🔬 condensed matter

From pink to brown: Interaction-driven noise color shift in the Aubry-Andr{é} Model

This paper demonstrates that while non-interacting Aubry-Andr{é} models exhibit brown-noise-like spectral fluctuations regardless of the phase, introducing interactions causes a distinct shift to pink noise in the metallic phase, thereby establishing interaction-driven noise color as a reliable indicator of the metal-insulator transition.

Original authors: M. Jiménez-Valdez, S. A. Montes-Camacho, E. J. Torres-Herrera

Published 2026-07-31

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: M. Jiménez-Valdez, S. A. Montes-Camacho, E. J. Torres-Herrera

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are listening to a symphony where every instrument plays a single, pure note. In the world of quantum physics, these notes are the "energy levels" of a system, like the specific frequencies a guitar string can vibrate at. Scientists have long been fascinated by how these notes arrange themselves. Sometimes, they are chaotic and push each other away, like a crowded dance floor where everyone needs personal space (this is called a "metallic" or "extended" phase). Other times, they clump together or stand still, ignoring each other, like a quiet library where books sit on shelves without interacting (this is an "insulating" or "localized" phase).

The big question is: how do we tell the difference between a chaotic dance floor and a quiet library just by listening to the notes? Physicists use a tool called "spectral statistics" to analyze the spacing between these energy notes. They look for patterns that resemble different types of "noise." Think of "pink noise" as the sound of a steady, rhythmic heartbeat or a gentle waterfall—a pattern that feels connected and alive. In contrast, "brown noise" sounds like a deep, rumbling static or the heavy thud of a drum with no rhythm—a pattern that feels random and disconnected. For decades, scientists have used these noise colors to diagnose whether a quantum system is conducting energy freely or trapping it.

Now, picture a specific musical instrument: a one-dimensional chain of tiny magnets (spin-1/2 particles) sitting in a strange, repeating-but-not-quite-repeating landscape (a quasiperiodic potential). This setup is known as the Aubry-André model. It's famous because, unlike most systems that need random messiness to stop conducting, this one stops conducting just by changing the strength of its landscape, even without any randomness. The big mystery was: if we turn on the "volume" of interactions between these tiny magnets, does the way they arrange their energy notes change in a way we can easily hear?

This paper, titled "From pink to brown: Interaction-driven noise color shift in the Aubry-André Model," takes a fresh look at this problem. The researchers, M. Jiménez-Valdez, S. A. Montes-Camacho, and E. J. Torres-Herrera, decided to listen to the "power-spectrum" of these energy notes. In simple terms, they took the sequence of gaps between the energy levels, turned it into a wave, and checked what color of noise it made.

Here is what they found, and it's a bit of a plot twist. When the tiny magnets don't talk to each other (the non-interacting case), the power-spectrum is stubborn. No matter how strong the landscape is, the noise always sounds like "brown noise." It's as if the system is pretending to be a quiet library even when it's actually a chaotic dance floor. The usual tools failed to spot the transition from metal to insulator in this silent scenario.

However, the story changes dramatically when the magnets start talking to their neighbors (the interacting case). Suddenly, the power-spectrum becomes a reliable detective. When the system is in a "metallic" state (where particles move freely), the noise shifts to "pink noise," revealing that rhythmic, connected heartbeat. But as the landscape gets stronger and the system turns into an "insulator" (where particles get stuck), the noise shifts back to "brown noise," the heavy, disconnected rumble.

The authors ran these scenarios on a computer simulation with a chain of 16 sites, which gave them a large enough set of energy levels (12,870 of them) to be statistically sure. They didn't just rely on the noise color; they also checked other clues, like how often energy levels repel each other and how spread out the particles are. These extra checks confirmed their story: the noise color shift from pink to brown is a clear sign of the metal-insulator transition, but only if the particles are interacting. Without those interactions, the noise stays brown and hides the transition.

In the end, this paper suggests that to hear the true story of how quantum systems switch from flowing to freezing, you have to listen to the interactions. If the particles are silent, the music sounds the same everywhere. But once they start chatting, the music changes color, giving us a new, vivid way to spot the boundary between a conductor and an insulator.

