Session DR2: Plasma Surface Interaction II

Atomistic scale modeling of plasma-material interactions for enhanced amorphous carbon layer performance in semiconductor manufacturing

With the advent of complex and sophisticated architectures in semiconductor device manufacturing, the performance of amorphous carbon layer (ACL) for a mask needs to be enhanced in terms of durability, residual stress, optical features, etc. However, due to the sensitive nature of ACL property to the process condition of plasma enhanced chemical vapor deposition (PECVD), it is challenging to tailor and further optimize the property of ACL thin film with experimental approaches. In this work, we conducted reactive molecular dynamics (MD) simulations to understand the fundamental property of ACL considering varying bond hybridization and hydrogen content. The elastic modulus, hardness and thermal expansion coefficient of ACL were examined and compared with experimental values. Moreover, PECVD process was modeled at the atomistic scale; we found a significant relationship between the incident energy of plasma species and the resultant layer characteristics such as structural conformation and residual stress evolution. The proposed computational framework provides the structure-property relationship of ACL and the resultant layer feature. Based on the information obtained from the atomistic scale simulations, it can be used as a concrete guidance for optimizing the ACL deposition process. Furthermore, the present method can be applied to investigate the fundamental nature of various plasma-material interactions for developing advanced equipment and process in semiconductor manufacturing.   

Hybrid Molecular Dynamics-Machine Learning Approach for Efficient Modeling of Particle Growth in Non-Thermal Plasma

The growth of particles in non-thermal plasma is a fascinating yet challenging problem to model accurately due to its non-equilibrium nature. Although it is feasible to model specific reactions occurring during particles' precursor growth, extending this approach becomes rapidly unfeasible. In this study, we propose a hybrid molecular dynamics-machine learning approach that significantly reduces computational requirements. To demonstrate the effectiveness of our approach, we examined the collisions between silane molecules using classical molecular dynamics simulations. By decoupling the internal energy from the collision speed, we conducted simulations that allow for rapid computation of results and uncertainties for any translational energy distribution. Using these simulations, we determined the probability of different reactions at various temperatures, information that was then used to train machine learning models that investigate the best inference of missing data from sampled conditions. The results indicate that machine learning can predict missing interactions, but caution must be exercised in the selection of molecular dynamics-generated data to achieve optimal accuracy and computational time reduction.   

Scale-Bridging Simulations for Plasma-Surface Interactions

Plasma-surface interactions play an integral role for a variety of processes (e.g., sputter deposition, plasma-enhanced catalysis) but combine length and time-scales that differ in orders of magnitude. They are therefore commonly accounted for by oversimplifying (model) assumptions (e.g., lookup tables with rate coefficients), often biased by experience and feasibility. Unbiased data-driven approaches are pursued in this work for the sputtering and the deposition of metals and metal nitrides in Ar and Ar/N₂ plasmas. Obstacles and key perspectives are highlighted for the data generation (i.e., hybrid reactive molecular dynamics / time-stamped force-bias Monte Carlo simulations) as well as for the selection of suitable artificial neural network architectures. The proposed architecture allows to simultaneously include both experimental and simulations data in the training set, despite the different physical properties that are accessible. The role of entrapped working gas atoms (i.e., Ar) for sputtering and thin film deposition is found to be negligible and significant, respectively. Ultimately, the temporal development of sputtering phenomena (e.g., emission of N₂ due to collisions with N split interstitials) and various thin film properties (e.g., composition, density) are predicted with atomic but still high physical fidelity for experimental process times of 45 minutes. Rare events (e.g., impingements of ions whose kinetic energies stem from the tail of the ion energy distribution function) can be of paramount importance to the steady-state and increase the time to reach equilibrium from a few seconds up to 30 minutes. The machine learning predictions took merely 34 GPU hours. In contrast, hybrid reactive molecular dynamics / time-stamped force-bias Monte Carlo simulations would last for more than approximately 8 million CPU years.   

