UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Development of a high-concentration hydrogen sensor and a gas sensor test bench Fan, Lilian Lai Yee 2007

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

Item Metadata

Download

Media
831-ubc_2007-0090a.pdf [ 20.53MB ]
Metadata
JSON: 831-1.0074519.json
JSON-LD: 831-1.0074519-ld.json
RDF/XML (Pretty): 831-1.0074519-rdf.xml
RDF/JSON: 831-1.0074519-rdf.json
Turtle: 831-1.0074519-turtle.txt
N-Triples: 831-1.0074519-rdf-ntriples.txt
Original Record: 831-1.0074519-source.json
Full Text
831-1.0074519-fulltext.txt
Citation
831-1.0074519.ris

Full Text

DEVELOPMENT OF A HIGH-CONCENTRATION HYDROGEN SENSOR AND A GAS SENSOR TEST BENCH  by  Lilian Lai Yee Fan B.A.Sc, The University of British Columbia, 2002  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE  in  THE FACULTY OF GRADUATE STUDIES  (Engineering Physics)  THE UNIVERSITY OF BRITISH COLUMBIA April 2007  © Lilian Lai Yee Fan, 2007  Abstract Resistive hydrogen gas sensors based on palladium-hydrogen (Pd-H ) interactions were fabricated and tested. 2  This thesis presents expected resistance characteristics for these sensors, and describes their fabrication process. Test results and analysis identified the H sensing mechanisms in these sensors. In particular, 2  linearity between sensor resistance and the approximate atomic hydrogen content in the Pd was established, and the estimated proportionality constant obtained is 2.7± 0.9 (a.u.). A low flow rate sensor test bench featuring a line-switching mechanism was designed and built also. A description of the test bench and its operation principles are discussed. Response times obtained from initial tests of the un-optimized system is 16s ± 9s, comparable to the system response time of a commercial gas sensor test station presently available. Recommendations for future work to optimize this test bench are discussed, as are recommendations for future sensor design and development.  ii  (  J  Table of Contents Abstract Table of Contents  :..  iii  List of Tables  vi  List of Figures  vii  Acknowledgements  -.  xii  Dedication Chapter 1  x  Introduction  1  1.1  Motivation and objectives  1  1.2  Applications  1  1.2.a  Manufacturing  2  1.2.b  Medicine  2  1.2.c  Environmental  3  1.2. d  Fuel regulation and monitoring  3  1.3  Recent research  3  1.3. a  Resistive sensors....  1.3.b  Semiconductor sensors  5  1.3.c  Other sensor structures  7  ;  3  1.3.C.1  Thermoelectric sensors  7  1.3.C.2  Optical sensors  8  1.3.C.3  Cantilever sensors  8  1.3.C.4  Pyroelectric sensors  9  Surface acoustic wave sensors  9  »1.3.C.5 1.4 Chapter 2  'i'  Sensor development  9  Theory and Simulation of Pd Mesowire Resistors  11  2.1  Resistive mesowire sensors  11  2.2  Basic sensor structure  11  2.3  Electrical resistance of Pd hydrides  12  2.3.a  Pressure-concentration (p-C) isotherms for the Pd-H system  12  2.3.b  Resistivity vs. [H]/[Pd]  14  2.3.c  Supersaturation  2.3.d  Magnitude of changes in resistance  2  ;  :  14 15  2.4  Percolation mechanism  15  2.5  Temperature effects  16 iii  2.5.a  Temperature coefficient of resistance in Pd hydrides  16  2.5.b  Pressure-concentration changes with temperature  16  2.5.c  Other temperature effects  IV  2.5.d  Temperature modelling  2.6 Chapter 3  Anticipated observations  3.1  Fabricated sensors  3.2  Process description  23 23 '.  26  Substrate preparation  26  3.2.b  Sensor patterning  29  3.2.c  Developing the resist  31  3.2.d  Sputtering and liftoff.  3.2.e  Packaging  :  39 ••  -45  Sensor testing  45  Development of a High Concentration H Test Bench  ...46  2  4.1  Application requirements  46  4.2  Description of the facility  46  Gas manifold  46  4.2.a. 1 Supply to MFC controllers  48  4.2.a.2  MFCs to LINEOUT:  49  4.2.a.3  Purge line and safety alarms  50  4.2.b  Testchamber  4.2.b. 1 Vacuum system  50 '.  •  (  51  4.2.b.2  Sensor mount  51  4.2.b.3  Operation of the gas manifold and test chamber  54  4.2. c  Data acquisition electronics  54  4.2.c.l  Sensor excitation and signal detection circuitry....  55  4.2.C.2  LabVIEW data logging program  55  Commissioning and testing with the test bench  57  4.3 4.3. a  Test bench commissioning  57  4.3.b  Testing with the test bench  57  Results and Analysis 5.1  r-  22  3.2.a  4.2.a  Chapter 5  •  Fabrication of Pd Mesowire Resistors  3.3 Chapter 4  17  •  61  Test results  5.1.a  Early sensors  5.1.b  Failure mechanisms  61 •  61 66 iv  5.1. c 5.2  Chapter 6  Sample response curves  68  Response analysis  73  5.2. a  Small signal step response  73  5.2.b  Background response  78  5.2.b. 1 Lattice and grain boundary effects  79  5.2.b.2  Percolation effects  79  5.2.b.3  Auxiliary effects  81  Conclusion  -82  6.1  Test bench development summary  82  6.2  Sensor development summary  83  6.3  Significance of results and comparison with literature  83  6.4  Additional recommendations  85  References  •  Appendices ..r.  87 .....95  Appendix A.  LabVIEW control panels and block diagrams  Appendix B.  Calibration charts for gas sensor test bench  95 100  v  List of Tables Table 1.1  Some resistive sensor characteristics  5  Table 1.2  Some semiconductor sensor characteristics  6  Table 3.1  JEOL 840 SEM Settings  29  Table 3.2  NPGS e-beam lithography program settings  30  Table 3.3  Equivalent line and area dose from NPGS program  30  Table 3.4  Gaussian best-fit parameters for Pd wire cross-section  43  Table 4.1  Mass flow controller characteristics  49  Table 4.2  Operation of line-selecting solenoid valves  49  Table 4.3  Test chamber physical data  51  Table 4.4  Average response times for commercial test station and test bench  59  Table 5.1  Reported (p-C) isotherm values of fi Pd hydrides from literature  .•.  73  Table 5.2  Summary of pressure interpolation constants for determining [H]/[Pd] ratios  73  Table 5.3  Average values for fit coefficients  77  Table B. 1 Conversion factors for MFCs  100  vi  List of Figures Figure 2.1  Top and cross-section view of basic sensor structure. Sensors developed in this study are variations of this geometry  Figure 2.2  11  Shape of the pressure-concentration (p-C) isotherm of hydrogen in Pd. The first rise corresponds to the a-hydride phase and follows a linear form whereas the second sharp rise is due to the P-hydride phase and follows a logarithmic form. The plateau corresponds to the ct-P transition region and shows the effect of hysteresis  Figure 2.3  13  R/Ro behaviour on temperature as reported by Sakamoto et al. Changes in plateau resistance are not monotonic with temperature  17  Figure 2.4 Heat Transport model of Pd mesowires with inset (not to scale) showing model structure. The Pd wire is mirrored across the boundary to the right, and is half the length of the simulation by symmetry Figure 2.5  18  Estimated temperature along Pd wire at various voltages. Modelled properties include conduction, convection and thermal effects on resistance. Thermal coefficient of 3.3 x 10" 3  /K is used  Figure 2.6 Estimated temperature along Pd wire at voltages varying from 20mVDC to 120mVDC  19 20  Figure 2.7 Model structure of Pd mesowire with a delaminated section (not to scale) for heat transport simulations  20  Figure 2.8 Temperature simulation of a delaminated Pd wire (over a 12.5|rm air gap). The structure of this simulation was illustrated in Figure 2.7. This figure clearly demonstrates the deleterious effect of delamination on sensor performance  21  Figure 2.9 Temperature simulation of a delaminated Pd wire (over a 12.5um air gap) for voltages between 0.02VDC and 0.12VDC Figure 3.1  21  Lithography and liftoff process: A) Coating with resist and patterning with electron beam; B) developing exposed resist; C) depositing the desired material for the pattern and D) removing the resist layer and leaving behind the patterned layer  23  Figure 3.2 Typical early design sensor showing many-wire Pd array across Au/Cr contacts ^  approximately 65pm apart  24  Figure 3.3 Later design sensor with five-wire Pd array. These sensors have improved liftoff yield and higher overall resistance than many-wire arrays  24  Figure 3.4 Sensor with U-shaped symmetric structure. Instead of patterning gaps along a wire by turning off the e-beam, these samples are patterned by having the e-beam travel half of the gap and then returning to the same side. A U shape was used on each side (symmetrically)  vii  to lessen overexposure of the resist at the turning points. Attempts at generating gaps in this manner were unsuccessful  25  Figure 3.5 Five-wire device sensor with crosslinked polymer platform. The presence of imperfections along these wires have less effects on the overall resistance than in other sensors developed  25  Figure 3.6 Au contact pads of a sample, shown with bonded Au wires. Crosses are alignment marks to ease navigation of electron beam during sample exposure  26  Figure 3.7 Mask for photolithography of Au/Cr contact pads for fabricated sensors. Pad sizes are ~lmm x 1mm. The patterns are rotated to align with the cleave planes of Si(l 11) wafers. ...27 Figure 3.8 Bilayer resist cross-section on SiC>2. The upper layer is PMMA, the middle layer isPMGI and the bottom is Si0 . Thicknesses of the resist layers are about 200nm each  28  2  Figure 3.9 Partial Liftoff of Pd from e-beam exposure of PMMA. This shows PMMA working as a positive resist at moderate electron-beam dose  32  Figure,3.10 Effects of overdeveloping the PMGI undercut layer. The PMMA layer becomes entirely delaminated from the substrate  33  Figure 3.11 PMMA development: IminOOsec, 1:3 MIBK:IPA. Very little development is observed at this point  33  Figure3.12 PMMA development: 2min00sec, 1:3 MD3K:IPA. Faint traces corresponding to developed lines begin to appear  34  Figure 3.13 PMMA development: 3min00sec, 1:3 MIBK:IPA  34  Figure 3.14 PMMA development: 4min00sec, 1:3 MIBK:IPA. Patterns in PMMA are now easily distinguishable  ......34  Figure3.15 PMMA development: 5min06sec, 1:3 MIBK:IPA. End of the MIBK development process. 3 5 Figure 3.16 PMGI development: 0min20sec 2.2% TMAH. Patterns begin to broaden  35  Figure 3.17 PMGI development: 0min40sec 2.2% TMAH. Cavities under the PMMA layer (dissolved PMGI) begin to appear  35  Figure 3.18 PMGI development: 1 minOOsec 2.2% TMAH. Mild delamination (circled) of the PMMA begins to appear  36  Figure 3.19 PMGI development: lmin20sec 2.2% TMAH. Further widening of PMGI cavity  36  Figure 3.20 Micrograph showing top view of sample cleaved in the center of the device. Delamination and debris from sample cross-sectioning are clearly visible  37  Figure 3.21 PMMA Resist peeling off is clearly visible from sample cross-sectioned area  37  Figure 3.22 Cross-section of developed PMMA/PMGI bilayer. The layer to the left is PMMA, which can be seen with a large flap/extension left behind from the cleaving process. The layer with the cavity is the PMGI layer, sandwiched between the PMMA and the Si0  2  38 viii  Figure 3.23 Chart showing the width of the developed PMGI cavities as a function of electron dosage. Only a few data points are available because of difficulties in obtaining intact cross- sections of polymer layers  38  Figure 3.24 Pd wires after liftoff from single layer resist  39  Figure 3.25 Pd wire after liftoff from bilayer resist. Pd wires fabricated from the bilayer process are thinner than with the single layer PMMA process  40  Figure 3.26 Secondary Electrons image of a thin Pd wire sputtered through bilayer resist  41  Figure 3.27 Electron counts associated with Si obtained from EDX measurements. Slightly lower Si count density is observed across the lower half of the image corresponding to the location ofthePdwire  '.  41  Figure 3.28 Electron counts associated with Pd obtained from EDX measurements. Pd is present as a very faint bar of higher count density across the lower half of the image! These measurements prompted further characterization of the bilayer samples with a field emission scanning electron microscope (FESEM) and an atomic force microscope (AFM), which determined unambiguously the presence of the Pd wires as a continuous thin film  41  Figure 3.29 FESEM micrograph of a Pd wire on oxide fabricated with single-layer resist  42  Figure 3.30 FESEM micrograph of a Pd wire on oxide fabricated with bilayer resist  42  Figure 3.31 Cross-section of Pd mesowire sputtered through bilayer resist. Fitted line follows the Gaussian form given by Wo+W^expHCx-W^/Ws) ) 2  :  43  Figure 3.32 Top view of Pd wire deposited through bilayer resist. Sample was exposed with l.OnC/cm electron beam dose, and subsequently developed in 1:3 MIBKTPA and MF-319 for 5m06s and lm20s respectively  44  Figure 3.33 Sputtered Pd wire widths with increasing e-beam dosage. Widths are normalized with the measured width of wires exposed with 4.0nC/cm electron dose  44  Figure 3.34 Packaged, bonded chips for sensor testing  45  Figure 4.1  47  Gas manifold and test chamber schematic for the test bench  Figure 4.2 Test chamber showing t est sensors, pressure transducers and vacuum pumps mounted on numerous ports Figure 4.3  Diagram of slit nozzle. Gas entering through the port at one side of the nozzle is distributed across a slit orifice close to the sample  Figure 4.4  50  52  Simulation model structure for slit nozzle. Gas enters through the circular inlet on the left and exits to the right through a slit. Samples would be located at the bottom surface (extrusion) of the model  53  Figure 4.5 Vertical section of simulated gas flow through a slit nozzle. Flow velocity is uniform across much of the sensor cavity. Sensors are attached at the centre of the bottom surface. ..53 ix  Figure 4.6 Horizontal section of simulated pressure in a slit nozzle. Pressure appears to be uniform across the sensor cavity. Sensors are attached at the centre of the bottom surface  53  Figure 4.7 Excitation and DAQ interface circuit to power resistive sensors and measure changes in  Figure 4.8  resistance  55  Sample measurement obtained from LabVIEW-controlled DAQ setup  56  Figure 4.9 Sample response of a wire sensor fabricated with bilayer resist at very high concentrations ofH  58  2  Figure 4.10 Sample response of a capacitor sensor tested with the test bench. Response of these sensors are compared with response of the same sensors tested with a commercial test station  ;  59  Figure 4.11 Sample response of the same sample as in Figure 4.10, tested in the Greenlight commercial sensor test station Figure 5.1  59  Resistance response from an intact array of many Pd wires. A very slow and gradual response was observed, in addition to a large hysteresis  62  Figure 5.2 Voltage across defect-ridden Pd wire array. Response is indicative of percolation effects. Wires that are open circuits in air become conductive at 100% H  62  2  Figure 5.3 Intact wire array sample corresponding to test results shown in Figure 5.1  63  Figure 5.4 Wire array with multiple defects corresponding to test results shown in Figure 5.2  63  Figure 5.5 Unstable response of the defect-ridden many-wire array of Figure 5.4. 0-100% H Step 2  response of this sensor was shown in Figure 5.2  64  Figure 5.6 Failure response of the defect-ridden sensor of Figure 5.4. This test was performed immediately after the sensor test shown in Figure 5.5  65  Figure 5.7  Small signal dependence of resistance on H concentration: very small changes  65  Figure 5.8  Sample with imperfections on the underlying oxide layer, as fabricated. The compromised ^  2  region is the darker patch of oxide between the contact pads, to the right  67  Figure 5.9 Failure of the sample of Figure 5.8. Portions of wires spanning oxide with imperfections can be seen to have melted. The adhesion between the wires and the oxide layer at these areas was compromised, which led to wire delamination and overheating  67  Figure 5.10 I-V curve for Pd wires with good adhesion to the oxide substrate. Trend line is a linear regression of I=0.008*V based on data up to 6.25VDC  68  Figure 5.11 Response from bilayer PMMA/PMGI resist sample. This sample is one with a platform structure. See text for circled property Figure 5.12 Step response portion of Figure 5.11 in detail. See text for circled properties  69 '.  69  Figure 5.13 Response from sample fabricated with the single PMMA resist process, 5 wire structure. See text for circled property >  70  Figure 5.14 Step response portion of Figure 5.13 in detail  70  Figure 5.15 Repeated testing of a platform sensor  71  Figure 5.16 Step response portion of Figure 5.15 in detail  71  Figure 5.17 Logarithmic fit of small signal resistance measured as a function of H concentration  74  Figure 5.18 Sample response along with small signal component for H concentrations over 10%  74  Figure 5.19 Sample background response left over after small signal component is removed  75  2  2  Figure 5.20 Reference resistances for samples analysed with respect to their initial resistance. Samples marked with t before their labels did not liftoff properly. Sample marked with * was an older, many-wire sensor with compromised signal-to-noise. These sensors (the three on the right) were not used to calculate averages  76  Figure 5.21 Coefficients (arbitrary normalized units) extracted from small signal response fitting. Again, samples marked with t before their labels did not liftoff properly. Sample marked with * was an older, many-wire sensor with compromised signal-to-nbise. These sensors (the three on the right) were not used to calculate averages.  .77  Figure 5.22 Percolation-like effects in resistance response. Change in resistance due to (a) expansion of Pd wire segments, forming an additional conductive path and (b) contraction of Pd segments  80  Figure A.l Main control program front panel (MultiChannelDAQ.vi) Figure A.2 Block diagram of main control program (MultiChannelDAQ.vi)  95 :  96  Figure A.3 Time reference case in main control program (MultiChannelDAQ.vi)  97  Figure A.4 Case for no plot data in main control program (MultiChannelDAQ.vi)  97  Figure A.5 Front Panel of module for parsing and extracting data to be logged (Array2DExtract01.vi)..97 Figure A.6 Block Diagram of module for parsing and extracting data to be logged (Array2DExtractO 1 .vi)  98  Figure A.7 Front Panel for plotting module. Plots scrolling data (MultiPlot2DArray.vi)  98  Figure A.8 Block Diagram for plotting module (MultiPlot2DArray.vi)  99  Figure A.9 Case for no data to plot in plotting module (MultiPlot2DArray.vi)  99  Figure A.lOCase for only one point to plot in plotting module (MultiPlot2DArray.vi)  99  Figure B.l Cross reference chart for N MFCs, received new from manufacturer (ASGE)  101  Figure B.2 Real flow conversion chart for an older H MFC  101  Figure B.3 Actual flow of older H MFC as the console setpoints vary  102  2  2  2  Figure B.4 Cross Reference curve for Cole Parmer MFC with ASGE MFC. Discrepancies of about 2slpm were observed between the two MFCs  102  xi  Acknowledgements I would like to express my thanks to the people who have made this thesis work possible. In particular, I would like to thank Dr. Jonathan Wu, now at the University of Windsor, for giving me the opportunity, support and encouragement to pursue this Master's degree. I thank Dr. Kevin Stanley, Dr. Cheng Hu, Dr. Eva Czyzewska, Tom Vanderhoek, and Dr. Yanghua Tang for their friendship, advice, and invaluable contributions to the thesis. Many thanks to Weimin Qian and Shuozhi Yang, who helped carry out much of the testing for this work.  I offer my gratitude to Dr. Andras Pattantyus-Abraham, Dr. Mario Beaudoin, John Yuen and my family for their unfailing support to solve all kinds of issues I have, whether in the cleanroom, laboratory, cubicle or at home.  •  To my supervisory committee and reviewers also I owe many thanks, for their willingness to read and provide useful suggestions on my thesis within such a tight schedule. Thank you Dr. Tiedje, Dr. Wu, Mario, Jun and Cheng!  Finally, I would like to thank the National Research Council of Canada - Institute for Fuel Cell Innovation for their support. It has indeed been a privilege to have the opportunity to work with so many admirable researchers at such an open and creative organization.  xii  To my parents.  xiii  Chapter 1  Introduction  '  With the commercialization of fuel cell and hydrogen technologies, there is a pressing need for small, fast, economical and efficient hydrogen (H ) monitoring devices. Interest in H sensors arises not only from safety 2  2  concerns, but also from their potential for monitoring and optimizing the operation of fuel system components. Numerous technology gaps exist, however, and development of sensors suitable for commercial applications such as embedded fuel cell and balance-of-plant* monitors remains a challenge that a number of research groups around the world are trying to address.  1.1 Motivation and objectives The goal of the present work is to develop a micro-sensor capable of detecting H concentration variations 2  between 90% and 100%) in nitrogen (N ) for on-line monitoring of fuel cell systems. The sensor should be 2  fast, with response and recovery times preferably below one second. The sensor's response should also be relatively stable and reliable. Its fabrication process should require little or no manual assembly to facilitate transfer of developed technologies to industry. To this end, resistive palladium (Pd) mesowire sensors that operate between 10% and 100% H have been successfully fabricated and tested. A test bench was also #  2  designed and assembled for H sensor testing with the objective of determining detailed time response of the 2  developed sensors. The time response characterization is not yet accomplished, but initial test bench calibration results have been obtained and are promising.  While the direct motivation for this work is fuelled by the emerging fuel cell industry, it is important to recognize that H sensors are also used in numerous other industries. The present chapter describes some 2  applications of H sensors and provides an overview of recent research reported in literature. Emphasis will 2  be placed on resistive devices based on Pd-H interactions, as these are most similar in nature to the sensors 2  developed in this work.  1.2 Applications Hydrogen is the lightest element known on earth. It is a flammable gas above 20.35K for concentrations between 4% and 75% in air, and has an auto-ignition temperature of 773K [1]. Atomic hydrogen (H) is a  * Balance-of-plant refers auxiliary systems needed for fuel cell operation, such as storage tanks, compressors and control electronics.  constituent of many common substances, and hydrogen gas (H ) is frequently used in a wide array of 2  industries and research.  1.2.a Manufacturing Many industrial processes use H as a raw material, while numerous others produce H as end products. Of 2  2  these, only a small portion is sold commercially, as most H requirements for industry are typically produced 2  and consumed on-site. H sensors are thus present in many facilities to monitor the safety and optimize the 2  efficiency of these industrial processes. H is primarily produced by reforming natural gas. The two largest chemical industries at present that use H 2  2  are the ammonia and petroleum refining (plastics) industries [2]. Other industrial users of H include 2  chemical plants that produce methanol, hydrogen peroxide, acids and metals.  A different niche manufacturing application for H sensors is its use as a tracer gas for locating leaks and 2  cracks in objects ranging from automotive exhaust pipes to food packaging. Average background levels of H  2  are typically on the order of 0.5ppm. As such, tracer gas mixtures of H in nitrogen (N ) may be easily 2  2  distinguished and located [3].  