Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Development of radial flow channel for improved water and gas management of cathode flow field in polymer… Friess, Brooks Regan 2012

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2012_spring_friess_brooks.pdf [ 5MB ]
JSON: 24-1.0072562.json
JSON-LD: 24-1.0072562-ld.json
RDF/XML (Pretty): 24-1.0072562-rdf.xml
RDF/JSON: 24-1.0072562-rdf.json
Turtle: 24-1.0072562-turtle.txt
N-Triples: 24-1.0072562-rdf-ntriples.txt
Original Record: 24-1.0072562-source.json
Full Text

Full Text

Development of Radial Flow Channel for Improved Water and Gas Management of Cathode Flow Field in Polymer Electrolyte Membrane Fuel Cell  by Brooks Regan Friess B. Applied Science, University of British Columbia, 2010  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTERS OF APPLIED SCIENCE  in  The College of Graduate Studies (Mechanical Engineering) THE UNIVERSITY OF BRITISH COLUMBIA  (Okanagan) January 2012 c Brooks Regan Friess, 2012  Abstract This thesis presents an innovative radial flow field design for PEMFC cathode flow plates. This new design, which is in the form of a radial field, replaces the standard rectangular flow channels in exchange for a set of flow control rings. The control rings allow for better flow distribution and use of the active area. The radial flow field was constructed with aluminum and plated with gold for superior surface and conductive properties. These materials were selected based on the results obtained from the performance of the standard flow channels of serpentine and parallel constructed of hydrophilic gold and hydrophobic graphite materials. The new flow field design provides a competitive performance compared to the current standard serpentine and parallel flow fields in a dry-air-flow environment. The polarization curves for a dry cathode reactant flow, however, shows excessive membrane drying with the radial design. Humidifying the air flow improves the membrane hydration while the fuel cell with the innovative radial flow field produces a higher limiting current density compared to other channel designs, even the serpentine flow field. The water removal and mass transport capacity of the radial flow field was proven to be better than parallel and serpentine. This performance increase was achieved while maintaining the pressure drop nearly half of the pressure drop measured in  ii  the serpentine flow fields. The initial results for this design show promising performance and further optimization and simplification of the design should improve the performance and allow for simpler manufacturing processes.  iii  Preface A paper based on the experimental work of this thesis has been submitted for publication. [Brooks R. Friess] and Hoorfar, M. (2011) Development of a Novel Radial Cathode Flow Field for PEMFC. International Journal of Hydrogen Energy. I conducted all the testing and wrote manuscript.  iv  Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iv  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  v  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vii  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  viii  Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xi  List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xiii  Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xiv  1  2  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  1.1  Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  1.2  Numerical Modeling . . . . . . . . . . . . . . . . . . . . . . . .  8  1.3  Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . .  11  1.4  Motivation/Organization . . . . . . . . . . . . . . . . . . . . . .  18  Flow Field Design Analysis . . . . . . . . . . . . . . . . . . . . . . .  20  v  2.1  Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . .  20  2.2  Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . .  22  2.3  Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .  31  2.4  Radial Flow Field . . . . . . . . . . . . . . . . . . . . . . . . . .  38  Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . . .  44  3.1  Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . .  47  3.2  Flow Field Design/Fabrication . . . . . . . . . . . . . . . . . . .  51  Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .  59  4.1  Surface Property and Architecture Results . . . . . . . . . . . . .  59  4.2  Performance Radial Flow Field . . . . . . . . . . . . . . . . . . .  67  4.3  Pressure Drop Measurement . . . . . . . . . . . . . . . . . . . .  71  Conclusions and Suggestions for Future Work . . . . . . . . . . . .  74  5.1  Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  75  5.2  Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  77  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  80  3  4  5  vi  List of Tables Table 2.1  List of assumptions used in the numerical model [12] . . . . .  23  Table 2.2  Model variable values . . . . . . . . . . . . . . . . . . . . . .  30  Table 4.1  Differential pressure measurements for parallel, double serpentine and radial flow fields . . . . . . . . . . . . . . . . . . . .  vii  72  List of Figures Figure 1.1  Fuel cell schematic in which (1), (2), (3), and (4) represent the flow field, GDL, catalyst, and membrane respectively . . . . .  3  Figure 1.2  Polarization curve (red line) and power curve (blue line) . . .  5  Figure 1.3  Flow field schematic for a parallel flow field . . . . . . . . . .  15  Figure 1.4  Flow field schematic for a serpentine flow field . . . . . . . .  16  Figure 1.5  Flow field schematic for an interdigitated flow field . . . . . .  17  Figure 2.1  Model geometry schematic . . . . . . . . . . . . . . . . . . .  23  Figure 2.2  Meshed sub-domain of the GDL model . . . . . . . . . . . .  29  Figure 2.3  Base model polarization curve . . . . . . . . . . . . . . . . .  31  Figure 2.4  Oxygen mass fraction inside the GDL . . . . . . . . . . . . .  33  Figure 2.5  Saturation level of the GDL . . . . . . . . . . . . . . . . . .  34  Figure 2.6  Modified model geometry schematic . . . . . . . . . . . . . .  35  Figure 2.7  Oxygen concentration of the modified geometry GDL . . . .  36  Figure 2.8  Saturation concentration of the modified geometry GDL . . .  37  Figure 2.9  Modified model polarization curve . . . . . . . . . . . . . . .  38  Figure 2.10 Schematics of previous radial flow field designs . . . . . . . .  40  Figure 2.11 Radial insert for proposed radial flow field . . . . . . . . . . .  41  Figure 2.12 Schematics of new radial flow field design assembly . . . . .  43  viii  Figure 3.1  Arbin c fuel cell test machine . . . . . . . . . . . . . . . . .  47  Figure 3.2  Fuel cell component layout . . . . . . . . . . . . . . . . . . .  50  Figure 3.3  Fuel cell fully assembled with alignment rods . . . . . . . . .  50  Figure 3.4  Parallel flow field structure . . . . . . . . . . . . . . . . . . .  51  Figure 3.5  Double serpentine flow field structure . . . . . . . . . . . . .  52  Figure 3.6  Photo of the radial back plate . . . . . . . . . . . . . . . . . .  53  Figure 3.7  Schematic dimensions of radial cutout . . . . . . . . . . . . .  54  Figure 3.8  Schematic for the major diameter of the radial insert . . . . .  55  Figure 3.9  Schematic for the major thickness of radial insert . . . . . . .  56  Figure 3.10 Mounting post for ring insert . . . . . . . . . . . . . . . . . .  57  Figure 3.11 Fully assembled radial flow field structure . . . . . . . . . . .  58  Figure 4.1  Polarization curve for saturated cathode and anode with parallel flow field architecture and graphite surface . . . . . . . . .  Figure 4.2  Polarization curve for saturated cathode and anode with parallel flow field architecture and gold surface . . . . . . . . . . .  Figure 4.3  62  Polarization curve for fully humidified cathode and anode with serpentine flow field architecture for a graphite surface . . . .  Figure 4.6  61  Polarization curve for dry cathode and fully humidified anode with parallel flow field architecture for gold surface . . . . . .  Figure 4.5  60  Polarization curve for dry cathode and fully humidified anode with parallel flow field architecture for graphite surface . . . .  Figure 4.4  60  63  Polarization curve for fully humidified cathode and anode with serpentine flow field architecture for a gold surface . . . . . .  ix  64  Figure 4.7  Polarization curve for dry cathode and fully humidified anode with serpentine flow field architecture for a graphite surface .  Figure 4.8  Polarization curve for dry cathode and fully humidified anode with serpentine flow field architecture for a gold surface . . .  Figure 4.9  66  66  Polarization curve for humidified anode side and dry cathode side using the radial flow field . . . . . . . . . . . . . . . . .  67  Figure 4.10 Polarization curves for humidified anode side and dry cathode side for all flow field designs . . . . . . . . . . . . . . . . . .  68  Figure 4.11 Polarization curve for humidified anode side and humid cathode side using the radial flow field . . . . . . . . . . . . . . .  69  Figure 4.12 Polarization curves for humidified anode side and humid cathode side for all flow field designs . . . . . . . . . . . . . . . .  70  Figure 4.13 Power density curves for saturated anode side and low humidity cathode side for all flow field designs . . . . . . . . . . . .  70  Figure 4.14 Power density curves for saturated anode side and saturated  Figure 5.1  cathode side for all flow field designs . . . . . . . . . . . . .  71  Future radial flow field design . . . . . . . . . . . . . . . . .  78  x  Nomenclature is the activation overpotential  ∆Vact,c R  is the gas constant  F  is the Faradays constant  i  is the cell current density  αcora  is the charge transfer coefficient  i0,cora  is the exchange current density is the ohmic resistance  Ri n  is the number of electrons in the reaction  iL  is the limiting current density  D  is a gas diffusion coefficient  δ  is the GDL thickness  CB  is the bulk fluid concentration  ρ  denotes the density kg/m3  ωi  is the mass fraction of speciesi  xj  is the molar fraction of species j  u  is the velocity vector (m/s)  p  is the pressure (Pa)  xi  Di j  is the i j component of the multicomponent Fick diffusivity  DTi  denotes the generalized thermal diffusion coefficient (kg/ (m · s))  Mj  is the molar mass of species j (kg/mol)  M  is the molar mass of the mixture (kg/mol)  T  is the molar mass of the mixture (K) the reaction rate kg/ m3 · s  Ri  is the dynamic viscosity  µ Ke f f  is the effective hydraulic conductivity  Pcap  is the capillary pressure  γ  is the saturation level  ε  is the porosity of the GDL  εe f f  is the effecitve porosity of the GDL is the fluid temperautre  Tf PHsat2 Ov  is the saturation vapor pressure  xii  List of Abbreviations PEMFC  Polymer Electrolyte Membrane Fuel Cell  GDL  Gas Diffusion Layers  MPL  Micro Porous Layer  PTFE  Polytetraflouroethelyne  xiii  Acknowledgments I offer my thanks and gratitude to the faculty, staff and my fellow researchers at the Advanced Thermofluidic Laboratory, who have inspired me to continue my work in this field. Thanks goes to UBC for providing me with the UGF award so that I could continue my research, and to my examining committee for helping prepare this dissertation. I owe particular thanks to Dr. Mina Hoorfar, whose energy and drive have given me the opportunity to develop new and innovative ideas and to Alexander Willer for his time and efforts in machining the final design. Special thanks are owed to my family and girlfriend, whose have supported me throughout my years of education.  xiv  Chapter 1  Introduction It will take a combined effort of academia, government, and industry to bring about the change from a gasoline economy to a hydrogen economy. The forces are building and progress is being made. It is of major importance that a change of this magnitude not be forced on unwilling participants, but that all of us work together for an economically viable path to change. — Geoffrey Ballard , World Hydrogen Energy Conference  1.1  Background  Over the past decade, Polymer Electrolyte Membrane Fuel Cell (PEMFC) have been under development to replace conventional combustion-based energy generation systems. The increasing costs of fossil fuels, and environmental damage caused by conventional systems, have led to further interest in PEMFCs as they provide power with zero emissions and a higher efficiency compared to combustion systems. Fuel cells are already being developed for small stationary power generation or for large power plants(>1MW) [1]. The PEMFC is a way of converting the chemical energy directly into the electrical energy without the use of combustion. 1  The basic principle of a PEMFC is that an electrolyte layer is sandwiched between two gas flow plates that carry the fuel gas at the anode (hydrogen) and the oxidant gas at the cathode (Air/Oxygen). Since there is a solid electrolyte separating the two reactant gas streams, no mixing of the fuel and oxidant occurs. This is different from the traditional combustion-based systems in cars and large power plants in which the fuel and oxidant mix before ignition, which results in production of work and heat. Fuel cells consist of four main components including flow field plates, Gas Diffusion Layers (GDL), catalyst layer, and a solid electrolyte membrane shown in Figure 1.1[1]. The flow field plates, usually the most outer layer of the fuel cell, are responsible for distributing reactant gases evenly throughout the cell, removing excess water efficiently, and supplying a conduction route for the electrons from the external circuit. The GDL has a similar function to the flow field plate as the GDL further distributes reactant gases to the active sites on the catalyst layer, whilst facilitating the removal of water and providing and conductive path for electrons. The difference between the GDL and flow field plates is that the GDL is made of a porous carbon paper (or woven carbon cloth) that has been impregnated with a hydrophobic agent, Polytetraflouroethelyne (PTFE). PTFE makes some of the pores hydrophobic which causes them to repel water and allow a constant path for reactant gases to travel through. The other pores will remain hydrophilic and will be paths for water to move from the catalyst layer into the flow field. Another modification that has been made to the GDL is the addition of a Micro Porous Layer (MPL) which is an amorphous mixture of carbon/graphite particles and PTFE. The MPL is generally on the catalyst side and improves water removal from the catalyst layer to the GDL. The catalyst used for PEMFC appli2  cations is platinum and allows for reaction of the hydrogen (proton) and oxygen to occur, providing a driving force that creates the electrical current. The electrolyte membrane is a fluorinated polymer that has been treated (see Ma et al. [2] for details) to allow for proton conductivity and prevent electrons passing through. In general, the most popular membrane is Nafion R designed by DuPont. As a proton conductive membrane, Nafion R is partially hydrophilic and requires a certain amount of water saturation to function properly. Thus, there is a very fine balance to remove only the excess water from the catalyst area and not dry out the membrane [1].  