SURFACE AND INTERFACE CHARACTERIZATION OF SiC
AND III-V NITRIDES
by
SEAN WESLEY KING
A dissertation submitted to the Graduate Faculty of
North Carolina State University in partial fulfillment of the
requirements for the Degree of Doctor of Philosophy
MATERIALS SCIENCE AND ENGINEERING
Raleigh
1997
APPROVED BY:
_____________________ _________________________
_____________________ _________________________
Chair of Advisory Committee
ABSTRACT KING, SEAN WESLEY. Surface and Interface Characterization of SiC and III-V
Nitrides. (Under the direction of Robert F. Davis.)
The effects of various wet chemical and chemical vapor processes on the surfaces
of (0001) silicon carbide (SiC), aluminum nitride (AlN), and gallium nitride (GaN)
have been investigated using x-ray photoelectron spectroscopy (XPS), Auger electron
spectroscopy (AES), low energy electron diffraction (LEED), electron energy loss
spectroscopy (EELS), temperature programmed desorption (TPD), and x-ray
photoelectron diffraction (XPD). XPS was also used to determine both the growth
mechanism of GaN on (0001) AlN as well as the heterojunction valence band
alignment between 2H-AlN and 6H-SiC at the (0001) interface. Polycrystalline
scandium nitride (ScN) films were also grown on 3C/6H-SiC (0001) substrates by gas
source molecular beam epitaxy (GSMBE).
Thermally oxidized surfaces of (0001)Si 6H-SiC were visually observed to be
hydrophilic after removal of the oxide with HF solutions. The hydrophilic nature of
this surface was correlated with a monolayer coverage of oxygen (hydroxide)
observed on the surface by AES and XPS. A completely dry ex situ cleaning process
based on UV/ozone (O3) oxidation for adventitious carbon removal, followed by HF
vapor exposure for oxide removal was demonstrated and observed to be equivalent to
conventional wet chemical processes. Removal of the monolayer coverage of oxygen
from the (0001)Si 6H-SiC surface via chemical vapor cleaning (CVC) in a flux of
silane (SiH4) was observed to produce higher purity surfaces compared to typical
thermal desorption techniques. Low energy electron diffraction showed that (0001)Si
6H-SiC surfaces prepared by chemical vapor cleaning exhibited a (3x3)
reconstruction. X-ray photoelectron spectroscopy indicated that the (3x3)
ii
reconstruction consisted of an incomplete bilayer of silicon terminating the SiC
surface. X-ray photoelectron diffraction indicated that the stacking sequence of the
silicon bilayer was similar to that of a faulted Si (111) stacking structure. Selective
removal/etching of silicon from (3x3) (0001)Si 6H-SiC surfaces by atomic H
processes was observed by both AES and XPS.
Wet chemical processes based on HF were found to produce AlN surfaces with
the lowest oxygen whereas HCl based wet chemical processes were to produce the
lowest oxygen coverages for GaN surfaces. This observation was correlated with
XPS and AES studies which showed fluorine (chlorine) termination of AlN (GaN)
surfaces after HF (HCl) wet chemical processes. Complete desorption of F from AlN
surfaces was not observed to occur until Tsub > 800°C. Desorption of Cl from GaN
surfaces was observed to be complete by 650°C. Annealing GaN surfaces in NH3 at
800°C was observed to remove both oxygen and carbon below the detection limits of
both AES and XPS.
Using XPS, it was determined that GaN growth on (0001) AlN by GSMBE
occurs via a three dimensional Stranski-Krastonov growth mechanism at Tsub < ≈
780°C. At higher temperatures (Tsub > 800°C), the growth mechanism switches to a
layer by layer/Frank van der Merwe two dimensional growth mechanism. The
change in growth mechanism with temperature was attributed to a change from a
fully hydrogenated growth surface to an incompletely hydrogenated growth surface at
higher temperatures.
Using x-ray photoelectron spectroscopy and published theoretical valence band
density of states, a type I heterojunction valence band discontinuity of 1.4 ± 0.3 eV
was determined for the (0001) 2H-AlN/6H-SiC interface.
iii
DEDICATION
In honor of my parents:
Carrol David King and Nanci Marie Curry King
In memory of my grandmother:
Hilda Bell
iv
BIOGRAPHY
Sean Wesley King, son of Carrol David and Nanci Marie King, was born in Seneca,
South Carolina November 12, 1968. Shortly thereafter, his family moved to Stokesdale,
North Carolina where he was blessed with a younger sister, Julie Virginia King. The King
family eventually settled in the small community of Green Spring, Virginia which is the
ancestral home of many generations of this family. Sean then went on to quietly graduate
from Abingdon High School and follow in the footsteps of his father by enrolling in
Engineering at Virginia Polytechnic Institute and State University. After a brief stint as an
Aerospace Engineering major, Sean eventually found his niche in Materials Engineering and
went on to graduate summa cum laude. Shortly after completion of his Bachelors, Sean
enrolled in the Materials Science and Engineering Ph.D. program at North Carolina State
University, with Professor Robert F. Davis as his advisor. The work herein represents partial
completion of this degree.
v
ACKNOWLEDGMENTS
I would like to first express my appreciation to my advisory committee chairman, Dr.
Robert F. Davis, for his continued patience and support during the course of this study. I
would also like to express my appreciation to Dr. Robert J. Nemanich for his endless
enthusiasm and guidance in this research. Appreciation is also extended Dr. S. M. Bedair
and Dr. D. Griffis for their contributions as members of my advisory committee.
The initial stages of this research involved an abundant amount of equipment and
experimental design which without the assistance and guidance of Drs. L. Rowland, J.
Sumakeris, and J. van der Weide would have never come to completion and to whom I am
indebted. Appreciation is also expressed to Dr. J. Yates Jr., and his students for the
numerous telephone conversations regarding various aspects of surface science equipment. I
would also like to thank Mr. A. Illingsworth and Mr. J. Emerick for helping me make my
equipment designs work despite their many flaws.
I would like to thank the Office of Naval Research for its financial support of this
research and the Department of Education for its assistance through a GAANN fellowship.
Thanks are also extended to Dr. W. Perry, Dr. A.D. Batchelor, Mr. M. Bremser, Mr. E.
Carlson, Mr. R. Therrien, and Mr. D. Bray for their help with various measurements and
analysis. Appreciation is also expressed to Dr. W.R.L. Lambrecht for his guidance and
collaboration in the heterojunction valence band discontinuity effort.
Perhaps the most rewarding aspect of this venture was the chance to work and
interact with many diversified people who comprise both the Davis Laboratories and the
NCSU Surface Science laboratory. During the course of this research, many of these people
vi
have assisted and encouraged me through the highs and lows and with whom special
friendships have been formed including: Dr. J. Barnak, Mr. M. Benjamin, Dr. L. Bergman,
Mr. R. Busby, Mr. R. Carter, Mr. J. Christman, Mr. H. Ham, Dr. R.S. Kern, Dr. Ja-Hum Ku,
Dr. T. Schnieder, Mr. A. Sowers, Dr. S. Tanaka, Mr. K. Tracy, Mr. B. Ward, Mr. W. Yang,
Mrs. H. Ying, and Mr. S. Wagoner.
Finally, I would like to recognize the unwavering support and patience of my parents
and family during the completion of this endeavor.
vii
TABLE OF CONTENTS
LIST OF TABLES .................................................................................................... xiv
LIST OF FIGURES .................................................................................................. xvi
1. Introduction.................................................................................................. 1
2. Wet Chemical Processing of (0001)Si 6H-SiC: Hydrophobic and
Hydrophilic Surfaces...................................................................... 5
2.1 Abstract............................................................................................. 6
2.2 Introduction....................................................................................... 7
2.3 Experimental Procedure..................................................................... 10
2.4 Results............................................................................................... 12
2.4.1. (0001)Si 6H-SiC, Oxidized.......................................... 12
2.4.2. (000-1)C, (11-20), & (10-10) 6H-SiC, Oxidized......... 20
2.4.3. (0001)Si 6H-SiC, As-polished..................................... 22
2.4.4. Si Passivation Layer.................................................... 26
2.5 Discussion.......................................................................................... 28
2.5.1. (0001)Si 6H-SiC, Oxidized ......................................... 28
2.5.2. (0001)Si 6H-SiC, As-polished..................................... 35
2.5.3. Si Passivation Layer.................................................... 37
2.6 Conclusions........................................................................................ 39
2.7 Acknowledgments.............................................................................. 39
2.8 References.......................................................................................... 40
3. Dry Ex Situ Cleaning Processes for (0001)Si
viii
6H-SiC Surfaces.............................................................................. 47
3.1 Abstract............................................................................................. 48
3.2 Introduction....................................................................................... 49
3.3 Experimental Procedure..................................................................... 51
3.4 Results............................................................................................... 54
3.4.1. Solvents and UV/O3 Oxidation................................... 54
3.4.2. HF vapor..................................................................... 58
3.5 Discussion.......................................................................................... 59
3.5.1. UV/O3 oxidation......................................................... 59
3.5.2. HF Vapor.................................................................... 63
3.6 Conclusions........................................................................................ 66
3.7 Acknowledgments.............................................................................. 66
3.8 References.......................................................................................... 67
4. Chemical Vapor Cleaning of (0001)Si, (000-1)C, (11-20), and (10-10)
6H-SiC surfaces............................................................................... 70
4.1 Abstract............................................................................................. 71
4.2 Introduction....................................................................................... 72
4.3 Experimental Procedure..................................................................... 75
4.3.1. Integrated Surface Prep. and Analysis System........... 75
4.3.2. Substrate Preparation.................................................. 78
4.4 Results............................................................................................... 79
4.4.1. (0001)Si 6H-SiC.......................................................... 79
4.4.2. (000-1)C 6H-SiC......................................................... 86
ix
4.4.3. (11-20) and (10-10) 6H-SiC........................................ 91
4.4.4. Low vacuum CVC/LPCVD clean............................... 92
4.5 Discussion.......................................................................................... 93
4.5.1. (0001)Si 6H-SiC.......................................................... 93
4.5.2. (000-1)C 6H-SiC......................................................... 96
4.5.3. (11-20) and (10-10) 6H-SiC........................................ 98
4.6 Conclusions........................................................................................ 99
4.7 Acknowledgments.............................................................................. 99
4.8 References.......................................................................................... 100
5. X-ray Photoelectron Diffraction of (3x3) and (√3x√3)R30°
6H-SiC (0001)Si surfaces............................................................... 105
5.1 Abstract............................................................................................. 106
5.2 Introduction....................................................................................... 107
5.3 Experimental...................................................................................... 116
5.4 Results............................................................................................... 118
5.4.1. (v3xv3)R30° (0001)Si 6H-SiC..................................... 127
5.4.2. (3x3) (0001)Si 6H-SiC.................................................. 130
5.5 Discussion.......................................................................................... 132
5.5.1. (v3xv3)R30° (0001)Si 6H-SiC..................................... 132
5.5.2. (3x3) (0001)Si 6H-SiC.................................................. 133
5.6 Conclusions........................................................................................ 136
5.7 Acknowledgments.............................................................................. 136
5.8 References.......................................................................................... 137
x
6. Interaction of Atomic Hydrogen with (3x3) 6H-SiC (0001)Si surfaces. 140
6.1 Abstract............................................................................................. 141
6.2 Introduction....................................................................................... 142
6.3 Experimental...................................................................................... 146
6.4 Results............................................................................................... 149
6.4.1. Interaction with rf atomic H........................................ 149
6.4.2. Interaction with thermal atomic H............................... 155
6.5 Discussion.......................................................................................... 159
6.6 Conclusions........................................................................................ 173
6.7 Acknowledgments.............................................................................. 173
6.8 References.......................................................................................... 174
7. Ex Situ and In Situ Methods for Oxide and Carbon Removal from
(0001) AlN and GaN Surfaces........................................................ 179
7.1 Abstract............................................................................................. 180
7.2 Introduction....................................................................................... 180
7.3 Experimental Procedure..................................................................... 185
7.3.1. Integrated Surface Prep. and Analysis System............ 185
7.3.2. Samples and Ex Situ Preparation................................. 188
7.4 Results............................................................................................... 189
7.4.1. Ex Situ Cleaning of AlN.............................................. 189
7.4.2. In Situ Cleaning of AlN.............................................. 200
7.4.3. Ex Situ Cleaning of GaN............................................ 209
7.4.4. In Situ Cleaning of GaN............................................. 221
xi
7.4.5. Ex Situ Cleaning of AlxGa1-xN................................... 227
7.5 Discussion.......................................................................................... 229
7.5.1. As Recieved and UV/O3 Surfaces............................... 229
7.5.2. Wet Chemical and HF Vapor Processing................... 234
7.5.3. Thermal Desorption and Capping Layers.................. 240
7.5.4. Chemical Vapor Cleaning and H Plasma.................... 245
7.6 Conclusions........................................................................................ 249
7.7 Acknowledgments.............................................................................. 249
7.8 References.......................................................................................... 250
8. X-ray Photoelectron Spectroscopy Analysis of GaN/AlN and AlN/GaN
Growth Mechanisms....................................................................... 258
8.1 Abstract............................................................................................. 259
8.2 Introduction....................................................................................... 260
8.3 Experimental...................................................................................... 262
8.3.1. Integrated Growth Analysis System........................... 262
8.3.2. Substrate and Thin Film Preparation........................... 264
8.3.3. Growth Mode Analysis............................................... 266
8.4 Results............................................................................................... 268
8.4.1. GaN Growth on (0001) AlN....................................... 268
8.4.2. AlN Growth on (0001) GaN....................................... 274
8.5 Discussion.......................................................................................... 276
8.5.1. GaN Growth Mechanism on (0001) AlN.................... 276
8.5.1.1 Strain Effects...................................... 276
xii
8.5.1.2. Hydrogen Desorption....................... 279
8.5.2. AlN Growth Mechanism on (0001) GaN.................... 287
8.5.3. Surface Reconstruction................................................ 287
8.6 Conclusions........................................................................................ 288
8.7 Acknowledgments.............................................................................. 289
8.8 References.......................................................................................... 289
9. Interface Chemistry and Electronic Structure for the
(0001) 2H-AlN/6H-SiC Interface....................................................293
9.1 Abstract............................................................................................. 294
9.2 Introduction....................................................................................... 294
9.3 Experimental...................................................................................... 295
9.4 Theory............................................................................................... 296
9.5 Results............................................................................................... 297
9.6 Discussion.......................................................................................... 302
9.7 Conclusions........................................................................................ 303
9.8 Acknowledgments.............................................................................. 304
9.9 Addendum.......................................................................................... 304
9.10 References.......................................................................................... 305
10. Dependence of (0001) GaN/AlN Valence Band Discontinuity on
Surface Reconstruction and Growth Temperature..................... 307
10.1 Abstract............................................................................................. 308
10.2 Introduction....................................................................................... 308
10.3 Experimental...................................................................................... 309
xiii
10.3.1. Thin Growth and Analysis........................................ 309
10.3.1. GaN/AlN ?Ev Analysis............................................. 311
10.4 Results............................................................................................... 313
10.5 Discussion.......................................................................................... 316
10.6 Conclusions........................................................................................ 320
10.7 Addendum.......................................................................................... 320
10.8 Acknowledgments.............................................................................. 322
10.9 References.......................................................................................... 322
11. Gas-Source Molecular Beam Epitaxy growth of ScN on (111)/(0001)
3C and 6H-SiC Substrates............................................................. 324
11.1 Abstract............................................................................................. 325
11.2 Introduction....................................................................................... 325
11.3 Experimental Procedures.................................................................... 327
11.4 Results............................................................................................... 329
11.4.1. Thermodynamics....................................................... 329
11.4.2. Growth....................................................................... 333
11.5 Discussion.......................................................................................... 337
11.6 Conclusions........................................................................................ 338
11.7 Acknowledgments.............................................................................. 338
11.8 References.......................................................................................... 338
xiv
List of Tables
Table 1.1. Selected materials properties of Si, GaAs, 6H-SiC, 2H-GaN and 2H-AlN...................................................................................... 3 Table 2.1. Binding energy (in eV) of core level positions from (0001)Si 6H-SiC as polished and oxidized surfaces......................................... 18 Table 2.2. Summary of XPS Si2p/O1s, Si2p/F1s, C/C, and C-C data for (0001)Si 6H-SiC surfaces............................................................. 19 Table 2.3. Summary of wetting characteristics of as polished and oxidized (0001)Si 6H-SiC and (111) Si.......................................................... 20 Table 2.4. Peak to peak height (pph) ratios for various 6H-SiC surfaces................ 22 Table 3.1. XPS core level binding energies (eV) from (0001)Si 6H-SiC surfaces after various exposures........................................................ 57 Table 3.2. The XPS core level intensity ratios from (0001)Si 6H-SiC after various treatments .................................................................... 58 Table 3.3. Summary of SiC-C1s/surface C1s and Si/O intensity ratios from XPS data.................................................................................... 62 Table 5.1. Expected forward scattering/focusing peaks from bulk terminated (111) Si, (111) 3C-SiC, and (0001)Si 6H-SiC................. 127 Table 5.2. Expected Si 2p and C 1s photoelectron diffraction peaks for adatom scattering in T4 and H3 positions......................................... 133 Table 5.3. Estimated forward scattering peaks for (3x3) reconstructed (111)/(0001) 3C/6H-SiC surfaces...................................................... 135 Table 6.1. Kinetic Parameters used to model hydrogen adsorption/ desorption from Si and C sites on SiC surfaces................................. 162
xv
Table 7.1. OKLL/NKLL and CKLL/NKLL AES pph ratios from OMVPE AlN surfaces given various wet chemical treatments following a UV/O3 oxidation............................................................. 195 Table 7.2. XPS core level positions and full width half maxima (Γ, FWHM) from a GSMBE AlN surface after dipping in 10:1 BHF and annealing at various temperatures...................................................... 197 Table 7.3. XPS core level positions from a 200Å GaN capping layer on AlN after annealing at various temperatures................................ 206 Table 7.4. Ratio of integrated intensity of Al 2p to Ga 3d and Ga 2p3/2.................. 206 Table 7.5. XPS core level positions for OMVPE GaN surfaces after various ex situ treatments (Γ= FHWM)............................................ 211 Table 7.6. AES pph ratios of UV/O3 and wet chemical processed OMVPE GaN surfaces...................................................................... 216 Table 7.7. XPS core levels from GSMBE GaN surfaces after various treatments... 222 Table 7.8. Ga and N core levels from GaN relative to the GaN VBM after various processes....................................................................... 226 Table 7.9. Bond energies of Cl, F, and H with Al, Ga, and N.................................. 235 Table 9.1. Valence-band maxima and core levels measured on the same Au 4f7/2 based reference scale............................................................ 301 Table 9.2. Various measurements of Si2p-Al2p from a AlN/6H-SiC...................... 305 Table 10.1. Published data for GaN/AlN ?Ev.......................................................... 309 Table 10.2. Al 2p and N 1s core levels referenced to AlN VBM............................. 315 Table 10.3. CL-VBM data for AlN and GaN reported by various investigators..... 317 Table 11.1. Properties of ScN................................................................................... 327 xvi
List of Figures
Figure 2.1. Particle accumulation on removal from solution for a hydrophilic surface and a hydrophobic surface..................................................... 9 Figure 2.2. AES survey spectra of (a) (0001)Si 6H-SiC after thermal oxidation and removal of the oxide with 10:1 HF, (b) (0001)Si 6H-SiC as polished surfaces after solvent cleaning........................................ 13 Figure 2.3. XPS spectra of the F 1s core level from an oxidized (0001)Si 6H-SiC surface after (a) oxide removal with 10:1 HF, followed by (b) rigorous rinsing in running DI water............................................ 14 Figure 2.4. XPS spectra of the O 1s core level from (a) (0001)Si 6H-SiC after removal of a 750Å thermal oxide with 10:1 HF, and (b) (111) Si after a dip in 10:1 HF..................................................... 15 Figure 2.5. Typical XPS spectra of the C 1s core level from (a) oxidized (0001)Si 6H-SiC after oxide removal with 10:1 HF, (b) as-polished (0001)Si 6H-SiC after solvent cleaning, and (c) as-polished (0001)Si 6H-SiC after RCA SC1.............................. 17 Figure 2.6. EELS spectra of (0001)Si 6H-SiC (a) oxidized followed by 10:1 HF, (b) as-polished with only solvent cleaning......................... 18 Figure 2.7. The AES survey spectra of various 6H-SiC surface orientations after removal of a thermal oxide using 10:1 HF............ 21 Figure 2.8. XPS spectra of the F 1s core level from as-polished (0001)Si 6H-SiC, (a) after solvent cleaning, (b) Piranha etch, and (c) RCA SC1............................................................................... 25 Figure 2.9. AES survey spectra of as-polished (0001)Si 6H-SiC after (a) solvent cleaning, (b) a 10:1 HF dip, (c) 5 min. Piranha etch, and (d) 30 min. RCA SC1 clean........................................................ 25 Figure 2.10. (a) AES spectrum from 20Å a-Si/(0001)Si 6H-SiC after a 10:1 HF dip. (b) after thermal desorption of Si passivation layer at 1100°C.................................................................................. 27 Figure 2.11. XPS spectra of Si 2p core level from silicon passivated
xvii (0001)Si 6H-SiC (a) before thermal desorption and
(b) after thermal desorption at 1100°C.............................................. 28 Figure 2.12. (a) Schematic of (0001)Si 6H-SiC surface after thermal oxide removal with 10:1 HF. (b) Schematic of as-polished (0001)Si 6H-SiC................................................................................ 29 Figure 2.13. (a) Schematic illustrating mechanism of hydrogen termination of silicon in HF solutions. (b) Schematic illustrating stability of F- or OH- termination of SiC in HF solutions rather than H termination. (c) Schematic illustration of crystal potential in SiC.... 33 Figure 2.14. Energy diagram of (a) Si and (b) SiC in aqueous solutions................. 34 Figure 3.1. (a) Schematic of UV/ O3 oxidation system. (b) Schematic of HF vapor procedure...................................................................... 53 Figure 3.2. XPS of the C 1s core level from (0001)Si 6H-SiC after (a) solvent cleaning, sequentially followed by (b) UV/O3, and (c) HF vapor exposures............................................................... 56 Figure 3.3. XPS of the F1s core level from (0001)Si 6H-SiC after (a) solvent cleaning, sequentially followed by (b) UV/O3, and (c) HF vapor exposures............................................................... 56 Figure 3.4. XPS of the Si 2p core level from (0001)Si 6H-SiC after (a) solvent cleaning, followed by (b) UV/O3, and (c) HF vapor exposures..................................................................... 57 Figure 3.5. Schematic illustrating the surface termination from UV/O3 and HF vapor exposures on (0001)Si 6H-SiC................................... 60 Figure 3.6. Schematic illustrating the mechanism for F- and OH- termination of (0001)Si 6H-SiC........................................................ 65 Figure 4.1. AES of (0001)Si 6H-SiC surfaces after (a) 200£ SiH4 at 750°C, (b) 200£ SiH4 at 820°C, and (c) 200£ SiH4 at 880°C....................... 80 Figure 4.2. LEED patterns from (a) HF dipped (0001)Si 6H-SiC, (b) (1x1) (0001)Si 6H-SiC, (c) (3x3) (0001)Si 6H-SiC, (d) (v3xv3)R30° (0001)Si 6H-SiC, (e) (1x1) (000-1)C 6H-SiC, (f) (11-20) 6H-SiC, (g) (10-10) 6H-SiC............................................ 83 xviii
Figure 4.3. XPS spectra of the Si 2p core level from a (3x3) reconstructed (0001)Si 6H-SiC surface.................................................................... 84 Figure 4.4. XPS spectra of the Si 2p core level from (3x3), (1x1) and (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces...................... 84 Figure 4.5. AES survey spectra from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared by (a) SiH4 CVC, and (b) thermal desorption....................................................................... 85 Figure 4.6. XPS spectra of the C 1s core level from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared by (a) SiH4 CVC, and (b) thermal desorption........................................ 85 Figure 4.7. AES survey spectra from (0001)C 6H-SiC surfaces prepared by (a) thermal desorption, (b) SiH4 CVC, and (c) SiH4/C2H4 CVC..... 87 Figure 4.8. EELS spectra from (000-1)C 6H-SiC surfaces after (a) annealing in UHV at 1050°C, (b) annealing in 2000£ SiH4 at 1050°C followed by (c) annealing in 2000£ C2H4 at 950°C.......................... 88 Figure 4.9. XPS spectra of the C 1s core level from (000-1)C 6H-SiC surfaces after (a) 2000£ SiH4 at 1050°C, (b) thermal desorption in UHV at 1050°C.............................................................................. 89 Figure 4.10. EELS spectrum from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surface prepared by annealing in UHV at 1050°C.. 89 Figure 4.11. EELS spectra of (3x3), (1x1), and (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared via SiH4 CVC............................. 90 Figure 4.12. EELS spectra from (000-1)C 6H-SiC surfaces after (a) 2000£ SiH4 at 1000°C, (b) 400£ C2H4 at 850°C, and (c) 800£ C2H4 at 850°C............................................................................................. 90 Figure 4.13. EELS spectra of (10-10) 6H-SiC (a) after annealing in SiH4 at 1000°C, and (b) then annealing in C2H4 at 850°C.......................... 92 Figure 5.1. Schematic illustrating forward focusing/scattering effects xix
