betaFeedback

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 γ\gamma-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,

H=∑j=1,2[aj†aj+η2σjz+ληXjσjx+ϵ(t)Xj]+J(eiθa1†a2+H.c.),(1)H = \sum_{j=1,2} \left[a_j^\dagger a_j + \frac{\eta}{2}\sigma_j^z + \lambda\sqrt{\eta}X_j\sigma_j^x + \epsilon(t)X_j\right] + J\left(e^{i\theta}a_1^\dagger a_2 + \mathrm{H.c.}\right), \tag*{(1)}

where aja_j annihilates a photon in cavity jj, Xj=aj+aj†X_j = a_j + a_j^\dagger, σjx,z\sigma_j^{x,z} are Pauli matrices of the two-level system in cavity jj, η\eta is the ratio of the qubit and cavity frequencies, and λ\lambda is the dimensionless light–matter coupling. We set ℏ=1\hbar= 1 and the cavity frequency to unity, and use η=12\eta= 12, λ=0.7\lambda= 0.7, and θ=π/2\theta= \pi/2 unless stated otherwise. Only the uniform bias ϵ(t)\epsilon(t) varies during a sweep. The bond phase θ\theta is held fixed and, as shown below, controls the flux Φ\Phi through the configuration loop. In the trimer, the hopping is oriented cyclically around the triangle.

An isolated cavity becomes superradiant for λ>1/2\lambda> 1/2 as η→∞\eta\to\infty [24], and its field is then bistable. Eliminating the two-level system adiabatically yields a double-well potential for the field quadrature xj=Xj/2x_j = X_j/\sqrt{2}, with a barrier of height 1.41 for our parameters; tunneling between the two wells is suppressed exponentially in η\eta (Appendix § 1.2.1). We denote the dressed states localized in the two wells of a cavity by ∣±⟩\lvert\pm\rangle and the four configurations of the dimer by ∣++⟩\lvert++\rangle, ∣−+⟩\lvert-+\rangle, ∣−−⟩\lvert--\rangle, and ∣+−⟩\lvert+-\rangle, 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 sj=±1s_j = \pm1 acquires the bias energy qϵ(s1+s2)q\epsilon(s_1+s_2). The projected bias coupling qq is approximately the magnitude of the field displacement ⟨Xj⟩\langle X_j\rangle in a localized state; at the first zero-bias ground-state crossing, Jg1J_{g1}, we find q=4.04q = 4.04. Reversing one field therefore releases the energy Δ(t)=2qϵ(t)\Delta(t) = 2q\epsilon(t), and a positive bias favors ∣−−⟩\lvert--\rangle. The single-reversal configurations ∣−+⟩\lvert-+\rangle and ∣+−⟩\lvert+-\rangle, 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 JeiθJe^{i\theta}, and the four configurations in the plane of the two fields. Arrows mark single flips with amplitude −t1eiΦ/4-t_1e^{i\Phi/4}, and dotted diagonals mark direct double flips with amplitude −t2-t_2. (b) Magnetization during the sweep at J2πJ_{2\pi} and Jg1J_{g1} for T=2.4×104T=2.4\times10^4 (solid, microscopic model; dashed, two independent cavities at J2πJ_{2\pi} and the block {∣++⟩,∣1+⟩}\{|++\rangle,|1+\rangle\} at Jg1J_{g1}, § 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 ∣++⟩|++\rangle, and dotted lines are the configuration energies. (d) Final P2P_2 (symbols, microscopic model; line, Eq. (11) with the first-order refinement of Appendix § 1.2.2) and microscopic P1P_1 (gray) versus JJ for T=2.4×104T=2.4\times10^4; the top axis gives the flux. All panels use η=12\eta=12, λ=0.7\lambda=0.7, θ=π/2\theta=\pi/2, and h=0.0125h=0.0125.

The dimer is prepared in its ground state at ϵ=−h\epsilon= -h, which lies close to ∣++⟩\lvert++\rangle, and the bias is swept linearly to +h+h, ϵ(t)=−h+2ht/T\epsilon(t) = -h + 2ht/T, with h=0.0125h = 0.0125 unless stated otherwise. At t=Tt = T we project the state onto four orthonormal dressed configurations constructed from the instantaneous eigenstates (Appendix § 1.2.1). This projection yields the probability PNP_N that exactly NN fields have reversed and the magnetization M=P++−P−−M = P_{++}-P_{--}. A second readout bypasses the dressed configurations altogether: the probability Pr⁡(x1<0,x2<0)\Pr(x_1 < 0, x_2 < 0) 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 θ=π/2\theta= \pi/2, Eq. (1) commutes with C4=Π1S12C_4 = \Pi_1S_{12}, where Π1=−σ1z(−1)a1†a1\Pi_1 = -\sigma_1^z(-1)^{a_1^\dagger a_1} is the parity of cavity 1 and S12S_{12} exchanges the two cavities; a finite uniform bias or a real hopping component Jcos⁡θ≠0J\cos\theta\ne0 breaks this symmetry. The operator C4C_4 rotates the plane of the two fields by π/2\pi/2 and permutes the configurations cyclically, ∣++⟩→∣−+⟩→∣−−⟩→∣+−⟩→∣++⟩\lvert++\rangle\to\lvert-+\rangle\to\lvert--\rangle\to\lvert+-\rangle\to\lvert++\rangle, so the zero-bias Hamiltonian restricted to these configurations is a circulant matrix. Adding the leading diagonal bias gives the ring Hamiltonian

