Wigner Surmise¶

Approximate unfolded nearest-neighbor level spacings in a random-matrix symmetry class by the exact small-matrix spacing law, capturing level repulsion with a simple normalized density.

Version: v2 · 2026-09-06 · History
Domain-specific #: 3115
Origin domain: physics
Subdomain: nuclear physics
Aliases: Wigner spacing surmise, Wigner distribution

Core Idea¶

The Wigner surmise approximates the distribution of unfolded nearest-neighbor spacings between correlated energy levels by using the exact spacing law of a 2×2 Gaussian random matrix in the relevant symmetry class. For the Gaussian orthogonal ensemble, the unit-mean density is p(s)=(π/2)s exp(−πs²/4), displaying linear level repulsion near zero and a Gaussian tail.^[1]

Although derived from the smallest matrix, the law closely approximates the large-matrix spacing distribution and became a practical signature of chaotic quantum spectra. It is not universal without qualification: levels must be unfolded to remove changing mean density, symmetry sectors must be separated, and other ensembles have different repulsion exponents. Integrable or uncorrelated spectra are often closer to exponential Poisson spacings.

Structural Signature¶

The ordered spectrum. Comparable levels are sorted within one symmetry sector.
The unfolding map. Local mean density is removed so spacings have unit mean.
The nearest-neighbor gaps. Adjacent unfolded levels yield nonnegative s values.
The ensemble symmetry class. Time-reversal and related symmetries select the repulsion exponent.
The 2×2 exact law. A tractable small random matrix produces a closed density.
The large-system approximation. The small-matrix law stands in for the closely matching asymptotic distribution.
The level-repulsion signature. Probability vanishes as s approaches zero.
The comparison baseline. Poisson or exact random-matrix statistics test the surmise's adequacy.

What It Is Not¶

Not the Wigner semicircle law. That describes global eigenvalue density, not local nearest-neighbor spacing.
Not exact for arbitrary large matrices. Its celebrated large-N use is an approximation.
Not valid on raw nonstationary spacings. Unfolding is required.
Not one formula for every symmetry class. Orthogonal, unitary, and symplectic ensembles differ.
Not evidence of chaos from a mixed spectrum. Unresolved symmetry sectors can distort spacing statistics.
Not a causal model of a nucleus. It is a statistical universality approximation.

Scope of Application¶

The surmise is literal in random-matrix and spectral analysis after symmetry separation and unfolding.

Compound-nuclear spectra. Comparing resonance spacings within fixed quantum numbers.
Quantum chaos. Contrasting chaotic level repulsion with integrable Poisson statistics.
Random matrices. Providing a simple proxy for exact spacing laws.
Mesoscopic physics. Characterizing correlated spectra under symmetry constraints.
Wave systems. Testing universal spacing behavior in acoustic or microwave spectra.
Pedagogy and diagnostics. Showing how symmetry class controls local eigenvalue repulsion.

Clarity¶

State the ensemble, symmetry sector, unfolding method, normalization to mean spacing, and whether the 2×2 law or an exact finite/large-N result is used. Report finite-sample uncertainty and compare with alternatives. Do not label the fit 'Wigner' while pooling distinct conserved quantum numbers.

Declare the symmetry class, unfolding convention, spacing normalization, and whether the object is a nearest-neighbor spacing or a spacing ratio. The familiar orthogonal, unitary, and symplectic forms have different small-spacing repulsion exponents and constants. Unit mean spacing is imposed after unfolding; raw spectra with varying density cannot be compared directly with the universal curve. The surmise is exact for the corresponding small matrix used in its derivation and approximate for large random matrices, even when the numerical agreement is excellent. It should not be called the Wigner semicircle law, which concerns global eigenvalue density. Finite-size effects, degeneracies, mixed symmetries, unresolved levels, and superposed independent sequences can change the observed distribution. A fit is evidence about correlations under a model, not a proof of a physical mechanism.

