , § 1.2 A). (c) Levels of the two blocks of Eq. (7) versus the bias energy; the shaded curve is the path of a sweep that starts in , and dotted lines are the configuration energies. (d) Final (symbols, microscopic model; line, Eq. (11) with the first-order refinement of Appendix § 1.2.2) and microscopic (gray) versus for ; the top axis gives the flux. All panels use , , , and .
The dimer is prepared in its ground state at , which lies close to , and the bias is swept linearly to , , with unless stated otherwise. At we project the state onto four orthonormal dressed configurations constructed from the instantaneous eigenstates (Appendix § 1.2.1). This projection yields the probability that exactly fields have reversed and the magnetization . A second readout bypasses the dressed configurations altogether: the probability that both fields are negative at the end of the protocol requires no effective model and corresponds directly to a measurement of the two field quadratures.
Tunneling couples the four configurations, and the form of this coupling at zero bias follows from a symmetry of Eq. (1). At zero bias and , Eq. (1) commutes with , where is the parity of cavity 1 and exchanges the two cavities; a finite uniform bias or a real hopping component breaks this symmetry. The operator rotates the plane of the two fields by and permutes the configurations cyclically, , so the zero-bias Hamiltonian restricted to these configurations is a circulant matrix. Adding the leading diagonal bias gives the ring Hamiltonian
where sets the single-flip magnitude, is the signed direct double-flip matrix element, and is the flux through the loop (Figure 1(a)). The bias term is the bias energy introduced above, measured from the single-reversal configurations. At zero bias the ring energies, measured from their mean and labeled by the eigenvalues of , are
Matching Eq. (3) to the lowest zero-bias quartet of Eq. (1) determines , , and , and the matrix elements of within the same quartet determine (Appendix § 1.2.1). All parameters of the ring are thus fixed at zero bias. At the hopping strength used below, for example, and ; Table I lists the parameters at the four reference points of the dimer.
The flux is more than a calibration parameter: it has a direct geometric meaning. In terms of the quadratures and their conjugate momenta , the hopping term contains , the generator of rotations in the plane of the two fields. Completing the square in the photon kinetic energy yields the perpendicular synthetic magnetic field
The flux through the loop can then be estimated as , where is the area enclosed by the four dressed-state centers. At this gives and reproduces the spectral flux to (Appendix § 1.2.1). Increasing from to raises through , , and [top axis of Figure 1(d)], so the hopping strength alone tunes the interference between the two orders of reversal.
Interference and control of field reversal
Blockade of complete reversal
Figure 1(b) contrasts two sweeps of duration that differ only in the hopping strength. At , where , the sweep follows the bias: both fields reverse with probability , and the magnetization swings from nearly to . At , where the two lowest zero-bias levels cross close to , the same sweep stops halfway. The final magnetization is only ; one field reverses with probability , whereas . The direct readout gives , a larger but still small value, because the dressed single-reversal configurations are not confined to a single quadrant of the field plane. Neither energetics nor frozen dynamics can account for this outcome. At the end of the sweep lies below the single-reversal configurations, more than 30 times , and the first reversal occurs with near certainty.
The explanation lies in the two orders of reversal [Figure 1(a)]. Starting from , a single flip does not select one cavity; it produces the single-reversal state
with amplitude , where . The final configuration , in turn, is reached by a single flip only from
Whether the second step can proceed therefore depends on the overlap
which is proportional to the sum of the amplitudes of the two paths and , carrying the phases . At a gauge transformation makes all single-flip amplitudes real and equal, and for Eq. (2) describes two independent cavities. At the two paths cancel and the two states are orthogonal; for , Eq. (2) then separates into the two blocks
in the bases and [Figure 1(c)]. The first flip from remains allowed, but the state it reaches has no matrix element to , and the second flip is blocked. This is the signature of an Aharonov–Bohm cage [[18, 19], with configurations in place of lattice sites and the two orders of reversal in place of the two paths around a plaquette, and it is the origin of the Aharonov–Bohm blockade.
