Eigen-Microstate Approach in Complex Systems

• Eigen-Microstate Approach is a data-driven method that employs spectral decomposition of covariance matrices to extract dominant collective fluctuation modes from high-dimensional ensembles.
• It identifies emergent phase transitions by detecting condensation of the leading eigen microstate, serving as a model-independent order parameter with clear finite-size scaling behavior.
• EMA has been applied across disciplines—from Ising models and Lennard-Jones fluids to Earth-system dynamics—providing robust quantification of critical phenomena.

The Eigen-Microstate Approach (EMA) is a data-driven method for extracting, classifying, and interpreting collective fluctuation patterns (“eigen microstates”) in high-dimensional ensembles of configuration snapshots from both equilibrium and non-equilibrium complex systems. EMA is based on the @@@@3@@@@ of empirical covariance (or correlation) matrices formed from sampled microstates—be they configurations of many-body fluids, Earth-system observables, or event-by-event fluctuations in heavy-ion collisions. The approach introduces the spectrum of eigen microstate weights or amplitudes as intrinsic indicators of critical phenomena. In particular, condensation of the leading mode signifies the onset of an emergent phase and provides a model-independent order parameter, which can be tracked via controlled finite-size scaling analysis. EMA has been successfully applied to Lennard-Jones fluids, Ising/Potts models, Earth-system fields, self-organized criticality, and particle–physics event ensembles, offering a general and robust framework for detecting and analyzing phase transitions across disciplines (Yang et al., 10 Jan 2026, Hu et al., 2018, Zhang et al., 2023, Chen et al., 2021, Liu et al., 28 Dec 2025, Guo et al., 31 Jan 2026, Guo et al., 23 Oct 2025).

1. Mathematical Framework of Eigen Microstates

EMA begins from an ensemble of $M$ microstates, each represented as a high-dimensional vector in the system’s configuration space. For many-body systems (e.g., Ising models, fluids), this is typically a spin, density, or occupation vector; for Earth-system fields, it is a spatial field snapshot; for event-by-event fluctuations, it is a vector of binned observables (e.g., momentum-density fields).

The core object is the normalized ensemble matrix $A$, whose columns are the centered and normalized microstate vectors. The covariance or correlation matrix $C$ is constructed as $C = A^TA$ or $C = AA^T$, depending on whether one studies microstate-microstate or spatial correlations. The eigenvalue problem

$C \, u_I = \lambda_I \, u_I$

yields a set of orthonormal eigen microstates $\{u_I\}$ and associated eigenvalues $\{\lambda_I\}$.

Alternatively, the singular value decomposition (SVD)

$A = U \Sigma V^T$

provides singular values $\sigma_I$ whose squares are the eigenvalues of $AA^T$ or $A^TA$ and whose left and right singular vectors form the spatial and sample modes, respectively.

The normalized weights (probabilities or amplitudes) of the eigen microstates are given by

$w_I = \lambda_I / \sum_{J} \lambda_J$

and satisfy $\sum_I w_I = 1$. In the absence of condensation, all $w_I \to 0$ as $M \to \infty$. Condensation occurs when $\lim_{M \to \infty} w_{I^*} > 0$ for some $I^*$, indicating macroscopic occupation of a principal fluctuation mode (Hu et al., 2018, Yang et al., 10 Jan 2026).

2. Interpretation and Role as Order Parameter

The largest eigen microstate amplitude, $w_1$ (or $\lambda_1$), plays the role of an intrinsic, model-free order parameter. Its condensation signals the emergence of a dominant collective mode associated with phase transitions or critical phenomena. In second-order transitions, $w_1$ grows continuously and exhibits finite-size scaling governed by universal exponents: $w_1(t, L) = L^{-2\beta/\nu} F[t L^{1/\nu}]$ where $t$ is the reduced control parameter (e.g., temperature or density deviation from criticality), $L$ is the system size, and $\beta, \nu$ are the standard order parameter and correlation length exponents (Yang et al., 10 Jan 2026, Hu et al., 2018). At criticality, $\ln w_1 \sim -(2\beta/\nu) \ln L$.

The ratio of subleading to leading amplitudes, $R = w_2 / w_1$, exhibits crossing points (fixed points) when plotted versus control parameters for different system sizes, providing a scaling-independent method to locate critical points (Hu et al., 2018, Yang et al., 10 Jan 2026). In non-equilibrium and Earth-system cases, this order-parameter-like behavior is also manifest via the condensation and scaling of leading eigen patterns (Zhang et al., 2023, Chen et al., 2021, Guo et al., 31 Jan 2026).

3. Practical Algorithm and Computational Workflow

The general workflow is as follows:

1. Sampling and Preprocessing:
   • Collect $M$ microstate samples in the relevant space (spin configurations, density fields, particle-number fluctuations, etc.).
   • Center and normalize each microstate vector (e.g., subtract temporal/spatial mean, divide by standard deviation or total variance).
2. Ensemble Matrix Construction:
   • Form the data matrix $A$ of dimensionality $N \times M$, where $N$ is system size.
   • (For event ensembles, construct an $M \times M$ covariance matrix in event space.)
3. Spectral Decomposition:
   • Compute the eigenvalue decomposition (or SVD) to extract eigen microstates and their weights/amplitudes.
4. Sorting and Significance Ranking:
   • Sort eigenvalues/amplitudes in descending order; rank modes by weight.
   • Optionally, test statistical significance via null models or explained variance (Chen et al., 2021).
5. Finite-Size Scaling Analysis:
   • Repeat analysis for multiple system sizes $L$, assembling $w_1(L)$, $R(L)$, etc.
   • Fit to FSS forms to extract critical exponents and locate transition points.
6. Mesoscopic Pattern Retrieval:
   • Analyze and visualize leading eigen microstates to reveal emergent spatial, temporal, or structural patterns inaccessible to single snapshots.
7. Generalization:
   • The workflow applies to both equilibrium and non-equilibrium snapshots, with only the field and normalization details differing by context (Yang et al., 10 Jan 2026, Chen et al., 2021, Guo et al., 23 Oct 2025).

