Anderson localization is what disorder does to a wave: enough random scattering and the wave stops propagating, trapping in place as a mode that falls off exponentially from a center. Put audio frequency on the lattice and disorder becomes a sustain control. Raise it and each band's energy stops decaying and holds, with the top of the spectrum freezing first as a boundary sweeps down the band.
Where it comes from
Disorder turns a traveling wave into a trapped one.
In 1958, P.W. Anderson asked what happens to an electron moving through a crystal when the crystal is not perfect. In an ordered lattice, the electron occupies extended Bloch states: waves that spread coherently across the whole material, which is why ordered metals conduct. Introduce randomness in the on-site energies, the local potential each site presents, and the picture changes. Scattered partial waves interfere with their own time-reversed paths, a coherent backscattering that builds with each scattering event. Past a threshold the interference is destructive enough to halt transport entirely. The eigenstates stop being extended and become localized, pinned near a center with an amplitude that decays exponentially with distance.
- \(\psi(x)\)
- The wavefunction amplitude at position x. A localized state is concentrated near its center, not spread across the lattice.
- \(x_0\)
- The localization center. The state is trapped here rather than propagating away from it.
- \(\xi\)
- The localization length. Small \(\xi\) is a tightly trapped state; as \(\xi\) grows the state spreads and approaches extended behavior.
The governing model is the tight-binding Anderson Hamiltonian. Each lattice site carries a random on-site energy, and a hopping term lets amplitude move to neighbors.
- \(\epsilon_i\)
- The on-site energy at site i, drawn at random from a distribution of width \(W\). This randomness is the disorder.
- \(t\)
- The hopping amplitude between neighboring sites. It is what lets a wave travel, the ordering force.
- \(\langle ij \rangle\)
- A sum over nearest-neighbor pairs. Transport is local: amplitude moves one bond at a time.
Localization is a contest between two terms. The hopping \(t\) spreads amplitude; the random spread of on-site energies \(W\) frustrates it. The dimensionless ratio \(W/t\) sets the outcome. When \(W/t\) is small, order wins and states stay extended. When \(W/t\) is large, disorder wins and states localize. The transition between the two regimes is the metal-to-insulator transition Anderson identified, an insulator produced not by a band gap but by disorder alone.
Dimensionality is decisive, and this is the result that earned the scaling theory of localization its place. In one and two dimensions, the scaling argument of Abrahams, Anderson, Licciardello, and Ramakrishnan shows that any nonzero disorder localizes every state: there is no conducting phase to defend. In three dimensions a genuine transition survives, with a critical disorder \(W_c\) separating extended from localized states. The 1D and 2D fragility is exactly the property that makes the phenomenon useful as a one-dimensional frequency lattice: along a single axis, disorder always wins eventually, so a disorder control is a monotonic path from mobile to trapped.
The boundary in energy
Not every state localizes at once. A mobility edge separates them.
In three dimensions, below the full Anderson transition, localized and extended states coexist in the same disordered sample at different energies. The boundary between them is a critical energy called the mobility edge, \(E_c\). States in the band tails, far from the band center, localize first; states near the center stay extended until disorder rises further. As you cross \(E_c\), the localization length diverges with a power law.
- \(E_c\)
- The mobility edge. On one side states conduct; on the other they are trapped. Disorder moves it.
- \(\nu\)
- The critical exponent. The localization length grows without bound as the energy approaches the edge.
- \(\xi(E)\)
- The localization length at energy E. It is finite on the localized side and diverges at the edge.
The mobility edge is the part of the physics that gives the audio version its signature. A freeze that swept the whole spectrum at once would be a sample-and-hold, a flat snapshot. A freeze gated by a moving boundary is something else: a front that advances through the band as the control rises, locking one region while another is still moving.
Localization is measured, not only modeled, and that is the point: it is a property of waves in disorder, not of electrons specifically. Wiersma and colleagues reported localization of light in a strongly scattering semiconductor powder. The same trapping has been measured for ultrasound in elastic networks and for matter waves in ultracold-atom systems released into engineered random potentials. Sound is a wave; a frequency lattice with disordered band tunings is a legitimate place for the same interference physics to act.
The audio isomorphism
The band lattice is the medium. Disorder is the control.
Anderson Freeze splits the input into a bank of perceptual bands, spaced on a Bark/ERB scale so the lattice sites are laid out the way the ear resolves frequency. Each band is a resonant feedback loop: a site that can hold energy. With the loops tuned coherently the energy moves and decays the way an extended state propagates and leaks. The DISORDER control detunes the loops, scattering each band against its neighbors. As disorder rises the bands stop sharing energy cleanly, the loop modes localize, and the spectrum holds in place. The decay that would normally end the sound is replaced by sustain, because a localized mode does not leak away.
