Systematic construction of quantum many-body scars in frustrated Rydberg arrays

This paper introduces a graph-theoretic framework that systematically identifies two distinct mechanisms for constructing quantum many-body scars in frustrated Rydberg atom arrays on arbitrary lattices, demonstrating their existence on hexagonal lattices and establishing scarring as a generic feature for encoding protected information beyond bipartite systems.

The Gist

Imagine a crowded dance floor where everyone is trying to move, but there's a strict rule: no two neighbors can dance at the same time. If you try to jump up (get excited), your neighbors must stay seated. This is the world of "Rydberg atom arrays," a type of quantum computer simulator.

Usually, when you start a dance on such a floor, the chaos spreads instantly. The system gets "scrambled," and the original pattern is lost forever. This is called thermalization—everything just turns into a hot, messy soup of random motion.

However, scientists have discovered a rare exception called Quantum Many-Body Scars. In these special cases, the system doesn't turn into soup. Instead, it remembers its starting move and keeps dancing in a perfect, repeating loop, like a record skipping on the same groove.

Until now, this "perfect looping" was only seen on simple, checkerboard-style dance floors (called bipartite lattices). The big question was: What happens on more complicated, "frustrated" dance floors where the rules make it impossible for everyone to be happy?

This paper says: Scarring still happens, but in two very different ways. The authors created a "map-making" toolkit (using graph theory) to find these special loops on any shape of dance floor.

Here are the two ways they found to keep the dance going:

1. The "Team-Up" Strategy (Type-I Scars)

The Problem: On a tricky floor (like a hexagon or a triangle), the "no neighbors dancing" rule creates a deadlock. It's too frustrating for the system to loop.
The Solution: The authors realized you can group atoms into little teams (like holding hands in a tight circle).

• The Analogy: Imagine the dance floor is made of small, tight circles of three people. The rule says only one person in the circle can stand up at a time.
• How it works: Instead of treating every single atom as an individual, the system treats each circle as a single unit. Even though the floor is messy, these "team units" can still form a perfect checkerboard pattern.
• The Result: The system finds a way to "pretend" the floor is simple again. It creates a special starting state where these teams coordinate perfectly, allowing the whole system to oscillate back and forth without getting stuck.
• The Bonus: On a hexagonal floor, they found an exponential number of these special starting patterns. This means you could potentially store a lot of information (bits) in these loops that won't get erased by the chaos.

2. The "Freeze and Dance" Strategy (Type-II Scars)

The Problem: Some floors are so frustrating that the "Team-Up" strategy doesn't work. The rules are too tight.
The Solution: Instead of trying to make the whole floor dance, the system freezes a large chunk of it and lets the rest dance freely.

• The Analogy: Imagine a dance floor where the middle section is locked down with heavy chains (the "frozen" part). The people in the middle can't move at all. Because they are frozen, they act as a buffer. They stop the dancers on the left side from bumping into the dancers on the right side.
• How it works: The "frozen" middle section (Sublattice C) pins the system in place. This isolation allows the two outer sections (Sublattices A and B) to swing back and forth like a pendulum, completely free from the chaos of the middle.
• The Result: This works on highly frustrated shapes (like a 3D pyramid structure) where the "Team-Up" strategy failed. The frustration that usually stops the dance actually helps by locking the middle section down, creating a safe zone for the oscillation.

Why This Matters

The paper proves that these "perfect loops" aren't just a fluke of simple shapes. They are a generic feature of these quantum systems.

• The Toolkit: The authors didn't just guess; they built a mathematical "search engine" (based on graph theory) that can scan any lattice shape and tell you: "Here is the perfect starting state to make this system loop."
• The Experiment: They showed that on a hexagonal floor, you can create a massive family of these loops. This suggests that quantum simulators (machines that use atoms to simulate physics) can be programmed to find these states and use them to keep information safe from thermalization.

In short: The paper shows that even in the most chaotic, rule-heavy quantum environments, you can engineer specific starting conditions to make the system "remember" its dance steps. Sometimes you do this by grouping atoms into teams (Type-I), and sometimes by freezing part of the system to let the rest swing freely (Type-II).

Technical Summary

Technical Summary: Systematic Construction of Quantum Many-Body Scars in Frustrated Rydberg Arrays

Problem Statement
Quantum many-body scars (QMBSs) represent a mechanism for avoiding thermalization in chaotic interacting systems, characterized by a small number of non-thermal eigenstates within a special subspace. While QMBSs have been extensively observed and understood in one-dimensional (1D) Rydberg atom arrays and various two-dimensional (2D) bipartite lattices (e.g., square, honeycomb), their existence in non-bipartite, frustrated lattices remained an open question. Previous experimental and theoretical efforts failed to identify signatures of scarring in non-bipartite geometries such as triangular or Kagome lattices. The central challenge lies in the lack of dimension-insensitive methods to find suitable scarred initial states in these complex geometries, as existing tools often rely on matrix product states, require known scar states as input, or are limited to Fock states.