HR=−t1eiΦ/4(∣−+⟩⟨++∣+∣−−⟩⟨−+∣+∣+−⟩⟨−−∣+∣++⟩⟨+−∣)−t2(∣−−⟩⟨++∣+∣+−⟩⟨−+∣)+H.c.+Δ(t)(∣++⟩⟨+−∣−∣−−⟩⟨−+∣),(2)H_R = -t_1e^{i\Phi/4}(\lvert-+\rangle\langle++\rvert+\lvert--\rangle\langle-+\rvert+\lvert+-\rangle\langle--\rvert+\lvert++\rangle\langle+-\rvert) -t_2(\lvert--\rangle\langle++\rvert+\lvert+-\rangle\langle-+\rvert)+\mathrm{H.c.} +\Delta(t)(\lvert++\rangle\langle+-\rvert-\lvert--\rangle\langle-+\rvert), \tag*{(2)}

where t1>0t_1 > 0 sets the single-flip magnitude, −t2-t_2 is the signed direct double-flip matrix element, and Φ\Phi 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 imi^m of C4C_4, are

Em=−2t1cos⁡(πm2+Φ4)−t2(−1)m.(3)E_m = -2t_1\cos\left(\frac{\pi m}{2}+\frac{\Phi}{4}\right)-t_2(-1)^m. \tag*{(3)}

Matching Eq. (3) to the lowest zero-bias quartet of Eq. (1) determines t1t_1, t2t_2, and Φ\Phi, and the matrix elements of X1+X2X_1+X_2 within the same quartet determine qq (Appendix § 1.2.1). All parameters of the ring are thus fixed at zero bias. At the hopping strength Jg1=0.0488J_{g1}=0.0488 used below, for example, t1=2.97×10−3t_1 = 2.97\times10^{-3} and t2=4.87×10−5t_2 = 4.87\times10^{-5}; 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 xjx_j and their conjugate momenta pjp_j, the hopping term contains −Jsin⁡θ(x1p2−x2p1)-J\sin\theta\left(x_1p_2-x_2p_1\right), 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

B=2Jsin⁡θ1−J2cos⁡2θ.(4)B = \frac{2J\sin\theta}{1-J^2\cos^2\theta}. \tag*{(4)}

The flux through the loop can then be estimated as BABA, where AA is the area enclosed by the four dressed-state centers. At Jg1J_{g1} this gives A=32.6A = 32.6 and reproduces the spectral flux to 0.13%0.13\% (Appendix § 1.2.1). Increasing JJ from 0.030.03 to 0.150.15 raises Φ\Phi through π\pi, 2π2\pi, and 3π3\pi [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 T=2.4×104T=2.4\times10^4 that differ only in the hopping strength. At J2π=0.0941J_{2\pi}=0.0941, where Φ=2π\Phi=2\pi, the sweep follows the bias: both fields reverse with probability P2=0.978P_2=0.978, and the magnetization swings from nearly +1+1 to −0.978-0.978. At Jg1J_{g1}, where the two lowest zero-bias levels cross close to Φ=π\Phi=\pi, the same sweep stops halfway. The final magnetization is only 1.65×10−31.65\times10^{-3}; one field reverses with probability P1=0.998P_1=0.998, whereas P2=3.07×10−5P_2=3.07\times10^{-5}. The direct readout gives Pr⁡(x1<0,x2<0)=1.15×10−3\Pr(x_1<0,x_2<0)=1.15\times10^{-3}, 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 ∣−−⟩|--\rangle lies 2qh=0.1012qh=0.101 below the single-reversal configurations, more than 30 times t1t_1, and the first reversal occurs with near certainty.

The explanation lies in the two orders of reversal [Figure 1(a)]. Starting from ∣++⟩|++\rangle, a single flip does not select one cavity; it produces the single-reversal state

∣1+⟩=12(eiΦ/4∣−+⟩+e−iΦ/4∣+−⟩)(5)|1+\rangle=\frac{1}{\sqrt{2}}\left(e^{i\Phi/4}|-+\rangle+e^{-i\Phi/4}|+-\rangle\right) \tag*{(5)}

with amplitude −g-g, where g=2t1g=\sqrt{2}t_1. The final configuration ∣−−⟩|--\rangle, in turn, is reached by a single flip only from

∣1−⟩=12(e−iΦ/4∣−+⟩+eiΦ/4∣+−⟩).|1-\rangle=\frac{1}{\sqrt{2}}\left(e^{-i\Phi/4}|-+\rangle+e^{i\Phi/4}|+-\rangle\right).

