---
title: Local Semicircle Law in Random Matrix Theory
url: https://www.emergentmind.com/topics/local-semicircle-law
type: topic
---

# Local Semicircle Law in Random Matrix Theory

The local semicircle law is a central result in random matrix theory, describing the precise local behavior of the empirical spectral distribution of large random matrices, most notably Wigner matrices, on scales much finer than the global (macroscopic) regime. It provides quantitative bounds demonstrating that, with high probability, the Stieltjes transform of the empirical spectral distribution, the entries of the resolvent, and more generally quadratic forms in the resolvent converge to their deterministic counterparts given by Wigner’s semicircle law, all the way down to scales matching or slightly above the typical eigenvalue spacing. This convergence underpins universality of local spectral statistics, eigenvalue rigidity, and complete eigenvector delocalization in a broad variety of random matrix ensembles.

## 1. Definitions and Model Setup

Let $H=(h_{ij})$ be an $N\times N$ Hermitian (or real symmetric) random matrix satisfying the “Wigner matrix” conditions:
- The upper-triangular entries are independent (up to symmetry), centered, and normalized: $\mathbb{E} h_{ij} = 0$, $\mathrm{Var}(h_{ij}) = 1/N$ for $i \neq j$, $\mathrm{Var}(h_{ii}) = 2/N$ [1110.6449].
- Higher moments are uniformly bounded, or satisfy subexponential decay [1601.04055, 1212.0164].

For such a matrix, the resolvent (Green’s function) at spectral parameter $z = E + i\eta$ ($\eta>0$) is $G(z) = (H - z I)^{-1}$, with normalized empirical Stieltjes transform $m_N(z) = N^{-1} \mathrm{Tr}\, G(z)$. The semicircle law has density $\varrho_{sc}(x)=\frac{1}{2\pi}\sqrt{(4-x^2)_+}$ and Stieltjes transform $m_{sc}(z)$. The Stieltjes transform is the unique solution to $m_{sc}(z) + 1/m_{sc}(z) + z = 0$, $\Im m_{sc}(z)>0$.

The spectral window in which the law operates is typically $|E| \leq \Sigma$ for some fixed $\Sigma>2$, and $\eta \in [N^{-1+\epsilon}, \Sigma]$ for arbitrary small $\epsilon>0$ [1601.04055].

## 2. Statement of the Local Semicircle Law

The prototypical local semicircle law has the following form. Let
\[
\Psi(z) = \sqrt{ \frac{ \Im m_{sc}(z) }{ N\eta } + \frac{1}{N\eta} }
\]
Then for all $z$ in the spectral domain and with very high probability,
\[
| m_N(z) - m_{sc}(z) | \prec \frac{1}{N\eta} 
\]
\[
\max_{i,j} |G_{ij}(z) - \delta_{ij} m_{sc}(z) | \prec \Psi(z)
\]
where $\prec$ denotes stochastic domination: for any $\varepsilon, D>0$, $\mathbb{P}( |X| > N^\varepsilon Y ) \leq N^{-D}$ for large $N$ [1601.04055, 1212.0164].

The same optimal bounds hold for a wide variety of Wigner-type ensembles, generalized Wigner matrices with inhomogeneous variances, random regular graphs, sparse Erdős–Rényi graphs [1503.08702, 1103.1919], random band matrices with sufficient bandwidth, ensembles with additional symmetries [1506.04683], and certain correlated and exchangeable ensembles (Curie–Weiss type) [1907.08782].

## 3. Methods of Proof and Key Technical Ingredients

The proof of the local semicircle law proceeds by an intricate blend of resolvent expansions, large deviation estimates, self-consistent equations, and stability analysis.

1. **Resolvent/Schur complement expansion:** Each diagonal entry $G_{ii}(z)$ satisfies
\[
G_{ii}(z) = \frac{1}{ H_{ii} - z - \sum_{k,\ell\neq i} H_{ik} G^{(i)}_{k\ell}(z) H_{\ell i} }
\]
where $G^{(i)}(z)$ is the resolvent of the minor with the $i$-th row and column removed [1601.04055].

