|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Sequential Afterglow Processing and Non-C ontact Corona-Kelvin Metrology of 4H-SiC by Eugene L. Short, III 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 Hoff, Ph.D. Kenneth Buckle, Ph.D. Richard Gilbert, Ph.D. Stephen Saddow, Ph.D. Sarath Witanachchi, Ph.D. Date of Approval: June 22, 2009 Keywords: silicon carbide, remote plasma oxidation, surface conditioning, thin films Copyright 2009, Eugene L. Short, III
Acknowledgments First and foremost I am pleased to express my gratitude to my major professor, Dr. Andrew Hoff, for his invaluable support, insigh t and guidance. It ha s been truly an honor working with him. I would like to tha nk my committee members Dr. Kenneth Buckle, Dr. Richard Gilbert, Dr. Stephen Saddow, and Dr. Sarath Witanachchi, and chairperson Dr. Scott Campbell for their time and input. I am greatly indebted to Dr. Elena Oborina for all of her helpful collaboration and advi ce throughout this wor k, particularly with non-contact electrical characte rization. I would also like to thank each and every member of the USF SiC research group for their t eamwork, including Dr. Helen Benjamin, Dr. Chris Frewin, Chris Locke, Norelli Schettini a nd Dr. Jeremy Walker. I am grateful to the staff at Semiconductor Diagnostics, Inc. who contributed technical support for the FAaST measurement tool. I am also indebted to Richard Everly and Robert Tufts at the USF Nanomaterials and Nanotechnology Research Ce nter for their assist ance with technical issues regarding the afterglow reactor and cl eanroom. Fred Stevie of North Carolina State University provided XPS measur ement services and useful discussion.
i Table of Contents List of Tables iii List of Figures iv Abstract ix Chapter 1. Introduction and Background 11.1.SiC material properties a nd device applications 21.2.Theory of oxidation 51.2.1.Deal Grove linear-parabolic model for thermal oxidation of Si 51.2.2.Model for thermal oxidation of SiC 81.3.SiO2/SiC structure formation and improvement efforts 101.3.1.Conventional thermal oxidation of SiC 101.3.2.An alternative approach: remote plasma processing 131.3.3.Surface conditioning 161.4.Capacitance-voltage character ization of oxide/semiconductor structures 171.4.1.Capacitance-voltage measurement fundamentals 181.4.2.Contact vs. non-contact metrology 22 Chapter 2. Experimental Approach 262.1.Afterglow chemical processing 262.1.1.Afterglow apparatus description and operation 272.1.2.Dielectric growth by remote plasma sequential processing 302.2.Non-contact corona-Kelvin metrology 332.2.1.Corona-Kelvin tool operation and basis of measurement 332.2.2.Oxide/4H-SiC structures: typical non-contact capacitance-voltage behavior 39 Chapter 3. Corona-Kelvin Capacita nce Metrology of Afterglow Oxide Films 423.1.Oxidation time and temperature results vs. surface conditioning 433.2.High-temperature annealing effects vs. surface conditioning 50 Chapter 4. Corona-Kelvin VCPD Transients on Conditioned 4H-SiC Surfaces 574.1.VCPD transient measurement protocol and interpretation 574.2.Surface conditioning impact on VCPD decay 684.3.(N2:H2)* afterglow treatment vari ations: time, temperature, durative stability 78
ii Chapter 5. X-ray Photoelectr on Spectroscopy of Conditioned 4H-SiC Surfaces 915.1.XPS measurement technique 915.2.XPS results on 4H-SiC surfaces 93 Chapter 6. Conclusion 1036.1.Results summary 1036.2.Future work 108 References 110 About the Author End Page
iii List of Tables Table 1.1.Selected material properties of SiC and Si semiconductors at 27C. 4 Table 1.2.Comparison of selected 4H-SiC thermal oxidation results from atmospheric pyrogenic steam and remote plasma processes. 14 Table 3.1.Net total oxide charge estimated from non-contact C-V characteristics of afterglow oxide films. 55 Table 5.1.XPS atomic percent and ratios of selected elements obtained on n-type 4H-SiC surfaces treated by RCA clean or (FG)* afterglow conditioning, before and after sputtering. 101
iv List of Figures Figure 1.1.Tetrahedral bonding arrangement found in a SiC crystalline lattice. 3 Figure 1.2.Transport steps assumed in mode ling thermal oxidation of Si (a) and SiC (b). 7 Figure 1.3.Electrical model for th e capacitances and potentials associated with an oxide/semiconductor structure. 19 Figure 1.4.Electron energy band diagrams representing accumulation (a), depletion (b), and flat-b and condition (c) of an oxide/semiconductor structure under applied bias. 20 Figure 1.5.Oxide C-V characteristic example on n-type semiconductor, illustrating the effects of C-V stretch-out and flat-band shifting. 21 Figure 1.6.Alternative techniques of genera ting a bias pote ntial across an oxide/ semiconductor structure: MO S contact (a), Hg-probe (b), and corona ion deposition (c). 23 Figure 2.1.Schematic diagram of the remote plasma afterglow apparatus. 27 Figure 2.2.Photographic image of the remote plasma apparatus furnace zone during operation, with vi sible chemo-luminescence of afterglow excited species. 28 Figure 2.3.Photographic images of the remote plasma apparatus microwave cavity, depicting the plasma discharge and afterglow during operation. 29 Figure 2.4.Example of a general afterglo w process flow temperature profile, including pre-oxidation surface treatment, oxide growth, and post-oxidati on annealing steps. 31 Figure 2.5.Non-contact CPD probe schematic. 35 Figure 2.6.Typical VCPD data obtained during co rona-Kelvin metrology of an oxidized p-type SiC epi-layer. 36
v Figure 2.7.Typical V-Q response obtained during corona-Kelvin metrology of an oxidized p-type SiC epi-layer. 37 Figure 2.8.Typical C-V characteristic ex tracted from corona-Kelvin metrology V-Q data on an oxidi zed p-type SiC epi-layer. 38 Figure 2.9.Typical non-contact C-V curves obtained sequentially at a single measurement site on oxidized p-type 4H-SiC. 40 Figure 3.1.Electrical thicknesses of afterglow oxide films grown in (O2:N2O:FG)* afterglow ambient for various time intervals on 4H-SiC substrates at 850C, so me of which were subjected to (FG)* surface conditioning at 600C. 44 Figure 3.2.Thickness uniformity of afterglow oxide films grown for various time intervals on 4H-SiC substrates at 850C, some of which were subjected to (FG) surface conditioning at 600C. 45 Figure 3.3.EOTs of afterglow oxide films grown for 15 min. on 4H-SiC substrates at temperatures between 600C and 800C, some of which were subjected to (FG) surface conditioning at 600C. 46 Figure 3.4.Electrical thicknesses of afterglow oxide films grown in (O2:N2O:FG)* afterglow ambient for various time intervals on p-type 4H-SiC substrates at 600C, some of which were subjected to pre-oxidation (FG)* surface conditioning at 600C. 48 Figure 3.5.Thickness uniformity of afterglow oxide films grown for various time intervals on p-type 4H-SiC substrates at 600C, some of which were subjected to pre-oxidation (FG)* surface conditioning at 600C. 49 Figure 3.6.Non-contact C-V characteristics of oxide films grown for 15 min. at 850C on p-type (a) and n-type (b) 4H-SiC substrates, some of which underwent pr e-oxidation (FG)* surface conditioning at 600C and/or pos t-oxidation Ar annealing at 1000C for 30 min. 52 Figure 3.7.Non-contact C-V characteristics of oxide films grown for 60 min. at 600C on p-type (a) and n-type (b) 4H-SiC substrates, some of which underwent pr e-oxidation (FG)* surface conditioning at 600C and/or pos t-oxidation Ar annealing at 950C for 30 min. 53
vi Figure 4.1.Depletion surface barrier transients obtained at multiple sites on RCA cleaned n-type 4H-SiC epi-wafer A after negative corona deposition. 59 Figure 4.2.Diagram of the charge compen sation mechanism associated with the temporal decay of surface barrier, depletion width, and space-charge density. 62 Figure 4.3.Illustration of electric field e nhanced carrier emission from localized states. 63 Figure 4.4.Depletion surface barrier transients obtained at multiple sites on RCA cleaned n-type 4H-SiC epi-wafer B after negative corona deposition. 65 Figure 4.5.VCPD transient decays with consecutive repetitions of corona deposition spaced at 3 min. intervals, obtained on RCA cleaned n-type 4H-SiC epi-wafer A. 67 Figure 4.6.Depletion VSB transients obtained at multiple sites on n-type 4H-SiC epi-wafers A (a) and B (b) following (N2:H2)* afterglow surface conditioning for 20 min. at 600C. 69 Figure 4.7.VCPD transient decays with consecutive repetitions of corona deposition spaced at 3 min. intervals, obtained on n-type 4H-SiC epi-wafer A after (FG)* surface treatment for 20 min. at 600C. 71 Figure 4.8.Depletion surface barrier decay s obtained on n-type 4H-SiC epi-wafers A (a) and B (b) af ter various surface conditioning treatments, including (N2:H2)* or (N2)* afterglow exposure and non-excited N2:O2 media at 600C for 20 min., DI water rinsing after (FG)* conditioning, a nd standard RCA cleaning. 73 Figure 4.9.Depletion surface barrier transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) af ter various surface conditioning treatments, plotted relative to initial measured voltage to aid visualization of VSB decay rates. 74 Figure 4.10.Final voltage values of depletion surface barrier transients obtained on n-type 4H-SiC epiwafers after various surface conditioning treatments. 75 Figure 4.11.Depletion VSB transients obtained on p-type 4H-SiC 1 cm2 sample comparing RCA clean to (FG)* afterglow surface conditioning for 20 min. at 600C. 77
vii Figure 4.12.VSB transient decays obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow treatment for various time intervals at 600C. 79 Figure 4.13.Final voltage values of depletion VSB transient decays obtained on n-type 4H-SiC epi-wafers following (FG)* afterglow treatment for various ti me intervals at 600C. 80 Figure 4.14.Uniformity of VSB transient decays obtained on n-type 4H-SiC following (FG)* afterglow treatment for various time intervals at 600C. 81 Figure 4.15.Depletion VSB transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditioning for 20 min. at treatment temperatures in the range 400C 800C. 83 Figure 4.16.Depletion VSB transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditioning for 20 min. at various treatment temperat ures, plotted relative to initial measured voltage to aid viewing of VSB decay rates. 84 Figure 4.17.Final voltage values of depletion VSB transients obtained on n-type 4H-SiC epi-wafers following (FG)* afterglow conditioning for 20 min. at treatme nt temperatures in the range 400C 800C. 85 Figure 4.18.Depletion surface barrier decay s obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditioning for 20 min. at 800C, and remeasured after 1 day intervals of time delay. 87 Figure 4.19.Depletion surface barrier decay s obtained on n-type (a) and p-type (b) 4H-SiC 1 cm2 samples following (FG)* afterglow conditioning for 20 min. at 600C, and remeasured after accumulation corona stress, 6 day time delay, and heating in cleanroom ambient. 89 Figure 5.1.XPS measurement schematic. 92 Figure 5.2.Electron energy band diagram illustrating photoemission of core level electrons in the XPS technique. 93 Figure 5.3.XPS spectral data (a) and atomic percent values (b) obtained on n-type 4H-SiC surfaces after RCA clean or (FG)* afterglow treatment. 94
viii Figure 5.4.XPS spectral data (a) and atomic percent values (b) obtained on n-type 4H-SiC surfaces as treated by (FG)* afterglow, and after sputtering. 97 Figure 5.5.XPS spectral data (a) and atomic percent values (b) obtained on n-type 4H-SiC surfaces as treated by RCA clean, and after sputtering. 98 Figure 5.6.XPS spectral data (a) and atomic percent values (b) obtained after sputtering of n-type 4H-SiC surfaces treated by RCA clean or (FG)* afterglow. 100
ix Sequential Afterglow Processing and Non-C ontact Corona-Kelvin Metrology of 4H-SiC Eugene L. Short, III ABSTRACT Silicon carbide (SiC) is a wide bandgap semiconductor with advantageous electrical and thermal properties making it attractive for high temperature and power applications. However, difficulties with oxide /SiC structures have posed challenges to the development of practical MOS-type de vices. Surface conditioning and oxidation of 4H-SiC were investigated using a novel sequential afterglow processing approach combined with the unique capabilities of non-contact corona-Kelvin metrology. The use of remote plasma assisted thermal oxidation facilitated film growth at low temperature and pressure with the flexibility of sequential in-situ processing options including preoxidation surface conditioning. Corona-Kelvi n metrology (C-KM) provided a fast, nondestructive method for electrical evaluation of oxide films and semiconductor surfaces. Non-contact C-KM oxide capacitance-voltage characteristics combined with direct measurement of SiC surfaces using C-KM de pletion surface barrier monitoring and XPS analysis of surface chemistry were interp reted relating the impact of afterglow conditioning on the surface and its influence on subsequent oxide thin film growth. Afterglow oxide films of thicknesses 50500 we re fabricated on SiC epi-layers at low growth temperatures in the range 600850 C, an achievement not possible using conventional atmospheric oxidati on techniques. The inclusi on of pre-oxidation surface conditioning in forming gas (N2:H2)* afterglow was found to produce an increase in
x oxide growth rate (1025%) and a significan t improvement in oxide film thickness uniformity. Analysis of depletion voltage tr ansients on conditioned SiC surfaces revealed the highest degree of surface passivation, uni formity, and elimination of sources of charge compensation accomplished by the (N2:H2)* afterglow treatment for 20 min. at 600700C compared to other conditioning varia tions. The state of surface passivation was determined to be very st able and resilient when exposed to a variety of temporal, electrical, and thermal stress es. Surface chemistry analysis by XPS gave evidence of nitrogen incorporation and a reduction of the C/Si ratio achieved by the (N2:H2)* afterglow surface treatment, which was tied to the improvements in passivation, uniformity, and growth rate observed by noncontact C-KM measur ements. Collective results were used to suggest a clean, unifo rm, passivated, Si-enriched surface created by afterglow conditioning of 4H-SiC as a sequential preparation step for subsequent oxidation or dielectric formation processing.
1 Chapter 1. Intr oduction and Background Single-crystal silicon carbi de (SiC) is a promising wide band-gap (WBG) semiconductor material for future power el ectronic device applications, but requires continued technological improvement in or der to realize practical metal-oxidesemiconductor field-effect transistor (MO SFET) devices. SiC possesses superior electrical and thermal properties compared to silicon (Si), while remaining one of the few semiconductor materials forming silicon dioxide (SiO2) as its native oxide. This allows thermal oxidation methods to be employed in the fabrication of MOSFET power devices. The 4H hexagonal form of SiC has great pote ntial due to its wider band-gap and large electron mobility compared to other commonl y investigated SiC polytypes. However, electrically active defects present in the SiO2/4H-SiC material syst em have drastically limited achievable channel carrier mobilities a nd threshold voltage stability in device research efforts to date. Remote plasma assisted thermal processi ng offers an advantageous and flexible alternative to the conventional atmospheric thermal oxidation approa ch. An afterglow chemical reactor is capable of forming oxide films on SiC at higher growth rates and at temperatures hundreds of degrees lower th an required in traditional atmospheric oxidation furnace processes. Oxide growth at lower furnace temperatures is attractive for several reasons including lowe r processing cost. Atomic and excited oxidant species generated in a plasma discharge are suspected to play critical role s in oxidizing reactions
2 at the interface. Furthermore, oxidation by the afterglow method lends itself to the power and flexibility of sequential in-situ processing, including pre-oxidation surface conditioning steps and post-oxidation an neals in specific and novel chemical environments. Electrical measurements of semiconductor and oxide charac teristics are critical to the development of any material process or device application requiri ng dielectric films and interfaces. In-line me trology provides significant advantages over other common measurement techniques requiring additional device fabrication or thin film application. The ability to perform electrical measurements without building test structures translates to immense savings in time and cost of production. In addition, a non-contact characterization technique capable of obt aining quick, non-destructive electrical measurements gives one the opportunity to strategically implement this metrology at selected points in a se quence of processes. This work entails an investigation of SiO2/4H-SiC structures by applying the unique capabilities of remote plasma afterglow th ermal processing and non-contact metrology to the growth and characte rization of oxide thin films on 4H -SiC, with a focus on the impact of pre-oxidation su rface conditioning. 1.1. SiC material properties and device applications Silicon carbide is a binary compound semi conductor material comprised of Si and carbon (C) covalently bonded in a crystal line lattice. The tetrahedral bonding arrangement depicted in figure 1.1 is the building bl ock of every SiC crystal, with C-Si bond lengths measuring 1.89 and adjacent C atoms separated by 3.08 . The SiC lattice is structured with alte rnating planes of Si and C atoms. Each of the 170 polytypes
3 of SiC is defined by its specific ordering of at omic bi-layers of Si and C planes. The SiC polytype nomenclature contains a number refe rring to the amount of bi-layers after which the stacking order repeats. Among the most commonly investigated forms of SiC are the 3C, 4H, and 6H-SiC polytype s, with bi-layer stackin g sequences of ABCABCA..., ABCBABCBA..., and ABCACBABCACBA..., re spectively. The C or H polytype suffixes refer to the cubic or hexagona l crystal structure of the unit cell. Figure 1.1. Tetrahedral bonding arrangement found in a SiC crystalline lattice. SiC materials possess a wide energy band-gap (Eg), high therma l conductivity ( K ), large breakdown voltage (VBD), extremely low intrinsi c carrier concentration (ni), and chemical inertness. The aforementioned pr operties are well-suited for electronic device applications requiring high power, high volta ge, and high frequency operation in hightemperature and corrosive environments. Th e 4H-SiC polytype has generated the most research interest due to its wider band-gap and large electron mobility compared to other commonly investigated SiC polyt ypes. Table 1.1 lists select ed semiconductor material
4 parameters of several SiC polytypes as well as Si. Note that 4H-SiC possesses 3 larger Eg than Si, and 19 orders of magnitude smaller ni than Si. Table 1.1. Selected material properties of SiC and Si semiconductors at 27C. Note that 4H-SiC has 3 larger energy band-gap a nd 19 orders of magnitude smaller intrinsic carrrier concentration compared to Si. 4H-SiC 6H-SiC 3C-SiC Si Eg (eV) 3.26 3.03 2.36 1.12 ni (cm 3) 5 10 9 1.6 10 6 1.5 10 1 1.45 1010 n @ Nd = 1016 cm 3 (cm2 V 1 s 1) 800 400 800 1430 VSAT (cm s 1) 2.5 107 2 107 2.5 107 1 107 VBD @ Nd = 1017 cm 3 (MV cm 1) 3 3.2 1.5 0.3 K (W cm 1 K 1) 4.9 4.9 3.2 1.31 r 9.66 9.7 9.72 11.9 SiC is particularly appealing for MOSFET power device applicati ons because it is one of the few semiconductors which form a native SiO2 oxide layer, this due to the presence of Si in the crystal lattice. Because of this important property, numerous attempts have been made to apply thermal SiO2 growth techniques to SiC materials, analogous to the SiO2/Si based technology which has achieved unequaled success. However, practical SiC-based devices to date have been junction-type as MOS structures of desired quality have not been realized. A number of challenging problems have contributed to this failure, some of which ar e linked to the SiC mate rial quality itself. Bulk crystal quality is poor since substrates are produced by a sublimation process at very high temperatures with high metal contaminat ion levels. Furthermore, growth of high-
5 quality defect-free epitaxial f ilms critical for device applications has not been achieved. Epitaxial processes either genera te or propagate defects from the substrate, resulting in a relatively low quality of starting material for oxidation. 1.2. Theory of oxidation Perhaps the most promising advantage that SiC holds over other WBG materials is its ability, like Si, to thermally oxidize to form SiO2. In an attempt to understand and model the SiC oxidation mechanism and kinetics, it is helpful to first consider the relatively simple oxi dation of Si . 1.2.1. Deal Grove linear-parabolic mode l for thermal oxidation of Si Oxidation of silicon is governed by the transport of oxidant molecules to the SiO2/Si interface and reaction with Si su rface atoms according to the relation Si + O2 SiO2 (1) Si oxidation proceeds in a three-step sequence: 1) gas-phase transport and adsorption of molecular oxygen to the oxide surface, 2) in-diffusion of oxidant molecu les through any existing oxide film, 3) reaction with Si at the buried SiO2/Si interface. The first step is assumed to be rapid and not rate-controlling. A diagrammatic representation of the va rious stages involved in the oxidation of Si is depicted in figure 1.2a. The flux of oxidant molecules from the gas phase to the oxide surface is F = h(Cg Cs) (2) where h is the gas-phase transport coefficient, Cg is the equilibrium concentration of oxygen gas molecules, and Cs is the oxidant molecule con centration at the outer surface
6 of the existing oxide film. The flux of oxi dant molecules diffusi ng inward through the oxide film is defined as F = D(Cs Ci) / X (3) where D is the effective diffusion coefficient of oxidant in the oxide film, Ci is the oxidant molecule concentration near the SiO2/Si interface, and X is the thickness of the existing oxide film. Finally, the flux associat ed with the interfacial oxidation reaction is expressed as F = kCi (4) where k is the rate constant of the forward reacti on (1). The oxide growth rate at the SiO2/Si interface can be described as dX/dt = F/N0 (5) where N0 is the number of oxidant molecules in corporated into a unit volume of SiO2. The general model for the thermal oxidation of Si developed by Deal and Grove  assumed the three series fluxes of oxidant mo lecules to be constant and identical in steady state condition. Equating oxidant fluxes (2-4) and solv ing the differential equation (5) with some approximations yields th e general quadratic form expressed as X2 + AX = B(t + ) (6) where B and B/A are parabolic and linear ra te constants, respectively, t is time of oxidation, and is related to an initial thickness X = Xi assumed at t = 0. The rate constants and initial thickness para meter were defined as follows: B 2(Cg/N0)D (7) B/A (Cg/N0)(k 1 + h 1) 1 (8) = (Xi 2 + AXi) / B (9)
7 For short oxidation times and thin films, the oxidation rate is limited by the reaction at the SiO2/Si interface, and equation (6) can be approximated by X (B/A)(t + ) (10) resulting in a linear thicknesstime dependence. Alternativ ely, for long growth times and thick films, the growth rate is controlled by oxidant in-diffusion through the oxide film, and equation (6) reduces to X2 Bt (11) resulting in a parabolic relation be tween thickness and growth time. The Deal Grove linear-parabolic model succ essfully predicts thermal oxide growth rates on Si over a wide range of temperatur es, times and thicknesses. However, for oxidation by O2 molecules in the thin initial grow th regime, experimentally observed growth rates and thicknesses are consiste ntly higher than predicted by the linearparabolic model. (a) (b) Figure 1.2. Transport steps assumed in modeli ng thermal oxidation of Si (a) and SiC (b).
