David A. Harrington

A microfluidic fuel cell with flow-through porous electrodes.

A microfluidic fuel cell architecture incorporating flow-through porous electrodes is demonstrated. The design is based on cross-flow of aqueous vanadium redox species through the electrodes into an orthogonally arranged co-laminar exit channel, where the waste solutions provide ionic charge transfer in a membraneless configuration. This flow-through architecture enables improved utilization of the three-dimensional active area inside the porous electrodes and provides enhanced rates of convective/diffusive transport without increasing the parasitic loss required to drive the flow. Prototype fuel cells are fabricated by rapid prototyping with total material cost estimated at 2 USD/unit. Improved performance as compared to previous microfluidic fuel cells is demonstrated, including power densities at room temperature up to 131 mW cm-2. In addition, high overall energy conversion efficiency is obtained through a combination of relatively high levels of fuel utilization and cell voltage. When operated at 1 microL min-1 flow rate, the fuel cell produced 20 mW cm-2 at 0.8 V combined with an active fuel utilization of 94%. Finally, we demonstrate in situ fuel and oxidant regeneration by running the flow-through architecture fuel cell in reverse.

Detection of Membrane Drying, Fuel Cell Flooding, and Anode Catalyst Poisoning on PEMFC Stacks by Electrochemical Impedance Spectroscopy

Membrane drying, fuel cell flooding, and anode catalyst poisoning by carbon monoxide are investigated on Hydrogenics production-type proton exchange membrane fuel cell (PEMFC) stacks similar to the stacks used in Hydrogenics HyPM 10-kW fuel cell power modules. Changes in fuel cell voltage and impedance with time are presented for each type of fault, the fuel cell stacks being controlled in galvanostatic mode. This study shows that these PEMFC stack faults can be differentiated by their impedance responses while fuel cell voltage monitoring alone is insufficient to distinguish between failure types. Membrane drying leads to an increase in the fuel cell impedance magnitude and phase angle at all frequencies studied. Fuel cell flooding leads to an increase in the impedance magnitude at low frequencies (f < 10 Hz) and to a decrease in the impedance phase angle at frequencies less than 100 Hz. Anode catalyst poisoning by CO is characterized by an increase in the fuel cell impedance magnitude at frequencies less than a few hundred Hz. For this fault, the impedance phase angle decreases within a large frequency range and is characterized by a minimum value appearing at 20-25 Hz at moderate current density.

Hydrogen Peroxide as an Oxidant for Microfluidic Fuel Cells

We demonstrate a microfluidic fuel cell incorporating hydrogen peroxide oxidant. Hydrogen peroxide (H 2 O 2 ) is available at high concentrations, is highly soluble and exhibits a high standard reduction potential. It also enables fuel cell operation where natural convection of air is limited or anaerobic conditions prevail, as in submersible and space applications. As fuel cell performance critically depends on both electrode and channel architecture, several different prototype cells are developed and results are compared. High-surface area electrodeposited platinum and palladium electrodes are evaluated both ex situ and in situ for the combination of direct H 2 O 2 reduction and oxygen reduction via the decomposition reaction. Oxygen gas bubbles produced at the fuel cell cathode introduce an unsteady two-phase flow component that, if not controlled, can perturb the co-laminar flow interface and reduce fuel cell performance. A grooved channel design is developed here that restricts gas bubble growth and transport to the vicinity of the cathodic active sites, enhancing the rate of oxygen reduction, and limiting crossover effects. The proof-of-concept microfluidic fuel cell produced power densities up to 30 mW cm -2 and a maximum current density of 150 mA cm -2 , when operated on 2 M H 2 O 2 oxidant together with formic acid-based fuel at room temperature.

Powerful Insight into Catalytic Mechanisms through Simultaneous Monitoring of Reactants, Products, and Intermediates

