In textbook solid-state physics, topology and localization are deeply intertwined through the lens of perfect crystal symmetry. By taking the Fourier transform of a periodic lattice, we map electronic wavefunctions into momentum space (the Brillouin zone), where we can calculate integer topological invariants like the Chern number. If this invariant is non-zero, it acts as a mathematical obstruction: it is physically impossible to construct a smoothly varying, exponentially localized basis of real-space wavefunctions—known as Wannier functions—for the occupied states.
But what happens when we remove the safety net of perfect periodicity? Real-world materials are plagued by disorder, amorphous structures, and strong interactions that completely destroy momentum space.
A growing body of research is rebuilding this bridge from the ground up. By shifting the focus from momentum space to real-space operators, researchers are proving how spatial localization dictates topological triviality, while simultaneously discovering exotic phases where localization and topology coexist in unexpected ways.
The Real-Space Dictionary: Wannier Functions Without Lattices
To understand topology without a lattice, we must look at how wavefunctions decay in real space. In a periodic system, a trivial topology means you can localized your electrons into neat atomic-like orbitals. In non-periodic systems, we can attempt to do the same by constructing generalized Wannier functions.
Instead of using Bloch states, the strategy relies on the projected position operator. Imagine taking the coordinate operator $X$ and projecting it onto the subspace of all occupied electronic states. If the system is a trivial insulator, the eigenvalues of this projected operator form well-defined clusters, corresponding to the centers of localized electronic wavefunctions.
In Stubbs et al. (2022), Kevin D. Stubbs and Alexander B. Watson formalized this intuition by introducing an "iterated projected position" algorithm. They proved that as long as there are uniform spectral gaps in these projected position operators, one can systematically construct exponentially localized Wannier functions in two dimensions and higher, even completely bypassing translational symmetry.
Periodic System: Momentum Space (Brillouin Zone) ──> Chern Number
│ (Obstruction)
▼
Non-Periodic: Projected Position Operators ──> Spatial Localization
This real-space framing allows for rigorous proofs of the localization-topology dichotomy. In a single-authored work, Stubbs (2024) established that if a non-periodic insulator admits a generalized Wannier basis with sufficient algebraic localization (specifically, decaying faster than $1/r^{2+\delta}$), its real-space topological invariant—the Chern marker—must vanish. This mathematically cements the principle that sufficient spatial localization implies topological triviality, generalizing a foundational theorem of solid-state physics to disordered media.
The Exception: Topologically Localized Insulators
While localization often implies triviality, the interplay between disorder and topology can produce startling paradoxes. In standard Anderson localization, strong disorder scatters particles so severely that all bulk wavefunctions become exponentially localized, turning the material into an conventional insulator with zero conductance.
However, Lapierre et al. (2022) uncovered a new class of matter that defies this boundary: Topologically Localized Insulators (TLIs).
The authors demonstrated that a system can be subjected to enough disorder to fully localize every single bulk state at all energy levels, yet still preserve a non-trivial topological invariant. Because the bulk is entirely localized, it cannot conduct electricity. Yet, the system still hosts robust, delocalized boundary states that yield a precisely quantized Hall conductance. This challenges the traditional paradigm by showing that complete bulk localization does not inherently force the entire system to become topologically trivial.
Standard Anderson Insulator: Disorder ──> Bulk Localized ──> Trivial Boundaries
Topologically Localized (TLI): Disorder ──> Bulk Localized ──> Quantized Boundary States
A similar expansion of topological boundaries occurs when systems are driven out of equilibrium. In periodically driven (Floquet) quantum systems, traditional conservation laws often break down. Nathan et al. (2021) showed that even when particle numbers are not conserved—such as in spin systems or superconductors—a hierarchy of many-body topological invariants can emerge. These Correlation-Induced Anomalous Floquet Insulators (CIAFIs) leverage multi-particle correlations to generate quantized magnetization responses, operating outside the boundaries of standard static equilibrium topology.
New Tools for Generalized Invariants
As the definitions of topology expand beyond clean, periodic crystals, calculating these invariants becomes a massive computational and mathematical challenge. Without momentum space, traditional integration techniques fail.
To address this, researchers are turning to geometric data analysis and machine learning. For instance, the concept of persistence can track how topological features survive across different scales. Mémoli et al. (2020) developed the theory of the Persistent Laplacian, a spectral operator whose nullity precisely maps to persistent Betti numbers. By utilizing Schur complements, the authors provided an efficient algorithmic pathway to compute topological persistence directly from data networks, bridging spectral graph theory and topology.
On the simulation front, identifying the topology of complex, multi-band disordered systems from raw quantum states is notoriously difficult due to "gauge freedom"—the arbitrary phase factor associated with wavefunctions. To solve this, Huang et al. (2025) engineered Gauge Equivariant Neural Networks.
Raw Hamiltonian Data ──> Gauge Equivariant Layers ──> Invariant Chern Number Prediction
By embedding gauge symmetry directly into the neural network's architecture, the authors ensured that the model's predictions are fundamentally immune to phase fluctuations. Remarkably, when trained only on simple, topologically trivial data, these networks successfully generalize to identify complex topological phases (such as non-zero Chern numbers) in both 2D and 4D systems, providing a scalable tool to chart the vast landscape of non-periodic topological matter.
Key Foundational Work
Hierarchy of many-body invariants and quantized magnetization in anomalous Floquet insulators (2021) — contributed 1 concept elements to this direction.
Persistent Laplacians: properties, algorithms and implications (2020) — contributed 1 concept elements to this direction.