in x-ray photoelectron diffraction experiments.................................. 108 Figure 5.2. Schematics illustrating various adatom adsorption sites for (v3xv3)R30° reconstructions on (111)/(0001) surfaces. (a) Top down view along [000-1], (b) Side view along [11-20]........ 111 Figure 5.3. Model proposed by Kulalov et al for the (3x3) reconstructed (0001)Si 6H-SiC surface. (a) Top down view along [000-1], (b) side view along [11-20], and (c) [10-10]...................................... 113 Figure 5.4 Model proposed by Kaplan for the (3x3) reconstructed (0001)Si 6H-SiC surface. (a) Top down view along [000-1], (b) side view along [11-20]................................................................ 114 Figure 5.5. Top down view of model proposed by Li and Tsong for the (3x3) reconstructed (0001)Si 6H-SiC surface......................... 115 Figure 5.6. XPD pattern from (2x1) Si (100) along the [110] azimuth.................... 118 Figure 5.7. Si 2p x-ray photoelectron diffraction pattern along [1-21]/[10-10] azimuths from (a) (7x7) Si (111), (b) (3x3) 6H-SiC (0001)Si, and (c) (v3xv3)R30° (0001)Si 6H-SiC............................................... 119 Figure 5.8. Si 2p x-ray photoelectron diffraction patterns from (a) (3x3) 6H-SiC (0001)Si along [01-10], (b) (3x3) 6H-SiC (0001)Si along [10-10], and (c) (v3xv3)R30° 6H-SiC (0001)Si along [01-10]........................................................................ 120 Figure 5.9. Si 2p x-ray photoelectron diffraction patterns along [-110]/[11-20] from (a) (7x7) Si (111), (b) (3x3) 6H-SiC (0001)Si, and (c) (v3xv3)R30° 6H-SiC (0001)Si...................................................... 121 Figure 5.10. C 1s XPD patterns from (v3xv3)R30° 6H-SiC (0001)Si along (a) [01-10], and (b) [11-20]................................................................ 122 Figure 5.11. C 1s XPD patterns from (3x3) 6H-SiC (0001)Si along (a) [10-10], (b) [11-20], and (c) [10-10] azimuths................................................ 123 Figure 5.12. Schematic illustrating the differences in stacking along the [111]/[0001] direction for 3C and 6H-SiC........................................ 125 xx
Figure 5.13. Schematic illustrating expected forward scattering/focusing peaks in XPD along the [11-20] azimuth of 3C/6H-SiC............................. 126 Figure 5.14. Schematic illustrating expected forward scattering/focusing peaks in XPD along the [10-10] azimuth for 3C/6H-SiC............................ 126 Figure 6.1. AES of (3x3) reconstructed (0001)Si 6H-SiC (a) before remote H plasma, and (b) after remote H plasma.......................................... 152 Figure 6.2. XPS of the Si 2p core level from (3x3) reconstructed (0001)Si 6H-SiC (a) before remote H plasma, and (b) after remote H plasma............................................................................................ 152 Figure 6.3. TPD of (1x1) 6H-SiC (0001)Si after remote H plasma exposure (1 min., 20 W, 15 mTorr, and 450°C), (ß = 1°C/sec.)........................ 153 Figure 6.4. TPD of (a) sample heating stage after outgassing, and (b) molybdenum plate after remote H plasma exposure.................... 153 Figure 6.5. XPS of the C1s core level from (0001)Si 6H-SiC (a) before remote H plasma, (b) after remote H, and (c) after annealing at 1000°C....... 154 Figure 6.6. TPD of Si (111) after room temperature exposure to 2000£ H2 with rhenium filament at > 1700°C (ß=1°C/sec.).............................. 156 Figure 6.7. AES of (3x3) 6H-SiC (0001)Si (a) before atomic H exposure and (b) after room temperature exposure to 2000£ H2 with hot filament at > 1700°C.......................................................................... 157 Figure 6.8. TPD of (0001)Si 6H-SiC after room temperature exposure to 2000£ H2 with rhenium filament at > 1700°C (ß=1°C/sec.).............. 158 Figure 6.9. TPD of (3x3) 6H-SiC (0001)Si after cooling to 300°C in 10-6 Torr SiH4 (ß=1°C/sec.)............................................................... 159 Figure 6.10. Mono-hydride surface coverage on silicon sites of SiC....................... 164 Figure 6.11. Di-hydride surface coverage on silicon sites of SiC............................. 165 xxi
Figure 6.12. Hydrogen surface coverage on carbon sites of SiC based on kinetic data of Hamza....................................................................................165 Figure 6.13. Hydrogen surface coverage on carbon sites of SiC based on kinetic data of Thomas et al........................................................................... 166 Figure 6.14. Percent dissociation of H2 into H as a function of temperature........... 168 Figure 7.1. AES survey spectra of OMVPE AlN: (a) as received, (b) solvent cleaned and 20 min. UV/O3 exposure, and (c) 3 min. dip in 10:1 buffered HF (BHF)............................................................................ 192 Figure 7.2. XPS of O 1s core level from bulk AlN wafer (a) as recieved, (b) UV/O3 exposure, and (c) 10:1 BHF............................................. 192 Figure 7.3. XPS of Al 2p core level from bulk AlN wafer after a 10:1 BHF dip..... 193 Figure 7.4. XPS of N 1s core level from bulk AlN wafer after a 10:1 BHF dip....... 193 Figure 7.5. AES of (0001) OMVPE AlN after UV/O3 oxidation and oxide removal with (a) 1:1 NH3OH:H2O2, (b) 1:1 HCl:DI, (c) 10:1 BHF, (d) RCA SC1, and (e) RCA SC2.............................................. 195 Figure 7.6. Close up of AlLVV from AES of OMVPE AlN after (a) UV/O3 and (b) 10:1 BHF...................................................................................... 196 Figure 7.7. XPS of the F 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C............................ 198 Figure 7.8. XPS of the C 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C............................ 198 Figure 7.9. XPS of the O 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C............................ 199 Figure 7.10. XPS of Na 2p from polycrystalline AlN surface after (a) etching in NaOH and (b) an RCA clean............................................................. 200 xxii
Figure 7.11. TPD of m/e- (a) 18, (b) 20 and (c) 38 from 10:1 BHF dipped AlN (ß = 20°C/min.)......................................................................... 202 Figure 7.12. XPS of F 1s core level from polycrystalline AlN wafer cleaned in 10:1 BHF (a) before and (b) after remote H plasma exposure at 450°C (15 mTorr, 20W)................................................................. 204 Figure 7.13. AES survey spectra of OMVPE AlN after: (a) UV/O3 and 10:1 BHF dip and (b) remote H plasma at 450°C..................................... 204 Figure 7.14. XPS of O 1s from (0001) GSMBE AlN after (a) annealing at 1000°C and (b) annealing in a 0.1 ML/sec flux of Al at 1000°C....... 205 Figure 7.15. The XPS O 1s core level from an AlN surface (a) before and (b) after annealing in a SiH4 flux....................................................... 206 Figure 7.16. XPS of O 1s from In capping layer on OMVPE AlN (a) as recieved, (b) after annealing at 600°C, and (c) after annealing at 750°C............................................................................................. 207 Figure 7.17. XPS of O 1s core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as recieved, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) > 1000°C................................. 208 Figure 7.18. XPS of C 1s core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as recieved, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) >1000°C.................................. 208 Figure 7.19. XPS of Ga 2p3/2 core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as recieved, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) >1000°C.................................. 209 Figure 7.20. AES survey spectra from GSMBE GaN after (a) 1 day in air on laminar flow bench, (b) UV/O3 oxidation, and (c) 5 min. etch in 1:1 HCl:DI.............................................................................. 210 Figure 7.21. AES survey spectra from (001) GaAs (a) before and (b) after UV/O3. 211 Figure 7.22. XPS of C 1s core level from (0001) OMVPE GaN xxiii
after (a) ultrasonification in trichloroethylene, acetone, and methanol, and (b) UV/O3 exposure.................................................... 212 Figure 7.23. AES survey scan of OMVPE GaN after a 24 hr UV/O3 exposure with 1L/sec. flowing O2..................................................................... 213 Figure 7.24. XPS of O 1s core level from (0001) OMVPE GaN (a) after solvent cleaning, (b) after UV/O3 oxidation for 25 min., and (c) after UV/O3 oxidation for 24 hr. with 1000 L/sec flowing O2..... 214 Figure 7.25. AES of (0001) OMVPE GaN after UV/O3 oxidation and oxide removal with (a) 1:1 HCl:DI, (b) 10:1 BHF, and (c) 1:1 NH3OH:H2O2 (spectra normalized to NKLL)......................... 217 Figure 7.26. AES of (0001) OMVPE GaN after UV/O3 oxidation followed by (a) RCA SC1 and (b) RCA SC2 (spectra normalized to NKLL)......... 217 Figure 7.27. AES of (0001) OMVPE GaN (a) HCl:DI dip, (b) UV/O3 oxidation, and (c) HF vapor cleaning.................................................................. 219 Figure 7.28. XPS of F1s core level from (0001) GaN after UV/O3 oxidation followed by (a) 10:1 BHF vapor clean, and (b) a DI rinse................ 219 Figure 7.29. XPS of Ga2p core level from (0001) OMVPE GaN after UV/O3 oxidation and (a) BHF vapor clean and (b) DI rinse and NH3OH:H2O2 clean........................................................................... 220 Figure 7.30. XPS of N 1s core level from (0001) OMVPE GaN after UV/O3 oxidation and (a) BHF vapor clean and (b) DI rinse and NH3OH:H2O2 clean........................................................................... 220 Figure 7.31. m/e- 69 (Ga) signal from GSMBE GaN as a function of surface temperature............................................................................ 223 Figure 7.32. AES of (0001) OMVPE GaN after (a) HCl vapor clean, and annealing at (b) 300°C, (c) 450°C, and (d) 600°C............................. 223 Figure 7.33. AES survey spectra from GSMBE GaN (a) exposed to UV/O3, and after annealing in: (b) UHV at 650°C, 20 min., (c) NH3 at 650°C, 20 min, and (d) NH3 at 800°C, 25 min................................... 224 xxiv
Figure 7.34. AES survey spectra from GSMBE GaN after various sequential scans (a) 1, (b) 3, (c) 8, and (d) 9 scans............................................. 224 Figure 7.35. UPS of OMVPE GaN after (a) rinsing in methanol, (b) 1:1 HCl:DI, and (c) annealing in NH3 at 800°C, (d) as grown (2x2) GSMBE GaN........................................................................... 225 Figure 7.36. AES survey spectra from OMVPE GaN (a) after oxide removal with 1:1 HCl:DI, (b) 100°C H plasma exposure, and (c) 450°C H plasma exposure............................................................ 227 Figure 7.37. AES of AlGaN (a) as recieved, (b) after solvent cleaning in trichloroethylene, acetone, methanol, and isopropanol, (c) 1:1 NH3OH:H2O2, (d) 1:1 HCl:DI, and (e) 10:1 BHF...................................................................................... 228 Figure 7.38. AES of AlGaN after (a) UV/O3 exposure and (b) HF vapor oxide removal.................................................................................... 228 Figure 7.39. XPS of Al 2p core level after UV/O3 oxidation and (a) HF vapor oxide removal, and (b) DI rinse and NH3OH:H2O2..... 229 Figure 7.40. Schematic illustrating alignment of Cl- and F- ions with VBM of GaN and AlN in 1:1 HCl:DI and 10:1 HF respectively................... 237 Figure 8.1. Attenuation of Al 2p core level from AlN buffer layer as a function of overlying GaN film thickness for Tsub=650°C................ 270 Figure 8.2. LEED diffraction patterns from (a) (1x1) (0001) GaN, and (b) (2x2) (0001) GaN......................................................................... 270 Figure 8.3. SEM micrographs at 10 kX from GaN films grown in GSMBE at (a) 650°C, (b) 750°C, and (c) 800°C.................................................. 271 Figure 8.4. Photoluminesence (PL) at 4K of NH3-GSMBE GaN grown at (a) 650°C and (b) 800°C..................................................................... 272 Figure 8.5. Attenuation of Al 2p core level from AlN buffer layer as a function of overlying GaN film thickness for Tsub=800°C................ 273 xxv
Figure 8.6. Attenuation of Ga 3d and 3p core levels from OMVPE GaN as a function of overlying GaN film thickness for Tsub = 800°C.............. 275 Figure 8.7. Hydrogen surface coverage of Ga sites as a function of temperature and flux.............................................................................................. 283 Figure 8.8. Hydrogen surface coverage of N sites as a function of temperature and flux.............................................................................................. 284 Figure 8.9. Hydrogen surface coverage of Al sites as a function of temperature and flux.............................................................................................. 284 Figure 9.1. XPS spectra (arbitrary units) and theoretical densities of states (in states per unit cell per eV) of 6H-SiC.......................................... 298 Figure 9.2. XPS spectra (arbitrary units) and densities of states (in states per unit cell per eV) of 2H-AlN......................................................... 300 Figure 10.1. UPS Spectra of (2x2) (0001) AlN surface............................................ 313 Figure 10.2. UPS spectra of (2x2) (0001) GaN surface............................................ 315 Figure 11.1. ScN equilibrium diagram computed using HSC.................................. 330 Figure 11.2. AlN equilibrium diagram computed using HSC.................................. 330 Figure 11.3. GaN equilibrium diagram computed using HSC.................................. 331 Figure 11.4. GaN equilibrium diagram computed using experimental data of Thurmond and Logan, and Munir and Searcy............................... 331 Figure 11.5. InN equilibrium diagram computed using HSC and experimental data of Jones and Rose....................................................................... 332 Figure 11.6. AES survey scan of ScN film grown at 800°C on (111) 3C-SiC......... 334 Figure 11.7. XPS spectrum of N 1s and Sc 2p3/2,1/2 2p3/2,1/2 core levels xxvi
from a ScN film grown at 800°C...................................................... 335 Figure 11.8. UPS spectrum of ScN film grown at 800°C on (111) 3C-SiC............. 335 Figure 11.9. TEM of ScN film grown at 800°C on (111) 3C-SiC............................ 336 Figure 11.10. TEM of 800°C ScN on 5000Å (0001) GaN/AlN/6H-SiC.................. 336
xxvii
1. Introduction
Surfaces are the starting points for all semiconductor processes and hence represent
the foundation on which all subsequent microelectronic device fabrication processes are built.
Accordingly, proper surface preparation is an absolutely critical first step in all
semiconductor processes in order to achieve optimum results. The consequences of improper
surface preparation/cleaning are numerous leading to less than optimized results through
increased epitaxial defect densities (dislocations and stacking faults), lower dielectric
breakdown voltages, higher densities of interface states, variations in threshold voltage,
increased Ohmic contact resistance, and variations in Schottky barrier heights. All of these
effects manifest themselves in the bottom line through decreased device performance and
yield. In turn, these observations, have led to numerous studies of the surfaces of
technologically important semiconductors surfaces such as Si, Ge, GaAs, and InP. These
studies have shown that the entire chemical, electrical, and structural state of a semiconductor
surface must be considered when developing surface processes [1-4].
For the above reasons, the author has chosen to examine the surfaces and interfaces
of SiC, GaN, and AlN. These compounds are wide band gap semiconductors and are of
significant technological importance due to their extreme materials properties (see Table 1.1)
[5,6]. As documented in Table 1.1 [7], AlN and SiC are extremely strong materials with high
melting points and have accordingly long been of interest as high temperature structural
ceramics [5]. However, SiC exhibits both n and p type conductivity which with its large
band gap and high temperature stability has gained it significant consideration as a possible
semiconductor for high temperature microelectronic devices. The high thermal conductivity,
large breakdown voltage, and moderate electron and hole mobilities of SiC has also made it
of interest for possible high power and high frequency device applications. Due to the
insulating properties of AlN and moderately close lattice matching to SiC (?a/ao = 0.9%),
AlN/SiC quantum well structures and MISFET device applications have also been recently
proposed and/or fabricated [5].
However many of the same extreme materials properties exhibited by SiC are also
exhibited by GaN (see Table 1.1). Accordingly, GaN is also of interest for use in high
temperature, high power, and high frequency device applications [5,6]. In contrast to SiC,
however, GaN possesses a direct band gap which can be varied from the visible region into
the deep UV through alloying with InN and AlN. Accordingly, GaN and III-V nitride
materials are of considerable interest for numerous blue/UV optoelectronic device
applications [5,6]. The recent demonstration of a InGaN quantum well blue laser diode by
Shuji Nakamura of Nichia Chemical Co. is a highlight of the recent advances achieved in the
III-V nitride field [8].
2
Table 1.1. Selected materials properties of Si, GaAs, 6H-SiC, 2H-GaN and 2H-AlN [7].
Si GaAs 6H-SiC GaN AlN Eg (eV) 1.1 (i.) 1.43 (d) 3.01 (i) 3.40 (d) 6.2(i) Tmelt (°C) 1415 1238 2830 2200 2850 µn (cm2/V sec) 1350 8500 600 2000 µp (cm2/V sec) 480 400 40 150 14 E (GPa) 130 85.5 480 300 400 σT (W/cm K) 1.5 0.5 4.9 1.3 2.0 Vsat (cm/sec) 1x107 3x107 2x107
2.5x107 EB (MV/cm) 0.25 0.3 2.5 2.0 a (Å) 5.43 5.65 3.08 3.19 3.11 c (Å) 15.12 5.185 4.982 (α 106/K) 6.0 3.59 4.2 5.59 4.2 FJ (V/sec) 5x1011 13x1011 125x1011 159x1011
(EB vsat/2p)
3
In the following chapters of this text, a detailed examination of both ex situ and in situ
surface cleaning processes for (0001) surfaces of 6H-SiC, AlN, and GaN is presented. The
presented findings are based on results obtained from a variety of surface analytical
techniques used to characterize these surfaces including: Auger electron spectroscopy (AES),
x-ray photoelectron spectroscopy (XPS), low energy electron diffraction (LEED), electron
energy loss spectroscopy (EELS), temperature programmed desorption (TPD), and x-ray
photoelectron diffraction (XPD). Emphasis was first placed on understanding the nature of
the native contaminants present on these surfaces during ambient exposure. Processes were
accordingly then developed for the removal of these contaminants and the effects of these
processes on these surfaces were investigated. In addition, the surface processes involved in
the growth of GaN and AlN in gas source molecular beam epitaxy were also examined and
modeled. Finally, the heterojunction valence band discontinuity between AlN and SiC at the
(0001) interface was also examined.
1. W. Kern, RCA Review, 39, 278 (1978). 2. W. Kern, J. Electrochem. Soc., 137, 1887 (1990). 3. T. Ohmi, J. Electrochem. Soc., 143, 1957 (1996). 4. D.E. Aspnes and A.A. Studna, Appl. Phys. Lett., 39, 316 (1981). 5. R.F. Davis, Proc. of IEEE, 79, 702 (1992). 6. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10 1237 (1992). 7. J.H. Edgar, Ed., Properties of Group III Nitrides, Inspec, London (1994). 8. S. Nakamura, S. Masayuki, and S. Yasunobo, Appl. Phys. Lett., 68, 3269 (1996).
4
2. Wet Chemical Processing of (0001)Si 6H-SiC:
Hydrophobic and Hydrophilic Surfaces
To Be Submitted for Consideration for Publication
to the
Journal of the Electrochemical Society.
by
Sean W. King, Robert J. Nemanich, and Robert F. Davis,
Department of Materials Science and Engineering
North Carolina State University,
Raleigh NC 27695
5
2.1 Abstract
The wetting characteristics of oxidized and as-polished (0001)Si 6H-SiC (on and off
axis) surfaces in various acids and bases were compared to that of (111) Si. Auger electron
spectroscopy, x-ray photoelectron spectroscopy, and low energy electron diffraction were
used to characterize the chemical state and order of these surfaces. It was observed that the
oxidized SiC surfaces were hydrophilic after oxide removal with a 10:1 HF solution and were
terminated with approximately a monolayer of OH, CO, CH, and F species. In contrast, as-
polished SiC surfaces, were observed to be hydrophobic and covered with a thin (5-10Å)
contamination layer composed primarily of C-C, C-F, and Si-F bonded species. Removal of
this contamination layer using an RCA SC1 etch or Piranha clean resulted in a disordered
hydrophilic SiC surface. Similar effects were observed for other orientations of 6H-SiC (i.e.
(000-1)C, (11-20), and (10-10)). A 20Å amorphous Si capping layer was identified as an
alternative passivation layer for producing hydrophobic SiC surfaces.
6
2.2. Introduction
Preparation of clean, structurally well ordered surfaces is an important first step in all
semiconductor microelectronic fabrication processes [1-3]. Experience gained in silicon
technology has shown that the criteria for surface cleanliness must include removal of not
only native oxides and organic contaminants but also metallic impurities, particulate
contaminants, adsorbed molecules, and residual species left by previous processes [1-3]. The
effects of incomplete removal of all of these various contaminants by non-optimized surface
cleaning processes has been observed to consequently result in decreased device performance
and yield [4,5]. This is primarily through the generation of increased densities of electrical
defects (higher interface state densities, lower breakdown fields, increased leakage currents)
[6-13] and structural defects (dislocations and stacking faults) in epitaxial layers [14-25]. For
these reasons, an intensive effort has been made to understand the nature and source of
surface contaminants accumulated during silicon microelectronic processing [26-69].
Accordingly, these studies have developed a wide range of wet and dry (ex situ and in situ)
surface cleaning processes specifically optimized for silicon surfaces
[1,2,26,28,29,34,40,51,54-58].
In the case of SiC, numerous studies have been concerned with the nature and
removal of native oxides and other contaminants on silicon carbide surfaces [70-97].
However, few of these studies [87-92] have actually considered and/or investigated the effect
of wet chemical processes on SiC surfaces and any differences that might exist between
silicon and silicon carbide wet chemical processing. Therefore due to a lack of such detailed
studies for silicon carbide surfaces, many of the wet chemical processes optimized for silicon
7
have been implemented in SiC device processing [70-97]. This is in despite of the fact that
differences should be expected in the surface chemistry of these two semiconductors based
on the extreme chemical inertness of SiC compared to Si [98,99]. Thus given the plethora of
examples of the deleterious effects of non-optimized surface cleaning processes, it is evident
that further optimization of silicon carbide surface cleaning processes should result in a
reduced level of electrical and structural defects in silicon carbide device structures. Further
reduction in epitaxial and electrical defects should assist the development of SiC into the
semiconductor of choice for high speed, high frequency, and high temperature electronic
devices as well as the substrate of choice for III-V nitride heteroepitaxy [99,100].
In this paper, we investigate the effect on silicon carbide surfaces of various wet
chemical processes common to silicon. Emphasis is placed on HF processes which typically
serve as one of the last steps in silicon wet chemical processing and is primarily used for
surface oxide removal [40-69]. This process has become common due to the fact that it
produces a hydrogen terminated/hydrophobic silicon surface which is stable against
oxidation in air for several hours/days [56,63-69]. In this paper, we closely examine the
effect of HF based chemical processes on (0001)Si 6H-SiC surfaces. In this study, we have
observed that HF processing of oxidized (0001)Si 6H-SiC and other surface orientations (i.e.
(000-1)C, (11-20), and (10-10)) always leaves a hydrophilic SiC surface. Based on x-ray
photoelectron and Auger electron spectroscopy, we have additionally determined that HF
processing leaves a SiC surface terminated predominantly with O (or OH) rather than H.
This is in contrast with HF processing of silicon surfaces which is known to leave a
hydrophobic hydrogen terminated silicon surface [56,63-69].
8
The importance of controlling the wetting characteristics of surfaces in silicon
technology has been shown to not only be important from a surface termination point of view
but also important in reducing particulate and metallic contamination during wet chemical
processing [40-51]. In principle, wet chemical processing of hydrophobic surfaces should
lead to lower levels of particulate contamination due to a "snow plow" effect as the surface
traverses the liquid/air interface (see Figure 2.1). However in reality, particles have been
empirically observed to be more attracted to hydrophobic surfaces then hydrophilic surfaces
during wet chemical processing. This observation may be related to the fact that hydrophilic
surfaces tend to acquire a negative charge in solution which electrically repels particles in
solution which also have a negative charge [3, 41-43]. Hydrophobic surfaces on the other
hand are more passivated and develop less of a charge (or positive charge) in ionic solutions
and therefore negatively charged particles in the solution are attracted to the hydrophobic
surface [3, 41-43].
(a) (b)
9
Figure 2.1. Particle accumulation on removal from solution for (a) a hydrophilic surface and (b) a hydrophobic surface.
In the case of SiC, we have previously noted that the hydrophilic nature of SiC
surfaces in wet chemical processing leads to trapping of wet chemicals in the micro pipes of
SiC wafers and the unwanted removal/outgassing and incorporation of these chemicals in
following processes [101]. For this reason and those mentioned above, we have additionally
investigated other alternative processes and passivation layers which could lead to a
hydrophobic (0001)Si 6H-SiC surface. In particular, we have identified the as as-polished
(0001)Si 6H-SiC surface as being hydrophobic. However, XPS, AES and low energy
electron diffraction (LEED) analysis indicated this surface was composed of a thin (≈ 5-10
Å) contamination layer composed mainly of CFx and SiFx species. Removal of this
contamination layer using RCA SC1 (1:1:5 NH3OH:H2O2:H2O) or Piranha (7:3
H2SO4:H2O2) cleaning results in a defective and disordered surface. In contrast, we have
found that a hydrophobic SiC surface can be obtained using an amorphous 20Å Si capping
layer. The excess silicon may be removed by thermal desorption resulting in a structurally
well ordered SiC surface as evidenced by a (3x3) LEED pattern.
2.3. Experimental Procedure
On axis and vicinal n-type (typically Nd=1018/cm3) (0001)Si 6H-SiC wafers were
used in these experiments. These wafers were supplied by Cree Research, Inc. with or
without a 500-1000Å thermally grown oxide. Both the oxidized and as-polished, as-received
10
6H-SiC substrates were ultrasonically cleaned/degreased in trichloroethylene, acetone, and
methanol each for 10 min. prior to any other wet chemical treatments. The oxide was
removed from the thermally oxidized wafers using a 10 min. dip in a 10:1 HF solution,
followed by rinsing in de-ionized (DI) water and N2 blow drying. The wetting
characteristics of this surface and the as-polished surfaces were then investigated by
immersion in other acid/base solutions. The wet chemistries examined included 100:1 HF,
10:1 HF, 1:1 HF, 10:1 buffered HF (7:1 NH4F:HF), 30:1 buffered HF, 40% NH4F, 38%
HCl, 70% HNO3, Piranha Etch (7:3 H2SO4:H2O2@120°C), RCA SC1 & SC2 (1:1:5
NH3OH:H2O2:H2O@85°C) and 1:1:5 HCl:H2O2:H2O@85°C), 100% HC2H3O2, and 40%
KOH. These chemistries were chosen for examination primarily due to their extensive use in
the silicon microelectronics industry. Except where noted, after all wet chemical treatments
the samples were rinsed in DI water (18 M?) and blown dry with N2. At this point, the wafer
was visually inspected to determine whether a hydrophobic or hydrophilic surface had been
retained or obtained. To assist in the comparison to silicon, (100) and (111) Si (on and off
axis) wafers were dipped in the same acid/base solution immediately after the SiC wafer and
the wetting characteristics of these surfaces were compared. All wet chemicals were of
CMOS grade purity (J.T. Baker).
SiC wafers with a 200Å amorphous Si capping layer were prepared in the following
manner. First, an oxidized (0001)Si 6H-SiC wafer was given a 10 min. dip in 10:1 HF, DI
rinsed followed by N2 drying, and loaded into a Si-Ge MBE where it was then degassed at
450°C and annealed at 1000°C for 20 min. in a 10-6 Torr SiH4 flux. This produced an
oxygen free, Si rich (1x1) SiC surface. Next, an electron beam was used to evaporate/deposit
≈ 200Å of amorphous Si onto the SiC surface at room temperature. The a-Si/SiC sample was 11
then removed from vacuum for subsequent wet chemical processing as described above for
oxidized and as-polished SiC surfaces.
Surfaces prepared as described above were further subjected to surface analysis in an
integrated ultra-high vacuum (UHV) system incorporating the following analytical
techniques: x-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES),
electron energy loss spectroscopy (EELS), and low energy electron diffraction (LEED). The
details of these systems and the transfer line have been previously described [102]. After
each wet chemical treatment, the SiC wafer was mounted to a molybdenum sample holder
and loaded into a load lock for subsequent analysis by AES, XPS, EELS, and LEED. The
XPS analysis was performed using a VG CLAM II electron energy analyzer and an Al anode
(hν=1486.6 eV) at 20 mA and 12 kV. The AES spectra were obtained using a Perkin Elmer
CMA, a beam voltage of 3 keV and an emission current of 1 mA. The EELS spectra were
also obtained using the CMA but with a 100 eV electron beam and an emission current of 1
mA. The LEED was performed using Princeton Scientific rear view optics, a beam voltage
of approximately 115 eV, and an emission current of 1 mA. Calibration of the XPS binding
energy scale was performed by measuring the position of the Au 4f7/2 and shifting the
spectra such that the peak position occurred at 83.98 eV.
2.4. Results
2.4.1. (0001)Si 6H-SiC, Oxidized
12
In contrast to (100) and (111) Si, a 10:1 HF solution was visually observed to wet the
surface of (0001)Si 6H-SiC as this wafer was withdrawn from the HF solution after a 10 min.
dip (Note: this is sufficiently long to remove the 750Å thermal oxide since the SiO2 etch rate
with 10:1 HF is ˜ 10 Å/sec [60-62]). The 10:1 HF solution was noticed to slowly pool up on
the SiC surface over time. However after rinsing in DI water, the SiC surface was observed
to be clearly wetted by H2O (i.e. no pooling up of the H2O). Figure 2.2(a) displays an AES
survey spectrum taken from an off axis (0001)Si 6H-SiC wafer after removal of the 750Å
thermal oxide with a 10 min. 10:1 HF dip (Note: similar results were obtained from on axis,
(0001)Si 6H-SiC as well). Si, C, and significant amounts of oxygen were detected by AES.
