Nonreciprocal Landau-Zener Tunneling

• Nonreciprocal Landau-Zener tunneling is a class of quantum transitions where reversing the field or sweep direction yields unequal tunneling probabilities.
• It arises from geometric corrections, interference effects, and non-Hermitian dynamics that modify both the tunneling exponent and phase accumulation.
• These phenomena enhance our understanding of quantum geometry, exceptional points, and nonlinear band dynamics across diverse physical platforms.

Nonreciprocal Landau-Zener tunneling denotes a class of Landau-Zener transitions in which reversal of the sweep direction, electric field, or source-to-target channel does not leave the tunneling probability or its transport consequences invariant. In contemporary usage, the term covers several distinct mechanisms rather than a single universal model: geometric corrections from the shift vector in noncentrosymmetric crystals, interference among repeated tunnelings during Bloch oscillations under strong DC fields, exceptional-point-controlled dynamics in non-Hermitian systems with unequal off-diagonal couplings, direction-dependent tunneling in magnetic lattices, and interaction-induced asymmetry in nonlinear two-level models (Kitamura et al., 2019, Terada et al., 2024, Wang et al., 2022, Cheng et al., 4 Aug 2025).

1. Standard Landau-Zener theory and the meaning of nonreciprocity

For the standard reciprocal two-level problem,

$$H(t)=\frac12 \begin{pmatrix} vt & \Delta \ \Delta & -vt \end{pmatrix},$$

the transition probability is the usual Landau-Zener expression. In the solid-state formulation under a DC electric field, one equivalent form is

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$

with $\Delta$ the minimal gap and $v$ the speed of passage through the avoided crossing in $k$-space (Kitamura et al., 2019). In this reciprocal setting, the transition is symmetric under reversal of the sweep protocol in the absence of additional symmetry-breaking ingredients.

Nonreciprocity enters when this symmetry is lost. Depending on the system, the asymmetry may appear as $P(+E)\neq P(-E)$, as different forward and backward sweep probabilities, or as unequal probabilities for opposite transfer channels such as $P_{L\to R}\neq P_{R\to L}$. The mechanism can be geometric, spectral, dissipative, nonlinear, or interference-based rather than purely kinematic.

Setting	Microscopic source of asymmetry	Characteristic signature
Noncentrosymmetric insulator	Shift vector and Berry-connection difference	$P(+E)\neq P(-E)$
Strong-field Bloch dynamics	Multi-tunneling interference with geometric phase	Oscillatory nonreciprocal DC current
Non-Hermitian two-level system	Unequal couplings and exceptional points	Finite adiabatic tunneling
Magnetic lattice in Hall configuration	Gauge-field-induced direction dependence	Transverse current from LZ events
Nonlinear droplet model	Population-dependent gap modulation	$P_{L\to U}\neq P_{U\to L}$
Three-state impurity-mediated transfer	Left-right tunneling asymmetry	$P_{L\to R}\neq P_{R\to L}$
Time-varying circuit analog	Nonreciprocal coupling without spectral change	Same LZ probability as reciprocal case

These cases show that “nonreciprocal Landau-Zener tunneling” is a family of phenomena rather than a single theorem: some realizations modify the tunneling exponent itself, some alter the phase between repeated events, and some change the adiabatic topology through exceptional points or looped bands (Terada et al., 2024, Wang et al., 2022, Maksimov et al., 2013, Cheng et al., 4 Aug 2025, Raikh, 2022, Cheng et al., 28 Nov 2025).

2. Geometric origin in noncentrosymmetric band structures

The clearest geometric formulation arises in noncentrosymmetric crystals, where a DC field drives interband tunneling and the Bloch wavefunctions carry a gauge-invariant shift vector. In the adiabatic basis, the relevant geometric objects are the intraband Berry connection

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$0

and the shift vector

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$1

with

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$2

Within the Dykhne-Davis-Pechukas treatment, the generalized tunneling probability for a single interband process is

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$3

The first contribution is the usual dynamical Landau-Zener exponent, while the second is a purely geometric correction. Because the shift-vector contribution changes sign under field reversal, the theory yields $$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$4 in noncentrosymmetric crystals (Kitamura et al., 2019).

This formulation gives a concrete physical interpretation of nonreciprocity. The shift vector measures the intracell coordinate shift accompanying an interband transition. In the real-space picture presented for the same theory, tunneling through a classically forbidden region of thickness $$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$5 is modified to an effective thickness $$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$6, so the exponent becomes direction dependent when the field is reversed. The asymmetry is therefore not introduced phenomenologically; it is encoded in the geometry of the Bloch states themselves.

The one-dimensional Rice-Mele ferroelectric model provides an explicit example. For

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$7

the asymptotic tunneling probability in the limit $$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$8 is

$$P_{LZ}=\exp\!\left[-\frac{\pi \Delta^2}{e\hbar E v}\right],$$9

The field-even term reproduces the conventional Zener exponent, whereas the field-odd exponential factor is the nonreciprocal correction generated by the shift vector. In this sense, nonreciprocal Landau-Zener tunneling in noncentrosymmetric insulators is a quantum-geometric transport effect rather than merely an asymmetric avoided crossing (Kitamura et al., 2019).

