THE DEVELOPMENT OF OPTICAL MEASUREMENT TECHNIQUES FOR GAS SPECIES AND SURFACE TEMPERATURE ON A PLANAR SOFC METHANE-STEAM REFORMER by James Edward Appleby Saunders A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Mechanical Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2011 © James Edward Appleby Saunders, 2011 ii Abstract This thesis presents the development of an experimental apparatus and methods to allow the application of gaseous Raman spectroscopy to the challenging and original application of a small-scale, high-temperature methane/steam reformer developed to be representative of the technologies used in solid oxide fuel cell (SOFC) applications. The research is placed in the context of global energy trends and SOFC’s, with specific reference to the challenges related to directly internally reforming medium-temperature SOFC’s and the case for the development of non-intrusive measurement techniques for gas species and temperature is made. The practical aspects of the development of the broadband 308 nm Raman system are examined and previous works in this area are highlighted. The excitation light source is evaluated, the use of a liquid potassium hydrogen phthalate filter as a means to reduce Rayleigh line effects is demonstrated, and background fluorescence suppression through polarization of the 308 nm light source is presented. The arrangements of the experimental set-up, gas supply, metering, and humidification are shown, as are the optical arrangements for laser sheet formation and light collection. A description of the calibration experiments, procedures, and methodologies that are used to define the normalised relative differential Raman scattering cross sections of the major species of interest in this study is presented. The observation of an unexpected leakage of air into the reformer is described and a hypothesis is presented to explain the ingress of air. Finally, results are presented that describe the response of the optically-accessed reformer to variations in; operating temperature, humidification factor, total volume flow rate, methane volume flow rate, and the methane residency time within the reformer channel. From these results it was possible to conclude that increased reformer temperature increased reaction rate, increased gas residency time in the channel increased hydrogen production, and iii reactant streams with higher inlet mole fractions of methane resulting in increased reaction rates and amounts of hydrogen production. The performance of the reformer rig and the suitability of optical diagnostic techniques to the application of a SOFC scale reformer are discussed. iv Preface A version of Chapter 4 has been published. Saunders, J. E. A. and Davy, M. H.  Broadband 308 nm Vibrational Raman Spectroscopy Using a Potassium Hydrogen Phthalate (KHP) Liquid Filter and Polarization Fluorescence Suppression. Review of Scientific Instruments, vol. 81, 013108. I conducted all apparatus design, implementation, data collection and analysis with supervision and editing from Dr. M. Davy. v Table of Contents Abstract............................................................................................................................ ii Preface ............................................................................................................................ iv Table of Contents............................................................................................................v List of Tables ..................................................................................................................xi List of Figures ...............................................................................................................xii Nomenclature ..............................................................................................................xxii Abbreviations .............................................................................................................xxiv Acknowledgements ....................................................................................................xxv Dedication...................................................................................................................xxvi 1 Steam Reformation, Reforming Solid Oxide Fuel Cells and Existing Optical Diagnostic Techniques...................................................................................................1 1.1 Steam Reformation ............................................................................................1 1.2 Steam Reformation for Medium and High Temperature Fuel Cells....................2 1.2.1 External Reformation.................................................................................4 1.2.2 Indirect Internal Reformation .....................................................................6 1.2.3 Direct Internal Reformation........................................................................8 1.3 The Monitoring of SOFC Phenomena Using Optical Diagnostics ....................11 1.4 Gaseous Raman Spectroscopy and IR Thermometry......................................13 1.5 Methane/Steam Reformer Proof of Concept ....................................................14 1.6 The Aims of this Thesis ....................................................................................15 1.7 Thesis Structure ...............................................................................................15 vi 2 Raman Spectroscopy and Radiation Thermometry..........................................17 2.1 The Electromagnetic Spectrum and Raman Spectroscopy..............................17 2.2 Molecular (Internal) Energy Levels...................................................................18 2.3 Photon Molecule Interaction.............................................................................19 2.4 The Raman Spectrum ......................................................................................25 2.5 Making Raman Measurements ........................................................................29 2.6 Raman Signal Intensity ....................................................................................32 2.7 Signal-to-Noise Ratio (SNR) ............................................................................34 2.8 Radiation Thermometry Applied to the Surface of a SOFC Cell ......................37 2.9 Radiation Models..............................................................................................39 2.10 Surface Roughness..........................................................................................41 2.11 Directional Emissivity .......................................................................................42 2.12 Effects of Ambient Temperature, Pressure and Gas Composition...................44 2.13 Application of IR Thermometry to Optically Accessible Reformer ....................47 3 Optical Reformer Design .....................................................................................49 3.1 The Design Task ..............................................................................................50 3.2 Raman Spectroscopy Design Considerations..................................................50 3.3 IR Thermometry Design Considerations ..........................................................55 3.4 Optical Requirements of the Reforming Test Section ......................................56 3.5 Practical Requirements of the Reforming Test Section....................................58 3.6 Optically Accessible Heater..............................................................................61 3.7 Gas Flow in the Optically Accessible Reformer................................................67 3.8 Optical Accessibility of the Reformer Channel .................................................74 3.9 Optical Accessibility of the Heater ....................................................................79 3.10 Optically Accessible Reformer Channel Support .............................................81 vii 3.11 Apparatus Modifications ...................................................................................85 3.11.1 Optical Window Reactivity .......................................................................86 3.11.2 Top Window Seating Force .....................................................................88 4 Preliminary Raman Spectroscopy Set Up..........................................................92 4.1 Introduction.......................................................................................................92 4.2 Broadband 308 nm Vibrational Raman Spectroscopy Using a Potassium Hydrogen Phthalate (KHP) Liquid Filter and Polarization Fluorescence Suppression .....................................................................................................93 4.2.1 Introduction..............................................................................................94 4.2.2 Experiment ..............................................................................................97 4.2.3 Results and Discussion .........................................................................101 4.2.4 Conclusions ...........................................................................................109 4.2.5 Acknowledgements ...............................................................................110 4.3 Durability of Laser Beam Polarizer .................................................................110 4.4 Glan Polarizer.................................................................................................113 5 Description of Experimental Apparatus...........................................................116 5.1 Gas Supply, Metering and Treatment ............................................................116 5.2 Excitation Source, Beam Polarization and Light Sheet Formation.................122 5.3 Light Sheet Optical Setup...............................................................................128 5.4 Raman Signal Collection, Low Pass Filtration, Spectrometer and Image Capture...........................................................................................................133 5.5 Spectrometer and Intensified Charge Coupled Device (ICCD) ......................137 6 Experimental Test Matrix and Description of Data Collection Methodology 140 6.1 Test Matrix Design .........................................................................................140 viii 6.2 The Experimental Limitations of Gas Humidification ......................................141 6.3 Efficiency of Data Collection...........................................................................143 6.4 Range of Experimental Variables...................................................................145 6.5 Experimental Procedure.................................................................................147 6.6 Execution of Experimental Matrix ...................................................................149 6.7 Number of Images Collected..........................................................................150 7 Data Processing, Results and Discussion.......................................................153 7.1 Image Processing...........................................................................................153 7.2 Review of Raman Theory...............................................................................157 7.3 Relative Normalized Differential Raman Scattering Cross Sections ..............159 7.4 Raman Line Summing....................................................................................163 7.5 Raman Signal Calibration...............................................................................165 7.6 Differential Raman Scattering Cross Section for Water .................................172 7.7 Ingress of Air ..................................................................................................184 7.8 Molecular Flow ...............................................................................................193 7.9 Relative Normalized Differential Raman Scattering Cross Section for Methane .......................................................................................................................197 7.10 Data Processing and Quantification of Image to Image Variation ..................201 7.11 Spatial Variation in Mole Fraction Perpendicular to Anode Surface...............204 7.12 Calibration Data..............................................................................................206 7.13 The Effects of Temperature............................................................................211 7.14 The Effects of Humidification Factor ..............................................................221 7.15 The Effects of Total Volume Flow Rate..........................................................233 7.16 The Effect of Methane Volume Flow Rate......................................................247 7.17 Infrared Radiation (IR) Thermometry Results.................................................254 ix 7.18 Experimental Results Summary .....................................................................260 8 Summary, Claims of Originality and Future Work ..........................................263 8.1 Summary ........................................................................................................263 8.2 Claims for Originality ......................................................................................266 8.3 Recommendations for Future Work................................................................267 8.3.1 Apparatus in Current Configuration .......................................................267 8.3.2 Future Work with Modified Apparatus ...................................................268 References...................................................................................................................271 Appendices..................................................................................................................280 A Optical Reformer Flow Rate and Mean Stream Velocity Calculations ..........280 A.1 SOFC Fuel Consumption ...............................................................................280 A.2 Mean Stream Velocities .................................................................................282 A.3 Reynolds Number Calculations ......................................................................283 B FLUENT Simulation............................................................................................285 B.1 Reformer Inlet /Outlet and Channel Configurations .......................................285 B.2 FLUENT Results ............................................................................................287 B.3 Conclusions from FLUENT Simulations .........................................................289 C Molar Absorption Calculation ...........................................................................290 C.1 Beer-Lambert Law..........................................................................................290 C.2 Molar Absorptivity Calculation ........................................................................290 D Experimental Schedule......................................................................................293 D.1 Chronological Experimental Schedule ...........................................................293 D.2 Experimental Test Points ...............................................................................294 x D.3 Rotameter Set Points .....................................................................................295 D.4 Calibration Rotameter Set Points ...................................................................295 E Molar Flow Conductance...................................................................................296 E.1 Conductance of an Aperture ..........................................................................296 F Approximation of the Magnitude of Air in the Channel ..................................299 G Polarizer and Image Setup Procedures............................................................307 G.1 Polarizer Setup...............................................................................................307 G.2 Image Setup ...................................................................................................311 H Description of Experimental Apparatus Operation.........................................316 H.1 Anode Reduction............................................................................................316 H.2 Method of Experimental Apparatus Operation ...............................................317 H.3 Polarizer and Image Set Up ...........................................................................320 H.4 Data Collection ...............................................................................................321 H.5 Heater Cool Down ..........................................................................................325 I Raman Cross Section Description ...................................................................326 xi List of Tables Table 2.1 Vibrational frequencies, Stokes spectral positions of Q branches and normalized vibrational cross sections for the gaseous species involved in the steam methane reformation reactions produced by a 308nm excitation source. ...............29 Table 6.1 Experimental test matrix................................................................................147 Table 7.1. Variations in published relative normalized differential Raman scattering cross sections. † Using Q-branch N2 normalization, * The third vibrational frequency of methane is separated from the first by 100 cm-1....................................................160 Table 7.2 Metered channel inlet mole fractions. ...........................................................198 Table 7.3 Metered CH4, N2, and H2O mole fractions vs. measured mole fractions and associated 5 errors at 600 and 700 °C. .................................................................213 Table D.1 Chronological experimental schedule...........................................................293 Table D.2 Experimental test points. ..............................................................................294 Table D.3 Rotameter set points ....................................................................................295 Table D.4 Calibration rotameter set points....................................................................295 xii List of Figures Figure 1.1 Fuel cell schematic with anode, cathode and overall FC reaction. ..................3 Figure 1.2 Schematic of external SOFC steam reformation arrangement. .......................5 Figure 1.3 Generic diagram of a cell with fuel and air channels. ......................................6 Figure 1.4 Generic diagram of a cell stack highlighting cell and interconnect interaction. 7 Figure 1.5 Schematic of indirect internal reforming (IIR)...................................................7 Figure 1.6 Schematic (not to scale) of the reformation and electrochemically active regions of a methane fuelled SOFC...........................................................................9 Figure 2.1 A selection of possible photon/molecule absorptions and emissions. Adapted from Zhao et al. [Zhao et al. 2001]...........................................................................21 Figure 2.2 An example of spectral lines of a diatomic molecule produced through Raman scattering [Zhao et al., 2001] ...................................................................................26 Figure 2.3 Photograph showing cross section of SOFC anode supported cell from presentation given by B. Borglum, Versa Power Systems: “Development of Solid Oxide Fuel Cells at Versa Power Systems”, 2009 Fuel Cell Seminar, 17 November 2009, Palm Springs, CA...........................................................................................38 Figure 2.4 Schematic showing the pathway of radiation focused onto the detector head from the surface of interest. .....................................................................................45 Figure 3.1 Coaxial and orthogonal light in arrangements relative to the axis of the collection optics, with ray paths shown for the region of interest. ............................51 Figure 3.2 Optical arrangement for one dimensional Raman spectroscopy. ..................53 Figure 3.3 IR/primary Raman optical axis shown with potential secondary Raman axis in relation to anode surface and arbitrary gas flow direction. ......................................57 Figure 3.4 Schematic of horizontal and vertical clam shell heater arrangements...........62 xiii Figure 3.5 Spacing of semicircular ceramic radiant heaters to allow second optical axis .................................................................................................................................64 Figure 3.6 Schematic of optical axes orientations achieved through spaced semicircular radiant heaters .........................................................................................................66 Figure 3.7 Schematic of predominant inlet and outlet velocity directions for channel with a single inlet and outlet ............................................................................................68 Figure 3.8 Reformer channel with plenum chambers and inlet/outlet passages.............70 Figure 3.9 Schematic quasi fully developed flow as a product of multiple inlets and outlets outlet.............................................................................................................71 Figure 3.10 Partial cut away of optically accessible heater showing and resultant spatial restrictions for gas tube routing................................................................................73 Figures 3.11a and b. (a) Cutaway of heater showing gas routing (b) The heating element arrangement. Note that the plenum plugs had not yet been fitted and the third middle tube was that through which the control thermocouples were run. ..............74 Figure 3.12 Approximate locations of potential window seating areas ...........................76 Figure 3.13 Cross section of channel showing face sealing surface and obstruction of collection optic .........................................................................................................77 Figure 3.14 Cross section of channel showing alternate sealing layout with external seating force necessary for window perimeter seal .................................................78 Figures 3.15a and b. (a) A stainless steel spring to provide a lateral seating force to the top window. (b) Sprung top window fasteners. Note the first iteration front window attachment system using threaded fasteners. This was replaced with a sprung system prior to testing..............................................................................................79 Figure 3.16 Optical heater top window assembly spacing the semi-cylindrical heating elements ..................................................................................................................80 xiv Figure 3.17 Front window assembly and showing the optical accessibility afforded to the reformer channel......................................................................................................81 Figure 3.18 Schematic of reformer channel support design intent..................................82 Figure 3.19 Optically accessible reformer channel rail assembly with support structure 83 Figure 3.20 Reformer rail assembly after being removed from heater and sitting on a servicing stand. Note the integrated front access panel, front window and rail support, gas piping and insulation. The right hand support is equivalent to that used to support the rail when situated in the heater. ........................................................85 Figures 3.21a and b. (a) Residue deposited onto the surface of the window. (b) Milky discolouration around edge of window.....................................................................86 Figure 3.22 Photograph illustrating the following: 1. ceramic window gasket. 2. fully reduced anode sample as indicated by uniform grey colouration. 3. updated seating force arrangement (see below §3.11.2). ..................................................................87 Figure 3.23 Photograph showing the partial oxidation of the anode surface ..................88 Figure 3.24 Schematic of the use of an external mass to provide the reformer top window seating force. ..............................................................................................90 Figure 3.25 Suspended mass and pivot system to produce window seating force. The mass consisted of a block of mild steel weighing approximately 300 grams. ..........91 Figure 4.1 Setup. Experimental setup of laser, optics and detection system. Reprinted with permission of the American Institute of Physics. ..............................................98 Figure 4.2 Channel profile. Test channel showing anode surface in relation to excitation beam and optical windows. Reprinted with permission of the American Institute of Physics.....................................................................................................................99 Figure 4.3 Light sheet. Light sheet and beam waist in relation to anode surface. Reprinted with permission of the American Institute of Physics. ...........................100 xv Figure 4.4 Raman signal of air at 21 ˚C, 101 kPa: (a) unfiltered (b) WG320 glass filter, (c) KHP liquid filter. Note: the horizontal feature present at the top of image (c) is scattered light from the surface of the anode which was not fully eliminated from entering the spectrometer when masking the entrance slit. Reprinted with permission of the American Institute of Physics. ...................................................103 Figure 4.5 Filter comparison. Comparison of unfiltered and filtered Raman spectra of air (O2:324 nm, N2:332 nm) at 21 ˚C, 101 kPa - Top: six laser shots per image, spectrum produced by averaging over ten images and ten pixel rows. Bottom: 300 laser shots per image, spectrum produced by averaging over ten images and ten pixel rows. Reprinted with permission of the American Institute of Physics. .........104 Figure 4.6 Raw Raman spectra. Spectral curve produced from averaging ten images, each consisting of 300 laser shots. Reprinted with permission of the American Institute of Physics. ................................................................................................106 Figure 4.7 Polarization suppression. Effects of fluorescence suppression on spectral curve. Reprinted with permission of the American Institute of Physics..................109 Figure 4.8. Melles Griot 03 PBS 117 polarizer damage................................................111 Figure 4.9 Newport 10SC16PC.23 polarizer damage...................................................112 Figure 4.10 Comparison of spectra collected using excitation sources of opposing polarities at 21 ˚C, 101 kPa - 300 laser shots per image, spectrum produced by averaging over 20 images and seven pixel rows. ..................................................114 Figure 5.1 Schematic of gas routing, metering, humidification and post reformation treatment................................................................................................................117 Figure 5.2 Methane, hydrogen and carbon monoxide gas sensor and rotameters....... 118 Figure 5.3 Lab scale humidifier .....................................................................................119 Figure 5.4 Exhaust gas treatment .................................................................................122 Figure 5.5 Diagram of the laser beam steering, polarization and shaping.................... 124 xvi Figure 5.6 Beam masks situated immediately after the periscope (R) and polarizer mask (L)...........................................................................................................................125 Figure 5.7 Polarizer masking element...........................................................................126 Figure 5.8 Beam reduction sheet slit.............................................................................127 Figure 5.9 Optics mounting frame surrounding the optical heater ................................130 Figure 5.10 Light sheet optical traverse ........................................................................131 Figure 5.11 (a) Worm screw arrangement (b) scale indicating relative position of traverse ...............................................................................................................................132 Figure 5.12 Schematic of light collection system ..........................................................133 Figure 5.13 Ray tracing diagram for the collection optics. ............................................135 Figure 6.1 Schematic showing the positions physical data collection points on the non- operating anode half cell........................................................................................144 Figure 6.2 Background subtracted hydrogen portion line counts vs. number of images averaged. At position “55 mm” along anode surface, gas mixture 50/50, H2/N2. .152 Figure 7.1 Un-cropped, 30 image averaged Raman spectra detailing additional light sources from anode surface (top) and sapphire window (bottom).........................153 Figure 7.2 Un-cropped, 30 image averaged Raman spectra with numbered portions..154 Figure 7.3 Typical wavelength vs. a raw Raman signal (from portion 3), a background approximation and a processed Raman signal......................................................155 Figures 7.4a and b. Graphs of typical average pixel counts collected from within the channel at identically metered hydrogen and nitrogen flow rates illustrating decreasing pixel count with increasing longitudinal channel position. 50/50 H2/N2 blend at 600 ºC. .....................................................................................................168 Figure 7.5. Reduced apparent solid angle of the collection optics due to the Raman interrogation volume traversing to the rear of the reformer channel. .....................169 xvii Figures 7.6a and b. Graphs of portions 2, 3, and 4 and calculated average for hydrogen and nitrogen. 50/50 H2/N2 blend at 600 ºC.............................................................171 Figure 7.7 Averaged species counts vs. longitudinal channel position. 100% hydrogen calibration experiments (600 ºC, average signal counts from portions 2,3, and 4). ...............................................................................................................................173 Figure 7.8 Un-calibrated H2O/N2 ratios from 100% hydrogen calibration experiments (average signal counts from portions 2,3, and 4)...................................................175 Figures 7.9a and b. Average relative normalized differential Raman scattering cross sections of water at 100% hydrogen operating condition ......................................178 Figures 7.10a and b. Nitrogen and hydrogen proportionality curves. ...........................181 Figure 7.11a and 8.11b. Average relative normalized differential Raman scattering cross sections of hydrogen at 600 and 700 °C respectively............................................183 Figure 7.12 Oxidation and warping of the anode material when exposed to a partially oxidising environment. The direction of gas flow was from left to right. The “V” formation would suggest that a fully developed flow profile was present in the channel. .................................................................................................................186 Figures 7.13a and b. Water mole fraction vs. channel position for the four calibration mixtures (data fitted with 2nd order polynomial to illustrate approximate trends)....188 Figure 7.14 Graph of species mole fractions calculated using the methane and hydrogen COx formation models. Note the difference in COx values between the hydrogen and methane models as well as the relative insensitivities of the other species to this difference.........................................................................................................200 Figure 7.15 Plot of mole fraction vs. longitudinal channel position at the operating condition of test point four for the three middle portions of the channel. ............... 205 Figure 7.16 Calibration data: 80/20 nitrogen/hydrogen mixture, 600 and 700 °C. ........207 Figure 7.17 50/50 Calibration data: nitrogen/hydrogen mixture, 600 and 700 °C. ........208 xviii Figure 7.18 Calibration data: 80/20 nitrogen/hydrogen mixture, 600 and 700 °C. ........208 Figure 7.19 0/100 Calibration data: 0/100 hydrogen mixture, 600 and 700 °C. ............209 Figure 7.20 Effects of temperature (test point 1, 600 and 700 °C)................................215 Figure 7.21 Effects of temperature (test point 2, 600 and 700 °C)................................215 Figure 7.22 Effects of temperature (test point 3, 600 and 700 °C)................................216 Figure 7.23 Effects of temperature (test point 4, 600 and 700 °C)................................216 Figure 7.24 Effects of temperature (test point 5, 600 and 700 °C)................................217 Figure 7.25 Effects of temperature (test point 6, 600 and 700 °C)................................217 Figure 7.26 Effects of temperature (test point 7, 600 and 700 °C)................................218 Figure 7.27 Effects of temperature (test point 8, 600 and 700 °C)................................218 Figure 7.28 Percentage humidification of the bubble type humidifier vs. water bath temperature. The three set points of 68, 78 and 95 °C used for the experiments presented in this thesis are also shown. ................................................................220 Figure 7.29 Effects of humidification factor (test points three, commanded humidification factor of 2, and one, commanded humidification factor of 3 at 600 °C). ................223 Figure 7.30 Effects of humidification factor (test points three, commanded humidification factor of 2, and one, commanded humidification factor of 3 at 700 °C). ................224 Figure 7.31 Effects of humidification factor (test points seven, commanded humidification factor of 2, and eight, commanded humidification factor of 3 at 600 °C). ..............225 Figure 7.32 Effects of humidification factor (test points seven, commanded humidification factor of 2, and eight, commanded humidification factor of 3 at 700 °C). ..............226 Figure 7.33 Effects of humidification factor (test points two, commanded humidification factor of 2, and five, commanded humidification factor of 3 at 600 °C). ................228 Figure 7.34 Effects of humidification factor (test points two, commanded humidification factor of 2, and five, commanded humidification factor of 3 at 700 °C). ................229 xix Figure 7.35 Effects of humidification factor (test points four, commanded humidification factor of 2, and six, commanded humidification factor of 3 at 600 °C)...................230 Figure 7.36 Effects of humidification factor (test points four, commanded humidification factor of 2, and six, commanded humidification factor of 3 at 700 °C)...................231 Figure 7.37 Effects of total volume flow rate (test points four, total volume flow rate = 0.6 SLPM, and three, total volume flow rate = 1.0 SLPM, at 600 °C). .........................235 Figure 7.38 Effects of total volume flow rate (test points four, total volume flow rate = 0.6 SLPM, and three, total volume flow rate = 1.0 SLPM, at 700 °C). .........................236 Figure 7.39 Effects of total volume flow rate (test points six, total volume flow rate = 0.6 SLPM, and one, total volume flow rate = 1.0 SLPM, at 600 °C). .....................238 Figure 7.40 Effects of total volume flow rate (test points six, total volume flow rate = 0.6 SLPM, and one, total volume flow rate = 1.0 SLPM, at 700 °C). .....................239 Figure 7.41 Effects of total volume flow rate (test points seven, total volume flow rate = 0.6 SLPM, and two, total volume flow rate = 1.0 SLPM, at 600 °C). .....................240 Figure 7.42 Effects of total volume flow rate (test points seven, total volume flow rate = 0.6 SLPM, and two, total volume flow rate = 1.0 SLPM, at 700 °C). .....................241 Figure 7.43 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points four, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point two, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 600 °C).................243 Figure 7.44 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points four, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point two, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 700 °C).................244 Figure 7.45 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points six, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point five, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 600 °C).................245 xx Figure 7.46 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points six, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point five, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 700 °C).................246 Figure 7.47 Effect of methane volume flow rate (test points three, methane volume flow rate = 0.06 SLPM, and two, methane volume flow rate of 0.09 SLPM, at 600 °C).248 Figure 7.48 Effect of methane volume flow rate (test points three, methane volume flow rate = 0.06 SLPM, and two, methane volume flow rate of 0.09 SLPM, at 700 °C).249 Figure 7.49 Effect of methane volume flow rate (test points one, methane volume flow rate = 0.06 SLPM, and five, methane volume flow rate of 0.09 SLPM, at 600 °C).250 Figure 7.50 Effect of methane volume flow rate (test points one, methane volume flow rate = 0.06 SLPM, and five, methane volume flow rate of 0.09 SLPM, at 700 °C).251 Figure 7.51 Effect of methane volume flow rate (test points four, methane volume flow rate = 0.06 SLPM, and seven, methane volume flow rate of 0.09 SLPM, at 600 °C). ...............................................................................................................................253 Figure 7.52 Effect of methane volume flow rate (test points four, methane volume flow rate = 0.06 SLPM, and seven, methane volume flow rate of 0.09 SLPM, at 700 °C). ...............................................................................................................................254 Figure 7.53 IR temperature measurements (test points 1 – 8, 600 °C) ........................255 Figure 7.54 IR temperature measurements (test points 1 – 8, 700 °C) ........................256 Figure 7.55 Differences in IR temperature measurements between test points 1 – 8 and 100% H2 case (at heater command temperature of 600 °C)..................................258 Figure 7.56 Differences in IR temperature measurements between test points 1 – 8 and 100% H2 case (at heater command temperature of 700 °C)..................................259 Figure B.1 Diagram of overall channel dimensions and inlet / outlet ............................285 Figure B.2 Diagram showing station positions at which velocity profiles were plotted as well as the stepped reformer channel configuration ..............................................286 xxi Figure B.3 Plot of velocity vs. vertical position for full height channel...........................287 Figure B.4 Plot of velocity vs. vertical position for full height channel...........................287 Figure B.5 Plot of velocity vs. vertical position for a channel of height 2 mm with single inlet / outlet.............................................................................................................288 Figure B.6 Plot of velocity vs. vertical position for a channel of height .........................288 Figure F.1 De-absorption and air ingress plus hydrogen leak models..........................300 Figure F.2a and b. Air evolution rate for scenario A. 100% hydrogen calibration condition, 0.5 SLPM total flow rate. .......................................................................302 Figure F.3a and b. Air and hydrogen leak rate for scenario B. 100% Hydrogen calibration condition, 0.5 SLPM total flow rate. .......................................................................303 Figure G.1 Diagram of the laser beam polarization, shaping and steering ................... 309 Figure G.2 GLP6320 Glan laser polarizer showing light in/out ports ............................310 Figure G.3 The collection volume, as defined by the relative positions of the light collection optics and the spectrometer slit being filled by the light sheet............... 311 Figure G.4 Regions of signal collection.........................................................................313 Figure G.5 Light sheet and collection volume dimensions............................................314 Figure H.1 Diagram of apparatus gas routings with specific note given to the bypass valve and backflow prevention valve. ....................................................................323 xxii Nomenclature fueln& Molar fuel rate mol . s R~ Universal gas constant 8.31447 kJ . kmol -1 . K-1 m& Mass flow rate kg . s-1 V& Volume flow rate m3 . s U Mean stream velocity m . s-1 μ Dynamic viscosity kg . m-1 . s-1 λ Wavelength m α Spectral absorptivity ⎟⎠ ⎞⎜⎝ ⎛ Ω∂ ∂σ Vibrational cross section m2 A Area m2 a Spectral linear absorption b Black body co speed of light in a vacuum 2.9979 × 10-34 m . s-1 D Hydraulic diameter m E Spectral emissivity Energy J Electric field V. m-2 ƒ# F-number G Gibb’s free energy kJ . mol-1 . K-1 h Planck’s constant 6.6261 × 10-34 J . s H Enthalpy kJ . mol-1 . K-1 I Current density A . cm-2 I Irradiance W . m-2 Ir,i Grayscale intensity of ith species IΩ Radiant intensity W . sr-1 xxiii k Boltzmann’s constant 1.3087 × 10-23 J . K-1 L Spectral radiance W . m-2 . sr-1 . µm-1 L Radiance W . m-2 . sr-1 l Sampling extent m m Molar mass g . mol-1 M Radiant emission W . m-2 n Number of moles Ncell Number of cells ni Number density ith species mol . m-3 P Pressure Pa Wetted perimeter m p Induced dipole moment C . m Pi Incident light power W Pr,i Raman signal power ith species W Qi Quantum efficiency of camera r Spectral reflectance S Entropy kJ . K-1 T temperature K U Internal energy J α Polarizablity N . C-1 ε Emissivity η Efficiency ν Frequency Hz ρ Density kg . m-3 σ Stefan-Boltzmann constant 2.99792 ×108 J . m-2 . K-4 . s-1 Ω Solid angle sr xxiv Abbreviations CCD Charge-Coupled Device COx Oxides of carbon DIR Directly Internally Reforming EIS Electrical Impedance Spectroscopy FC Fuel Cell FTIR Fourier Transform Infra Red ICCD Intensified Charge-Coupled Device ICE Internal Combustion Engine ID Internal Diameter IIR Indirect Internally Reforming IR Infra Red KHP Potassium Hydrogen Phthalate LIF Laser Induced Fluorescence PEM Proton Exchange Membrane / Polymer Electrolyte Membrane RCF Refractory Ceramic Fibre ROI Region Of Interest RTD Resistance Temperature Detector SLPM Standard Litres Per Minute (298.15 K, 101.325 kPa) SNR Signal-to-Noise Ratio SOFC Solid Oxide Fuel cell TCE Thermal Coefficient of Expansion UF Utilization Factor UV Ultra Violet VRS Vibrational Raman Spectroscopy YAG Yttrium Aluminium Garnet xxv Acknowledgements I would like to thank the following for their financial support of this project; Advanced Systems Institute of British Columbia, the Canadian Foundation for Innovation, the British Columbia Knowledge Development Fund, NSERC, and Auto 21. I also wish to thank Versa Power Systems for their hardware donation and technical support, and in particular, Casey Brown and Scott Thompson. The project simply would not have been possible without the supervision of Dr. Martin Davy and the advice given to me by Dr. Kendal Bushe, particularly in the final stages of the research. In addition Dr. Steve Rogak has always provided advice, enthusiasm, and an alternative view throughout my work at UBC. I would also like to thank Dr. Gary Schajer for his encouragement and ‘long view’ which was invaluable at times of waning moral. The research and support staff in the department of Mechanical Engineering at UBC have been a pleasure to work and interact with, in particular Barbara Murray and Yuki Matsumura of the office, Markus Fengler and Roland Genshorek of the machine shop, and Glenn Jolly and Sean Buxton of the instrumentation department. I would also like to thank my fellow graduate students who I have worked with through the years; particular thanks go to Dave Gorby, Dave Williams, Malcolm Shield, Andrew Mezo, Eric Kastanis, Jean Logan, Mahdi Salehi, and Amir Aliabadi. All of who have given me advice and inspiration along the way. In closing I’d like to thank my family, which has tirelessly supported me throughout my education. I would also like to thank Mike Taylor for helping to give me a unique perspective of my research and life in Canada. Finally, I must thank my wife, without whose love and support this project would never have been completed; words cannot describe my gratitude to you for putting up with me during this undertaking. xxvi Dedication A PhD, from the perspective of anyone not working on one, is akin to being involved in an obscure and complex variant of poker, in a pitch-dark room, with blank cards, for infinitesimal stakes, with a Dealer who won’t tell you the rules, and who smiles all the time. - Adapted from Terry Pratchett and Neil Gaiman’s Good Omens Proof that you don’t always have to be an expert to know what you are talking about. To my wife, family and friends; thank you for helping me see this through. 1 1 Steam Reformation, Reforming Solid Oxide Fuel Cells and Existing Optical Diagnostic Techniques An outline of steam reformation and gas shift reactions is presented and applications for the produced hydrogen discussed. Reformation regimes in the context of SOFC’s are described, with the advantages and disadvantages of each technique highlighted. The use of optical diagnostic techniques for research into SOFC’s is reviewed and summarised. Gaseous Raman spectroscopy and IR thermometry are described and the potential value, to SOFC developers and computer modellers, of experimental data gathered using these techniques outlined. The technical challenge of implementing gaseous Raman spectroscopy into an optically restricted, high temperature environment is presented and the use of a methane/steam reformer so as to prove the concept is justified. The chapter concludes with the stated aims of the thesis and an outline of the thesis structure. 1.1 Steam Reformation The steam reformation of fossil fuels is a process in which steam and hydrocarbons are heated in the presence of a catalyst to produce hydrogen and carbon monoxide. Steam reformation is primarily performed to produce bulk hydrogen; the uses of which include the production of ammonia and as a fuel for fuel cells. The reaction for the steam reformation of hydrocarbons is shown in Equation 1.1 [Dicks, 1998]. )()2/()()()( 22 gHnmgnCOgOnHgHC mn ++→+ (1.1) The carbon monoxide is then reacted with steam via a gas shift reaction to produce carbon dioxide and further hydrogen, Equation 1.2. 2 )()()()( 222 gHgCOgOHgCO +→+ molkJH o /42−=Δ (1.2) Typically it is natural gas that is reformed, the major constituent of which is methane (approximately 95% [Union Gas, 2010]). Therefore, the reformation reaction in terms of methane is shown in Equation 1.3. )(3)()()( 224 gHgCOgOHgCH +→+ molkJH o /206+=Δ (1.3) The overall steam reformation and gas shift reaction of methane is given by Equation 1.4. )(4)()(2)( 2224 gHgCOgOHgCH +→+ molkJH o /164+=Δ (1.4) It can be seen that overall the reformation and gas shift reactions are endothermic and as such are typically performed at temperatures in excess of 700 °C in the presence of a nickel catalyst. Steam reformation on an industrial scale is an established process but is also an area of continuing research [Adris et al. 1994; Prasad et al., 2002; Johnsen et al., 2007]. However, steam reformation is also being investigated as a method of in-situ hydrogen production for fuel cells, particularly medium and high temperature fuel cells. 1.2 Steam Reformation for Medium and High Temperature Fuel Cells Fuel cells are electrochemical energy conversion devices that take fuel and air streams and catalytically combine them to produce electrical energy. Theoretically, fuel cells have higher energy conversion efficiencies than thermal energy conversion devices and as such are being actively developed as a practical means of energy conversion. 3 The basic operation of a fuel cell, in this case a Solid Oxide Fuel Cell (SOFC), is illustrated by Figure 1.1. Figure 1.1 Fuel cell schematic with anode, cathode and overall FC reaction. In the case of a SOFC, hydrogen on the anode side of the cell is combined with oxygen ions which have been transported through the electrolyte from the cathode side of the cell. The electrolyte, while permeable to oxygen ions is impermeable to gases and is non-conductive to electrons which are transmitted via an electrical circuit from which electrical energy can be extracted. It is the electrolyte that dictates the categorisation of the fuel cell and, as an example, a proton exchange membrane or polymer electrolyte membrane (PEM) fuel cell operates with the same hydrogen and air gas streams, but hydrogen ions are transmitted across the electrolyte at temperatures ranging from 70 – 90 °C. In the case of SOFC’s the electrolyte is a ceramic which exhibits the ability to A Anode H2 + O2- → H2O + 2e- Cathode O2 + 4e-→ 2O2- O2- H2 O2 Electrolyte Overall Reaction H2 + ½O2 → H2O e- e- 4 transmit oxygen ions only at considerably higher temperatures than that of a PEM, broadly categorised as high temperature; 800 – 1100 °C, medium temperature; 500 – 800 °C and low temperature; < 500 °C. It is important to note that the SOFC reaction, while not self initiating is self sustaining and it actually necessitates some degree of cooling. Therefore, SOFC’s have excess thermal energy available as a result of their operation. Conversely hydrogen, the typical fuel of fuel cells, while a relatively abundant element [Kaye et al., 2005] is not widely available in its elemental state. As such, fuel cell research is closely associated with methods of hydrogen production. One such hydrogen production technique is steam reformation and in light of the excess, high temperature, thermal energy available through SOFC operation the in-situ reformation of hydrocarbons at the point of energy conversion is an area of research. In addition to the SOFC providing thermal energy, the creation of water on the anode side of the cell also has the potential to provide sufficient steam for the reformation/gas shift reactions. Hydrocarbon fuels also offer practical advantages with respect to pure hydrogen in that they do not need to be stored at such high pressures or low temperatures in order to achieve practical levels of energy density. There are three regimes through which steam reformation can be performed using the excess thermal energy of a SOFC and these are external reformation, indirect internal reformation (IIR) and direct internal reformation (DIR). 1.2.1 External Reformation External reformation is described as such because the steam reformation process takes place away from the electrode and electrolyte assembly. For this reason only heat transferred from the SOFC stack is used in the reformation reaction, meaning that the steam necessary for the reformation process must be provided separately. This 5 technique however is the simplest to implement as the reformer has minimal restrictions placed upon it in terms of materials (both for construction and SOFC reaction catalyst), configuration and size. The priority is therefore efficient heat transfer from the fuel cell stack and a schematic of the arrangement is shown below in Figure 1.2. Figure 1.2 Schematic of external SOFC steam reformation arrangement. An advantage of external reformation is that the fuel cell stack is functionally unaffected by the presence of the reformer, allowing the design and operation to be tailored specifically to the needs of the SOFC. However, it can be seen that if the fuel cell stack under consideration is planar (as opposed to tubular) an external reformer arrangement has to be found that complements the cooling requirements of the SOFC. And as previously stated none of the water produced by the SOFC reaction can easily be used by the reformer, requiring the fuel stream to be fully humidified by an additional process. Air Fuel Cell stack Unused gases out Methane and steam Hydrogen and CO2 Reformer Heat transfer CO2 and steam 6 1.2.2 Indirect Internal Reformation A planar fuel cell stack consists of a series of repeating units. Each unit consists of the ‘cell’ or electrode and electrolyte assembly plus fuel and air channels which distribute the respective gases across the cell and collect the current from the surfaces of the electodes. A schematic of the repeating unit is shown in Figure 1.3 Figure 1.3 Generic diagram of a cell with fuel and air channels. In a stack, Figure 1.4, the anode and cathode current collectors are situated adjacent to one another and for this reason the anode and cathode current collectors are referred to as the interconnect. Air flow Fuel flow Anode Side Cathode Side 7 Figure 1.4 Generic diagram of a cell stack highlighting cell and interconnect interaction. The schematic of Figure 1.4 illustrates that a considerable percentage of the stack volume consists of the interconnect - approximately two thirds – and it is within the interconnect that IIR takes place, Figure 1.5. Figure 1.5 Schematic of indirect internal reforming (IIR). Repeating Unit CH4,H2O Air CO2,H2O CO2,CO,H2 Reforming Section Fuel Channel Air Channel Cell 8 IIR has several advantages over external reformation. The first is that by integrating the reformation inside the fuel cell stack, heat transfer is significantly simplified. Secondly, as with external reformation the catalyst used for the reformation process can be tailored directly to the need [Aguiar, et al., 2004]. Thirdly, as illustrated by Figure 1.5, water does not need to have been supplied to promote the gas shift reaction, Equation 1.4, as water produced from the fuel cell reaction can be utilised. This marginally reduces the amount of external steam generation required. At the same time the disadvantage is that significant humidification of the fuel stream is still necessary. In addition, the volume of an IIR stack with respect to a non-IIR stack of an equivalent number of cells is increased, which will reduce power density and increase cost, manufacturing etc. Finally, the electron path length between cells of an IIR stack is increased, resulting in greater losses through electrical resistance. 1.2.3 Direct Internal Reformation The anode side of the cell primarily exists to promote the combining of hydrogen and oxygen ions to form water and free electrons. However, a secondary function of the anode layer can be to mechanically support the electrolyte, the thickness of which (approx. 10 μm [Leng et al., 2004]) is minimised in order to lower the resistance to ionic conduction [Singhal et al., 2003; Souza et al., 1997]. This mechanical support provided by either electrode is particularly critical when the stack undergoes large changes in temperature such as start up and cool down [Dicks, 1998; Aguiar et al., 2004]. A mechanically supporting anode layer can be up to 1 mm in thickness, however the electrochemical reaction of the fuel cell can be shown to only extend 10 μm beyond the electrolyte-anode interface [Brown et al., 2000], which leaves approximately 90% of the anode structure unused by the electrochemical fuel cell reaction. However, the anode is 9 constructed of the same ion conducting ceramic as the electrolyte and an electron conducting metal to both catalytically promote the combining of hydrogen with oxygen ions as well as providing continuity to the electrical circuit of the cell. In SOFC’s the anode is typically constructed of a cermet consisting of a mixture of yittria stabilised zirconia and nickel, the same catalyst traditionally used in the reformation of hydrocarbons for bulk hydrogen production. Therefore, an anode supported electrolyte has a relatively large, catalytically rich region which is able to host the reformation and gas shift reactions, Figure 1.6. Figure 1.6 Schematic (not to scale) of the reformation and electrochemically active regions of a methane fuelled SOFC. An advantage of DIR is that the electrochemical reaction produces water directly adjacent to the reformation region and so provides much of the water needed by the reformation and gas shift reactions. This significantly reduces humidification Electrolyte Cathode Electrochemically active region O2- CH4 + H2O Structural / reformation catalyst region H2 H2O CO O2- H2 CO2 H2O CO2 + H2O Air Unused gases O2 Anode 10 requirements of the inlet gas stream. Equally, the fuel channels and anode side of the cell require little to no physical modification with respect to an externally reforming, or pure hydrogen fuelled, SOFC arrangement and therefore does not negatively affect the volumetric power density of the stack. However, as can be seen from Equation 1.3, the overall reformation/gas shift reaction is strongly endothermic, which results in cooling of the electrode and electrolyte at the anode entrance and results in thermal stresses [Dicks, 1998, Aguiar et al., 2004]. It has also been observed that the methane is completely consumed “a very short distance” beyond the anode inlet [Achenbach et al., 1994b]. The catalyst loading of a DIR anode is therefore a compromise between promoting the electrochemical reaction and mitigating against thermal stresses resulting from the reformation reaction. DIR fuel cells are therefore seen as a potential method of adding minimal complication to SOFC architecture when performing in-situ reformation to produce hydrogen. However, the advantages of DIR are offset by the increased thermal stresses, the extra complexities of additional chemical reactions on the anode side of the cell and the conflicting catalytic requirements of the reformation and electrochemical reactions. Another factor affecting the strategy of hydrocarbon reformation at the point of energy conversion is that of impurities in the fuel. One such impurity is sulphur which has been shown to poison nickel catalysts at concentrations of 10 ppm [Singhal et al., 2003] and in the case of hydrogen sulphide as low as 0.05 ppm [Matsuzaki et al., 2000]. It is therefore important to have a full understanding of this and other phenomena occurring in the electrode and gas channel in order to successfully and reliably implement DIR and other fuel reforming strategies. 