8 1.2.2. Model for thermal oxidation of SiC Oxidation of SiC is somewhat more comple x than Si due to the additional role of carbon in the growth kinetics. SiC oxidation is governed by the reaction SiC + 1.5O2 SiO2 + CO (12) The SiC oxidation process can be descri bed as a sequence of five steps: 1) gas-phase transport of molecular oxyge n and adsorption to the oxide surface, 2) in-diffusion of oxidant molecu les through the existing oxide film, 3) reaction with Si and C at the buried SiO2/SiC interface, 4) out-diffusion of vol atile reaction products (i .e. CO) through the oxide, 5) desorption and removal of CO products to the gas phase, where the first and last steps are assumed to be fast and not rate-limiting. The last two steps, not present in the oxidation of Si, add complexity to the SiC oxidation mechanism. Figure 1.2b visualizes the transport stages invo lved in the thermal oxi dation of SiC. The Deal Grove model cannot be directly applie d to SiC oxidation sin ce it does not account for the out-diffusion and removal of CO products However, a similar approach has been implemented to examine SiC oxidation kinetics . As before, the steady state in-flux (FO2) of oxidant molecules through the gas phase and SiO2 film is expressed as FO2 = hO2(Cg O2 Cs O2) = DO2(Cs O2 Ci O2) / X (13) Similarly, the steady state flux (FCO) describing the out-diffu sion and removal of carbon products is FCO = DCO(CiCO CsCO) / X = hCO(CsCO CgCO) (14) where coefficient and concentration subscr ipts are used to distinguish between O2 and CO molecules. The flux (FR) corresponding to the interf acial oxidation reaction is
9 FR = kfCi O2 krCiCO (15) where kf and kr are the forward and reverse rate cons tants of the oxidati on reaction (12). Again, the growth rate of the SiO2 layer is defined as dX/dt = FR /N0 (16) Under the steady state condition, the transp ort and reaction fluxe s are related as FR:FO2:FCO = 1:1.5:1 (17) After combining equations, the solution to (16) is in the same quadratic form as (6), although the coefficients A and B differ from th e Deal-Grove model. As with Si, the SiC oxidation kinetics exhibit lin ear and parabolic growth regimes corresponding to the interface reaction or diffusion processes being the rate-controlling step. In the interface reaction limited case, the linear ra te constant is approximated by B/A (Cg O2 /N0)kf (18) In the diffusion limited case, there are two possi ble extremes. If oxidant in-diffusion is the rate-controlling step, then the pa rabolic rate constant reduces to B (Cg O2 /1.5N0)DO2 (19a) Alternatively, if CO out-diffusion is the ra te-controlling step, then the parabolic rate constant can be approximated as B (Cg O2kf /N0kr)DCO (19b) It is also possible that both O2 and CO diffusion processes play comparable roles in controlling the growth rate. In such a case, the parabolic rate constant is expected to obey the more general relation B = N0 1(Cg O2kf CgCOkr)(1.5kf /DO2 + kr /DCO) 1 (19c)
10 The issue whether O2 or CO species are responsible for the diffusion-limited growth regime is currently disputed. In fact, this is not the only unresolved matter in a general attempt to understand the intricacies of the SiC oxidation mechanism. There is disturbing variation among reported oxidation rates on Si C. Furthermore, there is overwhelming experimental evidence to suggest that atmos pheric SiC oxidation is anisotropic in nature, i.e. the growth rate depends strongly on crystal orientation . For instance, 4H-SiC oxidizes almost an order of magnitude slow er on the (0001) Si-face compared to the (000 1) C-face in atmospheric furnaces, a phenomenon that is not predicted by current models [3,5,6]. The SiC oxidation mechanism is considerably more complicated than that of Si. Despite numerous studies, presen t understanding of the ex act kinetics of SiC oxidation remains only educated speculation. 1.3. SiO2/SiC structure formation and improvement efforts Numerous studies have focused on appl ying conventional oxidation methods to thermally form SiO2 films on SiC material. Although Si oxidation technology has been advanced and refined over the decades, there remains vast room for improvement in the growth of both SiC crystal material and oxi de layers with quality interfaces before practical MOSFET power devices can be achieved on SiC. Typical SiO2/SiC structures exhibit a broad range of electri cal defects. Much empirical work has been performed in an attempt to reduce the amount of defects, with limited success. 1.3.1. Conventional thermal oxidation of SiC Oxide films on SiC are chemically di fficult to form and require growth temperatures hundreds of degrees higher than Si in standard atmospheric furnace
11 processes. This is due partly to the fact that Si-C bonds require higher energy to break than Si-Si bonds (290 vs. 218 kJ /mol). In fact, typical Si C thermal oxidation rates are roughly an order of magnitude slower than thos e of Si at the same growth temperature. Furthermore, 4H-SiC has a sma ll lattice constant (3.08 ) co mpared to that of Si (5.43 ), a property which causes a large amount of compressive strain to develop at the interface during oxide growth. As a result, an abrupt SiO2/SiC interface is not energetically favorable. Instead, the interface consists of a wide defect-filled region transitioning between the SiC lattice and stoichiometric SiO2. As mentioned previously, a principle factor that complicates the formation of SiO2 films on SiC is the presence of carbon in the semiconductor mate rial, which ideally should be removed from the system by out-di ffusion of CO reaction products. However, it is generally believed that not all of the carbon products generated by the interface oxidation reaction actually out-diffuse and deso rb into the gas phase, but rather some residual carbon is incorporated into the interfacial transiti on region or even the oxide bulk. This residual carbon is a primary su spect for the large am ounts of defects and carrier traps which have thus far hindered SiO2/SiC technology. As a result of the strained lattice mism atch and residual carbon inherent in SiO2/4HSiC structures, the interfacial transition re gion likely contains silicon sub-oxide (SiOx x < 2) and silicon oxy-carbide (SixOyCz) components as well as other structural and carbonrelated defects, based on numerous interf ace studies [7-21]. The thickness of the SiO2/4H-SiC transition region is believed to be on the order of 50 , compared to an abrupt ~5 oxide interface on Si. Not surprisingly, oxidation of SiC produces significantly higher interface defect densities than those achieved on Si [7,8]. To date,
12 the precise structure and chem ical composition of the SiO2/4H-SiC transitional region is unresolved. Standard atmospheric oxidation of SiC typi cally involves an ambient of either dry oxygen (O2), water vapor (H2O) or pyrogenic steam (O2 + H2) at growth temperatures between 1000C and 1300C. Below 950C, no thermal oxide growth is believed to occur on SiC in dry or wet oxygen ambient unde r standard atmospheric conditions. SiC oxidation is generally followed by a re-oxidation annealing st ep in dry or wet oxygen at a temperature around 950C [7,8,22,23]. The low temp erature is chosen so that no further oxidation occurs at the interface, and no a dditional carbon-containi ng reaction products are generated as a result. Du ring re-oxidation, oxidant molecu les are suspected to further react with carbon in the inte rface or oxide and the resul ting oxy-carbide species undergo out-diffusion through the oxide, desorbing from the oxide surface to the gas phase. Reoxidation anneals at 950C may also allow th e oxide to relax and relieve compressive stress at the interface, especially consideri ng that the viscosity transition temperature of SiO2 is around 960C . Although some improvements in interface and oxide quality have been achieved by re-oxidation, residual carbonrelated defects and silicon sub-oxides still plague the defect-filled interfacial tr ansition region. Various postoxidation anneals [7,8,25-35] have been studied in an attempt to reduce interface trap densities (Dit). The anneals are typically performed at non-oxidizing temperatur es, similar to re-oxidation annealing, and have included a variety of ambients such as nitrous oxide (N2O), nitric oxide (NO), nitrogen (N2), ammonia (NH3), hydrogen (H2), and argon (Ar), with mixed results. Anneals in NO appear to have been the most effective to date in reducing or passifying
13 defects in the interfacial regi on, yielding relatively lower Dit values particularly near the conduction band edge in n-type 4H-SiC . The action of atomic nitrogen is believed to be responsible for these positive, however limited, results. Compositional studies have established that the nitridation anneals (excluding NH3) incorporate nitrogen in the interfacial region only, not in the oxide bulk. Despite concentrated research efforts, SiO2/SiC structures formed by conventiona l atmospheric oxidation processes and anneals contain high levels of electrically ac tive defects which are detrimental to device performance and have thus far stymied the great potential of SiC mate rials for field-effect power applications. 1.3.2. An alternative approach: remote plasma processing Plasma-assisted growth of oxide films at lo w pressures is an appealing alternative to standard atmospheric processes. The principa l advantage of such an approach is that a significant portion of the energy input required to drive a chemical process can be gained from electrons in a plasma discharge, inst ead of from thermal energy at the ambient process temperature. Since the production of reactive precursors, intermediates, or the final products are less dependent on thermal energy input, plasma-assisted processes can be performed at reduced temperatures which translates to production cost savings. The low pressures, reduced temperatures, and a dditional reactive species generated by plasma discharge imply different, and likely more co mplex, reaction kinetics than those in the traditional atmospheric oxidation model. Plasma-assisted processes at temperatures as low as 400C have been employed to grow oxide films on Si using oxygen radical s as the oxidizing species [36-38]. The reduction in growth temperature achieved is quite remarkable considering that the
14 thermal oxidation rate of Si is essentia lly zero below roughly 600C in atmospheric furnace processes. Oxide film growth has also been successf ully demonstrated on SiC processed in flowing afterglow of a remote plasma containing oxygen species [15,39-43]. High oxidation rates at low pressures have been obtained with growth temperatures hundreds of degrees below typical atmospheric processes. As an example (table 1.2), consider a 45 minute pyrogenic steam oxidation at 1100C wh ich produced only 180 of oxide on 4HSiC in an atmospheric furnace process  whereas 10 minutes of oxidative afterglow exposure at 1 Torr pressure and 850C grew 165 of oxide film (figure 3.1). These results illustrated some of the advantages of a plasma-assisted oxidation approach which achieved 4 higher growth rate despite o ccurring at 250C lower oxidation temperature and 3 orders of magnitude lower process pr essure (proportional to growth rate per equation 13). Table 1.2. Comparison of selected 4H-SiC thermal oxidation results from atmospheric p yrogenic stea m and remote plasma processes. The afterglow oxidation process demonstrated 4 higher growth rate de spited occurring at lower temperature an d p ressure. Oxidative ambient Pressure (Torr) Temperature (C) Time (min.) Thickness () Growth rate (/min.) pyrogenic steam 760 1100 45 180 4 oxygen afterglow 1 850 10 165 16.5 A study of oxidative removal of organic materials  discovered a low 0.5 eV activation energy (EA) for atomic oxygen reac ting with either polym eric or graphitic
15 carbon. Excited singlet molecular oxygen (O2*) also exhibited an EA of 0.5 eV for reaction with polymeric carbon. Interestingly, O2* reacted instantaneously at room temperature with graphitic ca rbon, yielding an immeasurable EA (essentially zero). An additional investigation of photo -resist stripping  found high etch rates of organic (i.e. containing C) polymers when exposed to a fl ow of oxygen microwave plasma afterglow. The oxygen radicals and excited species pr oduced in an oxygen plasma discharge serve a critical function in the afterglow oxidation of SiC. Na mely, they participate in the oxidizing reaction at the SiO2/SiC interfacial region by break ing Si-C bonds and forming Si-O bonds, which are added to the amorphous dielectric layer, and C-O products which out-diffuse through the oxide film. The prof iciency of plasma-generated reactive oxygen species in attacking and removing residual carb on at or near the interfacial region is suggested to be one of the main factors c ontributing to the high afterglow oxidation rates achieved on SiC. The aggressive action of O radicals and excited molecules toward carbon observed in the aforementioned studies gives solid support to this theory and further illustrates why a remote plasma proce ssing approach is particularly suited to face the challenge of growing quality SiO2/SiC film structures. In addition to using species generated by oxygen plasma discharge to grow oxide films, nitrogen radicals have also been used for treating oxides on both Si and SiC in the form of remote plasma nitridation a nneals. One investigation employed a N2 remote plasma treatment to nitride a thin SiO2 intermediate layer prior to HfO2 dielectric growth on Si, resulting in enhanced thermal stabilit y, resistance to oxygen diffusion during rapid thermal annealing, and lower leak age . In another study atomic N was used to form an ultrathin oxy-nitride dielectric film on Si with increased unifo rmity and reliability
16 . A study performed on SiC fo und that a remote plasma nitr idation anneal resulted in an improvement of interface properties of thin oxides on 4H and 6H-SiC . The positive influence of atomic N in removing or passifying interfacial defects on SiC warranted an investigation of the role of N radicals in the afterglow oxidation method. This was achieved by including a nitrogen-containing source ga s in the afterglow surface conditioning and oxidation media. 1.3.3. Surface conditioning Knowledge and control of th e SiC surface condition prior to oxidation is extremely important since the surface chemistry, morphol ogy, structure, and electrical state have a combined impact on oxidizing reac tions and formation of the SiO2/SiC interface. Several surface science studies have investigated the effects of remote plasma hydrogen and nitrogen treatments on 4H and 6H-SiC su rfaces at temperatur es between 200C and 750C and pressures of 0.21.0 Torr [48-50]. Bare untreated SiC surfac es were found to be typically terminated with a thin (~15 ) contamination layer containing Si-O, Si-F, CF, and C-C bonds. Oxygen and fluorine residuals were present following a standard RCA  wet cleaning procedure. This was prin cipally due to the inefficacy of hydrofluoric acid (HF) at terminating SiC surfaces with hydrogen, unlike the near-ideal hydrogen passivation of Si surfaces obtai ned by submersion in HF. An in-situ hydrogen cleaning performed by H2 remote rf plasma was found to sele ctively interact wi th residual oxygen which was removed as volatile H2O. Relatively clean, atomically flat and terraced SiC surfaces were achieved at lower temperat ures, and surface roughne ss was observed to increase with hydrogenation temperature. The ability to perform plasma-assisted
17 hydrogen cleaning at such low temperatures (200C 750C) was a huge advantage compared to temperatures around 1500C required for thermal H2 etching of 4H-SiC. SiC surfaces treated with atomic nitrogen generated by an N2 remote rf plasma were modified due to incorporati on of nitrogen into the SiC surface region, forming an ultrathin nitrided (SiNx) overlayer. The nitridation pr ocess proceeds, similarly to oxidation, with nitrogen adsorption and su rface coverage, in-diffusion through any existing nitride layer, reaction at the buried SiNx/SiC interface, out-diffusion of volatile CNx products, and desorption to the gas phase. Th e C site is preferred for N substitution, and this N-for-C exchange resu lts in Si-N being the stable bonding configuration at the interface. The chemisorption and reaction of nitrogen at the SiC surface induces a charge transfer between the adsorbate and semic onductor which alters the intrinsic surface charge due to structural defects and impurities. This results in a modification of the surface state density, band-bending, and electronic properties of the nitrided SiC surface, in addition to the chemical and structur al alterations caused by nitride overlayer formation. The findings of these SiC surf ace studies motivated the application of a combination of H2 and N2 remote plasma treatments of SiC surfaces prior to afterglow oxidation, and an examination of the influen ce of remote plasma surface conditioning on subsequent SiO2/SiC interface formation and film growth. 1.4. Capacitance-voltage characterizatio n of oxide/semiconductor structures Capacitance-voltage (C-V) characteristics reveal much information regarding the quality of dielectric films and interfaces on semiconductors. C-V measurements are the standard means of metrology by which to ev aluate oxide electrical performance and extract various parameters of merit. Severa l methods of obtaining CV characteristics are
18 possible, but all utilize the same basic principle. Every C-V measurement technique is capable of applying a bias voltage acr oss the oxide/semiconductor structure and extracting the total structure capac itance as a function of potential. 1.4.1. Capacitance-voltage measurement fundamentals The total capacitance (CTOT) of an oxide/semiconductor structure is the series combination of the dielectric capacitance (COX) and the capacitance due to any space charge region in the semiconductor (CSC). Interface traps can contribute a parasitic capacitance (CIT) in parallel with CSC. The equivalent total capacitance of the oxide/ semiconductor structure is expressed as CTOT 1 = COX 1 + (CSC + CIT) 1 (20) The total applied voltage (VB) used to electrically bias the structure under test is distributed between a potential drop acr oss the dielectric insulating layer (VOX), and a surface potential barrier asso ciated with the semiconduc tor surface and space-charge region (VSB). Figure 1.3 portrays a schematic repres entation of this simple electrical model for an oxide/semiconductor structure. For the purpose of illustration, consider CTOT of an oxide film grown on negativelydoped semiconductor material, with electrons serv ing as majority carriers. With a large positive applied bias (VB >> 0), a positive electric field develops across the oxide (VOX > 0), and electrons are accumulated at th e semiconductor surface inducing a positive surface barrier (VSB > 0). As a result, electron energy bands are bent downward in the semiconductor near the interface (figure 1.4a). Assuming an ideal structure without interface traps, the measured capacitance in accumulation (CACC) will be that of the oxide
19 Figure 1.3. Electrical model fo r the capacitances and potentials associated with an oxide/semiconductor structure. layer alone since no space-charge region exists in the semiconductor. With the measured capacitance normalized per unit area, CACC = COX = 0r /tOX (21) where 0 is the permittivity of vacuum, r is the relative permittivity of the dielectric, and tOX is the oxide film thickness. For a large negative applied bias (VB << 0), a negative electric field develops across the dielectric (VOX < 0), and electrons are repelled from the interface into the semiconductor, which become s depleted of majority carriers in the surface region. The resulting negative surface barrier (VSB < 0) corresponds to electron energy bands bending upward in the semic onductor near the inte rface (figure 1.4b). At the transition between accumulation and depl etion of majority carriers near the semiconductor surface is a state termed "fla t-band" because no bending occurs in the electron energy bands (figure 1.4c). Unde r ideal assumptions (i.e. without any contribution from charged defects), the a pplied bias at the flat-band condition (VFB) is equal to a relatively sma ll workfunction difference ( MS) between the semiconductor and the probe or gate metal used in the particular C-V measurement.