Methodology
The authors introduce a graph-theoretic framework to systematically identify scarred initial states on arbitrary lattices. They model the Rydberg atom array as a graph $G(V, E)$, where vertices represent atoms and edges represent blockade interactions. The core of their approach involves partitioning the vertex set $V$ into three disjoint subsets: $A$, $B$, and $C$.

1. Algebraic Construction: The framework constructs an approximate $su(2)$ algebra using raising and lowering operators ($J_+$, $J_-$) defined over these subsets. The Hamiltonian is proportional to $J_x$, while $J_z$ and $J_y$ define the basis for the scarred subspace. If the algebra closes (or is approximately closed), the ground state of $J(\theta) = \cos(\theta)J_z - \sin(\theta)J_y$ exhibits coherent oscillations (revivals) rather than thermalizing.
2. Partitioning Rules: The task of finding a scar reduces to finding a partition $\{A, B, C\}$ that satisfies specific geometric and connectivity conditions. The authors propose two distinct mechanisms:
   • Type-I Scarring: Designed for lattices with mild frustration. The method maps the physical lattice to an effective bipartite lattice by grouping vertices into cliques (subsets $s_j$ where all atoms block each other). The quotient graph formed by these cliques must be bipartite. This generalizes the standard Néel state to include locally entangled states (e.g., W-states within cliques) to overcome frustration.
   • Type-II Scarring: Designed for lattices with strong frustration. This mechanism utilizes a non-empty "dead" sublattice $C$ to pin part of the system. The conditions require that $A$ and $B$ only have neighbors in $C$, while $C$ is strongly connected and sufficiently "blocked" by $A$ and $B$ to suppress its own dynamics. This allows $A$ and $B$ to oscillate almost freely, effectively decoupled by the frustration-induced pinning of $C$.

The authors employ the Knuth dancing links (DLX) algorithm to systematically search for valid clique covers (for Type-I) and maximum independent sets (for Type-II) in specific lattice geometries.

Key Contributions and Results

• Generalization to Non-Bipartite Lattices: The framework successfully identifies scarred trajectories in non-bipartite geometries where previous methods failed.
• Type-I in Shastry-Sutherland and Hexagonal Lattices:
  • In the Shastry-Sutherland lattice, the authors identify a dimer cover that maps to an effective square lattice, revealing strong revivals in return fidelity despite the absence of clear outliers in entanglement entropy.
  • In the hexagonal lattice, the method uncovers an exponential family of scarred initial states ($2^{N_y-1}$ for a torus with $N_y$ rows). These states encode information protected from thermalization, preserving $N_y-1$ bits of information.
• Type-II in Quasi-2D Lattices:
  • The authors demonstrate that while pure triangular and Kagome lattices resist both Type-I and Type-II scarring, their quasi-2D counterparts (composed of tetrahedra, such as the asanoha and pyrochlore lattices) support Type-II scarring.
  • Numerical simulations on the quasi-2D asanoha lattice show that the $C$ sublattice dynamics are highly suppressed, allowing $A$ and $B$ to oscillate with high fidelity.
• Numerical Verification: The paper provides extensive numerical simulations (return fidelity, entanglement entropy, and sublattice occupation) confirming the existence of these scars in the PXP model on various lattices.

Significance and Claims
The paper establishes that quantum many-body scarring is a generic feature of Rydberg systems beyond one dimension, extending to frustrated, non-bipartite lattices. The authors claim that their graph-theoretic framework provides an experimentally accessible route to systematically probing non-thermal dynamics.

Key claims regarding the nature of scarring include:

1. Dual Mechanisms: Scarring in frustrated systems arises via two distinct mechanisms: Type-I, which uses local entanglement to overcome mild frustration, and Type-II, which exploits strong frustration to pin a sublattice and stabilize oscillations in the remainder.
2. Information Protection: The exponential family of scars found in the hexagonal lattice suggests a mechanism for encoding information that is robust against thermalization.
3. Experimental Relevance: The identified initial states are short-range entangled (SRE) and can be prepared via annealing or adiabatic ramping from product states, making them suitable for current Rydberg atom quantum simulators.

The authors remain modest regarding the universality of their findings, noting that scarring was not found in pure triangular or Kagome lattices, suggesting a "Goldilocks" zone of frustration is required. They also acknowledge that the robustness of these scars away from the ideal PXP limit and the potential coexistence of Type-I and Type-II scars remain open questions. The work invites further study into the links between maximum independent sets and QMBSs in other constrained models.