USF Libraries
USF Digital Collections

Non-contact characterization of dielectric conduction on 4H-SiC

MISSING IMAGE

Material Information

Title:
Non-contact characterization of dielectric conduction on 4H-SiC
Physical Description:
Book
Language:
English
Creator:
Benjamin, Helen N
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Corona-Kelvin metrology
Field emission
Voltage decay
Non-contact SiLC
Trapped charge
Dissertations, Academic -- Electrical Engineering -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Consistent charge or defect control in oxide grown on silicon carbide (SiC) continues to be difficult to achieve and directly impacts the electrical performance of SiC-based metal oxide semiconductor (MOS) devices. This research applied non-contact Corona-Kelvin metrology to investigate the charge transport in oxides grown on n-type 4H-SiC epitaxial substrates. The cost and engineering science impact of this metrology are significant as device fabrication is avoided leading to quick determination of electrical characteristics from as-grown oxide films. Non-contact current-voltage (I-V) measurements of oxide on SiC were first demonstrated within this work and revealed that Fowler-Nordheim (F-N) current emission was the dominant conduction mechanism at high electric fields.Oxides on SiC were grown at atmospheric pressure (thermal oxides) or at a reduced pressure (afterglow oxides) ambient and examined using non-contact charge-voltage (Q-V), capacitance-voltage (C-V), equivalent oxide thickness (EOT), and I-V methods. The F-N conduction model was modified to address charge trapping and effective barrier effects obtained from experimental oxide films. Trap densities determined with this metrology were used to show that the F-N model including their density and position was adequate for thermal oxides on SiC but not for afterglow films. Data from the latter films required further modification of the theory to include a chemical effect of the oxide growth process on the effective conduction band offset or barrier. This work showed that afterglow chemistry was able to vary the effective conduction band offset from 2.9 eV, typical of thermal oxidation of SiC, up to 3.2 eV.Stress induced leakage current (SILC), an excess above the F-N base current resulting from prolonged current through the dielectric films, was also investigated. Multiple point SILC testing was used to identify statistical effects of process variations and defects in as-grown oxide films on SiC. These results open the possibility to improve oxide manufacture on SiC using methods common in the silicon IC industry. This work demonstrated the first non-contact F-N current determination in oxides on SiC and showed both charge trapping and chemical dependencies of as-grown films. Future studies may extend the findings of this work to further improve this important dielectric-semiconductor system.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Helen N. Benjamin.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 142 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002061864
oclc - 528827476
usfldc doi - E14-SFE0002984
usfldc handle - e14.2984
System ID:
SFS0027301:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Non-Contact Characterization of Dielectric Conduction on 4H-SiC by Helen N. Benjamin A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Andrew M. Hoff, Ph.D. Stephen E. Saddow, Ph.D. Scott W. Campbell, Ph.D. Richard A. Gilbert, Ph.D. Sarath Witanachchi, Ph.D. Date of Approval: April 30, 2009 Keywords: Corona-Kelvin Metrology, Field Emission, Voltage Decay, Non-Contact SILC, Trapped Charge Copyright 2009, Helen N. Benjamin

PAGE 2

Dedication I would like to dedicate this manuscript to Mom, Dad, Marcia, Hilma, Alicia, Maurice, and family. Through their constant prayers and positive words of encouragement, I was able to overcome the challenges that I encountered throughout my pursuit of this degree. Also, to my fianc, Leonard, thank you for believing in me and providing a listening ear throughout this endeavor.

PAGE 3

Acknowledgments I am grateful toAndrew Hoff, Ph.D.,for giving me the opportunity to work in this field of non-contact metrology. I am also indebted toElena Oborina, Ph.D.,for not only providing me with scientific criticismof my work but also helping me with data acquisition. To Sasha Savtchouk,Ph.D., and John DAmico, Ph.D.,atSemiconductor Diagnostics,Inc. (SDI), thanks for providingtechnical support for theFAaST-230tool. I would also like to thank Eric Persson for teaching me how to analyze statistical data. To my committee members, StephenSaddow,Ph.D., RichardGilbert,Ph.D., Sarah Witanachchi,Ph.D., and ScottCampbell, Ph.D., I greatly appreciate your time, suggestions,and support of this manuscript. To Robert Tufts and Richard Everly at Nanomaterials and Nanomanufacturing Research Center (NNRC), thanks for the metrology training. Finally, I would like to thank all my friends and colleagues, Gene Short, AuraPolo, NorelliSchettini, AlexandraOliveros, Chris Frewin, and ChrisLocke for being my socialoutlet.

PAGE 4

i Table of Contents List of Tables.....................................................................................................................iv List of Figures....................................................................................................................vi Abstract..............................................................................................................................xi Chapter 1. Introduction.......................................................................................................1 1.1.Research Objectives and Motivation...............................................................1 Chapter 2. Overview of Electrical Stress Testing Methodology of Dielectrics..................4 2.1.Standard Measure of Oxide Quality................................................................4 2.2.Overview of 4H-SiC Substrates.......................................................................5 2.2.1. Silicon Carbide (SiC) Oxidation Theory..........................................8 2.3.Current Ramp Test and Voltage Ramp Test..................................................10 2.4.Conduction Mechanisms of Dielectric-Semiconductor System....................11 2.4.1. Poole-Frenkel Conduction..............................................................12 2.4.2. Fowler-Nordheim Conduction........................................................14 2.5.Constant Current Stress Technique................................................................20 2.6.Time Dependent Dielectric Breakdown (TDDB)..........................................21 2.7.Contact Stress Induced Leakage Current (SILC) Technique.........................27 2.8.Non-Contact Stress Induced Leakage Current (SILC) Technique................30 2.8.1. Corona-Kelvin Metrology...............................................................30 2.8.2. Capacitance-Voltage Measurements...............................................33 2.8.3. Equivalent Oxide Thickness (EOT) Measurement.........................38 2.8.4. Current-Voltage Measurements......................................................40 2.8.5. Stress-Induced Leakage Current Method........................................42 2.9.Statistical Issues in Device-Based Measurements.........................................45 2.10.Chapter Summary........................................................................................46

PAGE 5

ii Chapter 3. Experimental Procedures.................................................................................48 3.1.Non-Contact Instrumentation.........................................................................48 3.1.1. Film Analysis and Substrate Testing (FAaST) 230 and Components..............................................................................48 3.1.2. Ion-Drift Spectrometer....................................................................51 3.2.Pre-Oxidation Cleaning Procedure................................................................52 3.3.Oxidation Sequence/Interval Experiments....................................................53 3.3.1. Brief Description of Afterglow (AG) Furnace................................54 3.3.2. Afterglow (AG) Oxidation Parameter Variations...........................55 3.3.3. Thermal Oxidation..........................................................................60 Chapter 4. Experimental Results and Discussion.............................................................61 4.1.Measurement Test Conditions.......................................................................61 4.2.Non-Contact Corona Current Stress..............................................................61 4.3.Potential Factors Affecting the Accuracy of the Contact Potential Difference..........................................................................68 4.4.Fowler-Nordheim Conduction in Afterglow Oxide and Thermal Oxide on 4H-SiC.............................................................................69 4.4.1. Self Adjusting Steady State (SASS) Voltage on Oxides................82 4.4.2. Significant Parameters: Effective Mass in the Oxide and Barrier Height................................................................83 4.4.3. Comparison with Fowler-Nordheim Literature Data on Oxide-n-type 4H-SiC Devices...........................................87 4.4.4. Influence of Trapped Charge in the Oxide.....................................90 4.4.5. Fowler-Nordheim Equation Modification for Oxide-4H-SiC............................................................................94 4.5.Distribution of Trapped Charge in Oxide on 4H-SiC....................................97 4.5.1. Diluted Hydrofluoric Acid Etch Rate on Oxide-4H-SiC................97 4.5.2. Influence of Dehydration Procedure on Oxide-4H-SiC................100 4.5.3. Effective Trapped Charge Distribution.........................................102 4.6.Non-Contact Stress Induced Leakage Current (SILC) Analysis.................106 4.6.1. Oxidation Process Influence on Leakage Current........................107 4.6.2. Oxide Consistency: Statistical Distribution..................................109

PAGE 6

iii Chapter 5. Conclusion.....................................................................................................118 5.1.Summary of Research Contributions...........................................................118 5.2.Future Work.................................................................................................121 References.......................................................................................................................123 Appendices......................................................................................................................131 Appendix A: Afterglow Oxide Recipes..............................................................132 About the Author ...End Page

PAGE 7

iv List of Tables Table 2.1: Properties of 4H-SiC and Si at room temperature (300K) [14-16]...................7 Table 2.2: Chemical reactions during dry thermal oxidation of SiC [20]..........................9 Table 2.3: Commonly used effective oxide mass for Fowler-Nordheim calculations......................................................................................................18 Table 3.1: Annealed afterglow (AG) recipes....................................................................56 Table 3.2: Non-annealed afterglow (AG) recipes.............................................................57 Table 4.1: Stress current measurement parameters and conditions..................................63 Table 4.2: Stress measurement protocol...........................................................................65 Table 4.3: Oxidation process chemistry variation............................................................73 Table 4.4: Calculated effective barrier for annealed 4H-n-type experimental samples.......................................................................................76 Table 4.5: Calculated effective barrier for non-annealed 4H-n-type experimental samples.......................................................................................76 Table 4.6: Reported effective barrier height for 4H-SiC MOS devices at room temperature.........................................................................................89 Table 4.7: Etch rate comparison between 4H-SiC and Si wafers at 24 0 C........................98

PAGE 8

v Table 4.8: Effective trapped charge calculation parameters...........................................105 Table 4.9: Stress measurement protocol for SILC measurements..................................110 Table A.1: AG I_A Recipe.............................................................................................132 Table A.2: AG I_B Recipe..............................................................................................132 Table A.3: AG II Recipe.................................................................................................133 Table A.4: AG III Recipe................................................................................................134 Table A.5: AG IV Recipe...............................................................................................135 Table A.6: AG V Recipe.................................................................................................136 Table A.7: AG VI Recipe...............................................................................................137 Table A.8: AG VII Recipe..............................................................................................138 Table A.9: AGW I Recipe..............................................................................................139 Table A.10: AGW II Recipe...........................................................................................140 Table A.11: AGW III Recipe..........................................................................................140 Table A.12: AGW IV Recipe..........................................................................................141 Table A.13: AGW V Recipe...........................................................................................141 Table A.14: AGW VI Recipe..........................................................................................142

PAGE 9

vi List of Figures Figure 2.1: A simple MOS capacitor..................................................................................4 Figure 2.2: Four carbon atoms covalently bonded with a silicon atom [12]......................6 Figure 2.3: A 4H-SiC three-dimensional lattice structure [13]...........................................6 Figure 2.4: Location of oxide charges after a thermal oxidation of Si [24]......................10 Figure 2.5: Example of Poole-Frenkel plot for SiO 2 on Si...............................................13 Figure 2.6: Energy band diagrams depicting a MOS device without stress (a) and under Fowler-Nordheim tunneling (b)......................................15 Figure 2.7: An example of a Fowler-Nordheim plot........................................................17 Figure 2.8: An example of a gate voltage-time characteristic of a MOS device obtained by CCS testing [33].....................................................20 Figure 2.9: Weibull cumulative distribution for a population fraction failing by time [47]........................................................................................23 Figure 2.10: An example of a Weibull plot [48]...............................................................24 Figure 2.11: A normal cumulative distribution function for a population failing by y [47]............................................................................................27 Figure 2.12: A sketch illustrating an nMOS (a) without stress and (b) after SILC.........................................................................................28

PAGE 10

vii Figure 2.13: An example ofSILC modes [3]...................................................................29 Figure 2.14: A sketch of Kelvin probe measurement.......................................................31 Figure 2.15: Energy band diagram after the deposition of corona charges......................35 Figure 2.16: Non-contact C-V characteristics of (a) a 150 thermal oxide on a p-type Si and (b) a 400 afterglow oxide on an n-type 4H-SiC................37 Figure 2.17: Non-contact C-V characteristic of a 499 afterglow oxide on an n-type 4H-SiC measured in the dark (circles) and in the light (asterisks)............................................................................39 Figure 2.18: Oxide voltage decay after corona charging [76]..........................................41 Figure 2.19: Energy-band diagram during non-contact SILC..........................................43 Figure 2.20: Example of determining thickness with the Fowler-Nordheim current density [64].......................................................................................44 Figure 3.1: Schematic of the measurement apparatus based on ref. [64].........................49 Figure 3.2: Sketch of the (a) wire enclosure, (b) discharge distance to the wafer, and (c) aperture diameter.....................................................................51 Figure 3.3: RCA cleaning procedure................................................................................53 Figure 3.4: Sketch of afterglow furnace system [77]........................................................54 Figure 3.5: Measurement sequence to calculate J F-N ........................................................57 Figure 3.6: Etch pattern and quadrant designation on the substrate.................................58 Figure 3.7: Measurement sequence to calculate trapped charge.......................................59

PAGE 11

viii Figure 3.8: Measurement sequence to obtain SILC data acquisition................................60 Figure 4.1: Current density versus V cpd for various corona currents................................62 Figure 4.2: Theoretical ideal D it =0 (before stress) and D it 0 (after stress) [33].............64 Figure 4.3: Capacitance-voltage characteristics versus corona stress..............................66 Figure 4.4: Poole-Frenkel plot comparing theoretical and experimental oxides..............70 Figure 4.5: Fowler-Nordheim plot for various thermal oxide thicknesses on Si..............72 Figure 4.6: Oxidation process chemistry influence on the effective barrier height for SiO2 /4H-SiC systems (see Table 4.3 for oxidation process representation)...................................................................................74 Figure 4.7: Fowler-Nordheim plot of AG annealed oxides..............................................77 Figure 4.8: Fowler-Nordheim plot of thermal annealed oxides........................................78 Figure 4.9: Fowler-Nordheim plot of non-annealed AG oxides.......................................80 Figure 4.10: Comparison of experimental data and predicted Fowler-Nordheim lines.................................................................................81 Figure 4.11: Steady-state voltage experimental oxide comparison..................................83 Figure 4.12: Variation of M ox on the predicted Fowler-Nordheim characteristics................................................................................................84 Figure 4.13: Theoretical Fowler-Nordheim plot varying the effective barrier height with Mox of 0.36......................................................86

PAGE 12

ix Figure 4.14: Example of current-voltage characteristics of a 4H-SiC MOS capacitor with a 500 gate oxide measured at room temperature [52]..............................................................................87 Figure 4.15: Fowler-Nordheimconduction comparison of contact versus non-contact measurements............................................................................90 Figure 4.16: Charge trapped within tunneling regime in SiO 2 /Si system: (a) negative charge trapping, (b) positive charge trapping [92]..................................................................92 Figure 4.17: Example of trapped charge outside or inside the Fowler-Nordheim tunneling regime in a 500 oxide..................................93 Figure 4.18: Capacitance-voltage characteristics of AG IV and Thermal TA oxides before and after stress...................................................95 Figure 4.19: Non-triangular Fowler-Nordheim plot fitted to experimental oxides......................................................................................96 Figure 4.20: Oxide thickness after diluted HF etching for oxide AG IV_EA..................99 Figure 4.21: Oxide thickness after diluted HF etching for oxide AG IV_EB................100 Figure 4.22: The influence of a dehydration method on the Fowler-Nordheim plot (a) AG IV_EA and (b) AG IV_EB.......................101 Figure 4.23: Fowler-Nordheim plot of thinner oxides (unfilled geometric shapes) compared to thick oxides (filled geometric shapes)...................................103 Figure 4.24: Absolute trapped charge versus oxide thickness........................................104 Figure 4.25: Measurement site positions........................................................................107 Figure 4.26: Effect of stress fluence on a 150 thermal oxide grown on a p-type Si substrate...............................................................................108

PAGE 13

x Figure 4.27: Current density versusoxide field at point (10, 10) for AG I_A oxide........................................................................................110 Figure 4.28: Probability plot for AG III oxide at each cumulative time.........................111 Figure 4.29: Probability plot of AGW I oxidation process run repeatability.................112 Figure 4.30: Probability plot of AG I_A oxidation processrunrepeatability................113 Figure4.31: Probability plot of AG II oxidation process run repeatability....................115 Figure 4.32: Probability plot of AG III oxidation process run repeatability..................116 Figure 4.33: Leakage current sites greater than 1x10 -7 A/cm 2 .........................................117

PAGE 14

xi Non-Contact Characterization of Dielectric Conduction on 4H-SiC Helen N. Benjamin ABSTRACT Consistentcharge or defect control in oxide grown on silicon carbide (SiC) continues to be difficult to achieve and directly impacts the electrical performance of SiC-based metal oxide semiconductor (MOS) devices. This research appliednon-contact Corona-Kelvin metrology to investigatethe charge transport inoxides grown on n-type 4H-SiC epitaxial substrates.The costand engineering science impactof this metrology are significant as device fabrication is avoided leading to quick determination of electrical characteristics from as-grown oxide films. Non-contact current-voltage (I-V) measurements of oxide on SiC were first demonstrated within this work and revealed that Fowler-Nordheim (F-N) current emission was the dominant conduction mechanism at high electric fields. Oxides on SiC were grownat atmospheric pressure (thermal oxides) or at a reduced pressure (afterglow oxides)ambientandexaminedusingnon-contact chargevoltage (Q-V), capacitance-voltage (C-V), equivalentoxide thickness (EOT),and I-V methods. The F-N conduction model was modified to address charge trapping and effective barrier effects obtained from experimental oxide films. Trap densities determined with this metrology were used to show that the F-N model including their

PAGE 15

xii density and position was adequate for thermal oxides on SiC but not for afterglow films. Data from the latter films required further modification of the theory to include a chemical effect of the oxide growthprocess on the effective conduction band offset or barrier. This work showed that afterglow chemistry was able to vary the effective conduction band offset from 2.9eV, typical of thermal oxidation of SiC, up to 3.2eV. Stress induced leakage current (SILC), an excess above the F-N base current resulting from prolonged current through the dielectric films, was also investigated. Multiple point SILC testing was used to identify statistical effects of process variations and defects in as-grown oxide films on SiC. These results open the possibility to improve oxide manufacture on SiC using methods common in the silicon IC industry. This work demonstrated the first non-contact F-N current determination in oxides on SiC and showed both charge trapping and chemical dependencies of as-grown films. Future studies may extend the findings of this work to further improve this important dielectricsemiconductor system.

PAGE 16

1 Chapter 1.Introduction 1.1.Research Objectives and Motivation Silicon carbide (SiC), awide band gap semiconductor, is anideal candidate for the development of the next generation high power, frequency, and temperature device applications. The quality of an oxide can directly influence the electrical performance of metal oxide semiconductor (MOS) devices. Process induced charges in the oxide, which can either be neutral, positive, or negatively charged, impact the reliability and integrity of gate oxides in MOS devices. The commercialization of silicon-based MOS devices is due to over four decades of researchbasedon the control of chargesat the silicon dioxide/silicon (SiO 2 /Si) interfaceor in the bulk of the oxide. For example, one popular treatment to reduce charges in the oxideis the addition ofa post annealing processafter oxidation and metallization of the gate. The realization of commercialized SiC-based power devices, controlled by MOSFETs (field effect transistors), iscontingent on the control of charges in the oxide. Currently, an optimized oxidation process, for this binary semiconductor, to reduce charges in the oxide remains under investigation. Extensive analysis to obtain significant consistent statistical data for SiO 2 /SiC characterization is costly and time consuming using standard measurement techniques accepted for SiO 2 /Si characterization.

PAGE 17

2 The aim of this research was to investigate the application of non-contact stress induced leakage current (SILC) testing for oxides grown on 4H-SiC epitaxial substrates. Non-contact SILC testing has been successfully established for oxide characterization on Si substratesto detect the vulnerability of an oxide film to breakdown or become conductive.Oxide breakdown induced by SILC testing on SiO2 /Si systems hasbeen generally characterized bytrap-assisted tunneling. In the advent of integrated-circuit (IC) chip miniaturization, onepopular characterizationtechniquedone onthin oxides grown onSi isSILC testing[1-5]. This testing has also been used to characterize thick oxides on Si [6, 7]. Currently,oxides grown onSiC substratesfor MOSFETs used in power devices are nominally 500thick. The gate oxide is predominately characterized by current-voltage measurements to determinesuch parameters asitsFowler-Nordheim tunnelingcharacteristics and its time-to-breakdown characteristics.Prior to this research, SILC testing on SiO 2 /SiC systems has not beenextensivelyreported in literature. Since this testingtechnique does not exist forthe characterization ofoxides on4H-SiC substrates, it is intended to have the samecharacterization success established for SiO2 /Si systems. Specifically, this work investigatedthe details and variations in the conduction mechanism of an oxidegrownon n-type4H-SiC substratesusing a modifiednon-contact characterization tool. Stress induced leakage current isthe excesscurrent in addition to theFowler-Nordheim currentfor oxide thicknesses greater than 50. It was essential to establish the Fowler-Nordheim current as the dominateconductionmechanism for each experimental oxide. The testing methods developed identified the Fowler-Nordheim current on each oxide.