The food industries, in particular, use H sensors in various ways. Many food processing facilities modify 2  hydrogenation levels in fats for attaining desired baking/cooking results. H sensors are reportedly used for 2  monitoring cooking progress, and are used in studies for rapid identification of prior irradiation in certain foods [4, 5]. J  1.2.b Medicine H sensors have also found their way into medical laboratories, where "Hydrogen breath tests" are used for 2  diagnosis of gastrointestinal symptoms. Sugars and carbohydrates are usually digested and absorbed in the stomach and small intestines, before the remaining foodstuff reaches the colon. When these sugars are not absorbed properly, large amounts of these sugars may become accessible^to certain anaerobic bacteria in the colon. These bacteria will digest them, releasing H as one of the by-products. Some of the H is absorbed 2  2  into the blood stream and taken to the lungs where the gas is exhaled. Identification of specific medical disorders is done through analyzing the timing, frequency and amount of H exhaled. Examples include tests to identify patient intolerance to particular sugars (such as lactose) [6]. At 2  2  present, breath samples are generally trapped and analyzed with gas chromatography where H -concentration 2  changes down to ppm levels are monitored [7].  1.2.c Environmental H sensors are also used in atmospheric and environmental studies. Mass spectrometry is used to determine 2  hydrogen and deuterium concentrations in the earth's lower atmospheres, shedding light on studies of natural H uptake in soil [8]. Other uses of H sensors in environmental studies include their use in space missions to 2  2  provide key information on the presence and volume of water or ice on planetary bodies and are aiding scientists around the world study climate changes on the planet [9, 10, 11]  1.2. d Fuel regulation and monitoring More in line with the objectives of this thesis are H sensors in fuel monitoring. H detectors play a crucial 2  2  role in monitoring the safety, operation and fuel quality of H powered technologies such as space shuttles 2  and fuel cells. Environmental concerns over the use of fossil fuels and advancements in fuel cell technologies in recent years lead to ever increasing use of H as an alternative, environmentally conscious means of energy 2  delivery.  Specifically, mounting demands for device miniaturization in the fuel cell industries [12] is driving the development of tiny, novel sensors capable of fast and accurate monitoring of H environments. The majority 2  of H sensor research in literature are interested in developing sensors for fuel cell applications. Leak sensors 2  optimized for low concentration detection are needed for safety monitoring, while embedded sensors that distinguish a wide range of concentrations are needed for performance monitoring. Successful commercialization of H fuel cells depends on the appropriate use of sensing devices in related areas such as 2  fuel storage, fuel,processing (reforming) and transportation technologies.  The performance of commercially available H sensors follows developments in H sensor research closely. 2  2  The following section surveys of some recent experimental studies in the field.  1.3 Recent research 1.3. a Resistive sensors Resistive measurements of thick and thin-film semiconducting materials is a common method to determine gas concentrations. Longstanding research on semiconducting materials such as Sn0 [13, 14, 15, 16] 2  3  continue to bring progress in this area, as does recent work with materials such as Ti0 [17, 18] and W 0 2  3  [19]. Emphasis of these research endeavours is placed (for the most part) on detecting minute concentrations of H . Oxide resistive sensors operate on the basis of an increase in excess electrons from absorbed H 2  2  molecules, thus lowering the overall resistivity of the material. These sensors typically have simple structures, but suffer from poor selectivity and high operating temperature requirements.  Documented resistance changes in Pd-H systems predate semiconducting oxide sensors research by almost a 2  century, with an excellent reference summarizing studies on H interactions with bulk Pd written back in 1967 2  by F.A. Lewis [20]. By far the majority of work on resistive H sensors to this day use Pd-based materials as 2  the active sensing element. Pd as a sensor material is used as catalyst powders (thermoelectric sensors), thin films, gate films (in semiconductor sensors), nanowires and nanoclusters, among others. Some of these sensors will be described briefly in the following paragraphs. A description of the sensing mechanisms present in Pd-based resistive sensors will be presented in the following chapter.  Stability and degradation has always been a prime concern for Pd-based H sensors. Pd-Si devices are known 2  to degrade with the formation of palladium silicides [21]. Blister formation and delamination routinely occur in Pd thin film devices exposed to high concentrations of H [22]. Methods of stabilizing Pdfilmsinclude 2  using an adhesion layer between the Pd and the substrate, or alloying Pd with impurities [23, 24, 41].  In general, the introduction of alloy impurities in Pd increases in the upper concentration-sensitivity limit of the device, and improves the mechanical stability of the Pd. These changes are typically at the expense of reduced sensitivity at lower concentrations and, more importantly, slower time response [40].  In the past decade, a few researchers have investigated Pd-based resistive sensors operating at concentrations up to 100% H . 2  Examples include sensors featuring Pd-Ni alloy films evaporated on Si N substrates [25]. 3  4  These Pd-Ni sensors feature good mechanical stability at Ni contents over 8%, and exhibit reasonable dynamic response (film resistance) at high H concentrations. Comparable results from Pd-Ni alloy films 2  were obtained on A1203 substrates in a different study [26].  Thick-film resistive pastes were also studied as H sensing materials. This class of sensors include thick film 2  sintered glass binders with Pd or Pd-Ag alloy [27, 28, 29]. The response behaviour of these and the Pd-Ni film sensors mentioned earlier are similar, showing an increase in resistance upon exposure to H . Response 2  time however for these thick-film sensors are longer, with reports typically on the order of one to several minutes.  4  Interesting, novel sensing mechanisms are emerging as the use of micro and nanofabrication technologies become increasingly common. One such example is a sensor structure that operates based on the swelling and contraction of Pd clusters or grains. Palladium Mesowire Arrays (PMA) [30, 31] are electrodeposited Pd wires between 50-600nm in width. 'Gaps' along these wires close due to Pd lattice expansion in the presence of H , leading to a large change in the overall resistance of the wires. Sensors based on this activation 2  mechanism were tested with H concentrations between 0% and 10% in air. While signal saturation occurred 2  as the H concentration approached 5%, the sensor enjoys excellent response and recovery times from tens of 2  milliseconds up to several seconds.  Similar sensing mechanism was observed in other nanoscale Pd sensors, including Pd nanowire bundles electrodeposited into a sacrificial porous alumina host [32], discontinuous films of Pd clusters close to their percolation limit [33], and Pd nanowires electrodeposited onto patterned Si microchannels [34, 35].  Other resistive sensors in recent developments include Pd-coated rare-earth thin films supported on a microhotplate [36] and Pd-coated polysilicon mesoswires [37]. A number of resistive sensor structures are listed in Table 1.1 with representative characteristics.  Table 1.1 Sensor  Material/Structure  Type  Some resistive sensor characteristics Ref.  Response/  Operation  Concentration  Recovery  Temperature  range  time(s)  (K)  0.31%  100s  473-873K  [-14]  H  2  Oxide  Sn0  Oxide  Ti0 thermal  0.001%- 1%  2-10s  423-573K  [18]  Oxide  W03, Pt-W03, Au-  0.125%-1.25%  200s  300-573K  [19]  < 100%,  20s/50s  383K  [25]  2  2  W03 Metal  Pd/Ni-Ti-Si3N4  increasing R Metal  Pd mesowires  0.5%-10%  20ms-5s  300K  [30]  Metal  Pd/rare earth/hotplate  0.02% - 30%  < ls-200s  N/S  [36]  1.3.b Semiconductor sensors Semiconductor sensors encompass a large portion of H sensor research. Included within this class of gas 2  sensors are Schottky metal-semiconductor (MS) sensors, metal-insulator-semiconductor (MIS) and metalinsulator-metal (MIM) tunneling devices [23]. Pd-gate MOS devices were first reported by Lundstrom et al. 5  [38, 39]. In general, adsorption of H changes the work function of the gas-sensitive metal and causes shifts 2  in the barrier height and flat-band voltages for semiconductor devices. In the case where Pd is used as the active metal, further changes are typically observed from atomic hydrogen migration to the Pd/dielectric interface, which causes charge redistribution [40].  The majority of state-of-the-art research into semiconductor sensors focus on lowering the H detection limit 2  and minimizing the response and recovery times at low concentrations. As such, only a handful of recent research in this category of H sensors provided experimental results for H concentrations over 1%, among 2  2  which even less reported results close to 100% [41]. A list of some semiconductor sensors are given in Table 1.2 to illustrate general operating conditions and properties for these sensors.  Table 1.2  Some semiconductor sensor characteristics  Material/Structure  Sensor Type  Pd/Ni capacitors  MOS  Pd-InGaAs-AlGaAs  MOS  Ref.  Response/  Operation  Concentrations  Recovery  Temperature  range  time (s)  (K)  0.01 torr -  Not  300K  [41]  700 torr  Specified  0.0014%-  22-400s  300-460K  [42]  0.001%-10%  2s-1500s  305K-393K  [43]  H  2  T  0.997%  HEMT Pt-commercial Ta205 gate  FET  FET Schottky  Pd-Nb205  0.001%-0.8%  5s-500s  300-473K  [44]  Schottky  Ru-SiC  <2%  12s/ 150s  673K  [45]  pHEMT  Pd-AlGaAs-InGaAs-GaAs  < 0.4%  25-600s  300K  [46]  Pt-TaSix-Si0 -SiC  1%  <10ms/  673-1073K  [47]  573-673K  [55]  1  MIS  2  <150ms p ZnO-n ZnO  P-N  0.05%, 0.1%  146s/86s  Semiconductor H sensors are very sensitive at low concentrations, though this sensitivity typically drops off 2  with signal saturation at H concentrations greater than several percent. Response and recovery times of these 2  sensors range from several seconds to hours, and typically improve with higher operating temperatures [47,  Atmospheric pressure is 760 torr acronym for pseudorhorphic H i g h Electron M o b i l i t y transistor  6  48]. Some microsensors described in literature are now designed and built with integrated heaters [65]. It is useful to note however that response times on the order of several hundred milliseconds were reported from the early work of Poteat et al. at 760Torr H at room temperature [49, 50]. As such, long response and 2  recovery times attributed to semiconductor-based sensors may be characteristic only to low concentration detection.  A number of groups developed semiconductor sensors using Ru, Pt, and Ir as the active metal [45, 47, 51,52, 53, 54, 47, 48]. Pt especially is commonly used in recent semiconductor H sensor work. While devices 2  based on these metals are much more stable mechanically than Pd-based sensors, they exhibit less sensitivity to H [40], and are typically responsive to 0 , CO, N as well as numerous other hydrocarbon-based gases. 2  2  2  Other novel materials in recent semiconductor H sensor research include doped ZnO p-n junctions [55] and 2  porous silicon [56, 57], Results from these novel sensors are promising, albeit less refined than results from conventional semiconductor sensors.  1.3.c Other sensor structures Thermoelectric sensors  1.3.C.1  Thermoelectricity, also known as the Seebeck effect, is induced when a temperature gradient between two points in a material generates a potential difference between the two points. Early work using thermoelectric voltages as the basis for H sensing was done with Pt or Pd coated Sn0 pellets [58]. Changes in temperature 2  2  across Sn0 are produced by catalytic H oxidation on the metals' surface. 2  2  In recent years, thermoelectric Hydrogen Sensors (THS) exhibiting good linear response from 0.5ppm to 10% H in synthetic air was reported in the literature [59, 60, 61, 62, 63, 64, 65]. These sensors consist of Pt 2  catalysts deposited on a portion of the surface of thermoelectric Li- and Na- doped NiQ and SiGe films. The exothermic reaction of H dissociation on the Pt surface generates a temperature gradient across a 2  thermoelectric film, which in turn produces a voltage across the sensor. Recent research into miniaturizing the device using MEMS techniques resulted in the integration of the sensor (as a catalyst-loaded membrane) with a micro-hotplate [65]. The rise time and recovery time of these sensors are reported to be on the order of 50s or less. Cross sensitivity exists to a significant extent as the Pt catalyst is sensitive to numerous other gases [66].  7  1.3.C.2  Optical sensors  The majority of optical sensors are Pd-based and measures changes in the metal's optical or mechanical characteristics. Optical sensors feature "remote" detection capability that minimizes the potential for generating electrical sparks in an H2-rich environment. For the same reason, however, these sensors typically have complicated measurement setups (both electrical and optical), and require significant operating power to drive both the excitation and detection circuitry.  Recent research in optical sensors include studies on the transmittance changes of a Pd/PVDF polymer bilayer structure in the presence of H [67]. Photodiode measurements of laser excitation through thin Pd/PVDF 2  films are compared with a reference beam, and transmittance is shown to increase when such films are exposed to H . Measurements were made between 0.2% and 100%) H in N and excellent Pd adhesion and 2  2  2  mechanical stability was reported. Sensor sensitivity, however, tends to drop off as H concentration 2  increases beyond 10% H . Sensor response times are on the order of 50s at 3% H , and studies were done to 2  2  integrate these sensors in a dual photopyroelectric/optical-transmittance H sensor [68]. Transmittance 2  measurements to detect H were also done with Pd/ V O films on glass substrates between 0 and 10% H 2  2  s  2  [69]. Signal saturation occurred at around 6% H and response and recovery times were on the order of 20s 2  were at 4% H in N . 2  2  Other research on optical sensors include work studying surface plasmon resonances in thin film structures deposited at the end of a fibre optic cable [70, 71], plasmon-induced losses along Pd-coated fibres [72, 73, 74, 75], transmittance studies of fibres of various cladding materials [76, 77, 78], reflectivity and transmission measurements of Pd mirrors at the end of optical fibres [79, 80, 81]. Among these studies, a few were tested up to 100%H [80, 76]. 2  1.3.C.3  Cantilever sensors  Some groups have developed micro-cantilever sensors for H detection. These cantilevers are predominantly 2  driven by the physical expansion of Pd metal in the presence of H , and deflections are measured with optics 2  [82, 83, 84, 85] or capacitance [86] typically.  In recent reports, H -sensitive sensors were fabricated using 2  arrays of cantilevers coated with Pd-Ni alloys [86, 87, 88]. These cantilevers operate by detecting variations in the deflection-sensitive capacitance between the cantilever and a stationary baseplate. Test results for these sensors exhibit good sensitivity range for H concentrations up to 90% in N and in air. Typical response 2  2  times ranged from 60s for higher concentrations to hours for smaller concentrations.  8  Optical cantilever sensors include thin glass substrates coated with Pd or Pd-Ag alloy [83]. Changes in the curvature of these sensors were observed in a vacuum chamber at H pressures up to 600Torr between 20°C 2  and 50°C. The response time for the Pd-coated cantilevers at H pressures under 30Torr (4% partial pressure) 2  is approximately 1 minute, while recovery times are typically at least 10 minutes.  1.3.C.4  Pyroelectric sensors -  Pyroelectricity is the potential difference generated by certain anisotropic solids that become spontaneously and temporarily polarized when a temperature change is induced within the material [40]. Thermal waves and interference patterns are induced within PVDF fdms by ac laser excitation. Signal differences are observed and compared between Pd coated fdms and reference electrodes [89, 90]. Standing wave patterns could also be modified / matched by varying the relative intensity of the excitation laser. Variations in the pyroelectric signal intensity due to difference in the optical absorptances of the electrodes were measured upon the sensor's exposure to H . Typical operating results from these sensors saturate at H concentrations greater 2  2  than 1.5% in N , though good signal-to-noise is still discernible down to approximately 200ppm in N . 2  1.3.C.5  2  Surface acoustic wave sensors  Recent research on Surface Acoustic Wave (SAW) H sensors are typically also Pd-based. SAW sensors 2  measure perturbations of a propagating acoustic wave along the surface of a piezoelectric crystal [91, 92]. Perturbations could be in the form of a change in amplitude, phase or frequency. In the presence of H , SAW 2  sensor elements (again, typically Pd) change in density and elasticity, modifying wave propagation velocities along the sensor when compared to reference electrodes. Some SAW sensor structures studied include thin palladium / metal-phthalocyannine (Pd/CuPc and Pd/NiPc) bilayers on a LiNbOs piezoelectric substrate. Changes in the wave frequency at H concentrations between 0.5%. and 4% were reported, with response and 2  recovery times on the order of several hundred seconds.  1.4 Sensor development Despite the great variety of sensors in current state-of-the-art research, it is difficult to identify one single sensor that is superior across all concentrations of H in terms of its selectivity, sensitivity, speed, stability 2  (chemical and physical), complexity, and efficiency. In the present study, resistive Pd structures are fabricated and their response to concentrations of H between 10% and 100% will be discussed. We choose to 2  study these sensors on the basis of the following observations from literature:  9  ^  1. Pd is known to be selective to H gas. Pd is mildly cross-sensitive to a number of other chemical 2  species such as 0 , N , H S and CO. Its response to H , however, is by far the dominant response. 2  2  2  2  2. Although the pre-saturation operating range for the majority of H sensors described in the literature 2  are limited to concentrations less than several percent, it is clear that resistive changes in Pd can be detected across all concentrations of H . 2  3. The response and recovery speed of sensors reviewed range from several seconds to hours. Thickfilm Pd resistors are characteristically slow, as the sensing mechanism is rate-limited by the diffusion of hydrogen atoms into the bulk material. In this respect, Pd-based semiconductor sensors excel as their operation are more dependent on changes in surface.charges. In recent years, however, use of micro and nanofabrication technologies has resulted in faster resistive Pd devices, as device dimensions and diffusion lengths diminishes. 4. There is simplicity in using resistive sensors. In a monitoring and diagnosis system, it is likely that other sensors and electronics are used to provide key information on the system. It is ideal to use simple detection mechanisms that do not require complex electronics and optics to power and extract sensor signals. Complex excitation and detection circuitry are typically expensive in terms of the spatial requirements, power consumption, and manufacturing costs. Optical and pyroelectric sensors, for example, require externally powered devices (lasers, photodiodes) in addition to the usual signal processing circuitry. Many semiconductor sensors require elevated operating temperatures. For devices that will ultimately be embedded in and integrated with portable fuel cell components, it is not practical to consider devices that consume a lot of power. 5.  There are numerous mechanisms in the interaction of Pd and hydrogen atoms that translate to changes in the Pd's electrical resistance. These mechanisms have the potential to be integrated together to compensate for deficiencies in one another, optimizing the overall response.  Given the above considerations, a study of resistive H sensors was performed-. Details of the study are 2  described in the following chapters. A description of the theoretical characteristics of Pd in the presence of H will be presentedfirst(Chapter 1), followed by an account of the processes involved in fabricating Pd 2  mesowire resistors (Chapter 1). A digression from this sensor work follows with a description of a test bench developed for testing fast transients of H gas sensors (Chapter 1). This is followed by test results of sensors 2  developed in this work (Chapter 1), obtained from a commercial gas sensor test station. The results chapter describes observations of changes and failure modes in Pd mesowire resistors upon exposure to H . An 2  analysis of changes in the electrical resistance of these sensors will also be presented. Finally, the significance and relevance of these results, along with recommendations for future work, will be discussed in the concluding chapter (Chapter 6).  10  Chapter 2  Theory and Simulation of Pd Mesowire Resistors  2.1 Resistive mesowire sensors Sensors are devices that respond to environmental variations with changes in their physical properties. As discussed in the introductory chapter, there is a wide variety of materials and structures that are sensitive to the presence of H gas. The present study investigates the theoretical background relevant to the operation of 2  resistive Pd sensors.  2.2 Basic sensor structure The basic structure of sensors we developed consists of thin Pd wires deposited across two contact pads on oxidized Si substrates, as shown diagrammatically in Figure 2.1. The resistance of these Pd wires can be measured by applying a potential difference across the contacts and measuring the corresponding current. A detailed description of the sensors developed in this work will be given at a later chapter devoted to sensor fabrication (Chapter 1).  Top View Au/Cr  Au/Cr t Pd wires  Side View  Pd  Au/Cr contacts •SiO,  Figure 2.1  Top and cross-section view of basic sensor structure. Sensors developed in this study are variations of this geometry.  1 1  2.3  Electrical resistance of Pd hydrides  There is a large quantity of literature on the Pd-H system. It is well known that atomic hydrogen solubility in 2  Pd is very high, and that spontaneous dissociation of H molecules and chemisorption of hydrogen atoms at 2  Pd surfaces occur readily. The absorption of hydrogen atoms into Pd causes physical expansion in the Pd lattice, causing changes .its electronic states and physical characteristics. Two phases of hydrides exists in the Pd-H system. The a-hydride corresponds to low concentrations of hydrogen atoms in Pd, while the P2  hydride corresponds to high atomic hydrogen content and is characterized by large dislocation densities in the Pd host. Resistance measurements of bulk Pd in the presence of H have been measured since the mid2  nineteenth century. Several conventions were used to represent the hydrogen content in Pd, a traditional one being [H]/[Pd] atomic ratios, the concentration of hydrogen atoms in solid (Pd) solution. The electrical resistance of bulk Pd exposed to H is traditionally described as a roughly linear function of atomic content. 