Both cases reduce to Landau–Zener problems, which makes the contrast in Figure 1(b) quantitative. We measure the sweep rate by and define the adiabaticity parameter , which equals at for . At and , each field undergoes its own passage with coupling , and an infinite sweep reverses both with probability
which equals 0.974 at . The exact solution of the same two-level problems over the finite sweep gives 0.978 and follows the microscopic magnetization throughout the sweep [dashed line in Figure 1(b)]. At and , the sweep starts in and never leaves the block , whose lower level connects to [shaded curve in Figure 1(c)]. The infinite-sweep result is then
At the finite positive endpoint, the main correction is an admixture of , with ; at this gives , close to the microscopic .
The microscopic model does not meet the two ideal conditions and at the same hopping strength. At , where , the direct amplitude is , and the two blocks remain coupled. At zero bias their coupling equals half the signed splitting of the two lowest levels,
and it vanishes at the crossing of these levels, where the flux detuning compensates . The ideal cancellation therefore occurs not at half flux but slightly beyond it. As the bias passes through zero, the lower levels of the two blocks cross with relative slope , and the Landau–Zener formula gives the population transferred to the block that ends in ,
Evaluated with taken from the zero-bias spectrum at each , and with the bias dependence of the coupling included to first order (Appendix § 1.2.2), this expression reproduces the microscopic across the windows near and near the corresponding crossing close to , as well as between them, with a median relative deviation of where [Figure 1(d)].
The blockade requires an intermediate sweep rate. The sweep must be slow enough for the first reversal, , yet fast enough to suppress transfer between the blocks, . At fixed away from the crossing, a longer sweep eventually permits the second reversal, so the window in narrows as , consistent with the narrowing seen in the microscopic sweeps (Appendix § 1.2.2). Even at , where the zero-bias coupling vanishes, a small residual remains, because the sweep spans a finite range of bias. An amplitude acquired near the crossing can interfere with the admixture of at the endpoint; the estimate in Appendix § 1.2.2 accounts for the scale of this residual and explains why the smallest need not occur exactly at .
Coherence of the single-reversal state
Populations alone cannot distinguish the blockade from a sweep that reverses one field at random, since an incoherent mixture of and would yield the same . In the ideal ring, however, the sweep prepares , an equal-weight coherent superposition of two configurations with the same net displacement, whose field
part has the structure of an entangled coherent state [45, 46, 47]. For the single sweep in Figure 1(b), the microscopic populations are and , giving . It remains to show that the relative phase between them survives the sweep and can be read out.
To probe this phase, we extend the protocol by a pulse and a return sweep [Figure 2(a)]. At we sweep from to , apply opposite local biases for , and sweep back; each sweep has , , and . The pulses shift and in opposite directions and imprint a relative phase between them, which we characterize by the nominal phase , the value for uncoupled cavities with bias coupling (Appendix § 1.2.3); the ring gives the imprinted phase . At ideal half flux, this phase rotates into . The return sweep then traverses each block backward: returns to , whereas is carried on to . The result is the ideal adiabatic fringe
Figure 2. The single-reversal state is a coherent superposition: a differential pulse and a return sweep turn it into complete reversal with a probability set by the pulse phase. (a) Protocol: a sweep to , opposite bias pulses with nominal phase (Appendix § 1.2.3), and a sweep back to ; each sweep lasts , the pulses last , and . (b) Final from the dressed configurations (filled circles) and from the joint signs of the two fields (crosses). Open circles and the flat row of crosses show both readouts after dephasing of the configurations before the pulse. The black line gives the finite-sweep estimate of Eq. (21), and the gray horizontal line marks half its maximum.
The return sweep thus converts a relative phase into a population, as in Landau–Zener–Stückelberg interferometry [5, 48]. At a finite sweep rate, two imperfections shape the fringe. Incomplete block passages lower its maximum to , and the direct double flips, which mix and at finite bias, give it a nonzero minimum of at most 0.033 (Appendix § 1.2.3). The microscopic fringe, which ranges from 0.0276 to 0.9812, is consistent with both estimates (Figure 2(b)).
