Phase formation prediction in magnetron sputtered Cu(Ti)Zn thin films: Numerical vs experimental approaches

In this work, we evaluated the ability of three numerical methods to predict the phase formation in Cu–Zn binary and Cu–Ti–Zn ternary alloy thin films deposited by DC-magnetron sputter deposition. Molecular dynamics (MD) simulations were carried out to simulate the growth of the alloy film and study the organization at the atomic level. A Machine Learning (ML) approach trained with a recently published bulk HEA (high-entropy alloy) database was used to determine the presence of an amorphous phase, solid solutions, or/and intermetallics. Finally, CALPHAD (CALculation of PHAse Diagrams) thermodynamic modeling allows one to simulate the phase diagrams. Crystalline phases formed in experimental films were investigated by grazing incidence x-ray diffraction (GIXRD). Comparison with CALPHAD results highlights that for pure Ti or binary Cu–Zn films, the thermodynamically stable phases are formed in the films. Less agreement was found at low or high percentage of Ti introduced in the Cu–Zn system, and drastic differences were observed for elemental compositions close to equimolarity. In those cases, the out of equilibrium nature of the magnetron sputtering deposition technique is evidenced. The very limited agreement between the GIXRD and ML approach is explained by the available database, which is exclusively based on bulk alloys. Elemental composition of the alloy does not itself determine the stabilized phases: elaboration techniques are to be taken into account too. MD simulations bring information on a possible segregation of the Zn element to the surface and grain boundaries. A very good agreement is evidenced between the calculated and experimental diffraction patterns.