2  The slopes corresponding to different phases of the Pd hydrides differ from one study to another however [20].  ;  Concentrations of H gas at atmospheric pressure between 10% and 100% correspond to [H]/[Pd] ratios of 2  0.68 to 0.73, well into the (3-hydrides region. Reported slopes of normalized resistance (R/Ro) vs. [H]/[Pd] in the P region, on H absorption, vary from 0.87 to 8.3 depending on the study [20, 95]. 2  2.3.a Pressure-concentration (p-C) isotherms for the Pd-H system 2  It is generally easier, however, to measure the H concentration in the vicinity of the sensor rather than 2  measuring the atomic content of the material. Empirical equations describing pressure-concentration (p-C) isotherms can be used to convert H concentration in the sensor's surroundings to an estimate of the atomic 2  [H]/[Pd] ratio. A number of studies have been reported in the literature that measure these isotherms, and exact numeric results vary depending on the method and the form of the material used (eg. Pd black, Pd wire) [20, 93, 94]. Despite some differences in reported plateaux pressures and a-P phase boundaries, several generalizations can be made.  A  It is generally agreed that H absorption in Pd for small [H]/[Pd] ratios at room temperature follows a 2  Sievert's Law relationship:  [H]/[Pd]  (2.1)  = S-p[H ]  U2  2  where [H]/[Pd] is the atomic concentration, S is an empirical constant and p[H ] is the partial pressure of H . 2  2  The region over which this relation is valid corresponds to the a-region of the Pd hydride. 12  As [H]/[Pd] concentrations enter the two-phase region, the (p-C) isotherm enters a plateau, with atomic concentration of hydrogen in solution increasing quickly over small increases in the ambient H pressure. As 2  hydrogen content increases beyond this o>P transition into the p region, [H]/[Pd] content increases much more slowly with H pressure. The P-hydride isotherm can be approximated with the following equation: 2  [H]/[Pd]  (2.2)  = A + B\np[H ] 2  N  where A and B are constants. The pressure at which the phase transformation occurs is a function of temperature and absorption history. It is generally agreed that hysteresis occurs in the absorption/desorption isotherms. A diagram showing the relation between the (p-C) isotherm and hydrogen content in Pd is shown in Figure 2.2. The presence of the hysteresis poses a limitation for hydride resistance measurements to identify H gas concentrations corresponding to the phase transition. From the introductory chapter, one may 2  observe that sensors reported in literature using pure Pd generally operate at low H concentrations. Sensor 2  studies at higher concentrations typically feature Pd alloys such as Pd-Ni, in which the P-hydride is suppressed [25].  Pressure-Concentration Isotherm (Not to scale)  100 a.  Desorption  CM I  -Absorption  a. o  0.1 0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  H/Pd Atomic Ratio Figure 2.2 Shape of the pressure-concentration (p-C) isotherm of hydrogen in Pd. The first rise corresponds to the a-hydride phase and follows a linear form whereas the second sharp rise is due to the P-hydride phase and follows a logarithmic form. The plateau corresponds to the a-p transition region and shows the effect of hysteresis.  13  Our study of Pd resistive sensors is interested in looking at the behaviour of H in Pd at atmospheric pressure. 2  This corresponds to a solubility limit of approximately [H]/[Pd]=0.73 [93]. The ratio of H partial pressure to 2  atmospheric pressure can be considered directly as the H gas concentration in a H /N gas mixture. 2  2  2  2.3.b Resistivity vs. [H]/[Pd] Electrical resistivity measurements of Pd hydrides are typically normalized with the initial resistance (R ) of 0  the Pd metal. Early studies of changes in resistance as a function of atomic content show ah almost linear relationship, piecewise for each phase. The slope of resistance over atomic content (R/Ro vs. [H]/[Pd]) rises less steeply in the phase transition than in the a region, and depending on the study, either levels out or continues rising in the P-hydride region [20]. More recently, Sakamoto et ah [95, 96] reported Pd resistance data on H absorption with R/R vs. [H]/[Pd] 2  0  slopes that change drastically between different phases of the hydrides, to the extent where it follows the form of the (p-C) isotherms (Figure 2.2) very closely. In their results, resistance increases sharply in the a region, hardly changes over the phase transition and increases sharply again in the p region. At 323K, the slope of R/Ro vs. [H]/[Pd] relation drops off again within the p region (between 40% and 100% H ), albeit not as level 2  as over the o>P transition. On desorption, the flatter slope observed during absorption from 40% to 100% H  2  persists, causing a large hysteresis with subsequent desorption and re-absorption within the P and o>P transition region. The authors attributed this behaviour to P-induced lattice strains, as similar effects are observed with heavily cold-worked Pd [97]. The presence of a substantial'hysteresis will challenge direct application of resistive Pd sensors over the o>p transition.  N  Another recent study in the a-hydride region reports a correlation between changes in resistance and the surrounding H pressure [98]. The data was approximated with a linear fit to the square root of the H 2  2  pressure, which corresponds roughly to a linear relationship with the atomic [H]/[Pd] ratio according to Sievert's Law (Equation (2.1)). Deviations from the trend were consistently observed however at both ends of the fit for all three sets of data given.  2.3.c Supersaturation Another effect observed in some Pd-H studies is the presence of supersaturated a-hydrides at the initial onset 2  of the o>P transition. It was observed in some studies that a local maxima in pressure exists in the (p-C) isotherm, immediately before the a-P plateau [20]. This effect is found to be attributed to freshly annealed Pd samples, and is non-existent in cold-worked Pd samples or samples which have undergone previous a-P transformations [97]. Moreover, this supersaturation effect was found to account for variations in the 14  [H]/[Pd] ratios corresponding to the onset of the o>p transition, and changes the slope of the (p-C) isotherm in the a region.  2.3.d Magnitude of changes in resistance R/Pvo values for effects discussed in the above section varies from 1.0 to about 1.8 over gas concentrations between 0% and 100% H , regardless of the variety of R/Ro vs. [H]/[Pd] slopes reported. Observed 2  differences in various studies are best described with their reported slopes of the phase transition region. A number of groups claim increases in R/Ro of about 0.5 (>60% of the R/Ro range) [20], while another reports increases of less than 0.1 (-13% of the range) [95].  2.4 Percolation mechanism A different mechanism affecting the electrical characteristics of Pd in the presence of H is observed for semi2  continuous Pd fdms and wire segments near their percolation threshold. Percolation threshold refers to the point where the first continuous path appears across a film or line of growing Pd clusters. The overall electrical resistance of a metallic material changes over this non-conductive/conductive transition. In the case of Pd-H systems, studies have successfully shown Pd mesowires [30] and films [33, 99] that detect the 2  presence of H based on this mechanism. 2  Pd clusters formed via Joule heating of thin film Pd [33], and as-deposited Pd clusters on siloxane selfassembled monolayers [99], showed increased conductivity in the presence of H . Pd clusters swell in size 2  due to lattice expansion, directly generating conductive paths in some cases and narrowing inter-cluster gaps to lower the electron tunnelling barrier in others. The former effect is most evident around 1-2% H as the a2  P hydride transition occurs. The latter effect is dominant for lower H concentrations, and is useful for 2  reflecting small lattice changes within the a-region.  Pd mesowires about 150nm in diameter were also shown to exhibit percolation effects [30, 31]. These wires were electrodeposited on highly oriented pyrolytic graphite and subsequently transferred onto cyanoacrylate for H testing. In this case, aggregates and wire segments are formed after the wires' initial H exposure, as 2  2  hydrogen atoms desorbs from the Pd and the lattice contracts. Electrical conductivity is restored when the wire segments re-connect upon subsequent H exposures. 2  The effective resistance for an electron tunnelling across two points through a barrier is the inverse of the electron's transmission coefficient, and grows exponentially as the distance between these two points 15  increases for a constant electron potential. A number of groups have derived related formalisms and approximations to the conduction of electrons in semi-continuous media [100, 101]. It is sufficient for the purposes of this work to refer to physical observations of tunneling effects in the context of H absorption into 2  Pd metal. R/Ro values of 0.95 were reported for discontinuous Pd films about 1.6nm thick, exposed to 0.26% H at room temperature [101]. Similar work on (smaller) monolayer-promoted Pd clusters showed 2  comparable R/Ro values at approximately 0.025% H [99]. 2  j  1  The overall changes in resistance from percolation effects are much larger as H concentrations increase, with 2  reported film resistance down to half of their original values in one study for H pressures up to 90% [33]. 2  Examples reported in the mesowire studies showed R/Ro values down to 0.25 upon exposure to 10% H [30]. 2  2.5 Temperature effects 2.5.a Temperature coefficient of resistance in Pd hydrides Temperature plays a significant role in the electrical characteristics of Pd. The melting point of Pd metal is 1828K [102], and the temperature coefficient of resistance for bulk palladium is approximately 3.8 x 10" /K 3  [20]. This coefficient is related to changes in resistance according to Equation (2.3).  —  L  (2.3)  = a AT r  Where AR is the change in resistance due to temperature effects, R is the initial resistance, a is the T  0  T  temperature coefficient of resistance and AT is the change in temperature.  v  The temperature coefficient of Pd hydrides is a function of atomic content of hydrogen, increasing to ~ 4.2 x 10- /Katthe a region and dropping off as hydrogen concentration increases, down to about 1.8 x 10"/K at 3  3  [H]/[Pd]=0.7 [20], or 100% H . 2  2.5.b Pressure-concentration changes with temperature The increase in hydride resistance with temperature is offset by changes in the overall hydride content however. (p-C) studies generally agree on an upward shift and narrowing of the a-P transition plateau as temperature increases. Plateau pressure is reported to increase beyond 1 atmosphere by 433K [94]. This  16  translates to significant changes in the concentration and phase of the Pd hydrides present at a given H  2  pressure.  2.5.C Other temperature effects In the case of Sakamoto's study [96], which presented R/Ro vs. [H]/[Pd] relationships similar in shape to the (p-C) isotherm shown in Figure 2.2, changes in temperature between 323K and 433K during H absorption 2  causes changes in the height of the R/R vs. [H]/[Pd] plateau by up to 0.6 (75% of the range of R/R at 0  0  atmospheric pressure). Little change in the width of the plateau was observed however. Plateau height jumps from about R/R = 1.15 to 1.75 between 323K and 348K, then slowly decreases as temperature continues to 0  increase up to 433K. A diagram illustrating this behaviour is shown in Figure 2.3.  Reported R/Ro on Hz absorption (Not to scale)  I  1  Plateau height decreases  /1  / 1  from 323K to 433K  i  Plateau height increases  r  from 323K to 348K  l i ji  0  r  1 0.1  0.2  0.3  0.4  0.5  0.6  0.7  0.8  H/Pd Atomic Ratio Figure 2.3  R/R behaviour on temperature as reported by Sakamoto et al. Changes in plateau resistance are not monotonic with temperature. 0  The thickness of the Pd metal plays a role in its temperature coefficient, dropping off as film thicknesses decrease to less than lOnm [103]. Temperature also affects percolation films, where resistances (again) vary non-monotonically with temperature [101]. These temperature effects are exemplified in the variety of temperature dependencies observed from a number of Pd resistive sensors [36, 29].  2.5.d Temperature modelling Since the Pd structures studiedin this work behave as resistors across an applied voltage, it is instructive ,to estimate the extent to which their temperature increase during operation. To this end, a two-dimensional 17  COMSOL Multiphysics simulation was done to model the heat profile along a Pd mesowire 200nm wide, lOOnm thick and 60um long resting on thick Si02. Resistivity is set up as a function of resistance in the models. The initial wire resistance used in the simulation is 325Q, and is estimated from  where p is the resistivity, approximately 10.8 x 10" (Q, m) for bulk Pd at 300K [102], L is the length and A is the cross-section area of the modeled wire. This simulation was constructed and set up by Kevin Stanley at the National Research Council of Canada - Institute for Fuel Cell Innovation (NRC-IFCI). The model is shown in Figure 2.4 with labelled regions in the inset (not to scale).  171x10  COMSOL Multiphysics model  4  5  4  250urn  ^  Air  ^ 30um  30um  Air  3  i  Pd  Oxide  2  JOOnrX  7  600nrrY  Si 001  250um  X  ^ 1  1  0  i  1  2  3  4  s m*io  4  Figure 2.4 Heat Transport model of Pd mesowires with inset (not to scale) showing model structure. The Pd wire is mirrored across the boundary to the right, and is half the length of the simulation by symmetry.  Hydrogen content is not taken into account in this modelling study. Parameters modelled include: 18  1. Temperature changes from Joule heating, where the heat source is the applied voltages across the wire, 2.  Resistance changes from temperature differences (using a thermal coefficient of 3.3 x 10" /K); 3  3. Radiative, convective and conductive heat dissipation.  Plots of the temperature profile along the wire with voltages stepping from 0.5VDC to 7.0VDC are shown in Figure 2.5. From the simulation, the estimated voltage that brings the wire to its melting point is about 6.8VDC at a maximum temperature of 1825K.  Temperature along Pd wire 0.5V 1.0V 1.5V 2.0V - * - 2.5V * - 3.0V 3.5V 4.0V 4.5V 5.0V <•>••• 5.5V 6.0V 6.5V 7.0V —3—  —3—  —X—  0.5 1 1.5 2 2.5 3 Arc-length (m, edge to mid-span) i o " 5  x  Figure 2.5  Estimated temperature along Pd wire at various voltages. Modelled properties include conduction, convection and thermal effects on resistance. Thermal coefficient of 3.3 x 10"/K is used. 3  At excitation voltages below 0.1 VDC, temperature variations within the Pd wire is less than 0.34K. The temperature of the wire at mid-span increases from 300K to 301.7K as the applied voltage approaches 0.1 VDC. This is shown in Figure 2.6.  19  Temperature along Pd wire  302.5  ^0.02V - a - 0.04V -•-0.06V 0.08V -*-Q.10V * 0.12V  302  & ^301.5 3  E --0-  E 301  «-  300.5 _j  g  300 0  i  i  i  i  r#  —:—  ^  ;  0.5 1 1.5 2 2.5 3 Arc-length (m, edge to mid-span) 1 0 5  X  Figure 2.6  Estimated temperature along Pd wire at voltages varying from 20mVDC to 120mVDC  Effects of a changing temperature coefficient on temperature was investigated and a decrease of 0.006K was noted at an applied voltage of 0. IV over the coefficient increase from 1.8x10" /K to 4.2 x 10" /K. 3  3  By far the dominant course of heat dissipation in our Pd wires is through conduction to the oxide layer. It is likely, therefore, that localized heating will occur in the event that portions of the wire delaminates from the oxide layer. A similar simulation was done for a structure in which the layer beneath the wire is sectioned into oxide and air partitions to simulate the presence of a delaminated wire section. A diagram of the model is shown in Figure 2.7, and simulated temperatures from a voltage sweep of 0.2VDC to 1.0VDC are shown in Figure 2.8.  30pm  A  Air  Figure 2.7  r  250pm  30pm  Pd  100nm  Oxide  Air  Oxide 40pm  '  •  1*4  M  gj 12.5pm  600nm  7.5pm  250pm  Model structure of Pd mesowire with a delaminated section (not to scale) for heat transport simulations.  20  Temperature along delaminated Pd wire  Figure 2.8 Temperature simulation of a delaminated Pd wire (over a 12.5um air gap). The structure of this simulation was illustrated in Figure 2.7. This figure clearly demonstrates the deleterious effect of delamination on sensor performance.  The simulations above consist of a 12.5um air gap beneath the Pd wire, located lOum from the edge of the wire. Given that Pd is prone to delaminate to some extent in the presence of H , even moderate voltages 2  applied to probe the wire sensors may cause a substantial temperature effect in the sensors' response. Figure 2.9 shows plots of the same delaminated wire model for a range of smaller voltages. Applying 0.1V across the wire will result in an estimated temperature change of about 1 IK for a delaminated length of 12.5(rm.  Temperature along delaminated Pd wire  0  0.02V 0.04V 0.06V 0.08V ^O.IOV ..... 0.12V  0.5 1 1.5 2 2.5 3 Arc-length (m, edgetomid-span) i o " x  Figure 2.9  5  Temperature simulation of a delaminated Pd wire (over a 12.5um air gap) for voltages between 0.02VDC and 0.12VDC. 21  As discussed earlier, temperature increases will generally cause an upward shift in Pd and Pd hydride resistance. For a temperature increase of 1 IK, an estimate of  /\R /RQ — T  0.04 at 0% H can be obtained 2  (Equation (2.3)). Changes in resistance due to variations in the hydrogen solubility, however, are less predictable. These modelling results are important for selecting a suitable excitation voltage that provides good signal-to-noise while maintaining reasonable uniformity in sensor temperature.  2.6 Anticipated observations Given the multitude of mechanisms occurring in Pd-H systems that directly affect their electrical 2  characteristics, it is useful to summarize some anticipated changes in resistance we expect to observe in the sensors developed for this thesis. Relative resistance increase (R/Ro) of up to 1.8 is expected when sensor wires are exposed to H concentrations between 0% and 100% due to hydride resistance changes. This change 2  in resistance is expected to be non-linear with H gas concentration, and the shape of the relationship is 2  expected to be different for H concentrations corresponding to different phases of the Pd hydrides. A 2  hysteresis is expected to be present over the a-P hydride transition, and some irreversible changes in resistance may be observed due to supersaturation and cold-work-like effects. For sensors in which a large number of defects are present, resistance changes in the form of large step decreases may also occur due to percolation effects. Temperature effects onthe Pd resistance are expected to be negligible, as sensors will be probed with voltages less than lOOmV to ensure reasonable temperature uniformity. Test results obtained from sensors developed in this thesis will be evaluated with the above theoretical considerations in Chapter 1.  22  Chapter 3  Fabrication of Pd Mesowire Resistors  A description of the sensors developed in this thesis is presented in this chapter. Several sensor designs were fabricated, all of which consist of Pd wire structures spanning across two metallized contact pads on an insulator. A diagram showing the top view and cross-section of the general structure of these sensors was shown earlier in Figure 2.1. Actual sensor designs fabricated are variations of this basic structure.  3.1 Fabricated sensors PMMA (Polymethylmethacrylate) is a polymeric material that is sensitive to electron bombardment. It is a common electron-beam (e-beam) resist that hardens (crosslinks) upon exposure to high electron dosages. At moderate dosages, PMMA decomposes into chains of lower molecular weight fragments and becomes soluble in certain solvents (developers) in which it was originally insoluble. PMGI (polydimethylglutarimide), similarly, is an e-beam sensitive resist that uses a different developer and remover than that of PMMA. PMGI is often used as an undercut layer for liftoff processes. An e-beam lithography and liftoff process consists of A) coating a substrate with e-beam resist and using a beam of electrons to create patterns on the resist, B) developing the resist to obtain 'windows' that clears through the resist layer to the substrate, C) depositing a layer of material on the substrate and D) removing the resist layer, leaving behind window patterns of the desired material. A diagram illustrating the general idea behind the lithography and liftoff process is shown in Figure 3.1.  A  C  B  D  Figure 3.1 Lithography and liftoff process: A) Coating with resist and patterning with electron beam; B) developing exposed resist; C) depositing the desired material for the pattern and D) removing the resist layer and leaving behind the patterned layer.  Our present study uses the above process to fabricate H sensors. The earliest sensor design consisted of about 2  30 parallel wires spaced approximately 1pm apart. None of these samples withstood the PMMA development process. Subsequent designs are shown from Figure 3.2 through Figure 3.5. In these designs, spacing between wires is wider, and two additional sensor designs were fabricated. The first design, shown in Figure 3.4, is a symmetric U-shaped structure that may potentially produce a gap across the Pd traces close enough 23  together to form a mechanical switch from lattice expansion. The second is a platform structure, shown in Figure 3.5, which was originally designed to investigate crosslinked P M M A as a negative resist. A s w i l l be discussed at a later chapter, there is insufficient data to distinguish differences in H response among these 2  sensors designs.  SE  Figure 3.2  08-May-06  061243 WD13.1nmi 2 0 . O k V x l ^ V °25UIII  Typical early design sensor showing many-wire Pd array across Au/Cr contacts approximately 65um apart.  S4700 5.0kV 6.1mm x1.10kSE(U) 3/21/07  Figure 3.3  '  50.0um  '  Later design sensor withfive-wirePd array. These sensors have improved liftoff yield and higher overall resistance than many-wire arrays.  24  F i g u r e 3.4 Sensor w i t h U - s h a p e d s y m m e t r i c structure. Instead o f p a t t e r n i n g gaps a l o n g a w i r e by t u r n i n g off the e-beam, these samples are patterned by h a v i n g the e-beam t r a v e l h a l f o f the gap and then r e t u r n i n g to the same side. A U shape was used on each side ( s y m m e t r i c a l l y ) to lessen overexposure o f the resist at the t u r n i n g points. A t t e m p t s at generating gaps i n this m a n n e r were unsuccessful.  F i g u r e 3.5  F i v e - w i r e device sensor w i t h c r o s s l i n k e d p o l y m e r p l a t f o r m . T h e presence o f imperfections a l o n g these wires have less effects on the o v e r a l l resistance than i n other sensors developed.  