Whether the second step can proceed therefore depends on the overlap

⟨1−∣1+⟩=cos⁡(Φ/2),(6)\langle1-|1+\rangle=\cos(\Phi/2), \tag*{(6)}

which is proportional to the sum of the amplitudes of the two paths ++→−+→−−++\rightarrow-+\rightarrow-- and ++→+−→−−++\rightarrow+-\rightarrow--, carrying the phases ±Φ/2\pm\Phi/2. At Φ=2π\Phi=2\pi a gauge transformation makes all single-flip amplitudes real and equal, and for t2=0t_2=0 Eq. (2) describes two independent cavities. At Φ=π\Phi=\pi the two paths cancel and the two states are orthogonal; for t2=0t_2=0, Eq. (2) then separates into the two blocks

H±=(±Δ−g−g0)(7)H_{\pm}=\begin{pmatrix}\pm\Delta&-g\\-g&0\end{pmatrix} \tag*{(7)}

in the bases {∣++⟩,∣1+⟩}\{|++\rangle,|1+\rangle\} and {∣−−⟩,∣1−⟩}\{|--\rangle,|1-\rangle\} [Figure 1(c)]. The first flip from ∣++⟩|++\rangle remains allowed, but the state it reaches has no matrix element to ∣−−⟩|--\rangle, 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 v=dΔ/dt=4qh/Tv=d\Delta/dt=4qh/T and define the adiabaticity parameter ε=g2/v\varepsilon=g^2/v, which equals 1.451.45 at Jg1J_{g1} for T=2.4×104T=2.4\times10^4. At Φ=2π\Phi=2\pi and t2=0t_2=0, each field undergoes its own passage with coupling t1=g/2t_1=g/\sqrt{2}, and an infinite sweep reverses both with probability

P2=(1−e−πε)2,(8)P_2=\left(1-e^{-\pi\varepsilon}\right)^2, \tag*{(8)}

which equals 0.974 at J2πJ_{2\pi}. 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 Φ=π\Phi=\pi and t2=0t_2=0, the sweep starts in ∣++⟩\lvert++\rangle and never leaves the block H+H_{+}, whose lower level connects ∣++⟩\lvert++\rangle to ∣1+⟩\lvert1+\rangle [shaded curve in Figure 1(c)]. The infinite-sweep result is then

P1=1−e−2πξ2,P2=0.(9)P_{1}=1-e^{-2\pi\xi^{2}},\qquad P_{2}=0. \tag*{(9)}

At the finite positive endpoint, the main correction is an admixture P0≈g2/Δh2P_{0}\approx g^{2}/\Delta_{h}^{2} of ∣++⟩\lvert++\rangle, with Δh=2qh\Delta_{h}=2qh; at Jg1J_{g1} this gives 1.73×10−31.73\times10^{-3}, close to the microscopic 1.68×10−31.68\times10^{-3}.

The microscopic model does not meet the two ideal conditions Φ=π\Phi=\pi and t2=0t_{2}=0 at the same hopping strength. At Jπ=0.0481J_{\pi}=0.0481, where Φ=π\Phi=\pi, the direct amplitude is t2=4.80×10−5t_{2}=4.80\times10^{-5}, and the two blocks remain coupled. At zero bias their coupling equals half the signed splitting of the two lowest levels,

Γ=E0−E32=gsin⁡Φ−π4−t2,(10)\Gamma=\frac{E_{0}-E_{3}}{2}=g\sin\frac{\Phi-\pi}{4}-t_{2}, \tag*{(10)}

and it vanishes at the crossing Jg1J_{g1} of these levels, where the flux detuning Φ−π=0.0464\Phi-\pi=0.0464 compensates t2t_{2}. 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 vv, and the Landau–Zener formula gives the population transferred to the block that ends in ∣−−⟩\lvert--\rangle,

P2=1−exp⁡(−2πΓ2/v).(11)P_{2}=1-\exp\left(-2\pi\Gamma^{2}/v\right). \tag*{(11)}

Evaluated with Γ\Gamma taken from the zero-bias spectrum at each JJ, and with the bias dependence of the coupling included to first order (Appendix § 1.2.2), this expression reproduces the microscopic P2P_{2} across the windows near Jg1J_{g1} and near the corresponding crossing Jg2=0.137J_{g2}=0.137 close to Φ=3π\Phi=3\pi, as well as between them, with a median relative deviation of 1.2%1.2\% where P2>0.01P_{2}>0.01 [Figure 1(d)].

The blockade requires an intermediate sweep rate. The sweep must be slow enough for the first reversal, e−2πξ2≪1e^{-2\pi\xi^{2}}\ll1, yet fast enough to suppress transfer between the blocks, 2πΓ2/v≪12\pi\Gamma^{2}/v\ll1. At fixed JJ away from the crossing, a longer sweep eventually permits the second reversal, so the window in JJ narrows as T−1/2T^{-1/2}, consistent with the narrowing seen in the microscopic sweeps (Appendix § 1.2.2). Even at Jg1J_{g1}, where the zero-bias coupling vanishes, a small residual P2P_{2} remains, because the sweep spans a finite range of bias. An amplitude acquired near the crossing can interfere with the admixture of ∣−−⟩\lvert--\rangle at the endpoint; the estimate in Appendix § 1.2.2 accounts for the scale of this residual and explains why the smallest P2P_{2} need not occur exactly at Jg1J_{g1}.

