Twisted-Bilayer Optical Lattices

• Twisted-bilayer optical lattices are formed by superimposing two rotated periodic lattices, resulting in moiré superlattices with tunable flat bands and miniband structures.
• They enable diverse experimental realizations using ultracold atoms, photonic crystals, and synthetic coupling schemes to emulate complex quantum phenomena.
• These platforms reveal rich physics including angle-dependent band folding, dynamic interlayer interactions, and fractal energy spectra relevant for quantum simulation.

Searching arXiv for recent and foundational papers on twisted-bilayer optical lattices and closely related moiré optical platforms. Twisted-bilayer optical lattices are moiré platforms obtained by superimposing two periodic lattices with a relative rotation angle $\theta$, so that a long-wavelength superlattice emerges on top of the microscopic lattice scale. In ultracold-atom realizations, the two “layers” may be physical layers, state-dependent optical lattices addressing different internal states, or synthetic layers coupled by microwave or Raman transitions; in nanophotonics, the same geometry appears in twisted bilayer photonic crystals. Across these settings, the defining phenomena are moiré minibands, angle-tunable band folding and hybridization, quasi-flat or flat bands, and collective states ranging from superfluids and Mott-like phases to Bose-glass-like and superconducting regimes (Meng et al., 2021, González-Tudela et al., 2019, Tang et al., 2023, Zeng et al., 2024, Salamon et al., 2022).

1. Geometry and moiré superlattices

For two identical square lattices of lattice constant $a$, a relative rotation by $\theta$ is represented by

$$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$

and the long-wavelength moiré period is

$L_m=\frac{a}{2\sin(\theta/2)}.$

In the cold-atom experiment on Bose–Einstein condensates, this formula gives $L_m\simeq 4.35\,\mu\mathrm{m}$ for $a=395\,\mathrm{nm}$ and $\theta=5.21^\circ$, in excellent agreement with the in-situ images (Meng et al., 2021). Related square-lattice treatments distinguish commensurate from incommensurate twists: one formulation defines commensurate angles by $\cos\theta=(2mn)/(m^2+n^2)$, giving a finite moiré cell of area $A_M=|m^2+n^2|a^2$, while another uses $a$0, so that the rotated lattice closes on itself after forming a periodic supercell (Paul et al., 2023, Zeng et al., 2024).

The reciprocal-space description is equally central. In twisted-bilayer photonic crystals with square Bravais lattices, the two reciprocal lattices are $a$1 and $a$2, with the second obtained by rotation $a$3. Their interference produces a moiré reciprocal scale

$a$4

with $a$5. For small $a$6, $a$7 and $a$8, so moiré scattering appears as a fine grid of momentum-space spots, while the real-space moiré pattern is directly visible in optical or electron microscopy (Tang et al., 2023).

This geometry is the common substrate for the entire subject. Whether the lattice carries atoms, photons, or synthetic internal-state amplitudes, the twist introduces a new long length scale and a reduced mini-Brillouin zone in which folding, hybridization, and flat-band formation become possible.

2. Realizations and Hamiltonian frameworks

A direct cold-atom realization uses two spin-dependent optical lattices, with the internal states acting as the two layers. In second-quantized form, the two-component field $a$9 obeys

$\theta$0

where $\theta$1 are the spin-dependent lattice potentials and $\theta$2 is the microwave-induced interlayer coupling. In this setting, atoms in each spin state only feel one set of the lattice, while the interlayer coupling is independently controlled by microwave coupling between the spin states (Meng et al., 2021).

A conceptually different route emulates twistronics without physically twisting two layers. In the synthetic-bilayer proposal based on coherently coupled internal states, the intralayer dynamics is conventional nearest-neighbor hopping on a square lattice, but the interlayer Raman coupling carries both a spatially dependent Peierls phase and a spatially modulated envelope,

$\theta$3

The enlarged cell is then set by $\theta$4, and the moiré Brillouin zone is reduced accordingly, even though neither a physical bilayer nor literal twist is directly realized (Salamon et al., 2019).

