Power spectrum of magnetic relaxation in spin ice: anomalous diffusion in a Coulomb fluid

By employing high-frequency alternating-field susceptibility measurements on Dy${}_2$Ti${}_2$O${}_7$, this study corrects previous underestimations of the anomalous diffusion exponent $b(T)$, establishes its deviation from Brownian motion up to 20 K, and reveals its sample-dependent nature within the dense Coulomb fluid of magnetic monopoles.

The Gist

Imagine a special type of crystal called spin ice (more precisely, a material named Dy₂Ti₂O₇). Within this crystal, tiny magnetic particles, known as "spins," behave like a chaotic crowd trying to find space in an overcrowded room. They wish to follow a specific rule: for every group of four seats, two people must face inward and two outward. However, since the seats are arranged in a complicated triangular pattern, it is impossible for everyone to be perfectly comfortable simultaneously. This creates a state of "frustration."

In this frustrated crowd, the smallest disturbances look like magnetic monopoles. Do not think of entire magnets, but rather isolated "North" or "South" poles that can move freely, like individual people walking through the crowd.

The Puzzle: "Pink Noise" versus "Red Noise"

Scientists have listened to the "noise" generated by these moving monopoles. In physics, noise is not just static crackling; it possesses a pattern.

• Brownian Motion (Red Noise): If these monopoles were simply wandering randomly, like a drunk person in the fog, the noise would follow a specific, predictable pattern (a power law with an exponent of b = 2).
• Anomalous Diffusion (Pink Noise): However, previous experiments suggested that something strange was happening. The noise looked different, with an exponent b closer to 1.2 or 1.5. This implied that the monopoles were not just wandering randomly; they were navigating a complex, "fractal" landscape (like a labyrinth with holes within holes), making their movement "slower" or more constrained than simple random walking.

The Problem: A Measurement Error

The article points out a major flaw in these earlier measurements. The scientists who found the "strange" noise used a method that sampled data in tiny time intervals.

• The Analogy: Imagine trying to record a high-speed race car with a camera that takes photos very slowly. If the car moves too quickly between photos, the camera might "alias" the image, making the car appear to move in a strange, jerky manner or at the wrong speed.
• The Reality: The earlier noise measurements missed the very fast, high-frequency movements of the monopoles. Due to this "aliasing," the data appeared flatter than it actually was, leading scientists to calculate a lower "b" value (around 1.2) and assume the monopoles were trapped in a complex labyrinth.

The New Discovery: The "High-Speed Camera"

The authors of this article decided to examine the same crystal with a different tool: AC susceptibility (alternating field susceptibility).

• The Analogy: Instead of taking slow, bumpy photos (noise measurement), they used a high-speed camera capable of capturing movement up to 1 million times per second (1 MHz). This is much faster than previous methods, which only went up to about 100,000 times per second.
• The Result: When they viewed the data with this "high-speed camera," the picture changed. The exponent b was actually much closer to 2 (the value for simple random walking) than previously thought.
  • At low temperatures (around 2 K), b is approximately 1.8.
  • As the temperature rises to 20 K, b smoothly approaches 2.

What This Means for the Monopoles

The article concludes that these magnetic monopoles in the temperature range between 2 K and 20 K are not stuck in a complex, fractal labyrinth. Instead, they behave much more like a dense fluid, where they bump into each other and move in a way very close to the standard of random walking (Brownian motion).

• The Image of the "Dense Fluid": Imagine the monopoles as an overcrowded dance floor. They bump into each other and interact strongly (a "Coulomb fluid"), but they are not navigating a strange, hole-riddled labyrinth. Their movement is complex due to the crowd, but it follows the standard rules of random motion.
• The Image of the "Fractal": The idea that they reside in a fractal labyrinth might still hold true at very low temperatures (below 1 K), where the crowd thins out and they move very slowly. But in the "warm" zone (2–20 K), the labyrinth image was likely an illusion caused by the measurement device being too slow to detect the rapid movements.

A Note on Sample Differences

The researchers also noted that the exact numbers changed slightly depending on which specific crystal sample they tested. This suggests that tiny defects or impurities in the crystal (like a few people in the crowd wearing the wrong shoes) can influence how the monopoles move. However, the main trend—that the movement is closer to simple random walking than previously assumed—held true for all samples.

Summary

In short, this article corrects a measurement error. It tells us that the magnetic monopoles in spin ice do not do anything exotic or fractal over a wide temperature range; they essentially perform a very busy, overcrowded version of standard random walking. The "strange" behavior observed in earlier studies was likely just a trick of the measuring instruments.