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 ∣−+⟩\lvert-+\rangle and ∣+−⟩\lvert+-\rangle would yield the same P1P_{1}. In the ideal ring, however, the sweep prepares ∣1+⟩\lvert1+\rangle, 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 P+−=0.4657P_{+-}=0.4657 and P−+=0.5326P_{-+}=0.5326, giving P1=0.9983P_{1}=0.9983. 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 Jg1J_{g1} we sweep from −h-h to +h+h, apply opposite local biases for τ=160\tau=160, and sweep back; each sweep has h=0.015h=0.015, T=104T=10^{4}, and ξ=0.854\xi=0.854. The pulses shift ∣−+⟩\lvert-+\rangle and ∣+−⟩\lvert+-\rangle in opposite directions and imprint a relative phase between them, which we characterize by the nominal phase φ=4qloc∫ϵp dt\varphi=4q_{\mathrm{loc}}\int\epsilon_{p}\,\mathrm{d}t, the value for uncoupled cavities with bias coupling qloc=4.02q_{\mathrm{loc}}=4.02 (Appendix § 1.2.3); the ring gives the imprinted phase φ′=(q/qloc)φ\varphi'=(q/q_{\mathrm{loc}})\varphi. At ideal half flux, this phase rotates ∣1+⟩\lvert1+\rangle into cos⁡(φ′/2)∣1+⟩−sin⁡(φ′/2)∣1−⟩\cos(\varphi'/2)\lvert1+\rangle-\sin(\varphi'/2)\lvert1-\rangle. The return sweep then traverses each block backward: ∣1+⟩|1_+\rangle returns to ∣++⟩|++\rangle, whereas ∣1−⟩|1_-\rangle is carried on to ∣−−⟩|--\rangle. 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 +h+h, opposite bias pulses ±ϵp\pm\epsilon_{p} with nominal phase φ\varphi (Appendix § 1.2.3), and a sweep back to −h-h; each sweep lasts T=104T=10^{4}, the pulses last τ=160\tau=160, and h=0.015h=0.015. (b) Final P2P_{2} 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.

P2(φ′)=sin⁡2(φ′/2).(12)P_2(\varphi') = \sin^2(\varphi'/2). \tag*{(12)}

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 1−2e−2πξ2=0.9801 - 2e^{-2\pi\xi^2} = 0.980, and the direct double flips, which mix ∣1+⟩|1_+\rangle and ∣1−⟩|1_-\rangle 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, Pr⁡(x1<0,x2<0)\operatorname{Pr}(x_1 < 0, x_2 < 0) 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 ∣++⟩|++\rangle or on to ∣−−⟩|--\rangle, 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 π/2\pi/2 on each bond, J=0.0475J = 0.0475, η=12\eta= 12, and λ=0.7\lambda= 0.7 [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 π\pi. We apply three alternating uniform sweeps of duration T=2×104T = 2 \times10^4 and amplitude h=0.03h = 0.03, separated by phase-gradient pulses of duration τ=160\tau= 160 [Figure 3(b)], and compare two protocols that are identical up to the second pulse. Repeating the gradient yields three reversed fields with probability P3=0.986P_3 = 0.986, while the single-reversal probability is P1=2.43×10−4P_1 = 2.43 \times10^{-4}. Reversing only the second gradient changes what the same final sweep does: it now leaves a single reversed field, with probability P1=0.994P_1 = 0.994, and P3P_3 falls to 5.39×10−45.39 \times10^{-4} [Figure 3(c)].

Figure 3. A phase gradient controls the next reversal step in a trimer with flux π\pi on each face of its configuration cube. (a) Occupations of the momentum modes fkf_k (filled, occupied) after each step, for a second pulse identical to the first (blue) or reversed (orange); inset, the three cavities with hopping phase π/2\pi/2 on each bond. (b) Bias (top) and configuration magnetization MM (bottom) during three sweeps of duration T=2×104T = 2 \times10^4 with h=0.03h = 0.03 (solid, microscopic model; dashed, adiabatic toggling of f0f_0); horizontal lines mark N=0,…,3N = 0,\ldots,3. (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 π\pi 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 nj=cj†cj=1n_j = c_j^\dagger c_j = 1 when the field in cavity jj is reversed, the sign of the string around a face reproduces its flux π\pi, and after a gauge transformation the single-flip terms in the equal-amplitude limit become (Appendix § 1.2.3)

−t1∑j=13(cj+cj†)=−g3(f0+f0†),g3=3t1.(13)-t_1 \sum_{j=1}^{3}(c_j+c_j^\dagger)=-g_3(f_0+f_0^\dagger),\qquad g_3=\sqrt{3}t_1. \tag*{(13)}

Here t1t_1 denotes the trimer single-flip scale, and the operators fk=3−1/2∑je−2πikj/3cjf_k = 3^{-1/2}\sum_j e^{-2\pi i k j/3}c_j (k=0,1,2k = 0,1,2) annihilate the three momentum modes. Because the bias and the zero-bias configuration energies depend only on the number N=∑kfk†fkN = \sum_k f_k^\dagger f_k of reversed fields, a uniform sweep is a Landau–Zener passage that changes the occupation of the uniform mode f0f_0 and leaves the other two modes untouched. The zero-bias energies nonetheless differ between shells: the N=0,3N = 0, 3 configurations lie 0.0736 above those with N=1,2N = 1, 2, which shifts the 0↔10 \leftrightarrow1 and 2↔32 \leftrightarrow3 resonances to ϵ/h≃−0.305\epsilon/h \simeq-0.305 and +0.305+0.305, respectively, while the 1↔21 \leftrightarrow2 resonance remains at zero Figure 3(b). In either sweep direction, an adiabatic passage fills f0f_0 if it is empty and empties it if it is occupied, adding or removing one reversed field. At T=2×104T = 2 \times10^4 the first sweep reverses one field with probability 0.997, in agreement with 1−e−2πg32/v31-e^{-2\pi g_3^2/v_3}, where v3=4qloch/Tv_3 = 4q_{\mathrm{loc}}h/T.