A third widely used framework is a tight-binding bilayer with Gaussian interlayer hybridization. For non-interacting atoms on two twisted square lattices,

$\theta$5

where $\theta$6 is intralayer hopping, $\theta$7 the bare interlayer coupling, $\theta$8 the Gaussian width of the Wannier orbitals, and $\theta$9 a uniform bias between layers (Paul et al., 2023). Closely related cold-atom proposals emphasize that the intralayer tunneling $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$0 and interlayer hopping $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$1 can be controlled almost entirely independently, which is a recurring distinction from solid-state twisted bilayers (González-Tudela et al., 2019).

These formulations show that the subject is not tied to a single microscopic implementation. The common structure is a twisted geometry plus an independently tunable interlayer coupling, but the coupling may be microwave-induced, Raman-modulated, purely density–density, or photonic near-field hybridization.

3. Minibands, flat bands, topology, and fractality

In the earliest cold-atom bilayer proposals, diagonalization in the enlarged moiré unit cell already revealed the characteristic single-particle phenomenology: minibands, strong angle dependence of bandwidths, and van-Hove singularities. For the square-lattice model at the representative commensurate angle $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$2, increasing $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$3 beyond a critical value $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$4 yields two isolated flat bands split off from the bulk spectrum, while the density of states develops pronounced van-Hove peaks (González-Tudela et al., 2019).

Synthetic moiré supercells can generate the same structures in analytically tractable form. In the square synthetic bilayer, the “magic” choice $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$5 produces an emergent Lieb lattice in each $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$6-sector, with first-order bands

$$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$7

Second order in $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$8 gives a finite flat-band width, and the flatness ratio scales as $$R_{\theta}=\begin{pmatrix}\cos\theta&-\sin\theta\ \sin\theta&\cos\theta\end{pmatrix},$$9, reaching $L_m=\frac{a}{2\sin(\theta/2)}.$0 for $L_m=\frac{a}{2\sin(\theta/2)}.$1 (Salamon et al., 2019). In the fermionic synthetic-twist model, the same modulation produces a six-fold manifold around energy $L_m=\frac{a}{2\sin(\theta/2)}.$2; the residual bandwidth $L_m=\frac{a}{2\sin(\theta/2)}.$3 is minimal for $L_m=\frac{a}{2\sin(\theta/2)}.$4, and a band inversion occurs at $L_m=\frac{a}{2\sin(\theta/2)}.$5 (Salamon et al., 2022).

Hexagonal and honeycomb generalizations make contact with Dirac physics and with the topology of twisted graphene. In spin-twisted optical lattices, all twist angles $L_m=\frac{a}{2\sin(\theta/2)}.$6 can become magic that support gapped flat bands (Luo et al., 2020). In twisted bilayer honeycomb optical lattices for ultracold atoms, a weak interlayer-coupling “critical coupling” produces an isolated topological flat band at the Dirac-point energy; for $L_m=\frac{a}{2\sin(\theta/2)}.$7, one finds $L_m=\frac{a}{2\sin(\theta/2)}.$8, flat-band width $L_m=\frac{a}{2\sin(\theta/2)}.$9, and gap $L_m\simeq 4.35\,\mu\mathrm{m}$0. Wilson-loop winding distinguishes these weak-coupling flat bands from the trivial flat bands that proliferate in the strong-coupling regime (Sui et al., 13 Jun 2025).

A further extension is the fractal spectrum of twisted bilayer optical lattices. In the large-detuning limit, a square twisted bilayer optical lattice maps to a generalized Hofstadter model with long-range hopping and effective flux

$L_m\simeq 4.35\,\mu\mathrm{m}$1

The resulting spectrum as a function of $L_m\simeq 4.35\,\mu\mathrm{m}$2 exhibits a butterfly-like self-similar structure controlled purely by geometric moiré twisting rather than by a real magnetic field (Wan et al., 2024). This suggests that moiré geometry in optical lattices can serve not only as a flat-band mechanism but also as a route to fractal single-particle spectra.