20 (a) (b) (c) Figure 1.4. Electron energy band diagrams repr esenting accumulation (a), depletion (b), and flat-band condition (c) of an oxide/s emiconductor structure under applied bias. VFB = MS (ideal) (22) Under the depletion condition, a space charge region exists in the semiconductor whose added capacitance CSC in series with COX results in a lower measured CTOT. As VB becomes increasingly negativ e, the semiconductor depletion region widens and the energy bands are bent further upward. The decreasing CSC associated with a widening space-charge region causes CTOT to continually decrease towa rd lower capacitance values as the the semiconductor is further depleted (CDEP). A typical C-V response of an oxide on n-type semiconductor material is depicted in figure 1.5, illustrating the parameters VFB, CFB, CACC, and CDEP. When considering a non-ide alized oxide structure with electrically active def ects present, the general effects on a C-V measurement are basically twofold. First, any interface traps will cause a stretch-out of the C-V curve around the flat-band as the structure is swept between the extremes of accumulation and depletion conditions. This stretch-out occurs because some charge is trapped in the process of filling or emptying interface states distribut ed at various energy levels throughout the
21 band-gap, rather than further accumulating or depleting the semiconductor (depending on sweep direction). Second, any oxide trappe d charge or fixed charge will induce a horizontal translation of the C-V curve al ong the voltage axis, e ffectively shifting VFB from its theoretical value. A certain amount of applied bias is re quired to supply the charge needed to compensate for the ch arged oxide defects and achieve flat-band condition. The flat-band voltage shift ( VFB) due to net oxide charge (QTOT) is related to COX : VFB = MS + VFB = MS + QTOT /COX (23) The effects of DIT stretch-out and a flat-band shift due to negative QTOT on n-type C-V curves are illustrated in figure 1.5. Figure 1.5. Oxide C-V characteristic exampl e on n-type semiconductor, illustrating the effects of C-V stretch-ou t and flat-band shifting.
22 The saturated value of cap acitance measured with the semiconductor surface strongly accumulated is typically used to extract the electric al equivalent oxide thickness (EOT) of a dielectric film. EOT is the equivalent thickness of SiO2 which would yield a measured CACC. Although permittivities of amorphous SiO2 can fall in the range r = 3.7 5.1, a general value of r = 3.9 is assumed for stoichiometric SiO2. Hence, EOT is extracted from CACC by re-arranging (21) as follows: EOT = 3.9 0 /CACC (24) 1.4.2. Contact vs. non-contact metrology Any method for measuring CV oxide characteristics requires a means to apply a biasing potential across the oxide/semiconducto r structure. This is conventionally achieved by the deposition of metal (e.g. Al) or poly-crystalline silicon (poly-Si) conducting films on the oxide surface in orde r to fabricate gate contacts for MOS capacitor test structures (figur e 1.6a) or MOSFET devices. A voltage applied through an electrical probe contacting the gate serves to bias the oxide/semiconductor structure when the substrate is grounded. Determination of capacitance is possi ble using appropriate measurements of voltage, current, and/or impedance parameters. The requirement of device fabrication fo r contact C-V measurements means that these techniques are invasive, destructive, and limit the possibility of continued processing following determination of oxide qualit y. Test wafers are usually recycled or discarded following oxide metrology, which a dds an enormous burden of production cost and time. The additional process time is par ticularly cumbersome in the research and development stage, when quick feedback is crucial to the optimiza tion of novel material processes.
23 (a) (b) (c) Figure 1.6. Alternative techni ques of generating a bias potential across an oxide / semiconductor structure: MOS contact (a), Hg -probe (b), and corona ion deposition (c). A liquid mercury (Hg) probe is capable of providing a temporary electrical contact to an oxide surface (figure 1.6b). Thus, Hg-probe C-V measurements allow oxide characterization without test structure fabr ication. However, the Hg-probe technique leaves residual Hg metal contamination on the oxide surface following measurement. Due to the problem of mercury contaminat ion, Hg-probe measurements should be classified as destructive in nature. Subs equent fabrication steps cannot be performed without jeopardizing oxide and process cleanliness. An in-line metrology technique based on the deposition of corona ions on a surface and monitoring of the structure potentia l with a non-contacting probe provides many advantages over contact measurement met hods. The corona-Kelvin metrology (C-KM) approach [42,43,52-59] is fast, non-destructiv e, and can be applied to obtain oxide electrical characteristics at selected point s in a sequence of processes. The C-KM
24 technique is based on the use of charged ions (CO3 or H3O+(H2O)n ) generated by corona discharge in air, and deposited on a sample surf ace (figure 1.6c). Th e corona ions energy is reduced by ambient collisions such that they are non-damaging when arriving at the sample surface , and can be completely removed without residual contamination by rinsing in de-ionized (DI) water. The de posited corona surface charge accomplishes electrical biasing of an oxi de/semiconductor structure, analogous to the gate contact of MOS C-V measurements. Monitoring of the deposited charge, combined with potential determination by a non-contacting voltage probe, produces charge and voltage information useful for extracting many semi conductor and oxide parameters of merit, including the capacitance respons e. Thus, the C-KM method is a truly non-invasive and non-damaging technique capable of in-line electrical mon itoring of dielectrics and semiconductors, and a valuable tool for obta ining quick C-V characteristics of oxide films on semiconductors. The focus of this work is the growth and characterization of oxide thin films on 4HSiC using the unique capabilities of remote plasma afterglo w processing technology and in-line C-KM, with emphasis on the role of semiconductor surface conditioning prior to oxidation. A general attempt has been made to apply the oxidation mechanism knowledge and process technology that have been developed successfully for the SiO2/Si system to the oxidation of the WBG comp ound semiconductor SiC, with less than satisfactory results. Although SiC is an attractive candidate for power and other applications, most importantly the 4H polytype, numerous defects exist in SiO2/4H-SiC structures, and the precise inte rfacial chemistry, structure, and origin of defects is not presently understood. Remote plasma afterglo w processing at low pressure and reduced
25 temperatures offers an advantageous, flexib le and effective altern ative to conventional atmospheric furnace processes for growing oxide films on SiC. High growth rates at reduced temperatures are achievable, a nd the possibility exists for sequential in-situ processing steps, including surface treatment prior to oxidation which could have a significant impact on the SiO2/SiC interface formation and oxide growth process. A noncontact corona-Kelvin metrology technique may be used as a quick, non-destructive means for performing electrical characterizat ion of semiconductors and experimentally grown oxide films, and to evaluate structur e quality at various points in a process sequence due to its non-invasive nature.
26 Chapter 2. Experimental Approach The experimental methods used in this work included 4H-SiC surface conditioning and growth of dielectr ic films and utilized a remote pl asma-assisted seque ntial processing approach in an afterglow chem ical reactor. The resulting oxide/SiC structures were characterized using non-contac t corona-Kelvin metrology capacitance data to evaluate key oxide parameters such as film thickne ss, flat-band voltage, uniformity, and trapped charge. In addition, C-KM depletion voltage transien ts and X-ray photoelectron spectroscopy (XPS) analysis of the c onditioning stage preceding oxidation yielded electrical and chemical information regardi ng the effects of remote-plasma treatment on the SiC surface. The compiled results were used to investigate the effects of strategically selected process variations, and gain a be tter understanding of the afterglow surface treatment and oxidation of SiC. 2.1. Afterglow chemical processing As introduced previously, th e use of an afterglow chemical reactor for remote plasma processing offered an advantageous alternative to a conve ntional atmospheric oxidation furnace, and facilitated oxide film gr owth at temperatures hundreds of degrees lower than possible with thermal energy input alone. The added flexibility of sequential in-situ processing capability combined with a wide variety of possible chemical ambients provided important tools for controlling and manipulating the ini tial surface condition, interface formation, film growth and post-oxidation environment.
27 2.1.1. Afterglow apparatus description and operation The afterglow chemical reacto r  used in this work operated as a 1 Torr vacuum furnace system with a flowing reactive ambien t including excited and atomic gas species generated by microwave plasma discharge upstream from the heated substrates. A schematic representation of the afterg low apparatus is shown in figure 2.1. Figure 2.1. Schematic diagram of the re mote plasma afterglow apparatus. A resistive heating furnace surrounds a 6-in. diameter fused silica tube and maintained the substrates at a desired temper ature, up to 1200C. The temperature inside the furnace zone was monitored at multiple points simultaneously using thermocouples spaced along a profile rod. Substrate 4H-SiC wafers (typically 3-in. diameter) were positioned vertically by slots in a quartz load ing boat contained in the growth zone. The substrate or wafer area was perpendicular to gas flow, with the (0001) Si-face directed downstream by convention. A protocol of su rrounding the SiC substrates with additional
28 Si shielding wafers on either side was en forced to reduce turbulence and increase uniformity of film growth. A rotary-vane mechanical pump backing a series roots blower evacuated the growth chamber to 1 Torr total pressure with a comb ined flow of source gases approximately 4 standard liter atmospheres per minute (slam). This enabled a high mass flux of reactants in the wafer region and a short transit time of neutral species fr om generation in the microwave plasma to the furnace zone. Preci se mixtures of desired source gases were generated using an array of ma ss flow controllers. The sy stem pressure was monitored both by a capacitance monometer and thermoc ouple on the exhaust side of the furnace zone, and controlled by adjusting an exhaust valve which altered the rate of pumping. Figure 2.2. Photographic image of the remote plasma apparatus furnace zone during operation, with visible chemo-lumines cence of afterglow excited species. Figure 2.2 contains a photographic image of the furnace porti on of the afterglow apparatus during operation. The visible chem o-luminescence eminating from the fusedsilica enclosure may be observed entering a nd exiting the furnace growth zone. This luminescence was caused by photon emission proc esses associated w ith electrons in excited states returning to lower energy orbitals. As witnessed in the image, the lifetimes
29 of excited afterglow species were much l onger than required to transport downstream from the plasma discharge and react with substrates in the furnace zone. The core of the afte rglow reactor was a multi-mode excitation cavity surrounding a quartz tube containing flowing source gases upstream from the furnace zone. A remote continuous-wave 2.45 GHz microwav e source acted to drive th e excitation cavity through a series of waveguide sections and an injecting rod inserted an adjustable distance into the cavity interior. The microwave excitation established inside th e cavity generated a plasma discharge in the flowing gaseous species. A forward power around 1 kW was typically required to maintain a stable plas ma state. The forward and reverse power and coupling between the microwave source and cavity were tuneable by varying the highvoltage supply power, waveguide tuning st ubs, and cavity injec tion rod distance. Figure 2.3. Photographic images of the remote plasma apparatus microwave cavity, depicting the plasma discharg e and afterglow during operation. The microwave plasma discharge created a rich collection of ex cited molecular and atomic gaseous species, ions, electrons a nd photons. Charged species (i.e. ions and
30 electrons) were contained inside the excita tion cavity using an RF choke formed by a grounded metallic ring at the cav ity exit. A right-angle bend in the quartz tubing between the cavity and furnace prevented photon radiat ion, particularly damaging ultra-violet (UV) frequencies, from entering the grow th zone. Some residual amount of UV radiation, however, did travel down the walls of the quartz tubing towa rd the growth zone and was blocked by a fitting that joins the quart z plasma tube and furnace tube. Thus, the only plasma-generated species which were permitted to reach the growth zone were neutral molecules and radicals, some being in excited electronic states. Photographic images of the microwave cavity, plasma disc harge, and afterglow are shown in figure 2.3. 2.1.2. Dielectric growth by remote plasma sequential processing The ability to apply alternate chemistries and temperatures in sequential in-situ processing steps make the afterglow method a powerful tool for growth and improvement of oxide/4H-SiC structures. The afterglow chem ical reactor has the capability to provide pre-oxidation surface conditioning, oxidative growth, and post-oxidation annealing environments in a continuous process. This flexibility does not exist in conventional thermal oxidation methods using standard chemistry. Typical afterglow oxidation pro cesses used in this work (f igure 2.4) consisted of the following general sequence schedule example: 1) load wafers under N2 flow at temperature Tload (600C), 2) surface conditioning step at relatively low temperature Tcond (600C), 3) oxidation step at a higher temperature Toxid (850C), 4) inert annealing step at temperature Tanneal (1000C), 5) unload wafers under N2 flow at a temperature Tunload (600C).
31 Furnace temperature ramps were usually performe d with an inert ambient flow (e.g. Ar). Please note that steps 2-4 are optional, and the temperature, duration, and chemistry of each step may be varied as desired. Figure 2.4. Example of a general afterglow pr ocess flow temperatur e profile, including p re-oxidation surface treatment, oxide gr owth, and post-oxidation annealing steps. Prior to loading in the afterglow furnace, wafers were subjected to a rigorous wet cleaning procedure including piranha (2:1 H2SO4:H2O2) and dilute HCl rough cleaning, followed by a standard RCA  fine cleaning process usin g basic and acidic solutions of hydrogen peroxide to remove particles, organics, and metals from the semiconductor surface. The wet cleaning protocol ended w ith submersion in dilute HF to ensure removal of any oxide layer formed during ch emical cleaning. However, this step was known to leave residual flurine and oxygen cont aminants on the SiC surface , a fact to consider during subse quent afterglow processing. The primary surface treatment used in this work was chemistry from a plasma discharge of 5% H2 in balance N2, a source gas mixture labeled with the common name
32 "forming gas" (FG). Other conditioning trea tments included the afterglow ambient from pure N2 remote plasma, as well as a non-excited N2:O2 7:1 gas mixture. Treatment temperatures of 400C 800C and durations of 2.5 20 min. were investigated as variations of the standard treatmen t which occurred at 600C for 20 min. The surface conditioning step attempted to take advantage of the cleaning and nitriding effects of radicals and excited species. Reactive hy drogen along with N2* and N radicals were suspected to be available for reaction with the SiC su rface. The roles of hydrogen and nitrogen in a remote plasma treat ment could be crucial to controlling the SiC surface in preparation for subsequent oxidizing processes. Oxidative chemistries consisted of a remote plasma discharge of O2, mixed with other additive gases. Previous studies have shown that the addition of a small percentage of nitrous oxide (N2O) to an oxygen discharge resulted in an increase in atomic oxygen production and plasma stability . Hence, the use of an excited (O2:N2O)* 4:0.3 oxidation media induced a signifi cant increase in growth rate compared to remote plasma processes with pure O2 discharge . An additional growth rate increase was observed when adding FG to the (O2:N2O)* plasma. The standard oxida tion chemistry used in this work was (O2:N2O:FG)* 3:0.23:0.5. Oxidation temperatures of 600C 850C and durations of 10 90 min. were investigated. This work also examined the impact of post-oxidation anneals on resulting oxide quality, as a function of surface condition prio r to oxidation. A typical annealing step utilized non-excited Ar gas, absent any ener gy from a remote plasma discharge, at a high temperature such as 1000C for 60 min. durati on. High-temperature a nneals in inert Ar ambient were used to analyze the stabil ity of afterglow-fo rmed oxides to high-
33 temperature stresses which are required fo r subsequent fabrication steps (e.g. thermal annealing of deposited poly-Si gate film or silicide formation). Following every oxidation process, the wa fers were unloaded from the reactor under N2 flow, and allowed to cool. An HF va por etch was typically used to remove oxide from the wafer backside (i.e. C-face) to ensure electrical contact between the substrate and measurement chuck. The wafers were rinsed in DI water and dried under N2 flow following backside etching. This pro cedure left the frontside oxide surface in an assumedly repeatable and constant condition following every growth experiment. This was an important factor related to subse quent non-contact C-KM metrology, which relies on precise control of surface charge. 2.2. Non-contact corona-Kelvin metrology Electrical characterization of processes and oxide/SiC struct ures fabricated in this work included capacitance and voltage tran sient measurements performed by the noncontact corona-Kelvin method. The C-KM technique provided quick, non-invasive, electrical feedback by combining corona ion deposition and non-contact potential monitoring. This important in-line metr ology technique, now common in the Si integrated circuit industry, has been adopted to facilitate m easurement of SiC materials. 2.2.1. Corona-Kelvin tool operati on and basis of measurement The modified Semiconductor Diagnostics, In c. Film Analysis and Substrate Testing (FAaST) 230  tool used in this work was capable of performing a variety of semiconductor and dielectric measurements. As previously introduced, the tool achieved electrical biasing of the measured structures by depositing non-damaging ions in the form
34 of carbonate or hydrated hydronium (CO3 or H3O+(H2O)n ) on the sample surface from a high-voltage ( 5-10 kV) corona discharge in air. The stru cture potential was monitored in a non-contact fashion by a contac t potential difference (CPD) probe. The non-contact voltage probe applied the Kelvin met hod  of acquiring the CPD between a reference electrode a nd the grounded sample substrate (VCPD). Using a Monroe configuration , mech anically vibrating shutters acted to periodically vary the capacitance between the electrode and substrate (C0). The variable capacitance induced an alternating current (Jac) in the electrode which was proportional to VCPD and is expressed as Jac = (VDC + VCPD) C0sin( t) (25) where is the frequency of shutter vibration, and VDC is an external bias voltage applied to the reference electrode. Measurement electronics adjusted VDC to achieve the null current condition (Jac = 0) in which case the sum term on the rhs of equation (25) was zero, and hence the applied bias was equal to VCPD: VCPD = VDC (26) Non-contact measurement of an oxide /semiconductor struct ure yielded a VCPD approximated as: VCPD = MS + VOX + VSB (27) where, as previously, MS is the workfunction differenc e between the electrode and substrate, VOX is the oxide voltage, and VSB is the surface potential barrier. Note that in the absence of an oxide film, the measured VCPD should equal the surface barrier VSB, offset by the small constant MS (< 1V). A schematic illustration of the CPD probe apparatus is shown in figure 2.5.
35 Figure 2.5. Non-contact CPD probe schematic. In the FAaST tool, the test sample was held by vacuum on a motorized chuck, enabling multiple point measurement and wafe r mapping. The ion source and CPD probe were positioned on a shuttle mechanism whic h facilitated the charge and measurement cycle. After the corona source deposited a precise and monitored dose of corona charge ( QC), the adjacent CPD probe shifted over the same surface site to allow immediate VCPD measurement. The first VCPD reading occurred less than 1 sec. after the corona charge deposition, and the VCPD transient was monitored for a specified time interval following the initial reading. The VCPD voltage transient gave useful information regarding carrier transport and charge compensation proce sses when acquired over a significant time interval following a large corona charge dose pulsing the semic onductor into strong
36 depletion. This is the topic of discussion in chapter 4 where the electr ical state of the SiC surface following conditioning treatment s was examined using depletion VCPD transient decays. For the purposes of oxide film characterization, capacitance data were obtained using an alternating sequence of incremental corona applications and VCPD determination. After each corona charge dose ( QC) was deposited, the VCPD was monitored for 2.5 sec. Repetition of the charge and measure cycle result ed in a set of voltage transients such as those depicted in figure 2.6 obtai ned on an oxidized p-type SiC epi-layer. In the example measurement, the structure initially had a net negative corona charge density (QC < 0) on the surface, and was swept from accumulation to depletion condition as repeated doses of positive corona charge were deposited on the oxide surface. Figure 2.6. Typical VCPD data obtained during corona-K elvin metrology of an oxidize d p -type SiC epi-layer. Repeated doses of positive corona char ge were used to sweep the structure from accumulation to depletion.