Using XPS, further analysis of the (0001)Si 6H-SiC surface after a 10:1 HF dip, also
revealed the presence of significant amounts of fluorine, ≈ 1/4 ML (see Figure 2.3(a)), on the
surface. The fluorine was not detected by AES due to a lower sensitivity to fluorine and
possible electron beam stimulated desorption effects [103-107]. By analogy to intentionally
fluorinated silicon surfaces where the F 1s peak was located at 685.9-686.2 eV [103-107], the
F 1s peak at 686.8 eV observed from the oxidized (0001)Si 6H-SiC surface after the HF dip
was likewise attributed to Si-F bonded fluorine. The fluorine surface coverage was initially
observed to vary from 0-1/4 ML, but further investigation revealed that the fluorine surface
coverage was highly dependent on the DI rinsing procedure. In fact, fluorine was not
detected by XPS for HF dipped surfaces rigorously rinsed in DI water (see Figure 2.3(b)).
13
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
Si
C
NO
F
Figure 2.2. AES survey spectra of (a) (0001)Si 6H-SiC after thermal oxidation and removal of the oxide with 10:1 HF, (b) (0001)Si 6H-SiC as polished surfaces after solvent cleaning.
680 682 684 686 688 690 692
(a)
(b)
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Figure 2.3. XPS spectra of the F 1s core level from an oxidized (0001)Si 6H-SiC surface after (a) oxide removal with 10:1 HF, followed by (b) rigorous rinsing in running DI water.
14
Comparison of the relative intensities of the O KLL AES transition and the XPS O 1s
core level from oxidized (0001)Si 6H-SiC and (100) and (111) Si surfaces after oxide
removal with 10:1 HF indicated a significantly larger surface concentration of oxygen on SiC
surfaces in comparison to Si surfaces (see Figure 2.4). The intensity of these two
peaks/transitions were estimated to correspond to oxygen surface coverage's of ≈ 3/4±1/4
ML for (0001)Si 6H-SiC and < 1/10 ML for (100) and (111) Si. This is in agreement with
the observed hydrophilic and hydrophobic nature of these surfaces respectively. More
detailed analysis of the (0001)Si 6H-SiC surface using photoemission from the O 1s core
level indicated the presence of two peaks. The larger peak at 532.1 eV is indicative of Si-O
or Si-OH bonded oxygen, and the smaller peak at 533.5-533.9 eV is indicative of C-O
bonding [30,31,56,57]. The intensity of the C-O O 1s bonding peak like the F 1s Si-F
bonding peak was observed to depend on the DI rinsing procedure with no DI rinsing
resulting in the observance of the most C-O bonded oxygen at the SiC surface. The presence
of C-O bonded oxygen on the (0001)Si surface was further supported by photoemission from
the C 1s core level which detected two C 1s peaks, one at 282.8 eV indicative of C-Si bonds,
and one at 284.7 eV indicative of a mixture of C-H and C-O bonds [19,20] (see Figure
2.5(a)). In contrast, EELS did not detect the presence of any p-p* transitions (≈ 3-6 eV)
typically observed from organic molecules and contamination (see Figure 2.6(a)). As
mentioned by Mizokawa et al [89], the failure of EELS to detect adventitious surface carbon
could be due to electron stimulated desorption.
15
526 528 530 532 534 536 538 540
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Si-O
C-O
Figure 2.4. XPS spectra of the O 1s core level from (a) (0001)Si 6H-SiC after removal of a 750Å thermal oxide with 10:1 HF, and (b) (111) Si after a dip in 10:1 HF.
Due to the inherent asymmetry of the XPS Si 2p core level arising from the
unresolved Si 2p3/2,1/2 doublet, it was difficult to accurately determine if a Si-O Si 2p
bonding peak actually existed. Assuming a FWHM of 1.45 eV for the main Si-C Si 2p peak
at 100.7 eV, a second peak at ˜ 102.2 eV could be fitted to the spectrum, though the line
width of this peak was quite narrow < 1.4 eV. However, the presence of significant amounts
of Si-O bonded oxygen on the (0001)Si surface was supported by the Si KLL line shape in
AES (see Figure 2.2) which as noted by Mizokawa is similar to that of oxidized silicon
surfaces [89]. The intensity of the Si-O O 1s or Si 2p bonding peaks were not observed to
depend on the DI rinsing procedure with the O 1s (Si-O)/Si 2p (Si-C) intensity ratio
remaining essentially constant and independent of wafer cut (i.e. off or on axis). Finally,
these surfaces displayed intense (1x1) LEED patterns with broad dots which were clearly
16
visible at beam energies (Ep) as low as 60-100 eV. The AES, XPS, and EELS results from
the oxidized (0001)Si 6H-SiC surface are summarized in Tables 2.1 and 2.2.
280 282 284 286 288 290 292
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(c)
(b)
(a)
Figure 2.5. Typical XPS spectra of the C 1s core level from (a) oxidized (0001)Si 6H-SiC after oxide removal with 10:1 HF, (b) as-polished (0001)Si 6H-SiC after solvent cleaning, and (c) as-polished (0001)Si 6H-SiC after RCA SC1.
17
-30 -25 -20 -15 -10 -5 0 5
Cou
nts (
arb.
uni
ts)
Electron Loss Energy (eV)
(a)
(b)
š-š *
Figure 2.6. EELS spectra of (0001)Si 6H-SiC (a) oxidized followed by 10:1 HF, (b) as-polished with only solvent cleaning.
Table 2.1. Binding Energy (in eV) of core level positions from (0001)Si 6H-SiC as polished and oxidized surfaces after various treatments. The full width half maxima (Γ) of the peaks are also indicated.
Treatment Si 2p, Γ O1s, Γ C1s, Γ F1s, Γ N1s, Γ Oxidized 10:1 HF 100.7, 1.4 532.1, 1.9 282.8, 1.1 284.7, 2.1 As polished Solvents 100.5, 1.4 531.9, 2.3 282.6, 1.1 685.8, 1.9 398.2, 3.0 283.7, 2.7 687.5, 2.5 286.0, 4.5 10:1 HF 100.5, 1.4 531.6, 2.4 282.5, 1.1 685.6, 1.9 398.2, 3.0 283.6, 2.7 687.3, 2.5 286.0, 4.4 H2SO4:H2O2 100.4, 1.4 531.5, 2.5 283.5, 1.1 686.0, 3.0 398.2, 3.0 283.6, 2.7 RCA SC1 100.3, 1.4 531.3, 2.8 282.4, 1.1
18 283.9, 2.4
Aqua Regia 100.5, 1.4 531.6, 2.3 282.6, 1.1 685.5, 1.7 398.2, 3.0 283.7, 2.7 686.9, 3.2
able 2.2. Summary of XPS Si2p/O1s, Si2p/F1s, C/C, and C-C data (uncorrected for
Treatment Si2p/O1s Si2p/F1s C/C C-C
Tsensitivity factors) for (0001)Si 6H-SiC surfaces.
Oxidized
1.4 ∞ 6.65 1.9 eV ed
.2 1.6 1.1 1.3 eV
10:1 HF As Polish Solvents 110:1 HF 1.8 1.8 0.9 1.1 eV H2SO4:H2O2 0.9 4.0 1.1 1.1 eV RCA SC1 1.0 ∞ 2.2 1.5 eV Aqua Regia 1.2 3.8 1.2 1.1 eV
19
Following the 10:1 HF dip, the thermally oxidized treated (0001)Si 6H-SiC surfaces
were dipped in a variety of different acids and bases and the surface wetting characteristics
visually noted. The results are summarized in Table 2.3 and illustrate that this surface was
found to be hydrophilic in all acids and bases investigated. In all cases, the SiC surface was
observed to retain a monolayer coverage of oxygen. For the oxidized (0001)Si 6H-SiC
surface, the surface coverage of oxygen was not found to change appreciably with dipping
time (1-24 hr.), HF concentration (1:1 - 1000:1), composition (HF-NH4F), or pH (1-10).
Table 2.3. Summary of wetting characteristics of as polished and oxidized (0001)Si 6H-SiC nd (111) Si.
(0001)Si 6H-SiC (0001)Si 6H-SiC (111) Si &
a
Treatment As Polished Thermally Oxidized a-Si Passivated 6H-SiC
None Hydrophobic Hydrophilic Hydrophilic10:1 HF Phobic Philic Phobic 38% HCl Phobic Philic Phobic 70% HNO3 Phobic Philic Philic RCA SC1 Philic Philic Philic RCA SC2 Phobic Philic Philic Piranha Philic Philic Philic Aqua Regia Phobic Philic Philic Acetic Phobic Philic PhobicNH4F Phobic Philic Phobic KOH Phobic Philic Philic
2.4.2. (000-1)C, (11-20), and (10-10) 6H-SiC surfaces
Thermally oxidized surfaces of other orientations of 6H-SiC such as (000-1)C, (11-
20), and (10-10) were also investigated. After removal of the thermal oxide with 10:1 HF,
these surfaces were also observed to be hydrophilic in all acids and bases investigated and
very little difference was observed between these surfaces and those with the (0001)Si
orientation. Figure 2.7 displays a series of AES spectra obtained from all four orientations
investigated. The spectra were obtained after removal of the thermal oxide with 10:1 HF
(note that all spectra are normalized to the C KLL Auger transition). Table 2.4 lists the Si/C,
O/Si, and O/C pph ratios (uncorrected for differences in sensitivity) calculated for each
surface. Table 2.4 indicates that the O/C ratio for all the different orientations of 6H-SiC
surfaces is centered around 0.3. Given that two of these surfaces are polar ((0001)Si and
20
(000-1)C) and the others non polar ((11-20) and (10-10)), it is surprising that they would
exhibit this similarity. Further, the Si/C pph ratio for the (0001)Si, (11-20), and (10-10)
surfaces are all centered around 0.6 which is surprising in that the ideal (0001)Si surface
would be terminated exclusively with Si, whereas the (10-10) and (11-20) are non polar
surfaces ideally with equal numbers of carbon and silicon atoms at the outermost surface.
The Si/C ratio as shown in Table 2.4 for the (000-1)C surface, however, is half that found for
the (0001)Si surface. The lower Si/C pph ratio for the (000-1)C face is expected based on the
differences in polarity for these two surfaces.
30 130 230 330 430 530 630 730
Si LVV
O KLL
C KLL
N KLL
dN(E
)/dE
(b)
Electron Energy (eV)
(c)
(d)
2.7. The AES survey spectra of various 6H-SiC surface orientations after reml oxide using 10:1 HF. (a) (0001)Si, (b) (000-1)C, (11-20), and (d) (10-10).
(a)
Figure oval of thermaa
21
Table 2.4. Peak to peak height (pph) ratios for Various 6H-SiC surfaces (uncorrected for ensitivity factors).
(0001)
s
Si (000-1)C (11-20) (10-10) Si/C 0.6 0.3 0.6 0.6
.4 O/Si 0.5 1.0 0.4 0 O/C 0.3 0.3 0.2 0.3
2.4.3. (0001)Si, As-Polished
As-polished (0001)Si 6H-SiC on and off axis surfaces which were not oxidized were
observed to be hydrophobic as received without any further processing. Figure 2.2(b)
displays an AES survey spectrum obtained from an as-polished, off axis (0001)Si 6H-SiC
surface after ultrasonic degreasing in trichloroethylene, acetone, and methanol for 10 min.
each. In contrast to hydrophilic SiC surfaces in which a thermally grown oxide had been
removed with 10:1 HF, traces of N and F were detected by AES from the as-polished SiC
surfaces, as shown in Figure 2.2(b). The presence of F and N on the surfaces of these as-
polished SiC wafers was also subsequently confirmed by XPS analysis. A more detailed
analysis of the F 1s spectrum from the as-polished SiC surface revealed the presence of two F
1s peaks, one at 685.6 eV indicative of Si-F bonding [103-107] and a second peak at 687.3
eV indicative of C-F, N-F, or SiFx bonding [106, 107] (see Figure 2.8(a)). The N 1s peak
was observed at 398.2 eV and is indicative of Si-N bonding [108,109]. The surface
concentration of F and N based on the XPS data was estimated to be ≈ 1ML and < 1/10 ML
22
respectively. These surface concentrations were not observed to change appreciably with
subsequent HF processing.
Like the SiC surfaces which had undergone a thermal oxidation treatment, XPS
Figures 2.2(a) and 2.2(b), the oxygen surface coverage is
cleaning.
detected the presence of significant amounts of surface carbon from the solvent cleaned
surface as polished/unoxidized (0001)Si 6H-SiC surface. Figure 2.5(b) shows an XPS
spectrum of the C 1s core level obtained from this surface. This spectrum was fitted to three
peaks centered at 282.6, 283.7, and 286.0 eV indicative of C-Si, C-C, and C-F bonding
respectively [106,107] (see Table 2.1 for Si2p, O1s, C1s, N1s, and F1s core level positions).
Additionally, a loss peak at 5-6 eV indicative of the p-p* transition of graphite like carbon
was also detected in EELS (see Figure 2.6(a)). Further, a (1x1) diffraction pattern was barely
discernible in LEED at Ep=100eV and only clearly discernible at Ep ≈ 180 eV. This is
indicative of either a thin contamination layer as suggested by the non-SiC C 1s peaks
observed or due to a disordered surface from subsurface defects or damage produced by the
polishing treatment [96].
As can be determined from
lower for the solvent cleaned, as-polished (0001)Si 6H-SiC surface than the HF dipped
oxidized SiC surface. However, the oxygen surface coverage based on the AES O KLL and
XPS O 1s intensities still is ˜ 1/2 ML. The binding energy of the O 1s core level from the as
polished SiC surface is slightly larger than that from a SiC surface which had undergone an
oxidation treatment 531.1 (ox) vs. 531.6 eV (no-ox). The differences in binding energy are
most likely due to band bending. Small amounts of a C-O O 1s peak was also detected from
the as polished SiC surfaces after wet chemical processing, but was not detected after solvent
23
After dipping in 10:1 HF for 10 min., the as-polished surface was still observed to be
hydrophobic when rinsed in DI water. F, N, and surface carbon were also still detected by
acids and bases, but would become hydrophilic in
XPS, AES, and EELS. As indicated in Figure 2.9(a), the HF dip did remove some oxygen
from the surface and the XPS Si2p/O1s ratio was found to increase after the HF dip (see
Table 2.2), but the oxygen surface coverage was still ≈ 1/2 ML. The C1s(Si-C)/C1s(C-C)
ratio was found to decrease after the HF dip indicating that the HF dip left more surface
carbon on the surface. Additionally, the adventitious C1s peak was observed to shift from
283.9 eV to 283.6 eV (see Table 2.1).
After the 10:1 HF dip, the as-polished SiC surface was observed to remain
hydrophobic when dipped in various
extended dip/etches in RCA SC1 or H2SO4:H2O2 (Piranha etch). In some instances, this
hydrophilic surface could be made hydrophobic again by boiling in Aqua Regia (3:1
HCl:HNO3) for 5-10 min. Subsequent XPS analysis revealed complete removal of fluorine
from hydrophobic (0001)Si 6H-SiC surfaces which had been permanently converted to
hydrophilic surfaces by prolonged immersion in RCA SC1 (see Figure 2.8(c)). In cases
where hydrophilic (0001)Si 6H-SiC surfaces were observed to be converted back to
hydrophobic surfaces by a boiling Aqua Regia treatment, XPS revealed incomplete removal
of fluorine from the surface by the RCA SC1 or Piranha etch treatment (see Figure 2.8(b)).
In addition, the third C1s peak observed at 286.0 eV was observed to track with the fluorine
coverage possibly suggesting that this surface is terminated with a contamination layer of
fluorocarbons. As a further note, none of the wet chemical processes employed here were
successful in converting hydrophilic SiC surfaces to hydrophobic surfaces for surfaces which
had undergone a thermal oxidation treatment followed by oxide removal with HF. A
24
summary of the wetting characteristics observed for the as received SiC surface after various
wet chemical treatments is provided in Table 2.3.
680 682 684 686 688 690 692
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
C-FSi-F
Figure 2.8. XPS spectra of the F 1s core level from as-polished (0001)Si 6H-SiC, (a) after solvent cleaning, (b) Piranha etch, and (c) RCA SC1.
25
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(c)
Si
C
NO
F
(d)
(b)
(a)
Figure 2.9. AES survey spectra of as-polished (0001)Si 6H-SiC after (a) solvent cleaning, (b) a 10:1 HF dip, (c) 5 min. Piranha etch, and (d) 30 min. RCA SC1 clean.
2.4.4. Si Passivating Layer
As the hydrophobic nature of as received/polished SiC surfaces seemed to be related
to a fluorocarbon contamination layer, a Si capping layer was investigated as a more
controllable alternative hydrophobic passivation layer for (0001)Si 6H-SiC surfaces. SiC
wafers terminated with an amorphous 200Å Si passivation layer were observed to be
hydrophobic after dipping in various HF and NH4F solutions. To investigate thinner Si
passivation layers, the 200Å Si passivation layer was thinned by repeated UV/O3 oxidation
and 10:1 HF dips. The Si passivation layer was observed to maintain a hydrophobic SiC
surface down to a Si thickness of ≈ 20Å, below which the SiC surface became increasingly
hydrophilic. However, this effect could be partially related to non-uniformity's in the
thickness of the Si passivation layer. The wetting characteristics of this Si passivated 26
(0001)Si 6H-SiC surface in other acids and bases after an HF dip were additionally observed
to be similar to those of silicon (see Table 2.3).
Figure 2.10(a) displays an AES spectrum from a 20Å Si/(0001)Si 6H-SiC surface
after a 10:1 HF dip and as can be seen lower oxygen and non-SiC carbon levels were
observed. This passivation layer was easily removed in vacuum by annealing at 1100°C for
5 min. prior to epitaxy. Figure 2.10(b) shows that a silicon rich, oxygen free, well ordered
surface characterized by a (3x3) LEED pattern was obtained. Figure 2.11 shows an XPS
spectrum of the Si 2p core level from the same surface and further illustrates the loss of the
silicon passivation layer by the reduction in the Si-Si bonding peak at 99.5 eV after the
1100°C anneal. As previously noted [101], the silicon passivation layer also resulted in
lower outgassing rates in vacuum due to lower levels of wet chemicals trapped in micro pipes
in the SiC wafer. Additional advantages and applications of the silicon capping layer for SiC
surfaces will be more fully discussed in the following section.
27
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
Si
C
O
Figure 2.10. (a) AES spectrum from 20Å a-Si/(0001)Si 6H-SiC after a 10:1 HF dip. (b) after thermal desorption of Si passivation layer at 1100°C.
95 97 99 101 103 105 107
(a)
(b)
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Si-Si
Si-C
Figure 2.11. XPS spectra of Si 2p core level from silicon passivated (0001)Si 6H-SiC (a) before thermal desorption and (b) after thermal desorption at 1100°C. 28
2.5. Discussion
2.5.1. (0001)Si 6H-SiC, Oxidize
The data presented above illustrates that a significant fraction of the (0001)Si 6H-SiC
surface is covered with oxygen and adventitious carbon (≈1-2 ML) after oxide removal using
10:1 HF or other HF/NH4F compositions (see Figure 2.12(a)). As essentially all ex situ
cleaning practices are limited by the ambient in which they are conducted, the presence of
small quantities of contaminants such as oxygen and carbon may be expected. However, the
amounts detected on (0001)Si 6H-SiC after removal of a thermal oxide with HF are several
times (5-10X) larger than those detected on silicon in our laboratory under the same
processing conditions. Additionally, the observation of a hydrophilic surface as opposed to a
hydrophobic surface further illustrates that the chemistry and interactions occurring at Si and
SiC surfaces in HF are clearly different.
CO
CH
x
FOH
OH
OH
OH
C=O
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OHF F CO
C=O
C= O
(a)
(b)
10Å C-C, C-F, Si-F Contamination Layer
29
Figure 2.12. (a) Schematic of (0001)Si 6H-SiC surface after thermal oxide removal with 10:1 HF. (b) Schematic of as-polished (0001)Si 6H-SiC.
The relatively large levels of oxygen remaining on the surface of SiC after an HF dip
are particularly surprising given that concentrated HF is known to etch silicon oxide at rates
as high as 1000Å/sec [2]. Clearly something is prohibiting the HF from removing the last
monolayer of oxygen or hydroxide from the SiC surface. One possible explanation for the
differences in oxygen coverage between Si and SiC is that the extra oxygen observed from
SiC surfaces is due to more strongly bound oxygen located at the steps of the SiC surface and
which are bonded to both Si and C. If this scenario were true, one would expect to see in
XPS and AES a difference in the O/Si ratio between on axis and off axis (0001)Si 6H-SiC
surfaces. However, this was not observed and it was found that the O/Si ratios were
essentially the same for both on and off axis wafers. Additionally, excess oxygen trapped at
SiC steps would simply not be large enough of an effect to explain the observed 1 ML
surface coverage.
Another possible explanation for the observed differences between Si and SiC is that
the oxygen is bonded to only carbon. Evidence of C-O bonding at the SiC surface is seen in
the C 1s and O 1s XPS spectra with peaks at 284.7 eV and 533.9 eV respectively. However,
XPS shows most of the oxygen to occur predominantly at 532.1 eV which is clearly
indicative of Si-O or Si-OH bonding. This is further supported by the Si LVV line shape in
Figure 2.1 which is also indicative of Si-O bonding.
In order to explain the apparent oxygen or OH termination of oxidized (0001)Si 6H-
SiC surfaces after an HF dip, it is necessary to first consider why hydrogen termination of Si
is achieved with HF and secondly to account for the polar and ionic nature of (0001) SiC 30
surfaces and HF solutions respectively. In the case of silicon, it was originally suggested that
HF processes produced Si-F terminated surfaces due to the relatively large bond strength of
Si-F compared to Si-H bonds ( 6 eV vs. 3.5 eV respectively) [59]. However, subsequent IR,
TPD, and HREELS analysis showed that HF processed silicon surfaces were in fact
terminated largely with hydrogen with < 1/10 ML fluorine coverage [66-69]. Trucks et al
[63] explained the hydrogen termination as being a result of the instability of Si-F bonds due
to the polarization of Si-Si backbonds by the strongly heteropolar Si-F bond (see Figure
13(a)). Polarization of the Si backbonds leaves these bonds susceptible to the strongly
polarized H+F- molecule which can then attack the backbond and fluorinate the silicon
surface atom and hydrogen terminate the nearby atom. This continued scenario eventually
leads to removal of the Si surface atom by complete fluorination (i.e. SiF4) leaving behind
only Si-H species. The stability of the hydrogen terminated silicon surface in HF can be
explained by comparison of the electronegativites of Si, H, and F which are 1.9, 2.2, and 4.0
respectively [110]. Due to the similarities in electronegativities of Si and H, the Si-H bond is
non polar (as compared to Si-F) and is therefore not attacked by HF. Trucks et al [63] has
additionally shown that the reverse reaction Si-H + HF => Si-F + H2 is energetically
unfavorable.
In the case of (0001)Si SiC surfaces, the underlying Si-C bonds are already polarized
(see Figure 2.13(b)) similar to Si-F bonds on Si due to the differences in electronegativity
between Si and C (1.9 and 2.6 respectively). Due to the stacking sequence along the [0001]
direction in SiC, the polarization of the Si-C bonds leads to the build up of a large crystal
potential/field which must be canceled in order to stabilize the crystal (see Figure 2.13(c))
[111]. In vacuum, cancellation of this field is achieved by desorption of surface atoms which
31
produces a compensating charge that cancels the internal field [111]. However in an ionic
solution such as HF, cancellation of this field can be achieved by simply adsorbing ionic
species of opposite polarity/charge. For the (0001)Si SiC surface, this would require the
adsorption of negatively charged ions such OH- or F- instead of hydrogen which is exactly
what is observed. Termination of the SiC surface with OH- is also supported by the recent
high resolutions EELS (HREELS) studies of Starke et al [90] which were able to identify the
O-H stretch from HF processed (0001) 6H-SiC substrates.
By analogy to silicon [56,65-69], the observation that the fluorine coverage on SiC
surfaces after an HF dip depended on the DI rinsing procedure suggests that the fluorine is
located at defects sites on the SiC wafer. During the DI rinse, the surface fluorine is removed
from the surface by saturation of these defect sites with OH- instead. However, we did not
observe a change in the O/Si ratio with or without DI rinsing. Therefore, this suggests that F
either substitutes for H in OH (i.e. Si-OF) instead of bonding directly with Si at the surface
(i.e. Si-F) or the fluorine coverage is << 1/4 ML such that a change in the O/Si ratio can not
be detected.
Another way at looking at this is to consider the quasi band diagram shown in Figure
2.14. In this figure, we consider the liquid (HF) to have a quasi Fermi level which is aligned
with the Fermi level of Si and/or SiC [112]. All of the ionic species in the HF solution have
an energy somewhere below this quasi Fermi level and which is dependent on the
electronegativity of the specie [3, 112]. For Si it has been postulated that the energy position
of the H+ ion in the HF solution is located very closely to the Si VBM where effective charge
transfer can occur and hence is the reason for hydrogen termination of silicon by HF [3,112].
SiC, however, has a larger bandgap and its VBM should lie below that of Si and hence
32
termination of the SiC surface with species which lie at lower energy levels such as OH- and
F- is expected.
F
Si
Si Si Si
Si
Si Si Si
F F
H
F
F
F FSi
Si
Si Si Si
FF
F
H H
Si Si Si
H H H
δ−
δ−
δ+ H - Fδ−δ+
δ−
H - Fδ−δ+H - Fδ−δ+
δ−
(a)
Si
C C Cδ−
δ+
F
Si
C C C
δ−
δ−
δ+OH-
Si
C C Cδ−
δ+
(b)
(c)
σ+σ−
σ+σ−
σ+σ−
σ+σ−
[0001]
Ε
ΟΗ−
ΟΗ
−
ΟΗ
−
ΟΗ
−
ΟΗ
−
ΟΗ
−
ΟΗ
−
ΟΗ
−
Figure 2.13. (a) Schematic illustrating mechanism of hydrogen termination of silicon in HF solutions. (b) Schematic illustrating stability of F- or OH- termination of SiC in HF solutions rather than H termination. (c) Schematic illustration of crystal potential in SiC.
33
Si10:1 HF
H+
OH-
F-
K+
SiC 10:1 HF
OH-
F-
SO42-
O3-
(a) (b)
EF
EF
Figure 2.14. Energy diagram of (a) Si and (b) SiC in aqueous solutions.
The OH- termination of SiC wafers is important not only from a chemical purity point
of view but also from the viewpoint of particle contamination during wet chemical
processing. In silicon microelectronics, it has been found that the zeta potential of silicon
surfaces in acids is negative and that particles in the solution are charged positively [3,41-43].
As a result, particles are electrostatically attracted to silicon surfaces in acid processing and
consequently acid processes generally result in higher levels of particle accumulation [3, 41-
43]. To account for this, silicon surfaces have to be cleaned in basic (high pH) solutions for
particle removal where both Si surfaces and particles acquire the same negative charge and
electrostatically repel one another [3, 41-43]. Due to the OH termination of SiC, one would
expect SiC surfaces however to exhibit a zeta potential pH dependence more similar to SiO2.
In fact, this is exactly what has been observed in zeta potential measurements on SiC
powders [113-115]. The importance of this is that the zeta potential pH dependence of SiO2
34
and SiC particles are very similar [3,41-43]. Therefore, particles should be repelled from SiC
surfaces in both acids and bases.
2.5.2. (0001)Si 6H-SiC, As-Polished
The above results have shown that the (0001)Si 6H-SiC surfaces, which are not
oxidized after polishing, are hydrophobic before and after dipping in 10:1 HF. However,
only slightly smaller amounts of oxygen were detected from the as-polished surfaces after
even an HF dip. As previously shown in Figure 2.4, the oxygen surface coverage for both as-
polished and oxidized SiC surfaces is 5-10 times higher than that for hydrophobic hydrogen
terminated Si surfaces. So clearly, the hydrophobic, as-polished SiC surface observed in this
study is not related to the hydrophobic, hydrogen terminated silicon surface. This is evident
simply from the observation that the as-polished (0001)Si 6H-SiC surface is hydrophobic in
HNO3 and H2SO4 whereas the hydrogen terminated (111) Si surface is not. As the most
significant difference observed between the oxidized and unoxidized SiC surfaces is the
identification of large amounts of fluorine (≈ 1-2 ML) on the as-polished surface, the
hydrophobic nature of this surface is most likely related to fluorine termination (C-F or Si-F)
instead of hydrogen termination.
