Bilayer Atomic Arrays: Structure & Function

• Bilayer atomic arrays are two-layer systems with controllable interlayer coupling that defines geometric registry and influences collective modes and quantum responses.
• They span diverse platforms—from crystalline materials to ultracold gases and trapped ions—each tailoring stacking, twist, and reconstruction to achieve specific electronic and mechanical properties.
• Design principles leverage tuning of interlayer interactions, including twist angles and nonlinear density effects, to enable flat-band engineering, improved quantum interfaces, and advanced fabrication methods.

Bilayer atomic arrays are systems in which two atomically resolved, or effectively layer-resolved, arrays are coupled strongly enough that interlayer registry, spacing, symmetry, and collective response become defining variables of the physics. The term spans several distinct but technically connected realizations: crystalline two-dimensional materials such as $h$-BN, silicene, graphene-based moiré structures, and twisted double bilayer graphene; amorphous oxide bilayers such as vitreous silica; synthetic bilayers of ultracold atoms and free-space emitter arrays; and trapped-ion bilayers engineered either by self-organization or by two-plane electrode design (Wilson et al., 2013, Sakai et al., 2015, Tian et al., 2024, Ben-Maimon et al., 2024, Hawaldar et al., 2023). Across these platforms, the central problem is not merely the existence of two layers, but the way in which interlayer coupling selects stable geometries, reorganizes collective modes, modifies transport and optical selection rules, and, in several cases, creates functionality unavailable in a monolayer.

1. Scope and defining features

Bilayer atomic arrays are not restricted to perfectly crystalline, physically separated sheets. In vitreous silica bilayers, two amorphous monolayers of corner-sharing tetrahedra are joined through a shared central oxygen plane; the resulting structure is topologically disordered, yet it develops a global reflection plane and a well-defined bilayer geometry (Wilson et al., 2013). In ultracold gases, a bilayer can also be realized in an effective sense: a two-component Bose-Einstein condensate in spin-dependent optical lattices can behave as two rotated layers, even though the layers are internal-state components of the same gas rather than separate material sheets (Tian et al., 2024).

The common structural ingredients are therefore broader than literal sheet stacking. They include a two-layer degree of freedom, a controllable or emergent interlayer coupling, and collective observables that distinguish symmetric from antisymmetric, commensurate from incommensurate, or coherent from incoherent interlayer motion. This broader definition also clarifies why some architecturally related neutral-atom systems are adjacent to, but not themselves, strict bilayer arrays. The dual-element Rb/Cs array with up to 512 trapping sites is explicitly a same-plane interleaved architecture rather than a true vertically separated bilayer, and the Yb/Rb dual-type proposal is likewise a hybrid dual-array processor rather than a fixed two-plane device (Singh et al., 2021, Zhang et al., 21 Mar 2025). A plausible implication is that bilayer atomic-array design is best understood as a functional category—separated subsystems with controlled interlayer coupling—rather than as a single geometric template.

2. Structural stability, reconstruction, and mechanical degrees of freedom

In weakly bonded crystalline bilayers, stacking energetics and interlayer spacing are controlled by distinct physical mechanisms. For bilayer $h$-BN, periodic local second-order Møller–Plesset perturbation theory identifies AA$'$ as the ground-state stacking, with AB only $0.12\ \text{meV/atom}$ higher in the bilayer, while AA and A$'$B are much less favorable. The equilibrium spacing is $c=3.34\ \text{\AA}$, and the minimum sliding barrier is $3.4\ \text{meV/atom}$. The study attributes stacking preference mainly to electrostatics and short-range chemical effects, while London dispersion is essential for the correct interlayer distance and overall binding (Constantinescu et al., 2013). This establishes a general principle for bilayer atomic arrays: the interaction that fixes vertical cohesion need not be the same interaction that selects lateral registry.

