Abstract:We consider large non-Hermitian random matrices $X$ with complex, independent, identically distributed centred entries and show that the linear statistics of their eigenvalues are asymptotically Gaussian for test functions having $2+\epsilon$ derivatives. Previously this result was known only for a few special cases; either the test functions were required to be analytic [Rider, Silverstein 2006], or the distribution of the matrix elements needed to be Gaussian [Rider, Virág 2007], or at least match the Gaussian up to the first four moments [Tao, Vu 2016; Kopel 2015]. We find the exact dependence of the limiting variance on the fourth cumulant that was not known before. The proof relies on two novel ingredients: (i) a local law for a product of two resolvents of the Hermitisation of $X$ with different spectral parameters and (ii) a coupling of several weakly dependent Dyson Brownian Motions. These methods are also the key inputs for our analogous results on the linear eigenvalue statistics of real matrices $X$ that are presented in the companion paper [Cipolloni, Erdős, Schröder 2019].
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper attempts to solve the Central Limit Theorem (CLT) for linear eigenvalue statistics of non - Hermitian random matrices. Specifically, the paper considers large non - Hermitian random matrices \(X\) with complex, independent, identically distributed and zero - mean elements, and shows that the linear statistics of its eigenvalues are asymptotically Gaussian when the test function has \(2+\epsilon\) - order derivatives.
### Background and motivation
1. **Strong correlation of eigenvalues**:
- The eigenvalues of a random matrix form a strongly correlated point process. One manifestation of this strong correlation is that the fluctuations of the linear statistics are abnormally small, making the eigenvalue process significantly different from the Poisson point process.
2. **Limitations of known results**:
- Previous results were limited to several special cases:
- The test function must be analytic [73],
- The distribution of matrix elements must be Gaussian [74],
- Or at least the first four moments match those of the Gaussian distribution [82], [56].
3. **New contributions**:
- This paper not only extends these results but also finds the exact dependence of the limiting variance on the fourth cumulant, which was previously unknown.
### Main results
1. **Central Limit Theorem**:
- For an \(n\times n\) non - Hermitian random matrix \(X\) with complex, independent, identically distributed and zero - mean elements, the linear statistic \(L_n(f)\) of its eigenvalues is asymptotically a complex Gaussian random variable \(L(f)\), where:
\[
L_n(f):=\sum_{i = 1}^n f(\sigma_i)-\mathbb{E}\left[\sum_{i = 1}^n f(\sigma_i)\right]
\]
It has zero expectation and the variance \(V_f\) is given by the following formula:
\[
V_f=\frac{1}{4\pi}\langle\nabla g,\nabla f\rangle_{L^2(D)}+\frac{1}{2}\langle g,f\rangle_{\dot{H}^{1/2}(\partial D)}+\kappa_4\left(\frac{1}{\pi}\int_D g(z)d^2z-\frac{1}{2\pi}\int_0^{2\pi}g(e^{i\theta})d\theta\right)\left(\frac{1}{\pi}\int_D f(z)d^2z-\frac{1}{2\pi}\int_0^{2\pi}f(e^{i\theta})d\theta\right)
\]
where \(\kappa_4\) is the fourth cumulant of a single matrix element.
2. **Multivariate Central Limit Theorem**:
- For multiple test functions \(f^{(1)},\ldots,f^{(p)}\), the vector \((L_n(f^{(1)}),\ldots,L_n(f^{(p)}))\) is asymptotically a multivariate complex Gaussian distribution with zero expectation and covariance given by \(C(f,g)\) as described above.
### Methods and techniques
1. **Local law**:
- Use the local law of the product of two resolvent operators, where one resolvent operator corresponds to a different spectral parameter of \(H_z\).
2. **Coupling weakly - dependent Dyson Brownian motions**:
- Deal with problems on the microscopic scale by coupling multiple weakly - dependent Dyson Brownian motions.
### Conclusion
The paper successfully proves the Central Limit Theorem for linear eigenvalue statistics of non - Hermitian random matrices and finds the exact dependence of the limiting variance on the fourth cumulant, which is helpful for understanding.