Global Adiabatic Criterion

• Global Adiabatic Criterion is a framework that replaces pointwise adiabatic conditions with a global measure based on mean and variance of nonadiabatic factors.
• It mathematically bounds infidelity through inequalities involving the minimum energy gap and distributions of nonadiabaticity, optimizing evolution protocols.
• Practical applications in photonic SSH arrays, Fock-state lattices, and Rydberg synthetic lattices demonstrate its role in designing fast, robust, and compact adiabatic devices.

The Global Adiabatic Criterion (GAC) denotes a class of global fidelity conditions for finite-time adiabatic dynamics in which the relevant control variable is not the pointwise maximum of instantaneous nonadiabatic coupling, but a protocol-wide measure built from the mean and variance, or equivalently the root-mean-square value, of a nonadiabatic factor. In recent work on photonic topological pumping, Fock-state lattices, and Rydberg synthetic lattices, GAC is used to bound infidelity or leakage and to design faster adiabatic protocols by optimizing the temporal or spatial distribution of nonadiabaticity rather than forcing uniformly slow evolution everywhere (Xiao et al., 29 Dec 2025, Wu et al., 2 Jun 2026, Guo et al., 8 Jul 2026). Earlier analysis of adiabatic passage did not use the term by name, but it emphasized that high-fidelity state-to-state transfer is a separate criterion from strict instantaneous eigenstate following, thereby providing an important conceptual precursor (Benseny et al., 2020).

1. From instantaneous adiabaticity to global fidelity

The conventional adiabatic condition is local. In the language used for adiabatic passage, “the non-adiabatic coupling due to the time dependence of the eigenstate basis should be much weaker than the Bohr frequencies between the energy eigenstates,” or, equivalently,

$$\left|\langle n(t)|\dot m(t)\rangle\right| \ll |E_n(t)-E_m(t)|.$$

This criterion constrains the evolution at each instant and is therefore a worst-case condition on the entire trajectory (Benseny et al., 2020).

The later GAC literature argues that this local condition is often too restrictive for finite-time control. In the photonic SSH setting, the standard instantaneous adiabatic condition is described as a pointwise requirement that the nonadiabatic coupling from the target state to every other mode be small for all positions $z\in[0,L]$, but the paper explicitly replaces this by a global fidelity bound (Xiao et al., 29 Dec 2025). In the Fock-state-lattice formulation, the same shift is made in time rather than space: the central quantity is the total infidelity accumulated over the full protocol, not an instantaneous violation at a single time (Wu et al., 2 Jun 2026).

The 2020 reanalysis of adiabatic passage provides the same physical message in a different form. It shows that the exact evolving state may deviate from the adiabatic eigenstate at intermediate times, that the deviation can reach a maximum in the middle of the protocol, and that it can then decrease significantly toward the end. The relevant adiabaticity criterion for many applications is therefore “high-fidelity state-to-state transfer,” not strict pointwise adiabatic following (Benseny et al., 2020).

2. Mathematical structure of the criterion

In its most explicit form, GAC is a global inequality involving a minimum gap and the statistics of a nonadiabatic factor. For accelerated topological pumping in photonic SSH waveguide arrays, the central inequality is

$$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$

where $L$ is the total evolution length, $\Delta E_{\min}$ is the minimum gap over the entire evolution, $\overline{Q}_{\rm non}$ is the spatial mean of the total nonadiabaticity factor, $\sigma_Q^2$ is its variance, and $\varepsilon_c=1-\mathcal F_c$ is the target infidelity threshold. The same work also introduces the fluctuation-suppression acceleration criterion

$\sigma_Q \ll \overline{Q}_{\rm non},$

which requires spatial smoothness of nonadiabaticity and permits a larger mean nonadiabaticity while preserving fidelity (Xiao et al., 29 Dec 2025).

For fast topological photon transfer in Fock-state lattices, the nonadiabaticity factor is written as

$$Q(t)= \frac{|\langle\psi_{\text{bright}}|\partial_t\psi_{\text{dark}}\rangle|}{\Delta E(t)} = \frac{\sqrt{N}|\dot\theta(t)|}{\sqrt2G(t)},$$

with

$z\in[0,L]$0

Its temporal mean and variance are

$z\in[0,L]$1

and the sufficient condition takes the form

$z\in[0,L]$2

The infidelity upper bound therefore scales as $z\in[0,L]$3 (Wu et al., 2 Jun 2026).

