The document discusses using boron-doped carbon nanotubes as a catalyst for oxygen reduction in proton exchange membrane fuel cells. Density functional theory and Nudged Elastic Band calculations were used to study oxygen dissociation across a boron-doped (5,5) single-walled carbon nanotube. The most favorable reaction path was determined to have an activation barrier of 0.5 eV, lower than previous studies using a single boron atom. This is likely due to cooperative effects of having three boron atoms and electron transfer from boron to oxygen. While the 0.5 eV barrier is still relatively high, solvent effects in real systems could further reduce it. Boron-doped carbon nanotubes

1. OXYGEN DISSOCIATION ACROSS A BORON-DOPED CARBON
NANOTUBE: FUEL CELL CATALYSIS
Introduction
Fuel Cell Catalysis
In the recent years, increased interest in proton exchange membrane fuel cell (PEMFC) technology has resulted from both economic and
environmental factors. Unfortunately, current implementation of PEMFCs exhibit low efficiencies and high cost. Consequently, research into
the advancement of under-efficient , high cost, system components is necessary to the future of the PEMFCs. This research focuses on the
catalysis of the reduction/oxidation (redox) reactions that facilitate charge transfer. More specifically, this research investigates the structure
and reaction path of a feasible catalyst for the oxygen reduction reaction (ORR). A general schematic of a PEMFC and its half reactions are
shown in Figure 2. Note the catalyst electrodes that facilitate the redox half reactions. A typical ORR catalyst (designed for the NASA Apollo
Lunar Missions in the 1960’s) is platinum loaded carbon. Unfortunately, platinum catalyst efficiencies are rather low, ranging from 20% to
30%.
The low efficacy is a product of many factors; one problem being platinum’s strong affinity for carbon monoxide (CO). This is known as CO
contamination. CO contamination is a result of substantial d-π back bonding that takes place between Pt’s d-orbital and CO’s π orbital. The
permanent occupation of a Pt coordination site by CO greatly reduces the likelihood of reactant species experiencing redox catalysis by
platinum. Consequently, the PEMFC experiences time dependent drift, which is, a steady degradation in efficiency throughout its lifespan.
Also, not all of the platinum is exposed to the reactants because it is clumped in a disorganized manner on the carbon support.4
These
factors, combined with the high cost of rare metals, render platinum loaded carbon impractical for wide scale usage. Fuel cell redox catalysis
is the focus of this research, more specifically, the oxygen reduction reaction catalysis.
Carbon Nanotubes
Recently we have seen that carbon nanotubes (CNTs) show great potential for being a feasible catalyst support system. They appear to be
more mechanically and chemically stable than a conventional carbon support system, possess high electric conductivity, lack cracks, and have
larger surface areas. Also, the activity of the CNT allows for doping. Doping allows for other reactive elements to be covalently bonded to
the CNT surface which greatly alters the chemistry of the system. These factors make CNTs an appealing alternative for carbon black support
in PEMFCs.
The carbon nanotube is essentially an infinite hexagonal carbon network that is rolled cylindrically. They are synthesized by various
techniques; direct-current arc discharge, laser ablation, thermal and plasma enhanced chemical vapor deposition (CVD) and recently
developed self-assembly of single crystals of SWCNTs, are among the most popular methods. A carbon nanotube can be synthesized
differently to create varied chirality. The manner in which graphene sheets are formed determines their types of activity. Figure 1 shows the
main classes of nanotubes. The zigzag CNTs are semiconducting and the armchair CNTs are metallic. The helical CNTs have demonstrated
ferromagnetic properties and display high magnetization at room temperature.
Platinum-doped CNTs have been utilized successfully in lab settings to produce higher efficiencies, but the fact that they rely on expensive
metals renders them a less feasible catalyst.6
Also, the platinum-doped CNTs are subject to CO contamination which can become a serious
setback in the lifespan of a PEMFC. With this, exploration into the usage of nonmetal catalyst is essential. Nitrogen and boron are
electronically similar to carbon and can replace carbon in the CNT, leaving the CNT structure intact but altering the chemistry significantly.