The above sensor structures are fabricated via e-beam lithography and liftoff of Pd. These structures are patterned between and spans across etched Au contact pads on oxidized Si substrates. The fabrication process of these sensors will be described in the following section.  25  3.2  Process description  3.2.a Substrate preparation 50.8mm (2") p-type Si(l 11) wafers about 279um ± 25u.m thick (Polishing Corporation of America) are RCA cleaned and thermally oxidized at the Institute of Micromachine and Microfabrication Research at the SFU School of Engineering Science. The oxide layer for samples fabricated vary from 200nm to 600nm in thickness and serves as an insulating barrier on which the sensors are built. A thin layer of Cr is sputtered briefly onto the oxidized wafer as an adhesion layer for the subsequent Au deposition (SFU). Thicknesses between 174.5nm and 213.5nm are measured for the Au layer using a travelling-probe profdometer (AlphaStep) at the UBC Nanofabrication Facility. The metal layer is photolithographically patterned and subsequently etched in aqua regia to produce contact pads for the sensor. The earliest version of these contacts consists of pairs of 1.8mm x 1,975mm rectangular pads spaced 80um apart. Later versions are smaller and are designed with alignment marks to facilitate sample navigation during the e-beam writing process. An SEM image of the latest version of the sensor contacts is shown in Figure 3.6, and the corresponding mask used to fabricate them is shown in Figure 3.7.  Figure 3.6  Au contact pads of a sample, shown with bonded Au wires. Crosses are alignment marks to ease navigation of electron beam during sample exposure.  26  F i g u r e 3.7  M a s k for p h o t o l i t h o g r a p h y o f A u / C r contact pads for fabricated sensors. P a d sizes are • l m m x l m m . T h e patterns are rotated to align w i t h the cleave planes o f S i ( l l l ) wafers.  Dots between the contacts shown in the SEM image originate from imperfections on the acetate masks as received from the print shop (Digital People Inc.). Resolution of the printer was set at 3600dpi, which corresponds well to the edge roughness observed (<10um) and is sufficient for our work. Masks are printed as positives as they are used directly in the mask aligner to expose portions of the photoresist to be etched. Patterns are rotated to align with one of the cleave planes of Si(l 11) wafers.  The earliest samples fabricated were done on Cr sputtered contacts. Subsequently, Au films on a Cr adhesion layer were used instead of Cr to permit bonding with conventional ceramic packages. About half of the 27  wafers prepared this way are subsequently spin-coated with a 200nm or 400nm layer (4,000rpm and 2,000rpm for 45s respectively) of polydimethylglutarimide (PMGI SF5 from MicroChem) and heated to a temperature between 458K and 473K on a hotplate for 5min. Fabrication work up to this point, aside from design specifications, film characterization and lithography mask development, is performed by Eva Czyzewska of the NRC-IFCI at the SFU microfabrication facilities.  Both PMGI-coated and bare substrates are subsequently spin-coated at UBC with a 200nm layer of polymethylmethacrylate (PMMA C4 MW950k from MicroChem Corp). Existing data for the thickness as a function of spin rate was not available for PMMA on PMGI polymer. Cleaved cross-sections of bilayer samples confirm that the resist thicknesses are very close to expected values for spinning on bare Si. A crosssection SEM image of the bilayer resist is shown in Figure 3.8.  ^ M l l F i g u r e 3.8  B i l a y e r resist cross-section on S i 0 . T h e u p p e r l a y e r is P M M A , the m i d d l e layer is P M G I a n d the bottom is S i 0 . Thicknesses o f the resist layers a r e about 200nm each. 2  2  Wafers are subsequently diced into dies of 3 to 4 samples each, about 6mm x 8mm in size. Unused dies are kept in wafer boxes stored in a lithography room to minimize resist degradation. Selected dies (chips) are individually exposed in an e-beam lithography SEM at UBC (JEOL 840 with NPGS e-beam writing control software).  28  3.2.b Sensor patterning Patterning the chips typically requires navigation of an e-beam over the chip, thus exposing a path. Care must be taken to prevent accidental exposure of critical areas on the chip, as well as unexpected electrical 'shorts' across different sections of the sample.  Patterns with an overall size of approximately 90um x 90um are'drawn and saved as DesignCAD fdes. Pattern settings such as the pattern name, electron dosage, line spacing, line widths and beam currents are entered in the Nanometer Pattern Generation System (NPGS) program. NPGS calculates the beam 'dwell time' for each point based on the entered data, and controls the e-beam accordingly. For this study, dosages are entered as "line doses" with centre-to-centre spacing of 2.98nm. Centre-to-centre spacing refers to the distance between exposure points making up a line, and should ideally be set to about half of the smallest feature size. One of the original goals of this work was to introduce gaps on the order of several nm along the wire patterns. As will be shown at a later paragraph, the resolution needed to achieve these gap sizes using direct writing is likely beyond the capability of the instrument. The small centre-to-centre spacing was kept in later work for consistency.  PMMA and PMGI are both electron-beam sensitive resists. There are a number of studies reported in literature that investigate electron-PMMA interactions. Energy dissipation of incident electrons on PMMA peaks at a characteristic depth from the surface depending on the electron flux, resist density and the incident beam potential [104, 105]. An accelerating voltage of 20kV was used for patterning our sensors, which provides a fairly uniform beam exposure throughout the resist layer(s). The e-beam is first optimized on reference Au clusters, after which sample heights are adjusted to the working distance of the e-beam. Some SEM and NPGS program user settings are listed in Table 3.1 and Table 3.2.  T a b l e 3.1  J E O L 840 S E M Settings  Accelerating Voltage  20kV  Working Distance  15mm  Filament Current  240uA  Emission Current  50uA ,  Physical Aperture  4  Probe Current  3 x 10""A  Measured Beam Current  lOpA  (Faraday Cup) Chamber Vacuum  < 1.0 x lO-'torr 29  Table 3.2  NPGS e-beam lithography program settings lOOOx (90um x90um  Magnification  writing area) Center-to-center  30A  (  Distance Line Spacing  500A  Coarse Beam  1A  Measured Beam  As measured at  Current  Faraday Cup  Line Dose  0.5 to 4nC/cm  Electron dose in lithography refers to the amount of electrons per spatial unit that a sample is subject to, and is typically reported in literature as charge per unit area (eg. C/m ). Doses entered in the NPGS program can 2  be a point, line or area dose (fC, nC/cm, and uC/cm respectively) while the program itself calculates the 2  corresponding dwell time of the beam at each point. Equivalent area dose calculated by NPGS for the lithography settings used in the present work are tabulated in Table 3.3.  Table 3.3  Equivalent line and area dose from NPGS program  Line Dose  Area Dose (NPGS)  0.5nC/cm  100u€/cm  1 .OnC/cm  200uC/cm  3.0nC/cm  600uC/cm  lO.OnC/cm  2000uC/cm  2  2  2  2  Hoole et al. [106] observed and estimated an area vs. line dose relationship described by Equation (3.1), assuming a Gaussian distribution for each point along a line. D line  (3.1)  C7V7T  where D  area  and Du„ are the equivalent area and line dose, and o the Gaussian width. Using the values listed e  in Table 3.3, expected exposed PMMA linewidth can be estimated by the full width at half maximum (FWHM) of a Gaussian curve, approximately 67nm in our case. Given that  FWHM  = 2V21n2cr  (3.2) 30  The calculated 'expected' linewidth is a clear indication that patterning gaps on the order of several nm along the wires is not possible with the present setup. Finer features may be achievable on the instrument at a higher magnification at the expense of a smaller pattern writing area.  3.2.C Developing the resist Our observed line widths of fully developed lines exposed at 3nC/cm of electron dose are on the order of 150nm. Finer lines were observed earlier in the development process, though samples were generally developed for a longer time to ensure the PMMA is cleared through to the substrate. Later trials with development times show that PMMA is indeed cleared through much earlier. As such, it is quite conceivable that existing equipment may achieve developed line patterns on the same order of magnitude as the FWHM value estimated earlier. A thinner PMMA resist layer (such as PMMA in anisole instead of chlorobenzene solvent) may also help to produce thinner lines. Off-focusing of the electron beam also plays a role, since the focus plane of each sample is currently estimated from focusing on adjacent Au edges. We expect this effect to be small compared to the line-widening effects from extended resist development, as our focusing method was checked against spot focusing on the PMMA on several occasions and the resulting sample height from both methods were in good agreement.  The present development process for sensor fabrication uses 1:3 ratio of MIBK:IPA (methyl isobutyl ketone: iso-propyl alcohol) for developing the PMMA resist. It is well known that moderately-exposed PMMA becomes soluble in MIBK due to degradation of the polymer into fragments of lower molecular weight. Exposure to much higher doses, however, causes the polymers to crosslink and become a very tough material that typically requires plasma cleaning to remove. Confirmation that the resist is working as a positive resist with our e-beam settings is clear after Pd deposition, where developed lines can be seen in the form of grooves. An SEM image showing partial liftoff of a Pd pattern is shown in Figure 3.9.  31  Figure 3.9  Partial Liftoff of Pd from e-beam exposure of PMMA. This shows PMMA working as a positive resist at moderate electron-beam dose.  One set of samples was fabricated using PMMA both as a positive and a negative resist. Line patterns were exposed at the usual dosage, after which a vertical line with a much larger dose was written down the centre of the pattern, forming a platform at the centre of the gap. An example of this structure was shown earlier in Figure 3.5.  Exposed PMMA samples are sonicated in the MIBK:IPA solvent mixture for 3-4 minutes. Bilayer resist samples use PMGI as the 'undercut' layer. These samples are further developed with sonication in 2.2% tetramethylammonium hydroxide (TMAH, Rohm & Haas Electronic Materials MF319). As described in datasheets from the supplier, the development rate of the PMGI layer depends largely on the amount of leftover solvent in the resist, and is consequently dependent on the heat treatments and storage history of the sample. As a result, no correlation was observed for development rates across samples on different chips. It is clear, however, that for a given chip, the undercut development rate of the PMGI layer increases with electron dosage and/or the size of the PMMA opening. It was observed that the structural integrity of the PMMA is compromised as undercut cavity widths grow beyond lum. An example of patterned PMMA delaminating from the substrate is shown in Figure 3.10. It was concluded that the PMGI development progress should be monitored to make sure resist structures remain intact.  32  Figure 3.10 Effects of overdeveloping the PMGI undercut layer. The PMMA layer becomes entirely delaminated from the substrate.  Development progress for a sample with bilayer resist is shown in the following images. E-beam dosage for this sample is 4.0nC/cm, and is chosen for the best image contrast. Similar progression is seen for samples exposed with dosages down tol .OnC/cm (in 1 .OnC/cm decrements). Exposure at 0.5nC/cm was not sufficient to completely clear samples with a 200nm layer of PMMA developed in 1:3 MIBK:IPA for 5minutes.  Progression of MIBK development of PMMA exposed at 4nC/cm  50nm Figure 3.11 PMMA development: IminOOsec, 1:3 MIBK:IPA. Very little development is observed at this point.  33  F i g u r e 3.12 P M M A development: 2min00sec, 1:3 M I B K : I P A . F a i n t traces c o r r e s p o n d i n g to developed lines begin to appear.  50nm F i g u r e 3.13 P M M A development: 3min00sec, 1:3  MIBK:IPA.  50nm F i g u r e 3.14 P M M A development: 4min00sec, 1:3 M I B K : I P A .  Patterns i n P M M A are now easily distinguishable.  34  F i g u r e 3.15 P M M A development: 5min06sec, 1:3 M I B K : I P A .  E n d o f the M I B K development process.  TMAH development of PMGI layer following development of PMMA  F i g u r e 3.16 P M G I development: 0min20sec 2 . 2 % T M A H . Patterns begin to b r o a d e n .  «  M  50nm  F i g u r e 3.17 P M G I development: 0min40sec 2 . 2 % T M A H . C a v i t i e s u n d e r the P M M A layer (dissolved P M G I ) begin to a p p e a r . 35  F i g u r e 3.18 P M G I development: IminOOsec 2 . 2 % T M A H . M i l d d e l a m i n a t i o n (circled) o f the P M M A begins to appear.  50nm F i g u r e 3.19 P M G I development: l m i n 2 0 s e c 2 . 2 % T M A H . F u r t h e r w i d e n i n g o f P M G I cavity.  In the above images, the gradual emergence of written patterns can be observed. The bright regions on the two sides of each image are edges of the Au contact pads. The edge roughness is attributed to the mask used to pattern these contacts, which is simply a piece of acetate film printed at a local print shop.  MIBK development appears to be generally consistent with e-beam dosages, and the development / linewidening rate is greater for lines written at higher exposure. A control piece was cleaved off and kept after each development process, and a method was developed to cleave samples through the 60um gap in order to view the cross-sections. Unfortunately, even when cleaves run successfully through the gap between the contacts without extensive debris being generated, delamination and peeling of the resist layers occurred. Attempts at grinding samples held in epoxy with standard metallurgical techniques were also largely unsuccessful due to the excessive debris generated at patterned areas. It is possible that air trapped in the 36  small channel features was not completely removed before the epoxy cured, leading to structural damage during the grinding and polishing process. Images of samples cleaved through the pattern gap are shown in Figure 3.20 (top view) and Figure 3.21 (side view).  Figure 3.21 PMMA Resist peeling off is clearly visible from sample cross-sectioned area. There was one sample that produced three consecutive, successful cross-sections in which the progression of PMGI cavity sizes as e-beam dosage varies can be observed (despite PMMA peeling). An SEM image of one 37  of these cross-sections is shown in Figure 3.22, followed by a chart showing the increase in cavity width as ebeam dosage increased.  F i g u r e 3.22 Cross-section o f developed P M M A / P M G I b i l a y e r . T h e l a y e r to the left is P M M A , w h i c h can be seen w i t h a large flap/extension left b e h i n d f r o m the cleaving process. T h e layer w i t h the cavity is the P M G I layer, s a n d w i c h e d between the P M M A a n d the S i 0 . 2  Width of developed PMGI Cavity 1 0 0 0  i  Dosage (nC/cm) F i g u r e 3.23 C h a r t s h o w i n g the w i d t h o f the developed P M G I cavities as a function o f electron dosage. O n l y a few data points are a v a i l a b l e because o f difficulties i n o b t a i n i n g intact cross-sections o f p o l y m e r layers.  38  3.2.d  Sputtering and liftoff  Exposed and developed chips are brought to SFU Micromachining facilities where a 200nm ± 30nm thick Pd layer is sputtered on. Thicknesses are measured from 'control' glass slides that were sputtered at the same time. Measurements were made using an Alpha-step and a Talysurf profilometer (located at UBC and NRCIFCI respectively), and vary from 170nm to 235nm. One series of bilayer resist samples were sputtered with a thicker layer of Pd (measured to be approximately 290nm ± 26nm).  Liftoff for samples with a single layer of PMMA resist is done by sonicating samples in acetone for about 3minutes. Bilayer PMMA/PMGI resist samples are first soaked in l-methyl-2-pyrrolidinone (Shipley Remover 1165) for 30minutes, then briefly sonicated in the Remover and rinsed in DI water. Problems were encountered during the liftoff process for a number of samples in which residue resist layers remain despite extended immersion in the remover solvent. Contamination is suspected to be responsible as the same protocols at a different site (UBC) with dedicated glassware for the liftoff process did not encounter the same problems.  Examples of successful sensor structures fabricated were shown at the beginning of the chapter in Figure 3.2 through Figure 3.5. Gaps directly patterned along wire structures could not be seen in the resist during development, nor were they visible after the Pd liftoff process. A comparison of post-liftoff Pd wires shows a marked difference between wires fabricated with a single PMMA layer resist and those fabricated with PMMA/PMGI bilayer resist. An example of each is shown in Figure 3.24 and Figure 3.25.  Figure 3.24 Pd wires after liftoff from single layer resist 39  F i g u r e 3.25 P d w i r e after liftoff from b i l a y e r resist. P d wires fabricated from the b i l a y e r process are t h i n n e r than w i t h the single layer P M M A process.  Figure 3.24 and Figure 3.25 are tilted projections of Pd wires fabricated from PMMA-only liftoff and PMMA/PMGI liftoff respectively. The former yields wires that are rougher and thicker (close to the full deposition thickness). In single-layer PMMA samples, removal of the PMMA layer causes the Pd fdm to fracture at the wire edges (as sputtering is not directional). There are a number of samples where entire patterns were removed, presumably when the tensile strength of the fdm overcomes adhesion of the Pd fdm to the SiC»2 substrate. From the processing point of view, liftoff from a single layer resist at feature sizes down to lOOnm is not a reliable process if a non-directional deposition method is used. Later samples are fabricated with bilayer resist.  Pd wires deposited through bilayer resist windows are thin and relatively wide. These wire structures have a very smooth and gradual height profde. Liftoff of the bilayer samples originally appeared to have removed all patterns when viewed through an optical microscope! A more careful inspection of the samples showed that there were indeed Pd structures remaining. Energy Dispersive X-ray spectroscopy (EDX) measurements were made to confirm the presence of Pd (Hitachi S3500N at NRC-IFCI) and are shown in the following Figures.  40  F i g u r e 3.26 S e c o n d a r y E l e c t r o n s image o f a t h i n P d w i r e sputtered t h r o u g h b i l a y e r resist.  F i g u r e 3.27 E l e c t r o n counts associated w i t h S i obtained f r o m E D X measurements. S l i g h t l y lower S i count density is observed across the l o w e r h a l f o f the image c o r r e s p o n d i n g to the location o f the P d w i r e .  F i g u r e 3.28 E l e c t r o n counts associated w i t h P d obtained f r o m E D X measurements. P d is present as a v e r y faint b a r o f h i g h e r count density across the l o w e r h a l f o f the image. These measurements p r o m p t e d further c h a r a c t e r i z a t i o n of the b i l a y e r samples w i t h a field emission s c a n n i n g electron microscope ( F E S E M ) a n d an atomic force microscope ( A F M ) , w h i c h d e t e r m i n e d u n a m b i g u o u s l y the presence o f the P d wires as a continuous thin  film.  41  Subsequent investigation with a Field Emission Scanning Electron Microscope (FESEM) produced better quality images. It was determined that these wires are intrinsically difficult to image due to the lack of a clear Pd / oxide boundary. The wires' height is believed to be self limiting from closure of the PMMA windows as sputtered Pd accumulates at the window edges. Top views of Pd wires fabricated with single-layer and bilayer resist, obtained from a high resolution FESEM (UBC Bioimaging Facility), are shown in Figure 3.29 and Figure 3.30 for comparison.  S4700 5.0kV 6.1mm x90.0k SE(U) 3/21/07  i  i  I  500nm  i  T  Figure 3.29 FESEM micrograph of a Pd wire on oxide fabricated with single-layer resist.  Figure 3.30 FESEM micrograph of a Pd wire on oxide fabricated with bilayer resist. 42  The Pd height profde of a bilayer sample was obtained using an Atomic Force Microscope (AFM). This profde was obtained by Andras Pattantyus-Abraham from UBC Physics and Astronomy. A plot of the wire cross-section is shown in Figure 3.31.  0  500  •  r  i  100  150  Width (nm)  200  250  300  Figure 3.31 Cross-section of Pd mesowire sputtered through bilayer resist. Fitted line follows the Gaussian form given by W +W,*exp(-((x-W AV ) ). 2  0  2  3  The profde is Gaussian and is fitted with the function  x-W  2  h = W,+W -e  (3.3)  x  where h is the height, W are constants and x is the lateral distance across the wire. A fitted curve is shown as n  a continuous line on the same plot. The height is approximately 90nm and the FWHM width is around 260nm. Best-fit parameters used to calculate these values are listed in the following Table.  Table 3.4  Gaussian best-fit parameters for Pd wire cross-section.  Constant  Value  w  -0.32  W  91  0  ]  W  2  w  3  1400 150 43  Edge-to-edge widths for wires from bilayer resist samples are obtained from FESEM images. The image contrast had to be manually adjusted to identify the threshold where Pd is detected. These widths, as expected, increase with e-beam dosage. An example of these images is shown in Figure 3.32, and a summary of these widths is provided in Figure 3.33.  S4700 5.0kV 6.1mm x90.0k SE(U) 3/21/07  i  i  i  i  i  i  i  i  i  SOOnm  i  i  F i g u r e 3.32 T o p view o f P d w i r e deposited t h r o u g h b i l a y e r resist. S a m p l e was exposed w i t h l . O n C / c m electron beam dose, a n d subsequently developed in 1:3 M I B K : I P A a n d M F - 3 1 9 for 5m06s a n d l m 2 0 s respectively.  Pd width vs Dosage 1.2 C5  =  •g c E o c  1  H  0.8 0.6  TJ  D_  0.4 2  3  Dosage (nC/cm) F i g u r e 3.33 S p u t t e r e d P d w i r e w i d t h s w i t h i n c r e a s i n g e-beam dosage. W i d t h s a r e n o r m a l i z e d w i t h the measured w i d t h o f wires exposed w i t h 4 . 0 n C / c m electron dose. 44  The performance of some of these sensors will be described in detail at a later chapter. The thicker sensors fabricated with single-layer liftoff tend to delaminate readily in the presence of H . Delamination sometimes 2  leads to localized heating and causes sensor failure. The thinner, bilayer liftoff sensors are more robust, but have sometimes been observed to cluster and form disconnected beads upon exposure to H . These results 2  will be described in Chapter 1.  3.2.e  Packaging  Chips that are successfully fabricated are mounted onto 24pin ceramic DIP packages (Global Chip Materials) using conductive double-sided tape. Contact pads on the samples are gold-bonded to bonding pads using a wedge wirebonder (WestBond 7400A) at the UBC Nanofabrication Facility. Figure 3.34 shows a photograph of two packaged samples, ready for H testing. 2  Figure 3.34 Packaged, bonded chips for sensor testing.  