PHYSICAL REVIEW B 86, 064202 (2012) Time-resolved study of the crystallization dynamics in a metallic glass Michael Leitner, * Bogdan Sepiol, and Lorenz-Mathias Stadler † Universit¨ at Wien, Fakult¨ at f ¨ ur Physik, Strudlhofgasse 4, 1090 Wien, Austria Bastian Pfau TU Berlin, Institut f ¨ ur Optik und Atomare Physik, Hardenbergstraße 36, 10623 Berlin, Germany (Received 26 March 2012; revised manuscript received 11 May 2012; published 7 August 2012) We report a study of the atomic-scale dynamics in a metallic glass of composition Zr 65 Ni 10 Cu 17.5 Al 7.5 by x-ray photon correlation spectroscopy. Our results show a continuous slowing down of the dynamics from the pristine state of the sample until crystallization. We propose a phenomenological model in a framework of thermally activated dynamics with a decreasing attempt rate dependent on the sample state that quantitatively describes our results, giving an activation energy of E A = 1.95 ± 0.10 eV. This allows us to conclude that atomic motion and crystallization are manifestations of the same process, with the time scale of crystallization on the order of 100 local atomic rearrangements. This rules out the notion of equilibrium diffusion in a relaxed glassy state in this system. DOI: 10.1103/PhysRevB.86.064202 PACS number(s): 66.30.hh, 64.70.dg, 61.05.cf, 81.05.Kf I. INTRODUCTION For metallic systems in their conventional form, the crystalline phase, there is often a large gap between the theoretical and the actual values of parameters of technical importance, e.g., tensile strength or corrosion resistance. This is for the most part due to the unavoidable imperfections of the crystalline order such as grain boundaries, which are weak spots where cracks or corrosion can begin to degrade the integrity of the sample. Amorphous metals, on the other hand, are a very promising class of materials from this perspective. 1 They have no grain boundaries, so the actual mechanical strengths are close to their theoretical values, and due to the stability of the amorphous phase, it is also much harder for the corroding agents to make inroads. Preparing a metallic sample in the amorphous state is in principle not difficult. If a melt is cooled rapidly enough, the atoms do not have enough time to arrange in a crystal lattice and become frozen in the amorphous state. Due to the requirement of a high cooling rate, the achievable size of the amorphous samples is limited, however. In recent years, an ever-increasing number of systems has been discovered, where the necessary cooling rates allow the preparation of bulk samples. These are the so-called bulk metallic glasses, due to the above reasons one of the main emerging fields in materials science. With the invention of sophisticated processing techniques, their industrial application seems imminent. 2 Bulk metallic glasses consist in most cases of three or more components, and while they often share a typical majority component (such as Zr or Pd), the stability of the amorphous phase at a given composition can still only be ascertained experimentally. The excess free enthalpy of the amorphous phase compared to the crystalline ground state is the driving force leading to crystallization at elevated temperatures. Apart from the difference in free enthalpy, the crystallization rate is governed by atomic motion. In contrast to the case of crystalline media, where diffusion happens as a rule via thermal vacancies, in metallic glasses the situation is not so clear (see, e.g., Refs. 3 and 4 for a review). The predominant view is that here diffusion is mediated by highly collective processes, involving tens of atoms, 5,6 but it has also been proposed that just as in crystals, the random motion of less dense regions, termed quasivacancies, is responsible for mass transport. 7,8 The calorimetric glass transition seems to be connected to a change in the dynamics 9 and, moreover, it is well conceivable that different mechanisms are responsible for the diffusion of the respective constituents. 10 The main experimental method for measuring diffusion is the radiotracer method. This method measures the spreading out of concentration gradients of radioactive isotopes and directly gives the diffusion constant, which is a macroscopic quantity. However, detailed information on the diffusive processes on their fundamental scale, i.e., how the atoms move, can only be inferred indirectly, for example through measure- ments of the isotope effect 8,11 or the pressure dependence. 12,13 First-hand information on the processes on the atomic scale would therefore be highly desirable. Such information can be obtained by atomistic methods such as quasi-elastic neutron scattering (QENS). This method has been applied successfully for the study of melts, 14 but it can not resolve the very slow relaxations in solid metallic glasses. Conversely, relaxations on the time scale of minutes to hours are the natural domain for the recently established method of atomic-scale x-ray photon correlation spectroscopy (aXPCS), 15,16 which can be seen as the x-ray counterpart of QENS in the time domain. It works by scattering coherent x-rays at the sample and correlating over time the fluctuations in the scattered radiation at wave-vector transfers correspond- ing to atomic distances. Diffusivities as low as 10 −23 m 2 s −1 can be resolved, corresponding to less than 10 atomic jumps during the experiment. As such, it is the most powerful method for studying slow diffusive dynamics in solids. Here, we employ this method for the first direct experimental investigation of the atomic-scale dynamics in a metallic glass, in our case the Inoue alloy 17 Zr 65 Ni 10 Cu 17.5 Al 7.5 . As metallic glasses are not strictly in thermodynamic equi- librium, their state is not only a function of the instantaneous temperature, but also of their preparation and subsequent thermal history. At intermediate temperatures, high enough 064202-1 1098-0121/2012/86(6)/064202(7) ©2012 American Physical Society