PAGE 18

3 It was observed thatthevalue for theeffective barrier height and the effective mass in the oxidewere influenced by oxidation processconditions. These two parameters are important because they define the Fowler-Nordheim curve characteristic. The effective barrier height in this caseis related to the conduction band offset of the oxide-semiconductor interface. Trapped charges in the oxidealsostrongly influenced the Fowler-Nordheim characteristicsand as a result, a variation of the Fowler-Nordheim equation was addressed. The location of the trapped charge and its centroid was further analyzedto fit the experimental data to amodified Fowler Nordheim tunneling equation. Afterestablishing this current, seventeen sites on variousoxides were subjected to non-contact SILC testing and analyzed. The goal wastouse theeffectiveSILC value as an indicator to identify weak spots around the oxide surface. It was shown that this unique characterization method for oxides onn-type4H-SiC has the potential to be used as a measure of oxide reliability. In addition, it enabled afast assessment of an oxidation process in the absence of fabricated capacitors or transistors.

PAGE 19

4 Chapter 2.Overview of Electrical Stress Testing Methodologyof Dielectrics 2.1.Standard Measure of Oxide Quality The reliability of an oxide grown on SiC substrates dictates itssuccess in IC chips. Gate oxide of Si-based devices, such as capacitorsor field effect transistors, have been extensively researchedover four decades and summarized by D.J. Dumin, D.J. Dimaria, J.H. Sathis and others [3, 8-11]. A simple MOScapacitortest deviceis composed ofa semiconductor, a gate oxide, and a gate (see Figure 2.1). The gate is either ametalplateor doped poly-silicon, where a voltage or current can be applied. Figure 2.1: A simpleMOS capacitor. Oxide breakdown is the wear out mechanism in the oxide after losingits insulating or resistive properties in the presenceof high or low electric fields duetothe breakage of SiO 2 bonds in its lattice structure. There are two types of oxide breakdown: destructive (hard breakdown (HBD)-irreversible)and non-destructive breakdown (soft breakdown (SBD)-reversible). These breakdowns can be due to intrinsic failures, such asthe quality of the oxide, extrinsic failures, such as electrical stress, or both. The quality of the oxide SiO2 Sigate V or I

PAGE 20

5 corresponds to its processingcondition (e.g. free from metal contamination), its uniformity, its surface roughness, and any other form ofmicro-defects. The ability to monitor an oxide behavior overarange ofelectric fieldsin the oxidecan be performed using either a current or voltage ramp test. 2.2.Overview of4H-SiC Substrates Silicon carbide(SiC)crystal growth is not manufactured by the highly engineered Czochralski (CZ) crystal growth method, but from the seeded sublimation growth method [12]. By comparison with the CZ method for silicon substrates, SiC substrates are defective and require a high quality epitaxial layer due to underlying crystal defects such as, open-core screw dislocations (micro-pipes) per cm 2 and low angle boundaries. As a result, these substrates are 100 times more expensive than silicon substrates. Silicon carbidesubstrates are orientated between 3 to 8 degrees off-axis and have two terminated surfaces, silicon and carbon. In the Miller notation, which describes crystallographic directions in a unit cell, the silicon face is on the (0001) plane while the carbon face is on the (000 1 ) plane. The numbers in the parentheses represent the coordinates x, y, z, and c axes in the unit cell. The c-axis corresponds to the stacking direction. The fabrication of devices mostly occurs on the siliconface. The basis of every SiC crystal comprises of a silicon atomsurrounded by four carbon atoms forming a tetrahedral structure (see Figure 2.2). Likewise, each carbon atombonds to four nearest-neighbor silicon atoms. The distance between aneighboring silicon or carbon atomis approximately3.08.

PAGE 21

6 Figure 2.2: Four carbon atoms covalently bonded with a silicon atom[12]. Out of 200 SiC crystal systems, one of the commonly used SiC crystal geometries or polytypes for MOS-based power devices is 4H-SiC. This crystal geometry is depicted as 4H-SiC because it has a hexagonal crystal structure, whose repetitive stacking sequence of four layers of silicon-carbon atoms, is arranged as ABCBABCB to complete one unit cell. The position of A, B, and C each represent a layer of atoms. Figure 2.3illustrates a 4H-SiC three-dimensional latticestructure[13]. Figure 2.3: A 4H-SiC three-dimensional lattice structure[13]. C Si C C C

PAGE 22

7 The 4H-SiC substrates used for this study were 8 0 off-axis and were doped with nitrogen. Nitrogen impurities in the crystal lattice depictedthat its conduction type was n-type because of the addition ofnegative charge carriers. Table 2.1 lists the properties of 4HSiCand Si. These properties make 4H-SiC substratesattractive candidatesfor power switching devices. Table 2.1: Properties of 4H-SiC and Si at room temperature (300K) [14-16]. Crystal Structure 4H-SiC Si Band gap Energy (eV) 3.25 1.12 Instrinsic Carrier Concentration (cm -3 ) 2x10 -8 1x10 10 Breakdown Field (V/cm) 1.5-4x10 6 3-4x10 5 Bulk Electron Mobility (cm 2 /Vs) ( to c-axis) (|| toc-axis) ~1050 ~800 ~1350 Thermal Conductivity 3.3 1.5 Saturation Velocity (cm/s) 2.2x10 7 1x10 7 Lattice Constant () a=3.073 c=10.05 a=5.43 As seen in Table 2.1, thecritical breakdownfieldof 4H-SiCis an order of magnitude higher than silicon,making it conducive for highpower blocking or switching inpower devices. These properties continue to make this binary semiconductor a promising candidate for power devices due to its capability to block high voltages, switch at high frequencies, and operate at high temperatures.

PAGE 23

8 2.2.1.Silicon Carbide (SiC)Oxidation Theory The two types of oxidation growth method used in this study were atmospheric thermaloxidation and afterglowoxidation. An overview of the oxide growth kinetics on SiC is presented. Silicon carbide is the only binary compound semiconductor whose native oxide is silicon dioxide (SiO 2 ). Similarto oxidation of Si, thermal oxidation of SiC can be either in dry O 2 pyrogenic H 2 O, or both. The growth kinetics of SiO 2 on SiC is not the same as on Si dueto the presence ofcarbon atoms.Theoxidation growth rate on SiC is an order of magnitude slower than Si under the same conditions [17-22]. One reason attributed to this phenomenon is the oxidation of carbon. Based on the model for the oxidation of Si, themodel for theoxidation of SiC is expressed as [17]) ( 2 t B AX X (2.1) CO r O f CO r O f D K D K h K h K A 2 2 5.1 5.1 1 (2.2) CO r O f CO r O f D K D K N C K C K B 2 2 5.10 * (2.3) where X is the oxide thickness, t is the oxidation time, is the initial thickness, B is a parabolic rate constant, B/A is a linear rate constant, K f is the rate constant of the forward reaction, K r is the rate constant of the reverse reaction,O2 is oxygen, CO is carbon monoxide,h is the gas-phase transport coefficient, D is the diffusion coefficient, C* is the equilibrium concentration, and N 0 is the number of oxidant molecules incorporated into a

PAGE 24

9 unit volume of the oxide layer. Table 2.2 lists the chemical reactions at the SiC interface duringthe growth of adry thermaloxide. Table 2.2: Chemical reactions during dry thermal oxidation of SiC[20]. Sequence Chemical Reaction Primary C SiO O SiC CO SiO O SiC 2 2 2 2 2 3 Secondary CO O C SiO C CO SiC 2 2 3 2 2 2 From the chemical reactions, the oxidized carbon is out-diffused in the form of a gas. The amount of carbon remaining in the oxide, as a result of incomplete carbon oxidation, leads to the accumulation of carbon or carbon clustersatthe SiO2 /SiC interface. These clusters form a density of interface states or traps near the interface [23]. Other reported sourcesof traps near the interface are due to carbon or silicon interstitials and stable carbon pairs [21]. The four most cited types of charges associated withSiO2 /Si systems and more so inSiO2 /SiC systemsare fixed oxide charge (Qf ), mobile oxide charge (Q m ), oxide trapped charge (Q ot ), and interface trapped charge (Q it ). Figure 2.4illustrates the location of these charges after athermal oxidation of Si[24].

PAGE 25

Figure 2 4 : Location of oxide charges after a thermal oxidation of Si The reduction and control A hydrogen anneal at 450 state s (D it ) from above 10 this anneal does not reduce D oxide charges in SiO 2 / SiC Nitrous oxide (N2 O) anneal, A The reduction of D it at the SiO 2.3. Current Ramp Test and Voltage Ramp Test Current and voltage r integrity of a gate oxide [25 of electric fields to reveal intrinsic or extrinsic failures. reached observed as a spike in the current, insulating properties The 10 : Location of oxide charges after a thermal oxidation of Si [24] reduction and control of these charges are done typically by a post oxidation anneal hydrogen anneal at 450 0 C on an oxide grown on Si reduced the density of interface ) from above 10 11 /cm 2 eV to 10 10 /cm 2 eV [14] Contrary to SiO this anneal does not reduce D it at the SiO 2 /SiC interface Proposed anneals to decrease SiC systems are Argon (Ar) anneal, Nitric oxide (NO) anneal, O) anneal, A mmonia (NH 3 ) anneal, or Re oxidation (Re at the SiO 2 /SiC interface remains a n ongoing challenge. Current Ramp Test and Voltage Ramp Test Current and voltage r amp t est techniques enab le a quick assessment [25 27] These tests are used to analyze an oxide under a range of electric fields to reveal intrinsic or extrinsic failures. When a critical charge density observed as a spike in the current, the oxide is no longer able to sustain its The electric field in the oxide cannot be directly measured. [24] oxidation anneal reduced the density of interface Contrary to SiO 2 /Si systems Proposed anneals to decrease are Argon (Ar) anneal, Nitric oxide (NO) anneal, (Re ox) anneal. n ongoing challenge. assessment of the These tests are used to analyze an oxide under a range critical charge density is the oxide is no longer able to sustain its cannot be directly measured. I t can

PAGE 26

11 be calculated from the voltagedropacross the oxide(Vox )divided by the oxide thickness (t ox )(see equation 2.4). ox ox ox t V E (2.4) Both of these techniques reveal how the current conduction in an oxide over a range of electric fields can provide information about the charge-to-breakdown (Qbd ), the breakdown voltage (V bd ), and the time-to-breakdown (T bd ). The main disadvantageusing such techniques isthe inability to establishthe potential defect mechanismleading to thebreakdown of the oxide. In the following sections, other techniques, such as constant current stress (CCS), constant voltage stress (CVS), time dependent dielectric breakdown (TDDB), and stress induced leakage current (SILC),are standard measurements used to examinethe defect mechanisms leading to oxide breakdown. These device-based measurement methodsare described with the intention of applying similar technique principlesfor the investigation of oxides on 4HSiC wafers using non-contact Corona-Kelvin metrology. 2.4.Conduction MechanismsofDielectric-SemiconductorSystem In this research study, oxidesgreater than 80were stressed usingaconstant coronacurrentto induce anelectric field, greater than 5MV/cm, inthe oxide. In this field range, the current may conduct asFowler-Nordheim emission or Poole-Frenkel emission.

PAGE 27

12 2.4.1.Poole-Frenkel Conduction In 1938, Poole-Frenkel conduction, named after Horace H.Poole and Yakov Frenkel, isa process due to the field-enhanced thermal ionization of electrons from charged or neutral traps. Therefore, the current flow is due to the contribution of traps in the bulk of the dielectric [28]. The Poole-Frenkel plot is defined as ln(J pf /E) versus E where the current density(Jpf ) based on the Boltzmann approximation, isgiven by [29] kT qE q CE J r pf 0 exp (2.5) where is the ionization potential in eV of the Coulombic traps in the oxide, Cis the proportionality constant related to the density of the trap centers, k isBoltzmanns constant, Tis thetemperature, is thefactor, which varies between one and two depending on the relative concentration of acceptor traps (1) or donor traps (2) within the oxide, 0 is thepermittivity of vacuum, r is thedielectric constant, qis theelectronic charge, and Eis theelectricfield inthe dielectric. The ionization potentialis the energy requiredfor a trap charge to overcome the influence of the trapping center in the absence ofan electric field.The Poole-Frenkel plot yields a straight line and is expressed as[29] kT q C E kT q E J r pf ln ln 0 3 (2.6) The slope of this line is used to assess whether or not the dielectric conduction was due to Poole-Frenkel.

PAGE 28

13 For a SiO 2 /Si system, the Poole-Frenkel current density was calculated based on an ionization potential of 1eV, of 1,and a proportionality constant of 1x10-13 These parameters were reported valuesin literature for silicon dioxide [29]. Figure 2.5: Example of Poole-Frenkel plot for SiO 2 on Si. Based on these parameters, the slope of this line is 0.015 (see Figure 2.5). According to Equation 2.6, this slope decreasesif there are less acceptor traps within the oxide ( =2) to 0.0074. In this study, the slope was the only parameter used to check the occurrence of Poole-Frenkel conduction inan experimental oxide.

PAGE 29

14 2.4.2.Fowler-Nordheim Conduction In 1928, Sir Ralph Fowler and Lothar W.Nordheim introduced a theoryto explain field emission from a metal into vacuum. The Fowler-Nordheimcurrent density characteristic, based on the free-electron gas model and tunneling probability by the Wentzel-Kramers-Brillouin (WKB) approximation method, was related to the electric field at the surface of an emitter. Thiscurrent density was expressedas [30] F y v B ckT ckT y t AF J fn )) ( ( exp ) sin( ) ( 1 2 2 (2.7) h q A 8 3 (2.8) hq m B 3 ) ( 2 8 3 (2.9) hqF m y t c 2 )) (( 4 (2.10) where hisPlancks constant, qis theelectronic charge, Fis theelectric field, is the barrier height, mis thefree-electron mass, t(y) and v(y) are correction factors, kis Boltzmanns constant, and Tis thetemperature. The correction factors, t(y) and v(y), rendered the image potential rounding effect on the top of the barrier [30, 31]. Forty years later, the Fowler-Nordheim equation was modified for emission from a metal into silicon oxide. Fowler-Nordheim conductionoccurs when electrons tunnelfrom the semiconductor conduction band into the oxide conduction band,through the deformation of the potential barrier at the oxide-semiconductor interface. As the applied voltage

PAGE 30

15 across the oxide exceedsthe semiconductor electron affinity,the potential barrier changes from an impenetrable rectangular barrier to a penetrable triangular barrier [32]. Figure 2.6 illustratesthis phenomenon withan energy band diagram of a MOS device. Figure 2.6: Energy band diagramsdepicting a MOS devicewithout stress (a) and under Fowler-Nordheim tunneling(b). The semiconductor electron affinity potential (eV) is defined as the potential difference betweenthe vacuum level andthebottom of theconduction band [33]. Based on photoemission measurements, the electron affinity for Si is given as 4.05eV[16]and 4HSiC is given as 3.62eV [15]. In the modifiedFowler-Nordheimequation,the free electron mass (m) was replaced bythe effective mass ofan electron in theoxide band gap(mox ). Coefficients A, B, and c were modified from Equation 2.8 to 2.10as [34] E c =Conduction band energyE f = Fermi energy E v =Valence band energyI m =Imageforce barrier loweringqXs =Si Electron AffinityE c E f Oxide Metal Silicon(a)E v qV oxV>0 e (b)I m qX s Vacuum Level Vacuum Level

PAGE 31

16 ox b m h m q A 8 3 (2.11) qh m B b ox 3 ) ( 2 8 3 (2.12) ox b ox qhE m y t c 2 )) (( 4 (2.13) where E ox is theelectric fieldin the oxide(defined by Equation 2.4)and b is the effective barrier height. This effective barrier height takes into account barrier height lowering and quantization of electrons at the semiconductor surface [33]. Internal electron photoemission is a standard measurement method used to determine the band offset of a semiconductor-dielectric interface. Using photoemission measurements,it was reported that the effective barrier height represented the conduction band offset in an oxide-semiconductor system[34]. Unless otherwise noted, the electric field in the oxide for the Fowler-Nordheim plot is represented as Ethroughout the document. The FowlerNordheim plot is defined as ln(J/E 2 ) versus (1/E), which yields a straight lineexpressed as ln(A) E 1 B E J ln 2 (2.14) The slope of this line corresponds to negativeB and the intercept of the line is equal to thenatural log of A. From the Fowler-Nordheimplot, the effective mass in the oxide and the effective barrier height are obtained by simultaneously solving Equation 2.11 and 2.12. An example of a Fowler-Nordheim plot is shown by Figure 2.7.

PAGE 32

17 Figure 2.7: An example of a Fowler-Nordheim plot. In Figure 2.7, the image force barrier loweringeffects on the current density ledto a parallel shiftof the lineresulting inslightly higher values of J/E2 The modified Fowler-Nordheim equation was simplified further, whenthelow temperature approximation was used and the image-force barrier lowering was ignored (because of its negligible effect on the tunnelingdistance(see Figure 2.7)) to [33, 35] ox b ox ox b ox FNE m m x E m m x J 3 7 2 6 10 83 .6 exp 10 54 .1 (2.15)

PAGE 33

18 For an ideal case, Equation 2.15assumed that charges trapped in the oxide were negligible. It was also derived under the following conditions: the electrons in the emitting electrode can be described by a free Fermi gas; electrons in the oxide have a single effective mass (m ox ); and the tunneling probability is derived by taking into account the component of the electron momentum normal to the interface only [33, 35]. The ratio of m ox /m is represented as M ox throughout the remainder of the document. Table 2.3shows a few commonly used effective mass in the oxide, (Mox ), reported for Sibased MOS devices using either aFranz or parabolic dispersion relation. Table 2.3:Commonly used effective oxide mass for Fowler-Nordheim calculations. Reference Effective Mass in the Oxide(Mox ) Effective Barrier Height [36] 0.36 3.15eV [34] 0.42 3.25eV [35] 0.50 2.9eV These values may change based on a particular oxidation process. In this instance,the effective barrier heightand Mox calculated from the experimental Fowler-Nordheim plot, for a lightly nitride oxide on Si,was 3.05eV and 0.38, respectively [37]. Non-contact Corona-Kelvin metrology, used to obtain Fowler-Nordheim characteristics for anun-metallizedgate oxide on Si,wasfirstpioneered by R. Williams and M. Woods in 1973 [38]. The fundamental principlesof this metrology will be described in Section 2.8. The measured current density (J exp ) was expressed as[39, 40]

PAGE 34

19 dt t dE J ox r )(0 exp (2.16) ox o ox t V t V t E ) ( ) ( (2.17) where 0 is the permittivityof vacuumand r is the dielectric constant, V(t) is the surface voltage decay data following the cessation of corona charge, and V 0 is the potential difference between the Kelvin-probe and the oxide, priorto corona charge deposition. Equation 2.15 and Equation 2.16were combined and the solution was given as [39, 40] ) ( ln 1 1 o ox ox t t AB B E (2.18) 0 exp 1 E B AB t o (2.19) where E 0 is the initial field for the time after the cessation of corona charges. The Fowler-Nordheim plot was defined as log(t+t o ) versus 1/E ox which was analogous to ln(J/E 2 ) versus (1/E). The electric field in the oxide was calculated based on Equation 2.4. Any presence of oxide charges in the oxidewas observed by a shift between the C-V curves taken before and after Fowler-Nordheim current stress. Fowler-Nordheim characteristic for un-metallized oxide thicknesses ranging from 500 to 2600 on Si was reported to have a barrier height of 2.9eV and Mox of 0.48[39]. These values are comparable to Si-based MOS devices (see Table 2.3).