Technical Summary: Interaction-driven noise color shift in the Aubry-André Model

Problem Statement
The study addresses the characterization of metal-insulator transitions in one-dimensional systems subject to quasiperiodic potentials, specifically the Aubry-André (AA) model. While the non-interacting AA model is a paradigmatic example of a disorder-driven transition without randomness, the inclusion of interactions introduces complexity, potentially leading to many-body localization (MBL). Standard spectral diagnostics, such as level spacing distributions and level number variance, are widely used to distinguish between ergodic (metallic) and localized phases. However, the authors note that these standard indicators may fail to capture the specific noise-like structures present in spectral fluctuations, particularly in the non-interacting limit where the transition is known to occur. The central problem is whether the power-spectrum of energy-level fluctuations can serve as a sensitive diagnostic for this transition, and how inter-particle interactions influence this sensitivity.

Methodology
The authors investigate a system of spin-1/2 particles on a one-dimensional lattice of size $L=16$ (yielding a Hilbert space dimension $N=12,870$ in the $S^z=0$ sector) governed by a Hamiltonian containing nearest-neighbor interactions and an on-site quasiperiodic Harper potential. The interaction strength is controlled by the anisotropic parameter $\Delta$ (where $\Delta=0$ corresponds to the non-interacting case and $\Delta \neq 0$ introduces Ising-like interactions). The potential strength $h$ is varied to traverse different phases: extended (metallic), critical, and localized (insulating).

The primary diagnostic tool employed is the power-spectrum ( $P_k$ ) of the sequence of consecutive level spacings ( $\delta_n$ ). This is defined as the squared modulus of the Fourier transform of the unfolded level spacings. Theoretical benchmarks are established using Random Matrix Theory (RMT): Poisson statistics (characteristic of integrable or localized systems) are expected to yield $1/k^2$ decay (brown noise), while Gaussian Orthogonal Ensemble (GOE) statistics (characteristic of chaotic/ergodic systems) are expected to yield $1/k$ decay (pink noise).

These results are cross-validated using:

Level spacing distribution ( $P(s)$ ): Comparing against Wigner's surmise (GOE) and Poisson distributions.
Level number variance ( $\Sigma^2(\ell)$ ): Assessing long-range spectral rigidity.
Participation Ratio (PR): Measuring the spatial extent of energy eigenstates.
Porter-Thomas distribution: Analyzing the statistical distribution of eigenstate components to identify chaotic vs. localized behavior.

Key Results

Non-Interacting Case ( $\Delta = 0$ ):
- The power-spectrum fails to detect the metal-insulator transition. Regardless of the potential strength $h$ (whether in the metallic, critical, or localized regime), the power-spectrum consistently exhibits a $1/k^2$ decay, characteristic of brown noise.
- Level spacing distributions and level number variances also remain Poisson-like across all regimes, showing no level repulsion.
- Despite the failure of spectral statistics, the structure of energy eigenstates (via Participation Ratio and Porter-Thomas distribution) successfully distinguishes between extended and localized phases.
Interacting Case ( $\Delta \neq 0$ ):
- The inclusion of interactions dramatically alters the spectral statistics.
- Metallic Phase ( $h \approx 0.2$ ): The power-spectrum displays a $1/k$ decay, consistent with pink noise, and the level spacing distribution follows Wigner's surmise (GOE).
- Localized Phase ( $h \approx 3.0$ ): The power-spectrum reverts to $1/k^2$ decay (brown noise), and level statistics become Poisson-like.
- Critical/Intermediate Phase ( $h \approx 1.2$ ): The system shows a transition toward Poisson statistics, though the power-spectrum decay exponent is already close to 2 for the system sizes studied.
- Eigenstate structure analysis confirms that only the interacting model exhibits thermalization (extended states following GOE statistics) in the metallic regime, whereas the non-interacting model remains non-ergodic even in the extended phase.

Significance and Claims
The paper claims that the power-spectrum of level spacings is a robust indicator of the metal-insulator transition only when interactions are present. In the absence of interactions, the power-spectrum is "blind" to the transition, exhibiting brown-noise-like behavior across all regimes due to the specific nature of the non-interacting spin-1/2 representation.

The primary contribution is the demonstration that interactions are the driving force that makes the power-spectrum sensitive to the transition, shifting the noise color from brown (localized) to pink (metallic). The authors conclude that while spectral statistics (power-spectrum, level spacing) are effective for diagnosing the transition in interacting systems, they are insufficient for the non-interacting AA model. Conversely, diagnostics based on the structural properties of energy eigenstates (Participation Ratio and component distributions) remain effective indicators of the transition in both interacting and non-interacting scenarios.

More like this