Characterization of Two-Surface Multipactor with Non-Sinusoidal RF Fields

Multipactor is a generally undesirable nonlinear discharge occurring in vacuum RF components and systems in which an electron avalanche is sustained through secondary electron emission from metallic or dielectric surfaces in the presence of a high frequency RF field. We present the characterization of two-surface multipactor discharge with a non-sinusoidal Gaussian-type RF field waveform [1,2]. We employ Monte Carlo [3-5]  and 3D electromagnetic Particle-in-Cell (PIC) [3] simulations to examine the effects of the amplitude, power, and full width half maximum (FWHM) of the Gaussian electric field on multipactor susceptibility and time dependent physics [3,5].

 

We observe that the threshold peak RF voltage and the threshold time averaged RF power per unit area for multipactor development increase with a Gaussian-type electric field compared to those with a sinusoidal electric field. The threshold peak RF voltage and RF power for multipactor increase as the FWHM of the Gaussian profile decreases. Decreasing the FWHM of the Gaussian electric field can reduce the electron population corresponding to the strength of the multipactor at saturation, at fixed time-averaged input power. 

[1] D.-Q. Wen et al., Physics of Plasmas 26, 093503 (2019).

[2] D.-Q. Wen et al., Appl. Phys. Lett. 121, 164103 (2022).

[3] A. Iqbal et al., Phys. Plasmas 29, 012102 (2022).

[4] A. Iqbal et al., Phys. Plasmas 26, 024503 (2019).

[5] A. Iqbal et al., Physical Review E 102, 043201 (2020).   

Atomic oxygen surface recombination in glow discharge plasmas

Surfaces interact with either active discharges or their afterglow in most plasma processes, via heterogeneous surface kinetics. These processes can affect both plasma and surface properties. In particular, in oxygen-containing discharges, the adsorption and recombination of atomic oxygen on reactor surfaces determine the gas composition, the availability of O for important volume reactions (e.g.: CO₂ + O → CO + O₂; CO + O + M → CO₂ + M) and eventually the flux of reactive oxygen species (ROS) towards target surfaces. The wall loss frequencies of O atoms have been measured in the positive column of O₂ and CO₂ glow discharges in a Pyrex tube (borosilicate glass), for several pressures, currents and wall temperatures. However, the surface mechanisms determining recombination are not fully known yet. In this work the LoKI global model is employed to self-consistently simulate the volume kinetics in the plasma and the surface kinetics of O atoms in the conditions of the experiments. The simulation results are compared with experimental measurements, describing the experimental dependence of the atomic oxygen recombination probability on pressure, current, gas temperature and wall temperature. Moreover, the newly developed model allows to identify the most important recombination mechanisms for each condition.   

Overcoming Surface Inactivation through Catalytic Design in Low-Temperature Plasma-Catalytic Conversion of Methane to Hydrogen

In this talk, we demonstrate how catalytic surfaces can be designed in plasma environments to convert methane (CH₄) to hydrogen (H₂) directly while limiting surface inactivation. Conventional methane conversion on catalysts forms layers of carbon (i.e., coking) that reduce catalytic activity and the rate of hydrogen production. In plasma environments, we show that coking does not result in complete catalytic inactivation, but instead, self-assembles into carbon filaments that grow off the surface. We employ a combination of in- (i.e., microscopy) and ex-situ diagnostic techniques (i.e., profilometry) to characterize the growth of carbon in real time on catalysts that are enveloped in plasmas. We utilize density functional theory (DFT) and microkinetic models to show how the design of catalysts can synergize with vibrational excitations of methane to break scaling relations and enable the use of more coke-resistant catalysts. Experiments are repeated to validate these predictions and show how the design of catalysts can further change how carbon is deposited and localized on surfaces. In each case, we quantify H₂ production near catalytic surfaces through argon (Ar) actinometry with optical emission spectrometry (OES) and gas chromatography to evaluate the process efficiency. This research offers insights into the relationship between plasmas and catalysts in environments where surface morphologies change dynamically.