Nucleation and Growth of the Supercooled Liquid Phase Control Glass Transition in Bulk Ultrastable Glasses A. Vila-Costa , 1 J. R` afols-Rib´ e , 1,* M. González-Silveira , 1 A. F. Lopeandia , 1 Ll. Abad-Muñoz, 2 and J. Rodríguez-Viejo 1,† 1 Group of Nanomaterials and Microsystems, Physics Department, Universitat Autònoma de Barcelona, Bellaterra 08193, Spain 2 Instituto de Microelectrónica de Barcelona—Centre Nacional de Microelectrònica, Campus UAB, Bellaterra, Barcelona 08193, Spain (Received 3 August 2019; accepted 7 January 2020; published 21 February 2020) We report the anomalous bulk transformation of vapor deposited stable glasses into the liquid state. The transformation proceeds through two competing parallel processes: partial rejuvenation of the stable glass and nucleation and growth of liquid patches within the glass. The kinetics of the transformation extracted from heat capacity curves after isothermal runs is dominated by the heterogeneous nucleation and growth process that initiates at preexisting seeds and propagates radially at a velocity proportional to the alpha relaxation time. Remarkably, the distance between the activation seeds is independent of temperature within experimental uncertainty and amounts to several micrometers, a value in close agreement with the crossover length for TPD glasses. We speculate the initiation sites for the transformation of the glass into the supercooled liquid are localized regions of lower stability (or density). DOI: 10.1103/PhysRevLett.124.076002 After more than a century of intense research the glass transition is still poorly understood and many different theories have been devised to understand its phenomenol- ogy. The glass transition measured at the laboratory has no identifiable structural signatures but is the result of an impressive reduction of the dynamics of the supercooled liquid by 12–14 orders of magnitude in a relatively small temperature window [1]. The kinetic nature of the glass transition and the existence of memory effects that result in hysteresis upon cooling and heating add complexity to the analysis. One of the first attempts to comprehend the influence of the thermal history on the properties of the glass was the phenomenological description by the Tool- Narayanaswami-Moynihan model [2] that introduced mean-field equations of motion for the fictive temperature (T f ) of the glass. Later on, other approaches such as kinetic constraint models [3] or random first-order theory (RFOT) [4] have been able to accurately reproduce calorimetric measurements with variable heating and cooling rates [5,6]. These theories are based on active regions of high mobility that catalyze the mobility of the nearby low-mobility (inactive) regions leading to propagating fronts. The differ- ence in mobility is linked to a local field, spatially resolved, of the relaxation times (τ) or the local fictive temperatures in the glass. Previous experimental work has shown the transformation of highly stable thin film glasses into the supercooled liquid SCL proceeds through a moving front that propagates from surfaces to interfaces at a constant velocity that depends on the relaxation time of the liquid and on the stability of the glass [7,8]. If the film is thick enough or if the front is suppressed, the transformation occurs mainly in the interior of the glass [7,9]. The bulklike transformation of stable glasses has not been experimentally analyzed in detail yet. The only experimental work addressing it was in the original discovery of the front transformation where the kinetics of the bulk process was identified as Avrami type [7]. The transformation of stable glasses by moving fronts starting at surfaces, at inner regions of high mobility, or at nucleation sites has been discussed in detail in several theoretical and computational works [6,10–13]. Gutierrez and Garrahan [10] used a three dimensional East model with soft constraints to recreate the front and bulk transformation in stable glasses. Wolyness et al. [6,11] described the existence of moving fronts within RFOT by an analogy to a combustion process. On the other hand, Jack and Berthier used a triangular plaquette model to identify the transformation of stable glasses with a nucleation-and- growth process having large distances between nucleation events [12]. In a more recent work, the swap methodology [14] has enabled Berthier and co-workers to produce simulated glasses with stabilities that compare well to the highly stable glasses created in the laboratory by vapor deposition [15,16]. These simulated stable glasses transform into the liquid by moving fronts and if made thick enough a competing transformation between bulk and front or only a bulk process is observed [13]. Here, we report the kinetics of the bulklike transforma- tion into the supercooled liquid of highly stable glasses of N;N - -bisð3-methylphenylÞN;N - -bis(phenyl)-benzidine (TPD), glass transition temperature, T g ¼ 333 K, grown PHYSICAL REVIEW LETTERS 124, 076002 (2020) 0031-9007=20=124(7)=076002(5) 076002-1 © 2020 American Physical Society