4. Interactions and collective phases

The bosonic many-body problem first became concrete in the experiment with atomic Bose–Einstein condensates loaded into twisted-bilayer optical lattices. In that system, in-situ absorption images show fringe-like density modulations at the moiré period, while time-of-flight images display both primary Bragg peaks at $L_m\simeq 4.35\,\mu\mathrm{m}$3 and moiré peaks at $L_m\simeq 4.35\,\mu\mathrm{m}$4. The persistence of the moiré-peak visibility over $L_m\simeq 4.35\,\mu\mathrm{m}$5 confirms the superfluid phase in the twisted lattice, and ramping $L_m\simeq 4.35\,\mu\mathrm{m}$6 or $L_m\simeq 4.35\,\mu\mathrm{m}$7 yields a sequential loss of moiré coherence followed by loss of primary-lattice coherence, corresponding to entry into SF-II and then MI (Meng et al., 2021).

Interactions need not merely dress a pre-existing static moiré potential; they can generate one. In the interaction-induced bilayer Bose–Hubbard model with vanishing interlayer tunneling, the interlayer term $L_m\simeq 4.35\,\mu\mathrm{m}$8 creates a self-consistent “dynamical” moiré potential

$L_m\simeq 4.35\,\mu\mathrm{m}$9

This feedback mechanism supports SF$a=395\,\mathrm{nm}$0–SF$a=395\,\mathrm{nm}$1, MI$a=395\,\mathrm{nm}$2–MI$a=395\,\mathrm{nm}$3, MI$a=395\,\mathrm{nm}$4–SF$a=395\,\mathrm{nm}$5, SF$a=395\,\mathrm{nm}$6–MI$a=395\,\mathrm{nm}$7, “interlocked” MI, and self-localized or Bose-glass-like phases, including irregular density patterns without extrinsic disorder (Zeng et al., 2024).

Clusterization provides another organizing principle. In the cluster Gutzwiller treatment of interacting bosons, strongly hybridized pairs of sites are promoted to explicit two-site clusters, and the resulting phase structure contains Mott-like insulators, superfluids, and Bose-glass-like states defined by mobile pockets that fail to percolate. Notably, even at commensurate twist angles a BG-like phase appears because of clusterization, and in the incommensurate case mobility islands may occur even without inter-layer hopping solely due to inter-layer interactions (Paul et al., 16 Jul 2025).

For attractive fermions, the flat-band problem becomes a pairing problem. In synthetic twisted bilayer lattices, quasi-flat bands strongly enhance the pairing gap when they lie at the Fermi surface, and tuning the ratio $a=395\,\mathrm{nm}$8 switches superconducting correlations from intra-layer to inter-layer pairing, with a re-entrant superconducting valley near $a=395\,\mathrm{nm}$9 for $\theta=5.21^\circ$0 (Salamon et al., 2022). In spin-twisted hexagonal optical lattices, weak attractive interaction drives a Larkin–Ovchinnikov state with finite-momentum pairing, reflecting the interplay between flat bands and inter-spin interactions in the single-layer spin-twisted geometry (Luo et al., 2020).

5. Transport, ergodicity, and driven dynamics

Single-particle transport in twisted-bilayer optical lattices is strongly geometry dependent. For commensurate twist angles and $\theta=5.21^\circ$1, transport is dominated by channels: the directly coupled sites form two quasi-flat bands at $\theta=5.21^\circ$2, while the central bands propagate with approximate channel dispersion

$\theta=5.21^\circ$3

so that $\theta=5.21^\circ$4 and $\theta=5.21^\circ$5 (Paul et al., 2023). For incommensurate twistings, the interlayer coupling acts as an effective disorder strength: the spectrum remains ergodic up to $\theta=5.21^\circ$6, above which part of the eigenspectrum becomes multifractal, and a similar transition occurs at $\theta=5.21^\circ$7. The accompanying wavepacket dynamics changes from ballistic or diffusive expansion with $\theta=5.21^\circ$8 to strongly suppressed expansion with $\theta=5.21^\circ$9 (Paul et al., 2023).