PAGE 35

2.5. Constant Current Stress Technique Constant current stress (CCS) test is also known as a current limited or compliance limited stress test method predetermined density of coulombs per second voltage, measured as a function of time, drops abruptly. This abrupt voltage drop is indicative of oxide breakdown (see Figure 2.8). In CCS testing, the current flowing through the oxide is by Fowler adjusted to achieve the d esi the voltage drop across the increases, the electric field in the oxi become trapped within the oxide. constant as this is related to the supplied current. of trapped charges form s Figure 2 8 : An example of a gate voltage by CCS test ing [33] 20 Constant Current Stress Technique Constant current stress (CCS) test is also known as a current limited or limited stress test method [10] It is a measurement technique where a predetermined density of coulombs per second is appl ied to the oxide until the gate voltage, measured as a function of time, drops abruptly. This abrupt voltage drop is indicative of oxide breakdown (see Figure 2.8). In CCS testing, the current flowing through the oxide is by Fowler Nordheim conduction. During CCS, the electric field is esi red current stress factor. T he power supply raises or lowers the voltage drop across the device to maintain the desired current As the flux of charges increases, the electric field in the oxi de increases and these charges may or may not become trapped within the oxide. The generation rate of trapped charges is assumed to be constant as this is related to the supplied current. Subsequently, the c ontinuous formation a c onductive path leading to oxide breakdown. : An example of a gate voltage time characteristic of a MOS device obtained Constant current stress (CCS) test is also known as a current limited or is a measurement technique where a ied to the oxide until the gate voltage, measured as a function of time, drops abruptly. This abrupt voltage drop is indicative of oxide breakdown (see Figure 2.8). In CCS testing, the current flowing During CCS, the electric field is he power supply raises or lowers As the flux of charges de increases and these charges may or may not The generation rate of trapped charges is assumed to be ontinuous formation time characteristic of a MOS device obtained

PAGE 36

21 In Figure2.8, the device sustained an oxide field greater than 13 MV/cm for one minute until reaching an oxide breakdown at15MV/cm. The oxide breakdown mechanism duringCCStesting is related to animpact ionizationprocessand trap creation process [41, 42]. Impact ionization is aprocess where point defects in the oxide arecaused by trapped holes recombiningwith free electrons injected either from the substrate or the gate. Unlike impact ionization, trap creation is a process where electrons with energies greater than 2eV liberate hydrogenous species, such as atomic hydrogen radicals, from thesubstrate. These species maycause interface traps or oxide traps at or near the SiO 2 /Si interface [43]. Non-contact CCS testing was also performed on oxidized Si substrates with the intention to achieve oxide breakdown. This result will be discussed in Chapter 4. Another electrical stress test used to predict the lifetime of an oxide is known as time dependent dielectric breakdown (TDDB). 2.6.Time Dependent Dielectric Breakdown (TDDB) Time dependent dielectric breakdown (TDDB) technique applies either a constant current or voltage stress to a gate oxide at agiven temperatureand monitors the time it takesforthe oxide to breakdown. Elevated temperatures are used to accelerate testing as a means to quickly inducedevice failure. Agate oxide failure is defined when it surpasses a predefinedleakage current limit or drops below a predefined voltage [33]. UsingtheTDDB technique, percolation and other physical models were developed to postulatethe defect mechanismsleading tooxide breakdown.

PAGE 37

22 Fundamentally, all models are basedon the concept thata criticaldefect density leads to oxidebreakdown [8, 9, 43-45]. Other models relating the electric fieldin the oxidetothe time-to-breakdown withtemperature were also developed [3, 46]. Constant voltage stress(CVS)testis a measurement technique where a constant potential is applied to a MOS deviceand the gate current is measured as a function of time. The power supply adjusts the currentto maintain aconstant electric field in the oxide. The breakdown mechanism is similar toa CCStest. In atypical CVStest, a series of capacitors are tested simultaneously using contact probes onagrounded measurement chuck. As each gate oxide is stressed, defects are created at localized areas within the oxide. The time it takes for each capacitor to electrically short or fail is monitored and either the first detectable or final breakdown is recorded. The reliability of the gate oxide is then evaluated by plotting the cumulative probability versus time-tofailure or time-to-breakdown for each group of capacitors. The Weibull distribution, which predicts the reliability of a product and models the failure rate, is commonly used for TDDB statistical data acquisition. The cumulative failure probability in the Weibull distribution for oxide breakdown is described as [45] ) / ( 1 ) ( x e x F (2.20) )) 1 ln( ln( F W (2.21) where F is the cumulative failure probability, x is charge or time, is the charge or time where 63.2% of the samples failed, is the slope parameter, and W is the Weibit or

PAGE 38

23 Weibull parameter. The Weibull cumulative distribution F(x) for various slope values is plottedin Figure2.9[47]. Figure 2.9: Weibull cumulative distribution for a population fraction failing by time [47]. Thisslope parameter is obtained by plotting the Weibit versus ln(x). Plotting the log of the time-to-breakdown versus the appliedelectric fieldin the oxidealso yields the constant, which isreferred toas the electric-field acceleration factor, since this factor isproportional to 2 /1ox E [46].

PAGE 39

24 Figure 2.10: An example of a Weibull plot[48]. Figure 2.10illustrates a TDDB distribution for thirteen 100 m diameter MOS capacitors tested ata constant voltage stress of 37volts at 2500 C [48]. Astraight line represents that the devices had a similar breakdown mechanism. The kink and bend on the curve is usually interpreted as the occurrence of a different breakdown mechanism. In Figure 2.10, there are two bends in the curve, one at 31% and the other at 72%. Early failures in the first group, below 20%, were probably due toextrinsic failures. In the second group, from 25% to 70%,another type of failure mode occurred. The last group lasted for a longer time and failed with a different type of failure mechanism. Reliabilityengineers isolate the type of failure modes and rectifythe first two types of failures. The goal is to have the capacitors sustain longer time durations to shift the distribution to the right.

PAGE 40

25 To characterize the defect density for all oxide thicknesses, stress times, and electric fields, anequation fitted on the Eyring formulation was defined by[3] o T t t kTf E x N 012 .0 32 .0 exp 10 5.121 (2.22 ) where kT is the thermal energy in eV, and f(t/t o ) is the normalized time dependence ofthe trapgeneration. The reported gate oxide breakdown mechanism in SiC-based n-type MOS or nMOS devices after TDDB testingis as follows: 1.After testing 48 nMOS capacitors at room temperature, the slope of the Weibull distribution plot of the charge-to-breakdown indicatedthat the oxide breakdown was due to either wafer properties, such as the epitaxial surface roughness and metal impurities, or intrinsic failures. [49] 2.Two breakdown modes, edge breakdown and dislocation-related breakdown, were reportedafter testing 120 4H-SiC-based nMOS capacitors at room temperature. Oxide degradation was found to be caused by the dislocation defect density. [50] 3.After testing 40 6H-SiC based nMOS capacitors at 1450 C-305 0 C, the TDDB results revealed a high field acceleration factor at fields greater than 7MV/cm. The oxide was reliable if the electric field in the oxide was kept below 5MV/cm at 150 0 C. The temperature wasfound to be inversely proportional to the oxide reliability. The oxidebreakdown mechanism was due to intrinsic failures. [51] 4.After testing 4H-SiC nMOS capacitors and double implanted MOSFETsat 200 0 C, TDDB results showed that the failure mechanism varied at oxide fields

PAGE 41

26 above 6MV/cmdue tocurrent tunneling or impact ionization. After TDDB testing, it was also confirmed,through the use of electron-beam-induced current (EBIC) measurement,thatthe oxide breakdown mechanism was not due tothe epitaxial SiC film defects under the oxide as reported in 1 and 2. Furtheranalysis is ongoingto identify the defects responsible for oxide breakdown. [52] Inthis research study, seventeensiteswere measured on variousoxidesurfacesby SILC testingto obtain similar statistical analysis. The normal or Gaussian distribution function was used to assess the data distribution because ofthe short time durations used forthis analysis. The function depicting the population fraction failing by a given parameter, such as time or voltage, is given as [47] dy y y F y 2 5.0 2 2 1 exp ) 2( ) ( (2.23) where is the population mean, is the population standard deviation, and y is the parameter causing failure, i.e. voltage. A normal cumulative distribution function plot is illustrated in Figure 2.11[47].

PAGE 42

27 Figure 2.11: A normal cumulative distribution function for a population failing by y [47]. The experimental datasetwas sorted in ascending order before theprobability graphwas plotted. The formula used to generate the plotting positionis expressed as 1 100 n i P i (2.24) where P i is the percent cumulative probability, i is the number of measured sites (1,2,3,etc.), andn is the test sample size. The distribution resultswill be shown in Section 4.6. 2.7.Contact Stress Induced Leakage Current (SILC) Technique Contact stress induced leakage current (SILC) is referenced in this study as the technique tested withMOS devices. Stress induced leakage currentis defined as the increase of oxide leakage current after high field stress compared withthe current prior to stress.Itencompasses time dependent components,such astransient and steady-state components [53, 54]andis usually observed atelectric fields between4 to 8MV/cm [33]. The conduction in oxide thicknesses greater than or equal to 50at these fields is

PAGE 43

28 commonlydue to Fowler-Nordheim currentcharacteristics. As previously mentioned, Fowler-Nordheim tunneling occurs when electrons tunnel from the semiconductor conduction band into the oxide conduction band through a triangular barrierat the oxidesemiconductor interface. Stressing the oxide, beyond the Fowler-Nordheim tunneling regime, generates defects in the oxide. As a result, the current flows through the oxide by trap-assisted tunneling [3, 6, 54-58]. It is widely known that as an oxide degrades, under electrical stress, local defects or traps generated within the oxide structure create an increase inthecurrent density, which lead to thermal runaway [3]. Figure 2.12illustratesan energy band diagram of annMOS deviceduring SILC. In this tunneling regime,generated trapped charge sites,containing neutral electron traps, act as stepping stones for tunneling carriers, thereby assisting more current leakage to flow through the oxide [53]. Figure 2.12: A sketch illustrating an nMOS (a) without stress and (b) after SILC. E c =Conduction band energyE f = Fermi energy E v =Valence band energyE c E f Oxide Metal Silicon(a) E v qVox V>0 electrons(b) traps-

PAGE 44

29 These defects represent electron traps, interface traps, positively charge donor like states, or neutral electron traps [11, 45]. Neutral traps may be positively or negatively charged by acquiring a hole or an electron, respectively. This type of SILC has been classified as type-A SILC, trap-to-trap tunneling ortrap-assisted tunneling (TAT) [3]. There are three modes of SILC: type-A, type-B, and type-C. The conduction mechanism for B-mode SILC occurs after partial breakdown, while the conduction mechanism for C-mode SILC occurs after final breakdown. An example ofthese three SILC modes is illustratedin Figure 2.13[3]. Figure 2.13: An example of SILC modes [3]. Based on various annealing kinetics after SILC measurements and charge-to-breakdown measurements performed on MOSFETs, the type of trap generation from SILC was reported to differ from the traps created by charge-to-breakdown[59]. Out of the three

PAGE 45

30 SILCmodes, A-mode SILC will be investigated in this researchusing non-contact SILC testing. This is becausethe current injected in the oxide didnot lead to hardbreakdown. 2.8.Non-ContactStress Induced Leakage Current (SILC) Technique Non-contact stress induced leakage current testing characterizes an oxide in the absence of fabricated devices. The definition of SILC remains the same asincontact SILC. The fundamental principle for this metrology will be discussed in the following sections. 2.8.1.Corona-Kelvin Metrology Early in the 1970s, non-contact Corona-Kelvinmetrology was introduced to investigate thermal oxides grown on silicon substrates in the absence of a metal or doped poly-silicongate [60]. Fowler-Nordheim tunneling characteristicson oxide thicknesses greater than 500wereexamined using corona ions. These ionic charges were deposited on the surface of the oxide for a givenamount of time. After the cessation of charge, the contact potential difference (V cpd )at the surface of the oxidewas monitored and recorded based on the Kelvin method. Corona ions, generated from air, are formed by applying a high DC voltage (kV) to adischarge electrode. The composition of these ions is dependent on the environmental air condition, such as the relative humidity. A positive voltage creates positive corona charges which may either be composed of H + (H 2 O) n ions [61]or [H 3 O] + (H 2 O) n ions[62]. A negative voltage creates negative corona ions composed of n 2 3 O) (H CO n 2 2 O) (H O ,or O [61, 63].

PAGE 46

31 Depending on the voltage polarity, these ioniccharges generatean electric field in the oxide when they are deposited on the oxide surface. They areremoved from the oxide surface by rinsing the substrate with de-ionized (D.I.) water or neutralized by depositing the opposite ionicpolarity [60]. The technique of the Corona-Kelvin method utilizes a Kelvin probeto determine the contact potential difference (V cpd )at the surface of the oxide after the deposition of corona ions. The probe consists of a 2-4mm diameter metalplate referredto as the reference electrode. It isconnected to a lock-in amplifier and other relevant electronics for noise to signal enhancement(see Figure 2.14). The distance between the probeand the surface of the oxide is nominally less than a fraction of a millimeter. Figure 2.14: A sketch of Kelvin probe measurement. The probe vibratesverticallyat alow frequencyto induce an alternating current (Jac ) through the circuitry. This currentis defined as t C V V J o cpd b ac sin (2.25) where V b is aDCbias voltage andCo is the capacitance between the substrate and the reference electrode. To extract V cpd from Equation 2.25, the DC bias voltage is variedto Kelvin Probe Electronics V b + SiO2 Wafer Vibrating Ref. electrode Corona ions

PAGE 47

32 null the electric field,created between the electrode and the substrate. The contact potential difference (V cpd ) is then determined when the DC bias voltage renders the displacement current to zero. At this instance, the bias voltage value is equal to V cpd The contact potential difference is expressed as sb ox ms b cpd V V V V (2.26) where ms is the semiconductor-reference electrode work function difference, V ox is the voltage drop across the oxide, and V sb is the voltage drop across the surface barrier or space charge region. The use of ions generatedby corona discharge followed by measuring the contact potential difference at the oxide surface was classified as Corona Oxide Characterization of Semiconductor (COCOS) metrology in the late 1990s[5, 64-67]. The integrity ofthin oxides on Sihas been investigated using thismeasurement approach [68, 69]. Commercial toolsusing this metrology for whole wafer characterizationare manufactured by Semiconductor Diagnostics,Inc. (SDI), Tampa, Florida [70]. Characterization parameters include acquiring the flat band voltage, interface trap density, total dielectric charge, SILC value, andthecapacitance of anas-grownoxide film [69, 71]. Biasing the surface of the oxide with corona charges is the fundamental criterion to perform non-contact capacitance-voltage (C-V) measurements and currentvoltage (I-V) measurements.

PAGE 48

33 2.8.2.Capacitance-Voltage Measurements Capacitance-voltage (C-V) measurements detect the influence of interface trap charges and other charge defects within the oxide. In addition, C-V measurements are used to determine other characteristics of the oxide such as the oxide thickness, the oxide breakdown field, the work function difference between the semiconductor and the gate electrode, and the oxide conductivity [72]. Capacitance-voltage plots characterize the total capacitance of an oxide-semiconductor system withinfour ideal regionsof operation: accumulation, flat band, depletion, and inversion regions. For a low frequency (<1kHz) measurement, these regions are definedas follows[73] 1.Accumulation: the concentration of majority carriers near the oxidesemiconductor interface is larger than in the bulk of the semiconductor. The oxide capacitance(Cox ) is ox g r ox t A C 0 (2.27) wheretox is the oxide thickness and A g is the area of the gate. 2.Flat band: equilibrium state in the bulk of the semiconductor (no band bending). The flat band capacitance(Cfb )is s ox s ox fb C C C C C (2.28) where C s is the semiconductor capacitance. 3.Depletion: the concentration of majority carriers near theoxide-semiconductor interface is smaller thaninthe bulkof the semiconductor. Thecapacitance(C)is

PAGE 49

34 d ox d ox C C C C C (2.29) where C d is the depletion layer capacitance. 4.Inversion: the concentration of minority carriers exceeds the majority carriers near the oxide-semiconductorinterface. Thecapacitance is equal to the oxide capacitance, C ox If no oxide charges are present within the oxide, the flat band voltage is equal to the semiconductor-metal work function. In thepresence of oxide charges, the flat band voltage is defined as[33] ox oxt ms fb C Q V (2.30) where Q oxt includes the interface trapped charge, the fixed oxide charge, the oxide trapped charge, and the mobile oxide charge. UnlikeSi-based MOS systems,SiC-based MOS systemsat room temperaturecannot form an inversion region due to the low thermal generation rateof minority carriers. One way this region can be formed is through the introduction of these carriersinto the substratebyan ion implantation of a doped source region. In the non-contact low frequency or quasi-static C-Vmethod, a pulse of corona charge Q c(n) is deposited on the surface of the oxide. The semiconductoris placedinto a region where the concentration of electrons near the oxide-semiconductor interface is larger than in the bulk of the semiconductor. This region is referred to as the accumulation region for an n-type semiconductor or the inversion region for a p-type

PAGE 50

35 semiconductor. The deposited corona charges are imaged in the semiconductors space charge region or surface barrier and in any traps in the oxide, which can exchange charge with the semiconductor [66]. The energy band diagram illustrating these charges,with respect to the reference electrode, is seen in Figure 2.15. Figure 2.15: Energy band diagram after the deposition of corona charges. Immediately, after the deposition of corona chargeson the surface of the oxide, atime varying contact potential difference V cpd(n)(t) is then measured(see Equation 2.26).This voltage is measured with a precision of 0.1mV. Afterwards,seriesof corona charge quantities, Q c(n+1) are deposited on the surface of the oxide to slowly transition the semiconductor from accumulation to depletionto inversion. Each increment of charge is followed by the measurement of the contact potential differenceas a function of time, V cpd(n+1) (t). The total capacitance, C tot is derived from )1 ( ) ( )1 ( ) ( n cpd n cpd n c n c cpd c tot V V Q Q dV dQ C (2.31) qVox4HSiCE c E f E v Oxide ElectrodeqVsb q f s q f m Q it Q f Q otQ m Q c E i

PAGE 51

36 where n and n-1 represent the immediate and immediate past increments of charge application. In Equation 2.31, the change of the contact potential difference ( V cpd ) eliminates the contribution of ms and is expressed as SB ox cpd V V V (2.32) Figure2.16shows an example of anon-contact C-V characteristicof a150 thermal oxide on a p-type Sisubstrateand a 400 afterglow oxideon an n-type 4H-SiC substrate. These C-V characteristicsweremeasuredin the dark.

PAGE 52

37 Figure 2.16: Non-contact C-V characteristics of(a) a 150 thermal oxide on a p-type Si and (b) a 400 afterglow oxide on an n-type 4H-SiC. Themostcommonly usedcalibrated corona charge,deposited during C-V measurements, was 0.016 C/cm 2 In Si, this density of charge does not damage the oxide[74]. (a)(b)C ox = C inv C ox C ox C d C fbC fbC d

PAGE 53

38 2.8.3.Equivalent Oxide Thickness (EOT) Measurement Capacitance-voltage characteristicsof an oxide measured in either the dark or lightwas used to calculate its equivalent oxide thickness. The equivalent oxide thickness was determined in the region where the semiconductor was placed into accumulation. The C-V characteristic of the oxide, measured in light, is obtained with the same quantity of charge and step increments used in the dark. When the oxide is measured under illumination, the light generates excess electron-hole pairs in the semiconductor depletion layer or space charge region. As a result, the voltage drop across the surface barrieris minimizedto a negligible value. When the oxide is measured in the dark, the voltage drop across the surface barrier isminimized to a negligible valuein accumulation. In bothcases, the oxide capacitance isaccurately determinedsincethe measured voltage is onlytakenacross the oxide(see Equation 2.33). acc ill cpd c acc dark cpd c ox dV dQ dV dQ C (2.33) where C ox is the oxide capacitance measured in thedark or light. The equivalent oxide thickness is then calculated,in reference to SiO2 ,as ox o r C EOT (2.34) where r is the dielectric constant and 0 is the permittivity of vacuum. The EOT can also be extracted from the slope of the deposited corona charge and the measured voltage drop across the oxidewithin the accumulation region.

PAGE 54

39 Equation 2.34is also referred to as the capacitance equivalent thickness (CET). It has been reported that EOT obtained for oxides grown on 4H-SiC substrates is comparable to Ellipsometer measurements [75]. Figure2.17illustrates a non-contact C-V measurement characteristicof a 499 afterglow oxide on an n-type 4H-SiC substratemeasured inthe dark and in the light, respectively. Figure 2.17: Non-contact C-Vcharacteristicof a 499 afterglow oxideon an n-type 4HSiCmeasured in the dark (circles) andin the light (asterisks). In Figure 2.17, the C-V curve taken in the light does not deplete because of theexcess minority carriers generatedby the light.

PAGE 55

40 2.8.4.Current-Voltage Measurements Prior tonon-contact current-voltage(I-V) measurement, a blanket of positive corona charge (~10 -7 C/cm 2 ) is deposited overthe entire surface of the oxide to minimize lateral charge spreadingat a measurement site. Current-voltagemeasurements are performedunder illuminationto measure only the voltage drop across the oxide. A quantity of corona ionsis deposited on the surface of the oxide to generatean electric field in the oxide. As ionic charges buildup on the surface of the oxide, theoxide voltage increases linearly with time. The voltage drop across the oxide remains linear until the injection of electrons from the substrate occurs by Fowler-Nordheim tunneling. The onset of the Fowler-Nordheim voltage is equal to the conduction band offset voltage. As the flux of corona ions induces tunneling of electrons from the semiconductor into the oxide conduction band, the buildup of ionic charges on the surface of the oxide stops when the electron current density reaches the corona ionic current density, which is controlled by the power supply [64, 67]. After the cessation of charge, the corona charge decayis expressed as dt dV C dt dQ cpd c (2.35) This condition is referred to as a Self-Adjusting Steady State (SASS) condition [65, 76]. Thecurrent density, Jexp is calculated from the measured dielectric voltage decay given by

PAGE 56

41 exp ill cpd ox r o dt dV t J (2.36) Based on the classical non-contact Fowler-Nordheim analytical equation(see Equation 2.18), theoxide voltage decay was fitted toa logarithmic time dependencecurve as shown in Equation 2.37. t t b a b dt dV dt dV ill cpd ox 2 1 1 1 ln (2.37) where a 1 and b 1 are constantsfrom the logarithmic curve fitting and arerelated to the Fowler-Nordheim coefficients, A and B.[39, 76]. An example ofvariousoxide voltage decays with respect to apositive corona current density, in A/cm 2 on Siis illustrated in Figure 2.18. Figure 2.18: Oxide voltage decay after corona charging [76].