Figure 1.1: Fuel cell schematic in which (1), (2), (3), and (4) represent the flow field, GDL, catalyst, and membrane respectively The fuel cell is fed with the oxidant and fuel in either gaseous or liquid form. The electrolyte layer allows only positive ions (protons) to pass through the membrane area to combine with the oxidant stream. The negative ions (electrons) are then 3  forced through an external circuit providing electrical current before they react with the positive ions and the oxidant on the other side of the electrolyte membrane. In PEMFCs, the fuel is hydrogen and the oxidant is oxygen or air. The anode reaction (Equation 1.1) shows the separation of the hydrogen gas into protons and electrons. 2H2 → 4H + + 4e−  (1.1)  The cathode reaction (Equation 1.2) shows the combination of the proton, electron, and oxygen, creating water and heat as by-products. O2 + 4H + 4e− → 2H2 O  (1.2)  This provides the overall reaction equation for PEMFCs as the combination of hydrogen and oxygen into water (Equation 1.3).  2H2 + O2 → 2H2 O  (1.3)  The obvious advantage of this system is that the only exhaust created by the system is water. This is a deviation from usual energy systems that are based on the heat of the reactants and produce environmentally harmful by-products such as NOx and CO2 . Since the process in the PEMFC system is not similar to the heat engine process, its efficiency is not limited by the Carnot efficiency. In essence, the ideal efficiency of the hydrogen fueled PEMFC can be as high as 83% though no fuel cell has reached this efficiency due to the losses, called overpotentials, occurring within the cell. These overpotentials produce irreversible voltage losses throughout the current range of the PEMFC. The polarization curve (see Figure 1.2) presenting the 4  voltage as a function of the current of the system shows three different sources of the overpotentials. These three main loss sections have been labeled in Figure 1.2  Figure 1.2: Polarization curve (red line) and power curve (blue line) as I for activation losses, II for ohmic losses, and III for concentration losses [1] (mass transport). The activation losses can be described as the amount of the voltage loss required to start the reaction. Activation losses occur on both the anode and cathode sides; however, the cathode activation losses are significantly larger, and hence, the activation losses in the anode can be neglected. The cathode activation losses can be described by Equation 1.4 [1].  ∆Vact,c =  i RT ln αc F i0,c  (1.4)  In this equation, ∆Vact,c is the activation overpotential in the cathode, R and T are the gas constant, and temperature, respectively, αc is the charge transfer coefficient for the cathode, F is the Faradays constant, i and i0,c are the cell current density and the exchange current density of the cathode side, respectively. The anode acti-  5  vation loss is presented in (Equation 1.5) which has a similar form as the cathode activation loss. ∆Vact,a =  RT i ln αa F i0,a  (1.5)  Activation losses increase rapidly at lower current densities, but reach a steady state early in the polarization curve. Ohmic losses (∆Vohm ), on the other hand, present a linear drop in the voltage due to resistance (Ri ) to ion transfer in the network of the fuel cell and can be described by Equation 1.6.  ∆Vohm = iRi  (1.6)  The two main ion resistant sources in PEMFC are the electron conducting structure (usually the flow field plates) and the electrolyte membrane. Even if graphite is used in the PEMFC structure the resistance of the electron conducting material is usually much smaller than the proton resistance in the electrolyte membrane [1, 3]. The ohmic resistance is significant in the medium current density range of the fuel cell. Finally, the concentration losses are prevalent at high current densities and are the limiting factor for the fuel cell current density. These losses occur when there is no longer enough reactants reaching the reactant sites to maintain the reaction. Concentration losses (∆Vconc ) can be described by Equation 1.7 and Equation 1.8 which are heavily dependent on the bulk concentration (CB ) in the cell assuming that diffusion is the limiting factor in cell performance. This can be caused by two phenomena: i) stagnation point formation near the reactant sites, and ii) excess  6  flooding in the cell preventing reactants from reaching the reactant sites.  ∆Vconc =  iL RT ln nF iL − i  iL =  nFDCB δ  (1.7)  (1.8)  In these equations, iL is the limiting current density, n is the number of electrons, D is the gas diffusion coefficient, and δ is the GDL thickness. In order to extend the performance of the fuel cell it is important to increase the current density where concentration losses become prevalent. This must be achieved by improving the transport of reactant gases to the activation sites and water management. Despite their advantages, PEMFCs have yet to gain full acceptance for competitive mobile applications such as computers and automotive industries [1, 3]. In order for PEMFC technology to gain a foothold in these competitive industries the performance, size, and cost of the system have to be significantly improved [4]. The difficulties currently attributed to PEMFC designs can be overcome if the power density of the cell can be improved. This can be achieved in two ways: i) the thickness of each cell would need to be reduced allowing for more cells in a stack, and ii) the amount of power that can be drawn from each cell would need to be increased. The current limiting factor in the fuel cell performance is water management (related to the concentration losses). In essence, the ionic conductivity of the electrolyte is dependent on the hydration level of the membrane, however, excessive water vapor condensation, due to lengthy operation or large output current, forms micro-droplets that cover the active sites on the catalyst layers, fill the pores of the GDL, and block the access of the reactant gas to the reaction site. Typically,  7  this stage, referred to as flooding, is the origin of the limiting current PEMFCs. For this reason it would be desirable to reduce the effect of flooding preventing the dramatic drop seen in the mass transport section of the fuel cell polarization curve (see section III in Figure 1.2). The GDL has been shown to greatly improve the overall performance of fuel cells. However, it is the main structure that becomes flooded at high current densities. Flow fields are the last layer in the fuel cell that removes water from the GDL and the cell. As result, the most important factor to study and understand is the interaction that occurs between the GDL and the flow field. The result of this study can lead to the improvement of the performance of a single cell by using innovative materials and architectures enhancing water removal from the cell and preventing flooding. Several numerical and experimental efforts have been made [5–7] in the past to understand the water transfer mechanism inside the network of the fuel cell, especially in the GDL and cathode flow field. Some of these efforts are listed in the following sections.  1.2  Numerical Modeling  Currently it is very difficult, if not impossible, to characterize effectively every process occurring inside the PEMFC while the cell is running. It has become extremely important for researchers to develop numerical models that can effectively show mass transport and water management. The subject of PEMFC modeling has been heavily studied over the past decade and the models described here represent only some of the attempts that have been made to better characterize the performance of different components of PEMFC. There were three types of models that were researched for modeling fuel cell phenomena inside the flow channel and GDL sub-domains: i) droplet-based models [5, 8–11], ii) diffusion-based multi8  phase models [12–15], and iii) water transport pore network models [16, 17]. Each of these types of models describe specific phenomena occurring in the fuel cell. For instance, the droplet-based models can be used to predict water motion through the flow channel but cannot be used neither to study water transfer in other layers (such as the GDL) nor to produce performance curves. The diffusion-based models do not take into account surface properties of the flow channel but can predict fuel cell performance. The pore network models are the most accurate method for predicting water motion in the GDL layer, but are complicated and hard to integrate into other models that could produce performance curves. The model developed by Kim et al. [5] uses the volume of fluid (VOF) method to predict the water droplets propagation down a channel. This method also has the advantage of being able to account for wall surface properties by varying the hydrophobicity of the wall boundaries. This model uses the Navier Stokes equations for two-phase flows that allow for the definition of a discrete water droplet moving based on the gas flow and body force induced by the surface tension of the walls [5]. Another numerical study that uses a VOF method was conducted by Choi and Son [8] who added a defined water sources that effectively simulate a pore in the GDL structure. The addition of the pore model to the VOF simulation predicts the detachment point of the droplet from the GDL to the flow channel which is a better simulation for the actual phenomena occurring in PEMFC. The model developed by Golpaygan and Ashgriz [9] also uses a VOF model, but instead of looking at the droplet motion the study focused on the deformation of the droplet caused by the shear stresses imparted by the gaseous flow. All of these droplet-based models can predict droplet motion down a channel; however, they do have significant limitations when the overall fuel cell system is analyzed. As mentioned above, the 9  droplet-based models do not take into account water development inside the GDL pores. Also, these models have predicted that a hydrophobic channel would perform better as a flow field material. A hydrophobic wall seems to create slug flows that can cause significant pressure and performance fluctuations [18]. Hydrophilic channels, on the other hand, can avoid such flows by evenly spreading out the liquid water. Another modeling approach that strictly deals with one-phase and two-phase flows are diffusion-based models. An example is the model developed by Siegel [12] which is a non-isothermal diffusion based model using a combination of the StefanMaxwell equation, Darcy liquid water equation, and a conduction module from COMSOL. The model assumes that all of the water produced at the active area is in a liquid state and no convective heat fluxes were included. This model can give a qualitative analysis of the effect of the channel geometry on water removal. However, it does not accurately take into account the actual motion of the liquid water or the interaction of the flow channel and GDL. Pore network modeling has recently been implemented to understand the water removal mechanism through the GDL. One of the initial attempts was developed by Gostick et al. [16] who modeled the GDL as cubic network of pores and throats based on pore space geometry/topology. In this model, the capillary pressure is calculated based on the Young-Laplace equation. The model then uses a pressure difference between saturated and dry pores to decide whether the fluid should travel into the neighboring pores. This type of model will also take into account convection and diffusion processes that happen inside the porous GDL by using the properties of the GDL (pore size, pore distribution, contact angle) and the water generation rate at the catalyst layer [16]. The major disadvantage of this model 10  is that it requires great understanding of the surface properties of the GDL. This model also runs into problems when it becomes necessary to link the pore network model to a flow channel-based model in order to study the interaction between the GDL and the flow field. Other models have been developed to optimize the flow field structure based on several different factors [19]. The common feature for all the models has been the use of a maximum current as the desired output, of course, the design factors for different models differ mildly depending on the model and goals of the study [20–24]. Usually, the models tend to use the GDL thickness, channel width, and and gas channel height as the main optimization factors. Most models use a interdigitated or straight flow channel structures (described in the following section) to characterize the fuel cell performance ([20–22]). However, the 3-D models developed to study the geometric effect of the flow field on the fuel cell performance have used serpentine flow fields as the numerical sub-domain [23, 24]. The 3-D models in particular are intriguing due to the ability to show how different flow field structures can influence the PEMFC performance. However, these types of models are complex and computationally expensive.  