Window measurements of simulations in random systems

Simulation studies of random and frustrated systems are plagued by severely slow dynamics and strong finite-size effects. We introduce a window-measurement method as a practical solution to these difficulties. In the method, we only observe physical quantities within a subsystem located in the whole system and scale them with the correlation length estimated in the subsystem. Both equilibrium and nonequilibrium data with different system sizes and different window sizes fall onto a single scaling function.We confirmed the validity of the method using a three-dimensional ±J Heisenberg spin glass model. The spin-glass and chiral-glass transition temperatures were estimated to be very close to each other.