PAGE 57

42 In Figure 2.18, theelectric field in the oxide is proportional to the corona current density. For longer time durations, the precision of the SASS voltage increases. The electric field in the oxide can be calculated with further accuracy by including ms ox ms ill cpd ox t V E (2.38) In SiO 2 /Si systems, holesinjected from the Si valence bandisnegligible duetoahigher potential barrierof 4.73eVat the oxide-semiconductor valence band. Non-contact SILC testing is basedon non-contact I-V measurementprinciples.Itsmethodology will be discussed in the nextsection. 2.8.5.Stress-Induced Leakage Current Method A controlled ionic current is used to perform non-contact SILCtesting. Before stressing the oxide to calculate SILC values, asufficient quantity of corona chargesis deposited for a few secondsto produce substrate injection of electrons. This quantity of charges induces an electric field in the oxidegreater than5MV/cm. The contact potential difference ismeasured ina 20-second interval following a 5-second charge deposition. The calculated current densityisfittedtoa predictedFowler-Nordheim current. It isassumed that this small charge fluence doesnot give rise to defects above the Fowler-Nordheim current. Fluence isdefined asthe product of the calibrated ionic current and the deposition time. Ithas unitsof coulombs per cm2 (C/cm 2 ). After confirming that the experimental current was by Fowler-Nordheim conduction, the corona deposition time wasincreased. The oxide is stressed beyond the Fowler-

PAGE 58

43 Nordheim tunneling regime andelectron traps or defects are created within the oxide,as shown in Figure 2.19. Figure 2.19: Energy-band diagram during non-contact SILC. The voltage decayismonitoredand the total experimental current densityis calculated based on Equation 2.36. The total experimental current density (J t ) is expressed as J J J N F SILC t (2.39) In Equation 2.39, the component of the calculated currentdue to SILC is subtracted from the Fowler-Nordheim current. The SILC density is typically one or more orders of magnitude above the Fowler-Nordheim current. Therefore, the total calculated current is essentially equal to the SILC value. However, the thickness is not calculated from EOT but fromthemeasuredFowler-Nordheimcurrent after the 5-second charge deposition. Since the Fowler-Nordheim current density is dependent on the electric field in the oxide, qVox electronstrapsQ c SemiconductorSiO2

PAGE 59

44 the current density versus the oxide electricfieldis plotted to determine the oxide thickness. The oxide thicknessis adjusted to fitthetheoretical Fowler-Nordheim current curvefor a given oxide-semiconductor system. This thickness adjustment is accurate if the effective mass in the oxide and effective barrier height values areprecisely known. The theoretical Fowler-Nordheim currentcurve is based onMox of0.36and an effective barrier height of 3.15eV. These values are the default values foratmospheric thermal SiO 2 /Si systemsdefined in the FAaST 230 tools software. Figure 2.20 illustrates an example of this method for a thermal SiO 2 /Si system. Figure 2.20: Example of determining thickness with the Fowler-Nordheimcurrent density [64].

PAGE 60

45 This determined thickness also matched the independently measured EOT value for this thermal oxide grown on a p-type Si wafer. Once the thickness is determined, SILC valuesarecalculated. Aconstant oxide field value is then chosen to compare the SILC densities attheir respective stress fluence. An increase in the density of the leakage current indicatesa localized weak spot in the oxidestructure, which may besusceptible to more defect formation. A weak spot maybe due to process contamination. Preliminary,non-contact SILC testing was first used to characterize two oxide growth methods, pyrogenic steam oxide and afterglow oxide, on n-type 4H-SiC substrates [77]. The results revealed that the afterglow oxideshowed less susceptibility to defect formation at a field of 6MV/cm and a charge fluence of 1.55mC/cm2 2.9.Statistical Issues in Device-Based Measurements Even thougha new technique using dual voltage and time integration was developed to characterize long-term TDDBmeasurements[44], thereexisttwo major disadvantagesin obtaining TDDB distributions. First, it is time consuming when testing various voltages, currents, and temperatures. This data acquisition may take months or even years to report. Second, TDDB testingrequires a large group of MOS devicesto be tested consecutively to obtain statistical distributions. These distributionspredict the lifetime of the oxide and the intrinsic or extrinsic mechanism responsible for breakdown. Statistical data acquisition of an oxidation process requires the correct mathematical expression for the time-to-breakdown (t bd ) or time-to failure (TF), which is dependent on the electric field induced in the oxide. Based on either the Anode Hole Injection model ((1/E)-model) or the Thermochemical model (E-model), the time-to-failure can be expressed as [3]

PAGE 61

46 E-model: ox E kT Q TF 1 ) ln( (2.40) (1/E)-model: ox E G kT Q TF 1 ) ln( 2 (2.41) where and G represent the field acceleration parameters in the two respective models, Q 1 is the thermal activation energy required for bond breakage, and Q 2 is the thermal activation energy associated with the current-induced hole injection into the oxide. Since the electric field cannot be directly measured usingthecurrent state-of-art metrology techniques, theaccuracy of thevoltage and thickness measurementsarecritical. Any errors in the thickness can lead toanerroneous field calculation (see Equation 2.4) resulting inan over or under estimation of the lifetime of the oxide. Probable causes of anoxide thickness error from a C-V measurement are: an incorrect value of the gate area and an incorrect dielectric constant value. The cost to predict the lifetime of an oxide grown on SiC canbeexpensive,asthese substrates aremore costlythan Si substrates. 2.10.Chapter Summary The reliability of an oxide can be assessed using electrical stress testing methods. Thereportedfailure mechanisms ofoxides forSi-basedand SiC-basedMOS devices, under high electric fields,werepresentedusing TDDB measurements. An exact defect mechanism leading to the breakdownof an oxide forSiC-based MOS devices has not been reported. The conduction mechanisms, Fowler-Nordheim and Poole-Frenkel emission, were explained to analyze the conduction obtained in the experimental oxides

PAGE 62

47 using non-contact SILCtesting. The principletechnique of non-contact C-V measurementsand non-contact SILCtesting werealso introduced.

PAGE 63

48 Chapter 3.Experimental Procedures 3.1.Non-Contact Instrumentation TheIon-Drift (ID) Spectrometer tool and theFilm Analysis and Substrate Testing (FAaST) 230 tool were the non-contact metrology tools utilized to investigate the oxide/4H-SiC interface. In the following section, a brief description of each tool will be presented. 3.1.1.Film Analysis and Substrate Testing (FAaST) 230and Components The commercial Film Analysis and Substrate Testing (FAaST) 230toolwas modified to investigate oxides grown on 4H-SiC substrates. One unique modification implemented to the tool was the use of a UV light ( =375nm) diode instead of a green light diode. The light generates free minority carriersto eliminate the surface barrier voltage drop across the 4H-SiC substrates space charge layer. The components within the tool, whichwas utilized to perform non-contact C-Vmeasurements and non-contact SILCmeasurements, are illustrated below.

PAGE 64

49 Figure 3.1: Schematic of the measurement apparatus based on ref.[64]. In Figure 3.1, the charging station located on the leftcomprised ofa needle type corona discharge electrode and light diodes, and the measurement stationlocated on the right included a Kelvin probe and light diodes. The probe consistedof a platinum reference electrode plate and the vibrating frequency was set to 1600Hz.Within the corona dose generation modulation, a high DC voltage between + 5KVto + 10KV supplied the corona discharge electrode to create corona ions. The probe and the corona source were positioned about 500 m above the oxide surface. The distance of the corona discharge electrode above the wafer was automatically controlled by the computer within a calibrated range to achieve the desired corona dose. The corona charge deposition on the dielectric surface was measured either by the instrument directly or calibrated. The density of the charge deposition was influenced by the level of thevoltage, the charging time, the distance between the corona sourceand the oxidesurface, and the charge interaction with any surface chargeson the oxide[78]. The ion production could beset Wafer Moving Chuck Motorized arm Corona dose generation Kelvin Probe Electronics Discharge electrode Light diodeVibrating Ref. electrodeSiO2 Corona ions

PAGE 65

50 from 10 -7 to 10 -4 A/cm 2 Afterthecessation of charge deposition, in less than a coupleof seconds, the Kelvin probe determined the contact potential difference (Vcpd )at the deposition site with a precision of 0.1mV(refer to section 2.8.1). The probe transfer to thissite was achieved by amechanical actuator. The voltagedecay measurement acquisition wasterminated after a predefined time. The diodes were used to illuminate the oxide surface. As previously mentioned, when the dielectric is measured under illumination, the light generatesexcess electronhole pairs in the semiconductor depletion or space charge layer, which minimizedthe voltage drop across the surface barrier to a negligible value. The change ofthe contact potential difference was equalto thechange in thevoltage drop across the oxide. Subsequently, thechange in thevoltage drop across the surface barrier was obtainedby [66] ill cpd dark cpd sb V V V (3.1) ox sb dark cpd V V V (3.2) ox ill cpd V V (3.3) For non-contact C-Vmeasurements, the userinputthedesired charge value in units of q/cm 2 for a particular number of incremental steps. The softwarethencalculatedthe necessary corona current and timeneeded to generatethis charge density. On the contrary, for non-contact SILCmeasurements, the userinput a value from + 0.2 to + 0.8 in the corona current parameter.

PAGE 66

51 This parameter was proportional to the power supplyand it directly producedacalibrated ioniccorona current in the A/cm 2 range. The corona deposition timewas also entered by the user. The measurement cycle dependedon the number of stress fluence (corona current multiplied by thedeposition time) level(s)requested. 3.1.2.Ion-Drift Spectrometer TheIon-Drift Spectrometer was an experimental toolbased on the FAaST 230 tools contact potential difference measurement principles. One of the major differences between this tool and the FAaST 230 tool was the corona discharge electrode. The corona source encompassed a wire discharge electrode and its enclosure (see Figure 3.2). Figure 3.2: Sketch of the (a) wire enclosure, (b) discharge distance to the wafer, and (c) aperture diameter 4cm 2cm1.5 cm 1.5 cm Opening in enclosure Wire enclosure Small aperture 6 mm Large aperture 14 mm5 mm wafer Wires positions 5 mm19 mm 9 mm aperture (a)(b) (c)

PAGE 67

52 The opening at thetop of the wire enclosure allowed air to flow into the enclosure so corona ions maybe created. The diameter of the apertures opening, which was placed at the bottom of the enclosure, controlled the amount of corona charges deposited on the oxide surface. The user interface was similar to the one used in the FAaST 230 tool for non-contact SILCmeasurements. The userinput a value from + 0.2 to + 0.8 in the corona current parameter and the corona deposition time in seconds. This tool was used as a training toolfor three purposes: 1. to understand the non-contact Corona-Kelvin measurement method; 2. to verify the current limitation from the corona source; and finally, 3.to indicate the influence ofconstantcoronacurrent stress on C-V measurements. 3.2.Pre-Oxidation Cleaning Procedure Prior to oxidation, if the wafers were previously oxidized, the wafers were submerged in a (10:1) H 2 O:HF solution for 5 minutes to remove the oxide. This was followed by de-ionized (D.I.) water rinse cycles and a (2:1) H 2 SO 4 :H 2 O 2 or piranha clean for 10 minutes. The wafers were then rinsed and dipped in a (1:1) H 2 O:HCL solution for 10minutes and rinsed. Afterward, the wafers followed a standard cleaning procedure for bare silicon. Standard clean 1 (SC1) and standard clean 2 (SC2), developed by the Radio Corporation of America[79], removed organic and metallic surface contamination from the wafer. Figure 3.3 illustrates the RCA cleaning procedure used in this study.

PAGE 68

53 Figure 3.3: RCA cleaning procedure. The D.I. water used for cleaning and rising had aresistivity of18Mohm-cm. The HF dip after SC1 and SC2 removed anythin layer ofchemical oxide formed during the standard cleans. Subsequently, the wafers were oxidized using either an afterglow (AG) furnace system or an atmospheric furnace system. 3.3.Oxidation Sequence/Interval Experiments The substrates used in this work were commercially available80 off-axis n-type 4H-SiC epitaxial wafers.These 3-inch wafers had ann-type epitaxial layer, grown on the (0001) Si face, with a nitrogen doping concentration of approximately 5x10 15 /cm 3 This concentration was determined by a non-contact doping profiling method [80]. These wafers were reused or recycledallowingsubstrate cost-savings to the project and the characterization of a variety of oxidation processes. HF Dip10:1 H 2 O:HF 22 0 C 5 minutes SC1 6:1:1H 2 O:H 2 O 2 :NH 4 OH 65 0 C 10 minutes HF Dip2 0:1 H 2 O :HF 22 0 C 5 0 seconds SC2 6:1:1H 2 O:H2 O 2 :HCL 65 0 C 10 minutes HF Dip2 0:1 H 2 O:HF22 0 C 5 0 secondsRinse Rinse Rinse Rinse Rinse and N2 Dry

PAGE 69

54 3.3.1.Brief Description of Afterglow(AG)Furnace The afterglow(AG)or remote plasma vacuum furnace system has been shown to oxidize 4H-SiC at a lower oxidizing temperature, time, and pressure [75, 77]than possible for this material when compared toan atmospheric furnace system. The afterglow system is illustrated in Figure 3.4. Only the primary gas inlet was used. The gases available were Argon (Ar), Oxygen (O 2 ), Nitrous Oxide (N 2 O), Nitrogen (N 2 ), and Forming gas (H 2 :N 2 ). Forming gas (FG)was a mixture of 4-5% H2 : 96-95% N 2 A microwave plasma source generated neutral atomic and excited molecular species of O2 N 2 O, and FG. The furnace idle temperature and pressure in a N 2 ambient was 400 0 C at 0.3 Torr. To load the wafers, the system was brought up to atmospheric conditions, where the wafers werecentered in the growth zone of the furnaceand placed under vacuumin a N2 ambient. Figure 3.4:Sketch of afterglow furnace system [77].

PAGE 70

55 In a non-excited media, the temperature was stabilizedto the oxidation temperature. The standard oxidation was performed at 850 0 C for 90 minutes at one Torr in the remote plasma excited mixture of O 2 : N 2 O: FG with a gas flow rate of (3: 0.2: 0.5) liters/minute. Following oxidation, the wafers were either annealed at a specific temperature and time in a non-excited media or unloaded. Argon was the standard gas used to stabilize the temperatureeither between anneals or prior to unloading. The wafers were then unloaded in a N 2 ambient at a specific temperature. 3.3.2.Afterglow(AG)Oxidation Parameter Variations Thirteenafterglow oxide recipes were reportedin this work. Three major measurement sequenceswill be presented along with their purpose. A detailed listing of each afterglow oxide recipe can be found in Appendix A. The oxidation parameter variations included surface pre-conditioning, oxidation time variation, and post-oxidation annealing. The objectiveof the first stage in this studywas toestablish and verify the conduction mechanism of annealed andnon-annealedoxides,under a high oxide field (greater than 5MV/cm),as Fowler-Nordheim conduction. The annealed AGrecipes werevaried by the gas mediamixtureand temperatureused for post-oxidation annealing (see Table 3.1).Thepost-oxidation annealsused for comparison wereArgon (Ar) anneal, Re-oxidation (Re-ox) anneal, Oxygen-nitrousanneal, orOxygen-FG anneal.

PAGE 71

56 Table 3.1: Annealed afterglow (AG) recipes. AG Anneal Recipe Anneal Description Duration Time (min) I_A Re-ox anneal 60 I_B Re-ox anneal 25 II Ar anneal 60 III Re-ox and Ar anneal 120 IV Oxygen-FG anneal 120 V Re-ox anneal 360 VI Oxygen-nitrous anneal 120 The AG oxides referred to as AG I_B, AG IV,AG V,and AG VI weremodified in the following manner: in AG I_B, the oxidation time was reduced from 90 minutes to 40 minutes; and in AG IV, AG V,and AG VI, instead of Ar, the same anneal gas ambient was used to stabilize the temperature to the anneal temperature. These minor variations were included to observe the influence of growth thickness and annealvariationon the Fowler-Nordheim characteristics. The variation in the post-oxidation anneal was to verify the influence of charges in the oxide on the Fowler-Nordheim plot,since it was reported that thedensity of interface states isdecreasedbypost-oxidation anneals [14]. Table 3.2 lists themodification within thenon-annealed recipes.

PAGE 72

57 Table 3.2: Non-annealed afterglow (AG) recipes. Non-Annealed AG recipes Modified Process Unload Temperature AGWI N/A 600 0 C AGWII No FG pre-treatment 600 0 C AGWIII 10 minutes of oxidation 850 0 C AGWIV No FG pretreatment and 10 minutes of oxidation 850 0 C AGWV Oxidation mixture media 600 0 C AGWVI Oxidation mixture media 600 0 C Theinfluence of the surface pre-conditioning, the oxidation growth time,the variation of the oxidation mixture media,and the unloading temperature on the Fowler-Nordheim characteristics wereobserved for the non-annealed AG oxides. The measurement sequence is illustrated in Figure 3.5. Figure 3.5: Measurement sequenceto calculate JF-N Non contact CV C= dQ c / dV cpd Calculate J F N C ox ( dV ox / dt ) Determine EOT= e r e 0 /C ox

PAGE 73

58 Theresults of the first stageprompted theobjective for the second stageof measurements. In this stage,the trapped charge distribution within AGoxideswas examined. The trapped charge distribution may existin the bulk of the oxide or at the oxide interface. AfterglowIVoxidewas the primary afterglow oxide process recipe used for this experiment. To perform this task, the oxidized substrates were etched in a diluted HF solution bath of 100:1 (H 2 O:HF)at 240 C. Figure3.6shows the etch patternand quadrant created on asubstrate. Figure 3.6: Etch patternand quadrant designationon the substrate. The wafers were then dehydratedusing three methods:a hotplate at 2000 C for 5 to 60 minutes, a rapid thermal processing (RTP) anneal in Ar at 800 0 C for 2 minutes, and an afterglow non-excitedAr annealat6000 C for 20 minutes. The measurement sequence is illustrated in Figure 3.7. Reference B F E C A PrimaryD

PAGE 74

59 Figure 3.7: Measurement sequenceto calculate trapped charge. Finally, the objective for stage threewas to analyze the potential breakdown mechanism andassessthe growth oxide consistency usingSILCtesting at seventeen sites on the wafer. Afterglow recipes AGWI, AG I_A, AG II, and AG III were repeated at least three times on the same substrate. Theinfluence of the environmental/processing conditions, which included unintentional contamination from the cleaning procedure or the furnace, wasobservedby the leakage current measurements. The measurement sequence is illustrated in Figure 3.8. Non contact CV C= dQ c / dV cpd Stress Oxide for 15s calculate J15s Calculate Trapped Charge Q t =( D V FB )C ox Calculate JF N C ox ( dV ox / dt ) [J=Cox ( dV ox / dt )] Determine EOT= e r e 0 /C ox Non contact CV C= dQ c / dV cpd

PAGE 75

60 Figure 3.8: Measurement sequenceto obtain SILC data acquisition. 3.3.3.Thermal Oxidation To compare the AG oxides to atmospheric oxides, n-type 4H-SiC oxidized samples, Thermal TA and Thermal TBwere obtained from two separate industrial laboratories. The oxide on these 3-inch diameter wafers were grown using an atmospheric furnace. Oxidation occurred at a temperature greater than 1100 0 C followed by anitric oxide (NO)anneal at a temperature not exceeding 11500 C. Non contact CV C= dQ c / dV cpd Non contact SILC Calculate J F N C ox ( dV ox / dt ) J SILC= J afterstress J beforestress[J=Cox ( dV ox / dt )] Determine EOT= e r e 0 /C ox

PAGE 76

61 Chapter 4.Experimental Results and Discussion 4.1.Measurement Test Conditions To ensure proper wafer backside contact with the measurement chuck, the oxide was stripped from the carbon face of eachwafer using hydrofluoric (HF) vapor. Following the 5-minute HF vapor etch, the wafers were subjected toD.I. water rinse cycles and dried with a nitrogen air gun. Since the relative humidity(R.H.)affected the amount of corona discharge ionsgenerated in air, all measurements performed on the FAaST 230 tool were takenat an R.H. level between 45% and 50%. The room temperature ranged from 19 to 20 0 C. On the contrary, the Ion Drift Spectrometer was not in a controlled environment and as a result,the relative humidity and temperature was recorded for each measurement. Unless otherwise noted, the oxide field range, chosen to calculate thepredicted Fowler-Nordheim characteristics, was based on thefield range of theexperimental data. 4.2.Non-ContactCoronaCurrent Stress Two sets of experiments were performed on the Ion-Drift Spectrometer:constant coronacurrentstressand corona deposition time versus C-Vcharacteristics. To have a known reference, thermal oxides on p-type (100) Si wafers were used for these sets of experiments. The objective of the first experiment was to determine the current limitation of the measurement tool.