Bilayer silicene shows a different regime, in which idealized stacking labels are not sufficient descriptors of the stable structures. Extensive total-energy minimization produced 24 local minima for $1\times1$, $2\times2$, and $\sqrt{3}\times\sqrt{3}$ cells, but phonon calculations showed that 14 of them were dynamically unstable. The decisive relaxation pattern is atomic protrusion, a collective out-of-plane reconstruction that stabilizes the bilayer and selects the preferred superperiodicity. Only 10 structures were dynamically stable, and the lowest-energy phase was hex-OR-$h$0 at $h$1, with an HSE gap of about $h$2 (Sakai et al., 2015). One common misconception is that bilayer stability can be inferred from stacking names alone; silicene shows instead that reconstruction pattern and periodicity can be the essential variables.

Mechanical bilayering also changes long-wavelength fluctuation physics. Monte Carlo simulations of bilayer graphene found that the effective bilayer bending rigidity is approximately twice the monolayer value in the coherent regime, with a crossover from correlated to uncorrelated out-of-plane fluctuations at wavevectors shorter than about $h$3. The in-plane thermal expansion coefficient changes sign near $h$4, substantially lower than in monolayer graphene, and the out-of-plane expansion coefficient is $h$5 (Zakharchenko et al., 2010). This implies that bilayer atomic arrays generically possess both an in-phase height mode and an out-of-phase thickness mode, with a finite scale beyond which the two layers cease to fluctuate as a single effective membrane.

In twisted double bilayer graphene, reconstruction is controlled by both the inner twisted interface and the stacking character of the two outer bilayers. The inner layers relax much like twisted bilayer graphene, forming enlarged AB/BA domains and shrunken AA regions. The outer layers, however, are not passive: AB-stacked outer bilayers tend to follow the relaxation of the adjacent inner layer, whereas AA-stacked outer bilayers move oppositely to avoid higher-energy AA registry (Liang et al., 2020). This shows that in multilayer bilayer-derived arrays, nearby untwisted layers can actively reshape the moiré-scale atomic texture.

3. Incommensurability, moiré coupling, and local electronic response

A general language for incommensurate bilayer atomic arrays is provided by reciprocal-space coupling rather than supercell construction. Starting from a distance-dependent interlayer tight-binding Hamiltonian, the interlayer matrix element can be written in terms of a generalized Umklapp condition,

$h$6

which states that interlayer hybridization occurs when Bloch components from the two layers share the same true in-plane momentum (Koshino, 2015). This formulation applies to arbitrary lattice structures, lattice mismatch, and arbitrary rotation angles, including large-angle twisted bilayers that do not admit a useful long-wavelength moiré description. Small-angle moiré continuum models then emerge as a special limit rather than a separate theory.

In twisted double bilayer graphene, this momentum-space viewpoint is complemented by a strong stacking dependence of the low-energy bands. All three studied systems—AB/AB, AA/AA, and AB/AA—develop flat bands at small twist angles, but the low-energy manifolds are qualitatively different. AB/AB exhibits an isolated four-band manifold and a magic angle near $h$7; AA/AA instead retains shifted-Dirac-cone character with ultraflat segments but without an isolated four-band sector; AB/AA combines shifted-cone and tBLG-like features and becomes strongly entangled at small twist angles. A further notable result is that rigidly separating the outer layers from the inner twisted pair reveals that hybridization with the outer layers produces additional flattening of the inner-layer flat-band manifold (Liang et al., 2020). This suggests a broader design principle: flat-band engineering in layered atomic arrays can proceed not only by twist-angle control, but also by modifying the electronic role of adjacent non-twisted layers.

At the atomic scale, bilayer registry and layer polarization can also localize nonlinear tunneling phenomena to specific sites. In Bernal bilayer graphene on SiC, scanning tunneling spectroscopy found negative differential resistance near $h$8, but only on the $h$9 site of the top layer. The effect was traced to near-gap van Hove singularities created by a substrate-induced transverse electric field and localized on specific sublattices in specific layers; defects could suppress the effect through intervalley interference (Kim et al., 2013). Bilayer arrays therefore support not only band-topological distinctions between stackings, but also highly local, registry-sensitive nonlinear response.

4. Nonlinear and self-organized ultracold bilayers

In ultracold atoms, bilayer atomic arrays need not be statically imposed by two optical lattices. A two-component Bose-Einstein condensate can realize a nonlinearity-induced dynamical self-organized twisted-bilayer lattice in which one component acts as a quasi-static layer and the other evolves in the interaction-dressed potential

$'$0

The crucial point is that the second layer enters through the density pattern $'$1, so the interlayer coupling is nonlinear and density-generated rather than a linear single-particle tunneling or Raman term (Tian et al., 2024).