Manages Complexity¶

One elementary density captures a difficult many-level correlation and turns spectral irregularity into a testable histogram or likelihood. Its compactness makes it robust for reconnaissance. The same convenience can hide preprocessing and approximation error; unfolding, missing levels, mixed sequences, and correlated histogram bins often dominate the inference.

Exact large-ensemble spacing laws can be represented through integral operators and limiting processes that are costly to evaluate or explain. The Wigner surmise compresses much of their local behavior into a normalized one-variable density with the correct qualitative ingredients: vanishing probability at zero spacing, a symmetry-dependent repulsion power, and a rapidly decaying tail. This makes it a useful benchmark for histograms and first-pass reasoning. The compression also hides structure. Unfolding removes global density information, nearest-neighbor spacing discards longer-range correlations, and one fitted parameter can mask a mixture of sequences. Strong use therefore pairs the surmise with declared preprocessing and additional statistics. The abstraction manages complexity by preserving the local-correlation signature while explicitly surrendering exact finite-ensemble detail.

Abstract Reasoning¶

Select a homogeneous spectral sequence.
Estimate and remove the smooth counting function.
Normalize adjacent gaps to unit mean.
Choose the random-matrix symmetry class.
Compute the matching Wigner-surmise density.
Compare data with Wigner, Poisson, and if needed exact ensemble predictions.
Quantify missing-level and finite-sample sensitivity.
Restrict conclusions to the resolved symmetry and energy window.

Knowledge Transfer¶

The strict parent is Approximation: a low-dimensional exact model serves as a remarkably accurate surrogate for a harder large-system distribution. Probability is related, but the identity lies in the small-to-large approximation. Spectral uses outside random-matrix prerequisites are analogies unless they preserve unfolding and symmetry class.

Approximation is the strict parent because a tractable exact result for a small random matrix stands in for a harder limiting spacing distribution. The transferable pattern is solve a symmetry-preserving minimal case → normalize shared invariants → use it as a high-quality surrogate. Transfer is warranted only when the minimal case preserves the mechanism that dominates the target statistic. It fails if agreement in one marginal distribution is treated as full ensemble equivalence. The domain residual includes random-matrix symmetry, spectral unfolding, level repulsion, nearest-neighbor order, and unit-mean normalization.

Examples¶

Canonical¶

For a real symmetric 2×2 Gaussian matrix, changing variables from matrix entries to mean level and spacing yields the GOE unit-mean density p(s)=(π/2)s exp(−πs²/4). The factor s suppresses nearly coincident levels. Larger GOE matrices do not have exactly this law, but their nearest-neighbor distribution is numerically very close.^[1]

Mapped back: 2×2 Gaussian ensemble → eigenvalue gap → normalization → exact repulsive density → large-N surrogate.

Applied / In Practice¶

A quantum-billiard study separates modes by symmetry, unfolds each sequence, and compares nearest-neighbor spacings with Wigner and exponential baselines. A Wigner-like fit after separation supports random-matrix-type spectral correlations; the same fit before separation would be uninterpretable because independent symmetry sequences can cross.

A researcher obtains a sequence of measured levels whose average density changes across the observation window. The levels are unfolded with a declared smooth density estimate, adjacent spacings are normalized to mean one, and the histogram is compared with the appropriate surmise and with an uncorrelated Poisson benchmark. A deficit of very small spacings supports level repulsion, but residual disagreement prompts checks for missing levels, symmetry mixing, and finite sample size. The report does not claim exact random-matrix membership from one curve. It treats the surmise as a compressed local diagnostic and preserves the preprocessing choices needed to reproduce the comparison.

Mapped back: measured spectrum → symmetry split → unfolding → spacing distribution → qualified chaos diagnostic.

