Physical Vapor Deposition (PVD) equipment is complex to design and operate, as are the processes to run them to achieve a desired result. Therefore, there is motivation to find ways to reduce the expense and improve results in thin film coating operations. Simulation software of many types has been developed to predict the performance of PVD tools and to gain insight into film properties. If the simulation captures the physics of the process, the virtual result approximates the actual behavior of the real system. The virtual result is in a sense a double – a digital twin – of the physical system. While there exist various levels of sophistication in digital twin configurations, the aim is the same in them all: aid decision-making and control in complex thin film deposition processes. The subject of digital twins is much larger than we need to cover here (McKinsey expects a $73.5 billion market by 2027 [1] but we touch on the ways thin film deposition can be enhanced in simple forms and in complex schemes by such tools. - Before a piece of equipment is engineered, a virtual representation can be created to address important questions: what components must be used and how should they be arranged in the vacuum chamber to obtain coatings on the substrates of interest with the throughput and film uniformity required for the application? The virtual realization of the core components (sputtering guns, evaporators, substrate stages, ion guns, etc.) come into being before the physical equipment, retiring much of the risk, particularly in new designs. - For existing equipment, the virtual counterpart can be used to predict the outcome of a process for the components in a particular configuration. As an example, is the required thickness uniformity met at a certain gun tilt angle and target-substrate distance combination in sputtering with a gun of a certain size? This is an example of controlling a macroscopic property. Microscopic properties such as film porosity or deposition into 3D features (e.g. trench-filling) can also be predicted, allowing the researcher to get a virtual head start on optimizing a film for its use. In this case it is an aid in experiment planning. Software such as Virtual Coater ™ [2] (created by an AJA customer – see below) can predict these properties. - Virtual tools can readily provide exposure to real-world equipment for educational purposes. Virtual Reality (VR) models of equipment give students the opportunity to interact with a representation of tools that would be inaccessible otherwise. Our customer at Cornell University developed an entire VR Digital Classroom [3] for this. - A digital twin in the fullest sense is connected to the real-world equipment via sensors that monitor the tool operation. The actual status and progress of the physical process is mirrored in the twin. This enables it to report status and allow decision-making midstream. A coupled Artificial Intelligence (AI) could further run simulations and models, command its physical counterpart to carry out processes, and then monitor progress in real time with little to no human intervention. At whatever level of complexity – virtual machines or digital twins – these are tools for saving time, saving money, supporting decision-making, and furthering integration with AI for autonomous discovery.