This construction supports both commensurate and incommensurate moiré structures. For the commensurate Pythagorean angle $'$2, the effective potential is spatially periodic; for the non-Pythagorean angle $'$3, it becomes aperiodic. The dynamical distinction appears directly in the mean-square displacement of the evolving wave packet,

$'$4

with $'$5 for the commensurate case and $'$6 for the incommensurate case, where $'$7 saturates (Tian et al., 2024). The associated band-structure analysis attributes the localization in the incommensurate case to flattening of the lowest bands.

This platform corrects another common misconception: moiré bilayers are not necessarily static laser-written superpositions. In the Gross–Pitaevskii setting, the bilayer potential is assembled dynamically during time evolution, and the moiré supercell is partly an emergent property of the interacting medium. A plausible implication is that bilayer atomic arrays can be interaction-defined objects as much as geometry-defined ones.

5. Free-space quantum-optical bilayer arrays

Bilayer atomic arrays of emitters support input–output phenomena that are impossible in a single layer because they provide two layer-parity scattering channels. In a square bilayer lattice with atoms at $'$8, the scattered fields from symmetric and antisymmetric collective excitations obey

$'$9

Interference between these even- and odd-parity channels allows a bilayer to absorb a single photon incident from one side only, to store it in subradiant out-of-plane modes, and later to emit it into an arbitrary forward/backward superposition. For a $0.12\ \text{meV/atom}$0 array with $0.12\ \text{meV/atom}$1 and $0.12\ \text{meV/atom}$2, the storage efficiency reaches about $0.12\ \text{meV/atom}$3 for a Gaussian pulse and about $0.12\ \text{meV/atom}$4 for a time-reversal-matched exponential pulse, exceeding the monolayer one-sided limit of $0.12\ \text{meV/atom}$5 (Ballantine et al., 2021).

Free-space bilayers also support nonlocal subradiant states when the two layers are spatially separated rather than closely stacked. Two distant single-layer arrays at separations satisfying $0.12\ \text{meV/atom}$6 support bright and dark parity combinations with decay rates

$0.12\ \text{meV/atom}$7

The dark state is a Bell-like nonlocal excitation shared coherently between the arrays, and for $0.12\ \text{meV/atom}$8 the finite-array scaling is $0.12\ \text{meV/atom}$9 (Guimond et al., 2019). In this setting, bilayering functions as a mechanism for directional free-space coupling, entanglement distribution, and high-fidelity state transfer rather than as a material stacking problem.

A distinct quantum-interface regime appears for superwavelength arrays, where a single layer couples poorly to the target mode because higher diffraction orders are radiative. In a bilayer, unwanted diffraction orders can destructively interfere while the normal-incidence mode remains coupled. For the Bragg-symmetric case, the interface efficiency is

$'$0

and optimized two-layer arrays obey the universal finite-size law $'$1 (Ben-Maimon et al., 2024). A subsequent generalization to non-Bragg spacings showed that the efficiency of a bilayer quantum interface is fully determined by reflection and transmission, and that non-symmetric bilayers can outperform Bragg-constrained designs, with examples showing up to a factor of 5 reduction in inefficiency and exact cancellation of multiple diffraction orders at suitable $'$2 values (Ben-Maimon et al., 15 Apr 2026). Bilayer atomic arrays therefore act as interference-engineered free-space interfaces whose performance is governed by parity, diffraction geometry, and interlayer phase rather than by cavity confinement.

6. Trapped-ion bilayers and reconfigurable quantum-processing architectures

Trapped ions realize bilayer atomic arrays in two complementary ways. One is architectural: a bilayer ion trap design based on two perpendicularly rotoreflected linear surface traps in separate planes creates a trapping region between the planes and allows two-dimensional transport by handing an ion from one layer’s linear channel to the orthogonal channel of the other layer. In a concrete Peregrine-based example with $'$3 plane separation, the RF null of one active trap shifts to $'$4 above that surface, leaving the two nulls $'$5 apart. Stability of the transfer is analyzed through the Mathieu equation and confirmed numerically, while the design avoids the RF-routing complications and pseudopotential bumps of standard planar junctions (Nop et al., 2023).