In the nitrogen doped system, nitrogen changes the chemical bonding environment and is capable of increasing the binding energy of
diatomic oxygen. Oxygen can bind to either nitrogen or carbon but quantum mechanical calculations suggest that the carbon atoms adjacent
to the nitrogen respond to the nitrogen’s electronegativity with a relatively high positive charge density. With this, it is understood that
redox cycling reduces carbon which invites adsorption of diatomic oxygen to return to a formal oxidation state.
This research investigates the possibility of using boron-doped (5, 5) single-walled carbon nanotubes (B3SWCNTs) as a catalyst. More
specifically, we studied the effect of boron-doping on the oxygen reduction reaction (ORR) computationally via density functional theory
(DFT). Structure design and doping sites play a critical role in the overall efficacy; subsequently the aim of this research is to manipulate the
boron-doped SWCNT to improve catalytic effect on the ORR and eliminating the use of precious metal catalyst.
Matt Powell and Hee-Seung Lee
Department of Chemistry and Biochemistry
University of North Carolina at Wilmington
Computational Method
Density Functional Theory
The computation consisted of electronic structure and Nudged Elastic Band (NEB) calculations. The electronic structure calculations were
performed to understand electronic properties of geometrically optimized systems as well as generating reactant and product for subsequent
NEB calculations. NEB calculations were utilized to achieve mechanistic information on the minimum energy path (MEP) of the overall
reaction. All DFT calculations were performed with the QuantumEspresso package. We chose Perdew-Wang91 exchange-correlation (XC)
functional and Vanderbilt ultra-soft pseudopotential. This combination of XC functional and pseudopotential has been used in many studies
of metal doped CNT systems. The energy cut-off and cell size were optimized to ensure the convergence. We used periodic boundary
condition with an orthorhombic super-cell accommodating 6 unit cells of (5, 5) SWCNT (total 120 carbon atoms), where three carbon atoms
are replaced by boron atoms. The convergence threshold for geometry optimizations (including NEB) was 4.0x10-4
hartree/bohr.
Electronic Structure
Nudged Elastic Band Calculation
NEB calculations are necessary to finding the minimum energy path (MEP) of O2 dissociation across the B3SWCNT (the ORR). In essence, NEB
calculation estimates a series of images that would take place between two geometrically optimized reactants and products, then does
electronic structure calculations on each of the estimated intermediate images. After the intermediate images are optimized, they are said to lie
on the MEP. Of the many electronic structure calculations, the figure below shows the product/reactant pairs (image 1 and 12 in MEP plots)
selected for further NEB calculations. Reactant species was determined as a compromise between computational cost and the desire to
maximize reaction mechanism information. Products were chosen based on stability.
Figure 2. This is a schematic of a typical PEMFC. Note that a precursor to the reaction is the dissociation of H2 at the
anode and and O2 at the cathode.
a) b) c) d)
Figure 1. The above figure displays different CNT conformations. From left: SWCNTs a) armchair, b)
zigzag, c) helical, and d) MWCNT. Geometry Optimization
Figure 3 shows the geometrically optimized (5, 5) B3SWCNT. This structure will act as a starting reactant. The green junctions are
carbon atoms and the blue junctions are boron atoms. The active site is the collection of the 3 boron atoms known as the boron
cluster. Note that the system is infinitely long due to the periodic boundary condition, despite the image seeming finite. Geometry
optimizations were also achieved for possible physisorbed, chemisorbed, and dissociative adsorption states of diatomic oxygen (O2).
The calculations showed that oxygen preferentially binds to boron atom.
Band Structure
Figure 4 and 5 compare the band structure of a pristine (5,5) SWCNT, a singly doped (5, 5) BSWCNT, and the boron cluster doped
B3SWCNT . Energy zero is set to be the Fermi level of the system. The Fermi level can be thought of as the top of the occupied electron
energy levels and contains the most loosely held electrons that will more likely be involved in charge transfer. K-points (x-axis) can be
thought of as quantum numbers associated with infinitely large periodic system. In Figure 4, a pristine (5, 5) SWCNT and a singly
doped BSWCNT show band crossing at the Fermi level. Band crossing at the Fermi level represents a metallic system, whereas small
gap is a characteristic of semiconductor, and a large gap is an insulator. Figure 5 shows the band structure of the B3SWCNT. It
appears that adding multiple boron atoms to the SWCNT does perturb the metallic nature of the (5, 5) SWCNT, but not enough to
alter the metallic nature because band crossing at the Fermi level is still exhibited.