The same fringe appears in the field signs themselves, without reference to the dressed configurations. Evaluated at 25 pulse phases, has a contrast, defined as the difference between its maximum and minimum, of 0.951, close to the dressed-configuration contrast of 0.9536. Dephasing the configurations before the pulse removes the fringe: the field-sign probability stays near half the coherent maximum, and its contrast drops to 0.006 (Figure 2(b)). A local phase can thus steer the same return sweep either back to or on to , and the loss of contrast when the two reversal paths are distinguished is the expected signature of which-path dephasing [49, 50].
Stepwise reversal in a trimer
If a local pulse can decide whether a return sweep completes the reversal, a phase imprinted on an intermediate state should also be able to steer the next step of a longer sequence. We test this idea with three cavities on a triangle, with a hopping phase on each bond, , , and [Figure 3(a)]. The eight field configurations now form a cube. Each face describes reversals of two fields with the third held fixed, and at this working point every face carries a flux of . We apply three alternating uniform sweeps of duration and amplitude , separated by phase-gradient pulses of duration [Figure 3(b)], and compare two protocols that are identical up to the second pulse. Repeating the gradient yields three reversed fields with probability , while the single-reversal probability is . Reversing only the second gradient changes what the same final sweep does: it now leaves a single reversed field, with probability , and falls to [Figure 3(c)].
Figure 3. A phase gradient controls the next reversal step in a trimer with flux on each face of its configuration cube. (a) Occupations of the momentum modes (filled, occupied) after each step, for a second pulse identical to the first (blue) or reversed (orange); inset, the three cavities with hopping phase on each bond. (b) Bias (top) and configuration magnetization (bottom) during three sweeps of duration with (solid, microscopic model; dashed, adiabatic toggling of ); horizontal lines mark . (c) Final microscopic probabilities for identical (blue, left) and reversed (orange, right) second gradients; open circles mark the unit-probability outcomes of ideal mode toggling.
The mode picture of Figure 3(a) accounts for this behavior. In the dimer, a flux of blocked the second reversal; on the faces of the cube, the same flux makes the configurations equivalent to three fermion modes. In a Jordan–Wigner representation [51] with when the field in cavity is reversed, the sign of the string around a face reproduces its flux , and after a gauge transformation the single-flip terms in the equal-amplitude limit become (Appendix § 1.2.3)
Here denotes the trimer single-flip scale, and the operators () annihilate the three momentum modes. Because the bias and the zero-bias configuration energies depend only on the number of reversed fields, a uniform sweep is a Landau–Zener passage that changes the occupation of the uniform mode and leaves the other two modes untouched. The zero-bias energies nonetheless differ between shells: the configurations lie 0.0736 above those with , which shifts the and resonances to and , respectively, while the resonance remains at zero Figure 3(b). In either sweep direction, an adiabatic passage fills if it is empty and empties it if it is occupied, adding or removing one reversed field. At the first sweep reverses one field with probability 0.997, in agreement with , where .
The gradient pulses, in turn, move occupation between modes. A pulse that imprints the phase on configurations in which the field in cavity is reversed shifts each occupied mode from to modulo three (Appendix § 1.2.3). The protocol of Figure 3 then reads as a short sequence of mode operations. The first sweep fills , the first pulse moves that occupation to , and the second sweep fills again. Repeating the gradient shifts the occupied modes from to , leaving empty, and the final sweep fills it to complete the reversal. Reversing the gradient instead shifts to ; now is occupied, and the same sweep empties it, leaving a single reversed field. For identical pulses, the microscopic is close to , the prediction for three successful passages.