11 1.3 The Monitoring of SOFC Phenomena Using Optical Diagnostics As discussed in the previous section, reforming SOFC’s and particularly DIR SOFC’s, have a potential fuel handling advantage over SOFC’s fuelled by pure hydrogen. However, it can also be seen that there are potential disadvantages and added complexities to DIR. In order to optimise the performance of the SOFC, reformation/gas shift reactions and the chemical/thermal interactions between the two it is necessary to have a complete understanding of the phenomena taking place. SOFC performance can be monitored using conventional techniques such as thermocouples, gas sampling probes and the monitoring of the cell’s electrical performance. However, particularly with regards to SOFC’s the high temperatures make the use of physical probes non-ideal and as such computer modelling has been used extensively to gain insight into SOFC performance characteristics. Equally however, SOFC computer models are subject to constant improvements and refinements so as to more accurately predict observations made experimentally, making experimental validation an important part of computer model development. Any technique that can offer further insight into SOFC operation and performance is of value, with one area being the use of optical diagnostic techniques. Optical techniques have been found to be advantageous in the context of SOFC’s because they require no physical contact with the medium being measured, requiring only an optical line of sight and in some cases a light source, which partially eliminates issues such as electrical and thermal conductivity. One technique used was that of Fourier transform infrared (FTIR) emission spectroscopy [Lu et al., 2002] on the electrode surfaces. Surface Raman spectroscopy was however able to determine the chemical composition of the electrode surfaces more accurately [Cheng et al., 2007; Choi et al., 2006; Choi et al., 2008; Maher et al., 2008; Pomfret et al., 2006; Pomfret et 12 al., 2007; Pomfret et al., 2010a]. Surface Raman spectroscopy has also been used to monitor the poisoning and contamination of the electrode, specifically the formation of solid carbon on the anode surface [Eigenbrodt et al., 2010; Pomfret et al., 2008]. Critically however, these techniques have, to date, only been employed on button cells which are a miniaturised version of a full scale planar SOFC. Button cells are circular SOFC’s in the order of tens of millimetres in diameter and allow researchers to quickly and relatively cheaply prototype aspects of SOFC design such as catalyst loadings, electrode porosities etc. Another advantage of button cells is that they require simplified gas sealing arrangements, gas manifolding and current collectors with respect to full scale planar SOFC’s which can measure hundreds of millimetres across. The work on surface Raman spectroscopy has also, to date, been carried out using a Raman microscope arrangement, necessitating the close proximity of the detector head to the surface being interrogated. In addition to Raman spectroscopy studies into temperature variation due to fuel type and electrical load have been performed using IR thermography [Brett et al., 2007; Ju et al., 2008; Pomfret et al., 2010b]. The thermography studies have, to date, also used button SOFC’s and produced global temperature profiles of the electrode surface. Surface Raman spectroscopy and thermography have provided valuable information with respect to in-situ changes in electrode composition and temperature of an operating button cell SOFC and helped to refine computer models. However, one relative unknown is the precise gaseous composition of the gas in the gas channels above the electrodes. As has been described, this information is difficult to obtain to a high degree of spatial resolution using conventional sampling techniques. Equally, the surface Raman spectroscopy and thermography have, to date, not been used in a SOFC configuration equivalent to that found in large scale planar SOFC’s. 13 1.4 Gaseous Raman Spectroscopy and IR Thermometry Gaseous Raman spectroscopy is similar to surface Raman spectroscopy in that it allows the chemical composition of a gaseous sample to be determined without any physical contact with the medium being sampled. It is therefore a technique with the potential to determine the chemical composition of the electrode gas channels and produce experimental information of value to SOFC developers and computer modellers. Gaseous Raman spectroscopy can also be implemented without using a specialized Raman microscope as used in the works previously cited. This significantly removes the experimental constraint imposed by the comparatively short focal length of the Raman microscope and so enables gaseous reacting flows other than that of button SOFC’s to be observed. The greater freedom of experimental architecture is significant when considering the chemical composition of gaseous flows in DIR SOFC’s; which not only host the electrochemical reaction but the reformation reactions also. In addition to these reactions phenomena such as carbon deposition can occur by the mechanism of the Boudouard reaction, Equation 1.5 and the reduction of carbon dioxide by hydrogen, Equation 1.6. CCOCO +⇔ 22 molkJH o /172−=Δ (1.5) OHCOHCO 222 +→+ molkJH o /414−=Δ (1.6) Equally, if optical access can be achieved to allow Raman spectroscopy it is conceivable that sufficient optical access exists to allow additional non-contact measurements such as temperature to be made. However, rather than seek a global surface temperature profile possible through thermography a device such as a 14 pyrometer or IR thermometer can be used to allow ‘point’ measurements to be made and reduce the need for large scale optical access. As will be discussed, localised temperature measurement can potentially improve the accuracy of the gaseous Raman spectrometry technique. Gaseous Raman spectroscopy is however non-trivial. Unlike surface Raman spectroscopy the gaseous medium from which the Raman signal is derived is significantly less dense, which results in a proportional decline in Raman signal intensity. In addition, the relatively high temperature of the SOFC means that blackbody radiation, depending on the wavelength of the light source, could potentially obscure the relatively weak Raman signal. As such, a light source of suitable wavelength must be selected, however, as will be discussed in the following chapter this brings with it additional challenges. The task of achieving optical access while maintaining a gaseous flow field that more closely represents that which would be found in an unmodified SOFC is also non-trivial and requires careful experimental design and implementation. 1.5 Methane/Steam Reformer Proof of Concept With the proposal of a novel application of gaseous Raman spectroscopy and SOFC’s at the pre-commercialisation stage the decision was made to focus efforts on the implementation and refinement of the technique, rather than attempt to operate a SOFC and implement the diagnostic technique in such a challenging environment. As such, the reacting gas flow of a planar methane/steam reformer was selected with which to prove the concept of gaseous Raman spectroscopy in an optically constrained, high temperature environment. A methane/steam reformer was also chosen because the catalytic surface used to host the reactions was similar to that of the anode side of a planar SOFC. This makes the results of interest to DIR SOFC’s developers but should 15 simplify the subsequent transferral of the technique to a fully operational SOFC. The work presented in this thesis is however applicable to any optically constrained, high temperature reacting flow, with SOFC’s being one of several possible applications. 1.6 The Aims of this Thesis In light of the above, it was perceived that there was a need for accurate experimental data to be extracted from a SOFC type environment in order to assist high temperature fuel cell modelling and fuel cell development in general. Conventional in-situ techniques such as thermocouple temperature measurements and exhaust gas sampling have been joined by advanced optical diagnostic techniques such as surface Raman spectroscopy and thermography. It was therefore anticipated that applying gaseous Raman spectroscopy to a SOFC type environment would generate data of value to SOFC developers and modellers. These techniques would have to provide the ability to measure chemical composition and temperature without significantly effecting the overall operation of the fuel cell. Also, it should be done in a way that yields data at a sufficiently high resolution in both one and two dimensions to make the adoption of the techniques suitably advantageous to SOFC researchers and developers. 1.7 Thesis Structure The aim of the thesis is to provide the context and background to the research undertaken, as well to present the results and findings of the author’s original work. As such, Chapter 1 provides the context for the research with respect to SOFC’s, and the challenges facing SOFC development. It also outlines the previous work carried out using optical diagnostic techniques on button SOFC’s. 16 Chapter 2 explores the fundamental principles of Vibrational Raman spectroscopy and Radiation thermometry. This chapter is intended to provide the non- expert reader with a basic understanding of these techniques so they might appreciate the subsequent apparatus design decisions made as a result of the (predominantly) optical requirements of the techniques. Readers of this work who are already familiar with the operating principles and experimental applications of Vibrational Raman Scattering and Radiation thermometry may choose to move directly from the end of Chapter 1 to Chapter 3. Chapters 3 and 4 cover the design and development of the experimental apparatus while Chapters 5 and 6 describe the experimental apparatus and test matrix design. Key design parameters and important decisions taken during the development process are discussed and justified. Chapters 5 and 6 are also intended to serve as a guide to future replication of the apparatus and experiments. Chapter 7 details the experimental results which includes the process of calibration, detection of experimental errors and the effects of the experimental variables. Chapter 8 summarises the work presented, highlights original aspects of the work, and highlights areas of future work. 17 2 Raman Spectroscopy and Radiation Thermometry In this chapter the quantum and classical principles of Raman spectroscopy are reviewed. Raman Scattering is differentiated from other scattering and absorption phenomena and the categories of Stokes and Anti-Stokes Scattering are presented. An overview of the important Raman lines related to this research is presented and a methodology is outlined that allows quantitative analysis of gaseous chemical species concentrations to be made. Similarly, the factors and relationships to be considered when selecting an excitation wavelength are addressed and the importance of minimising the signal-to-noise ratio (SNR) is presented. The chapter then moves to describe the fundamental principles of radiation thermometry and its suitability to measure the temperature of a SOFC electrode. The effects of surface finish, directional emissivity, experimental environment, and background radiation are considered and the need for calibration of a device is presented if accurate measurements are to be obtained. The chapter ends with the conclusion that infrared radiation thermometry is a valuable non-contact temperature measurement tool that is well suited to the application of measuring the surface temperature of a SOFC electrode. 2.1 The Electromagnetic Spectrum and Raman Spectroscopy In order to apply the technique of Raman spectroscopy it is necessary to have a detailed understanding of the interactions between electromagnetic radiation and the body of interest and specifically an understanding of the atomic interactions that occur between the incident radiation and the molecule(s) of interest. In simple terms the electromagnetic spectrum is an energy scale, with the shorter wavelengths (higher frequencies) having greater energy than longer wavelengths. This is described by Planck’s equation: 18 hvE = (J) (2.1) Where: E = is the energy (J) h = Plank’s constant, 6.6262 x10-34 (J . s) v = frequency (Hz) Quantum physics tells us that the energy of a wave is quantized, consisting of discrete packets of energy referred to as photons. It is the manner in which these photons interact with the atoms of a material that can provide valuable information about the material itself. The study of these phenomena is called photo-physics and goes far beyond the scope of this thesis. Therefore what follows is a condensed explanation of the mechanics involved in Raman spectroscopy with a focus on the limitations of the technique and the considerations that need to be taken into account when and applying it to a practical problem. 2.2 Molecular (Internal) Energy Levels A molecule has three methods of internal energy storage; these are electronic, rotational and vibrational. Each one of these parameters is quantized, which results in the overall internal energy of a molecule only being able to sum to one of a number of discrete energy levels particular to that molecule. Therefore, two molecules consisting of different elements will have different allowable energy levels; thus, dictating how internal energy is stored, added to and subtracted from the molecule. It is the differences between the allowed energy levels of different molecules that allow different compounds to be distinguished from one another. 19 Rotational energy is the energy stored due to the frequency at which a molecule rotates. Vibrational energy is the storage of potential energy by the stretching, compressing or bending of the atomic bond between two atoms of the molecule. The rotational and vibrational energies are in fact related to one another, with changes in the vibration energy affecting the rotational moment of inertia and vice versa. The electronic energy of a molecule is primarily determined by the orbit of its electron(s) which directly affect its angular momentum. As mentioned earlier, quantum mechanics and photo-physics have been, and are, the focus of extensive research. The accuracy and completeness of the model used when performing Raman spectroscopy is dependant on the application and information sought from the experiments being performed. Entire books, such as the Handbook of Raman Spectroscopy [Lewis et al., 2001], discuss at great detail the observable energy states of molecules in order to validate theories and demonstrate the sensitivities of specialised spectroscopy equipment. However, in research fields such as combustion and chemical kinetics, this degree of resolution is typically not required. Accordingly, further discussion of the internal energy mechanisms of molecules is considered unnecessary at this time, other than to say that more intricate interactions do exist but that they have no significant effect on the Raman spectroscopy measurements made in this thesis. 2.3 Photon Molecule Interaction With the energy storage mechanisms of a molecule outlined it is now possible to consider their relation to Raman spectroscopy. Raman spectroscopy involves electromagnetic radiation (photons) interacting with the material of interest. More specifically, it is necessary to consider how a single photon interacts with a single 20 molecule of the material of interest. It can be seen that each photon has a discrete amount of energy and each molecule has a finite number of energy storage mechanisms and in turn a discrete number of energy levels. This leads to a finite number of possible outcomes when a photon and molecule interact which can be divided into several different categories, a number of which are presented in Figure 2.1. 21 Figure 2.1 A selection of possible photon/molecule absorptions and emissions. Adapted from Zhao et al. [Zhao et al. 2001] The first interaction shown is that of Rayleigh scattering. In the simplest terms this is the elastic scattering of light; that is to say that the energy of the incident photon is the same as the photon scattered by the molecule. This is the most likely process when considering the scattering of photons by particles that are smaller in size to the Ground Electronic State Virtual State Excited Electronic State J+2 J V’=0 V”=1 Rayleigh Raman (Stokes) Raman (Anti-Stokes) Fluorescence Where: J = rotational quantum number V= vibrational quantum number V”=vibrational quantum number of ground electronic state V’=vibrational quantum number of excited electronic state J-1 J-1 J-1 J J+2 J+2 J V”=0 22 wavelength of the incident photons (in contrast to Mie Scattering, which will not be considered here). Two points should be noted at this time; the first being the ‘interim’ energy level reached by the molecule after absorbing the photon. This energy state is described as the virtual state because the energy level reached cannot necessarily be supported by the allowable rotational, vibrational and electronic internal energy storage mechanisms, which as discussed earlier are subject to strict quantum rules. The molecule can therefore not truly absorb the photon resulting in a very rapid interaction between the photon and molecule in the order of femtoseconds (10-15s) [Zhao et al., 2001]. This rapid interaction is the second point to note and is a factor that differentiates between Raman Scattering and photon emission through fluorescence, which will be discussed later in this section. The next interaction shown in Figure 2.1 is that of Raman Scattering, or more specifically Stokes Scattering. Raman is distinguished from Rayleigh because it is the inelastic scattering of light; that is to say that the scattered photon is of a dissimilar energy to that of the incident photon. In the case of Stokes Scattering, the scattered photon will have a lower energy (lower frequency) than the incidnet photon. The mechanisms dictating the final internal energy level of the molecule are complex, but in summary it is the rotational/vibrational combination that is; firstly, the most feasible given the harmonic frequencies and vibrational modes the molecule is able to occupy, and secondly, the most statistically likely final energy state achievable. As will be discussed in §3.4, many different final energy levels are both feasible and achievable, but in all likelihood only a small number (usually one or two) distinguish themselves as being statistically more likely to occur than other final energy states. It is these statistically significant energy levels and the scattering resulting from these that are of most interest to a researcher seeking information about chemical species and species concentrations. 23 Anti-Stokes Scattering is identical to Stokes Scattering in that it is an inelastic scattering of light, albeit by the scattering of a photon of greater energy than the incident photon. This is only possible however when the molecules initial energy state is sufficiently high before the initial photon/molecule interaction. Only with sufficient initial internal energy can the molecule impart energy to the scattered photon and still conform to the quantum internal energy distribution rules. Because a molecule naturally seeks a minimum electronic, rotational and vibrational energy state it is not always possible to observe Anti-Stokes scattering. For this reason Stokes Scattering is the more frequently observed form of scattering and provides the largest signal available to the experimentalist considering which form of vibrational Raman Scattering to observe. However, Stokes Scattering is challenging to observe as the signal is around three to five orders of magnitude smaller than the Rayleigh signal [Zhao et al., 2001; Stevenson et al., 1996]. Thus, the use of Raman techniques necessitates careful consideration of signal-to-noise ratios (SNR) during experimental design and procedures. Before providing the description of fluorescence it is necessary to clarify an aspect Raman scattering and that is that Stokes Raman Scattering, as seen from Figure 2.1, involves an elevation in vibrational energy level of the molecule. This may also be accompanied by a change in the rotational energy too, but as will be discussed later in this chapter, rotational excitation results in very small wavenumber shifts. These are of little use for practical applications although extensive literature does exist on this specialised area of spectroscopy as detailed by Laserna, Lewis, and Weber [Laserna et al., 1996; Lewis et al., 2001; Weber, 1979]. The final photon/molecule interaction that will be considered here is that of fluorescence. Fluorescence occurs when a photon/molecule interaction elevates the electronic state of the molecule directly to a higher electronic state. The incident photon was of sufficient energy that the molecule can be said to have genuinely absorbed the 24 photon rather than scattering it, with or without an energy exchange. However, despite the molecule being in a stable energetic state from the perspective of the energy storage mechanisms of the molecule, it is crucially no longer at equilibrium with its surroundings. Therefore, the molecule seeks to return to equilibrium and in doing so emits a photon. It is the absorption mechanism of fluorescence which requires the excitation source’s wavelength to coincide with the absorption frequency of the species of interest [Zhao et al., 2001] and it is this frequency requirement that inhibits the practical excitation of stable and diatomic molecules as this would necessitate excitation sources deep into the UV. The different mechanisms of photon/molecule interaction of Raman Scattering and fluorescence result in significantly different observable signals from the two techniques. The signal intensity of fluorescence can be several orders of magnitude larger than the Raman signal [Zhao et al., 2001; Stevenson et al., 1996], so Fluorescence (and the avoidance there of) is of critical importance to the Raman experimentalist. In addition, each technique has the ability to detect specific chemical species but the higher photon energies needed to raise the electronic state of stable and diatomic molecules necessitates extremely short wavelengths, making fluorescence impractical with the current excitation sources available. Also, for fluorescence to be able to detect multiple chemical species multiple excitation sources would be required which would substantially increase experimental complexity and expense. For this reason the stronger fluorescence signal is typically used to analyse the concentrations of more easily excited radicals which are produced during combustion processes. In addition, combustion is a comparatively noisy process, in the SNR sense, and so a stronger signal is required in this application which Raman is not able to provide [Alger et al., 2004]. Alternatively if the spatial concentration measurements of stable species need to be measured it is sometimes possible to substitute a more easily 25 fluoresced species for non-reactive fluid flows or mix a secondary seeding agent and approximate the location of the original species of interest [Bombach, 2002]. With respect to the gas phase reactants and products of reformation (H2O, CH4, CO, CO2 and H2) fluorescence is not easily achieved and ideally the concentration of each species would like to be ascertained. Raman spectroscopy is therefore deemed a more suitable technique in this particular application as it allows the simultaneous measurement of numerous chemical species. Care must however be taken in selecting an excitation source so as not to unintentionally cause fluorescence which could lead to unfavourable SNR ratios. 2.4 The Raman Spectrum As discussed in §2.3 the energy exchange that occurs during the photon/molecule interaction is governed by quantum theory but it is not limited to a unique pathway. Therefore, a sample of molecules, identical in both elemental composition and energetic configuration will inelastically scatter photons from a monochromatic source at a number of different frequencies. Initially this may seem to make Raman Scattering an unusable diagnostic tool to distinguish between and quantify the concentrations of various chemical species and compounds. However, the strict quantum rules make the frequencies at which photons are scattered distinct and whilst a number of spectral lines are produced there is a line of sufficient intensity which distinguishes and quantifies a species with respect to dominant spectral lines produced by other chemical species. This dominant line or branch is termed the Q-branch. Figure 2.2 below shows a representative diagram of the spectral lines resulting from the scattering of light from a diatomic molecule. 26 Figure 2.2 An example of spectral lines of a diatomic molecule produced through Raman scattering [Zhao et al., 2001] Figure 2.2 illustrates the presence of multiple branches in addition to the Q- branch labelled O and S. The Q-branch is the result of a purely vibrational change in energy level of the molecule, thereby imparting or subtracting the equivalent amount of energy from the incident photon before the scattering takes place. The O and S branches are the result of a rotational and a vibrational change in the energy level of the molecule interacting with the photon. Changes in rotational energy are small in comparison to changes in vibrational energy levels. This means that for inelastic scattering occurring purely through rotational energy changes only a small amount of energy is imparted to or given up by the molecule to the scattered photon. As such the resulting scattered photons will be of a frequency similar to that of the incident photon. This results in a spectral line close to the elastically scattered Rayleigh line, but up to 5 orders of magnitude smaller, which raises issues with differentiating and accurately measuring the Raman signal. This difficulty in Wavenumber (cm-1) Rayleigh Line S c a tt e r In te n s it y ( W /s r) O O S S Q Q Stokes Anti-Stokes 27 distinguishing between the weak pure rotational Raman lines and intense Rayleigh line make it unsuitable for use as an engineering diagnostic tool to distinguish chemical species. It is however useful when studying the fundamental structures and chemical behaviours of complex compounds [Brodersen, 1979]. The more energetic occurrence of vibrational Raman scattering is of greater practical benefit to researchers using Raman spectroscopy as a diagnostic tool. The greater change in energy results in greater shifts in frequency of the scattered photons, making it more easily distinguishable from Rayleigh scattering. Changes in rotational and vibrational energy can and do occur and it is these interactions that result in the spectral lines such as O and S, but these lines are not as intense as those of pure vibrational Raman scattering. As mentioned in §2.2 the rotational and vibrational energies of a molecule cannot be considered entirely independently of one another. A different vibrational energy, or quantum number, can affect the energy stored by a given rotational quantum number, or more specifically the distribution of the energy stored. However, for diatomic molecules, certainly for linear ones, the change in energy distribution from one vibrational quantum number to another is small [Eckbreth, 1988]. This means that for a sample of linear molecules at thermal equilibrium there will be a significant proportion at the same vibrational quantum number and encompassing a range of rotational quantum numbers. If this sample of molecules is then elevated in vibrational quantum number as a result of interacting with photons from a monochromatic light source, the subsequent change in rotational energy distribution will be small. Therefore the energy imparted to or given up by each photon, regardless of the rotational energy level of the molecules of the sample, will be the same. This produces a spectral line (the Q-branch) that represents a purely vibrational change in the internal energy of the molecule. Other lines such as O and S will also occur, but these represent the rotational and vibrational transitions. The 28 vibration/rotation interaction does however become more significant, particularly for species such as hydrogen [Eckbreth, 1988] at higher temperatures as will be discussed in §7.6. The above description of molecular energy storage mechanisms have been validated by the spectral analysis of pure samples subjected to excitation by monochromatic sources of electromagnetic radiation. Thus, if equipped with a suitable list of vibrational frequency shifts for a selection of chemical species it is them possible to detect which species are present by spectral analysis through Raman spectroscopy, Table 2.1. 29 Species Vibrational Frequency 1 (cm-1) Stokes Spectral Position of Q branch (nm) Σi(Q)1 CH4 2914 338.4 8.55 CO2 1388 321.7 1.13 CO 2143 329.8 0.93 H2 4155 353.2 3.86 H2O 3652 347.0 3.51 N2 2331 331.8 1.00 Table 2.1 Vibrational frequencies, Stokes spectral positions of Q branches and normalized vibrational cross sections for the gaseous species involved in the steam methane reformation reactions produced by a 308nm excitation source. Note that the significance of the final column, the normalised relative differential Raman Scattering cross section will be discussed in the following section. 2.5 Making Raman Measurements Up until this point, Raman Scattering has been considered from the perspective of quantum physics. There is however a classical approach which explains Raman from a more intuitive perspective. Very simply, the explanation revolves around considering a molecule as a negatively charged ‘cloud’ of electrons around a positively charged nucleus. Crucially for a molecule in equilibrium, the ‘centre’ of the negative electron cloud is either at the same location as the positive centre of charge of the nucleus (an isotropic molecule) or some distance and direction from it (anisotropic molecules). In the 1. [Schrötter, 1982] “Linear Raman Spectroscopy: A State of the Art Report” Σi(Q) : Average normalised relative differential Raman Scattering cross section 30 case of anisotropic molecules, the charge imbalance causes the molecule to be polarised; i.e., have a region of positive charge and negative charge. As the electric field of the electromagnetic wave interacts with a molecule there is a physical displacement of the electron cloud. This results in the ‘centre’ of the negative charge cloud changing location with respect to the positive charge of the nucleus. In turn, this produces a polarisation in the case of an isotropic molecule or a change of polarisation in the case of an anisotropic molecule. Examples of isotropic molecules are the diatomic molecules discussed in the previous chapter with a water molecule being an example of an anisotropic molecule on account of the dipole moment present. The principle of Raman spectroscopy, with respect to the induced displacement of the equilibrium positions of the centres of the positive and negative charges of a molecule as a result of interaction with an electromagnetic wave, is the same for both isotropic and anisotropic molecules. This resulting displacement is referred to as the induced dipole moment and its relationship with respect to the strength of the electric field is given in Equation 3.2 [Baranska et al., 1987; Stevenson et al., 1996; Eckbreth, 1988]. Ep α= (C . m) (2.2) Where: p = induced dipole moment (C . m) α = polarisability (C . m2 . V-1) E = Electric field (V . m-1) From Equation 2.2 it can be seen the induced dipole moment is a function of the polarisability and the electric field, with the increase of either of these terms resulting in a greater induced dipole moment. It is the size of the induced dipole moment that dictates 31 the intensity of the Raman signal and for this reason increasing the magnitude of the electric field interacting with the molecule is advantageous, as discussed further in §2.6. However, the polarisability of the molecule undergoing the interaction is an aspect which is beyond the control of the experimentalist. In the simplest terms the polarisability of a molecule describes the propensity of the molecule to have an induced dipole imposed upon it, with some molecules electron clouds being more easily displaced than others. Another way of considering this mechanism is the likelihood of a molecule to participate in the process of dipole displacement and as such can be considered to be the cross section of a molecule (literally a molecule with a greater cross section is more likely to interact with the electromagnetic wave). A more rigorous description of molecular cross section is presented in Appendix I, but for small solid angles the power of Raman scattering can be described using Equation 2.3 below [Eckbreth, 1988; Zhao et al., 2001]: ησ lΩ⎟⎠ ⎞⎜⎝ ⎛ Ω∂ ∂= nPP ir (W) (2.3) Where: Pr = Raman signal power (W) Pi = Incident light power (W) n = number density of the scattering species (m-3) ⎟⎠ ⎞⎜⎝ ⎛ Ω∂ ∂σ = vibrational cross section (m2 . sr) Ω = solid angle (sr) l = the sampling extent (m) η = the efficiency of the collection optics 32 With Equation 2.3 it is now possible to either determine the Raman cross section of a molecule, or its concentration, if all other parameters are known. As a final note in this section it should be noted that when using an unpolarised excitation source the intensity of the Raman scattering observed is independent to the orientation at which it is observed. 2.6 Raman Signal Intensity As discussed previously in §2.3 the Raman signal is extremely weak in relation to Rayleigh scattering and fluorescence. It is therefore important to optimise the Raman signal by carefully designing the experimental set up, as there will always be some degree of unavoidable background noise. Accordingly it is necessary to have a good understanding of the factors that determine the intensity of the Raman signal. Fortunately, the study of the Raman is a mature one (its discovery having been made in 1928) and the governing equation for signal strength is well known, Equation 2.4 [Eckbreth, 1988]. nnmnmnm Npc I 2403 0 2 )( 2 ννε π −=Ω (W) (2.4) Where: IΩnm = radiant intensity of Raman signal (W) n = initial state m = final state ε0 = permittivity of free space, 8.85 x10-12 (C. V-1. m-1) c = speed of light in a vacuum, 2.9979 x 10-34 (m . s-1) 33 ν 0 = incident frequency (s-1) ν nm = frequency shift (s-1) pnm = induced dipole moment (C . m) Nn = Number of scattering molecules in state n (m-3) From Equation 2.4 it can be seen that there are two major parameters that affect the Raman signal intensity. The first is that the Raman signal increases to the power of four in relation to the difference between the excitation frequency and the frequency shift. This means that the shorter the wavelength excitation source that can be used, the greater the Raman signal; with shorter wavelengths having significant gains in signal intensity. It is for this reason that visible and UV lasers have found favour amongst high temperature and combustion researchers as Nn is typically small in gaseous, low pressure applications. The second point to note from Equation 2.4 is the dependency of radiant signal intensity on the induced dipole moment, pnm. The effect of the dipole moment is somewhat intuitive, in that the more intense the incident light the greater the dipole moment produced and thus the greater the Raman signal. Therefore, an increase in signal can be achieved by both an increase in the number of incident photons and photon flux per unit time [Eckbreth, 1988]. Put simply, the greater the intensity of the excitation source the larger the signal. There is however an upper limit to this trend as there comes a point at which the medium through which the photon beam passes begins to break down. In the case of laser excitation, the final beam configuration is often achieved by focusing the beam down to a waist at some intermediate position. A very large electric field is produced at the point where the beam converges that can result in a spark being produced [Bombach, 2002]. Should this occur in the region of interest it may 34 produce erroneous results as well as potentially damaging the optics and experimental apparatus. 2.7 Signal-to-Noise Ratio (SNR) A discussion of Raman spectroscopy is not complete without consideration being given to sources of experimental noise and to the measures that can be taken to reduce or eliminate them. Experimental noise is of such significance in Raman spectroscopy because the Raman signal is relatively weak. It is thus necessary to have a good understanding of possible noise sources and have strategies with which to address them. Stevenson [Stevenson et al., 1996] splits the total noise into three categories, Equation 3.8: sbdtot NNNN ++= (2.5) Where: Ntot = total noise Nd = dark noise Nb = blank noise Ns = signal noise The dark noise is the noise recorded when no light or signal is falling on the detector. This noise is therefore purely a function of the detector (electronic noise in the camera in the case of a charged coupled device (CCD)). Although dark noise can be quantified it cannot be overcome with improved experimental design or procedures. 35 Blank noise, a term more often associated with spectroscopy of chemical solutions held within a sample container, is the noise produced by the container itself. In engineering applications of Raman spectroscopy this is the additional signal created by the confines of the experiment; be it the combustion chamber of a turbine or reciprocating internal combustion engine or the interior of a catalytic reformer. Due to the weak nature of the Raman signal, any spurious sources of electromagnetic radiation that enter the spectrometer can overwhelm the signal and cause all useful information to become lost in noise. It is therefore important to limit this noise, a large source of which is reflected, unshifted incident laser light. Reflected laser light can be reduced by the use of beam collectors or ‘dumps’ which are employed to capture laser light once it has passed through the region of interest. However, if such devices are not permissible it becomes necessary to filter the Raman signal before it enters the spectrometer. A more detailed evaluation and analysis of spectral filters is provided in Chapter 4 describing the experimental design and procedure (§4.2.1). Another potential source of potential blank noise is the optical windows through which the Raman signal must pass. Again, reflected radiation (typically of a far greater intensity than that of the Raman signal) is the largest potential noise source, reinforcing the need for beam capture where applicable and/or efficient spectral filtering. In addition, some window materials themselves can be caused to fluoresce by the excitation source. Again careful attention should be paid to the type of material used in optical windows for Raman spectroscopy applications. Signal noise is the result of the medium being measured. This is of critical concern to experimentalists observing combustion phenomena because of the noisy nature of the combustion process. The many and rapid chemical reactions taking place are a potential source of chemiluminescence; when through the course of the chemical reactions molecules become electronically excited and spontaneously emit a photon through fluorescence. This is referred to as luminosity and can only be reduced by 36 selecting an excitation source which produces a signal that is sufficiently strong and spectrally discrete from background luminosities. Similarly, for the combustion researcher, particulate incandescence is an important consideration. Particle incandescence is where particulates, already at an elevated temperature due to the combustion process are elevated still further by the energy imparted to them by the laser excitation source. At these high temperatures black/grey body radiation can be of sufficiently large intensity to be a significant noise source. However it should be stressed that these particular phenomena are of main concern to experimentalists studying combustion. The steady state catalytic methane/steam reformer is a more conducive environment in which to use Raman measurement techniques. Particulates, primarily the product of incomplete combustion, will not be present with respect to the reformation and gas shift reactions. Any solid carbon that is produced, due to insufficient methane/steam ratios, will likely not be produced in the gaseous flow field [Hecht et al., 2005]. The other main contributor to signal noise is that of the optical collection system. This is distinct from the window fluorescence mentioned earlier. This is primarily because optical collection systems collect light in front of and behind the depth of field of the system [Eckbreth 1988]. This source of noise can be minimised by tuning the collection optics and selecting lenses which correctly match the requirements of the spectrometer being used. The excitation beam orientation with respect to the optical axis of the collection system needs to be carefully considered, §3.2. It is also possible to suppress the amount of spurious light produced by polarizing the excitation beam, §4.2.2, and utilising the polarized quality of the Raman signal when viewed orthogonally with respect to the excitation beam [Hecht, 1997]. A full description of the collection optics and their effects on the efficiency of the collection system can be found in Chapter 5. 37 From the above breakdown of noise sources it can be seen that many obstacles stand in the way of detecting the comparatively weak Raman signal. There is however one final area in which signal detection can be enhanced and this is signal averaging. The methane/steam reformer is particularly well suited to using averaging methods due to the steady state nature of the process. It is possible to obtain more accurate measurements by combining the signals of several excitation events, §6.5, in conjunction with background noise subtraction, §7.1. Several weak Raman signals can be combined to produce a representative sample from which a more accurate assessment of the species concentrations can be made. The signal averaging technique is of particular significance for regions with low concentrations of the species of interest. A detailed description of the optical processing procedure is found in §7.1. 2.8 Radiation Thermometry Applied to the Surface of a SOFC Cell Before any discussion can take place into the principles and practicalities of radiation thermometry the application in which the technique is to be used must first be understood. In this instance, Infrared Radiation (IR) thermometry is to be used to measure the temperature of a SOFC anode. Figure 2.3 shows the three main layers of a SOFC “cell”. The first layer is the cathode and it is in this layer that the oxygen atoms combine with electrons to produce the oxygen ions which go on to pass through the electrolyte. This process takes place at the triple phase boundary, with these regions existing on both electrodes. The triple phase boundary is named as such because it is the region where the gas phase, electronic conductor and ionic conductor converge. In the case of the cathode, the reactions taking place at the triple phase boundary are rate limiting. Therefore, the electrode is porous - to not only allow the diffusion of gas - but also to increase the number of potential triple phase sites [Singhal et al., 2003]. 38 Figure 2.3 Photograph showing cross section of SOFC anode supported cell from presentation given by B. Borglum, Versa Power Systems: “Development of Solid Oxide Fuel Cells at Versa Power Systems”, 2009 Fuel Cell Seminar, 17 November 2009, Palm Springs, CA. The next layer labelled in Figure 2.3 is the electrolyte and it is a dense, non porous and relatively thin layer of the cell (in the order of 10μm). The electrolyte for this study is constructed from yttria stabilised zirconia (YSZ) and its purpose is to be impermeable to gas while the ceramic lattice allows the movement of oxygen ions when at a suitably elevated temperature. As discussed in Chapter 2, the electrolyte thickness in a medium temperature SOFC is kept as thin as possible so as to reduce ionic resistance to oxygen ions. This reduction in thickness necessitates that the electrolyte is supported by a secondary structure; in this case, the anode. In an anode supported cell, the anode layer is typically the thickest layer and therefore the layer that provides the necessary mechanical robustness to the structure. The anode used for this research is constructed of a nickel cermet; a mixture of nickel Cathode Electrolyte Anode 39 and YSZ. Like the cathode the construction is porous to allow gas diffusion, increase the availability of triple phase boundary sites and provide catalytic sites on which the adsorption reformation reactions can take place [Wilson et al., 2006]. It should however be understood that the nickel and zirconia are purely mixed and in no way chemically bonded. This is of importance as it means that the anode cannot be classified as either a metallic conductor or a dielectric (electrically insulating non-metal) and the significance of this will be discussed further in §2.10 and §2.11. It should also be noted that surface finish and indeed ‘roughness’ of the anode is dissimilar to metallic surfaces typically found in other engineering applications such as those found when a component has been machined, sand cast or moulded. With this description of the physical nature of a SOFC cell (more specifically the anode) it is now possible to examine the potential challenges of implementing IR thermometry measurements. 2.9 Radiation Models The basis of most surface radiation models is black body radiation, as described by Plank’s Law, which states that the spectral radiance of a black body can be determined only by the temperature of the body’s surface and the wavelength at which the body is being observed [Nutter, 1985]: ( ) ⎟⎠ ⎞⎜⎝ ⎛ − = 1 , 25 1 , T cb e cTL λ λ λ λ (W . m-2 . sr-1 . μm-1) (2.6) Where: hcc o 2 1 2= , 1.191062 x 108 (W . µm4 . m-2 . sr-1) 40 khcc o /2 = , 1.438786 x 104 (µm . K) co = velocity of light in a vacuum, 2.9979 x 108 (m . s-1) h = Plank’s constant 66.6261 x 10-34 (J . s) k = Boltzmann constant, 1.3087 x 10-23 (J . K-1) Grey body radiation can be described as being equivalent to a body filled with black body radiation, part of which is reflected by the surface and reabsorbed into the interior of the body. The derivation can be found in the excellent review paper of Nutter [Nutter, 1985]. ( ) ( )λλ rE −= 1 (2.7) Where: ( )λE = emitted radiation ( )λr = reflected radiation Thus, the emissivity of an opaque grey body can be described in terms of its reflectivity. In addition, the emissivity of a body is a function of the wavelength at which it is being observed. Crucially however, only when the body is of appreciable thickness can it be described as opaque and behave as if it were filled with black body radiation. As stated in [Siegel et al., 1992], “Surface radiation properties, especially in the case of metals, are normally determined in the first few wavelengths below the surface. For visible and near infrared radiation, which generally dominate radiation heat transfer, the peak wavelength is in the order of 1.0 to 10.0 μm. It follows that for wavelengths of this order of magnitude the radiation surface properties for metals are determined within the first 0.01 mm of depth”. 41 The key part of Siegel’s statement is that the metallic material properties with respect to radiation are determined within 0.01 mm of the surface when being observed in the visible/near infrared portion of the electromagnetic spectrum. In the case of a SOFC electrode this is important due to the presence of the interface below the surface of the electrode between the electrode and electrolyte. Whilst it is without question that electrodes are thin, they are still typically between one and two orders of magnitude thicker than 0.01 mm as anything less than that would reduce the availability of triple phase boundaries as previously discussed in §1.2.3. 2.10 Surface Roughness Surface roughness has a significant effect on the apparent emissivity of a material. A sample with a rough surface will appear more emissive than a perfectly smooth sample of the same material. At thermal equilibrium it can be seen that absorption is equal to emissivity. A rough, opaque surface that does not absorb all the incident radiation that falls upon it can then only reflect the radiation away. An irregularly rough surface is however unlikely to perfectly reflect the radiation away but rather directs a portion of it back onto itself; thus allowing greater absorption of the overall incident radiation and leading to an increased emissivity. Much work has been done to quantify these effects, as indicated by both Nutter and Siegel and co-workers [Nutter, 1982; Siegel et al., 1992] and yet no single model exists that can completely characterise them. Significant variables are; the randomness of the surface roughness, the magnitude of the irregularity, the average slope of the irregularity, the magnitude of the irregularity of the surface in relation to the wavelength of the incident radiation, the surface finish of the material, the angle at which the radiation from the surface is observed, and again whether the surface being observed is 42 that of a dielectric or metal. When considering all these variables it can be seen that the porous, highly irregular and mixed composition of the SOFC anode makes the use of a model to predict emissivity ill advised. Whilst there have been studies on the emissivity of zirconia [Tanaka et al., 2001; Ferriere et al., 2000; Wilson, 1965] and nickel [Birkebak et al., 1965; Toyloukian et al., 1970] as far as this author can ascertain no studies have been carried out on surfaces equivalent to that of a SOFC anode. That is to say that emissivity values have not been determined at the same nickel/zirconia mixture ratio, microstructure (porosity and zirconia/nickel particle size), observation temperature, and wavelength. It is clear that all of the above parameters and characteristics could be measured and incorporated in a model to predict emissivity. Similarly the nickel/zirconia mixture ratio could be varied and empirically related to the observed emissivity. But in all likelihood this would promote a false understanding of the variability of emissivity and not be sufficiently robust if additional parameters were changed (e.g., temperature, microstructure). Calibration of the IR thermometer would therefore seem a more efficient and scientifically valid process than trying to predict the effects of surface roughness by using an inadequate model for a very complex engineering problem. 2.11 Directional Emissivity As mentioned in the previous section, the angle at which the surface being measured is viewed has an effect on the observed emissivity. For the interested reader previous work has been done on characterising the behaviour of dielectrics and conductors with different surface roughness parameters and optical constants [Keegan et al., 1965; Spangenberg et al., 1965]. To provide a complete analysis of these studies is beyond the scope of the present thesis but summarizing these works allows the following simple conclusions to be made. When considering perfectly smooth metal and 43 dielectric surfaces, emissivity typically has a dependence on the angle at which the surface is observed. However the relationship between angle of observation and emissivity is not identical for both metals and dielectrics. For dielectrics the maximum emissivity is observed normal to the surface, with the minimum (approaching zero) observed at a tangent. Metals are significantly emissive when viewed normal to the surface but emissivity increases to a maximum between 75-80° off normal approaching zero at a tangent as with a dielectric [Nutter, 1982; Siegel et al., 1992]. Whilst these observations are perfectly true from a theoretical standpoint, there are problems applying them practically. Firstly, the dependence of emissivity on the angle of observation is given with respect to perfectly smooth surfaces, whereas the surface of a SOFC electrode is porous and highly irregular. Secondly, the anode can neither be described as a metal nor a dielectric. Assuming the emissivity of the discontinuous metal and dielectric portions of the anode still behaved as ideally smooth surfaces it could be argued that an observation angle should be chosen that is a compromise between 75-80° off normal for the metal and normal for the dielectric. However, as the emissivity of a dielectric and metal both approach zero at 90° and in the case of the dielectric decrease rapidly beyond 80°, small changes in angle of observation could result in significant changes in observed emissivity. The angle of observation is also limited by practical considerations in the context of what is achievable within the constraints of the test rig. As will be discussed in Chapter 3, limitations exist as to the amount of optical accessibility that can be introduced into the heater body. Similarly, the optical requirements of Raman spectroscopy strongly dictate the window orientations that can be practically achieved. The effects of windows on the emitted radiation must also be considered with respect to absorbance, reflectance and refraction. To mitigate these effects it would seem most prudent to make all observations normal to the surface of the window(s) and sample. 44 This would minimise absorbance by minimising path length through the windows, reaction gases and heater atmosphere (see §2.12), utilise the angle of maximum transmissivity of the windows, minimise reflectance, and reduce spatial measurement errors by eliminating refraction. Measurements of spectral radiance made normal to the sample surface are therefore considered the best compromise and will implicitly contain contributions from both the metallic and dielectric components of the surface as neither can be considered zero at this angle. At the same time experimental errors will be minimised. 2.12 Effects of Ambient Temperature, Pressure and Gas Composition Until this point the discussion has focused on the possible variations in IR signal due to the surface properties of the observed material and the angle at which the observations are being made. Attention must now turn to the environment through which the radiation has to travel before reaching the detector head. In the case of an optically accessible SOFC/reformer the IR radiation, once radiated from the electrode surface, must pass through the following: the reacting gas mixture in the stack/test section, the test section window, the ambient atmosphere of the heater (in which the SOFC/reformer test section resides), the heater window, and finally the ambient atmosphere in which the IR detector head sits, Figure 2.4. 45 Figure 2.4 Schematic showing the pathway of radiation focused onto the detector head from the surface of interest. Each feature through which the radiation passes will have an effect on the intensity of radiation which falls upon the detector head. The two windows for example will have a certain level of transmissivity, the value of which can theoretically be obtained from the manufacturer. What is less clear however is whether this transmissivity is a function of temperature; with the test section window approaching the temperature of the surface of interest while the heater window is at a substantially lower temperature due to its exposure to ambient laboratory conditions? Similarly, the gaseous medium of the test section, heater body and laboratory atmosphere will each respectively absorb a fraction of the radiated IR radiation. This phenomenon has long been understood and is described by the Beer-Lambert law otherwise know simply as Beer’s law. This states that the absorption of a medium is equal to the molar absorptivity (ε) multiplied by the Laboratory ambient atmosphere Heater ambient atmosphere Reacting gas mixture Heater window Test section window 46 molar concentration (b) of the medium itself and the path length through the medium (c), Equation 2.8. ( ) bcA ελ = (2.8) At first glance it would seem feasible to calculate the respective absorbencies of each portion of the radiations path to the detector head and to correct the signal accordingly. However, whilst it is possible to determine path lengths and at the very least provide bounds for expected molar concentrations, the required molar absorptivities are, to the author’s knowledge, not available. The unavailability of this information is due, in part, to the wavelength at which it is required – in the work presented here the infrared region over which the IR thermometer germanium sulphide photo-detector collects radiation from 1.8 μm and 3.0 μm, while the available data from literature is concentrated in the visible and UV portions of the electromagnetic spectrum. In part, the lack of appropriate data is also influenced by the author’s species of interest. Molar absorptivity data for inorganic species such as nitrogen, hydrogen, water vapour and simple organic species such as methane and oxides of carbon are rare [CRC Handbook of Chemistry and Physics, NIST WebBook]. The use of Beer’s Law for the correction of IR temperature measurements in the present application is further complicated by the nature of the reformation process itself. For Beer’s law to be applicable a number of criteria have to be fulfilled and these are: the incident radiation must consist of parallel rays, the incident radiation should ideally be monochromatic, and the absorbing medium must be homogeneously distributed throughout the interaction volume [Skoog et al., 1982]. It can clearly be seen that the rays are not parallel and the emitted radiation is not monochromatic. Perhaps more significantly, due to the consumption and evolution of reaction gases over the length of 47 the test section (methane, steam, carbon monoxide, carbon dioxide and hydrogen, from Equations 1.3 and 1.4) the adsorption species cannot be described as homogeneous. In order to assess the potential difficulties posed by the reformation reaction, a Beer’s Law calculation was performed to approximate the effects of absorption given a 6 mm high channel containing carbon dioxide at a mole fraction of 0.2 (see Appendix C). The effect of this concentration of carbon dioxide was an absorption of 5% of the radiation passing through it, which while not an insignificant amount, should be considered in relation to the other constituents of the channel gas stream. Nevertheless, should high levels of accuracy be required from IR temperature measurements during the process of reformation the author suggests that the effect of the changing gas composition in the channel be considered. From the above, it can be seen that the absorptive effects of the channel and heater atmospheres will cause the radiation collected to appear to have been emitted from a grey body. It can also be see that in order to maximise the IR signal, care must be taken to select a window material that exhibits a high transmittance to IR wavelengths. Sapphire is often the material of choice, since despite its high cost it offers excellent resistance to elevated temperature and allows approximately 85% of incident radiation to pass through it between 0.3 μm and 3.0 μm whereas traditional optical materials such as fused silica suffer reduced transmissivity in the IR region [Musikant, 1985]. Of course, the overall transmissivity of sapphire is still dependant on path length and this is considered in the experimental apparatus design as presented in Chapter 3. 