Tailoring the waveforms to specific applications on Advanced Energy's DRP power supplies
Advanced Energy's Customer Solution Lab is a hub for plasma or vacuum-based equipment and thin-film deposition research and development, and materials characterization, with a focus on training and product demonstrations. The primary focus of the lab is plasma-based thin film coatings for commercial electronics applications, optical coatings, flat panel display development, and more.
T.B.D. AEI - more marketing text?
Advanced Energy has developed the Dynamic Reverse Pulsing (DRP) technique to eliminate various pain points of standard bi-polar sputtering. The asymmetrical waveforms of DRP power supplies help to reduce arcing, decrease substrate heating, and make the processes more stable and reproducible. Furthermore, retrofitting DRP into coaters with dual magnetrons is very straightforward.
The DRP supplies increase the number of parameters that the process engineer can control (asymmetry in duration, asymmetry in voltage, frequency). The correct choice of these parameters is sometimes non-trivial and may depend on the coater layout and other process conditions.
Therefore, the goal set in this project was to develop a simulation model that reproduces and explains the unique properties of DRP-driven plasma. The main use-case for such a model is finding optimum DRP configuration for a given coater layout and part geometry.
Results and Benefits
The numerical model leveraged in this project
Elucidated the plasma dynamics during the DRP pulse and explained why DRP-powered processes are more robust as opposed to bipolar pulsing.
Advanced Energy was able to quantify the magnitude and energy of ion bombardment on a dielectric substrate, which is the factor determining film density and compactness.
Helps to design waveforms which are tailored to a specific application, coater layout and part size.
Demonstrated that PlasmaSolve's hybrid model for sputter plasmas can be applied to waveform tailoring.
Animation of the plasma distribution (above) and ion flux (below) during a DRP pulse in a drum coater with 2 rotary cathodes.
The project was a perfect use-case for PlasmaSolve's hybrid sputter plasma model. The model as such is implemented in COMSOL Multiphysics but to provide faithful and quantitatively correct results, it needs to be interfaced with MatSight Target Erosion App developed by PlasmaSolve.
The model solves the drift-diffusion-reaction equations for various plasma species (e-, Ar+, Ar*, N2+, N+, N ...), the energy equations for electrons and the low pressure gas and the Poisson equation for the time-dependent electric field. So the plasma distribution, ion current and electron current distributions are obtained practically from first principles.
Because the model is so general, it allows us to perform simulation-driven engineering by testing various DoEs in a simulation. The parameters that can be varied include but are not limited to:
The target-substrate distance
Anode position in the coater
Voltage magnitude and duration of the SPUTTER phase
Voltage magnitude and duration of the POSITIVE phase
The performance indicators are application specific, but most commonly include maximizing ion flux to substrate, preventig erosion of adjacent components or ensuring stable coating properties over a range of power levels or part geometries. The simulation produces a list of most promising DoEs that are then verified in the lab.
Example of an ideal DRP waveform - it is asymmetrical both in terms of duration and in terms of voltage.
Simulated target current for the DRP power supply.
Simulated ion current to a dielectric substrate for the DRP power supply.
Refractive index of deposited SiO2
How DRP helps to achieve better process reproducibility
It has been experimentally observed that the DRP power supplies are more robust in terms of material properties - with BP, the refractive index of SiO2 changes considerably when the power level of the cathode is changed while with DRP, it remains perfectly constant. (see upper plot on the left - here we need a hi-res image or data fro Philipp)
To maintain constant film properties at modified process conditions (e.g. power), one should ensure that the following properties are kept the same:
The atomic composition of the coating
The ion-to-neutral ratio at the surface
The incident energy of the bombarding ions
All these quantities can be obtained from PlasmaSolve's hybrid model of sputter plasma.
In the bottom chart, we show the ion-to-neutral ratio at the surface and the incident ion energy obtained from the simulation. It is apparent that for DRP, these quantities are much more stable, compared to BP where they are a stron function of applied power. This explains the improved process stability and robustness.
Gas Flow Simulations
The pressure in the PECVD system varies accross several orders of magnitude - from 100 Pa to a little under 1 Pa. For gas flow simulation, this presents certain challenges because there are two methods one could use - neither of them being a perfect fit.
The DSMC (direct-simulation Monte-Carlo) method would be warranted due to the low pressure limit where the flow stops to follow the rules of the continuum. However, it is computationally unfeasible at the higher pressures, above 10 Pa.
A compressible fluid simulation with slip boundary conditions is about 200x more computationally efficient compared to DSMC, but its accuracy in the low pressure region is unknown.
For that reason, PlasmaSolve simulated an AGC coater using both the types of models and validated them against pressure readings on two distinct locations in the low pressure region.
The study revealed that the fluid simulation is very well applicable to a glass-coating PECVD reactor and both the simulation techniques showed matched the measurement very well.
The validated code was then used for several engineering studies, typically pertaining to gas flow uniformity.
Benchmarking flow simulation with measurements
Chemical composition of plasma as a function of time - only the 10 most abundant reactive neutrals and 10 most abundant positive ions are shown.
Understanding the Chemistry
The plasma-phase chemistry of the precursor - TMDSO (tetramethyldisiloxane) was previously unknown. PlasmaSolve worked with a tremendouns number of literature resources to formulate a complete plasma-kinetic system for this complex compound. The resulting system contains nearly 70 species and over 1200 plasma-chemical reactions.
The scheme was migrated into PlasmaSolve's Global Plasma Model, which is a computationally efficient tool describing the plasma in a volume-averaged approximation but with all the possible chemical processes that could take place.
Once this respectable task is completed, the Global Plasma Model can be used for
Discovering favorable reaction pathways in the system.
Adjusting coater settings to favor a concrete reaction pathway.
Providing necessary input into film-growth simulations