The gradient pulses, in turn, move occupation between modes. A pulse that imprints the phase 2π(j−2)/32\pi(j-2)/3 on configurations in which the field in cavity jj is reversed shifts each occupied mode from kk to k+1k+1 modulo three (Appendix § 1.2.3). The protocol of Figure 3 then reads as a short sequence of mode operations. The first sweep fills f0f_0, the first pulse moves that occupation to f1f_1, and the second sweep fills f0f_0 again. Repeating the gradient shifts the occupied modes from {0,1}\{0,1\} to {1,2}\{1,2\}, leaving f0f_0 empty, and the final sweep fills it to complete the reversal. Reversing the gradient instead shifts {0,1}\{0,1\} to {2,0}\{2,0\}; now f0f_0 is occupied, and the same sweep empties it, leaving a single reversed field. For identical pulses, the microscopic P3=0.986P_3 = 0.986 is close to 0.9973≃0.9910.997^3 \simeq0.991, 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 η=12\eta= 12 and h=0.0125h = 0.0125. (a) Microscopic P2P_2 (symbols) near Jg1J_{\mathrm{g1}} for three sweep durations; lines give the phase average ∣Ac∣2+∣Ae∣2|\mathcal{A}_{c}|^2 + |\mathcal{A}_{e}|^2 from Eqs. (18) and (19), and the dotted horizontal line marks P2=0.01P_2 = 0.01. (b) Full width of the interval below that threshold. End caps give the inner and outer widths resolved by the microscopic JJ grid, and the line is the static prediction of Eq. (17).

Configuration ring and microscopic readout

At fixed x=X/2x = X/\sqrt{2}, the lower eigenvalue of the two-level Hamiltonian in Eq. (1) gives the Born–Oppenheimer potential

U(x)=x22−η21+8λ2x2/η.(14)U(x)=\frac{x^2}{2}-\frac{\eta}{2}\sqrt{1+8\lambda^2x^2/\eta}. \tag*{(14)}

For λ>1/2\lambda> 1/2 the potential has two minima at ±x0\pm x_0, separated by a barrier. Their positions and the barrier height are

x02=η(16λ4−1)8λ2,U(0)−U(x0)=η(4λ2−1)216λ2.x_0^2=\frac{\eta(16\lambda^4-1)}{8\lambda^2},\qquad U(0)-U(x_0)=\frac{\eta(4\lambda^2-1)^2}{16\lambda^2}.

For η=12\eta= 12 and λ=0.7\lambda= 0.7 the barrier height is 1.41. Writing x=ηux=\sqrt{\eta}u gives U=ηUˉ(u)U=\eta\bar{U}(u), so the semiclassical tunneling action grows linearly with η\eta. The corresponding displacement estimate, 2x0=4.17\sqrt{2}x_0=4.17, is close to qloc=∣⟨e∣X∣o⟩∣=4.02q_{\mathrm{loc}}=|\langle e|X|o\rangle|=4.02 for the lowest isolated-cavity parity doublet, whose relative phase we choose such that (∣e⟩±∣o⟩)/2(|e\rangle\pm|o\rangle)/\sqrt{2} have mean fields ±qloc\pm q_{\mathrm{loc}}.

Let EmE_m denote the lowest energy in the sector C4=imC_4=i^m, measured from the mean of the quartet. Inverting Eq. (3) gives

t1eiΦ/4=14[E2−E0+i(E1−E3)],t_1e^{i\Phi/4}=\frac{1}{4}[E_2-E_0+i(E_1-E_3)],
t2=12(E1+E3).(15)t_2=\frac{1}{2}(E_1+E_3). \tag*{(15)}

The branch of Φ\Phi is fixed by continuity from J=0J=0. For the uniform bias, let DD be the matrix of X1+X2X_1+X_2 in the same quartet; matching Tr⁡D2\operatorname{Tr}D^2 to its ring value gives q=(Tr⁡D2/8)1/2q=(\operatorname{Tr}D^2/8)^{1/2}. Table I lists the resulting parameters and distinguishes the half-flux point JπJ_\pi from the first crossing Jg1J_{g1}. The hopping strength Jmin⁡(T)J_{\min}(T) at which the final P2P_2 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 Hs′,s=⟨s′∣H∣s⟩H_{s',s}=\langle s'|H|s\rangle. Their product around + ⁣+→+ ⁣−→− ⁣−→− ⁣+→+ ⁣++\!+\to+\!-\to-\!-\to-\!+\to+\!+ is invariant under rephasing of the localized states, and its argument agrees modulo $2\piwiththespectralfluxatwith the spectral flux atJ=0.01, 0.02, and 0.03 for θ=π/2\theta=\pi/2. Second, for the geometric estimate in Sec. II, the quadratic photon Hamiltonian has momentum terms pTQp/2+pTRxp^{\mathsf T}Qp/2+p^{\mathsf T}Rx, with Q11=Q22=1Q_{11}=Q_{22}=1, Q12=Q21=Jcos⁡θQ_{12}=Q_{21}=J\cos\theta, R12=−R21=Jsin⁡θR_{12}=-R_{21}=J\sin\theta, and R11=R22=0R_{11}=R_{22}=0. Completing the square gives the vector potential −Q−1Rx-Q^{-1}Rx and the uniform field in Eq. (4). Taking the four centers (⟨x1⟩,⟨x2⟩)(\langle x_1\rangle,\langle x_2\rangle) of the dressed states as vertices gives A=32.595A=32.595 at Jg1J_{g1}, and the estimate BABA differs from Φ\Phi by 0.13% there and by 0.65% at Jg2J_{g2}.