4. Applications Across Physical, Statistical, and Earth Systems

EMA has been applied to a wide range of systems:

Classical Statistical Ensembles: For Ising and Potts models, condensation of one or several leading weights signals the ferromagnetic or symmetry-broken phase. The FSS of $w_1$ recovers universal exponents; the ratio $w_2/w_1$ cleanly locates $T_c$ (Hu et al., 2018, Liu et al., 28 Dec 2025).

Fluids and Liquids: In Lennard-Jones fluids, the leading mode amplitude $\lambda_1$ serves as an order parameter for the liquid-gas transition, with direct finite-size scaling analysis determining $T_c$, $\rho_c$, and critical exponents $\beta, \nu$ in excellent agreement with the Ising universality class (Yang et al., 10 Jan 2026).

Earth System and Climate Data: Applied to global ozone mass mixing ratio fields, EMA separates dominant annual-cycle patterns from interannual or trend signals (ENSO, QBO modes) and quantifies their variance contributions. Modes are interpreted via correlation with physical indices (Chen et al., 2021).

Self-Organized Criticality and Avalanche Models: EMA quantifies the absorption-to-critical transition in sandpile models (BTW, Manna) by the condensation of the leading spatial/temporal eigenmodes. FSS of $w_1$ and associated exponents reveal universality class distinctions (Zhang et al., 2023).

Event-by-Event Fluctuations in Heavy-Ion Collisions: Treating each event as a microstate, EMA exposes the emergence of fractal, critical-patch patterns in the leading eigen microstate. The largest eigenvalue functions as an order parameter for critical fluctuations in the search for the QCD critical point (Guo et al., 31 Jan 2026, Guo et al., 23 Oct 2025).

5. Eigen-Microstate Entropy and Complexity Metrics

Beyond leading-mode condensation, EMA provides a rigorous definition of the entropy of statistical ensembles in terms of eigen microstate probabilities. The eigen microstate entropy (Shannon–Boltzmann form)

$S_{\mathrm{EM}} = -\sum_{I} w_I \ln w_I$

quantifies ensemble complexity and captures precursor signals to phase transitions. At criticality, $S_{\mathrm{EM}}$ exhibits singular scaling, $$S_{\mathrm{EM}} \sim \ln L + L^{-2\beta/\nu} \tilde{S}(t L^{1/\nu})$$, mirroring the nonanalytic behavior of thermodynamic entropy and order parameters (Liu et al., 28 Dec 2025). In non-equilibrium systems, a significant increase in $S_{\mathrm{EM}}$ has been observed to precede phase transitions, such as biomolecular condensate formation and El Niño onset, preceding conventional signals (Liu et al., 28 Dec 2025).

6. Critical Phenomena, Universality, and Limitations

EMA provides a unifying, model-agnostic framework for detecting phase transitions across both equilibrium and non-equilibrium settings. The condensation of a leading eigen microstate captures the breaking of symmetry and emergence of long-range order, while finite-size scaling of its weight recovers critical exponents characteristic of the underlying universality class (Hu et al., 2018, Yang et al., 10 Jan 2026).

Universality and scaling are validated by FSS collapse in simulations of Ising, Potts, and lattice gas models. Similarly, in self-organized criticality and heavy-ion collisions, scaling of eigen microstate amplitudes and fractal invariance across lattice resolutions are consistent with theoretical expectations.

EMA assumes sufficient ensemble sampling and appropriate normalization/scaling by system size. It is insensitive to short-range correlations and background noise, isolating only genuinely collective modes in the leading eigenvectors. Care must be taken in binning/discretization to avoid signal degradation, and the assignment of order parameter role to $w_1$ presupposes that collective fluctuations manifest as dominant principal components (Guo et al., 31 Jan 2026, Guo et al., 23 Oct 2025).

7. Significance, Perspectives, and Extensions

The Eigen-Microstate Approach constitutes a general-purpose, data-driven methodology for phase transition detection, spatial/temporal pattern extraction, universality analysis, and complexity quantification in complex systems. It requires no prior assumptions on the order parameter or nature of the transition, leveraging only the ensemble structure of observed or simulated microstates. Its key operational advantages include background filtering, statistical independence of extracted modes, applicability to both equilibrium and non-equilibrium settings, and direct interpretability in terms of phase structure and criticality (Yang et al., 10 Jan 2026, Liu et al., 28 Dec 2025).

EMA has been extended to a diverse set of systems, including molecular fluids, climate fields, neuronal cultures, and high-energy physics event data. The emerging focus includes cross-disciplinary applications (e.g., in network dynamics, biological aggregates, climate tipping), refinements for real-time analysis, and incorporation of temporal dynamical correlations (e.g., via dynamic mode decomposition).

Its proven utility in critical point searches, early warning detection, and universal scaling analysis establishes EMA as a central tool in the study of collective phenomena in high-dimensional stochastic systems.