Molecular Diffusion Lab

Hi everyone! My name is Sophia Lauf (J) and I work with fellow students Naina Kaimal (J), Sam Marquez (J), G Missaka (S), and Hunter Whaples (S) at the Molecular Diffusion Laboratory under the guidance of Prof. Lape. We study nanoparticles suspended in polymer membranes in order to better understand and design novel materials for separating gas mixtures. We hope this research can advance all kids of separation technologies including the emerging field of Carbon Capture!

Two major properties of polymer membranes are permeability, which is the product of diffusivity and solubility representing the ease of gas penetration, and selectivity, which is the ratio of permeability between two penetrants representing separation performance. The ideal separation membrane would be both highly permeable and highly selective for a given gas, meaning it allows that gas to pass very easily without allowing contaminants to pass. Unfortunately, in the poymeric membranes that the Molecular Diffusion Lab studies, there is a tradeoff between selectivity and permeability as shown by the Robeson plot below.

The Robeson Plot. An ideal separation membrane would be in the top right, highly permeable and highly selective. http://dns2.asia.edu.tw/~ysho/YSHO-English/1000%20CE/PDF/J%20Mem%20Sci320,%20390.pdf

Interestingly, a 2002 study found that the introduction of silica nanoparticles into glassy polymer membranes to form polymer nanocomposites increased both permeability and selectivity. Our lab seeks to understand and further quantify this occurrence that we have called “The Merkel Effect.”

Back when we had access to the lab, we were able to create membranes with silica nanoparticle fillers following something called the Stöber process. The Stöber process synthesizes differently sized silica nanoparticles.

Creation of these films proved to be quite the task. Despite the ease in making the solution for the films, drying them and letting them solidify into a functional membrane was difficult. Some membranes dried too quickly and were crispy instead of malleable, while others formed large bubbles as they dried, which was also nonideal.

Before the Pandemic we spent lots of time in a chemical engineering lab synthesizing carefully sized silica nanoparticles using a technique called the Stöber process. We incorporated these nanoparticles into casts of various polymer membranes in order to reproduce the Merkel Effect and study the resulting permeability and selectivity performance.

Although synthesizing nanoparticles sounds like it should be the most intimidating part, casting membranes proved to be quite the task. While many chemical procedures require good timing, measuring, or lab technique,, we found that membrane casting was quite the artform. Some membranes dried too quickly and were crispy instead of malleable, while others formed large bubbles as they dried. Casting the perfect smooth membrane takes remarkable patience!

Rearcher Naina Kaimal with a failed cast. 

Without access to a lab space this summer, the research group pivoted to computational work that we could develop remotely. Of course, the burning question is how do you study the material performance of membranes without making the well … the membranes? The method our lab uses is called Molecular Dynamics (MD) Simulation.

Using the open source software GROMACS along with a lot of MATLAB and Python we were able to construct a simplified model of the silica nanoparticle and polymer that we normally would have synthesized in the lab. Of course, simulating the motion of millions of atoms is beyond what any normal computer is capable of. Therefore, we defined groups of repeating atoms into equivalent “beads” making the models coarser but also allowing them to be much larger, and most importantly, runnable.

All atom models of silica nanoparticle (left) and coarse grain silica nanoparticle (right).

The GROMACS Procedure puts a particle and polymer in a box and then runs energy minimization, which results in an image like the following:

To run a simulation we place nanoparticle molecules and polymer chains randomly into a  box and then compress the box to 100 bar (similar to atmospheric pressure on the surface of Venus!). We then decompress the box which allows molecules to settle into a low energy state representative of the positions they might take in real life. 

Animation of model compression and energy minimization step.

What certainly doesn’t look like real life, however, is how the model seems to cut off on one side and resume on the other. This “pac-man” quality is called the periodic boundary condition and helps the software emulate as if it extended infinitely in all directions. This is great computationally, but visuals are far easier to understand without it.

Animation of simulation block without periodic boundary condition.

We then insert a gas such as butane into the model and track the motion of each gas particle over the duration of several nanoseconds. Based on their displacement we can characterize their motion and even determine an approximation for the diffusivity constant. We can see from the animation that gas particles get trapped in cages for long periods and then suddenly jump long distances before getting trapped in another cage.

Distance of an gas penetrant from origin (left) and animation of its motion (right)

We can also examine the pore size distribution and fractional free volume for different theoretical membranes to better understand how including the nanoparticle changes these calculable quantities. Interestingly, preliminary results observe the existence of larger pores when adding a nanoparticle that do not exist when simulations do not have a nanoparticle.

Observation of large pore size spike with the addition of a nanoparticle.

Our plan for the rest of the summer and continued remote research is to further these models in a variety of ways. First, we plan to develop computational tools to better analyze solubility, polymer configuration, and eventually estimate permeability in much the same way we could in the lab. Second, we plan to design new nanoparticles and new polymer membranes to more rapidly inform the types of materials we should make and test once we can get back in the lab.

While certainly I was not planning on researching remotely this summer, the great developments we’ve been able to make with computational modeling is super exciting! With a great team, good models, and a wonderful advisor, the only thing left to hope for is to get back into the lab safely and start using our results to make real membranes. Until then, GROMACS reminds you: "If you thought that science was certain - well, that is just an error on your part." (Richard Feynman)