3.3 Sensor testing Sensors fabricated in the manner described above are tested in H gas by measuring the resistance across the 2  sensor contact pads. It was generally observed that sensors fabricated using PMMA/PMGI bilayers produced better quality results - both in terms of mechanical stability and electrical signal-to-noise. Results of these tests will be discussed in Chapter 1  45  Chapter 4  Development of a High Concentration H Test Bench 2  4.1 Application requirements A facility for testing high-concentration H sensors is ideally one where H is either isolated from 2  2  oxidants and sources of heat, or sufficiently diluted and purged quickly to ensure that explosive or flammable mixtures will not build up. The equipment should provide accurate measurement and control of gas concentrations, and allow gas mixtures to be tested. Pressure, humidity, temperature, vibrations, flow velocities and the presence of other chemical species are key environmental parameters that affect sensor response. H is known to cause embrittlement in numerous materials. Materials for a H test 2  2  facility should therefore take into consideration its chemical compatibility with H , and the mechanical or 2  structural function of each component.  4.2 Description of the facility A H sensor test bench was designed and assembled to provide accurate control over gas concentrations 2  with minimal system response time. Initial sensor test results have been obtained on this test bench, but are mostly limited to that of a few previously characterized Pd-MOS capacitor sensors. The development of this apparatus was nonetheless a major component of the thesis work, and is thus described here for completeness. At the present time, the test bench controls the composition of H /N mixtures accurately 2  2  where N is present as a trace gas. The test bench is divided into 3 modules that can be used 2  independently of one another: the gas (mixing) manifold, the test chamber (TC), and the data acquisition electronics (DAQ).  4.2.a Gas manifold The gas manifold module takes N and H gases from the building supply and routes them to a network of 2  2  mass flow controllers (MFCs). A valve panel downstream from the MFCs allows the user to selectively mix the monitored gases in two separate lines, LI and L2. These lines afe electrically controlled downstream with solenoid valves in which one of either LI or L2 is open to the test chamber (TC) at any one time.  This line-switching mechanism is designed to allow gas mixtures to stabilize before being  supplied to the test chamber, and warrants accuracy of the system response time and gas concentration. A separate line connected to N and compressed air is routed to the test chamber directly through a high 2  flow rate MFC for test chamber purging (up to 50 Standard Litres per Minute (slpm)). A schematic of the gas manifold and test chamber modules is shown in Figure 4.1. 46  V  Pressure Regulator J\j  Check Valve  Junction  |^j  Ball Valve (lever)  Flash Arrester  Gas Filler  TO e re a  S3 cya  3 3  ]  3-Way Ball Valve  •ifi  ReliefValve  Cap Solenoid Valve  0  Thermocouple  fpl  Pressure Tranaduc'ei.Gauge  (MFC|  ^  a  s  s  ^  o  w  Conlrdler .  3-Way Solenoid Valve  Gale Valve  TC  Tesl Chamber  RV  ReliefValve  PV  Purge Valve  PRV  Purge ReliefValve  II,V, NjV  Hydrogen, Nitrogen Venl Valves  NC.NO  Normally Closed, NormallyOpen  L1V.L2V  Line 1, Line 2 Venl Valves (Co EXHAUST!  L1S.L2S  Solenoid Valves, electrically connected loswap either L] or L2 to L1NE0UT or EXHAUST  TEST" C H A M B E R  BENCH MANIFOLD  | Q. BS 3  a. re sr S3  March 12,2007 manifold schematic v4.dwg  ABBREVIATIONS  LEGEND  ,, (RACK) (RACK) JPPLY PV V1 1 II, SUPPLY P  II,JN RV2 ^ A >  &I-H&—a  1  ™  T  RA  H,MAiN  SNC NC T  IIUMIDiFIER  r  ^jn  m  i  A  LIS  U V  T  LINEOUT  TNC TC OVERPRESSURE  (I! T 1  EXHAUST  VENT O 3"  re 3  S3  (RACK) N, SUPPLY  3" ST  re C re 3 re ET  N.1N  •nil—o  re'  CET— rH©H  r-tir-  0  RVT  0,  TESTOIAMBER - K M  TC PURGE  N,l  (RACK) (RACK) NP , URGE PV5  0,  PURGE QP MAIN RELIEF RESET  (RACK) AIR SUPPLY  M:.  TC RELIEF RESET  L^-T>M TC  -1(3)1-  EXHAUST I  PURGE -|MFC|  j\J-  4.2.a.l  Supply to MFC controllers  Main gas shutoff valves are provided by the building's gas supply racks. Present gas services to the supply rack include H , N , compressed air and domestic water, along with an actively ventilated copper 2  2  vent for combustible gases and a separate vent for oxidants. Provisions (additional labelled lines) exist for other gases such as 0 , CO, C0 , C H and Heliox . These gases are not presently used in the test 8  2  2  4  bench but are readily available to the system if required. A three-way valve is installed at the rack connecting N supply to the H lines to facilitate system-wide purging. 2  2  All lines that may potentially carry H (such as the H SUPPLY lines to the test bench and the exhaust 2  2  lines) are stainless tubing or grounded, flame-resistant hose. N supply lines are connected via Teflon 2  tubing. Flexible supply and exhaust lines allow for some mobility for the test bench during installation and servicing. All other tubing, fittings and wetted components used in the gas manifold are made of brass, stainless steels suitable for H service [107], and H compatible sealing materials such as EPDM, 2  2  VITON and TEFLON [108, 109]. The test bench uses Swagelok® fittings and valves, Burkert solenoid valves rated for potentially explosive environments, and mass flow controllers from Advanced Specialty Gas Equipment.  Building supplied gases are first passed through step down regulators and relief valves to limit pressure throughout the test bench to below 275.8kPa (40PSI). An emergency (push button) shutoff solenoid valve is installed in the H supply line to provide an easily accessible means to terminate H flow. 2  2  N and H mains are each split into three feed lines, each feed line is controlled separately by a pressure 2  2  regulator and MFC combination. Some characteristics of the MFCs used on the test bench are.listed in  Table 4.1. The step down regulators immediately upstream from each MFC stabilizes pressure fluctuations in the line that may be caused by disturbances from other equipment connected to the building gas supply. This translates to a more stable output from the MFCs.  s  Heliox is a mixture of He and 0 gas 2  48  (  Table 4.1  Mass flow controller characteristics  H (test gas)  N or Air (test gas)  N (purge gas)  Maximum flow  500sccm  20sccm  50slpm  Minimum flow  lOsccm  0.4sccm  l.Oslpm  2  2  2  (alarm at 12slpm) Accuracy  ±1.0sccm  ±0.2sccm  ±0.8slpm  Repeatability  ±0.25% reading  ±0.25% reading  ±0.13slpm  4.2.a.2  MFCs to LINEOUT  Downstream from the N and H MFCs, the feed lines merge at a mixing panel into two mixed-gas lines 2  2  LI and L2 (see Figure 4.1). Feed lines may be shut off manually using ball valves on the mixing panel. Additional gas-handling devices such as a humidifier may be installed in-Hne with LI. LI and L2 are each fitted with a flash arrestor, a temperature probe and a pressure transducer (from Specialty Gas Equipment, Valax Manufacturing and GP:50 respectively) before being routed to two 3-way solenoid valves which are mechanically and electrically coupled to one another. The coupled operation of these valves is outlined in Table 4.2.  Table 4.2  Operation of line-selecting solenoid valves  Switch Position  LI power  L2 power  To LINEOUT  To Exhaust  ON: (L2)  On  On  L2  LI  OFF: (LI)  Off  Off  LI  L2  The coupled solenoid valves are used as a gas line-selector, allowing gas flows to stabilize in each line before being 'switched' to enter LINEOUT, the supply line for the test chamber (see Figure 4.1).  Solenoid valves heat up significantly when they are powered on. This is not ideal for gas sensor testing as gas temperatures will tend to increase over the course of the experiments. They are, however, fast, readily available, and are easy to integrate with electronics for data acquisition if needed. Solenoid latch valves would have been ideal, but ones that satisfy our safety considerations and budget could not be easily acquired. As a result, the above line-selector mechanism was chosen to minimize heating from  ** Data according to manufacturer specifications  49  these valves. One line is open to the test chamber at any one time unless mechanical overrides on the solenoid valves are activated. In this configuration, valves can be left in the OFF position when testing LI, and can be manually overridden if L2 must be left on for testing over an extended period of time.  4.2.a.3  Purge line and safety alarms  A separate N and compressed air supply line is connected to a MFC that is set to deliver 20slpm of N or 2  2  compressed air for purging while samples are tested in localized, high concentrations of H inside the test 2  chamber. Two combustible gas monitors (Sierra Monitor Corporation) set to alarm at 400ppm H are 2  installed in the immediate vicinity of the test bench and test chamber, and an alarm buzzer is attached to the purge MFC to warn users of insufficient purge flow rates during testing.  4.2.b Test chamber The test chamber consists of a vacuum-fitted stainless steel drum. Pressure transducers, valves and other fittings for mounting and testing sensor specimens are attached to the body of the drum via numerous ports. A photo of this chamber is shown in Figure 4.2. The test chamber can be operated at pressures ranging from 0.027Pa (2 x 10" torr) to 120kPa (3PSIG). Some physical data for the test chamber and its 4  affiliated components are listed in Table 4.3.  Figure 4.2  Test chamber showing t est sensors, pressure transducers and vacuum pumps mounted on numerous ports. 50  Table 4.3 Approximate Volume  Test chamber physical data 6.2 x 10 cm  .  Purge inlet ball valve flow coefficient C  4  3  0.6  TT  v  TC Check valve maximum C / cracking pressure  1.68 / 2.3kPa differential (0.333PSI)  Overpressure outlet plug valve C  1.6  v  TT  v  Vacuum vent device delay / vent timer  15s/4min  Cavity size of 24pin test chip packages  0.965cm x 0.965cm x 0.1143cm  4.2.b.l  n  Vacuum system  ^  The test chamber is fitted with a Varian Turbomolecular Pump and a Multi-gauge controller that provides pressure readout from a number of pressure transducers. Repeatedly evacuating the test chamber to below 0.03Pa (2 x 10" torr) and backfilling with N to close to atmospheric pressure minimizes the 4  2  possibility of having pockets of air present inside the test chamber. The pressure transducers attached to the test chamber ( P  T a  and P  TC  2  in Figure 4.1) are not suitable for H service, and must only be used 2  during the initial evacuation procedures. A gate valve between the Turbo pump and the test chamber isolates the test chamber from potential leaks across the Turbo and Mechanical pumps. The system is also fitted with a vent device which vents the pump side of the gate valve to atmospheric pressure after the pump stops to prevent backflow of oil into the system.  4.2.b.2  Sensor mount  The sensor mount fitting was designed by Tom Vanderhoek at NRC-IFCI, and was built with support from shop personnel there. The fitting features an 8-wire electrical feedthrough and three gas ports to:  1. Route mixed gas (LINEOUT) to mounted sensors 2.  Monitor pressure (mechanical gauge) in the test chamber during sensor testing  3. Exhaust H gas mixtures from the test chamber to the H vents 2  2  The fitting is designed with an optional cap that closes off the fitting from the rest of the chamber and allows the fitting to be used as a nested chamber. This option is not used in the present study as all experiments are performed at close to atmospheric pressure. A check valve and a manual plug valve (TC  DatafromSwagelok Company 51  CHECK and TC OVERPRESSURE in Figure 4.1) connect the test chamber to the EXHAUST line. These are installed to keep the system from holding substantial positive pressure.  On the inside of the fitting is a spring-loaded test socket for inserting standard Dual In-line Packaged chips up to 24pins. A maximum of 8 pins can be connected at any one time to the electrical feedthrough wires for connection to external circuitry. A short piece of polyurethane hose with an inner diameter of 1.59mm (0.0625") connects the gas inlet (LINEOUT) to a slit nozzle (rectangular orifice) mounted on the sensor. A diagram illustrating the nozzle and sensor configuration is shown in Figure 4.3. CFD simulations of gas flow through a slit nozzle were done by Yanghua Tang at NRC-IFCI. The model structure is shown in Figure 4.4. Simulations using estimated inlet gas velocity of 1 m/s blowing into the sensor cavity show flow and pressure uniformity across the sensor surface. Results of these simulations are illustrated in Figure 4.5 and Figure 4.6.  (side view)  Gas In  Gas entrance through slit  Gas Out 4 (To Test Chamber)  1 i s i i l i '.-lev,-;  Sample in DIP (end riew} Figure 4.3  Diagram of slit nozzle. Gas entering through the port at one side of the nozzle is distributed across a slit orifice close to the sample.  52  4mm  the centre of the bottom Figure 4.4 Simulation model structure for slit nozzle. Gas enters through the circular inlet on the left and exits to the right through a slit. Samples would be located at the bottom surface (extrusion) of the model.  U - m/s 0BS72  -0 0856  Figure 4.5 Vertical section of simulated gas flow through a slit nozzle. Flow velocity is uniform across much of the sensor cavity. Sensors are attached at the centre of the bottom surface.  P - IM/m^ 50  -2<H -23.63  Figure 4.6  Horizontal section of simulated pressure in a slit nozzle. Pressure appears to be uniform across the sensor cavity. Sensors are attached at the centre of the bottom surface.  53  4.2.D.3  Operation of the gas manifold and test chamber  The test chamber serves as a large N reservoir in which high-concentration H gas mixtures from the gas 2  2  manifold are introduced through a small slit nozzle (see Figure 4.3), blowing directly onto test sensors before exiting and entering the rest of the chamber. This enables gas sensors to be tested in a small, localized 'pocket' of test gas within an otherwise inert reservoir. Test gas mixtures entering through this nozzle are controlled by H and N (or Air) MFCs described in section 4.2.a. 1, while purge N is forced 2  2  2  into the test chamber at 20slpm through a separate inlet (TC PURGE in Figure 4.1).  A check valve (TC CHECK) is installed close to the H inlet and a plug valve (TC OVERPRESSURE) is 2  installed at the top of the test chamber to pipe exhaust gases (a mixture of purge N and test H ) to the 2  2  building H -safe vents. The test chamber is maintained at a slight positive pressure to ensure air from the 2  outside does not migrate into the test chamber while H testing is in progress. The purge MFC alarm is 2  set to buzz at N flow rates below 12slpm based on the following calculation: 2  . /  \  max(a>„)  Where min((p „ ) is the lowest acceptable purge flow rate, max(cp ) is the maximum H flow rate, and p  LEL  H  rge  H  2  is the lower explosive limit of H in air, about 4% H . 2  2  By maintaining this minimum flow rate, and by having exhaust outlets close to areas where H is likely to 2  be present at higher concentrations (close to the H inlet and near the top of the test chamber), we ensure 2  that incoming H gas will be sufficiently diluted and will not accumulate to dangerous levels inside the 2  test chamber.  4.2.c Data acquisition electronics H sensors inside the test chamber can be accessed electrically using 8 TEFLON insulated feedthrough 2  cables. A number of options are available for powering the sensors and recording their response characteristics. A voltage-logging program was written in LabVIEW to monitor and record voltage readings from a data acquisition card (DAQ, National Instruments PCI 607IE) to ASCII files, and a simple resistive circuit was built to power the sensors.  /  54  \ 4.2.C.1  Sensor excitation and signal detection circuitry  A simple, battery-operated resistive circuit is used to power test sensors and connect them with the data acquisition board. The main function of the circuit is to provide and limit the power applied to the sensors, and protect the data acquisition board from unexpected shorts and other sensor failures. This DC excitation circuit uses rechargeable NiMH batteries to power the sensors. Rechargeable batteries are also used to power current amplifiers that measure current through the sensors. The use of batteries and shielded cabling lowers the peak-to-peak noise detected by at least an order of magnitude. Background noise in current measurements of this data acquisition unit was around ± 0.6nA.  A schematic for the circuit used for sensor excitation and measurement is shown in Figure 4.7.  • DAQProbe  Battery .Sensor.  100  CurrentArnp (o) . . . , 0  Figure 4.7  Excitation and DAQ interface circuit to power resistive sensors and measure changes in resistance.  Resistance values in the circuit are modified based on the sensors' initial resistance to optimize the detected signals. Typical values of R used to test early versions of fabricated sensors with low resistances are on the order of 10Q. DAQProbe and CurrentArnp are probes connected either directly or via a current amplifier to the DAQ.  4.2.C.2  LabVTEW data logging program  A LabView data logging program was developed to monitor and record time and voltage measurements from a National Instruments PCI-6071E data acquisition card. The main objective for developing this program is to create a DAQ module that allows users to modify the rate at which data is logged without the need to stop the acquisition, reconfigure, and restart settings on the DAQ board. This was deemed necessary to maintain an accurate time-stamp on the recorded data. The program records time measurements extracted from the DAQ hardware, which should be accurate to the DAQ card's  specifications and not dependent on the control computer's system clock or processing load. Such a module is useful for logging data in systems where both transient and steady-state information exist at intermittent intervals. Gas sensor testing is a classic example of such a system.  The resulting module's control panel permits users to enter the file name to be saved, the desired sampling frequency of the DAQ hardware (i.e. the maximum sampling frequency), and the number of points sampled for each data point to be logged and saved. Users may also enter the edit mode and increase the number of physical channels to monitor. At the present time, no processing is performed on the acquired (raw) data, though simple signal processing functions can be easily added into the program module that extracts the logged data.  The program outputs a text file containing a header specifying the device, date, start time and scan rate of the DAQ hardware. A scrolling chart on the control panel allows users to monitor the measured signals. The acquisition start time is displayed on the control panel, as is a numerical monitor of the amount of data 'backlogged' in the acquisition board buffer. This monitor is important for ensuring that the acquisition settings do not overwhelm the computer's available memory space.  An example of data acquired using this DAQ setup is shown in Figure 4.8. This sensor was mounted in a commercial gas sensor test station (Greenlight Power Systems) and is electrically connected to a simple voltage divider. The voltage signal measured across the sensor is normalized with the open circuit (supply) voltage. This sensor was fabricated with a single layer of PMMA resist, and is marked by a large number of defects along the Pd wires. The fabrication process for this sensor' was described in Chapter 1, and observations on its test results will be described later in Chapter 1.  Sensor Response (W04_3N25AB) Run2 f  1.004  100% 80%  g, 1 000 [ii"" irfijiptjnfn|tiiT >  •a  60%  0.996  I  - Signal  M  -H2  |  0.992  Z  0.988 -I 0% 0.984  Figure 4.8  20  40  60 80 Time (s)  100  120  140  Sample measurement obtained from LabvTEW-controIled DAQ setup  Screen outputs of the LabVIEW control panels and the source-code diagrams are included in Appendix A.  4.3 Commissioning and testing with the test bench 4.3.a Test bench commissioning At the time of writing, the H sensor test bench had just commenced operation. Existing MFCs were 2  calibrated against new units from the manufacturer, and resulting calibration charts are included in Appendix B. Relief valve settings were calibrated using an N MFC with a readout accuracy of 0.2sccm 2  and a digital pressure gauge (Cole Parmer) with an accuracy of 0.69kPa differential (0.1PSI). This calibration was necessary to minimize leakage across the relief valves below 380kPa (40PSIG). The test bench is leak tested to 380kPa (40PSIG). There are known, small leaks across two valves in the system into the H exhaust vent. The leak rate is sufficiently small and is satisfactory for H testing to proceed. 2  2  The volume of hose connecting the LINEOUT solenoid valves (LIS and L2S, refer to Figure 4.1) to the test sample is estimated to be 221cm ± 4cm by measuring the volume of gas flow through the line (using 3  3  MFCs) to increase the line pressure by 101.3kPa. The use of building-supplied flame-resistant hoses and present equipment fixation requirements at the laboratory account for the use of such a large volume. At present, estimated system response time with H MFCs fully open (500sccm) is approximately 27s. This 2  test volume will drop substantially when alternative provisions and modifications are made for the test bench.  »  4.3.b Testing with the test bench There are numerous ways of mixing gases for sensor testing [110]. The test system described in this chapter uses a dual-line buffer system where a stabilized mixed-gas line is fed to the test chamber while another is routed to the exhaust for concentration and flow rate 'preparation'. Coupled solenoid valves are then used to switch the two lines. MFC settings can be chosen to match the total flow in each line. This results in less disturbances in the test gas. (  Total flow from each line serves as the starting point from which the flow rates of component gases are calculated and set, that is,  <Pi=C O t  (4.2)  i.  where 57  (4:3)  (p is the flow setpoint of constituent gas /, C, is the gas' concentration and O is the total flow rate in the ;  line. The present selection of MFCs is optimized for two types of tests: one in which small changes in high concentrations of H gas can be delivered (useful for fine resolution gas testing at high 2  concentrations), and another in which high concentrations of H gas flow alternates with low H 2  2  concentration flow.  Initial sensor response data from this test station is shown in Figure 4.9 and Figure 4.10. Changes in resistance were observed using this test bench for one of our wire sensors described earlier in this thesis (Figure 4.9). Testing was performed by Weimin Qian and Cheng Hu at NRC-IFCI. Results from the test bench are compared to response obtained from a commercial high flow rate test station (Greenlight Power Technologies), an example of which is shown in Figure 4.11. Figure 4.10 and Figure 4.11 are response curves obtained from the same acquisition unit testing the same sample. The sample was a MOS sensor obtained from a different project [111]. Average response times for both test facilities are obtained from testing and comparing the response of a number of MOS sensors, and results are tabulated in Table 4.4. The observed response time for the test bench is less than our original estimated time by about 10s, which suggests an inaccurate test volume measurement that can be attributed to system leaks.  