3. Multi-tunneling interference under strong DC fields

A more elaborate strong-field regime arises when an electron repeatedly traverses the Brillouin zone and experiences multiple Landau-Zener events during Bloch oscillations. In the two-band description

$\Delta$0

a static electric field is introduced through the Peierls substitution $\Delta$1, so successive avoided crossings are linked by coherent evolution across a Bloch cycle. In this setting, the central effect is not only the probability of a single tunneling event, but also the interference between different tunneling paths (Terada et al., 2024).

For two successive crossings, the total lower-band return amplitude is written as

$\Delta$2

where

$\Delta$3

is the dynamical phase accumulated between crossings and $\Delta$4 is a geometric shift phase. In the formulation of the nonequilibrium steady state,

$\Delta$5

so the average shift vector over the Bloch cycle enters directly into the interference condition.

This produces an oscillatory occupation of the upper band. To lowest order in the tunneling probability $\Delta$6 and for $\Delta$7,

$\Delta$8

The resulting intraband DC current acquires an oscillatory correction

$\Delta$9

Because $v$0 when the Brillouin-zone-averaged shift vector is nonzero, the positions of the oscillatory current peaks differ for positive and negative fields. The paper identifies this as a strongly nonreciprocal DC response generated by multi-tunneling interference rather than by a single isolated LZ event (Terada et al., 2024).

This regime sharpens the distinction between two notions of nonreciprocity. In the single-crossing theory, the shift vector modifies the tunneling exponent itself. In the strong-field Bloch-oscillation regime, the same geometric quantity also controls the phase relation between successive tunneling amplitudes. The nonreciprocity then appears in the field dependence of interference maxima and minima, and the response is amplified with increasing electric-field intensity (Terada et al., 2024).

4. Non-Hermitian couplings, exceptional points, and adiabatic breakdown

A second major lineage of nonreciprocal Landau-Zener tunneling is based on non-Hermitian two-level systems with unequal off-diagonal couplings. In the linear model with time-dependent bias $v$1,

$v$2

the parameter $v$3 measures nonreciprocity. The instantaneous eigenvalues are

$v$4

For $v$5, the spectrum develops exceptional points at

$v$6

where eigenvalues and eigenvectors coalesce (Wang et al., 2022).

The adiabatic limit in this non-Hermitian problem differs qualitatively from the Hermitian Landau-Zener theorem. A state can closely follow an instantaneous eigenstate until an exceptional point is encountered, but adiabatic tunneling need not vanish as $v$7. In the linear theory, the adiabatic tunneling probabilities take the values $v$8 and $v$9 depending on the initial state and sweep direction once the evolution crosses the relevant exceptional point. By contrast, for $k$0 no exceptional point is crossed in the linear model and the adiabatic tunneling probability remains zero (Wang et al., 2022).

Adding nonlinear self-interaction produces a richer structure. With

$k$1

the dynamics can be mapped onto a classical Josephson Hamiltonian in canonical variables $k$2 and $k$3. The nonlinear eigenstates appear as fixed points; for sufficiently large $k$4 and weak nonreciprocity, a swallow-tail loop of four real eigenvalues emerges around $k$5, bounded by nonlinear exceptional points. In that regime, the adiabatic tunneling probability is controlled by the area

$k$6

enclosed by the homoclinic orbit, yielding

$k$7

For strong nonreciprocity, the nonlinear interaction can be completely suppressed in the adiabatic probabilities over the parameter ranges $k$8 with $k$9, or $P(+E)\neq P(-E)$0 with $P(+E)\neq P(-E)$1, where the linear exceptional-point values are exactly recovered (Wang et al., 2022).

Colored noise adds a further extension. In the stochastic non-Hermitian model, the level bias is perturbed by an Ornstein-Uhlenbeck process with

$P(+E)\neq P(-E)$2

The noisy system exhibits stochastic fluctuations of instantaneous levels, and sufficiently large nonreciprocity can trigger multiple exceptional points. The detailed analysis shows that colored noise generally enhances tunneling in the nonadiabatic regime, can break reciprocity of Landau-Zener tunneling in the adiabatic limit for $P(+E)\neq P(-E)$3, and becomes ineffective in the adiabatic limit for large nonreciprocity $P(+E)\neq P(-E)$4, where the tunneling probability is set by exceptional points and is insensitive to noise (Cao et al., 24 May 2026).

5. Platform-dependent realizations and contrasts

In a square lattice with a perpendicular gauge field, Landau-Zener transitions beyond the single-band approximation acquire a direction dependence through the magnetic flux. In the supplied formulation, the magnetic field modifies both the effective gap and the sweep rate, so that

$P(+E)\neq P(-E)$5

For vanishing gauge field, repeated interband tunneling is associated with electric breakdown. In the Hall configuration, this breakdown is absent in its common sense; instead, Landau-Zener tunneling produces a finite current orthogonal to the electric field, with

$P(+E)\neq P(-E)$6

Here nonreciprocity is tied to the gauge field and transverse Hall drift rather than to non-Hermitian coupling or Berry-shift asymmetry alone (Maksimov et al., 2013).