The mapping is term-for-term, and that is what makes it an isomorphism rather than a metaphor. The on-site energies of the Hamiltonian are the per-band loop tunings. The width \(W\) of their random spread is the DISORDER control. The hopping \(t\) is the coupling between adjacent bands. Extended states are a spectrum that still moves and decays; localized states are a frozen spectrum. The same \(W/t\) contest that decides metal versus insulator decides mobile versus frozen here.
High frequencies localize first. A high-frequency band has a shorter wavelength, so a fixed amount of detuning is a larger fraction of its scale, and its localization length \(\xi\) is correspondingly smaller. Short \(\xi\) means it reaches the trapped regime at lower disorder. The freeze front therefore enters at the top of the band and sweeps down as disorder rises: the treble locks first, then the front descends through the mids toward the bass. You hear the top of the spectrum lock in place while the low end is still alive, then the front advances downward. One honest note on naming. In a real disordered solid the mobility edge lives in energy, and the band tails localize before the band center, not the top of the spectrum before the bottom. The treble-first sweep here is a design choice, legible and musical, grounded in the genuine fact that shorter-wavelength bands localize sooner. It is consistent with localization, not a literal transcription of the energy-space edge, which is why the marker in the lab is labeled the freeze front rather than the mobility edge.
The CRYSTAL control sets the density-of-states shape, the distribution of how many band modes sit at each frequency and how their resonances are laid out across the lattice. In condensed matter the density of states determines where modes are available to localize; here it sculpts the texture and grain of the frozen bed, from smooth to edged to steep. It is the lattice geometry the disorder acts on.
One honest boundary: this is a structural transfer of the mechanism, not a real-time solve of the Anderson Hamiltonian. The plugin does not diagonalize a random matrix every block and read off its eigenstates. It builds a DSP structure whose governing behavior is the same contest between coupling and disorder, tuned so that the audible result, sustain that enters treble-first behind a descending front, is the result the physics predicts. The equation is the design, not a decoration bolted onto an unrelated effect.
Interactive lab
Raise the disorder. Watch the treble lock first.
Each column is a frequency band, low on the left and high on the right. At low disorder the bands carry extended traveling waves: broad, moving energy that would decay if you stopped feeding it. Raise DISORDER and the bands collapse, one region at a time, into sharp exponential peaks, \(e^{-|x|/\xi}\), that hold in place. The dashed marker is the freeze front, the audio analog of the mobility edge. It enters from the right and sweeps left through the band as disorder grows: everything above the front is frozen, everything below it is still mobile. The readouts track the frozen fraction, the front frequency on a 60 Hz to 16 kHz log axis, and the smallest localization length \(\xi\) in the band, a normalized model figure rather than a solved eigenstate width, which shrinks as the trapping tightens.
A second freeze
Measure a state often enough and it cannot evolve.
Anderson localization is one way to stop a wave. The quantum Zeno effect is a different one, and Anderson Freeze offers it as a second mode. Misra and Sudarshan formalized it in 1977: a quantum system that is measured continuously cannot leave its initial state. Each measurement projects the system back, and in the limit of frequent measurement the cumulative survival probability approaches one.
- \(\tau\)
- The interval between measurements. The shorter it is, the more often the state is re-pinned.
- \(\Delta H\)
- The energy spread driving evolution. Over a short interval it produces only a tiny, second-order departure.
- \(P\)
- The survival probability over total time T. As \(\tau \to 0\) the exponent grows and P approaches one: the state is frozen by observation.
The structural idea that carries into audio is measurement rate. In Zeno mode the engine holds the spectrum not through disordered feedback but by re-sampling it: per-band looping micro-snapshots that re-pin the captured spectrum on a fast cycle. Observe often and the sound cannot drift; observe less often and motion returns. This was a real engineering finding, not a flourish. An early build held each band with a one-sample sample-and-hold, which is a DC staircase at the measurement rate and buzzed at low frequencies. The fix was a per-band looping micro-buffer with an integer-cycle window and a crossfaded capture, which is the faithful reading of "re-pin the state" and removed the buzz.
The two freezes sound different because they trap by different physics. Localization holds with the resonant character of a feedback bed: it has body and a hint of ring, because the held energy still lives in tuned loops. Zeno holds flatter and harder, because the spectrum is pinned to a snapshot rather than sustained by resonance. The choice between them is a sound-design decision, and the difference is audible, which is the bar every Chiral mapping has to clear.
Localization
Disorder traps the band loops. A resonant bed with body and ring; high frequencies lock first behind a descending edge.
Zeno
Frequent re-sampling pins the spectrum. A flatter, harder hold with less resonant character, set by measurement rate.
The choice
Localization for a living, resonant freeze. Zeno for a still, pinned snapshot. Two physics, two textures, one capture.
Motion in tempo
A frozen spectrum does not have to be a dead one.
The risk with any infinite sustain is that it becomes a static pad and stops being musical. Anderson Freeze answers that with a living freeze. With SYNC engaged, the held spectrum refreshes on the tempo grid through a level-conserving crossfade: on each grid tick the old loop fades down across a raised-cosine window while the live input fades in on the same window, so the bed morphs and breathes in tempo instead of sitting still.
