Abstract
The distributions of singular thermodynamic quantities, on an ensemble of -dimensional quenched random samples of linear size near a critical point, are analyzed using the renormalization group. For much larger than the correlation length , we recover strong self-averaging (SA): approaches a Gaussian with relative squared width . For we show weak SA ( decays with a small power of ) or no SA [ approaches a non-Gaussian, with universal -independent relative cumulants], when the randomness is irrelevant or relevant, respectively.
- Received 1 August 1996
DOI:https://doi.org/10.1103/PhysRevLett.77.3700
©1996 American Physical Society