Positive specific-heat critical exponent of a three-dimensional three-state random-bond Potts model

Whereas the stability of a pure critical system is determined by the sign of its specific-heat critical exponent alpha according to Harris' criterion, whether a d-dimensional dirty system should satisfy nu >= 2/d and alpha < 0 or not has been a controversial issue for several decades, where nu is its correlation-length critical exponent. Here, contrary to recent analytical and numerical results, we find for the three-dimensional three-state random-bond Potts model whose pure version exhibits a first-order phase transition a random fixed point whose nu < 2/d and alpha > 0 using a finite-time scaling combining with extended dynamic Monte Carlo renormalization-group method. This suggests further studies are still needed to clarify the issue in three-dimensional systems. (C) 2012 Elsevier B.V. All rights reserved.