Origin of end-of-aging and subaging scaling behavior in glassy dynamics

Linear response functions of aging systems are routinely interpreted using the scaling variable t(obs)/t(w)(mu), where t(w) is the time at which the field conjugated to the response is turned on or off, and where t(obs) is the "observation" time elapsed from the field change. The response curve obtained for different values of tw are usually collapsed using values of mu slightly below one, a scaling behavior generally known as subaging. Recent spin glass thermoremanent magnetization experiments have shown that the value of mu is strongly affected by the form of the initial cooling protocol [G.F. Rodriguez et al., Phys. Rev. Lett. 91, 037203 (2003)], and even more importantly [G.G. Kenning et al., Phys. Rev. Lett. 97, 057201 (2006)], that the tw dependence of the response curves vanishes altogether in the limit t(obs) >> t(w). The latter result shows that t(obs)/t(w)(mu) scaling of linear response data cannot be generally valid, thereby casting some doubt on the theoretical significance of the exponent mu. In this work, a common mechanism is proposed for the origin of both subaging and end of aging behavior in glassy dynamics. The mechanism combines real and configuration space properties of the state produced by the initial thermal quench which initiates the aging process.