eduKate Learning Manual
Science | Edge Cases Science | Physical World
Understand → Observe → Explain → Test → Transfer → Go Deeper
Anderson Localization
How Disorder Can Stop a Wave From Spreading
Wait, What? More Randomness Can Stop Transport Without Absorbing the Wave
Suppose a wave moves through a messy material containing many irregular scatterers. A simple intuition says the wave should bounce around randomly, diffuse slowly and eventually spread everywhere.
Quantum and wave physics allow a sharper outcome. If scattering remains coherent, the many possible paths interfere. For sufficiently strong disorder, that interference can suppress transport so strongly that the wave remains confined to a finite region.
the wave is not trapped by a wall and not removed by absorption; its own multiply scattered paths interfere in a way that prevents diffusion.
Quick Answer
Anderson localization occurs when coherent multiple scattering from disorder creates enough destructive interference to suppress long-range wave transport. In a localized regime, an eigenstate can decay approximately exponentially away from a localization centre:
|ψ(r)| ∝ exp(−r/ξ)
where ξ is the localization length.
The phenomenon was introduced for electrons in disordered solids but is a wave-interference effect, so related localization appears with light, sound, microwaves and ultracold matter waves. In three dimensions, disorder can produce a mobility edge: states on one side are extended while states on the other are localized.
Nature Physics — Measurement of the Mobility Edge for 3D Anderson Localization →
Nature Physics (2026) — Observation of Exact Quantum Critical States →
What You Will Learn
- Why ordinary random-walk diffusion is not the whole story for coherent waves.
- How multiple scattering creates interfering paths.
- Why time-reversed paths reinforce return probability.
- What localization length means.
- How an extended state differs from a localized state.
- What a mobility edge is.
- Why dimensionality matters.
- How decoherence weakens localization.
- Why localization is not absorption.
- How experiments distinguish localization from slow diffusion.
- Why quasiperiodic and non-Hermitian systems require careful naming.
- How modern 2026 experiments probe critical and mobility-edge states.
Part 1 — The Naive Model: Disorder Just Makes Diffusion Slower
Classical diffusion treats repeated scattering as a random walk. Each collision changes direction, and after many collisions the probability cloud spreads with a width that grows roughly like the square root of time.
This assumes the different routes can be added as probabilities.
Coherent waves require us to add amplitudes first. Their phases matter, and different histories can cancel or reinforce.
Part 2 — One Destination Can Have Thousands of Paths
A wave entering a disordered medium can scatter from object A, then C, then B; or B, then C, then A; or follow countless other sequences.
The total amplitude at one point is a coherent sum over these paths.
Most pairs of paths have unrelated phase and partly cancel when averaged. Some special pairs are strongly correlated.
Part 3 — Return Paths Are Special
Consider a multiple-scattering path that eventually returns near its starting point. Its time-reversed counterpart visits the same scatterers in reverse order.
In a time-reversal-symmetric setting, these two paths acquire the same phase. They therefore interfere constructively for return.
This increases the probability of remaining near the starting region and produces the phenomenon known as weak localization before full Anderson localization sets in.
Part 4 — Stronger Interference Can Stop Diffusion
As disorder increases, repeated returns and interference corrections become stronger.
Beyond a threshold in appropriate dimensions and models, the diffusion coefficient effectively flows toward zero at large scales. A wave packet no longer expands indefinitely.
The resulting stationary states can be exponentially localized.
Part 5 — Localization Length Is the Size of the Trapped Wave
The localization length ξ tells us how far a localized state extends before its amplitude becomes exponentially small.
Strong localization means small ξ. Near a localization transition, ξ can become very large and may diverge at the critical point.
This gives a higher-resolution distinction than simply saying “trapped” or “not trapped.”
Part 6 — Mobility Edge: Same Material, Different Energies, Different Fate
In three-dimensional Anderson models, not every energy must localize at the same disorder strength.
A mobility edge separates localized states from extended states in energy.
