A square array’s corners can decide where an ordered pattern breaks

An October 14 preprint investigates boundary-induced frustration in square Rydberg arrays using Aquila experiments and tensor-network comparisons. In the examined striated regime, odd and even array sizes support different boundary compatibility, with competing corner-seeded domains and defects. The study also distinguishes prepared dynamics from equilibrium behavior at larger detuning. This is a many-body ordering result, not evidence of defective atoms or a claim that every square antiferromagnetic lattice is geometrically frustrated. The boundaries and the particular ordered pattern together create the incompatibility being studied.^[1]

The Aquila experiments compare 13-by-13 and 14-by-14 arrays. In the even array, competing corner-seeded patterns produce a four-domain state. The paper also identifies a diagonal two-domain equilibrium regime numerically, but the experimental preparation ramp remains trapped in the four-domain regime. Keeping those outcomes separate shows what the apparatus observed and what the calculation suggests could become accessible.^[1]

A one-dimensional illustration makes the role of parity transparent. Consider an alternating pattern A–B–A–B and suppose both ends prefer A. A five-site chain can use A–B–A–B–A and satisfy both ends. A six-site continuation ends in B, so it cannot satisfy both preferences while preserving the same perfect alternation. This invented string example is not the Rydberg Hamiltonian or a derivation of its two-dimensional domains. It shows how changing size by one can change compatibility even when the local pattern rule stays the same.

In two dimensions, several corners can seed different domain orientations or translations. Their influence competes as the ordered regions grow inward. A domain wall records where the choices meet. Its location and symmetry can carry information about both the equilibrium energy landscape and the route taken during preparation. Measuring only an average excitation density can conceal this structure. Spatially resolved correlations and repeated images of the pattern are useful because they distinguish a reproducible boundary-selected interface from a collection of unrelated local errors.

Preparation dynamics need not find the lowest-energy arrangement on the accessible timescale. A ramp can remain in a symmetry sector or a metastable configuration even when another state becomes energetically favored. Faster and slower ramps can therefore yield different patterns without changing the final Hamiltonian. This makes equilibrium and time-dependent classical comparisons complementary. The former asks what the system prefers energetically; the latter asks what the specified preparation can reach. Interpreting the hardware requires both questions when the evolving state is sensitive to the path through parameter space.

Finite arrays also make averages over symmetry-related outcomes subtle. If individual runs choose different domain orientations, their average can look more symmetric than any single realization. Conversely, an apparently preferred orientation can reflect a small systematic imbalance in the apparatus. A useful analysis should examine the distribution of configurations and correlations, not only the mean image. Controlled changes to boundary conditions can test whether a pattern follows the intended geometry or an unintended spatial bias, giving the observed ordering a clearer physical explanation.

The boundary contribution does not automatically vanish in every experimentally accessible system. Near a transition or in a constrained ordering regime, its influence can extend far into a finite sample. Increasing size can change the balance rather than merely adding more of the same bulk. This is why a quantum simulator's geometry is part of the modeled problem. Selecting dimensions solely because they fit the available hardware can unintentionally favor or frustrate the order being sought, altering the scientific question before measurement begins.

The study turns that sensitivity into a tool for exploring programmable matter. The next useful experiments would vary corner conditions, size and ramp history in a coordinated way, then compare the resulting domain statistics with equilibrium and dynamical predictions. Readers could see whether an observed defect is an intended signature of competing boundary constraints. That perspective makes the apparatus's edges scientifically productive: they can reveal how local ordering rules, global geometry and preparation history combine to determine the state that a finite quantum simulator actually produces.