As previously noted, two fluorine peaks were detected by XPS from the as-
polished/unoxidized (0001)Si 6H-SiC surface after solvent cleaning (see Figure 2.8). The
first peak was located at 685.8 eV and attributed to Si-F bonding based on previous
examinations of fluorinated silicon surfaces [103-107]. The second F 1s peak was detected at
687.5 eV and attributed to either C-Fx, N-Fx, or SiFx bonding [106,107]. Based on the
observation of a third broad C 1s peak at 286.0 eV (FWHM=4.1eV), we suggest that the
35
second F 1s peak is due to C-Fx bonding. Any SiFx species present on the SiC surface after
polishing would have been removed by the DI rinse or HF dip. It is possible that the nitrogen
detected by XPS and AES is simply nitrogen dopants which have segregated or which have
been preferentially left at the SiC surface during polishing. Hence, we propose that the
nitrogen is primarily bonded to Si (i.e. Si-N) at the SiC surface
As a LEED pattern was only observable from as-polished SiC surfaces at high beam
energies of ≈ 200 eV, we suggest that the as-polished SiC surface is terminated with a thin (≈
5-10Å) contamination or disordered/defective layer. This contamination layer is composed
primarily of a mixture of C-C, C-F, Si-F bonded species and is directly responsible for the
hydrophobic nature of the as-polished SiC surfaces. Once this contamination layer has been
completely removed by oxidation of the C-C and C-F bonds, the F 1s and C 1s peaks at 685-
687 and 286.0 eV respectively disappear and the C-C C 1s peak at 283.6 eV shifts to 283.9
eV indicative of more C-O bonding. After oxidation, this surface is irreversibly hydrophilic.
As far as why some hydrophilic as-polished SiC surfaces can be reverted back to being
hydrophobic by a boiling Aqua Regia treatment, the authors speculate that the Aqua Regia
treatment somehow replaces C-O bonds with C-H bonds.
As for the source of the fluorine detected on the as-polished SiC surfaces, the authors
offer two explanations. One possible source of the fluorine are the Fluoroware containers in
which the SiC wafers are shipped from Cree. The containers are made of natural
polypropylene which has a rather high outgassing rate for particles [116]. It has been
previously noted by others working on Si [117], that SiO2 films stored in these containers for
sufficiently long times can become apparently hydrophobic due the large concentrations of
hydrophobic particles found on them. However, we do not think that the wafers investigated
36
here were so contaminated. Therefore we think that the only other source of fluorine should
be Cree's polishing or etching procedure. Unfortunately, Cree has not been willing to divulge
or share any information concerning this. We do note, however, that we have observed
similar contamination layers from Si surfaces etched in CF4 RIE systems [118].
At this point, it is worth considering which surface would be better to work with: the
hydrophilic SiC surface or the as-polished hydrophobic SiC surface. The hydrophobic as
polished SiC surface has the advantage of minimizing trapping of wet chemicals in micro
pipes and reduced particulate contamination from wet chemical processing. However, the
results indicate that a thin contamination layer is responsible for the hydrophobic nature of
as-polished SiC surfaces. The hydrophilic SiC surface produced by thermal oxidation
followed by oxide removal with HF is most likely the more appropriate surface for devices
due to the better crystallinity and surface order. Thermal oxidation of as-polished SiC
surfaces not only removes the thin contamination layer but also oxidizes and removes much
of the subsurface damage present after polishing [96].
2.5.3. Si Passivation Layer
37
We now consider a more controllable passivation layer based on a 20-200Å
amorphous Si capping layer. As described in the results section, an amorphous Si capping
layer behaves similarly in acids and bases to Si (111) surfaces and may be easily removed in
situ by annealing in UHV at 1100°C. Hence, the a-Si passivation allows the application of
knowledge developed from ex situ processing of silicon. The use of this capping layer was
also found to result in lower outgassing rates in vacuum due to the production of a
hydrophobic surface preventing liquids from being trapped in micro pipes. The authors
would like to point out that the a-Si capping layer has the added advantage that it can
potentially be incorporated into already existing processing routes. Potentially, the Si
capping layer could be deposited during cooling from SiC thin film CVD epitaxy. Rupp et al
[119], have already currently demonstrated the ability to control the surface stoichiometry of
SiC epitaxial films by controlling the gas phase composition in their LPCVD system during
cooling. Therefore, the deposition of a 20-200Å amorphous or polycrystalline layer of Si
during cooling after SiC epitaxial growth is possible. Another advantage of the Si capping
layer is that it may be oxidized or nitrided to form the oxide/insulator for MOSFET/MISFET
structures. The advantage here is that the Si capping layer would protect the SiC surface
from various metallic and other contaminants during processing which could effect the
quality of the SiC/SiO2 interface. In instances where a hydrophilic surface is needed the Si
capping layer can be easily made hydrophilic by immersion in HNO3 or H2SO4.
38
2.6. Conclusions
In conclusion, it has been found that removal of a thermal oxide from (0001)Si 6H-
SiC surfaces using 10:1 HF results in a hydrophilic surface terminated primarily with Si-OH
and C-O species. Other crystallographic orientations ((000-1)C, (11-20), and (10-10)) were
also observed to be hydrophilic after oxide removal with 10:1 HF. In contrast, non-
oxidized/as polished (0001)Si 6H-SiC surfaces were observed to be hydrophobic as received.
This surface was observed be to terminated by a thin (5-10Å) disordered contamination layer
composed mainly of C-C, C-F, and Si-F bonded species. Removal of this contamination
layer using RCA SC1 or Piranha etch converted the surface to hydrophilic. As an alternative
passivation layer, a 20-200Å a-Si capping layer was demonstrated to produce a hydrophobic
SiC surface. This a-Si passivation layer was easily removed in situ via thermal desorption at
1100°C.
2.7. Acknowledgments
The authors would like to thank Cree Research Inc., for supplying the SiC wafers
used in these studies. This research was sponsored by the Office of Naval Research and
through the Department of Education via a an Electronic Materials/GAANN fellowship.
39
2.8. References 1. W. Kern, RCA Review, 39, 278 (1978). 2. W. Kern, J. Electrochem. Soc., 137, 1887 (1990). 3. T. Ohmi, J. Electrochem. Soc., 143, 1957 (1996). 4. G.R. Srinivasan and B.S. Meyerson, J. Electrochem. Soc., 134, 1518 (1987). 5. G.R. Srinivasan, J. Cryst. Growth, 70, 201 (1984). Electrical Defects related to cleaning 6. E.H. Snow, A.S. Grove, B.E. Deal, and C.T. Sah, J. Appl. Phys., 36, 1664 (1965). 7. E. Yon, W.H. Ko, and A.B. Kuper, IEEE Trans. on Electron Devices, ED-13, 276 (1966). 8. J.V. Dalton and J. Drobek, J. Electrochem. Soc., 115, 865 (1968). 9. J. Ruzyllo, A.M. Hoff, D.C. Frystak, and S.D. Hossain, J. Electrochem. Soc., 136, 1474 (1989). 10. S.R. Kasi, M. Liehr, P.A. Thiry, H. Dallaporta, and M. Offenberg, Appl. Phys. Lett., 59, 108 (1991). 11. L.J. Huang and W.M. Lau, Appl. Phys. Lett., 60, 1108 (1992). 12. T. Ohmi, T. Imaoka, T. Kezuka, J. Takano, and M. Kogure, J. Electrochem. Soc., 140, 811 (1993). 13. C. Sheng, D. Gong, X. Wei, F. Lu, Q. Wang, H. Sun, and X. Wang, Jpn. J. Appl. Phys., 33, 2276 (1994). Structural Defects from cleaning 14. B.A. Joyce, J.H. Neave, and B.E. Watts, Surface Science, 15, 1 (1969). 40
15. J.H. McFee, R.G. Swartz, V.D. Archer, S.N. Finegan, and L.C. Feldman, J. Electrochem. Soc., 130, 214 (1983). 16. S. Nagao, K. Higashitani, Y. Akasaka, and H. Nakata, J. Appl. Phys., 57, 4589 (1985). 17. B.S. Meyerson, E. Ganin, D.A. Smith, and T.N. Nguyen, J. Electrochem. Soc., 133, 1232 (1986). 18. A.J. Pidduck, D.J. Robbins, A.G. Cullis, D.B. Gasson, and J.L. Glasper, J. Electrochem. Soc., 136, 3083 (1989). 19. A. Miyauchi, Y. Inoue, M. Ohue, N. Momma, T. Suzuki, and M. Akiyama, J. Electrochem. Soc., 137, 3257 (1990). 20. A. Miyauchi, Y. Inoue, T. Suzuki, and M. Akiyama, Appl. Phys. Lett., 57, 676 (1990). 21. M. Racanelli, D.W. Greve, M.K. Hatalis, and L.J. van Yzendoorn, J. Electrochem. Soc., 138, 3783 (1991). 22. Eaglesham, G.S. Higashi, and M. Cerullo, Appl. Phys. Lett., 59, 685 (1991). 23. C. Galewski, J. Lou, and W.G. Goldham, IEEE Trans. Semicond. Manfact., 3, 93 (1990). 24. M.K. Sanganeria, M.C. Ozturk, G. Harris, K.E. Violette, I. Ban, C.A. Lee, and D.M Maher, J. Electrochem. Soc., 142, 3961 (1995). 25. F.K. LeGoues, MRS Bulletin, 21, 38, (1996). Silicon cleaning and oxidation 26. D.E. Aspnes and A.A. Studna, Appl. Phys. Lett., 39, 316 (1981). 27. B.F. Phillips, D.C. Burkman, W.R. Schmidt, and C.A. Peterson, J. Vac. Sci. Technol. A, 1, 646 (1983). 28. N. Yabumoto, K. Saito, M. Morita, and T. Ohmi, Jpn. J. Appl. Phys., 30, L419 (1991). 29. W.A. Cady and M. Varadarajan, J. Electrochem. Soc., 143, 2064 (1996). 30. G. Hollinger and F.J. Himpsel, Appl. Phys. Lett., 44, 93 (1984). 31. G. Hollinger and F.J. Himpsel, J. Vac. Sci. Technol. A, 1, 640 (1983). 32. C.R. Inomata, H. Ogawa, K. Ishikawa, and S. Fujimura, J. Electrochem. Soc., 143, 2995 (1996). 33. A. Licciardello, O. Puglisi, and S. Pignataro, Appl. Phys. Lett., 48, 41 (1986). 41
34. P. Singer, Semicond. Inter., 10, 8 (1995). 35. A. Bousetta, C.C. Hsu, and Y. Wang, Surface Rev. Lett., 2, 171 (1995). 36. T. Konishi, T. Yao, M. Tajima, H. Ohshima, H. Ito, and T. Hattori, Jpn. J. Appl. Phys., 31, L1216 (1992). 37. T. Konishi, K. Uesugi, K. Takaoka, S. Kawano, M. Yoshimura, and T. Yao, Jpn. J. Appl. Phys., 32, 3131 (1993). 38. W. Ranke, Surface Science, 369, 137 (1996). 39. M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and M. Ohwada, J. Appl. Phys., 68, 1272 (1990). Metal, Particles, Oxidation of Silicon 40. T. Ohmi, T. Imaoka, I. Sugiyama, and T. Kezuka, J. Electrochem. Soc., 139, 3317 (1992). 41. M. Itano, T. Kezuka, M. Ishii, T. Unemoto, and M. Kubo, J. Electrochem. Soc., 142, 971 (1995). 42. M. Itano, F.W. Kern, M. Miyashita, and T. Ohmi, IEEE Trans. Semicond. Manufact., 6, 258 (1993). 43. M. Itano, F.W. Kern, R.W. Rosenberg, M. Miyashita, I. Kawanabe, and T. Ohmi, IEEE Trans. Semicond. Manufact., 5, 114 (1992). 44. T. Ohmi, T. Isagawa, M. Kogure, and T. Imaoka, J. Electrochem. Soc., 140, 804 (1993). 45. H. Mishima, T. Yasui, T. Mizuniwa, M. Abe, and T. Ohmi, IEEE Trans. Semicond. Manufact., 2, 69 (1989). 46. H. Morinaga, M. Suyama, and T. Ohmi, J. Electrochem. Soc., 141, 2834 (1994). 47. T. Ohmi, M. Miyashita, M. Itano, T. Imaoka, and I. Kawanabe, IEEE Trans. Electron Devices, 39, 537 (1992). 48. H. Morinaga, T. Futatsuki, T. Ohmi, E. Fuchita, M. Oda, and C. Hayashi, J. Electrochem. Soc., 142, 966 (1995). 49. T. Ohmi, T. Isagawa, T. Imaoka, and I. Sugiyama, J. Electrochem. Soc., 139, 3336 (1992). 50. H. Mishima, T. Ohmi, T. Mizuniwa, and M. Abe, IEEE Trans. Semicond. Manufact., 2, 121 (1989). 42
51. M. Miyashita, T. Tusga, K. Makihara, and T. Ohmi, J. Electrochem. Soc., 139, 2133 (1992). HF related processing of silicon 52. C.H. Bjorkman, J.L. Alay, H. Nishimura, M. Fukada, T. Yamazaki, and M. Hirose, Appl. Phys. Lett., 67, 2049 (1995). 53. Y. Morita and H. Tokumoto, Appl. Phys. Lett., 67, 2654 (1995). 54. B.S. Meyerson, F.J. Himpsel, and K.J. Uram, Appl. Phys. Lett., 57, 1034 (1990). 55. L. Li, H. Bender, T. Trenkler, P.W. Mertens, M. Meuris,W. Vandervorst, and M.M. Heyns, J. Appl. Phys., 77, 1323 (1995). 56. M. Grundner and H. Jacob, Appl. Phys. A, 39, 73 (1986). 57. M.R. Houston and R. Maboudian, J. Appl. Phys., 78, 3801 (1995). 58. T. Takahagi, I. Nagai, A. Ishitani, H. Kuroda, and Y. Nagasawa, J. Appl. Phys., 64, 3516 (1988). 59. B.R. Weinberger, G.G. Peterson, T.C. Eschrich, and H.A. Krasinski, J. Appl. Phys., 60, 3232 (1986). 60. S. Verhaverbeke, I. Teerlinck, C. Vinckier, G. Stevens, R. Cartuyvels, and M.M Heyns, J. Electrochem. Soc., 141, 2852 (1994). 61. H. Proksche, G. Nagorsen, and D. Ross, J. Electrochem. Soc., 139, 521 (1992). 62. J.S. Judge, J. Electrochem. Soc., 118, 1773 (1971). 63. G.W. Trucks, K. Raghavachari, G.S. Higashi, and Y.J. Chabal, Phys. Rev. Lett., 65, 504 (1990). 64. T. Suzuki ans S. Adachi, Jpn. J. Appl. Phys., 33, 5599 (1994). 65. E. Yablonovitch, D.L. Allara, C.C. Chang, T. Gmitter, and T.B. Bright, Phys. Rev. Lett., 57, 249 (1986). 66. G.S. Higashi, R.S. Becker, Y.J. Chabal, and A.J. Becker, Appl. Phys. Lett., 58, 1656 (1991). 67. G.S. Higashi, Y.J. Chabal, G.W. Trucks, and K. Raghavachari, Appl. Phys. Lett., 56, 656 (1990).
43
68. V.A. Burrows, Y.J. Chabal, G.S. Higashi, K. Raghavachari, and S.B. Christman, Appl. Phys. Lett., 53, 998 (1988). 69. Y.J. Chabal, G.S. Higashi, K. Raghavachari, and V.A. Burrows, J. Vac. Sci. Technol. A, 7, 2104 (1989). (001) 3C-SiC 70. R. Kaplan, J. Appl. Phys., 56, 1636 (1984). 71. M. Dayan, J. Vac. Sci. Technol. A, 4, 38 (1986). 72. J.M. Powers, A. Wander, P.J. Rous, M.A. Van Hove, and G.A. Somorjai, Phys. Rev. B, 44, 11159 (1991). 73. V.M. Bermudez and R. Kaplan, Phys. Rev. B, 44, 11149 (1991). 74. M Balooch and D.R. Olander, Surface Science, 261, 321 (1992). 75. M. Dayan, Surface Science Letters, 149, L33 (1985). 76. T.M. Parrill and Y.W. Chung, Surface Science, 243, 96 (1991). 77. B. Jorgensen and P. Morgan, J. Vac. Sci. Technol. A, 4, 1701 (1986). 78. V.M. Bermudez, J. Appl. Phys., 66, 6084 (1989). (0001) 6H-SiC 79. A.J. van Bommel, J.E. Crombeen, and A. van Tooren, Surface Science, 48, 463 (1975). 80. F. Bozso, L. Muehlhoff, M. Trenary, W.J. Choyke, and J.T. Yates, Jr., J. Vac. Sci. and Technol. A, 2, 1271 (1984). 81. R. Kaplan and T.M. Parrill, Surface Science Letters, 165, L45 (1986). 82. R. Kaplan, Surface Science, 215, 111 (1989). 83. S. Nakanishi, H. Tokutaka, K. Nishimori, S. Kishida, and N. Ishihara, Applied Surface Science, 41/42, 44 (1989). 84. A.N. Andreev, M.M. Anikin, A.L. Syrkin, and V.E. Chelnokov, Semiconductors, 28, 577 (1994). 85. J.A. Dillon, Jr., R.E. Schlier, and H.E. Farnsworth, J. Appl. Phys., 30, 675 (1959). 86. J.M. Powers and G.A. Somorjai, Surface Science, 244, 39 (1991).
44Silicon Carbide
87. L.M. Porter, R.F. Davis, J.S. Bow, M.J. Kim, R.W. Carpenter, R.C. Glass, J. Mater. Res., 10, 668 (1995). 88. H. Tsuchida, I. Kamata, and K. Izumi, Jpn. J. Appl. Phys., 34, 6003 (1995). 89. Y. Mizokawa, S. Nakanishi, O. Komoda, S. Miyase, H.S. Diang, C. Wang, N. Li, and C. Jiang, J. Appl. Phys., 67, 264 (1990). 90. U. Starke, Ch. Bram, P.R. Steiner, W. Hartner, L. Hammer, K. Heinz, K. Muller, Appl. Surf. Sci., 89, 175 (1995). 91. M.E. Lin, S. Strite, A. Agarwal, A. Salvador, G.L. Zhou, M. Teraguchi, A. Rockett, and H. Morkoc, Appl. Phys. Lett., 62, 702 (1993). 92. A.N. Anrdeev, M.M. Anikin, A.L. Syrkin, and V.E. Chelnokov, Semiconductors, 28, 377 (1994). 93. M. Ghamnia, C. Jardin, D. Kadri, and M. Bouslama, Vacuum, 47, 141 (1996). 94. A.O. Evwaraye, S.R. Smith, M. Skowronski, and W.C. Mitchell, J. Appl. Phys., 74, 5269 (1993). 95. G. Ramis, P. Quintard, M. Cauchetier, G. Busca, and V. Lorenzelli, J. Amer. Ceram. Soc., 72, 1692 (1989). 96. W. Qian, M. Skowronski, G. Augustine, R.C. Glass, H.M. Hobgood, and R.H. Hopkins, J. Electrochem. Soc., 142, 4290 (1995). 97. S.F. Avramenko, V.V. Vainberg, E.F. Venger, S.I. Kirillova, V.S. Kiselev, V.E. Primachenko, and V.A. Chernobai, Semiconductors, 28, 572 (1994). 98. R.F. Davis, Advances in Ceramics, 23, 477 (1987). 99. R.F. Davis, G. Kelner, M. Shur, J. Palmour, J.A. Edmond, Proc. of the IEEE, 79, 677 (1991). 100. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10, 1237 (1992). 101. S.W. King, M.C. Benjamin, R.S. Kern, R.J. Nemanich, and R.F. Davis, Proc. of the MRS, 423, 563 (1996) 102. J. van der Weide, Ph.D. dissertation, NCSU (1994). 103. T.J. Chuang, J. Appl. Phys., 51, 2614 (1980). 104. C.D. Stinespring and A. Freedman, Appl. Phys. Lett., 48, 718 (1986).
45105. J.H. Thomas and L.H. Hammer, J. Vac. Sci. Technol. B, 5, 1617 (1987).
106. T.J. Chuang, H.F. Winters, and J.W. Coburn, Surface Science, 2, 514 (1978). 107. J.W. Coburn, H.F. Winters, and T.J. Chuang, J. Appl. Physics, 48, 3532 (1977). 108. P.H. Holloway and H.J. Stein, J. Electrochem. Soc., 123, 723 (1976). 109. K. Okada, K. Fukuyama, and Y. Kameshima, J. Am. Ceram. Soc., 78, 2021 (1995). 110. B. Douglas, D.H. McDaniel, and J.J. Alexander, Concepts and Models of Inorganic Chemistry, p. 74, John Wiley & Sons, New York, 1983. 111. A. Zangwill, Physics at Surfaces, p. 104-109, Cambridge University Press, New York, 1988. 112. A.W. Adamson, Phyical Chemistry of Surfaces, p. 218-238, John Wiley & Sons, Inc., New York, 1990. 113. B. Popping, A. Deratani, B. Sebille, N. Desbois, J.M. Lamarche, and A. Foissy, Colloids and Surfaces, 64, 125 (1992). 114. P.K. Whitman and D.L. Feke, J. Am. Ceram. Soc., 71, 1086 (1988). 115. L.S. Cerovic and S.K. Milonjic, J. Am. Ceram. Soc., 78, 3093 (1995). 116. Fluoroware data sheet. 117. Personal communication, Mike Jolly, Semitool, Inc. 118. J.P. Barnak, H.Y. Ying, S.W. King, and R.J. Nemanich, to be published. 119. R. Rupp, P. Lanig, J. Volkl, and D. Stephani, Proc. of the MRS, 423, 253 (1996).
46
3. Dry Ex Situ Cleaning Processes for (0001)Si 6H-SiC Surfaces
To be Submitted for Consideration for Publication
to the
Journal of the Electrochemical Society
by
Sean W. King, Robert J. Nemanich, and Robert F. Davis.
Department of Materials Science and Engineering
North Carolina State University
Raleigh, NC 27695
3.1. Abstract
47
A completely dry ex situ cleaning process based on UV/O3 oxidation for surface
carbon contamination removal and HF vapor exposure for oxide removal has been
demonstrated for (0001)Si 6H-SiC surfaces. Using x-ray photoelectron spectroscopy (XPS)
analysis for comparison, this cleaning procedure has been demonstrated to reduce residual
surface carbon and oxide contamination to levels equivalent to or better than conventional
wet chemical ex situ processing. Specifically, XPS showed that (0001)Si 6H-SiC surfaces
exposed to UV generated ozone were oxidized resulting in the formation of both carbon and
silicon oxides. This was clearly illustrated in XPS by the formation of a broad Si-O Si 2p
peak at 102.4 eV (FWHM=2.1 eV) and a shift in the surface C1s peak from 283.6 to 284.2
eV. A reduction in the amount of surface carbon was evidenced by an increase in the ratio of
the SiC C1s peak to the surface C1s from 0.8 to 2.7 after the UV/O3 treatment. Removal of
the UV/O3 silicon oxide via exposure to the equilibrium vapor from a 10:1 buffered HF
solution was deduced from the absence (below the XPS detection limit) of the Si-O Si 2p
peak at 102.4 eV. Significant amounts of fluorine remained on the surface after the HF vapor
exposure suggesting that the process results in a fluorine terminated surface.
48
3.2. Introduction
For SiC to succeed as the semiconductor/substrate of choice for high frequency, high
temperature, high power devices and as a substrate for III-N heteroepitaxy, a considerable
reduction in defects (line, planar, point, etc.) must be achieved [1,2]. Following Si
technology, where surface cleaning and preparation are critical first steps in all processes [3-
5], a continued reduction in defects in SiC should be expected as a result of improved SiC
wafer surface cleaning techniques. In Si technology for example, improper removal of
surface contamination and oxides prior to Si homoepitaxy has been shown to result in an
increase in the density of line and planar defects in epitaxial films from < 104/cm2 to >
1010/cm2 [6-8]. The increased defect densities were in turn found to correspond with a
reduction of device yield [6]. In the case of heteroepitaxy, studies on SixGe1-x alloy growth
on Si (100) have additionally shown that surface defects produced in the Si substrate by
residual organic/carbon contamination act as the preferred sites for misfit dislocation
generation [9]. These examples clearly illustrate that surface preparation and cleaning should
be equally important to the control of defects in both homo and heteroepitaxial growth of SiC
and III-V nitrides on (0001) 6H-SiC.
Due to a limited number of studies concerned with ex situ SiC cleaning practices [10-
13], most SiC ex situ wet chemical processing has been based on processes specifically
developed for and employed in Si technology [13,14]. SiC ex situ cleaning/surface
preparation has typically consisted of some variation of solvent degreasing, organic
contaminant removal using RCA or Piranha cleans, and finishing with oxide removal using
HF based solutions [10-14]. An important assumption underlying the use of these procedures
49
is that the SiC surfaces should behave similar to silicon surfaces in these wet chemicals. In a
previous study we have provided examples of where this assumption fails, specifically with
regard to oxide removal from SiC surfaces using an HF dip process. In Si technology, oxide
removal with a dilute HF etch is known to generate a hydrophobic, hydrogen terminated
surface, stable against oxidation in air for several hours [15-20]. However, we have
previously shown that SiC surfaces are inherently hydrophilic after oxide removal with HF
due to a preference for OH termination [21]. The hydrophilic surface allows water and HF to
become trapped in micropipes in the SiC wafer which can lead to large concentrations of
oxygen and fluorine at the SiC-dielectric interface if not properly outgassed. In order to
produce a hydrophobic surface, passivation layers based on silicon or fluorocarbons were
required [21].
An alternative to the use of passivation layers to form hydrophobic SiC surfaces
would be to develop a completely dry cleaning process and thus remove the need for wet
chemical processing. In Si and GaAs technology, dry removal of carbon contaminants from
surfaces using UV/O3 oxidation has become an alternative to wet chemical processing [22-
30]. In the UV/O3 oxidation process, UV radiation from a Hg lamp (specifically the 184.9
nm line) is used to photo excite molecular oxygen (O2) and generate ozone (O3) [22].
Additionally, the 253.7 nm line of Hg assists in removal of carbon contaminants as the light
is adsorbed by most hydrocarbons and excites C-H and C-C bonds [22]. Removal of the
UV/O3 generated oxide is typically achieved by wet chemical processing and/or in situ
thermal desorption [25,26,28,30]. However, Iyer et al [32] have shown that the equilibrium
vapor from an HF solution can be alternatively used to remove the oxide from a silicon wafer
via a dry process. Thus the combined use of UV/O3 oxidation for removal of carbon
50
contaminants and HF vapor exposure for oxide removal represents a completely dry cleaning
process. Use of completely dry processing techniques may eliminate the need for large
quantities of expensive high purity chemicals while simultaneously reducing the costs for
disposal of these toxic materials [32-33].
In this paper, we demonstrate for the first time a completely dry cleaning process for
(0001)Si 6H-SiC surfaces which is based on the combined use of UV/O3 oxidation and HF
vapor cleaning. This clean has been found to be equivalent to or better than typical wet
chemical processes in terms of residual surface carbon and oxide contamination levels as
measured by XPS. The combined UV/O3-HF vapor treatment eliminates the need for a
hydrophobic SiC surface and avoids the use of exotic passivation layers.
