PhD

I pursued my PhD in the group of Prof. Lutz Mädler. at Leibniz Institute for Materials Engineering-IWT, University of Bremen, Germany. I developed computational models in DSMC to simulate surface reaction.

The baseline of the surface reaction model is the paper by Pesch et al. (2015). Here, they descibe a DSMC algorithm to model molecular diffusion, molecular collision, adsorption, desorption and reaction of CO and O₂ in a mesoporous Pd layer. My goal is to make the algorithm more computationally efficient and more modular.

Surface reactions are by default difficult to code. A number of surface parameters such as adsorption coefficient, desorption and reaction barrier interact in complicated ways. So, the initial code was written to see the feasibility of writing a microkinetic model that can decouple each event. The easiest way to do this is to represent each molecule in the gas by a single simulation molecule. But, DSMC is computationally exhaustive and the process becomes tedious. That's where I come in!

I first started with the existing equivalent parcel method existing in DSMC. It groups molecules into parcels and executes these parcels as one. Essentially, the molecular diffusion process becomes less expensive and that makes the code faster. But, soon I discovered that having a fixed parcel size would never work. The reactants and products are present in different concentrations and various concentration gradients exist in the nanoreactor. This calls for a method with different parcel sizes possible.

I called this the variable parcel method. The idea was inspired by this paper by Nanbu et al. (2000). There is a modified molecular collision process defined in the paper which helps to conserve mass in the system. After adopting this, I have made the model twice as efficient. Not to forget, that the DSMC simulation still takes days to compute a mere few microseconds, and not weeks. Regardless, the simulation provides valuable insights in to mass transfer and chemical kinetics in such small reactors which might be often overlooked.

Further, we collaborated with the Chemical Physics department at Chalmers to validate my model. They have created this cool microfluidic chip with nanochannels. Read more here. Each nanochannel then has nanoparticles embedded in it. The inlet and outlet gases are monitored. The temperature of the chip is maintained using a small heater. With all these experimental values for the outlet gas concentration available at different temperature, we improved our model to agree well with the experiments. To our surprise, it did not involve any significant changes in the physical model. All the improvements could be easily explained. One of the improvements entailed lowering the activation energy. In literature, small activation energies are often due to a polycrystalline nature of the crystal. Although we had no prior knowledge of this, our simulation helped us reach the same conclusion as the experiments. Our model showed excellent agreement with increasing temperature.

The model contributed to understanding mass transfer inside the nanofluidic channels. We were able to allocate the particle with the maximum reactivity. We were able to gather information about local surface coverages of the molecules. This is something that our collaborators were curious about. They normally use plasmonic sensing to determine gas concentration. Since, the particle size is so small and Pd is not very effective under plasmonic sensing, the model was able to give some answers. Later, when they come up with some novel technique to achieve the same with experimental techniques, we will be able to know if we were correct or not! For now, our results are based on callibrating with the macroscopic data available from the spectrophotometer at the outlet. We were also able to obtain the local gas concentration. Currently, the manuscript for this work is publiished.

Like I said, the model is the simplest implementation of a very complicated process. We are continuosly working to add features to the model that can be easily turned on/off. The most recent being the thermodynamics of the surface reaction. It is really cool to see how a simulation method that was developed to model air flow around giant spaceships is now effectively used to model gas flow in nanosized systems.

Rightly said, The only way to discover the limits of the possible is to go beyond them into the impossible. –Arthur C. Clarke