Structural Tensions¶

Tiny model vs. large-system accuracy. The 2×2 law works far beyond its derivation. Diagnostic: Is approximation error relevant at the available precision?
Universal local law vs. system-specific unfolding. Universality appears only after a modeled density is removed. Diagnostic: How sensitive is the result to unfolding?
Level repulsion vs. symmetry mixing. Independent sectors can cross even when each repels internally. Diagnostic: Were all conserved labels resolved?
Simple histogram vs. finite-sample inference. Visual agreement can overstate evidence. Diagnostic: Are uncertainty and missing levels modeled?
Autonomous surmise vs. generic approximation. Approximation travels; random-matrix spacings define this case. Diagnostic: Does the surrogate come from the matching 2×2 ensemble?

Structural–Framed Character¶

The Wigner surmise is structural-leaning. The random-matrix implication is formal; ensemble selection and unfolding are modeler choices tied to physical symmetry. It is evaluatively neutral and observer-independent once those choices are fixed. It remains domain-specific because it concerns local eigenvalue statistics of particular ensembles.

Symmetry class, small-matrix derivation, unfolded nearest-neighbor spacing, normalization, repulsion exponent, and approximate large-ensemble role are structural. The physical spectrum, binning, sample size, numerical fitting method, and selected energy window are framed. Constants change with normalization, so a formula without its scale convention is incomplete. The same curve can be used descriptively across several systems, but the interpretation must remain conditional on completeness and symmetry. This framing separates the autonomous surmise from any one application in nuclear, atomic, condensed-matter, or dynamical spectra.

Structural Core vs. Domain Accent¶

The skeleton is tractable small exact model → normalized law → high-dimensional surrogate → residual check. The accent is Gaussian ensembles, spectral unfolding, nearest-neighbor levels, symmetry class, and repulsion. Removing these yields generic approximation.

The portable core is preserve a dominant invariant in a minimal solvable model → normalize → use the result as a surrogate for a harder distribution. The random-matrix accent is ordered unfolded eigenvalues, nearest-neighbor spacings, symmetry class, level-repulsion exponent, and ensemble asymptotics. Removing these features leaves Approximation or model reduction. Retaining only a fitted curve without the small-matrix and symmetry rationale leaves an empirical distribution family rather than the Wigner surmise. The residual therefore includes both its formula and why that formula is expected to track local spectral correlations. Exact agreement is never part of the recognition test for the large-ensemble use. The approximation's value lies in preserving the characteristic local shape with exceptional simplicity.

Approximation is the strict parent because the exact 2×2 spacing law is used as a good-enough representation of the harder large-matrix law. Probability describes the output but not the small-model surmise relation.

The prospective workspace queue contains one strict upward edge to prime:approximation. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Wigner Surmise Domain-specific

Parents (1) — more general patterns this builds on

Wigner Surmise is a kind of Approximation Prime

Approximation is the strict parent because the exact 2×2 spacing law is used as a good-enough representation of the harder large-matrix law.

Hierarchy path (1) — routes to 1 parentless root

Wigner Surmise → Approximation → Representation → Abstraction

Neighborhood in Abstraction Space¶

Wigner Surmise sits in a sparse region of the domain-specific corpus (87^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Stochastic Fields & Random-Matrix Dynamics (6 abstractions)

Nearest neighbors

Dyson Brownian Motion — 0.86
Correlation Dimension — 0.81
Monotone Likelihood Ratio Property — 0.79
Hyper-Wiener Index — 0.78
Hunt–Szymanski Algorithm — 0.78

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Wigner semicircle distribution. Global eigenvalue density rather than local spacing.
Exact GOE spacing distribution. The true large-ensemble law, closely approximated but not equaled by the surmise.
Poisson spacing law. Exponential no-repulsion baseline for uncorrelated levels.
Brody distribution. A phenomenological interpolation between Poisson and Wigner-like behavior.
Level-spacing ratio statistic. An unfolding-resistant statistic based on adjacent gaps.

References¶

[1] Madan Lal Mehta, Random Matrices, 3^rd ed. (Elsevier, 2004), chapters 1 and 7. registry ↩a ↩b