Figure 5.Band structure (left) and density of state DOS (right) plots.
The band crossing at the Fermi level shows that the addition of 3
boron atoms does not alter the metallic nature of the (5, 5)
SWCNT. The DOS shows electron population below the Fermi level.
Figure 4.Band structure of pristine (5, 5) SWCNT (left) and
boron-doped (5,5) SWCNT. Note the band crossing at the
Fermi level, indicating the metallic nature of these systems.
Figure 3. Geometrically optimized, (5, 5) BSWCNT with integrated boron cluster .The blue areas
are boron atoms and the green portion represents carbon atoms.
Path I, II, and III
Path I. Ea = 2.1 eV Path II. Ea
= 0.88 eV Path III. Ea
= 0.62 eV
These plots show the change in energy as the reaction proceeds. The first and last point on the MEP represent the geometrically optimized
product/reactant pairs. The ten points in between the first and last are the intermediate images generated by NEB calculations. Path I had
the highest activation energy, 2.1 eV, out of the four reactions and had one transition state at the maximum energy. Path II had an activation
energy of 0.88 eV. Of the four reaction paths, Path II was the only that showed a stable intermediate state (image 8). Path III showed
decreased activation energy equal to 0.62 eV. It had one transition state then experienced a steady decline to products.
Path IV
Conclusion
The above figures show the MEP for Path IV and the 10 intermediate images (2-11) that lie on the MEP between reactant and product
structures. When taking activation energy and total system energy into consideration, Path IV appears optimal out of the four. There is one
transition state and no stable intermediates which suggest the pathway’s simplicity. The present results with boron cluster doped CNT
(B3SWCNT) shows substantial improvement over the previous results with only one boron (B1SWCNT). The lowest activation barrier for oxygen
dissociation on B1SWCNT was found to be 0.81 eV. Other reaction paths considered previously for B1SWCNT had activation barrier over 1.2 eV.
The lower activation energies we observed with B3SWCNT in the present study is most likely due to the fact that B-C bond is weaker than C-C
bond. In Path IV, both oxygen atoms are attached to B-C bonds and the B-C bonds are partially broken, leading to significant distortion of
overall CNT structure. On the other hand, with B1SWCNT, at least one of the two oxygen atoms is always attached to C-C bond. The reaction
paths with B1SWCNT are also more complicated than those studied in this work with B3SWCNT, typically having more than one transition state.
2 3 4 5 6
7 8 9 10 11
Ea
= 0.50 eV
Oxygen dissociation on boron doped carbon nanotube (B3SWCNT) is studied with the density functional theory (DFT) and nudged elastic band
(NEB) calculations. From a series of NEB calculations, the most likely reaction path is determined to be path IV, where two oxygen atoms are
adsorbed on two crystallographically identical boron atoms, followed by the insertion of each oxygen atom into B-C bond. The activation barrier
for this path is 0.5 eV, which is a substantially smaller than the previous results with (5, 5)-SWCNT doped with only a single boron atom. Atom
projected density of state (PDOS) calculation shows that oxygen molecule is negatively charged, whereas boron atoms are positively charge,
implying electron transfer from boron to oxygen. In fact, the adsorbed molecular oxygen has 2
O2
-
character, which is highly reactive. However, all
three boron atoms are equally charged, indicating cooperative effect of boron. This should have contributed to the lower activation barrier
found in the present B3SWSCNT system.
The activation barrier of 0.5 eV (~11 kcal/mol) is still relatively large, but our calculations are done for gas phase systems. It is expected that
the presence of solvent reduces the activation barrier significantly. Therefore, as a future work, it would be interesting to see the role of solvent
in oxygen dissociation on the surface of doped carbon nanotube. Regardless, the PEMFC is a feasible future fuel source that holds the potential
to augment current energy systems from personal electronics to automobiles.