PAGE 77

62 Eachoxide was stressed with negative corona charges until the current density (following such stress) reached between10-6 A/cm 2 and10-5 A/cm 2 Negative charges were used to place the semiconductor into accumulation to induce substrateinjection ofelectrons. Two oxides, 320 and 365 thick, were stressed. The large aperture was chosen since more corona charges were deposited on the oxide surface. As shown in Figure 4.1, the current density range of 10 -6 to 10 -5 A/cm 2 was notachieved even though the electric field in the oxidewas greater than 13MV/cm. The electric field breakdownfor silicon dioxide ranges from 10 to 14MV/cm. Table 4.1 shows the measurement conditions and parameters. Figure 4.1: Current density versus V cpd for various corona currents.

PAGE 78

63 Table 4.1: Stress current measurement parametersand conditions. Oxide Thickness () Room Temp ( 0 C); R.H.(%) Corona Current Parameter Corona Deposition Time (s) Pulse Duration (s) Oxide Electric Field (MV/cm) 320 23; 68 -0.3 15 105 14 365 25; 70 -0.8 3 57 13 365 24; 66 -0.5 10 110 14 Various corona currents and times were tested to obtain the desired breakdown field. As seen in Figure 4.1, the highest current density achieved was approximately 0.01 A/cm 2 Even though more charges were deposited at the maximum current parameter, the voltage potential,along with the current density, leveled at a particular value. It was concluded that the corona source was current limited. The objective ofthe second experiment was to observethe influence ofthecorona deposition time versus C-Vcharacteristics. Using standard device based contact measurements, capacitance-voltage plots can beused to illustrate the influence ofthe interface trapdensity(see Figure 4.2).

PAGE 79

64 Figure 4.2: Theoretical ideal D it =0 (before stress) and D it 0 (after stress) [33]. In Figure 4.2, the minimum capacitance increased as a result of the formation of interface trap defects. The same concept was predicted to occur once the oxide was stressedwith corona ions. The measurementprotocol included: 1. obtain theinitial C-V, 2. stress the oxide using negative corona charges for 10s, and 3. re-measure the C-V. Step two was repeated for 10s, 11s, 12s, 13s, and 14s. The total cumulative time was 70s. Table 4.2 reiterates the measurement procedure.

PAGE 80

65 Table 4.2: Stress measurement protocol. Measure Time (s) Cumulative Time (s) C-V 1 st Stress 10 10 C-V 2 nd Stress 10 10+ 10= 20 C-V 3 rd Stress 11 20 + 11= 31 C-V 4 th Stress 12 31+ 12= 43 C-V 5 th Stress 13 43+ 13=56 C-V 6 th Stress 14 56+ 14= 70 The corona current parameter was set at -0.8. Each C-V measurement was done at an incremental charge of ~0.03 C/cm 2 Below illustrates the results fortwo different sites on a 150 thick oxide.

PAGE 81

66 Figure 4.3: Capacitance-voltage characteristics versus corona stress. (a) (b)

PAGE 82

67 In Figure 4.3, as the time of stress increased up to thecumulative time of 70s, the minimum capacitance and depletion width increased. The minimum capacitance is the value at which the curve depletes before weak inversion occurs. This phenomenonmay be due to the creation of traps or a high charge density on the oxide surface. At the depletion region,the total capacitance without interface trap charges is equal to the series connection of the oxide capacitance and the depletion layer capacitance (see equation 4.1). it d ox it d ox t C C C C C C C (4.1) where C ox is the oxide capacitance, C d is the depletion capacitance, and C it is the interface trap capacitance. As C it increased, the total capacitance in the depletion region also increased. Since the charge deposited on the oxide surface cannot be directly measured, it was calculated using a calibrated capacitor. The results from this capacitor revealed that the deposited charge was in the range of micro-coulombs. As a result of the experiments, the Ion-Drift Spectrometer tool could be used for non-contact C-V measurements but the location of the tool was not ideal for non-contact SILC measurementson 4H-SiC oxidized wafers. The relative humidity and room temperature varied on a daily basiswhich created inconsistencies on the quantity ofionic charges. Hence, the deposited charges could not be efficiently calibrated for a sequence of repeatable measurements. In summary, it was shown that the corona source is current limited and highcorona charge fluencesinduced trapped charges in the oxide.

PAGE 83

68 4.3.Potential Factors Affecting the Accuracy ofthe Contact Potential Difference The contact potential difference (V cpd ) was determined by the Kelvin methodas stated in Section 2.8.1. A Kelvin probe is connected to a lock-in amplifier and other relevant electronics. As the reference electrode vibratedperiodicallyat a defined frequency, thelock-in amplifier detecteda displacement current defined by Equation 2.25. TheDC bias voltagevarieduntil this current approachedzero. The precision of the determined V cpd was within 0.1mV. According to Equation 2.25, the height of the probe and the frequency of the sinusoidal signal affect the determined V cpd Internal calibrationswereperformed on known SiO2 /Si systems to verify the accuracy of the measurement. The height of the probe from the surface of the substratewas calibrated between 200 m and 500 m. After corona deposition on the surface of the oxide, V cpd wasmeasured. There wereessentially three unknown components of Vcpd, the semiconductor-reference electrode work function difference, the voltage drop across the oxide and the voltage drop across thesemiconductorsurface barrier (see Equation 2.26). The work function varied based on the semiconductor doping levels and condition of the probes surface. It primarilycauseda shift in thecalculatedelectric field. This work function offseteffect was not observedwhen determining the capacitance and current density because the change in the contact potential difference was calculated (see Equation 2.32).The voltage drop across the oxide and the voltage drop across the semiconductor surface barrier were determined by following the procedure described in Section 3.1.1.

PAGE 84

69 The contact potential difference was measured as a function of time. A logarithmic regression analysis of the inverse of the measured V cpd versus time was performed to determine thecharacteristic of the curve,which was related to the predicted Fowler-Nordheim current density equation coefficients. The corresponding R-squared value determined if the datafittedtothis function. The R-square value ranged from 0.997 to 1 for all measured data. The current density was then calculated at each increment of time from the fitted voltage decay and the oxide capacitance (see Equation 2.36). 4.4.Fowler-Nordheim Conduction in Afterglow Oxide and Thermal Oxide on 4H-SiC After the pre-chargingtreatment, all measured oxides were given a high dose of positive corona chargefora 5-seconddurationunder illumination. The substrate was placed into deep accumulation. The corona current parameterwas set to 0.8. This low charge fluence, 40 C/cm 2 createda high oxide field for the occurrence of Fowler Nordheim tunneling and established the maximum voltage drop across the oxide. This maximum voltage was knownas VSASS. Theoxide voltage (Vox )decaywas monitored for 20 seconds after the cessation of corona charging. The average EOT value for each oxidized wafer was obtained from their accumulation capacitance. This thickness was used in the calculation ofthe experimentalFowler-Nordheimcurrent density.The valence band offset in reference to SiO 2 for 4H-SiC is 3.05eV as oppose to 4.73eV for Si[14]. Substrate injection of holes, from thisbinary semiconductor,was energetically not feasible. The injection of holes wouldrequire enough energy to goagainst the direction of theelectricfield in the oxidecreatedby a quantumof positive corona charge.

PAGE 85

70 In this regards, only substrate injection of electrons was considered to occur, whena quantumof positivecorona chargewasdeposited on the surfaceof the oxide. Before this current was analyzed, each oxide was compared to the Poole-Frenkel slope for SiO 2 to verify that the conduction was indeedbyFowler-Nordheimconduction. In Figure 4.4, the slope of the experimental oxides, AG IV (0.023) and Thermal TA (0.029), did not match the Poole-Frenkel theoretical slope of 0.015 accepted for silicon dioxide. Figure 4.4: Poole-Frenkel plot comparing theoretical and experimental oxides. The theoretical Poole-Frenkel current density was calculated based on an ionization potential of 1eV and a proportionality constant of 1x10-13 .Both of these parameters were reportedin literature for silicon dioxide [29]. The theoretical slope assumed that a

PAGE 86

71 significant number of acceptor traps were present within the oxide at room temperature and as a result was set toone. According to Equation 2.6, the slope decreases, if there were less acceptor traps within the oxide ( =2), to 0.0074. This slope wasnot remotely close to the experimental slopes. The proportionality constant calculated from the intercept of the Poole-Frenkel plot for the experimental oxides indicatedthat the density oftrap centers was inthe range of 10-28 to 10 -24 Due to the mismatch between the slope of the theoretical line and the experimental lines, it was concluded that the conduction through the experimental oxides was not due to Poole-Frenkel. Although, the line for Thermal TA intersected thetheoretical Poole-Frenkel line, the parameters which affected the slope, the dielectric constant and the value of could not rectify the angle between the lines. The Fowler-Nordheim characteristics wereanalyzed for eachAGoxideand thermal oxide of4H-SiC substrates. The experimentalFowler-Nordheim characteristics of oxidized Si samples, with oxide thicknesses greater than 50, hadan overall good match to the predicted Fowler-Nordheimcurrent characteristic (see Figure 4.5). As seen in Figure 4.5, the Fowler-Nordheim characteristic is independent ofoxide thickness.

PAGE 87

72 Figure 4.5: Fowler-Nordheim plot for various thermal oxide thicknesses on Si. In this study, the effective barrier height was establishedbytwo methods. The first methodutilized the measured current (Jexp )fromI-Vmeasurements. Itwas then fitted to the predictedtheoretical Fowler-Nordheim current characteristics defined in the software and the thickness was determined. The default theoretical Fowler-Nordheim current characteristic for SiO 2 /Si systems was based on M ox of 0.36and an effective barrier height of 3.15eV. The change implemented in the software for SiO2 /4H-SiC systems, with regards to the default values, was the effective barrier height. It was changed from 3.15eV, the conduction band offset for SiO2 /Si [36]to 2.7eV, the conduction band offset for SiO 2 /4H-SiC [81]. The effectivemass in the oxide was not changed since it was previously shown that AG oxide or thermal oxide grown on 4H-SiC substrates had a

PAGE 88

73 relative dielectric constant of 3.9 [75]. However, thethicknessvaluefrom the I-V measurementdid not match theindependently measured average EOT valuefrom C-V measurements. This indicated that trapped chargesand the oxidationprocesschemistry influenced themeasured voltage. To optimize the measured Fowler-Nordheim curve to the theoretical Fowler-Nordheim curve, the effective barrier height was varied untilthe thicknessfitted to+ 5% of the independently measured average EOT value. Table 4.3 along with Figure 4.6 shows the effect of oxidation process chemistry on the effective barrier height for SiO 2 /4H-SiC systems. The x-axis on the graph represents the oxidation process for a thermal oxide and AG oxides (annealed and non-annealed). Table 4.3: Oxidation process chemistry variation. Afterglow oxidation at 850 0 C, 90 min. in excited media: (O 2 :N 2 O:FG) with variation of FG flow Process FG Flow Parameters of Post Oxidation Anneal (l/min) Temp ( 0 C) Time (min) FG (l/min) N 2 O (l/min) O 2 (l/min) TH TB Thermal atmospheric oxide, outside vendor AG I_A 0.5 850 60 0.5 0.2 3 AG IV 0.5 950 120 1 3 AG V 0.5 900 360 0.5 0.2 3 AG VI 0.5 950 120 0.3 4 AG VIII 0.5 AG IV after RCA clean and anneal (2h, 1100 0 C, Ar) AGW V 1 AGW VI 1.3

PAGE 89

74 Figure 4.6: Oxidation processchemistryinfluence ontheeffective barrier heightfor SiO 2 /4H-SiC systems (see Table 4.3 for oxidation process representation). In Figure 4.6, the estimated effective barrier height value slightly varied depending on the oxidation process conditions. It is shown that a post-oxidation anneal in (O 2 :FG)slightly decreased the estimated effective barrier height,whencompared toa post-oxidation anneal in (O 2 :N2 O). Comparing AGW V and AGW VI, the estimated effective barrier height slightly increased to 3.2eV for a variant of the oxidation growth mixture,when compared to 3.1eV fora slight FG increasein the oxidation growth mixture. The oxidation of 4H-SiC at a temperature greater than 1100 0 C following a NO post-oxidation anneal at a similar temperature (Thermal TB) revealed a reduced estimated effective barrier height of 2.9eV, when compared to AG oxidationof 4H-SiC at a temperature of 850 0 C. The AGVIII oxide isthe same asAG IVoxide,except for the addition of a 2hour 1100 0 C Ar anneal afterSC1 and SC2 cleanswithout HF dip. This oxide had a lower estimated barrier height value of 3.03eV compared to AG IVoxide.

PAGE 90

75 Similar evidence of the oxidation process influence on the effective barrier height has been reported in literature on SiO 2 /Si systems. The effective barrier height increased up to 0.2eV due to microscopic atomic structural changes in an oxide grown on silicon within a controlled chemical environment [82]. This method was used to calculate the effective SILC values for a given oxidation process. In the secondmethod, Fowler-Nordheim current was calculated using Equation 2.36. From the Fowler-Nordheim plot,the experimental slope value for each oxidewas extracted andthe effective barrier height was calculated using Equation 2.12. The Fowler-Nordheim analysisreported in literaturefor SiO2 /SiC systems assumesMox =0.42 and calculates the effective barrierheight using Equation 2.12. As a result, two effective mass inthe oxide values were compared: Mox =0.36, used for Fowler-Nordheim characteristicsintheFAaST 230 tool;andMox = 0.42, usedcommonlyfor SiO2 /SiC systems[83-87]. Table 4.4 and Table 4.5showthe effective barrier height calculated based on an assumed effective mass in the oxide.

PAGE 91

76 Table 4.4: Calculated effective barrier for annealed 4H-n-type experimental samples. Oxidation Process EOT () Oxide Electric Field (MV/cm) Experimental Slope (V/cm) b based on 0.36 (eV) b based on 0.42 (eV) AG I_A 515 6.1to 6.9 -1.81E+08 2.70 2.56 AG I_B 320 6.2 to 7.0 -1.91E+08 2.79 2.65 AG II 500 6.0 to 6.8 -1.87E+08 2.75 2.61 AG III 554 6.1 to 6.9 -1.82E+08 2.70 2.57 AG IV 520 6.1 to 6.9 -1.81E+08 2.70 2.56 AG V 489 6.1 to 6.9 -1.76E+08 2.65 2.51 AG VI 480 6.3 to 7.0 -1.97E+08 2.84 2.70 Thermal TA 517 5.3 to 5.9 -1.81E+08 2.69 2.55 Thermal TB 439 5.5 to 6.1 -1.87E+08 2.75 2.61 Thermal TC 318 6.1 to 6.9 -1.87E+08 2.75 2.62 Table 4.5: Calculated effective barrier for non-annealed 4H-n-type experimental samples. Oxidation Process EOT () Oxide Electric Field (MV/cm) Experimental Slope (V/cm) b based on 0.36 (eV) b based on 0.42 (eV) AGWI 511 6.1 to 6.9 -1.83E+08 2.71 2.57 AGWII 433 6.2 to 6.9 -1.89E+08 2.77 2.63 AGWIII 160 5.6 to 6.2 -1.70E+08 2.58 2.45 AGWIV 135 5.7 to 6.3 -1.71E+08 2.59 2.46 AGWV 412 6.0 to 6.7 -1.88E+08 2.76 2.62 AGWVI 310 6.2 to 7.0 -1.84E+08 2.72 2.59

PAGE 92

77 In Figure 4.7,the Fowler-Nordheim plot is influenced by post-oxidation annealing, since there wasa slightvariationin the slopes of the oxides. Figure 4.7: Fowler-Nordheim plot of AG annealed oxides. Although the oxidation process conditions varied, the AG oxides Fowler-Nordheim slope was comparatively independent ofoxide thickness.The slight parallel shift, between the various annealed experimental AG oxides,indicatedthat the density of the oxideand trapped charges in the oxideaffectedthe intercept and slope of the FowlerNordheim plot. The offset between the experimental lines and the theoretical line could be due to a variant of the effective mass in the AG oxides.

PAGE 93

78 From the experimental data, the effective mass in the AG oxide could not beaccurately extracted from theslope and intercept of experimental Fowler-Nordheim plots. As seen in Equation 2.15, the effective mass in the oxide is includedin the pre-exponential and exponential factor of the Fowler-Nordheim current density equation. Any slighterror in the intercept of the line results in a significant change of M ox This was not a concern for the measurements on SiO 2 /Si systems. Therefore, it was possible that the intercept for an AG oxide on 4H-SiC was affected by the quantization of the energy level of the electrons tunneling through the oxide. To confirm this hypothesis, Fowler-Nordheim plot of thermal oxides measured under the same conditions was compared to the theoretical Fowler-Nordheimline (see Figure 4.8). Figure 4.8: Fowler-Nordheim plot of thermal annealed oxides.

PAGE 94

79 All of the experimental thermal oxides were parallel to the theoretical line,with Thermal TA being in close proximity to this line. The offset between the experimental thermal oxides and the theoretical line, however, was still present. There arethree possible explanations. The first possibility was based on the assumption that traps created during Fowler-Nordheim stress on oxidized n-type 4H-SiC werenot negligible as compared to oxidized Si samples. The second possibility was based on the assumption that the dielectric constant of oxide grown on 4H-SiC wasnot3.9as acceptedfor thermal silicon dioxides [16]. However, the secondpossibility was discredited. Asimulation of this parameter revealed that the dielectric constantneeded to beincreased more than 100% of itscurrent valueto match the theoretical intercept value. Finally, the third possibility was based on thefact that thevalue of the effective mass in the AG oxidecould be higher than 0.42m. In additionto the AG annealed oxides, the non-annealed 90-minute oxides also exhibited a similar oxide field rangecompared to the annealed oxides (see Figure 4.9).

PAGE 95

80 Figure 4.9: Fowler-Nordheim plot of non-annealed AG oxides. In Figure 4.9, the ten minute non-annealed oxides, AGWIII and AGWIV shifted toward thetheoretical line. This variance further indicated that the Fowler-Nordheim characteristics had a dependence on theoxidation of n-type 4H-SiC substrates. The effect of the FG pre-treatment was not significant on the Fowler-Nordheim plot except by thickness comparability. The FG pre-treatment led to a higher growth rate and better oxide thickness uniformity throughout the wafer. This evidence was seen through C-V measurements of five measurementsites on the wafer taken in the dark and light, respectively. These regions were located on the waferstop, right, center, left, and bottom point sites. The offset was still evident with non-annealed oxides, which confirmed the firstand thirdpossibilitiesconjectured earlier.

PAGE 96

81 As seen in Table 4.4 and Table 4.5, using an assumed M ox of 0.36 instead of 0.42, the experimental effective barrierheight valueis in proximity to 2.7eV.However,in Figure 4.10, the theoretical line with M ox of 0.36hasa greater offset when compared to the experimental oxides. Figure 4.10: Comparison of experimental data and predicted Fowler-Nordheim lines. The predicted Fowler-Nordheim linesimplied that AG oxides had an effective barrier height value greater than 2.7eV, contraryto thevalue found using internal electron photoemission study [81]. To further investigate this offset, the influence on the selfadjusting steady state voltage (V SASS ), the effective mass in the oxide, effective barrier height, and trapped charges in the oxide will be analyzed in the proceeding sections.

PAGE 97

82 4.4.1.Self Adjusting Steady State (SASS) Voltage on Oxides The self-adjusting steady state voltage, V SASS ,was determined for each oxide after the Fowler-Nordheim conduction. The corona current density was set to 8 A/cm 2 Each oxide was further stressed with the same high dose of positive corona chargein100s increments. It was reported that the SASS voltage correlated to the band offset in a dielectric-Si system [76]. The SASS voltage reported for SiO 2 /Si systems was taken at 1.2 seconds. In this study, the SASS voltage for SiO 2 /4H-SiC systems was taken at 1.7 seconds duetothe later stabilization of the voltage measurement acquisition. The total stress cumulative times were 5s, 105s, 205s, 305s, and 405s on AG oxides and up to 505s for thermal oxides. Figure 4.11illustrates the VSASS trend for all experimentalannealed oxides.