37 When combined with the precise monitoring of corona charge doses, the voltage data were used to generate the V-Q depende nce such as that shown in figure 2.7, where the structure VCPD is given as a function of total surface corona charge density QC. Each data point corresponds to a single positive corona charge dose as the oxide/p-SiC structure was swept from accumulation to depl etion. The V-Q relationship lies at the foundation of the corona-Kelvin metrology technique. Figure 2.7. Typical V-Q response obtained during corona-Kelvin metrology of an oxidized p-type SiC epi-layer. Repeated dos es of positive corona charge were used to sweep the structure from accumulation to depletion. The structure capacitance (C) was extracte d from the relationship between the QC increment and the change in VCPD: C = QC / VCPD (28) which is simply the inverse slope of the V-Q curve. A plot of C vs. VCPD such as that in figure 2.8 revealed the capacitance behavior of the oxide/semiconductor structure over a
38 voltage range depending on th e polarity of deposited QC and number of charge doses. The C-V response, as extracted from the Vt and V-Q data, was us ed for electrical evaluation of oxide film parameters. Figure 2.8. Typical C-V characteristic ex tracted from corona -Kelvin metrology V-Q data on an oxidized p-type SiC epi-layer. Repeated doses of positive corona charge were used to sweep the structure fr om accumulation to depletion. The corona-Kelvin tool, as it was confi gured for SiC measurements, possessed the capability to perform non-contact C-V measurem ents with the sample either in ambient darkness or under strong illumination provided by a UV ( = 370 nm) diode. The value of CACC measured on an illuminated oxide/SiC samp le was used to extract the electrical film thickness. The UV diode generated photon s with energy of 3.4 eV just larger than the 4H-SiC band-gap (3.26 eV). The oxide was transparent to the photons, which were absorbed in the SiC and generated electron -hole pairs, thus el iminating any residual space-charge in the semiconductor. This ensu red the oxide film was the sole contributor
39 to the measured structure capacitance, with the oxide potential remaining unaffected by the illumination process. 2.2.2. Oxide/4H-SiC structures: typical noncontact capacitance-voltage behavior Figure 2.9 contains typical non -contact C-V results demonstrating the electrical behavior of afterglow oxide thin films on 4H-SiC. The curves shown were measured sequentially at a single site on an oxidized p-type epi-layer. Prior to any corona charge application, the structure was in a slightly depleted condition, with a small initial VCPD less than a volt in magnitude, implying a sm all amount of positive oxide charge. The polarity of QC for the first sweep was chosen to dr ive the structure toward accumulation, negative for the example p-type semiconductor. The initial sweep into accumulation was a shallow sloped ramp, as many surface states and near-interface traps were charged. Following the initial sweep, the SiC was slig htly accumulated, and a second sweep of opposite polarity was performed to bring the structure into deplet ion (positive for p-type). The process of alternating negative and posit ive sweeps was repeated several times to obtain a series of C-V curves, each one stressing the struct ure deeper into accumulation. The subsequent sweeps directed toward accumulation did not display the dramatic stretch-out that was visible dur ing the initial sweep, indicating a satiation of the majority of interface traps. Also, the VFB of each sweep toward deple tion was generally larger in magnitude than the sweep toward accumulation that preceded it. Furthermore, as the structure was stressed deeper and deeper into accumulation at each subsequent C-V measurement, the VFB increased in magnitude. The VFB saturated at a maximum magnitude when the structure was st ressed to very strong accumulation.
40 Figure 2.9. Typical non-contact C-V curv es obtained sequentially at a single measurement site on oxidized p-type 4H-S iC. The measurement order of positive an d negative sweeps is indicated in the legend. All except the last sweep (illuminated) were measured in darkness. This behavior of permanent VFB shifting due to accumulation stress was believed to be caused by charging of transitional or near -interface traps, also called border traps. These border traps were probably related to th e wide, defect-filled transition region at the interface, but they did not behave as typi cal interface traps whic h should fill and empty during each alternating sweep in to accumulation and depletion, causing stretch-out of the C-V curve. On the contrary, these border tr aps, once filled with majority carriers during accumulation, retained their charge even when the semiconductor returned to depletion, causing a permanent VFB shift toward accumulation (i.e. larger magnitude VFB). It was possible to recover the majority of border trap s to their unfilled states by sweeping into depletion with the structure under strong UV ill umination. In other words, after a C-V
41 measurement in light, the struct ure behaved similarly to its initial condition prior to any corona charging. The measured CACC value with the structure under strong illumination was used to extract the electrical thickness value. This was more crucial on p-type than on n-type samples, because both the p-type and n-type 4H-SiC material used in this work were epitaxial films grown on heavily doped n-type bulk substrates. Considering the p-type epitaxial layer, a parasitic series capacitance existed due to space charge at the buried p/n+ epi/bulk junction, which effectively lowered the measured total structure capacitance. Under illumination, the buried sp ace charge region was eliminated and the measured capacitance rose to its expected value of COX, as witnessed in the figure. In addition to detailed electrical measurements at single points, simple multiple-point C-V scans were performed both in dark and illuminated ambi ent conditions in order to investigate the uniformity of EOT and VFB parameters across the wafer. This work used various figures of meri t, obtained primarily by non-contact coronaKelvin metrology, to analyze the effects of sp ecific process variations in remote plasma surface conditioning and oxidation of SiC materi al. The role of nitrogen and hydrogen in pre-oxidation surface treatments were emphasized, with the aid of VCPD depletion voltage transient data and XPS analysis of conditioned surfaces. The effect of annealing as a test of stability to high temperature stresses was also considered. Metrology parameters including EOT, VFB, flat-band shifting, uniformity, temp erature stability, surface barrier decay, and surface chemistry were used as t ools for elucidating the mechanisms involved in the plasma-assisted surface conditioning, ox ide interface formation, and film growth on 4H-SiC.
42 Chapter 3. Corona-Kelvin Capacita nce Metrology of Afterglow Oxide Films A variety of experiments have been pe rformed to investig ate oxide/4H-SiC structures formed by the afterglow method. Non-contact C-V charact eristics obtained by corona-Kelvin metrology were used to measure the impact of process variations on growth rate, charge trapping, uni formity, and stability to temp erature stress. Experiments were designed to focus on the role of nitrogen and hydrogen in the remote plasma ambient during surface treatment prior to oxidation. Results of oxide electrical characterization are discussed in this chapte r. A set of supplemental surface analysis experiments analyzed the conditioned SiC surfaces directly, without any subsequent oxide film growth. Surface metrology, addre ssed in chapters 4 and 5, consisted of noncontact C-KM depletion voltage transients fo r electrical evaluation of the surfaces, while XPS analysis served to provide additional chemical information. Thus the combined results from both oxide and semiconductor C-KM electrical characterization, as well as XPS chemical analysis, were drawn upon to discuss the impact of remote (N2:H2)* plasma surface conditioning on the afterglow ox idation mechanism, interfacial reactions, chemistry and structure regarding the important oxide/4H-SiC system. A remote plasma processing approach, using selected variations of surface preparation, oxidation, and annealing steps, wa s used to prepare oxide thin films on nand p-type epitaxial Si-face (0001) 4H-SiC 8 off-axis 3-in. wafers. All oxidations were performed in the remote plasma chemical re actor at 1 Torr total pressure as heated
43 substrates were exposed to a flow of rich oxidizing (O2:N2O:FG)* afterglow ambient described previously. The oxi dation intervals used were 10 90 min. at temperatures 600C 850C, with resulting EOTs of 50 500 . Non-contact C-KM capacitance measurements were performed at multiple site s on each wafer to extract oxide electrical parameters and examine uniformity. 3.1. Oxidation time and temperature results vs. surface conditioning In order to highlight the importance of initial surface chemistry, structure, and charge on subsequent oxide interface formation and growth mechanism, experiments were performed which incorporated remote plasma conditioning steps prior to afterglow oxide film growth. All surface treatment step s occurred at 600 C for 20 min. duration in microwave-excited forming gas plasma (F G)* afterglow ambient containing reactive hydrogen and nitrogen species. For comparis on, some processes included pre-oxidation exposures to the standard low-temperature ramp media, a non-excited N2:O2 7:1 ambient, in order to simulate the same process temperature profile. Figure 3.1 depicts electrical thickness values extracted from non-contact C-V curves obtained at multiple sites on both nand p-type 4H-SiC epitaxial layers oxidized for various growth intervals at 850C by afterglow of (O2:N2O:FG)* plasma discharge. Thickness values, determined by EOT extraction from CACC, were averaged over multiple points on each wafer. As witnessed in the figure, (FG)*-treated surfaces resulted in a growth rate increase, roughly 20%, du ring subsequent oxide film formation. Clearly, the surface conditioning step in (F G)* afterglow has a significant influence on the initial oxide/4H-SiC interfacial reactions and afterglow oxidation mechanism. As
44 Figure 3.1. Electrical thicknesses of afterglow oxide films grown in (O2:N2O:FG)* afterglow ambient for various time interval s on 4H-SiC substrates at 850C, some o f which were subjected to (FG) surface conditioning at 600C. EOT is expressed as the average from multiple measurement sites. The oxide growth rate increased by roughly 20% when the SiC surface was remote plasma-treated prior to oxidation. previously introduced, it was possible that the aggressive action of hydrogen species in the (FG)* afterglow media were responsib le for removing residual contaminants following wet chemical cleaning. Hydrogen coul d also act to create a higher degree of order and passivation of the SiC surface, bette r suited for subsequent oxidizing reactions and interface formation. It was also possible that the excited and atomic nitrogen species within the (FG)* remote plasma ambient aided in formation of a thin nitrided overlayer or nitrogen-rich surface region which altered, chem ically, mechanically, and electrically, the initial stages of afterglow oxidation. Nitrogen could act to remove or passivate C clusters in the surface layers, effec tively reducing the presence of C-related surface defects and creating a Si-rich surf ace region. The removal of car bon from the surface region might
45 allow initial interfacia l reactions to proceed at a faster rate, since the high-energy C-Si bond is an important factor contributing to the slow oxide growth rate on SiC. Figure 3.2. Thickness uniformity of afterg low oxide films grown for various time intervals on 4H-SiC substrates at 850C, so me of which were subjected to (FG)* surface conditioning at 600C. The EOT standard deviation is expre ssed as a percentage of the average of EOT values from multiple meas urement sites. EOT deviation decreased roughly five-fold when the SiC surface was re mote plasma-treated prior to oxidation. Another significant influence of the (FG)* surface conditioning step on the subsequent oxide interface formation and el ectrical properties was a stark improvement of oxide uniformity, both of EOT and VFB parameters. The SiC surfaces which were subjected to remote (FG)* plasma ambient prior to oxidation exhibited a much smaller degree of variation in their C-V characteristics obtained at multiple sites across the wafer. An example of this uniformity effect is gi ven in figure 3.2, where the electrical thickness standard deviation is expressed as a percenta ge of the average EOT value from multiple measurement sites. The oxide /4H-SiC structures formed w ith (FG)* surface conditioning
46 prior to oxidation demonstrated a roughly fi ve-fold reduction in thickness deviation. The observed uniformity improvement was sugg ested to result from the nitrogen and hydrogen reactive species helpi ng to create a clean, smooth, and passivated SiC surface prior to oxidizing interfacial reactions, similar to previous studies which conditioned the surface prior to epitaxia l growth . These results il lustrated the positive effects of the pre-oxidation (FG)* afterglow treatment on preparing the SiC surface for uniform oxide film formation. Figure 3.3. EOTs of afterglow oxide films grown for 15 min. on 4H-SiC substrates at temperatures between 600C and 800C, some of which were subjected to (FG)* surface conditioning at 600C. EOT is expressed as the average over multiple measurement sites. The oxide growth rate increased by roughly 10-15% when the SiC surfaces were conditioned via (FG)* afte rglow prior to oxidation. Thus, pre-oxidation surface conditioning in (N2:H2)* afterglow facilitated an increase in oxide growth rate and improvement in film uniformity for various oxidation time intervals and film thicknesses at 850C. Additional oxidations at temperatures other
47 than 850C were required to examine the dependence of surface conditioning effects on oxidation temperature. A series of oxidations at temperatures 600C 800C were performed using (O2:N2O:FG)* remote plasma ambient fo r growth intervals of 15 min. Some 4H-SiC surfaces were conditioned by (F G)* afterglow for 20 min. at 600C prior to film growth. As before, EOT values were extracted from non-contact C-KM capacitance curves, and averaged over multiple wafer sites. The oxide thickness results shown in figure 3.3 corresponded to approxima tely a 10-15% growth rate increase on both nand p-type 4H-SiC substrates wh en the surfaces were conditioned by (FG)* remote-plasma before oxidation. Over the temperature range examined, the (FG)* surface conditioning effect on oxide film thickness was generally observed as a small increase in EOT. The magnitude of growth rate increase was observed to diminish at lower oxidation temperatures however, most noticeably at 600C. This was somewhat misleading since the resulting oxide films grow n for only 15 min. at this low temperature were extremely thin, around 50 in thickness. Hence, a roughly 10% growth rate increase translated to only a 5 difference in EOT. A more detailed series of experiments were performed to examine the (FG)* afterglow conditioning effects at the relatively low oxidation temperature of 600C. Films were grown during oxidation intervals of 15 60 min. in the standard (O2:N2O:FG)* remote-plasma chemistry. Some of the p-type 4H-SiC subtrates were conditioned for 20 min. in (N2:H2)* afterglow at 600C preceding the oxidation stage. Average EOTs were extracted from C-KM capacitance measuremen ts and are depicted in figure 3.4. The thickness results observed at 600C were consis tent with previous findings, although not as pronounced as the conditioning effects s een at 850C oxidation temperatures. The
48 (FG)* afterglow conditioned 4H-SiC surfaces e xhibited an increased film growth rate, roughly 10%, during subsequent oxidation intervals at 600C. Figure 3.4. Electrical thicknesses of afterglow oxide films grown in (O2:N2O:FG)* afterglow ambient for various time intervals on p-type 4H-SiC substrates at 600C, some of which were subjected to pre-oxidation (FG)* surface conditioning at 600C. EOT is expressed as the average from multiple measurement sites. The oxide growth rate increased by roughly 10% when the SiC surf ace was remote plasma-treated prior to oxidation. The uniformity of oxide film thickness was also examined for various oxidation time intervals at 600C. The standard deviations of EOT values, expressed as percentages of the average oxide thickness obtained from multiple wafer sites, are visible in figure 3.5. Oxide films grown on afterglo w-conditioned surfaces showed a slight decrease in thickness deviation, compared to oxides on RCA-treated surfaces. The observed improvement in oxide film uniformity was not nearly as significant as that found at 850C oxidation temperature.
49 Figure 3.5. Thickness uniformity of afterg low oxide films grown for various time intervals on p-type 4H-SiC s ubstrates at 600C, some of which were subjected to preoxidation (FG)* surface conditioning at 600C. The EOT standard deviation is expresse d as a percentage of the average of EOT values over multiple measurement sites. EOT deviation showed a slight decrease when the SiC surface was remote plasma-treated prio r to oxidation. The action of reactive hydroge n and nitrogen afterglow chemistry on 4H-SiC surfaces thus resulted in a significant an d positive impact on growth rate and film uniformity during subsequent oxidation for a wi de variety of growth temperatures, time intervals, and oxide thicknesses. As determined by C-KM capacitance metrology, an increase in average film thickness and decr ease in EOT deviation was observed when surfaces were conditioned via (FG)* remote pl asma prior to the oxidation stage. These results were in agreement with the suggestion that (N2:H2)* afterglow expos ure facilitated the preparation of a cleaner, smoother, or dered, passivated SiC surface with reduced surface defects and Si-enriched surface chemistry, compared to conventional wet chemical pre-furnace cleaning methods. Su ch a semiconductor surface was better suited
50 for in-situ oxide interface formation and film growth as part of a sequential afterglow processing approach. Additional electrical and chemical analysis of (FG)*-conditioned surfaces were needed to discuss these surface treatment effects in further detail (chapters 4 and 5). 3.2. High-temperature annealing effects vs. surface conditioning Another figure of merit for examining th e electrical behavior of oxide film structures was the flat-band voltage position of the C-V characteristic. VFB position was related to the net oxide charge and film th ickness according to equation (23). Thus VFB gave an indication of the net total amount of charge incorpor ated into the oxide bulk and near-interfacial regions. It is desired that the flat-ba nd position be stab le under electrical stresses (device operation) a nd temperature stresses (pos t-oxidation processing). As shown in section 2.2.2, VFB shifting to larger magnitude va lues occurred during electrical testing of oxide/4H-SiC structures, and the VFB position saturated to a maximum value after strong accumulation of the semiconductor. A consistent measurement protocol was implemented, in which this max VFB position was used for comp aring C-V characteristics of different films and processes. It was of particular interest to examine the stability of VFB position during hightemperature inert annealing of oxide films. The purpose of such anneals was to mimic the temperature stresses invol ved during typical post-oxidation processing such as polysilicon gate annealing or silicide forma tion. Based on preliminary findings, it seemed possible that pre-oxidation afterglow surf ace conditioning might have an impact on VFB shifting behavior during post-oxidation annealin g at high temperature. However, these results were inconclusive since they were drawn from a scattering of experiments
51 performed at different times, with differing conditions such as ramp media, load and unload temperature, oxidation time, etc. This of course warranted a specific and consistent series of experi ments to investigate the VFB response to high-temperature postoxidation annealing, as a function of surface conditioning. Afterglow oxidations were performed at both 600C and 850C in standard (O2:N2O:FG)* media. At 600C the 4H-SiC substrates were oxidized for 60 min. followed by optional Ar annealing at 950C fo r 30 min. At 850C the oxidation interval was 15 min. followed by an optional 30 min. Ar anneal at 1000C after furnace ramp-up. Preceding the oxidation stage, semiconducto r surfaces were conditioned for 20 min. at 600C in forming gas (N2:H2)* 19:1 remote-plasma afterglow chemistry or non-excited N2:O2 7:1 media. The resulting C-V charact eristics obtained by non-contact coronaKelvin metrology are displayed in figures 3.6 and 3.7 for 850C and 600C oxidation temperatures, respectively. From the VFB shifting behavior observed, it wa s apparent that high-temperature annealing did have a significant, yet undesira ble, impact on flat-band position. All oxide films which underwent post-oxidation Ar anne aling exhibited C-V cu rves with larger VFB magnitudes compared to non-annealed films. This corresponded to more negative VFB values on p-type and more posi tive values on n-type SiC. The increase in flat-band voltage implied that high-temperature stress leads to defect formation and charge incorporation, and thus a higher amount of net oxide charge, QTOT. The general effect of VFB increase after annealing seemed to have little dependence on surface conditioning or oxidation temperature.
52 a) b) Figure 3.6. Non-contact C-V characteristics of oxide films grown for 15 min. at 850C on p-type (a) and n-type (b) 4H-SiC substr ates, some of which underwent pre-oxidatio n (FG)* surface conditioning at 600C and/or pos t-oxidation Ar annea ling at 1000C for 30 min. All annealed films demonstrated significant VFB shifting to larger magnitude values. Also, films grown on (FG)*-treated surf aces demonstrated slightly larger VFB values.