3.3. Experimental Procedure
As polished on axis, n-type (typically Nd=1018/cm3) (0001)Si 6H-SiC wafers
supplied by Cree Research, Inc. were used in these experiments. Selection of the as polished
wafers for examination was based on previous investigations which showed these surfaces to
be terminated with a thin (˜5-10Å) contamination layer of C-C, C-F, and Si-F bonded species
[21]. Prior to UV/O3 oxidation, each wafer was first ultrasonically degreased in
trichloroethylene, acetone, and methanol for 10 min. each. The experimental system
employed for the UV/O3 exposures described in this study employed a high intensity Hg
lamp positioned in close proximity (≈ 1 cm) to the SiC wafer (see Figure 1). In order to
increase the concentration of generated O3 (i.e. to increase the oxidation rate), the UV/O3
box was purged with 1 L/sec O2 during the UV exposure. Further details of this process
51
have been described previously [22,26]. The HF vapor exposures were achieved by simply
positioning the SiC wafer in ambient air within approximately 5 mm of a 10:1 buffered HF
solution for times ranging from 5-30 min. (see Figure 1). Condensation of HF on the SiC
surface was not observed for exposures of this length.
Surfaces prepared in the above manner were subjected to surface analysis in an
integrated ultra-high vacuum system incorporating the following analytical techniques: x-ray
photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES), electron energy loss
spectroscopy (EELS), and low energy electron diffraction (LEED). Details of this system are
given elsewhere [34]. After each treatment above, the SiC wafer was mounted to a
molybdenum sample holder and loaded into the load lock for subsequent analysis by AES,
XPS, EELS, and LEED. The XPS analysis was performed using the Al anode (hν=1486.6
eV) at 20 mA and 12 kV. The AES spectra were obtained using a beam voltage of 3 keV and
an emission current of 1 mA. The EELS spectra were obtained using a 100 eV electron beam
and an emission current of 1 mA. The LEED was performed using rear view optics, a beam
voltage of approximately 115 eV, and an emission current of 1 mA. Calibration of the XPS
binding energy scale was performed by measuring the position of the Au 4f7/2 core level
(from a ~1 µm thick Au film) and shifting the spectra such that the peak position occurred at
the accepted value of 83.98 eV [15].
52
b.)
HF
Hg UV Lamp
hν=185nm 254 nm
O3
SiC
a.)
Air or O2
Figure 3.1. (a) Schematic of UV/ O3 oxidation system. (b) Schematic of HF vapor procedure.
53
3.4. Results
3.4.1. Solvents and UV/O3:
Figures 3.2(a) and (b) show XPS spectra of the C 1s core level obtained from a as-
polished, solvent cleaned (0001)Si 6H-SiC surface followed by a UV/O3 oxidation treatment.
As is indicated in Figure 3.2(a), a broad C 1s feature was obtained. Previous analysis of this
spectrum [21] revealed the presence of three C 1s peaks centered at 282.5, 283.6, and 286.0
eV. The most intense peak at 282.5 is associated with carbon bonded to silicon in SiC.
Based on the large FWHM (full width half maximum) of 2.6 and 4.5 eV, the latter two peaks
were respectively assigned to a mixture of C-C and C-Hx (283.6 eV), and C-Fx (286.0)
bonded carbon [35-37]. The presence of C-Fx species was further supported by the XPS
spectra of the F 1s core level (see Figure 3.3(a)). The spectra from the as polished (0001)Si
6H-SiC surface after solvent cleaning shows two F 1s peaks located at 685.4 and 687.2 eV.
These two peaks have been assigned to Si-F [40-42] and C-F [35-37] bonding respectively.
The presence of a thin "fluorocarbon" contamination layer is also indicated by the inability to
obtain a LEED pattern from these surfaces at beam energies (Ep) < 200 eV.
Following a 2 hour UV/O3 exposure, the C-Fx C 1s peak was observed to disappear
and the C1s peak associated with C-C and C-Hx was observed to shift from 283.6 to 284.2
eV (see Figure 3.2(b) and Table 3.1). In contrast, the SiC C1s peak showed an increase in
intensity and shifted by only 0.1 eV to 282.6 eV. In addition, the ratio of the C1s peak
intensities associated with the SiC and the surface carbon (uncorrected for sensitivity factors)
was observed to increase from 0.8 to 2.7 (see Table 3.2). This result indicates that the
UV/O3 process removes the contamination layer via oxidation. The shift and reduction in
the C-C and CHx C1s peak is consistent with the formation of C-O bonding at the surface
and removal of some surface carbon via desorption of CO and CO2 [22,24,37]. Removal of
54
the contamination layer was also supported by the complete disappearance (below the XPS
detection limit) of the F 1s peak at 687.2 eV after the UV/O3 treatment (see Figure 3.3b).
Only a slight trace of the lower binding energy F 1s peak was detected, and it was observed
to shift by 0.5 eV to 685.9 eV (see Figure 3.3(b)).
The formation of silicon oxides on the SiC surface is displayed in the XPS of the Si
2p core level from the (0001)Si 6H-SiC surface before and after UV/O3 treatment (see
Figures 3.4(a) and (b)). As shown in Figure 3.4(a), a single Si 2p peak was detected before
UV/O3 oxidation. The line shape of this Si 2p peak is asymmetric suggesting the possibility
of a Si-O bonding peak on the higher BE (binding energy) side. Unfortunately,
deconvolution of this peak is complicated by the fact that the Si 2p peak is an unresolved
doublet (i.e. Si 2p3/2,1/2 ) and fitting a second peak to this spectrum showed only a small
peak at 102.2 eV with a FWHM narrower than the substrate peak (i.e. 1.1 vs. 1.4 eV). As
such, it was not possible to conclusively detect a Si-O peak prior to the UV/O3 exposure.
However after the UV/O3 exposure, a broad Si-O peak centered at 102.4 eV (FWHM=2.1
eV) was clearly observed (see Figure 3.4(b)). The width of the Si 2p peak at 102.4 eV
indicates that silicon in +2,+3, and +4 oxidation states is bonded to the oxygen (i.e. SiOx)
[16,38,39]. Based on the attenuation of the Si-C Si 2p peak, the thickness of the SiOx layer
was estimated to be < 20Å.
55
280 282 284 286 288 290
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
Figure 3.2. XPS of the C 1s core level from (0001)Si 6H-SiC after (a) solvent cleaning, sequentially followed by (b) UV/O3, and (c) HF vapor exposures.
680 682 684 686 688 690 692
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
Si-F C-F
56
Figure 3.3. XPS of the F1s core level from (0001)Si 6H-SiC after (a) solvent cleaning, sequentially followed by (b) UV/O3, and (c) HF vapor exposures.
96 98 100 102 104 106 108
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
Figure 3.4. XPS of the Si 2p core level from (0001)Si 6H-SiC after (a) solvent cleaning, followed by (b) UV/O3, and (c) HF vapor exposures.
Table 3.1. XPS core level binding energies (eV) from (0001)Si 6H-SiC surfaces after various exposures. The full width half maxima (Γ) are also included.
Si 2p, Γ C 1s, Γ O 1s, Γ F 1s, Γ Solvents 100.4, 1.5 282.5, 1.1 531.6, 2.3 685.4, 1.8 283.6, 2.6 687.2, 2.7 286.0, 4.5 UV/O3 100.5, 1.4 282.6, 1.1 532.1, 2.4 685.9, 1.9 102.4, 2.1 284.2, 2.1 HF Vapor 100.5, 1.5 282.6, 1.1 531.8, 2.3 685.8, 1.9 283.8, 2.8
57
Table 3.2. The XPS core level intensity ratios from (0001)Si 6H-SiC after various treatments (uncorrected for differences in sensitivity factors). The first ratio represents the C1s peaks attributed to carbon in SiC and to adventitious or surface carbon.
SiC C1s/surface C1s Si2p/O1s Si2p/F1s Solvents 0.8 1.1 5.4 UV/O3 2.7 0.3 10.6 HF Vapor 1.7 1.3 0.5
3.4.2. HF vapor
After oxidation of the (0001)Si 6H-SiC surface using a UV/O3 exposure, removal of
the thin silicon oxide layer was achieved by a 30 min. exposure of the SiC surface to a vapor
from a 10:1 buffered HF solution. As shown in Figure 3.4(c), the higher binding energy Si
2p peak centered at 102.4 eV was not detectable after the HF vapor exposure. Shorter HF
vapor exposures were observed to result in an observable Si-O Si2p peak but at larger
binding energies (≈104 eV). After the HF vapor treatment, the amount of surface carbon was
monitored from the intensity ratio of the SiC C1s to the surface carbon C1s peaks. The ratio
was observed to decrease from 2.7 to 1.7 after the HF vapor exposure indicating an increase
in the amount of surface carbon contamination. The surface C1s peak was likewise observed
to shift back to 283.8 eV and displayed a large increase in FWHM from 2.1 to 2.8 eV.
However, the SiC to surface carbon intensity ratio of 1.7 after the HF vapor treatment is still
much larger (i.e. less surface carbon) than the 0.8 value that was found after the solvent
cleaning process. In addition to an increase in surface carbon after the HF vapor treatment,
the amount of fluorine on the surface was also observed to increase significantly (see Figure
3.3(a)). Prior to the HF vapor treatment, the Si/F ratio was equal to 10.6. However, after the
HF vapor treatment the Si/F ratio decreased to 0.56. Further the binding energy of the F 1s
58
was not observed to shift but remained centered at 685.9 eV which suggests that fluorine is
bonded only to silicon atoms at the surface.
3.5. Discussion
3.5.1. UV/O3 Oxidation
In the above section, it was demonstrated that exposure of (0001)Si 6H-SiC surfaces
to ozone generated by a Hg UV lamp oxidized and removed adventitious and CFx bonded
carbon from the SiC surface (see Figure 3.5(a) and (b)). This resulted in an increase of the
SiC/non-SiC carbon ratio from 0.8 to 2.7. This result is in agreement with previous studies
of UV/O3 oxidation of Si and GaAs surfaces which have also shown a reduction of carbon
contaminants [22-30]. Despite the long exposure (2 hr.), some adventitious carbon or surface
carbon, however, was observed to be present on the SiC surface after the UV/O3 treatment.
Some of this surface carbon is likely due to recontamination of the SiC surface during sample
transfer and mounting in a laboratory ambient prior to insertion into vacuum. However,
contamination levels of this magnitude are usually not observed from silicon wafers cleaned
in the same environment. Alternatively, the remaining surface carbon could be due to carbon
trapped in the SiOx layer and/or carbon bonded to both silicon and oxygen at the SiC/SiOx
interface. We also note that the studies of Fominski et al [24] and Baunack and Zehe [30]
report incomplete removal of carbon contaminants from Si surfaces using O3 generated from
a Hg lamp. In fact Fominski et al [24] found that it was necessary to employ deeper UV
radiation from a D2 lamp and immerse the wafer in an O2/NF3/H2 gas mixture.
59
F F F F F F FOH
OH
OH
OH
F F F F F F FOH
OH
F
5-10Å C-C, CHx, and CFxcomtamination layer
- 20Å SiOx layer
UV/O3
HF Vapor
Figure 3.5. Schematic illustrating the surface termination from UV/O3 and HF vapor exposures on (0001)Si 6H-SiC.
The shift in the position of the surface C 1s peak from 283.6 to 284.2 eV with UV/O3
oxidation is consistent with the oxidation of C-C, CHx, and CFx bonds to form CO. For HF
dipped Si wafers, it has been previously determined that residual carbon contaminants with C
1s peaks positions of 284.6, 286.3, and 288.4 eV are composed mostly of C-H2, C-O, and O-
C=O bonded carbon respectively which illustrates the trend to higher binding energies for C-
60
O bonds [43]. Unfortunately, a direct comparison between the surface C1s peak position and
bonding configuration for both Si and SiC surfaces is complicated by the probable existence
of 0.5-1.0 eV of band bending at the SiC surface due to surface Fermi level pinning.
Although the UV/O3 oxidation treatment was not completely successful in removing
all of the non-carbidic carbon from the SiC surface, a comparison of this technique with wet
chemical processes does show the utility of the technique. In a previous study [21], we
examined the effect that standard wet chemical treatments such as RCA SC1 and Piranha
etch have on the removal of the same carbon surface contamination observed in this study.
Table 3.3 provides a direct comparison of the SiC C1s/surface C1s and Si2p/O1s intensity
ratios for each treatment. As can be seen, the UV/O3 treatment provides the highest SiC
C1s/surface C1s ratio of all the treatments examined. We also note that Afanas'ev et al [33]
has recently found UV/O3 oxidation to be a useful cleaning or pre-oxidation procedure prior
to thermal oxide growth for p and n-SiC/SiO2 MOS structures. In comparison to RCA
cleaned SiC samples, they observed that the UV/O3 pre oxidation treatment resulted in a
reduction of defects (fast interface states) and a decrease in positive charge at the p-SiC/SiO2
interface from 2x1012/cm2 to 6-8x1011/cm2. They suggested that the reduction in positive
charge by the UV/O3 treatment was due to the removal of carbon clusters (i.e. C-C bonding)
that remain on the SiC surface after the growth of epitaxial layers and which are not removed
by RCA cleaning or the thermal oxidation process itself. We note that this suggestion is
supported by our observation that UV/O3 oxidation removes the non-SiC carbon (C-C, CHx,
and CFx) from SiC surfaces.
Finally, we would like to emphasize the ability of a room temperature UV/O3
treatment to grow or form thin SiOx layers ( < 20Å) on SiC. In a separate study [21], we
investigated the ability to oxidize SiC surfaces using other wet chemical treatments
commonly employed to form passivating oxides on silicon. As shown in Table 3.3, the
Si2p(Si-C)/O1s intensity ratio of 0.3 resulting after a UV/O3 exposure is much lower than the
61
≈ 1.0 obtained after wet chemical treatments such as boiling Aqua Regia or RCA SC1. In
fact, the Si2p(Si-C)/O1s intensity ratio after the RCA SC1 clean is not significantly different
from Si2p(Si-C)/O1s ratio of 1.1 observed from solvent cleaned SiC surfaces. However, this
observation is consistent with the known inability of any of these acids to etch SiC.
Therefore, the ability of UV/O3 to grow a thin (10-20Å) passivating oxide is an added
benefit over conventional wet chemical processing. Also, the ability to form the passivating
oxide at room temperature is an additional bonus compared to thermal oxidation of SiC
which typically requires temperatures of 1000-1200°C [45]. Finally, it is noted that the Si 2p
spectrum obtained from SiC treated by UV/O3 oxidation (see Figure 4b) bares a striking
resemblance to the Si 2p spectra obtained from a SiC sample previously exposed to a total
fluence of 9x1021 oxygen atoms (i.e. O instead of O2) during low earth orbit on a satellite
[46]. This suggests that UV/O3 oxidation could be used to simulate the operational
conditions of SiC devices in outer space and other harsh oxidizing environments.
Table 3.3. Summary of SiC-C1s/surface C1s and Si/O intensity ratios from XPS data (uncorrected for differences in sensitivity factors).
Treatment SiC C1s/surface C1s Si2p/O1s Solvents 0.8 1.1 Piranha 1.1 0.9 RCA SC1 2.2 1.0 Aqua Regia 1.2 1.2 UV/O3 2.7 0.3 HF vapor 1.3 1.2
3.5.2. HF Vapor
62
As Figures 3.4(b) and 3.4(c) show, the equilibrium vapor from an HF solution alone
can be used to effectively remove thin silicon oxide layers from SiC surfaces. Though Figure
3.4(c) shows the complete removal (below detection limits) of the higher BE Si 2p peak at
102.4 eV, some oxygen was observed to remain on the SiC surface (probably in the form of
suboxides or hydroxides of silicon and carbon i.e.C3-Si-O(H) and Si3-C-O). Table 3.3
shows that the resulting Si/O XPS intensity ratio after the vapor treatment was observed to
increase from 0.3 to 1.3. The Si/O intensity ratio of 1.3 compares well with the value of 1.4
obtained from a SiC surface after removal of a thermal oxide using a 10:1 HF dip [21].
These results indicate that HF vapor exposure is equally as effective as an HF dip in
removing surface silicon oxides from SiC surfaces. It should be noted, however, that the
silicon oxide etch rates for HF vapor and an HF dip are substantially different. In the HF
vapor case, a 30 minute exposure was required to remove only 10-20Å of surface oxide
resulting from a UV/O3 treatment, whereas in the HF dip case, only 10 minutes were
required to remove 1000Å of thermal oxide.
Unfortunately, the SiC C1s/surface C1s intensity ratio was observed to decrease
from 2.7 to 1.7 after the HF vapor treatment. Some of this increased surface carbon may be
attributable to the ambient exposure during and after the vapor treatment. An in situ HF
vapor exposure, however, could eliminate this recontamination. Takayuki et al [25] has
previously demonstrated the removal of native oxides on Si (001) using photoexcited fluorine
gas. Our results presented here suggest that an in situ HF vapor exposure should work as
well.
The authors also note that the HF vapor exposure results in significant amounts of
fluorine on the (0001)Si 6H-SiC surface (see Figure 3.5c). The observed fluorine coverage
following the HF vapor treatment was 3-4 times larger than that previously observed from
(0001)Si 6H-SiC wafers dipped in 10:1 HF and blown dry (without a de-ionized water rinse).
The fluorine surface coverage approaches that of 1/2 to a full monolayer. As the peak
63
position of the F 1s core level after the HF vapor treatment remains essentially unchanged at
685.9 eV (i.e. Si-F bonding), this suggests that the HF vapor treatment leaves a Si-F
terminated SiC surface [40-42]. These results are in contrast with those of Iyer et al [31] for
Si (100) in which no fluorine was detected by XPS and hydrogen termination was confirmed
by TPD. However, in a previous study on HF wet chemical processing of SiC [21], it was
argued that OH- termination would be preferred for (0001)Si 6H-SiC as opposed to hydrogen
due to the polarity of the Si-C bond. Termination of the (0001)Si SiC surface with OH- tends
to cancel the dipole created by the Si-C bond whereas termination with H does not (see
Figure 3.6). Similarly, F- ions (from HF vapor) could also cancel this dipole and in the HF
vapor exposure the F- ions are more readily available than OH- ions. Therefore, fluorine
termination of SiC surfaces should be expected after HF vapor processes as opposed to either
H+ or OH- termination.
64
Si
C C Cδ−
δ+
F
Si
C C C
δ−
δ−
δ+OH-
Si
C C Cδ−
δ+
σ+σ−
σ+σ−
σ+σ−
σ+σ−
[0001]
Ε
(a)
(b) ΟΗ
−
ΟΗ−
ΟΗ−
ΟΗ−
ΟΗ−
ΟΗ−
ΟΗ−
ΟΗ−
Figure 3.6. Schematic illustrating the mechanism for F- and OH- termination of (0001)Si 6H-SiC.
3.6. Conclusions
65
In conclusion, a completely dry process which removes carbon contamination from
(0001)Si 6H-SiC surfaces via UV/O3 oxidation and removes surface oxides via HF vapor
exposure has been demonstrated. Based on the levels of surface carbon and oxide
contaminants, this dry cleaning procedure has been found to be equivalent to or better than
other standard wet chemical processes. In contrast to silicon, the HF vapor exposure was
observed to result in a fluorine terminated SiC surface as opposed to a hydrogen terminated
SiC surface. This process was observed to leave some residual adventitious carbon. The
residual carbon was largely attributed to recontamination in the laboratory ambient.
However, this effect may be eliminated by either in situ vapor phase cleaning or well
controlled mini environments.
3.7. Acknowledgments:
The authors would like to thank Cree Research, Inc., for supplying the SiC wafers
used in these studies. This research was sponsored by the Office of Naval Research and
through the Department of Education via an Electronic Materials/GAANN fellowship.
3.8. References: 1. R.F. Davis, G. Kelner, M. Shur, J. Palmour, J.A. Edmond, Proc. of the IEEE, 79, 677 (1991). 66
2. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10, 1237 (1992). 3. W. Kern, RCA Review, 39, 278 (1978). 4. W. Kern, J. Electrochem. Soc., 137 (6) 1887 (1990). 5. T. Ohmi, J. Electrochem. Soc., 143, 1957 (1996). 6. G.R. Srinivasan and B.S. Meyerson, J. Electrochem. Soc., 134, 1518 (1987). 7. B.S. Meyerson, E. Ganin, D.A. Smith, and T.N. Nguyen, J. Electrochem. Soc., 133, 1232 (1986). 8. M.K. Sanganeria, M.C. Ozturk, G. Harris, K.E. Violette, I. Ban, C.A. Lee, and D.M Maher, J. Electrochem. Soc., 142, 3961 (1995). 9. F.K. LeGoues, MRS Bulletin, 21, 38 (1996). 10. L.M. Porter, R.F. Davis, J.S. Bow, M.J. Kim, R.W. Carpenter, R.C. Glass, J. Mater. Res., 10, 668 (1995). 11. H. Tsuchida, I. Kamata, and K. Izumi, Jpn. J. Appl. Phys., 34, 6003 (1995). 12. Y. Mizokawa, S. Nakanishi, O. Komoda, S. Miyase, H.S. Diang, C. Wang, N. Li, and C. Jiang, J. Appl. Phys., 67, 264 (1990). 13. U. Starke, Ch. Bram, P.R. Steiner, W. Hartner, L. Hammer, K. Heinz, K. Muller, Appl. Surf. Sci., 89, 175 (1995). 14. M.E. Lin, S. Strite, A. Agarwal, A. Salvador, G.L. Zhou, M. Teraguchi, A. Rockett, and H. Morkoc, Appl. Phys. Lett., 62, 702 (1993). 15. B.S. Meyerson, F.J. Himpsel, and K.J. Uram, Appl. Phys. Lett., 57, 1034 (1990). 16. M. Grundner and H. Jacob, Appl. Phys. A, 39, 73 (1986). 17. Y.J. Chabal, G.S. Higashi, K. Raghavachari, and V.A. Burrows, J. Vac. Sci. Technol. A, 7, 2104 (1989). 18. G.S. Higashi, R.S. Becker, Y.J. Chabal, A.J. Becker, Appl. Phys. Lett., 58, 1656 (1991). 19. G.S. Higashi, Y.J. Chabal, G.W. Trucks, and K. Raghavachari, Appl. Phys. Lett., 56, 656 (1990). 20. M. Houston and R. Maboudian, J. Appl. Phys., 68, 3801 (1995).
6721. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to J. Electrochem. Soc.
22. J.R. Vig, J. Vac. Sci. Techonl. A, 3, 1027 (1985). 23. M. Tabe, Appl. Phys. Lett., 45, 1073 (1984). 24. V.Y. Fominski, O.I. Naoumenko, V.N. Nevolin, A.P. Alekhin, A.M. Markeev, and L.A. Vyukov, Appl. Phys. Lett., 68, 2243 (1996). 25. T. Takahagi, I. Nagai, A. Ishitani, H. Kuroda, and Y. Nagasawa, J. Appl. Phys., 64, 3516 (1988). 26. J.A. McClintock, R.A. Wilson, and N.E. Byer, J. Vac. Sci. and Technol., 20, 241 (1982). 27. R.F. Kopf, A.P. Kinsella, and C.W. Ebert, J. Vac. Sci. Technol. B, 9, 132 (1991). 28. M. Suemitsu, T. Kaneko, and M. Miyamoto, Jap. J. Appl. Phys., 28, 2421 (1989). 29. S.J. Pearton, F. Ren, C.R. Abernathy, W.S. Hobson, and H.S. Luftman, Appl. Phys. Lett., 58, 1416 (1991). 30. S. Baunack and A. Zehe, Phys. Stat. Solid A, 115, 223 (1989). 31. S.S. Iyer, M. Arienzo, and E. de Fresart, Appl. Phys. Lett., 57, 893 (1990). 32. R. Iscoff, Semiconductor International, 7, 58 (1993). 33. W.A. Cady and M. Varadarajan, J. Electrochem. Soc., 143, 2054 (1996). 34. J. van der Weide, Ph.D Dissertation, NCSU. 35. T.J. Chuang, H.F. Winters, and J.W. Coburn, Surface Science, 2, 514 (1978). 36. J.W. Coburn, H.F. Winters, and T.J. Chuang, J. Appl. Phys., 48, 3532 (1977). 37. V.S. Smentkowski, J.T. Yates, Jr., X. Chen, and W.A. Goddard, III, Surface Science, 370, 209 (1997). 38. G. Hollinger and F.J. Himpsel, Appl. Phys. Lett., 44, 93 (1984). 39. G. Hollinger and F.J. Himpsel, J. Vac. Sci. Technol. A, 1, 640 (1983). 40. T.J. Chuang, J. Appl. Phys., 51, 2614 (1980). 41. C.D. Stinespring and A. Freedman, Appl. Phys. Lett., 48, 718 (1986). 42. J.H. Thomas III, and L.H. Hammer, J. Vac. Sci. Technol. B, 5, 1617 (1987).
68
43. A. Miyauchi, Y. Inoue, M. Ohue, N. Momma, and T. Suzuki, J. Electrochem. Soc., 137, 3257 (1990). 44. V.V. Afanas'ev, A. Stesmans, M. Bassler, G. Pensi, M.J. Schulz, and C.I. Harris, Appl. Phys. Lett., 68, 2141 (1996). 45. J.W. Palmour, Ph.D. Dissertation NCSU. 46. G.N. Raikar, J.C. Gregory, W.D. Partlow, H. Herzig, and W.J. Choyke, Surface and Interface Analysis, 23, 77 (1995).
69
4. Chemical Vapor Cleaning of (0001)Si, (000-1)C, (10-10) &
(11-20) 6H-SiC Surfaces
To be Submitted for Consideration for Publication
to the:
Journal of Applied Physics
by
Sean W. King, R. Scott Kern, *Mark C. Benjamin, John P. Barnak,
*Robert J. Nemanich, and Robert F. Davis,
Department of Materials Science and Engineering and
*Department of Physics
North Carolina State University
Raleigh, NC 27695
70
4.1. Abstract
A chemical vapor cleaning (CVC) procedure based on annealing in fluxes of SiH4
and C2H4 has been demonstrated for (0001)Si, (000-1)C, (11-20), and (10-10) 6H-SiC
surfaces. In comparison to SiC surfaces prepared by thermal desorption techniques, SiH4
CVC prepared surfaces were found to be of higher purity, free of both oxides and C-C
bonded carbon/graphite. For the (0001)Si orientation, the SiH4 CVC procedure was found to
produce (3x3) reconstructed surfaces which consisted of an incomplete bilayer of silicon on
top of the SiC surface. Reconstructed (√3x√3)R30° (0001)Si 6H-SiC surfaces could be
prepared by annealing the (3x3) SiH4 CVC surface in UHV at 1050°C. In contrast, no
reconstructions were observed for SiH4 CVC prepared (000-1)C, (11-20), and (10-10) 6H-
SiC surfaces. The SiH4 CVC procedure was found to be particularly effective in preventing
and removing graphite formation from (000-1)C surfaces. The stoichiometry of CVC
prepared (000-1)C and (11-20) and (10-10) surfaces was easily controlled via a second
exposure to C2H4.
71
4.2. Introduction
Preparation of clean, structurally well ordered surfaces is an important first step in all
semiconductor microelectronic fabrication processes [1-3]. Surface cleaning prior to epitaxy
is particularly important as improper removal of surface oxides and organic contaminants has
been shown in Si homoepitaxy to result in an increased density of stacking faults and
dislocations from < 104/cm2 to > 1010/cm2 [4-15]. In fact, studies on the heteroepitaxial
growth of SixGe1-x alloys on Si have shown that residual oxide and organic contaminants
act as the preferred sites for nucleation of misfit dislocations [15]. Epitaxial defects from
improper surface cleaning are important as they have been shown to cause a decrease in
device performance and yield [16-22]. Accordingly, the control of defects in epitaxial films
is clearly related to surface cleaning and preparation. These observations are of paramount
importance to SiC as a range of structural and electrical defects are currently prohibiting it
from becoming the semiconductor/substrate of choice for high power, high frequency, and
high temperature electronic devices and III-V nitride heteroepitaxy [23-25]. Thus based on
analogy to silicon technology, it is clear that improved surface cleaning procedures should
result in decreased epitaxial defect densities. In fact, Powell et al [26] and Burk and Rowland
[27] have previously shown surface pretreatments to be instrumental in the control of
polytype and reduction of interfacial Al in atmospheric CVD and VPE growth of SiC.