Technical Summary

Technical Summary: Power Spectrum of Magnetic Relaxation in Spin Ice: Anomalous Diffusion in a Coulomb Liquid

Problem Statement
Previous measurements of magnetization noise in the spin-ice material $\text{Dy}_2\text{Ti}_2\text{O}_7$ revealed a "pink noise" power spectrum $S(f, T) \propto f^{-b(T)}$ below 4 K, interpreted as evidence for the diffusion of magnetic monopole excitations in a fractal landscape. However, a significant discrepancy exists in the literature regarding the anomalous exponent $b(T)$. While neutron spin-echo experiments establish $b=2$ (characteristic of Brownian motion) for temperatures above 20 K, earlier noise-based measurements suggested that $b(T)$ decreases from $\approx 1.5$ at 1 K to $\approx 1.2$ at 4 K, showing no clear approach to the expected value $b=2$ even at higher temperatures. This contribution addresses the inconsistency between noise-derived estimates of $b(T)$ and the established high-temperature behavior, as well as the theoretical expectation of a transition to Brownian motion around 20 K.

Methodology
To clarify the discrepancy, the authors applied alternating current (AC) susceptibility measurements instead of passive noise spectroscopy. This approach was chosen because, in the temperature range where the Fluctuation-Dissipation Theorem holds, linear response measurements provide unbiased estimates of the spectral density $S(f, T)$ up to the highest measurable frequencies.

• Experimental Setup: Measurements were performed on a single crystal of $\text{Dy}_2\text{Ti}_2\text{O}_7$ aligned along the [111] direction. The study utilized a Quantum Design Physical Properties Measurement System (PPMS) with an ACMS option for frequencies $10^1 \le f \le 10^4$ Hz, as well as a custom-built high-frequency magnetic susceptometer for frequencies up to $f \sim 10^6$ Hz.
• Data Processing: The imaginary part of the intrinsic magnetic susceptibility, $\chi''(f, T)$, was measured over a temperature range from 2 K to 28 K. A demagnetization factor ($N=0.367$) was applied to correct for sample shape dependence.
• Analysis: The power spectrum $S(f, T)$ was derived from $\chi''(f, T)$ using the classical Fluctuation-Dissipation Theorem. The exponent $b(T)$ was extracted by fitting the data to a Biltmo-Henelius function, which describes the distribution of relaxation times, rather than relying solely on simple gradient fits of log-log plots.

Main Results

1. Correction of the Anomalous Exponent: The study demonstrates that previous noise-based measurements significantly underestimated the value of $b(T)$. The AC susceptibility data reveal that $b(T)$ is consistently much closer to 2 than previously reported.
2. Temperature-Dependent Evolution: The exponent $b(T)$ evolves smoothly from anomalous values at low temperatures toward the Brownian motion limit of $b=2$. The data confirm a transition to $b=2$ at approximately 20 K, consistent with neutron scattering results and theoretical expectations for the dense Coulomb liquid regime.
3. Identification of Systematic Errors: The authors attribute the underestimation of $b(T)$ in earlier noise studies to "aliasing" effects. Due to discrete time sampling and the associated discrete Fourier transform, high-frequency power-law tails in $S(f, T)$ can be distorted, leading to a "flattened" high-frequency tail in log-log plots and an artificially low exponent.
4. Sample Dependence: Although the general trend of $b(T)$ approaching 2 is robust, the absolute values of $b(T)$ exhibit sample dependence. Measurements on different samples with varying quality and isotopic composition yielded values dropping as low as 1.48, although the temperature dependence remained similar. This suggests that crystal defects influence the monopole population and diffusion, with the specific correlation (e.g., slower diffusion versus lower $b$) appearing more complex than previously assumed.
5. Relaxation Dynamics: The thermal evolution of the relaxation rate $\tau(T)$ suggests that effectively Arrhenius-activated dynamics (phonon-assisted spin flipping) only become relevant above approximately 11 K, consistent with earlier susceptibility and neutron studies.

Significance and Claims
The contribution establishes the correct form of the temperature-dependent exponent $b(T)$ characterizing monopole diffusion in the dense Coulomb liquid state of spin ice. By extending measurements to $10^6$ Hz, the authors clarify that the system behaves as a highly correlated state where multiple dynamic processes combine, leading to deviations from Brownian motion ($b=2$) up to approximately 20 K.

The authors modestly claim that their results do not exclude the validity of the "fractal landscape" picture at very low temperatures ($T < 1$ K), where the monopole gas is dilute and relaxation time scales are slow; in this regime, systematic errors due to aliasing may be reduced. However, for the regime between 1 K and 20 K, the study corrects the understanding of monopole diffusion, showing it is closer to Brownian motion than previously assumed. The work underscores the necessity of applying spectral filtering methods (as proposed by Kirchner) in noise measurements to avoid systematic errors and emphasizes that the spectral power density reflects a combination of all dynamic processes (including correlated quantum mechanical tunneling and phonon-assisted flipping) rather than a single mechanism.