Conclusion
We have shown that the amplitudes for the two orders in which a pair of bistable fields reverses can interfere, and that this interference can block complete reversal. In coupled Rabi cavities, the four field configurations form a loop whose flux is set by a single complex hopping amplitude. Near half a flux quantum, a slow sweep reverses only one field even though the final bias favors reversing both, and a four-state model calibrated at zero bias predicts how the window of this Aharonov–Bohm blockade narrows with sweep time.
The blocked state is not a random reversal of one field. Its relative phase survives the sweep and can be read out from the joint field signs after a return sweep, so the coherence of the blockade is accessible to field-quadrature measurements. In a trimer, a phase imprinted between sweeps determines whether the same final sweep adds or removes a reversed field. Because the blockade is set by the phase of a single bond, it may also be relevant to other arrays of bistable units swept by a common bias, such as the oscillator networks proposed for annealing [8, 9].
Acknowledgments
This work was supported by the National Natural Science Foundation of China (Grant Nos. 11804240 and 12305032) and the Natural Science Foundation of Zhejiang Province (Grant No. LCQ23A050003). OpenAI Codex was used to assist with language editing, checking analytical derivations and numerical convergence, and preparing the data archive.
Data availability
The numerical data underlying Figure 1–Figure 4 and Table I, together with the simulation and plotting scripts, are available on Zenodo [52].
Figure 4. Finite-sweep test of the blockade at and . (a) Microscopic (symbols) near for three sweep durations; lines give the phase average from Eqs. (18) and (19), and the dotted horizontal line marks . (b) Full width of the interval below that threshold. End caps give the inner and outer widths resolved by the microscopic grid, and the line is the static prediction of Eq. (17).
Configuration ring and microscopic readout
At fixed , the lower eigenvalue of the two-level Hamiltonian in Eq. (1) gives the Born–Oppenheimer potential
For the potential has two minima at , separated by a barrier. Their positions and the barrier height are
For and the barrier height is 1.41. Writing gives , so the semiclassical tunneling action grows linearly with . The corresponding displacement estimate, , is close to for the lowest isolated-cavity parity doublet, whose relative phase we choose such that have mean fields .
Let denote the lowest energy in the sector , measured from the mean of the quartet. Inverting Eq. (3) gives
The branch of is fixed by continuity from . For the uniform bias, let be the matrix of in the same quartet; matching to its ring value gives . Table I lists the resulting parameters and distinguishes the half-flux point from the first crossing . The hopping strength at which the final is smallest for a given sweep time can differ from both.
Two independent checks support this calibration of the flux. First, the flux also follows from the phases of the directed matrix elements . Their product around is invariant under rephasing of the localized states, and its argument agrees modulo $2\piJ=0.01, 0.02, and 0.03 for . Second, for the geometric estimate in Sec. II, the quadratic photon Hamiltonian has momentum terms , with , , , and . Completing the square gives the vector potential and the uniform field in Eq. (4). Taking the four centers of the dressed states as vertices gives at , and the estimate differs from by 0.13% there and by 0.65% at .
| 0.04814 | 1.000 | 2.978 | 4.796 | 4.037 | 1.45 | 0.996 | ||
| 0.04884 | 1.015 | 2.972 | 4.874 | 4.037 | 1.45 | 0.998 | ||
| 0.09407 | 2.000 | 2.424 | 4.264 | 4.079 | 1.18 | 0.0220 | 0.978 | |
| 0.13671 | 3.008 | 1.765 | 4.143 | 0.85 | 0.985 |
Table 1. Dimer calibration at , , and . and are defined by their flux, and at and the two lowest zero-bias levels cross. The last three columns refer to a forward sweep with and .
For the microscopic dimer we keep 40 photon states per cavity, checked against 32, and 48 for the final static calibration. The propagation basis is built from instantaneous eigenstates. Within the space of the 64 lowest zero-bias states, we take the six lowest eigenvectors at each of five biases from to , and the dominant left singular vectors of this combined set span a fixed basis of dimension 24. For the pulse protocol, the sampling also includes the differential bias, and the basis has 32 vectors. Each time step applies a midpoint exponential. For the trimer we keep 40 photon states and 20 dressed levels per cavity and propagate in the basis of the 192 lowest zero-bias eigenstates. Uniform sweeps conserve the cyclic momentum, whereas the gradient pulses couple the three momentum sectors; each sweep uses 15 960 midpoint steps and each pulse 320.