2.13 Application of IR Thermometry to Optically Accessible Reformer The previous sections of this chapter have shown that the surface of an SOFC anode represents a suitable environment on which to perform IR thermometry. What is 48 more, with the temperature variation over the anode surface predicted in certain instances to be in the range of 50 – 100 °C [Kang et al., 2009; Achenbach, 1994a], the potential benefits of measuring surface temperature with respect to greater insight into in- channel processes and the quantification of Raman measurements are clear. It has also been shown that the absolute temperature values reported by an IR thermometer are affected by numerous variables relating to the material properties of the surface and the medium through which the radiation passes that are, in the case of the author’s set- up, either unknown or are not readily quantifiable. However, it should be recognised that the benefit of an experimental technique which can measure the magnitude of the temperature variation is of equal importance to the ability to measure the temperature accurately. Should high accuracy be required then it has been shown that it would be more efficient to calibrate out the effect of the unknown variables rather than model their effects. To attempt to model the effects would be undoubtedly problematic and would likely result in the user having false confidence in the accuracy of the temperatures recorded. Calibration experiments should therefore be performed with the IR thermometer using a reference temperature measurement, such as thermocouples (which would not necessarily have to be on an active part of the electrode). Absolute temperature measurements from IR thermometers should only be reported in the presence of a robust calibration procedure. Calibration was not able to be performed in the work presented here, for reasons discussed in §7.17. 49 3 Optical Reformer Design This chapter narrates the major design considerations that took place during the design, manufacture and commissioning of the optical reformer rig. This equipment was produced specifically for the work presented in this thesis and was designed for the specific needs of the experimental techniques of Raman spectroscopy and infra red radiation thermometry (see Chapter 2). The author was aware from the beginning of the project that the application of Raman spectroscopy in the challenging environment of a fuel cell reformer was a non-trivial task and would therefore require several design iterations before a feasible, repeatable experimental technique could be achieved. Accordingly, the initial design of the apparatus was intended to be as simple and flexible as possible so as to accommodate the expected advances in experimental technique made through the course of the research. What follows in this chapter is then a step-by- step discussion in the requirements of the optical techniques and how this influenced the design of both the reforming section and heater. The reformation reaction itself is examined and considered from the perspective of an operational SOFC and consideration is given as to how best to reproduce this environment without an equivalent gas supply and the heat generated by the fuel cell reaction. The practical implications of introducing optically transmissive windows into the reformer and heater environment given the widely differing thermal coefficients of expansion of the window and stainless steel reformer body are examined. The chapter continues with a description of the main issues encountered when high temperature testing was first attempted in the rig. Finally, the author’s solutions to these issues, which ultimately enabled the optically accessible reformer test section to produce sustainable, repeatable reformation reactions, are discussed. 50 3.1 The Design Task With Raman spectroscopy and IR thermometry selected as the diagnostic techniques of choice for the study of methane/steam reformation, the challenge was to design a suitable experimental apparatus that would allow their application in a suitably realistic operating environment. A full description of the final experimental layout will be presented in Chapter 6. The purpose of the following sections of this chapter is to detail and discuss the major factors that were considered during the design process in order to make key decisions. While neither Raman spectroscopy nor IR thermometry are new diagnostic techniques, their application to a reformer with characteristics similar to that of a SOFC is, to the author’s knowledge, a wholly original aspect of the current research. The challenge facing the author was to design an apparatus that could sustain a reformation reaction while providing appropriate optical access for the chosen experimental techniques and a stable platform from which these techniques could be performed in a repeatable manner. The author’s key design objectives can therefore be described as follows; to provide: An optically accessible reforming section A heating apparatus with optical accessibility An experimental platform that enables repeatable measurements to be made 3.2 Raman Spectroscopy Design Considerations Raman spectroscopy is an active technique. It requires an excitation source - in the research presented here a 308 nm broadband excimer laser - and light collection optics combined with a monochromator/spectrometer with which to analyse the 51 scattered light from the region of interest. Thus, from the perspective of optical accessibility, Raman spectroscopy requires paths for both ‘light in’ as well as ‘light out’ of the region of interest. There are two typical experimental set ups for Raman spectroscopy, which are distinguishable by the relative axes of the light into and out of the region of interest as shown in Figure 3.1. Figure 3.1 Coaxial and orthogonal light in arrangements relative to the axis of the collection optics, with ray paths shown for the region of interest. There are advantages and disadvantages to coaxial and orthogonal (with respect to the optical axis) ‘light in’ to the region of interest. Some factors may be considered purely from the perspective of optical efficiency and signal to noise ratios (SNR’s). These issues are covered in detail in the exhaustive text by Eckbreth [Eckbreth, 1988]. Summarising this work, Eckbreth shows that for coaxial light collection, the region of interest (ROI) – in this case the depth of field of the optical system – is completely filled with laser light, so a large amount of light can be collected from the ROI, which is a important consideration when dealing with the relatively weak Raman signal (§2.3). Depth of field Spectrometer aperture Orthogonal laser light Coaxial laser light x z ROI Plano-convex spherical lenses 52 However, a coaxial optical system also results in light being collected from infinity to approximately half the focal length of the lens closest to the region of interest. The collection of spurious laser light therefore becomes a serious issue with the coaxial arrangement. Coaxial light collection also inhibits high spatial resolution for similar reasons as contributions from Raman signals are collected over a large spatial extent. In comparison, the orthogonal arrangement’s spatial resolution is limited to the spatial width of the light source as it crosses the optical axis. Spatial and spectral resolution will be covered more extensively in the following chapter covering experimental procedure, §5.2. From a practical perspective, the coaxial system has one significant advantage over an orthogonal light source, but it also has several disadvantages. The advantage is that, theoretically, only one optical access port is required - although as mentioned above, when collection optics are collecting light out to infinity it is often advantageous to terminate the beam path using a beam dump, which may necessitate a secondary point of optical access. A significant disadvantage of the set up is that reflecting the light through 90° in front of the collection lens is, in practice, not as simple as the diagram (Figure 3.1) suggests. The difficulty of light reflection is compounded if one-dimensional Raman spectroscopy is being performed, as issues are encountered concerning the obstruction of the collection optics by the reflecting medium. The use of an orthogonal light source on the other hand decouples the light collection process from any optical treatment that might be necessary to the ‘light in’ such as shaping and polarization (§4.2.2). With the anticipated difficulty of the coaxial system and the experimental necessity of achieving good spatial resolution the orthogonal ‘light in’ set up was selected for the present work. The other major factor that needs to be considered by the experimentalist wishing to implement Raman spectroscopy (using either the coaxial or orthogonal 53 arrangement) is the arrangement of the collection optics. Figure 3.2 shows a collection optics system consisting of two plano-convex spherical lenses which image the ROI onto the slit of a monochromator/spectrometer. By imaging the Raman signal from the ROI it is possible to not only determine gaseous chemical species concentrations but also how these vary in a single spatial direction. The direction in which the spatial variation is measured depends on the orientation of the spectrometer slit itself, in the case of Figure 3.2 the y-direction. Figure 3.2 Optical arrangement for one dimensional Raman spectroscopy. One dimensional Raman spectroscopy as described above can be seen to be a valuable tool, as a single measurement has the potential to not only detect multiple species simultaneously but also it has the potential to provide information as to the distribution of these species in a single direction (subject to the ROI being excited by a monochromatic light source). However, one dimensional Raman spectroscopy adds a further complication because it takes the already weak Raman signal and distributes it over the length of the spectrometer slit - further reducing the intensity of the signal and Spectrometer λ y x y z Plano-convex spherical lenses Image produced by spectrometer Spectrometer entrance slit 54 incurring the losses associated with passing through a spectrometer. In order to maximize the efficiency of the light passing through the spectrometer the focusing lens (that which is closest to the spectrometer) must have a ƒ# that is equal (or slightly less) than that of the spectrometer. By matching these numbers the light neither over nor under fills the spectrometer and maximum throughput is achieved. Matching of the spectrometer ƒ# with that of the focusing lens has significant implications to the experimental design. If the reader returns their attention to Figure 3.1 it can be seen that if the two plano-convex lenses have equal diameters (for maximum collection efficiency) and identical focal lengths (and thus ƒ#’s) then the image projected on the spectrometer slit will be equal in size to the light collected from the ROI. If the focal length of the collection optic - the lens closest to the region of interest - has a focal length less than the focusing optic, the image projected onto the spectrometer slit will be magnified. Likewise, the reverse is true, so a longer focal length will result in a ‘minified’ image on the spectrometer slit, which could potentially result in issues with resolution. An in-depth discussion of resolution, both optically and with respect to the intensified charge coupled device capturing the spectrometer output will follow in §5.5, but for now the limitations imposed on apparatus design by the collection optic needs to be considered. In this thesis a spectrometer with a ƒ# of 4 was used, which when used in conjunction with plano-convex lenses of diameter of 50 mm (the largest ‘off the shelf’ optic available for use in the UV) dictates that a focusing optic be used that has a focal length of 200 mm. In turn, for no reduction in image size a collection optic of focal length 200 mm would have to be used implying that the ROI be 200 mm from the collection optic. These considerations provide a limiting value on the size of heating apparatus, the size of the reforming section, and go on to have additional implications for the orientation of the Raman collection axis. 55 3.3 IR Thermometry Design Considerations IR thermometry is distinct from that of Raman spectroscopy in that it is a passive technique which does not require an excitation source of electromagnetic radiation. As a passive technique the optical requirements are somewhat simpler than those of Raman spectroscopy, albeit subject to the considerations discussed in Chapter 2. The technique is sufficiently established that IR Thermometers are available “off the shelf”, and in the research presented here a thermometer was purchased from Omega Vanzetti, a division of Omega Engineering. Vanzetti offers a range of IR Thermometer configurations. For this work a configuration was selected that consisted of an optical head, a fibre optic cable, and a housing containing the photo detector, an optical chopper and the necessary electronics required to convert the optical signal into an electrical output. Significantly, the optical head of the chosen thermometer (P/N OS1562) can be configured to collect infrared radiation at a variety of focal lengths and spot sizes (the surface area over which radiation is collected). Broadly speaking smaller spot sizes are obtained by using shorter focal lengths, and as such the desired spatial resolution is the driving force behind detector head selection. However, additional factors also need to be considered, with one of these being the temperature the optical head will experience when making measurements. Typically, detector head assemblies are rated to a maximum allowable operating temperature; in the case of the Omega Vanzetti detector head this is 150 °C and it can be inferred that the heating effect will be a function of the temperature, thermal mass, and proximity to the surface being measured. In the context of a methane/steam reformer, the temperature and thermal mass of the surface being observed are to some extent variables beyond the control of the apparatus designer. This makes the selection of the focal length a function of desired spatial resolution, maximum permitted operating 56 temperature, and any physical limitations placed upon the location of the detector head by other experimental apparatus. Physical limitations at this stage are the least defined and will be the subject of further discussion in §3.6, which considers both the overall size of the heater and its size in relation to the reforming section. As will be shown in §3.5, the focal length selected was ultimately a compromise between resolution and physical constraints concerning the position of the detector. These considerations resulted in an optical head being selected that has a spot size of 0.036” (0.91 mm) and a focal length of 8” (203.2 mm). To have selected a smaller spot size would have imposed too great a restriction on the size of the optical heater and potentially required an overly complicated cooling regime to be implemented. A final requirement of IR thermometry is that the surface of interest must be optically visible to the detector head. The Omega Vanzetti IR Thermometer uses a germanium sulphide photo detector which collects IR radiation between the wavelengths of 1.8 μm and 3.0 μm. As discussed in §2.12 optical windows must therefore have good transmissivity at these wavelengths, in addition to having suitable qualities for other optical diagnostic techniques such as being able to transmit at wavelengths necessary for Raman spectroscopy. It was also noted that optical path lengths through partially transmissive materials should be minimized in order to limit signal attenuation. Beam paths that are perpendicular to the transmissive materials minimize these absorption effects. This is also seen to satisfy the requirements imposed by the directional emissivity of the anode material as previously discussed in §2.11. 3.4 Optical Requirements of the Reforming Test Section The chosen techniques of Raman Spectrometry and IR thermometry require that optical accessibility be provided perpendicular to the surface of interest (IR thermometry) 57 and that two optical paths, one orthogonal to the other, are present (one dimensional Raman spectrometry). The perpendicular path requirement is an inflexible constraint which dictates that an optically transmissive material be placed at some distance parallel to the anode surface. This transmissive material, or window, must allow the passage of IR radiation out of the test section while at the same time making up some part or all of the roof of the gas channel above the anode. As will be discussed shortly, the very act of introducing a window into the extreme environment of a reformer is not without its challenges and, as such, the number of instances of this having to occur should be minimised. It is therefore advantageous to combine the optical window necessitated by IR thermometry with that required by one of the optical paths of Raman spectroscopy. With one optical axis specified (the IR/primary Raman optical axis) it becomes necessary to evaluate the position of the second axis, Figure 3.3. Figure 3.3 IR/primary Raman optical axis shown with potential secondary Raman axis in relation to anode surface and arbitrary gas flow direction. IR / primary Raman optical axis Potential secondary Raman optical axes Arbitrary gas flow direction Anode surface x y z 58 In order to determine the most advantageous secondary Raman axis it is necessary to consider additional information such as the nature of the reactions occurring over the anode surface. The most important of which is the reformation reaction (Equation 1.3) occurring between the point at which the gas stream comes into contact with the reactive surface and according to Achenbach and co workers, a short distance downstream from the leading edge of the anode surface [Achenbach et al., 1994b]. Also of interest are the subsequent gas shift reaction (Equation 1.2), the potential Boudard reaction (Equation 1.5) and reduction of CO (Equation 1.6) which occur primarily downstream of the reformation reaction zone. The anode surface over which these reactions take place must be able to support these reactions and enable the diagnostic techniques to measure the progress of these reactions as accurately and insightfully as possible. As such, it is the axis along the direction of flow of the reformer that is of most interest to this research. Similarly, the spatial distributions of gaseous species concentrations perpendicular to the reacting surface are also of interest. The second optical axis is therefore dependant on the orientation and size of the anode surface itself while at the same time, as was discussed in §3.2, the axis must permit the use of a collection optic that at the very least has a focal length similar to that of the focusing optic. With these parameters identified the practical factors affecting the reforming section design can be considered. 3.5 Practical Requirements of the Reforming Test Section The focus of the design narrative until this point has been on the requirements of the experimental techniques; Raman spectroscopy and IR thermometry. From the perspective of a practical scientific apparatus for the proposed research other factors must be considered. Crucially, a sustainable reformation reaction must be generated 59 over the reformer surface. Without the electrochemical reaction of an actual fuel cell to thermally sustain this reaction this implies that the experimental apparatus must incorporate an external heating source. There are also practical considerations as to how the reformation reaction is supplied with fuel. In order to ensure that the experimental measurements are of maximum relevance the gaseous fuel flow field must be reasonably similar to that of a planar solid oxide fuel cell (SOFC). Finally, some consideration must be given to the fact that the experimental apparatus is in itself experimental and will therefore be subject to many cycles of heating, and assembly and disassembly, which as previously stated is in stark contrast to pilot scale SOFC’s which are typically single use items. As previously mentioned, the chemical reactions occurring in a methane/steam reformer, occur predominantly in the direction of flow of the gases over the anode surface. These reactions are of primary interest. Of secondary interest is the spatial distribution of gaseous chemical species perpendicular to the anode surface. Variations across the anode surface, perpendicular to the direction of gas flow are of less interest to the work presented here as there are unlikely to be any significant asymmetries in temperature and species concentrations. This is in contrast to those asymmetries predicted to exist in cross flow SOFC arrangements. The one dimensional nature of the flow across the anode surface has implications to the experimental design when considering what size methane/steam reformer to implement. For the research presented in this thesis the SOFC anode supported half cell samples were donated by Versa Power Systems who at the time of the apparatus design were developing SOFC stacks using individual cells measuring 100×100 mm2. However, since the second dimension is not expected to yield any significant information of interest, it was not considered necessary to produce a reforming section of this size for this research. For this reason, combined with the expense of producing bespoke half cell samples, a 60 standard 100×100 mm2 half cell was cut into two. This results in a 50×100 mm2 sample over which to host the reformation reactions and which still allowed a 100 mm run over which to observe gaseous chemical species distributions similar to those found in a full size SOFC. Dividing the original half cell samples in two has further benefits for the design of the experimental apparatus; reducing the half cell size results in a smaller thermal mass that has to be brought to test point temperatures and a smaller heating enclosure is required which, as was discussed in §3.3, significantly simplifies the execution of IR thermometry by allowing shorter focal lengths, smaller spot sizes and improved resolution. A reduction in half cell surface area also results in smaller overall gas flow rates and, in turn, a reduction in humidification capacity, which significantly reduces experimental cost. With the reactive area of the reformer now fixed it was necessary to give consideration to the nature of the gas flow over the surface. Typically in a SOFC, the gases are fed over the respective electrodes from manifolds that are equal in width to that of the cell. This system is used so that the reformation and fuel cell electrochemical reactions occur as uniformly as possible across the reactive surface and catalysts are utilized equally. Basic fluid dynamics theory states that viscous fluid flow in a channel will assume a velocity profile with a mean stream velocity, a maximum velocity on the stream centreline, and a minimum velocity at the wall. It is therefore impossible to achieve a truly uniform flow field across the reactive surface. However, as mentioned previously, so as to ensure the maximum relevance of the experimental results, the design of the experimental apparatus must provide a flow field as similar as possible to that found in a fully functioning SOFC. It is therefore insufficient to simply provide a single circular inlet and outlet for the gas flow, as this produces a flow field that bears no similarity to that of a functioning SOFC. However, as will be shown shortly, other factors 61 that affect the gas routing for an optically accessible reformer section must be considered before the manifolding arrangements for the channel can be specified. 3.6 Optically Accessible Heater As mentioned in the previous section, without the electrochemical reaction of a SOFC present to sustain the endothermic reformation reaction it is necessary to provide the reformer rig with an external heating source. In fact, high temperature heaters are common in medium/high temperature fuel cell research as the SOFC reaction is not self initiating. This is because the ion conducting ceramic electrolyte present in SOFC’s only exhibits this property at elevated temperatures. Radiant heaters are one of the simplest heating solutions and have become the industry standard heating solution for this application. The radiant heaters used in SOFC test stations comprise of a heating coil arranged in a serpentine pattern embedded in an insulating material. When a current is passed through the coil they heat up and emit IR radiation. Two typical heater arrangements are used; a box heater and a cylindrical heater. A box heater is typically a five sided box made up of series of flat radiant heating panels. The box is then lowered over the component to be heated which either sits on a floor of refractory bricks or similar insulating material, with gases/monitoring equipment routed up to the component from below. Alternatively, two semi-circular ceramic heating elements are combined to make up a cylinder, in the centre of which is placed the object to be heated. The second arrangement is more commonly used because it allows easy access to the inner portion of the heater, by way of a clamshell arrangement, and, what is more, the radiation is more evenly distributed over the object being heated. 62 Figure 3.4 Schematic of horizontal and vertical clam shell heater arrangements Figure 3.4 shows a schematic of two clam shell heater arrangements with arrows showing the relative motion of the two hinged semicircular heating components when accessing the inner portion of the heater during test set up or maintenance. The cylindrical heating arrangement is useful because either end of the cylinder can be used to route gas lines or monitoring equipment, such as thermocouple probes. Care must however be taken to block these ends with insulation during testing otherwise an effect called ‘chimneying’ occurs, where lower density hot air leaves the hot zone to be replaced by cold air. This not only increases the energy necessary for heating but can also result in non-uniformity of heating of the component. It should be noted that neither the box nor cylindrical heating arrangements immediately lend themselves to optical techniques that use orthogonal ‘light in’ techniques; as the secondary axis cannot immediately be accommodated. Therefore, both the box and cylinder radiant heating methods would have to be modified to incorporate the secondary optical axis required by the author. In fact the box arrangement would have to be modified two fold as it cannot easily incorporate even a single optical axis, whereas at least one end of the cylinder could potentially allow an B Hinge Hinge A 63 unimpeded light path. To modify either heating arrangement to allow optical access requires the opening of a gap somewhere in one or more of the heating panels. While this is not impossible for the box arrangement, it would require either a bespoke heating panel(s) to be manufactured or the creation of a box side out of a ‘patchwork’ of heating panels around an optical access port. This box heater approach was discounted for several reasons; the first being that multiple panels could result in difficulties in temperature regulation, with different panels having different electrical loading characteristics. Furthermore, the effect of removing a section from a flat panel has a more pronounced effect on the radiant energy falling upon the component to be heated (an effect that is less pronounced in the cylindrical arrangement as will be shown shortly). Finally, the inherent action of lowering the heating box over the component of interest has ramifications for the repeatability and stability of optical components placed around the heater. As has been touched upon and will be discussed further, it is critical to have a physically stable platform on which to mount optical components in a repeatable manner. Having to move optics every time access to the optical reformer is required introduces additional variables into the position of the optics, which in turn has ramifications for repeatability and potentially spatial resolution and accuracy. These issues are by no means insurmountable, but are more easily overcome using a cylindrical heater arrangement. Based upon the above considerations a cylindrical heating arrangement was chosen as a practical starting point on which to base an optical heater. One optical axis was then chosen to be the cylindrical axis of the original heater. The second orthogonal optical axis required for the author’s work was achieved by spacing the two half ‘shells’ a small distance from one another, as shown in Figure 3.5. 64 Figure 3.5 Spacing of semicircular ceramic radiant heaters to allow second optical axis A cylindrical heating arrangement has another advantage over that of a box type heater. As shown earlier in Figure 3.4, the orientation of the cylindrical heater’s axis could feasibly be either horizontal or vertical; however, an important aspect of heater design is to ensure that the size of the object being heated is matched to the size of the heater. This is to ensure that the heater does ‘hunt’ for a temperature set point by overshooting the required temperature due to the components thermal mass being considerably smaller than the heating power available. As such, with the reformer size selected as 50 ×100 mm the heater would have to have a minimum internal radius of 2.5” (≈ 64 mm) without consideration to the physical size of the reformer channel itself. The length of the heater would then have to be sufficient to incorporate the height of the reformer channel, which, as a first estimate would be approximately 25 mm. A cylindrical heater with these relative dimensions is atypical and, to the author’s knowledge, not available. A heater with a horizontal orientation overcomes this problem by allowing stock cylindrical heaters to be used with internal radii of 3” (≈ 76 mm) and a standard length of 10” (≈ 254 mm) to be used which once again simplifies construction and reduces costs. Optical axis 65 Assuming that a horizontal cylindrical heater arrangement is selected and that a second optical axis is provided by separating the two ‘half shells’, it remains for the orientation of the spacing between the two heaters to be specified. It is clear from Figure 3.5 that the spacing could also be either horizontal or vertical (or indeed positioned at any other chosen angle to the horizontal). The advantage of one orientation over the other is one of practicality when considering the reforming surface. As will be discussed in §3.7 the reforming surface is being used outside its typical operating environment, that is to say without a cathode layer applied to one side and without being bonded to adjacent interconnects and so on. Also, as has already been discussed, the reforming section must allow easy disassembly. This meant that it was not ideal to mechanically attach the reforming surface to the test section channel, not least because of the risk of over constraining the fragile component. As the apparatus is not subject to any vibrations it was therefore sufficient to primarily locate the reforming surface in the bottom of the reforming channel using the force of gravity as the seating force. A horizontal reforming surface therefore required a vertical gap between the two semicircular heaters, Figure 3.6. In the present work, this gap will define the primary optical axes of both the Raman and IR measurement techniques. 66 Figure 3.6 Schematic of optical axes orientations achieved through spaced semicircular radiant heaters The cylindrical heating arrangement also dictates the direction of orientation of the secondary Raman optical axis, which having defined the primary axis must now follow the axis of the cylindrical heater. The alternative would be to modify the ceramic heater, which would incur the difficulties and expense previously discussed with flat panel modification. It should also be noted that the cylindrical heater arrangement enables the distance between the reforming surface and the top of the heater to be less than the 8” focal length of the IR thermometer. More significantly it can be seen that the distance between the front of the cylindrical heater and the front of the reformer is minimal, which is critical when selecting the focal length of the collecting lens for the Raman signal collection system, with the corresponding effect on spatial resolution. One feature of the cylindrical heater configuration that could not be carried over to the optically accessible arrangement was that of hinged access to the heating volume. There were several reasons for this; one was that any articulation of the heaters would have interfered with the platform on which optics were to located as well as inhibiting the placement of additional insulation placed around the heater body. In addition, with the subsequent gas routing and reforming section support structure (see §3.10) there was Optical axis Optical axes 67 no advantage to maintaining hinged access. In fact, the loss of this articulation simplified heater construction and reduced movement around the optical window situated in the top of the heater. 3.7 Gas Flow in the Optically Accessible Reformer With the main physical characteristics of the reforming surface, optical heater and optical axes determined, it was possible to design the optically accessible reformer. As described previously the design was now required to incorporate two optical windows, provide a suitable environment for the reforming surface, generate a gaseous flow field similar to that found in a fully functional SOFC, and allow gases to be routed into and out of the heater in a practical manner. The properties of the flow over the reforming surface have already been touched upon (§3.5), but critically it was required that the desired flow profile be achieved in as short a channel length as possible so as not to make the reforming channel too long, or more specifically, of such a length so as not to place the far portion of the reforming surface beyond the focal length of the collecting optic of the light collection system. It can be calculated (Appendix A) that a 50×100 mm2 cell will consume up to 0.2 SLPM of methane gas, with the largest ratio of steam to methane resulting in approximately 1.0 SLPM of gas flowing over the reforming section in the most extreme case. If the volume flow rate of 1.0 SLPM were passed through a standard ¼” tube, inner diameter 3.3 mm (a standard tube size, readily available, and with an abundance of easily sourced fittings), at 800°C the mean stream velocity would be 7.2 ms-1. When this is compared to the mean stream velocity for the same volume flow rate passing through a duct of width 50 mm and height 7 mm, the mean stream velocity would be 0.2 ms-1. This comparison of mean stream velocities suggests, as would be expected, that there would be a transitional region as the flow 68 exits the ¼” tube and fully develops into that found in a SOFC/reformer channel. A FLUENT simulation of a channel with single ¼” inlet and outlet ports supported this supposition and predicted a velocity profile that developed as it passed over the reforming surface, as well as gas velocities in directions other than the principal axis of the reformer channel, Figure 3.7 (detailed results, Appendix B). Figure 3.7 Schematic of predominant inlet and outlet velocity directions for channel with a single inlet and outlet Fluid dynamics theory also suggests that given a sufficiently long channel a fully developed flow will develop but this necessitates regions before and after the reformer surface for the flow to transition from the localised flow fields around the single inlet/outlet ports. This places the reforming surface further away from the collecting optic of the Raman collection system and therefore an alternate method of supplying and removing gas from the channel was sought to avoid this problem. It is also worth noting that the seeking of a fully developed flow field over the reforming surface is motivated by more than just wanting to make the flow as similar to that of a SOFC electrode. A ‘uniform’ flow field, that changes as little as possible over the length of the reformer Inlet Outlet Plan view of channel 133 mm 54 mm 69 surface also makes any scientific information generated of wider interest to SOFC developers and numerical modellers as the results will depend less on specific channel geometries and be more generic for one dimensional flow patterns. As previously stated a typical SOFC has gases fed onto either side of the cell via manifolds. However, for a manifold to work effectively is must be carefully designed to develop the flow from the inlet state to a desired velocity profile, be that by means of a diffuser, baffles or other such device. When these techniques are considered for the optical reformer they either reintroduce the problem of increasing the separation between the reacting surface and collecting optic, produce a diffuser/channel assembly that requires a larger heater (with potential ramifications for IR Thermometer focal lengths) or obstruct one of the optical axis of the Raman spectroscopy technique. One gas delivery method that was considered was that of a plenum chamber, from which a series of passages lead into the reformer channel floor at a minimum practical distance in front of the reacting surface, Figure 3.8. 70 Figure 3.8 Reformer channel with plenum chambers and inlet/outlet passages One crucial difference between the plenum chamber and multiple inlets to that of a single inlet is that the combined cross sectional area of the former was greater than that of the latter, thus reducing localised velocities. In addition the process of adding and removing fluid from the channel was more evenly distributed in a spatial sense when compared to the single inlet and outlet configuration. The plenum chamber combined with multiple inlets/outlets resulted in greatly reduced distances being required for a quasi fully developed flow to be achieved. Again this scenario was simulated using Plenum Chambers Multiple inlets/outlets Inlet/outlet passages Gas in Gas out Plenum Chambers 71 FLUENT (see Appendix B) supporting the theory behind the design intent. While it was not a sensible use of resources to experimentally validate this result it would appear, at the very least, that this approach has significant advantages over a single inlet/outlet port arrangement. A factor which lends credence to the simulation results is that the flow inside SOFC’s, and fuel cells in general is generally considered to be laminar, as supported by Reynold’s number calculations (see Appendix A). This increases confidence in the simulation results as the largest uncertainties in purely computational fluid dynamics simulations (as opposed to reacting flows and diffusion mechanisms) typically results from the use of turbulence models. The approximate inlet and outlet configuration of the author’s reformer channel is shown in Figure 3.9. Figure 3.9 Schematic quasi fully developed flow as a product of multiple inlets and outlets outlet The multiple inlet/outlet and plenum arrangement was not only selected to produce the desired flow profile, but also as a consequence of the restricted packaging constraints of the heater. For the reforming section to be heated as uniformly as possible the reforming surface is aligned to the central axis of the cylindrical heater configuration. Inlets Outlets Plan view of channel 133 mm 54 mm 72 With the cylindrical heaters only having an internal radius of 3”, it is necessary to route gas lines to and from the reformer in this restricted space. This challenge is compounded by the issue of connecting ¼” tubing to the reformer body. The tubing system used in the construction of this apparatus was Swagelok™, which was selected because of its modularity, relatively high temperature rating and stainless steel construction which itself has a good tolerance to elevated temperatures. As will be discussed shortly, the material from which to construct the body of the reformer was chosen to be 316 stainless steel for many of the same reasons the Swagelok™ system was selected. The issue however, was that it is necessary to make a gas tight connection between the tube and reformer body. Ordinarily this would be accomplished by either using a rubber face seal or a tapered pipe thread. In this instance a rubber seal was not suitable, due to the high temperatures involved and a tapered fitting was unsuitable for a number of reasons. In a tapered fitting the seal is made by deforming the threads, however stainless steel suffers from a phenomenon referred to as ‘cold welding’ which is likely to occur when forcing two stainless steel surfaces over one another. This phenomenon can occur at very low loads and so a thread seal is not guaranteed. In addition, a tapered thread could result in a permanent connection, which could have potentially hampered any subsequent modifications to the reformer channel should they have been required. The chosen solution was to use a fitting with a straight thread and a well prepared mating surface with which to make a face seal. This arrangement has the advantages that the fittings can be removed and the thread re-cut (if necessary) should a modification to the reformer body be required in the future. However, the use of straight thread fittings further reduced the space available in which to locate the plenum chamber, particularly as there was a restricted amount of space beneath the reformer channel due to the proximity of the heaters, Figure 3.10. 73 Figure 3.10 Partial cut away of optically accessible heater showing and resultant spatial restrictions for gas tube routing As can be seen in Figures 3.11a, the combination of straight thread Swagelok™ fittings with the minimum bend radius of the ¼” stainless steel tubing necessitated the gas lines to run below the reformer in the space between the two semi-cylindrical Straight thread Swagelok™ fittings Semicircular ceramic heater Reformer channel centered in heater Restricted gas routing space 74 heaters. However, it should be noted that there was some advantage to the chosen piping arrangement as in running the pipe work in this manner allows the reformer channel to be loaded into the front of the heater without disassembly, Figures 3.11a and b. This was ultimately the design intent as it yielded the significant benefits of allowing the reformer channel to be removed from the heater without disturbing the optics surrounding the heater. (a) (b) Figures 3.11a and b. (a) Cutaway of heater showing gas routing (b) The heating element arrangement. Note that the plenum plugs had not yet been fitted and the third middle tube was that through which the control thermocouples were run. 3.8 Optical Accessibility of the Reformer Channel The most difficult aspect of the reformer channel was the integration of optical windows to allow the optical diagnostic techniques ‘access’ to the reforming channel and the anode surface. The technical challenge of providing optical access was compounded 75 by the requirement that the windows be easily removable as physical access to the channel was an ongoing requirement. It was also apparent that the widely differing thermal coefficients of expansion (TCE) of the window material (UV grade sapphire, TCE = 7.7 x10-6 m/m K) and reformer channel (316 stainless steel, TCE = 16.0 x10-6 m/m K) required the window attachment technique to incorporate this large differential in material expansion properties. Over a length of 100 mm and temperature differential of 750°C the difference in expansion is approximately 0.6 mm. More important, whatever window location technique ultimately employed, was the requirement of providing some degree of gas seal so as to promote the flow of gas over the anode material. Gas sealing is an ongoing area of research in SOFC development and as such the details of the techniques are well protected intellectual property. Frustratingly, the techniques under development, as previously discussed, are usually intended to be permanent; thus, making disassembly impossible and the sealing methods not applicable to the requirements of the optical reformer. At the same time, the author’s remit was to demonstrate the feasibility of Raman spectroscopy in a high temperature environment and as such the performance of the reformer was not considered a priority. Small gas leaks, which in a practical setting would reduce the efficiency of a reformer, were not considered to be an overriding priority in the context of developing an experimental test rig for proof of concept. Working on the premise that a 100% gas seal was not necessary at this point in the development of the optical rig and measurement techniques and in the knowledge that the high operating temperatures of the rig precluded all conventional low temperature sealing techniques such as O-rings etc. a simple sapphire-to-steel seal between two well finished surfaces was decided upon. In the case of the reformer, the sapphire windows naturally had a high quality surface finish and the stainless steel surfaces were lapped to provide a surface free from machining marks. Given the expected relative movement between the window and reformer 76 channel body (due to thermal expansion) it was noted that the ‘top’ and ‘bottom’ surfaces of the window would be the most suitable locations at which to implement face seals, Figure 3.12 (to implement a seal around the ‘perimeter’ of the window would be highly problematic due to the relative displacements occurring between room temperature and reformer operation ≈ 700 – 800 °C). It was thought that by sealing on either the top or the bottom surfaces of the window relative movement between the window and reformer body could be allowed while still maintaining some level of contact seal between the two surfaces. Figure 3.12 Approximate locations of potential window seating areas A considerable challenge was the perpendicular orientation of the windows, §3.2. While seemingly a simple requirement, providing the necessary seating surfaces for two orthogonal windows presents a considerable challenge if both optical paths were to remain significantly un-obscured. Note that maintaining as large an un-obscured viewing area as possible through the reformers ‘front’ window is of particular importance in relation to light collected from deep within the reformer channel. In this case the Top window seal area Bottom window seal area Perimeter window seal area Optically clear region of window 77 apparent solid angle of the collection optic becomes a critical factor in determining the success of an experiment, as will be shown later in §6.5. In the case of a channel of which two sides are replaced by windows, there exists a region in which both windows require a corresponding surface to be at the same location in order to maintain the surface seal regime, Figure 3.13. Given that the channel height in the author’s reformer rig is only to be six millimetres high (measured above the anode surface), the amount of material necessary to provide the respective seating surfaces for the windows at the front and top of the channel would severely compromise optical accessibility. An alternative solution therefore had to be found. Figure 3.13 Cross section of channel showing face sealing surface and obstruction of collection optic In order to maintain the simple and easily implemented window face seal solution a sealing surface had to be incorporated into the channel that did not obscure an optical axis. The seating surface was therefore moved from within the channel itself and adapted to become a portion of the channel wall. By doing this the front window had a surface with which to seal against around the four edges of the window. The top window on the other hand only had three surfaces in the same plane as the window surface with Reduced collection optic solid angle Required sealing surface 78 which to seal against. To complete the seal, a surface was provided for the edge of the window to seal against, Figure 3.14. Figure 3.14 Cross section of channel showing alternate sealing layout with external seating force necessary for window perimeter seal However, the use a perimeter seal required that a seating force be provided in order for the two surfaces to remain in contact. This was initially achieved using a spring formed from a piece of stainless steel sheet, Figures 3.15(a). An additional seating force on the window surface seals was also provided using sprung clips fabricated from stainless steel sheet. Perimeter seal for one edge of top window Face seal for front window Required seating force 79 (a) (b) Figures 3.15a and b. (a) A stainless steel spring to provide a lateral seating force to the top window. (b) Sprung top window fasteners. Note the first iteration front window attachment system using threaded fasteners. This was replaced with a sprung system prior to testing. 3.9 Optical Accessibility of the Heater In comparison to the reformer channel, the practical task of providing optical accessibility to the heater was considerably less challenging. The only design challenge was to locate the windows at the required positions while at the same time not over constraining the relatively brittle material when undergoing expansion and contraction during heater start up and cool down. Sealing the windows was not a major concern as the interior of the heater did not need to be gas tight to the same extent as that of the reformer. However, it was important that no large openings existed in order to limit ‘chimneying’ effects which in turn could result in localised cooling of the reformer body. Continuing the theme of the heater body, the enclosures for the windows were designed from 316 stainless steel sheet metal. In the case of the top heater window – that which enables optical access to the top reformer channel window – the window assembly also provided the structure with which to space the semi-cylindrical heating 80 elements Figure 3.16. The ‘T’ cross section of the window assembly was designed as such so that optical devices or beam shaping components could be placed as close to the reformer channel/anode surface as possible without being confined by the relatively narrow 40 mm spacing between the heating elements. Figure 3.16 Optical heater top window assembly spacing the semi-cylindrical heating elements In contrast to the top window of the heater, the front window assembly had no secondary structural purpose, with the window housing only having to enable an optical path through the 50 mm of insulation between the heater outer surface and interior. An interesting aspect of the window assembly was that it was to also be removed ‘en masse’ from the heater body when accessing the reformer channel. This was a result of the tight packaging constraints which prohibited the use of a fixed heater front window while at the same time allowing the reformer channel to be easily removed. The window was also recessed deep into the heater to try and limit the thermal gradient that it was exposed to – with one side of the window being at a temperature approximately equal to the heater set point and potentially the other being exposed to air at ambient conditions. 81 By recessing the window the air around the window exterior is somewhat stagnated, resulting in a less abrupt temperature change across the sapphire, Figure 3.17. Figure 3.17 Front window assembly and showing the optical accessibility afforded to the reformer channel Both the top and front heater window assemblies were constructed using stainless steel fasteners in conjunction with stainless steel nuts that were press fit into the sheet metal. This allowed components a degree of adjustability during the construction process and ensured the windows were not subjected to any mechanical stresses. However, it should be noted that once the window assemblies were subjected to the higher operational temperatures of the heater (700 – 800 °C) the fasteners suffered from a considerable degree of spalling – the flaking of the metal surface – resulting in the assemblies essentially becoming a single component. 3.10 Optically Accessible Reformer Channel Support As has been discussed in previous sections of this work, the effects of thermal expansion due to the large changes in temperature experienced by components between room and operational temperature were of major concern during the design 82 process. In no area was this considered to be more important than in the support and positioning of the reformer channel. At almost 150 mm in length this channel expands by almost a millimetre between room temperature and 800 °C. To over constrain this expansion would result in either the deformation and/or displacement of the structure. Such displacements could potentially have had serious ramifications with respect to the accuracy of the optical set up and the positioning of the region of interest within the reformer channel. Careful attention was therefore paid to the design of the reformer channel support structure. The chosen design solution was intended to avoid over constraining the structure while at the same time ensuring that the movement of the reformer section due to the process of thermal expansion, which was accepted to be inevitable, was controlled and repeatable. From the perspective of the minimising movement of the channel with respect to the spectrometer it was considered desirable to have the expansion occur, in effect, in only one direction. In order to achieve this, the support structure was arranged in such a way so as to be located outside the heater which ensured expansion was limited to the channel only. A schematic of the arrangement is shown below, Figure 3.18. Figure 3.18 Schematic of reformer channel support design intent Thermal expansion Hot zone Simple pinned joint 83 The realisation of the schematic shown in Figure 3.18 can be seen below in Figure 3.19. On the left hand side is the support structure which fulfilled the role of the pinned joint, but in this instance the rotation provided by the pin was replaced by the flexure of the sheet metal for small angles. The reformer channel was integrated into the reformer channel ‘rail’; a structurally stiff assembly which leads into and out of the heated region of the heater and through the 50 mm of insulation making up the heater walls. On the right hand side the reformer channel rail rests upon a functionally slender column, constructed from a flat piece of sheet metal. Thermal expansion was accommodated either by the slippage of the reformer rail over the right hand support, or in actuality the deflection of the slender column which could easily deflect around its point of attachment at its base. Figure 3.19 Optically accessible reformer channel rail assembly with support structure In addition to its primary function of limiting the movement of the reformer channel (with respect to the collection optics) due to thermal expansion, there is an Thermal expansion Primary support structure Flexible support structure 84 additional benefit to the support structure and rail mechanism described in Figure 3.19 with respect to the servicing and maintenance of the rig. By simply supporting the reformer rail at one end the entire assembly including the left hand support structure can be slid out of the heater (to the left in Figure 3.19), providing access to both the heater and reformer. In addition to the fixed rail support, the front panel (including the front optical heater window) is also removed. The only preparation necessary for this action is the disconnection of the gas supply which also feeds through the front access panel. It is important to note that the gas tube connections were situated outside of the hot region so as to limit thermal degradation effects. The heater with the reformer rail removed can be seen in Figures 3.11b, note the slot through which the optical rail is inserted in the centre of the image. All of the removable elements of the test rig can be seen in Figure 3.20. 85 Figure 3.20 Reformer rail assembly after being removed from heater and sitting on a servicing stand. Note the integrated front access panel, front window and rail support, gas piping and insulation. The right hand support is equivalent to that used to support the rail when situated in the heater. 3.11 Apparatus Modifications The previous sections of this chapter have presented a summary of the main design considerations and decisions made before the manufacture and subsequent high temperature testing of the apparatus. However, it will come as no surprise to the informed reader that a series of issues were encountered through the course of commissioning the apparatus that necessitated some redesign and modification. Some of these modifications were trivial and were merely a second iteration of a pre-existing design philosophy; however two modifications – both associated with the seating and sealing of the reformer windows - were of sufficient importance to merit further detailed discussion. These two issues were; an unanticipated reactivity between the sapphire 86 and stainless steel at high temperatures and an insufficiently large seating force necessary for the perimeter face seal of the top reformer window. 