1.3  Experimental Studies  Since the development of the first PEMFC system, a prodigious amount of research has been conducted on the membrane and electrodes [25–27]. The key improvemnts occuring in the development of fuel cells are: i) the development of a polymer membrane which had lower resistance but produced four times as much power as the Nafion membrane [28], and ii) the reduction of the amount of platinum by a factor of ten [29]. Fewer improvements in the areas of material and structure of the 11  GDLs and flow fields despite their importance in water management. This reluctance stems mainly from the structural complexity of the GDL as well as complex interaction between the GDL and flow field which has not been well characterized. The main interest of the proposed research is the improvement of the material and structure of the flow field to enhance water removal at the interface of the GDL/flow field and finally from the cell. Current PEMFC flow field architectures have been around for a number of years and seem to have reached their limits. Recently, a number of studies have been conducted on the possibility of using alternative materials to build the flow field plates [30–33]. In general, the industry standard flow fields are made of graphite. The reason for using this type of material is that graphite is completely corrosion resistant. Graphite also has a low contact resistance. This becomes very important as the flow field thickness is reduced and hence the through plane resistance becomes less significant than the contact resistance [32]. Despite the advantages of graphite, it has relatively high porosity which limits the thickness of the flow field plate. The possible alternatives are: i) graphite composites and ii) metallic plates. Graphite composite plates are made of graphite particles that are molded into a shape by mixing them with a resin matrix through heat and compression. This process has the advantage of reducing porosity and makes mass production fast and effective. The drawback of this process is that the conductivity of the composite plate is lower than that of the standard graphite plates. Metallic plates are much easier to machine and are not porous allowing for reducing the thickness of plates. The disadvantage of metallic plates is corrosion problems which contribute to cell degradation and high contact resistance. The corrosion problems can be resolved by addition of alloying elements that create an oxide layer at the surface of 12  the material. Surfaces oxides tend to be non-conductive unless the correct alloying elements are present. One way of eliminating the surface oxide and corrosion problems is using metal coatings such as gold. The gold coating process can be cost effective as long as the gold thickness is kept less than 2µm [32]. Gold provides a hydrophilic surface (i.e., the contact angle of water on gold is less than 20◦ deg.) in contrast to the graphite plates that have a relatively hydrophobic contact angle (85◦ deg.)[1]. The surface properties of the flow field play an important role in water removal. There are still debates in the literature about the right choice of the material for the flow field. Some studies have concluded that a hydrophobic material can be beneficial in water removal from the flow field [5, 34], while other studies have shown experimentally that hydrophilic materials can actually provide better performance in terms of water removal from the surface of the GDL [18, 35, 36]. The research conducted by Lu et al. [18] is of particular interest because it is one of the few studies that experimentally shows the effects of surface properties on water flow for different types of flow fields. In this study, the Concus-Finn condition (θ + α < π/2 where θ is contact angle and α is the channel corner half angle) was used as an argument to why hydrophilic materials could improve water management. In essence, if the Concus-Finn condition is met then the liquid water will wick into the channel corner and be removed by the film flow effect instead of gas shearing that is the common in droplet flows. This is preferable because gas shearing force is proportional to the projected area of the droplet and the gas velocity. Removing the water in the flow field using gas shearing could be effective when either the droplet is large enough to be removed at a low gas flow rate or a high gas flow rate is applied to remove the small droplets. When the water droplet reaches a critical size (which is large enough to be removed using low air flow) there could 13  be large pressure fluctuations in the flow field producing instabilities in the fuel cell performance. Increasing the gas flow rate, on the other hand, would increase the compressor work causing larger parasitic losses. Thus, the film flow is preferred in PEMFC because it provides improved water removal while avoiding excessive parasitic losses or pressure fluctuations. This conclusion has been supported by the experimental results presented in Lu et al. [18]. Lu et al. [18] also tested different flow channel cross-sectional shapes (square, inverse trapezoidal, and sinusoidal bottom) to see if these designs would provide any changes in the water removal pattern. The square cross-sectional channel is one of the most common flow field patterns in research because it is easily machined while inverse trapezoidal and sinusoidal bottom flow channels are similar to stamped or molded flow fields used by industry. The experimental results showed that a sinusoidal-bottom channel could improve performance, however, these types of channels are difficult to produce unless the plate is manufactured using a stamping process. Clearly the effectiveness of hydrophobic or hydrophilic channels is still not fully understood. Few of the aforementioned papers have completed experimental performance tests. Instead the conclusions were based of numerical models (describing water removal/accumulation) or experimental visualization of the two-phase flow profile in the channels. These papers have provided an initial understanding as to why a hydrophobic or hydrophilic flow channel could provide improved performance. The next logical step is to provide an experimental performance curve for both types of flow channel surface properties in order to provide a more definitive performance study. Besides the material of the flow fields, the geometry of the plates are also highly important for water removal [37–39]. In general, there are three main flow field de14  Figure 1.3: Flow field schematic for a parallel flow field signs that are currently considered as industry standards: i) parallel (Figure 1.3), ii) serpentine (Figure 1.4), and iii) interdigitated (Figure 1.5) [40]. Each of the three conventional designs has particular strengths, but none of them provide the necessary combination of mass transport, water removal, and flow distribution that will provide the best performance for high current densities or for longer operational periods. Parallel or straight flow field designs are the simplest design. However, there are significant problems attributed to this design, especially if the length and number of channels are chosen improperly [41]. Parallel flow fields tend to have a very low pressure drop across the flow field which is beneficial in reducing parasitic power losses created by compressors, but can be harmful for flow distribution and water management [40]. In essence, there is not a significant force to remove water droplets formed inside the channels due to the low pressure drop associated 15  Figure 1.4: Flow field schematic for a serpentine flow field with the parallel flow field design. Instead, the droplet is likely to block the entire channel, causing stagnation areas where no reaction occurs, and hence degrading the overall performance of the cell. Another problem with the low pressure drop of the parallel design is when they are used in stacks where all of the cells are connected via a common inlet. Since the pressure drop is low the majority of the flow tends to travel through the first cell and the subsequent cells are starved for reactants. Serpentine flow fields are seen as an industry standard for PEMFC because of their high water removal rate and performance at high current densities [42]. This is achieved due to the relatively high pressure drop across the field that is created by the bends throughout the channel length. The pressure drop around the bends creates a force that removes water from the flow field and causes a convective flow  16  Figure 1.5: Flow field schematic for an interdigitated flow field of reactant gas into the GDL, providing better reactant flow to the active area and better water removal from the GDL pores. In general, increasing the number of bends will improve the power output and water management capabilities of the fuel cell [40]. The long flow path of the serpentine flow field also causes large reactant gradients with a low reactant concentration at the end of the flow field [43]. The main problem with serpentine fields is that they produce a very large pressure drop across the flow field causing large parasitic losses [44]. Interdigitated flow fields (see Figure 1.5) are very similar to parallel channels. The major difference is that there is no direct connection between the inlet and outlet channels. Instead the reactant gas is forced through the GDL to the reactive area and then back into the exit channel. This provides an effective water removal mechanism at the cathode side of the fuel cell. Any excess water existed in the GDL will be quickly removed 17  while the droplets are still small. This prevents the formation of water slugs that can block off the entire channels. Despite the strong performance of the interdigitated flow field, the pressure drop in these types of channels are even larger than that in the serpentine flow fields [40].  1.4  Motivation/Organization  The purpose of this thesis is to understanding the important factors affecting flooding at high current densities. In particular, the interaction between the GDL and flow field will be studied to elucidate the architectural features of the flow field that produce effective water removal from the GDL. With this information an improved flow field plate design will be developed in an attempt to produce better performance under a variety of operating conditions. The new flow field design will then be tested against industry standard designs of parallel and serpentine. The interdigitated design was not included in this study due to its relatively high pressure drop (compared to serpentine) and marginal performance improvement (compared to parallel). The thesis will begin with discussion on a numerical model that was developed to gain a basic insight into the interaction between the GDL and flow fields. The model, is developed using COMSOL multi-physics software. It is based on a diffusion model that shows the water content of the GDL for a given set of initial and boundary conditions [12]. After presenting the numerical modeling results, the experimental studies conducted with the standard flow field designs of parallel and serpentine will be presented. These experimental studies were performed to find a baseline performance and pressure value which will be compared to to the performance of the new flow field design. In order to provide a more complete 18  analysis, the surface properties are also analyzed in these experimental studies to not only characterize the role of the flow field material on the water removal but also to find an appropriate material for the new design. Finally, the experimental studies conducted using the new flow field design will be presented. The comparison between the performance of the new flow field design and those of the industry standard will provide an insight into water management of PEMFCs.  19  Chapter 2  Flow Field Design Analysis For the purpose of this research a diffusion-based numerical model was implemented to study the interaction between the gas diffusion layer (GDL) and flow field. This model was developed based on the work presented in Siegel [12]. The model is based on the Stefan-Maxwell equation and is implemented in COMSOL. For a more rigorous explanation on solving the Stefan-Maxwell equation the reader is referred to works done by [45–47]. This chapter will discuss in detail the formulation of the model, the modification/simplifications implemented, and the results of the model. Finally, the results will be analyzed to show which architectural features of the flow field will provide efficient water removal from the GDL.  2.1  Model Formulation  In general, the Stefan-Maxwell equation is used to predict diffusion of different components including the oxygen, nitrogen, and water vapor through the porous  20  GDL Equation 2.1. n  ∇ ρωi u − ρωi ∑ Dei jf f ∇x j + (x j − ω j ) j=1  ∇P P  = Ri  (2.1)  where ρ is the density, ωi is the mass fraction of species i, u is the velocity vector, Dei jf f is the effective binary diffusion of species i into species j, x j is the molar fraction of species j, P is the local pressure, and Ri is the species reaction rate. This equation is combined with the Darcy equation shown below(Equation 2.2).  n˙ H2 Ol = −  dPcap ρH2 Ol e f f − K µH2 Ol dγ  ∇γ  (2.2)  where µ is the dynamic viscosity, K e f f is the effective hydraulic conductivity, Pcap is capillary pressure, and γ is the saturation level. The Darcy equation describes liquid water transport. The Stefan-Maxwell and Darcy equations are balanced using a switching function (Equation 2.3) that will predict whether the water is evaporating or condensing based on the saturation level of the gaseous phases in the model.  m˙ H2 Ov.l =kc  MH2 O ε e f f ωH2 Ov (ωH2 Ov P − PHsat2 Ov )g RT f  (2.3)  + ke εγρH2 Ol (ωH2 Ov P − PHsat2 Ov )(1 − g) In this equation, g is the switching function defined based on the humidity, pressure, and saturation pressure of the system (Equation 2.4), ε is the porosity of the GDL, ε e f f is the effective porosity of the GDL due to water saturation, T f is the  21  fluid temperature, and PHsat2 Ov is the saturation vapor pressure. 1+ g=  |ωH2 Ov P−PHsatOv | 2 ωH2 Ov P−PHsatOv 2  (2.4)  2  PHsat2 Ov can be found using a semi-empirical Antoine equation (Equation 2.5)[12] PHsat2 Ov  A  = 100 · 10  B C+ T f −273.15  (  )  (2.5)  The combination of these equations with the properly chosen boundary conditions provides insight into the interaction between the GDL and the flow field while the system is running. In this thesis, the fully derived Maxwell-Stefan equation will be used to model the diffusion-based process as implemented below.  