Electrospray ionization mass spectrometry (ESIMS) has become a valuable tool in the mechanistic study of organometallic catalytic reactions. Analysis is fast, intermediates at low concentrations can be detected, and complex mixtures are tractable. The family of palladium-catalyzed C C bondforming reactions are the most studied by ESIMS. Although the majority of these investigations have focused on the structural identification of short-lived or low-concentration intermediates, some recent studies have monitored the intensities of intermediates or reactants and products over time. 14] However, no one has yet shown this technique to be capable of providing robust kinetic information for reactants, products, by-products, and low-abundance intermediates simultaneously and under standard reaction conditions. We show herein how powerful this information can be in leading reaction design. The copper-free Sonogashira (Heck alkynylation) reaction is widely used in the synthesis of natural products, pharmaceuticals, and novel materials, but the mechanism is not well understood. Ideally, the reaction should be observed under typical reaction conditions for meaningful information to be obtained about the mechanism, because under such conditions anions and bases 18] as well as alkynes are thought to act as ligands for palladium, with complex effects on the reaction efficiency. In most cases, a large excess of an amine base is required to promote reaction; however, the exact role of the base is in question. Dieck and Heck and Amatore et al. suggested a carbopalladation mechanism in which the terminal alkyne undergoes carbopalladation and the base consumes the H formed during the b-hydride elimination that forms the product. Ljundahl et al. prefer a deprotonation mechanism in which deprotonation of the terminal alkyne by the amine occurs from the cationic intermediate [Pd(Ar)(PR3)(NR’3)(HC CR’’)] or the neutral intermediate [Pd(Ar)(PR3)(X)(HC CR’’)], depending on the electronic nature of the alkyne. An anionic mechanism has also been proposed in which [Pd(PR3)2X] and [Pd(PR3)(X)(Ar)(CCR’’)] intermediates feature. The identity of palladium-containing intermediates has been proposed on the basis of electrochemical or NMR spectroscopic data but not through direct observation. Charged tags are required for the detection of species otherwise invisible to ESIMS, 24] an idea first introduced by Adlhart and Chen; we used an aryl iodide functionalized with a phosphonium hexafluorophosphate salt, [p-IC6H4CH2PPh3] [PF6] . This tag provides very low detection limits owing to its high surface activity, and the noncoordinating counterion reduces ion pairing. The bulky nature of the charged group ensures that the ionization efficiency is largely insensitive to the remaining structure of the ion, so the intensity of the various ions correspond very closely to their real concentration (see the Supporting Information). ESIMS data on reaction progress collected under typical reaction conditions by using pressurized sample infusion (PSI) compare well with H NMR and UV/Vis spectroscopic data (Figure 1). The number of data points is much higher for

Simulation of anodic Pt oxide growth

A microscopic model of the growth of the thin anodic oxide film on Pt is presented. The slow step is migration of a Pt atom from the metal lattice to become a Pt(II) species. Pt(II) species can then rapidly diffuse across the surface. Sites of different types, e.g. terrace atoms or step atoms, have different reactivity, dependent on their number and type of bonds. Monte Carlo simulations are used to predict the kinetics and the surface reconstruction. The treatment of the structures of the oxide and metal is oversimplified by the assumption that they both have simple cubic unit cells with identical lattice parameters. The model predicts correctly the observed direct logarithmic growth law for potential step transients, without any change in behavior at full monolayer coverage. For cyclic voltammetry, it predicts an anodic peak on the first cycle, and plateau behavior for subsequent cycles. The reversible component is found, and interpreted in terms of a reversible reconstruction aided by the rapid surface diffusion. The height distributions predicted by this model agree with recent structural measurements, but the exact topography of the surface after multiple cycles is not reproduced.

Surface phases of Ni(110) induced by adsorption of deuterium

The surface phases induced by D2 adsorption on Ni(110) have been investigated by LEED, thermal desorption spectroscopy (TDS), Rutherford backscattering (RBS) and nuclear reaction analysis (NRA). The data were correlated via accurate work function (Δφ) measurements. The (2 × 1) and (1 × 2) phases saturate at coverages (θ) of 1.0 and 1.5 monolayer respectively. Nickel atoms are not displaced laterally from their bulk like positions for 0 0.01 nm. The fraction of the surface in the (1 × 2) phase increases linearly with D-coverage from 0% at θ = 1.0 to 100% at θ = 1.5. The streaked (1 × 2) phase formed by cooling in D2 or by heating the (1 × 2) phase above ~ 220 K is not simply a disordered (1 × 2) phase. Only 20% of the Ni atoms have displacements large enough to be in the (1 × 2) phase. The sticking coefficient (S) is ~ 0.35 at 175 K atθ → 0 and after decreasing to ~ 0.1 by θ ≈ 0.95 either levels out or increases around θ = 1.0 at which coverage, the (1 × 2) phase nucleates. The desorption of D2 occurs in 3 states with increasing temperature: α state, associated with the (1 × 2) → streaked (1 × 2); β1, associated with desorption from the (1 × 1) surface and β2 associated with desorption from the streaked (1 × 2) phase. The fraction of the adsorbed deuterium desorbing in the various states depends on the kinetics of the (1 × 2) → (1×1) and (1 × 2) → streaked (1 × 2) phase transitions and can be altered by trace amounts of impurities such as adsorbed oxygen.