Interacting bosons in aperiodic moiré potentials show a related but distinct connectivity physics. In the effective single-layer Bose–Hubbard description motivated by twisted bilayer and quasicrystal optical lattices, the percolation probability $\cos\theta=(2mn)/(m^2+n^2)$0 distinguishes a globally connected superfluid from a Bose glass of disconnected SF islands, while the inverse participation ratio tracks density localization. At low filling, increasing $\cos\theta=(2mn)/(m^2+n^2)$1 can induce a BG$\cos\theta=(2mn)/(m^2+n^2)$2SF transition; at higher filling, the same interaction can produce BG$\cos\theta=(2mn)/(m^2+n^2)$3SF$\cos\theta=(2mn)/(m^2+n^2)$4BG reentrance. Quenches across the phase boundary sharply alter both $\cos\theta=(2mn)/(m^2+n^2)$5 and the IPR, whereas intra-phase quenches do not (Ding et al., 5 Mar 2025).

Periodic driving adds another control axis. In Floquet-engineered twisted bilayer hexagonal optical lattices, modulation of the on-site interaction renormalizes tunneling as $\cos\theta=(2mn)/(m^2+n^2)$6. The density-wave dynamics separates into four stages—preparation, excitation, pattern-forming, and nonlinear heating—and at $\cos\theta=(2mn)/(m^2+n^2)$7 and $\cos\theta=(2mn)/(m^2+n^2)$8 a D$\cos\theta=(2mn)/(m^2+n^2)$9 moiré quasicrystal pattern emerges around $A_M=|m^2+n^2|a^2$0. A $A_M=|m^2+n^2|a^2$1 change in $A_M=|m^2+n^2|a^2$2 near $A_M=|m^2+n^2|a^2$3 completely alters the emerging quasicrystal pattern, indicating extreme frequency sensitivity (Fan et al., 27 Aug 2025).

6. Photonic analogues and broader scope

The moiré principle extends beyond matter waves to on-chip photonics. In the optical twisted-bilayer photonic crystal, each layer is a free-standing silicon-nitride membrane of thickness $A_M=|m^2+n^2|a^2$4, perforated with a square lattice of air holes of period $A_M=|m^2+n^2|a^2$5 and radius $A_M=|m^2+n^2|a^2$6, with the two membranes separated by an SU-8 spacer fixing $A_M=|m^2+n^2|a^2$7. An effective Hamiltonian in a reciprocal-vector basis,

$A_M=|m^2+n^2|a^2$8

captures single-layer lattice scattering and interlayer near-field coupling. Experimentally, wavelength-resolved back-focal-plane imaging directly visualizes moiré-folded parabolic bands and full $A_M=|m^2+n^2|a^2$9–M–X–$a$00 dispersions up to $a$01, with angle tuning from $a$02 to $a$03 shifting the band centers in $a$04-space and producing a weak parabolic dependence of the band-edge frequency on $a$05 (Tang et al., 2023).

Two common misconceptions are corrected by the optical-lattice literature. First, a twisted-bilayer optical lattice does not require a literal mechanically twisted bilayer: synthetic layers with spatially patterned Raman coupling reproduce moiré minibands, Dirac cones, and quasi-flat bands without a physical twist (Salamon et al., 2019). Second, moiré physics in these systems does not require a static single-particle interlayer tunneling; it can arise self-consistently from interlayer density–density interactions, producing a genuinely dynamical moiré potential and phases not present in conventional twisted bilayers (Zeng et al., 2024).

Taken together, these results define twisted-bilayer optical lattices as a family of highly tunable moiré simulators rather than a single model. Their parameter space includes twist angle, lattice symmetry, lattice depth, interlayer coupling, interaction strength, bias, detuning, drive amplitude, and drive frequency; their observables include in-situ moiré density modulation, momentum-space diffraction, band spectroscopy, Wilson-loop topology, transport anisotropy, percolation, and multifractality. This suggests that the field occupies an intermediate position between condensed-matter twistronics, quantum simulation with ultracold atoms, and moiré nanophotonics, with each implementation isolating different pieces of the same underlying geometric mechanism.