2.2  Implementation  The numerical model described below is a modified version of the model developed in Siegel [12] and has been implemented in COMSOL multiphysics. COMSOL has a built in Stefan-Maxwell module which allows for an easier implementation. The main difference between Siegel [12] and this model is that the model described here does not take into account heat transfer. This simplification is added to the basic assumptions that have been put forth for this model in Table 2.1. The isothermal assumption does not affect the overall saturation pattern in the GDL and was hence left out. This is echoed in the results shown in Siegel [12]. The basic geometry for this model is based on a two dimensional architecture of the GDL (Figure 2.1). The initial model was developed to include a narrow section of the GDL structure including a current collector section (IV) and a flow channel section (II). The 2D 22  Table 2.1: List of assumptions used in the numerical model [12] The model is a 2D GDL model The model assumes constant pressure for the gas phase and an ideal gas mixture Only steady state behavior Tafel approach is used to calculate the current density GDL has constant physical properties The contact resistance is not modeled The activation over-potential is constant and uniform  slice was taken perpendicular to the flow direction through the flow channel in order to provide information on water movement above the current collector and the flow channel. The remaining contact surfaces of the GDL are two impermeable  Figure 2.1: Model geometry schematic wall boundaries (I and V) perpendicular to the catalyst boundary (III) where the reaction occurs and water is produced. The central white section is the sub-domain of the GDL where the Stefan-Maxwell equation (Equation 2.1)and the Darcy Law (Equation 2.2) will be implemented to solve for the gaseous diffusion transport and to predict liquid water motion, respectively. In order to use the Stefan-Maxwell equation for oxygen (O2 ), water vapor (H2 Ov ), and nitrogen (N2 ) the mass fractions of each species needs to be solved to calculate the diffusion direction and velocity. The mass fractions for oxygen and water vapor can be found by taking the divergence of the mass flux for oxygen (Equation 2.6) and water (Equation 2.7) 23  species.. n  ∇ −ρ · ωO2 · ∑ DeOf2 fj · ∇x j + (x j − ω j ) · i=1  ∇P P  = − (ρ · u · ∇ωO2 )  n  f ∇ −ρ · ωH2 Ov · ∑ DeHf2 O · ∇x j + (x j − ω j ) · vj i=1  ∇P P  (2.6)  = (2.7)  − (ρ · u · ∇ωH2 Ov ) − M˙ H2 Ov i The nitrogen mass fraction can be found from the oxygen and water vapor mass fractions because the combination of them should add to unity Equation 2.8.  ωN2 = 1 − ωO2 − ωH2 Ov  (2.8)  The multi-component diffusion coefficient used in the Maxwell-Stefan equation can be calculated using Equation 2.9 which provides a measure of how the different gases or water vapor will diffuse through dry pores.  Di j = k ·  T f1.5 1 3  1 3  2  ·  P · υi + υ j  1 1 + Mi M j  (2.9)  In this equation, υi or j is the kinematic viscosity of species i or j. In order to account for pores that are partially or fully saturated with liquid water, the binary diffusion needs to be modified using an effective porosity term ε e f f as shown in Equation 2.10. Dei jf f = Di j · ε e f f  24  1.5  (2.10)  The effective porosity can be calculated using the standard porosity ε and a saturation factor γ. Combining these two factors provides an equation (Equation 2.11) for effective porosity that will vary with the saturation level of the GDL structure.  ε e f f = ε · (1 − γ)  (2.11)  The last step in defining the single-phase, multi-component model is the calculation of the velocity vector that will predict the diffusion of gas or vapor species. This can be calculated by multiplying the molar flux of nitrogen n˙ N2,di f f by the inverse of the mass fraction of nitrogen ωN2 and the mixture density ρ as Equation 2.12 as implemented by Siegel [12]  u=−  1 ωN2 · ρ  · n˙ N2,di f f  (2.12)  where the molar flux of nitrogen can be describe by Equation 2.13. n  n˙ N2,di f f =∇ − ρ · ωN2 · ∑ DeNf2 fj · j=1  ∇M M · ∇ω j + Mj M  + (2.13)  ∇P (x j − ω j ) · P To modify the above single phase model into a two-phase model, two additions need to be made: first, an equation that governs the evaporation/condensation process needs to be developed. Second, a governing equation that predicts the water motion needs to be added. The phase change (condensation/evaporation) model (Equation 2.3) uses constants for evaporation (ke ) and condensation (kc ) along with a switching function (g) (Equation 2.4), water saturation pressure (PHsat2 Ov )(Equation 2.5),  25  and gas constant (R) to predict the water phase change. The mass balance for water can be written as the change in the liquid water flux in the system being equal to the evaporation/condensation rate as shown in Equation 2.14.  ∇ · n˙ H2 Ol = m˙ H2 Ov,l  (2.14)  The liquid water flux can be described by the following equation Equation 2.15  n˙ H2 Ol = −  dPcap ρH2 Ol · Ke f f · − µH2 Ol dγ  · ∇γ  (2.15)  where Pcap is the capillary pressure and K e f f is the effective permeability defined as Equation 2.16. Ke f f = K · γ  (2.16)  One of the simplification made in this models is the assumption that the gradient of the capillary pressure over saturation remains constant.  −  dPcap dγ  = const  (2.17)  In reality it is well known that the capillary pressure gradient with respect to saturation is not constant. There are approximations that have been developed to predict the capillary pressure gradient [48]. The correlations found from these studies are based on approximations that do not use water and were originally designed for use in porous rocks and sand. In all likelihood, these approximations will introduce errors that are unknown do to the inaccuracies in the correlations. Assuming a constant capillary pressure gradient may introduce errors into the modeling sys-  26  tem, errors which are ultimately well understood, this thesis has decided to use the assumption of constant capillary pressure gradient. Using the above equations (Equation 2.6-Equation 2.17) the GDL system can now be completely defined and modeled as long as proper boundary conditions are instituted. The boundary conditions for this model are relatively simple because there are only five surfaces and many of them have repeated boundary conditions (i.e. insulation boundary conditions). For simplicity, the boundary conditions will be marked based on Figure 2.1. At the boundaries I and V the gradient of saturation are set to zero (see Equation 2.19) and a zero flux condition is applied (see Equation 2.18 were ni is the mass flux vector for species i and n is the normal vector to the boundary [49]) because they are considered impermeable walls.  ni · n|I,V = 0 ∂γ ∂x  =0  (2.18)  (2.19)  I,V  At the boundary II the GDL is open to the channel air flow which provides humidified air and removes liquid water. For this reason, mass fractions for oxygen and water vapor are set to initial conditions and the saturation level is set to zero (Equation 2.20 through Equation 2.22)  ωO2 |II = ωO2 ,o  (2.20)  ωH2 Ov |II = ωH2 Ov ,o  (2.21)  γ|II = 0  (2.22)  27  The boundary III is the catalyst area where there is a flux of oxygen out of the sub-domain modeling oxygen consumption (Equation 2.23), and liquid water flux into the sub-domain modeling water production (Equation 2.24). Ic · MO2 4·F  (2.23)  Ic 1 · + αdrag · MH2 O F 2  (2.24)  n˙ O2 |III = −  n˙ H2 Ol |III =  In Equation 2.24 Ic is the cathode local current density, F is Faraday’s constant, and αdrag is the osmotic drag factor that predicts the number of water molecules pulled from the anode side by the proton transport through the membrane. Ic can be defined using the Tafel equation Equation 2.25  Ic |III = −  Sa · d · Io 4·F  · (1 − γ) ·  F·ηc ωO2 · e R·T ωO2 ,o  (2.25)  where Sa is the specific surface area, d is the catalyst layer thickness, Io is the reference current density, and ηc is the cell overpotential. The cell overpotential is defined as ηc = VOC −Vcell where VOC is the open circuit voltage and Vcell is the cell voltage. The cathode local current density can be used to find the average cathode current density, Ic,AV G , by integrating the local current over the entire catalyst length (Lx ) and then normalizing it by the length of the catalyst layer (Equation 2.26).  Ic,AV G |III =  1 · Lx  28  Lx 0  Ic · dx  (2.26)  The water vapor mass fraction gradient at the boundary III is also set to zero to make sure there is no flux of water vapor out of the system. ∂ ωH2 Ov ∂x  =0  (2.27)  III  The boundary IV is assumed as an insulation boundary due to the current collector blocking all reactants from passing through this site. Thus, the saturation gradients is set to zero and an impermeable boundary is set to prevent any flux in the ydirection . ni · n|IV = 0 ∂γ ∂y  =0  (2.28) (2.29)  IV  The values for the variables in the previously stated equations are shown in Table 2.2. The final implementation of this model was completed in COMSOL multiphysics package , i.e., a commonly-used commercial multi-physics simulator. A linear triangular mesh system was used to discretize the sub-domain area of the GDL, along with, second-order Lagrange elements or shape functions. The air inlet boundary over the channel area is a difficult calculation boundary that requires a finner mesh in order to guarantee convergence in this area. In order to achieve this  Figure 2.2: Meshed sub-domain of the GDL model mesh requirement boundary, II is selected in the COMSOL free mesh parameters 29  Table 2.2: Model variable values Name k F R vN2 vO2 vH2 O MO2 MH2 O MN2 ωO20 ωH2 O0 T0 p0 η Sa δ i0c αDrag dPcap dγ  Expression 3.16x10−8 [Pa m2 /s] 96485 [C/mol] 8.314 [J/mol K] 17.9x10−6 16.6x10−6 12.7x10−6 32 [g/mol] 18 [g/mol] 28 [g/mol] 0.1447 0.3789 353 [K] 101 [kPa] 0.05 : 0.05 : 0.8 [V ] 1x107 [m2 /m3 ] 10 [µm] 1 [A/m2 ] 3 22.95  Description MS diffusivity pre-factor Faraday’s constant Gas Constant Molar diffusion volume, N2 Molar diffusion volume, O2 Molar diffusion volume, H2 O Molar mass, O2 Molar mass, H2 O Molar mass, N2 Inlet mass fraction, O2 Inlet mass fraction, H2 O Temperature Pressure Overpotential Specific surface area Active-layer thickness Exchange current density Drag number, H2 O Gradient of capillary pressure with respect to saturation  settings and the larges element touching boundary II is set to 9e-6. This allows for a finer mesh around the boundary that will become more coarse until it reaches the standard density of the rest of the sub-domain providing improved calculations without significant increase in computational effort. The final meshed sub-domain consisted of 813 elements. An example of the discretized sub-domain has been shown in Figure 2.2. These equations were solved using a direct linear solver UMFPACK. A parametric iteration for ηC was taken for values of 0.05 to 1 with increments of 0.05 volts in between. The solver was iterated for each parametric value until relative tolerance of 10−6 is reached.  30  2.3  Model Results  This section will describe the results that were obtained using the aforementioned numerical model and provides the theories that were developed from the numerical data which led to further insights into the improvement of the flow field architecture. The model’s preliminary results were obtained for the same base geometry as described in Siegel [12] in order to verify the model. Modifications were then made to provide further information on the properties that would make a higher performance (in terms of water removal) flow field design. The first result that was extracted from the model was a polarization curve that can be used to verify the model is working correctly. The polarization curve is shown in Figure 2.3 and is shown in the same format that Siegel [12] used for the polarization curve. This was done to allow for a direct comparison between the model  Figure 2.3: Base model polarization curve polarization curves shown in this thesis and the one shown in [12]. Comparing the  31  two curves shows that model developed here provides the similar current values for -0.1 overpotential and and limiting current density. The small discrepancy in the calculated values is probably due to the isothermal assumption used in this thesis. The voltage drop also appears to be the same for both polarization curves. Based on this information, it appears that the results provided by the model presented in this thesis should provide an accurate replicate of the model developed by Siegel [12] and the results can be used to provide data in developing the new flow field design. The key results obtained from the model are the mass fraction distribution of oxygen providing a measure of the cell current and the mass fraction distribution of liquid water presenting the features facilitating water transport through the GDL. The images shown in Figure 2.4 present the oxygen concentration within the GDL for different overpotential (ηc ) values. The average current (shown with units of Amps/cm2 to stay consistent with [12]) values have also been included to allow for comparison of cell performance at each overpotential value. At initial overpotential values the entire GDL has a strong concentration of oxygen throughout the modeled area and remains high until approximately ηc = 0.5 where the concentration begins to drop slowly around the areas near the catalyst and the current collector. The overpotential value of 0.6 begins to show a more dramatic drop in the oxygen concentration at the top right side of the GDL. Further increase in the overpotential begins to form a stagnation point at the top right area of the GDL at around ηc = 0.7 which then begins to spread across the catalyst layer as the overpotential reaches 0.9V , causing reactant starvation and ending the reaction. The saturation level of the GDL is shown in Figure 2.5 for different overpotential values, and mirrors the results shown in Figure 2.4. The saturation levels are very 32  (a) ηc = −0.1, Ic,AV G = 14.3  (b) ηc = −0.2, Ic,AV G = 80.0  (c) ηc = −0.4, Ic,AV G = 2200.0  (d) ηc = −0.6, Ic,AV G = 17850.0  (e) ηc = −0.8, Ic,AV G = 27260.0  (f) ηc = −1.0, Ic,AV G = 27790.0  Figure 2.4: Oxygen mass fraction inside the GDL low at low overpotential values but at the higher values the GDL quickly becomes saturated. This saturation effect is significantly greater over the current collector area. This show that reducing the current collector width, or perhaps breaking the collector up into blocks, would allow the reactant air to remove liquid water, leaving fewer stagnation areas. The effect of breaking up the current collector into blocks could be studied by partially reducing the current collector width in the the 33  (a) ηc = −0.1, Ic,AV G = 14.3  (b) ηc = −0.2, Ic,AV G = 80.0  (c) ηc = −0.4, Ic,AV G = 2200.0  (d) ηc = −0.6, Ic,AV G = 17850.0  (e) ηc = −0.8, Ic,AV G = 27260.0  (f) ηc = −1.0, Ic,AV G = 27790.0  Figure 2.5: Saturation level of the GDL model and calculating the new concentration and saturation patterns. An example of this geometry is shown in Figure 2.6 and uses the same boundary conditions as the model stated above. If the oxygen concentration and saturation curves are analyzed for the new geometry it can be shown that separating the the current collector into small blocks could improve water management and gas distribution. The oxygen concentration plots 34  Figure 2.6: Modified model geometry schematic of the new geometry (Figure 2.7) shows similar phenomena as the original geometry. The major difference is that the stagnation areas appear to be smaller and the growth of the stagnation area is prevented until very high overpotential values. This is particularly noticeable at an overpotential value of 0.6 Volts. Clearly the geometry with the smaller current collector is only starting to form a stagnation area where the original model already had a large depletion