Random matrix theory: similarities and distinctions
For an NxN random matrix, the top eigenvalues behave as
The scaling with N suggests a correspondence:
If such correspondence holds, for the free energy of m non-intersecting paths, we would get
To reproduce the m^2 for free energy, we would need for eigenvalues of to behave as
Similar scaling with N=>t as expected for DPRM, but different scaling with eigenvalue index.
Sherry Chu, MIT thesis (2019) (offline) chapter
Density of states: The scaling with eigenvalue index is a reflection of the scaling of density of states at the edge:
(Sidenote: while the product of tridiagonal random matrices is not symmetric, all eigenvalues remain real and positive!)
Evolution of Log(eigenvalues) in "time":
Open questions:
Transition between and .
Transition between edge and middle Log(eigenvalues).
"Particle-hole" symmetry between smallest and largest Log(eigenvalues).
Is the arrangemet of Log(eigenvalues) hyperuniform?