JJΦ/π\Phi/\pit1 (10−3)t_1\ (10^{-3})t2 (10−5)t_2\ (10^{-5})qqξ\xiP1P_1P2P_2
JπJ_\pi0.048141.0002.9784.7964.0371.450.9962.03×10−32.03 \times10^{-3}
Jg1J_{g1}0.048841.0152.9724.8744.0371.450.9983.07×10−53.07 \times10^{-5}
J2πJ_{2\pi}0.094072.0002.4244.2644.0791.180.02200.978
Jg2J_{g2}0.136713.0081.765−1.553-1.5534.1430.850.9858.61×10−68.61 \times10^{-6}

Table 1. Dimer calibration at η=12\eta=12, λ=0.7\lambda=0.7, and θ=π/2\theta=\pi/2. JπJ_\pi and J2πJ_{2\pi} are defined by their flux, and at Jg1J_{g1} and Jg2J_{g2} the two lowest zero-bias levels cross. The last three columns refer to a forward sweep with h=0.0125h=0.0125 and T=2.4×104T=2.4\times10^4.

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 −h-h to +h+h, 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 LL be the matrix of product well states and VV an orthonormal basis of instantaneous eigenstates. If V†L=UΣW†V^\dagger L=U\Sigma W^\dagger, the dressed configurations are the columns of VUW†VUW^\dagger, and the squared singular values measure the overlap of the product states with the selected subspace before orthogonalization. For the dimer, VV contains the four lowest eigenstates. At the trimer endpoints ∣ϵ∣=0.03|\epsilon|=0.03, it contains the eight eigenstates with the largest total overlap with LL, and the smallest squared singular value is at least 0.888. The probabilities PNP_N retain their normalization to the full state. In the trimer we compute M=∑N=03(1−2N/3)PNM=\sum_{N=0}^{3}(1-2N/3)P_N, which reduces to 1−2⟨N⟩/31-2\langle N\rangle/3 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 Φ=π\Phi=\pi as a fixed basis. In each ideal block of Eq. (7) the level separation is Ω=(Δ2+4g2)1/2\Omega=(\Delta^2+4g^2)^{1/2}, and the lower levels of H+H_+ and H−H_- have energies (Δ−Ω)/2(\Delta-\Omega)/2 and (−Δ−Ω)/2(-\Delta-\Omega)/2, respectively; their relative slope therefore has magnitude vv. Their weights on ∣++⟩|{+}{+}\rangle and ∣−−⟩|{-}{-}\rangle are (1−Δ/Ω)/2(1-\Delta/\Omega)/2 and (1+Δ/Ω)/2(1+\Delta/\Omega)/2. Projecting a flux detuning and a direct double flip onto this pair gives the interblock matrix element

V−+(Δ)=ΓΦ−2gt2Ω,ΓΦ=gsin⁡[(Φ−π)/4].(16)V_{-+}(\Delta)=\Gamma_\Phi-\frac{2gt_2}{\Omega},\qquad\Gamma_\Phi=g\sin[(\Phi-\pi)/4]. \tag*{(16)}

The factor 2g/Ω2g/\Omega arises because the direct flip couples both pairs of opposite configurations. At zero bias V−+V_{-+} reduces to Γ\Gamma of Eq. (10). Near the second crossing the same construction applies with ΓΦ=−gsin⁡[(Φ−3π)/4]\Gamma_\Phi=-g\sin[(\Phi-3\pi)/4] and Γ=(E2−E3)/2\Gamma=(E_2-E_3)/2.

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 2πg2/v≫12\pi g^2/v\gg1 and 2πΓ2/v≪12\pi\Gamma^2/v\ll1. Linearizing Γ\Gamma at JgJ_g gives the full width for P2≤PcP_2\le P_c,

w=2∣∂JΓ∣v2πln⁡11−Pc.(17)w=\frac{2}{|\partial_J\Gamma|}\sqrt{\frac{v}{2\pi}\ln\frac{1}{1-P_c}}. \tag*{(17)}

At Jg1J_{g1}, ∣∂JΓ∣=0.0685|\partial_J\Gamma|=0.0685, and with Pc=0.01P_c=0.01 Eq. (17) gives w=3.39×10−3w=3.39\times10^{-3} at T=2.4×104T=2.4\times10^4, inside the microscopic grid bracket [3.2, 3.6] ×10−3\times10^{-3}. For the shortest sweep, T=6×103T=6\times10^3, the predicted width 6.77×10−36.77\times10^{-3} lies just below the microscopic bracket [6.8, 7.2] ×10−3\times10^{-3} Figure 4(b).