3.11.1 Optical Window Reactivity The reactivity of the sapphire window with the stainless steel reformer channel was manifested as a residue on the surface of the window itself, accompanied by a milky discolouration of the window surface at points were it contacted the reformer channel, Figures 3.21a and b. (a) (b) Figures 3.21a and b. (a) Residue deposited onto the surface of the window. (b) Milky discolouration around edge of window. The precise cause of the discolouration was not determined. It is most likely a chemical reaction between the two materials brought about by the elevated temperatures at which the components were operating. However, while the cause was not known, the ramifications were clear; the degradation of the face sealing surfaces and a reduction in the transmissivity of the window. While the reduction in transmissivity was restricted to the outer edges of the window, the author was concerned the region could 87 extend over time and ultimately affect the optical properties of the window at locations were optical transmissivity was essential. Degradation of the window and reformer channel surface finish was however an immediate cause for concern as surface imperfections would only reduce the effectiveness of the already imperfect surface seals. Given that the face seals were used specifically because alternatives were not available the solution to the reactivity issue was unlikely to result in an improvement in sealing. The priority however was to stop the reaction between the window and reformer channel while maintaining a reacting flow with which to validate the Raman technique. The author’s chosen solution was to separate the two surfaces and accept that a portion of the gas would escape to the surroundings. A high temperature 1/32” ceramic paper gasket (Rascor 300 Series, Cotronics Corportation) was therefore introduced at the window channel interface, Figure 3.22. Figure 3.22 Photograph illustrating the following: 1. ceramic window gasket. 2. fully reduced anode sample as indicated by uniform grey colouration. 3. updated seating force arrangement (see below §3.11.2). 1. 2. 3. 88 The ceramic paper used was the thinnest available and it was considered that the gas losses incurred would be relatively small in comparison to the channel flow rate due to the low pressurisation of the system. This assumption was confirmed as a seal was detectable, i.e. a noticeable pressure increase when the exit of the channel was blocked, but also by the reduction of the anode material (Appendix H), which visually indicated a reducing atmosphere present in the reformer channel. However, the use of the gasket was not without its problems, specifically from the perspective of quantifying the results produced by Raman spectroscopy. A complete discussion of the experimental consequences of introducing the gasket is presented in §7.7. 3.11.2 Top Window Seating Force As described previously, by using face seals, particularly that of the perimeter face seal, Figure 3.14, a seating force was required in order to guarantee contact between the two surfaces. This was initially achieved using stainless steel sheet metal springs, Figures 3.15a, which gave satisfactory results for the first thermal cycle after rebuilding the assembly. However, shortly after beginning a second thermal cycle the reduced the anode sample suffered partial oxidation, which was observed as a green colouration at the far extent of the anode sample, Figure 3.23. Figure 3.23 Photograph showing the partial oxidation of the anode surface Direction of gas flow Reduced colouration Oxidised colouration 89 The oxidation, presumably the result of oxygen in the gas stream, occurred despite only mixtures of hydrogen and nitrogen being passed through the channel. Subsequently the source of the oxygen was found to be an air leak in the vicinity of the perimeter face seal. It was surmised that during cooling of the reformer channel assembly the perimeter of the window would loose contact with the corresponding reformer channel surface via a process of ‘thermal inch worming’. While the stainless steel sheet window springs were conceived to counteract this effect, the material properties of the metal, particularly that of elasticity, were altered by exposure to the high temperatures. It was therefore necessary to provide a seating force via an alternate method. Using a technique similar to that used for the reformer rail assembly, §3.10, the chosen solution was to remove the source of the seating force from the high temperature environment and place it outside of the heater body. As such a mass was suspended via a series of linkages constructed from stainless steel rod (0.75 mm diameter) and arranged to provide a constant, temperature independent, seating force, Figure 3.24. However, while the theory behind the solution was simple, the practical implementation was complicated by the need for the seating force to be distributed evenly and to not impede the existing apparatus or optical paths. 90 Figure 3.24 Schematic of the use of an external mass to provide the reformer top window seating force. A suspended mass, as opposed to a spring, was selected to provide the seating force because it provided a constant force regardless of potential displacements due to thermal expansion. In order to distribute the load evenly an ‘L’ section was formed from stainless steel sheet metal and laid along the back of the top window, Figure 3.22. This alleviated point loads on the window which would have been present if the stainless steel rod had been used to apply the force directly. In actuality, a pulley arrangement was considered overly complicated, particularly as it would have required both the permanent location of pulleys at the front of the heater which was not designed to be load bearing and equally, pulleys would have required the use of a flexible wire or cord, the latter being inadvisable due to the exposure to high temperatures. Instead a pivoting structure was constructed from stainless steel sheet, which eliminated the need for pulleys, required only stainless steel rod to be used, needed no permanent attachment to the reformer or heater body, and left the optical path unobstructed. Mass Seating force Pulley Hot zone 91 Figure 3.25 Suspended mass and pivot system to produce window seating force. The mass consisted of a block of mild steel weighing approximately 300 grams. Note: An added advantage of the external mass system was that it was sufficiently remote from hot surfaces that an additional force could be exerted on the mass by the operator. This proved useful both on warm up, but more importantly, cool down to ensure the seating force continued to overcome any displacements due to material contraction. After the system was introduced no further issues with oxidation were observed. Mass Pivot 92 4 Preliminary Raman Spectroscopy Set Up 4.1 Introduction Through the course of designing the experimental apparatus and implementing Raman spectroscopy into the environment of a high temperature methane reformer a number of challenges were encountered. At each challenge an appropriate solution had to be found to allow the project to proceed further and to enable a Raman signal to be detectable from within the reformer channel. When a repeatable Raman signal was discernable from inside the reformer channel it was determined that a sufficient number of techniques and advances had been made to merit a journal paper. The paper was published in the Review of Scientific Instruments1 and it is reproduced, with permission, in its’ entirety in §4.2 below. The two most significant issues encountered were the contributions the Rayleigh line “tail” and fluorescence from the surrounding solid surfaces producing a signal across the spectral range of interest and which overwhelmed the gaseous Raman signal. The subsequent solutions to these challenges (a liquid potassium hydrogen phthalate low pass filter and fluorescence suppression through polarization) were in themselves not novel. However, the implementation of these technical solutions in the context of a broadband, 308 nm XeCl excimer laser was novel, and thus of interest to the wider scientific community. 1 Saunders, J. E. A. and Davy, M. H.  Broadband 308 nm Vibrational Raman Spectroscopy Using a Potassium Hydrogen Phthalate (KHP) Liquid Filter and Polarization Fluorescence Suppression. Review of Scientific Instruments, vol. 81, 013108. I conducted all apparatus design, implementation, data collection and analysis with supervision and editing from Dr. M. Davy. 93 After the publication of the paper an additional technical challenge was encountered, which was the long term durability of the polarizer used to polarize the laser beam. A brief summary of this issue is included after the paper, along with a discussion examining the rational for the differences in experimental setup and technique, between that proposed in the paper and that which was subsequently adopted. 4.2 Broadband 308 nm Vibrational Raman Spectroscopy Using a Potassium Hydrogen Phthalate (KHP) Liquid Filter and Polarization Fluorescence Suppression Broadband XeCl excimer lasers operating at 308 nm are not currently used in the field of gas phase Vibrational Raman Spectroscopy (VRS). An explanation as to why alternative wavelengths, and in particular tuneable, narrowband lasers are currently used in VRS is presented in addition to demonstrating a set-up which makes the XeCl laser a viable alternative when considering excitation sources for VRS. A solution of potassium hydrogen phthalate is shown to be a practical low pass liquid filter and to reduce substantially the effects of Rayleigh scattering on collected Raman spectra. The use of a commercial beam polarizer is also shown to be effective in suppressing background fluorescence that otherwise necessitates the use of expensive tuneable, narrowband lasers when performing VRS with sources of background fluorescence. Finally, an unconventional excitation beam arrangement is shown to produce viable Raman spectra from which species concentrations and distributions can be determined. © 2010 American Institute of Physics 94 4.2.1 Introduction Vibrational Raman Spectroscopy (VRS) is a valuable, and now well-established measurement, technique that is commonly applied to determine species concentrations within reacting gaseous flows. The principal advantage that VRS offers over the complementary, noncontact measurement technique of laser-induced fluorescence (LIF) is that VRS is able to detect multiple chemical species simultaneously (although LIF is able to detect trace quantities of chemical species that VRS cannot). The noncontact, in situ nature of VRS can also be advantageous over conventional gas sampling where the introduction of a sampling probe is impractical. However, the implementation of VRS is nontrivial, requires specialized equipment, and careful experimental design. Perhaps the most significant equipment consideration facing the would-be experimentalist is the choice of excitation light-source. A review of the literature suggests that narrowband, tuneable KrF excimer lasers (operating at 248nm) has become the de- facto standard excitation sources for VRS studies in combustion [Grünefeld et al., 1995; Grünefeld et al., 1994; Pitz et al., 1990; Wehrmeyer et al., 1992; Reckers et al., 1993; Hagris, 1981; Knapp et al., 1997; Blotevogel et al., 2004; Ipp et al., 2000; Hartfield et al., 1997; Nandula et al., 1994; Tolboom et al., 2003]. The widespread use of KrF excimer lasers as an excitation source for VRS in combustions research is similarly noted in the review papers of Hassel and Linow [Hassel et al., 2000] and Rothe and Andresen [Rothe et al., 1997]. ArF (193 nm) and XeCl (308 nm) excimer lasers have also been used for VRS combustion but must less frequent. The choice of a UV laser is typically justified with respect to signal intensity (in terms of both power and flux); the Raman signal is proportional to the fourth power of the excitation frequency, i.e. the higher the excitation frequency, the greater the signal magnitude [Long, 1977; Eckbreth, 1988; Hassel, 1993]. Given that high signal strength is 95 desirable, the high-powered, narrowband, tuneable KrF excimer laser is used in an effort to mitigate the many drawbacks associated with performing Raman spectroscopy in the UV. Two issues are of particular concern when performing VRS with an UV excitation source, the first of which is fluorescence, typically from OH, O2, or NO in combustion studies. The second is that the Raman signal is considerably weaker than that produced by Rayleigh scattering, necessitating the attenuation of the latter so that the Raman signal can be properly evaluated. However the choice of low-pass filters able to operate in the UV is somewhat limited, as will be discussed later. By using a tuneable, narrowband laser it is sometimes possible to avoid exciting fluorescence completely (the laser is tuned to avoid the excitation bands of the fluorescing species) [Hassel, 1993; Lipp et al., 1993]. Where this is not possible, the mostly polarized nature of the light emitted from narrowband, tuneable lasers make it possible to subtract the polarization insensitive background fluorescence from the polarization sensitive Raman signal [Grünefeld et al., 1995; Reckers et al., 1993; Knapp et al., 2000; Egermann et al., 2004]. Fluorescence subtraction does however require that there be no variation between the Raman spectra and fluorescence spectra subtracted from it. This therefore makes Raman spectroscopy with background fluorescence subtraction difficult to implement in unsteady flows unless the spectra are collected simultaneously from a single laser shot [Grünefeld et al., 1994]. If single shot Raman is not possible, the background fluorescence technique can only be implemented for steady flow applications or when the spectra/fluorescence is being collected from quasistationary objects. Attenuation of Rayleigh scattered light is achieved through the use of a suitable low-pass filter positioned in front of the spectrometer's entrance slit. However, the shorter the excitation wavelength, the smaller the spectral separation between the excitation line and the Raman signal on a wavelength scale [Hassel, 1993]. Hence, at 96 shorter wavelengths, the greater is the necessity for low-pass filters with, sharp, precisely placed cutoffs in order to avoid loss of information regarding species with small Raman shifts. Conventional low-pass filters that exhibit suitable cut-off behaviour at longer wavelengths, such as dielectric stack edge and holographic notch filters, either do not exhibit the correct cutoff for use with 308 nm excitation sources or, as is the case of holographic notch filters, are manufactured using substrates that do not transmit in the UV range [Kleimeyer et al., 1996; Mckenzie, 1993]. The choice of UV laser wavelength for Raman spectroscopy is therefore limited to those for which suitable low-pass filters can be found. The literature suggests that Schott glass filters BG24 and UG11 offer acceptable (although not ideal) performance for ArF and KrF wavelengths respectively [Rothe et al., 1997], with the use liquid butyl acetate filters also reported for KrF applications [Grünefeld et al., 1994; Hargis et al., 1981; Appel et al., 2002]. At this point the authors pause to reflect that, aside from the reasons presented above, narrowband excitation is not specifically required in many applications outside of fundamental research in the fields of physics and chemistry. In these fields, the fine spectral resolution facilitated by the use of narrowband lasers allows rotational energy levels to be observed and thus can provide useful information about molecular structure. However, in applications where only Q-branch Raman responses are of interest (such as species detection in a gaseous mixture), this higher resolution is not required. In the present work, we seek to demonstrate the feasibility of an alternative (and relatively low- cost) approach to VRS using a broadband, 308 nm XeCl excimer laser, a potassium hydrogen phthalate (KHP) liquid filter, and polarization fluorescence suppression. The choice of the XeCl excimer as an excitation source for VRS is, in itself, not unique – although their reported use is somewhat rare [Rothe et al., 1997]. In an early work, Hassel [Hassel, 1993] used a tuneable, narrowband XeCl laser to perform Raman measurements in flames. However, the significant interference from broadband 97 fluorescence from OH radicals and polycyclic aromatic hydrocarbons in the flame was reported [Hassel et al., 2000; Hassel. 1993; Lipp et al., 1993]. Alger and Wooldridge [Alger et al., 2004] used XeCl laser excitation and VRS in a study of residual gas fraction in an internal combustion engine - regrettably though, Alger and Wooldridge make no mention of the laser line width or the issue of background fluorescence in their work. Despite this relative lack of popularity for the XeCl laser as a light source for VRS, the present authors believe that there is merit in the choice of 308nm excitation. Provided that any interfering fluorescence signal can be adequately suppressed, we believe that the 308nm XeCl laser offers a good compromise position between the high Raman signal strength associated with excitation in the deep UV and the excellent spectral separation of the Raman lines associated with excitation at longer wavelengths. 4.2.2 Experiment The broadband 308nm Raman system described in this work was developed against the backdrop of a larger study of methane steam reformation for solid oxide fuel cell (SOFC) applications – the detailed results of which will be published at a later date, Figure 4.1. 98 Figure 4.1 Setup. Experimental setup of laser, optics and detection system. Reprinted with permission of the American Institute of Physics. Test channel: A stainless-steel channel of height 6 mm, width 54 mm and length 135 mm was machined to hold a 100 x 50 mm2 SOFC anode sample. The anode, a yittria stabilized zirconia cermet with a nickel catalyst, is a 1 mm thick porous structure with voids in the order of 1 μm. A 5 mm thick UV grade sapphire window is used to form the roof of the channel. A second window, also of UV grade sapphire, is positioned at one end of the channel as shown in Figure 4.2. Imaging Spectrograph and ICCD Polarizing Beam Splitter Test Channel 308nm XeCl Laser 300mm cylindrical lens f=300mm Spherical lenses f=200mm x y z KHP liquid filter 99 Figure 4.2 Channel profile. Test channel showing anode surface in relation to excitation beam and optical windows. Reprinted with permission of the American Institute of Physics. Laser characteristics and beam shaping: A Lambda Physik COMPexPro 102 XeCl broadband laser (centred at 308 nm with a line width of 1.0 nm) was used to excite air at room temperature and pressure in the test channel. The beam exited the laser with an energy of 200 mJ and a profile of 10 mm x 24 mm (repetition rate 20 Hz, pulse duration of 20 ns). The beam was focused using a non coated cylindrical lens of focal length 300 mm in order to increase beam energy in the interrogation volume and improve spatial resolution (making the spatial resolution a function of beam position rather than spectrometer slit width). A slit mask of width1.5 mm was used to produce a beam that was slightly in excess of the interrogation volume width, while also allowing fine tuning of the beam position in the z direction and reducing the amount of unnecessary background scattered light (Figure 4.3). The shaped beam was directed into the channel through the top window such that it was perpendicular to the anode with the beam waist focused approximately 10 mm below the anode surface. The resulting Excitation beam Scattered light Anode surface Sapphire Windows Body of channel x z y 100 excitation volume was a symmetrical trapezoid, 6 mm in height with widths of 1.0 and 0.6 mm respectively, projected through 1.5 mm. Figure 4.3 Light sheet. Light sheet and beam waist in relation to anode surface. Reprinted with permission of the American Institute of Physics. Beam polarization: Due to the unpolarized nature of the broadband laser the beam had to be polarized so that a scheme for background fluorescence suppression could be implemented; in this case a Melles Griot 03 PBS 117 polarizing cube beamsplitter with a 308 nm coating was used. The polarizer sat in the beam path before the shaping optic and it could be rotated through 90 degrees in order to vary the orientation of the E vector of the beam. Light collection, edge filter, spectrometer and CCD: The signal collection system consisted of two plano-convex spherical lenses (curvatures facing each other) of 50 mm diameter with wavelength specific anti-reflection coatings and focal lengths of 300 and 200 mm, respectively The ƒ# of the second optic was specifically selected to Slit mask Sapphire window Beam waist 1.5 mm . x y z x z y 101 match that of the spectrometer. A single stage 300 mm spectrometer (Acton SP-2356) with ƒ# of 4 was used, fitted with a 1800 g/mm grating. The slit width of the spectrometer was set to 0.75 mm which was a compromise between spectral resolution and light collection capability. Scattered light from the excitation volume exits the channel through the front window; thus, the signal is collected orthogonal to the direction of laser propagation and parallel to the anode surface. Prior to entering the spectrometer, the collected light is passed through a liquid filter (path length 10 mm) comprising of a UV grade cuvette (Newport Oriel P/N 13960) filled with a 37.5 mg/mL aqueous solution of potassium hydrogen phthalate (KHP). Images of the resulting spectra were recorded using a PCO DiCam-Pro intensified CCD (ICCD) camera with a 1280 x 1024 pixel (8.6 x 6.9 mm2) resolution giving a spectral range of 28 nm. “On-chip” integration over multiple laser pulses was used to maximize the contrast of each CCD “exposure”, while the camera was operated with a 70 ns intensifier gate to minimize the effect of background light. 4.2.3 Results and Discussion Effectiveness of KHP liquid filter: The lack of a readily available low-pass filter with a suitably sharp cut-off for 308 nm excitation has likely contributed to the low adoption of XeCl lasers as an excitation source for VRS. Alger and Wooldridge [Alger et al., 2004] employed a WG320 Schott glass filter for Rayleigh line suppression while performing VRS measurements of the H2O/ N2 ratio in the in-cylinder gases of an internal combustion engine. However, the results of their research suggest that the cut- off of the WG320 filter is not sufficiently sharp as to allow Raman signals close to the excitation wavelength – such as the CO2 line at 320.7 nm – to be resolved with a suitable 102 signal-to-noise rayio (SNR). In the present work, the authors have taken a different approach using a liquid filter of potassium hydrogen phthalate KHP. The use of a KHP filter was originally proposed by Kleimeyer et al. [Kleimeyer et al., 1996] who, having developed the filter for use with the first H2 anti-stokes shifted 3rd Nd: yttrium aluminum garnet harmonic at 309 nm, suggested that it might, “with little or no modification”, be equally suitable for a 308 nm excitation source. However, to the authors' knowledge, the validity of this claim has not been tested within the literature. Accordingly, one of the primary goals of the present work was to demonstrate that the KHP liquid filter is suitable for Rayleigh line suppression at 308 nm and to assess its effectiveness in comparison with the WG320 glass filter previously used by Alger and Wooldridge [Alger et al., 2004]. 103 Figure 4.4 Raman signal of air at 21 ˚C, 101 kPa: (a) unfiltered (b) WG320 glass filter, (c) KHP liquid filter. Note: the horizontal feature present at the top of image (c) is scattered light from the surface of the anode which was not fully eliminated from entering the spectrometer when masking the entrance slit. Reprinted with permission of the American Institute of Physics. Figure 4.4 presents an example of the raw images recorded by the spectrometer system for the unfiltered and filtered cases of air at 21 ˚C, 101 kPa while Figure 4.5 shows the filtered and unfiltered Raman spectra derived at the same conditions. Even without the additional benefits of fluorescence suppression, it is clear that the SNR’s of the O2 and N2 Q branches are substantially improved and the peaks made more 104 apparent by filtering with either the WG320 or KHP liquid filters. Significantly, while the authors were not able to evaluate either filter directly at 308 nm in the present work due to saturation of the ICCD, even for single laser pulse, a comparison of the raw images and the derived spectra near the ‘tail’ of the excitation line clearly show the beneficial effect of the KHP filter compared with the WG320 filter at wavelengths shorter than around 323 nm. This is especially beneficial should carbon dioxide be present in the process being studied with Q-branches at 321.7 and 320.7 nm. The authors also note that fluorescence of the KHP filter has not been encountered to date. Figure 4.5 Filter comparison. Comparison of unfiltered and filtered Raman spectra of air (O2:324 nm, N2:332 nm) at 21 ˚C, 101 kPa - Top: six laser shots per image, spectrum produced by averaging over ten images and ten pixel rows. Bottom: 300 laser shots per image, spectrum produced by averaging over ten 105 images and ten pixel rows. Reprinted with permission of the American Institute of Physics. Orthogonal to surface beam orientation: As discussed earlier, the present work is presented within the context of the development of a measurement technique for use in SOFC applications. This intended application imposes severe restrictions upon the available optical access and thus poses significant technical challenges. In a typical gaseous Raman experimental set up the excitation beam’s path into and out of the region of interest is strictly controlled such that the beam does not impinge on surfaces that might either fluoresce or produce sources of light stronger than that produced by Raman scattering of the gaseous flow. The orthogonal interaction between the excitation beam and the anode surface demonstrated in the present work is therefore not a common optical configuration placing, as it does, the region of interest – the gaseous flow over the anode surface – directly adjacent to a significant source of scattering and fluorescence – the surface. During the course of the present work, it was found that the position of the beam waist with respect to the anode surface was a critical factor in controlling the relative strength of the gaseous and surface Raman signals. By positioning the focal point of the beam below the surface, the surface Raman signal was reduced to the extent that it no longer obscured the Raman signal produced by the gas above it. This beam focussing has the additional benefit of reducing the beam energy at the anode surface (and through the window) thus helping to avoid potential laser damage. In addition to reducing the effects of Raman scattering from the anode surface (by positioning the beam waist below the surface as described above), significant care was necessary to avoid saturating the ICCD due to Raman and Fluorescence signals from the anode surface and the leading and trailing faces of the channel window. This 106 was be achieved by carefully masking the spectrometer slit in the spatial direction. Note that the porous nature of the anode, as opposed to that of a pure metallic surface e.g. stainless steel, also helps reduce the effect of elastically scattered light. With these sources of spurious light eliminated and the surface Raman reduced by beam focussing, a viable Raman spectra is produced from the gaseous region of interest, Figure 4.6, albeit with a significant contribution from background fluorescence. However, the result shown in Figure 4.6 does suggest that geometries which cannot accommodate optical paths traditionally used by Raman spectroscopy can be interrogated successfully and that, specifically, the confined volumes present in the gas channels of an SOFC are a feasible environment in which to perform VRS. Figure 4.6 Raw Raman spectra. Spectral curve produced from averaging ten images, each consisting of 300 laser shots. Reprinted with permission of the American Institute of Physics. 107 Fluorescence suppression: It is apparent from the results shown in Figure 4.6, that the level of background noise present in the spectrum is still significant with respect to the magnitude of the O2 and N2 lines. This noise, which in this instance is primarily fluorescence excited at the anode surface, not only renders the determination of the true O2 and N2 signal intensities difficult; it prohibits lower intensity spectral events from being observed. Several methods of correcting for the background fluorescence in VRS spectra have been proposed within the literature. These range from the simple subtraction of an intensity level derived from a linear curve fit through selected points at the estimated ‘base” of the observed Raman peaks [Hassel, 1993] to more complex methods that utilize the polarization characteristics of the Raman and fluorescence signals [Grünefeld et al., 1995; Reckers et al., 1993; Knapp et al., 2000; Egermann et al., 2004]. The most commonly reported polarization technique for fluorescence suppression in the field of combustion research is that commonly attributed to Grünefeld et al. [Grünefeld et al., 1994]. Grünefeld’s method, which assumes a high degree of linear polarization in the laser radiation, involves the discrete detection of both polarization components (perpendicular and parallel to the E vector of the laser radiation) of the collected light. The signal with the vertical polarization will contain the Q branch Raman signal, a contribution from the O and S branches and the fluorescence, while the signal with the horizontal polarization will contain the fluorescence signal, the contribution from the O and S branches and some small fraction of the Q branch signal. With appropriate calibrations applied to the raw data the Q branch intensity may be obtained [Grünefeld et al., 1995]. Several alternative implementations of Grünefeld’s method are reported in the literature. In his initial work, Grünefeld employed a single spectrometer/ICCD camera 108 set-up to capture each of the polarization components in turn, rotating the E vector of the laser source through 90° using a half-wave plate as required [Grünefeld et al., 1995; Grünefeld et al., 1994]. Subsequent studies have employed a glan prism and two spectrometer/ICCD camera systems allowing simultaneous capture of the two polarization components [Egermann et al., 2004]. Hartfield et al., [Hartfield et al., 1997] demonstrated a novel alternative to the two camera system, using a calcite crystal to split the unpolarized signal beam into vertical and horizontal components within the spectrometer in such a way as to allow both components to be captured side-by-side on a single ICCD chip. Note that regardless of the chosen collection technique, a highly polarized laser source is specified in all of the referenced works. In the present work the authors adopt a slightly different approach to the previously referenced works, combining an unpolarized laser source with an external polarizing device (Melles Griot 03 PBS 117). The polarizer is then rotated through 90° to provide the two polarization components as required by the technique. Taking this approach reduces the overall system cost as a relatively inexpensive broadband laser is employed with a single one spectrometer and ICCD camera. This does however mean that “single” shot measurements are not possible, making this set up unsuitable for unsteady flows. The results of the applied fluorescence suppression method are shown in Figure 4.7. It can be see that a significant component of the background noise is removed greatly increasing the SNR’s of the O2 and N2 peaks. 109 Figure 4.7 Polarization suppression. Effects of fluorescence suppression on spectral curve. Reprinted with permission of the American Institute of Physics. 4.2.4 Conclusions An unconventional experimental arrangement for performing VRS at 308 nm using a broadband XeCl excimer laser has been demonstrated under cold (non- reforming) and non-flow conditions in a model reformer section of a SOFC. The results of this work support the following conclusions: • Aqueous solutions of KHP are effective liquid filters for the suppression of the 308 nm Rayleigh line from XeCl lasers. The KHP filter offers markedly superior performance against the WG320 Schott glass alternative at wavelengths shorter than 323 nm. 110 • Polarization of unpolarized 308 nm broadband laser light through the use of an external polarizing device is a practical alternative to the use of expensive, highly-polarized, narrowband, tuneable lasers, enabling fluorescence suppression techniques to be applied. • Subject to sufficient attenuation of the Rayleigh line and the polarization suppression of any interfering fluorescence signal, geometries that cannot accommodate the optical paths traditionally used by Raman spectroscopy–such as the confined volumes present in the gas channels of an SOFC–can be interrogated successfully using VRS. • The broadband XeCl excimer laser shows excellent potential as an excitation source for VRS of reacting flows. 4.2.5 Acknowledgements The authors gratefully acknowledge past and present financial assistance from the Advanced Systems Institute of British Columbia, the Canada Foundation for Innovation, the British Columbia Knowledge Development Fund, NSERC, and Auto21 in the completion of this work. Hardware donations and technical support from Versa Power Systems are also acknowledged. 4.3 Durability of Laser Beam Polarizer On determining that the experimental setup and technique was able yield a suitable Raman signal from within the reformer channel, focus shifted to optimizing the signal and refining the experimental procedure. It was through the course of this effort, which subjected the polarizer to high numbers of consecutive laser pulses, that the polarization technique ceased to produce the results reported by Saunders and Davy 111 [Saunders and Davy 2010]. Upon further investigation the source of the problem was found to lie with the Melles Griot 03 PBS 117 polarizer, Figure 4.8. Figure 4.8. Melles Griot 03 PBS 117 polarizer damage The failure of the component necessitated the cause of the failure to be understood and it was determined that the polarizer had been incorrectly specified by the manufacturer. It was therefore necessary to source a new polarizer of sufficient robustness to withstand both the high laser powers and high photon energies experienced at 308 nm. Accordingly, an alternative polarizer was sourced from the Newport Corporation (P/N 10SC16PC.23), which (on paper at least) was specified to be suitable for the 308 nm XeCl excimer laser application. However, after only approximately 1000 laser shots the polarizer was also observed to have been damaged, Figure 4.9. i.e. once again the chosen component failed to meet the manufacturers advertised performance standards. 112 Figure 4.9 Newport 10SC16PC.23 polarizer damage On further investigation, it was determined that both the Melles Griot and Newport polarizer’s achieve polarization through the use of a polarity selective coating situated at the interface of two optically transmissive prisms, which are bonded together to form a cube. Ultimately, both manufacturers surmised that it was the bonding material between the prisms that was unable to tolerate the repeated excitation experienced as a result of repeated laser pulses. Once again, an alternative polarization device had to be sourced and assessed for suitability for the author’s experiments. Two candidate devices were selected from the optical equipment supplier Delta Photonics the first being a plate polarizer, P/N PPB- 5009U-308, with a coating specific to 308 nm and the second being a Glan laser polarizer, P/N GLP-63320, with an air space, thus avoiding the problem of insufficiently durable bonding cement. The operating principle of the Delta Photonics polarizers differ from the Melles Griot and Newport polarizers. Both the plate polarizer and the Glan 113 polaizer utilize the principle of Brewster’s angle, an angle of approximately 56° between an incident beam and an air/substrate interface at which the electric vector of polarity parallel to the surface is reflected while the perpendicular vector is transmitted. Nevertheless, the plate polarizer, while not visibly suffering any damage, was found to not be suitable for the author’s purpose. However, the Glan polarizer was able to reproduce, at least partially the results observed using the original Melles Griot polarizer. 4.4 Glan Polarizer The Glan polarizer was observed to be polarizing the laser beam as physical rotation of the polarizer – i.e. varying the electric vector/polarity of the beam - resulted in significant reduction of the Raman signal, as expected and as described by Hecht, [Hecht et al. 1997]. However, the Raman signal was not completely eradicated, Figure 4.10. Nor was the background component of the two spectra of equal magnitude. Despite efforts to correct this – such as optimizing polarizer orientation (Appendix G) and compensating for differences in beam energy of the opposing polarities (due to the beam being ‘partially’ polarized on leaving the laser) - neither the Raman signal for the fluorescence only orientation could be eliminated nor the background signal made equal in magnitude for both polarizer orientations. Fluorescence suppression through polarization using the Glan polarizer did not therefore deliver the benefits shown when using the Melles Griot polarizer, the results from which could not be repeated due to the lower than specified damage threshold of the Melles Griot polarizer as has been previously discussed. In fact, the benefits of using the Glan polarizer were found to be offset by a reduced overall signal level, necessitating more laser shots per image. It was for this reason that fluorescence suppression through opposing polarity background 114 subtraction was abandoned and an alternative method of background subtraction applied, §7.1. Figure 4.10 Comparison of spectra collected using excitation sources of opposing polarities at 21 ˚C, 101 kPa - 300 laser shots per image, spectrum produced by averaging over 20 images and seven pixel rows. While a full scale investigation into why the Glan polarizer did not repeat the earlier results was not performed, several factors could have contributed to this result. The first being the relatively large beam divergence of the laser (between 1 and 3 mrad). With this divergence it is impossible for the entire extent of the laser beam to be at an angle of incidence of 56° with the working surface of the polarizer. Therefore a greater proportion of the opposing polarity of light will pass through the polarizer, resulting in the 115 Raman signals clearly visible in the theoretically fluorescence only signal. In addition, any Raman signal derived from the anode surface itself (a phenomenon not examined in the work presented here) could itself be influenced by beam polarity, potentially affecting the magnitude of the broad background signal. The use of the Glan polarizer was not without merit however because it produced a beam that was composed primarily of light orientated to elicit a Raman response from the gaseous phase, and so served to improve the ratio of signal to noise. That is to say that light incapable of eliciting a Raman signal was prohibited from passing to the collection volume, thus reducing the amount of fluorescence and thus background signal. Use of the polarizer was therefore found to still be critical to improving the quality of the Raman signal produced, even if the ultimate method of doing so was not that which was originally conceived. 116 5 Description of Experimental Apparatus The following is a description of the experimental apparatus, as used to perform the experiments described in Chapter 6. Key features of components are described as well as operational limitations which may form the basis of improvements/suggestions for future work. Primarily this chapter is intended to allow the experimental conditions to be replicated and repeated independently. The description will first discuss the reforming gas supply, metering and treatment before and after the reforming section. Focus will then switch to the excitation source and follow the path of excitation light to the region of interest before concluding with the path of the scattered light as it passes through collection system, low pass filter, spectrometer and finishing at the intensified charge coupled device (ICCD). 5.1 Gas Supply, Metering and Treatment Four gases were needed in order to carry out the actions of anode sample reduction, Raman signal calibration and methane/steam reforming tests. These gases were hydrogen, nitrogen, methane and steam. The first three gases were supplied by Praxair, with hydrogen and nitrogen being of the 5.0 product grade entitled ‘research’ and the methane of grade 3.7 entitled ‘ultra high purity’. The purity of the gases was 99.999 and 99.97% respectively, with maximum impurities in the order of 50 ppm. In addition the methane was assayed to guarantee the absence of sulphur, the presence of which would result in catalyst poisoning, a problem specifically highlighted by Versa Power Systems, who have considerable experience in methane steam reformation in the context of solid oxide fuel cells (SOFC’s) [Brown et al., 2006]. The high pressure cylinders of hydrogen, methane and nitrogen had the pressure regulated to 50 psi (gauge), a requirement of the back pressure rotameters used, Figure 5.1. 117 Figure 5.1 Schematic of gas routing, metering, humidification and post reformation treatment In addition to supplying the reforming channel, a portion of nitrogen was used to displace the air surrounding the reforming section inside the heater by supplying gas at a rate of 0.25 SLPM to the heater body. A stream of nitrogen was also separated from the main supply line that would not be passed through the humidifier and was referred to as N2 CH4 H2 Exhaust gas burner Humidifier Deionised water, 5-15psi (gauge) Check valve Pressure relief valve Pressure gauge Backflow prevention valve Reformer gas rotameters Reformer gas mixing tube 50 psi N2 purge N2 ‘dry’ Heater body Two way bypass valve Water separator Reformer Sparger TC 50 psi 50 psi TC = thermocouple Heated and insulated tubing 118 dry nitrogen or N2 “dry”, the purpose of which will be described fully in §6.2. All gases were metered using rotameters purchased from MUIS Controls of Edmonton, Alberta, Canada, a distributor of rotameters from original equipment manufacturer Aalborg. The reformation gases themselves were mixed using a three tube mixing arrangement with a dedicated mixing tube, Figure 5.2. As mentioned the rotameters were back pressure compensated to avoid fluctuation in metered flow rates as a result of down stream pressure variations. The tube specifications, float materials, flow rates, and settings are listed in Appendix D. Figure 5.2 Methane, hydrogen and carbon monoxide gas sensor and rotameters Once the gases have been metered they may either be humidified or bypassed around the humidifier and passed directly to the reforming section. The lab scale humidifier is custom built and of the bubbler type, specified to humidify up to 1.0 SLPM of dry gases and impart a factor of between 0.02 and 0.82 of water vapour into the gas 119 stream that is being humidified per unit volume passed through it, Figure 5.3. This is achieved by bubbling the gas into a sufficiently large water bath so that the gas leaving is saturated with water vapour. By varying the temperature of the water bath the partial pressure of water vapour can be controlled (as found in steam tables) and the factor of humidity controlled. This process was verified by measuring the wet and dry bulb temperatures of the gas exiting the humidifier and found to be the same, thus indicating that the gas stream was saturated. Figure 5.3 Lab scale humidifier It was necessary to only fill the humidifier with deionised water as impurities can contaminate the anode sample which could potentially introduce an experimental error, complicating comparison of experimental data. The humidifier was also fitted with a pressure gauge recording the pressure of the exiting gas stream so that the humidification factor can be corrected as the water vapours partial pressure is a function of both temperature and pressure. That being the case, the experimental set up used 120 was not able to operate above atmospheric pressure, thus making pressure correction unnecessary. Temperature control of the humidifier was achieved using a Eurotherm 2132 digital controller and a K-type thermocouple. The water level was automatically regulated via a Solartron Mobrey VT05 level sensor, water solenoid and a deionised water supply at a pressure of between 5-15 psig. A pressure relief valve was retrofitted to allow for unattended operation and prohibit the vessel pressure from progressing to unsafe levels in the event of a blockage downstream of the exit. From the perspective of humidification it is necessary to insulate and heat the tubing lines exiting the humidifier to above 100 °C to ensure that liquid water did not condense out of the gas stream and reduce the steam content of the gas. On leaving the humidifier the gas either passes directly to the reforming channel or is mixed with the N2 “dry” stream. The purpose of the dry stream was to allow experimental flexibility and improve productivity as discussed in §6.2. Two features that should be described are the heating of the line and the use of an isolation valve to close the line when there was no N2 “dry” flowing. The heating of the N2 “dry” line from the exit of the rotameter until it joins the humidified gas path was necessary as it was important that the temperature of the combined nitrogen and humidified gas stream does not fall below 100 °C for the same reasons outlined above. Similarly, an isolation valve prohibits the dead volume of the N2 “dry” tube from becoming a region of water condensation or gas leakage. The length of tubing along which the N2 “dry” and humidified gases mix was sufficiently long (≈3 m), tortuous (5 x 90° bends) and the quantities of N2 “dry” added were sufficiently small (see Appendix D), that the gas stream could be considered homogenous on reaching the reforming channel. At this point in the apparatus the gases, at approximately 120 °C, pass into the heater body and on to the reforming channel itself, a complete description of which is 121 included in §3.7. The temperature of the gases entering the reformer channel was not explicitly measured or controlled, but such was the gas routing, the temperature of the gases at the leading edge of the non-operating half cell surface could be considered similar to the heater set point. This was because of a short length of gas inlet tubing inside the heater, Figures 3.11a, the inlet plenum, Figure 3.8, and the approximately 15 mm run between the inlets and reforming surface. This assertion is supported by the IR thermometry results, §7.17, which suggest a maximum 20 °C temperature difference across the entire non-operating half cell in non-reforming conditions, which was consistent with spatial temperature variation due to heater geometry and localised cooling resulting from the non-insulated front window. On leaving the channel the exhaust gases consisted of nitrogen, unconsumed water vapour, unconsumed methane, reformed hydrogen and the reformation by-product carbon dioxide. Due to the continued water content the exhaust line was also heated until an appropriate water separator could be incorporated into the gas line. This consisted of a closed vessel into which the ¼” stainless steel tubing was passed, out of which exited a 5/8” copper pipe in a near vertical orientation. The copper pipe passed the exhaust gases to the exhaust gas burner situated directly below a roof mounted fume hood and was unheated, Figure 5.4. This arrangement resulted in the pipe requiring neither heating nor insulation as it was of a sufficiently large diameter that blockages were not an issue due to the condensate running down the walls of the pipe to the separator. The exhaust gas burner consisted of an off the shelf natural gas pilot light assembly which served as an ignition source which burnt off any flammable gases before entering the buildings fume extraction system. The pilot light had a thermocouple actuated safety shut off valve in the event the flame became extinguished. Should the exhaust gas burner and extraction system fail a methane/hydrogen sensor was installed which would activate an audible alarm. A final safety measure was a carbon monoxide 122 sensor which would trigger an audible alarm should carbon monoxide be formed and enter the laboratory space. Figure 5.4 Exhaust gas treatment 5.2 Excitation Source, Beam Polarization and Light Sheet Formation A Lambda Physik COMPexPro 102 XeCl broadband laser was used as the excitation source. This produced a beam with a spectral line width of 1.0 nm and the Pilot Light Exhaust gas pipe 123 physical dimensions 10 x 24 mm2 (h x v). The COMPexPro is a pulsing laser with each pulse having a duration of 20 ns and a maximum repetition rate of 20 Hz, although it was operated at a repetition rate of 10 Hz due to polarizer durability. Similarly, the laser is rated to produce a beam with a beam energy of 200 mJ, but was used at the reduced energy of 169 mJ for the experiments presented in this work. The lasing phenomena is the result of laser gases, in this case XeCl, being stimulated by an electric discharge between two electrodes. Laser beam energy is maintained by increasing the magnitude of the electric discharge as the laser gases degrade with respect to the number of laser pulses produced by those gases. When the point is reached where the electric charge can no longer be increased it is necessary to either perform a full or partial replacement of the laser gases to return the beam energy to that specified by the user. For the purposes of the experiments presented here the laser was operated in a partial gas replacement mode where a quantity of halogen gas was periodically injected into the laser tube to maintain beam energy. On exiting the laser, the beam was steered to the region of interest, reduced in size, polarised, and focused to increase the energy density of the beam and improve spatial resolution of the technique, Figure 5.5. 124 Figure 5.5 Diagram of the laser beam steering, polarization and shaping Beam steering was achieved using a periscope consisting of two Newport Corporation 18E20Al.2 elliptical aluminum broadband mirrors which raised the beam to the working height. The aluminum mirrors in the periscope were not specifically for 308 nm use, but as will be discussed shortly absolute beam energy was not a concern. The relatively high beam divergence (3 x 1 mrad, v x h) of the beam caused the beam to expand in profile to around 50 x 20 mm2 by the time it left the periscope. This was problematic as not only were the beam dimensions impracticable, with difficulties sourcing optics, but stray unused light was potentially a safety concern. It was therefore Side View Plan View Beam Shaping Schematic y x z Periscope y x z End View y x z Polarizer beam mask Polarizer Shaping optic 45 °Mirror Slit mask Anode surface Polarizer beam mask Slit mask Optical platform traverse Beam waist 125 necessary to clip the light beam to make it more practical and manageable. This was achieved using a number of simple cardboard masks, Figure 5.6. Figure 5.6 Beam masks situated immediately after the periscope (R) and polarizer mask (L) As secondary effect of beam clipping was that the edges of the beam became more defined, a considerable advantage when levelling and aligning the beam with the datum beam, an important part of the experimental set-up process, §G.1. The more important effect of masking was the reduction in the beam energy the optics were exposed to. This was of critical significance to the polarizer, the most sensitive component of the optical system. As discussed in §4.3. the task of polarizing the laser beam was non-trivial. This was predominantly due to the polarizer being unable to tolerate the high power of the excimer laser. The high power was a result in the combination of the high energy associated with the ultraviolet portion of the electromagnetic spectrum, the size of the Masks Datum Beam 126 beam and the short duration of the laser pulses. Clearly the photon energy could not be mitigated, but the beam energy, size and the repetition rate could be. Therefore, the beam energy was kept to below 169 mJ, below the rated power of the laser, the repetition rate was fixed at 10 Hz, below the max rep. rate of the laser, 20 Hz, and the beam profile was reduced to the minimum required. The limiting factor in beam profile was the aperture of the polarizer itself, which was 20 mm, Figure 5.7. Reduction of the beam repetition rate decreased the localised thermal heating effects on the polarizer due to the large number of consecutive laser pulses. Figure 5.7 Polarizer masking element After passing through the polarizer mask the beam passed through a FOCtek GLP6320 Glan laser polarizer. The beam was then steered through 90° into a vertical orientation by using a Newport Corporation 20D20AL.2 aluminum circular mirror. The mirror was mounted in 2 axis gimball to allow for exact adjustment of beam orientation ensuring that orthogonality between the excitation beam and the collection optics was Polarizer Mask 127 optimized. Mounting of the gimbal onto an optical rail required a bespoke bracket to be fabricated so that the axis of the beam path could be aligned to that of the subsequent beam shaping optic. The bracket was also necessary to minimise the moment applied to the optical rail carriage as a result of the unconventional vertical orientation of the optical rail. After the beam is steered through 90° the beam passes through the final masking element, the width of which was dictated by the size of the collection volume as described in §G.2. A slit size of 1.5 mm was selected to be marginally larger than the ‘width’ of the collection volume in order to maximise the signal being collected, Figure 5.8. To have used a significantly larger light sheet resulted in excess light around the collection volume, which had the consequence of increased fluorescence and spuriously scattered light around the ROI. This fluorescence and spuriously scattered light had the consequence of negatively effecting the signal to noise ratio. Figure 5.8 Beam reduction sheet slit Light sheet slit 128 All remaining light was then passed through the final beam shaping optic which was a non-coated, UV grade, cylindrical lens of focal length 300 mm. The purpose of this optic was to reduce the beams x-wise dimension from 20 mm to 1.0 mm before passing through the reformer channel. This served the purpose of not only increasing the beams intensity, but also increasing the spatial resolution of the technique in the x-direction. As discussed in §G.2, the spectrometer slit size and light collection lens combination result in a collection volume with an approximate length of 9.0 mm. A beam profile of 1.5 x 20 mm2 would have significantly overfilled the collection volume in x-direction, resulting in inefficient signal collection outside the bounds of the collection volume. In addition, the collected Raman signal would be the result of the integration of light scattered from across the length of the beam profile, reducing spatial resolution in the x-wise direction. By focusing the beam to a dimension of between 1.0 and 0.6 mm, Figure G.5. the spatial resolution in the x-direction is improved to 1.0 mm. It also allows the spatial resolution to be a function of beam position rather that the combination of beam and collection volume position. Focusing of the beam is not without its dangers, as the laser intensity at the beams waist can rapidly reach levels that either result in dielectric breakdown or damage the optical windows or anode sample. For this reason the beam waist was situated approximately 10 mm below the anode surface which resulted in laser intensities of a sufficiently low level that did not damage the reformer window or anode material. 5.3 Light Sheet Optical Setup §5.2 described the fundamental optical components required to produce a suitable light sheet for one dimensional Raman spectroscopy. However, the practical 129 challenges of generating a light sheet situated above a high temperature, optical heater require further description. As mentioned in the preceding section it was the position of the light sheet, not the size of the collection volume that dictated the spatial resolution of the technique in the x-direction. It was therefore necessary to produce a light sheet assembly that was adjustable, able to translate over a distance of approximately 100 mm, be repeatable and consistent. Thermal considerations dictated that optical component mounting points be isolated from components which experienced significant temperature variation during heater operation, as any variation in physical position would make light sheet alignment and repeatability problematic. For this reason an external frame was constructed around the optical heater which experienced little to no temperature variation during heater operation. The frame was also of a more robust structure on which to mount optics than the sheet metal construction of the heater. It could also be adjusted to ensure that the optics mounting platform was level in the xz plane, Figure 5.9. 