The dressed configurations used in the readout are obtained by symmetric orthogonalization [53]. Let be the matrix of product well states and an orthonormal basis of instantaneous eigenstates. If , the dressed configurations are the columns of , and the squared singular values measure the overlap of the product states with the selected subspace before orthogonalization. For the dimer, contains the four lowest eigenstates. At the trimer endpoints , it contains the eight eigenstates with the largest total overlap with , and the smallest squared singular value is at least 0.888. The probabilities retain their normalization to the full state. In the trimer we compute , which reduces to for unit weight in the octet, as labeled in Figure 3(b). The field-sign readout uses the joint negative-half-axis projector for the two oscillator quadratures, summed over the qubit states.
The trimer results are converged well beyond the precision quoted in the text. All eight endpoint probabilities in Figure 3(c) change by less than 0.01% when the basis is enlarged from 144 to 192 states or the number of time steps is doubled, and they are similarly insensitive to reducing the photon cutoff to 32 states or enlarging the local dressed basis to 24 levels per cavity.
Blockade window and finite-sweep leakage
We use the bright states at as a fixed basis. In each ideal block of Eq. (7) the level separation is , and the lower levels of and have energies and , respectively; their relative slope therefore has magnitude . Their weights on and are and . Projecting a flux detuning and a direct double flip onto this pair gives the interblock matrix element
The factor arises because the direct flip couples both pairs of opposite configurations. At zero bias reduces to of Eq. (10). Near the second crossing the same construction applies with and .
The width of the blockade window follows from the two conditions of § 1.2 A: the sweep must be adiabatic for a single reversal and diabatic at the interblock crossing, which in the asymptotic Landau–Zener limit read and . Linearizing at gives the full width for ,
At , , and with Eq. (17) gives at , inside the microscopic grid bracket [3.2, 3.6] . For the shortest sweep, , the predicted width lies just below the microscopic bracket [6.8, 7.2] Figure 4(b).
The curve in Figure 1(d) retains the bias dependence of Eq. (16). With at the crossing, the lower-level energy difference is , and integrating over an unbounded sweep gives, to first order and up to an overall phase,
where is the Hankel function of the second kind and as . Figure 1(d) uses , a continuation of the weak-transfer result that recovers Eq. (11) in this limit.
A second amplitude arises near the endpoint. For , the two predominantly single-reversal levels have separation and coupling , so their mixing angle at the positive endpoint is . To first order in , integrating the relative phase gives an admixture . Its overlap with , together with the direct flux contribution, yields the endpoint amplitude, up to the relative propagation phase,
Near the first crossing the two amplitudes interfere, and their relative dynamical phase is sensitive to the sweep duration: at it changes by about for a change in . The lines in Figure 4(a) therefore show their phase-averaged probability. At the two extreme relative phases give and , bracketing the microscopic value . The same phase dependence allows to move to either side of .
Phase pulses in the dimer and trimer
During each pulse the uniform bias is held at its endpoint. In the dimer, the pulse adds to cavity 1 and to cavity 2; for elapsed pulse time we take
The opposite shifts of and give the relative phase , with at . At ideal half flux the pulse populates and with amplitudes and , and only connects to on the return sweep.
Two effects set the fringe envelope at finite . Each passage fails with probability , giving to leading order. In addition, the single-reversal state lags behind the endpoint rotation by , and the two sweeps contribute lags of equal magnitude. Averaging over their relative phase gives the curve plotted against the nominal phase in Figure 2(b),
Here . For the dephased control, we apply just before the pulse, using the four bias dressed-configuration projectors and their complement; the final dressed readout uses the instantaneous basis at .