Test Bench Response (W05_2N24AB) 149.6 -— Resistance  .g. 149.4  — H2 %  .c O — 149.2 a>  o £  |  149.0  98%  .1  + 96%  g  94%  g CM  92%  "55 0)  * 148.6 148.8  100%  0  u c  X  90% 200 400 600 800 1000 1200 1400 1600 1800 2000  Time (s) Figure 4.9  Sample response of a wire sensor fabricated with bilayer resist at very high concentrations of H . 2  58  Test bench sample response 100 ° Capacitance response — H2 %  80  rr 9.50E-11 g  e 60  9.30E-11  c  (0  o  9.10E-11 H  O  8.90E-11  40  Q-  g o o  CM  + 20 8.70E-11 -| 8.50E-11 200  400  600  800  1000  Time (s)  1200  1400  1600 1800  Figure 4.10 Sample response of a capacitor sensor tested with the test bench. Response of these sensors are compared with response of the same sensors tested with a commercial test station.  Greenlight Test Station sample response 100  1.61E-10 1.6E-10 1.59E-10 £1.58E-10  81.57E-10 |l.56E-10 21.55E-10 O1.54E-10 1.53E-10 1.52E-10 1.51E-10 600  Time (s) Figure 4.11 Sample response of the same sample as in Figure 4.10, tested in the Greenlight commercial sensor test station.  Table 4.4  Average response times for commercial test station and test bench Greenlight Test Station  Test Bench  Total Flow rate  lOOslpm  400sccm  Average Response Time  13s ± 7s  16s ± 9s  59  The primary difference between the operation of the sensor test bench and the commercial test station is the amount of manual operation required, and the range of flow rates and flow concentrations achievable. The Greenlight test station features automatic control and adjustment of gas flow rates in a mixture given the total flow and corresponding % of constituent gases. For the test bench, however, individual flow rates for gas mixtures in each line (out of two) must be 'prepared' in advance by entering flow setpoints manually into the flow controller. When the resulting mixed gas stream has stabilized in one line, it can be switched over to flow to the test chamber for sensor testing using a line-selector.  The Greenlight test station is designed for high flow rate applications, however, and is unable to provide accurate concentration changes of less than 2 % of total flow. Minimum flow requirements for the MFCs also limit the use of the Greenlight in trace concentration testing, as in the case of H leak tests (trace 2  concentrations of H ), and fuel stream monitoring tests (trace concentrations of N or air). Outside of 2  2  these concentration ratios, the Greenlight test station offers excellent versatility and flexibility for sensor testing. It is exactly at these concentrations that the operation of our test bench is optimized. In this way, the two test facilities complement one another in terms of testing capabilities. When modifications to minimize the test bench system response times are implemented, the test bench will also provide a means to measure much faster sensor response than what is possible with the present systems.  60  Chapter 5  Results and Analysis  This chapter reports on test results for gas sensors of which the fabricated by the process described in Chapter 1. A test bench, described in Chapter 1, was developed as part of the thesis work. Due to time constraints, however, the sensors developed have not been extensively tested with this system. Instead, sensors were tested in a commercial high flow rate sensor test station (Greenlight Power Technologies G7830 Gas Sensor Test Station). Sensor testing is performed by Shuozhi Yang and Cheng Hu at NRCIFCI. Sensors are tested in H /N mixtures at a total gas flow of lOOslpm at atmospheric pressure and 2  2  room temperature. The data acquisition unit used in conjunction with this test station is a commercial . impedance analyser (Solartron SI 1287 and SI 1260), although some early testing was performed with a LabVIEW controlled DAQ station described in the previous chapter. The DAQ station was developed to capture sub-millisecond changes in DC signals that is beyond the output time-resolution of the Solartron impedance analyser. After initial tests of these sensors, however, it became apparent that the physical response of the Greenlight test station is much slower, allowing the more refined Solartron unit to be used for later tests.  5.1 Test results 5.1.a Early sensors The original plan for sensor development was to investigate the use of an alternative top-down processing method to generate mesowires that operate based on the percolation mechanism described in the introductory chapters. This was desirable since the use of top-down fabrication techniques permit greater ease of sensor integration with electronics and other system-components. In addition, it has the potential to facilitate simpler alloying of the sensing material. Sensors fabricated early on consisted of arrays of 20 to 30 wires, and were fabricated using a single layer of PMMA as the e-beam resist. It was clear from initial testing that the sensors were indeed sensitive to H , and two general types of response were 2  obtained. These are shown in Figure 5.1 and Figure 5.2. Images of sensors corresponding to these tests are shown in Figure 5.3 and Figure 5.4  61  Signal Response (W04_4N01CD) 10  100 80  E  .c  60  0)  o c re *-» t/>  5 '> a>  7 4  1000  2000  3000  Time (s) Figure 5.1  o  c  40  <u o c o O  20  X  0  4000  5000  Resistance response from an intact array of many Pd wires. A very slow and gradual response was observed, in addition to a large hysteresis.  Sensor R e s p o n s e (W04_3N25AB) Run2 1.004  -i  1  -r 100%  1  1.000 cs  80%  c  60%  £  •4-*  > 0.996 •a at N  =  E Z  -— Signal  s  — H2  0.992  o  40%  0.988 0.984  Figure 5.2  T  4- 20%  20  40  60 80 Time (s)  100  o  120  140  o  CM I  0%  Voltage across defect-ridden Pd wire array. Response is indicative of percolation effects. Wires that are open circuits in air become conductive at 100% H . 2  62  Figure 5.3  Intact wire array sample corresponding to test results shown in Figure 5.1  Figure 5.1 shows the response of a sensor with few defects along the Pd wires (Figure 5.3), and Figure 5.2 shows one with many defects (Figure 5.4). The former shows resistance measurements recorded by the Solartron AC Impedance Analyser, using a probe voltage of lOOmVpp at 1kHz. The latter shows DC voltage measurements from the DAQ station, which can be easily configured to determine whether sensors exhibiting infinite resistance would become conductive when exposed to H . Voltages are 2  normalized to the open circuit voltage. Resistance measurements from intact sensor wires consistently show a large hysteresis, in which the final resistance after purging of H remains at much higher values 2  63  than prior to the wires' exposure to H . Observations and analysis of these response curves will be 2  described in later sections.  Defect-ridden sensors, though exhibiting behaviour that resemble percolation-type operation (lattice expansion in wires causing wire segments to connect and become conductive, refer to section 2.4), generated response that were too unstable, and tended to fail readily during H tests. SEM images of 2  these sensors after H tests typically showed portions of the sensor wires had melted. Response showing 2  this instability and breakdown from the same sensor as that in Figure 5.2 is shown in Figure 5.5 and Figure 5.6, tested immediately afterwards.  Sensor R e s p o n s e (W04_3N25AB) Run3 1.000  100%  0.998  80% Signal  £  0.996 -j  • H2 concentration  60%  •a N  re 0.994  40%  o  20%  E Z  0.992  £  c u c o U  0%  0.990 100  Figure 5.5  c o  200  300  400  500  Time (s)  600  700  800  900  Unstable response of the defect-ridden many-wire array of Figure 5.4. 0-100% H Step response of this sensor was shown in Figure 5.2 2  64  S e n s o r Failure (W04_3N25AB) Run4 100%  11  80%  o 0.998  co  c o  CO +-»  £  60%  0.996  ^  c  TJ Q> N  V  're 0.994  E i—  Signal  o z  -H2  0.992  o  O O  40%  CM X  20% 0%  0.99 200  400  600  800  1000  1200  1400  Time (s)  Figure 5.6  Failure response of the defect-ridden sensor of Figure 5.4. This test was performed immediately after the sensor test shown in Figure 5.5  Response from intact wires, on the other hand, seemed sluggish, and changes at high H concentrations 2  were not easily discernible through the system noise. Despite the large background however, some of these sensors exhibited a secondary, smaller response that followed changes in concentration closely. An example of this small signal response is shown in Figure 5.7. This sample consists of an array of 32 Pd wires that were each approximately 1 pm in width, 200nm thick and spaced 2um apart.  Sensor R e s p o n s e (W04_4N01AB) Run2 (detail) 100  6.8  ^  0)  o c re (A 0)  + 80 g_  6.75  c o 60  *  + 40  c o c o U  6.7  6.65 4  20  CM X  0  6.6 1500  2000  2500  3000  3500  4000  4500  5000  5500  6000  Time (s)  Figure 5.7  Small signal dependence of resistance on H concentration: very small changes. 2  65  To amplify these changes in resistance, the overall resistance of the wire sensors must be much higher. This early work led to subsequent sensor designs which consist of 5 or less patterned wires across the contact pads. Bilayer resists were also used to achieve thinner wires to increase the overall resistance and improve the signal-to-noise. The resulting response of these sensors will be described in section 5. I.e.  5.1.b Failure mechanisms The defect-ridden sensor in Figure 5.4 was fabricated with a single PMMA resist layer. The sensor was patterned with a gap running down the centre perpendicular to the wires. This sensor illustrates a typical example of early sensors that showed good initial H2 response: they generally correspond to ones where a large number of defects exist along the wires. In the case of the sensor shown, some of these defects were introduced by patterning lOOnm gaps during the e-beam lithography process, while others were simply results of poor liftoff and processing. The nature and the origin of these defects point to two failure modes of these sensors:  -»  1. Pd lattice does indeed expand and contract upon exposure and removal from H . However, the 2  extent to which subsequent expansion occurs is insufficient to reconstruct a continuous, conductive wire. This failure mode is the most common among all sensors tested. Sensors that suffered from this mechanism were generally functional immediately after their H tests, but 2  became open circuits the following day, presumably after hydrogen atoms have desorbed completely from the a-hydrides region and the Pd host lattice has contracted. This may be due to the release of residual tensile stresses in the sputtered Pd during lattice expansion and contraction, causing the contracted wire segments to be shorter than pre-H exposure lengths. 2  2. Defects that were not the result of direct patterning point to poor adhesion of the Pd film to the oxide surface. These defects may have originated from strong (thick) Pd layers that stripped portions of the patterned wires during liftoff, or they may simply have resulted from poor surface quality of the oxide.^Proper adhesion of the wire to the surface of the substrate is critical since delamination from the substrate causes the sensor to overheat and melt at much lower probing voltages. This was illustrated by heat transport models described in section 2.5.d. A sensor that shows the effects of overheating is shown before voltage measurements in Figure 5.8 (intact), and after sensor excitation in Figure 5.9 (damaged). Irregularities on the oxide surface are clearly seen in this example, and correspond to the exact locations where overheating from delamination occurs.  66  Figure 5.8  Sample with imperfections on the underlying oxide layer, as fabricated. The compromised region is the darker patch of oxide between the contact pads, to the right.  SE  Oa-May-06  061241 WD13.3mm 20. O k W l !2k° °2°5um  Figure 5.9 Failure of the sample of Figure 5.8. Portions of wires spanning oxide with imperfections can be seen to have melted. The adhesion between the wires and the oxide layer at these areas was compromised, which led to wire delamination and overheating.  A third failure mode is suspected to exist for these sensors, in the case where the Pd lattice expansion in the presence of H2 causes wire structures to buckle (and consequently delaminate). This failure mechanism may explain the response behaviour of sensors that become open circuits during the initial exposure to H (but were otherwise intact when tested in air). 2  67  Thick wire structures from single-layer resist liftoff that had good adhesion to the substrate were tested to withstand excitation voltages up to 7VDC, comparable to that predicted by temperature simulations discussed earlier. Subsequent work with bilayer resist samples resulted in a similar voltage limit before changes in the wires' resistance were observed. Figure 5.10 shows a Joule heating I-V curve for a 25wire sample fabricated with bilayer resist.  Joule heating of Pd wires 0.08 0.07 0.06 < 0.05  I 0.04 1—  § 0.03 0.02 0.01 0  10  12  Voltage (V) Figure 5.10 I-V curve for Pd wires with good adhesion to the oxide substrate. Trend line is a linear regression of I=0.008*V based on data up to 6.25VDC.  5.1.c Sample response curves Response characteristics of the bilayer resist sensors are generally similar to that of the single-layer resist samples. The sensors themselves generally survive 2-3 rounds of testing. This compares favourably to the single-layer resist samples, all of which failed after repeated H cycling during their first H test. 2  2  General response shape is similar across all wire sensors fabricated using the bilayer resists regardless.of structure. These sensors appear to be less prone to the delamination-caused overheating mentioned earlier, except for the sensor structure where a Pd platform exists between the contact pads, shown earlier in Figure 3.5. Failure points for this set of platform sensors were generally at the edge of the thick, Pd coated, crosslinked PMMA platform, where portions of the connecting Pd wires are not in contact with the oxide layer. 1kHz AC probing voltages are kept at lOOmVAC to minimize temperature effects while 68  maintaining reasonable signal-to-noise. Typical response curves for these sensors with their corresponding W concentrations are shown in the following Figures. 2  Sensor Response (W05_2N16AB) 100  370 360 H  80  c o k_ •*-» C <d  40 £  "o u  + 20 300  1000  2000 3000 Time (s)  4000  CM  5000  Figure 5.11 Response from bilayer PMMA/PMGI resist sample. This sample is one with a platform structure. See text for circled property  Figure 5.12 Step response portion of Figure 5.11 in detail. See text for circled properties  Resistance Reponse (W07_2D2AB)  Time (s) Figure 5.13 Response from sample fabricated with the single PMMA resist process, 5 wire structure. See text for circled property  Resistance Reponse (W07_2D2AB) (detail)  770  100 80  o  a) u c co  60  750  + 40  740  20  Signal -H2  730  a> u  c .—  2500  3000  3500  4000  o  c  *9,  —  a> O) O  "D X  0  720 2000  c O  en  a) or  c o  4500  Time (s) Figure 5.14 Step response portion of Figure 5.13 in detail.  70  Figure 5.15 Repeated testing of a platform sensor.  Several general observations could be made from the sensors' response data and tests:  1. The shape of the response curves are similar regardless of geometry of Pd wires. 2. H response of these Pd structures appear to be a superposition of two or more response 2  mechanisms (marked with their corresponding letters in the Figures):  a. A large step increase between 0% and 10% H (covering the entire [H]/[Pd] range for a2  hydrides and portions of the (3-hydrides), followed by a decay that disappears after the first few 0-100% H cycling, 2  b. An increase in resistance as H decreases from 20% to 0%H , causing a hysteresis or (in 2  2  some cases) permanent change in the sensors' resistance. c.  A smaller step response that follows closely the changes in H concentration. 2  d. In a few of the sensors tested, a step decrease in resistance as H concentration increase. 2  3.  Starting resistances of each test for a given sample increase from repeated H testing  4.  All concentration steps beyond 10% H should be well into the disordered fj-hydrides-only region  2  2  according to established (p-C) isotherms, but R/Ro values are much less than reported resistance ratios for P-hydrides. 5. Time response of bilayer resist samples is several times faster than response of single-layer resist samples. Further testing is needed to confirm and determine this relationship, as only two singlelayer resist samples were tested with the present test protocols, and both had relatively poor signal-to-noise. 6.  Signal-to-noise is low for samples with many wires in parallel. This is expected since placing wires in parallel lowers the overall resistance of the wires. It is difficult to detect resistance changes when the overall resistance is very low, as these changes become comparable in magnitude to system noise.  7. The platform sensor design shown in Figure 3.5 was observed to be more robust and reliable than non-platform wire sensors, in that effects from imperfections along the wiresare averaged out over sections of parallel wires that are in series with one another. This maintains a large overall resistance across the sensor, and translates to better signal-to-noise that is less dependent on sensor imperfections.  Close inspection of the small signal step response led to conclusions that this change in resistance could be logarithmic. This is of interest since the (p-C) isotherm in our region of interest (high [H]/[Pd] ratios) i can be approximated by a logarithmic relation. Table 5.1 shows some values extracted from (p-C) isotherms reported in literature.  72  Table 5.1  Reported (p-C) isotherm values of p Pd hydrides from literature  Concentration  Reference  Temperature  [H]/[Pd]  (% H at atmospheric pressure) ratio  (K)  2  1%  0.62  293K  [94]  10%  0.68  293K  [94]  100%  0.73  293K  [94]  10%  0.63  323K  [95]  100%  0.68  323K  [95]  By inspection of reported data in literature, [H]/[Pd] values for H concentrations of over 10% in an inert 2  carrier gas can be estimated with the above data by  [H]/[Pd]  = A + Bln[p]  (5.1)  where [P] is % H from 0 to 100. Table 5.2 summarizes values for these coefficients at 293K and 323K. 2  Table 5.2  Summary of pressure interpolation constants for determining [H]/[Pd] ratios  Temperature (K)  A  B  293 K  0.621± 0.006  0.025 ± .003  323K  0.58 ± 0 . 0 2  0.022 ±.007  We will see in the following analysis section that resistance changes, in [3-hydrides follow a similar logarithmic behaviour.  5.2  Response analysis  5.2.a Small signal step response The small signal step response from each sample was extracted by measuring the step sizes in the overall response, integrating them, then using fitted values to generate estimated resistance step responses from H concentrations. The estimated response is further translated such that the response has a reference 2  value of 0 for all H concentrations at or below 10%. R , ps is the portion of normalized resistance 2  s  e  attributed to the small signal response. A sample fit of this small signal response is shown in Figure 5.17 and is fitted with the function  73  tf,  = C l + C2-ln[P]  (5.2)  +  ^ 30  1 40  1 50  1 60  1  1  70  80  :  |  |  90  100  H2 Concentration (%) Figure 5.17 Logarithmic Fit of small signal resistance measured as a function of H concentration. 2  The resulting small signal step response fit is shown with a vertical offset in Figure 5.18 over the normalized response. This signal can be subtracted from the overall response. The leftover background is shown in Figure 5.19 and is mostly smooth for the majority of sensors tested. Deviations exist at datapoints immediately after each concentration change, as the delay in system time response was not taken into account during the subtraction. This delay is typically 20s and corresponds roughly to the test station's system response time.  0  1000  2000 3000 Time (s)  4000  ' Figure 5.18 Sample response along with small signal component for H concentrations over 10%. 2  74  I 1000  I 2000'  I 3000 Time (s)  I 4000  I 5000  Figure 5.19 Sample background response left over after small signal component is removed.  More observations of the response were made over the course of fitting data in this manner:  1. The background resistance change is attributed to more than one irreversible response mechanism that may include physical damage (delamination and melting) during H absorption. The 2  traditional R/RQ ratio, as such, is not a meaningful normalization for comparison of fitted coefficients across different samples. Instead, the stabilized resistance at 100% H for each 2  sample, just prior to desorption, was chosen as the reference resistance (R  re  f)  with which the  sensor response is normalized. Figure 5.20 shows the R f/Ro ratios for data presented in this re  analysis. The average value of this ratio is 1.2 ± 0.2, and is extracted from the data simply for rough comparisons with literature reported values. This ratio is clearly much lower than the reported maximum R/Ro values of 1.8 reported in literature for atmospheric H . The samples 2  marked with t before the identifying labels on the x-axis representfilm-likesensors that were the result of poor (bilayer) liftoff. The sample marked with * represents a many-wired sample fabricated with single-layer resist. Results from these sensors are not included in the average.  75  Reference Resistance/Initial Resistance  Test Samples Figure 5.20 Reference resistances for samples analysed with respect to their initial resistance. Samples marked with f before their labels did not liftoff properly. Sample marked with * was an older, many-wire sensor with compromised signal-to-noise. These sensors (the three on the right) were not used to calculate averages.  2.  Small signal step response was set to 0 at concentrations below 10% H , as the [H]/[Pd] vs. (In 2  p[H ]) approximation (Equation (2.2)) is not valid at lowr [H]/[Pd] ratios in the P-hydride region. 2  There is, as a result, an additional uncertain (but constant) portion of the background response attributed to the P-hydrides. 3.  The observed resistance increase during desorption through 20% to 0% H clearly exhibits a 2  hysteresis (about -5,.5 x 10" for the sample shown) in the a and a-P regions. 3  4.  Absorption response of resistance for P-hydrides can be considered linear to the hydrogen atomic ratio, but not so for desorption. This is in clear agreement with Lewis' summary [20] and less apparent with Sakamoto's [95] results, in which the slope of R/Ro decreases greatly at close to 40% H . It is important to note however that Sakamoto's results were reported for testing at 2  323K, while sensors were tested at room temperature for the present study.  Fit coefficients extracted are summarized in Figure 5.21 and appear to be roughly constant among samples analysed. The average value of C/.and C2 are listed in Table 5.3. As in the case for the R /R-o re  plot shown in Figure 5.20, averages are taken only from bilayer resist samples that went through successful liftoff, and exclude samples marked with | and * on their labels. 76  Fitted Coefficients of < R s t e = C1 + C2 Liri[P]> P S  .5. c c> o o O  0 02 -i 0 01 0-0 01 -0 02 -0 03 -0 04 -0 05 -0 06 CQ < CD  21 c CN 3 i ce: o  O  T - CN  Z  CN  in o  c  3 i Q:  CD  T— T—  ~Z. <= CN if) o  O  a o  CQ <  T - CN  T - CO  CN  3  CN  Q  Q < CD  3  $  CD CN  m p  CD  ^  i  CN  IT) g  5  Z  LU T -  c  3  I tt. m  co  _  co p  §  p  X  CQ < CN  a  T -  CN  C  P  co  CQ <  O _ p  P  Test Samples  Figure 5.21 Coefficients (arbitrary normalized units) extracted from small signal response fitting. Again, samples marked with f before their labels did not liftoff properly. Sample marked with * was an older, many-wire sensor with compromised signal-to-noise. These sensors (the three on the right) were not used to calculate averages.  