72
In a previous set of studies [27,28], we have shown that typical wet chemical
cleaning processes leave a monolayer coverage of oxygen and adventitious surface carbon on
(0001)Si 6H-SiC surfaces which must be removed in situ prior to epitaxial growth. Removal
of the oxide and adventitious carbon in situ has been achieved via a variety of techniques
including thermal desorption [30-46], ion bombardment/sputtering [47-50], ECR H2 plasma
cleaning [51], and annealing in a Si flux [55-62]. As others have shown [34,35,55], thermal
desorption of the monolayer of oxide from SiC occurs at ≈ 1000°C which is ≈ 200°C higher
than that required for silicon [63]. Additionally, thermal desorption of oxygen results in the
loss of silicon from the SiC surface and the formation of C-C bonding which may lead to
graphite bonded structures on the surface [33-36,38,39]. This is primarily due to the fact that
surface oxides on Si and SiC desorb as SiO instead of O2. This desorption depletes the SiC
surface of silicon leaving behind excess carbon which may form graphite bonded structures
[38,39]. In the case of plasma cleaning, an ECR H2 plasma clean has been previously shown
to be useful for removing C-C, C-F, and C-O bonded contaminants, but the technique is
inefficient or incomplete with regard to removing Si-O [51]. As we [52] and others [53,54]
have shown in separate studies, atomic H also selectively removes Si from the SiC surface
producing a carbon rich surface. Ion bombardment or sputtering inherently induces surface
damage and disorder which must be removed via high temperature annealing. This will also
lead to a loss of silicon and the possibility graphitic bonding at the surface [49].
An alternative to the above techniques is to anneal a SiC wafer in a flux of material
whose oxide or suboxide is more volatile than SiO2. The advantage of this technique is that
the oxide can be chemically removed at temperatures 100-200°C lower than simple thermal
desorption and without physically bombarding or damaging the surface with high energy
73
ions. In the case of silicon, fluxes of Ga [63], Ge [64], GeH4 [65], SiH4 [66,67], and Si2H6
[68] have been successfully used to chemically reduce and remove oxides on silicon surfaces.
In the case of SiC, Kaplan and Parril [55-57] have demonstrated that evaporated Ga or Si can
also be used to reduce and remove surface oxides from SiC surfaces at ≈ 850°C which is ≈
100-200°C lower than the temperature necessary to thermally desorb the surface oxide from
SiC surfaces [34,35]. The technique of Kaplan [55] has been recently implemented by Fissel
et al [62] in solid source MBE growth of SiC on (0001)Si 6H-SiC. . Unfortunately however,
the approach used by Kaplan [55] and Fissel [62] is most compatible with MBE growth
techniques. An alternative to this approach would be to use a gas source of silicon such as
SiH4 or Si2H6 which would allow this technique to be extended to lower vacuum with
techniques.
In this paper, we demonstrate that for preparation of silicon terminated (0001)Si 6H-
SiC surfaces, essentially similar results to evaporated silicon can be achieved using a gas
source of silicon such as SiH4. This chemical vapor cleaning (CVC) cleaning procedure
was found to be compatible with both GSMBE and LPCVD growth processes as well. We
additionally demonstrate that the combined sequential use of SiH4 and C2H4 can be used for
the preparation of carbon terminated (000-1)C 6H-SiC surfaces and non-polar (11-20) and
(10-10) 6H-SiC surfaces of varying stoichiometry. Direct comparison to thermal desorption,
shows the CVC technique to produce the purest SiC surfaces free of oxide and carbon-carbon
bonded contaminants.
74
4.3. Experimental Procedure
4.3.1. Integrated Surface Preparation and Analysis System.
All experiments described below were conducted using a unique ultra high vacuum
(UHV) configuration which integrates several completely independent UHV surface
preparation, thin film growth and surface analysis systems via a 36 ft. long transfer line
having a base pressure of 9x10-10 Torr (additional details of the transfer line, and many of
the associated systems are provided in Refs. 69-71). The experiments described in this paper
employed the SiC atomic layer epitaxy (ALE), Auger electron spectroscopy (AES), electron
energy loss spectroscopy (EELS), low energy electron diffraction (LEED), x-ray
photoelectron spectroscopy (XPS), and remote H2/SiH4 plasma CVD systems. A brief
description of these systems is provided below.
The SiC ALE system consisted of a UHV chamber with a base pressure of 3x10-10
Torr and was equipped with a residual gas analyzer (RGA) and a variety of gas dosers. The
RGA (a 0-200 amu quadrapole gas analyzer from Hiden Analytical Ltd.) was housed in a
separate differentially pumped cylindrical chamber (similar in design to that of Smentkowski
and Yates [72] ) which had a 0.5 cm diameter orifice at the head of the RGA for TPD
experiments and an approximately 50 cm2 "sunroof" for monitoring residual gases in the
75
system. The sample heating stage in the ALE system consisted of a wound tungsten heating
filament positioned close to the back of the sample and mounted on a boron nitride disk [69].
A W/6%Re-W/26%Re thermocouple was employed to measure the temperature of the
backside of the wafer. Surface temperatures and heating profiles to 1100°C were easily
achieved using a programmable microprocessor and 20 amp SCR power supply. Actual
surface/sample temperatures (i.e. those reported herein) were recorded using an infra-red
pyrometer with a spectral response of 0.8 to 1.1 µm and a emissivity setting of 0.5. The
estimated experimental accuracy for the latter temperatures was estimated to be ± 25°C.
Gas sources in the ALE system included SiH4 (99.995%), C2H4 (99.99%), and H2
(99.995%). Mass spectroscopic analysis of the as received silane using the RGA revealed
that the primary resolvable impurities were H2O, CO2, and Si2H6 in concentrations of 180
ppm, 31 ppm, and 170 ppm respectively. Impurities such as CO and O2 were expected but
difficult to resolve due to overlap with the silane cracking patterns. However, we estimate
that the level of these impurities were at least below 600 ppm or better. No further
purification was deemed necessary and the silane was used in this purity. Similar purity was
found for the ethylene and hydrogen. Sample exposure to SiH4 and/or C2H4 was obtained
using "molecular beam" dosers similar to the design of Bozack et al [73]. Collimation of
SiH4 or C2H4 into a molecular beam focused onto the sample was achieved with this doser
using a 13 mm diameter x 2 mm thick glass capillary array with a ten micrometer pore size
(Galileo Electro Optics Inc.). Thought, the doser to sample distance was fixed at 2", no
attempts were made to accurately measure the flux of SiH4 or C2H4 and hence all exposures
are quoted as Langmuirs (£ = 10-6 Torr sec).
76
XPS experiments were performed in a stainless steel UHV chamber (base pressure =
2x10-10 Torr) equipped with a dual anode (Mg/Al) x-ray source and a 100 mm
hemispherical electron energy analyzer (VG CLAM II). All XPS spectra reported herein
were obtained using Al Kα radiation (hν = 1486.6 eV) at 12 kV and 20 mA emission
current. XPS analysis typically required less than 1 hour during which time the pressure
never increased above 9x10-10 Torr. Calibration of the binding energy scale for all scans
was achieved by periodically taking scans of the Au 4f7/2 and Cu 2p3/2 peaks from
standards and correcting for the discrepancies in the measured and known values of these two
peaks (83.98 and 932.67 eV, respectively [74]). Curve fitting of most data was performed
using the software package GRAMS 386. A combination Gaussian-Lorentzian curve shape
with a linear background was found to best represent the data. The Auger electron
spectrometer and the low energy electron diffraction optics were mounted on a six way cross
off the transfer line and pumped through the transfer line. In the AES analysis, a 3 keV, 1
mA beam was used. Each Auger electron spectrum was collected in the undifferentiated
mode and numerically differentiated. In LEED an 80 eV, 1 mA beam was used.
The remote plasma CVD system consisted of a metal seal stainless steel vacuum
chamber pumped by a 330 l/s turbomolecular pump. The base pressure of this system was
4x10-9 Torr. The process gases flowed through a quartz tube mounted at the top of the
chamber, and the plasma is excited by rf (13.56 MHz) applied through a copper coil wrapped
around the quartz tube. The sample was located 40 cm below the center of the rf coil. An
inline Nanochem purifier and filter was used for point of use purification of hydrogen and
silane. Sample heating in the plasma system was conducted using a sample heater similar in
design to the one previously described in the ALE system. The plasma system was also
77
equipped with a differentially pumped 0-100 amu RGA which allowed direct analysis of the
purity of the process gases. RGA analysis of the hydrogen and silane (both 99.999% purity)
used in these experiments after in situ purification revealed that the impurity level of these
gases were below the baseline of the system (<1 ppm)
4.3.2. Substrate Preparation:
Various orientations of 6H-SiC substrates supplied by Cree Research Inc. were
examined in this research including: on axis and vicinal (4° off axis toward (11-20))
(0001)Si, on axis and vicinal (000-1)C, (11-20), and (10-10). The size of these substrates
ranged from ≈ 1.5 cm2 for (0001)Si and (000-1)C orientations to 5 mm2 for the (11-20) and
(10-10) samples. All substrates were nitrogen doped n-type with a carrier concentration, Nd-
Na, equal to ≈ 1018/cm3. The off axis (0001)Si and (000-1)C substrates were provided with
an ≈ 1 µm n-type epitaxial layer (Nd=5x1017/cm3) by Cree Research Inc. Cree additionally
oxidized 500-1000Å of the surface of all SiC wafers by dry oxidation. This oxide was
removed via a 10 min. dip in 10:1 HF.
After removal of the thermal oxide, the unpolished back side of each wafer was
coated with tungsten via rf sputtering to increase the heating efficiency of the SiC, as the
latter is partially transparent to the infrared radiation emitted from our tungsten filament
heaters. Originally platinum was used as the refractory metal but was later found to be 78
unsuitable. Typically, platinum was found to react with the back of the wafer forming Pt-
silicides which evaporated at the temperatures used in these experiments. Further Pt was
observed to diffuse through micropipes in the wafer to the polished side of the wafer also
forming platinum silicides which were detected by AES and XPS. Platinum contaminated
surfaces were observed to exhibit (2x2), (4x4), and (5x5) LEED patterns which are
uncharacteristic of pristine SiC surfaces [55].
After coating the backside of the SiC wafer with tungsten, the SiC wafers were
ultrasonically rinsed in trichloroethylene, acetone and methanol each for 5 min. and then
exposed to the vapor from a 10:1 buffered HF solution for 10 min. The wafers were then
mounted to a 1" diameter Mo disk using Ta wire. Each wafer/Mo assembly was then
fastened to a ring shaped Mo sample holder using Ta wire and inserted into the transfer line
load lock for further experimentation.
4.4. Results
4.4.1. (0001)Si 6H-SiC
Figure 4.1 shows a series of AES spectra acquired from a vicinal (0001)Si 6H-SiC
surface exposed to 200£ SiH4 at various different temperatures in the SiC ALE system
Similar results were obtained from on axis (0001)Si 6H-SiC surfaces. Comparison between
Figures 4.1(a) and 4.1(b) shows that the SiH4 exposure at a substrate temperature of 750°C
results in very little change in the amount of surface oxide on SiC. A 200£ exposure at
820°C (Fig. 4.1(c).) results in an incomplete reduction of the surface oxide. Removal of the
79
oxygen to below the detection limits of both AES and XPS was achieved at temperatures of
880°C and greater. This temperature is 100-200°C lower than the temperature needed to
simply thermally desorb the oxide in UHV. Using these conditions, sharp (1x1) LEED
patterns were obtained whereas prior to the CVC clean a (1x1) pattern with broad diffraction
spots were obtained (see Figure 4.2(a),(b)). The changes in the Si and C concentrations on
the surface were examined from the ratio of the peak-to-peak heights (pph) of the AES
signals or the ratio of the XPS peak heights. The SiC surfaces prepared in this manner were
slightly silicon rich as observed by the increase in the AES Si LVV/C KLL pph ratio from <
1 to ≈ 3 after the CVC treatment. A similar increase from < 1 to ˜ 1.1 was observed for the Si
2p/C 1s ratio in XPS.
30 150 270 390 510 630
dN(E
)/dE
(c)
(b)
Electron Energy (eV)
(a)
Si C
O
Figure 4.1. AES of (0001)Si 6H-SiC surfaces after (a) 200£ SiH4 at 750°C, (b) 200£ SiH4 at 820°C, and (c) 200£ SiH4 at 880°C.
80
If a larger a SiH4 exposure (> ˜ 500£) was used at Tsub > 900°C, a second surface
reconstruction, the (3x3), is observed with LEED (see Figure 4.2(c)). In this case, the Si
LVV/C KLL pph ratio was observed to increase from ≈ 3 to 5 suggesting that excess silicon
was deposited on the SiC surface. The excess Si was confirmed by XPS analysis of the Si 2p
core level from the (3x3) surface where two Si 2p peaks were observed at 99.5 and 101.3 eV.
The two different peaks are indicative of Si-Si and Si-C bonding respectively (see Figure
4.3). More detailed analysis of the intensity of these two peaks based on the attenuation of
the Si-C Si 2p peak indicated that the (3x3) reconstruction corresponded to a silicon coverage
of ≈ 1.5 monolayers (i.e. an incomplete bilayer). Higher SiH4 exposures (>2000£) in
temperature range of 900-1050°C were not typically observed to result in additional
reconstructions past the. However, annealing of the (3x3) surface at 1000°C in UHV did
lead to the loss of the excess Si resulting in the surface reverting back to the (1x1) LEED
pattern. Still further annealing, resulted in the observation in LEED of a third surface
reconstruction, the (√3x√3)R30° (see Figure 4.2(d)). Figure 4.4 displays XPS spectra of the
Si 2p core level from (0001)Si surfaces exhibiting the three different SiC reconstructions
observed in this study. As can be seen, the intensity of the Si-Si peak/shoulder decreases in
going from the (3x3) to the (1x1) to the (√3x√3)R30° surfaces. Further, the Si LVV/C KLL
pph ratio in AES was observed to decrease from ≈ 4-5 for the (3x3) surface to 3 and 2
respectively for the (1x1) and (√3x√3)R30° surfaces. Accounting for the 2:1 difference in
sensitivity to Si and C in AES [55], the (√3x√3)R30° surface appears to be the closest to bulk
stoichiometry. Similarly results were observed in XPS were the Si2p/C1s ratio was observed
to decrease from ≈ 1.3 for the (3x3) surface to 1.1 and 0.9 for the (1x1) and (√3x√3)R30°
surfaces. Other reconstructions past the (√3x√3)R30° such as a (6√3x6√3)R30° or (6x6)
[45,60] were not observed with extended annealing in the temperature range of 1000-1100°C.
These results are in excellent agreement with the results of Kaplan [55] who found that the
81
surface oxide could be removed at 850°C and (3x3) and (√3x√3)R30° reconstructions could
be obtained via annealing in evaporated silicon and/or UHV
For comparison purposes, (√3x√3)R30° 6H-SiC (0001)Si surfaces were also prepared
by thermal desorption at 1000°C. A comparison of the AES survey spectra acquired from
(√3x√3)R30° surfaces prepared by thermal desorption and SiH4 CVC is presented in Figure
4.5. It is clearly evident in Figure 4.5 that the Si LVV/C KLL pph ratio is < 1 for the thermal
desorption surface and > 1 for the CVC prepared surface. This suggests that the
(√3x√3)R30° surface prepared by thermal desorption is more carbon rich (or Si deficient)
than the (√3x√3)R30° surface prepared by CVC. This observation was additionally
supported by XPS analysis of the C 1s core level from these two surfaces (see Figure 4.6).
The spectra display a second C-C C 1s peak in the thermal desorption spectrum which is not
observed from the CVC surface. It should also be pointed out that in this, study removal of
oxygen below the detection limits of AES and XPS was rarely achieved using thermal
desorption.
82
(a) (b)
(c) (d)
83
(e) (f) (g)
Figure 4.2. LEED patterns from (a) HF dipped (0001)Si 6H-SiC, (b) (1x1) (0001)Si 6H-SiC, (c) (3x3) (0001)Si 6H-SiC, (d) (v3xv3)R30° (0001)Si 6H-SiC, (e) (1x1) (000-1)C 6H-SiC, (f) (11-20) 6H-SiC, (g) (10-10) 6H-SiC.
96 98 100 102 104 106
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Si-Si
Si-C
Figure 4.3. XPS spectra of the Si 2p core level from a (3x3) reconstructed (0001)Si 6H-SiC surface.
84
95 97 99 101 103 105 107
(1x1)(¦3x¦3)(3x3)
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Figure 4.4. XPS spectra of the Si 2p core level from (3x3), (1x1) and (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces.
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
Si
C
N O
Figure 4.5. AES survey spectra from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared by (a) SiH4 CVC, and (b) thermal desorption.
85
280 282 284 286 288
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 4.6. XPS spectra of the C 1s core level from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared by (a) SiH4 CVC, and (b) thermal desorption. 4.4.2. (000-1)C 6H-SiC
Oxide removal from (000-1)C 6H-SiC surfaces via annealing in SiH4 in the SiC ALE
system was observed to exhibit a similar temperature dependence to (0001)Si surfaces with
temperatures > 850°C being generally required for complete oxide removal. However in
contrast to (0001)Si, (3x3) reconstructions were never observed from CVC cleaned (000-1)C
surfaces despite the observation of Si LVV/C KLL pph ratios as large as 3-5 in AES and Si-
Si bonded Si 2p peaks in XPS (see Figure 4.2(e)). Further, we were not able to obtain
(√3x√3)R30° reconstructions from CVC cleaned (000-1)C surfaces despite extended
annealing (≈ 1 hr.) at 1100°C. These observations may be related to the trace amounts of
nitrogen consistently observed from the (000-1)C wafers after CVC cleaning or thermal
86
desorption (see Figure 4.7(a)). Similar persistent traces of nitrogen have also been observed
from (000-1)C 6H-SiC surfaces by Bermudez [61] and have been attributed to preventing
reconstruction of the SiC surface.
Comparisons between SiH4 CVC and thermal desorption cleaned (000-1)C surfaces
were also made. Firstly, traces of oxygen were observed from (000-1)C surfaces prepared by
annealing in UHV at 1050°C for 15 min. which was similar to (0001)Si surfaces. However
for the (000-1)C orientation, significant amounts of graphitic C-C bonding were observed to
form on the (000-1)C surface after annealing in UHV at 1050°C to thermally desorb the
surface oxide. The graphite formation was clearly observed in EELS spectra via a loss peak
at 6 eV (see Figure 4.8(a)) and a second C 1s peak at 284.5 eV in XPS (see Figure 4.9(a)).
This observation is to be contrasted to (0001)Si surfaces where a second C 1s peak was
observed in XPS from (√3x√3)R30° surfaces prepared via thermal desorption but a 6 eV loss
peak in EELS was not observed (see Figure 4.10). However for (000-1)C surfaces which had
undergone a 1050°C SiH4 CVC treatment, no graphite bonded C was observed in EELS (see
Figure 4.8(b)) and no C-C bonding C 1s was observed in XPS. In fact, the SiH4 CVC
treatment was also found useful for removing (i.e. converting to SiC) graphite formed on
(000-1)C surfaces via thermal desorption.
87
30 130 230 330 430 530 630 730
(a)
(b)
(c)
dN(E
)/dE
Electron Energy (eV)
SiC
Figure 4.7. AES survey spectra from (0001)C 6H-SiC surfaces prepared by (a) thermal desorption, (b) SiH4 CVC, and (c) SiH4/C2H4 CVC.
One interesting characteristic of the SiH4 CVC treated (000-1)C surfaces was an
inability to regraphitize this surface via high temperature annealing in UHV. Despite
annealing at 1100°C for extended periods of time (≈ 1-2 hr.), we were unsuccessful in
obtaining a 6 eV loss peak in EELS and/or a C-C bonding C 1s peak in XPS. Additionally, a
Si LVV/C KLL pph ratio of > 1 was maintained for all SiH4 CVC treated (000-1)C surfaces
annealed at 1100°C. The Si LVV/C KLL ratio of > 1 and similarities between the EELS
spectra of SiH4 CVC treated (0001)Si (see Figure 4.11) and (000-1)C (see Figure 4.8(b))
surfaces suggests that despite high temperature annealing, SiH4 CVC treated (000-1)C
surfaces are still silicon terminated. In order to reduce the Si LVV/C KLL ratio to < 1 (i.e.
carbon termination), it was found necessary to expose the SiH4 CVC treated (000-1)C
surfaces to C2H4.(see Figure 4.7(c)). However it was observed that C2H4 exposure at
88
temperatures > 950°C resulted in regraphitization of the (000-1)C surfaces (see Figure
4.8(c)). In order to reduce the Si LVV/C KLL ratio to < 1 and avoid graphite formation, it
was found necessary to decrease the sample temperature to ≈ 850°C (see Figure 4.12).
Unfortunately in this case, sharp (1x1) LEED patterns were not obtained.
-35 -30 -25 -20 -15 -10 -5 0 5
(a)
(b)
(c)
Cou
nts (
arb.
uni
ts)
Loss Energy (eV)E
last
ic P
eak
Figure 4.8. EELS spectra from (000-1)C 6H-SiC surfaces after (a) annealing in UHV at 1050°C, (b) annealing in 2000£ SiH4 at 1050°C followed by (c) annealing in 2000£ C2H4 at 950°C.
89
280 282 284 286 288
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 4.9. XPS spectra of the C 1s core level from (000-1)C 6H-SiC surfaces after (a) 2000£ SiH4 at 1050°C, (b) thermal desorption in UHV at 1050°C.
-40 -35 -30 -25 -20 -15 -10 -5 0
Cou
nts (
arb.
uni
ts)
Energy Loss (eV)
Ela
stic
Pea
k
Figure 4.10. EELS spectrum from (v3xv3)R30° reconstructed (0001)Si 6H-SiC surface prepared by annealing in UHV at 1050°C.
90
-40 -35 -30 -25 -20 -15 -10 -5 0
Cou
nts (
arb.
uni
ts)
Energy Loss (eV)
(a)
(b)
(c)
Ela
stic
Pea
k
Figure 4.11. EELS spectra of (3x3), (1x1), and (v3xv3)R30° reconstructed (0001)Si 6H-SiC surfaces prepared via SiH4 CVC.
-35 -30 -25 -20 -15 -10 -5 0 5
(a)
(b)
(c)
Cou
nts (
arb.
uni
ts)
Electron Energy Loss (eV)
Ela
stic
Pea
k
Figure 4.12. EELS spectra from (000-1)C 6H-SiC surfaces after (a) 2000£ SiH4 at 1000°C, (b) 400£ C2H4 at 850°C, and (c) 800£ C2H4 at 850°C.
91
4.4.3. (10-10) & (11-20) 6H-SiC
Oxide removal from (10-10) and (11-20) 6H-SiC surfaces via a SiH4 CVC clean was
observed to exhibit a similar temperature dependence to the (0001)Si and (000-1)C surfaces.
The LEED patterns displayed by these surfaces were observed to intensify and sharpen with
the SiH4 clean, but no reconstructions were observed to occur as a result of the clean or
deposition of excess silicon (see Figure 2(f),(g)). As with the (0001)Si and (000-1)C, the
SiH4 clean resulted in Si LVV/C KLL ratios > 1 which could not be reduced below one by
annealing at 1100°C. However, the Si LVV/C KLL ratio could be reduced to < 1 by
annealing in C2H4 at 900-1000°C. In this case, a 6 eV loss peak characteristic of graphite
formation was not observed from the (10-10) and (11-20) 6H-SiC surfaces (see Figure 13).
However, a C-C bonding C 1s bonding peak could be seen in XPS for large C2H4 exposures.
Similar to (0001)Si and (000-1)C, thermal desorption of the oxide from (11-20) and (10-10)
surfaces at 1050°C did result in incomplete thermal desorption and the observation of a C-C
bonded C 1s peak in XPS.
92
-35 -30 -25 -20 -15 -10 -5 0 5
(b)
(a)
Cou
nts (
arb.
uni
ts)
Electron Energy Loss (eV)
Ela
stic
Pea
k
Figure 4.13. EELS spectra of (10-10) 6H-SiC (a) after annealing in SiH4 at 1000°C, and (b) then annealing in C2H4 at 850°C.
4.4.4. Low vacuum CVC/LPCVD clean
In order to simulate conditions in lower vacuum processes (i.e. LPCVD and
OMVPE), (0001)Si 6H-SiC wafers were exposed to various fluxes of a 1% SiH4/H2 mixture
in the plasma CVD system. In this case, it was found that oxygen free, (3x3) reconstructed
(0001)Si 6H-SiC surfaces could be attained via annealing the SiC wafer in 70 sccm H2 and
0.01 sccm SiH4 at 900°C for 1 min. No other differences were observed between (0001)Si
6H-SiC surfaces prepared in low fluxes of SiH4 in the GSMBE and higher the fluxes of
H2/SiH4 in the plasma CVD.
93
4.5. Discussion
4.5.1. (0001)Si 6H-SiC
The results presented here demonstrate that the SiH4 CVC clean for (0001)Si 6H-SiC
yields surfaces similar to those prepared by the Si evaporation technique of Kaplan [55]. Our
Si 2p XPS spectra indicate an ≈ 1.5 monolayer coverage of Si (i.e. incomplete bilayer) for the
(3x3) reconstructed surface generated via the SiH4 CVC clean. This result is consistent with
the previously reported AES results and incomplete bilayer model of Kaplan [55] for the
(3x3) reconstruction. The (3x3) reconstruction has also been recently observed by Kulakov
et al [59] and Li and Tsong [60] using STM. Based on their observation of only one maxima
in the (3x3) unit cell, Kulakov et al [59] proposed a slight modification to the model of
Kaplan for the (3x3) reconstruction. In the model of Kulakov et al [59], the (3x3)
reconstruction still consists of an incomplete Si bilayer but with 1 adatom, 3 rest atoms, and 7
atoms directly over Si atoms in SiC as opposed to the 2 adatoms, 6 rest atoms and 8 second
layer Si atoms in the model of Kaplan [55]. In contrast, Li and Tsong [60] have proposed
that the (3x3) reconstruction consists of a 4/9 monolayer coverage of Si-C tetrahedra
organized in a (3x3) pattern. Our observation of a Si-Si bonding Si 2p peak in XPS with an ≈
1.5 monolayer coverage is clearly more consistent with the (3x3) models proposed by Kaplan
and Kulakov et al [59]. In a separate x-ray photoelectron diffraction study [75], we also
provide data which additionally supports the (3x3) model of both Kaplan and Kulakov et al
[59]. Finally, we note that similar Si-Si bonding Si 2p peaks have also been observed in XPS
94
from (3x2) [76] and c(4x2) [77] reconstructed (100) 3C-SiC surfaces prepared via annealing
in an evaporated Si flux.
For the (√3x√3)R30° reconstruction, it is interesting to note the differences in surface
stoichiometry observed between the two (√3x√3)R30° reconstructed surfaces prepared by
thermal desorption and SiH4 CVC. As previously noted, the Si LVV/C KLL pph ratio in
AES was < 1 for the (√3x√3)R30° surface prepared by thermal desorption whereas the SiH4
CVC (√3x√3)R30° surface was observed to have a Si LVV/C KLL ratio > 1. Additionally, a
C-C bonding C 1s peak was observed in XPS for the thermal desorption (√3x√3)R30°
surface. The observation of C-C bonding for (√3x√3)R30° (0001)Si surfaces prepared by
thermal desorption has been observed in several other studies [33,36,38-41]. For
(√3x√3)R30° (0001)Si surfaces prepared via annealing in an evaporated Si flux, Kaplan has
previously reported the observation of a 6 eV loss peak indicative of graphite in EELS.