130 Figure 5.9 Optics mounting frame surrounding the optical heater The requirement of the light sheet to traverse through 100 mm was significant as this was beyond the range of most “off the shelf” optical traverses. Compounding this was the requirement of the light sheet to be consistent in position and dimension across this entire range of movement. The 45° mirror had to be mounted to the traverse, as it was this components movement would result in the beams movement across the anode. However, in order to guarantee the position of the beam waist in relation to the anode surface remain constant, the position of the focusing optic in relation to the anode surface must also be constant, thus necessitating its location on the traverse. The inclusion of the polarizer on the traverse platform was a result of the relatively large divergence rate of the beam which would have caused the size of the beam striking the 45° mirror to have fluctuated significantly at the extremes of the traverses range. Therefore, a traversing platform which consisted of the aforementioned components, in Optics mounting frame 131 addition to an optical rail for adjustability was located above the heater body, Figure 5.10. Figure 5.10 Light sheet optical traverse As discussed in §G.2. the light sheet position had to be adjustable in the z- direction to ensure the collection volume was completely filled and thus ensure the maximum collection of the Raman signal. Initially the light sheet slit above the focusing optic was the method of adjustment, but this did not allow sufficient resolution as well as different portions of the beam being used. This potentially introduced variations in beam energy, particularly with the reduction in the polarizer aperture away from the centre line. The solution was to traverse the entire optical platform in the z-direction using two sets of worm screws, Figure 5.11a. Light sheet traverse 132 Figure 5.11 (a) Worm screw arrangement (b) scale indicating relative position of traverse The worm screw arrangement also allowed the rotation of the platform around the y-axis in addition to its translation, which aided in the adjustability and alignment of the polarizer with the laser beam. Pure translation of the platform was ensured by equal rotation of the screws and the position of the platform could be controlled within +/- 0.125 mm by rotating the screws partial turns. This exceeded the resolution given by the positional rules mounded to the frame, Figure 5.11(b), and fulfilled the light sheet positioning requirement with respect to the size of the collection volume. An additional advantage of the traversing optical platform assembly was the ability to slide the assembly away from above the heater optical window which allowed detailed examination of the reformers optical window gaskets. (a) (b) 133 5.4 Raman Signal Collection, Low Pass Filtration, Spectrometer and Image Capture A schematic of the light collection system is shown below, Figure 5.12. Figure 5.12 Schematic of light collection system The emphasis of the light collection system was somewhat different to that of the light sheet system in that it was conceived to maximise light collection efficiency. Whereas the light sheet system used uncoated lenses and broadband mirrors resulting in significant energy loses at each component, the light collection optics used were coated to minimise reflections and maximise signal transmission to the spectrometer, such was the relative weakness of the Raman signal. The traversing collection lens and fixed focussing lens were plano-convex spherical lenses of focal lengths 300 mm and 200 mm respectively, both of diameter 50 mm. The ƒ# of the focussing lens was specified to match that of the spectrometer. This ensures that the spectrometer is neither y x z y x z Anode surface Channel Spectrometer slit Traversing collection optic Fixed focusing optic “Top” and “bottom” slit masks Cuvette 134 over nor under filled and thus maximizing optical efficiency. Under filling does not fully utilize the capability of the device whereas over filling could potentially result in light exiting the spectrometer which has not undergone diffraction due to reflections from internal components. The role of the collection optic was to collimate the scattered light by positioning the light sheet at the focal length of the lens. The focussing lens should then focus the collimated light onto the slit of the spectrometer. Theoretically, as long as the lenses are aligned, the light collected at the focal point of the collection optic will remain focussed at the spectrometer slit, irrespective of the distance between the two lenses. It was this property that allowed the focussing optic, spectrometer and camera to remain fixed with only the collection optic repositioned in response to the repositioning of the light sheet [Eckbreth, 1988]. The validity of this statement is only true for light collected directly on the optical axis centre line. For light collected off the axis centre line a greater lens separation resulted in reduced effective solid angles, but this effect was negligible for small lens separations. While the theoretical operation of the collection/focus lenses was well understood, the practicalities were less obvious. The first issue was the lack of visible light with which to confirm that the beam was being focused on the spectrometer slit. Secondly, the focal length of a lens is dependant on the wavelength of the light passing through it. Therefore, even careful initial positioning of the optics, at the manufactures specified focal lengths derived using visible light, did not guarantee collimation and waist of the focussing optic being located a the spectrometer slit. It was therefore important to consider the following: for a given lens combination, such as that in Figure 5.12, there will be a position in front of the collection lens at which an object can be placed that will result in the image of that object being projected at the entrance slit of the spectrometer. This point does not have to lie at the focal length of the collection lens, but is rather a 135 function of the relative position between the two lenses, given the alignment of the two lenses and spectrometer slit. It is for this reason that a thorough ‘sweep’ of the light sheet position is necessary in order to find this position given that no visible alignment technique could be used, §G.2. Similarly, while perfect alignment between the lenses and spectrometer was strived for in order to optimize solid angles and collimation, small misalignments still resulted in high collection efficiencies. Once the collected light had passed through the lenses it was clipped using masks positioned to cover the top and the bottom of the spectrometer slit. From the ray tracing diagram, Figure 5.13, the size of the collection volume could be seem to be dependant on the width of the slit aperture and the combination of the focal lengths of the collection optics. For the collection/focussing optic combination, with focal lengths 300/200 mm respectively, in combination with a slit width of 0.75 mm, the width of the collection volume was 1.125 mm (300/200 x 0.75 mm). Figure 5.13 Ray tracing diagram for the collection optics. Depth of field Spectrometer aperture x z Collection volume Plano-convex spherical lenses 136 The same treatment can also be applied to the slit height; therefore, with a slit height of 14 mm the collection volume would be 21 mm in height. However, the confines of the channel restrict the region of gaseous generated Raman signal to a height of six millimetres. Beyond six millimetres the optical system continues to collect light, but from sources such as elastic scattering, fluorescence and Raman scattering from solid surfaces such as the window and anode. In the strictest sense, elastically scattered light should have no effect on the spectral images as the spectrometer should completely separate out any light from outside the spectral range over which the spectrometer is set to operate. Fluorescence and solid Raman signals were however a problem, as the number density of molecules were many times greater than the gas phase, resulting in very strong signals in relation to the gaseous Raman signal. Again, this was not theoretically a problem as one dimensional Raman spectroscopy should be able to differentiate between light signals derived over the spatial extent of the light sheet in the y-direction. Practically however, issues were encountered with saturation of the ICCD due to the intensity of the surface derived signals, which could potentially permanently damage the device. Similarly, a high signal count on a CCD pixel can result in the inflation of adjacent pixel counts due to “bleeding” of the signal. It was therefore necessary to mask the region of the ICCD that would otherwise collect light from the solid surfaces within the reforming channel. As already discussed, the spectrometer slit should be located on the image plane produced by the lens combination. It is therefore only on this plane that complete masking can occur. The spectrometer used - an Acton, single stage, SP-2356, 300 mm spectrometer - had only slit width adjustment, which therefore necessitated an additional method with which to adjust slit height. However, the execution of the liquid filter imposed limitations on the position of the masks with the effect that they could only be mounted a distance away from the spectrometer entrance slit. In addition, the masking 137 of the slit was required to be adjustable. This was due to a small misalignment between the collection systems optical axis and the orientation of the reformer channel in the horizontal plane. The result of this misalignment was that as the light sheet was moved along the channel the position of the collection volume changed with respect to its projected image on the ICCD. If a fixed mask had been used, light emanating from one of the solid surfaces would have become increasingly un-masked as the light sheet was traversed along the channel. Conversely, a portion of the gaseous Raman signal would have become increasingly obscured. It was for this reason that the spectrometer slit masks were needed to be adjustable and were required to be repositioned at each physical testing position in the reforming channel, as described by the experimental procedure, §G.2. The imperfect masking, with respect to the masks not being located on the imaging plane, resulted in contributions of light from the solid surfaces to continue to be collected. This was because the masks only served to reduce the effects of the light emanating from the window and anode surfaces in order to avoid ICCD saturation. By extension the imperfect masking also lead to a portion of the gaseous Raman signal to be obscured, the effects of which will be discussed in §6.11. An improved system of slit making is an area in which potential improvements in the diagnostic technique may be found, as discusses in §8.3. 5.5 Spectrometer and Intensified Charge Coupled Device (ICCD) Light which has been collected, masked and filtered from the collection volume entered the spectrometer via the spectrometer slit, set at a width of 0.75 mm. The slit size is not arbitrary as it affects both the resolution at which the spectrometer separates the light, as well as the more fundamental action of regulating the amount of light which passes through the spectrometer. Spectral resolution is a function of slit width, with 138 smaller widths resulting in greater resolution. A narrower slit results in a reduced throughput of light, lowering the overall magnitude of the spectral image; thus setting the slit size was a compromise between resolution and throughput. Spectral resolution was not a priority of the technique, particularly in conjunction with the broadband XeCl laser, which places a limit on the spectral resolution possible to no better than 1 nm. However, two factors limited how wide the slit could be set. The first was non gaseous Raman signals from the surfaces from within the channel. These regions were a significant source of non Raman light, and so were the limiting factor in terms of maximum counts on the ICCD. Efforts were made to reduce these signals through masking, but due to reasons already outlined these were imperfect. In addition, by reducing spectral resolution, secondary spectral effects such as the third vibrational mode of methane (Raman shift 3017 cm-1, 339.6m) begin to be completely obscured by the Raman Q- branch. Therefore the slit width of 0.75 mm was a compromise arrived at by trial and error. The spectrometer was set to 343 nm for all spectral images collected and utilized a grating with 1800 g/mm (grooves per millimeter), which gave a spectral range of 28 nm. The spectral image projected at the exit of the spectrometer was collected using a PCO DiCam-Pro ICCD consisting of an intensifier in conjunction with a 1280(h) x 1024(v) pixel, 8.6 x 6.9 mm2 CCD, producing 12/bit images. Due to the “on-chip” integration of the image over multiple laser pulses it was necessary to operate the CCD in a gated mode. That is to say the CCD is collecting signal counts over the entire duration of the image capture process. However, the intensifier was only triggered during the period of the laser pulse; effectively closing the ‘shutter’ in front of the CCD while no Raman signal could be being generated. This arrangement was orchestrated using a pulse generator to simultaneously trigger the laser pulse, camera intensifier and CCD. Laser pulse duration was 20 ns, camera delay 790 ns and the intensifier energised for 139 70 ns per pulse to account for variations in timings. The period of camera delay was optimised by performing a simple series of trial runs and selecting the delay resulting in the maximum signal count. The intensifier was also operated at gain of 100%. In addition to optimising the camera timings, the CCD was also operated using a ‘super-pixel’ regime. This was where a number of pixels were grouped together to form a larger pixel. The advantage of this was that the readout noise (a pixel count of approximately 80) was only incurred for a single super-pixel rather than for each pixel from which the super-pixel was constructed, thereby increasing the signal-to-noise ratio. For the work presented here the pixels were binned in groups of 64 (8 x 8 pixels). The reasons behind this choice were as follows. By increasing the size of the super-pixel the signal to noise ratio is increased and, for particularly low light applications, the utilization of the dynamic range of the CCD improved. However, the opposite effect is that the resolution of the camera is reduced. In case of spectroscopy this is not such a significant effect as image fidelity is not of principle concern. There is however a point, even when using a broadband light source, of line width 1 nm, that some spectral definition is required. Similarly, with respect of spatial resolution the advantage of one dimensional Raman spectroscopy will begin to be lost. This has particular ramifications if, as in the work presented here, there is a variation in the position of the image on the CCD with respect to channel position. This in turn may result in Raman signals from the edges of the gaseous flow field being ‘lost’ in the large signal count of a pixel also receiving light originating from a solid surface, limiting the spatial resolution when comparing images from different positions from within the channel. Each super-pixel therefore represented a domain approximately measuring 0.16 mm x 0.175 nm. The camera software also enabled regions of the CCD to be ‘turned off’ and as such the full width of the CCD was used, but only a vertical portion of 352 pixels (44 super-pixels). 140 6 Experimental Test Matrix and Description of Data Collection Methodology In this chapter the rationale behind the experimental test matrix is presented, incorporating the need to run the reformer within the boundaries of a semi-realistic operating envelope, over a variety of conditions to demonstrate the suitability of Raman spectroscopy, and in such a way so as to promote efficient data collection and minimise unnecessary experimental errors. The limitations of the humidification system are examined and the efforts to minimise these effects discussed. The experimental procedure itself is then described, with focus on possible areas of experimental error and variation. Similarly, the justification for non-randomization of experimental test points is presented. Finally, the number of spectral images over which single test point is averaged is experimentally validated and considered in relation to the background image subtraction technique used. 6.1 Test Matrix Design The primary aim of the experimental work described in this thesis was to run the fuel cell reformer rig under a number of widely differing operational conditions in order to determine if Raman spectroscopy is a practical diagnostic tool for this application. From a purely academic perspective however, the operational parameters of interest for the reformer were chosen to be the quantity of methane being reformed, the overall volume flow rate of the reforming gas, the quantity of steam available to the reformation and gas shift reactions, and the global reformation temperature. The following subsections describe the development of the test matrix that was followed during this work. As will be shown, there are two major limiting factors restricting the extent of the reformer’s operational window. These are the maximum 141 amount of gas that the bubbler type humidifier is able to treat in its current configuration, and the operational ranges of the rotameters, which are a compromise between the minimum and maximum flow rates that can be metered accurately by a single rotameter tube. These factors, in turn, limit the amount of methane that can be reformed, and the amount of excess water vapour that can be supplied to the channel. Additional factors such as the ability of the humidifier to reach set points were also considered from the perspective of developing a practical and efficient experimental matrix. 6.2 The Experimental Limitations of Gas Humidification The amount of gas that could be humidified was limited by the maximum flow capacity of the humidifier, which was 1.0 SLPM. In turn, the maximum temperature at which the humidifier could be operated was 95 °C which corresponded to a humidity factor of 0.818 (where the humidity factor is the relative amount of water vapour, by volume, which can be added to a dry gas stream). Therefore, the theoretical maximum flow rate was calculated to be 1.818 SLPM (assuming the gases and water vapour mix as ideal gases). However, the process of selecting an appropriate level of humidification for the rig is complicated by the problem of providing sufficient water for the reformation and gas shift reactions (Equations 1.3 and 1.4), the overall reaction for which is given by Equation 6.1. )(4)()(2)( 2224 gHgCOgOHgCH +→+ molkJH o K /4.191973 +=Δ (6.1) Equation 6.1 states that for every mole of methane, two moles of water are required for the reformation and gas shift reactions to go to completion which, due to the ideal gas law expression pTRnV /~= is also the volumetric ratio of reactant gases. 142 Insufficient water would prohibit carbon monoxide from being converted into carbon dioxide and reduce the amount of hydrogen produced by the reformation process. However, as described above, the humidifier is only able to operate at a humidity factor of 0.818 resulting in a methane/water ratio of 1:0.818 by volume - far below the necessary ratio of 1:2. From the perspective of an operating SOFC this is not a significant issue, as the fuel cell reaction itself results in the in the production of one mole of water for every mole of hydrogen consumed, the overall reaction given by Equation 6.2. )(2)()(2 222 gOHgOgH →+ (6.2) Note that production of water through the mechanism of Equation 6.2 relies on hydrogen being present, i.e., the water producing reaction can not occur unless some amount of methane reformation has already taken place. Thus, it might be considered that the primary purpose of humidification in direct internally reforming SOFCs is to ensure that the reformation can take place the moment the gases come into contact with the anode surface. In the author’s reformer rig, with the maximum humidification possible through the bubble type humidifier limited to a factor of 0.818, it was necessary to use a second ‘inert’ humidified gas stream to ‘carry’ additional water to the reaction zone. Nitrogen was therefore used as a ‘carrier gas’ in these experiments with its sole purpose being to transport water to the reaction zone and otherwise play no part in the reformation and gas shift reactions other than to have the effect of diluting the reacting gases. The use of this ‘carrier gas’ was an additional factor in determining the flow rates through the channel, placing further limitations on the operating conditions at which the reforming channel was able to operate. As will be discussed further this arrangement is not ideal, 143 but was the best that could be achieved with the humidification technique available. An area of future work will be to improve the humidification technique to avoid the complications of using a carrier gas. 6.3 Efficiency of Data Collection As will be discussed shortly, §6.5, the time required to collect data at each operating condition and measurement location was approximately 30 minutes. Therefore, for each test point a 30 minute window was allocated – where the 30 minute window does not include time for heater start up, cool down or time taken for the humidifier or reformer channel to reach steady state. In the final test matrix twelve data sets were to be collected at seven physical positions in the channel (0, 10, 20, 30, 50, 70, 87 mm from the front edge of the anode surface), Figure 6.1, at two different temperatures (600 and 700 °C). The time taken for the humidifier to reach a new set point was at least 30 minutes, or substantially longer if the humidifier needed to reach a lower water bath temperature. Therefore, if the humidifier had been required to reach a new set point for each condition it would have restricted operation to only four test points per day of running, causing data collection to run to 42 continuous days. However, if the humidification of several test points could be met by a single humidifier setting the quantity of data collected per day could be substantially increased. It was for this reason an additional stream of nitrogen was routed to the reformer channel which bypassed the humidification process. In doing this a single humidifier set point - in actuality providing greater humidification than theoretically necessary - could be adjusted using a dry nitrogen stream which ensured the overall humidification and total flow volume flow rates of the experimental test matrix were satisfied, while eliminating a unique humidifier setting for each experimental point. 144 Figure 6.1 Schematic showing the positions physical data collection points on the non-operating anode half cell As described previously, for a given quantity of methane, an amount of nitrogen was passed through the humidifier to provide the required amount of water for the reformation/gas shift reactions. Therefore, by varying the amount of nitrogen passing through the humidifier, the amount of water supplied to the reaction zone could be easily and quickly controlled allowing the methane/water ratio to be easily studied as an experimental variable. However, in changing the amount of nitrogen flowing through the humidifier, the total volume flow rate to the reformer would of course change, which would likely make isolating purely the effect of the methane/water ratio problematic. Accordingly, a nitrogen line bypassing the humidifier was added to the experimental system, Figure 5.1. so that the total flow rate to the reformer could be held constant by the addition of a ‘dry’ nitrogen stream. The ‘dry’ nitrogen line was also used when test points required similar, but not identical, levels of humidification. By increasing the humidification factor of the humidifier for the test point requiring lower humidification, a reduced amount of humidified nitrogen would be needed to carry the water to the 100 mm 50 mm 0 mm 10 mm 20 mm 30 mm 50 mm 70 mm 87 mm Direction of gas flow 145 reaction zone, this reduction in total flow volume flow rate was then compensated for by the ‘dry’ nitrogen stream. Introduction of the nitrogen bypass stream resulted in only three humidifier set points being necessary for eight experimental test points. As a consequence, a single physical position in the reformer could be interrogated in a single day, greatly reducing experimental time. 6.4 Range of Experimental Variables As discussed in the introduction to this chapter, the primary variables of concern in this study were methane flow rate, overall volume flow rate, humidity ratio, and reformer temperature. From Appendix A.1, it can be seen that for a ‘worst case’ scenario, attempting to calculate the highest flow rates over the anode surface, a methane flow rate of 0.2 SLPM was calculated for a 50 x 100 mm2 anode at a current density of 1.0 Acm-2 and fuel utilisation factor of 0.5. However, this calculation (which does not include the effect of the nitrogen carrier gas) goes on to predict an absolute worst case scenario total flow rate of 1.0 SLPM if the methane is over-humidified by factor of two. Accordingly the potential exists for the volume flow rate in the author’s reformer rig to be far in excess of that found in any practical SOFC. In an effort to contain the overall flow rate in the rig to a more realistic value the fuel utilisation factor was therefore increased to an unattainable, but more realistic 1.0 - bringing the methane requirement to 0.1 SLPM. However, even this reduction in methane flow rate restricted the lower bounds of methane flow rate as humidification required too much nitrogen to flow, even with the humidifier operating at maximum humidity. The upper methane flow rate was therefore restricted to 0.09 SLPM and total volume flow rate to 1.0 SLPM. 146 The humidity factor, the amount of water supplied with respect to the ideal ratio as stipulated by Equation 6.1, was also constrained by the maximum humidification able to be delivered by the humidifier. While the ideal ratio of methane to steam is 1:2 this could only be increased to 1:3 before overall flow rates became excessively inflated by subsequent nitrogen flow rates. Selecting the lower bound of total volume flow rate was also restricted by humidification. By reducing total volume flow rates the amount of carrier gas able to be used was also restricted, resulting in levels of humidification that were unobtainable. The lower bound of methane flow rate was not constrained and selected by the author to be 0.06 SLPM, as it was thought that to use flow rates any lower than this value would result in species concentrations of no practical interest to SOFC researchers. Reformer temperature was restricted by several factors. The first was the molecular density at high temperatures, which resulted in spectral images of considerably lower intensity than at 21 °C. At 700 °C the molecular density is approximately only 30% of that at 21 °C which yields approximately the same percentage reduction in image intensity. In addition, there was a desire to gather a complete set of data without any unforeseen equipment for component failures, so the channel was operated a slightly conservative upper temperature of 700 °C. The lower temperature was primarily concerned with providing an environment that was still sufficient to sustain a reformation reaction. For this reason the lower operating temperature was chosen to be 600 °C. Accordingly, the final experimental matrix selected was as follows: 147 Variable Upper Value Lower Value Methane Flow Rate 0.09 SLPM 0.06 SLPM Total Flow Rate 1.0 SLPM 0.6 SLPM Humidity Factor Methane : Water 1:3 1:2 Reformer Temperature 700 °C 600 °C Table 6.1 Experimental test matrix The operating conditions specified by the 2 x 2 x 2 x 2 experimental matrix shown in Table 6.1 were repeated at each of seven measurements locations within the reformer channel. These measurement positions were, respectively, 0, 10, 20, 30, 50, 70 and 87 mm along the channel, where “0 mm” was determined to be the beginning of the reactive surface of the anode sample. Note that the position of 87 mm was the maximum extent at which the collection optic could be positioned on the optical rail. 6.5 Experimental Procedure A practical and detailed description of the operation of the experimental apparatus is contained in Appendix H, with what follows a description to provide a record of the experimental procedure itself. In addition, aspects of the experimental procedure are presented with respect to potential experimental errors and the reduction thereof. Once the heater, humidifier, laser and camera have been brought to their respective operating temperatures (see §H.2 and H.4) the process of Raman signal calibration could begin. This was necessary because, while the relative normalized cross sections of the various chemical species present during the reforming process remain constant for a given temperature, the intensity of the signals varied with respect to the 148 physical position in the channel at which the signal was derived. It was therefore necessary to calibrate the Raman signal response for hydrogen and nitrogen at varying mole fractions, a process that would also prove useful in quantifying the effect of air leaks (§7.6). The calibration process consisted of passing dry hydrogen and nitrogen mixtures ratios of ratios were 20/80, 50/50, 80/20 and 100/0 respectively, at a total volume flow rate of 0.5 SLPM. At each mixture ratio, 30 images, consisting of 500 laser pulses per image were collected. The order at which these test points were carried out was always that as listed above as this would allow the anode to experience a pure hydrogen atmosphere for a minimum of 30 minutes prior experimental testing, which served to remove any oxidation that my have occurred during the previous days cool down procedure. With the calibration process complete it was necessary to begin passing humidified gases over the anode surface. Again, a pure hydrogen gas stream was used to perform this operation as it would serve to preferentially react with any oxygen that may have entered the humidifier system since it was last used, thus preserving the reduced nature of the anode. At this stage the humidifier has been brought to the necessary temperature for the first test point, but without gases flowing through it. As the gases are unheated the commencement of gas flow through the humidifiers water bath results in a temporary drop in temperature before the humidifier heater controller can return to the set point. Similarly, an increase in gas flow rate will result in humidifier temperature variation, so it is important to begin passing the gas mixture of the first test point through the humidifier as soon as possible so that steady state operating conditions can be reached in a timely manner. 149 6.6 Execution of Experimental Matrix Ideally the full test matrix (16 operating conditions at seven measurement locations) would have been executed entirely randomly in an effort to eradicate systematic error. However, factors such as the humidifier temperature response times, time taken to reposition optics and slit masks, hydrogen and nitrogen Raman signal calibration, and the repeatability of the optical set up as a whole prohibited this approach in the author’s experiments. Instead it was decided to cycle through the experimental test matrix at each physical position for a given temperature. And while this was a questionable practice from a purely statistical perspective it brought with it several advantages as well as making data collection more efficient. The main advantage was the consistency in the positions of the light sheet and collection optics. As described in the image setup procedure, §G.2, the process of optimising the position of the light sheet was a laborious exercise. While every effort was made to implement a systematic and repeatable procedure for the positioning of the light sheet and spectrometer slit masking there were invariably variations in the final optics and masking positions. It was for this reason that calibration of the Raman signals of hydrogen and nitrogen at each physical position was so important to account for these variations and eliminate their effects. By working through the reformer operating conditions at each physical position the effects of variation in optics positions were further eliminated by ensuring that the optical set up was identical to allow direct comparisons between test points. Consistency in experimental procedure also minimised variations in the method of arriving at each operating condition. By identically cycling through the calibration procedure and test matrix, each operating condition was preceded by exactly the same process. Therefore, the humidifier was allowed to reach steady state for each test point 150 in an identical way, eliminating the contribution of this parameter to the procedural error. It should however be noted that the aim of this thesis is not to prove a scientific hypothesis in the traditional sense. If however the aim of this thesis were to prove the hypothesis that a new design of humidifier was more effective, or reached operational steady state more quickly, a statistically robust procedure would have to be applied. As it is, the aim of this thesis was to explore the suitability of Raman spectroscopy to high temperature reformer/SOFC applications and the removal of component or operational variations is an advantage in this aim. 6.7 Number of Images Collected One experimental variable that had to be taken in to consideration was the number of spectral images to average over. This was important because a single image was subject to a significant contribution from shot to shot variation. The causes of this variation were fluctuations in laser power, the noise associated with using an ICCD, and the very nature of the Raman scattering process. An additional factor, distinct from that of data collection was the effect of the background subtraction technique introduced during data processing. Laser beam power was regulated via the laser control software which in turn regulated the charge voltage necessary to produce each laser pulse. No additional laser power meter was used and so no average power reading could be recorded for series of laser pulses and corrected accordingly. While this could be seen as a deficiency in the technique, the image repetition study, which will subsequently be discussed, suggested that variation in beam power was not a concern. The use of an ICCD for the capture of weak light sources inherently introduced a noise component which resulted in significant shot to shot variation. Image averaging 151 was therefore an essential process to minimise these effects and it was necessary to investigate the number of images over which averaging should take place. Raman scattering from a finite excitation volume also had the potential to introduce shot to shot variations due to its stochastic nature. This is because Raman scatting requires molecules to be of at a specific vibrational energy level to elicit the Raman response. The number of molecules occupying the energy level will fluctuate to an extent, and such is the efficiency of the light collection and intensification system that this can potentially have an effect on the final pixel count recorded. This effect was particularly noticeable when generating spectral images from low numbers of laser pulses. For single digit laser pulse spectral images it could be seen that elevated pixel counts had been recorded in the spectral and spatial locations predicted for the gaseous species being excited, but by no means was the distribution of the signal uniform in intensity or location. Fortunately, with 500 laser pulses being required to generate a spectral image of sufficient intensity, the effects of variation in laser power and Raman signal were minimised. Variations due to ICCD noise could only be eliminated by image averaging, which in turn would further reduce the affect of laser power and Raman scattering variations. The background subtraction technique was also affected by noise present in the spectral image, primarily that introduced by the ICCD. This was because the background subtraction technique consisted of a morphological structuring element, as discussed in §7.1, that returned the lowest pixel count present along its length as the value for the pixel being evaluated. For spectral curves containing a large component of noise the minimum pixel count will be significantly lower than the trend indicated by the spectral signal’s curve. This in turn leads to an under predicted background signal to be subtracted from the spectral curve. Only through image averaging will the effect of noise 152 be increasingly eliminated, thus improving the background subtraction techniques’ ability to accurately quantify the background level. It is for this reason that when initially averaging over a small number of images the species line count is artificially inflated due to insufficient background subtraction taking place, Figure 6.2. Figure 6.2 Background subtracted hydrogen portion line counts vs. number of images averaged. At position “55 mm” along anode surface, gas mixture 50/50, H2/N2. Despite consistent line count values being observed through the averaging of approximately 20 images and greater, the decision was made to average over 30 images as this increased the robustness of the technique and more closely satisfied the central limit theorem. 153 7 Data Processing, Results and Discussion 7.1 Image Processing The raw experimental results from each of the test points examined in this study comprise of thirty 12-bit images, each measuring 160 x 44 pixels (w x h) and each representing the integrated Raman signal from 500 laser pulses. From these thirty images a single ensemble averaged image is generated; thus reducing influence of image-to-image variation. The raw images encapsulate a spectral range of approximately 28 nm and a physical height of approximately 6 mm. Note that the physical height that is imaged is greater than the physical height of the channel which is 6 mm minus the height of the anode material (≈ 1 mm) this results in the raw images showing the effects of light scattered from the anode surface and window (see Figure 7.1). Figure 7.1 Un-cropped, 30 image averaged Raman spectra detailing additional light sources from anode surface (top) and sapphire window (bottom) Accordingly, each ensemble averaged image was cropped prior to analysis to ensure that the Raman signal, particularly at the anode surface interface, was free from influence from light scattered by the surface itself. The cropped image was subsequently N2 331.8 nm CH4 338.4 nm H2O 347.0 nm H2 353.2 nm 154 divided vertically into five equal height portions or segments each of which was analyzed separately in an effort to observe potential concentration gradients perpendicular to the anode surface, Figure 7.2 illustrates the vertical subdivision of the images. The height of each segment is seven pixels, representing a physical distance of 1.2 mm in the test section. It is important to note that the presented images are inverted with respect to the physical orientation of the channel and that the uppermost portion (portion one) therefore refers to the region adjacent to the anode surface and portion five to the region adjacent to the window surface. Similarly, it should be kept in mind that the horizontal axis has no spatial meaning and only represents variations in the spectral dimension. Figure 7.2 Un-cropped, 30 image averaged Raman spectra with numbered portions The final stage of image processing was to approximate the background noise present in the raw Raman signal and then to subtract the background signal from the raw data. The act of background subtraction is non-trivial as the background signal represents a substantial portion of the Raman signal, so errors in estimating this term will result in large inaccuracies, particularly for species of low molar concentrations or small Raman cross sections. Figure 7.3 represents a typical Raman spectra recorded during the course of the research presented herein. From the figure it can see that the magnitude of the 24 nm 7 mm 4. 2. 1. 3. 5. 155 background component of the signal varies with respect to the wavelength, further complicating the task of background subtraction. Methods to reduce or eliminate background noise at the point of data collection have previously been discussed in §5.4. If, as discussed therein, these techniques are not sufficiently effective in removing the background signal or are inapplicable due to low signal to noise ratios then other correction methods must be applied. One approach that is commonly reported in the literature is to approximate the background signal by a mathematical function before subsequently subtracting the predicted noise levels from the raw Raman signal. Hassel [Hassel, 1993] having to contend with a significant contribution from the unfiltered 308 nm excitation line, used the function y = a / (x - b) + c to approximate the noise level. In contrast, Bombach [Bombach, 2002] used a polynomial function interpolated by Nevilleis algorithm. Figure 7.3 Typical wavelength vs. a raw Raman signal (from portion 3), a background approximation and a processed Raman signal 156 In the present work, a morphological structuring technique was used to approximate the background noise level in preference to using a polynomial fit. Interestingly, this method was necessitated by the absence of any significant tail from the excitation line – the effectiveness of the potassium hydrogen phthalate (KHP) low pass filter [Kleimeyer et al., 1996; Saunders et al., 2010] described in §4.2.3 causing the background noise to have insufficient curvature as to ensure a sufficiently good fit from the chosen polynomial. The poor fit of the polynomial was extenuated by the broad width of the Raman lines, a natural consequence of the broadband excitation source, which results in limited background noise data being available for the fitting algorithm. In contrast, the length of the chosen morphological structuring element (a rectangle) was sufficiently large (20 pixels) as to span the width of the Raman lines but was still able to follow the general contour of the background signal. This element was traversed one pixel at a time along each portion with the central pixel being assigned the value of the minimum value contained by the structuring element. Repetition of this process in the opposite direction ensured no directional bias. Close inspection of the resulting background curve (as shown in Figure 7.3 for example) demonstrates that this technique is not ideal due to the linear nature of the background beneath the Raman peaks; however, the method does have the advantage of being consistent. The primary issue with the polynomial method was that the curve assigned beneath the Raman lines would vary in inflection in the spectral direction. One Raman line would therefore have a larger noise term subtracted than another due to variations in inflection in the polynomial. This also left the possibility that variations in the fitted background polynomial curve would result in variations between portions, physical positions in the channel, and potentially reformer operating conditions. Variations in approximated background noise would have contributed further to experimental errors already present in the data. 157 7.2 Review of Raman Theory It will be appreciated that the primary purpose of the image processing and background noise reduction routines described above is to provide a consistent base from which the concentrations of each species of interest might be quantified. At this point then it is useful to review the appropriate Raman theory. It follows from §3.5 (Equation 2.3) the power of the Raman signal from the ith species for small solid angles is given by: i i iiir nPP ησ lΩ⎟⎠ ⎞⎜⎝ ⎛ Ω∂ ∂=, (W) (2.3) Where; Pr,i = Raman signal power of the ith species Pi = Incident light power ni = number density of the ith species ∂σ ∂Ω ⎛ ⎝ ⎜ ⎞ ⎠ ⎟ i = vibrational cross section of the ith species Ω = solid angle l= the sampling extent (measurement volume) ηi = the efficiency of the collection optics at the wavelength of the ith species It follows that if the irradiated region is of constant dimension and is subject to constant incident light power and is interrogated by a fixed set of optics, then the power of the Raman signal from the ith species is equal to some constant value multiplied by the number of molecules of that species in the sampling extent: 158 Pr ,i = Ai ni (W) (7.1) Where; Pr, i = Raman signal power of the ith species Ai = Constant of proportionality for ith species = Pi ∂σ ∂Ω ⎛ ⎝ ⎜ ⎞ ⎠ ⎟ i Ωlηi ni = number density of the ith species in measurement region In this thesis work, Raman power is implied from the intensity level (counts) of a 12-bit grayscale image captured by an integrated CCD camera therefore Equation 7.1 should be modified to include a term for the quantum efficiency of the camera system at the wavelength of the ith species such that; Ir,I = Bini (Counts) (7.2) Where; Ir,I = the grayscale intensity of Raman signal from the ith species Bi = Constant of proportionality for ith species = Qi Ai Qi = Quantum efficiency of ICCD camera at wavelength of ith species emission ni = number density of the ith species in measurement region Since the Raman signals for respective species are spectrally discreet it is theoretically possible to experimentally determine a constant of proportionality for each species of interest and thereby directly determine the number of molecules present in the interrogation volume. However, this approach would require individual calibration experiments to be performed for all species of interest, which – depending on the nature of the experiment – might be considered impractical. Therefore, a popular approach to 159 the quantification of Raman spectra is to reference the Raman signal strength of the species of interest to the signal strength of a common reference species, which is by convention nitrogen, using published data for the relative normalized differential Raman scattering cross section [Eckbreth, 1988; Schrötter et al., 1979; Schrötter, 1982; Bombach, 2002; Fenner et al., 1973; Long, 1977; Zhao et al., 2001]. 7.3 Relative Normalized Differential Raman Scattering Cross Sections On initial inspection it appears as if this task is trivial, as the most critical pieces of information – the relative normalized differential Raman scattering cross sections for the Q-branch of the key gaseous species examined in this work – have been reported extensively within the literature [Eckbreth, 1988; Schrötter et al., 1979; Schrötter, 1982; Fenner et al., 1973]. In the most widely quoted work of this type, Schrötter et al. [Schrötter et al., 1982] referenced upward of 30 sources at various wavelengths for commonly found molecules and compounds. In principal, with this information known it should only remain to quantify interrogation volume, optical efficiency, and any variation in quantum efficiency for specific Raman shifts. However, closer inspection of the data compiled by Schrötter et al. and further cross referencing with figures published by Eckbreth [Eckbreth, 1988] and the smaller (six source) literature review of Fenner et al. [Fenner et al., 1973] show significant variation in the relative normalized differential Raman scattering cross sections values reported for the Q-branch of the three major species of interest in this work, Table 7.1 refers. 160 Species Schrotter et al.  Eckbreth  Fenner et al.  N2 1.0 1.0(Q) 1.0 H2 3.4 – 4.2 3.1† 1.6 (Q(1)) 2.3-2.8 (sum) CH4 6.8 – 9.5 7.5† (v1) 5.0† (v3)* 6.0 – 8.1(v1) H2O 2.5 – 7.7 7.8† (Q) N/A Table 7.1. Variations in published relative normalized differential Raman scattering cross sections. † Using Q-branch N2 normalization, * The third vibrational frequency of methane is separated from the first by 100 cm-1 In many respects the wide variation in the reported relative normalized differential Raman scattering cross sections as shown in Table 7.1 is surprising. As has been discussed in previous chapters of this work, Vibrational Raman Spectroscopy (VRS) is the result of the energy level of a molecule being raised to a virtual state, before returning to a vibrational energy level either greater or less than the initial state. It is the pure increase/decrease in vibrational energy level, combined with a weak rotational/vibrational interaction that causes all instances of vibrational excitation, regardless of rotational energy sate, to experience an effectively equal shift in wavenumber. This inelastic scattering of light produces the relatively strong Q-branch signal, the spectral position and magnitude of which enables species identification. From this definition it would appear there is little scope for ambiguity when presenting results. However, Table 7.1 shows not only large variations in the reported values, but also inconsistent notation when publishing such data. The relative normalized differential Raman cross section of nitrogen is by convention unity; as all other species are 161 normalized by nitrogen’s cross section. Eckbreth [Eckbreth, 1988] however publishes two figures for nitrogen at identical Raman shifts and makes the distinction between them by using a “Q” in parenthesis. This is presumably to indicate this is the value recorded for the Q-branch, however no explanation is given for the other larger value. Eckbreth then goes on to quote values for hydrogen, methane and water, but somewhat confusingly gives no indication as to what branch or vibrational mode is reported for hydrogen, repeats the Q notation for water and then quotes cross sections for the respective vibrational modes of methane. Fenner et al. [Fenner et al., 1973] on the other hand specifies that the cross section of nitrogen is that of the Q-branch of the first vibrational mode. When quoting values for hydrogen the word ‘sum’ is included, but no further information is provided about the bounds over which the cross section is presumably summed over. A second figure for hydrogen, which was experimentally determined by Fenner et al. and which is denoted as (Q(1)), to indicate the Q-branch for the first vibrational mode, is distinct from the other values listed as determined by other researchers previously. For methane Fenner et al. quotes values for the first vibrational mode, with no mention of the Q- branch. Schrötter et al. [Schrötter et al., 1979] state that all values are Q-branches of the vibrational bands (derived from multiple excitation wavelengths), but inconsistently switches between terminologies even when presenting the same numerical values for identical species. It is therefore doubtful that all data compiled by Schrötter et al. can be relied upon to relate to Q-branch cross sections, which in turn could suggest that variations in presented values may also be a function of inconsistent spectral line summing as indicated by Fenner et al. [Fenner et al., 1973]. It is clear from the above that the selection of relative normalized Raman cross sections from the literature requires significant care. Moreover, it also clear that cross sections appropriate to the 162 present experimental set-up (particularly with respect to the broad spectral width of the excitation source) are unlikely to exist within the literature. The lack of definitive database of Q-branch Raman cross sections warrants further discussion. In general this can be attributed to two major factors. The first, and most general reason why an accurate database of Q-branch Raman cross sections does not exist is that from the perspective of an experimentalist wishing to use Raman spectroscopy as a diagnostic tool, there appears to be little demand for more accurate data. The second major factor, which directly affects the accuracy of the data that has been published, is that there are no standards outlining the conditions in which Raman cross sections should be recorded: variation therefore occurs due to temperature differences between experiments [Eckbreth, 1988], and differences in the spectral resolution and the type of excitation source employed (specifically the spectral width). As shown by Bendtsen [Bendtsen, 1974] the Q-branch of nitrogen is in the order of 10cm-1 wide. Given a 308 nm excitation source, this equates to a spectral width of approximately 0.1 nm. Equally, the first 10 rotational lines surrounding the Q-branch occupy approximately 0.9 nm (80 cm-1), which necessitates a sufficiently spectrally slender excitation line and high resolution spectrometer arrangement if the Q-branch is to be isolated. If either of these two requirements is not met, the reported figure for the Q-branch cross section may be inflated due to contributions of the rotational lines or spectral smearing effects. With these two sources of inaccuracy outlined and other factors such as contributions from additional spectral lines (e.g. the third vibrational energy level of methane) the large variations in the reported Raman cross section are perhaps more easily understood. With respect to the research presented in this thesis the spectrometer/camera arrangement is capable of recording approximately a 28 nm spectral range. Therefore, 163 even without pixel binning and the requirement to use larger than ideal spectrometer slit widths, which improve intensity counts at the expense of spectral resolution, the maximum possible spectral resolution would be in the region of 0.022 nm/pixel. However, given the limited power of the scattered light the optical collection system has had its resolution reduced to 0.175 nm/pixel. Perhaps most importantly though, the use of a broadband excitation source of width 1.0 nm effectively precludes the possibility of accurately reporting an isolated Q-branch Raman cross section free of any contribution from adjacent rotational lines. It is clear from the above that the values for Q-branch Raman scattering cross section determined from the author’s experimental arrangement will likely vary from those values published in the literature due to the novelty of the chosen excitation light source and experimental apparatus. Note that such variation is expected without having given consideration to additional errors such as background noise, excitation line suppression and variations in temperature between the author’s experimental setup and the diagnostic set-up that generated the literature data. However, while values of Raman cross section derived from a particular diagnostic set up are not necessarily accurate, they are generally precise. This therefore makes calibration a vital part of the author’s experimental technique and while the values of relative normalized differential Raman cross section published in literature are not expected to provide an exact match to the author’s results they do provide an indication of where the experimentally determined cross sections for key species might be expected to lie. 7.4 Raman Line Summing As indicated in §7.3 above, the preference in the literature is to quote relative normalized differential Raman cross sections for the vibrational Q-branch scattering. 164 However, it is now clear that the literature data cannot be fully trusted due to inconsistencies in notation, methodology and spectral resolution. Equally, it has been shown the author’s experimental apparatus is unable to achieve the spectral sensitivity, particularly at the shorter UV wavelengths, to distinguish between the spectrally narrow Q-branch and the adjacent rotational lines. The spectral range over which to sum the Raman lines is therefore somewhat flexible. In the work presented in this thesis the Raman signal is summed over three pixels centered on a species ‘centre line’, the position of which is kept consistent throughout the entire data processing procedure. The reasons for using three pixels are as follows: Additional signal averaging – despite the final spectra being the result of averaging 30 individual images, and the spectra for each portion being the average of seven rows of the averaged image, there is still some small scale variation in the signal intensity. By using 3 pixels this variation, and the effect on the final Raman signal count, is further reduced. Total incorporation of the Q-branch – the combined effect of the spectrometer, CCD size and pixel binning results in a spectral resolution of 0.175 nm/pixel. Bendtsen [Bendtsen, 1973] showed that the Q-branch of nitrogen was in the order of 0.1 nm in width, so easily bound by the width of a single pixel. However, the relative coarseness of the collection device means that the spectral position of this spectral bin cannot be adjusted; it is therefore possible that the spectral boundary of a single pixel could intersect the Q-branch. Summing over three pixels reduces this likelihood while also ensuring that the contribution from the Q-branch is always collected despite the existence of small variations in Raman line positions due to minor changes in the alignment of the collection optics. Exclusion of secondary effects – In the case of species such as methane, the close spectral proximity between the first and third vibrational lines leads will cause wide 165 spectral summing of the Raman lines to incorporate the effects of not just of the Q- branch and the rotational lines in the immediate vicinity but also an additional vibrational line. While theoretically the magnitude of this second spectral line should also be proportional to the species concentration, it adds an additional variable that is difficult to quantify without an additional study into this specific effect. Therefore, to avoid any possible complications in this area spectral summing is restricted to a small band around the centre line of the Raman shift. Increased distinction between Raman signal and background noise – in the case of species of low molar concentrations it was determined that a clearer signal could be derived from a narrower spectral bin. In the case of hydrogen (Raman line of 353.2 nm for 308 nm excitation) in Figure 7.3, the small absolute signal will suffer from an increased relative noise effect the larger the spectral band over which it is summed over. By keeping the spectral band small, the relative signal to noise effects will be minimized. 