PAGE 98

83 Figure 4.11: Steady-state voltage experimental oxide comparison. In Figure 4.11,comparing comparable thicknesses, the AG oxides had a higher steady state voltage than thethermal oxides. This indicatedthat the AG oxides had a higher tunneling barrier height. At the highest stress time, the AG oxides tend to slightly decrease in voltagewhencompared to the thermal oxides, which was attributed to the post-oxidation anneal treatment. 4.4.2.Significant Parameters: Effective Mass in the Oxide and Barrier Height To examinethe shift in the Fowler-Nordheim plot between the experimental oxides and the predicted Fowler-Nordheim characteristics, only AG IV and Thermal TA oxides will be used since their thickness was comparable. Thepossibility,which

PAGE 99

84 assumed that the traps induced in the Fowler-Nordheim regime werenegligible, is discussed in the following paragraph. The effective mass in the oxide and effective barrier height from the theoretical curve were varied to fit the experimental oxides. First, theeffective mass in the oxide was increased and the effective barrier was kept constant at2.7eV (see Figure 4.12). Figure 4.12: Variation of M ox on thepredictedFowler-Nordheim characteristics. For the AG oxides,Mox was approximately 0.57and for the commercial thermal oxides, it was about 0.44. Further increase in M ox did not lead to an exact match for the AG oxides. There are threeprobable causes: the trapped chargeinfluence on the oxide field, the oxide structures incorrect effective barrier height, or both. The effective mass in the

PAGE 100

85 oxide increases as a result of theelectronsdispersionduring tunneling. Theoretically, the effective mass in the oxide is based either on a Franz dispersion relation or parabolic dispersion relation [34, 36, 88]. These dispersion relations areattributedto the tunneling electron energy momentum in the oxide band gap. This momentum includes the electron wave vector of the tunneling electron andthe total tunneling electron energy from the semiconductor conduction band edge. In Table 2.3, the standard accepted value for M ox on SiO 2 /Si system ranged from 0.36 to0.5. AMox of 0.42is the typicalvaluechosenfor SiO 2 /SiC systems since it is assumed that electrons tunneling inthe oxide band gap have similar dispersion relationship. Determining the dispersion relation for SiO2 /SiC was beyond thescope of this study. To demonstrate the influence of varying the effective barrier height on the Fowler-Nordheim plot, the theoretical line was evaluated over the same oxide field using a constant M ox of 0.36with an effective barrier height of 2.7eV, 2.9eV, 3.15eV,and 3.25eV, respectively (see Figure4.13).

PAGE 101

86 Figure 4.13: Theoretical Fowler-Nordheim plot varying the effective barrier height with M ox of 0.36. In Figure 4.13, as the effective barrier height increased, the intercept and the offset between the lines decreased. The slopes of the lines were nominally in the same range ~10 8 V/cm. The apparentreductionor increasein the calculated effective barrier height for the experimental oxides comparedto2.7eV (the theoretical effective barrier height for SiO 2 /4H-SiC)was attributed to the influence of trapped charge in the oxide. Trapped charges wereconfirmed byanobserved shift of the flat band voltage onre-measured C-V measurements.

PAGE 102

87 4.4.3.Comparison with Fowler-Nordheim Literature Data on Oxide-n-type 4H-SiC Devices Figure 4.14 illustrates an example of atypical measurement performed on 4HSiC-MOS devices to analyze the Fowler-Nordheim conduction. The voltage is swept at a defined rate (dV/dt) and the current is measured. A spike in the current indicatedthe breakdown of the oxide. Figure 4.14: Example of current-voltage characteristicsof a 4H-SiC MOS capacitor with a 500 gate oxide measured at room temperature [52]. The displacement current,below the Fowler-Nordheim voltageknee,changed toa conduction current from the conduction band,above this voltage knee. The area of interest fornon-contact I-Vmeasurements was within theFowler-Nordheim voltage Displacement current measurement floor= C(dV/ dt) Oxide BreakdownF N KneeArea of interest to non contact IV

PAGE 103

88 knee. The maximum electric field attained during non-contact I-V measurements was approximately 7MV/cm. Equation2.15is the standard Fowler-Nordheim equation used historically to analyzeSiO2 /SiC devices. Onlyafew research groups have explored the occurrence of Fowler-Nordheim tunnelingregimeon SiC-based MOS devices with respect to temperature [85-87]. It was concluded that higher temperatures decreased the effective barrier height. At room temperature, the conduction band offset at the SiO2 /4H-SiC interface was determined to be 2.7eV using internal electron photoemission (IPE) [81]. As a result, the theoretical Fowler-Nordheim plot for SiO 2 /4H-SiC systems is based on the effective barrier height of 2.7eV and on an assumed effective mass in the oxide. This M ox iscommonly chosen as0.42and is taken from measurements done onSiO2 /Si systems [34]. Table4.6includes the effective barrier height values reported for 4H-SiC MOS devices after Fowler-Nordheim injection of electrons from 4H-SiC into the oxide conduction band. These values were calculated from the Fowler-Nordheim plot slope using Equation 2.12with an assumed Mox of 0.42.

PAGE 104

89 Table 4.6: Reported effective barrier height for 4H-SiC MOS devices at room temperature. Reference Gate oxide T ox () Experimental Slope(V/cm) Effective BarrierHeight (eV) Oxide Electric Field (MV/cm) [83] 400 -2.06E+08 2.78 6.9 to 8.3 [85] 230 -1.90E+08 2.64 7.7 to 10 [86] 93 -1.73E+08 2.48 6.3 to 7.1 [84] 670 -1.96E+08 2.70 7.1 to 10 [87] 225 -1.68E+08 2.43 6.3 to 10 In Table 4.6, although the oxidation process conditions in each report varied, the FowlerNordheim slope was comparativelyindependent of oxide thickness. Figure 4.15 illustrates the comparison betweendevice-based contact measurements and non-contact measurements. The Fowler-Nordheim characteristic of the un-metallized thermal oxide is comparable to metallized oxides.

PAGE 105

90 Figure 4.15: Fowler-Nordheimconduction comparison of contact versus non-contact measurements. The current density for the experimental oxides is lower than the metallized oxides due to the current limitation of the non-contact measurement system. The inconsistency observed by these studies in the effective barrier height value indicates the sensitivity of thefabrication process. It was postulated to be duetothe effect ofthepost-oxidation annealing condition used, whichinfluenced the quantity ofcharges in the oxide. 4.4.4.Influence of Trapped Charge in the Oxide To reiterate, Fowler-Nordheim conduction occurs when electrons flow into the oxide conduction band through a triangular potential barrier. If oxide charges are located within theFowler-Nordheim tunneling regime, the centroid of these trapped charges

PAGE 106

91 influences the voltage drop across the oxide and changes the shape of the potential barrier from triangular toanon-triangularbarrier. In the presence of these charges within the Fowler-Nordheim tunneling regime, the oxide field (F)is given as[89-91] ox r t ox ox t x Q t V F 1 0 (4.2) where Q t is the trapped charge in the oxide, x is the centroid of the charge distribution measured with respectto thesemiconductor/SiO2 interface, t ox is the oxide thickness, V ox is the voltage drop across the oxide, r is the dielectric constant, and 0 is the permittivity of vacuum. Figure 4.16 illustrates the influence of charge trapped within the FowlerNordheim tunneling regime in SiO 2 /Si system[92].

PAGE 107

92 Figure 4.16: Charge trapped within tunneling regime in SiO 2 /Si system: (a) negative charge trapping, (b) positive charge trapping[92]. In Figure 4.16, the gradient of the potential in the oxide changed depending on the polarity ofthe trapped charge. The modified oxide field is lowered as predicted by Equation4.2due to trapped charge. To account for this effect, the Fowler-Nordheim current equation was modified and denoted as J non [91] ox non ox non ox ox non hE D qm C hm E mq J 3 2 8 exp 8 2 2 (4.3) 2 x E Q E Q C ox b t ox ox t b non (4.4)

PAGE 108

93 2/3 3 x E Q E Q D ox b t ox ox t b non (4.5) Equation 4.3assumes that the trapped sheet charge is uniformly distributed within the oxide. To illustrate the effect of this modification, an example of atheoretical FowlerNordheim plot with trapped charges within or outside theFowler-Nordheimtunneling regimeis shown in Figure 4.17. Figure 4.17: Example of trapped charge outside or insidethe Fowler-Nordheim tunneling regimein a 500 oxide. As seen in Figure 4.17, charges within theFowler-Nordheimtunneling regime increased the slope of the line. The angle between the theoretical line without trapped charges and the theoretical line with trapped charges,withintheFowler-Nordheimtunneling distance,

PAGE 109

94 also increased as the oxide field decreased. This angle is controlled by the location of this trapped charge centroid. 4.4.5.Fowler-Nordheim Equation Modification for Oxide-4H-SiC The influence ofthe trapped charge in anexperimental oxide was verified by remeasuring its C-Vcharacteristics after stress. A flat band voltage shift marked the presence of trapped charges in the oxide. The offset between the Fowler-Nordheim theoretical line and the experimental lines could onlybe resolved with the assumption thattrapped charges were within theFowler-Nordheimtunneling regime. Equation 4.3 was fitted to the experimental oxides in the attempt to minimize or eliminate this offset. The quantity of the trap charge was estimated by ox final fb initial fb t C V V Q (4.6) where initial fb V and final fb V are the flat band voltage taken beforeFowler Nordheim conductionand afteramoderate stress, respectively, and Cox is the oxide capacitance. Figure 4.18 illustrates the C-V characteristics of the experimental oxides before and after stress.

PAGE 110

95 Figure 4.18: Capacitance-voltagecharacteristics of AG IV and Thermal TA oxides before and after stress. The trappedcharge density for Thermal TAoxideand AG IVoxidewas evaluated to be 4.07x10 -7 C/cm 2 and -7.06x10 -8 C/cm 2 respectively. A Mox of 0.42 did not provide a reasonable overlay to the experimental oxides. Since the calculated effective barrier height value was in proximity to 2.7eV (the conduction band offset for SiO2 /4H-SiC systems), M ox was chosen to be 0.36for the non-triangulartheoreticalFowler-Nordheim current characteristic(see Table 4.4 and Table 4.5).

PAGE 111

96 Figure 4.19: Non-triangular Fowler-Nordheim plot fittedto experimental oxides. As seen in Figure 4.19, to obtain a nominal overlay to the non-triangular FowlerNordheim theoretical line, the location of the charge centroid was adjusted to 25 for Thermal TA oxide. However, the afterglow oxide did not provide an overlay at the minimum distance of 10. Contrary to Thermal TAoxide, the non-triangular FowlerNordheim model parameters could not intersect the AG IVoxidewithout a change in the barrier height from 2.7eV to 3.1eV. This further indicatedthat the AG oxide structure hasfundamental differences from the thermal oxide structure.Thisincrease in the effectivebarrier height for the AG oxidewas also an indication that the AG oxidation growth method modifiedthesurface of the interface. It has been reported that oxidation of 4H-SiC with atomic nitrogen and hydrogen at a temperature of 750 0 C revealed a

PAGE 112

97 disordered Si-rich surface [22]. An investigation of the Fowler-Nordheim derivative also suggested that in the presence of negative trapped charge, the exponential oxide field dependence and the effective barrier height valueincreased[89].The non-triangular Fowler-Nordheim theoretical line only intersected the experimental lines, which revealed that the exact location of the trapped charge centroid is a critical fitting parameter. The perfect match will depend on the exact charge value and location of the charge centroid within the oxide. To potentially obtain the location of the trapped charge centroid and the charge distribution within an AG oxide, an AG oxide thickness was etched back and analyzed. 4.5.Distribution of Trapped Charge in Oxide on 4H-SiC The oxide recipe used for this experiment was AG IVoxide. This oxide was etched back to obtain various thicknesses. Prior to testing, the wafers were subjected to two dehydration treatment: a hotplate anneal and a high temperature Ar anneal. After the average EOT was obtained by non-contact C-Vmeasurements in light, theFowlerNordheim analysis was examinedfor each quadrant (see Figure 3.6). Subsequently, the C-Vcharacteristicfor each quadrant was re-measured to estimate the trappedcharge within each thickness. 4.5.1.Diluted Hydrofluoric Acid Etch Rate on Oxide-4H-SiC In Figure 3.6, the corresponding time for etch quadrant is as follows: region A was etched for 2 minutes; region B was etched for 5 minutes; region C was etched for 3 minutes; region D was not etched; region E was etched for 8 minutes; and region F was etched for 10 minutes. AfterglowIVoxidewas repeated twice for this experiment. This

PAGE 113

98 oxide was also grown on n-type Si wafers. Before each wafer was etched, the uniformity of the oxide was checked by measuring 5 sites on the wafer (top, bottom, right, left, and center). For the 4H-SiC wafers, non-contact C-V measurements determined the average EOT on each site. For the Si wafers, the average oxide thickness was determined by Ellipsometer measurements. This is because afterglow oxidation process was too harsh for the silicon surface,as evident through noisy non-contact C-V characteristics. The difference between the two oxidation process runs, AG IV _EA and AG IV_EB,was the excitation plasma power level. The firstplasma runwas set to approximately 900Wand the secondplasma runwas set about 1100W. A linear regression analysis was then performed to determine the etch rate. In Table 4.7, the etch rate,along with the corresponded R-squared value,is reportedfor each oxidation process run. Table 4.7: Etch rate comparison between 4H-SiC and Si wafers at 24 0 C. Oxidation Run Etch Rate (/min) for Si Etch Rate (/min) for 4H-SiC AG IV_EA 41 (R 2 =0.937) 46 (R 2 =0.939) AG IV_EB 32 (R 2 =0.996) 36 (R 2 =0.992) The oxide on 4H-SiC etched at aslightly faster rate than the oxide on Si. The etch rate variation between Si and 4H-SiC differed only by a few angstroms. In the first oxidation process run, AG IV_EA, the R-squaredcorrelationdid not give a relativelygood fit to the data but gave a better fit for the second oxidation run. This was attributed to the as-

PAGE 114

99 grown oxideuniformity and density structure. Figure 4.20 and Figure 4.21illustrates the linear regression fit for each oxidation process run. Figure 4.20: Oxide thickness after diluted HF etching for oxide AG IV_EA.

PAGE 115

100 Figure 4.21: Oxide thickness after diluted HF etching for oxide AG IV_EB. 4.5.2.Influenceof Dehydration Procedure on Oxide-4H-SiC The AGIV_EA oxide on 4H-SiC was dehydrated usingtwo methods: on a hotplateat 2000 C for 5 minutes and a rapid thermal processing (RTP) anneal in Ar at 800 0 C for 2 minutes. Afterglow IV_EB oxide on 4H-SiC was also dehydrated using a hotplate at 200 0 C for one hour and an afterglow Ar anneal at 600 0 C for 20 minutes. Following each dehydration method, the experimental Fowler-Nordheim tunneling current was analyzed. The influence of moisture on the electric fieldin the oxideis demonstrated in Figure 4.22. The random behavior of the voltage, influencingthe electric field in the oxide,was attributed to the presence of water molecules absorbed on the surface of the oxide (seeFigure 4.22(b)). The effect of theabsorbed moisture was

PAGE 116

101 more prominent on the Fowler-Nordheim slope of the thinnest oxide. This behavior was corrected by the high temperature Ar anneal treatment, which eliminated the error in the measured voltage drop across theoxideforboth oxidation processes. Figure 4.22: The influence of a dehydration method on the Fowler-Nordheim plot (a) AG IV_EA and (b) AG IV_EB. (a) (b)

PAGE 117

102 4.5.3.Effective Trapped Charge Distribution Afterglow IV_EB oxide was used toinvestigate the trapped charge distribution and determine its centroid within the oxide. After obtaining the Fowler-Nordheim regime, the oxide was given a small charge fluence of 0.12mC/cm2 before re-measuring the C-V in each quadrant. The effective trapped charge in the oxide was calculated using Equation 4.6. To further study the influence of trapped charge on the electrical signature of the Fowler-Nordheim plot, the thickness in four quadrants were etch downfrom: 376 to 265; 287 to 178; 209 to 164; and 125 to 79. These oxide thicknesses were subjected to a 5-minute 1:1 (H 2 O:HNO 3 ) diluted nitric clean followed by another 5-minute 1:1 (H 2 O:HCL) diluted hydrochloric clean. Between each solution clean, the wafer was rinsed withD.I.water. Afterwards, the wafer was annealed in Ar in the afterglowsystem for 20 minutes at 6000 C. The same measurement procedure was repeated. In Figure 4.23,thevariationof the slopes of the thinner oxidesindicated that the distribution of the trapped charge did not appear to be uniformly distributed in the oxide. In addition,etching and cleaningconditionsalso influenced the Fowler Nordheim characteristic.

PAGE 118

103 Figure 4.23: Fowler-Nordheim plot of thinner oxides(unfilled geometric shapes) compared to thick oxides (filled geometric shapes). The thinnest oxide, 79, exhibiteda straight line buthada different slope as compared to the thicker oxides. The parallel shifts ofthe second etched regions indicated the influence of further hydrogenation onthe oxide surface. This effect wasseen in the distribution of the calculated effective trappedcharge. Figure 4.24illustrates the behavior of the effectivetrapped chargein the etched AG oxide and un-etched AG oxide.

PAGE 119

104 Figure 4.24: Absolute trapped charge versus oxide thickness. In Figure 4.24, the un-etched 500 oxide showed a similar quantity of trapped charges in the oxide. Two possible trends were predicted as thickness decreased. As the thickness was etched back, the density of the trapped charges tendsto decrease as the thickness decreased to 175. Below thisthickness value, the density of the trapped charge did not followeither of thepredictedtrends. A higher trapped charge valuewas also noted for an un-etched 146AG oxide (AG VII). Further investigation to explain this phenomenon for oxide thicknesses less than 175 is ongoing. Table 4.8summarizes the measurement parametersfor AG IV_EBsuch as the flat band voltage, current at the SASS voltage,andthe effective trappedcharge, in an attemptto explain the trapped charge distribution.

PAGE 120

105 Table 4.8: Effective trapped charge calculation parameters. Oxide Thickness () Initial V fb (V) Final V fb (V) Q t ( C/cm 2 ) Initial E ox at V SASS (MV/cm) Final E ox at V SASS (MV/cm) F-N current at V SASS ( A/cm 2 ) Final current at V SASS ( A/cm 2 ) 495 0.6 2.82 -0.150 6.75 6.73 0.060 0.054 406 0.43 1.57 -0.097 6.90 6.89 0.058 0.057 376 0.53 2.72 -0.200 6.70 6.78 0.064 0.059 287 0.4 1.62 -0.150 6.83 6.91 0.058 0.056 265 0.7 1.18 -0.063 7.21 7.14 0.053 0.051 209 0.38 1.09 -0.120 6.66 6.91 0.056 0.052 178 0.85 1.06 -0.041 6.49 6.98 0.058 0.059 164 0.62 1.75 -0.240 7.20 7.13 0.051 0.056 125 0.38 1.46 -0.300 7.02 6.94 0.050 0.123 79 0.49 0.78 -0.130 6.75 5.95 0.085 0.118 In Table 4.8, although these oxides weresubjected to the same charge doses, the initial flat band voltage varied independent of thickness. However, this behaviorwas not evident before the etching process. The flat band voltage before the etching process was about 0.9 volts on 5 sites of the wafer (center, bottom, top, left, and right). The possible reason for this variation in the flat band voltage istheinfluence of a hydrogenated oxide surface. The high temperature, Ar anneal, may have left the oxide with shallow trap centers. However, the electric field in the oxide remained at a constant value of ~7 MV/cm before and after stress except for the thinnest oxide region. The increase in the current density after stress,for thicknesses below 175, was attributed toanincrease in

PAGE 121

106 the trapped charge density. As a result of this experiment, the trapped charge centroid could not be determined due the influence of unintentional trap creation within the oxide. The influence of these charges was evident through C-V characteristics. Charge trapping and de-trapping effect impacted the accuracy of the contact potential measurement. 4.6.Non-ContactStress Induced Leakage Current(SILC)Analysis A controlled ionic current of 8 A/cm 2 was used for all measurements. Noncontact stress induced leakage current (SILC)testing was performed to characterize four oxidation process recipes, AGWI, AG I_A, AG II, and AG III. The difference between these recipes was the post-oxidation anneal: AGWI was not annealed; AG I_Aincluded a re-oxidation anneal; AG IIincluded an Ar anneal; and AG III included both a reoxidation anneal and an Ar anneal. The impact of the process conditions, which included unintentional contamination from the cleaning procedure orfromthe furnace environmental conditions, was investigated. To obtain statisticaldataacquisition on an oxide, seventeen test sites were evaluated on the 3-inch diameter substrates (see Figure 4.25).