However, in our case for (√3x√3)R30° surfaces prepared by the SiH4 CVC process, we did
not detect either a clear 6 eV loss peak in EELS or a C-C bonding C 1s peak in XPS
The above noted differences between the (√3x√3)R30° reconstructed (0001)Si
surfaces is important as various adatom models for this reconstruction have been proposed
based on recent STM investigations and theoretical calculations. Recent STM images by
Owman and Martensson [44,45] and Li and Tsong [60] have shown that the (√3x√3)R30°
(0001)Si surfaces prepared by thermal desorption are similar in nature to the group III
adatom (√3x√3)R30° Si (111) reconstructed surfaces [78,79]. These STM investigations
[44,45,60] have shown the existence of a 1/3 monolayer coverage of adatoms in threefold
symmetric sites. Unfortunately, neither investigation was able to resolve the chemical nature
95
of the adatom (i.e. silicon or carbon) or the adatom site (i.e. T4 or H3). However, the first-
principles total-energy calculations of Northrup and Neugebauer [80] indicate that Si
adatoms are preferred over C adatoms and that T4 sites are preferred over H3 sites. In
contrast, the quantum mechanical cluster calculations of Badziag [81,82] indicate that the
structure of the (√3x√3)R30° reconstruction on (0001) SiC surfaces should consist of
hydrogenated C3 triangles (resembling cyclopropane) saturating Si dangling bonds at the
surface. Based on the differences in surface stoichiometry, it is tempting to conclude that the
(√3x√3)R30° surface prepared by thermal desorption in this study and others [33,36,38-
41,43-45] is most appropriately described by the model of Badziag [81,82]. Similarly, for the
(√3x√3)R30° surface prepared by a SiH4 CVC, the model of Northrup and Neugebauer [80]
perhaps best describes the structure of this reconstructed surface. In the case of the (1x1)
reconstruction observed between the transformation from (3x3) to (√3x√3)R30°
reconstructions, the authors feel that this surface is disordered and is likely composed of a
mixture of (√3x√3)R30° and (3x3) reconstructions.
Finally additional proof of the effectiveness of the SiH4 CVC cleaning procedure has
been recently demonstrated by Kern [58] using secondary ion mass spectroscopy (SIMS)
analysis. In this case, SIMS analysis of the interface between a 3C-SiC epitaxial film grown
by GSMBE on a 6H-SiC wafer which had been cleaned by this procedure revealed absolutely
zero oxygen at the interface [58]. For comparison purposes the detection limit of AES and
XPS for O is typically 0.1 at%. SIMS on the other hand has a sensitivity limit in the ppm
range.
4.5.2. (000-1)C 6H-SiC
96
As noted above, we were unable to observe any surface reconstructions from the
(000-1)C surface for wafers undergoing either thermal desorption or SiH4 CVC treatments.
This is in contrast to the results of Nakanishi et al [33] where (3x3) reconstructions were
observed in LEED from the (000-1)C surface after thermal desorption at temperatures of
900-1100°C. Using STM, Li and Tsong [60] have additionally observed (3x3) and
(√3x√3)R30° reconstructions from (000-1)C surfaces prepared via annealing in an
evaporated Si flux. As previously noted by Bermudez [61], our inability to observe any
reconstructions from the (000-1)C surface with LEED may be related to the presence of
nitrogen on these surfaces. The presence of the nitrogen may be related to segregation of
nitrogen dopants to surface due to volatilization of silicon from the SiC surface leading to the
observed graphite formation. However, the authors have observed larger concentrations of
nitrogen on (0001)Si surfaces which have displayed (3x3) and (√3x√3)R30° reconstructions.
In this case, the source of nitrogen was attributed to reaction of the (0001)Si surface with
residual ammonia (NH3) in the system. Alternatively, the inability to observe surface
reconstructions from the (000-1)C surface with LEED may be related to an increased surface
roughness for the (000-1)C surface relative to the (0001)Si surface. This could occur due to
the lack of an optimized surface polishing procedure for the (000-1)C face. In the case of
Nakanishi et al [33], the (000-1)C surfaces used were the natural faces of Acheson crystals
which are speculated to be naturally atomically smooth. As STM samples a smaller region of
the surface compared to LEED, this may also explain the ability of Li and Tsong using STM
97
[60] to observe (3x3) and (√3x√3)R30° reconstructions from (000-1)C surfaces which have
undergone the same polishing/surface treatment as in our study.
The observation of graphite in EELS from (000-1)C surfaces prepared via thermal
desorption at 1050°C and not from (0001)Si surfaces similarly prepared is consistent with the
observations of Muehlhoff et al [32] where carbon segregation for the (000-1)C surface was
observed to occur ≈ 300°C lower than that for the (000-1)Si face. However, the inability to
regraphitize (000-1)C surfaces via annealing after a SiH4 CVC treatment was particularly
surprising, given the relative ease with which this occurred when SiH4 CVC treated (000-1)C
surfaces were exposed to C2H4. These results clearly indicate that the addition of a
monolayer of Si somehow stabilizes the (000-1)C surface against graphitization. Regardless
though, this surface termination is highly unstable in the presence of excess or free carbon
and immediately forms graphite. These results indicate that for (000-1)C surfaces a silicon
rich surface should be maintained during high temperature cleaning and growth in order to
avoid graphite formation. The ability to produce such a surface is an advantage for SiH4
CVC over thermal desorption treatments. Finally, the authors note that C2H4 has been
previously used for the preparation of carbon terminated c(2x2) reconstructed (001) 3C-SiC
surfaces [83,84]. In our, case it was found necessary to use low temperature C2H4 exposures
(≈ 850°C) in order to suppress graphite formation. However, surfaces prepared in such a
manner exhibited poor (1x1) LEED patterns indicating a disordered surface.
98
4.5.3. (11-20) and (10-10) 6H-SiC
The (11-20) and (10-10) orientations of 6H-SiC are non-polar surfaces with an equal
number of carbon and silicon atoms at the outermost surface. Accordingly, one may expect
these surfaces to exhibit properties intermediate to those of the (0001)Si and (000-1)C
orientations. In our limited studies here, though, we have observed the (11-20) and (10-10)
orientations to behave almost exactly like that of the (0001)Si face of 6H-SiC. However as
these are the first reported examinations of these orientations, the results reported here only
emphasize the need for a more detailed examination of the these surfaces. In particularly, a
more detailed examination of the electronic structure of the (11-20) and (10-10) surfaces
should be most beneficial as it should lead to a better understanding of the polar (0001)Si and
(000-1)C orientations.
99
4.6 Conclusions
A chemical vapor cleaning (CVC) procedure based on annealing in fluxes of SiH4
and C2H4 has been demonstrated for (0001)Si, (000-1)C, (11-20), and (10-10) 6H-SiC
surfaces. In comparison to SiC surfaces prepared by thermal desorption techniques, SiH4
CVC prepared surfaces were found to be of higher purity, free of both oxides and C-C
bonded carbon/graphite. For the (0001)Si orientation, the SiH4 CVC procedure was found to
produce (3x3) reconstructed surfaces which consisted of an incomplete bilayer of silicon on
top of the SiC surface. Reconstructed (√3x√3)R30° (0001)Si 6H-SiC surfaces could be
prepared by annealing the (3x3) SiH4 CVC surface in UHV at 1050°C. In contrast, no
reconstructions were observed for SiH4 CVC prepared (000-1)C, (11-20), and (10-10) 6H-
SiC surfaces. The SiH4 CVC procedure was found to be particularly effective in preventing
and removing graphite formation from (000-1)C surfaces. The stoichiometry of both (000-
1)C and (11-20) and (10-10) surfaces were easily controlled via exposure to C2H4.
4.7. Acknowledgments
The authors would like to thank Cree Research, Inc. for supplying the wafers used in
these experiments. This research was supported by the Office of Naval Research and by the
Department of Education through an Electronic Materials/GAANN fellowship.
100
4.8 References General Surface Cleaning References 1. W. Kern, RCA Review, 39, 278 (1978). 2. W. Kern, J. Electrochem. Soc., 137, 1887 (1990). 3. T. Ohmi, J. Electrochem. Soc., 143, 1957 (1996). Structural Defects in Epitaxy from cleaning 4. B.A. Joyce, J.H. Neave, and B.E. Watts, Surface Science, 15, 1 (1969). 5. J.H. McFee, R.G. Swartz, V.D. Archer, S.N. Finegan, and L.C. Feldman, J. Electrochem. Soc., 130, 214 (1983). 6. S. Nagao, K. Higashitani, Y. Akasaka, and H. Nakata, J. Appl. Phys., 57, 4589 (1985). 7. B.S. Meyerson, E. Ganin, D.A. Smith, and T.N. Nguyen, J. Electrochem. Soc., 133, 1232 (1986). 8. A.J. Pidduck, D.J. Robbins, A.G. Cullis, D.B. Gasson, and J.L. Glasper, J. Electrochem. Soc., 136, 3083 (1989). 9. A. Miyauchi, Y. Inoue, M. Ohue, N. Momma, T. Suzuki, and M. Akiyama, J. Electrochem. Soc., 137, 3257 (1990). 10. A. Miyauchi, Y. Inoue, T. Suzuki, and M. Akiyama, Appl. Phys. Lett., 57, 676 (1990). 11. M. Racanelli, D.W. Greve, M.K. Hatalis, and L.J. van Yzendoorn, J. Electrochem. Soc., 138, 3783 (1991). 12. Eaglesham, G.S. Higashi, and M. Cerullo, Appl. Phys. Lett., 59, 685 (1991). 13. C. Galewski, J. Lou, and W.G. Goldham, IEEE Trans. Semicond. Manfact., 3, 931 (1990). 14. M.K. Sanganeria, M.C. Ozturk, G. Harris, K.E. Violette, I. Ban, C.A. Lee, and D.M Maher, J. Electrochem. Soc., 142, 3961 (1995). 15. F.K. LeGoues, MRS Bulletin, 21, 38, (1996).
101
Surface Cleaning and Device Performance/Yield 16. G.R. Srinivasan and B.S. Meyerson, J. Electrochem. Soc., 134, 1518 (1987). 17. G.R. Srinivasan, J. Cryst. Growth, 70, 201 (1984). Electrical Defects related to cleaning 18. J.V. Dalton and J. Drobek, J. Electrochem. Soc., 115, 865 (1968). 19. J. Ruzyllo, A.M. Hoff, D.C. Frystak, and S.D. Hossain, J. Electrochem. Soc., 136, 1474 (1989). 20. S.R. Kasi, M. Liehr, P.A. Thiry, H. Dallaporta, and M. Offenberg, Appl. Phys. Lett., 59, 108 (1991). 21. L.J. Huang and W.M. Lau, Appl. Phys. Lett., 60, 1108 (1992). 22. T. Ohmi, T. Imaoka, T. Kezuka, J. Takano, and M. Kogure, J. Electrochem. Soc., 140, 811 (1993). Applications of SiC and III-N's 23. R.F. Davis, Advances in Ceramics, 23, 477 (1987). 24. R.F. Davis, G. Kelner, M. Shur, J. Palmour, J.A. Edmond, Proc. of the IEEE, 79, 677 (1991). 25. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10, 1237 (1992). Control of Polytypes and Surface Pretreatments 26. J.A. Powell, J.B. Petit, J.H. Edgar, I.G. Jenkins, L.G. Matus, J.W. Yang, P. Pirouz, W.J. Choyke, L. Clemen, and M. Yoganathan, Appl. Phys. Lett., 59, 333 (1991). 27. A.A. Burk, Jr., and L.B. Rowland, Appl. Phys. Lett., 68, 382 (1996). Ex Situ Cleaning of SiC 28. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to J. Electrochem. Soc. 29. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to J. Electrochem. Soc. Thermal Desorption 30. A.J. van Bommel, J.E. Crombeen, and A. van Tooren, Surface Science, 48, 463 (1975). 31. F. Bozso, L. Muehlhoff, M. Trenary, W.J. Choyke, and J.T. Yates, Jr., J. Vac. Sci. and Technol. A, 2, 1271 (1984). 32. L. Muehlhoff, M.J. Bozack, W.J. Choyke, and J.T. Yates, Jr., J. Appl. Phys., 60 2558 (1986). 102
33. S. Nakanishi, H. Tokutaka, K. Nishimori, S. Kishida, and N. Ishihara, Applied Surface Science, 41/42, 44 (1989). 34. M. Dayan, J. Vac. Sci. Technol. A, 4, 38 (1986). 35. Y. Mizokawa, S. Nakanishi, O. Komoda, S. Miyase, H.S. Diang, C. Wang, N. Li, and C. Jiang, J. Appl. Phys., 67, 264 (1990). 36. U. Starke, Ch. Bram, P.R. Steiner, W. Hartner, L. Hammer, K. Heinz, K. Muller, Appl. Surf. Sci., 89, 175 (1995). 37. L.M. Porter, R.F. Davis, J.S. Bow, M.J. Kim, R.W. Carpenter, R.C. Glass, J. Mater. Res., 10, 668 (1995). 38. L.I. Johansson, F. Owman, and P. Martensson, Surface Science, 260, L483 (1996). 39. L.I. Johansson, F. Owman, and P. Martensson, Phys. Rev. B, 52, 13793 (1996). 40. T. Tsukamoto, M. Hirai, M. Kusaka, M. Iwami, T. Ozawa, T. Nagamura, and T. Nakata, Surface Science, 371, 316 (1997). 41. C.S. Chang, I.S.T. Tsong, Y.C. Wang, and R.F. Davis, Surface Science, 256,354 (1991). 42. M.A. Kulakov, P. Heuell, V.F. Tsvetkov, and B. Bullemer, Surface Science, 315, 248 (1994). 43. Y. Marumoto, T. Tsukamoto, M. Hirai, M. Kusaka, M. Iwami, T. Ozawa, T. Nagamura, and T. Nakata, Jpn. J. Appl. Phys., 34, 3351 (1995). 44. F. Owman and P. Martensson, Surface Science, 330, L639 (1995). 45. F. Owman, P. Martensson, J. Vac. Sci. & Technol., 14, 933 (1996). 46. L.B. Rowland, R.S. Kern, S. Tanaka, and R.F. Davis, J. Mater. Res., 8, 2753 (1993). Sputtering/Ion Bombardment 47. R. Kaplan, J. Appl. Phys., 56, 1636 (1984). 48. M Balooch and D.R. Olander, Surface Science, 261, 321 (1992). 49. S.V. Didziulis, J.R. Lince, P.D. Fleishauer, and J.A. Yarmoff, Inorg. Chem., 30, 672 (1991). 50. J.M. Powers and G.A. Somorjai, Surface Science, 244, 39 (1991). ECR H2
103
51. M.E. Lin, S. Strite, A. Agarwal, A. Salvador, G.L. Zhou, M. Teraguchi, A. Rockett, and H. Morkoc, Appl. Phys. Lett., 62, 702 (1993). 52. S.W. King, M.C. Benjamin, J.P. Barnak, R.J. Nemanich, and R.F. Davis, to submitted to Journal of Applied Physics. 53. M.D. Allendorf and D.A. Outka, Surface Science, 258, 177 (1991). 54. Y. Kim and D.R. Olander, Surface Science, 313, 399 (1994). Evaporated Si Flux Clean 55. R. Kaplan and T.M. Parrill, Surface Science Letters, 165, L45 (1986). 56. R. Kaplan, Surface Science, 215, 111 (1989). 57. T.M. Parrill and Y.W. Chung, Surface Science, 243, 96 (1991). 58. R.S. Kern, Ph.D. Dissertation, NCSU (1996). 59. M.A. Kulakov, G. Henn, B. Bullemer, Surface Science, 346, 49 (1996). 60. L. Li, and I.S.T. Tsong, Surface Science, 351, 141 (1996). 61. V.M. Bermudez, Applied Surface Science, 84, 45 (1995). 62. A.Fissel, B. Schroter, and W. Richter, Appl. Phys. Lett., 66, 3182 (1995). Chemical Reduction of oxides on Si 63. S. Wright and H. Kroemer, Appl. Phys. Lett, 36, 210 (1980). 64. J.F. Morar, B.S. Meyerson, U.O. Karlsson, F.J. Himpsel, F.R. McFeely, D. Rieger, A. Taleb-Ibrahimi, and J.A. Yarmoff, Appl. Phys. Lett., 50, 463 (1987). 65. M. Racanelli, D.W. Greve, M.K. Hatalis, and L.J. van Yzendoorn, J. Electrochem. Soc., 138, 3783 (1991). 66. H. Hirayama, R. Tatsumi, A. Ogura, and N. Aizaki, Appl. Phys. Lett., 51, 2213 (1987). 67. H. Hirayama and T. Tatsumi, J. Appl. Phys., 66, 629 (1989). 68. K. Saito, T. Amazawa, and Y. Arita, J. Electrochem. Soc., 140, 513 (1993). Experimental References 69. Jacob van der Weide, Ph.D. Dissertation (1994). 70. S.W. King, R.S. Busby, R.J. Nemanich, and R.F. Davis, submitted to Surface Science. 104
71. S.W. King, M.C. Benjamin, J.P. Barnak, R.J. Nemanich, and R.F. Davis, submitted to Journal of Applied Physics. 72. V.S. Smentkowski and J.T. Yates Jr., J. Vac. Sci. Technol. A, 7, 3325 (1989). 73. M.J. Bozack, L. Muehlhoff, J.N Russel Jr., W.J. Choyke, and J.T. Yates, Jr., J. Vac. Sci. Technol. A, 5, 1 (1987) 74. XPS Handbook, Perkin Elmer. XPD of (3x3) 6H-SiC (0001)Si 75. S.W. King, R.S. Busby, R.J. Nemanich, and R.F. Davis, submitted to Surface Science. (3x2) and c(4x2) 3C-SiC (100) 76. V.M Bermudez and J.P. Long, Appl. Phys. Lett., 66, 475 (1995). 77. M.L. Shek, Surface Science, 349, 317 (1996). Group III adatom (√3x√3)R30° Si (111) 78. X. Chen, T. Abukawa, S. Kono, Surface Science, 356, 28 (1996). 79. Y. Taguchi, M. Date, N. Takagi, T. Aruga, M. Nishijima, Applied Surface Science, 82/83, 434 (1994). (√3x√3)R30° Theory 80. J.E. Northrup and J. Neugebauer, Phys. Rev. B., 52, R17001 (1995). 81. P. Badziag, Surface Science, 337, 1 (1995). 82. P. Badziag, Surface Science, 352, 396 (1996). C2H4 - (001) 3C-SiC 83. J.M. Powers, A. Wander, P.J. Rous, M.A. Van Hove, and G.A. Somorjai, Phys. Rev. B, 44, 11159 (1991). 84. V.M. Bermudez and R. Kaplan, Phys. Rev. B, 44, 11149 (1991).
105
5. X-ray Photoelectron Diffraction from (3x3) and (√3x√3)R30° (0001)Si
6H-SiC Surfaces.
To be Submitted for Consideration for Publication
to
Surface Science
by
Sean W. King, *Richard S. Busby, *Robert J. Nemanich, and Robert F. Davis
Department of Materials Science and Engineering
*Department of Physics
North Carolina State University
Raleigh, NC 27695
106
5.1. Abstract
High resolution (±1°) x-ray photoelectron diffraction (XPD) patterns were obtained
along high symmetry azimuths of (3x3) and (√3x√3)R30° reconstructed (0001)Si 6H-SiC
surfaces. The data obtained were compared to previously reported XPD patterns from (7x7)
Si (111) as well as models proposed for the (3x3) and (√3x√3)R30° 6H-SiC reconstructions.
Forward scattering features similar to those observed from (7x7) Si (111) were also observed
from (√3x√3)R30° 6H-SiC (0001)Si surfaces. However, additional features not observed in
(7x7) Si (111) were observed in the (√3x√3)R30° 6H-SiC XPD patterns which were
attributed to the substitution of carbon atoms for silicon atoms on the diamond FCC lattice.
Unlike (1x1) and (7x7) Si (111) surfaces, differences were observed between the XPD
patterns of (3x3) and (√3x√3)R30° SiC (0001)Si surfaces. The most significant difference
observed between the (3x3) and (√3x√3)R30° reconstructions was the equivalence of the [10-
10] and [01-10] azimuths in the (3x3) structure. The differences between the (3x3) and
(√3x√3)R30° XPD patterns were attributed to the presence of an incomplete bilayer of Si the
(3x3) surface. The (3x3) SiC XPD patterns observed in this study are consistent with a
faulted Si bilayer stacking sequence recently proposed based on STM observations.
107
5.2. Introduction
X-ray photoelectron diffraction (XPD) is an exciting new technique for probing the
local atomic structure of metal and semiconductor surfaces with atomic specificity [1-3].
XPD experiments essentially consist of performing angle dependent x-ray photoelectron
spectroscopy (XPS) measurements. Anisotropies in the angular dependence of the intensity
of emitted photoelectrons in XPS are created by interference between other emitted
photoelectron waves and/or scattering by nearest neighbor atoms. For the high kinetic
energies (≈ 1 keV) typically employed, XPD spectra are dominated by forward scattering (or
focusing) of the emitted photoelectron by the potential of the atomic nucleus of nearest
neighbor atoms. This effect which can be viewed as a zeroth order approximation to XPD
creates intensity enhancements along crystallographic and surface-adsorbate bond directions
(see Figure 5.1). Accordingly, this technique has been successfully employed in the
determination of surface adsorption sites for various atoms and molecules on metals and
semiconductors as well as for studying a number of different epitaxial growth systems [1-6].
Most recently, further developments and enhancements of this technique have actually
resulted in the demonstration of holographic images of various metal [7-10] and
semiconductor surfaces [11]. In this paper, we apply XPD to study the atomic structure of
(3x3) and (√3x√3)R30° reconstructed (0001)Si 6H-SiC surfaces.
108
Emitter Scatterer
Intensit
y
Lost
Intensity Gained
Figure 5.1. Schematic illustrating forward focusing/scattering effects in x-ray photoelectron diffraction experiments.
SiC is a wide band gap compound semiconductor (Eg (6H-SiC) = 3.0 eV) which is of
considerable importance to the development of high temperature, high frequency, and high
power electronic devices [12]. The ability to develop SiC into the material of choice for
these applications, however, has been currently limited by the inability to control the types
and densities of a variety of line, planar, and macroscopic defects in SiC wafers and films
[12]. By analogy to a more thoroughly investigated material such as silicon [13-16], it is
conceivable that many of these defects originate and/or nucleate at defects on the SiC
109
surface. Therefore in order to understand how these defects originate, a detailed
understanding of the atomic structure of the SiC surface is needed. For the (0001) surface of
6H-SiC, this has in part been provided by many recent scanning tunneling microscopy (STM)
studies [17-24] which have identified a variety of different surface reconstructions ranging
from (3x3), (√3x√3)R30°, (9x9), (6x6), and (6√3x6√3)R30°.
By analogy to the group III adatom (√3x√3)R30° Si (111) reconstructed surfaces [25-
28], many have proposed that (√3x√3)R30° 6H-SiC (0001) surface reconstructions are due to
bulk terminated (0001) 6H-SiC surfaces with a 1/3 ML (monolayer) coverage of silicon or
carbon adatoms in the T4 position (see Figure 5.2) [29,30]. Recent STM investigations by
Owman and Martensson [19] and Li and Tsong [21] have been able to confirm the three fold
symmetric unit cell, but unfortunately were unable to resolve the chemical identity of the
adatom or determine the exact position of the adatom (i.e. T4 or H3). However, Owman and
Martensson [19] were able to determine that the reconstruction was not composed of a
mixture of Si and C adatoms or a mixture of T4 and H3 sites (i.e. single adatom on a single
site). These findings by Owman and Martensson [19] are complementary to the theoretical
results of Northrup and Neugebauer [31]. Their recent supercell calculations using the
density functional method have shown that for (v3xv3)R30° (111) 3C-SiC surfaces, Si
adatoms are preferred over C adatoms and that the T4 site is favored over the H3 site by both
Si and C adatoms. These results are consistent with previous calculations by Northrup which
showed the T4 site to be preferred in Si (111) (√3x√3)R30°:Si adatom geometries [32
In contrast, semi-empirical, self consistent quantum mechanical cluster calculations
by Badziag [33,34] show that for the (0001)Si (√3x√3)R30° reconstructed surface a
hydrogenated triangle of C atoms (i.e. similar to cyclopropane) centered on the T4 position is
110
energetically more favorable than C or Si adatoms hydrogenated or unhydrogenated. This
hydrogenated C3 model is similar in nature to the milk stool model deduced for (√3x√3)R30°
Si (111):Sb surfaces where the Sb atoms form trimers centered on the T4 site [35,36]. The
validity of this model is somewhat questionable since it would require hydrogen to not
desorb from the SiC surface until temperatures of > 1150°C at which point the (√3x√3)R30°
reconstruction disappears. In defense of his model, Badziag points out that for diamond,
hydrogen desorbs at temperatures of 1000 and 1150°C for the (111) [37] and (100) [38]
surfaces which is consistent with the observed stability of this reconstruction. However,
Allendorf and Outka [39] observed two hydrogen desorption peaks from polycrystalline SiC
surfaces at ≈ 700 and 850°C which is well below the temperature at which this reconstruction
is observed to occur.
111
View Along [11-20] or [1-10][0001], [111]
[10-10], [1-21]
T4 H3b.)
H3
T4
<10-10>a.)
<11-20>
C atomSi atom
Figure 5.2. Schematics illustrating various adatom adsorption sites for (√3x√3)R30° reconstructions on (111)/(0001) surfaces. (a) Top down view along [000-1], (b) Side view along [11-20].
112
For the (3x3) (0001)Si 6H-SiC surface, Kaplan [29] originally proposed a model
based on AES data for a SiC surface terminated by a bilayer of silicon. Based on analogy to
the (7x7) Si (111) DAS model, Kaplan proposed a (3x3) unit cell which consisted of two
adatoms, six rest atoms (three dimers), and eight silicon atoms in the second layer positioned
approximately directly over the silicon atoms of the SiC substrate. However, the recent STM
results of Kulakov et al [22] detected only one maxima (i.e. one adatom) in the (3x3) unit cell
which is in contrast to the model proposed by Kaplan which would predict two maxima.
Based on this discrepancy, Kulakov et al [22] proposed a modified structure which was
consistent with the AES results of Kaplan and their STM data. The model for the (3x3)
surface proposed by Kulakov et al [22] consists of a unit cell with 1 adatom, 3 rest atoms, and
7 silicon atoms located approximately on top of the silicon atoms of the SiC surface (see
Figure 5.3). This model includes 3 dimers and three dangling bonds (two unsatisfied Si
bonds from the SiC substrate, and 1 dangling bond from the adatom) compared to the 4
dangling bonds in the model by Kaplan [29]. However, Kulakov et al [22] did observe
stacking faults in their (3x3) reconstructed surface which had a structure essentially like that
of the (3x3) model proposed by Kaplan [29] (see Figure 5.4). Using STM, Li and Tsong [21]
also confirmed the presence of one maxima in the (3x3) unit cell but, in contrast, concluded
that the (3x3) reconstruction consisted of only 4/9 ML coverage of silicon for the (0001)Si
6H-SiC surface. Accordingly they attributed the (3x3) surface to extra Si-C tetrahedra on the
surface distributed in a 3x3 pattern (see Figure 5.5) rather than a bilayer of silicon.
113
[11-20]
[10-10]
a.)
View Along [10-10] or [1-21] [001], [111]
[11-20], [1-10]
b.)
View Along [11-20] or [1-10] [0001], [111]
[10-10], [1-21]
C atomSi atomSi Dangling
Bond
c.)
114
Figure 5.3. Model proposed by Kulalov et al [22] for the (3x3) reconstructed (0001)Si 6H-SiC surface. (a) Top down view along [000-1], (b) side view along [11-20], and (c) [10-10].
[11-20]
[10-10]
View Along [11-20] or [1-10] [0001], [111]
[10-10], [1-21]
C atomSi atomSi Dangling
Bond
a.)
b.)