2. **Self-consistent equation:** By averaging over $i$, an approximate quadratic equation for $m_N(z)$ is derived. For Wigner case,
\[
1 + z m_N(z) + m_N(z)^2 = O(\Psi(z))
\]
which is compared via stability estimates to the deterministic equation for $m_{sc}(z)$ [1601.04055].

3. **Large deviation control:** Quadratic forms in the random entries are bounded uniformly in high moments by subexponential or appropriately controlled algebraic tails [1212.0164, 1601.04055].

4. **Fluctuation averaging:** Due to independence and centering, averages of fluctuation errors (e.g., $\frac{1}{N}\sum_i Q_i$ for certain quadratic forms $Q_i$) are smaller by a factor $(N\eta)^{-1/2}$ than individual terms, a crucial input to descend the error from $O(\Psi(z))$ to $O((N\eta)^{-1})$ in $| m_N - m_{sc}|$ [1601.04055, 1212.0164].

5. **Multiscale bootstrapping:** Control at large $\eta$ is iteratively propagated to the fine scale $\eta\sim N^{-1+\epsilon}$ by continuity in $\eta$ and careful tracking of probability estimates [1601.04055].

6. **Stability analysis:** The nonvanishing derivative of the self-consistent equation ensures that the deviation $| m_N(z) - m_{sc}(z) |$ remains tightly controlled once errors in the Schur complement are small compared to the stability radius $(\kappa+\eta)^{1/2}$, where $\kappa = ||E|-2|$ [1110.6449].

7. **Isotropic law and extension:** For any deterministic unit vectors $v,w$, the isotropic extension controls
\[
|\langle v, (G(z) - m_{sc}(z)I ) w \rangle| \leq (\log N)^{C} \Psi(z) \|v\| \|w\|
\]
with high probability, in particular under vanishing third moment or further spectral restrictions [1110.6449].

## 4. Regimes, Optimality, and Extensions

The local semicircle law is optimal down to the scale $\eta \gtrsim N^{-1+\epsilon}$ (for arbitrary small $\epsilon>0$), corresponding to spectral windows containing $N\eta \sim N^{\epsilon}$ eigenvalues—barely more than one. Near the spectral edge and for more general ensembles (e.g., with nontrivial variance profiles), additional factors reflecting spectral stability or the inhomogeneity parameter $M$ are introduced [1212.0164, 1311.2016]. The law remains valid:
- For generalized Wigner ensembles with varied variances, with the deterministic bound $1/(M\eta)$ replacing $1/(N\eta)$ [1212.0164].
- In sparse or dependent models—Erdős–Rényi graphs, random $d$-regular graphs, Curie–Weiss ensembles, and band matrices under suitable conditions [1103.1919, 1503.08702, 1907.08782].
- For finite moment conditions ($4+\delta$ moments suffice; see ongoing refinement to minimal fourth moment with logarithmic corrections) [1602.03073, 1510.07350, 1904.08723].

In deformed Wigner models $H = W + \lambda V$ (with $V$ diagonal, independent of $W$), a local version of the deformed semicircle law holds: the spectral density converges locally to the free convolution of the semicircle law and the law of $V$ [1302.4532].

At the spectral edge, dedicated combinatorial and moment methods yield local law and rigidity of order $n^{-2/3+\epsilon}$ for GUE/GOE and Gaussian $\beta$-ensembles [1111.1351, 1112.2016].

## 5. Applications: Universality, Rigidity, and Delocalization

The local semicircle law has several key corollaries:

- **Eigenvector Delocalization:** All eigenvectors are completely delocalized: with high probability, $\max_{j,k}|u_j(k)|^2 \leq C (\log N)^C / N$ [1601.04055, 1110.6449, 1511.00862, 1904.08723].
- **Eigenvalue Rigidity:** The random eigenvalues $\lambda_j$ are close (within $O(N^{-2/3} (\min(j,N+1-j))^{-1/3} (\log N)^C)$) to their classical locations under the semicircle law; the same holds with appropriate modifications for generalized and deformed models [1601.04055, 1212.0164, 1511.00862].
- **Universality of Local Spectral Statistics:** Together with Green function comparison, the local law enables detailed control and universality results for $n$-point correlation functions and eigenvalue gap distributions, conditional on matching low-order moments [1601.04055, 1212.0164, 1110.6449].