region. A smaller or broken up current collector should allow for better mass transport properties helping to extend the amount of the current the PEMFC can produce. The saturation concentration plots (Figure 2.8) shows similar patterns to the simpler model. The difference with the modified geometry is that the flooding regions appear to become smaller at the medium overpotential range (Figure 2.8b) and the flooding at the higher overpotentials (Figure 2.8c) is less dramatic. Clearly the largest saturation value shown in Figure 2.8f is 0.698 where as the largest saturation level in Figure 2.5f is 0.708. This may not seem like a significant difference but essentially this means that at the same overpotential value of 1.0 Volts the smaller current collector design has fewer blocked pores in the GDL allowing for continued mass transport and higher current densities. This is echoed by the polarization curve for the provided by the modified geometry model (shown in Figure 2.9). The modified  35  (a) ηc = −0.1, Ic,AV G = 15.5  (b) ηc = −0.2, Ic,AV G = 89.0  (c) ηc = −0.4, Ic,AV G = 2606.0  (d) ηc = −0.6, Ic,AV G = 25150.0  (e) ηc = −0.8, Ic,AV G = 39400.0  (f) ηc = −1.0, Ic,AV G = 40050.0  Figure 2.7: Oxygen concentration of the modified geometry GDL geometry polarization curve shows a similar trend to the base model polarization curve (Figure 2.3). The major difference is that for each current value in the modified geometry model the overpotential value is less than that of the base model. This means that the modified geometry would have a larger voltage value for every current value on the polarization curve. Based on the data derived from the simple and extended geometry models, initial ideas for possible flow field architectures 36  (a) ηc = −0.1, Ic,AV G = 15.5  (b) ηc = −0.2, Ic,AV G = 89.0  (c) ηc = −0.4, Ic,AV G = 2606  (d) ηc = −0.6, Ic,AV G = 25150  (e) ηc = −0.8, Ic,AV G = 39400.0  (f) ηc = −1.0, Ic,AV G = 40050.0  Figure 2.8: Saturation concentration of the modified geometry GDL can be postulated. One important component would be to ensure that the current collector areas are broken up and not a continuous line. This allows the reactant gas to reach the activation sites at higher overpotentials. The added benefit is that the GDL has a greater exposure to the inlet flow which facilitates evaporation of water from the external surface of the porous GDL. This is an initial step toward the identification of the flow field properties affecting the performance of PEMFC. 37  Figure 2.9: Modified model polarization curve However, as stated in Chapter 1 this type of diffusion based model does not take into account the surface properties of the flow field. Therefore, to provide a better overall picture of the flow field performance, the architectural information from the numerical model will be added to an experimental study conducted to analyze the effect of the surface properties of the flow field on the PEMFC performance.  2.4  Radial Flow Field  One design that can be produced with a small field of current collectors and symmetrical flow is a radial flow field. Radial flow field designs have been attempted before [44, 50], but have not been considered as a truly viable replacement for parallel or serpentine flow field designs. The two radial flow field designs that have been developed in [44, 50] are shown in Figure 2.10a and Figure 2.10b. Each of these designs shows promising attributes that could provide improved performance  38  but still suffer from flaws that cause significant problems when implementing them in PEMFCs. For instance, the design reported in [50] is an interesting design improving mass transport (see Figure 2.10a). This particular design is a combination of a radial flow direction and an interdigitaded flow control system with no direct connections between the inlet and outlet of the flow field. While improving mass transport, this design produces a large pressure drop that may offset performance improvements similar to an interdigitated design. The radial flow design developed in [44], on the other hand, avoids the large pressure drop (compared to the previous design) using defined channels that radiate outwards from a central inlet (see Figure 2.10b). The radial channels are linked with rings that are used to keep the pressure drop even across the channel length. The major problem with this design is that it has multiple outlets that are not desirable for fuel cells as they increase complexity and can create problems for water removal and pressure distribution when one of the outlets is clogged with water. In essence, both designs use defined rectangular channels that flow from the center of the plate to the outer rings. The idea of using defined channels is a carryover from the conventional designs such as serpentine, parallel, and interdigitated flow fields. However, in radial style channels, a defined rectangular channel forces the bulk of the reactant gas to be concentrated in the channel and does not make use of the large triangular current collecting area. Performance of a radial flow field could be improved if the overall design allowed for the reactant gas to be distributed across a larger portion of the active area. In this research, a new radial flow field is introduced which improves the gas distribution and overall performance of the fuel cell. The major difference between the proposed radial flow design and the previous versions in [44, 50] is that the con39  (a) Interdigitated Radial Channel  (b) Multiple Exit Radial Channel  Figure 2.10: Schematics of previous radial flow field designs  40  ventional rectangular channels are completely removed. Instead, a series of rings are machined into the radial insert (see Figure 2.11) to maintain flow distribution and the desired pressure drop. The design has the inlet for the gas centered in the cell flowing through the center of the back plate (Figure 2.12c) and through the radial insert before spreading out radially along the control rings. The rings are de-  Figure 2.11: Radial insert for proposed radial flow field signed to not only offset the diffuser effect caused by the increasing cross-sectional flow area, but also produce a convective force into the GDL around the current collectors improving water removal in this area. Once the reactant gas has reached the edge of the radial insert, the major portion of the pressure drop in the system has been achieved and the flow easily moves down the gap that is present between 41  the back plate and the insert. The flow then travels back towards the center of the back plate before reaching the ring exhaust (outer ring cut in Figure 2.12c) and leaves the cell. The advantage of using control rings is that a large amount of the active area is open to the reactant flow; the pressure drop is concentrated to the areas near the current collectors where convective flow into the GDL is preferred; and the current collector center has a shorter distance to an open area compared to conventional or radial flow field designs presented in [44, 50]. The convective flow and short center distance to the open area resolve the flooding issue that commonly exists over the current collector area. Since the new radial design allows for the better use of the active area, the overall mass transport properties of the system are also improved. The evaporation of liquid water from the GDL is also improved because of the large area that is exposed to the reactant gas and the short distance that the reactants have to travel across the GDL before reaching the gas outlet. The combination of these design factors provide much better mass transport and water management properties compared to many of the conventional flow field designs. This new radial system can also have a secondary water outlet machined into the bottom of the plate (see point C in Figure 2.12c). This would make the system even more stable over long runs. When excess liquid water in the system begins to build up, the water outlet can be opened momentarily without any loss in performance.  42  (a)  (b)  (c)  Figure 2.12: Schematics of new radial flow field design assembly 43  Chapter 3  Experimental Procedure In the previous chapter an initial numerical analysis was conducted to gain insights into the flow field architectural features that improve water removal from the GDL. To complete the analysis, the most suitable materials and flow field designs that contribute to the performance of the flow field must be identified experimentally. Numerical models have currently shown that hydrophobic materials, which tend to form the water into spherical droplets, would provide the best water removal performance from the flow field [5]. However, as stated in Chapter 1, the flow channel models like [5] do not take into account the interaction between the flow field and the GDL. If the interaction between the GDL and flow field is considered in the model, there will be less of a surface tension gradient between the partially hydrophobic GDL and the hydrophobic flow field. This lack of the surface tension gradient could prevent water from efficiently moving out of the GDL and into the flow channel area, creating increased flooding regions in the GDL structure. However, if the flow field is constructed of a hydrophilic material, which causes water to spread out into a thin film layer, the liquid water would be more likely 44  to leave the GDL due to the surface energy gradient mentioned above. Another possible problem associated with hydrophobic flow field materials can be slug formation. The current numerical models such as [5] assume only droplets of water form in the flow field. However, if a large amount of water accumulates in the flow field structure, it could form a slug instead of individual droplets. The slug can completely block the cross-sectional area of the flow field. This is particularly disastrous in parallel style channels in which the entire length of a blocked channel is inaccessible to new reactants as it does not have enough pressure drop to remove the slug. Hydrophilic materials, on the other hand, will tend to form films instead of slugs which can slightly block the air flow in a channel rather than completely blocking the reactant gas flow. Finding flow plate materials that provide a significantly different surface energy while providing high conductivity and corrosion resistance, is an area that has been studied extensively in the fuel cell research community [32, 33, 51]. One of the materials that is mostly used for fuel cell systems is graphite which has a relatively hydrophobic surface (contact angle of water on a graphite surface is 85◦ deg.) [1]. Graphite is non-corrosive and can easily be machined into the simple structures of the serpentine, interdigitated and parallel flow fields. Thus, in this study, graphite is used as the hydrophobic material. Finding a hydrophilic material that provides comparable conductivity and strong corrosion resistivity, while still being able to be easily machined, proved to be difficult. Originally aluminum plates were chosen because they can be easily machined; however, aluminum corrodes significantly in the fuel cell environment. Aluminum can also leach ions into the membrane, blocking pores for proton transfer [32]. This ion leaching becomes a significant problem especially if the aluminum is in direct contact with the mem45  brane. To counter this problem, in this work, the aluminum plates are gold plated to prevent corrosion and ion leaching. Another experimental analysis that must be conducted in parallel with the surface property tests is the study of the effect of flow field architectures on water removal. It has been reported that one of the reasons for the large performance improvement in the serpentine flow field is the relatively large pressure drop across the channel length compared to that in the parallel flow field [40, 42]. This increased pressure drop allows the serpentine design to expel water from the flow field area more efficiently which results in better gas distribution. This high pressure drop also produces a convective flow into the GDL removing droplets more efficiently from the GDL and keeps the droplets at a sufficiently small size to avoid large slug formation. Although a larger pressure drop across the flow field is desirable to enhance water removal it can cause an increase in parasitic losses. Parasitic losses due to the pressure drop are negligible for miniature flow fields. However, if the system is in a large stack or has a large active area, the imposed losses due to the pressure drop can become significant. The large pressure drop in the serpentine flow fields makes parallel flow fields attractive for large stacks. In this research parallel, serpentine and the new radial fields (introduced in Chapter 2) will be tested in terms of the performance and pressure drop . The latter will be obtained using a differential pressure measurement system determining the total pressure drop across the flow field. In this chapter, the experimental setup and the design and fabrication of the three flow fields studied here will be described in detail.  