Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $X$ be $n \times n$ matrix whose matrix elements are independent identically distributed normal variables with zero mean and variance of $\frac{1}{2}$. Then $$ A = \frac{1}{2} \left(X + X^\top\right) $$ is a random matrix from GOE ensemble with weight $\exp(-\operatorname{Tr}(A^2))$. Let $\lambda_\max(n)$ denote its largest eigenvalue. The soft edge limit asserts convergence of $\left(\lambda_\max(n)-\sqrt{n}\right) n^{1/6}$ in distribution as $n$ increases.

Q: I am seeking to get an intuition (or better yet, a simple argument) for why the largest eigenvalue scales like $\sqrt{n}$.

share|improve this question
add comment

2 Answers

up vote 2 down vote accepted

Consider a Frobenius norm squared of $A$: $$ \|A\|^2 = \sum_{i,j} a_{i,j}^2 = 2 \sum_{i<j}a_{i,j}^2 + \sum_{i} a_{ii}^2 = \sum_{i=1}^n \lambda_i^2 $$ The expected value of $\|A\|^2$ is easy to find: $$ \mathbb{E}\left( \|A\|^2 \right) = 2 \sum_{i<j} \mathbb{E}\left(a_{i,j}^2\right) + \sum_{i} \mathbb{E}\left( a_{ii}^2 \right) = 2 \sum_{i<j} \frac{1}{4} + \sum_{i} \frac{1}{2} = \frac{n(n+1)}{4} < n \mathbb{E}\left(\lambda_\max(n)^2\right) $$ this establishes $\lambda_\max(n)$ scales at least as $\sqrt{n}$

share|improve this answer
add comment

The scaling follows from the Wigner semicircle law. Proof of the Wigner semicircle law is outlined in section 2.5 of the review "Orthogonal polynomials ensembles in probability theory" by W. König, Probability Surveys, vol. 2 (2005), pp. 385-447.

share|improve this answer
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.