53 a) b) Figure 3.7. Non-contact C-V characteristics of oxide films grown for 60 min. at 600C on p-type (a) and n-type (b) 4H-SiC substr ates, some of which underwent pre-oxidatio n (FG)* surface conditioning at 600C and/or po st-oxidation Ar annealing at 950C for 30 min. All annealed films demonstrated significant VFB shifting to larger magnitude values. Also, films grown on (FG)*-treated surf aces demonstrated slightly larger VFB values.
54 It should be noted that when surfaces were treated by (FG)* afterglow before oxidation, the resulting VFB was slightly larger compared to films grown on non-treated surfaces. This effect was visible at both 600C and 850C oxidation temperatures and for both annealed and as-grown films. Also, the absolute value of VFB and the magnitude of shifting after anneal were both smaller on n-type than p-type samples, which is a general property of all our afterglow oxide films on 4H-SiC. Furthermore, VFB magnitudes and shifts after annealing were sl ightly smaller for films grown at 600C compared to 850C. Oxide film thickness variations between e xperiments can account at least partially for some of these observations. At each grow th temperature, the film thicknesses from different experiments were similar, but not identical. Due to this variation in EOT between experiments, there was a correspondi ng variation in oxide capacitance, which was inversely proportional to thickness per equa tion (21). The capacitance curves in the figures have been plotted normalized to COX for ease of viewing. However, VFB position was also thickness-dependent. For a fixed level of QTOT, a thicker oxide film corresponds to a smaller COX and hence a larger VFB per equation (23). As discussed in section 3.1, films grow n on (FG)*-conditioned surfaces were slightly thicker. This result wa s consistent with the increased VFB observed on afterglowtreated surfaces. Likewise, th e oxides grown at 850C were thicker than those formed at 600C, which was also consistent with the larger absolute values and shifts of VFB observed in films grown at the higher temperature. Film thickness variations were not the sole cause of changes in VFB position. The other contributing factor was the net total oxide charge mentioned previously. Simple estimates of QTOT were calculated from the C-V measurements using CACC and VFB values,
55 Table 3.1. Net total oxide charge estimated from non-contact C-V characteristics o f afterglow oxide films. Post-oxidation hightemperature Ar annealing caused a significant increase in QTOT, regardless of pre-oxidation surface conditioning or oxidatio n temperature. QTOT (1012 cm 2) p-type 4H-SiC n-type 4H-SiC TOX = 850CTOX = 600CTOX = 850C TOX = 600C no anneal (N2:H2)* 6.7 6.6 4.5 4.9 6.2 6.4 4.4 5.2 anneal Ar (N2:H2)* 12 12 7.4 7.1 11 13 7.0 6.8 and are listed in table 3.1. These QTOT estimates took into account any thickness variations. By far the most evident trend was the increase in QTOT caused by hightemperature Ar annealing of oxide film s. This was consistent with the VFB shifting to larger values during annealing. Also, the oxides on n-type material had somewhat smaller QTOT compared to p-type, which was consistent with the smaller VFB values generally exhibited by films on n-type compared to p-type 4H-SiC. On the contrary, both pre-oxidation surface treatment and choice of ox idation temperature seemed to have very little observable impact on QTOT. Based on these results, it was concluded that the effects of conditioning 4H-SiC surfaces in (N2:H2)* afterglow chemistry did not have a significant impact on oxide charge incorporation during oxidation and an nealing processes. Furthermore, hightemperature post-oxidation annealing di d cause a significant increase in VFB and QTOT values compared to as-grown dielectric films. The observed oxide charge increase might have resulted from interfacial rearrangement and viscous oxide fl ow at the elevated annealing temperatures. Desp ite posing a challengi ng hurdle to the development of 4H-
56 SiC device processing and manuf acture, this problem did not seem to be closely related to the focus on surface conditio ning effects in afterglow oxidation processing, and did not warrant further discussion within the scope of this work. Hence, the issue of VFB shifting during high-temperature annealing remain s to be addressed in future work. A variety of afterglow oxidation experime nts were performed on 4H-SiC at growth temperatures of 600C 850C for 10 90 min. intervals resulting in oxide films 50 500 thick. An emphasis was placed on th e effects of surface conditioning via (N2:H2)* afterglow to prepare the substrate for oxida tion. Corona-Kelvin capacitance metrology was used to evaluate electrical oxide parameters. General tre nds of growth rate increase and film uniformity improvement were witn essed to be caused by pre-oxidation (FG)* surface treatment. Tentative hypotheses of cleaning, nitridation, surface passivation, defect reduction, and carbon re moval were suggested as mechanisms to explain the observed effects of hydrogen and nitrogen species on the surfaces. The following chapters treat a deeper investigation of conditioned 4H-SiC surfaces, in which the metrology tools of C-KM depletion VCPD transient characterization and XPS analysis were used to supply additional electrical a nd chemical evaluation of surface treatments.
57 Chapter 4. Corona-Kelvin VCPD Transients on Conditioned 4H-SiC Surfaces Afterglow conditioning and other surface trea tments on 4H-SiC were investigated in an attempt to further understand the im pact of (FG)* exposur e as a pre-oxidation surface preparation technique. Non-contact VCPD voltage transient measurements were used to provide important and direct electr ical information regarding surface barrier and charge compensation processes after depletion of a bare semiconductor by corona charge deposition. Corona-Kelvin depletion volta ge monitoring was used as a non-contact method for electrical evaluati on of 4H-SiC surfaces after va rious conditioning treatments. A measurement protocol wa s established and the VCPD voltage decays on depleted 4H-SiC surfaces were interpreted in terms of a char ge compensation model. Forming gas (FG)* afterglow treatments were compared with othe r selected types of surface conditioning. Treatment time and temperatur e were addressed, as well as the durative stability of conditioned surfaces in retaining their state following treatment. 4.1. VCPD transient measurement protocol and interpretation The chosen method for electrical characte rization of semiconductor surfaces in this work combined corona charge deposition and non-contact voltage measurement, as with the oxide film characterization discussed previo usly. However, in the case of the bare semiconductor, a single corona charge pulse was deposited on the surface. A large corona dose and appropriate ch arge polarity were chosen in order to achieve strong depletion of the semiconductor. Following ch arge deposition, the depletion voltage was
58 observed over a significant time interval. The first VCPD reading was executed almost immediately after the corona charge pulse delayed by approximately one second while the CPD probe was shifted to the deposition site Without subsequent charging steps, the VCPD decayed in magnitude from the first measured value. Recalling equation (27), in the absence of an oxide layer, and neglecting the miniscule constant MS (< 1V), the measured depletion voltage was practically equivalent to the semiconductor surface potential barrier: VCPD VSB (29) Thus the observed VCPD transient corresponded to a decay of the depletion surface barrier from its initial value following the corona charge pulse. The VSB decay was associated with charge compensation processes in the se miconductor. In partic ular, some means of carrier generation/emission and transport allowed VCPD to change after the initial spacecharge region width and surface barrier voltage were established following corona application. The basic VCPD transient measurement protocol is summarized in the following two sequential steps: 1) deposit large corona charge pulse to force semiconductor in to strong depletion. 2) monitor VCPD over time as the surface barrier decays from its initial value. For all VCPD transient metrology described in this wo rk, a standard corona pulse size was consistently used to achi eve an areal density of QC = 1.51012 q cm 2, or equivalently 2.410 7 C cm 2, on the measurement site. Corona ions, CO3 or H3O+(H2O)n were deposited on nor p-type SiC material, respectively, ap propriate for depleting the semiconductor. Following the initial VCPD measurement (V0), the voltage was continually
59 monitored over time as additional readings were repeated every 200 ms for a total interval of 180 s. Thus each voltage curve was comprised of 900 data points obtained during a full 3 minutes af ter charge deposition. Figure 4.1. Depletion surface barrier transients obtained at multiple sites on RCA cleaned n-type 4H-SiC epi-wafer A after negative cor ona deposition. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Decay of VSB and depletion region width was attribut ed to charge compensation processes due to carrier emission from traps in the presence of high electric field. As an example, consider the VCPD decay curves depicted in figure 4.1. The SiC material examined was the Si-face of an n-type (Nd = 11015 cm 3) 4H-SiC 8 off-axis 3in. epi-wafer obtained from Cree, Inc. and grow n with their standard epitaxial process. This epitaxial wafer was identified as wafer "A ". Prior to measurement, the SiC surface was only processed by wet cleaning using standard piranha and RCA solutions and ending with a dilute HF dip and DI rinse. VCPD transients were obtained at 5 different wafer sites, organized in a cross pattern with radius 25 mm, and referenced to the major
60 flat as the bottom direction on the wafer. Some small variation in starting value (V0), final value (Vsat), and decay rate ( VCPD) were observed over wafer A (figure 4.1). However, the general trend of VCPD decay was clearly evident. The first measured value of depletion surface potential barr ier was around an average of V0 = 150 V, and slowly decayed during 3 minutes to an average of 16 V. The observed temporal decay of VSB was associated with a corresponding decrease in space-charge region width, pr oportional to the square root of the surface barrier: W2 = 2 0rVSB(qNd) 1 (30) where W is depletion region width, 0 is permittivity of vacuum, r is relative permittivity of the semiconductor, VSB is surface potential barrier, q is elemental charge, and Nd is dopant concentration. Th e space-charge density QSC was directly proportional to the depletion region width: QSC = qNdW (31) and the maximum value of QSC must equal the surface corona charge density QC minus any compensating charge generated in the se miconductor. After the space-charge region and surface barrier were initially established, associated with the deposition of corona charge on the semiconductor surface, any decay in VSB and W was due to the emission/generation and transport of charge carriers resulting in the compensation of surface corona charge. Otherwise, electros tatic equilibrium would remain unperturbed, resulting in a constant surface barrier. Given th e fact that these measurements were performed at room temperature, and in am bient darkness, direct band-to-band carrier generation was highly improbable due to the la rge band-gap of 3.27 eV. It was similarly
61 improbable that electron-hole pair generation assisted by mid-gap generation centers was a dominating factor. The principal mechanism of temporal charge compensation was suggested to revolve around the emission of charge carriers from deep-level traps concentrated in the semiconductor surface region. As majority ca rriers were emitted from traps into the conduction or valence energy bands (nor p-do ped, respectively) they were swept into the semiconductor, driven by the electric fiel d present at the surf ace, and eventually recombined with dopant ions in the space-ch arge region. The release and movement of majority charge away from the surface wa s responsible for compensation of corona surface charge, and thus the observed decay of surface potential barrier height and corresponding depletion region width shrinkage. The proposed charge compensation mechanis m is simply depicted in figure 4.2 by way of two snapshot diagrams. In the illustration, a dose of negative corona charge was deposited on an n-type SiC surface, which repe lled majority carriers (electrons) into the semiconductor. Due to the surface electric field penetrating into the semiconductor, a space-charge region of positively ionized dopant atoms was established (figure 4.2a) with a corresponding depletion charge density QSC and surface barrier voltage VSB. However, the probability of carrier emission from trap s in the surface region was also enhanced by the electric field. Any emitted electrons we re driven into the semiconductor by the surface E field and recombined with dopant ions in the space-charge region. Regardless of whether the filled traps were neutral or charged impurities, the emission of electrons resulted in a net positive charge increas e in the surface region and compensation of negative surface corona charge. Consequentl y, the depletion region width, space-charge
62 a) b) Figure 4.2. Diagram of the charge compen sation mechanism associated with the temporal decay of surface barrier, deple tion width, and space-cha rge density. Fieldenhanced carrier emission from surface trap s and recombination with dopant ions in the space-charge region result in the compensati on of surface corona charge. The depletio n region width and surface barrier decay from th eir initial values established at coron a deposition (a) to smaller values after charge emission from surface traps (b).
63 density, and surface potential ba rrier decreased in order to maintain charge neutrality (figure 4.2b). Without considering the input of carrier emission from traps, the depletion spacecharge density equaled the deposited corona charge. However, because of charge emission processes, the space-charge de viated from the ideal according to: QSC = QC Qdef (32) where QSC is depletion space-charge density, QC is surface corona charge density, and Qdef is the effective density of compensating charge liberated from defects over time. (a) (b) (c) Figure 4.3. Illustration of elect ric field enhanced carrier em ission from localized states. In the presence of electric field, Poole-Fren kel emission (b) or phonon-assisted tunneling mechanisms (c) can increase the probability of carrier emission compared to the basic thermionic emission of carriers with out E field stimulation (a) . Thus the observed decay in the surface potenti al barrier was suggested to be a result of charge compensation from field-stimulated carrier emission processes. In the absence
64 of electric field, carriers may be emitted from traps by thermionic emission (figure 4.3a), using only thermal energy from the environment to provide the difference between the trap level and conduction band (T). However, thermionic emission was improbable to occur at room temperature from deep levels with T >> kT 25mV, since the probability of emission had an inverse exponential depe ndence on trap energy nor malized to thermal ambient energy: re ~ exp(T /kT) (33) where re is emission rate, T is relative trap energ y, and kT is thermal energy. In an electric field, carrier emission from traps may be enhanced in two ways. The first mechanism, known as Poole-Frenkel emi ssion , occurs due to Coulomb barrier lowering in the presence of an electric field (figure 4.3b). Poole-Frenkel emission processes are only possible when the filled trap is in a charged state, because barrier lowering does not occur if the defect is neutral. Also, PooleFrenkel emission is significant at relatively low electric fields. At higher E field strengths, phonon-assisted tunneling becomes the more dominant emi ssion mechanism (figure 4.3c) . The enhanced degree of energy band-bending at high fields allows carriers to tunnel from the trap level directly into the conduction band with nonnegligible probability. When partially assisted by energy input from ther mal vibrations, the probability of tunneling increases because the energy barrier that the electron mu st tunnel through is less. The surface electric fields resul ting from corona ion deposition in VCPD transient metrology were relatively high. The standard corona dose of QC = 2.410 7 C cm 2 corresponded to an electric field of 269 kV/cm at the surface, according to the relation:
65 E = QC / 0r (34) Regarding the transition between phonon-assist ed tunneling and Poole-Frenkel emission as competing processes, values below the orde r of 1 kV/cm were considered to lie within the low-field regime . Hence, the fiel d strengths typically used in this work (hundreds of kV/cm) warrant the conclusion that phonon-assisted tunneling was the more dominant mechanism for field-stimulated ca rrier emission. Thus, the observed temporal decay of surface potential barrier in depleti on was suggested to be a result of charge compensation due to field-enhanced carrier emission from deep-level surface traps. Figure 4.4. Depletion surface barrier transients obtained at multiple sites on RCA cleaned n-type 4H-SiC epi-wafer B af ter negative corona deposition. VCPD was monitore d during 180 sec. following corona application of density QC = 2.410 7 C cm 2. The inset is zoomed in on the low-voltage portion of curves. The variation of initial measure d voltage V0 and the faster rates of VSB decay indicated higher concentrations of chargeemitting defects and non-uniform epitaxial quali ty on wafer B compared to wafer A (see figure 4.1).
66 Defect concentration variati ons, particularly near the SiC surface, were expected to have an observable impact on VCPD transient behavior. Shallow states (T < kT) were suggested to empty immediately following corona deposition, mu ch faster than the initial voltage reading V0. Deep-level traps were expected to contribute to the slow VSB decay observable over the 3 min. transient measurement. Thus, both the initial voltage and rate of surface barrier decay depended on def ect levels and concentrations in the semiconductor surface region. Figure 4.4 shows voltage decays obtained on another n-type 4H -SiC epi-layer, identified as wafer "B", which was grown w ith a non-standard epitaxial process by Cree, Inc and doped identical to wafer A (Nd = 11015 cm 3). In comparison to material A, epi B had higher defect levels and more variation of defect conc entrations across the wafer. This was clearly evident when comparing the VCPD decay behavior between figures 4.1 and 4.4. On average, wafer B dem onstrated much lower initial V0 readings, and much faster rates of VCPD decay, than the standard epitaxial material A. However, the bottom measurement point (0, 25mm) on B had a high V0 and slow VSB decay, indicating a high degree of non-uniformity of charge-emitting def ects across the wafer. In fact, the bottom of wafer B behaved comparably to the be st points on A. In general, the VCPD transient was fairly uniform across wafer A. The cente r point demonstrated the largest deviation from the other wafer sites. The faster voltage decay in the center implied higher concentrations of deep-level charge-emitting tr aps compared to the radial sites near the outside of the wafer. The majority of VCPD transient data presented in this work consisted of a single decay curve obtained immediately following the initial corona charge pulse deposited on
67 Figure 4.5. VCPD transient decays with consecutiv e repetitions of corona depositio n spaced at 3 min. intervals, obtained on RCA cl eaned n-type 4H-SiC epi-wafer A. Each curve is the average of multiple wafer sites. VCPD was monitored during 180 sec. following each of 3 corona applications of density QC = 2.410 7 C cm 2. The rate of VSBdecay was very consistent among repeated measurements. a freshly treated surface. However, sim ilar transient behavior was observed with additional charge pulses and voltage measur ements. Figure 4.5 gives an example of consecutive charge and decay measurements on wafer A after an RCA clean. Each curve in the figure was the average of transients from multiple wafer sites. Three separate corona pulses of QC = 2.410 7 C cm 2 were spaced at intervals of 3 minutes, and the VCPD decays were recorded after each deposition. The average rates of decay and starting values were quite similar. A very s light voltage increase was observed at each subsequent measurement. Although most of the depletion space-charge had dissipated due to charge compensation during the 3 min. of measurement, some small quantity of
68 space-charge still remained when the next corona pulse and measurement were executed. Thus the depletion region width was slightly larger after the next corona deposition. 4.2. Surface conditioning impact on VCPD decay The observed slow VCPD decay after depletion charging was suggested to result from carrier emission from surface deep-level trap s, recombination with depletion region space-charge, and compensation of surface cor ona charge. As previously introduced, defects in the 4H-SiC surface region were at tributed to a conglomeration of various dangling bonds, chemical states, hydrocarbons vacancies, C-clusters, other C-related defects, hydroxyls, oxygen, fluorine and othe r residual contaminants, etc. Hence, VSB transient decay behavior was used as an in dication regarding concen trations of chargeemitting surface defects, and served as a valuable figure of merit when comparing the effectiveness of different surface conditioning treatments in the reduction and passivation of surface defects. A series of surface conditioning experiments were performed to compare the standard pre-oxidation (FG)* afterglow trea tment with other conditioning treatments. The baseline afterglow treatment consisted of an RCA clean, furnace loading, temperature ramp-up in N2:O2 7:1, 20 min. of exposure to (N2:H2)* 19:1 afterglow at 600C, followed by immediate unloading, cooli ng, and characterization of the treated surface. Other treatment ambients included 20 min. exposures to pure (N2)* afterglow or non-excited N2:O2 7:1 media at 600C. Surfaces rinsed in DI water following the baseline (FG)* afterglow condi tioning were also examined, as well as the previously introduced RCA cleaned surfaces.
69 a) b) Figure 4.6. Depletion VSB transients obtained at multiple sites on n-type 4H-SiC epiwafers A (a) and B (b) following (N2:H2)* afterglow surface conditioning for 20 min. a t 600C. VCPD was monitored during 180 sec. followi ng corona applica tion of density QC = 2.410 7 C cm 2. Rates of VSB decay were extremely slow, implying a much smalle r concentration of charge-emitting surface defect s on (FG)*-treated surfaces vs. RCA clean (compare to figures 4.1, 4.4 for wafers A, B).
