The National Institute for Computational Sciences

University of Tennessee supercomputer helps researchers understand fundamental properties of PEM fuel cell

Researchers use Kraken to simulate proton transport in PEM fuel cells

by Caitlin Rockett

Using the Cray XT5 supercomputer known as Kraken, a team lead by University of Tennessee-Knoxville’s (UTK) Stephen Paddison is studying fundamental properties of the electrolyte of proton exchange membrane (PEM) fuel cells —currently being tested as clean energy conversion devices for automobiles—from the bottom up.

Figure: Simulations show that fluorinating the inner carbon nanotube walls stabilizes the protonic charge carrier as an Eigen ion (top), while bare walls favor Zundel ion formation (bottom). Carbon, fluorine, hydrogen, oxygen, and sulfur atoms are depicted in grey, green, white, red, and yellow respectively.

“Our project focuses on materials, specifically the polymer electrolyte membrane that acts as a central component of this type of fuel cell,” explained Paddison, a professor of chemical engineering at UTK. Funded by the U.S. Army Research Laboratory and the U.S. Army Research Office, Paddison’s team is looking at the way in which protons are transported through the central component of a PEM, studying the functionality of the material at the molecular level.

Housed at the National Institute for Computational Sciences at Oak Ridge National Laboratory and managed by UTK for the National Science Foundation, Kraken is the world’s fastest supercomputer managed by academia. At 1.03 petaflops—over 1 thousand trillion calculations per second—Kraken is ranked third fastest on the Top500 list of the world’s most powerful computing systems.

Inside a PEM fuel cell

PEM fuel cells operate using hydrogen (H2) fuel and oxygen (O2) from the air to produce electricity. Hydrogen fuel is oxidized (loses electrons) on one side of the membrane while oxygen is reduced (gains electrons) on the opposite side.

Hydrogen gas is pumped through field flow plates to an electrode (the anode) on one side of the fuel cell. Oxygen from the air is simultaneously pumped through separate field flow plates to another electrode, the cathode, on the other side of the fuel cell. At the anode a catalyst—usually platinum—splits hydrogen into positively charged hydrogen ions (protons) and negatively charged electrons. The proton exchange membrane allows only the protons to pass through to the cathode, creating an electrical current as the electrons are forced to travel an external circuit to the cathode. In addition to separating the electrodes, the membrane is also impervious to gases, preventing the stream of hydrogen from mixing with the oxygen until they reach the cathode. Once the protons and electrons reach the cathode via their separate paths, they recombine with oxygen to form water which flows out of the fuel cell.

PEMs operate at temperatures between 80 and 110 degrees Celsius, approximately. This relatively low temperature regime (other types of fuel cells operate at temperatures between 600 and 1,000 degrees Celsius) allows these devices to produce electricity quickly. However, a single PEM fuel cell only produces about 1 volt of electricity, so multiple cells must be ‘stacked’ to produce enough electricity to power a car. Accordingly, PEMs must be chemically, thermally, and mechanically stable to withstand oxidation processes, heat, and mechanical strain.

Perfluorosulfonic acid (PFSA) ionomers (an ion containing polymer, a large molecule composed of repeating structural units—a backbone—typically connected by covalent chemical bonds) have been used as the electrolyte in fuel cells since the mid-1960s, but, according to Paddison, the manner in which they conduct protons is incompletely understood.

“We decided to separate some of what is really a very complex problem into some very simple things to try to solve with computations from first principles,” explained Paddison. And that’s where Kraken comes into play.

Modeling the PEM

Water management plays a decisive role in the performance of a PEM fuel cell. Water must be introduced to the hydrogen stream at the anode; however, too much water will flood the membrane, but too little water allows the membrane to dry out and crack. Either scenario reduces proton energy production significantly. For the ease of managing the device, developers would like do away with the addition of water completely, but it’s not so simple.

“Water is critical to formation of charge carriers,” explained Paddison. A charge carrier is an unbound particle carrying an electric charge, such as hydrated protons. “With proton conductivity, we’re interested in how the protons move under a gradient that exists between the anode and cathode,” continued Paddison. “They don’t move as bare protons, they move via a complex mechanism within the absorbed water within the PEM.” In the presence of water, PFSA membranes exhibit fine phase separation, meaning they develop a network of very small (in the range of a few nanometers, or billionths of a meter) hydrophilic (“water-loving”) and hydrophobic (“water-hating”) channels. These hydrophilic regions contain the hydrated protons and water, while the hydrophobic regions consist of the aggregated backbone of the polymer. It is the hydrophilic channels that mediate proton transport, but the hydrophobic channels are essential, endowing mechanical, and to some extent, chemical and thermal, stability to the membrane. These tiny regions create a large number of structural variations, making understanding the factors governing proton transport difficult. Despite much research, there is no real consensus among the scientists active in PEM fuel cell research concerning how water accumulates in the material to mediate proton transport.

“So with the desire to understand proton transport in a PEM at very low water content and not be hindered by the ambiguity in the hydrated morphology of the membrane, we decided to begin with a regular and predefined morphology,” said Paddison.

After significant computational studies on actual PFSA membranes—trying to model and understand the connection between transport, morphology, structure, and hydration—Paddison’s team created a simplified hypothetical structure. Their goal was to focus on elucidating how proton transport is affected by specific structural and physical aspects of the membrane structure and morphology. The hydrophobic polymer backbone was replaced with a well-defined carbon nanotube which could be decorated with the acidic groups of a PFSA membrane and then filled with prescribed amounts of water.

The effects of fluorine were of particular interest to the team as the backbone of PFSA polymers are comprised of polytetrafluoroethylene—more commonly known as Teflon. By either decorating the carbon nanotube with fluorine atoms or leaving the walls of the tube bare, the team was able to study the effects of fluorine on proton transfer between sulfonic acid groups with differing separation and degrees of hydration – all properties that may be varied in real PFSA membranes.

To do this, Paddison’s team used Kraken to run simulations using the Vienna Ab Initio Simulations Package (VASP) code. The term “ab initio” is a Latin term that translates literally into “from the beginning.” To model atomic movement, computational scientists typically use applications referred to as molecular dynamics (MD). MD simulations require the definition of a potential function, a description of the terms by which the particles in the simulation will interact. VASP is a type of MD application, but as an ab initio application, VASP does not require any prior assumptions as to the interactions between water molecules, sulfonic acid groups, and protons, as the electron density is instantaneously relaxed to its ground state at each time step of the MD. “The real benefit of VASP is that it allows for the breakage and formation of both covalent and hydrogen bonds,” said Paddison. Paddison’s team was able to simulate between 180 and 250 atoms in a single system with 256 processors on Kraken and achieve linear scaling.

“The first part of the transfer of a proton is dissociation from an acidic group to either another acidic group or water,” he explained. “What we’ve found is that the fluorine may not have a big impact on the actual dissociation of the proton, but it does have a very definite effect on the state of the proton at hydration levels of three water molecules per sulfonic acid group.”

This assessment goes back to the concept of the charge carrier—protons in the case of this project. Paddison’s team determined that fluorine does influence the nature of the charge carrier, implying that the fluorine backbone may play an important role in the efficient transfer of protons within a PEM fuel cell operating at low relative humidity.

Paddison and his colleagues know that this is only a first step toward improving hydrogen fuel cell technology, but it’s a fundamental first step.

“Simulations and computational research are featured more and more in the development and understanding of existing and novel materials,” said Paddison. “The hydrogen fuel cell is deemed to be a viable, clean alternative as an energy converter for a number of applications, and since we lack real understanding of several important features, we’re working to answer questions using computations that will provoke novel experiments and materials.”