Contribution to the study of aging in disordered systems

We study general mechanisms that lead to aging behavior of dynamics in random environments. Aging is observed in the behavior of correlation functions that never become independent of the age of the system. A universal approach to this problem was developed over the past decades: the behavior correlation functions can be linked to the long-time behavior of the clock process, which is the total time elapsed along the trajectory of the random motion. An elegant approach to studying clock processes was proposed by Gayrard (2010,2012). Here, the clock process is viewed as a partial sum process whose increments are dependent random variables and then convergence criteria, due to Durrett and Resnick (1978), are employed. This method was further developed by Bovier and Gayrard (2013). We extend the methods of Gayrard (2012) and Bovier and Gayrard (2013) and use our methods to study the aging behavior of various models. In the first part we establish criteria for the convergence of clock processes on sequences of finite graphs to extremal processes and improve results on extremal aging obtained by Ben Arous and Gun (2012). The second part deals with dynamics that are defined on infinite graphs. We introduce sufficient conditions for the clock process to converge to a subordinator and establish the existence of a normal aging regime in Bouchaud's asymmetric trap model on $Z^d$, for $d\geq 2$. In the third part of this thesis we consider Bouchaud's asymmetric trap model for $d\geq 3$, and its symmetric version for $d=2$. We prove the existence of an super-aging regime