Quantum Systems Defy Random Matrix Predictions

When scientists study complex quantum systems, they often rely on random matrix theory—a mathematical framework that predicts how energy states should be distributed in chaotic systems. But new research shows that real quantum systems don't always follow these predictions, revealing fundamental differences between mathematical models and physical reality that could reshape our understanding of quantum chaos.

The researchers discovered that while random matrix ensembles predict specific statistical patterns for quantum states, actual physical systems like the quantum kicked rotor and XXZ spin chain follow different paths toward ergodicity—the state where quantum systems explore all available states equally. The study analyzed multifractal dimensions, which measure how quantum states spread across different configurations, finding that real systems approach the expected behavior much more slowly and along system-specific trajectories.

To investigate these differences, the team compared two types of systems: random matrix ensembles (mathematical models) and physical quantum systems. For the random matrix cases, they examined circular orthogonal ensemble (COE) and circular unitary ensemble (CUE), which represent systems with and without time-reversal symmetry. For physical systems, they studied the quantum kicked rotor—a paradigmatic model of quantum chaos—and the XXZ spin chain, a many-body quantum system. The researchers calculated multifractal dimensions D_q, which quantify how quantum states distribute themselves across different configurations, analyzing how these values change with system size.

The data revealed striking deviations. For random matrix ensembles, the analytical predictions provided lower bounds for the typical values of multifractal dimensions, with D_q approaching 1 (indicating full ergodicity) as system size increases—but this approach was remarkably slow, following logarithmic corrections. For example, in CUE systems, D_1 = 1 - γ/ln N where γ ≈ 0.577 is Euler's constant. However, even these mathematical models showed significant state-to-state fluctuations, with the distribution of D_q values becoming increasingly broad for larger q values.

When examining physical systems, the quantum kicked rotor closely followed random matrix predictions, with its multifractal dimensions matching CUE results across different system sizes (N=400, 2000, 10000). But the XXZ spin chain told a different story. Its mid-spectrum eigenstates—those expected to behave most ergodically—showed significant deviations from GOE predictions. While these states still approached D_q = 1 in the large-size limit, they followed a system-specific path that differed from random matrix expectations. The researchers quantified this weak ergodicity through an effective Hilbert space fraction, finding that these states occupy only a finite fraction of the available configurations compared to random matrix states.

These findings matter because random matrix theory underpins much of our understanding of quantum chaos and thermalization—the process by which isolated quantum systems reach equilibrium. The discovery that real many-body systems only weakly follow these predictions suggests that standard random matrix classes may not be optimal models for describing mid-spectrum eigenstates of non-integrable Hamiltonians. This could impact how scientists model quantum thermalization in isolated systems and understand the statistical properties of energy eigenstates in complex quantum materials.

The research acknowledges several limitations. The finite-size effects are remarkably strong, requiring exponentially large sample sizes to restore agreement between numerical computations and analytical predictions. The study also found that statistical errors become significant beyond a certain q value that scales with ln N, meaning that obtaining reliable results for larger q values would require averaging over exponentially many realizations. Additionally, the work examined only specific parameter regimes and system types, leaving open questions about how these behaviors might change in other chaotic systems or near integrability points.