In the trimer, let record whether the field in cavity is reversed, and let flip that bit. The Jordan–Wigner operators then anticommute, and the strings give a product of hopping signs of around each face, the same Wilson phase as the microscopic single-flip amplitudes. A rephasing of the cube vertices therefore yields Eq. (13) for equal single-flip magnitudes. In the microscopic trimer these magnitudes differ by between adjacent shells; this shell dependence preserves the selection of , and direct flips of two or three fields give only small corrections.
The trimer pulse adds to cavity the bias
where for the forward gradient. A reversed field in cavity thus acquires the phase relative to the unreversed configuration, and the pulse acts as
The opposite gradient shifts to instead, which produces the second outcome in Figure 3(c).
References
References
- [1]L. D. Landau, Zur Theorie der Energieübertragung. II, Phys. Z. Sowjetunion 2, 46 (1932). Phys. Rev. 51, 652 (1937).
- [2]C. Zener, Non-adiabatic crossing of energy levels, Proc. R. Soc. London A 137, 696 (1932).DOI
- [3]E. C. G. Stueckelberg, Theorie der unelastischen Stösse zwischen Atomen, Helv. Phys. Acta 5, 369 (1932).
- [4]E. Majorana, Atomi orientati in campo magnetico variabile, Nuovo Cimento 9, 43 (1932).DOI
- [5]S. N. Shevchenko, S. Ashhab, and F. Nori, Landau–Zener–Stückelberg interferometry, Phys. Rep. 492, 1 (2010).DOI
- [6]T. Kadowaki and H. Nishimori, Quantum annealing in the transverse Ising model, Phys. Rev. E 58, 5355 (1998).DOI
- [7]T. Albash and D. A. Lidar, Adiabatic quantum computation, Rev. Mod. Phys. 90, 015002 (2018).DOI
- [8]H. Goto, Bifurcation-based adiabatic quantum computation with a nonlinear oscillator network, Sci. Rep. 6, 21686 (2016).
- [9]S. Puri, C. K. Andersen, A. L. Grimsmo, and A. Blais, Quantum annealing with all-to-all connected nonlinear oscillators, Nat. Commun. 8, 15785 (2017).DOI
- [10]D. Loss, D. P. DiVincenzo, and G. Grinstein, Suppression of tunneling by interference in half-integer-spin particles, Phys. Rev. Lett. 69, 3232 (1992).DOI
- [11]J. von Delft and C. L. Henley, Destructive quantum interference in spin tunneling problems, Phys. Rev. Lett. 69, 3236 (1992).DOI
- [12]A. Garg, Topologically quenched tunnel splitting in spin systems with Kramers’ degeneracy, Europhys. Lett. 22, 205 (1993).
- [13]W. Wernsdorfer and R. Sessoli, Quantum phase interference and parity effects in magnetic molecular clusters, Science 284, 133 (1999).
- [14]S. Brundobler and V. Elser, S-matrix for generalized Landau–Zener problem, J. Phys. A 26, 1211 (1993).DOI
- [15]V. N. Ostrovsky and H. Nakamura, Exact analytical solution of the N-level Landau–Zener-type bow-tie model, J. Phys. A 30, 6939 (1997).DOI
- [16]N. A. Sinitsyn, Solvable four-state Landau–Zener model of two interacting qubits with path interference, Phys. Rev. B 92, 205431 (2015).
- [17]V. Y. Vyacherny, A. N. Sinitsyn, and C. Sun, Dynamic spin localization and γ-magnets, Phys. Rev. B 100, 224304 (2019).