The curve in Figure 1(d) retains the bias dependence of Eq. (16). With t=0t=0 at the crossing, the lower-level energy difference is −vt-vt, and integrating V−+(vt)e−ivt2/2V_{-+}(vt)e^{-ivt^2/2} over an unbounded sweep gives, to first order and up to an overall phase,

Ac=2πv[ΓΦ−t2r(ξ)],\mathcal{A}_c=\sqrt{\frac{2\pi}{v}}\left[\Gamma_\Phi-t_2r(\xi)\right],
r(ξ)=−iπ2ξei(ξ2+π/4)H0(2)(ξ2),(18)r(\xi)=-i\sqrt{\frac{\pi}{2}}\xi e^{i(\xi^2+\pi/4)}H_0^{(2)}(\xi^2), \tag*{(18)}

where H0(2)H_0^{(2)} is the Hankel function of the second kind and r→1r \to1 as ξ→∞\xi\to\infty. Figure 1(d) uses 1−exp⁡(−∣Ac∣2)1 - \exp(-|\mathcal{A}_{c}|^2), a continuation of the weak-transfer result that recovers Eq. (11) in this limit.

A second amplitude arises near the endpoint. For Δh=2qh≫g\Delta_h = 2qh \gg g, the two predominantly single-reversal levels have separation 2g2/Δ2g^2/\Delta and coupling −t2-t_2, so their mixing angle at the positive endpoint is βh=12arctan⁡(t2Δh/g2)\beta_h = \frac{1}{2}\arctan(t_2\Delta_h/g^2). To first order in t2t_2, integrating the relative phase 2ξ2ln⁡(Δ/Δh)2\xi^2\ln(\Delta/\Delta_h) gives an admixture b≃βh2iξ2/(1+2iξ2)b \simeq\beta_h 2i\xi^2/(1 + 2i\xi^2). Its overlap with ∣−−⟩|--\rangle, together with the direct flux contribution, yields the endpoint amplitude, up to the relative propagation phase,

Ae≃t22g12iξ21+2iξ2−ΓΦΔh.(19)\mathcal{A}_e \simeq\frac{t_2}{2g_1}\frac{2i\xi^2}{1 + 2i\xi^2} - \frac{\Gamma_\Phi}{\Delta_h}. \tag*{(19)}

Near the first crossing the two amplitudes interfere, and their relative dynamical phase is sensitive to the sweep duration: at T=2.4×104T = 2.4 \times10^4 it changes by about 2π2\pi for a 1%1\% change in TT. The lines in Figure 4(a) therefore show their phase-averaged probability. At Jg1J_{\mathrm{g1}} the two extreme relative phases give P2≃8.0×10−6P_2 \simeq8.0 \times10^{-6} and 5.7×10−55.7 \times10^{-5}, bracketing the microscopic value 3.07×10−53.07 \times10^{-5}. The same phase dependence allows Jmin⁡(T)J_{\min}(T) to move to either side of Jg1J_{\mathrm{g1}}.

Phase pulses in the dimer and trimer

During each pulse the uniform bias is held at its endpoint. In the dimer, the pulse adds +εp+\varepsilon_p to cavity 1 and −εp-\varepsilon_p to cavity 2; for elapsed pulse time 0≤s≤τ0 \le s \le\tau we take

εp(s)=φ2qlocτsin⁡2(πs/τ).(20)\varepsilon_p(s) = \frac{\varphi}{2q_{\mathrm{loc}}\tau}\sin^2(\pi s/\tau). \tag*{(20)}

The opposite shifts of ∣−+⟩|{-+}\rangle and ∣+−⟩|{+-}\rangle give the relative phase φ′=(q/qloc)φ\varphi' = (q/q_{\mathrm{loc}})\varphi, with q/qloc=1.0039q/q_{\mathrm{loc}} = 1.0039 at Jg1J_{\mathrm{g1}}. At ideal half flux the pulse populates ∣1+⟩|1+\rangle and ∣1−⟩|1-\rangle with amplitudes cos⁡(φ′/2)\cos(\varphi'/2) and −sin⁡(φ′/2)-\sin(\varphi'/2), and only ∣1−⟩|1-\rangle connects to ∣−−⟩|--\rangle on the return sweep.

Two effects set the fringe envelope at finite TT. Each passage fails with probability e−2πξ2e^{-2\pi\xi^2}, giving Pmax⁡≃1−2e−2πξ2P_{\max} \simeq1 - 2e^{-2\pi\xi^2} to leading order. In addition, the single-reversal state lags behind the endpoint rotation by ∣α∣=∣βh−b∣≃∣βh∣/1+4ξ4|\alpha| = |\beta_h - b| \simeq|\beta_h|/\sqrt{1 + 4\xi^4}, 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),

P2est(φ)=Pmax⁡[sin⁡2(φ/2)+2∣α∣2cos⁡2(φ/2)].(21)P_2^{\mathrm{est}}(\varphi) = P_{\max}\left[\sin^2(\varphi/2) + 2|\alpha|^2\cos^2(\varphi/2)\right]. \tag*{(21)}

Here ∣α∣=8.3×10−3|\alpha| = 8.3 \times10^{-3}. For the dephased control, we apply ϕ↦∑aPaϕPa\phi\mapsto\sum_a P_a\phi P_a just before the pulse, using the four bias dressed-configuration projectors and their complement; the final dressed readout uses the instantaneous basis at −h-h.

In the trimer, let nj=0,1n_j = 0,1 record whether the field in cavity jj is reversed, and let FjF_j flip that bit. The Jordan–Wigner operators cj=(∏l<j(−1)nl)Fjnjc_j = (\prod_{l<j}(-1)^{n_l})F_j n_j then anticommute, and the strings give a product of hopping signs of −1-1 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 2.2%2.2\% between adjacent shells; this shell dependence preserves the selection of f0f_0, and direct flips of two or three fields give only small corrections.

