Climbing to Everest Base Camp was a reflective journey, filled with amazing Sherpas, tasty Nepali chai, and the bitter cold of the great Everest summer. As I completed this experience, many things caught my eye that I wasn't accustomed to seeing at home.
Do you ever travel to a forest or mountain isolated from the entirety of civilization, and see houses there? Where do they get their milk? Where do their kids go to study? How do they get internet? These are thoughts we all had but never realized the truth of. Their lives are hard.
We departed from Lukla, the beginning of our journey, and arrived at an isolated town, Phakding, our first night stop. That's where the scene truly caught my attention. Around twenty mules collectively carried over 100 gas cylinders, trekking up a steep hill to reach the remote village. A local that my father and I approached explained the oddity of what we had just witnessed. Energy was a huge problem up in Everest, as the volatile weather made it hard for traditional renewables, such as solar or hydro, to function. Basic privileges such as living in a warm house or having gas to cook food were a challenge.
I further dug into this and discovered the science behind why solar was a problem. Everest sunlight is direct, as the snow's albedo effect and minimal pollution allow for strong sunlight to pass through, making it perfect for solar. Right? The issue is the temperature, as the volatility of hot to cold makes it hard for inverters to transform DC to AC. Traditional silicon becomes inefficient, and the problem arises.
My internship at the Stanford WBG Lab showed me the solution. I've been focused on wide-bandgap semiconductors, a material with a higher band gap, allowing it to operate at much higher voltages, frequencies, and temperatures than traditional silicon. I'm specifically focusing on diamond, as diamond substrates can handle extreme power, making them ideal for efficient solar inverters and grid applications. But there's a problem: we don't fully understand what happens at the atomic scale during chemical-mechanical polishing (CMP), the critical step in fabricating the surface for diamond WBG semiconductors, which is where my work comes in.
What is CMP?
CMP is a polishing process to planarize surfaces in semiconductor fabrication. It uses slurry chemicals (such as H2O2) and mechanical polishing techniques, which are critical for the construction of wide band-gap materials such as SiC and diamond.
However, the current challenge is that CMP performance depends on surface chemistry at the atomic scale, which, as earlier stated, is not fully understood. The objective of my research is to understand bond dynamics during CMP, as it can assist in material defect identification and enable optimization of CMP slurries and processes for WBG devices.
The Research Process
1. Running ReaxFF Simulations
Now, to begin with the research. We began by running ReaxFF reactive force field simulations. But what is ReaxFF? Reactive force field molecular dynamics simulations allow the simulation of large-scale bond-breaking and formation without pre-defined bond connectivity. Compared to other MD simulations, such as Lennard-Jones Potentials, ReaxFF runs faster and is able to model chemical reactions at a macroscale in nanoseconds. That's exactly what we need!
Through simulating diamond substrate (the material that was to be polished), diamond abrasive (the material contacting the substrate during polishing), and the H2O2 slurry (the chemical liquid which planarizes the substrate), LAMMPS trajectory and bond data were generated (LAMMPS is the simulation software that was used).
2. Writing MATLAB Filter Functions
The next step was writing filter functions in MATLAB that identify specific functional groups based on atomic bonds within the simulation. For example, Carbonyl groups (C=O) are detected when carbon is double-bonded to oxygen under specific conditions. A similar logic followed for alcohol (R-OH) and carboxyl (COOH) groups. This step essentially quantifies what's happening chemically during polishing, which was the original motive of the research.
3. Visualizing with OVITO
The final step is to visualize the filter functions, which was done using OVITO (an atomic visualization software that takes LAMMPS data as input to produce visual atomic simulations). We color-coded atoms based on functional groups (red for oxygen, gray for abrasive carbon, etc.) and transferred the MATLAB logic to the OVITO pipeline, allowing us to filter specific functional groups as the simulation ran in real time. For detailed workflow diagrams, logic conditions, and OVITO visualizations, check out the full presentation linked above.
Impact and Applications
The OVITO framework with the MATLAB filter functions is adaptable for any simulation environment, as researchers can test different slurry compositions, pressures, and temperatures without expensive lab trials every time. Better surface chemistry understanding means smoother wafer surfaces and more predictable fabrication, which wide-bandgap semiconductors need to actually compete with silicon in renewable energy applications.
This was my first time working with molecular dynamics simulations at a large scale, and the learning curve was steep. Processing LAMMPS datasets and writing MATLAB filters forced me to really understand what was happening chemically, not just accept it as a black box. The coolest part was watching the OVITO visualizations come together (generating virtual atoms made me feel like a genius) and seeing surface reactions unfold in real-time.
Looking Forward
As I enter college, my long-term goal is to make WBG semiconductors economically viable for renewable energy infrastructure, which tackles the energy access challenges I witnessed at Everest Base Camp. Understanding CMP at the atomic scale gets us closer to that reality, as perhaps in the future, I can finally see mules resting and resilient solar panels powering life up in the heights of Everest.