7.5 Raman Signal Calibration The previous sections of this chapter have shown that prior calibration of the Raman signal from the reformer apparatus is an essential process if the data produced during the reformation process is to be related back to species concentration in a quantitative or semi-quantitative manner. Ideally, calibration experiments should be performed for all of the species of interest – in this case, nitrogen, methane, water, and hydrogen – at temperatures representative of the operating temperatures of the reformer. However, herein lies a significant problem: the nickel cermet that comprises the anode must be fully reduced in order to offer the maximum number of catalytic sites on which to host the reformation and gas shifts reactions. This requires that the anode is 166 not exposed to non oxidizing atmospheres at temperatures above 200 °C. Accordingly, it is not possible to directly calibrate the Raman signals of methane and water (steam) in the reformer at the temperatures of interest. Other methods must be applied to determine appropriate values of normalized differential Raman cross sections for these species in this application as will be described shortly. Calibration experiments for hydrogen and nitrogen were directly integrated into the experimental procedure for the investigation of methane/steam reformation in the apparatus. The calibration procedure itself consisted of flowing mixtures of nitrogen and hydrogen across the anode surface at various mixture ratios prior to performing any reforming test at any given temperature and physical location within the reformer. Four different hydrogen/nitrogen mixture ratios were supplied to the reformer during each calibration experiment at a constant total gas flow rate of 0.5 SLPM. The chosen mixture ratios, listed in the order in which they were supplied to the reformer, were 20/80, 50/50, 80/20 and 100/0 for hydrogen/nitrogen respectively. This process of increasing/decreasing the respective mole fractions of hydrogen/nitrogen to obtain a calibration was similar to the technique used by Fenner et al.  who varied the molar concentrations of the species of interest by varying the pressure of the gas of interest and then used the linear relationship between the magnitude of the Raman signal vs. pressure (molecule number density) to calculate the Raman cross sections of each species. Note that by increasing the concentration of hydrogen flowing across the anode surface (from 20% – 100%) throughout the calibration process the author’s chosen calibration method ensures the existence of a fully reduced anode surface prior to each reformation test. Figures 7.4a and b present the background corrected raw data (counts) for, respectively, hydrogen and nitrogen for the case of a nominally 50/50 mixture (% mole) of hydrogen and nitrogen at a reformer temperature of 600ºC against horizontal position 167 within the reformer channel (measured with respect to the front edge of the anode surface). Data is shown for all 5 of the vertical portions of the recorded spectral image as described in §7.1. (a) Hydrogen 168 (b) Nitrogen Figures 7.4a and b. Graphs of typical average pixel counts collected from within the channel at identically metered hydrogen and nitrogen flow rates illustrating decreasing pixel count with increasing longitudinal channel position. 50/50 H2/N2 blend at 600 ºC. Figures 7.4a and b are typical of the data collected during the calibration experiments, which exhibit many common trends worthy of discussion. It can be seen that the signal count from portion 1, which is the portion closest to the anode surface, is significantly lower than the signal count at portions 2-5 inclusive for both species. This is a feature seen in all calibration data, i.e. for all 4 hydrogen/nitrogen ratios at both temperatures of interest (600 and 700 ºC). Similarly, the data from portion 5 – the uppermost portion nearest the window – consistently shows the lowest signal strength of all of the imaged portions at the entrance of the reformer section – thereafter steadily transitioning to the have highest signal strength at the furthest recorded position within the channel (87 mm). The other notable feature evident in Figures 7.4a and b is that, 169 regardless of the chosen portion, there is a significant decrease in signal strength with increasing distance along the reformer channel. Figure 7.5. Reduced apparent solid angle of the collection optics due to the Raman interrogation volume traversing to the rear of the reformer channel. The decrease in signal strength with distance into the channel is not unexpected. Recalling Equations 7.1 and 7.2, the intensity of the measured Raman signal has some dependency on the solid angle of the imaging set-up. However, as the location of the interrogation volume moves further back into the reformer channel the apparent solid angle of the collection optics is reduced as shown in Figure 7.5. The importance of maintaining the alignment between the collection optics and the axis of the reformer channel may also be inferred from Figure 7.5, particularly with respect to signals collected from close to the channel boundaries. It is considered likely therefore that the significantly differing behaviour of the portion 1 and portion 5 signals compared with the Reduced solid angle due to channel height restrictions Reformer channel 170 relatively consistent behaviour of the signals from portions 2, 3, and 4 are due, at least in part, to some misalignment of the optics with the channel and inefficient clipping of the image on the spectrometer slit, §5.4. Accordingly, it was decided to remove the portion 1 and portion 5 data from the calculation of average signal values to be used in the calibration process. Figures 7.6a and b below show the portion 2, 3, and 4 intensity data and the calculated average value for H2 and N2 from the 50/50 H2/N2 blend at 600 ºC. (a) Hydrogen 171 (b) Nitrogen Figures 7.6a and b. Graphs of portions 2, 3, and 4 and calculated average for hydrogen and nitrogen. 50/50 H2/N2 blend at 600 ºC. With respect to the extent of the decrease in signal intensity with increasing penetration into the channel, it is important to note that the experimental data does not fit a smooth curve – as is evident in the nitrogen data shown in Figures 7.6b for example. This behaviour is an artefact of the experimental technique. At each individual channel position at which the laser excitation occurs, the collection optics are repositioned. With each adjustment of the optics the masking elements placed in front of the entrance slit of the spectrometer which stop saturation of the ICCD due to light scattered from the anode/window surfaces must also be repositioned. It follows from Equations 2.3, 7.1, and 7.2, which contain terms for the extent of the measurement volume, incident light power, and collection optics efficiency - all of which will be affected to some degree by the repositioning of the light sheet and collection optics, that the absolute values of the 172 collected signals at different physical positions within the reformer channel are not directly comparable. The effects of optics misalignment and variations in the position of masking elements on the Raman signal are readily accounted for by performing calibration experiments at each physical position within the channel. Accordingly, the hydrogen/nitrogen calibration experiments described previously in this section were performed at every measurement location prior to operating the reformer in reforming mode. Nonetheless, some restrictions remain on the comparison of results from different physical locations in the reformer channel. Most notably, since it is not possible to directly calibrate the Raman signals of methane and water - for the reasons described previously in this section - it is not possible to make any comparison on the basis of absolute species concentration. However, comparison on a mole fraction basis is possible. By collecting calibration data at four different mixture ratios the linear response of the respective species Raman signals could be observed and the Raman cross section derived, regardless of the effectiveness of the background subtraction technique. Conversely, four calibration points incurred a significant time penalty as the data was collected using the same procedure as for reformation data. Each image was the result of 500 laser pulses at a repetition rate of 10 Hz; thus, a single calibration point is the result of averaging 30 images, taking approximately 30 minutes to collect and the full calibration process taking approximately two hours. 7.6 Differential Raman Scattering Cross Section for Water Figure 7.7, below, shows the experimental results for the 100% hydrogen calibration experiments performed at 600 ºC at all physical locations within the channel 173 (average signal counts from portions 2, 3, and 4). The significant decrease in the hydrogen signal with increasing distance into the channel is immediately seen; however, the most startling feature of this figure is the existence of significant Raman signals from the nitrogen and water lines (along with trace values of methane). This result was unexpected and prompted a detailed investigation into the source of these signals. Note that similar results were also seen at 700 ºC. Figure 7.7 Averaged species counts vs. longitudinal channel position. 100% hydrogen calibration experiments (600 ºC, average signal counts from portions 2,3, and 4). Initially, it was assumed that the nitrogen and water signals found in the 100% hydrogen test cases were the result of impurities, most likely nitrogen and oxygen, in the 174 supply gases. Accordingly, samples were taken directly from the supply bottles for analysis via gas chromatography. Note that the gas chromatograph used to perform the analysis was not able to detect water vapour. Therefore, a separate test, in which gas samples were flowed across a desiccant material (silica gel), which was weighed before and after the gas flow, were performed to check the gases for water content. No impurities or water were found in the supply gases. Further examinations were performed with gases sampled from the supply lines after the humidifier and at the entrance of the reformer channel under both hot and cold operating conditions. Again, on sampling the gases at the reformer entrance, no traces of impurities were found. At this point the examination turned to the nitrogen and water signals in the reformer. Figure 7.8 shows the ratio of H2O/N2 through the reformer for both the 600 and 700°C test cases. 175 Figure 7.8 Un-calibrated H2O/N2 ratios from 100% hydrogen calibration experiments (average signal counts from portions 2,3, and 4). The results shown in Figure 7.8 are interesting. In both cases, it can be seen that the H2O / N2 ratio is essentially constant along the channel. Assuming that the source of the water vapour is the oxidation of hydrogen (and that under these conditions that all of the available oxygen is consumed) then this would suggest that there was ingress of a gas of constant O2/N2 ratio into the channel during operation. Note that other potential sources of oxygen in the channel were also considered during the investigation of these results. Most notably, it was suggested by one observer that oxygen could have become trapped in the anode material itself, such that it was gradually released into the gas stream during the experiments. However, this scenario was considered unlikely and the hypothesis rejected due to the continuous nature of the experiments. The 100% hydrogen experiments were performed after three other hydrogen/nitrogen blends had 176 been tested in the rig, i.e. the 100% hydrogen tests occurred after oxygen free gases had been passed through the channel for a minimum of 90 minutes, in addition to the start up procedure. It was considered extremely unlikely that oxygen could have been trapped in the anode material for this length of time. In addition, gas samples were drawn from before and after the reformer section, at room temperature, and analysed using gas chromatography. The results of these tests once again showed gases free from oxygen content prior to the reformer section, but with an oxygen content on leaving the reformer section. It could therefore be determined, that at the very least, if oxygen was being released from the anode sample, it did so in a process independent of temperature. While the presumed gas leak into the reformer channel was initially regarded as a serious problem, subsequent consideration showed that it did, in fact, provide a substantial benefit in the context of the authors’ project - which is essentially the development of tools and techniques for future study of the reformer operation. As was described earlier in the chapter, published values of relative normalized differential Raman scattering cross sections for the species of interest in this work differ substantially. Accordingly, the author has indicated the desirability of calibrating individual experimental set-ups to determine appropriate Raman cross sections. Note that this is particularly important in the case of the author’s experiments where, contrary to convention, a broadband laser is used for excitation. By assuming a fixed value for the O2/N2 ratio of the gas leaking into the reformer during the 100% hydrogen experiments and assuming that all of the O2 is then consumed to form H2O then a relative normalized differential Raman scattering cross section for water can be determined for the author’s experiments. As was earlier described in §5.1, a very low flow of nitrogen was used as a purge gas for the oven surrounding the reformer section. Accordingly, the exact composition of 177 the gas surrounding the reformer is unknown. However, for the purposes of this phase of this work, i.e. development and proof of concept, it was assumed that the composition of the gas leaking into the reformer section was 79% nitrogen and 21% oxygen by volume (although the author recognizes that there is likely a higher percentage of nitrogen in the actual gas ingress). Thus, for one mole of nitrogen/oxygen blend entering the reformer we will have 0.79 mole of nitrogen and 0.21 moles of oxygen, which in turn produces 0.42 moles of water vapour if fully utilized. This gives a target value of 0.532 for the H2O/N2 ratio on a molar basis. Fitting this value to the mean of all data points at 600 and 700 ºC yields values of 2.59 and 2.54 respectively for the relative normalized differential Raman scattering cross sections of water in the author’s experiments, Figures 7.9a and b. Note that although these values are lower than that given by Eckbreth [Eckbreth, 1988], they do fall within the range of values presented by Schrötter et al. [Schrötter et al., 1979] (Table 7.1). 178 (a) 600 °C (b) 700 °C Figures 7.9a and b. Average relative normalized differential Raman scattering cross sections of water at 100% hydrogen operating condition 179 With the value of the relative normalized differential Raman scattering cross section of water calculated and the relative normalized differential Raman scattering cross section of nitrogen, by convention, unity it remained to determine the relative normalized differential Raman scattering cross section of hydrogen. This could not be determined using the same technique as used for water as the ratio of hydrogen to nitrogen was subject to the magnitude of the air leak, itself a function of channel position. It was therefore necessary to use an additional parameter to enable the unknown – the relative normalized differential Raman scattering cross section of hydrogen – to be determined. In this case it was the fact that the mole fractions of the gas mixture components must sum to unity. Therefore, while the amount of nitrogen, hydrogen and water could not be directly determined, given that the relative normalized differential Raman scattering cross sections of the components are constant at a given temperature and pressure, it was possible to calculate a relative normalized differential Raman scattering cross section of hydrogen to satisfy this condition. The above technique would however require the vibrational cross sections, as opposed to the relative normalized differential Raman scattering cross sections, to be known. This information was not available as the precise number of molecules present in the excitation volume was not known, hence the use of relative normalized differential Raman scattering cross sections and using nitrogen as the standard. But it was not possible to achieve equivalent quantities of hydrogen and nitrogen in the channel due to the error associated with the air leak. For this reason one final piece of information was used, which was that the Raman signals recorded for nitrogen and hydrogen (at various unknown mixture ratios) could be assumed to still exhibit linear proportionality. This was felt to be a reasonable assumption as the effect of additional nitrogen and water vapour to the gas mixture would only serve to make direct mole fraction calculations impossible, 180 not alter the fundamental behaviour of the Raman signal intensity versus species concentration. Using the above assertions and assumptions the relative normalized differential Raman scattering cross section of hydrogen was determined in the following manner. The process was once again iterative so that a relative normalized differential Raman scattering cross section of hydrogen could be adjusted until the solution converged. Each physical position in the channel was considered independently, as the magnitude of the air leak and apparent optical solid angle restricted comparisons with other physical positions in the channel. The Raman spectra at each mixture calibration mixture ratio was then converted to a relative number density with respect to the intensity of the nitrogen line and the respective relative normalized differential Raman scattering cross sections. This was done so that each species could be assigned a mole fraction equivalent to the relative number density divided by the total relative number density (the combined relative number densities of the three species present) at that calibration condition which by default would sum to unity. This process was repeated at each physical position in the channel and the mole fractions of both nitrogen and hydrogen were plotted against their respective Raman signals so that a first order polynomial could be fitted, Figures 7.10a and b. 181 (a) Nitrogen at 600 °C Hydrogen at 600 °C Figures 7.10a and b. Nitrogen and hydrogen proportionality curves. 182 The plots of Figures 7.10a and b are particularly satisfactory: the data shows a high degree of linearity despite the variation in mole fraction at differing positions in the channel due to the ingress of air. Similarly, the intercept of the curves being approximately zero suggest good agreement with theory for the process used. With the Raman intensity curves vs. mole fraction known for each position in the channel it was possible to compare the slopes of the hydrogen and nitrogen curves at each position and determine the relative normalized differential Raman scattering cross section of hydrogen, Figure 7.11a and b. (a) 600 °C 183 (b) 700 °C Figure 7.11a and 8.11b. Average relative normalized differential Raman scattering cross sections of hydrogen at 600 and 700 °C respectively. As with the relative normalized differential Raman scattering cross section of water the calculated value can be seen to vary with physical position in the channel. This variation can be explained by experimental and image processing errors, particularly at positions deep in the channel that suffered a large reduction in signal intensity. Of particular note however is the variation in relative normalized differential Raman scattering cross section with temperature. In the case of the relative normalized differential Raman scattering cross section of water, the variation between the two values can be considered negligible, especially in the context of the experimental error. However, for hydrogen the difference in relative normalized differential Raman scattering cross section at 600 and 700 °C is noticeable. This variation is due to the comparatively 184 large rotational vibrational interaction constant of hydrogen [Eckbreth, 1988]. The interaction constant for Nitrogen is 0.017 [Herzberg, 1945] and for water 0.0582 [Voitsikhorskaya et al., 1972] but for hydrogen this figure increases to 1.42 [Namioka, 1964]. So in addition to the general reduction in Raman cross section with temperature [Eckbreth, 1988] the hydrogen molecule also exhibits increased spectral spread in Raman shift as a result of different rotational energy levels being excited. The Raman signal is therefore effectively distributed over a wider spectral extent, resulting in a reduced Raman signal over the author’s spectral sampling extent. 7.7 Ingress of Air During the processing of the calibration data, the source of air, or rather nitrogen and water vapour (the product of catalytic oxygen combustion), was initially not clear. This was primarily because while the reformer channel was not specifically pressurized, the very fact that the gas supplied to the reformer channel proceeded to pass through the exhaust piping, the exit of which was at atmospheric pressure, would suggest that the channel was at a pressure greater than atmospheric - albeit by a small increment. Conversely, viscous fluid dynamics does not offer a mechanism by which air can enter the channel given the incremental over-pressurization of the channel. For this reason the supply gases, specifically hydrogen and nitrogen were sampled for signs of atmospheric contamination using gas chromatography. On finding the gases to be free of impurities the tubing leading to the reformer section was examined and checked for leaks (both by a pressurized leak test and gas sampling at various positions up to the reformer inlet). Only on sampling the gas exiting the reforming section could evidence of air be found in the gas stream. 185 The seemingly contradictory situation led the author to consider several scenarios; the first being that the testing process was flawed. However, repetition of the gas chromatography tests replicated earlier results and suggested the testing process be free from errors. Alternatively, it was considered that a substantial opening exists in the reformer channel, essentially exposing the channel to the atmosphere and the Raman spectroscopy results are merely representative of the mixing of the supplied gases with the atmosphere. This too would seem not to be the case, as gas sampling at the reformer exit was achieved by connecting the tube exiting the reformer to a sample bag (the gas not being drawn into the sample bag, rather forced into the bag through the action of back pressure) which would suggest no alternative, preferential gas path available. In addition, exposure of the anode sample to a significant oxidizing atmosphere has been observed to result in the oxidisation of the nickel content of the anode which in turn is accompanied by the anode reverting to its unreduced green colouration. In one set-up experiment, for example, a substantial air leak was subsequently found to be present due to a loose fitting in the inlet gas system. During this test the anode sample was observed to oxidise heavily, ultimately resulting in the warping of the anode sample, Figure 7.12. However, during the reformation tests the continued grey colouration and lack of deformation of the anode sample would suggest at the magnitude of the air ‘leak’ to be substantially less than that experienced through the presence of a loose gas fitting. 186 Figure 7.12 Oxidation and warping of the anode material when exposed to a partially oxidising environment. The direction of gas flow was from left to right. The “V” formation would suggest that a fully developed flow profile was present in the channel. The next hypothesis considered and rejected was for nitrogen and oxygen to be absorbed onto the anode’s cermet material when in the presence of air and then de- absorbed when in the presence of a reducing atmosphere (presumably driven by concentration gradient) but this hypothesis seemed unlikely on several accounts. The first was that the phenomenon has, to date, not been reported in the literature. Secondly, the operating instructions for correct handling of the anode material state to heat it to the desired operating temperature in a non reducing environment and then to dwell at the temperature for an hour while flowing a ‘safe mixture’ (96% nitrogen, 4% hydrogen) over the anode surface to ensure complete reduction. It is therefore considered unlikely that during the 100% hydrogen phase of the calibration process, during which time the anode has been at temperature for approximately two hours, the material will still be de- 187 absorbing nitrogen and oxygen. Similarly, if de-absorption is driven by concentration gradients the calibration process only exposes the anode to substantially lower than atmospheric gaseous nitrogen concentrations from the 50/50 nitrogen/hydrogen mix onwards, which is half an hour before the 100% hydrogen condition. Whereas the inlet stream has been free from oxygen for a period of approximately two hours, which would suggest a greater proportion of the total absorbed oxygen would have been expunged by this point. Thus the oxygen/water content of the stream would be expected to have begun to decrease by the 100% hydrogen condition. This however is not the case, as water production - from the perspective of mole fraction - can be seen to increase with hydrogen content, Figures 7.13a and b. (a) 600 °C 188 (b) 700 °C Figures 7.13a and b. Water mole fraction vs. channel position for the four calibration mixtures (data fitted with 2nd order polynomial to illustrate approximate trends). Figures 7.13a and b are also interesting from the perspective of the increased mole fraction of water observed with respect to increasing percentage of hydrogen in the channel bulk flow. If oxygen was being liberated from the anode material, and if this process was severely rate limited such that it took place over hours rather than seconds or minutes it is unlikely the change in hydrogen content of the bulk flow would have such a marked effect. However, if the system is considered from the perspective of diffusion, where hydrogen leaves the channel and air ingresses through ceramic gasket (a topic discussed in more detail in the following section) the variation in mole fraction of water becomes more easily understood. As increasing the concentration of hydrogen in the bulk flow will result in increased diffusion fluxes (Equation 7.5) for both hydrogen leaving 189 the channel and nitrogen/oxygen entering as a result of increasing the molar concentration gradient across the gasket. It is this combination of diffusive effects that causes the mole fraction of water to appear to increase with respect to bulk flow composition as, when reporting in mole fractions, it is only possible to report the composition of a sample rather than absolute molar concentrations. At this point the author offers a brief note of clarification about the use of mole fractions. Due to the likelihood of gases leaking from the channel and, at this stage of the discussion the potential ingress of air it is not possible to report absolute values of water production, only the amount of water vapour relative to other constituent gases in the channel. It should however be stated that if nitrogen and oxygen were being evolved from the surface of the anode and gases were only leaking from the channel the mole fraction figures reported could be directly compared to one another assuming the composition of the leaked gases were comparable to that in existence at that location in the channel. As will be discussed shortly it is not the authors’ belief that oxygen and nitrogen are being evolved or that gas leaking from the channel is the only mechanism of gas transport in the channel, making direct determination of water vapour production problematic. Returning to the hypothesis of air absorption/de-absorption, the increasing mole fraction of water demonstrated by Figures 7.13a and b suggest that an increased amount of oxygen is de-adsorbed and reacts with hydrogen at the 100% hydrogen condition than at smaller hydrogen/nitrogen ratios. As has already been described the anode material had been operated in a low oxygen environment from the moment the reformer was at 200 °C or above for a minimum of 90 minutes before the 100% hydrogen case. And while it has already been considered unlikely that anode material continues to de-adsorb oxygen after this length of time it is considered equally unlikely 190 that the amount of oxygen de-absorbed will increase with time, as the amount of oxygen absorbed must be finite. It can also be noted that the amount of water detected with respect to channel position does not support the de-absorption hypothesis. As illustrated by the measurement location schematic in Figure 6.1, the first measurement location is on the leading edge of the anode sample. Therefore, the amount of water detected at 0 mm should not only be comparatively small but also be the result of catalytic combustion of hydrogen with the de-adsorbed oxygen up until that point. Even if the positional accuracy of the excitation volume was +/- 5 mm (in actuality closer to +/- 0.5 mm) it would mean the amount of oxygen de-adsorbed until that point was the result of 5 mm of anode material. It could therefore be expected that by the last observation point (at 87 mm) around 17 times more water resulting from the de-absorption of oxygen would be detected. From Figures 7.13a and b it can be seen that the 0 mm levels of water are not negligible and even for the 100% hydrogen case the increase in water content is only approximately two times greater than that measured at the first observation point. For the de-absorption of oxygen hypothesis to be feasible the rate at which the anode material de-absorbs would have to be non-linear with respect to channel position and it is considered unlikely this is the case. An additional consideration is that the concentration gradient between the ‘absorbed’ oxygen in the anode structure and the gas stream has been at a maximum for the entire high temperature period of operation. It would then seem reasonable to suggest that a greater amount oxygen would have been de-absorbed with respect to nitrogen. This is clearly not reflected in the oxygen/nitrogen ratios, Figure 7.8, which predict an entirely feasible relative normalised differential Raman cross section for water vapour and in turn hydrogen (especially when considering the elevated temperature). Equally, as will be shown in §7.14 excess water is still in evidence during reformation 191 operation, suggesting that oxygen is still present in the channel up to seven hours after start up. Another peculiarity with this hypothesis, as discussed earlier, is the continued presence of oxygen at the exit of the reforming channel at room temperature. This would suggest that if the anode material exhibits such gas absorption properties it is not a function of temperature, further making its failure to be reported in the literature less likely. Another hypothesis suggested to the author was that instead of oxygen reacting with hydrogen to form water, water was actually present in the channel itself. As an example it was cited that for high vacuum experiments, experimental assemblies are baked out for days at temperatures approaching 300 °C in order to remove any water present. Firstly, the experiment presented here was not performed at a high vacuum which removes one of the mechanisms by which water could potentially be ‘drawn’ out of materials or crevices. Secondly, not only was the reformer channel and anode material operated at temperatures greatly in excess of 300 °C, but was only ever exposed to gaseous flows containing water vapour at the experimental set points of 600 and 700 °C. During heat up, cool down and calibration testing the gases passed through the channel were dry, giving opportunity for any residual water present to be removed through an evaporative process. This was confirmed by the desiccant test described previously, where the mass of the desiccant over which gasses were passed after passing through the channel weighed marginally less than before the test was commenced. This somewhat unexpected result was rationalised by the desiccant not being totally dry before the test but becoming more so after exposure to the dry gases for approximately two hours. Given the inconsistencies of the above hypotheses to explain the presence of oxygen it led the author to consider the principle of Occam’s razor and conclude that a) oxygen must be present in the channel, and b) an explanation must exist which does not 192 require a new phenomenon to be discovered or require the contradiction of a well established principle of fluid mechanics. The authors’ explanation of the presence of air is based on the following: An interesting feature of the ingress of air is that it increases in magnitude the further along the channel the Raman signals are collected, Figures 7.13. This suggests that the ingress of air is not originating from a single entrance point in the reformer; rather, it suggests that air continues to enter the channel at many points along the channel via a mechanism independent of channel location. Besides the unproven (or reported) phenomenon or nitrogen/oxygen de-absorption from the anode surface the other feature in the channel which extends its entire length is the ceramic paper gasket which separates the stainless steel reformer body from the sapphire window, as discussed in §3.11.1 This gasket was incorporated into the reformer assembly so as to avoid previously observed damage to the window as a result of chemical reactivity between the two components at the high operating temperatures of the rig. Thus, the use of the gasket was a response to an unexpected operational issue and while the gasket was expected to be the source of some leakage it was assumed that the leakage would be in the direction of the channel to the atmosphere. It was also assumed that due to the relatively low over-pressurisation of the channel the quantity of gas escaping from the channel would: a) be relatively small and b), provided the gases leaving the channel did not do so preferentially, would have little effect on the overall gas composition. What was not initially considered however was that the gasket could enable a mechanism by which air could ingress into the channel against the pressure gradient. 193 7.8 Molecular Flow Molecular flow is the transportation or flux of a fluid due to the statistical likelihood that a molecule will strike or pass through an area of interest. In most engineering situations molecular flow is negated by viscous flow occurring in the opposite direction or considered negligible in the contrast to fluid flux as a result of viscous flow. However, in specific cases molecular flow is no longer negligible and in some cases dominant. The occasions when molecular flow becomes significant is specified by the Knudsen number, Equation 7. 3. l Kn λ= (7.3) Where; λ = mean free path of the molecules (m) l = linear dimension of the fluid conduit (m) If the Knudsen number is greater than unity, i.e. the mean free path is equivalent or greater than the dimension of the conduit under consideration, the flow can be considered purely molecular. If however the Knudsen number is substantially smaller than unity the fluid flow is considered viscous. In region between these boundaries, defined as 10-2 < Kn < 0.5 by Hucknall and Morris [Hucknall and Morris, 2003] the flow is considered intermediate such that both the viscous and the molecular fluid flow mechanisms must be considered. To determine if molecular flow should be a consideration in the author’s analysis of the air ingress problem it was necessary to calculate the approximate linear dimension of conduit which could entertain an intermediate flow regime. Using Equation 7.3 and applying the limiting value supplied by Hucknall and Morris it can clearly be seen 194 that the linear dimension to trigger an intermediate flow regime would be equal to or less than 100 times the mean free path of the molecule - where the mean free path of the molecule is a function of molecular diameter, pressure, and temperature, Equation 7.4 [Roth, 1990]. Pd kT πλ 22= (7.4) Where: λ = mean free path of the molecules (m) k = Boltzmann’s constant, 1.380650 x10-23 (J . K-1) T = Temperature (K) d = molecular diameter (m) P = Pressure, 101.325 (kPa) Taking the diameter of the molecule of interest to range between 2.76 x10-10 m for hydrogen and 3.14 x10-10 m for nitrogen [Roth, 1990] the mean free path can be approximated to between 400 to 300 x10-9 m. Significantly, the elevated temperature at which the reformer operates increases the mean free path by a factor of approximately four compared to standard room temperature. With these values the linear dimension of the conduit must be in the region of 40 x10-5 m for molecular flow to be a consideration. A complication to this analysis is that the above dimensions relate either to an orifice or tube of short relative length. In reality the flow path encountered by gases passing through the ceramic gasket material will be tortuous and far from uniform in length or dimension which makes the Knudsen number analysis less applicable. This was confirmed by observation of the gasket material under a microscope in which the 195 spacing between the fibres was seen to be of the order of five to ten fibre diameters. The gasket material itself is constructed from refractory ceramic fibres (RCF) which range in diameter from 0.5 x10-6 m to 10 x10-6 m [Refractory Ceramic Fibers Coalition, 2010]. It was also noted that the microscope could only observe the surface of the material and so the spacing between fibres was likely to be less in the middle of the material. Consideration therefore needs to be given to whether the porous, tortuous nature of the ceramic gasket is more or less likely to lead to a viscous flow regime. It is the author’s assertion that the predicted tortuous nature of the randomly oriented fibrous structure of the gasket material will limit a viscous flow regime and equate the molecular flow to that of long slender tube with numerous elbows. The existence of elbows can be seen to reduce the molecular conductance of the tube [Roth, 1990], where each molecule/surface interaction has the possibility of either ejecting the molecular back in the general direction from where it had just travelled from, or alternatively continuing on in the approximate direction the molecule was previously travelling. So while the flow regime is almost still certainly intermediate, the molecular effects continue to be significant. A final consideration must be that of diffusion; specifically that defined by Fick, Equation 7.5. x DJ ∂ ∂−= φ (mol m-2 s) (7.5) Where: J = diffusion flux (mol . m-2 . s-1) D = diffusion coefficient (m2 . s-1) φ = molar concentration (mol . m-3) x = position (m) 196 Equation 8.5 serves to explain the driving force behind the oxygen and nitrogen’s incentive to enter the low air concentration environment of the channel. What is more the low hydrogen environment outside the reformer channel also serves as equal incentive for hydrogen to diffuse out from the reformer channel, in addition to any pressure differential driven viscous effects. The above summary of the flow regimes potentially in existence in the ceramic gasket is echoed by a paper investigating the “Backstreaming of Impurity Gases Through a Leak in a Pressurized Vessel” published by a group interested in the effect of leaks in cryogenic systems [Dauvergne and Vandoni, 1998]. In this paper it was determined that leaks of linear dimension smaller than ~ 1 x10-5 m were insensitive to over-pressurization with the resulting fluid flux primarily a result of molecular flow. Similarly, the amount of impurity gas entering the leaking vessel was found to be approximately equal to that of the cryogenic gas leaving for leaks of linear dimension less than 1 x10-5 m. And while the magnitude of the leak was extremely small (and dependant on the length of the tube) specific reference is made to the cumulative effect of a multiple leaks, such as that found in a porous medium. Importantly however, the referenced paper provided a mechanism for the ingress of air and suggested that gases may pass in opposite directions until the leak is of sufficient size for viscous effects to dominate. With those two findings and the large cumulative area of the gasket material around the perimeter of the reformer windows the magnitude of the air leak could be approximated. A brief analysis of the magnitude of the leak is contained in Appendix F. 197 7.9 Relative Normalized Differential Raman Scattering Cross Section for Methane With the relative normalized differential Raman scattering cross sections of hydrogen and water determined, as described above, the final barrier to producing semi- quantitative results from the reformer rig in operation was the determination of a suitable value for the normalized differential Raman scattering cross section of methane. As has been previously discussed, it is not possible to directly calibrate the normalized differential Raman scattering cross section of methane in the rig. Accordingly, it was decided to simply assign a value to the methane cross section that satisfied the ‘inlet’ conditions as stipulated by the experimental test points, §D.2 (the inlet conditions, in terms of mole fractions of the gaseous mixture, were derived from the ratios of gases metered to the anode surface, Table 7.2, while the measured inlet mole fractions were derived from the Raman measurements made at “0 mm”). However, as has been noted previously, the use of normalized relative differential Raman cross sections (as opposed to absolute vibrational Raman cross sections) restricts the quantification of results to the presentation of species mole fractions, as opposed to the absolute number density of a species. This approach then requires that all products of the methane steam reformation reaction, which includes oxides of carbon (COx) (Equations 1.3 and 1.4) are considered. It is therefore necessary to have some measure of COx levels in the rig before a value of the Raman cross section for methane can be assigned. 198 Table 7.2 Metered channel inlet mole fractions. The author’s experimental set-up is unable to detect COx directly due to the low molar concentrations present in relation to the high quantities of humidified nitrogen used to transport water to the reaction zone and the fact that the hydrogen phthalate edge filter only allows partial transmission at the Raman shift of CO2 (321.7 nm). Thus, calculation of species mole fractions during reformation requires the quantities of carbon monoxide and dioxide present in the rig to be estimated from other available data. In theory, this estimation of COx could be made in two distinct methods: COx can be estimated in relation to the amount of hydrogen evolved from the reformation reaction. However, this technique was discounted for two reasons: the first being that COx is produced from two reactions which do not yield equal amounts of hydrogen, the ratio of COx to H2 evolved is therefore potentially non-linear. Secondly, the amount of hydrogen present is not purely dependant on that evolved by the reformation reaction, as shown by the increase in the mole fraction water with respect to channel position – a result of catalytic hydrogen combustion – observed in the calibration tests, Figures 7.13a and b. Test Point CH4 Mole Fraction N2 Mole Fraction H2O Mole Fraction 1 0.06 0.76 0.18 2 0.09 0.73 0.18 3 0.06 0.82 0.12 4 0.10 0.70 0.20 5 0.09 0.64 0.27 6 0.10 0.60 0.30 7 0.15 0.55 0.30 8 0.15 0.40 0.45 199 Thus, relating COx production to hydrogen production will tend to under predict by the amount of catalytic hydrogen combustion which takes place (this error is in turn amplified by a factor of four which relates the number of moles of hydrogen to carbon dioxide produced in the overall reformation reaction, Equation 6.1). The alternative estimation method is to relate the amount of COx produced to the amount of methane consumed, with the ratio of species in this case being one to one. This method was not without potential sources of errors - such as the reduction in methane concentrations as a result of catalytic combustion. However the concentrations of methane used in the experiments were such that combustion was of little concern in this context. Moreover, the product of methane combustion is carbon dioxide which is again produced at a ratio of one-to-one. Most importantly however, the error present when estimating the COx from methane concentration is not amplified by a factor as is the case when COx levels are estimated from hydrogen production. Using this method COx levels were estimated by taking the relative number density of methane at “0 mm” as the initial amount of methane present in the channel, with the difference in relative number density of methane at all subsequent physical positions being equivalent to the amount of COx produced. The weakness in this method is that it cannot incorporate any losses of methane due to reformation before “0 mm” (see test points 4, 6 and 7 where significant signals from hydrogen were detected at “0 mm”), catalytic oxidation, and any methane leakage from the channel. Summarising the two estimation methods, estimating COx concentrations from the production of hydrogen will under-predict COx levels by the amount of hydrogen ‘lost’ from the channel through the mechanisms of catalytic combustion and gas leakage. In contrast, the methane COx method will over-predict COx levels by the amount of methane lost through gas leaks and will under-predict by the amount of COx produced prior to the “0 mm” measurement. On balance, the methane COx prediction method was 200 selected to be used in the final stage of data processing due, in part to the opposing nature of the associated inaccuracies, and in part due to the amplification of error associated with the hydrogen based method. Nevertheless, the affect of the chosen COx estimation model on the results for non-COx species is relatively small as is shown in Figure 7.14, below. Finally, the author notes that if the methane COx estimation method be repeated in future experiments that an improvement to the technique would be to make a Raman measurement at a position sufficiently far up stream of the reactive surface to enable a true reference of pre-reaction methane concentration to be determined. Figure 7.14 Graph of species mole fractions calculated using the methane and hydrogen COx formation models. Note the difference in COx values between the 201 hydrogen and methane models as well as the relative insensitivities of the other species to this difference. Having fixed a method for accounting for COx a value was assigned to the relative normalized differential Raman scattering cross section for methane so as to broadly satisfy the inlet conditions as described in Table 7.2 (unsurprisingly no single value could be used to satisfy all eight test points - an explanation for this can, in part, be attributed to the inaccuracies of humidifying the gas stream, a topic that will be covered in more detail in §7.13). The values chosen for the relative normalized differential Raman scattering cross section of methane were 7.2 and 7.1 for 600 and 700 °C respectively, which once again are noted to fall within the bounds of values previously reported in the literature, as demonstrated previously in Table 7.1. 7.10 Data Processing and Quantification of Image to Image Variation With values for the relative normalized differential Raman scattering cross section of the four major species determined for the author’s experimental set-up it was possible to process the species signal counts and produce a final set of results. What follows therefore is a brief description of the data processing technique, distinct from image processing technique as previously described in §7.1. Given the non-closed nature of the channel due the ingress of air and presumed gas leak through the window gasket material it was not possible to report species concentrations in absolute number densities. The use of mole fractions was therefore adopted with which to report the relative composition of the gas in the excitation volume. The definition of a mole fraction relates the number of molecules of a substance per unit volume to the total number of molecules in that same unit volume. In this case, the mole fraction was determined from the pixel, or number count, recorded on the ICCD from the 202 collection volume for a given chemical species in relation to the total number count collected for all species from that volume. The number count of each species recorded on the ICCD was therefore normalised in relation to the nitrogen count (by convention) and corrected for the respective relative normalized differential Raman scattering cross section. At this point a normalised number density term for COx was added and all terms summed to give a total normalised number density for that particular experimental test point and physical position within the channel. Finally, the relative normalised number density was calculated, that being the number density of an individual species in relation to the total normalised number density, and equated to the mole fraction of that species in relation to the excitation volume. During image processing the spatial extent of the Raman signal was divided into five rows, otherwise referred to portions so as to distinguish between the pixel rows on the ICCD. This was done in an attempt to observe any spatial variation in gas species concentration above the anode surface. When reporting individual portion data the mole fractions are calculated from relative number densities in that portion alone. When reporting data from several portions combined together, the mole fractions are calculated over that total region. The distinction between individual portion and combined portion reporting is continued when plotting confidence intervals for the calculated mole fractions. It should however be noted that the confidence intervals reported are derived from the variation in the 30 images over which each test point is averaged rather than an attempt to determine the absolute experimental error between the reported mole fraction and the actual mole fraction present in the channel for a given experimental condition and physical position. The inability to determine the precision of the technique is in part due to the unknown amount of air entering the channel. In addition, the current experimental set up is not able to determine precise values of the relative normalized differential 203 Raman scattering cross sections, which makes the reporting of absolute error bars impossible. Therefore, with semi-quantitative results the confidence intervals plotted indicate to what level of accuracy the experimental technique can determine the mole fraction of a given species given the variation observed over the 30 images collected. It should also be noted that the number count of each species in each portion, as discussed in §7.4, is the result of summing over 7 x 3 pixels. By summing over a group of pixels the effects of pixel-to-pixel variation are reduced, such that the effects of individual large or small pixel counts is offset against the variation of the other pixels in the group. As such, the variance of a single portion over 30 images is greater than that which is seen for the combination of 2 or more portions due to the larger pixel summing effect. An additional consideration due to the results being reported in terms of mole fractions is that the variance is no longer independent and it was therefore necessary to calculate the propagation of uncertainty [Goodman, 1960]. In calculating the propagation of uncertainty the number counts of each species were not considered to be correlated and therefore the covariance terms were considered zero. Similarly, while there is obviously an inaccuracy in the calculated values of the relative normalized differential Raman scattering cross sections the variation is considered zero as the values used throughout remain constant and have no effect on the uncertainty calculations. The variance of the COx normalised number density term was also ignored because there was no directly measured variation in the term, and in addition the number densities were such that its effect was negligible. Care was also taken to use the correct uncertainty propagation expressions when calculating the variance as a product of means (normalised number densities) and single values (total normalised number density and mole fractions). The confidence intervals are plotted at certainty level of 95% over 30 observations. No confidence intervals are plotted for the in the x-axis as the 204 position of the light sheet optics was not altered from image to image or between operating conditions. The physical position of the platform on which the light sheet optics were mounted was determined using a millimetre scale and as such the positional accuracy can be approximated to +/- 0.25 mm. 7.11 Spatial Variation in Mole Fraction Perpendicular to Anode Surface A review of the processed experimental results plotted in terms of the various species mole fractions at the respective portions above the anode surface with respect to longitudinal channel position show that there is negligible spatial variation perpendicular to the anode surface, Figure 7.15. Figure 7.15 displays the results of test point four, which exhibited one of the fastest rates of the evolution of hydrogen (thus consumption of methane) of the eight test points. Nitrogen is not plotted for the sake of clarity, it having a relatively constant mole fraction of approximately 0.6 and the mole fractions of portions one and five are not shown due to their irregularity (see §7.5). Portion two is that which is closest to the anode surface and portion four is closest to the window. 205 Figure 7.15 Plot of mole fraction vs. longitudinal channel position at the operating condition of test point four for the three middle portions of the channel. The lack of spatial resolution perpendicular to the anode surface is not to say that the technique has insufficient resolution to detect variations in mole fraction should they be present, merely that in this instance any such variation is not large enough to be detected within the confidence intervals calculated. Several possible reasons exist for undetectable mole fraction variation. One is that while the experimental variables of temperature, humidity factor, methane flow rate, and total flow are seen to have an observable effect on the methane/steam reformation reaction, as will be shown shortly, the overall rate of the reformation reaction is not as rapid as suggested by Achenbach [Achenbach, 1994b]. The lack of spatial variation in mole fraction perpendicular to the anode surface may merely then be due to an insufficiently dynamic reaction occurring with which to generate spatial variation. The presumed slow rate of reaction is likely the 206 combined result of the relatively low temperature of reaction, the dilution effects of the water carrying nitrogen gas, and a larger channel height than might be expected in a fully functional SOFC reformer. Therefore, the lack of observable spatial variation in species mole fraction does not necessarily highlight the techniques inability to detect variation. Another factor in the lack of detectable spatial variation perpendicular to the anode surface is the failure of portions one and five to produce consistent data. As in the case of Raman signal calibration discussed in §7.5, the inconsistency is the result of inefficient spectrometer slit clipping (§5.4) in order to reduce spurious light from the anode and window surfaces. As such, all subsequent results presented in this thesis are the result of averaging the middle three portions to give a global representation of the progress of the reformation reaction with respect to longitudinal channel position. In future work more focus must be placed on perfecting a spectrometer slit clipping technique in the spatial direction. 7.12 Calibration Data The calibration data was pivotal in determining the relative normalized differential Raman scattering cross sections of water and hydrogen (§7.6); but the data also offers an insight into the general characteristics of the experimental apparatus and reformer channel. This is because the nitrogen/hydrogen mixtures represent the least chemically active flow regime passing over the anode surface. These conditions therefore allow observations to be made about the accuracy of the rotameters and the consistency of the experimental technique, while also allowing conclusions to be drawn regarding the effectiveness of the window gasket without having to consider the added effects of a 207 simultaneously occurring reformation reaction. Plots of these calibration points are presented in Figure 7.16 to Figure 7.19. Figure 7.16 Calibration data: 80/20 nitrogen/hydrogen mixture, 600 and 700 °C. 208 Figure 7.17 50/50 Calibration data: nitrogen/hydrogen mixture, 600 and 700 °C. Figure 7.18 Calibration data: 80/20 nitrogen/hydrogen mixture, 600 and 700 °C. 209 Figure 7.19 0/100 Calibration data: 0/100 hydrogen mixture, 600 and 700 °C. Figure 7.16 highlights several points of note, the first being the consistency of the technique as demonstrated by the small confidence intervals of the plotted points. While the trends of the curves for nitrogen and hydrogen are not perfectly flat as would have been ideal - due to both gas metering error and the ingress of air - the variation is minimal. The plot also shows a trend repeated throughout the experimental results of the species with the highest number count, that is the highest count recorded on the ICCD, as also having the largest confidence interval. This observation was confirmed by reviewing the raw experimental data. It should also be noted that in the nitrogen/hydrogen 80/20 case, nitrogen will be the most intense spectral line. However, with all other species having a greater relative normalized differential Raman scattering cross section the larger confidence intervals do not necessary coincide with the largest reported mole fraction. 