70 Depletion voltage decays on epi-layers A (figure 4.6a) and B (figure 4.6b) were obtained after the baseline (FG)* afterglo w treatment for 20 min. at 600C. The VCPD transient response on (FG)*-conditioned surfaces was dramatically improved in comparison to the same materials after RCA cleaning (figures 4.1, 4.4). An extremely slow rate of VSB decay was observed on the standard epi-wafer A. Averaged across the multiple wafer sites, the initial VCPD value around 161V decayed an average of only 3V during the 3 min. measurement interval. Th e drastic reduction in surface barrier decay rate implied very few sources of compensating charge that were evident during the measurement, attributed to lower levels of deep-level traps on the surfaces prepared via the (FG)* afterglow baseline process. Alt hough a comparatively faster rate of decay occurred in the center point, this was cons istent with the observed decays after RCA cleaning, which indicated that the center of wafer A had higher concentrations of chargecompensating defects relative to the other measurement sites. However, an ideal surface barrier around 20 0V was expected if all the corona charge were perfectly imaged in the space-charge region. The measured V0 value of 161V corresponded to roughly 2.5108 C cm2 of compensating charge density emitted after corona depostion but before initial VCPD measurement. The majority of this initial charge compensation was attributed to shallow trap levels and crystal defects, and were not strongly affected by surface condition. The VCPD decays observed on wafer B after (FG)* conditioning were also somewhat faster compared to wafer A. An average of 12V of decay over 3 min. was observed on wafer B, compared to 3V on wafer A. The faster decay rates and larger voltage distribution were also consiste nt with the RCA clean results, implicating the presence of
71 higher surface defect co ncentrations and non-uniform epitaxi al quality on wafer B. The bottom point on wafer B exhibited very slow decay, only 0.3V during 3 min. of transient measurement. This was slower than even the best points on wa fer A, which decayed 0.9V and 1.0V over 3 min. It was concluded th at the wafer B bottom point had extremely low concentrations of surface defect sources for charge compensation, while the rest of the tested wafer areas had hi gher surface defect densities. Figure 4.7. VCPD transient decays with consecutiv e repetitions of corona depositio n spaced at 3 min. intervals, obtained on n-type 4H-SiC epi-wafer A after (FG)* surface treatment for 20 min. at 600C. Each curve is the average of multiple wafer sites. VCPDwas monitored during 180 sec. following each of 3 corona applications of density QC = 2.410 7 C cm 2. Rate of VSB decay was very consistent among repeated measurements, while the increase in initial measured voltage V0 after each consecutive corona depositio n was due to remaining depletion spacecharge from previous measurement. The results of repeated VCPD transient measurements on (FG)*-conditioned surfaces are shown in figure 4.7 as averages of multip le sites on wafer A. Three consecutive charge and decay intervals were executed. After each consecutive charge pulse, the rates
72 of voltage decay were similar. However, the starting VCPD value increased significantly, reaching a V0 of 242V after the 3rd charging. Due to the slow rate of VSB decay on the (FG)*-treated surfaces, much of the depletion space-charge re mained at the end of the 3 min. measurement interval. Th e subsequent deposition of an other corona pulse caused an increase in space-charge density and depl etion region width, resulting in a higher VSB value at the beginning of the next transient. The increase in V0 upon subsequent measurements was not nearly as dramatic on the RCA cleaned surfaces (figure 4.5) because of the much higher rates of VSB decay due to higher concentrations of chargeemitting surface defects. Evidently, a large reduction of surface charge-emitting defects was accomplished by conditioning of 4H-S iC surfaces in (N2:H2)* afterglow, as supported by a large decrease in depletion VSB decay rate relative to RCA cleaned surfaces. Other treatments were examined to compare to the st andard (FG)* conditioning process. VCPD transient measurements were obtained on wafer A (f igure 4.8a) and wafer B (figure 4.8b) after each treatment. Curves shown are averages from multiple wafer sites. For ease of visualization, the same results are shown in figur e 4.9 with voltage values plotted relative to the initial VCPD reading, as: VCPD = VCPD V0 (35) where VCPD is the relative voltage, VCPD is the absolute voltage, and V0 is the initial reading of VCPD following charge deposition. Plotting voltages offset by the initial value gave a clearer comparison of VSB decay rates, without the added visual confusion of varying V0 values.
73 a) b) Figure 4.8. Depletion surface barrier decays obtai ned on n-type 4H-SiC epi-wafers A (a) and B (b) after various surface c onditioning treatments, including (N2:H2)* or (N2)* afterglow exposure and non-excited N2:O2 media at 600C for 20 min., DI water rinsing after (FG)* conditioning, and standard RCA cleaning. Each curve is the average o f multiple wafer sites. VCPD was monitored during 180 sec. following corona application o f density QC = 2.410 7 C cm 2.
74 a) b) Figure 4.9. Depletion surface barrier transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) after various surface conditioning treatments, plotted relative to initial measured voltage to ai d visualization of VSB decay rates. Each curve is the average o f multiple wafer sites. VCPD was monitored during 180 sec. following corona application o f density QC = 2.410 7 C cm 2. Forming gas (N2:H2)* afterglow treated surfaces exhibite d the slowest rates of VSB decay among all treatments including pure (N2)* afterglow.
75 Figure 4.10. Final voltage values of depleti on surface barrier transients obtained on ntype 4H-SiC epi-wafers after various surface conditioning treatments. Each Vsat value is the wafer average of the fina l voltage value obtained af ter 3 minutes of decay. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Forming gas (N2:H2)* afterglow treated surfaces exhibited the highest Vsat values among all treatments including pure (N2)* afterglow. Figure 4.10 contains the Vsat values which corresponded to the average final voltage values obtained after 3 minut es of measurement. The Vsat value in figure 4.10 was exactly the last voltage measured after 180 se c. of decay in figure 4.8. A higher value of Vsat indicated less charge compensati on and improved surface passivation. An examination of the VCPD decays revealed that forming gas (N2:H2)* afterglow was the most effective at reducing concen trations of charge-emitting surface defects, indicated by the slowest decay rate and highest Vsat of the treatment options. However, surface conditioning in pure (N2)* afterglow at 600C for 20 min. also achieved a very low rate of VSB decay, although the observed rate of 9V over 3 min. (wafer A) was not as slow as the 3V of decay obt ained after (FG)* afterglow treatment. Evidenced by the electrical surface behavior, the 5% H2 present in the FG mixture did have an impact on
76 cleaning and passivation of the semiconducto r surface, perhaps due to contaminant removal and termination of dangling bonds and chemical surface states. Hence the combination of N2 and H2 afterglow was more effective at surface defect reduction and passivation than pure N2 alone. Following the standard (FG)* afterglow c onditioning, rinsing of the surface in DI water did result in faster VSB decay rates relative to as -treated surface s, although the difference in transient behavior was not larg e. In fact, (FG)*-c onditioned surfaces after DI rinse still performed comparable or slightly better than the pure (N2)* afterglow treatment. Based on this observation, one mi ght suggest that DI rinsing removed the majority of the bond termination and defect pa ssivation accomplished by the inclusion of H2 in the afterglow media. Simply put, the water rinse possibly removed some hydrogen from the surface. Exposing 4H-SiC surfaces to non-excited N2:O2 7:1 at 600C for 20 min. seemed to accomplish little more than the basic RCA clean. Depletion VSB decay rates were very fast and Vsat values were low following N2:O2 treatment, implying that concentrations of surface defects for charge compensation were comparable to RCA cleaned surfaces. Thus the standard furnace temperature pr ofile and exposure to nitrogen-containing ambient were not sufficient in and of themse lves to produce any sign ificant reduction of surface charge-emitting defects. The afterglow species of N2 and H2 microwave discharge were suggested as critical to the bond termination, defect passivation, and contaminant removal effects which resu lted in the observed lowering of VSB decay rate.
77 Figure 4.11. Depletion VSB transients obtained on p-type 4H-SiC 1 cm2 sample comparing RCA clean to (FG)* afterglow surface conditioning for 20 min. at 600C. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. (FG)*-treated p-type material demons trated a much slower rate of VSB decay than RCA cleaned surfaces. The majority of surface conditioning experiments for the purpose of VCPD transient evaluation in this work were investigated on n-type 4H-SiC material For completeness, a small selection of experiments included p-type samples in order to confirm that the general trends of conditioning effects were not exclusive to n-doped material. Figure 4.11 gives an example of VCPD transient behavior on a 1 cm2 p-type 4H-SiC sample, comparing the RCA cleaned surface to the standard (N2:H2)* afterglow treatment. A dose of 2.410 7 C cm 2 of positive corona ion density was used to pulse the p-type semiconductor into depletion, followed by VSB transient monitoring. Consistent with findings on n-type SiC, the RCA cleaned surface s showed very fast decay rates, while the (FG)*-conditioned surfaces demonstrated extremely slow rates of VSB decay. In the case
78 of ptype material, the depletion surf ace potential barrier was positive. Both VCPD transients began around the same V0 value of 117V. The surf ace barrier on RCA cleaned surfaces decreased very rapidly, with appr oximately 100V of decay during the 3 min. measurement, while the (FG)*-treated surf ace demonstrated only 2.3V of decay over the identical time interval. Thus on both p-t ype and n-type 4H-SiC semiconductors, (FG)* afterglow conditioning provided an effective means of surface passivation and reduction of charge-emitting defects, and produced the slowest rates of depletion VSB decay among the various treatments investigated. 4.3. (N2:H2)* afterglow treatment variations: time, temperature, durative stability The use of forming gas (N2:H2)* afterglow was demonstrated to be a superior method of surface passivation and charge-emitting defect elimination. The slowest rates of VSB decay and highest Vsat were achieved on (FG)*-conditioned 4H-SiC surfaces among other selected treatments including pure (N2)* afterglow, non-excited media, and wet cleaning. Further investigation was requi red to reveal how the choice of treatment time and temperature impacted the proficiency of the (FG)* surface conditioning. For this purpose, an additional set of experiments incorporated variations of time and temperature to compare to the standard ba seline parameters of 20 min. exposure at 600C. Depletion VCPD transient decays were used to ev aluate the effectiveness of each (FG)* treatment variation in passivating surfac e charge-emitting defects. Also, a separate set of experiments examined the stability of the (FG)*-treated surface in retaining its condition as a function of delay time and other post-treatment stresses. Silicon carbide surfaces were conditioned for time intervals of 2.5 20 min. in (FG)* afterglow media at 600C furnace temperat ure. All experimental conditions were
79 a) b) Figure 4.12. VSB transient decays obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow treatment for various time intervals at 600C. Each curve is the average of multiple wafer sites. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Slower rates of VSB decay were achieve d after longer trea tment times.
80 Figure 4.13. Final voltage values of depletion VSB transient decays obtained on n-type 4H-SiC epi-wafers following (FG)* afterglo w treatment for various time intervals a t 600C. Each Vsat value is the wafer average of the final voltage value obtained after 3 minutes of decay. VCPD was monitored during 180 sec. following corona application o f density QC = 2.410 7 C cm 2. Higher Vsat values and improved su rface passivation were achieved after longer treatment time. held constant except for the time of treatment. Investigation of depletion VSB transients revealed a distinct correlation between treatment time and resulting decay rate and uniformity. Initial V0 values were similar, but the rate of VCPD decay increased as the treatment time was decreased, as seen in figures 4.12a and 4.12b for wafers A and B, respectively. The decays curves shown were averages of multiple wafer sites. The Vsat values corresponding to the final voltages after 180 s ec. of decay are s hown in figure 4.13 as a function of treatment time. The la rgest concentration of surface chargecompensating defects was observed after the sh ortest treatment time of 2.5 min., resulting in an average 33V of decay on wafer A duri ng the measurement interval ending with a Vsat of 127V compared to only 3V of decay and Vsat of 158V after the standard 20 min.
81 treatment time. Of the treatment durations studied, the longest treatment of 20 min. (FG)* afterglow provided the best surface passivation and charge-emitting defect reduction. Figure 4.14. Uniformity of VSB transient decays obtained on n-type 4H-SiC following (FG)* afterglow treatment for various time intervals at 600C. The standard deviation o f VCPD was calculated from multiple wafer sites, and averaged over the 3 min. measurement interval. The deviation values shown were calculated from the data of the average VCPDdecays in figure 4.12. The highest uniformity of surface condition was achieved at the longest treatment time of 20 min. (FG)* afterglow. To examine the uniformity of the treated surface condition as a function of treatment time, the standard deviation of VCPD was calculated and averaged over the 3 min. measurement interval. Figure 4.14 contai ns the uniformity calculated values which displayed a trend of decreasing VCPD deviation with increasi ng (FG)* treatment time on both wafers A and B. Thus the longest treatment time of 20 min. (FG)* afterglow
82 resulted in the lowest deviation of depletion VSB decay behavior across the wafer and hence the highest uniformity of surface condition achieved during treatment. Forming gas afterglow conditi oning processes were perfor med at temperatures of 400C 800C to examine the dependence of surface passivation effects on thermal energy input. In each experiment, 4H-SiC wafers were RCA cleaned, loaded in the afterglow furnace under N2 flow at 400C, ramped up in non-excited N2:O2 7:1 media to a specified treatment temperature (400 800C), conditioned in (N2:H2)* 19:1 afterglow for the standard 20 min. interval immediately unloaded under N2 flow, cooled in cleanroom ambient, and characterized. Resulting VCPD transients after (FG)* conditioning of 4H-SiC surfaces at different temperat ures are depicted in figures 4.15a and 4.15b for wafers A and B, respectively. Some temperature dependence was apparent in the rates of VSB decay. The VCPD curves were also plotted relative to V0, to facilitate easier visualization of decay rates (figures 4.16a and 4.16b, wafers A and B, respectively). The final voltage values of the depletion VCPD decays are displayed in figure 4.17 as a function of treatment temperature. Apparently, a thermal energy threshold existed between conditioning temperatures 500C and 600C. Forming gas (FG)* afterglow surface treatment performed at the low temperatures 400C and 500C result ed in much higher rates of VSB decay and lower Vsat values, in fact an order of magnitude fa ster than decays obtained from treatment temperatures 600C 800C. Evidently, a temperature of at least 600C was needed to provide the required thermal activation of surface passivation processes which occur during the (N2:H2)* afterglow conditioning of 4H -SiC. However, increasing the temperature to 700C or 800C produced quite similar depletion VSB transients compared
83 a) b) Figure 4.15. Depletion VSB transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditioning for 20 min. at treatment temperatures in the range 400C 800C. Each curve is the average of multiple wafer sites. VCPD was monitored during 180 sec. following co rona application of density QC = 2.410 7 C cm 2.
84 a) b) Figure 4.16. Depletion VSB transients obtained on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditioning for 20 min. at various treatment temperatures, p lotted relative to initial measur ed voltage to aid viewing of VSB decay rates. Each curve is the average of multiple wafer sites. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Forming gas (FG)* conditioning a t temperatures 600C or higher resulted in an order of magnitude slower rate of VSB decay, indicating a thermal energy threshold exis ts between 500C and 600C (FG)* treatment temperature.
85 Figure 4.17. Final voltage values of depletion VSB transients obtained on n-type 4H-SiC epi-wafers following (FG)* af terglow conditioning for 20 min. at treatment temperatures in the range 400C 800C. Each Vsat value is the wafer average of the final voltage value obtained after 3 minutes of decay. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Forming gas (FG)* conditioning a t temperatures 600C 800C resulted in higher Vsat values compared to treatments at 400C 500C, indicating a thermal energy threshold exists between 500C and 600C (FG)* treatment temperature. to 600C. It should also be noted that prior work investig ated the (FG)* conditioning of as-grown 4H-SiC epitaxial material at temperatures between 600C 1100C? . Atomic force microscopy revealed a minimum in resulting surface roughness after (FG)* conditioning between 600700C. Judging from these results, 600700C is an optimal choice of (FG)* afterglow conditioning temper ature in order to accomplish the combined effects of surface smoothing and defect passivation. Most of the VCPD transient results presented in this work were obtained after minimal delay following surface treatment, in an effort to ensure that any change in surface condition over time would not impact the characterization results. However,
86 certain measurements were performed with the specific intent of testing the stability of surface condition to time delay afte r (FG)* afterglow treatment. VCPD transient decays were measured on consecutive days following a 20 min. (FG)* surface treatment at 800C. VSB decays were obtained immedi ately after treatment, and at 24-hour intervals up to 3 days later. Following each measurement, wafers were exposed to light and stored in plastic Fluoroware ca ses until the next day. The surface s were not rinsed in DI water or disturbed in any other way betwee n measurements. The time-delayed VCPD decays are depicted in figures 4.18a and 4.18b for wafers A and B, respectively. Each curve shown is the average of multiple wafer sites. The observed rate of VSB decay was very consistent from day to day. The curve slope after the 3rd day of delay was practically identical to that obtained initially after the (FG)* surface treatment. The retention of VSB transient behavior over time provided str ong evidence to support the suggestion that (N2:H2)* afterglow conditioning of 4H-SiC achieved a stable and resilient state of surface passivation. Other aspects of surface condition stabil ity in addition to time delay were investigated on nand p-type 4H-SiC 1 cm2 samples. VCPD transients were recorded immediately after the standa rd (FG)* afterglow treatment for 20 min. at 600C. Following initial VCPD decay measurements, the surfaces were electrically stressed with a large corona pulse of opposite charge pola rity, to achieve accumulation of majority carriers at the semiconductor surface and elimin ate any space-charge remaining from the previous depletion voltage decay measuremen t. After accumulation stress and light exposure, the standard VCPD transient measurement was repeated with another depletion pulse of corona charge. Then the wafers were stored in plastic cases and measured again
87 a) b) Figure 4.18. Depletion surface barrier decays obtai ned on n-type 4H-SiC epi-wafers A (a) and B (b) following (FG)* afterglow conditi oning for 20 min. at 800C, and remeasure d after 1 day intervals of time delay. Each cu rve is the average of multiple wafer sites. VCPD was monitored during 180 sec. followi ng corona application of density QC = 2.410 7 C cm 2. The resulting surface condition prepared by (FG)* afterglow treatment was quite stable, showing consistent rates of VSB decay many days after treatment.
88 after 6 days of storage. Following VCPD transient measurement on the 6th day, the samples were heated for 60 min. on a hotplat e at 170C in cleanroom ambient. After heating, depletion surface barrier decays we re measured a final time. The resulting depletion VSB transient curves are shown in figur es 4.19a and 4.19b for nand p-type 4HSiC, respectively. The condition of (FG)*-treate d surfaces seemed to be very stable. All curves demonstrated similar starting volta ges and rates of decay. No significant deviation in VCPD depletion response was caused by a combination of accumulation electrical stress, thermal stre ss, air exposure, or time delay. It was concluded that the improved state of defect termination and surface passivation suggested to exist after (N2:H2)* afterglow treatment of 4H-SiC demonstr ated an enduring resilience and stability in retaining the surface conditioning eff ects while exposed to temporal, thermal, electrical, and environmental stresses. VCPD transient metrology was used as a tool for electrical ev aluation of 4H-SiC surfaces. The observed slow temporal de cay of the depletion surface barrier was attributed to charge compensation through fi eld-enhanced carrier emission from deeplevel surface traps and recombination with sp ace-charge dopant ions. Slower rates of VSB decay were suggested to be correlated wi th fewer charge-emitting surface defects. Surface conditioning treatments were examined to further investigate the effects of (FG)* afterglow exposure as a preoxidation surface preparation protocol. Of the various treatments, including wet cleaning, nonexcited thermal treatment, and pure (N2)* afterglow, the a superior effectiveness of charge-emitting defect reduction and surface passivation was achieved by forming gas (N2:H2)* afterglow exposure. A trend of decreasing decay rates and improvement in uniformity of surface condition was observed
89 a) b) Figure 4.19. Depletion surface barrier decays ob tained on n-type (a) and p-type (b) 4HSiC 1 cm2 samples following (FG)* afterglow conditioning for 20 min. at 600C, an d remeasured after accumulation corona stress, 6 day time delay, and heating in cleanroom ambient. Each curve is the average of multiple wafer sites. VCPD was monitored during 180 sec. following corona application of density QC = 2.410 7 C cm 2. Observed rates of VSB decay remained stable even when the (FG)*-treated surfaces were exposed to temporal, electrical, thermal, and environmental stresses.