46  3.1  Experimental Setup  The experiments were conducted to analyze the performance of each flow plate design in two extreme operating conditions and to measure the pressure drop across each channel to estimate parasitic losses that are caused by each of the flow design. All of the experiments were conducted using an Arbin c fuel cell test machine Figure 3.1. The test machine in its current test configuration has three gas flow lines,  Figure 3.1: Arbin c fuel cell test machine  47  two humidifiers, and two heaters. Two of the gas flow lines are for the reactant gases and subsequently have one humidifier and heater to allow for proper conditioning of the reactant gases. The third gas line is for a nitrogen purge line that will remove all reactant gases and water from the anode and cathode flow field. The test machine also contains a potentiometer and electronic load to allow for the measurement of the voltage and current produced by the fuel cell providing performance curves. The potentiometer has three settings (i.e., low, medium, and high) for the current measurement system and two settings for the voltage levels. The three current levels are for the ranges of 0.5 amps, 5 amps, and 50 amps while the voltage levels are for the ranges of 0-2 volts and 0-10 volts. The current range also has an automatic setting that will choose the correct range in order to obtain high resolution. For the particular set of tests conducted using a single cell system, the automatic range of current and 0-2 volt rang of voltage were used. Each flow field was tested using flow rate values of 0.06 slpm on the anode side and 0.14 slpm on the cathode side. The back pressure for the system was set to zero for every test completed in this thesis. The cell was also tested at two different humidification levels. The anode side was maintained at 100% humidity (dew point temperature 80◦ C) for all tests; however, each cathode plate was tested at both saturated (dew point temperature 80◦ C) and 9.5% humidity (dew point temperature 25◦ C) humidity. The tests with saturated and dry operational conditions were chosen to study mass transport and water removal properties of the flow field designs. These conditions are extreme but will provide extensive information on how the cell removes water from the fuel cell. The gas temperature was set to 80◦ C while the cell temperature was maintained at 65◦ C. Each flow field was run at 0.6 V for 4 hours to ensure that 48  the temperature, humidity, and membrane water content has been stabilized. The cell was then run through a serious of current ramps that started at 0 amps and was increased at a rate of 0.0003 amps/sec until a voltage of less than 0.05 V was reached. The cell was allowed to rest for 4 min; while the cell voltage recovered before the next current ramp was initiated. This cycle was repeated for four hours to ensure the results were consistent. The pressure drop measurements were conducted using a differential pressure measurement setup. The setup used a Pasco ScienceWorkshop interface that was connected to two pressure sensors that have a range of 0-10 kPa and a resolution of 0.005 kPa. One of the pressure sensors was connected to the inlet line just before it was connectes to the flow plate. The second sensor was connected just after the outlet line where it leaves the flow plate. Both sensors were connected using Swagelock T-joint connectors. These connections produce some pressure drop due to the expansion/contraction of the line around the T-joint, but should remain constant for all of the experiments. The two pressure sensors were then connected to the Pasco interface which measured the differential pressure based on the data obtained from the two pressure sensors. Fuel cell assembly procedures were strictly controlled to ensure that the changes in performance are due to the flow field design and surface properties rather than experimental errors. The cell assembly components are displayed in Figure 3.2 which include (from left to right) insulating outer plate, gold plated current collector plate (anode), anode flow field, membrane assemble encased in rubber gasket, cathode flow field, gold plated current collector plate (cathode), and insulator plate with assemble bolts. The cell was assembled using gloves to prevent any membrane contamination and dirt that will prevent sealing. Also, two alignment rods 49  Figure 3.2: Fuel cell component layout were used throughout the assembly procedure until the cell was completely assembled and torqued to the proper rated value (see points a and b in Figure 3.3). Once  Figure 3.3: Fuel cell fully assembled with alignment rods the cell is aligned properly, nuts are added to the mounting bolts and are torqued to 35 inch/pounds which is the standard torque value for the cell system described above, allowing for proper compression and sealing of all the cell components.  50  3.2  Flow Field Design/Fabrication  In this experimental study, the cathode flow field was tested using three different designs (parallel, double serpentine, and the new radial flow field) and different materials (graphite and gold); while for all the tests a graphite double serpentine flow field was used in the anode side. The rectangular channels for the parallel (Figure 3.4) and double serpentine (Figure 3.5) flow fields were machined 1 mm wide by 3mm deep. The inlet/outlet pipes were drilled into the side of the flow  Figure 3.4: Parallel flow field structure field and were connected directly to the rectangular channels of the parallel field (see dotted boxes in Figure 3.4) and to the inlet/outlet channels of the serpentine flow field (see dotted boxes in Figure 3.5). The active area for the parallel and double serpentine fields was held at 5 cm2 . The diameter of the radial flow field active area was set equal to the length of the parallel channels. This diameter was 51  Figure 3.5: Double serpentine flow field structure chosen to provide a similar flow length compared to the parallel and serpentine flow fields and to allow for the use of the same double serpentine flow field on the anode side. This produced a slightly smaller active area of 3.93 cm2 which was taken into account when calculating the current density. The back plate of the new radial design (see Figure 3.6)) consists of a center circle (23.2 mm shown in Figure 3.7) that is recessed (3.004 mm) into the plate to allow for the placement of the ring insert. The two cutouts in the center of the plate are the outlet (ring cutout shown in Figure 3.6 with the outer diameter of 6 mm and the inner diameter of 5 mm) and the inlet (the center hole with the inlet of 3.175 mm ) as shown in Figure Figure 3.6. The outlet ring is cut 3.70 mm into the material; while the inlet is cut 8 mm deep into the material. The offset in depth of the inlet and outlet is necessary so that the  52  Figure 3.6: Photo of the radial back plate outlet pipe that links the center hole to the edge of the plate will not cut through the outlet ring. The inlet and outlet pipes that run inside the plate come out at points a and b as shown in Figure 3.6. The main diameter and base thickness of the inset rings are 22.780 mm and 1.889 mm, respectively (see Figure 3.8 and (Figure 3.9). These diameter and thickness are specifically chosen to allow for a constant cross-sectional area adjusting the diffusion effect caused by the flow moving radially outwards. The rings control the pressure drop and ensure that most of the pressure drop occurs at the active area. Each control ring has a thickness of 0.5 mm and rise 0.9 mm from the base structure of the insert. The control rings have 1 mm by 0.4 mm slots cutout to reduce the pressure drop per ring and provide  53  Figure 3.7: Schematic dimensions of radial cutout a path for liquid water to move. The bottom of the insert has a post built up (see Figure 3.10) to be press fit into the back plate. There is also a small offset ring that allows for the insert to stay out of contact with the back plate and allow for air flow underneath the insert. The top ring is press fit into the back plate to ensure that there is no significant resistance between the ring insert and the back plate. The assembled structure is shown in Figure 3.11. Once finally assembled, the flow will travel through the center hole (point a in Figure 3.11) before traveling radially out (red arrows in Figure 3.11) towards the edge of the ring insert. Once the flow has reached the edge of the insert the flow travels down along the gap that exists between the ring insert and the back plate (point b in Figure 3.11). The flow then travels behind the back ring plate towards the outlet cutout in the back plate of the assemble. The velocity should increase as the gas flow moves towards the outlet 54  Figure 3.8: Schematic for the major diameter of the radial insert enhancing water removal behind the ring insert. Each flow field was CNC machined from solid blocks of aluminum or graphite at the UBC Okanagan Campus engineering machine shop. The aluminum plates were gold coated to produce superior electrical, corrosion, and surface properties. The gold plating was completed in two steps: first a nickel layer (∼ 10µm) was applied to prevent aluminum from bleeding through the metal coatings which would cause the membrane to be poisoned with the aluminum ions. The second layer applied was a thin layer of gold (< 3µm) to provide the correct surface and conductiing  55  Figure 3.9: Schematic for the major thickness of radial insert properties. Each flow field was tested using a Nafion 212 membrane which was hot pressed to an Etek GDL including a micro porous layer (MPL). Both the membrane and GDL were supplied by Lynntech Inc.  56  Figure 3.10: Mounting post for ring insert  57  Figure 3.11: Fully assembled radial flow field structure  58  Chapter 4  Results and Discussion The results obtained for two different materials (graphite and gold) and three different flow field designs (parallel, serpentine, and radial) show the effect of surface properties and architectures on mass transfer and water removal and demonstrate that the two attributes are not mutually exclusive. Multiple polarization curves are shown for each test to provide a measure of the stability of each flow plate. In essence, if the performance of the flow field is high but unstable the system will not be effective in real world applications.  4.1  Surface Property and Architecture Results  Figure 4.1 and Figure 4.2 shows the first comparison for surface energy conducted with the parallel flow field design with a fully humidified air flow in the cathode. The curves obtained using the graphite plate and fully humidified gas flows in both cathode and anode are very consistent over the entire experimental test. A slight instability appears just as the flooding effect becomes prominent, but the system converges to the limiting current of 0.26 Amps/cm2 . The performance of 59  Figure 4.1: Polarization curve for saturated cathode and anode with parallel flow field architecture and graphite surface  Figure 4.2: Polarization curve for saturated cathode and anode with parallel flow field architecture and gold surface  60  the gold-coated parallel flow field with fully humidified gas flows in both cathode and anode proves to be slightly less stable (seeFigure 4.2) as each test run has its own distinct path. However, the deviation between the runs is still small and should not introduce significant power fluctuation during constant running. This was tested during the warm up test during which little variation in current output at 0.6 Volts was observed for the parallel gold flow field. From the point of view of the limiting current density, all of the test curves obtained for the gold plated parallel flow field provide a higher limiting current density (averaging 0.34 Amps/cm2 ) than any of the curves obtained using the graphite parallel field. As expected, the parallel flow field architecture in particular benefits from the hydrophilic surface property. This is due to the fact that the low pressure drop in the parallel channels make it particularly susceptible to slug formations that are alleviated by a more hydrophilic surface. The above tests were repeated at the dry conditions for graphite and gold surfaces. The results are shown in Figure 4.3 and Figure 4.4. The po-  Figure 4.3: Polarization curve for dry cathode and fully humidified anode with parallel flow field architecture for graphite surface  61  Figure 4.4: Polarization curve for dry cathode and fully humidified anode with parallel flow field architecture for gold surface larization curves obtained for graphite at the dry air flow conditions show a more pronounced ohmic loss region (due to drying the membrane) and more instability deviation between different runs. Similar to the fully-humidified tests the gold flow field appears to be more unstable, showing significantly different limiting current densities among different runs. Once again the warm up showed that the gold plate does provide a consistent current output at 0.6 Volts. Finally, similar to the fully humidified condition, the gold flow field provides a higher limiting current density (0.6 Amps/cm2 ) compared to the graphite parallel flow field (producing the limiting current density of 0.54 Amps/cm2 ). In summary, the comparison between the polarization curves obtained for the humidified (Figure 4.1 and Figure 4.2) and dry (Figure 4.3 andFigure 4.4) air conditions (in the cathode side) shows that the performance of the gold plated parallel flow field is significantly better than that obtained for the graphite parallel flow field. In both the humidified and dry cases, the gold-plated flow field provides an improvement in the limiting current density  62  by approximately 0.1 − 0.05 Amps/cm2 which is a small performance increase but considerable compared to the overall limiting current of the cell (%10). Since the architecture is the same, the improved limiting current density must be due to the enhanced water management rather than improved mass transport. This is in contrast to the model predictions that were discussed in Chapter 2. The gold-plated flow field also shows a slight improvement in the ohmic loss region that has less of a slope compared to that observed for the case of the graphite flow field. A double serpentine architecture is also tested with the same two surface properties. The experimental polarization curves for the fully-humidified(Figure 4.5 and Figure 4.6) and dry (Figure 4.7 and Figure 4.8) tests show that a hydrophilic surface property does not perform well for the serpentine architecture. The two humidified curves show that the graphite polarization curves (Figure 4.5) are very consistent in the activation and ohmic loss regions but becomes much less stable in the flooding regime. The performance of the graphite double serpentine is significantly higher  Figure 4.5: Polarization curve for fully humidified cathode and anode with serpentine flow field architecture for a graphite surface  63  Figure 4.6: Polarization curve for fully humidified cathode and anode with serpentine flow field architecture for a gold surface compared to both the gold serpentine flow field. The graphite double serpentine was able to reach an average limiting current density of around 0.4 Amps/cm2 which is 0.06 Amps/cm2 greater than the gold parallel flow field. This performance is not mirrored by the gold double serpentine channel (Figure 4.6. In the fully-humidified tests, the performance curves of the gold serpentine plate are very erratic and it is difficult to draw many conclusions. It appears that between some of the runs that excess water was not removed and stayed in the flow field causing premature flooding in the following run. Based on the above results it can be concluded that a hydrophilic material is not appropriate for the serpentine design since this type of flow fields rely on a large pressure drop to push water around the curves of the flow channel. This increases the shearing force required to move the water and causes excessive flooding. This is in contrast to the parallel flow field design in which the straight flow path and low pressure drop allow a film flow to develop. This results in effective water removal. In both cases, there will be more  64  water build up in the flow