For nearly a century, the Ginzburg-Landau paradigm provided the definitive playbook for understanding phases of matter. To identify a phase or detect a transition, you looked for a local order parameter—a macroscopic observable, like the alignment of spins in a ferromagnet, that emerges when a system spontaneously breaks a symmetry.
Topological phases of matter shattered this paradigm. States like quantum spin liquids, topological insulators, and fractionalized systems break no local symmetries and possess no local order parameters. As noted by Hae‐Young Kee et al. in their study of Topological Spin Liquid on the Hyperkagome Lattice of Na4Ir3O8, these exotic states hide their order globally, embedding it within the long-range quantum entanglement of the system's wavefunction. Because traditional experimental and numerical probes cannot directly "see" this global structure, researchers have had to invent entirely new diagnostic toolkits.
To understand why a system has shifted into a topological phase, we must look beyond individual points in space and instead measure the global "twist" of the wavefunction, the structure of its entanglement, or the mathematical anomalies that arise when a phase changes.
The Entanglement Spectrum as a Wavefunction Fingerprint
If local measurements fail, how do we extract a phase's topological identity? The answer lies in partitioning the system and observing how quantum information is shared across the boundary.
When you divide a quantum system into two regions, $A$ and $B$, you can calculate the reduced density matrix of region $A$ by tracing out region $B$. The logarithms of the eigenvalues of this matrix form the entanglement spectrum. Rather than collapsing this data into a single number like entanglement entropy, analyzing the spectrum itself reveals robust topological signatures.
In a seminal paper, Entanglement spectrum of a topological phase in one dimension, Frank Pollmann et al. demonstrated that the Haldane phase of spin-1 chains is uniquely characterized by a universal double degeneracy across its entire entanglement spectrum. This degeneracy is not an accident of a specific Hamiltonian; it is robustly protected by global symmetries (such as inversion or time-reversal). If you deform the system without breaking these symmetries or closing the energy gap, the double degeneracy persists. It acts as a clear, wavefunction-based fingerprint that distinguishes the topological phase from a trivial one without relying on local symmetry breaking.
Extracting Chiral and Anyonic Signatures from the Bulk
While the entanglement spectrum provides a powerful diagnostic, standard metrics like topological entanglement entropy (TEE) hit a fundamental limit: they only capture the total quantum dimension of a phase. As Suk Bum Chung et al. showed in Topological quantum phase transition in an exactly solvable model of a chiral spin liquid at finite temperature, TEE can fail to differentiate between distinct phases, such as Abelian and non-Abelian chiral spin liquids, which may share identical dimensions but possess entirely different anyonic statistics. Furthermore, TEE cannot capture the chiral central charge ($c_-$)—a measure of the directional flow of energy along a system's edge—without relying on complex geometries or physical boundaries.
To bypass this limitation, Isaac H. Kim introduced a symmetry-independent tool called the modular commutator in Chiral Central Charge from a Single Bulk Wave Function.
+-------------------+
| Region A |
| +-------------+ |
| | Region B | |
| | +-------+ | |
| | | Reg C | | |
+--+--+-------+--+--+
By considering three adjacent subregions ($A$, $B$, and $C$) in the bulk of a single 2D ground-state wavefunction, the modular commutator measures the non-commutation of the modular Hamiltonians associated with these regions.
Why does this work? In a chiral system, the entanglement is fundamentally "directional." Swapping the quantum states of adjacent subregions accumulates a geometric phase that mirrors the physical propagation of edge modes. By quantifying this non-commutation, the modular commutator extracts the exact chiral central charge directly from bulk entanglement properties, eliminating the need to simulate a physical edge.
Similarly, to isolate the unique signatures of anyonic quasiparticles, Yarden Sheffer et al. developed a method in Extracting topological spins from bulk multipartite entanglement that utilizes permutations across multiple replicas of the system. This multipartite approach isolates the topological spin of anyons, providing a mathematical sieve to separate phases that look identical under simpler entanglement metrics.
Unsupervised Machine Learning as an Anomaly Detector
When dealing with novel or highly frustrated quantum models, we often do not know what mathematical invariant or hidden string order to look for. In these scenarios, unsupervised machine learning can discover phase transitions by treating them as structural anomalies in the data.
Consider how an autoencoder or a Generative Adversarial Network (GAN) operates. If you train a neural network on physical configurations or density matrices drawn exclusively from a single, well-understood phase, the network learns to compress and reconstruct those specific patterns with high fidelity.
In Detection of Berezinskii-Kosterlitz-Thouless transition via Generative Adversarial Networks, Daniele Contessi et al. applied this logic to the entanglement spectrum. Because a topological phase transition alters the global structure of the wavefunction, crossing the phase boundary introduces configurations that the network has never encountered. The GAN's reconstruction error suddenly spikes, acting as an automated anomaly detector that flags the transition point without requiring any prior knowledge of an order parameter.
[Physical States] ---> [ Trained Autoencoder ] ---> Low Reconstruction Loss (Same Phase)
[Topological State] -> [ Trained Autoencoder ] ---> HIGH Reconstruction Loss (Phase Transition)
This concept was expanded by Mohamad Ali Marashli et al. in Identifying quantum phase transitions with minimal prior knowledge by unsupervised learning. By feeding Reduced Density Matrices (RDMs) across a broad parameter space into an autoencoder, the authors demonstrated that the network naturally captures non-local features. When the system transitions into a topological or Berezinskii-Kosterlitz-Thouless (BKT) phase, the sudden shift in the RDM's entanglement structure causes a sharp peak in reconstruction loss, pinpointing the boundary.
Complementary approaches, such as the AC-GAN framework proposed by Jiewei Ding et al. in Rapid measurements and phase transition detections made simple by AC-GANs, leverage physical symmetries to drastically reduce the amount of training data needed, making unsupervised detection a highly efficient diagnostic tool.
Hidden Symmetries and Non-Local Indicators
Another way to detect topological order is to generalize our definition of symmetry. Traditional symmetries act on individual points in space (0-form symmetries). However, topological order can be protected or characterized by higher-form symmetries, which act on extended objects like lines or surfaces.
As explored by Ki‐Seok Kim et al. in Higher-form symmetries and d-wave superconductors from doped Mott insulators, determining whether these higher-form symmetries are spontaneously broken allows researchers to identify topological order in systems that appear completely featureless under standard local measurements.
This manifests physically as a hidden string order parameter, a concept originally understood in 1D chains and extended to more complex states like Pair Density Waves by Hao-Kai Zhang et al. in Pair density wave characterized by a hidden string order parameter. While a local measurement sees only random fluctuations because quantum fluctuations disorder the individual sites, a non-local string operator—which multiplies operators along an extended path—reveals a perfectly alternating, coherent pattern.
Finally, when even symmetry-based indicators lack the resolution to distinguish between subtle phases—such as differentiating a gapless Weyl semimetal from a gapped topological insulator, a limitation highlighted by Yuting Qian et al. in Weyl semimetals with S4 symmetry—mathematical frameworks like persistent homology offer a solution. As shown by Irene Donato et al. in Persistent homology analysis of phase transitions, tracking how the multi-dimensional "shape" and connectivity of a system's configuration space evolves allows researchers to map out phase boundaries in complex, disordered, or finite-size systems where traditional thermodynamic scaling breaks down.