The adsorption of water on Ni(110): Monolayer, bilayer and related phenomena

Abstract The adsorption of water of Ni(110) has been studied by nuclear reaction analysis (NRA), thermal desorption spectroscopy (TDS), LEED and work function measurements (Δφ). The major findings of this study are: (1) the saturation coverage of the first chemisorbed layer of water is slightly less than 0.5 water molecules per surface Ni atom or 0.5 ML (1 ML = 1 monolayer = 1.14 × 10 15 molecules cm −2 ) and the layer exhibits a c(2 × 2) LEED pattern; (2) this water desorbs in three separate desorption states; (3) the slightly less strongly bound, second layer of water can be distinguished from subsequent “ice” layers by a discrete work function change. These results are discussed in terms of a recently published model of Benndorf and Madey [C. Benndorf and T.E. Madey, Surf. Sci. 194 (1988) 63].

The interaction of hydrogen with a Pd(100) surface

The interaction of H2 with Pd(110) was studied by simultaneous measurements of work function (Δφ), thermal desorption spectra (TDS) and LEED. Upon adsorption at 130 K, (2x1), (1x2) and streak phases form in sequence, the first two being associated with maximum coverages of 1 and 1.5 ML respectively. The (1x2) phase, which is reconstructed, converts to the (2x1) at 180 ≤ T ≤ 210 K with the migration of hydrogen to subsurface sites. These ordered surfaces exhibit a broad, asymmetric desorption peak, which we believe consists of two distinct desorption states, which we label β1 and β2. At much higher exposures, low temperature desorption peaks α1 (∼160 K) and α2 (∼230 K) appear. The α1 peak originates from decomposition of layers of palladium hydride near the surface, the α2 from the desorption of hydrogen from the reconstructed (1x2) phase, with consequent lifting of the reconstruction. The α2 phase is thus a surface adsorbed species. The absence of a change in Δφ during β1 desorption (low temperature side of β peak) leads us to postulate that the β1 desorption is “fed” from subsurface sites which are either populated at 130 K or during the thermal desorption experiment. The β2 state exhibits second-order kinetics and a large work function change on desorption. It originates from surface adsorbed species. The normal isotope effect observed in filling the α1 states suggests that bulk diffusion is not limiting its rate of formation. A more likely explanation is layer by layer growth of palladium hydride, the driving force being lower for D2 than H2 at equal pressures because of the higher decomposition pressure of the deuteride.

An ac voltammetry study of Pt oxide growth

Abstract AC voltammetry of polycrystalline Pt in sulfuric acid solutions has been used to study the growth kinetics of the thin anodic Pt oxide film. Data were collected from 2 Hz to 50 kHz, one frequency per cycle, and were analyzed in the complex impedance plane. The faradaic process was modeled as a resistance parallel to the double-layer impedance, with a value approximately independent of potential in the do voltammetry plateau region. The equivalent circuit for the known growth law is derived and is shown to be a series RC combination. The capacitance was not detected but is expected to have a negligible effect in the measured frequency range. The value of the resistance found was consistent with the growth law found in other experiments. Evidence for additional faradaic elements in the equivalent circuit was inconclusive. We found no additional features in the impedance spectra at higher frequencies that could be associated with the fast electrosorption of OH suggested by other workers. The reversibility of the early stages of growth is therefore associated with structural reversibility rather than a fast process.

Platinum oxide growth kinetics for cyclic voltammetry

Abstract The oxide growth law dσox/dt = ao exp(cE) exp(- bσox), previously derived on theoretical grounds and shown to explain constant potential growth data for the anodic oxide on Pt, is used to predict the current—potential relationship for cyclic voltammetry. The current is predicted to rise to a plateau current whose value is proportional to the sweep rate, as observed experimentally. Experimental data from constant-potential experiments are used to find values of the parameters ao, b and c. These parameters are then used in the cyclic voltammetry equations and shown to explain correctly the experimental data. The results of sweep—hold—sweep experiments, in which the final oxide coverage is nearly independent of the hold time, are also shown to be consistent with the same rate law.