210 Figure 7.17 more clearly demonstrates the effect of the ingress of air, with the mole fraction of nitrogen being considerably larger than that of hydrogen, despite the gases theoretically being metered in equal portions, and increasing the further into the channel measurements are made. Also shown is the increase in confidence interval with increasing longitudinal position. This is because while the magnitude of image-to-image variation remains relatively constant the absolute Raman signal decreases with channel position due to the decreasing effective solid angle. This, in itself, will have no effect on the variance of the Raman signal but in calculating the propagation of uncertainty the normalised number density calculation includes a term in which the variance is divided by the number count; thus, increased uncertainty with smaller number counts. Figure 7.17 also shows the suggestion of gas metering error at the position of 30 mm. It is however worth considering that the rotameter settings were replicated on fourteen separate occasions and in the most part the mole fractions of the two gases at the two separate temperatures were in good agreement. However, at position 30 mm, temperature 600 °C, a greater proportion of nitrogen appears to be metered in relation to hydrogen. The cause of this inaccuracy could be as simple as an incorrect rotameter setting or a contaminant on the rotameter float or tube, but more importantly the diagnostic technique is able to detect this variation. Figure 7.18 highlights a more serious metering error, with the nitrogen mole fraction at 600 °C being consistently higher than that of 700 °C from 0 to 70 mm. This inconsistency is believed to have been caused by the nitrogen rotameter setting. The criteria for the selection of rotameters for this work was a compromise position between the desired experimental range and accuracy. The manufacturers quote an accuracy figure for the rotameters over the full scale of the tube with the exception of the first and last 10 mm of the tubes’ range. In this instance, the hydrogen/nitrogen 80/20 setting required the nitrogen tube to meter gas at 11 mm and the hydrogen at 123 mm (of 211 140 mm). While neither of these settings is in the unspecified region of the tube it is not unexpected that the accuracy of the gas metering is not as accurate as previously displayed in Figure 7.16 and Figure 7.17. Figure 7.19 showing 100% hydrogen confirms the presence of nitrogen and water in the reformer channel despite only hydrogen being metered to it. It also shows that the amount of water produced (and thus hydrogen consumed) is temperature dependant, with increased temperature resulting in increased water production. At 600 °C for the first three physical positions there is a greater, albeit statistically insignificant, mole fraction of nitrogen present than at 700 °C. No immediate explanation for this exists; with two possible explanations being that the hydrogen rotameter is operating at 144 mm (within 10 mm of the maximum scale reading) or that the window gasket underwent a ‘bedding in’ process during the first three thermal cycles. It is however impossible to confirm or discount this second explanation as the gasket was not studied for any time dependant or thermal cycle performance fluctuation. 7.13 The Effects of Temperature Figure 7.20 to Figure 7.27 show the final processed results for test points 1 to 8 respectively. Each figure shows data for both the 600 and 700 °C case. The data consistently shows that an increase in temperature results in an increased rate of the reformation reaction which is indicated by increased mole fractions of reaction products with respect to physical position in the channel. Of course, this is not an unexpected observation and, in fact, it would have been considered an anomaly if this were not the case. More significantly, the results clearly demonstrate that the author’s rig and experimental methods do allow the difference in reaction rate to be quantified in terms of the mole fractions of the reactants and products at a given physical position within the 212 channel. What is more, despite the low reactant mole fractions - due to the large quantity of nitrogen gas required to ‘carry’ water into the reformer channel - and the relatively low molecular densities, due to the elevated temperature, the technique is able to determine semi-quantitative mole fractions to a high level of confidence. The consistency with which the measurements report the inlet mole fractions at the two different temperatures is pleasing. Table 7.3 compares the (nominal) metered inlet mole fractions of CH4, N2, and H2O with the Raman measurement results at 600 and 700 °C. Methane Test Point Metered MF 600 °C Raman MF 600 °C Percentage diff MF 700 °C Raman MF 700 °C Percentage diff MF 1 0.06 0.059 -1.3 0.057 -5.7 2 0.09 0.087 -2.9 0.082 -9.2 3 0.06 0.062 3.9 0.064 7.4 4 0.1 0.101 1.4 0.099 -1.1 5 0.09 0.080 -11.0 0.086 -4.9 6 0.1 0.088 -11.6 0.089 -10.7 7 0.15 0.137 -8.5 0.136 -9.0 8 0.15 0.063 -57.9 0.067 -55.2 213 Nitrogen Test Point Metered MF 600 °C Raman MF 600 °C Percentage diff MF 700 °C Raman MF 700 °C Percentage diff MF 1 0.76 0.711 -6.4 0.723 -4.8 2 0.73 0.685 -6.2 0.693 -5.1 3 0.82 0.785 -4.2 0.783 -4.6 4 0.70 0.629 -10.1 0.626 -10.6 5 0.64 0.536 -16.3 0.515 -19.5 6 0.60 0.489 -18.6 0.488 -18.7 7 0.55 0.430 -21.9 0.445 -19.1 8 0.40 0.171 -57.2 0.154 -61.6 Water Test Point Metered MF 600 °C Raman MF 600 °C Percentage diff MF 700 °C Raman MF 700 °C Percentage diff MF 1 0.18 0.211 17.1 0.201 11.7 2 0.18 0.208 15.5 0.193 7.1 3 0.12 0.136 13.4 0.131 9.1 4 0.20 0.224 11.8 0.221 10.5 5 0.27 0.369 36.8 0.374 38.4 6 0.30 0.391 30.4 0.375 25.0 7 0.30 0.378 25.9 0.359 19.6 8 0.45 0.759 68.7 0.766 70.2 Table 7.3 Metered CH4, N2, and H2O mole fractions vs. measured mole fractions and associated 5 errors at 600 and 700 °C. 214 Table 7.3 highlights a number of features and trends in the experimental data. The first being the errors of the measured mole fractions which can be seen to form three distinct groups; these groups are test points 1-4, 5-7 and 8. A potential theory for this grouping is discussed in more detail below. The percentage errors for each species are approximately constant for each grouping, with the possible exception of the inlet mole fraction of water at test point seven, although in light of the consistency in the measurements of the mole fraction of other species and test points this could be explained as a statistical outlier. What is most pleasing however is the consistency in errors between the two temperatures, increasing the confidence in the values derived for the respective normalized differential Raman scattering cross sections. This also suggests increased confidence in the measurement technique itself, as the change in metered mole fraction of the respective species is, in most cases, accompanied by the equivalent change in measured mole fraction. The only minor exception to this trend is for the mole fractions of methane, but this is due to the normalized differential Raman scattering cross section of methane being derived so as to satisfy the metered inlet mole fraction of methane. Therefore, the percentage differences in measured methane mole fraction for test points 1-4 are merely the variation around the derived normalized differential Raman scattering cross section rather than a specific trend as observed for other species and groupings. 215 Figure 7.20 Effects of temperature (test point 1, 600 and 700 °C). Figure 7.21 Effects of temperature (test point 2, 600 and 700 °C). 216 Figure 7.22 Effects of temperature (test point 3, 600 and 700 °C). Figure 7.23 Effects of temperature (test point 4, 600 and 700 °C). 217 Figure 7.24 Effects of temperature (test point 5, 600 and 700 °C). Figure 7.25 Effects of temperature (test point 6, 600 and 700 °C). 218 Figure 7.26 Effects of temperature (test point 7, 600 and 700 °C). Figure 7.27 Effects of temperature (test point 8, 600 and 700 °C). 219 Interestingly, the results shown in Table 7.3 indicate that the discrepancy between the metered and experimentally determined inlet mole fractions of water (both at 600 and 700 °C) is related to, and increases with, humidifier set point. As will be recalled, the time taken for the humidifier to reach a set point was the limiting factor in determining number of operating conditions which could be examined in a single day. This limitation was overcome to some extent by varying the amount of nitrogen gas passed through the humidifier so that a number of test points could be performed at a single humidifier setting; albeit at the cost of the ability to randomise the order in which the experimental test points are performed. Three humidifier temperature set points (68, 78 and 95 °C) were used in order to perform a complete set of eight test points in the following groupings; test points one to four, five to seven, and eight. From Table 7.3 it can be seen that he percentage errors between the commanded and measured water mole fractions are grouped similarly with an approximate error of 12, 30, and 69% for the three groups respectively. On further examination, the discrepancy between the commanded (via the humidifier and nitrogen rotameter settings) and the experimentally determined (by Raman scattering intensity measurements) water mole fractions is thought to be caused by inaccuracies in the humidification process itself. As such, this result in itself provides further evidence of the value of the Raman measurements. The explanation for the observed inaccuracies in the metered water mole fraction is as follows: as described in §5.1 humidification is achieved by using a bubble type humidifier, the temperature of which dictates the partial pressure of water vapour and thus the water content of the gas passing through the water bath assuming saturation. The operating characteristics of the author’s humidifier, operating at atmospheric pressure is shown below, Figure 7.28. 220 Figure 7.28 Percentage humidification of the bubble type humidifier vs. water bath temperature. The three set points of 68, 78 and 95 °C used for the experiments presented in this thesis are also shown. From the figure it can be seen that the gradient of the humidification curve at the three set points used in this work increases substantially with set point temperature. Therefore, errors in set point temperature amplify errors in percentage humidification accordingly. In addition, the humidifier temperature controller only operates at a resolution of 1 °C, which when combined with the accuracy of the ‘K’ type thermocouple used by the controller (+/- 2 °C) can begin to account for the discrepancies recorded in the mole fractions of water. However, the discrepancy in test point eight is a factor of three greater than that of test points five to seven, which suggests an additional mechanism being present. In this instance the author suggests that there may be some localised nucleate boiling close to the heating elements in the water bath, which is 221 nominally at a bulk temperature of 95 °C. It is thought to be additional water vapour generated via this phase change process that contributes to the large water mole fractions observed at test point eight. Returning to the effects of temperature, it is interesting to note the plateauing of the hydrogen mole fraction curves that is seen in some of the results – most notably in the plots of test points two to five (Figure 7.21 to Figure 7.24) and test point seven (Figure 7.26) – at 700 °C and that is not present at 600 °C (the possible exception to this trend being test point four at 600 °C, Figure 7.23, which exhibits somewhat of a plateau in the hydrogen curve although the methane curve suggests it is still being consumed at the position of 70 mm). This plateauing of the hydrogen mole fraction curves suggests that the reformation reaction is approaching the thermodynamic limit before the end of the channel and can clearly be seen at the higher reformer temperature. The curves that do not exhibit a plateauing suggest that the gas composition at that furthest extent of the channel is kinetically limited. Note also that the subsequent decrease in the hydrogen mole fraction at the furthest extent of the channel seen at 700 °C at test points two, four and five (accompanied by an increase in the mole fraction of water) support the author’s previous assertion that there is oxygen present not only at the channel entrance but also at the furthest extent of the channel. 7.14 The Effects of Humidification Factor The humidification factor was controlled during the experiments such that the ratio of the number of moles of water to the number of moles of methane was either two- to-one or three-to-one, with the two-to-one ratio representing the theoretical minimum amount of water necessary for maximum hydrogen production from a given amount of methane. At this point it should be noted that the humidification plots shown in this 222 section of this work (Figure 7.29 to Figure 7.36) differ significantly from those which examine the effect of temperature on the reformation reaction in the previous section of this chapter (Figure 7.20 to Figure 7.27) in one important aspect: the mole fractions of water and nitrogen are, in the present case, not directly comparable as it is these two parameters that are varied in order to achieve the desired humidification factor. The mole fractions of methane and hydrogen however can be directly compared between the test point pairs of three and one, two and five, four and six, and seven and eight - although in light of the conclusions drawn in the previous section regarding the accuracy of the humidification process care must be taken not mistake variations between the methane and hydrogen curves that are a result of incorrect gas stream humidification rather than the effect humidification factor itself. Considering the uncertainty in the accuracy of the humidification process, the most reliable experimental pairing from which a trend or variation between the methane and hydrogen mole fraction curves with respect to humidification might be extracted is that of test points three (humidification factor of 2) and test point one (humidification factor of 3) as these are from the same humidifier operating point, Figure 7.29 and Figure 7.30. Likewise the data from the pairing of test points seven (commanded humidification factor of 2) and eight (commanded humidification factor of 3) must be discarded because of the large error in the humidification of test point eight as indicated by the data presented in Table 7.3. 223 Figure 7.29 Effects of humidification factor (test points three, commanded humidification factor of 2, and one, commanded humidification factor of 3 at 600 °C). 224 Figure 7.30 Effects of humidification factor (test points three, commanded humidification factor of 2, and one, commanded humidification factor of 3 at 700 °C). 225 Figure 7.31 Effects of humidification factor (test points seven, commanded humidification factor of 2, and eight, commanded humidification factor of 3 at 600 °C). 226 Figure 7.32 Effects of humidification factor (test points seven, commanded humidification factor of 2, and eight, commanded humidification factor of 3 at 700 °C). Considering then Figure 7.29 and Figure 7.30 which show the paired data from test points three (H.F. = 2) and one (H.F. = 3) at 600 and 700 °C respectively it is not possible to observe any significant effect on the reformation reaction by altering the humidification factor. Given the two humidification factors examined this is not a wholly unexpected result as, theoretically, there is sufficient water available to the reformation reaction at all times, unlike an operational SOFC where the production of water is a function of the SOFC reaction itself. In contradiction to Figure 7.29 and Figure 7.30, Figure 7.34 and Figure 7.36 (below) do suggest that there is some variation in reformer performance with respect to humidification factor. At positions 30 and 50 mm in Figure 7.34 and Figure 7.33 and 227 between 20 and 50 mm in Figure 7.36 there is a statistically significant difference in hydrogen mole fraction, with the lower humidification factor promoting a more rapid production of hydrogen. Although the author recognises that the comparison of test points from differing humidification set points is problematic for the reasons discussed earlier, there is an argument to suggest that this result might indeed be valid. Note that the observed variation in hydrogen mole fraction between test points at different humidification ratios is spatially discrete. This contrasts with the expected behaviour of the reformer in response to an error in humidification. An increased quantity of water present, i.e. an increased mole fraction of water, should result in the dilution and reduction of the hydrogen mole fraction of hydrogen across the entire length of the reformer channel. The results therefore suggest that the rate of the reformation reaction could be controlled in certain instances by the humidification factor, and thus water content, of the gas stream passing through the channel. 228 Figure 7.33 Effects of humidification factor (test points two, commanded humidification factor of 2, and five, commanded humidification factor of 3 at 600 °C). 229 Figure 7.34 Effects of humidification factor (test points two, commanded humidification factor of 2, and five, commanded humidification factor of 3 at 700 °C). 230 Figure 7.35 Effects of humidification factor (test points four, commanded humidification factor of 2, and six, commanded humidification factor of 3 at 600 °C). 231 Figure 7.36 Effects of humidification factor (test points four, commanded humidification factor of 2, and six, commanded humidification factor of 3 at 700 °C). The humidification plots also demonstrate further characteristics of the experimental rig. A common trend between Figure 7.29, Figure 7.30 and Figure 7.33 to Figure 7.36, above, is that the mole fraction of water never drops below a value of 0.07. At first glance, the inability for the mole fraction of water to reach or approach a value of zero even for humidification factors of two could be considered indicative of the reaction not reaching completion before the last observation point at 87 mm. However the plateauing of the hydrogen mole fraction observed at 700 °C as described in the previous section, would suggest that this not be the case. Further consideration of the results suggest that the persistence of water vapour through to the last observation point 232 is likely the result of both inaccurate humidification (particularly for test points five and above as indicated by the results shown in Table 7.3), and the catalytic combustion of methane and hydrogen, both of which result in the production of water vapour. Note that the combustion of methane will not only increase the water content in the channel, but will also increase the humidification factor by reducing the amount of methane present. Note also that the hypothesis of catalytic hydrogen combustion is also consistent with the unexpected decrease in hydrogen mole fraction previously noted at the downstream end of the reformer channel in several of the 700 °C cases (Figure 7.21, Figure 7.23 and Figure 7.24) A final observation that can be made from the humidification factor plots relates to the resolution of the measurements of the mole fraction of methane. As already mentioned, the mole fractions of all the species, with the exception of nitrogen, are greatly reduced due the requirement of transporting sufficient water into the channel via the nitrogen carrier gas. Changes in molar number density are therefore lost in the coarse resolution afforded by reporting in mole fractions. Nowhere is this more apparent than in Figure 7.34 and Figure 7.36 where the variation in hydrogen mole fraction is not mirrored by that of the methane mole fraction. The effect of humidification factor was only observable because the ratio of hydrogen production to methane consumption is four to one. However, if the gas stream were not diluted by such a large quantity of nitrogen the variation in hydrogen production and methane consumption would be more easily observed, as well as more accurately reproducing the gas stream of a functioning SOFC. Therefore in future work, not only should the humidification technique be improved in terms of accuracy but alternative humidification methods should be sought so that a more realistic gas stream with lower nitrogen mole fractions can be produced. 233 7.15 The Effects of Total Volume Flow Rate The experimental matrix was constructed such that the volume flow rate of methane was to be kept constant while the total volume flow rate was varied. This was done in an effort to determine if the same quantity of hydrogen would be produced if, in effect, the mole fraction of methane was varied. The analysis of this type of experimental variable is distinct from that for temperature and humidification factor in that both the mole fractions of the metered inlet gases and the predicted outlet mole fractions (if it is assumed the reaction goes to completion) will be different for the two operating conditions. Therefore, while the analysis of the temperature and humidification factor graphs is, at the most basic level, simply a matter of determining if the observed mole fractions are statistically different at the respective physical positions, this approach is no longer suitable for reviewing the total volume flow rate data. In this instance an appropriate metric – such as the percentage of methane consumed or hydrogen produced – must be established with which to gauge the progress of the reformation reaction. However, the consistent application of the chosen metric would then demand that the channel is either perfectly sealed or that any gas leaks comprised of a homogeneous mixture of the gases in the channel at that physical location, such that the remaining gas in the channel may be taken as truly indicative of the reformation reactions’ progress to that point. In the work presented herein these conditions are known not to hold. Not only is there a suspected gas leak from the channel, but the Raman results have also clearly shown that there is an uncontrolled ingress of air into the channel. The Raman results indicate that the presence of oxygen in the channel is causing catalytic oxidation of both reactant and product. Furthermore, the results show that the metered humidification factors for certain test point pairs are subject to substantial error. 234 In light of these observed, but un-quantified, uncertainties in the experimental boundary conditions detailed analysis of the hydrogen production in the rig and the assignment of an associated rate of reformation will not be performed in this work as it is clear any conclusions of this nature would be insufficiently robust as to merit publication. Instead observations of a more general nature will be made. Figure 7.37 and Figure 7.38 show the two most favourable experimental pairs of test points four (total volume flow rate 0.6 SLPM) and three (total volume flow rate 1.0 SLPM) at 600 and 700 °C respectively. Again this test point pair is selected for study as both operating points use the same humidification set point which therefore limits experimental error through humidification. As shown in Table 7.3, the difference between commanded and measured inlet conditions for the test points is typically in the region of 1-5% for methane, 5-10% for nitrogen, and 9-13% for water. 235 Figure 7.37 Effects of total volume flow rate (test points four, total volume flow rate = 0.6 SLPM, and three, total volume flow rate = 1.0 SLPM, at 600 °C). 236 Figure 7.38 Effects of total volume flow rate (test points four, total volume flow rate = 0.6 SLPM, and three, total volume flow rate = 1.0 SLPM, at 700 °C). Calculation of the theoretical mole fractions of the exhaust stream, assuming the reformation and gas shift reactions proceed to completion, suggests that the mole fractions of hydrogen will be 0.33 and 0.21 for test points four (0.6 SLPM) and three (1.0 SLPM) respectively. Neither test point at either temperature approaches these values, although as previously noted both methane and hydrogen are potentially consumed through catalytic oxidation. Therefore, additional indicators must be sought with which to gauge the progress of the reaction. Looking at the mole fractions of methane in Figure 7.37, the 600 °C case, it would appear neither test point completely consumes all the methane present, an observation mirrored by the continued upward trend in hydrogen mole fraction. In Figure 7.38 however not only does the mole fraction 237 of methane approach zero at 700 °C, but the hydrogen curves also exhibit the plateauing behaviour described previously. The mole fractions of hydrogen are 0.20 and 0.15 respectively for test points four (0.6 SLPM) and three (1.0 SLPM), although at position ‘70 mm’ hydrogen achieved a maximum of 0.22 for the 0.6 SLPM case. Clearly, these values are still considerably below those that are theoretically possible but significantly the ratio of hydrogen mole fractions between the two volume flow rate cases is approximately 2/3, which is equivalent to that calculated theoretically. The result therefore suggests that at 700 °C, decreasing the inlet mole fraction of methane from 0.10 to 0.06 does not alter the overall quantity of hydrogen produced. The pattern of incomplete reformation at 600 °C, complete reformation at 700 °C (as indicated by plateauing of the hydrogen curve and the mole fraction of methane approaching zero), and the approximate ratio of hydrogen exit mole fractions equating those calculated theoretically is repeated for test point pairs six (0.6 SLPM) and one (1.0 SLPM), and seven (0.6 SLPM) and two (1.0 SLPM), Figure 7.39 to Figure 7.42. It should however be reiterated that these plots contained the added variation of different humidification settings with the associated error previously cited. 238 Figure 7.39 Effects of total volume flow rate (test points six, total volume flow rate = 0.6 SLPM, and one, total volume flow rate = 1.0 SLPM, at 600 °C). 239 Figure 7.40 Effects of total volume flow rate (test points six, total volume flow rate = 0.6 SLPM, and one, total volume flow rate = 1.0 SLPM, at 700 °C). 240 Figure 7.41 Effects of total volume flow rate (test points seven, total volume flow rate = 0.6 SLPM, and two, total volume flow rate = 1.0 SLPM, at 600 °C). 241 Figure 7.42 Effects of total volume flow rate (test points seven, total volume flow rate = 0.6 SLPM, and two, total volume flow rate = 1.0 SLPM, at 700 °C). From the above plots it can be seen that the methane mole fraction curves appear to converge before the end of the channel. This would therefore suggest that for higher mole fractions of methane the reformation reaction appears to proceed more rapidly in a manner similar to that described by Achenbach [Achenback, 1994b]. However, caution is exercised when offering this theory as the measurement technique cannot distinguish differences between such small mole fractions with any statistical certainty. In addition, the variable is not merely the mole fraction of methane, but also the time taken for the gases to pass through the channel. For while the volume flow rate of methane is the same between test point pairs, the quantity of accompanying gases changes dramatically. Therefore in the above cases, higher mole fractions of methane 242 (the first listed test point of any given test point pair) also coincide with a longer period of residency in the reformer channel. It is therefore important to make a distinction between the combined effects of varying methane mole fraction and changing the residency time of methane within the reformer channel, and the effects of changing the methane mole fraction independent of residency time. It is with this clarification in mind that the ‘0 mm’ hydrogen mole fractions are again referred to. Without exception the greater ‘0 mm’ hydrogen mole fraction occurs with the combination of a higher methane mole fraction and longer residency time. But due to the fact that there are two variables being varied simultaneously it is clearly impossible to determine if one has a more significant effect than the other from this result alone. Accordingly, the next section of this chapter looks at the effect of variation of the methane volume flow rate at a fixed total volume flow rate, which has the effect of altering the methane mole fraction in isolation. Clearly, the evaluation of a specific volume flow rate of methane is not as simple as when considering the effect of temperature or humidification factor. In keeping the volume flow rate of methane constant both the methane mole fraction and residency time in the channel are altered, making evaluation of either variable problematic. Ideally these two parameters would have been evaluated in isolation. However, two experimental pairs do exist in the present test matrix where the inlet mole fractions of methane are approximately equal: these are test point pairs four and two, and six and five. Within these two pairs the volume flow rates of methane are 0.06 and 0.09 SLPM and the total volume flow rates 0.6 and 1.0 SLPM respectively, while the inlet mole fractions of methane are calculated to be 0.10 and 0.09. The humidification factors are: two for the test point pair four and two, and three for the pair six and five, Figure 7.43 to Figure 7.46. 243 Figure 7.43 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points four, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point two, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 600 °C). 244 Figure 7.44 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points four, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point two, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 700 °C). 245 Figure 7.45 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points six, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point five, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 600 °C). 246 Figure 7.46 Effect of methane ‘residency time’ with approximately constant inlet mole fraction (test points six, MFCH4 = 0.10, total volume flow rate = 0.6 SLPM, and test point five, MFCH4 = 0.09, total volume flow rate = 1.0 SLPM at 700 °C). Figure 7.43 to Figure 7.46 demonstrate several consistent trends. The first is that the lower total volume flow rate (thus longer residency time) condition always produces a larger ‘0 mm’ mole fraction of hydrogen, suggesting this feature is a function of residency time of the gas inside the channel rather than the inlet mole fraction of methane. This conclusion is supported by the observations discussed in the following section. It can also be seen that the longer residency time results in a higher spatial rate of hydrogen production at both experimental temperatures and humidification factors. In agreement with this observation, the plots also show the higher flow rate having a higher mole fraction of methane in the mid to latter part of the channel, despite the inlet mole fraction 247 of methane being marginally lower. Again however, the low overall mole fraction of methane makes this observation less pronounced with respect to hydrogen, especially at the locations of the channel inlet and outlet. Note: plots for the test point pair of eight and five are not presented in this section due to the large error associated with the humidification of test point eight. 7.16 The Effect of Methane Volume Flow Rate In §7.15 the experimental matrix was designed to observe the effect in constant methane volume flow rate and variable total volume flow rate. It was subsequently discovered that this configuration resulted in the simultaneous variation in mole fraction and residency time, making scientific observations problematic. However, somewhat fortuitously two experimental test point pairs allowed an approximate analysis of constant methane mole fraction with variable residency time as presented in Figure 7.43 to Figure 7.46 above. The alternative regime is then that of constant residency time (total volume flow rate) and variable methane mole fraction, and it is this scenario that was produced by varying the volume flow rate of methane. Once again, the test point pair that was deemed to be most statistically robust and least affected by experimental errors was that which utilised the same humidification set point; in this case test points three (methane volume flow rate of 0.06 SLPM) and two (methane volume flow rate of 0.09 SLPM), Figure 7.47 (600 °C) and Figure 7.48 (700 °C). 248 Figure 7.47 Effect of methane volume flow rate (test points three, methane volume flow rate = 0.06 SLPM, and two, methane volume flow rate of 0.09 SLPM, at 600 °C). 249 Figure 7.48 Effect of methane volume flow rate (test points three, methane volume flow rate = 0.06 SLPM, and two, methane volume flow rate of 0.09 SLPM, at 700 °C). In Figure 7.47 and Figure 7.48 the inlet mole fractions of methane are 0.06 and 0.09 respectively. Following on from the statement made in §7.15 regarding the magnitude of the ‘0 mm’ hydrogen mole fraction being a function of residency time it can be seen that with identical residency times yet different inlet mole fractions of methane the ‘0 mm’ hydrogen mole fractions are approximately equal. As might be expected, a greater quantity of hydrogen is produced with the larger methane intake mole fraction. It can also be seen, that as in previous observations, the reaction does not appear to reach completion in the 600 °C case and appears to be close to completion at between 30 and 50 mm for both test points at 700 °C. 250 The plots of test point pair one (methane volume flow rate of 0.06 SLPM, humidification factor of 3) and five (methane volume flow rate of 0.09 SLPM, humidification factor of 3) broadly support the observations made above, with the only noticeable difference being the apparent reduced rate of reaction at 700 °C, Figure 7.49 and Figure 7.50. Figure 7.49 Effect of methane volume flow rate (test points one, methane volume flow rate = 0.06 SLPM, and five, methane volume flow rate of 0.09 SLPM, at 600 °C). 251 Figure 7.50 Effect of methane volume flow rate (test points one, methane volume flow rate = 0.06 SLPM, and five, methane volume flow rate of 0.09 SLPM, at 700 °C). As been discussed previously the high level of uncertainty in the humidification levels supplied using different humidifier set point temperatures suggests that it would therefore be unwise to place too much emphasis on the apparent difference in completion point between the test point pairs of three and two, and one and five. It should also be considered that due to the over humidification of test points five to eight the mole fractions (with the exception of water) will tend to be smaller. Accordingly, the comparison of test points one and five will tend to see a smaller distinction between the mole fractions of methane and hydrogen. 252 The plots of test points four (methane volume flow rate of 0.06 SLPM, humidification factor of 2) and seven (methane volume flow rate of 0.09 SLPM, humidification factor of 2), Figure 7.51 and Figure 7.52, resemble those of test points three and two; the difference between the test point pairs in this case being the total volume flow rate with test points three and two representing a 1.0 SLPM total volume flow rate and test points four and seven a 0.6 SLPM total volume flow rate. Accordingly, the methane supplied to test points four and seven have a longer residency time in the channel than the methane supplied to test points three and two. It is wholly consistent with previous observations of a dependency of the hydrogen mole fraction at the “0” mm position with residency time, that a large mole fraction of hydrogen is seen in both the test point four and test point 7 cases, irrespective of the considerably different ‘0 mm’ mole fractions of methane. 253 Figure 7.51 Effect of methane volume flow rate (test points four, methane volume flow rate = 0.06 SLPM, and seven, methane volume flow rate of 0.09 SLPM, at 600 °C). 254 Figure 7.52 Effect of methane volume flow rate (test points four, methane volume flow rate = 0.06 SLPM, and seven, methane volume flow rate of 0.09 SLPM, at 700 °C). Note: plots for the test point pair of eight and six are not presented due to the large error associated with the humidification of test point eight. 7.17 Infrared Radiation (IR) Thermometry Results Infrared radiation thermometry measurements were performed at reformer operating conditions specified by test points 1 – 8. In addition a set of IR temperature measurements were collected at the 0.5 SLPM dry hydrogen condition, although this was purely from the perspective of comparison as - in contrast to Raman spectroscopy - no species specific signal calibration was required. The IR thermometer was operated at 255 an arbitrary emissivity setting of 0.92. IR temperature measurements were collected across the entire length of the anode surface at each operating condition and three measurements were collected at each physical position as a result of three passes of the anode surface by the detector head. The results of the IR temperature measurements are shown below, Figure 7.53 and Figure 7.54. Figure 7.53 IR temperature measurements (test points 1 – 8, 600 °C) 256 Figure 7.54 IR temperature measurements (test points 1 – 8, 700 °C) The most striking aspect of the IR thermometry results is the scale of the temperature variation across the anode (in the order of 30 – 35 °C) with the inlet conditions being significantly lower than the command temperature. While efforts were made to heat the reformer channel as uniformly as possible some variation in anode surface temperature was expected; however the magnitude of the temperature difference between the initial and final physical positions was more substantial than expected. With respect to this temperature difference, the inlet gases enter the heater body at approximately 130 °C and despite passing through the inlet tube situated inside the heater and the plenum chamber it is not inconceivable that the inlet gases themselves will have a cooling effect at the entrance of the channel. More significant however, is the temperature variation between the test points 1 - 8 and the 100% hydrogen reference condition. Here it can clearly be seen that the anode surface is 257 significantly cooler during methane reformation than for the non-reforming reference case (particularly at the 700 °C command temperature that has previously been shown to promote a more complete reformation reaction than the 600 °C case). At this point it should be stressed that the exact magnitude of the temperature variation is not known. For the exact magnitude to be known the IR results would need to be calibrated via an additional temperature measurement of the anode surface acting as a reference. This was not possible as a consequence of the issues associated with the ingress of air, which restricted the addition of thermocouple probes without further distorting the operational accuracy of the channel. In addition, the focus of the research shifted during the projects duration to Raman spectroscopy as a result of this requiring greater effort to perfect, being of greater academic interest and yielding results of greater research value. The IR thermometry does however indicate that there is a significant temperature variation between measurement locations, even if the exact temperature at which this occurs is unknown. These temperature variations have importrant implications for future Raman measurements, particularly with respect to improving experimental accuracy. As discussed in §7.6, the normalised relative differential Raman cross section of hydrogen varies substantially with temperature (2.65 and 2.30 for 600 and 700 °C respectively), making the variation in cross section significant over the 30 – 35 °C variation seen to occur along the anode surface. Incorporation of this temperature variation will significantly improve experimental accuracy and, if the ingress of air can be overcome, make fully quantitative results possible. The second meaningful observation which can be seen is the lack of any substantial temperature differences between test conditions 1 - 8. In light of the results obtained through Raman spectroscopy this is not unexpected given the overall similarities between operating conditions at any given temperature. As discussed the inability to operate the reformer at mole fractions approaching those found in fully 258 operational SOFC’s prohibited direct comparison of the results with the findings presented in the literature regarding rapid reformation at the channel entrance. The majority of operating conditions exhibited steady reformation occurring along the full length the reformer channel and this is reflected in the relatively tight grouping in IR measurements across the eight test points at each physical position. In this instance therefore, it cannot be definitively concluded as to whether IR thermometry is capable of detecting the variations in temperature that, given more rapid reformation reactions, are reported to be present. What is of significance however, is the variation between the 100% hydrogen case and the eight reformation operating conditions, Figure 7.55 and Figure 7.56. Figure 7.55 Differences in IR temperature measurements between test points 1 – 8 and 100% H2 case (at heater command temperature of 600 °C). 259 Figure 7.56 Differences in IR temperature measurements between test points 1 – 8 and 100% H2 case (at heater command temperature of 700 °C). Differences between the eight test points and the 100% hydrogen case are expected given the endothermic nature of the reformation reaction. Note that these differences can confidently be attributed to the reformation reaction and not the change in total volume flow rate as the variation in volume flow rate between the 100% hydrogen condition and test points 4, 6, 7, and 8 is only 0.1 SLPM, while the remaining test points are at 1.0 SLPM and do not exhibit the equivalent variation. It could equally be expected that the greater volume flow rates might be expected to increase the cooling effect whereas the opposite would appear to be true at both heater temperatures with test points 1 – 3 resulting in a marginally hotter surface temperatures. Drawing further conclusions from the temperature difference curves is however problematic given the secondary effects of inaccurate humidification factor as described previously. 260 The final observation is that there is a greater temperature difference between the eight test points and the 100% hydrogen case at 700 °C than there is at 600 °C. This again is not an unexpected observation, particularly in conjunction with the Raman spectroscopy results which demonstrated the effect of increased heater temperature being a more rapid reformation reaction. Given this greater rate of reaction, a greater amount of methane – and thus energy – is consumed from the surroundings, resulting in a larger observed temperature drop with respect to the non-reforming case. From the results presented it is clear that IR thermometry can detect temperature variations which are in line with those expected from a surface hosting reformation reactions. However, in light of the imperfections of the reformer channel, the inability to operate at higher mole fractions of methane, and the energies expended to demonstrate the feasibility of Raman spectroscopy, no further effort to improve the accuracy of IR thermometry has taken place. It should however be the focus of future work to investigate to what degree of accuracy temperature measurements can be made using IR thermometry , given reformation reactions more representative of those found in fully functioning SOFC’s. 7.18 Experimental Results Summary The overall quality of the results presented above has been extremely satisfactory. The consistency in the results indicates that the experimental apparatus, experimental procedure and data processing provide a setting in which to evaluate the reformation process and the effect of varying operating conditions on the chemical reaction. A methodology with which to determine the normalised relative differential Raman cross sections of the major chemical species involved in the reformation reaction has been demonstrated and shown to produce results that are consistent and in good 261 agreement with those expected given the experimental variable examined. Similarly, the effect of diminishing apparent solid angle, a result of collecting light from within the optical confines of a narrow channel has been calibrated out of the results produced. The technique has also demonstrated its potential by conclusively detecting the ingress of air into the reformer channel. Furthermore, the ability of the technique to report mole fractions along the spatial extent of the channel led to the mechanism of molecular flow to being proposed as the cause of the ingress of air. It is unlikely that this mechanism would have been considered credible without the evidence collected using Raman spectroscopy, particularly in contrast to the information that would have been available using exhaust gas sampling. However, it has also been shown that as a result of the ingress of air (and expected gas leaks resulting from the window gasket) the subsequent experimental results must be reported in terms of mole fraction rather than absolute molecular number density, making results at this time semi-quantitative. The quality of results was further improved through the consideration of reaction products not currently observable using the diagnostic technique. As such, the effects of COx have been incorporated into the reporting of species mole fractions and done so using the demonstrably more robust method of linking COx production to methane consumption rather than hydrogen production. This thorough data processing, combined with the repeatability of the experimental technique has resulted in extremely consistent results as indicated by the small confidence intervals in which the observed measurements lie. This further justifies the use of Raman spectroscopy in environments similar to that of a SOFC reformer and suggests it could be a useful tool with which to evaluate the performance of a SOFC and its components. Despite the consistency of the technique, when considering spatial variation in chemical species perpendicular to the anode surface none could be detected. However, this was surmised to be a result of a combination of the chemical reaction not being as 262 rapid as had been expected and the inability of the technique to make mole fraction measurements close to the solid surfaces inside the reformer channel. However, areas of the technique within which to make improvements have been identified, which in combination with improvements to the humidification process suggest there is still scope for the technique to yield one dimensional Raman measurements making it of further use to SOFC researchers. The cause for the relatively slow progress of the reformation reaction was attributed to the low mole fractions of methane as a result of water vapour having to be transported to the reformer channel via a nitrogen carrier gas. It was therefore concluded, that if purely reformation reactions are to be studied in the future an alternative to a bubble type humidifier should be found. This finding was emphasised by the inaccuracies in the performance of the humidifier which were observed throughout experimentation. Finally, the effects of the variables of temperature, humidification factor, gas residency, and methane mole fraction were observed in relation to the reformation reaction. Using the technique it could clearly be seen that the higher the reaction temperature, the faster the reaction occurred. In the case of humidification factor the results were less clear, with certain operating conditions showing some dependence on humidification factor. Increasing the residency time of methane in the reformer was also seen to increase the amount of hydrogen produced whereas reducing the mole fraction of methane led to a reduction in hydrogen production. It can therefore be concluded that Raman spectroscopy has been successfully applied to the high temperature environment of a SOFC reformer. In addition the results produced using the technique of Raman spectroscopy have demonstrated repeatability, a high level of consistency, and produced sensible results as would be expected of the experimental variables studied. 263 8 Summary, Claims of Originality and Future Work 8.1 Summary This thesis details the development of an experimental apparatus and methods to allow the application of gaseous Raman spectroscopy in a challenging and original application - specifically, a small-scale, high-temperature methane/steam reformer developed to be representative of the technologies used in solid oxide fuel cell (SOFC) applications. The contents of the thesis are summarized as follows: Context for the research is provided Chapter 1 where hydrocarbon steam reformation was described, different in-situ hydrogen production techniques for SOFC’s presented, and the specific challenges related to internally-reforming medium- temperature SOFC operation were detailed. Chapter 1 also included a review of the current state of SOFC development and the case for the development of non-intrusive measurement techniques for gas species and temperature. Chapter 2 presented fundamental material on Vibrational Raman Scattering and radiation thermometry. The author’s original contribution to the literature begins with Chapter 3 which details the design and development of a novel optically-accessible methane/steam reformer and the optically accessible heater assembly. The specific requirements of Raman spectroscopy and radiation thermometry were examined and were incorporated into a practical experimental arrangement that was able to sustain a suitable reformation reaction. The effects of thermal expansion were anticipated and incorporated into the apparatus design to allow satisfactory optical access and ensure experimental repeatability. Chapter 4 describes the development of a novel Raman measurement system. Gaseous Raman spectroscopy was performed in the optical reformer/heater rig using the spectrally uncommon 308 nm laser in the equally uncommon broadband 264 configuration (with both the unusual wavelength and line width of the laser source being of interest the scientific community of researchers using optical diagnostic techniques). This study was also performed using an unconventional (and in many ways non-optimal) orientation of the laser excitation source. The results presented in this chapter clearly demonstrate that the effects of light reflected and scattered from solid surfaces can be sufficiently suppressed to allow useful Raman signals to be collected; thus, suggesting that Raman spectroscopy is a viable technique for future fuel cell studies despite the likely limitations on optical access. Chapter 4 concludes with an examination of the unanticipated challenges encountered during the initial operation of the apparatus and a review of the subsequent design solutions. The work presented in Chapter 4 also demonstrated the successful use, and the benefits of, a liquid Potassium Hydrogen Phthalate (KHP) low pass filter over the coloured glass filters that have been used to date in conjunction with 308 nm excitation sources. The results presented in this chapter clearly showed that while, in an ideal situation, a narrowband and polarized excitation source would be preferential to the broadband laser used here, equivalent results could be obtained with substantially reduced equipment costs through the correct application of a sufficiently robust polarizer. The results also showed that an alternate background signal suppression technique was required for the particular application that is the focus of this thesis. The full experimental set-up for reformer testing and the rig and reformer operating procedures are detailed in Chapter 5. The arrangements for gas supply, metering and humidification are shown, as are the optical arrangements for laser sheet formation and light delivery. The optical set up and characteristics of the light collection and measuring systems (optics, spectrometer, and ICCD camera systems) are also shown. 265 Chapter 6 covered the design of the experimental test matrix, the limitations of the experimental apparatus on the range of experimental variables, and the optimisation of data collection. The calibration procedure with which the normalised relative differential Raman cross sections of the species of interest were determined as well as incorporating the effect of reduced effective solid angle with physical channel position was also described. The number of images collected and averaged over for each physical position was also justified. Chapter 7 describes the image processing techniques that are applied to the raw results, specifically the use of a morphological structuring element to eliminate the contribution of background signal. The calibration experiments, procedures, and methodologies that are used to define the normalized differential Raman scattering cross sections of the major species of interest in this study are detailed and discussed. During these experiments an unexpected leakage of air into the reformer was observed. A hypothesis is presented to explain the observed air ingress. Finally, results are presented that describe the response of the optically-accessed reformer to variations in; operating temperature, humidification factor, total volume flow rate, methane volume flow rate, and the methane residency time within the reformer channel. The performance of the reformer rig and the results of the reformer study are discussed and can be broadly summarised as follows: • The effect of increasing reformer temperature is observed to increase the rate of the reformation reaction. • The effect of varying the humidification factor (while maintaining sufficient steam for the gas shift reaction) had no significant observable effect on the rate of production of the reformation products • The effect of increasing total volume flow rate, which could otherwise be considered to be a measure of reactant residency time in the channel, 266 was found to reduce the progress of the reformation reaction with respect to physical position in the reformer channel. • The effect of increasing methane volume flow rate, which could otherwise be considered to be increasing the mole fraction of methane in the channel, was found to increase the rate of the reformation reaction with respect to physical position within the channel. 8.2 Claims for Originality This research project was conceived without precedent and as such did not continue on from or replicate the work of a pervious researcher in the context of an optically accessible methane/steam reformer. Moreover, as is evidenced by the relatively limited number of references contained in the author’s introductory chapters, only a small amount of literature was found that was directly applicable to this particular application. The project was therefore conceived, executed and refined with minimal external influence. The test and measurement systems that have been developed by the author during this study are novel both in design and application. The optically accessed heater and fuel cell reformer channel developed during this work are of unique design and all of the design considerations concerning optical access and excitation orientation leading to the final design specification were the product of independent thought as described by the author in this thesis. The application of a laser Raman measurement system to interrogate the gas flow of methane/steam reformation reaction in a fuel cell like environment is, to the author’s knowledge, the first study of its kind. There is novelty and originality in the set- up and configuration of the measurement system as shown by the publication of the bulk 267 of the material presented in Chapter 4 in Review of Scientific Instruments [Saunders and Davy, 2010]. The broadband 308 nm excitation source, while not unique in terms of its use in gaseous Raman spectroscopy has not been used extensively for the reasons presented therein. The author’s use of a KHP liquid filter for Raman spectroscopy at 308 nm is new to the literature (the filter was originally proposed by Kleimeyer et al. [Kleimeyer et al. 1996] for use at 309 nm) as is the published demonstration of the filter’s superiority to competing glass filters for this application. The author believes that the test and measurement capabilities of the developed reformer rig and Raman system are presently unique; and thus, that they offer fuel cell researchers an unparalleled opportunity to spatially interrogate high-temperature reacting gaseous flows using non-contact optical techniques. Accordingly, the data sets that have been produced in the rig with respect to the effects of temperature, humidification, methane residence time, and methane mole fraction on the reformation process are themselves an original contribution to the literature. 8.3 Recommendations for Future Work The scope of future work can be divided into two broad categories; those works which can be undertaken using the apparatus in its current configuration, and those works that could be performed following a series of modifications. 8.3.1 Apparatus in Current Configuration With the rig in it its present configuration the experimental parameters that can be studied are effectively limited to those that can be evaluated through direct comparison – notwithstanding these limitations, the reformer rig and the optical methods developed by the author in their present state do allow the study of several topics of 268 major importance to fuel cell researchers. For example, with benchmark performance curves for the optical reformer established it would be possible to study the effects of varying the catalytic surface in the channel. This is an area of research that is known to be of significant current interest. A more detailed study of the effect of temperature on reaction rate could also be performed. Studies into preferential catalytic surfaces, such as those under development for use in single chamber SOFC’s, could also be undertaken without modification to the rig. Temporal effects on the catalytic surface may also be observed, with an aim to gain insight into general temporal catalytic degradation or the effect of poisoning by species such as sulphur in the gas stream. The temporal effect of operating the reformer with insufficient water would also be of interest to researchers seeking insight into solid carbon formation and SOFC/reformer performance degradation. The type of fuel undergoing reformation could also be varied, ranging from other alkanes, blends of alkanes and other hydrogen carriers such as ammonia. 