- [18]J. Vidal, R. Mosseri, and B. Douçot, Aharonov–Bohm cages in two-dimensional structures, Phys. Rev. Lett. 81, 5888 (1998).DOI
- [19]S. Mukherjee, M. Di Liberto, P. Öhberg, R. R. Thomson, and N. Goldman, Experimental observation of Aharonov–Bohm cages in photonic lattices, Phys. Rev. Lett. 121, 075502 (2018).DOI
- [20]J. G. C. Martinez, C. S. Chiu, B. M. Smitham, and A. A. Houck, Flat-band localization and interaction-induced delocalization of photons, Sci. Adv. 9, eadj7195 (2023).DOI
- [21]I. I. Rabi, Space quantization in a gyrating magnetic field, Phys. Rev. 51, 652 (1937).DOI
- [22]D. Braak, Integrability of the Rabi model, Phys. Rev. Lett. 107, 100401 (2011).pubmed.ncbi.nlm.nih.gov/21981483
- [23]P. Forn-Díaz, L. Lamata, E. Rico, J. Kono, and E. Solano, Ultrastrong coupling regimes of light-matter interaction, Rev. Mod. Phys. 91, 025005 (2019).
- [24]M.-J. Hwang, R. Puebla, and M. B. Plenio, Quantum phase transition and universal dynamics in the Rabi model, Phys. Rev. Lett. 115, 180404 (2015).DOI
- [25]K. Hepp and E. H. Lieb, On the superradiant phase transition for molecules in a quantized radiation field: the Dicke maser model, Ann. Phys. (N.Y.) 76, 360 (1973).DOI
- [26]K. Baumann, C. Guerlin, F. Brennecke, and T. Esslinger, Dicke quantum phase transition with a superfluid gas in an optical cavity, Nature 464, 1301 (2010).DOI
- [27]D. Lv, S. An, Z. Liu, J.-N. Zhang, J. S. Pedernales, L. Lamata, E. Solano, and K. Kim, Quantum simulation of the quantum Rabi model in a trapped ion, Phys. Rev. X 8, 021027 (2018).DOI
- [28]R. Puebla, M.-J. Hwang, J. Casanova, and M. B. Plenio, Probing the dynamics of a superradiant quantum phase transition with a single trapped ion, Phys. Rev. Lett. 118, 073001 (2017).DOI
- [29]M.-L. Cai, Z.-D. Liu, W.-D. Zhao, Y.-K. Wu, Q.-X. Mei, Y. Jiang, L. He, X. Zhang, Z.-C. Zhou, and L.-M. Duan, Observation of a quantum phase transition in the quantum Rabi model with a single trapped ion, Nat. Commun. 12, 1126 (2021).
- [30]M. Schirò, M. Bordyuh, B. Öztop, and H. E. Türeci, Phase transition of light in cavity QED lattices, Phys. Rev. Lett. 109, 053601 (2012).pubmed.ncbi.nlm.nih.gov/23006171
- [31]Q.-X. Mei, B.-W. Li, Y.-K. Wu, M.-L. Cai, Y. Wang, L. Yao, Z.-C. Zhou, and L.-M. Duan, Experimental realization of the Rabi–Hubbard model with trapped ions, Phys. Rev. Lett. 128, 160504 (2022).DOI
- [32]Y.-Y. Zhang, Z.-X. Hu, L. Fu, H.-G. Luo, H. Pu, and X.-F. Zhang, Quantum phases in a quantum Rabi triangle, Phys. Rev. Lett. 127, 063602 (2021).DOI
- [33]D. Fallas Padilla, H. Pu, G.-J. Cheng, and Y.-Y. Zhang, Understanding the quantum Rabi ring using analogies to quantum magnetism, Phys. Rev. Lett. 129, 183602 (2022).
- [34]J. Zhao and M.-J. Hwang, Frustrated superradiant phase transition, Phys. Rev. Lett. 128, 163601 (2022).DOI
- [35]S. He and L.-W. Duan, Extensively degenerate superradiant phase in a giant-atom quantum Rabi ring (2026), arXiv:2610.09880 [quant-ph].arxiv.org/abs/2610.09880
- [36]J. Dalibard, F. Gerbier, G. Juzeliūnas, and P. Öhberg, Colloquium: Artificial gauge potentials for neutral atoms, Rev. Mod. Phys. 83, 1523 (2011).
