Skip to main content
added 10 characters in body
Source Link
Ben Deitmar
  • 1.3k
  • 1
  • 5
  • 15

Let $\lambda_1(A_n),...,\lambda_n(A_n)$ be the random eigenvalues of a random $(n \times n)$ matrix $A_n$. We can define the empirical spectral measure $\mu_n^{A_n}$ on $(\mathbb{C},\mathcal{B}(\mathbb{C}))$ by $\mu_n^{A_n} = \frac{1}{n} \sum\limits_{j=1}^n \delta_{\lambda_j(A_n)}$. For many interesting random matrix ensembles $(A_n)_{n \in \mathbb{N}}$ there are results of the form $$ n\big( \mu^{A_n}_n(f) - \mu(f) \big) \xrightarrow{n \rightarrow \infty}_{\mathcal{D}} X_f $$ for certain (sufficiently smooth) functions $f : \mathbb{C} \rightarrow \mathbb{R}$, a determinsitic measure $\mu$ and a Gaussian process $X_f$.

My question is if anyone knows if such a result is known for iid matrices, that is when the entries of $A_n$ are all independent identically distributed with mean zero and variance one$\frac{1}{n}$. By the Circular Caw we know, that in this case $\mu$ would need to be the uniform distribution on the unit sphere $B_1^{\mathbb{C}}(0)$.

An example for a random matrix ensemble, where such a result is known, is the Wigner ensemble. See Theorem 9.2 of Bai and Silversteins 'Spectral Analysis of Large Dimensional Random Matrices'.

Let $\lambda_1(A_n),...,\lambda_n(A_n)$ be the random eigenvalues of a random $(n \times n)$ matrix $A_n$. We can define the empirical spectral measure $\mu_n^{A_n}$ on $(\mathbb{C},\mathcal{B}(\mathbb{C}))$ by $\mu_n^{A_n} = \frac{1}{n} \sum\limits_{j=1}^n \delta_{\lambda_j(A_n)}$. For many interesting random matrix ensembles $(A_n)_{n \in \mathbb{N}}$ there are results of the form $$ n\big( \mu^{A_n}_n(f) - \mu(f) \big) \xrightarrow{n \rightarrow \infty}_{\mathcal{D}} X_f $$ for certain (sufficiently smooth) functions $f : \mathbb{C} \rightarrow \mathbb{R}$, a determinsitic measure $\mu$ and a Gaussian process $X_f$.

My question is if anyone knows if such a result is known for iid matrices, that is when the entries of $A_n$ are all independent identically distributed with mean zero and variance one. By the Circular Caw we know, that in this case $\mu$ would need to be the uniform distribution on the unit sphere $B_1^{\mathbb{C}}(0)$.

Let $\lambda_1(A_n),...,\lambda_n(A_n)$ be the random eigenvalues of a random $(n \times n)$ matrix $A_n$. We can define the empirical spectral measure $\mu_n^{A_n}$ on $(\mathbb{C},\mathcal{B}(\mathbb{C}))$ by $\mu_n^{A_n} = \frac{1}{n} \sum\limits_{j=1}^n \delta_{\lambda_j(A_n)}$. For many interesting random matrix ensembles $(A_n)_{n \in \mathbb{N}}$ there are results of the form $$ n\big( \mu^{A_n}_n(f) - \mu(f) \big) \xrightarrow{n \rightarrow \infty}_{\mathcal{D}} X_f $$ for certain (sufficiently smooth) functions $f : \mathbb{C} \rightarrow \mathbb{R}$, a determinsitic measure $\mu$ and a Gaussian process $X_f$.

My question is if anyone knows if such a result is known for iid matrices, that is when the entries of $A_n$ are all independent identically distributed with mean zero and variance $\frac{1}{n}$. By the Circular Caw we know, that in this case $\mu$ would need to be the uniform distribution on the unit sphere $B_1^{\mathbb{C}}(0)$.

An example for a random matrix ensemble, where such a result is known, is the Wigner ensemble. See Theorem 9.2 of Bai and Silversteins 'Spectral Analysis of Large Dimensional Random Matrices'.

moved meta info from title to tag, added tag
Link
YCor
  • 63.9k
  • 5
  • 187
  • 286

Reference Request: Spectral CLT for random matrices with iid entries

Source Link
Ben Deitmar
  • 1.3k
  • 1
  • 5
  • 15

Reference Request: Spectral CLT for random matrices with iid entries

Let $\lambda_1(A_n),...,\lambda_n(A_n)$ be the random eigenvalues of a random $(n \times n)$ matrix $A_n$. We can define the empirical spectral measure $\mu_n^{A_n}$ on $(\mathbb{C},\mathcal{B}(\mathbb{C}))$ by $\mu_n^{A_n} = \frac{1}{n} \sum\limits_{j=1}^n \delta_{\lambda_j(A_n)}$. For many interesting random matrix ensembles $(A_n)_{n \in \mathbb{N}}$ there are results of the form $$ n\big( \mu^{A_n}_n(f) - \mu(f) \big) \xrightarrow{n \rightarrow \infty}_{\mathcal{D}} X_f $$ for certain (sufficiently smooth) functions $f : \mathbb{C} \rightarrow \mathbb{R}$, a determinsitic measure $\mu$ and a Gaussian process $X_f$.

My question is if anyone knows if such a result is known for iid matrices, that is when the entries of $A_n$ are all independent identically distributed with mean zero and variance one. By the Circular Caw we know, that in this case $\mu$ would need to be the uniform distribution on the unit sphere $B_1^{\mathbb{C}}(0)$.