In the Rydberg synthetic Lieb lattice, GAC is constructed from a representative three-level $z\in[0,L]$4-type adiabatic passage. The local nonadiabatic factor is

$z\in[0,L]$5

with

$z\in[0,L]$6

Its mean and variance over the transfer window $z\in[0,L]$7 are

$z\in[0,L]$8

and the resulting conservative estimate of leakage is

$z\in[0,L]$9

In this formulation, the global character of GAC lies in the simultaneous control of the mean nonadiabatic burden and its temporal fluctuation (Guo et al., 8 Jul 2026).

3. Uniformity, fluctuation suppression, and endpoint behavior

The recent GAC papers identify fluctuation suppression, rather than mere gap preservation, as the decisive mechanism for fast adiabatic transfer. In photonic SSH pumping, the fluctuation-suppression acceleration criterion is introduced precisely because the root-mean-square nonadiabaticity, not only its mean, governs the fidelity bound. If fluctuations are sufficiently suppressed, the GAC reduces to a simpler inequality involving the mean nonadiabaticity alone, allowing shorter devices and faster modulation (Xiao et al., 29 Dec 2025).

The Fock-state-lattice analysis makes the same point more sharply. It states that the key to fast transfer is not a constant energy gap but the vanishing nonadiabaticity variance. For the power-law coupling family

$$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$0

the variance vanishes only for the sinusoidal shape $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$1, so that $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$2 is constant and $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$3. The paper then constructs an alternative constant-gap family and shows that a constant gap alone is not sufficient for fast topological photon transfer; the essential condition is uniformity of nonadiabaticity (Wu et al., 2 Jun 2026).

An allied mechanism appears in the 2020 adiabatic-passage analysis. There, the first-order deviation from the adiabatic basis is $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$4 during the evolution, but if the Hamiltonian changes smoothly and its rate vanishes near the beginning and end, the endpoint error can be $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$5 or smaller. The explicit endpoint requirement is

$$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$6

This suggests that global adiabatic performance depends not only on how large nonadiabaticity becomes, but also on how it is distributed and extinguished across the full protocol (Benseny et al., 2020).

4. Realizations in photonic and Fock-state topological pumping

Two recent implementations make GAC a concrete device-design principle rather than only a theoretical bound.

Platform	GAC design variable	Reported outcome
Photonic SSH waveguide arrays (Xiao et al., 29 Dec 2025)	Power-law coupling modulation with $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$7	$$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$8 versus $$L\,\Delta E_{\min}\sqrt{\overline{Q}_{\rm non}^2+\sigma_Q^2}\le \sqrt{\varepsilon_c},$$9 for the conventional scheme; $L$0 fidelity; $L$1; robust across a $L$2 nm bandwidth
Fock-state lattices in a two-mode Jaynes–Cummings system (Wu et al., 2 Jun 2026)	Sinusoidal profile within the power-law family, where $L$3	Predicted optimal transfer duration $L$4 ns for a five-photon state, far shorter than $L$5 ns used in the experiment, reducing time by over $L$6 while increasing transferred photons by $L$7

In the photonic SSH experiment, the accelerated design replaces the conventional smooth cosine modulation by scalable power-law coupling modulation. The central result is a fivefold reduction in the device length required for $L$8, together with approximately linear scaling in system size and broadband robustness from $L$9 nm to $\Delta E_{\min}$0 nm, experimentally tested at $\Delta E_{\min}$1 nm, $\Delta E_{\min}$2 nm, $\Delta E_{\min}$3 nm, and $\Delta E_{\min}$4 nm (Xiao et al., 29 Dec 2025).

In the Fock-state-lattice problem, GAC serves a different role. It explains why the previously demonstrated sinusoidal protocol was unusually fast. The paper argues that the earlier intuition based on a constant gap was incomplete, and that the true reason is the temporal uniformity of $\Delta E_{\min}$5. The result is both explanatory and prescriptive: it yields a practical duration estimate in the presence of decoherence and a simple linear scaling with photon number (Wu et al., 2 Jun 2026).