PAGE 122

107 Figure 4.25: Measurement site positions. 4.6.1.Oxidation Process Influence on Leakage Current Theleakage current calculation dependedon the oxide thickness, the dielectric constant, and the differentiation of the measured oxide voltage with respect to time (see Equation 2.36). An example of non-contact SILCdensityversus stress time is shown in Figure 4.26on a 150thermaloxide grown onSi. -40 -30 -20 -10 0 10 20 30 40 -40 -30 -20 -10 0 10 20 30 40

PAGE 123

108 Figure 4.26: Effect of stress fluence on a 150 thermal oxidegrownon a p-type Si substrate. In this example,a 150 thermal oxidegrownon a p-type Sisubstrate wasstressed for 100s and then in three 200s increments. The charge fluence ranged from 0.8mC/cm2 to 5.6mC/cm2 In Figure 4.26, it was observed that the SILC densityincreasedby two orders of magnitude from the lowest fluence to the highest fluence. The slope of the line is almost horizontal at 700s. This demonstratedthe limitation of the corona current source. At this current level, the trap density wasat its maximum. However, at a field of 5.9MV/cm, the difference between the 5s and 100s stress was minimal,indicatingthat the density of traps was comparable. As the stress fluence increased, the current flowing across the oxide also increased. Therefore, the density of trapped charge was

PAGE 124

109 proportional to the current density. Thisbehavior was expected tooccur on AG oxides grown on n-type 4H-SiC substrates. For non-contact SILC, the leakage current iscalculated from the excesscurrent above the Fowler-Nordheim currentcharacteristics. To fit the theoretical FowlerNordheim curvewith the assumption ofMox = 0.36, the barrier height value was adjusted until the thickness determined from the I-Vmeasurement was fitted to+ 5% ofthe independently measuredaverageEOT valueof each oxide. For the AGW I, AG I_A, AG II and AG IIIoxides, theestimatedbarrier height value wasdetermined to be3.1eV. As a result, the SILC values obtained through this optimization was known as effective SILC values. 4.6.2.Oxide Consistency: Statistical Distribution Each measurement site was stressedin 100s increments(see Figure 4.27) to observe any trendsin the effective SILC characteristics. The normal distribution was used to investigate theeffectiveSILC values. The fluence ranged from 0.8mC/cm2 to 3.2mC/cm2 assuming a steadycalibratedionic currentof 8 A/cm 2 fromthe corona source. Figure 4.27illustrates the typical plotauser sees when selecting each site after the measurement hasbeen performed.

PAGE 125

110 Figure 4.27: Current density versusoxidefield at point (10,10) for AGI_A oxide. The measurement sequence at each site was as follows:step 1:5s charge depositionand monitor the contact potential difference for 20 seconds;andstep 2: 100s charge deposition and monitor the contact potential difference for 60 seconds. Step two was repeated three more times. Table4.9summarizedthe stress measurement protocol. Table 4.9: Stress measurement protocol for SILC measurements. Charge Deposition Time Cumulative Stress Time 5s 5s (establish F-N conduction) 100s 100s 100s 100s +100s= 200s 100s 200s+ 100s= 300s 100s 300s + 100s= 400s

PAGE 126

111 Once the measurement sequence was completed, the wafer wasautomatically positioned to another measurement site. For each new site,the measurement sequencewasrepeated. Figure 4.28illustrates a typical probability plot with respect to the stress timeonAG III oxidetakenat 5.7MV/cm. Figure 4.28: Probability plot for AG III oxide at each cumulative time. In the above figure, there are three straight lines: one below 30%, the second between 30 to 70%, and the third above 80%,which indicatedweakareason the wafer. The leakage current increased two orders of magnitude fromthesites within the 30 percentile range to sites within the80 percentile range. This was attributed to an increase of defectsin the as-grown oxidefilm. Since the differences between the curves were minimal, the data analyzed for each oxide was taken at 400s for each repeatable run.

PAGE 127

112 The electric field chosen to perform data analysis was at 5.7MV/cm asit encompassed the stressed curves. Each oxidation processrun was repeated andanalyzed to determine if the effectiveSILC values were consistent on the same oxidation recipe. These oxidation recipes were repeated at least three times. The same cleaning procedure and oxidation plasma excitation power was kept relativelyconstantfor each run. The first three oxides, AGWI, AG I_A, and AG II, was repeated three times except for AG III which was repeated twice. This was because the microwave power supplywas replaced. One of the specifications of the new powersupply which differed from the old power supply was the magnetron head. Figure 4.29shows the probability plot of AGWI oxidation processrunrepeatability. Figure 4.29: Probability plot of AGWIoxidation processrunrepeatability.

PAGE 128

113 In Figure 4.29, AGWIoxide did not include a post-oxidation anneal. In the first process run,70% of the wafer sustaineda leakage current below 5x10-9 A/cm 2 Thesecond process run had five straight linesin the curve, which indicatedthat the cleaning procedure was not optimal andunintentionalcontamination occurred. The third process run was similar to the first process run except for two site areas, which showed a leakage current greater than 1x10 -8 A/cm 2 In summary, the optimum acceptable current for each region was defined at 1x10 -8 A/cm 2 Under this condition,80% of the wafer sites passed inthe first and thirdprocess run, and 40% of the wafer sites passed for the secondrun. Figure 4.30showsthe probability plot for AGI_Aoxidation processrunrepeatability. Figure 4.30: Probability plot of AG I_A oxidation processrunrepeatability.

PAGE 129

114 Figure 4.30shows the trend for an oxidation process run which included a 60-minute reoxidation anneal atthe same oxidation temperature of8500 C. Compared to the second and third process runs, the first process run had a higher current density andthe leakage current variedslightlythroughout the wafer. Subsequent process runs indicated that the cleaning process of the wafer improved. At a current leakage of 1x10 -8 A/cm 2 approximately 80% of the wafer sites passed in process runs two and three,and 50% of the wafer sites passed in the first process run. Although the third process run showed a lower leakage current at 80%, the curve had three kinks above 50%. These were attributed to weak spots in the as-grown oxidefilm. In summary, the re-oxidation anneal did not significantly lower the current leakage throughoutthe wafer. Figure 4.31 illustrates the probability plot of AG II oxidationprocessrunrepeatability.

PAGE 130

115 Figure 4.31: Probability plot of AG II oxidation processrunrepeatability. In Figure 4.31, this oxidation process included an Ar anneal for one hour at 950 0 C. All three process runs showed a comparable leakage current below 60% of the measurement sites. Above 60%, the firstprocess run showed approximately alinearincreasein the currentcompared to the others. Leaky sites were attributed to the possibility of unintentional process contamination. In conclusion, similar to the re-oxidation anneal, the high temperature Arpost oxidationannealshowedan improvement ineach subsequent processing runwithin 60% of the wafer area. Finally, Figure 4.32shows the repeatability of the oxidation process run, which combined both the Ar and re-oxidation post anneal.

PAGE 131

116 Figure 4.32: Probability plot of AG III oxidation processrunrepeatability. In Figure 4.32, 40% of the measurement siteson both process runssustained a leakage current of 3x10 -9 A/cm 2 whichis animprovementcompared to the other oxidation processes. However, between 40% and 80%, both process runs showed an opposite density of trap formation,as apparentbytheslope of their line. Above 80%, the leakage current increased linearly forboth process runs. Insummary, this post-oxidation process did not improve the oxide tendency to form defects throughout the wafer. The repetitionof the oxidation process runsshowed hownon-contact SILC testingcan be used to assess the oxidation process conditions andcleaning protocols. Figure 4.33shows the correspondingsites on the wafers which had leakage currents greater than 1x10 -7 A/cm 2 intheseexperiments.

PAGE 132

117 Figure 4.33: Leakage current sites greater than 1x10 -7 A/cm 2 Stricter cleaning protocols were implemented for subsequent oxidation process runsand a standard excitedplasma mixture was usedto clean the furnace tube prior to loadingthe wafers for oxidation process runs. Non-contact SILC has the potential to assessthe reliability of oxides grown on 4H-SiC substrates. AG WIAG I_A; AG IIAG IIAG IIAG III

PAGE 133

118 Chapter 5.Conclusion 5.1.Summary of Research Contributions Non-contact Corona-Kelvin metrology was used to investigate the charge transport in variousoxidesgrown onn-type4H-SiC substrates. The measurementsused were voltage-charge(V-Q)measurements, capacitance-voltage(C-V)measurements, equivalentoxide thickness(EOT)measurements, charge trapping(Qt )evaluation measurements, and current-voltage (I-V)measurements. Variations ofafterglow oxides and thermal oxides were compared. After the deposition of a charge fluence of 0.04mC/cm2 on the oxide surface, Fowler-Nordheim conduction was identifiedfor the first time. The electric field in the oxideat this regime wasgreater than 5MV/cmfor thick oxides. The experimental Fowler-Nordheim characteristicscompared to the classicalFowler-Nordheim characteristicsrevealedan offset between the plots. Thisoffset was greater forthickAG oxides as compared tothermal oxides. The AGoxidation variation parameters influenced the Fowler-Nordheim characteristics. The threedominant variationparameters werethe pre-conditioning treatment ofthesubstrate surfaceprior to oxidation,the oxidation growth time, and the post-oxidation anneal. The offsetbetween AGoxide thicknesses less than 160and the theoretical line in the Fowler-Nordheim plotwasmarginallyreduced. The effective barrier height and the

PAGE 134

119 effective mass in the oxidearethe twodefiningparameters in the current density equation describingthe Fowler-Nordheim conduction. The theoretical classical FowlerNordheim plot used an effective barrier height of 2.7eV and an effective mass in the oxide of 0.42m. Thiseffective barrier height value, which represented the conduction band offset at the SiO 2 /4H-SiC interface, was independently found usingstandard photoemission experimentation. The effective mass in the oxide was the common value used in 4H-SiC MOS devicesforFowler-Nordheim analysis. Tofit the experimental data to the effective barrier height of 2.7eV, the effective mass in the oxide was assumed to be 0.36m. Using this value, the calculated effective barrier height for boththermal and AGoxideswascomparable to the calculated effective barrier height value foundinliteratureusing 4H-SiC MOS devices. However, this effective mass in the oxide did not correct the offset between the classical FowlerNordheim curveand the experimental Fowler-Nordheim curves. As a result, a modified Fowler-Nordheim equation, which accounted for trapped charges and their centroid location within theFowler-Nordheimtunneling regime, provided a proximate overlay for the thermal oxides butnot for the AGoxides. An additional adjustment of the barrier height value of 3.1eV for the AGoxides was necessary to provide this overlay with a reasonable location of the trapped charge. This result was an indication that the AG oxidation growth method modifiedthe surface of the substrate and theoxide grown could besilicon enriched. Recently, it was found that a 146AG oxide on new 4H-SiC substrate revealed an effective barrier height of 2.75eV usingtheEOT optimization parameter.

PAGE 135

120 To further investigate the distribution of the trapped charge and their location in AGoxides, various AG oxide thicknesses,ranging from 490 to 79,determined that trapped charges are not uniformly distributed in the oxide. The density of trapped charges increased for oxidethicknessesless than 175 indicating that shallow traps are not neutral. These trapsinfluencedthe measured voltage drop across the oxide. The electric field in the oxide decreasedas a result of these traps. It was determined that hydrogenation of the oxide surfaceincreased the trapped charge density. The source of the hydrogenation of the oxide was from the diluted HF solution. The high temperature Ar anneal possibly led tounintentional surface roughness changes. This post-etching anneal was performed since thehotplate did notcompletelydehydrate thesurface of the oxide. Moisture in the oxideinfluencedthe value of the measured voltage. Determining the exact location of thetrappedcharge could not be achieved using the etch-back experimental procedure. Currently, only Fowler-Nordheim data on SiC-based MOS devices have been published in the literature but not SILC data. Non-contact SILCtestingonly caused area defects in various sites for a particular oxide growth recipe in the absenceof fabricated devices. Caution should be takenwhen defining the Fowler-Nordheim current. As a result, the effective SILC values were reported for a particular oxidation process. Various oxidation process parameters could not be compared with SILC testing because changing the effective barrier height, to fit thethicknessdetermined from I-V measurementsto EOT, impacts the electric field in the oxide. Subsequently, the FowlerNordheim current characteristicis changed.

PAGE 136

121 Taking this condition into consideration, non-contact SILC revealed the effectiveness of an anneal treatment. Normal probability plots of theeffectiveSILC value variation gave a quick assessment oftheoxide reliability. Stress times were proportional to the quantity of the defect formation. Constantcoronacurrent testing revealed that this test does not cause destructive breakdown ofthe oxide. Non-contact CV, EOT, and SILC show promise to be used for in-line monitoringof as-grown oxide films. Two benefits of applying in-linemonitoring techniques to SiC technology are the absence of device fabrication and the quick assessment of an oxide to identify process variations to defect.These techniquesmay enable the optimization of an oxidation growth process to effectively controloxide reliability,in an attemptto commercialize 4H-SiC-MOS devices. 5.2.Future Work The location of the trapped charge and it is distribution within the oxide need to be addressed on un-etched thin AG oxides. With this information, the Fowler-Nordheim equation, which includes the effects of trapped charge within theFowler-Nordheim tunneling regime, shouldbe re-examined. If thisset of experimentsdoes not adjust the Fowler-Nordheim experimental curves characteristicsto the theoretical curve characteristic, then the Fowler-Nordheim equation shouldbemodifiedfurther. The theory to determine the effective mass in the oxideand the image force effectalsorequire further examination. The image force of an electron tunneling through trap centers may decrease the intercept value.

PAGE 137

122 To confirm the structural changes duringtheAG oxidation of n-type 4H-SiC substrates, X-ray photoelectron spectroscopy (XPS), which measures the elemental composition of a surface, should be performed onAG oxide films. If AGoxidation of 4H-SiC is lightly nitrated, then it wouldfurther support the increase ofthe effective barrier height value. It has been reported that SiO 2 containing nitrogen grown on silicon varied the effective barrier height and the effective mass in the oxide [93]. Also, sequential oxidation process runsshould be performedwithvariation of oxidation parameters to improve AG oxide reliability,such as surface pre-conditioning, oxidation chemistry, and post-oxidation anneal. Longer and higher temperature post-oxidation anneals may result in the reduction of shallow trap charges in the oxide. Thispostoxidation anneal treatment mayprovide an effective barrier height close to 2.7eV. Statistical acquisition of dielectrics grown on 4H-SiC using non-contact SILC assessment can be used to acquire fundamental information about the oxide-substrate interface. Reliability curves can be developed from various dielectrics to compare differentoxidationprocess parameters.

PAGE 138

123 References [1] M. A. Alam, "SILC as a measure of trap generation and predictor of T BD in ultrathin oxides," IEEE Trans. Electron Devices, vol. 49, pp. 226-231, 2002. [2] D. J. Dumin and J. R. Maddux, "Correlation of stress-induced leakage current in thin oxides with trap generation inside the oxides," IEEE Trans. Electron Devices, vol. 40, pp. 986-993, 1993. [3] D. J. Dumin, Oxide Reliability : A Summary of Silicon Oxide Wearout, Breakdown, and Reliability. New Jersey: World Scientific, 2002. [4] R. Moazzami and C. Hu, "Stress-induced current in thin silicon dioxide films," in International Technical Digest on Electron Devices Meeting, 1992, pp. 139-142. [5] M. Wilson, J. Lagowski, A. Savtchouk, D. Marinskiyand L. Jastrzebski, "Study of stress-induced leakage current in thin oxides stressed by corona charging in air: relationship to GOI defects," Mater. Res. Soc. Symp. Proc., vol. 592, pp. 345-350, 2000. [6] E.Rosenbaum and L. F. Register, "Mechanism of stress-induced leakage current in MOS capacitors," IEEE Trans. Electron Devices, vol. 44, pp. 317-323, 1997. [7] E. F. Runnion, S. M. Gladstone, R. S. Scott Jr., D. J. Dumin, L. Lie and J. C. Mitros, "Thicknessdependence of stress-induced leakage currents in silicon oxide," IEEE Trans. Electron Devices, vol. 44, pp. 993-1001,1997. [8] D. J. Dimaria, "Defect production, degradation, and breakdown of silicon dioxide films," Solid-State Electron., vol. 41, pp. 957-965,1997. [9] S. Holland, I. C. Chen, T. P. Ma and C. Hu, "On physical models for gate oxide breakdown," IEEE Electron Device Lett., vol. 5, pp. 302-305, 1984. [10] S. Lombardo, J. H. Stathis, B. P. Linder, K. L. Pey, F. Palumbo and C. H. Tung, "Dielectric breakdown mechanisms in gate oxides," J. Appl. Phys., vol. 98, pp. 121301-1121301-36, 2005. [11] J. H. Stathis, "Reliability limits for the gate insulator in CMOS technology," IBM J. Res. Develop., vol. 46, pp. 265-286, 2002.

PAGE 139

124 [12] O. Kordina and S. E. Saddow, "Silicon carbide overview," in Advances in Silicon Carbide Processing and Applications S. E. Saddow and A. Agarwal, Eds. Boston: Artech House, 2004, pp. 1-27. [13] R. Madar, "Silicon carbide in contention," Nature, vol. 430, pp. 974-975, 2004. [14] J. A. Cooper Jr., "Silicon carbide MOSFETs," in Wide Energy Bandgap Electronic Devices F. Ren and J. C. Zolper, Eds. Singapore: World Scientific Publishing Co., 2003, pp. 339-394. [15] H. O. Olafsson, "Detection and removal of traps at the SiO 2 /SiC interface," Ph. D. Dissertation, Chalmers University of Technology, 2004. [16] S. M. Sze, Physics of Semiconductor Devices ,2nd ed.New York: John Wiley & Sons, Inc., 1981. [17] Y. Song, S. Dhar, L. C. Feldman, G. Chung and J. R. Williams, "Modified Deal Grove model for the thermal oxidation of silicon carbide," J. Appl. Phys., vol. 95, pp. 4953-4957, 2004. [18] I. Vickridge, J. Ganem, Y. Hoshino and I. Trimaille, "Growth of SiO 2 on SiC by dry thermal oxidation: mechanism," J. Phys. D: Appl. Phys., vol. 40, pp. 6254-6263, 2007. [19] C. Raynaud, "Silica films on silicon carbide: a review of electrical properties and device applications," J. Non-Cryst. Solids, vol. 280, pp. 1-31, 2001. [20] C. I. Harris and V. V. Afanas'ev, "SiO 2 as an insulator for SiC devices," Microelectron. Eng., vol. 36, pp. 167-174, 1997. [21] J. M. Knaup, P. Deak, T. Frauenheim, A. Gali, Z. Hajnal and W. J. Choyke, "Theoretical study of the mechanism of dry oxidation of 4H-SiC," Phys. Rev. B, vol. 71, pp. 235321-1-235321-9, 2005. [22] M. Losurdo, M. M. Giangregorio, P. Capezzuto, G. Bruno, A. S. Brown, T. Kim and C. Yi, "Modification of 4H-SiC and 6H-SiC (0001) Si Surfaces through the interaction with atomic hydrogen and nitrogen," J. Electron. Mater., vol. 34, pp. 457-465, 2005. [23] M. Bassler, G. Pensl and V. V. Afanas'ev, ""Carbon cluster model" for electronic states at SiC/SiO 2 interfaces," Diamond Relat. Mater., vol. 6, pp. 1472-1475, 1997. [24] B. E. Deal, "Standardized terminology for oxide charges associated with thermally oxidized silicon," IEEE Trans. Electron Devices, vol. 27, pp. 606-608, 1980. [25] D. Brisbin and Q. Wang. (2002, Evaluating oxide reliability with V-ramp and Jramp testing. [http://archive.evaluationengineering.com/archive/articles/0802wafer.htm ].

PAGE 140

125 [26] J. T. Yue, "Reliability," in ULSI Technology C. Y. chang and S. M. Sze, Eds. New York, NY: McGraw-Hill, 1996, pp. 656-704. [27] F. Jensen, Electronic Component Reliability : Fundamentals, Modelling, Evaluation, and Assurance. Chichester, England;: John Wiley & Sons Ltd., 1995. [28] S. Chakraborty, M. K. Bera, S. Bhattacharya and C. K. Maiti, "Current conduction mechanism in TiO 2 gate dielectrics," Microelectron. Eng., vol. 81, pp. 188-193, 2005. [29] W. R. Harrell and J. Frey, "Observation of PooleFrenkel effect saturation in SiO 2 and other insulating films," Thin Solid Films, vol.352, pp. 195-204, 1999. [30] R. H. J. Good and E. W. Muller, "Field emission," in Encyclopedia of Physics: Electron-Emission Gas Discharge I vol. 21, S. Flugge, Ed. Berlin: Springer-Verlag, 1956, pp. 176-231. [31] J. M. Houston, "The slope of logarithmic plots of the Fowler Nordheim equation," Phys. Rev., vol. 88, pp. 349, 1952. [32] R. A. Ashton, "Gate oxide thickness measurement using Fowler-Nordheim tunneling," Proc. IEEE Int. Conf. Microelectron. Test Struct., vol. 4, pp. 57-60, 1991. [33] D. K. Schroder, "Oxide and interface trapped charges, oxide integrity," in Semiconductor Material and Device Characterization ,2nd ed.Anonymous New York: John Wiley & Sons, Inc., 1998, pp. 337-419. [34] M. Lenzlinger and E. H. Snow, "Fowler-Nordheim tunneling into thermally grown SiO 2 ," J. Appl. Phys., vol. 40, pp. 278-283, 1969. [35] Z. A. Weinberg, "On tunneling in metal-oxide-silicon structures," J. Appl. Phys., vol. 53, pp. 5052-5056, 1982. [36] M. Depas, B. Vermeire, P. W. Mertens, R. L. Van Meirhaeghe and M. M.Heyns, "Determination of tunneling parameters in ultra-thin oxide layer poly-Si/SiO2/Si structures," Solid-State Electron., vol. 38, pp. 1465-1471, 1995. [37] C. Y. Ng, T. P. Chen, Y. Liu and C. Q. Sun, "Influence of nitrogen on tunneling barrierheights and effective masses of electrons and holes at lightly-nitrided SiO2 /Si interface," Journal of Applied Physics, vol. 96, pp. 5912-5914, 2004. [38] R. Williams and M. H. Woods, "High electric fields in silicon dioxide produced by corona charging," J. Appl. Phys., vol. 44, pp. 1026-1028, 1973.