The crossfade is level-conserving by construction, which is the engineering that makes it usable. The fresh injection on each tick exactly replaces the feedback energy the window removes, so the mean level of the bed is pinned to the static-hold level and engaging SYNC does not jump the volume. The seam is hidden because the window is a smooth bell rather than a hard edit. PULSE DEPTH sets the intensity of that motion, not its sharpness: at depth zero the freeze is a true static hold, and raising it makes the breathing more pronounced while the crossfade stays seam-free at every setting.
Practical reading: freeze a pad, switch MODE to taste, then engage SYNC and bring PULSE DEPTH up from zero. The hold starts to move in tempo without changing level. Tune the DIV grid to set how fast it breathes.
Controls map
Each control reads back to the physics, then to the sound.
The engine groups into the freeze itself and the gestures that drive it. The mapping below reads control, then the physics it carries, then what you hear.
| Control | Physics | Sound |
|---|---|---|
| DISORDER | The width \(W\) of on-site energy spread; the freeze amount. | Raises sustain. Treble locks first, then the freeze descends. |
| CRYSTAL | The density-of-states shape; the lattice geometry. | Texture and grain of the bed, from smooth to edged to steep. |
| MODE | Which physics freezes: localization or quantum Zeno. | Resonant bed versus flatter, pinned snapshot. |
| DAMPING | Loss applied to the held modes at the top. | Low-pass on the freeze; lower is a darker hold. |
| TONE | Spectral tilt of the held bed. | Negative darkens, positive brightens. |
| COLOR | Added harmonic character on the wet bed. | Zero is clean; higher adds grit and color. |
| THAW | Decay time once a freeze is released. | How fast the hold melts when you let go. |
| FREEZE | One-shot capture of the current spectrum. | The hero gesture: grabs and holds the instant. |
| GRAB | How hard FREEZE charges the captured bed. | From a light grab to a fat, blooming capture. |
| ARM | Transient-gated capture. | Freezes on the next hit, not on a button press. |
| SYNC | Level-conserving tempo refresh of the hold. | A living, breathing freeze instead of a static pad. |
| PULSE DEPTH | The living-freeze pulse amount under SYNC. | Zero is a static hold; higher breathes harder in tempo. |
| HARM | Pins the frozen bed to a pitch via ROOT and CHORD. | Tunes a noisy freeze into a chord that sits in key. |
| LEVEL | Trim on the whole frozen bed, pre-mix. | Sits the freeze under or over the dry signal. |
| DUCK | A mix utility, not localization: ducks the bed under the input. | The freeze tucks under a live part, then swells back in the gaps. |
Anderson Freeze ships every control above.
Intellectual honesty
What the isomorphism claims, and what it does not.
The thesis is structural isomorphism: the organizing principles of a physical phenomenon map onto the organizing principles of a DSP architecture, producing behavior a conventional effect does not. That claim has to be kept honest, so here is the line.
Faithful. The contest between coupling and disorder that decides extended versus localized states is the contest that decides mobile versus frozen here. The treble-first freeze front follows from wavelength scaling, a design choice consistent with localization rather than a literal transcription of the energy-space mobility edge. The Zeno mode genuinely re-pins the spectrum on a measurement cycle. CRYSTAL is the density-of-states the disorder acts on. These are the mechanism, transferred.
Not claimed. The plugin does not solve a condensed-matter Hamiltonian in real time, does not simulate a physical sample, and is not a proof of Anderson localization in a material. It is a DSP structure governed by the same equations, designed so the audible consequence matches the physics. And one control is explicitly outside the physics: DUCK ducks the frozen bed under the live input so a part can sit on top, then swells it back in the gaps. It is a mix utility and models no localization behavior. We label it as such rather than dress it up as science.
References
Source material.
- Anderson, P. W. (1958). "Absence of diffusion in certain random lattices." Physical Review 109(5), 1492-1505.
- Abrahams, E., Anderson, P. W., Licciardello, D. C., and Ramakrishnan, T. V. (1979). "Scaling theory of localization: absence of quantum diffusion in two dimensions." Physical Review Letters 42(10), 673-676.
- Lee, P. A., and Ramakrishnan, T. V. (1985). "Disordered electronic systems." Reviews of Modern Physics 57(2), 287-337.
- Wiersma, D. S., Bartolini, P., Lagendijk, A., and Righini, R. (1997). "Localization of light in a disordered medium." Nature 390, 671-673.
- Lagendijk, A., van Tiggelen, B., and Wiersma, D. S. (2009). "Fifty years of Anderson localization." Physics Today 62(8), 24-29.
- Misra, B., and Sudarshan, E. C. G. (1977). "The Zeno's paradox in quantum theory." Journal of Mathematical Physics 18(4), 756-763.
- Itano, W. M., Heinzen, D. J., Bollinger, J. J., and Wineland, D. J. (1990). "Quantum Zeno effect." Physical Review A 41(5), 2295-2300.