90 as treatment time was increased. A treatment temperature of 600C or above was determined to provide the required thermal activation of the (FG)* afterglow conditioning process. The process para meters of 20 min. exposure at 600700C were identified as sufficient for a standard base line (FG)* afterglow surface treatment of 4HSiC. The resulting state of surface passiva tion was determined to be quite stable. Conditioning effects as observed by VCPD transient metrology were retained over significant periods of time following treatment, and were stable even when surfaces were exposed to various electrical, thermal a nd environmental stresses. The action of hydrogen and nitrogen afterglow species were su ggested to be responsible for preparation of the SiC surface, with a combination of dangling bond termination, cleaning of contaminants, reduction of surface states, ni tridation or nitrogen incorporation, Sienrichment, and passivation or removal of C clusters and other C-re lated defects. The supplemental XPS metrology results of chapte r 5 provided crucial ch emical information to facilitate the continuation of an in depth analysis of the (N2:H2)* afterglow conditioning impact on 4H-SiC surfaces.
91 Chapter 5. X-ray Photoelectron Spec troscopy of Conditioned 4H-SiC Surfaces Chemical and elemental analysis and identi fication of SiC surfaces were crucial to the investigation of afterglow conditioning effects and the impact of such surface treatments on subsequent oxidation processi ng. Results of non-contact corona-Kelvin capacitance metrology revealed an increase in oxidation growth rate and improvement in film uniformity after prep aring 4H-SiC surfaces by (N2:H2)* afterglow prior to oxidation. As indicated by C-KM VCPD transient decay measurements, (FG)* afterglow conditioning resulted in the lowest densities of ch arge-emitting surface defects among the other surface treatments considered. In add ition to the primary metrology method of noncontact C-KM electrical measurements on se miconductor surfaces and oxide films, this work also incorporated XPS examination of 4H-SiC to determine the impact of (FG)* afterglow treatment on surface chemistry. 5.1. XPS measurement technique The X-ray photoelectron spectroscopy techni que essentially adapts a high-energy version of the photoelectric effect to identify binding energies of chemical species at the sample surface. High-energy photons (1 2 keV) impinge upon the sample surface from a monochromatic X-ray source. As the primary X -rays interact with co re-level electrons, it is possible for any electron to be ejected if it has a binding energy EB less than the incident photon energy h Only electrons originating from the top 5 50 surface region are ejected from the sample, limite d by the electron escape depth . The
92 emitted photoelectrons arrive at the spectrometer with energy Esp, where they are analyzed and counted. The electron energy at the analyzer will be related to the binding energy and the primary photon energy as follows: EB = h Esp q sp (36) where EB is the core electron binding energy referenced to the Fermi energy EF, h is the primary X-ray energy, Esp is the energy of the ejected photoelectron arriving at the spectrometer, and sp is the spectrometer work func tion. The resulting XPS signal contains electron counts per se cond as a function of binding energy. Since the binding energy of an electron is influenced by its ch emical state, XPS spectra allow determination of chemical compounds and elements in the sa mple surface. Software analysis of peak heights and widths with appropr iate correction factors allows density calculations such as atomic percent of various elements. A measurement schematic and illustration of the electronic processes involved in XPS metrology are depicted in figures 5.1 and 5.2. Figure 5.1. XPS measurement schematic.
93 Figure 5.2. Electron energy band diagram ill ustrating photoemissi on of core level electrons in the XPS technique. 5.2. XPS results on 4H-SiC surfaces XPS measurements were used to compare SiC surfaces prepared by (FG)* afterglow conditioning and RCA wet cleaning. XPS measurement service was provided by the lab of Fred Stevie at NCSU. XPS spectral data were analyzed using CasaXPS processing software . After standa rd RCA cleaning of n-type 4H-SiC 1 cm2 samples, some surfaces were conditioned by the baseline (FG)* afterglow process for 20 min. at 600C. Afterglow conditioned samples were unloaded from the furnace at 600C, allowed to cool in cleanroom ambient, a nd immediately shipped along with the RCA cleaned samples in Fluoroware cases for ne xt-day measurement. XPS spectra were obtained on as-received surfaces, and af ter 2 min. sputtering with 5 keV Ar+ source estimated to remove 4 of material.
94 a) b) Figure 5.3. XPS spectral data (a ) and atomic percent values (b) obtained on n-type 4HSiC surfaces after RCA clean or (FG)* afterglow treatment.
95 Figure 5.3a compares the survey spectral data of the XPS signal obtained on 4HSiC surfaces after RCA clean and (FG)* afterg low treatment. The XPS spectra shown in the figure were obtained on as -received surfaces after sample shipment, without any additional cleaning or sputtering. Atomic percen t values of the elements of interest were calculated in CasaXPS and ar e displayed in figure 5.3b. Peaks around binding energies 102, 285, 400, and 534 eV were prominent in th e XPS spectra, corre sponding to the Si 2p, C 1s, N 1s, and O 1s core levels, respectively. Unfortunately, hydrogen was undetectable due to limitations of the XPS technique. All peaks obtained were single component signatures, absent of any evidence of superposition of neighboring peaks with closely spaced energies. Since the XPS tec hnique is extremely surface sensitive , the results were used to identify chemical speci es in the first few m onolayers of material. The N 1s peak was only visible on the (F G)*-treated surface, and absent on the RCA cleaned surface. The presence of a sm all N 1s peak in the treated surface XPS spectra provided evidence of nitrogen in corporation during the (FG)* afterglow conditioning of 4H-SiC. The O 1s peak s howed a relatively higher intensity on the (FG)*-conditioned surface compared to the RCA cleaned surface. The presence of oxygen on the RCA cleaned surface was primarily a result of native oxide formation and oxygen adsorption after the wet chemical cl eaning procedure and during shipping. The larger percentage of oxygen observed on the (FG)*-treated surface was suggested to be a direct result of the afterglow processi ng unloading procedure. Following (FG)*treatment, heated samples were unloaded from the furnace and allowed to cool in cleanroom air. The 600C furnace unloading temperature was far below the previously mentioned threshold of 950C required for ther mal oxidation of SiC. Nonetheless, the
96 exposure of hot substrates to the oxygen cont ent in air supports an increased amount of oxygen adsorption at the surface. Percentage s of both C 1s and Si 2p after the (FG)* treatment were less than the RCA clean, primarily due to th e increase in both oxygen and nitrogen content observed on the treated surface. Since the total percentage of all species was by definition 100, an increase of the O and N peaks required a decrease in percentages of other species (C and Si) on the (FG)*-tr eated surface. Additional significance was placed on the changes in ratios of species rather than absolute atomic percentages. The C/Si ratio was 0.978 on the RCA cleaned surface, and decreased to 0.891 on the surface conditioned in (FG)* afterglo w. The reduction of C/Si ratio gave strong evidence in support of the removal of carbon from the SiC surface layers during (FG)* afterglow exposure. Carbon removal coul d contribute to the formation of a Si-rich surface region, but also could be part of the nitridation process responsible for the observed nitrogen incorporation at the surface Thus the XPS spectra were consistent with a surface chemistry enriched in both Si and N achieved during afterglow conditioning. Following XPS measurement of as-received surfaces, the samples were sputtered and remeasured. Sputtering for 2 min. with a 5 keV Ar+ source was estimated to remove roughly 4 of material, on the order of one monolayer. Figure 5.4 contains an XPS comparison of the (FG)*-treated SiC surface befo re and after sputtering. The significant changes in XPS spectra caused by sputtering were a decrease of both nitrogen and oxygen percents, and a correspond ing increase in Si and C content. Nitrogen atomic percent was reduced from 4.74 to 1.51 after sp uttering. The presence of the N 1s peak even after sputtering implied some degree of nitridation and incorpor ation of N below the
97 a) b) Figure 5.4. XPS spectral data (a ) and atomic percent values (b) obtained on n-type 4HSiC surfaces as treated by (FG)* afterglow, and after sputtering. Sputtering for 2 min. with 5 keV Ar+ source was estimated to remove 4 of material from the sample surface.
98 a) b) Figure 5.5. XPS spectral data (a ) and atomic percent values (b) obtained on n-type 4HSiC surfaces as treated by RCA clean, and af ter sputtering. Sputtering for 2 min. with 5 keV Ar+ source was estimated to remove 4 of material from the sample surface.
99 topmost monolayer(s) during (FG)* treatment. The N 1s peak could not be dismissed as simply adventitious contamination. Both C and Si percentages incr eased after sputtering, as required by the aforementioned decreases in nitrogen and oxygen content. However, the C/Si ratio remained almost unchanged at 0.894, compared to 0.891 before sputtering. Even after material removal by sputtering, the relative car bon content remained lower, suggesting that the (FG)* afterglow effectiv ely removed carbon at least from the top several monolayers. Figure 5.5 displays the XPS data obtained before and after sputtering of the RCA cleaned 4H-SiC surface. Again, the N 1s peak was not present on the RCA cleaned surface. The O 1s peak was reduced after sp uttering, with corresponding slight increases in both Si and C atomic percents. In this case, the C percentage increased more than the Si, yielding an increase in th e C/Si ratio from 0.978 to 1.057 after sputtering. This was partially attributed to rem oval of roughly a monolayer of native oxide during sputtering, thus decreasing the relative Si (a nd O) content in the surface layers. Figure 5.6 shows a comparison between th e RCA clean and (FG)* afterglow treated surfaces, both after sputtering. In fact, the XPS differences were generally similar to the comparison of the surfaces before sputtering (f igure 5.3). A small N 1s peak was visible on the (FG)*-treated sample after sputteri ng, which was not present on the RCA cleaned sample. The oxygen percentage was also hi gher for the (FG)* conditioning than the RCA clean. However, the Si percent on th e (FG)*-treated sample was essentially identical to the RCA clean. This required a reduction in carbon percent to offset the small increases in nitrogen a nd oxygen content on the (FG)*-treated sample. As a result,
100 a) b) Figure 5.6. XPS spectral data (a ) and atomic percent values (b) obtained after sputtering of n-type 4H-SiC surfaces treated by RCA cl ean or (FG)* afterglow. Sputtering for 2 min. with 5 keV Ar+ source was estimated to remove 4 of material from the sample surface.
101 the C/Si ratio was significantly higher for the RCA clean (C/Si = 1.057) compared to the afterglow treatment (C/Si = 0.894). Thus, afte r sputtering removal of ~4 of material, the (FG)* afterglow treated sample still demo nstrated strong indica tions of Si-enriched surface chemistry and some degree of nitrogen incorporation which were not evident on the RCA cleaned sample after sputtering. All atomic percentage values and ratios calculated by XPS analysis are summarized in table 5.1. Table 5.1. XPS atomic percent and ratios of selected elements obtained on n-type 4HSiC surfaces treated by RCA clean or (F G)* afterglow conditioning, before and afte r sputtering. Sputtering for 2 min. with 5 keV Ar+ source was estimated to remove 4 o f material from the sample surface. XPS atomic % RCA clean (FG)* afterglow surface sputtered surface sputtered Si 2p 44.33 45.39 38.50 45.61 C 1s 43.37 47.96 34.29 40.78 N 1s 4.74 1.51 O 1s 11.90 6.41 22.47 11.34 C / Si 0.978 1.057 0.891 0.894 O / Si 0.268 0.141 0.584 0.249 The observed effects of (FG)* treatment on surface chemistry were consistent with the results of chapter 3 demonstrating the im pact of surface conditioning on subsequent oxide film formation. An oxide film th ickness increase of 1020% was observed when the 4H-SiC surfaces were conditioned via (FG)* afterglow prior to oxidation. This increase in growth rate co rresponded to approximately 225 additional monolayers of oxide film growth (~3.3 /layer), dependi ng on growth time and temperature. The
102 preparation of a Si-rich surf ace region by afterglow conditioni ng, at least through the first several monolayers, might have contributed to additional oxide grow th considering that the presence of carbon inhibits the oxidation process on SiC due to the higher C-Si bond energy. The removal of carbon and incorporation of nitrogen were also consistent with depletion surface barrier decay measurements presented in chapter 4. Afterglow surface conditioning was effective at reducing the rate of VSB decay, a result which was attributed to a reduction in the am ount of charge emission from va rious surface defects. The XPS spectra obtained on (FG)*-condi tioned surfaces suggested the elimination and passivation of C-related defects in the surface region as a contributing factor in the reduction of surface charge-emitting defect concentrations.
103 Chapter 6. Conclusion 6.1. Results summary Surface conditioning and oxidation of the wide band-gap semiconductor 4H-SiC were investigated using a novel sequential af terglow processing approach combined with the unique capabilities of noncontact corona-Kelvin metr ology. The use of remote plasma assisted thermal oxidation facilitated fi lm growth at low temperature and pressure with the flexibility of sequential in-situ processing options in cluding pre-oxidation surface conditioning. Corona-Kelvin metrology provided a fast, non-destructive method for electrical evaluation of oxide films and semiconductor surfaces. Treatment in forming gas (N2:H2)* 19:1 afterglow was used to condition the SiC surface prior to remote plasma oxidation by (O2:N2O:N2:H2)* chemistry. The pre-oxidation (FG)* afterglow treatment step was found to have a significant impact on resulting oxide film thickness and thickness uniformity as de termined by non-contact C-KM oxide capacitance-voltage measurements (chapter 3) Direct measurement of SiC surfaces for various treatment conditions was accomplis hed using non-contact C-KM depletion surface barrier decay monitoring (chapter 4) and XPS analysis of surface chemistry (chapter 5). Results were interpreted relating the impact of afterglow conditioning on the surface and its influence on subsequent oxide th in film growth as part of a sequential afterglow oxidation approach to the promising SiC material.
104 The results of oxide film characterization were presented in chapter 3 for afterglow oxides grown at various temperatures ( 600850C) and time intervals (1090 min.). Prior to oxidation, an optional surface condition ing step exposed the 4H-SiC samples to (FG)* afterglow media for 20 min. at 600C. Oxide thickness (EOT) values were extracted from C-KM capacitance characteris tics. Preparation of the surface in (FG)* afterglow was found to produce an oxide growth rate increase (10-20% thicker films) and an improvement in oxide uniformity. The growth rate and uniformity improvements were more prominent at higher oxide grow th temperatures. Based on the surface measurement findings of chapters 4 and 5, pa st experimental resu lts, and supplemental information from the literature, the (F G)* afterglow conditioning of 4H-SiC was suggested to accomplish a combination of pa ssivation of surface states and defects, removal of carbon and eliminati on of C-related defects, Si -enrichment of the surface layers, cleaning and removal of surface cont aminants, nitrogen incorporation, and reduction of surface roughness. These cond itioning effects were among the factors contributing to the observed impact on subs equent oxide formation. The removal of carbon during surface conditioning was particul arly expected to enhance the oxide growth rate. Surface defect passivation, contaminant removal, and smoothing were suggested to improve the oxide film uniformity. Post-oxida tion high temperature inert Ar annealing was also used to test the therma l stability of oxide films and examine the possible impact of surface conditioning on said stability. Inert annealing at 950C or 1000C after oxide growth at 600C or 850C, respectively, was found to cause degradation of oxide qualit y evidenced by an increase in the flat-band voltage (VFB) and the calculated net total oxide charge (QTOT). Pre-oxidation (FG)* surface conditioning did
105 not have any noticeable impact on the QTOT values before or after annealing. Interfacial rearrangement or viscous oxide flow were s uggested as possible factors contributing to the oxide charge increa se during annealing. Non-contact electrical ev aluation of SiC surfaces af ter variations of (FG)* afterglow and other treatments were presente d in chapter 4. C-KM measurements were accomplished by deposition of a large pulse of majority charge (depleting the semiconductor) and subsequent monitoring of the VCPD transient. The observed VCPD decay after depletion charging wa s associated with a decay of the surface potential barrier (VSB). Depletion surface barrier decay was attributed to a charge compensation mechanism, suggested to consis t primarily of majority carrie r emission from traps in the surface region and recombination with dopant i ons in the space-charge region deeper in the crystal. The electric field resulting fr om surface corona charging was expected to assist in the carrier emission and transp ort processes. The charge compensation mechanism involved a net transport of majori ty charge away from the surface and into the semiconductor, with corresponding comp ensation of surface corona charge and decreases in depletion region width and surf ace barrier height. The observed rate of VSB decay was taken as a measure of surface charge -emitting defect concentrations. Also, the standard deviation of decay behavior averag ed from multiple wafer sites was used as a measure of uniformity of surface condition. Surface conditioning in (N2:H2)* afterglow was determined to result in the lowest rates of VSB decay and least deviation of decay behavior across th e wafer, when compared to other treatments including RCA clean, N2:O2 thermal treatment, pure (N2)* afterglow, and (N2:H2)* afterglow followed by DI water rins e, although the latter two were only
106 slightly worse than the (FG)* treatment. T hus, among the various treatments considered, (FG)* afterglow was concluded to achieve th e lowest concentrations of surface chargecompensating defects and highest uniformity of surface condition. Various treatment times ranging 2.520 min. of (FG)* afterglow we re investigated, re vealing a trend of decreasing VSB decay rate and improved uniformity with increasing treatment time. The longest treatment time of 20 min. yielded th e lowest levels of surface charge-emitting defects and best uniformity of surface cond ition. Various (FG)* afterglow treatment temperatures were also investigated. A te mperature threshold of conditioning effects was observed, with surfaces treated at 40 0500C demonstrating high rates of VSB decay, and treatments at higher temperatures 600800 C exhibiting very low decay rates and improved uniformity. Thus a (FG)* trea tment temperature 600C or above was concluded to be necessary to achieve eff ective conditioning and surface passivation. The state of the SiC surface achieved during (FG)* afterglow conditioning proved to be stable and resilient when exposed to a variety of post-treatment stresses, including temporal, thermal, electrical, and environmental. The effective passivation of surface chargeemitting traps and improved uniformity of surface condition accomplished by (FG)* afterglow treatment were suggested to contri bute to the oxide growth rate increase and greater film uniformity observed during subsequent SiC oxidation (chapter 3). X-ray photoelectron spectroscopy analysis of SiC surfaces was performed to evaluate the impact of (FG)* afterglow conditioning on surface chemistry. The XPS results of chapter 5 compared 4H-SiC su rface chemistry after RCA clean or (FG)* afterglow treatment. Surfaces were measured as-received and after sputtering estimated to remove 4 of material. Prominent bind ing energy peaks of Si 2p, C 1s, N 1s, and O
107 1s core levels were identified and atomic percents were computed. There was no indication of nitrogen on the RCA cleaned samp le, while a small N 1s peak was evident on the (FG)*-treated surface which implie d nitrogen incorporation during (FG)* afterglow conditioning. The N 1s peak was redu ced but still present after sputtering of the (FG)*-treated surface. O xygen was present after both treatments, attributed to native oxide and contaminant oxygen adsorption durin g post-treatment air exposure. A higher atomic percent of oxygen was found on the (F G)*-treated surface, presumably due to wafer unloading at 600C following afterglow conditioning and increased oxygen adsorption to the heated surfaces in air. Th e C/Si ratio, computed from atomic percents of C 1s and Si 2p peaks, demonstrated a significant decrease on the (FG)*-conditioned surface compared to the RCA clean. The lowe r C/Si was linked to the removal of carbon resulting in a Si-rich surface region duri ng afterglow conditioning. This silicon enrichment was suggested to be an important contributor to the in creased oxide growth rate observed after (FG)* surface conditioning in chapter 3. Furthermore, the removal of C-related defects combined with possible nitr ogen passivation of def ects were consistent with the reduction of surface defect concentrations achi eved during (FG)* afterglow conditioning in chapter 4. The formation of thermally grown oxi de films on 4H-SiC at low growth temperatures of 600850C was a unique cont ribution of this work. This was accomplished only by the use of afterglow chemistry, and would have been impossible using a conventional atmospheric oxidation appr oach. Also, this work was the first to develop an afterglow surface conditioning pr ocess on 4H-SiC specifically to improve subsequent thermal oxidation. This novel contribution was augmented by the
108 demonstration of the passivating nature of the afterglow conditioning process with respect to depletion charge co mpensation in 4H-SiC material. 6.2. Future work There exist numerous possibili tes for directions of future investigation based on this work. All oxidation experiment s presented in this work involved an RCA pre-furnace clean with an optional afterglow surface condi tioning step using baseline parameters of 20 min. (FG)* at 600C. Additional oxidations could be performed using pre-oxidation surface treatments with variations in media, temperature, and time, similar to those of chapter 4. The standard oxidative media used in this work was (O2:N2O:FG)* 3:0.23:0.5 afterglow. Experiments using (O2:N2O)* with varying amounts of FG would help in illuminating the role of FG during oxidation, and how that depends on starting surface condition. Further experimentation is required to improve the thermal stability of afterglow oxide thin films on 4H-SiC. The flatband shif ting and oxide net ch arge increase observed during high temperature inert annealing in ch apter 3 require improvement if oxide quality is to be stable during postoxidation thermal processing. In addition to C-KM capacitance charac terization, oxide films could also be analyzed by interpreting the depletion voltage transient de cay behavior, analogous to the measurements on surfaces in chapter 4. Change s in the charge transfer process could be correlated with initial surface c ondition and oxidation parameters. XPS analysis presented in this work was taken from SiC surfaces after RCA clean or (FG)* afterglow with the baseline treatment parameters of 20 min. exposure at 600C.