PAGE 141

126 [39] Z. A. Weinberg, W. C. Johnson and M. A. Lampert, "High-field transport in SiO 2 on silicon induced by corona charging of the unmetallized surface," J. Appl. Phys., vol. 47, pp. 248-255, 1976. [40]Z. A. Weinberg, "Tunneling of electrons from Si into thermally grown SiO2 ," SolidState Electron., vol. 20, pp. 11-18, 1977. [41] J. H. Chen, C. T. Wei, S. M. Hung, S. C. Wong and Y. H. Wang, "Breakdown and stress-induced oxide degradation mechanisms in MOSFETs," Solid-State Electron., vol. 46, pp. 1965-1974, 2002. [42] T. Tomita, H. Utsunomiya, T. Sakura, Y. Kamakura and K. Taniguchi, "A new soft breakdown model for thin thermal SiO 2 films under constant current stress," IEEE Trans. Electron Devices, vol. 46, pp. 159-164, 1999. [43] D. J. DiMaria, E. Cartier and D. Arnold, "Impact ionization, trap creation, degradation, and breakdown in silicon dioxide films on silicon," J. Appl. Phys., vol. 73, pp. 3367-3384, 1993. [44] Y. Chen, J. S. Suehle, C. Shen, J. B. Bernstein, C. Messick and P. Chaparala, "A new technique for determining long-term TDDB acceleration parameters of thin gate oxides," IEEE Electron Device Lett., vol. 19, pp. 219-221, 1998. [45] J. H. Stathis, "Percolation models for gate oxide breakdown," J. Appl. Phys., vol. 86, pp. 5757-5766, 1999. [46] I. C. Chen and C. Hu, "Accelerated testing of time-dependent breakdown of SiO 2 ," IEEE Electron Device Lett., vol. 8, pp. 140-142, 1987. [47] W. Nelson, Accelerated Testing: Statistical Models, Test Plans, and Data Analyses. New York: Wiley-Interscience Publication, 1990. [48] E. Persson, "Analysis of reliability test failure data: data processing and presentation," (private communication), 2006. [49] J.Senzaki, K. Kojima and K. Fukuda, "Effects of n-type 4H-SiC epitaxial wafer quality on reliability of thermal oxides," Appl. Phys. Lett., vol. 85, pp. 6182-6184, 2004. [50] J. Senzaki, K. Kojima, T. Kato, A. Shimozato and K. Fukuda, "Correlation between reliability of thermal oxides and dislocations in n-type 4H-SiC epitaxial wafers," Appl. Phys. Lett., vol. 89, pp. 022909-1-022909-3, 2006. [51] M. M. Maranowski and J. A. Cooper Jr., "Time-dependent-dielectric-breakdown measurements of thermal oxides," IEEE Trans. Electron Devices, vol. 46, pp. 520-524, 1999.

PAGE 142

127 [52] K. Matocha, G. Dunne, S. Soloviev and R. Beaupre, "Time-dependent dielectric breakdown of 4H-SiC MOS capacitors and DMOSFETs," IEEE Trans. Electron Devices, vol. 55, pp. 1830-1834, 2008. [53] D. J. DiMaria and E. Cartier, "Mechanism for stress-induced leakage currents in thin silicon dioxide films," J. Appl. Phys., vol. 78, pp. 3883-3894, 1995. [54] J. De Blauwe, J. Van Heudt, D. Wellekens, G. Groeseneken and H. E. Maes, "SILCrelated effects in flash E 2 PROM's-Part I: aquantitative model for steady-state SILC," IEEE Trans. Electron Devices, vol. 45, pp. 1745-1750, 1998. [55] B. De Salvo, G. Ghibaudo, G. Panabnakakis, B. Guillaumo and G. Reimbold, "Study of stress induced leakage current by using high resolution measurements," Microelectron. Reliab., vol. 39, pp. 797-802, 1999. [56] M. P. Houng, Y. H. Wang and W. J. Chang, "Current transport mechanism in trapped oxides: ageneralized trap-assisted tunneling model," J. Appl. Phys., vol. 86, pp. 1488-1491, 1999. [57] D. Ielmini, A. S. Spinelli, A. L. Lacaita, A. Martinelli and G. Ghidini, "A recombination-and trap-assisted tunneling model for stress-induced leakage current," Solid-State Electron., vol. 45, pp. 1361-1369, 2001. [58] K. Sakakibara, N. Ajika, M. Hatanaka, H. Miyoshi and A. Yasuoka, "Identification of stress-induced leakage current components and the corresponding trap models in SiO 2 films," IEEE Trans. Electron Devices, vol. 44, pp. 2267-2273, 1997. [59] L. Pantisano and K. P. Cheung, "Stress-induced leakage current (SILC) and oxide breakdown: are they from the same oxide traps?" IEEE Trans. Device Mater. Reliab., vol. 1, pp. 109-112, 2001. [60] M. H. Woods and R. Williams, "Injection and removal of ionic charge at room temperature through the interface of air with SiO 2 ," J. Appl. Phys., vol. 44, pp. 55065510, 1973. [61] S. Sakata, M. Inoue, T. Kushima and T. Okada, "Formation of clusters from clean room air by corona discharge," in The 39 Th Annual Technical Meeting of Instituteof Environmental Sciences, 1993, pp. 79-86. [62] M. Pavlik and J. D. Skalny, "Generation of [H 3 O] + (H 2 O) n clusters by positive corona discharge in air," Rapid Commun. Mass Spectrom., vol. 11, pp. 1757-1766, 1997. [63] J. D. Skalny, T. Mikoviny, S. Matejcik and N. J. Mason, "An analysis of mass spectrometric study of negative ions extracted from negative corona discharge in air," Int. J. Mass Spectrom., vol. 233, pp. 317-324, 2004.

PAGE 143

128 [64] J. Lagowski, M. Wilsonand A. Savtchouk, "Method for measuring stress induced leakage current and gate dielectric integrityusing corona discharge," USPatent 6538462B1, Mar. 25, 2003. [65] A. Savtchouk, J. Lagowski, J. DAmico, M. Wilson and L. Jastrzebski, "Steady state method formeasuring the thickness and the capacitance of ultrathin dielectric in the presence of substantial leakage current," USPatent 668062B2, Jan. 20, 2004. [66] M. Wilson, J. Lagowski, L. Jastrzebski, A. Savtchouk and V. Faifer, "COCOS (corona oxide characterization of semiconductor) non-contact metrology for gate dielectrics," in Characterization and Metrology for ULSI Technology, 2001, pp. 220-225. [67] M.Wilson, J. Lagowski, A. Savtchouk, D. Marinskiy, L. Jastrzebski and J. D'Amico, "COCOS (corona oxide characterization of semiconductor) metrology: physical principles and applications," in Gate Dielectric Integrity: Material, Process, and Tool Qualification D. C. Gupta and G. A. Brown, Eds. ASTM STP1382 West Conshohocken, Pa., 1999, pp. 74-90. [68] J. D'Amico, "Thin Silicon Dioxide Integrity: Advanced Techniques," Ph. D. Dissertation, University of South Florida, 2000. [69] P. Edelman, J. Lagowski, A. Savtchouk, M. Wilson, A. Aleynikov, D. Marinskiy and J. Navarro, "Full wafer non-contact mapping of electrical properties of ultra-thin advanced dielectrics on Si," Mater. Sci. Eng. B, vol. 91-92, pp. 211-215, 2002. [70] Semiconductor Diagnostics Inc. [http://www.sditampa.com ]. [71] P. Edelman, D. Marinskiy, C. Almeida, J. N. Kochey, A. Byelyayev, M. Wilson, A. Savtchouk, J. DAmico, A. Findlay, L. Jastrzebski and J. Lagowski, "Non-contact charge-voltage method for dielectric characterization on small test areas of IC product wafers," Mater. Sci. Semicond. Process., vol. 9, pp. 252-256, 2006. [72] E. H. Nicollian and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and Technology. New York: Wiley-Interscience Publication, 1982. [73] B. V. Zeghbroeck. (2004, Principles of Semiconductor Devices. [http://ecewww.colorado.edu/~bart/book/ ]. [74] M. S. Dautrich, P. M. Lenahan, A. Y. Kang and J. Conley J.F., "Noninvasive nature of corona charging on thermal Si/SiO 2 structures," Appl. Phys. Lett., vol. 85, pp. 18441845, 2004. [75] A. M. Hoff, E. Oborina, S. E. Saddow and A. Savtchouk, "Thermal oxidation of 4Hsilicon carbide using the afterglow method," Mater. Sci. Forum, vol. 457-460, pp. 13491352, 2004.

PAGE 144

129 [76] P. Edelman, M. Wilson, J. D'Amico, A. Savtchouk and J. Lagowski, "Bandoffset diagnostic of advanced dielectrics," J. Mater Sci: Mater Electron, vol. 19, pp. 73-78, 2008. [77] A. M. Hoff, "Growth and metrology of silicon oxides on silicon carbide," Mater. Res. Soc. Symp. Proc., vol. 815, pp. 189-198, 2004. [78] J. D'Amico, "Theory and application of non-contact methods for in-line reliability determination," inIntegrated Reliability Workshop Final Report, 2003IEEE International, pp.157-168. [79] W. Kern and D. A. Puotinen, "Cleaning solutions based on hydrogen peroxide for use in silicon semiconductor technology," RCA Rev., vol. 31, pp. 187, 1970. [80] A. Savtchouk, E. Oborina, A. M. Hoffand J. Lagowski, "Non-contact doping profiling in epitaxial SiC," Mater. Sci. Forum, vol. 457/460, pp. 755-758, 2004. [81] V. V. Afanas'ev, M. Bassler, G. Pensl, M. J. Schulz and E. S. von Kamienski, "Band offsets and electronic structure of SiC/SiO 2 interfaces," J. Appl. Phys., vol. 79, pp. 31083114, 1996. [82] H. Yamada, "Additional fluorine passivation to pyrolytic-N 2 O passivated ultrathin silicon oxide/Si(100) films," J. Appl. Phys., vol. 100, pp. 034507-1-034507-7, 2006. [83] R. K. Chanana, K. McDonald, M. Di Ventra, S. T. Pantelides, L. C. Feldman, G. Y. Chung, C. C. Tin, J. R. Williams and R. A. Weller, "Fowler-Nordheim hole tunneling in p-SiC/SiO 2 structures," Appl. Phys. Lett., vol. 77, pp. 2560-2562, 2000. [84] P. Friedrichs, E. P. Burte and R. Schomer, "Dielectric strength of thermal oxides on 6H-SiC and 4H-SiC," Appl. Phys. Lett., vol. 65, pp. 1665-1667, 1994. [85] H. Li, S. Dimitrijev, D. Sweatman and H. B. Harrison, "Analysis of Fowler Nordheim injection in NO nitrided gate oxide grown on n-type 4HSiC," Microelectron. Reliab., vol. 40, pp. 283-286, 2000. [86] R. Waters and B. V. Zeghbroeck, "Temperature-dependent tunneling through thermally grown SiO 2 on n-type 4H-and 6H-SiC," Appl. Phys. Lett., vol. 76, pp. 10391041, 2000. [87] A. K. Agarwal, S. Seshadri and L. B. Rowland, "Temperature dependence of Fowler-Nordheim current in 6H-and 4H-SiC MOS capacitors," IEEE Electron Device Lett., vol. 18, pp. 592-594, 1997. [88] B. Brar, G. D. Wilk and A. C. Seabaugh, "Direct extraction of the electron tunneling effective mass in ultrathin SiO 2 ," Appl. Phys. Lett., vol. 69, pp. 2728-2730, 1996.

PAGE 145

130 [89] R. Kies, T. Egilsson, G. Ghibaudo and G. Pananakakis, "A method for the assessment of oxide charge density and centroid in metal-oxide-semiconductor structures after uniform gate stress," Appl. Phys. Lett., vol. 68, pp. 3790-3792, 1996. [90] C. Papadas, P. Morfouli, G. Ghibaudo and G.Pananakakis, "Analysis of the trapping characteristics of silicon dioxide after Fowler-Nordheim degradation," Solid-State Electron., vol. 34, pp. 1375-1379, 1991. [91] S. J. Oh and Y. T. Yeow, "Voltage shifts of Fowler-Nordheim tunneling J-V plots in thin gate oxide MOS structures due to trapped charges," Solid-State Electron., vol. 32, pp. 507-511, 1989. [92] J.A. Lopez-Villanueva, J.A. Jimenez-Tejada, P. Cartujo, J. Bausells, and J.E. Carceller, "Analysis of the effects of constant-current Fowler-Nordheim-tunneling injection with charge trapping inside the potential barrier," J. Appl. Phys. vol. 70, pp. 3712-3720, 1991. [93] L. F. Mao, Z. O. Wang, J. Y. Wang and G. Y. Yang, "The effective mass issues in light nitride silicon oxide," Semicond. Sci. Technol., vol. 20, pp. 1078-1082, 2005.

PAGE 146

131 Appendices

PAGE 147

132 Appendix A: Afterglow Oxide Recipes Table A.1: AG I_A Recipe. Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pretreatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 ReOx Anneal 850 O 2 : N 2 O : FG (3:0.2:0.5) 60 7 Ramp Down 850 to 600 Ar (0.23) 75 8 Unload 600 N 2 (10) 5 TableA.2: AG I_B Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pretreatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 40 Yes 6 ReOx Anneal 850 O 2 : N 2 O : FG (3:0.2:0.5) 25 7 Ramp Down 850 to 740 Ar (0.23) 5 8 Unload 740 N 2 (10) 5

PAGE 148

133 Appendix A (Continued) TableA.3: AG II Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 Ramp Up 850 to 950 Ar (0.23) 24 7 Ar Anneal 950 Ar (0.23) 60 8 Ramp Dow 950 to 600 Ar (0.23) 90 9 Unload 600 N 2 (10) 5

PAGE 149

134 Appendix A (Continued) TableA.4: AG III Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 ReOx Anneal 850 O 2 : N 2 O: FG (3:0.2:0.5) 60 7 Ramp Up 850 to 950 Ar (0.23) 24 8 Ar Anneal 950 Ar (0.23) 60 9 Ramp Down 950 to 600 Ar (0.23) 90 10 Unload 600 N 2 (10) 5

PAGE 150

135 Appendix A (Continued) TableA.5: AG IV Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 Ramp Up 850 to 950 O 2 : FG (3:1) 20 7 ReOx Anneal 950 O 2 : FG (3:1) 120 8 Ramp Down 950 to 600 Ar (0.23) 90 9 Unload 600 N 2 (10) 5

PAGE 151

136 Appendix A (Continued) TableA.6: AG V Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 Ramp Up 850 to 900 O 2 : N 2 O :FG (3:0.2:0.5) 10 7 ReOx Anneal 900 O 2 : N 2 O : FG (3:0.2:0.5) 360 8 Ramp Down 900 to 600 Ar (0.23) 90 9 Unload 600 N 2 (10) 5

PAGE 152

137 Appendix A (Continued) TableA.7: AG VI Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 Ramp Up 850 to 950 O 2 : N 2 O (4:0.3) 20 7 ReOx Anneal 950 O 2 : N 2 O (4:0.3) 120 8 Ramp Down 950 to 600 Ar (0.23) 90 9 Unload 600 N 2 (10) 5

PAGE 153

138 Appendix A (Continued) TableA.8: AG VII Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 7 Yes 6 Ar Anneal 850 Ar (0.23) 60 7 Ramp Down 850 to 600 Ar (0.23) 75 8 Unload 600 N 2 (10) 5 HF Vapor Etch 9 Load 600 N 2 (10) 5 10 Ar Anneal 600 Ar (0.23) 20 11 Unload 600 N 2 (10) 5

PAGE 154

139 Appendix A (Continued) TableA.9: AGWI Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 6 Ramp Down 850 to 600 Ar (0.23) 75 7 Unload 600 N 2 (10) 5

PAGE 155

140 Appendix A (Continued) TableA.10: AGWII Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 850 Ar (0.23) 38 3 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 90 Yes 4 Ramp Down 850 to 600 Ar (0.23) 75 5 Unload 600 N 2 (10) 5 TableA.11: AGWIII Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 10 Yes 6 Unload 850 N 2 (10) 5

PAGE 156

141 Appendix A (Continued) TableA.12: AGWIV Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 Ramp Up 600 to 850 Ar (0.23) 38 4 Oxidation 850 O 2 : N 2 O : FG (3:0.2:0.5) 10 Yes 5 Unload 850 N 2 (10) 5 TableA.13: AGWV Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (3:0.2:1) 90 Yes 6 Ramp Down 850 to 600 Ar (0.23) 75 7 Unload 600 N 2 (10) 5

PAGE 157

142 Appendix A (Continued) TableA.14: AGWVI Recipe.Interval Description Temp ( 0 C) Gas Flow (l/min) Time (min) Excited Media 1 Load 400 N 2 (10) 5 2 Ramp Up 400 to 600 O 2 : N 2 (0.5:3.5) 28 3 FG Pre-treatment 600 FG (4) 20 Yes 4 Ramp Up 600 to 850 Ar (0.23) 38 5 Oxidation 850 O 2 : N 2 O : FG (2.5:0.15:1.3) 90 Yes 6 Ramp Down 850 to 600 Ar (0.23) 75 7 Unload 600 N 2 (10) 5

PAGE 158

About the Author Helen Benjaminwas born on the Caribbean island of Martinique. She attended Sts. Peter and Paul High School in St. Thomas, U.S.Virgin Islands, where she graduated as the Valedictorian of her class. She began her college education at the University of the Virgin Islands in Mathematics. After one year, she transferred to the Universityof South Florida, Tampa, FL, where shereceivedher Bachelor of Science and Masters of Science in Electrical Engineering. During the pursuit of her Doctoral degree, she received three prestigious awards: the Alfred P. Sloan Foundation Scholarship Award, the Integrative Graduate Education and Research Traineeship(IGERT)Award, and the National Science Foundation (NSF) Graduate Research Fellowship Award. Shegained teaching experience by being a laboratory instructor for the Integrated-Circuit (IC) Processing Lab for ayear. She authored one publication in Sensors and Actuators B andco-authored publications related to this dissertationand her Masters thesis.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200397Ka 4500
controlfield tag 001 002061864
005 20100225145448.0
007 cr mnu|||uuuuu
008 100225s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002984
035
(OCoLC)528827476
040
FHM
c FHM
049
FHMM
090
TK145 (Online)
1 100
Benjamin, Helen N.
0 245
Non-contact characterization of dielectric conduction on 4H-SiC
h [electronic resource] /
by Helen N. Benjamin.
260
[Tampa, Fla] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 142 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Consistent charge or defect control in oxide grown on silicon carbide (SiC) continues to be difficult to achieve and directly impacts the electrical performance of SiC-based metal oxide semiconductor (MOS) devices. This research applied non-contact Corona-Kelvin metrology to investigate the charge transport in oxides grown on n-type 4H-SiC epitaxial substrates. The cost and engineering science impact of this metrology are significant as device fabrication is avoided leading to quick determination of electrical characteristics from as-grown oxide films. Non-contact current-voltage (I-V) measurements of oxide on SiC were first demonstrated within this work and revealed that Fowler-Nordheim (F-N) current emission was the dominant conduction mechanism at high electric fields.Oxides on SiC were grown at atmospheric pressure (thermal oxides) or at a reduced pressure (afterglow oxides) ambient and examined using non-contact charge-voltage (Q-V), capacitance-voltage (C-V), equivalent oxide thickness (EOT), and I-V methods. The F-N conduction model was modified to address charge trapping and effective barrier effects obtained from experimental oxide films. Trap densities determined with this metrology were used to show that the F-N model including their density and position was adequate for thermal oxides on SiC but not for afterglow films. Data from the latter films required further modification of the theory to include a chemical effect of the oxide growth process on the effective conduction band offset or barrier. This work showed that afterglow chemistry was able to vary the effective conduction band offset from 2.9 eV, typical of thermal oxidation of SiC, up to 3.2 eV.Stress induced leakage current (SILC), an excess above the F-N base current resulting from prolonged current through the dielectric films, was also investigated. Multiple point SILC testing was used to identify statistical effects of process variations and defects in as-grown oxide films on SiC. These results open the possibility to improve oxide manufacture on SiC using methods common in the silicon IC industry. This work demonstrated the first non-contact F-N current determination in oxides on SiC and showed both charge trapping and chemical dependencies of as-grown films. Future studies may extend the findings of this work to further improve this important dielectric-semiconductor system.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Andrew M. Hoff, Ph.D.
653
Corona-Kelvin metrology
Field emission
Voltage decay
Non-contact SiLC
Trapped charge
690
Dissertations, Academic
z USF
x Electrical Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2984