The trimer pulse adds to cavity jj the bias

δεj(s)=[ϕjqlocτ]sin⁡2(πs/τ),0≤s≤τ,\delta\varepsilon_j(s) = \left[\frac{\phi_j}{q_{\mathrm{loc}}\tau}\right]\sin^2(\pi s/\tau), \qquad0 \le s \le\tau,

where ϕj=2π(j−2)/3\phi_j = 2\pi(j - 2)/3 for the forward gradient. A reversed field in cavity jj thus acquires the phase 2qloc∫0τδεj(s) ds=ϕj2q_{\mathrm{loc}}\int_{0}^{\tau}\delta\varepsilon_j(s)\,\mathrm{d}s=\phi_j relative to the unreversed configuration, and the pulse acts as

cj†↦eiϕjcj†,fk†↦e−4πi/3fk+1†.c_j^\dagger\mapsto e^{i\phi_j}c_j^\dagger,\qquad f_k^\dagger\mapsto e^{-4\pi i/3}f_{k+1}^\dagger.

The opposite gradient shifts kk to k−1k-1 instead, which produces the second outcome in Figure 3(c).

References

References

  1. [1]L. D. Landau, Zur Theorie der Energieübertragung. II, Phys. Z. Sowjetunion 2, 46 (1932). Phys. Rev. 51, 652 (1937).
  2. [2]C. Zener, Non-adiabatic crossing of energy levels, Proc. R. Soc. London A 137, 696 (1932).DOI
  3. [3]E. C. G. Stueckelberg, Theorie der unelastischen Stösse zwischen Atomen, Helv. Phys. Acta 5, 369 (1932).
  4. [4]E. Majorana, Atomi orientati in campo magnetico variabile, Nuovo Cimento 9, 43 (1932).DOI
  5. [5]S. N. Shevchenko, S. Ashhab, and F. Nori, Landau–Zener–Stückelberg interferometry, Phys. Rep. 492, 1 (2010).DOI
  6. [6]T. Kadowaki and H. Nishimori, Quantum annealing in the transverse Ising model, Phys. Rev. E 58, 5355 (1998).DOI
  7. [7]T. Albash and D. A. Lidar, Adiabatic quantum computation, Rev. Mod. Phys. 90, 015002 (2018).DOI
  8. [8]H. Goto, Bifurcation-based adiabatic quantum computation with a nonlinear oscillator network, Sci. Rep. 6, 21686 (2016).
  9. [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. [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. [11]J. von Delft and C. L. Henley, Destructive quantum interference in spin tunneling problems, Phys. Rev. Lett. 69, 3236 (1992).DOI
  12. [12]A. Garg, Topologically quenched tunnel splitting in spin systems with Kramers’ degeneracy, Europhys. Lett. 22, 205 (1993).
  13. [13]W. Wernsdorfer and R. Sessoli, Quantum phase interference and parity effects in magnetic molecular clusters, Science 284, 133 (1999).
  14. [14]S. Brundobler and V. Elser, S-matrix for generalized Landau–Zener problem, J. Phys. A 26, 1211 (1993).DOI
  15. [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. [16]N. A. Sinitsyn, Solvable four-state Landau–Zener model of two interacting qubits with path interference, Phys. Rev. B 92, 205431 (2015).
  17. [17]V. Y. Vyacherny, A. N. Sinitsyn, and C. Sun, Dynamic spin localization and γ-magnets, Phys. Rev. B 100, 224304 (2019).
  18. [18]J. Vidal, R. Mosseri, and B. Douçot, Aharonov–Bohm cages in two-dimensional structures, Phys. Rev. Lett. 81, 5888 (1998).DOI
  19. [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. [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. [21]I. I. Rabi, Space quantization in a gyrating magnetic field, Phys. Rev. 51, 652 (1937).DOI
  22. [22]D. Braak, Integrability of the Rabi model, Phys. Rev. Lett. 107, 100401 (2011).pubmed.ncbi.nlm.nih.gov/21981483
  23. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [34]J. Zhao and M.-J. Hwang, Frustrated superradiant phase transition, Phys. Rev. Lett. 128, 163601 (2022).DOI
  35. [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. [36]J. Dalibard, F. Gerbier, G. Juzeliūnas, and P. Öhberg, Colloquium: Artificial gauge potentials for neutral atoms, Rev. Mod. Phys. 83, 1523 (2011).
  37. [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. [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. [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. [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. [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. [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. [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. [44]Y. Aharonov and D. Bohm, Significance of electromagnetic potentials in the quantum theory, Phys. Rev. 115, 485 (1959).DOI
  45. [45]B. C. Sanders, Entangled coherent states, Phys. Rev. A 45, 6811 (1992).pubmed.ncbi.nlm.nih.gov/9907804
  46. [46]J. Joo, W. J. Munro, and T. P. Spiller, Quantum metrology with entangled coherent states, Phys. Rev. Lett. 107, 083601 (2011).DOI
  47. [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. [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. [49]M. O. Scully, B.-G. Englert, and H. Walther, Quantum optical tests of complementarity, Nature 351, 111 (1991).
  50. [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. [51]P. Jordan and E. Wigner, Über das Paulische Äquivalenzverbot, Z. Phys. 47, 631 (1928).
  52. [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. [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).

Paper details

vlnc:W_rtarx8hd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Codex

View statement
Looking for related papers…
Contents