field when a hydrophilic surface is used, but the parallel flow field has a direct outlet that the film flow will easily travel towards and out of the fuel cell. The difference between the graphite (Figure 4.7) and gold (Figure 4.8) double serpentine at the dry air condition is even greater than that shown in the fullyhumidified condition. The graphite double serpentine (Figure 4.7) has fair stability throughout the activation and ohmic loss regions. A significant instability appears in the mass transport region around 1.2 Amps/cm2 where a few of the test curves reach a premature limiting current of approximately 1.4 Amps/cm2 . The vast majority of the test curves reach the limiting current of approximately 1.65 Amps/cm2 which is by far the highest limiting current density of the conventional flow fields tested. The gold serpentine performance with dry inlet flow once again provides a very erratic performance. The stability is mildly improved when compared to the fully humidified case and some of the test curves reach a limiting current that are close to 0.6 Amps/cm2 . However, the overall performance of the gold serpentine flow field is to inconsistent to be competitive when compared tot he other flow fields tested. The above results suggest that the effectiveness of the hydrophilic surface in terms of water management greatly depends on the flow field architecture as there is a wide range in performance when such a material is used . It is very obvious that the hydrophilic surface is very effective in conjunction with a flow field that has no significant obstructions from the inlet to the outlet of the field. In an open type of flow fields, the water can form a film flow that will provide a strong water removal force. This type of water removal phenomena causes more water to be retained in the flow field but prevents slug formation which can cause performance instability. If the flow field has a number of bends or signif65  Figure 4.7: Polarization curve for dry cathode and fully humidified anode with serpentine flow field architecture for a graphite surface  Figure 4.8: Polarization curve for dry cathode and fully humidified anode with serpentine flow field architecture for a gold surface icant obstructions, the hydrophilic surface will cause excessive water build up in the obstructed area and cause premature flooding.  66  4.2  Performance Radial Flow Field  For the purpose of this thesis, the new radial design will be tested using only one of the materials (gold or graphite) mentioned thus far. Since the radial design has no sharp bends where water can get trapped and the flow structure is open, the radial design should behave more like the parallel flow field. For this reason the radial flow field was developed using a hydrophilic gold surface to provide the best performance. The polarization curves for the radial flow field show great performance compared to the previous results shown in Section 4.1. For the tests with the dry air flow at the cathode side, the fuel cell with the new radial flow field provides the limiting current density as high as1.7 Amps/cm2 (see Figure 4.9). In the dry tests,  Figure 4.9: Polarization curve for humidified anode side and dry cathode side using the radial flow field the cell suffers from excessive ohmic losses which are most likely due to drying the membrane. This loss is alleviated slightly after the current density of 1 Amps/cm2 . This validates the assumption of membrane drying . The curves for the dry tests also show a very small amount of mass transport loss near the end of the polar67  ization curve. Despite the large ohmic losses the cell performed well(producing a limiting current density of 1.7 Amps/cm2 which is the second best limiting current of all the flow fields tested as shown in Figure 4.10). The dry test also produces an  Figure 4.10: Polarization curves for humidified anode side and dry cathode side for all flow field designs unusually low open circuit voltage of 0.7 Volts. The reason for this phenomena is not entirely clear. The radial flow field does not show the same open circuit voltage in the fully-humidified tests which was conducted right after the dry test without taking the call apart between the two operating conditions; Therefor, it is unlikely that there is any physical damage imposed on the cell. The tests conducted with fully-humidified air flow in the cathode (see Figure 4.11) show a drastic improvement in performance compared to the previous flow field designs. In this case, the radial flow field provides limiting current densities as high as 0.8 Amps/cm2 (which is significantly better than even the graphite serpentine flow field shown above). The ohmic loss observed in the dry test does not appear once the cathode flow has been humidified. The increased humidity does 68  Figure 4.11: Polarization curve for humidified anode side and humid cathode side using the radial flow field not seem to effect the consistency of the radial flow field in low to mid current densities. The test curves are tightly grouped together until the cell begins to flood around 0.7 Amps/cm2 to 0.8 Amps/cm2 where the curves begin to diverge. This divergence is similar to the fully-humidified tests conducted with the graphite serpentine flow field, proving that the radial flow field is just as stable as the double serpentine design (see Figure 4.12). Not only is the radial flow field consistent and also produces the highest current densities during the saturated cathode flow tests. The radial flow field is not only consistent but also produces the highest current densities during the fully-humidified tests. This improved performance would suggest that the radial flow field can efficiently remove water, and means that the small drop near the end of the polarization curve of the dry test is likely due to mass transport restrictions not flooding. The performance improvement may seem small when the polarization curves are compared. However, the power densities curves (Figure 4.13 and Figure 4.14)  69  Figure 4.12: Polarization curves for humidified anode side and humid cathode side for all flow field designs show that even a small increase in the limiting current (using the radial flow field) can significantly enhance the power.  Clearly the serpentine graphite flow field  Figure 4.13: Power density curves for saturated anode side and low humidity cathode side for all flow field designs at the dry air condition provides the largest power density output (0.480 W/cm2 ) which is 19.8 % greater than the radial flow field run under the same condition. 70  Figure 4.14: Power density curves for saturated anode side and saturated cathode side for all flow field designs This improvement is reversed when the cell is run at the fully-humidified air flow at which the power density of the radial flow field was 0.419 W/cm2 (serpentine graphite provided 0.307 W/cm2 ) and provides a power increase of 36.5% compared to the double serpentine flow field with a graphite surface. This small power density increase can provide a significant performance boost considering larger power systems.  4.3  Pressure Drop Measurement  The results for the pressure drop measurements have been summarized in Table 4.1 and show that the parallel flow field has a very low pressure drop; whereas the double serpentine flow field has a pressure drop ten times larger. This trend, which is similar to that reported in Spernjak et al. [52], explains the enhanced water removal mechanism in the serpentine flow field [40, 42]. It also shows that despite the relatively poor performance of the parallel flow field the low pressure drop will prevent excessive parasitic losses. The radial flow field does not match the low 71  Table 4.1: Differential pressure measurements for parallel, double serpentine and radial flow fields Flow field architecture Parallel Double serpentine Radial  Differential pressure [kPa] 0.0302 0.3072 0.1899  pressure drop of the parallel flow field, but it is half of the pressure drop of the double serpentine while maintaining a high performance during the polarization tests. In conclusion changing the surface properties can cause the serpentine flow field structure to show a drastic decrease in performance to the point where the parallel flow field with the same surface coating provides a higher limiting current density. The design that provided the largest limiting current density is the graphite serpentine flow field at the dry-air flow. The combination of the high pressure drop and hydrophobic surface properties allow for fast liquid water droplet removal while the droplets are still small. However, the gold material improves the performance of the parallel flow structure because it creates a film of water that could travel straight down towards the outlet. Based on the numerical model and the experimental data that was gathered (Section 4.1) the new radial design was developed and tested. The new design produced significant performance gains at the dry-air condition compared to the parallel flow field. Despite this promising performance the dry test produced excessive ohmic losses and the limiting current density was still slightly smaller than the double serpentine with a graphite surface. When the cell was tested in a fully-humidified operating condition the ohmic loss problem associated with the radial flow field was relieved and its performance surpassed all  72  of the other tested designs This proves the water management capabilities of the new flow field design.  73  Chapter 5  Conclusions and Suggestions for Future Work In this thesis the numerical and experimental approaches were used to study the effect of surface energy, architecture, and pressure drop of the flow field in the cathode side on the the fuel cell performance. The numerical model developed based on the model presented in Siegel [12] was used to identify the design features affecting water removal from the interface of the GDL/flow channel. The experimental studies were conducted to study the effect of surface properties and designs to elucidate features in the flow field improving cell performance. Additional pressure drop measurements were conducted to provide a measure of parasitic loses caused by the cathode flow field. These studies and measurements provided insight into the design of a new flow field in the form of radial flow which proved to have better water removal capabilities compared to the current industry-standard flow fields including the graphite serpentine channels.  74  5.1  Conclusions  • Numerical model - The Maxwell-Stefan equation for multi-component diffusion was implemented along with Darcy equation for liquid water motion in the GDL sub-domain. This model was used to test two systems: i) a simple parallel flow field with equal channel and current collector widths, and ii) a simple parallel flow field including a current collector with the half of the width of the one included in the the original model. The model showed that the smaller the current collector area the smaller the saturation level in the GDL. This allows for improved reactant transport and better cell performance. • Effect of surface properties and flow channel geometry - double serpentine and parallel flow field designs were machined out of graphite and aluminum. The latter was gold-plated. The graphite and gold-coated plates presented hydrophobic and hydrophilic surfaces, respectively. The performance of the fuel cell was tested using each of these flow fields in the cathode side while the anode flow field was kept as a graphite double serpentine for all tests conducted using the Arbin test system. The tests were run with a fullyhumidified hydrogen gas flow on the anode side. Two conditions were used on the cathode side: fully-humidified and dry air flow. The comparison between the results obtained for the dry and fully-humidified cases showed that a hydrophilic surface for the parallel structure provides a small performance increase. The double serpentine architecture showed a very different performance when a hydrophilic surface was used: the performance of the hydrophilic serpentine flow field was much lower and less stable than the 75  hydrophobic serpentine. In addition to the performance tests, the pressure drop for each of the flow field designs were measured using a Pasco differential pressure measurement system to provide a measure of the parasitic loss for each flow field design. As it was expected the pressure drop in the double serpentine was significantly higher than that measured for the parallel design. • Radial flow field design - A new hydrophilic radial flow field was developed based on information gathered from the numerical model. The new structure uses flow control rings that have notches cut out to reduce pressure drop. This design has the advantage of using small current collector areas that provide better access for the gas flow to remove water from the GDL. The new radial design was machined out of aluminum and coated with gold since it presents an open flow design similar to the parallel flow field for which the gold-coated surface provided better performance than the graphite surface. This hydrophilic radial flow field was tested similar to the other plates. The results obtained for this new design showed promising performance in the dry test. However, the cell experienced a relatively large ohmic loss (compared to the graphite double serpentine field which showed the highest performance). This large ohmic loss was due to the large water removal rate from the new flow field which resulted in drying out the membrane. In the fully-humidified test, the ohmic loss was relieved and the new architecture provided the best performance among all the flow fields tested. The pressure drop measurement obtained for the radial flow field was larger than that obtained for the parallel flow field but was nearly half the pressure drop of the  76  serpentine flow field.  