Table 5.3  Average values for fit coefficients C2 (a.u.)  CI (a.u.)  (7.9 ± 2.3) x 10"  (-2.4 ± 0.9) x 10"  J  2  Coefficients CI and C2, in conjunction with the coefficients A and B from interpolating the p region (p-C) isotherm in Equation (2.2), relate the normalized resistance to atomic hydrogen content as follows:  [H/Pd]=C +C ~ A  B  (5.3)  where  77  (5.4)  and  It can be seen that the uncertainty in the background for the fitted step resistances will have an effect only in C . Taking into account the average R /Ro ratio of 1.2 ± 0.2, the slope of our R/R vs. [H]/[Pd] curve A  re  0  for P-hydrides then, at 293K, is around 2.7 ± 0.9. This result is higher than reported values in Lewis' summary of about 0.87 [20], and much less than the slope estimated from Sakamoto's results of 8.3 for 323K [95]. Our estimated errors are large, and fitted constants for the same sensor vary with each H test. 2  It is likely that a substantial portion of the errors stem from the need to subtract a background response which, at present, is not well characterized and understood. Other sources of error include changes and differences in temperature, H response of other sensor components, and the response of Pd to N or the 2  2  lack of 0 . Changes in the background resistance were observed during tests when the sensors are purged 2  with N . This was thought to be due to the removal of oxygen atoms from the Pd, and was assumed 2  constant in our analysis. It should be clear, however, that should the signal be the result of sensor crosssensitivity to N , this change in resistance will not be constant. Identification and characterization of these 2  effects will require further tests and studies.  5.2.D Background response The background portion of the response extracted from data analysis typically shows a large step increase followed by a slow decay that stabilizes in time scales on the order of 1000s. Fits of the response with 1/t and exponential decays did not converge well and were not consistent with the measured data. As such, further tests and analysis are needed to determine a mathematical form of the decay mechanism occurring here. The subsequent discussion will present some qualitative interpretations of the background response.  78  5.2.b.l  Lattice and grain boundary effects  The decay portion of the background response vanishes after repeated H cycling, and is likely to be 2  related to dislocations, stresses and cracks in the Pd. Resistivity is generally considered to be inversely proportional to the mean free path of electrons in a metal as follows [112]:  m P = —  v  r  T  •  <-> 5  6  where p is the resistivity, / is the mean free path (effective path) of an electron, e is the electron electronic charge, m, n and Vf are the electron mass, carrier density and Fermi velocity of the conduction electrons. A higher density of dislocations and grain boundaries are introduced in the form of deformations and cracks as the Pd lattice expands and contracts. This translates to a shorter mean free path for conduction electrons before they are scattered by these scattering centres. These scattering effects are amplified as physical dimensions of the conduction structure itself diminish and approaches dimensions comparable to the electrons' mean free path in the material. Microstructural changes from the formation of disordered P-hydrides are permanent upon repeated H  2  cycling [97], and accounts for non-repeatable peaks and decays in the background resistance when H is 2  introduced to a sample for the first time. Lattice stresses'and expansion of Pd hydrides in the stochiometric a region, however, is reversible and repeatable. Electron scattering from these lattice changes and associated volume-induced deformations likely accounts for the repeatable portion of the resistance hysteresis. These effects would be observed in the resistivity of the material at the a and a-P regions regardless of the absorption history of the sample.  5.2.b.2  Percolation effects  In several cases, a large step decrease in resistance was observed during H absorption. The resistances of 2  samples in these cases return to close to their original value on H desorption. The background for such a 2  sample is illustrated in Figure 5.22.  79  Background Reponse (W07 2D2AB)  Time (s)  Figure 5.22 Percolation-like effects in resistance response. Change in resistance due to (a) expansion of Pd wire segments, forming an additional conductive path and (b) contraction of Pd segments.  This behaviour likely originates from percolation effects occurring in the Pd wire structures. The magnitude of change in resistance from a change in the number of identical conducting wires in parallel is as follows:  Ai?  R  AR  N  r  (5.7)  where  N. =  (5.8)  In the above, «/and n, are the final and initial number of conducting wires. For a single wire to two identical wires transition, N reduces the normalized resistance by (a maximum of) 50%. r  80  5.2.b.3  Auxiliary effects  A different process that may contribute to the background response characteristics is the response of other materials present in the device. Although Pd is by far one of the most sensitive materials to H , it is well 2  known that characteristics of Au (such as the work function) and Si0 (such as barrier height at the Pd2  Si0 interface) also change in the presence of H . Such auxiliary processes are worthy of further 2  2  investigation to determine their relevance in our sensor structure.  81  Chapter 6  Conclusion  A resistive Pd nanowire sensor was fabricated to address application needs for on-line monitoring of H  2  fuel cell systems. The objective of the study was to design a device that monitors concentration changes between 90% and 100% H in N . The original work plan included developing a top-down fabrication 2  2  method for percolation-based wire sensors. Top down methods use existing silicon chip fabrication technologies, which facilitate integration of sensors with signal processing electronics. Modifications of material and sensor structure (for response tuning, for example) are also generally easier with these fabrication techniques.  6.1 Test bench development summary In parallel with the sensor development was the design and assembly of a test bench and data acquisition system capable of detecting fast sensor response, since commercial test stations presently available have a system lag time of about 13s. The test bench was commissioned and initial test data comparable to the commercial test station were successfully obtained. Due to time constraints, sensors developed in this work have not yet been characterized in this test bench, and a number of modifications to the test bench are still pending. These include the following:  1. Addition of an extra controller console to take advantage of the two additional gas supply lines to improve versatility of the gas mixing. Present use of these extra MFCs requires physical rewiring of cables and jumper settings on the controller console. 2.  Reducing gas volume between the line-selector and the test sample in order to reduce the system response time to ~ 1 s.  Initial test results obtained from the test bench showed comparable signal results to those obtained from the commercial gas test station. The average time delay observed in these sensor signals was about 16s ± 9s, while the same sensors tested in the commercial test station showed an average delay of about 13s ± 7s. Compared to the commercial test station, the test bench requires much more user involvement in gas routing and settings, as the majority of controls are manual. The test bench setup features a lineswitching mechanism which, when the system is optimized, has the potential to supply accurate concentration changes of H /N gas mixtures much more quickly than existing commercial test stations. 2  2  i  82  6.2 Sensor development summary Sensors developed in this thesis use pure Pd sputtered onto patterned PMMA or PMMA/PMGI resists for liftoff. The use of PMMA/PMGI bilayer resists resulted in less defects in sensor wire structures, and the developed fabrication protocols were more reliable than those using a single PMMA layer for liftoff.  Wire thicknesses of bilayer samples were on the order of 80nm and were much thinner than those fabricated with a single layer of PMMA (about 200nm). The use of bilayer resist introduces a 'deposition stop', in which metallization accumulating at the upper resist layer closes off the layer's opening, and no further thickness is added to the patterned cavity. Further work in this area may allow post-liftoff thickness of the structures to be accurately controlled and replicated depending on the electron beam dose and development times of the resist layers.  Multi-wire structures were fabricated and tested in a commercial H sensor test station. Results were 2  analysed as a superposition of a fast, small signal response attributed to p Pd hydrides, and a slow, large decay and hysteresis corresponding to the existence / coexistence of a-hydrides in the system. The small signal response can be modeled as a logarithmic function of absorbed hydrogen content, and the slope of the resulting R/R vs. [H]/[Pd] relationship is estimated to be 2.7 ± 0.9 for concentrations of H over 10%. 0  2  It is clear from our study that between 90% and 100% H , the small signal resistance response can be used 2  to monitor changes in H concentration. Despite the existence of a large and relatively slow 'background' 2  transient, the small signal response can be extracted quite easily and follows changes in concentration about as quickly as the present gas test system will allow. Further insights to the time-response characteristics of the sensors may be obtained when pending changes for the test bench are in place. These modifications will allow the system to trigger fast concentration changes and measure sensor response at acquisition rates on the order of 10,000samples/s.  6.3 Significance of results and comparison with literature Pd film resistance measurements are reported in the literature with system settling times varying from minutes to hours [95]. A number of Pd film sensor studies report observing faster response times for higher concentrations of H , but results for concentrations beyond the a-hydrides region are generally not 2  presented, making a direct comparison with these studies difficult. The presence of a large hysteresis over the hydrides phase transition is well documented, and partly accounts for the lack of Pd film sensor developments over concentrations above several percent H . At the present time, a number of groups are 2  investigating the percolation mode of operation for thin film Pd on a variety of substrates (such as 83  polymeric cluster-nucleation promoters [99], anodic alumina [113], etc.), while others are studying Pd alloy compositions to eliminate this phase change entirely such that only the a-hydride forms [25].  Of the resistance response observed in our thin Pd wire sensors, the P-hydride resistance provides the most consistent and repeatable means of determining H concentrations over 10% regardless of exact 2  geometries of the sensors. Comparatively, sensor mechanisms based on percolation is novel, and has a great potential for excellent sensitivities and robustness over low- to mid-H concentrations. Percolation 2  effects in a device, however, are strongly dependent on its fabricated geometry as well as its material properties. This may result in stringent fabrication tolerances in order that accurate response analysis can be performed in sensing applications. Normalized resistance of the P-hydrides at a given temperature, on the other hand, is a material properly that is relatively invariant across sensors regardless of geometry, making its use more forgiving, simple and economical for mass manufacturing. With advancements in signal processing technologies and fuzzy logic algorithms-that compensate for superimposed signals and sensor cross-sensitivities, extracting p-hydride response may prove to be a viable means of measuring high concentrations of H accurately. 2  Our study shows that geometry does not play a significant role in affecting hydride resistance. It does, however, affect the robustness of the sensors. Inserting platform structures spanning across sensor wires improves the robustness of the sensors by averaging out effects of imperfections over several parallel wire segments, while maintaining good signal-to-noise by having these parallel sections in series with one another.  Pd films are known to blister and delaminate from exposure to H . Delamination is a primary source of 2  failure for our wire sensors as well, especially after repeated exposure to H . The emphasis of the present 2  work has been on designing a reliable fabrication process that yields intact wire structures. More sensor tests will be needed to determine whether thin Pd wires in H exhibit better adhesion than thick Pd films. 2  As spatial dimensions of these sensors become smaller, the long response times associated with Pd resistive devices diminish and the initial resistances of our sensors increase. It is generally agreed that smaller physical dimensions translate to larger surface area to volume ratios (more surface coverage) and shorter diffusion lengths for atomic hydrogen migration. This is generally favourable for speeding up sensor response. In addition, we can expect at least a comparable (if not speedier) response for nanoscale hydride resistance measurements with percolation film measurements for high-concentration monitoring. The former involves chemical absorption processes that should be faster than the physical lattice expansion dominant in the latter mechanism. Although there is insufficient data to establish a 84  quantitative correlation between response times and our sensors' physical geometries, thicker Pd wires fabricated from single-layer PMMA liftoff exhibited slower reponse times in our tests, on the order of minutes, as opposed to about 20s for bilayer resist Pd wire sensors.  Wire sensors, as mentioned earlier, also feature higher initial resistance and improved signal-to-noise over comparable fdm sensors. Although characteristic parameters are described in terms of the ratio of resistances, wire sensors with high initial resistance were generally easier to work with. Since the total resistance of fdm and many-wired sensors are on the order of a few ohms, sensor signals (changes in resistance) are easily overwhelmed by background noise from the rest of the system! In one example, noise from a fdm sensor was on the order of half the resistance step size corresponding to the concentration increase from 80% to 100% H . Five-wired sensor devices however, have signal-to-noise 2  ratios that are far superior to those of fdm sensors. As such, it is recommended that wire based patterns such as those presented in this study or serpentine structures be used in place of fdm structures for resistive sensor fabrication.  6.4 Additional recommendations We observed that scaling down of the device caused amplification of dislocation and grain boundary effects, as a-P transition peaks are much sharper and stronger than those observed in many-wired and film samples. No conclusive fit was achieved for the background unfortunately, and further work in identifying physical (expansion) and chemical (absorption and desorption) processes involved in background response studies is recommended.  The sensors studied in this work may be successfully incorporated into sensor packages where processing circuitry exists to extract sensor data froni an array of different sensors' response. The sensor should be integrated with ones optimized for low H concentration measurements (corresponding to a and a-(3 2  hydrides). To achieve H monitoring and detection throughout all H concentrations with a single Pd2  2  based sensing material or structure however, one must address challenges caused by the a-P hydride transition. In this regard, promising progress is noted in Pd percolation films and Pd-alloy films. Some preliminary work on creating a Ni target mask has been performed for the present study, but results are not reported here as this work has progressed only as far as confirming the uniformity and accuracy in Pd/Ni content for resulting sputtered films. As existing reports in literature on alloy studies have already achieved sufficient sensitivity, continuation of alloy studies should place emphasis on reducing response time via design of the sensor geometry.  85  Future work to continue direct patterning of defects along Pd nanowires is not recommended unless access to higher resolution patterning instruments, or alternative deposition methods such as angled deposition systems become available. Studies into depositing wires of thicknesses close to the percolation limit can be pursued, however, especially due to its apparent ease of implemention with existing capabilities. Such future work will benefit from existing literature on nucleation promoters such as siloxane self-assembled monolayers [99]. Work recommended in this area includes controlling deposited wire thicknesses by controlling 'window' (pinhole) widths of developed PMMA in PMMA/PMGI bilayers, as such a technique may facilitate operation range 'tuning', allowing multiple sensors optimized at different concentrations to be fabricated on the same chip with the same fabrication processes.  The sensor work described in this document covers the development of fabrication protocols and records a number of observations in tested sensors. A linear relationship between normalized resistance to hydrogen atomic content in Pd was determined, which agrees in form with reports from the literature, though the numeric values derived from this study was different. Other observations were more qualitative, and several effects, also reported in the literature, were noted. This work represents a small contribution towards identifying the multitude of sensing mechanisms present in very thin Pd mesowires exposed to H gas. While the combination of these mechanisms makes direct implementation of these 2  sensors difficult, continued research and further testing to enhance or suppress selected mechanisms should allow development of robust, high-concentration H2 sensors for fuel system monitoring.  86  References 1.  . Praxair MSDS for Hydrogen gas, http://www.praxair.eom/praxair.nsf/O/eal 79e7fb21 df0858525653c00641810/$FILE/p4604e.pdf  2.  US Department of Energy, A multiyear plan for the hydrogen R&D program, (1999),  http://www.eere.energy.gov/hydrogenandfuelcells/pdfs/bk28424.pdf 3.  Sensistor Technologies, http://www.sensistor.se/e_index.htm  4.  C.H.S. Hitchcock, Determination of hydrogen as a marker in irradiated frozen food, Journal of  the Science of Food and Agriculture 80 (2000), 131-136. 5.  C.H.S. Hitchcock, Determination of hydrogen as a marker in irradiated eggshell, Journal of the  Science of Food and Agriculture 80 (2000), 137-139. 6.  J.W. Marks, Hydrogen breath test, MedicineNet, http://www.medicinenet.com  7.  P.M. Heacock, S.R. Hertzler and B.W. Wolf, Fructoseprefeeding  reduces the glycemic  response  to a high-glycemic index, starchy food in humans, Journal of Nutrition 32 (2002), 2601-2604.  8.  T. Rahn, J.M. Eiler, K.A. Boering, P.O. Wennberg, M.C. McCarthy, S. Tyler, S. Schauffler, S. Donnelly and E. Atlas, Extreme deuterium enrichment in stratospheric hydrogen and the global  atmospheric budget of H , Nature 424 (2003), 918-921. 2  9.  V.A. Krasnopolsky and P.D. Feldman, Detection of molecular hydrogen in the atmosphere of  Mars, Science 294(2001), 1914-1917. 10.  W.V. Boynton, W.C. Feldman, S.W. Squyres, T.H. Prettyman, J. Bruckner, L.G. Evans, R.C. Reedy, R. Starr, J.R. Arnold, D.M. Drake, P.A.J. Englert, A.E. Metzger, I. Mitrofanov, J.I. Trombka, C. d'Uston, H. Wanke, O. Gasnault, D.K. Hamara, D.M. Janes, R.L. Marcialis, S. Maurice, I. Mikheeva, G.J. Taylor, R. Tokar and C. Shinohara, Distribution near surface of Mars:  11.  of hydrogen in the  evidence for subsurface ice deposits, Science 297 (2002), 81 -85.  NASA Press release, Odyssey studies changing weather and climate on Mars, (2003),  http://marsprogram.jpl.nasa.gov/odyssey/gallery/science/PIA04907.html 12.  K.G. Stanley, E.K. Czyzewska, T.P.K. Vanderhoek, L.L.Y. Fan, K.A. Abel and Q.M.J. Wu, A hybrid sequential deposition fabrication  technique for micro fuel cells, Journal of  Micromechanics and Microengineering 15^(2005), 1979-1987. 13.  T. Seiyama, A. Kato, K. Fujiishi and M. Nagatani, A new detector for gaseous components using semiconductive thin films, Analytical Chemistry 34 (1962), 1502-1503.  14.  K. Wada and M. Egashira, Hydrogen sensing properties of Sn0 subjected'to surface chemical 2  modification with ethoxysilanes, Sensors and Actuators B, Chemical 62 (2000), 211-219. 15.  A. Katsuki and K. Fukui, H selective gas sensor based on Sn0 , Sensors and Actuators B, 2  2  Chemical 52 (1998) 30-37. 87  16.  C.H. Han, S.D. Han, I. Singh and T. Toupance, Micro-bead of nano-crystalline F-doped Sn0 as 2  a sensitive hydrogen gas sensor, Sensors and Actuators B, Chemical 109 (2005) 264-269. 17.  O.K. Varghese, D.W. Gong, M. Paulose, K.G. Ong and C.A. Grimes, Hydrogen sensing using titania nanotubes, Sensors and Actuators B, Chemical 93 (2003), 338-344.  18.  ^  Y.K. Jun, H.S. Kim, J.H. Lee and S.H. Hong, High H sensing behaviour of Ti0 films formed by 2  2  thermal oxidation, Sensors and Actuators B, Chemical 107 (2005), 264-270. 19.  S.J. Ippolito, S. Kandasamy, K. Kalantar-zadeh and W. Wlodarski, Hydrogen sensing characteristics of WO3 thin film conductometric sensors activated by Pt and Au catalysts, Sensors and Actuators B, Chemical 108 (2005), 154-158.  20.  F.A. Lewis, The Palladium Hydrogen System. Academic Press Ltd., London. (1967).  21.  H.I. Chen, Y.I. Chou and C.Y. Chu, A novel high-sensitive Pd/InP hydrogen sensor fabricated by electroless plating, Sensors and Actuators B, Chemical 85 (2002), 10-18.  22.  M. Armgarth and C. Ny lander, Blister formation in Pd gate MIS hydrogen sensors, IEEE Electron Device Letters EDL-3 (1982), 384-386.  23.  S. Okuyama, K. Umemoto, K. Okuyama, S. Ohshima and K. Matsushita, Pd/Ni-Al 0 -Al 2  3  tunnel  diode as high-concentration hydrogen sensor, Japanese Journal of Applied Physics 36 (1997), 1228-1232. 24.  K. Scharnagl, M. Eriksson, A. Karthigeyan, M. Burgmair, M. Zimmer and I. Eisele, Hydrogen detection at high concentrations with stabilised palladium, Sensors and Actuators B, Chemical 78 (2001), 138-143.  25.  R.C. Thomas and R.C. Hughes, Sensors for detecting molecular hydrogen based on Pd metal alloys, Journal of the Electrochemical Society 144 (1997), 3245-3249.  26.  Y.T. Cheng, Y. Li, D. Lisi and W.M. Wang, Preparation and characterization ofPd/Ni thin films for hydrogen sensing, Sensors and Actuators B, Chemical 30 (1996), 11-16.  27.  B.S. Hoffheins, L.C. Maxey, W. Holmes Jr., R.J. Lauf, C. Salter and D. Walker, Development of low cost sensors for hydrogen safety applications, Proceedings of the 1999 U.S. DOE Hydrogen Program Review. US Department of Energy. (1999).  28.  ' R.J. Lauf, C. Salter and R.D. Smith II, Development of low-cost hydrogen sensors, Proceedings of the 2000 U.S. DOE Hydrogen Program Review. US Department of Energy. (2000).  29.  