Ultracold-atom experiments have measured such a mobility edge by preparing atoms in a laser-speckle disorder potential and tracking whether different energy components remain confined or expand.
Part 7 — Dimensionality Changes the Story
Localization depends strongly on dimension, symmetry and the statistical structure of disorder.
For ideal non-interacting particles with uncorrelated disorder, scaling theory predicts particularly strong localization tendencies in one and two dimensions, while three dimensions allow a genuine disorder-driven localization transition with a mobility edge.
Real systems can deviate through correlations, long-range hopping, interactions, topology or open-system effects.
Part 8 — Why This Is Not Absorption
Absorption removes wave energy from the measured wave channel and converts it into heat, internal excitation or another form.
Anderson localization can occur in an ideal lossless wave equation. The energy or probability remains present but spatially confined by interference.
A real experiment may contain both absorption and localization, so scientists must separate the two quantitatively.
Part 9 — Why This Is Not a Band Gap
A photonic crystal or periodic solid can forbid propagation in certain frequency bands because periodic structure creates a band gap.
Anderson localization does not require periodicity. It is generated by disorder and interference.
A state can exist at an allowed energy yet still fail to transport because it is localized.
Part 10 — Coherence Is Essential
The interference correction requires stable phase relationships among multiple paths.
Strong inelastic scattering, thermal dephasing, time-dependent disorder or environmental noise can destroy phase memory and restore more classical diffusion.
This is why low temperatures are often important for electronic Anderson-localization experiments, while optical and acoustic analogues can preserve coherence in other ways.
Part 11 — How Do We Know a Wave Is Localized Rather Than Merely Slow?
A slowly diffusing packet can look stationary over a short observation time. That is not sufficient evidence.
- Measure wave-packet width over increasingly long times.
- Look for saturation rather than continued diffusive growth.
- Measure spatial profiles and test exponential tails.
- Change system size and test finite-size scaling.
- Extract localization length or inverse participation ratio.
- Vary energy to locate a mobility edge.
- Measure absorption independently.
- Vary coherence or dephasing and test whether transport returns.
Part 12 — Critical States Live Between Localized and Extended
At a localization transition, wavefunctions need not look like ordinary extended plane waves or simple exponential localized states.
They can be critical, with scale-dependent or multifractal structure.
Programmable quantum experiments reported in 2026 have directly probed such critical behaviour and anomalous mobility edges in quasiperiodic models.
This extends the educational boundary: “localized versus extended” is a useful first split, not the entire modern phase diagram.
Part 13 — Disorder-Free Localization Is a Naming Boundary
Modern systems can localize through quasiperiodicity, Stark potentials, topology or non-Hermitian effects without ordinary random disorder.
Those phenomena may share mathematical tools such as localization lengths and mobility edges, but they should not automatically be called classic Anderson localization.
The canonical Anderson job here is localization generated by coherent multiple scattering from disorder.
Failed Model → Better Model
| Naive model | Why it fails | Better model |
|---|---|---|
| Random scattering only slows diffusion. | Coherent paths interfere. | Add amplitudes and phase, not probabilities alone. |
| No transmission means absorption. | A lossless wave can remain localized. | Measure energy conservation and spatial confinement separately. |
| All energies behave alike. | 3D systems can have mobility edges. | Resolve transport versus energy. |
| Any localized wave is Anderson localized. | Band gaps, Stark localization and interactions can localize for other reasons. | Identify the actual localization mechanism. |
Observation vs Inference
- Observation: wave packets can stop expanding in sufficiently disordered coherent systems.
- Measurement: localized states show finite spatial extent and often exponential tails.
- Intervention: changing disorder or energy can cross a mobility edge.
- Inference: coherent multiple scattering suppresses diffusion through interference.
- Boundary: absorption, interactions, finite size and dephasing must be excluded or modelled before attributing confinement to Anderson localization.
Checkpoint Questions
- Why is classical random-walk diffusion incomplete for coherent waves?
- What is special about a path and its time reverse?
- What is localization length?
- What is a mobility edge?
- Why does dimensionality matter?