8.3.2 Future Work with Modified Apparatus Overcoming the ingress of air into the channel would however greatly improve the quality and scope of experimentation possible using Raman spectroscopy in relation to SOFC’s and methane/steam reformers. As such, it is recommended that arrangements for providing an impervious seal between the reformer window and channel should be investigated as a matter or priority. It is noted that the value of experimental data produced from the rig would also be increased if the methane mole fraction of the inlet stream could be increased to that which is comparable to an operational SOFC. Accordingly, an alternative method of humidification for the gas stream must be sought which removes the need to introduce nitrogen into the gas stream. Note that a substantial increase in methane mole fraction would have the 269 additional benefit of increasing the mole fractions of carbon monoxide and carbon dioxide produced in the rig - potentially making the measurements of these species possible. Note also that an increase in the mole fraction of methane would also be expected to increase the overall rate of reaction, particularly at the reformer entrance. The subsequent increase in rate of reaction would be expected to result in greater variations in anode surface temperature, which could make measurements made using radiation thermometry of greater interest than those which were observed. The detection of significant temperature variation could enhance the accuracy of a subsequent observations made using Raman spectroscopy, particularly for species such as hydrogen with highly temperature sensitive relative normalised differential Raman cross sections. Another area in which the apparatus might be improved is in that of channel height and the author acknowledges that the current channel height is somewhat large. However, a conservative value of this parameter was selected for this study so as to ease the evaluation of the Raman technique for this application. As such, it is recommended that the channel height be reduced in future studies to that which is more comparable to operational SOFC’s. Note however, in reducing channel height the light collection optics must be updated accordingly. Conventional wisdom states that the optics solid angle should be as large as possible (dictating a smaller ƒ# at constant focal length) in order to collect as much scattered light as possible. However large collection optic solid angles are quickly obscured by the physical restrictions of the channel when interrogating narrow channels. In theory, a collection optic with larger ƒ# should be used for smaller channel dimensions in order in order to maximise collected signal intensity at the furthest physical extent of the channel. However, by using a larger ƒ# collection optic the size of the image projected onto the spectrometer slit is reduced in size (the focus optic ƒ# being determined by the ƒ# of the spectrometer itself). It is therefore 270 recommended that a study be performed to determine the minimum channel height that could be investigated using the current spectrometer and to determine what greater capability would be achieved by using a spectrometer with longer focal length. Another spectrometer improvement would be to derive an improved method of restricting the height of the spectrometer slit so that Raman spectroscopy measurements may be performed reliably in close proximity to the anode surface. The large degree of optical access available in the author’s reformer rig should enable additional measurement techniques to be applied to the reformation reaction in future works. For example, the author suggests that observations of species adsorption could be obtained using surface Raman spectroscopy, which would complement the gaseous Raman studies and allow further understanding of the reformation reaction in the context of a SOFC reformer. Finally, it should be recognized that the ultimate aim of all future work should be the integration of gaseous Raman spectroscopy into a fully operational, directly internally reforming SOFC. To do so would enable further insights to be made into the interactions between the simultaneous reformation and fuel cell reactions and enable this technology to be improved and developed. 271 References Achenbach, E. [1994a]: Three-dimensional and time-dependant simulation of a planar solid oxide fuel cell stack, Journal of Power Sources, vol. 49 pp. 333-348 Achenbach, E. [1994b]: Methane/team reforming kinetics for solid oxide fuel cells, Journal of Power Sources, vol. 52 pp. 283-288 Adris, A.M., Lim, C.J., Grace, J.R : The Fluidized-Bed Membrane Reactor for Steam Methane Reforming: Model Verification and Parametric Study, Journal of Chemical Engineering Science, vol. 52, pp 1609-1622 Aguiar, P., Chadwick, D., Kershenbaum, L. : Dynamic Effects in Autothermal Systems Applications to a Coated-wall Internally Reformed Solid Oxide Fuel Cell, Trans IChemE, vol. 82(A2) pp. 259-266 Alger, T., Wooldridge, S. : Measurement and Analysis of the Residual Gas Fraction in an SI Engine Variable Cam Timing, SAE Technical Paper 2004-01-1356 Appel, C., Mantzaras, J., Schaeren, R., Bombach, R., Kaeppeli, B., Inauen, A. : An Experimental and Numerical Investigation of Turbulent Catalytically Stabilized Channel Flow Combustion of Hydrogen Air Mixtures over Platinum, Proceedings of the Combustion Institute, vol. 29 pp 1021-1029 Baranska, H., Labudzinska, A., Terpinski, J. : Laser Raman Spectroscopy Analytical Applications, Ellis Horwood Ltd., Chichester Bendtsen, J : The Rotational and Rotational-Vibrational Raman Spectra of 14N2, 14N15N and 15N2, Journal of Raman Spectroscopy, vol. 2 pp 133-145 Birkebak, R.C., Eckert, E.R.G. : Effects of Roughness of Metal Surfaces on Angular Distribution of Monochoromatic Reflected Radiation, Journal of Heat Transfer, vol. 87, no. 1, pp. 85-94 Blotevogel, T., Egermann, J., Goldlucke, J., Leipertz, A., Hartmann, M., Schenk, M., Berckmuller, M. : Developing Planar Laser-Induced Fluorescence for the Investigation of the Mixture Formation Process in Hydrogen Engines, SAE Technical paper, 2004-01-1408 Boersma, R.J., Sammes, N.M. : Computational analysis of the gas-flow distribution in solid oxide fuel cell stacks, Journal of Power Sources, vol. 63 pp. 215-219 Bombach, R. : Lecture notes on Laser Spectroscopy at the ERCOFTAC summer school "Introduction to Numerical and Experimental Methods in Combustion Research" ETH Zürich, March 15-21, 2002 [electronic resource, URL: http://cdg.web.psi.ch/ ], accessed 29 September 2008 Brown, C., Wood, T., Thompson, S. : In conversation with Versa Power Systems, Calgary. May 2006. 272 Brown, M., Primdahl, S., Mogensen, M. : Structure/Performance Relations for Ni/Yittria-Stabilized Zirconia Anodes for Solid Oxide Fuel Cells, Journal of Electrochemistry, vol. 147 pp. 475-485 Brett, D.J.L., Aguiar, P., Clague, R., Marquis, A.J., Schöttl, S., Simpson, R., Brandon, N.P. : Application of infrared thermal imaging to the study of pellet solid oxide fuel cells, Journal of Power Sources, vol. 166, pp 112-119 Brodersen, S. : High-Resolution Rotation-Vibration Raman Spectroscopy in Raman Spectroscopy of Gases and Liquids, Editor Weber, A., Springer-Verlag, New York Heidelberg Berlin Cheng, Z., Abernathy, H., Liu, M. : Raman Spectroscopy of Nickel Sulfide Ni3S2, Journal of Physical Chemistry, vol. 111, pp 17997-18000 Choi, S., Wang, J., Cheng, Z., Liu, M. : Surface Modification of Ni-YSZ Using Niobium Oxide for Sulfur-Tolerant Anodes in Solid Oxide Fuel Cells, Journal of the Electrochemical Society, vol. 155 (5), pp B449-B454 Choi, Y., Liu, M. : Functionally Graded Cathodes for Solid Oxide Fuel Cells, DOE Report, Contract No.: DE-FC26-02NT41572 Dauvergne, J. P., Vandoni, G. : Backstreaming of Impurity Gas Through a Leak in a Pressurized Vessel, Presented at ICEC 17, 14-17 July 1998, Bournemouth, UK, pp 209-212, Notes: XXIV, p895 Dicks, A.L. : Advances in catalysts for internal reforming in high temperature fuel cells, Journal of Power Sources, vol. 71 pp. 111-122 Eckbreth, A.C. : Laser Diagnostics for Combustion Temperature and Species (Second Edition), Gordon and Breach, London Egermann, J., Seeger, T., Leipertz, A. : Application of 266 nm and 355 nm Nd:YAG Laser Radiation for the Investigation of Fuel-Rich Sooting Hydrocarbon Flames by Raman Scattering, Journal of Applied Optics, vol. 43 No. 29 pp 5564-5574 Egermann, J., Göttler, A., Leipertz, A. : Application of Spontaneous Raman Scattering for Studying the Diesel Mixture Formation Process Under Near-Wall Conditions, SAE Technical Paper, 2001-01-3496 Eigenbrodt, B.C., Pomfret, M.B., Steinhurst, D.A., Owrutsky, J.C., Walker, R.A. : Direct, In Situ Optical Studies of Ni-YSZ Anodes in Solid Oxide Fuel Cells Operating with Methanol and Methane, Journal of Physical Chemistry Fenner, W.R., Hyatt, H.A., Kellan, J.M., Porto, S.P.S. : Raman Cross Section of Some Simple Gases, Journal of the Optical Society of America, vol. 63 pp73-77 Ferriere, A., Lestrade, L., Robert, J-F. : Optical Properties of Plasma-Sprayed ZrO2-Y2O3 at High Temperature for Solar Applications, Journal of Solar Energy Engineering, vol. 122 pp. 9-13 273 Gay-Desharnais, S.E., Routex, J-Y., Holtzapple, M., Ehsani, M. : Investigation of Hydrogen Carriers for Fuel Cell Based Transportation, SAE Technical Paper 2002-01- 0097 Gilbert, A., Baggott, J. : Essentials of Molecular Photochemistry, Blackwell Scientific Publications, Oxford, England Goodman, L. A. : On the Exact Variance of Products, Journal of the American Statistical Association, vol. 55, No. 292 pp 708-713 Grünefeld, G., Beushausen, V., Andresen, P. : Interfereence-free UV-Laser- Induced Raman and Rayleigh Measurements in Hydrocarbon Combustion Using Polarization Properties, Journal of Applied Physics, vol. B61 pp473-478 Grünefeld, G., Beushausen, V., Andresen, P., Hentschel, W. : Spatially Resolved Raman Scattering for Multi-Species and Temperature Analysis in Technically Applied Combustion Systems: Spray Flame and Flour-Cylinder In-Line Engine, Journal of Applied Physics, vol. B58 pp333-342 Hagris, P. J. : Trace Detection of N2 By KrF-Laser-Excited Spontaneous Raman Spectroscopy, Journal of Applied Optics, vol. 20 No. 1 pp149-152 Hartfield, R.J., Dobson, C., Eskridge, R., Wehrmeyer, J. : Development of a Technique For Separating Raman Scattering Signals from Background Emission With Single-Shot Measurement Potential, Presented at the 33rd AIAA/SAE/ASME/ASEE Joint Propulsion Conference and Exhibit, Seattle, WA, AIAA 97-9957 Hassel, E. P. : Ultraviolet Raman-scattering Measurements in Flames by the Use of a Narrow-band XeCl Excimer Laser, Journal of Applied Optics, vol. 32 pp 4058-4065 Hassel, E. P., Linow, S. : Laser Diagnostics for Studies of Turbulent Combustion, Measurement Science Technology, vol. 11 ppR37-57 Hecht, E., Zajac, A. : Opics, 3rd Edition, Addison Wesley, Massachusetts Hecht, E.S., Gupta, G.K., Zhu, H., Dean, A.M., Kee, R.J., Maier, L., Deutschmann : Methane reforming kinetics within a Ni-YSZ SOFC anode support, Applied Catalysis A: General, vol. 295 pp 40-51 Herzberg, G. : Molecular Spectra and molecular Structure. II. Infrared and Raman Spectra of Polyatomic Molecules. Van Nostrand Rienhold Co., New York, NY. Hofmann, P., Panopoulos, K.D., Fryda, L.E., Kakaras, E. : Comparison Between Two Methane Reforming Models Applied to a Quasi-two-dimensional Planar Solid Oxide Fuel Cell Model, Energy, vol. 34 pp 2151-2157 Hucknall, D. J., Morris, A. : Vacuum Technology Calculations in Chemistry. The Royal Society of Chemistry, Cambridge 274 Ipp, W., Egermann, J., Schmitz, I., Wagner, V., Leipertz, A., Hartmann, M., Schenk, M., : 2D Mapping and Quantification of the In-Cylinder Air/Fuel-Ratio in a GDI Engine by Means of LIF and Comparison to Simultaneous Results from 1D Raman Measurements, SAE Technical Paper 2001-01-1977 Janardhanan, V., Deutschmann, O. : CFD Analysis of a Solid Oxide Fuel Cell with Internal Reforming: Coupled Interactions of Transport, Heterogeneous Catalysis and Electrochemical Processes, Journal of Power Sources, vol. 162 pp. 1192-1202 Janardhanan, V.M., Heuveline, V., Deutschmann, O. : Performance analysis of a SOFC under direct internal reforming conditions, Journal of Power Sources, vol. 172 pp. 296-307 Jiang, L., Barnett, S.A. [2003b]: Operation of anode-supported solid oxide fule cells on methane and natural gas, Solid State Ionics, vol. 158 pp. 11-16 Jiang, Y., Virkar, A.V. [2003a]: Fuel Composition and Diluent Effect on Gas Transport and Performance of Anode-Supported SOFC’s, Journal of the Electrochemical Society, vol. 150 pp. A942-A951 Johnsen, K., Grace, J.R. : High-Temperature Attrition of Sorbents and a Catalyst for Sorption-Enhanced Steam Methane Reforming in a Fluidized Environment, Powder Technologies, vol. 173, pp 200-202 Ju, G., Reifsnider, K., Huang, X. : Infrared Thermography and Thermoelectrical Study of a Solid Oxide Fuel Cell, Journal of Fuel Cell Science and Technology, vol. 5, pp 031006 1-6 Kang, Y., Li, J., Cao, G., Tu, H., Li, J., Yang, J. : One-dimensional Dynamic Modelling and Simulation of a Planar Direct Internal Reforming Solid Oxide Fuel Cell, Product Engineering and Chemical Technology, vol. 17 pp 304-317 Kaye, G., Laby, T. : Tables of Physical & Chemical Constants National Physical Laboratory [electronic resource, URL: http://www.kayelaby.npl.co.uk/chemistry/3_1/3_1_3.html ], accessed 29 September 2008 Keegan, H.J., Schleter, J.C., Wiedner, V.R. : Effect of Surface Texture on Diffuse Spectral Reflectance, Part A: Diffuse Spectral Reflectance of Metal Surfaces, Symposium on Thermal Radiation of Solids (Katzoff, S. Editor), NASA SP-55, pp. 165- 169 Kleimeyer, J. A., Fister III, J. C., Zimmerman, J. Z., Harris, J. M. : Liquid Filters for UV Resonance Raman Spectroscopy, Journal of Applied Spectroscopy, vol. 50 pp1597 Knapp, M., Beushausen, V., Hentschel, W., Manz, P., Grünefeld, G., Andresen, P. : In-Cylinder Mixture Formation Analysis with Spontaneous Raman Scattering Applied to a Mass-Production SI Engine, SAE Technical Paper, 970827 275 Larminie, J., Dicks, A. : Fuel Cell Systems Explained (1st Edition), John Wiley & Sons, Ltd, England. Laserna, J. J. : Modern Techniques in Raman Spectroscopy, John Wiley and Sons, Chichester Leng, Y.J., Chan, S.H., Khor, K.A., Jiagn, S.P. : Performance evaluation of anode-supported solid oxide fuel cells with thin film YXZ electrolyte, International Journal of Hydrogen Energy vol. 29 pp. 1025-1033 Lewis, I.R., Edwards, H.G.M. : Handbook of Raman Spectroscopy: From the Research Laboratory to the Process Line, CRC Press. Lipp, F., Hassel, E. P., Janicka, J. : Comparison of UV Raman Spectroscopy with 248 nm and 308 nm For Determination of Flame Temperature and Concentrations, Proceedings of the Joint Meeting British and German Sections Combustion institute, Cambrige, (unpublished) pp 255-258 Long, D.A. : Raman Spectroscopy, McGraw-Hill, New York Lu, X., Faguy, P.W., Liu, M. : In Situ Potential-Dependent FTIR Emission Spectroscopy A Novel Probe for High Temperature Fuel Cell Interfaces, Journal of the Electrochemical Society, vol. 149, pp A1293-A1298 Mahan, J.R. : Radiation Heat Transfer. A Statistical Approach (1st Edition), John Wiley and Sons, New York Maher, R.C., Cohen, L.F., Lohsoontorn, P., Brett, D.J.L., Brandon, N.P. : Raman Spectroscopy as a Probe of Temperature and Oxidation State for Gadolinium- Doped Ceria Used in Solid Oxide Fuel Cells, Journal of Physical Chemistry, vol. 112, pp 1497-1501 Matsuzaki, Y., Yasuda, I. : The poisoning effect of sulfur-containing impurity gas on a SOFC anode: Part I. Dependence on temperature, time, and impurity concentration, Solid State Ionics, vol. 132 pp. 261-269 Mckenzie, R. L. : Rayleigh Rejection Filters for 193 nm ArF Laser Raman Spectroscopy, Optics Letters, vol. 18 No. 12, pp 995-997 Musikant, S. : Optical Materials: An Introduction to Selection and Application, vol. 6 Optical Engineering, Dekker, New York Namioka, T. : Absoprtion Spectra of H2 in the Vacuum Ultraviolet Region. II. The 8’-X, B”-X, D-X, and D’-X Bands, Journal of Chemical Physics, vol. 41 pp2141 Nandula, S. P., Brown, T. M., Pitz, R. W., Debarber, P. A. : Single-Pulse, Simultanerous Multipoint Multispecies Raman Measurements in Turbulent Nonpremixed Jet Flames, Optics Letters, vol. 19 pp 414-416 276 Nutter, G.D. : Radiation Thermometry-The Measurement Problem, Applications of Radiation Thermometry, ASTM STP 895, Richmond J.C., DeWitt D.P., Eds., American Society for Testing and Materials pp. 3-23 Pisani, L., Murgia, G., : An Analytical Model for Solid Oxide Fuel Cells, Journal of the Electrochemical Society, vol. 154 pp. B793-B801 Pitz, R. W., Wehrmeyer, J. A., Bowling, J. M., Cheng, T. : Single Pulse Vibrational Raman Scattering By a Broadband KrF Excimer Laser In a Hydrogen-Air Flame, Journal of Applied Optics, vol. 29, No. 15 pp 2325-2332 Pomfret, M.B., Owrutsky, J.C., Walker, R.A. [2010a]: In Situ Optical Studies of Solid- Oxide Fuel Cells, Annual Review of Analytical Chemistry, vol. 3, pp 151-174 Pomfret, M.B., Marda, J., Jackson, G.S, Eichhorn, B.W., Dean, A.M., Walker, R.A. : Hydrocarbon Fuels in Solid Oxide Fuel Cells: In Situ Raman Studies of Graphite Formation and Oxidation, Journal of Physical Chemistry, vol. 112, pp 5232-5240 Pomfret, M.B., Owrutsky, J.C., Walker, R.A. : High-Temperature Raman Spectroscopy of Solid Oxide Fuel Cell Materials and Processes, Journal of Physical Chemistry, vol. 110, pp 17305-17308 Pomfret, M.B., Owrutsky, J.C., Walker, R.A. : In Situ Studies of Fuel Oxidation in Solid Oxide Fuel Cells, Analytical Chemistry, vol. 79, pp 2367-2372 Pomfret, M.B., Steinhurst, D.A., Kidwell, D.A., Owrutsky, J.C. [2010b]: Thermal Imaging of Solid Oxide Fuel Cell Anode Processes, Journal of Power Sources, vol. 195, pp 257-262 Prasad, P., Elnashaie, S.S.E.H. : Novel Circulating Fluidized-Bed Membrane Reformer for the Efficient Production of Ultraclean Fuels from Hydrocarbons, Journal of Industrial and Engineering Chemistry, vol. 41, pp 6518-6527 Primdahl, S., Hendricksen, P.V., Larsen, P.H., Kindl, B., Mogensen, M. : Evaluation of SOFC Test Data, Singhal, S.C.; Yokokawa, H. (eds.), Electrochemical Society Proceedings, Pennington, PV2001-16 pp. 932-941 Reckers, W., Hüwel, L., Grünefeld, G., Andressen, P. : Spatially Resolved Multispecies and Temperature Analysis In Hydrogen Flames, Journal of Applied Optics, vol. 32 No. 6 pp907-924 Refractory Ceramic Fibers Coalition : [Electronic resource, URL: http://www.rcfc.net/characteristics.htm, accessed 15 August 2010. Roth, A. : Vacuum Technology (3rd Edition), North Holland, London Rothe, E. W., Andressen, P. : Application of Tunable Excimer Lasers to Combustion Diagnostics: A Review, Journal of Applied Optics, vol. 36 No. 18 pp 3971- 4033 277 Sánchez, D., Chacartegui, R., Munoz. A., Sánchez, T. : On the Effect of Methane Internal Reformation Modelling in Solid Oxide Fuel Cells, International Journal of Hydrogen Energy, vol. 33 pp 1834-1844 Saunders, J. E. A., Davy, M. H. : Broadband 308 nm vibrational Raman Spectroscopy of Gaseous Species Using a Potassium Hydrogen Phthalate Liquid Filter and Polarization Fluorescence Suppression, Review of Scientific Instruments, vol 81 013108 Schrötter, H. W., Klöckner, H. W. : Raman Scattering Cross Sections in Gases and Liquids in Raman Spectroscopy of Gases and Liquids, Editor Weber, A., Springer- Verleg, Berlin Schrötter, H.W. : Linear Raman Spectroscopy: A State of the Art Report in Non- Linear Raman Spectroscopy and Its Chemical Applications, Editors Keifer, W., Long, D.A., Nato Advanced Study Institute Series, Reidel, Dordrecht Siegel, R., Howell, J. : Thermal Radiation Heat Transfer (4th Edition), Taylor and Francis, New York Singhal, S.C., Kendall, K. : High Temperature Solid Oxide Fuel Cells: Fundamentals, Design and Applications (1st Edition), Elsevier, Oxford Skoog, D.A., West, D.M. : Fundamentals of Analytical Chemistry (4th Editoin), CBS College Publishing, Philadelphia Spangenberg, D.B., Strang, A.G, Chamberlin, J.L. : Effect of Surface Texture on Diffuse Spectral Reflectance, Part B: Surface Texture Measurements of Metal Surfaces, Symposium on Thermal Radiation of Solids (Katzoff, S. Editor), NASA SP-55, pp. 169-177 Stevenson, C.L., Vo-Dinh, T. : Signal Expressions in Raman Spectroscopy in Modern Techniques in Raman Spectroscopy, Editor Laserna, J.J., Wiley and Sons, Chichester. Sudaprasert, K., Travis, R.P., Martinez-Botas, R.F. : A computational fluid dynamics model of a solid oxide fuel cell, Proceedings of IMechE, Journal Power and Energy, vol. 219 part A pp. 159-167 Sudaprasert, K., Travis, R.P., Martinez-Botas, R.F. : A Study of Temperature Distribution Across a Solid Oxide Fuel Cell Stack, Journal of Fuel Cell Science and Technology, vol. 7 011002 Suwanwarangkul, R., Croiset, E., Fowler, M.W., Douglas, P.L., Entchev, E., Douglas, M.A. : Performance comparison of Fick’s, dusty-gas and Stefan- Maxwell models to predict the concentration overpotential of a SOFC anode, Journal of Power Sources, vol. 122 pp. 9-18 Souza, S., Visco, J., De Jonghe, L.C. : Reduced-Temperature Solid Oxide Fuel Cell Based on YSZ Thin-Film Electrolyte, Journal of Electrochemistry, vol. 144 pp. L35- L37 278 Tanaka, H., Sawai, S., Morimoto, K., Hisano, L. : Measurement of Spectral Emissivity and Thermal Conductivity of Zirconia by Thermal Radiation Calorimetry, Journal of Thermal Analysis and Calorimetry, vol.64 pp. 867-872 Teagan, P., Stobart, R. : Fuels for Fuel-cell Powered Vehicles, SAE Technical Paper 2001-01-0001 The Rough Guide to Energy : Eds. Lockett, A., Milton, M. (1st Edition), Rough Guides, London Timmermann, H., Fouquet, D., Weber, A., Ivers-Tiffée, E., Hennings, U. : Internal Reforming of Methane at Ni/YSZ and Ni/CGO SOFC Cermat Anodes, Fuel Cells vol. 6 pp 307-313 Tolboom, R. A. L., Dam, N. J., Sijtsema, N. M., Meulen, J. J. : Quantitative Spectrally Resolved Imaging through a Spectrograph, Optics Letters, vol. 28 No. 21 pp 2046-2048 Toth, R. A., Brown, L.R., Miler, C. F., Devi, V. M., Benner, D. C. : Spectroscopic Database of CO2 Line Parameters: 4300 – 7000 cm-1, Journal of Quantitative Spectroscopy and Radiative Transfer, vol 109 pp906-921 Toyloukian, Y.S., DeWitt, D.P. : Thermal Radiative Properties-Metallic Elements and Alloys, Vol. 7 IFI Plenum New York Union Gas : [Electronic resource, URL: http://www.uniongas.com/aboutus/About ng/composition.asp], accessed 19 January 2010. Voitsekhorskaya, O. K., Zuev, V. E., Ippolitov, I. I., Makushkin, Yu. S. : Calculation of the Spectral-Line Intensities of Water Vapour Taking Into Account the Vibration Rotation Interaction, Journal of Applied Spectroscopy, vol. 17 pp958 Weber, A. : High-Resolution Rotational Raman Spectra of Gases in Raman Spectroscopy of Gases and Liquids, Editor Weber, A., Springer-Verlag, New York Heidelberg Berlin Wehrmeyer, J. A., Cheng, T., Pitz, R. W. : Raman Scattering Measurements In Flames Using a Tunable KrF Excimer Laser, Journal of Applied Optics, vol. 31 No. 10 pp 1495-1503 Wilson, J.R., Kobsiriphat, W., Mendoza, R., Chen, H.Y., Hiller, J.M., Miller, D.J., Thornton, K., Voorhees, P.W., Adler, S.B., Barnett S.A. : Three-dimensional Reconstruction of a Solid-oxide Fuel-cell Anode [electronic resource, URL: http://www.nature.com/nmat/journal/v5/n7/full/nmat1668.html ], accessed 16 Jan 2008. Wilson, R.G. : Hemispherical Spectral Emittance of Ablation Chars, Carbon and Zirconia to 3700 K, Langley Research Centre, NASA TN D-2704 Wu, W., Ross, M. : Modelling of Direct Injection Diesel Engine Fuel Consumption, SAE Technical Paper 971142 279 Yakabe, H., Hishinuma, M., Uratani, M., Matsuzaki, Y., Yasuda, I. : Evaluation and modeling of performance of anode-supported solid oxide fuel cell, Journal of Power Sources, vol. 86 pp. 423-431 Zhao, H., Ladommatos, N. : Engine Combustion Instrumentation and Diagnostics, Society of Automotive Engineers, Inc., Warrendale, Pa. Zhu, H., Kee, R.J., Janardhanan, V.M., Deutschmann, O., Goodwin, D.G. : Modeling Elementary Heterogeneous Chemistry and Electrochemistry in Solid-Oxide Fuel Cells, Journal of The Electrochemical Society, vol. 152 pp. A2427-A2440 Zinovik, I., Poulikakos, D., : Modelling the Temperature Field in the Reforming Anode of a Button-shaped Solid Oxide Fuel Cell, Electrochimica Acta, vol. 54 pp 6234- 6243 280 Appendices A Optical Reformer Flow Rate and Mean Stream Velocity Calculations The optical reformer is to be operated under conditions (temperature and gas flow rates) similar to those found in a direct internally reforming (DIR) solid oxide fuel cell (SOFC). As such, it is possible to calculate the rate at which reacting gases (methane and steam) are supplied to anode side of the cell per unit of reacting surface area. This figure is dependent on the specific operating conditions of the SOFC which, depending on power output requirements could be continually fluctuating. It was therefore necessary to assume conditions that produced the highest volume flow rates which would in turn create the most challenging scenario for a uniform flow to develop over the reforming surface. Finally, mean stream velocity calculations for various tube and inlet dimensions at a reactant gas temperature similar to those found in a SOFC are presented. A.1 SOFC Fuel Consumption The amount of fuel utilised, or utilisation factor (U.F.) can be defined as follows [Larminie et al., 2000]: Fn nNI FU fuelcell⋅ ⋅⋅= &.. (A.1) Where: n = number of moles of electrons released per mol of fuel Ncell = number of cells in the fuel cell stack 281 fueln& = molar flow rate of fuel (mol . s-1) F = Faraday’s constant, 96485 (C . mol-1) I = current density (A . cm-2) For the optical reformer the half cell has an active area of 50 cm2. Assumptions have to be made for the current density, with a best/worst case scenario being considered where the current density is high, 1 Acm-2, and U.F. of 0.5 or 50%. The number of cells is one and the number of electrons electrochemically released per mol of methane is eight. Substituting these values into A.1 and re-arranging in terms of fueln& a flow rate in terms of standard litres per minute (SLPM) can be produced, equation A.2. s P TRnSLPM fuel 60 ~ ⋅⋅= & (A.2) Where: fueln& = molar flow rate of fuel (mol . s-1) T = ‘standard’ temperature, 298.15 (K) P = atmospheric pressure 101.325 (kPa) R~ = Universal gas constant, 8.31447 (kJ . kmol-1 . K-1) s = seconds, conversion factor (s) This gives an upper limit for the flow rate of methane of 0.2 SLPM. From the overall methane/steam reformation reaction, Equation A.3, it can be seen that two moles of water vapour are required per mole of methane gas; however occasions can arise where the ratio of steam to methane is twice this, giving four moles of steam per mole of 282 methane. Flow rates for a 50 × 100 mm2 half cell can therefore potentially, in a ‘worst case’ scenario be up to 1.0 SLPM. )(4)()(2)( 2224 gHgCOgOHgCH +→+ (A.3) A.2 Mean Stream Velocities With the maximum volume flow rate determined for the reformer it is possible to calculate some approximate mean stream velocities for a number of geometric arrangements. As a SOFC operates at temperatures around 800°C it is necessary to convert the volume flow rate at standard conditions to a mass flow rate that is independent of temperature. For simplicity the gas stream is considered to consist of a single component, which was arbitrarily chosen to be methane but could have easily been water vapour (which is its greater constituent). RT VPm & & = (kg . s-1) (A.4) Where: V& = volume flow rate of fuel (m3 . s-1) R = Gas constant for methane, 0.518 (kJ . kmol-1 . K-1) P = atmospheric pressure, 101.325 (kPa) T = ‘standard’ temperature, 298.15 (K) For a volume flow rate of 1.0 SLPM a mass flow rate of 1.09×10-5 kg.s-1 is calculated. From continuity the mean stream velocity can be expressed as shown in equation A.5. 283 A mU ρ &= (m . s-1) (A.5) Where: U = mean stream velocity (m . s-1) ρ = density = RT P as per ideal gas law (kg . m-3) A = cross sectional area of conduit (m2) T = temperature (K) For a ¼” Swagelok™ tube (I.D. 3.3 mm), cross sectional area = 8.55×10-6 m2, carrying gas at 800°C, the mean stream velocity is 7.16 ms-1. By comparison, a plenum chamber with 6 inlets or outlets, each of 2 mm in diameter, has a combined cross sectional area of 1.86×10-6 m2 giving a mean stream velocity of 3.25 ms-1. A.3 Reynolds Number Calculations With the maximum volume flow rate and thus mass flow rate determined it is possible to calculate Reynolds numbers in the reformer channel for the two extreme channel dimensions, 7 mm and 2 mm respectively. μ ρ DU=Re (A.6) Where: D = hydraulic diameter = P A4 (m) A = cross sectional area of conduit (m2) 284 P = wetted perimeter (m) μ = dynamic viscosity (kg . m-1 . s-1) Using mean stream velocities calculated using equation A.5, a channel of width 50 mm and a viscosity calculated for methane at 800°C using a generalized chart nondimensionalized for critical-point properties ( μ = 2.7×10-5 kg.m-1.s-1) the Reynolds numbers calculated for 2 and 7 mm high channels respectively are 14.0 and 14.7, indicating a laminar regime. 285 B FLUENT Simulation B.1 Reformer Inlet /Outlet and Channel Configurations With the maximum flow rates calculated (Appendix A) a selection of inlet and channel configurations were simulated using FLUENT. Two inlet configurations were simulated; the first with single, 3.3 mm I.D., inlet and outlet, the second with six, 1.0 mm I.D., inlets and outlets respectively, Figure B.1. Figure B.1 Diagram of overall channel dimensions and inlet / outlet The length of the channel was kept constant (133 mm) but a ‘step’ was introduced 12 mm along the channel floor that reduced the channel height to 2 mm for a distance of 116 mm in order to investigate the effect of a reduced channel heights should the channel be reconfigured in later studies, Figure B.2. 133 mm Multiple Inlets / Outlets Single Inlet / Outlet 54 mm 286 Figure B.2 Diagram showing station positions at which velocity profiles were plotted as well as the stepped reformer channel configuration As previously discussed in §3.5 the desired flow pattern was that of a ‘quasi’ fully developed flow similar to that present in a SOFC. In order to ascertain if a well developed flow was achieved the velocities parallel to the reactive surface were plotted at four stations along the channel surface, Figure B.2. The FLUENT simulation was performed a viscous laminar model, the fluid was specified as methane for ease of calculation, with a viscosity adjusted for a temperature of 800 °C. The boundary conditions were non-slip walls, mass flow inlets with flow perpendicular to the boundary. The outlet(s) were set to pressure outlets with flow also perpendicular to the boundary. The simulation was initialized from the inlet(s) with residuals set to 1×10-7 and the singe inlet/outlet, flat floor converged in 233 iterations. Inlet(s) Outlet(s) Inlet(s) Outlet(s) 15 mm 45 mm 105 mm 75 mm Stations Flat floor Stepped floor 7 mm 2 mm 287 B.2 FLUENT Results Figure B.3 Plot of velocity vs. vertical position for full height channel Figure B.4 Plot of velocity vs. vertical position for full height channel 288 Figure B.5 Plot of velocity vs. vertical position for a channel of height 2 mm with single inlet / outlet Figure B.6 Plot of velocity vs. vertical position for a channel of height 289 B.3 Conclusions from FLUENT Simulations It can be seen from the results plotted above and the three dimensional flow pattern produced by the FLUENT simulation that the multiple inlet/outlet configuration provides the quasi fully developed flow that was desired. This is shown by the absence in variation in velocity profile between the four stations along the reformer channel. Similarly, it can also be seen that the reduction in channel height has no detrimental effect on this quasi developed flow. 290 C Molar Absorption Calculation C.1 Beer-Lambert Law The Beer-Lambert law equates absorption, as a function of wavelength, to the product of molar absorptivity, ε (m2.mol-1), multiplied by the molar concentration, b (mol.m-3), of the medium itself and the path length through the medium, c (m), Equation C.1 [Gilbert and Baggott, 1991]. ( ) bcA ελ = (C.1) This expression may be re-written to show how the intensity of the radiation, I (W m-2), that has passed some distance through the absorbing medium will decrease exponentially with increasing path length with respect to the intensity of the incident radiation, Io, entering the medium. bc o eII ε−= (W . m-2) (C.2) In order to quantify the associated decrease in radiation intensity it is necessary to ascertain the aforementioned absorptivity, molar concentration of the absorptive media, and path length. C.2 Molar Absorptivity Calculation As mentioned in the main text there is a scarcity of information regarding the molar absorptivities of species in the portion of the wavelength at which the infrared (IR) thermometer operates, however Toth et al. [Toth et al., 2008] reports a figure for carbon 291 dioxide of 5.966 x10-20 cm2 molecule-1 for between 4300 – 7000 cm-1 which corresponds to a spectral range of 2.3 – 1.4 μm, which only partly covers the range of the IR thermometer (1.8 μm and 3.0 μm). This was however considered sufficient to give an indication of the absorption effects in this spectral region. It was also assumed that the channel was completely filled with a 100% carbon dioxide gas stream, which is not what would occur in reality, but would serve as indicative of absorptive effects. The path length of the radiation emitted from the anode surface is 6 mm, the height of the reformer channel. Finally it was necessary to calculate the molecular density of the gas at the reformers operating temperature, in this case 700 °C. Using the ideal gas law, Equation C.3. TR PVn ~= (mol . m-3) (C.3) Where: R~ = universal gas constant, 8.31447 (kJ . kmol-1 . K-1) T = temperature, 973 (K) V = volume (m3) P = atmospheric pressure,101.325 (kPa) Substituting these values into Equation C.3 gives a molecular density of 7.5 x1018 molecules cm-3, which in turn this returns an absorptivity of 0.25. However, in reality carbon dioxide at most makes up 20% of the reformer gases as the ideal overall reformation reaction shows, Equation 6.1. )(4)()(2)( 2224 gHgCOgOHgCH +→+ molkJH o K /4.191973 +=Δ (8.1) 292 The reduction in carbon dioxide concentration results in an absorptivity of 0.05, which while not negligible can easily be accounted for by a calibration procedure or factored into the IR thermometry results. In addition the water carrier gas in the reformer case would further significantly dilute the carbon dioxide. A similar process would occur in the SOFC case, with water vapour diluting the carbon dioxide and serving to offset the effects absorption from this specific species. However, the absorptivity of all anode gases would also have to be considered if this type of post processing of IR thermometry temperature data were to take place. 293 D Experimental Schedule D.1 Chronological Experimental Schedule Date Temperature Channel Position¹ Beam Energy /mJ Charge Voltage /kV Laser Counts /x103 13/05/10 600 °C 0 mm 169 22.0 11128 14/05/10 600 °C 10 mm 169 22.3 11316 15/05/10 600 °C 20 mm 169 22.7 11512 16/05/10 600 °C 30 mm 169 23.1 11709 17/05/10 600 °C 50 mm 169 23.3 11904 18/05/10 600 °C 70 mm 169 24.1 12096 19/05/10 600 °C 87 mm 169 24.5 12298 23/05/10 700 °C 0 mm 169 26.7 12492 24/05/10 700 °C 10 mm 169 24.1† 12687 25/05/10 700 °C 20 mm 169 24.4 12879 27/05/10 700 °C 30 mm 169 24.7 13069 28/05/10 700 °C 50 mm 169 25.5 13260 29/05/10 700 °C 70 mm 169 24.1† 13452 30/05/10 700 °C 87 mm* 169 24.5 13647 31/05/10 25 – 700 °C 10 mm 169 24.7 13672 ¹ Each channel position requires the repositioning of the light sheet, collection optic and adjustment of the spectrometer slit mask. † Partial Gas Replacement (PGR). PGR is the periodic injection of halogen gas into the laser tube when the charge voltage reaches a preset threshold. * New window gasket fitted prior to testing Table D.1 Chronological experimental schedule. 294 D.2 Experimental Test Points The following test points were run through in order (test point 1 to 8) due to the limitation of the humidifier, which cannot reduce temperature rapidly, so can only be increased throughout the course of a days testing schedule. Similarly, nitrogen bypass values were selected to enable multiple test points to utilise single humidifier set points for multiple test points. Test Point Total Volume Flow Rate /SLPM % Humidity Ratio of H2O to CH4 CH4 /SLPM N2 wet /SLPM N2 dry /SLPM 1 1.00 0.250 3.0 0.060 0.660 0.100 2 1.00 0.250 2.0 0.090 0.630 0.100 3 1.00 0.250 2.0 0.060 0.420 0.400 4 0.60 0.250 2.0 0.060 0.420 0.000 5 1.00 0.429 3.0 0.090 0.540 0.100 6 0.60 0.429 3.0 0.060 0.360 0.000 7 0.60 0.429 2.0 0.090 0.330 0.000 8 0.60 0.818 3.0 0.090 0.240 0.000 Table D.2 Experimental test points. 295 D.3 Rotameter Set Points The following rotameter tubes, floats and scale readings were used for the respective test points. Test Point CH4 /mm 042-15-N Sapphire N2 wet /mm 062-01-N S. Steel N2 dry /mm 032-31-N S. Steel 1 56 120 46 2 83 115 46 3 56 76 131 4 56 76 0 5 83 99 46 6 56 65 0 7 83 59 0 8 83 41 0 Table D.3 Rotameter set points D.4 Calibration Rotameter Set Points The following gas mixture ratios, flow rates, rotameter tubes, floats and scale readings were used for the respective calibration points. Gas mixture ratio, H2/N2 H2 /SLPM N2 /SLPM H2 /mm 045-15-N S.Steel N2 dry /mm 032-31-N S. Steel 20/80 0.1 0.4 34 73 50/50 0.25 0.25 85 43 80/20 0.4 0.1 123 11 100 0.5 0.0 144 0 Table D.4 Calibration rotameter set points 296 E Molar Flow Conductance In an intermediate flow regime both viscous and molecular effects must be considered. The precise behaviour of an intermediate flow is the subject of multiple parameters; primarily the rate of fluid flux due to the pressure differential either side of the conduit in question, molar concentrations and the mean velocity of the molecules in question. The accurate quantification of each of these parameters, particularly in the context of a ceramic gasket consisting of randomly orientated fibres, is difficult if not impossible. However, the parameter of most interest, the molecular flow of air into the reformer channel, can be determined for a simplified case and compared on an order of magnitude basis. Therefore, the molecular conductance of a finite area will be calculated for air and hydrogen, at the conditions present inside the optical heater, and then compared against those required by the analysis presented in Appendix F to asses the feasibility of molecular flow being responsible for the nitrogen and oxygen related products detected in the channel. E.1 Conductance of an Aperture For an aperture of negligible length the transmission probability, the probability that a molecule impinging the aperture area will pass through it, is considered to be unity, such that all molecules will pass through the aforementioned aperture. Tubes of finite length are characterised by transmission probabilities which are less than unity. This is due to molecules interacting with the interior surfaces of the tube with the resultant paths of the molecules being random and a portion of the molecules being ejected back out of the tube. The greater the length of the tube the fewer molecules can be expected to pass completely through it. However, as a first approximation an aperture 297 will suffice to determine the potential molecular conductance at the conditions inside the optical heater. The conductance of an aperture can be calculated as follows, Equation E.1. AcCap 4 = (m3 . s-1 per m2) (E.1) Where: apC = aperture conductance (m3 . s-1 per m2) c = average molecular velocity (m . s-1) A = aperture area (m2) The average molecular velocity is a function of the molecular weight of the molecule in question and the temperature, Equation E.2. M TRc π ~8= (m . s-1) (E.2) Where: R~ = universal gas constant, 8.31447 (kJ . kmol-1 . K-1) T = temperature (K) M = molecular mass (kg . mol-1) In the case of air, it can be treated as an ideal gas mixture having a molecular mass of 0.029 kg mol-1. At 600 °C this corresponds to an average velocity of approximately 800 m/s. By way of comparison the average molecular velocity of air at 20 °C is 465 m s-1. Similarly, for hydrogen at 600 °C the average molecular velocity is 298 3000 m/s which illustrates’ the widely differing behaviour of the substantially lower mass hydrogen molecule. Using these velocities in Equation E.1 gives molecular conductances of 200 and 750 m3 s-1 per m2 for air and hydrogen respectively, or more tangibly 20 and 75 L s-1 per cm2. An approximation of the total area of gasket material between at window/reformer channel interface suggests there is 0.85 cm2 of gasket material. In the absolute maximum case, if the gasket material were considered ‘free’ space this corresponds to flow rates of 1020 and 3800 litres per minute of air entering and hydrogen leaving the channel respectively. The analysis of Appendix F suggested that in the order of 0.2 SLPM of air would be required to enter the channel to result in the Raman spectra and corresponding mass fractions in the 0.5 SLPM, 100% hydrogen calibration case. This rate of air ingress would require a probability of transmission of 0.02%. As was also stated in Appendix F, the final gas composition of the gas channel could well be the result of a localised viscous leak upstream of the anode sample and a generalised ingress of air at all other window/reformer channel interface locations, further reducing the probability of transmission necessary to generate the Raman spectra recorded. 299 F Approximation of the Magnitude of Air in the Channel Previously it was assumed that in all likelihood the composition of gas leaking from the reformer would be equivalent to that inside the reformer – no gas species leaking preferentially. Similarly, it was assumed that the direction of the leak would be from the reformer channel into the surrounding atmosphere. While the review of molecular flow would suggest that the composition of the exiting gas is potentially a function of mean free path and molar concentration gradients, the composition of the gas leaving will continue to be considered equivalent to that passing through the channel due to the small variation in molecule size and the extreme concentration gradients of the 100% hydrogen case. However, the potential counter flow of gas through the gasket will be considered in one of the following calculations to determine the approximate magnitude of the ingress/evolution of air, whichever should be occurring. The analysis consists of considering a unit volume of pure hydrogen entering the channel and then becoming mixed with a volume of air, the source of which will either be evolved from the surface or through molecular flow through the window gasket. The oxygen content of air is assumed to, upon entering the channel in the gas phase, undergo complete catalytic combustion and produce water. This in turn will result in an overall gas composition recorded via Raman spectroscopy, given the relative normalised Raman cross sections that have previously been determined. The quantity of air required to produce this Raman spectra is then calculated accordingly. The two cases are represented schematically below, Figure F.1 300 Figure F.1 De-absorption and air ingress plus hydrogen leak models The distinction between scenario A and B is that a gas leak in case A will not fundamentally alter the overall composition of the gas in the channel. In scenario B, with H2 x(0.21 O2 + 0.79 N2) + H2 + x(0.21O2 + 0.79 N2) H2 – 0.42xH2+ 0.42xH2O + 0.79xN2 Scenario A: De-adsorption of Air H2 x(0.21 O2 + 0.79 N2) + H2 – xH2 + x(0.21O2 + 0.79 N2) H2 – 0.42xH2 - xH2 + 0.42xH2O + 0.79xN2 Scenario B: Ingress of air plus hydrogen leak xH2 - 301 an equivalent quantity of hydrogen leaving as air entering, the final composition is a function of both how much air ingresses the channel, as well as how much hydrogen leaves. The results of the two sets of calculations are displayed below in terms of an air evolution rate ‘x’ for scenario A and air and hydrogen leak rate ‘x’ for scenario B, Figure F.2a and b and Figure F.3a and b. Also, given the association of the ‘x’ term with every component in the final gas composition, each component’s Raman signal was used to predict the leak evolution/leak term. (a) 600 °C 302 (b) 700 °C Figure F.2a and b. Air evolution rate for scenario A. 100% hydrogen calibration condition, 0.5 SLPM total flow rate. 303 (a) 600 °C (b) 700 °C Figure F.3a and b. Air and hydrogen leak rate for scenario B. 100% Hydrogen calibration condition, 0.5 SLPM total flow rate. 304 Both Figure F.2a and b and Figure F.3a and b show good agreement between the three constituent species present in the channel gas. This is not unexpected, given the assumptions necessary to arrive at relative normalised Raman cross sections which makes the agreement a foregone conclusion. What the agreement does however indicate is the amount of error between the species line counts, which in all cases tends to increase the spread of the points the further into the channel data is recorded. The second striking feature is the magnitude of the air evolution/ingress occurring in the channel. In the most extreme case – 700 °C, scenario A – almost an equivalent amount of air to hydrogen flowing into the channel would be required to be evolved to produce the composition of the channel at 87 mm using the Raman spectra. It could also be argued that in reality an even greater quantity of gas would have to be evolved as by this point air as well as hydrogen has escaped through the gasket material. In addition, scenario A sees widely varying air evolution rates for the 600 and 700 °C cases. This is perhaps an indication that the air evolution hypothesis is not feasible, as it would suggest that the process is highly temperature dependant. Firstly, at room temperature air was detected in gas exiting the channel and secondly, one has to question if the continued elevation in temperature would continue to see the rapid increase in air evolution. While additional testing at a range of temperatures was not carried out to confirm or disprove the temperature dependence of air evolution it does raise a significant question. Another noticeable feature form all of the graphs is the initial quantity of air present in the channel at ‘0 mm’. This figure is approximately 0.1 SLPM of air for all cases. This is another surprising result for the air evolution case as up until this physical position in the channel there has been no anode material from which to evolve air from. However, if considering molecular flow through the gasket, there has been significant 305 opportunity for this to have taken place, specifically via the front window gasket and the large potential molecular flow zone at the front of the top window. For scenario B the variation in flow rates between the two temperature cases is less pronounced. In addition the magnitude of the leak is less extreme. This makes the molecular flow case somewhat more feasible. Another factor in favour of the validity of the molecular flow argument is that the source of the air is essentially infinite whereas the air evolution case must be considered finite. So not only does the air evolution case require a large quantity of air, it also suggests that it is evolved from an ever decreasing source. The magnitude of the hydrogen/air leak is still comparatively large for scenario B, which used a ratio of one to one for air entering and hydrogen leaving the channel. This equivalent leak rate is, in all likelihood, an over simplification; this is confirmed via diffusivity calculations. In reality an additional factor could act to lower the quantity of air required to enter the channel and still produce the Raman spectra recorded. It is highly likely that an intermediate flow regime exists, in which case over-pressurisation of the channel must be considered. With a pressure differential an even greater amount of hydrogen would be driven from the channel via a viscous regime, particularly from larger orifices at the window corners upstream of the anode surface. This would substantially reduce the amount of air required to enter the channel via a molecular flow mechanism as well as reducing over-pressurisation and thus viscous effects downstream of a larger leak site. However, in reality due to the relativistic nature of the Raman spectra, it is impossible to determine the amount of air required to enter the channel because the magnitude of any hydrogen leak is unknown, as is the ratio of hydrogen leakage to air ingress. Significantly, the evidence strongly suggests that the evolution or de-adsorption of air is not feasible, whereas, given a sizeable hydrogen leak upstream and a portion of 306 air passing via molecular flow regime into the channel the Raman spectra are extremely feasible. Given that the aim of this research was to determine if Raman spectroscopy was feasible in the high temperature environment similar to that encountered in an SOFC the precise magnitude of the air leak and thus inaccuracy in the experimental apparatus as currently designed is not the focus of the work presented in this thesis. All results were therefore presented under the proviso that an inadequacy in the experimental apparatus exists and that further work would be required to improve the quality of the results so that they may be directly related with those either computer simulated or derived from a non-leaking SOFC reforming device. 307 G Polarizer and Image Setup Procedures Due to the comparatively weak nature of the Raman signal it is critical to optimise the signal intensity and collection efficiency. In order to achieve this experimental setup is of paramount importance. This appendix therefore lays out in detail the procedure required to optimise polarizer orientation and the position of the collection volume with respect to the position of the light sheet. G.1 Polarizer Setup In the work presented, the role of the polarizer was distinct from that as described by Saunders and Davy [Saunders and Davy, 2010]. In this instance the polarizer was employed to reduce background fluorescence rather than eliminate it entirely. As discussed in §2.7, it is known that the light collected orthogonally from Raman scattering is polarized [Hecht et al. 1997]. Fluorescence does not exhibit this behaviour, and so any light emitted as a result of fluorescence (from the anode window surfaces) will be the result of any incident light, regardless of orientation. Therefore, to achieve maximum Raman scattering signal to fluorescence ratio it is advantageous to use a polarized light source. Light emitted from the Lambda Physick COMPexPro 102 XeCl broadband laser was not polarized, therefore requiring a secondary polarizing technique. A FOCtek GLP6320 Glan laser air spaced polarizer was purchased from Delta Photonics to polarize the beam, a non-trivial task due to the high photon energy at the comparatively short wavelength of 308 nm. The Glan polarizer utilised the effect of Brewster’s angle, the result of which is the selective reflection/transmission of the respective E vector components of the beam. A result of using the polarizer was a greater than 50% reduction in beam energy due to the ‘rejected’ laser light of the 308 opposing E vector and losses as a result of transmission through the polarizer itself. It was therefore critical that the polarizer was operated with maximum efficiency. Firstly, polarization by Brewster’s angle is sensitive to the incident angle of the beam being polarized and therefore it was essential that the beam be suitably aligned to the polarizer in order for the transmitted beam to contain as little of the opposing polarity as possible. Secondly, the resulting beam needed to be suitably oriented so that it was orthogonal to optical collection axis in order to maximise the Raman scattering with respect to fluorescence/spuriously scattered light. With these two factors in mind it was essential to perform the following actions when setting up and orienting the polarizer. Beam orientation to the polarizer and spectrometer: A technique had to be developed to determine the orientation of the beam. In this instance the beam was adjusted so that it was perfectly horizontal, where horizontal is taken to be a common datum, and that the optical axis was also sought to be horizontal, thus ensuring the beam was at the very least aligned in the horizontal plane, xz, Figure G.1. 309 Figure G.1 Diagram of the laser beam polarization, shaping and steering Alignment in the second plane, xy, was somewhat more difficult. For this a physical datum reference beam was erected which was aligned with the optical axis of the spectrometer and collection optics in the xy plane. This datum beam was situated adjacent to the beam path as it exited the periscope and served as a reference with which to align both the beam and the polarizer (the polarizer having already been horizontally aligned). Polarizer orientation: With the polarizer aligned in both the xz and xy (parallel to the spectrometer x axis) planes it was necessary to orient the polarizer, specifically the Side View Plan View Beam Shaping Schematic y x z Periscope y x z End View y x z Polarizer beam mask Polarizer Shaping optic 45 °Mirror Slit mask Anode surface Polarizer beam mask Slit mask Optical platform traverse Beam waist 310 crystal/air/crystal interface with the xz plane. The polarisers light paths were made up of light in, transmitted light out and reflected light out, Figure G.2. Figure G.2 GLP6320 Glan laser polarizer showing light in/out ports The orientation of the polarizer about the x-axis needed to be accurate and repeatable. By ensuring the light transmitted through the polarizer had the E vector oriented in the xy plane, maximum Raman signal could be produced relative to background noise/fluorescence. In this instance this was achieved by using optical targets with 0.5 degree increments situated several meters away (laboratory wall/roof) from the polarizer. Using the reflected light out the polarizer as a guide, the polarizer could be easily orientated by adjusting the rotation of the polarizer about the x-axis until the desired orientation was obtained. y x z y x z y x z Light in Polarized light out Crystal/air interface ° ° ° ° Orientation about x-axis Unused, opposing polarity, light out 311 G.2 Image Setup The experimental matrix was such that each day of data collection was at a different physical position in the channel for a given operational temperature. With each change in channel position it was necessary to move both the light sheet optical platform, Figure G.1, and the collection optic. In turn, the spectrometer slit masks also required adjustment to account for any variation in channel height between physical positions; that is to say that the plane of the collection optics and the reforming channel were not completely aligned. As mentioned in the description of the experimental apparatus §6.2 the light sheet was reduced in size through the use of a number of masks. The resulting (relatively thin) light sheet must therefore be aligned with the collection volume as defined by the collection optics so that the volume is optimally filled, Figure G.3. Figure G.3 The collection volume, as defined by the relative positions of the light collection optics and the spectrometer slit being filled by the light sheet Collection volume Spectrometer aperture x z Light sheet Collection optic Focusing optic 312 The position of the light sheet could be adjusted in the x and z directions relative to the collection volume, as defined in Figure G.3. The optimisation of the light sheet position was determined in series of set-up tests, performed at room temperature and using air as the ‘target’ gas. For the tests the positions of the optics were fixed while the position of the light sheet was moved through a grid pattern in the approximate location of the collection volume as dictated by the relative focal lengths of the collection optics. The initial grid pattern consisted of 6 positions in the x direction (at a spacing of 5.0 mm) and 5 positions in the y direction (at a spacing of 1.0 mm). A spectral image at each of the grid points was collected by the integration of 300 laser pulses and a central portion of the image was used as a comparison so as not to include spurious light scattered from the anode or window surfaces. The result was a series of N2 line counts that ranged from close to ‘zero’ (the background noise present in the spectrum) to line counts representing a significant signal. By plotting these values in a three dimensional ‘surface’ plot the approximate position of the collection volume could be seen as the ‘peak’ of the surface, where the Raman scattered light was collected most efficiently by the collection optics. It should however be noted that the N2 Raman signal is not necessarily ‘zero’ if the light sheet does not fall inside the collection volume shown above, as Raman scattered light will still be collected when the light sheet falls either in front of or behind the collection volume. However, while this light is still collected, it will be outside the depth of field, so in photographic terms ‘out of focus’, Figure G.4. In addition, the waist of the collection volume was the point at which miss alignment of the light sheet with the collection volume in the y direction became most apparent. This is because light would still be collected at increasing distances away from the optical axis centre line at positions in front of and behind the collection volume. Proximity to the optimum light sheet position was therefore characterized by the rapid reduction in signal strength when 313 moving away from the optical axis centre line, as these were the only regions where Raman signals could not be significantly collected. Figure G.4 Regions of signal collection With the combination of the collection optics (collection optic: focal length 300 mm, diameter 50 mm, focusing optic: focal length 200 mm, diameter 50 mm) and the spectrometer slit width, 0.75 mm, the collection volume had the dimensions of a diamond of length 9 mm, width 1.125 mm, projected through the height of the channel Figure G.5. The light sheet was a symmetrical trapezoid, 6 mm in height with widths of 1.0 and 0.6 mm, respectively, projected through 1.5 mm. Because of these relative dimensions, the light sheet ‘over filled’ the collection volume in the z-direction, but substantially ‘under filled’ the collection volume in the x direction, Figure G.5. The position of the light sheet with respect to the collection volume was therefore somewhat insensitive to small variations in both the x and z directions as close to optimum filling of the collection volume would still occur. Spatial resolution is therefore primarily a function of light sheet position. Spectrometer aperture x z No signal collected ‘Out of focus’ signal collected Optimum signal collection 314 Figure G.5 Light sheet and collection volume dimensions. Once a rough position of the collection volume was determined a finer grid was used to more accurately locate the collection volume by using a 5
UBC Theses and Dissertations
The development of optical measurement techniques for gas species and surface temperature on a planar… Saunders, James Edward Appleby 2011
Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 24-ubc_2011_spring_saunders_james.pdf [ 2.76MB ]
- JSON: 24-1.0071674.json
- JSON-LD: 24-1.0071674-ld.json
- RDF/XML (Pretty): 24-1.0071674-rdf.xml
- RDF/JSON: 24-1.0071674-rdf.json
- Turtle: 24-1.0071674-turtle.txt
- N-Triples: 24-1.0071674-rdf-ntriples.txt
- Original Record: 24-1.0071674-source.json
- Full Text