B.S. Hoffheins, R.J. Lauf and J.J. Felten, Thick film hydrogen sensor: design and materials, International Journal of Microcircuits and Electronic Packaging 18 (1995), 297-303.  30.  E.C. Walter, F. Favier and R.M. Penner, Palladium mesowire arrays for fast hydrogen sensors and hydrogen-actuated switches, Analytical Chemistry 74 (2002), 1546-1553.  31.  F. Favier, E.C. Walter, M.P. Zach, T. Benter and R.M. Penner, Hydrogen sensors and switches from electrodeposited palladium mesowire arrays, Science 293 (2001), 2227-2231. 88  32.  K.T. Kim, S.J. Sim and S.M. Cho, Hydrogen gas sensor using Pd nanowires  into anodizedalumina 33.  electro-deposited  template, IEEE Sensors Journal 6 (2006), 509-513.  O. Dankert and A. Pundt, Hydrogen-induced  percolation  in discontinuous films, Applied Physics  Letters 81 (2002), 1618-1620. 34. -^  Y. Im, C. Lee, R.P. Vasquez, M.A. Bangar, N.V. Myung, E.J. Menke, R.M. Penner and M. Yun, Investigation of a single Pd nanowire for use as a hydrogen sensor, Small 2 (2006), 356-358.  35.  G. Kaltenpoth, P. Schnabel, E. Menke, E.C. Walter, M. Grunze and R.M. Penner, Multimode detection of hydrogen gas using palladium-covered  silicon u-channels, Analytical Chemistry 75  (2003), 4756-4765. 36.  F. DiMeo Jr., I.S. Chen, P. Chen, J. Neuner, M. Stawasz and J. Welch, Micro-machined  thin film  hydrogen gas sensors, Proceedings of the 2002 U.S. DOE Hydrogen Program Review. US  . Department of Energy. (2002). NREL/CP-610-32405. 37.  A. Tibuzzi, C. Di Natale, A. D'Amico, B. Margesin, S. Brida, M. Zen and G. Soncini,  Polysilicon  mesoscopic wires coated by Pd as high sensitivity H sensors, Sensors and Actuators B, Chemical 2  83 (2002), 175-180. 38.  I. Lundstrom, S. Shivaraman, C. Svensson and L. Lundkvist, A hydrogen-sensitive  MOS field-  effect transistor, Applied Physics Letters 26 (1975), 55-57. 39.  K.I. Lundstrom, M.S. Shivaraman and C M . Svensson, A hydrogen-sensitive  Pd-gate MOS  transistor, Journal of Applied Physics 46 (1975), 3876-3881. 40.  A. Mandelis. and D. Christofides, Physics, chemistry and technology of solid state gas sensor  devices, Chemical Analysis. V. 125. J.D. Winefordner (ed). Wiley. New York. (1993). 41.  R.C. Hughes and W.K. Schubert, Thin films of Pd/Ni alloys for detection of high hydrogen  concentrations, Journal of Applied Physics 71 (1992), 542-544. 42.  C.C. Cheng, Y.Y. Tsai, K.W. Lin, H.I. Chen, W.H. Hsu, C.W. Hong and W.C. Liu, Characteristics  of a Pd-oxide-InO.49GaO.51P  high electron mobility transistor  (HEMT)-based  hydrogen sensor, Sensors and Actuators B, Chemical 113 (2006), 29-35.  43.  K. Tsukada, T. Kiwa, T. Yamaguchi, S. Migitaka, Y. Goto and K. Yokosawa, A study of fast response characteristics for hydrogen sensing with platinum FET sensor, Sensors and Actuators  B, Chemical 114 (2006), 158-163. 44.  T. Hyodo, J. Ohoka, Y. Shimizu and M. Egashira, Design of anodically oxidizedNb 0 2  5  films  as a  diode-type H-2 sensing material. Sensors and Actuators B, Chemical 117 (2006), SL359-366. 45.  S. Roy, C. Jacob, C. Lang and S. Basu, Studies on Ru/3C-SiC  Schottky junctions for high  temperature hydrogen sensors, Journal of the Electrochemical Society 150 (2003), H135-H139.  89  46.  K.W. Lin and R.H. Chang, Comprehensive study of pseudomorphic  transistor (pHEMT)-based  high electron mobility  hydrogen sensor, Sensors and Actuators B, Chemical 119 (2006), 47-  51. 47.  P. Tobias, P. Martensson, A. Goras, I. Lundstrom and A. Spetz, Moving gas outlets for the evaluation of fast gas sensors, Sensors and Actuators B, Chemical 58 (1999), 389-393.  48.  H. Wingbrant, I. Lundstrom and A.L. Spetz, The speed of response of MISiCFET  devices,  Sensors and Actuators B, Chemical 93 (2003), 286-294. 49.  T.L. Poteat and B. Lalevic, Transition metal-gate MOS gaseous detectors, IEEE Transactions on Electron Devices ED-29 (1982), 123-129.  50.  M. Yousuf, B. Kuliyev, B. Lalevic and T.L. Poteat, Pd-InP Schottky diode hydrogen sensors,  Solid-State Electronics 25 (1982), 753-758. 51.  A.L. Spetz, P. Tobias, L. Uneus, H. Svenningstorp,rL.G. Ekedahl and I. Lundstrom, High temperature catalytic metal field effect transistors for industrial applications  Sensors and  Actuators B, Chemical 70 (2000), 67-76. 52.  J. Kim, B.P. Gila, C R . Abernathy, G.Y. Chung, F. Ren and S.J. Pearton, Comparison and Pt/4H-SiC  53.  ofPt/GaN  gas sensors, Solid-State Electronics 47 (2003), 1487-1490.  B.P. Luther, S.D. Wolter and S.E. Mohney, High temperature Pt Schottky diode gas sensors on n-  type GaN, Sensors and Actuators B, Chemical 56 (1999), 164-168. 54.  K. Scharnagl, A. Karthigeyan, M. Burgmair, M. Zimmer, T. Doll and I. Eisele, Low temperature hydrogen detection at high concentrations:  comparison of platinum and iridium, Sensors and  Actuators B, Chemical 80 (2001), 163-168. 55. '  S.K. Hazra and S. Basu, Hydrogen sensitivity ofZnOp-n  homojunctions, Sensors and Actuators  B, Chemical 117 (2006), 177-182. 56.  V. Polishchuk, E. Souteyrand, J.R. Martin, V.I. Strikha and V.A. Skryshevsky, A study of hydrogen detection with palladium  modified porous silicon, Analytica Chimica Acta 375 (1998),  205-210. 57.  N. Gabouze, S. Belhousse, H. Cheraga, N. Ghellai, Y. Ouadah, Y. Belkacem and A. Keffous, C0  2  58.  and H detection with a CHx/porous silicon-based 2  sensor, Vacuum 80 (2006), 986-989.  J.F. McAleer, P.T. Moseley, P. Bourke, J.O.W. Norris and R. Stephan, Tin dioxide gas sensors: use of the seebeck effect, Sensors and Actuators 8 (1985), 251-257.  59.  W. Shin, M. Matsumiya, N. Izu and N. Murayama, Hydrogen-selective  thermoelectric gas sensor,  Sensors and Actuators B, Chemical 93 (2003), 304-308. 60.  W. Shin, K. Imai, N. Izu and N. Murayama, Thermoelectric thick-film hydrogen gas sensor  operating at room temperature, Japanese Journal of Applied Physics 2 (2001), L1232-L1234.  90  61.  M. Matsumiya, F. Qiu, W. Shin, N. Izu, N. Murayama and S. Kanzaki, Thin-film  Li-dopedNiO  for thermoelectric hydrogen gas sensor, Thin Solid Films 419 (2002), 213-217. 62.  W. Shin, M. Matsumiya, F. Qiu, N. Izu and N. Murayama, Li- and Na-doped NiO thick film for thermoelectric hydrogen sensor, Journal of the Ceramic Society of Japan 110 (2002), 995-998.  63.  F. Qiu, W. Shin, M. Matsumiya, N. Izu and N. Murayama, Hydrogen sensor based on RFsputtered thermoelectric SiGe film, Japanese Journal of Applied Physics Part 1 42 (2003), 1563-  1567. 64.  F.B. Qiu, M. Matsumiya, W. Shin, N. Izu and N. Murayama, Investigation of thermoelectric hydrogen sensor based on SiGe film, Sensors and Actuators B, Chemical 94 (2003), 152-160.  65.  M. Nishibori, W. Shin, L.F. Houlet, K. Tajima, T. Itoh, N. Izu, N. Murayama and I. Matsubara, New structural design of micro-thermoelectric  sensor for wide range hydrogen detection, Journal  of the Ceramic Society of Japan 114 (2006), 853-856. 66.  N. Sawaguchi, W. Shin, N. Izu, I. Matsubara and N. Murayama, Enhanced hydrogen of thermoelectric gas sensor by modification  selectivity  of platinum catalyst surface, Materials Letters 60  (2006), 313-316. 67.  A. Mandelis and J.A. Garcia, Pd/PVDF modulated optical-transmittance:  thin film hydrogen sensor based on  laser-amplitude-  dependence on H concentration and device physics, Sensors 2  and Actuators B, Chemical 49 (1998), 258-267. 68.  J.A. Garcia and A. Mandelis, Laser-amplitude-modulated  dualphotopyroelecrtric  / optical-  transmittance hydrogen sensor, Photoacoustic and Photothermal Phenomena: 10th International  Conference. American Institute of Physics. F. Scudieri and M. Bertolotti (eds). (1999), 533-535. 69.  P. Liu, S.H. Lee, H.M. Cheong, C.E. Tracy, J.R. Pitts and R.D. Smith, Stable Pd/V 0 2  Optical H  5  2  sensor, Journal of the Electrochemical Society 149 (2002), H7.6-H80. 70.  R.D. Smith, P. Liu, S.H. Lee, E. Tracy and R. Pitts, Interfacial stability of thin film fibre-optic  hydrogen sensors, Proceedings of the 2002 U.S. DOE Hydrogen Program Review. US  Department of Energy. (2002). 71.  R.Pitts, P. Liu, S.H. Lee and E. Tracy, Interfacial stability of thin film hydrogen sensors,  Proceedings of the 2001 U.S. DOE Hydrogen Program Review. US Department of Energy. (2001). 72.  X. Bevenot, A. Trouillet, C. Veillas, H. Gagnaire and M. Clement, Surface plasmon  resonance  hydrogen sensor using an optical fibre, Measurement Science & Technology 13 (2002), 118-124. 73.  J. Villatoro, D. Luna-Morendo and D. Monzon Hernandez, Optical fiber hydrogen sensor for concentrations  below the lower explosive limit, Sensors and Actuators B, Chemical 110 (2005),  23-27.  91  74.  M. Tabib-Azar, B. Sutaput, R. Petrick and A. Kazemi, Highly sensitive hydrogen sensors using palladium  coated fibre optics with exposed cores and evanescent field interactions, Sensors and  Actuators B, Chemical 56 (1999), 158-163. 75.  B. Sutapun, M. Tabib-Azar arid A. Kazemi, Pd-coatedelastooptic fibre optic bragg grating  sensors for multiplexed hydrogen sensing,>Sensors and Actuators B, Chemical 60 (1999), 27-34.  76.  S. Sumida, S. Okazaki, S. Asaka, H. Nakagawa, H. Murayama and T. Hasegawa, Distributed hydrogen determination with fiber-optic sensor, Sensors and Actuators B, Chemical 108 (2002),  508-514. 77.  -~  S. Sekimoto, H. Nakagawa, S. Okazaki, K. Fukuda, S. Asakura, T. Shigemori and S. Takahashi, A fibre-optic evanescent-wave hydrogen gas sensor using palladium-supported  tungsten oxide,  Sensors and Actuators B, Chemical 66 (2000), 142-145. 78.  S. Okazaki, H. Nakagawa, S. Asakura, Y. Tomiuchi, N. Tsuji, H. Murayama and M. Washiya, Sensing characteristics  of an optical fibre sensor for hydrogen leak, Sensors and Actuators B,  Chemical 93 (2003), 142-147. 79.  M.A. Butler, Micromirror  optical-fibre  hydrogen sensor, Sensors and Actuators B, Chemical 22  (1994), 155-163. 80.  X. Bevenot, A. Trouillet, C. Veillas, H. Gagnaire and M. Clement, Hydrogen leak detection using an optical fibre sensor for aerospace applications,  81.  Sensors and Actuators B67 (2000), 57-67.  I. Aruna, R. Mehta and K. Malhotra, Fast H recovery in Pd nanoparticle  layer based Gd  switchable mirrors: size-induced geometric and electronic effects, Applied Physics Letters 87  (2005),NIL_144-NIL_146. 82.  M. Tabib-Azar and B. Sutapun, Novel hydrogen sensors using evanescent microwave probes,  Review of Scientific Instruments 70 (1999), 3707-3713. 83.  S. Okuyama, Y. Mitobe, K. Okuyama and K. Matsushita, Hydrogen gas sensing using a Pdcoated cantilever, Japanese Journal of Applied Physics 39 (2000), 3584-3590.  84.  H.P. Lang, R. Berger, F. Battiston, J.P. Ramseyer, E. Meyer, C. Andreoli, J. Brugger, P. Vettiger, M. Despont, T. Mezzacasa, L. Scandella, H.J. Guntherodt, C. Gerber and J.K. Gimzewski, A Chemical Sensor Based on a Micromechanical  Cantilever Array for the Identification  of Gases  and Vapors. Applied Physics A, Materials Science & Processing 66 (1998), S61-S64. 85.  H.P. Lang, M.K. Bailer, R. Berger, C. Gerber, J.K. Gimzewski, F.M. Battiston, P. Fornaro, J.P. Ramseyer, E. Meyer and H.J. Guntherodt, An Artificial  Nose Based on a  Micromechanical  Cantilever Array, Analytica Chimica Acta 393 (1999), 59-65.  86.  D.R. Baselt, B. Fruhberger, E. Klaassen, S. Cemalovic, C.L. Britton Jr., S.V. Patel, T'.E. Mlsna, D. McCorkle and B. Warmack, Design and performance  of a microcantilever-based  hydrogen  sensor, Sensors and Actuators B, Chemical 88 (2003), 120-131. 92  87.  C.L. Britton, R.L. Jones, P.I. Oden, Z. Hu, R.J. Warmack, S.F. Smith, W.L. Bryan and J.M. Rochelle, Multiple-input  88.  microcantilever  sensors, Ultramicroscopy 82 (2000), 17-21.  C.L. Britton Jr R.J. Warmack, S.F. Smith, P.I. Oden, R.L. Jones, T. Thundat, G.M. Brown, W.L. v  Bryan, J.C. Depriest, M.N. Ericson, M.S. Emery, M.R. Moore, G.W. Turner, A.L. Wintenberg, T.D. Threatt, Z. Hu, L.G. Clonts and J.M. Rochelle, Battery-powered,  wireless MEMS sensors for  high-sensitivity chemical and biological sensing, IEEE 20th Anniversary Conference on  Advanced Research in VLSI. (1999), 359-368. 89.  C. Wang, A. Mandelis and J.A. Garcia, Pd/PVDF  photopyroelectric  purely-thermal-wave  thin film hydrogen sensor system based on  interference, Sensors and Actuators B, Chemical 60  (1999), 228-237. 90.  " A. Mandelis and C. Wang, A novel PVDF thin-film photopyroelectric  thermal-wave  interferometry, Ferroelectrics 236 (2000), 235-246. 91.  W.P. Jakubik, M.W. Urbanczyk, S. Kochowski and J. Bodzenta, Bilayer structure for hydrogen detection in a surface acoustic wave sensor system, Sensors and Actuators B, Chemical 82  (2002), 265-271. 92.  W.P. Jakubik, M.W. Urbanczyk, S. Kochowski and J. Bodzenta, Palladium  andphthalocyanine  bilayer films for hydrogen detection in a surface acoustic wave sensor system, Sensors and  Actuators B, Chemical 96 (2003), 321-328. 93.  E. Wicke, H. Brodowsky and H. Zuchner, Hydrogen in palladium and palladium  alloys.  Hydrogen in Metals II Application-Oriented Properties. Springer-Verlag Berlin Heidelberg. Germany. (1978). 94.  H. Frieske and E. Wicke, Magnetic susceptibility and equilibrium diagram of PdHn, Berichte der  Bunsen - Gesellschaft fur Physikalische Chemie 77 (1973), 48-51. 95.  Y. Sakamoto and I. Takashima, Hysteresis behaviour of electrical resistance of the Pd-H system  measured by a gas-phase method, Journal of Physics. Condensed Matter 8 (1996), 10511-10520. 96.  A.K.M. Fazle Kibria and Y. Sakamoto, Pressure-composition composition isotherms of a palladium-hydrogen  and electrical  resistance-  system, International Journal of Hydrogen  Energy 23 (1998), 475-481. 97.  J.F. Lynch, J.D. Clewley, T. Curran and T.B. Flanagan, The effect of the a-b phase change on the a phase solubility of hydrogen in palladium,  153-163. 98.  Journal of the Less-Common Metals. 55 (1977),  *  ,  A.L. Cabrera and R. Aguayo-Soto, Hydrogen absorption in palladium films sensed by changes in  their resistivity, Catalysis Letters 45 (1997), 79-83.  93  99.  T. Xu, M.P. Zach, Z.L. Xiao, D. Rosenmann, U. Welp, W.K. Kwok and G.W. Crabtree, Selfassembled monolayer-enhanced  hydrogen sensing with ultrathin palladium films, Applied  Physics Letters 86 (2005), 203104. 100.  D. Lenstra and R.T.M. Smokers, Theory of nonlinear quantum tunnelling resistance in one-  dimensional disordered systems, Physical Review B, Condensed Matter 38 (1988), 6452-6460. 101.  J.E. Morris, Effects of hydrogen absorption on the electrical conduction of discontinuous  palladium thin films, International Journal of Electronics 81 (1996), 441-447. 102.  CRC Handbook of Chemistry and Physics.  103.  S.M. Shivaprasad and M.A. Angadi, Temperature coefficient of resistance of thin palladium  films, Journal of Physics D, Applied Physics 13 (1980), LI 71-LI 72. 104.  J.S. Greeneich, Developer characteristics of poly-(methyl methacrylate) electron resist, Journal of  the Electrochemical Society: Solid-State Science and Technology 122 (1975), 970-976. 105.  T.E. Everhart and P.H. Hoff, Determination ofkilovolt electron energy dissipation vs penetration  distance in solid materials, Journal of Applied Physics 42 (1971), 5837-5846. 106.  A.C.F. Hoole, M.E. Welland and A.N. Broers, Negative PMMA limits and possibilities,  as a high-resolution  resist - the  Semiconductor Science and Technology 12 (1997), 1166-1170.  107.  NASA Safety Standard for Hydrogen and Hydrogen Systems. NSS 1740.16.  108.  Air Liquide compatibility charts,  109.  Posi-flatea butterfly valves. Seat Selection Guide, http://www.posiflate.com/pdf/seatguid.pdf  110.  O. Salyk, P. Castello and F. Harskamp, A facility for characterization  http://encyclopedia.airliquide.com/Encyclopedia.asp?GasID=36  and testing of hydrogen  sensors, Measurement Science & Technology 17 (2006), 3033-3041. 111.  C. Hu, X. Qu, J. Wu and G. Chapman, MOS hydrogen sensor array for 2D gas distribution  mapping, Conference of Metallurgists 2005. Calgary, Canada. (2005), 497-504. 112.  C. Kittel, Introduction to Solid State Physics - 7th Ed. John Wiley & Sons Inc., USA. (1996).  113.  D. Ding, Z. Chen and C. Lu, Hydrogen sensing of nanoporous palladium films supported by  anodic aluminum oxides, Sensors and Actuators B, Chemical 120 (2006), 182-186.  94  Appendices Appendix A.  LabVIEW control panels and block diagrams  Control panels and source diagrams of vi's and subvi's written in LabVIEW 7.1 for acquiring voltage signals are shown here.  max scans/sec  file path  10000 j \% C:\Lilian_Labview_060119\Controls\DAQData\t:est (#sampled/datapoint Saved Data Chart rJjlO  bean backlog 0~  scans/read 1000  start time 100:00:00.000 Pr JDD/MM/YYYY  error no error |  code  source  I  STOP  Figure A.l  Main control program front panel (MultiChannelDAQ.vi)  95  scan backJogl  -sl5] LpTJ |Read acquired data in Buffer 1 Llears | DAQ  is/read |  n o i o c t t  d D a D D D D D a a b ' B n o o ' D n n n n p ^ p q p t i a D a D D D D D D - g  [Build time (relative) array |  — Convert scan rate info to string for file header  [Hlf  BnP&flaflQrJ'QDaDDQDQDDOaDDQO  OD u DTP •DDQDDDDQQDDP  [Merge time & data into ZD array]  Discard excess datapomts (take only 1 point out of every #sampled/datapomt)  K create or replace \ w  MultiChannelDAQ.vi Lilian Fan Aug 18, 2005 Code to acquire data from multiple channels (default 3) at a variable rate. Modified from MultiChannelDAQ_10.vi. Inserted comments and rearranged blocks Program generates data with 1 time column only: i.e. does not take into account intrinsic acquisition delay between channel reads, Inputs; max Function: 1) 2) 3) 4) 5)  scans/sec [unsigned integer], #sampled/datapoint [unsigned integer], file path [path], stop [boolean] Acquires data to buffer from DAQ board channels specified on Block Diagram at rate <max scans/sec> Reads <scans/read> (set to equal <max scans/sec> for now) # of points from buffer (i.e. ~1 read per second) Writes header information with <max s c a n s / s e o , date, <start time> Generates time axis based on <start time> and <dt> (from DAQ board) Extract 1 data point out of every <#sarnpled/datapoint> points to plot and save to file (at 6 significant figs in Exp, notation) (this to minimize the amount of data storage space needed) 6) Repeat until <stop> is true During execution, the only control that can be changed is <#data/DAQ stream> and <stop>. Saved data is shown on <5aved Data Chart> and is updated once per read. <scans/read> indicator is pretty useless at the moment but is placed on front panel to make it easy to change to a control. <scan backlog> indicator is for monitoring how well the program is clearing the DAQ buffer. <start time> shows the time/date when the acquisition starts. <error> indicator shows if there are any errors with the DAQ board. Data in the following format is saved to file specified in <file path>: > # [date] // [time] / / Max scan rate: [max scans/sec] scans/sec > x_0 [tab] y i _ 0 [tab] y2_0 [tab] y3_0 [tab] > x _ l [tab] y l _ l [tab] y 2 _ l [tab] y 3 _ l [tab] > x_2 [tab] y l _ 2 [tab] y2_2 [tab] y3_2 [tab]  For reference, datafile size at 3 channels of 17,000 scans/sec (61.2ml time points) extrapolates to about 2.4Gb per hour,  Figure A.2 Block diagram of main control program (MultiChannelDAQ.vi)  96  Figure A.3 Time reference case in main control program (MultiChannelDAQ.vi)  1 False T£T  Figure A.4 Case for no plot data in main control program (MultiChannelDAQ.vi)  • 1 Buffer  array r.i'O • •> *  -in  j -;.o ^ffcT" ,: o r.''  r• 0  0  r  0  New Arra  New Buffer  jQ  Q  fj  0  1,0  0  0  o  0  0  ID  o  G  II 0  |0  0  1  J  0 0  0  0  •  . Q  o  0  (J  0 :  -foT^  0 0  Figure A.5 Front Panel of module for parsing and extracting data to be logged (Array2DExtract01.vi)  97  [at ray |  j jmt"jPuFferj  Eh:;  Length of Array |  tew Buffer"  ft  > mi)  EH  . R  Rate!  3  04  ^Jew Array]  Inputs: Y [2D array]. Buffer [2D array], Rate [unsigned integer] Outputs: New Y [2D array], New Buffer [2D array] Usage: Appends <Buffer> to the beginning of <Array>, Selects 1 column out of <Rate> # of columns in <Array> and build <New Array> from it. Generate <New Buffer> out of remaining elements to be appended to the following set of data. Rate must be > Q. Range set to coerce to 1 if Rate <= 0 , Figure A.6 Block Diagram of module for parsing and extracting data to be logged (Array2DExtract01.vi)  2D array  |i;  Cl  70  ,:o~_H  0  !:  I/P waveforms  |M0O:0O:0OPM  J HDD/MM/YYYY  I,,  1  I  dt  m f^i .uuuuuu O/P lAiflVftf nrm«;  to  1 00:00:00 PM  ; 1 DD/MM/YYYY  v||o  1 1  dt p . . 000000  Figure A.7 Front Panel for plotting module. Plots scrolling data (MultiPlof2DArray.vi)  98  Same as MultiPlot2DArray_02.vi, Lilian Fan, Aug 17, 2005 Plots a 2D array, works for multiple traces with the same x-axis. Input array format as follows: x_0 x_l x_2 .,, Y1_0 Y l _ l Yl 2 ... Y2_0 y2_l Y2_2 ... Figure A.8  Block Diagram for plotting module (MultiPlot2DArray.vi)  99  Appendix B.  Calibration charts for gas sensor test bench  Calibration charts for MFCs (Advanced Specialty Gas Equipment) used in the test bench are provided here. These MFCs are calibrated against one another, using readings from MFCs newly acquired from the manufacturer as reference (where applicable). MFCs to be calibrated are placed upstream and set the gas flow rate through reference MFCs that are fully open downstream. Calibration tests are done in N at 2  446kPa (50PSIG). Care was taken to ensure that an older H MFC is calibrated to be consistent with a 2  brand new H MFC. For reference however, equivalent flow of gas through an MFC that is calibrated for 2  a different gas remains linear and can be approximated with the ratio of the gases' specific heats. The approximating equation and factors are given by the manufacturer, as follows''*: 1  _ . Conv ersionFactor Output * — = Output Conv ersionF actor R  A  B  A  Equation B. 1  Table B.l  Conversion factors for MFCs  Gas  Conversion Factor  Hydrogen  1.01  Nitrogen  1.00  Air  1.00  A represents the gas for which the MFC is calibrated, and B represents the gas flowing through the MFC. A cross-reference chart for two brand new N MFCs, followed by the calibration curve for readings from 2  an older H MFC calibrated against a new unit, are shown in the following Figures. 2  {J  DatafromAdvanced Specialty Gas Equipment. Instructions for Mass Flow Sensors and Mass Flow Control  Modules. 100  Nitrogen MFC's X-Reference 20  ? u o  16  -  <£, 12 ^  3  O T3  y = 0.9667x-0.1501 R = 0.9999 2  0  10  20  15  25  N2 readout (seem) Figure B.l  Cross reference chart for N MFCs, received new from manufacturer (ASGE) 2  Actual Flow vs. H2 Flow Readout, 50PSI  y = 1.0078x + 37.854 R = 0.9999 2  100  150  200  250  300  350  400  450  500  H2 readout (seem) Figure B.2 Real flow conversion chart for an older H MFC. 2  Flow readings from MFCs used in the test bench generally follow closely the setpoints entered in their control console except for the older hydrogen MFC. The actual flow through this unit plotted against the console's setpoint is shown in the following Figure.  101  Actual Flow vs H2 Flow Setting, 50PSI 600  -i  0 -I 0  :  ,  1  100  200  :  1  '  1  300  r  400  :  ' 500  i 600  H2 Flow Setting (seem) Figure B.3 Actual flow of older H M F C as the console setpoints vary. 2  A Cole Parmer MFC is also used in the test bench for N purging. This MFC comes with its own readout 2  display. Cross reference measurements with another high flowrate MFC show discrepancies of about 2slpm and are shown in the following Figure. Purge flow rates are thus set to well above the estimated required minimum.  Advanced vs. Cole Parmer MFC N2 X-Reference  o -I 0  1  1  1  1  5  10  15  20  25  Cole Parmer 50SLPM MFC flow rate (slpm) Figure B.4 Cross Reference curve for Cole Parmer M F C with ASGE M F C . Discrepancies of about 2slpm were observed between the two MFCs.  102  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0074519/manifest

Comment

Related Items