The second realization is self-organized. In Penning traps, adding a quartic anharmonic electrostatic potential produces clean bilayer Coulomb crystals of $'$6 ions. For $'$7 ions and $'$8, a clean bilayer appears near $'$9, with a large interlayer separation of order $c=3.34\ \text{\AA}$0. The bilayer normal modes differ sharply from monolayer Penning crystals: axial and planar motion no longer decouple, $c=3.34\ \text{\AA}$1 modes acquire nonzero $c=3.34\ \text{\AA}$2-participation, and drumhead modes can become complex and chiral (Hawaldar et al., 2023). The optical-dipole-force phase difference between the two layers then becomes a direct control parameter for interlayer versus intralayer Ising couplings and for complex spin-exchange amplitudes.

Architecturally related neutral-atom work suggests a similar functional separation, even when the geometry is not a literal bilayer. The dual-element Rb/Cs platform demonstrated independent placement of single Rb and Cs atoms in arrays with up to 512 trapping sites and a continuous operation mode without off-time, while the Yb/Rb dual-type proposal uses single Yb atoms as data qubits and small Rb ensembles as ancillas, with simulated single- and multi-qubit operation fidelities of about $c=3.34\ \text{\AA}$3 and $c=3.34\ \text{\AA}$4, and readout fidelities exceeding $c=3.34\ \text{\AA}$5 within tens of microseconds (Singh et al., 2021, Zhang et al., 21 Mar 2025). Although these are not true bilayer demonstrations, they suggest that bilayer atomic arrays may be especially valuable when the two layers are assigned different quantum-information roles.

7. Fabrication, misconceptions, and open directions

Bilayer atomic arrays are also fabrication platforms. In twisted bilayer graphene, an aberration-corrected STEM can define reactive vacancies at chosen lattice positions while nearby source nanoparticles provide diffusing adatoms. Under these conditions, the electron beam creates the attachment sites and the bilayer surface captures foreign atoms spontaneously, enabling atomic patterning of circles and arrays with limited human interaction. The mean dose required to obtain an outcome in twisted bilayer graphene was $c=3.34\ \text{\AA}$6, much lower than the average dose above $c=3.34\ \text{\AA}$7 previously required to mill a hole in monolayer graphene, which the authors interpret as evidence that the bilayer suppresses vacancy migration. Around $c=3.34\ \text{\AA}$8, about half of events were dopant insertions and half holes; above $c=3.34\ \text{\AA}$9, only dopants were observed (Dyck et al., 2023). This shows that bilayer geometry can improve not just equilibrium physics, but also atomic-scale manufacturability.

Several recurring misconceptions are corrected by the literature. First, bilayer stability is not determined by stacking labels alone: bilayer silicene is governed by atomic protrusion, not just AA/AB nomenclature (Sakai et al., 2015). Second, bilayers need not be crystalline: the vitreous silica bilayer is an amorphous continuous random network with a reflection plane and a finite flexibility window (Wilson et al., 2013). Third, moiré bilayers need not be statically fabricated: in a two-component condensate, the twisted-bilayer potential is generated dynamically by nonlinear interactions (Tian et al., 2024). Fourth, the best free-space quantum interface need not satisfy Bragg symmetry: non-symmetric bilayer atomic arrays can suppress diffraction losses more effectively by tuning interlayer phases continuously (Ben-Maimon et al., 15 Apr 2026).

Taken together, these results indicate that bilayer atomic arrays are best viewed as a family of coupled two-layer many-body systems rather than as a single material class. Their defining variables are registry, parity, spacing, and the mechanism of interlayer coupling—electrostatic, dispersive, elastic, Coulombic, dipolar, Rydberg-mediated, or nonlinear density-generated. This suggests several converging research directions: interaction-defined moiré structures in cold atoms, parity-engineered free-space interfaces, mechanically and optically functional trapped-ion bilayers, and atomic-scale fabrication on layered van der Waals templates. The unifying theme is that adding a second layer does more than duplicate a monolayer; it creates new collective coordinates and new interference channels, and these often become the dominant design variables of the array.