109 XPS measurements of other surface conditions (media, temperature, time, etc.) would yield additional information regarding the dependence of surface chemistry on treatment variations. The majority of depletion voltage transients presented in this work incorporated a consistent charge pulse size of 2.4107 C cm2 followed by surface barrier monitoring. Further investigation might include explora tion of the V-Q relationship, using corona charge pulses of various sizes to characte rize the response of surface barrier decay behavior. Afterglow conditioning of 4H-SiC was show n to prepare a stable and passivated surface with reduced defects and contaminan ts and greater uniformity. Deposition of alternative dielectric films on such cond itioned surfaces could lead to achieving an improvement in interface quality compar ed to thermally grown oxides on 4H-SiC.
110 References 1. E. Irene, "Models for the Oxidation of Sili con", Crit. Rev. Solid State Mater. Sci. 14, 175 (1988). 2. B. Deal and A. Grove, "General Relations hip for the Thermal Oxidation of Silicon", J. Appl. Phys. 36, 3770 (1965). 3. Y. Song, S. Dhar, L. Feldman, G. Chung, and J. Williams, "Modified Deal Grove model for the thermal oxidation of silicon carbide", J. Appl. Phys. 95, 4953 (2004). 4. K. Christiansen, S. Christiansen, M. Albrecht, H. Strunk, and R. Heibig, "Anisotropic oxidation of silicon carbi de", Diam. Rel. Mater. 6, 1467 (1997). 5. N. Wright, C. Johnson, and A. O'Neill, "M echanistic model for oxidation of SiC", Mater. Sci. Eng. B 61-62, 468 (1999). 6. J. Costello and R. Tressler, "Oxidation Kinetics of Silicon Carbide Crystals and Ceramics: I. In Dry Oxygen", J. Amer. Cer. Soc. 69, 674 (1986). 7. C. Raynaud, "Silica films on silicon carbide: a review of electric al properties and device applications", J. N on-Cryst. Sol. 280, 1 (2001). 8. J. Cooper Jr., "Advances in SiC MOS Technology", Phys. Stat. Sol. A 162, 305 (1997). 9. F. Amy, H. Enriquez, P. Soukiassian, P. Storino, Y. Chabal, A. Mayne, G. Dujardin, Y. Hwu, and C. Brylinski, "Atomic Scale Oxidation of a Complex System: O2/ -SiC(0001)-(33)", Phys. Rev. Lett. 86, 4342 (2001). 10. F. Amy, P. Soukiassian, Y. Hwu, and C. Brylinski, "Si-rich 6Hand 4H-SiC(0001) 33 surface oxidati on and initial SiO2/SiC interface formation from 25 to 650 C", Phys. Rev. B 65, 165323 (2002). 11. E. Pippel, J. Woltersdorf, H. Olafsson, and E. Sveinbjornsson, "Interfaces between 4H-SiC and SiO2: Microstructure, nanochemistry, a nd near-interface traps", J. Appl. Phys. 97, 034302 (2005). 12. V. Afanasev, M. Bassler, G. Pensl, and M. Schulz, "Intrinsic SiC/SiO2 Interface States", Phys. Stat. Sol. A 162, 321 (1997).
111 13. V. Afanasev, F. Ciobanu, S. Dimitrijev, G. Pensl, and A. Stesmans, "Band alignment and defect states at SiC/oxide interfaces", J. Phys .: Cond. Mat. 16, S1839 (2004). 14. V. Afanasev and A. Stesmans, "Valence band offset and hole injection at the 4H-, 6H-SiC/SiO2 interfaces", Appl. Phys. Lett. 77, 2024 (2000). 15. G. Lucovsky and H. Niimi, "Remote pl asma-assisted oxidation of SiC: a low temperature process for SiC-SiO2 interface formation that eliminates interfacial Si oxycarbide transition regions", J. Ph ys.: Condens. Matter 16, S1815 (2004). 16. R. Buczko, S. Pennycook, and S. Pantelides, "Bonding Arrangements at the Si-SiO2 and SiC-SiO2 Interfaces and a Possible Origin of their Contrast ing Properties", Phys. Rev. Lett. 84, 943 (2000). 17. S. Wang, M. Di Ventra, S. Kim, and S. Pantelides, "Atomic-Scale Dynamics of the Formation and Dissolution of Carbon Clusters in SiO2", Phys. Rev. Lett. 86, 5946 (2001). 18. G. Soares, C. Radtke, I. Baumvol and F. Stedile, "Morphological and compositional changes in the SiO2/SiC interface region induced by oxide thermal growth", Appl. Phys. Lett. 88, 041901 (2006). 19. J. Dekker, K. Saarinen, H. Olafsson, and E. Sveinbjornsson, "Observation of interface defects in th ermally oxidized SiC using positron annihilation", Appl. Phys. Lett. 82, 2020 (2003). 20. R. Ghosh, S. Ezhilvalavan, B. Golding, S. Mukhopadhyay, N. Mahadev, P. Joshi, M. Das, and J. Cooper Jr., "Profiling of the SiO2 SiC Interface Using X-ray Photoelectron Spectroscopy", MR S Symp. Proc. 640, H3.7.1 (2001). 21. L. I. Johansson and C. Virojanadara "Synchrotron radiat ion studies of the SiO2/SiC(0001) interface", J. Phys.: Cond. Mat. 16, S3423 (2004). 22. M. Okamoto, M. Tanaka, T. Yatsuo, a nd K. Fukuda, "Effect of the oxidation process on the electrical characteris tics of 4H-SiC p-ch annel metal-oxidesemiconductor field-effect transistor s", Appl. Phys. Lett. 89, 023502 (2006). 23. H. Yano, F. Katafuchi, T. Kimoto, and H. Matsunami, "Effects of Wet Oxidation/Anneal on Interface Properties of Thermally Oxidized SiO2/SiC MOS system and MOSFET's", Trans. Electron Dev. 46, 504 (1999). 24. E. EerNisse, Appl. Phys. Lett. 35, 8 (1979).
112 25. G. Chung, C. Tin, J. Williams, K. McDona ld, M. Di Ventra, S. Pantelides, L. Feldman, and R. Weller, "Effect of nitric oxide annealing on the interface trap densities near the band edges in the 4H polytype of silicon carbide", Appl. Phys. Lett. 76, 1713 (2000). 26. K. McDonald, L. Feldman, R. Weller, G. Chung, C. Tin, and J. Williams, "Kinetics of NO nitridation in SiO2/4H-SiC", J. Appl. Phys. 93, 2257 (2003). 27. K. McDonald, R. Weller, S. Pantelid es, L. Feldman, G. Chung, C. Tin, and J. Williams, "Characterization and modeling of the nitrogen passivation of interface traps in SiO2/4H-SiC", J. Appl. Phys. 93, 2719 (2003). 28. G. Chung, J. Williams, K. McDonald, and L. Feldman, "4H-SiC oxynitridation for generation of insulating layers", J. Phys.: Condens. Matter 16, S1857 (2004). 29. C. Lu, J. Cooper Jr., T. Tsuji, G. Chung, J. Williams, K. McDonald, and L. Feldman, "Effect of Process Variations and Ambient Temperature on Electron Mobility at the SiO2/4H-SiC Interface", Trans. Electron Dev. 50, 1582 (2003). 30. K. Fujihira, Y. Tarui, M. Imaizumi, K. Ohtsuka, T. Takami, T. Shiramizu, K. Kawase, J. Tanimura, and T. Ozeki, "Cha racteristics of 4H-SiC MOS interface annealed in N2O", Solid State Electron. 49, 896 (2005). 31. K. Ellis and R. Buhrman, "Nitrous oxide (N2O) processing for silicon oxynitride gate dielectrics", IBM J. Res. Devel. 43, 287 (1999). 32. S. Dhar, S. Wang, J. Williams, S. Pantelides, and L. Feldman, "Interface Passivation for Silicon Dioxide Layers on Silicon Carbide", MRS Bulletin 30, 288 (2005). 33. G. Chung, C. Tin, J. Williams, K. McDona ld, M. Di Ventra, R. Chanana, S. Pantelides, L. Feldman, and R. Weller, "Effects of anneals in ammonia on the interface trap density near the band e dges in 4H-silicon carbide metal-oxidesemiconductor capacitors", A ppl. Phys. Lett. 77, 3601 (2000). 34. Y. Hijikata, H. Yaguchi, S. Yoshida, Y. Ishida, and M. Yoshikawa, "Effect of Ar post-oxidation annealing on oxide4H-SiC interfaces studied by capacitance to voltage measurements and photoemission sp ectroscopy", J. Vac. Sci. Tech. A 23, 298 (2005). 35. L. Lipkin, "Method of N2O annealing an oxide layer on a silicon carbide layer", Cree, US Patent 6610366 (2003). 36. A. Hoff and J. Ruzyllo, "Atomic oxygen and the thermal oxidation of silicon", Appl. Phys. Lett. 52, 1264 (1988).
113 37. K. Sekine, Y. Saito, M. Hirayama, and T. Ohmi, "Highly Relia ble Ultrathin Silicon Oxide Film Formation at Low Temperat ure by Oxygen Radical Generated in HighDensity Krypton Plasma", Trans. Electron Dev. 48, 1550 (2001). 38. M. Togo, K. Watanabe, M. Terai, T. Ya mamoto, T. Fukai, T. Tatsumi, and T. Mogami, "Improving the Quality of Sub-1.5nm-Thick Oxynitride Gate Dielectric for FETs with Narrow Channel and Sha llow-Trench Isolation using Radical Oxygen and Nitrogen", Trans. Electron Dev. 49, 1736 (2002). 39. E. Short, "Growth of Oxide Thin Film s on 4HSilicon Carbid e in an Afterglow Reactor", Thesis, Department of Electrica l Engineering, College of Engineering, University of South Florida (2006). 40. A. Tibrewala, "Thermal Oxidation of Single Crystal Silicon Carbide in Flowing Afterglow", Thesis, Department of Electri cal Engineering, College of Engineering, University of South Florida (2002). 41. A. Hoff, E. Oborina, S. Saddow, and A. Savtchouk, "Thermal Oxidation of 4HSilicon Carbide Using the Afterglow Me thod", Mater. Sci. Forum 457-460, 1349 (2004). 42. A. Hoff, "Growth and Metrology of S ilicon Oxides on Silicon Carbide", MRS SiC Symp. Proc. (2004). 43. A. Hoff and E. Oborina, "Fast Non-C ontact Dielectric Char acterization for SiC MOS Processing", ICSCRM, 1035 (2005). 44. J. Cook and B. Benson, "Application of EPR Spectroscopy to Oxidative Removal of Organic Materials", J. El ectrochem. Soc. 130, 2459 (1983). 45. J. Spencer, R. Borel, and A. Hoff, "Hi gh Rate Photoresist Stripping in an Oxygen Afterglow", J. Electrochem. Soc. 133, 1922 (1986). 46. J. Choi, S. Kim, J. Kim, H. Ka ng, H. Jeon, and C. Bae, "Effects of N2 remote plasma nitridation on the structural a nd electrical characte ristics of the HfO2 gate dielectrics grown using remote plasma at omic layer deposition methods", J. Vac. Sci. Tech. A 24, 900 (2006). 47. Y. Maeyama, H. Yano, T. Hatayama, Y. Uraoka, T. Fuyuki, and T. Shirafuji, "Improvement of SiO2/ -SiC Interface Properties by N itrogen Radical Treatment", Mater. Sci. Forum 389-393, 997 (2002). 48. M. Losurdo, G. Bruno, A. Brown, and T. Ki m, "Study of the temperature-dependent interaction of 4H-SiC a nd 6H-SiC surfaces with at omic hydrogen", Appl. Phys. Lett. 84, 4011 (2004).
114 49. M. Losurdo, M. Giangregorio, G. Br uno, A. Brown, and T. Kim, "Study of the interaction of 4H-S iC and 6H-SiC(0001)Si surfaces with atomic nitrogen", Appl. Phys. Lett. 85, 4034 (2004). 50. M Losurdo, M. Giangregorio, P. Capezzu to, G. Bruno, A. 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. Elec. Mater. 34, 457 (2005). 51. W. Kern and D. Puotinen, "Cleaning so lutions based on hydrogen peroxide for use in silicon semiconductor tec hnology", RCA Rev. 31, 187 (1970). 52. A. Savtchouk, E. Oborina, A. Hoff, and J. Lagowski, "Non-contact Doping Profiling in Epitaxial SiC", Mate r. Sci. Forum 457-460, 1349 (2004). 53. P. Edelman, A. Savtchouk, M. Wilson, J. D'Amico, J. Kochey, D. Marinskiy, and J. Lagowski, "Non-contact C-V measurements of ultra thin dielectrics", Eur. Physic. J. Appl. Phys. Proc. (2003). 54. P. Edelman, A. Savtchouk, M. Wilson, J. D'Amico, J. Kochey, D. Marinskiy, and J. Lagowski, "Non-Contact C-V Technique fo r high-k Applications ", Intern. Conf. Char. Metrol. ULSI Tech. Proc. (2003). 55. M. Wilson, J. Lagowski, A. Savtchouk, L. Jastrzebski, and J. D'Amico, "COCOS (Corona Oxide Characterization of Semic onductor) Metrology: Physical Principles and Applications", ASTM Conf. Ga te Diel. Oxide Integ. Proc. (1999). 56. M. Wilson, J. Lagowski, A. Savtchouk, L. Jastrzebski, J. D'Amico, D. DeBusk, and A. Buczkowski, "New COCOS (Corona Ox ide Characterization of Semiconductor) Method for Monitoring the Reliability of Thin Gate Oxides", M1-Diag. Techniq. Semicond. Mater. Dev. Symp., 196th Joint Intern. Mtg. Electrochem. Soc. Proc. (1999). 57. A. Hoff and D. DeBusk, M1-Diag. Techniq. Semicond. Mater. Dev. Symp., 196th Joint Intern. Mtg. Electroc hem. Soc. Proc. (1999). 58. A. Hoff, D. DeBusk, and R. Schanzer, SPIE Proc. 3884-17 (1999). 59. P. Edelman, A. Hoff, L. Jastrzebski, and J. Lagowski, SPIE Proc. 2337, 154 (1994). 60. M. Dautrich, P. Lenahan, A. Kang, and J. Conley, Jr., "Non-invasive nature of corona charging on thermal Si/SiO2 structures", Appl. Phys. Lett. 85, 1844 (2004). 61. J. Spencer, R. Borel, K. Linxwiler, and A. Hoff, "Microwave apparatus for generating plasma afterglows", US Patent 4673456 (1987). 62. Semiconductor Diagnostics, Inc., http://www.sditampa.com.
115 63. W. Thomson (Lord Kelvin), Phil. Mag. 46, 82 (1898). 64. D. Schroder, "Surface Voltage and Surface Photovoltage: History, Theory and Applications", Meas. Sci. Tech 12, R16 (2001). 65. S. Ganichev, E. Ziemann, W. Prettl, I. Yassievich, A. Istr atov, and E. Weber, "Distinction between the Poole-Frenkel and tunneling models of electric-fieldstimulated carrier emission from deep levels in semiconductors", Phys. Rev. B 61, 15 (2000). 66. J. Frenkel, "On Pre-Breakdown Phenomen a in Insulators and Electronic SemiConductors", Phys. Rev.54, 647 (1938). 67. K. Chang, T. Witt, A. Hoff, R. Woodin, R. Ridley, G. Dolny, K. Shanmugasundaram, E. Oborina, and J. Ruzyllo, "Surface Roughness in Silicon Carbide Technology", Cleaning Tech. a nd Dev. Manuf. IX, Electrochem. Soc., T200500103, 228 (2005). 68. D. Schroder, "Semiconductor Material a nd Device Characterization", 3rd Ed., J. Wiley, New York, NY (2006). 69. CasaXPS Processing Software, http://www.casaxps.com.
About the Author Eugene L. Short, III pursued his undergraduate studies at the California Institute of Technology in Pasadena, CA, and received his B.S. in Engineering and Applied Science in 2003. He then initiated his graduate studies at the Univ ersity of South Florida in Tampa, FL, where he earned an M.S. in El ectrical Engineering in 2006, and a Ph.D. in Electrical Engineering in 2009.
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 002068253
007 cr mnu|||uuuuu
008 100412s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003102
Short, Eugene L.
Sequential afterglow processing and non-contact Corona-Kelvin metrology of 4H-SiC
h [electronic resource] /
by Eugene L. Short.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 115 pages.
Dissertation (Ph.D.)--University of South Florida, 2009.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: Silicon carbide (SiC) is a wide band-gap semiconductor with advantageous electrical and thermal properties making it attractive for high temperature and power applications. However, difficulties with oxide/SiC structures have posed challenges to the development of practical MOS-type devices. Surface conditioning and oxidation of 4H-SiC were investigated using a novel sequential afterglow processing approach combined with the unique capabilities of non-contact corona-Kelvin metrology. The use of remote plasma assisted thermal oxidation facilitated film growth at low temperature and pressure with the flexibility of sequential in-situ processing options including pre-oxidation surface conditioning. Corona-Kelvin metrology (C-KM) provided a fast, non-destructive method for electrical evaluation of oxide films and semiconductor surfaces.Non-contact C-KM oxide capacitance-voltage characteristics combined with direct measurement of SiC surfaces using C-KM depletion surface barrier monitoring and XPS analysis of surface chemistry were interpreted relating the impact of afterglow conditioning on the surface and its influence on subsequent oxide thin film growth. Afterglow oxide films of thicknesses 50-500 angstroms were fabricated on SiC epi-layers at low growth temperatures in the range 600-850C, an achievement not possible using conventional atmospheric oxidation techniques. The inclusion of pre-oxidation surface conditioning in forming gas (N:H)* afterglow was found to produce an increase in oxide growth rate (10-25%) and a significant improvement in oxide film thickness uniformity.Analysis of depletion voltage transients on conditioned SiC surfaces revealed the highest degree of surface passivation, uniformity, and elimination of sources of charge compensation accomplished by the (N:H)* afterglow treatment for 20 min. at 600-700C compared to other conditioning variations. The state of surface passivation was determined to be very stable and resilient when exposed to a variety of temporal, electrical, and thermal stresses. Surface chemistry analysis by XPS gave evidence of nitrogen incorporation and a reduction of the C/Si ratio achieved by the (N:H)* afterglow surface treatment, which was tied to the improvements in passivation, uniformity, and growth rate observed by non-contact C-KM measurements. Collective results were used to suggest a clean, uniform, passivated, Si-enriched surface created by afterglow conditioning of 4H-SiC as a sequential preparation step for subsequent oxidation or dielectric formation processing
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Andrew Hoff, Ph.D.
x Electrical Engineering
t USF Electronic Theses and Dissertations.