5.2  Future Work  The pressure drop of the current radial flow field design is nearly half of the serpentine flow field. However, this pressure drop is still much greater than a parallel flow field but could be reduced by either increasing the size of the slots cutout of the radial rings or reducing the number of rings. Thus, depending on the application the pressure drop can be altered easily which shows the flexibility of the proposed design. Reducing the pressure shop without degrading the performance would be an asset but the biggest improvement would be to simplify the design for mass production. Future work for this design will include simplifying the flow filed plate architecture to enhance overall packaging and performance. The current radial flow field is a two-piece press-fit design developed to ensure even flow and pressure distribution. This relatively complicated design was successful in the experimental testing, but is too difficult and expensive for industry use. This design also requires a relatively thick flow field plate in order to fit all of the inlet and outlet pips at the back of the press-fit ring. This becomes problematic in the development of stacks in which the largest portion of the volume is occupied by the flow field plates. For this reason, most fuel cell manufactures attempted to produce very thin flow plates (¡5mm). In order to produce the proposed radial flow plate in such a small thickness the design needs to be altered by rerouting the inlet flow pipe through instead of into the flow plate (see Figure 5.1 point A). The outlet would have to be developed in a similar way forcing the placement of the outlet to be moved away from the center of the plate to the bottom (see Figure 5.1 point B). Since the placement of 77  the outlet is no longer centered the flow and pressure distribution will no longer be perfectly symmetrical. However, the effects of the non-symmetric design can be offset slightly as long as the pressure drop in the outlet is significantly less than the pressure drop across the flow control rings. This new flow field design can easily  Figure 5.1: Future radial flow field design be stamped using aluminum or Grafcell R , i.e. a flexible graphite material. This design also has the advantage of locating all of the water outlets at the bottom of the flow field allowing gravity to drain any excess water droplets out of the cell and expel them efficiently. The radial flow field design also needs to be optimized by adjusting the control rings and notches for a minimal pressure drop while still providing enhanced water management and performance. This system also needs to be optimized with regards to humidity requirements on the cathode side of the fuel  78  cell. The performance of the flow field using a dry cathode reactant flow showed the potential for this system to provide high current density values but was hampered by large ohmic losses. When the cathode reactant flow was fully humidified the limiting current density, on the other hand, was much smaller but the problem of the ohmic loss was alleviated. As a future study, an optimal humidity ratio, providing enough water to keep the membrane hydrated while still allowing the flow field to reach a better limiting current density, must be identified.  79  Bibliography [1] F. Barbir, PEM Fuel Cells: Theory and Practice, 1st ed. Academic Press, 2005. → pages 1, 2, 3, 5, 6, 7, 13, 45  New York:  [2] Z. Q. Ma, P. Cheng, and T. S. Zhao, “A palladium-alloy deposited nafion membrane for direct methanol fuel cells,” Journal of Membrane Science, vol. 215, no. 12, pp. 327–336, 4/15 2003. → pages 3 [3] E. T. S. Inc, Fuel Cell Handbook, 7th ed. Morgantown, West Virginia: U.S. Department of Energy, 2004. → pages 6, 7 [4] N. Akhtar, A. Qureshi, J. Scholta, C. Hartnig, M. Messerschmidt, and W. Lehnert, “Investigation of water droplet kinetics and optimization of channel geometry for pem fuel cell cathodes,” International Journal of Hydrogen Energy, vol. 34, no. 7, pp. 3104 – 3111, 2009. [Online]. Available: → pages 7 [5] H. Kim, J. H. Nam, D. Shin, T.-Y. Chung, and Y.-G. Kim, “A numerical study on liquid water exhaust capabilities of flow channels in polymer electrolyte membrane fuel cells,” Current Applied Physics, vol. 10, no. 2, Supplement 1, pp. S91–S96, 3 2010. → pages 8, 9, 13, 44, 45 [6] G. He, Y. Yamazaki, and A. Abudula, “A three-dimensional analysis of the effect of anisotropic gas diffusion layer(gdl) thermal conductivity on the heat transfer and two-phase behavior in a proton exchange membrane fuel cell(pemfc),” Journal of Power Sources, vol. 195, no. 6, pp. 1551–1560, 3/15 2010. → pages [7] L. Wang and H. Liu, “Performance studies of pem fuel cells with interdigitated flow fields,” Journal of Power Sources, vol. 134, no. 2, pp. 185 – 196, 2004. [Online]. Available: → pages 8  80  [8] J. Choi and G. Son, “Numerical study of droplet dynamics in a pemfc gas channel with multiple pores,” Journal of Mechanical Science and Technology, vol. 23, no. 7, pp. 1765 – 1772, 2009. [Online]. Available: → pages 8, 9 [9] A. Golpaygan and N. Ashgriz, “Effects of oxidant fluid properties on the mobility of water droplets in the channels of pem fuel cell,” International Journal of Energy Research, vol. 29, no. 12, pp. 1027 – 1040, 2005. [Online]. Available: → pages 9 [10] G. He, P. Ming, Z. Zhao, A. Abudula, and Y. Xiao, “A two-fluid model for two-phase flow in pemfcs,” Journal of Power Sources, vol. 163, no. 2, pp. 864 – 873, 2007. [Online]. Available: → pages [11] C. Fang, C. Hidrovo, F.-m. Wang, J. Eaton, and K. Goodson, “3-d numerical simulation of contact angle hysteresis for microscale two phase flow,” International Journal of Multiphase Flow, vol. 34, no. 7, pp. 690 – 705, 2008. [Online]. Available: → pages 8 [12] C. Siegel, “Two-dimensional, non-isothermal, two-phase flow inside the gas diffusion layer unit of the polymer-electrolyte-membrane fuel cell,” Proceedings of the second COMSOL user conference, 2006. → pages vii, 9, 10, 18, 20, 22, 23, 25, 31, 32, 74 [13] T. Berning, D. M. Lu, and N. Djilali, “Three-dimensional computational analysis of transport phenomena in a pem fuel cell,” Journal of Power Sources, vol. 106, no. 12, pp. 284–294, 4/1 2002. → pages [14] A. A. Kulikovsky, J. Divisek, and A. A. Kornyshev, “Modeling the cathode compartment of polymer electrolyte fuel cells: Dead and active reaction zones,” Journal of The Electrochemical Society, vol. 146, no. 11, pp. 3981–3991, 1999. [Online]. Available: → pages [15] W. M. Yan, C. Y. Soong, F. Chen, and H. S. Chu, “Effects of flow distributor geometry and diffusion layer porosity on reactant gas transport and performance of proton exchange membrane fuel cells,” Journal of Power Sources, vol. 125, no. 1, pp. 27–39, 1/2 2004. → pages 9 [16] J. T. Gostick, M. A. Ioannidis, M. W. Fowler, and M. D. Pritzker, “Pore network modeling of fibrous gas diffusion layers for polymer electrolyte 81  membrane fuel cells,” Journal of Power Sources, vol. 173, no. 1, pp. 277 – 290, 2007. [Online]. Available: → pages 9, 10 [17] J. Hinebaugh, Z. Fishman, and A. Bazylak, “Unstructured pore network modeling with heterogeneous pemfc gdl porosity distributions,” Journal of the Electrochemical Society, vol. 157, no. 11, pp. B1651 – B1657, 2010. [Online]. Available: → pages 9 [18] Z. Lu, C. Rath, G. Zhang, and S. G. Kandlikar, “Water management studies in pem fuel cells, part iv: Effects of channel surface wettability, geometry and orientation on the two-phase flow in parallel gas channels,” International Journal of Hydrogen Energy, vol. 36, no. 16, pp. 9864–9875, 8 2011. → pages 10, 13, 14 [19] M. Secanell, J. Wishart, and P. Dobson, “Computational design and optimization of fuel cells and fuel cell systems: A review,” Journal of Power Sources, vol. 196, no. 8, pp. 3690–3704, 4/15 2011. → pages 11 [20] M. Grujicic and K. M. Chittajallu, “Design and optimization of polymer electrolyte membrane (pem) fuel cells,” Applied Surface Science, vol. 227, no. 1-4, pp. 56–72, 4/15 2004. → pages 11 [21] M. Grujicic, C. L. Zhao, K. M. Chittajallu, and J. M. Ochterbeck, “Cathode and interdigitated air distributor geometry optimization in polymer electrolyte membrane (pem) fuel cells,” Materials Science and Engineering: B, vol. 108, no. 3, pp. 241–252, 5/15 2004. → pages [22] M. Grujicic and K. M. Chittajallu, “Optimization of the cathode geometry in polymer electrolyte membrane (pem) fuel cells,” Chemical Engineering Science, vol. 59, no. 24, pp. 5883–5895, 12 2004. → pages 11 [23] C.-H. Huang and J.-W. Lin, “Optimal gas channel shape design for a serpentine pemfc: Theoretical and experimental studies,” Journal of The Electrochemical Society, vol. 156, no. 1, pp. B178–B187, 2009. [Online]. Available: → pages 11 [24] X.-D. Wang, Y.-X. Huang, C.-H. Cheng, J.-Y. Jang, D.-J. Lee, W.-M. Yan, and A. Su, “Flow field optimization for proton exchange membrane fuel cells with varying channel heights and widths,” Electrochimica Acta, vol. 54, no. 23, pp. 5522–5530, 9/30 2009. → pages 11  82  [25] K. T. Adjemian, S. J. Lee, S. Srinivasan, J. Benziger, and A. B. Bocarsly, “Silicon oxide nafion composite membranes for proton-exchange membrane fuel cell operation at 80-140[degree]c,” Journal of The Electrochemical Society, vol. 149, no. 3, pp. A256–A261, 2002. [Online]. Available: → pages 11 [26] C. Yang, P. Costamagna, S. Srinivasan, J. Benziger, and A. B. Bocarsly, “Approaches and technical challenges to high temperature operation of proton exchange membrane fuel cells,” Journal of Power Sources, vol. 103, no. 1, pp. 1–9, 12/30 2001. → pages [27] S. Gamburzev and A. J. Appleby, “Recent progress in performance improvement of the proton exchange membrane fuel cell (pemfc),” Journal of Power Sources, vol. 107, no. 1, pp. 5–12, 4/20 2002. → pages 11 [28] J. M. Nail, G. Anderson, G. Ceaser, and C. J. Hansen, “The evolution of the pem stationary fuel cell in the u.s. innovation system,” U.S. national innovation system, 7/31 2000. → pages 11 [29] S. Litster and G. McLean, “Pem fuel cell electrodes,” Journal of Power Sources, vol. 130, no. 1-2, pp. 61–76, 5/3 2004. → pages 11 [30] D. J. Brett and N. P. Brandon, “Review of materials and characterization methods for polymer electrolyte fuel cell flow-field plates,” Journal of Fuel Cell Science and Technology, vol. 4, no. 1, pp. 29 – 44, 2007. [Online]. Available: → pages 12 [31] H.-C. Kuan, C.-C. M. Ma, K. H. Chen, and S.-M. Chen, “Preparation, electrical, mechanical and thermal properties of composite bipolar plate for a fuel cell,” Journal of Power Sources, vol. 134, no. 1, pp. 7–17, 7/12 2004. → pages [32] D. J. Brett and N. P. Brandon, “Review of materials and characterization methods for polymer electrolyte fuel cell flow-field plates,” Journal of Fuel Cell Science and Technology, vol. 4, no. 1, pp. 29 – 44, 2007. [Online]. Available: → pages 12, 13, 45 [33] D. Davies, P. Adcock, M. Turpin, and S. Rowen, “Bipolar plate materials for solid polymer fuel cells,” Journal of Applied Electrochemistry, vol. 30, pp. 101–105, 2000, 10.1023/A:1003831406406. [Online]. Available: → pages 12, 45  83  [34] J. P. Owejan, T. A. Trabold, D. L. Jacobson, M. Arif, and S. G. Kandlikar, “Effects of flow field and diffusion layer properties on water accumulation in a pem fuel cell,” International Journal of Hydrogen Energy, vol. 32, no. 17, pp. 4489–4502, 12 2007. → pages 13 [35] T. Tomimura, M. Shoyama, S. Mizutani, and R. Marubayashi, “Effect of hydrophilic properties of gas channel on freezing properties of pem fuel cells,” vol. 26, no. 1, Palm Springs, CA, United states, 2010, pp. 215 – 218. [Online]. Available: → pages 13 [36] F. Y. Zhang, X. G. Yang, and C. Y. Wang, “Liquid water removal from a polymer electrolyte fuel cell,” Journal of The Electrochemical Society, vol. 153, no. 2, pp. A225–A232, 2006. [Online]. Available: → pages 13 [37] A. Hakenjos, H. Muenter, U. Wittstadt, and C. Hebling, “A pem fuel cell for combined measurement of current and temperature distribution, and flow field flooding,” Journal of Power Sources, vol. 131, no. 12, pp. 213–216, 5/14 2004. → pages 14 [38] X.-D. Wang, Y.-Y. Duan, W.-M. Yan, and X.-F. Peng, “Effects of flow channel geometry on cell performance for pem fuel cells with parallel and interdigitated flow fields,” Electrochimica Acta, vol. 53, no. 16, pp. 5334–5343, 6/30 2008. → pages [39] S.-W. Cha, R. O’Hayre, S. J. Lee, Y. Saito, and F. B. Prinz, “Geometric scale effect of flow channels on performance of fuel cells,” Journal of The Electrochemical Society, vol. 151, no. 11, pp. A1856–A1864, 2004. [Online]. Available: → pages 14 [40] X. Li and I. Sabir, “Review of bipolar plates in pem fuel cells: Flow-field designs,” International Journal of Hydrogen Energy, vol. 30, no. 4, pp. 359–371, 3 2005. → pages 15, 17, 18, 46, 71 [41] N. S. Siefert and S. Litster, “Voltage loss and fluctuation in proton exchange membrane fuel cells: The role of cathode channel plurality and air stoichiometric ratio,” Journal of Power Sources, vol. 196, no. 4, pp. 1948–1954, 2/15 2011. → pages 15 [42] X.-D. Wang, Y.-Y. Duan, W.-M. Yan, and X.-F. Peng, “Local transport phenomena and cell performance of pem fuel cells with various serpentine flow field designs,” Journal of Power Sources, vol. 175, no. 1, pp. 397 – 407, 84  2008. [Online]. Available: → pages 16, 46, 71 [43] M. Ji and Z. Wei, “A review of water management in polymer electrolyte membrane fuel cells,” Energies, vol. 2, no. 4, pp. 1057 – 1106, 2009. [Online]. Available: → pages 17 [44] R. I. Perez, G. A. Hernandez, B. F. Elizalde, R. D. Juarez, and H. I. Almanza, “3d analysis of a new radial channel for pemfcs and comparison with a traditional channeled system,” Proceedings of the ASME 2010 IMECE, 2010. → pages 17, 38, 39, 42 [45] A. Weber and J. Newman, “Modeling gas-phase flow in porous media,” International Communications in Heat and Mass Transfer, vol. 32, no. 7, pp. 855–860, 2005. [Online]. Available: → pages 20 [46] R. B. Bird, “Transport phenomena,” Applied Mechanics Reviews, vol. 55, no. 1, pp. R1–R4, 2002. [Online]. Available: → pages [47] L. Monchick, R. J. Munn, and E. A. Mason, “Thermal diffusion in polyatomic gases: A generalized stefan—maxwell diffusion equation,” The Journal of Chemical Physics, vol. 45, no. 8, pp. 3051–3058, 1966. [Online]. Available: → pages 20 [48] J. T. Gostick, M. W. Fowler, M. A. Ioannidis, M. D. Pritzker, Y. M. Volfkovich, and A. Sakars, “Capillary pressure and hydrophilic porosity in gas diffusion layers for polymer electrolyte fuel cells,” Journal of Power Sources, vol. 156, no. 2, pp. 375–387, 6/1 2006. → pages 26 [49] COMSOL Multiphysics Chemical Engineering Module User’s Guide, 3rd ed., COMSOL, October 2007. → pages 27 [50] J. N. K. Vladimir Guraru, Frano Barbir, “Fuel cell collector plates with improved mass transfer channels,” Patent US 6 551 736, 04 22, 2003. [Online]. Available: USPAT6551736&id=ktcNAAAAEBAJ&oi=fnd&dq=Gurau+V,+et+al.+US+ Patent+No.+6,551,736,+2003&printsec=abstract#v=onepage&q&f=false → pages 38, 39, 42 [51] A. Woodman, E. Anderson, K. Jayne, and M. Kimble, “Development of corrosion-resistant coatings for fuel cell bipolar plates,” Proceedings of the AESF Annual Technical Conference, pp. 717 – 725, 1999. → pages 45 85  [52] D. Spernjak, A. K. Prasad, and S. G. Advani, “In situ comparison of water content and dynamics in parallel, single-serpentine, and interdigitated flow fields of polymer electrolyte membrane fuel cells,” Journal of Power Sources, vol. 195, no. 11, pp. 3553 – 3568, 2010. [Online]. Available: → pages 71  86  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items