In random regular graphs and Erdős–Rényi graphs with $pN\to\infty$, analogous statements imply optimal delocalization and semicircle law down to scale $N^{-1}$ (up to logarithmic corrections) [1503.08702, 1103.1919]. For correlated Curie–Weiss-type models with slow correlation decay, the same optimal entrywise law holds under de Finetti-type assumptions [1907.08782].

## 6. Variants, Generalizations, and Limitations

The local semicircle law extends to:
- Symmetry classes with additional constraints (e.g., fourfold symmetry) [1506.04683].
- Generalized variance profiles and doubly stochastic matrices, yielding optimal bounds for band matrices and covariance-type (Marchenko–Pastur) models at the hard edge [1212.0164, 1311.2016].
- Tridiagonal models for $\beta$-ensembles via moment and resolvent expansion methods [1111.1351, 1112.2016].
- Deterministic matrices, where the law quantifies deviation from the semicircle law purely in terms of deterministic stability parameters and matrix block expansions [1308.4668].

Limitations are present for matrices with heavy tails lacking the required moment conditions, or ensembles with a strong mean-field component, or insufficient independence. Rigorous extensions to $\eta \sim N^{-1}$ with no logarithmic loss remain open in low-moment/no subexponential scenarios [1602.03073, 1904.08723].

## 7. Summary Table of Core Results

| Ensemble                       | Entrywise Bound                          | Averaged Bound                     | Scale               | Authors / arXiv id      |
|-------------------------------|------------------------------------------|------------------------------------|---------------------|------------------------|
| Wigner matrix                 | $\Psi(z)$                                | $(N\eta)^{-1}$                     | $\eta \gtrsim N^{-1+\epsilon}$ | [1601.04055]          |
| General variance profile      | $\Psi(z)$ w/ $M$                         | $(M\eta)^{-1}$                     | $\eta \gtrsim M^{-1+\epsilon}$ | [1212.0164]           |
| $d$-regular graphs            | $(N\eta)^{-1/2} + d^{-1/2}$              | as above                           | $\eta \gtrsim d^{-1}/N$        | [1503.08702]          |
| Sparse Erdős–Rényi            | $q^{-1} + (N\eta)^{-1/2}$                | as above                           | $q \gtrsim (\log N)^{C}$       | [1103.1919]           |
| Deformed Wigner               | $(N\eta)^{-1/2}$                         | $(N\eta)^{-1}$                     | $\eta \gtrsim N^{-1+\epsilon}$ | [1302.4532]           |
| Fourth moment only            | $(N\eta)^{-1/2} \log N$                  | $(N\eta)^{-1} \log N$              | $\eta \gtrsim (\log N)^2/N$    | [1904.08723]          |

All error bounds hold with overwhelming or $\zeta$-high probability. $\Psi(z)$ denotes the typical control parameter $\Psi(z)=\sqrt{\Im m_{sc}(z)/N\eta+1/N\eta}$. Further details on conditions, technical hypotheses, and exceptions are found in the corresponding references.

## References

- Isotropic semicircle law and deformation: [1110.6449]
- Local semicircle law for random regular and sparse graphs: [1503.08702], [1103.1919]
- General Wigner and variance profile matrices: [1601.04055], [1212.0164], [1311.2016]
- Local law in bulk and edge for $\beta$-ensembles: [1111.1351], [1112.2016], [1104.3431]
- Moment assumptions and fourth moment law: [1602.03073], [1511.00862], [1510.07350], [1904.08723]
- Fourfold symmetry and correlated models: [1506.04683], [1907.08782]
- Deterministic local law: [1308.4668]
- Deformed Wigner: [1302.4532]

These results collectively constitute the modern, highly quantitative understanding of the local spectral statistics for large random matrix ensembles and their deterministic analogues.

Source: https://www.emergentmind.com/topics/local-semicircle-law