- [37]T. Ozawa, H. M. Price, A. Amo, N. Goldman, M. Hafezi, L. Lu, C. Rechstman, D. Schuster, J. Simon, O. Zilberberg, and I. Carusotto, Topological photonics, Rev. Mod. Phys. 91, 015006 (2019).DOI
- [38]J. Koch, A. A. Houck, K. Le Hur, and S. M. Girvin, Time-reversal-symmetry breaking in circuit-QED-based photon lattices, Phys. Rev. A 82, 043811 (2010).
- [39]P. Roushan, C. Neill, A. Megrant, Y. Chen, R. Bab-bush, R. Barends, B. Campbell, Z. Chen, B. Chiaro, A. Dunsworth, et al., Chiral ground-state currents of interacting photons in a synthetic magnetic field, Nat. Phys. 13, 146 (2017).DOI
- [40]S. Debnath, M. N. Linke, S.-T. Wang, C. Figgatt, K. A. Landsman, L.-M. Duan, and C. Monroe, Observation of hopping and blockade of bosons in a trapped ion spin chain, Phys. Rev. Lett. 120, 027301 (2018).DOI
- [41]T. Manovitz, Y. Shapira, N. Akerman, A. Stern, and R. Ozeri, Quantum simulations with complex geometries and synthetic gauge fields in a trapped ion chain, PRX Quantum 1, 020303 (2020).DOI
- [42]Y. Shapira, T. Manovitz, N. Akerman, A. Stern, and R. Ozeri, Quantum simulations of interacting systems with broken time-reversal symmetry, Phys. Rev. X 13, 021021 (2023).arxiv.org/abs/2205.11178
- [43]A. Celi, P. Massignan, J. Ruseckas, N. Goldman, I. B. Spielman, G. Juzeliūnas, and M. Lewenstein, Synthetic gauge fields in synthetic dimensions, Phys. Rev. Lett. 112, 043001 (2014).
- [44]Y. Aharonov and D. Bohm, Significance of electromagnetic potentials in the quantum theory, Phys. Rev. 115, 485 (1959).DOI
- [45]B. C. Sanders, Entangled coherent states, Phys. Rev. A 45, 6811 (1992).pubmed.ncbi.nlm.nih.gov/9907804
- [46]J. Joo, W. J. Munro, and T. P. Spiller, Quantum metrology with entangled coherent states, Phys. Rev. Lett. 107, 083601 (2011).DOI
- [47]C. Wang, Y. Y. Gao, P. Reinhold, R. W. Heeres, N. Ofek, K. Chou, C. Axline, M. Reagor, J. Blumoff, K. M. Sliwa, L. Frunzio, S. M. Girvin, L. Jiang, M. Mirrahimi, M. H. Devoret, and R. J. Schoelkopf, A Schrödinger cat living in two boxes, Science 352, 1087 (2016).DOI
- [48]W. D. Oliver, Y. Yu, J. C. Lee, K. K. Berggren, L. S. Levitov, and T. P. Orlando, Mach-Zehnder interferometry in a strongly driven superconducting qubit, Science 310, 1653 (2005).pubmed.ncbi.nlm.nih.gov/16282527
- [49]M. O. Scully, B.-G. Englert, and H. Walther, Quantum optical tests of complementarity, Nature 351, 111 (1991).
- [50]B.-G. Englert, Fringe visibility and which-way information: An inequality, Phys. Rev. Lett. 77, 2154 (1996).pubmed.ncbi.nlm.nih.gov/10061872
- [51]P. Jordan and E. Wigner, Über das Paulische Äquivalenzverbot, Z. Phys. 47, 631 (1928).
- [52]S. He and L. Duan, Data and code for “Aharonov–Bohm blockade of complete reversal in coupled Rabi cavities”, Zenodo (2026), doi:10.5281/zenodo.24328307.arxiv.org/abs/2611.23861
- [53]P.-O. Löwdin, On the non-orthogonality problem connected with the use of atomic wave functions in the theory of molecules and crystals, J. Chem. Phys. 18, 365 (1950).