5. Rydberg synthetic Lieb lattices and non-Abelian pumping

In the Rydberg synthetic-lattice implementation, GAC is used to design non-Abelian Thouless pumping in a finite Lieb lattice encoded in twelve selected microwave-coupled Rydberg levels. These levels form a three-cell structure with six degenerate zero-energy states, which define the working subspace for cyclic modulation, while the remaining bright states provide the dominant leakage channels at finite evolution time (Guo et al., 8 Jul 2026).

The construction proceeds by reducing each active pumping segment to a representative $\Delta E_{\min}$6-type transfer. For the first step of $\Delta E_{\min}$7, the relevant basis is $\Delta E_{\min}$8, with local Hamiltonian

$\Delta E_{\min}$9

and dark state

$\overline{Q}_{\rm non}$0

Within a Gaussian pulse family,

$\overline{Q}_{\rm non}$1

the delay parameter $\overline{Q}_{\rm non}$2 is scanned, and $\overline{Q}_{\rm non}$3 is identified as a favorable compromise among low mean nonadiabaticity, reduced temporal fluctuation, and high target-state population (Guo et al., 8 Jul 2026).

The selected timing is then transferred to the full twelve-level pumping dynamics. Composing two elementary pumping cycles in opposite temporal orders produces distinct projected population maps, exactly consistent with noncommuting matrix-valued adiabatic operations in the zero-energy subspace. Numerical simulation is performed with the Lindblad master equation including state-dependent Rydberg loss and representative perturbations. Within the same Gaussian pulse family, the GAC-selected timing gives higher target-state population than two literature-adapted Gaussian pulse schedules over the simulated parameter ranges, including scans with global amplitude errors, detuning gradients, static on-site disorder, static coupling disorder, and Rydberg decay (Guo et al., 8 Jul 2026).

A central clarification in this work is that GAC is not presented as a shortcut-to-adiabaticity scheme that adds new couplings. It is a timing-design rule for fast but still sufficiently adiabatic evolution inside a fixed pulse family (Guo et al., 8 Jul 2026).

6. Scope, ambiguity, and limits of the term

The supplied literature does not support a single, historically uniform meaning of “Global Adiabatic Criterion.” Some papers clearly formulate GAC as a named criterion with explicit bounds on fidelity or leakage (Xiao et al., 29 Dec 2025, Wu et al., 2 Jun 2026, Guo et al., 8 Jul 2026). Other papers discuss global or endpoint notions of adiabaticity without introducing the term. The global-AdS study interprets a maximal slope $\overline{Q}_{\rm non}$4 as an adiabaticity threshold for quasistatic tracking, but it explicitly does not formulate a formal “Global Adiabatic Criterion” as a theorem or universal principle (Carracedo et al., 2016). The SrTiO$\overline{Q}_{\rm non}$5 tunneling study uses the standard adiabatic condition

$\overline{Q}_{\rm non}$6

as a materials-scale organizing parameter and likewise does not introduce GAC as a named formal principle (Yoon et al., 2021).

There is also direct acronymic ambiguity. In mobile sensing, “GAC” denotes “GPS + Accelerometer + Compass,” an energy-efficient hybrid localization scheme unrelated to adiabatic dynamics or fidelity bounds (Youssef et al., 2010). For this reason, the meaning of GAC is field-dependent and cannot be inferred from the acronym alone.

A common misconception is therefore to treat GAC as a single system-independent law already standardized across disciplines. The available evidence is narrower. The recent topological and Rydberg papers present GAC as a universal design rule for fast, compact, and robust adiabatic devices, but the broader literature shows that the term remains localized to particular control problems and that adjacent papers often advance only GAC-like ideas rather than the formal criterion itself (Xiao et al., 29 Dec 2025). This suggests that GAC is presently best understood as a transferable framework for finite-time adiabatic design, centered on global control of mean nonadiabaticity and fluctuation, rather than as a universally codified theorem across all adiabatic phenomena.