- Why is localization not absorption?
- Why can dephasing weaken localization?
- How would you distinguish localization from very slow diffusion?
- Why is a photonic band gap not the same mechanism?
- What are critical states?
Answer Key
Open after attempting the questions
- Wave amplitudes have phase and interfere.
- In a time-reversal-symmetric system they acquire matching phase and enhance return probability.
- The characteristic spatial decay length of a localized state.
- An energy separating localized and extended states.
- Interference corrections and localization transitions scale differently with spatial dimension.
- Localization can conserve the wave energy or probability while preventing transport.
- It destroys the phase relationships needed for coherent path interference.
- Track packet width for long times and test finite-size scaling/exponential profiles.
- A band gap comes from periodic spectral structure rather than random-disorder interference.
- States at the localization transition with scale-dependent structure between ordinary extended and localized limits.
Primary Science Bridge
- waves can interfere;
- random obstacles change motion;
- energy can remain present even when transport stops;
- patterns can depend on phase, not just force;
- simple random-walk models have limits.
Secondary and JC Bridge
| Core idea | Higher-resolution route |
|---|---|
| Scattering | Coherent multiple scattering |
| Interference | Path amplitudes and phase |
| Transport | Diffusion coefficient and conductance scaling |
| Localization | Localization length and inverse participation ratio |
| Transition | Mobility edge and critical states |
| Real materials | Dephasing, absorption and interactions |
Unfamiliar Transfer Challenge
A microwave pulse enters a random network and after a short time its spatial profile stops broadening. The signal amplitude slowly falls.
Do not call this Anderson localization from the photograph alone. Measure absorption independently, test long-time packet width, change disorder strength, look for exponential spatial profiles and test whether phase-breaking perturbations restore transport.
Deep Science Window — Scaling Theory
Localization can be studied by asking how dimensionless conductance changes as the imagined sample size increases. If larger samples become relatively better conductors, states are extended; if conductance collapses with size, states localize. At a critical point the scale flow changes character, giving a compact way to understand why three dimensions can support a mobility-edge transition.
Deep Science Window — Universality and Multifractality
Near a localization transition, microscopic details can become less important than symmetry class and dimensionality. Critical eigenstates often display multifractal statistics: their intensity is distributed across many scales rather than filling space uniformly or decaying with one simple length.
Evidence Boundaries
- Localization ≠ absorption.
- Slow diffusion ≠ proven localization.
- Anderson localization ≠ every kind of wave trapping.
- Mobility edge ≠ universal in every dimension or model.
- Critical state ≠ ordinary exponential localized state.
- Non-interacting Anderson model ≠ many-body localization.
Manual Summary — KNOW → CONNECT → EXPLAIN → APPLY → CHECK
KNOW: disorder, coherent scattering, interference, localization length, mobility edge, dephasing.
CONNECT: repeated scattering to many coherent paths, path interference to enhanced return, and enhanced return to suppressed diffusion.
EXPLAIN: how disorder can halt transport without absorbing a wave.
APPLY: diagnose whether confinement in a new wave system is truly Anderson localization.
CHECK: separate localization from absorption, finite-size trapping and alternative localization mechanisms.
Teaching Guide for Parents, Tutors and Teachers
Begin with a classical random-walk picture, then introduce phase as the missing variable. The strongest learning move is to ask why “more scattering” can change from slower diffusion into no asymptotic diffusion at all.
- Review wave interference.
- Build a random-scattering picture.
- Compare probabilities with amplitudes.
- Introduce time-reversed paths and enhanced return.
- Define localization length.
- Add the mobility edge.
- Use absorption and dephasing as competing explanations.
- Finish with the critical-state and model-boundary discussion.
Independent check: later give a confined-wave experiment and require students to list at least three alternative mechanisms before naming Anderson localization.
Safety boundary: authentic electronic, optical and ultracold-atom localization experiments use specialist laboratories. Classroom teaching should use safe wave simulations, low-power acoustic/microwave analogues, or published datasets.