The time-varying electric-circuit analog provides a contrasting result. Two RLC channels with time-dependent capacitances and nonreciprocal coupling $P(+E)\neq P(-E)$7 can be mapped, after linearization and block-diagonalization, onto two nearly decoupled quantum Landau-Zener models. Near the crossing, the effective circuit Hamiltonian yields a gap

$P(+E)\neq P(-E)$8

and sweep rate

$P(+E)\neq P(-E)$9

The final remaining probability obeys

$P_{L\to R}\neq P_{R\to L}$0

and identical results hold for reciprocal and nonreciprocal couplings because the instantaneous spectrum does not depend on $P_{L\to R}\neq P_{R\to L}$1 and the nonreciprocal Liouvillian is similar to the reciprocal one. This case is important because it shows that nonreciprocal coupling need not imply a nonreciprocal tunneling probability (Cheng et al., 28 Nov 2025).

In nonlinear Bose-Einstein quantum droplets in shallow optical lattices, the asymmetry originates from interaction-modulated band structure. Near the Brillouin-zone edge, the dynamics reduce to

$P_{L\to R}\neq P_{R\to L}$2

with $P_{L\to R}\neq P_{R\to L}$3 and effective nonlinearity

$P_{L\to R}\neq P_{R\to L}$4

For $P_{L\to R}\neq P_{R\to L}$5, the adiabatic spectrum develops looped bands and multiple fixed points in the classical Josephson phase space. Hamilton-Jacobi analysis then yields different tunneling probabilities for lower-to-upper and upper-to-lower sweeps; the asymmetry is traced to self-interaction-driven modulation of the effective minimum gap, so that $P_{L\to R}\neq P_{R\to L}$6 even under the same sweeping protocol (Cheng et al., 4 Aug 2025).

A related multi-state extension appears in the impurity-mediated transfer between two quantum dots. With unequal dot-impurity tunnel couplings $P_{L\to R}\neq P_{R\to L}$7 and $P_{L\to R}\neq P_{R\to L}$8 and impurity energy $P_{L\to R}\neq P_{R\to L}$9, the three-state Landau-Zener problem becomes nonreciprocal when $P(+E)\neq P(-E)$0 and $P(+E)\neq P(-E)$1, producing

$P(+E)\neq P(-E)$2

The asymmetry disappears in the resonant “bow-tie” limit $P(+E)\neq P(-E)$3, in the far-detuned cotunneling regime, or when $P(+E)\neq P(-E)$4 (Raikh, 2022).

6. Conceptual distinctions and recurrent misconceptions

A common misconception is that nonreciprocal Landau-Zener tunneling is synonymous with non-Hermitian dynamics. The current literature does not support that identification. Nonreciprocity can be purely geometric, as in noncentrosymmetric crystals where the shift vector changes the tunneling exponent (Kitamura et al., 2019); it can be interference-based, as in strong-field Bloch oscillations where the shift vector controls the phase between successive tunnelings (Terada et al., 2024); it can be generated by a gauge field through direction-dependent magnetic Bloch dynamics (Maksimov et al., 2013); and it can arise from nonlinear interaction-induced gap modulation without any explicit non-Hermiticity (Cheng et al., 4 Aug 2025).

The opposite misconception is equally inaccurate: nonreciprocal couplings do not automatically produce nonreciprocal tunneling probabilities. The electric-circuit analog gives an explicit counterexample. There the coupling matrix is nonreciprocal, but the gap $P(+E)\neq P(-E)$5 and sweep rate $P(+E)\neq P(-E)$6 are unaffected, the Liouvillian is similar to the reciprocal case, and the final Landau-Zener probability is unchanged (Cheng et al., 28 Nov 2025). What matters is not the presence of asymmetry in the equations alone, but whether that asymmetry survives in the spectral, geometric, or interference quantities controlling the transition.

Another recurrent point concerns adiabaticity. In the reciprocal Hermitian problem, slow sweep suppresses tunneling. In several nonreciprocal settings, that expectation fails. Exceptional points produce finite adiabatic tunneling probabilities fixed by invariants or by the sink structure of the dynamics (Wang et al., 2022, Cao et al., 24 May 2026), while looped nonlinear bands generate separatrix crossings and nonzero adiabatic action jumps (Cheng et al., 4 Aug 2025). The adiabatic limit therefore does not, in general, restore reciprocity or eliminate transitions.

This suggests a useful classification of the field into three technical layers. One layer concerns modifications of the local tunneling exponent, as in shift-vector corrections. A second concerns phase accumulation between repeated events, as in multi-tunneling interference under Bloch oscillations. A third concerns changes in the global structure of instantaneous states, as in exceptional points and looped bands. Much of the recent literature can be organized by which of these layers carries the asymmetry. In that sense, nonreciprocal Landau-Zener tunneling serves as a meeting point for quantum geometry, non-Hermitian spectral theory, nonlinear adiabatic dynamics, and strong-field nonequilibrium transport (Terada et al., 2024, Wang et al., 2022).