Figure 5.4 Model proposed by Kaplan [29] for the (3x3) reconstructed (0001)Si 6H-SiC surface. (a) Top down view along [000-1], (b) side view along [11-20].
115
[11-20]
[10-10]
Figure 5.5. Top down view of model proposed by Li and Tsong [21] for the (3x3) reconstructed (0001)Si 6H-SiC surface.
In this paper, we report the first XPD patterns obtained from (0001)Si 6H-SiC
surfaces. The XPD patterns obtained from the (3x3) and (√3x√3)R30° (0001)Si 6H-SiC
surfaces are compared with those obtained from (7x7) (111) Si surfaces and the proposed
models for these SiC reconstructions. Based on this data, we are able to support only some
of the proposed models for these reconstructions. Estimates of the experimental inaccuracies
116
possible in XPS experiments due to anisotropies in the angular dependence of the intensity of
Si 2p and C 1s photoemission are also provided by this data.
5.3. Experimental
The experiments described in this paper were conducted in an integrated surface
analysis and growth system which has been previously described [40,41]. In this study, only
the XPS and SiC ALE systems were used. The n-type (Nd=1018/cm3), off axis (4° toward
{11-20}) (0001)Si 6H-SiC wafers used in this research were supplied by Cree Research, Inc.
with an ≈ 1 µm n-type 6H epilayer (Nd=1017/cm3) and a 1000Å thermally grown oxide.
The back side of the SiC wafer was sputter coated with tungsten after removal of the thermal
oxide with a 10 min. dip in 10:1 HF. The back side tungsten coating was necessary in order
to improve the heating efficiency of the SiC wafer by our tungsten filament heater as SiC is
transparent in the infra-red. Prior to insertion into the SiC ALE system, the SiC wafers were
given an ex situ clean consisting of ultrasonification in trichloroethylene, acetone, and
methanol for 10 min. each, followed by a 10 min. 10:1 buffered HF vapor clean to remove
any native oxides. The SiC wafer was then loaded into the SiC ALE system and annealed in
10-6 Torr SiH4 for 15 min. at 1050°C. This produced an oxygen free (3x3) reconstructed
surface. The (√3x√3)R30° reconstruction was generated by annealing the (3x3) surface in
UHV in the ALE system at 1050°C for ≈ 10 min. Additional details regarding our sample
preparation and procedure have been previously published [40-42,43].
After either the (3x3) or (√3x√3)R30° surface had been prepared, the SiC wafer was
transferred in situ to the XPS system. XPD patterns were acquired in this system by rotating
the SiC wafer about various polar and azimuthal angles using a computer driven goniometer 117
with five degrees of freedom (x,y,z,θ, and φ) while the positions of the x-ray source and
electron energy analyzer were fixed. Though the angular acceptance of the lens of the
electron energy analyzer (VG CLAMII) was ± 7°, an angular resolution of ≈ ±1° was
achieved by geometric constraints via grounding the lens and using smaller channeltron
acceptance slits. The SiC XPD patterns were acquired by monitoring the Si 2p and C 1s core
levels photoexcited by Al Kα radiation (hν = 1486.6 eV). The kinetic energy of these
photoelectrons (≈ 1380 and 1203 eV respectively) was sufficiently high that forward
scattering effects should be dominant and probe ≈ 20Å of the SiC surface. Polar scans along
high symmetry azimuths were acquired in increments of 0.9° from -35° to 70°. The wafer
flat provided by Cree was used to locate the various azimuths and is of the {10-10} family of
planes. The data presented is the raw angular distribution of the measured intensity of the Si
2p and C 1s core levels. No attempts were made to correct for background, variation in
sampling depth, or surface area seen by the electron energy analyzer. To ensure that the
system was operating properly, XPD spectra were first acquired from Si (100) and Si (111)
surfaces and compared with previously published high resolution XPS spectra for these
surfaces [44-48]. Figure 5.6 displays an XPD spectrum from a (2x1) Si (001) surface along
the [110] azimuth. As illustrated, sharp features with FWHM ≅ 3° were easily resolved and
were found to be in excellent agreement with previously reported results for this surface [48].
118
-40 -30 -20 -10 0 10 20 30 40 50 60 70
45Þ15Þ
20Þ23Þ
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
30Þ
60Þ
Figure 5.6. XPD pattern from (2x1) Si (100) along the [110] azimuth.
5.4. Results
Figures 5.7-5.11 show various Si 2p and C 1s XPD patterns obtained from (7x7) Si
(111), and (3x3) and (√3x√3)R30° 6H-SiC (0001)Si surfaces. These patterns were acquired
along high symmetry azimuths such as [10-10], [11-20], and [01-10] azimuths (note <1-
21>cub = <10-10>hex and <1-10>cub = <11-20>hex). As shown in these figures, peaks of
varying FWHM were observed. Generally speaking, diffraction structures of FWHM ≈ 10-
20° are consistent with forward scattering features associated with crystallographic directions
or nearest neighbor atoms [1-3]. Narrow peaks or wide peaks with fine structure of ≈ 3° are
usually due to scattering from more distant atoms or complex/higher order interference
phenomena [1-3].
119
-10 0 10 20 30 40 50 60 70
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
(a)
(b)
(c)
Figure 5.7. Si 2p x-ray photoelectron diffraction pattern along [1-21]/[10-10] azimuths from (a) (7x7) Si (111), (b) (3x3) 6H-SiC (0001)Si, and (c) (v3xv3)R30° (0001)Si 6H-SiC.
120
-10 0 10 20 30 40 50 60 70
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
(a)
(b)
(c)
Figure 5.8. Si 2p x-ray photoelectron diffraction patterns from (a) (3x3) 6H-SiC (0001)Si along [01-10], (b) (3x3) 6H-SiC (0001)Si along [10-10], and (c) (v3xv3)R30° 6H-SiC (0001)Si along [01-10].
121
-10 0 10 20 30 40 50 60 70
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
(a)
(b)
(c)
Figure 5.9. Si 2p x-ray photoelectron diffraction patterns along [-110]/[11-20] from (a) (7x7) Si (111), (b) (3x3) 6H-SiC (0001)Si, and (c) (v3xv3)R30° 6H-SiC (0001)Si.
122
-15 -10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
(a)
(b)
Figure 5.10. C 1s XPD patterns from (v3xv3)R30° 6H-SiC (0001)Si along (a) [01-10], and (b) [11-20].
123
-10 0 10 20 30 40 50 60 70
Cou
nts (
arb.
uni
ts)
Polar Angle (ÞTheta)
(a)
(b)
(c)
Figure 5.11. C 1s XPD patterns from (3x3) 6H-SiC (0001)Si along (a) [10-10], (b) [11-20], and (c) [10-10] azimuths.
124
In order to determine which peaks in the XPD spectra were due to forward scattering
effects, we next consider the atomic structure of the (0001)Si 6H-SiC surface. SiC is a
unique material that exhibits several different polymorphs which differ only in the stacking
sequence along the c axis. This particular phenomena is referred to as polytypism, and as
many as 256 different polytypes of SiC have been reported. However, only a few polytypes,
3C, 4H, 6H, and 15R are commonly observed. In the Ramsdell notation used to describe
these polytypes, the preceding number represents the number of Si-C bilayers needed to
repeat the stacking sequence along [111]/[0001] directions and the following letter describes
the crystal structure (i.e. C = cubic, H = hexagonal, and R = rhombohedral). In the case of
6H-SiC, the H is deceiving as 6H-SiC is actually 66.6% cubic and exhibits an ABCB'C'A'
stacking sequence which is similar to that of 3C-SiC differing only in the periodic stacking
fault in the 6H structure (see Figure 5.12). Accordingly, in a surface sensitive technique such
as XPD which effectively only samples the first 10Å of the surface, (0001) 6H and (111) 3C-
SiC should be essentially indistinguishable. Therefore for simplicity sake, we will treat the
(0001) 6H XPD spectra as if it were from (111) 3C-SiC. This is fortuitous as 3C-SiC and Si
and have similar crystal structures and therefore comparisons can be made between XPD
spectra from (111) Si and (111)/(0001) 3C/6H-SiC. Based on these considerations, the
expected peaks for forward scattering/focusing along certain crystallographic directions for
bulk terminated (111) 3C-SiC, and (0001)Si 6H-SiC surfaces are listed in Table 1 (see
Figures 5.13 and 5.14). Table 5.1 presents the forward scattering peaks expected from both
C 1s and Si 2p photoelectrons. In the case of Si (111), the C 1s forward scattering peaks
125
would be expected to appear in Si 2p XPD patterns as the carbon atoms have been replaced
by Si atoms.
3C-SiC
[111]
[1-21][1-10]
6H-SiC
[0001]
[10-10][11-20]
Stacking = ABCABC
Stacking = ABCB'C'A'
126
Figure 5.12. Schematic illustrating the differences in stacking along the [111]/[0001] direction for 3C and 6H-SiC.
View Along [11-20] or [1-10] [0001], [111]
[10-10], [1-21]
35.2Þ
54.7Þ70.5Þ
29.5Þ
0Þ
35.2Þ
Figure 5.13. Schematic illustrating expected forward scattering/focusing peaks in XPD along the [11-20] azimuth of 3C/6H-SiC.
127
58.5Þ
72.9Þ44.4Þ31.4Þ
View Along [10-10] or [1-21] [0001], [111]
[11-20], [1-10]
Figure 5.14. Schematic illustrating expected forward scattering/focusing peaks in XPD along the [10-10] azimuth for 3C/6H-SiC. Table 5.1. Expected forward scattering/focusing peaks from bulk terminated (111) Si, (111) 3C-SiC, and (0001)Si 6H-SiC surfaces along [10-10], [11-20], and [01-10] azimuths.
Si 2p [10-10] [11-20] [01-10] Scatterer 35.3° Si 54.7° Si 58.5° C 70.5° C-Si 72.9° C C 1s [10-10] [11-20] [01-10] Scatterer 29.5° Si 31.4° Si 35.3° C 44.4° Si 54.7° C 70.5° Si
5.4.1. (√3x√3)R30° (0001)Si 6H-SiC
128
For Si 2p XPD patterns obtained from (√3x√3)R30° 6H-SiC (0001)Si surfaces, most
of the expected forward scattering peaks were identified. In the [10-10] azimuth (see Figure
5.7(c)), a broad/intense peak at 36-37° was identified as expected. This peak was similar to
and consistent with forward scattering along the [011]/[10-11] crystallographic axes (i.e. the
Si-C atomic row). However unlike the (7x7) Si (111) surface, broad/intense peaks at 19 and
42° were symmetrically observed on both sides of the [011]/[10-11] forward scattering peak.
As mirror symmetry is expected about the [011] atomic row due to (100) glide planes [44],
these additional peaks are probably due to higher order interference phenomena or forward
scattering from larger emitter-scatterer distances. Similar to (7x7) Si (111), peaks of this
nature were also observed at 15 and 59° in Si 2p XPD patterns along the [10-10] azimuth.
However in contrast to (7x7) Si (111), a forward scattering peak at 70.5° was not observed.
In the case of Si (111), this peak is due to forward scattering along the (11-1) crystallographic
direction. The absence of this peak from (√3x√3)R30° 6H-SiC (0001)Si surfaces may be
related to the fact that (0001) and (000-1) are not equivalent directions in SiC but are
equivalent directions in Si.
C 1s XPD patterns obtained along the [10-10] azimuth showed a single sharp peak at
30° which is 5° off from the expected value of 35° for forward scattering along the [011]/[10-
11] atomic row (see Figure 5.10(a)). This discrepancy, however, may be related to the fact
that the 35° peak would be expected based on forward scattering by carbon atoms. Carbon
has a smaller atomic nucleus and should be expected to be a weaker scatterer. Therefore the
position of this forward scattering peak maybe determined more by silicon atoms along the
[011]/[10-11] row.
129
Like Si (111), asymmetry's were observed between Si 2p XPD spectra acquired along
<10-10> and <01-10> azimuths of (√3x√3)R30° 6H-SiC (0001)Si (i.e. <10-10> ≠ <01-10>).
In the [01-10] azimuth, the expected peak for forward scattering in the [100] direction was
observed at 55°C (see Figure 5.8(c)). Similar to Si 2p XPD patterns along [-12-1] azimuths
[44], peaks due to complex/higher order interference phenomena were also observed from
10-40° in the [01-10] azimuth of (√3x√3)R30° 6H-SiC (0001)Si surfaces. However, unlike
Si, symmetry was not observed about this peak.
In the [11-20] azimuth, a mosaic of broad diffraction peaks of equal intensity were
observed from Si 2p XPD patterns from the (√3x√3)R30° 6H-SiC (0001)Si surface (see
Figure 5.9(c)). Most of the peaks observed in the Si 2p XPD pattern from (7x7) Si (111)
observed along the [1-10] azimuth were also observed in the Si 2p XPD patterns along the
[11-20] azimuth from the (√3x√3)R30° 6H-SiC (0001)Si surface. However, the peak at
58.5° expected for forward scattering along the [-131] direction was observed to have a
volcano shape for (√3x√3)R30° 6H-SiC (0001)Si surface and a rounded shape for the (7x7)
Si (111) surface. In fact, rounded peaks from the Si (111) surface along [-110] were
observed to have a volcano shape in the [11-20] azimuth of the SiC surface (and vice versa).
The C 1s XPD patterns obtained along the [11-20] azimuth showed strong peaks at 25, 35,
and 52.5° (see Figure 5.10(b)). The sharpest/most intense peak at 35° is slightly off from the
expected value of 31.4° for scattering by a top layer carbon atom (see Figure 5.14).
However, this can be expected as the scattering carbon atom does not actually lie in the (10-
10) plane.
The maximum anisotropy in intensity observed in both the Si 2p and C 1s XPD
patterns was observed about the 0°/[0001] forward scattering peak. For Si 2p and C 1s the 130
maximum anisotropy (Imax-Imin)/Imax was ˜ 65 and 40 % respectively. Similar to (7x7) Si
(111), higher order diffraction effects were also observed at 10-15° on both sides of the Si 2p
and C 1s 0°/[0001] forward scattering peaks. However in contrast to Si (111), the Si 2p
0°/[0001] forward scattering peak from (0001)Si 6H-SiC surfaces does not exhibit a volcano
type shape but rather a flat sawtooth type shape (see Figure 5.8(a)). The C 1s 0°/[0001]
forward scattering peak from (0001)Si 6H-SiC does exhibit a volcano shape (see Figure
5.10(a)). A similar effect has been observed between Si 2p and C 1s spectra from (001) Si
and 3C-SiC [48-50]. The shape of this peak is strongly affected by the presence of scattering
atoms surrounding the [111]/[0001] direction. As silicon is the nearest neighbor atom to
carbon along the [0001] direction, scattering by these silicon atoms probably induces the
volcano shape observed in the C 1s XPD. For silicon atoms in SiC, carbon is the nearest
neighbor atom, but the scattering factor of carbon is much weaker, hence the sawtooth
structure. However in pure silicon and diamond, all the atoms are either silicon or carbon
and the volcano shape reappears [44-48,51]. The also explains many of the differences
between Si 2p XPD spectra from Si and SiC along the [11-20] azimuth. Finally, it should be
mentioned that the centroid of the Si 2p and C 1s 0°/[0001] forward scattering peaks were
observed to vary by ± 2°. As this variation was not observed from on axis (001) and (111) Si
substrates, the variation in the position of [0001] in the polar scans is probably related to the
fact that the SiC wafers were 4° off axis in the <11-20> direction and not related to drift in
our sample stage.
5.4.2. (3x3) (0001)Si 6H-SiC
131
In contrast to reconstructed and unreconstructed diamond and silicon surfaces [44-
48,51], significant differences were observed between the Si 2p XPD patterns from (3x3) and
(√3x√3)R30° 6H-SiC (0001)Si surfaces. For instance, the [011]/[101-1] forward scattering
peak at ≈ 33° was observed from [10-10] Si 2p XPD patterns from both (3x3) and
(√3x√3)R30° 6H-SiC surfaces (see Figure 5.7(b) and (c)). However, peaks centered
symmetrically at 29 and 46° about the [011]/[101-1] forward scattering feature were not
observed from the (3x3) surface (see Figure 5.7(b)) which is more similar to Si 2p XPD
patterns from Si (111) in this azimuth (see Figure 5.7(a)). Additionally, the sharp higher
order diffraction peaks at 9-18 and 58° observed from the (√3x√3)R30° surface were
observed to be more broad and less intense for the (3x3) surface. Similar differences were
also seen in the (3x3) C 1s XPD patterns in the [10-10] azimuth. In this case, a volcano
shaped peak centered at 35° was observed in [10-10] C 1s XPD patterns from the (3x3)
surface instead of the one sharp peak centered at 30° observed from the (√3x√3)R30° 6H-SiC
surface (see Figure 5.11(c)). Sharper peaks centered symmetrically about the 35° volcano
peak at 20 and 59° were also observed in the (3x3) [10-10] C 1s XPD patterns.
In the [11-20] azimuth, a mosaic of sharp higher order diffraction features were
observed in C 1s XPD patterns from the (3x3) surface instead of the single sharp peak
centered at 35° observed in the (v3xv3)R30° XPD patterns (see Figure 5.11(b)). However
for [11-20] Si 2p XPD patterns, there were not any significant differences between the (3x3)
and (√3x√3)R30° 6H-SiC surfaces (see Figure 9(b),(c)). Finally in Si 2p XPD patterns, the
0°/[0001] forward scattering peak was observed to be volcano shaped for the (3x3) surface
whereas this peak exhibited a sawtooth/rounded shape for the (√3x√3)R30° 6H-SiC surface
(see Figures 5.7(b),(c) and 5.8(b),(c)). The opposite, however, was observed in C 1s XPD
132
patterns where a volcano shaped 0°/[0001] forward scattering peak was observed from the
(√3x√3)R30° surface whereas a sawtooth/rounded peak was observed from the (3x3) surface
(see Figures 5.10 and 5.11).
Perhaps the largest differences between XPD of (3x3) and (√3x√3)R30° 6H-SiC
surfaces are found in the [01-10] azimuth. As can be seen in Figures 5.8(a),(b) and
5.11(a),(c), Si 2p and C 1s XPD patterns from the (3x3) 6H-SiC surface along the [01-10]
and [10-10] azimuths are identical. This is in complete contrast, to the (√3x√3)R30° 6H-SiC
surface in which the [10-10] and [01-10] azimuths were observed to be completely different
(i.e. asymmetric about [0001] along <10-10>). The equivalence of the [10-10] and [01-10]
directions for the (3x3) 6H-SiC surface suggests drastic changes in the surface structure of
the SiC which will be discussed further in the next section.
5.5. Discussion
5.5.1. (√3x√3)R30° (0001)Si 6H-SiC
As some peaks were observed in XPD patterns from (√3x√3)R30° (0001)Si 6H-SiC
which had not been previously observed from Si (111) surfaces, attempts were initially made
to see if these extra peaks could be assigned to forward scattering/diffraction due to silicon or
carbon adatoms on the surface. As previously mentioned, adatoms in T4 or H3 sites are
commonly believed to be the origin of the (√3x√3)R30° reconstruction. Table 5.2 lists the
additional expected forward scattering peaks due to Si or C adatoms based on various
different models proposed for the (√3x√3)R30° reconstruction [31-34]. Unfortunately, we
133
were not able to determine with any certainty whether any of these peaks truly existed.
Therefore, single scattering cluster simulations are probably necessary in order to determine
the exact structure of the (√3x√3)R30° reconstruction based on XPD data. However, the
authors note that Pirri et al [44] and Kuttel et al [51] have experienced similar difficulties in
distinguishing between XPD patterns from (7x7) and (1x1) Si (111) and (2x1) and (1x1)
Diamond (111) respectively.
Table 5.2. Expected Si 2p and C 1s photoelectron diffraction peaks for adatom scattering in T4 and H3 positions based on data of Northrop and Neugebauer [31] and Badziag [34] for (√3x√3)R30° (0001)Si 6H-SiC.
Si adatom, T4 Northrop & Neugebauer [31] <10-10> <11-20> <-1010> Si 2p C 1s Si 2p C 1s Si 2p C1s 66.9° 56.0 45.2° 65.2° 53.6° 52.1° 43.2° 20.1° 36.5° C adatom, T4 Northrop & Neugebauer [31] <10-10> <11-20> <-1010> Si 2p C 1s Si 2p C 1s Si 2p C1s 71.8° 59.1° 53.8° 55.3° 56.1° 22.3° 40.6° C3 Model Badziag [34] <10-10> <11-20> Si 2p C 1s Si 2p C 1s 68.8° 48° 33.3° 37.9° 27.3° 20.2° 17.0° 14.5° 9.5°
134
5.5.2. (3x3) (0001)Si 6H-SiC
As mentioned above, significant differences were observed between Si 2p and C 1s
XPD patterns from the (3x3) and (√3x√3)R30° 6H-SiC (0001)Si surfaces. Perhaps, the most
striking difference was the observed equality of the [10-10] and [01-10] azimuths for the
(3x3) reconstruction and the inequality of these azimuths for the (√3x√3)R30° reconstruction.
In order to gain further insight into the nature of these differences, comparisons were made to
previously proposed models for the (3x3) reconstruction based on recent STM images
[21,22]. Based on these models, a new set of forward scattering peaks were
calculated/estimated for the (3x3) reconstruction and which are presented in Table 5.3. As
discussed in the introduction, Li and Tsong [21] have proposed that the (3x3) reconstruction
is a result of 4/9 ML absorption of Si-C tetrahedra arranged in a (3x3) pattern (see Figure
5.5). As can be seen in Figure 5.5 and Table 5.3, this model does not predict equivalence of
the [10-10] and [01-10] azimuths. This model is also in disagreement with our previous
observation of a Si-Si Si 2p bonding peak in XPS which indicated a partial bilayer of silicon
on the (3x3) SiC surface [40].
Based on the previous AES data of Kaplan [29] and their STM data, Kulakov et al
[22] proposed a different model for the (3x3) reconstruction which consisted of an
incomplete bilayer of Si. We find this model for the (3x3) reconstruction to be in better
agreement with our observed XPS and XPD patterns. First, this is clearly consistent with
our observation of a Si-Si Si 2p bonding peak in XPS. As can be seen in Figure 5.3(c), this
model also specifically adds an additional Si-Si bilayer to the [011]/[10-11] atomic row. As
silicon has a larger nucleus it should be a more effective scatterer than carbon. Therefore the
135
addition of the Si-Si bilayer should enhance the intensity along the [011]/[10-11] chain due to
increased forward scattering. This is exactly what we observe in both our C 1s and Si 2p [10-
10] XPD patterns. Unfortunately, the model proposed by Kulakov et al fails to explain the
observed equality of our [10-10] and [01-10] Si 2p and C 1s patterns. However, we do note
that the model originally proposed by Kaplan [29] for the (3x3) reconstruction would explain
the equivalence of the [10-10] and [01-10] XPD patterns (see Figure 5.4). This is primarily a
result of the stacking fault in this structure which produces Si-Si bilayers oriented in both the
[10-10] and [01-10] directions. The presence of this structure on SiC surfaces has actually
been confirmed by Kulakov et al [22] were they observed faults or domains of different
orientation in STM images of the (3x3) surface. The stacking structure in these domains are
consistent with the model originally proposed by Kaplan for the (3x3) reconstruction [29].
Unfortunately however, at this stage it is difficult to determine with certainty whether our
observations of the equivalence between [10-10] and [01-10] in XPD are due to these
stacking faults.
Table 5.3. Estimated forward scattering peaks for (3x3) reconstructed (111)/(0001) 3C/6H-SiC surfaces based on models proposed by Kulakov et al [22] and Li and Tsong [21].
Si 2p Kulakov Li & Tsong [11-20] [10-10] [01-10] [11-20] [10-10] [01-10] 59.3° 66.2° 66.2° 59.3° 66.2° 48.6° 51.5° 53.7° 53.7° 52.6° 29.5° 42.9° 29.5° 48.6° 42.9° 38.2° 33.2° C 1s Kulakov Li & Tsong [11-20] [10-10] [01-10] [11-20] [10-10] [01-10] 59.4° 43.3° 38.0° 58.5° 43.3° 58.5°
136
39.3° 30.8° 30.8° 39.3° 30.8°
27.3° 21.3° 25.3° 25.3°
137
5.6. Conclusion
High resolution (±1°) x-ray photoelectron diffraction (XPD) patterns were obtained
along high symmetry azimuths of (3x3) and (√3x√3)R30° reconstructed (0001)Si 6H-SiC
surfaces. The data obtained were compared to previously reported XPD patterns from (7x7)
Si (111) as well as models proposed models for the (3x3) and (√3x√3)R30° 6H-SiC
reconstructions. Forward scattering features similar to those observed from (7x7) Si (111)
were also observed from (√3x√3)R30° 6H-SiC (0001)Si surfaces. However, additional
features not observed in (7x7) Si (111) were observed in the (√3x√3)R30° 6H-SiC XPD
patterns which were attributed to the substitution of carbon atoms for silicon atoms on the
diamond FCC lattice. Unlike (1x1) and (7x7) Si (111) surfaces, differences were observed
between the XPD patterns of (3x3) and (√3x√3)R30° SiC (0001)Si surfaces. The most
significant difference observed between the (3x3) and (√3x√3)R30° reconstructions was the
equivalence of the [10-10] and [01-10] azimuths in the (3x3) structure. The faulted (3x3)
structure proposed by Kulakov et al [22] was found to be consistent with the (3x3) XPD
patterns presented here.
5.7. Acknowledgments
The authors would like to thank Cree Research, Inc. for supplying the 6H-SiC wafers.
.8 References
The authors would also like to thank the REU program for partially sponsoring this effort.
Meaningful discussions regarding experimental setup with Dr. Egelhoff are also noted. The
research was supported by ONR under contract and the Department of Education via an
Electronic Materials/GAANN fellowship..
5
, Jr., Crit. Rev. in Solid State and Materials Sciences, 16 (1990) 213.
s. Rev. B, 33 (1986) 8837.
nce, 287/288 (1993) 495.
.L. Bischoff, L. Kubler, and D. Bolmont, Surface Science 291 (1993) 110.
.K. Saldin, and B.P. Tonner, Phys. Rev. Lett., 65 (1990) 1012.
14423.
rp, D.K. Saldin, and B.P. Tonner, Phys. Rev. B, 42 (1990) 9199.
L. Schlapbach, Surface Science, 264 (1992) 380.
im, and C.S. Fadley, Phys. Rev. Lett., 68 (1992) 650.
, M. Shur, J. Palmour, J.A. Edmond, Proc. of the IEEE, 79 (1991) 677.
1. W.F. Egelhoff 2. S.A. Chambers, Surface Science Reports, 16 (1992) 261. 3. S.A. Chambers, Adv. in Physics, 40 (1991) 357. 4. D.A. Wesner, F.P. Coenen, and H.P. Bonzel, Phy 5. M. Seelmann-Eggbert, G.P. Carey, R. Klauser, and H.J. Richter, Surface Scie 6. M. Diani, D. Aubel, J 7. G.R. Harp, D 8. B.P. Tonner, Z.L. Han, G.R. Harp, and D.K. Saldin, Phys. Rev. B, 43 (1991) 9. G.R. Ha 10. A. Stuck, D. Naumovic, H.A. Aebischer, T. Greber, J. Osterwalder, and 11. G.S. Herman, S. Thevuthasan, T.T. Tran, Y.J. K 12. R.F. Davis, G. Kelner 138
13. M.K. Sanganeria, M.C. Ozturk, G. Harris, K.E. Violette, I. Ban, C.A. Lee, and
D.M. Maher, J. Electrochem. Soc., 142 (1995) 3961.
. Soc. 133 (1986) 1232.
n and B.S. Meyerson, J. Electrochem. Soc., 134 (1987) 1518.
avis, Surface Science, 256 (1991) 354.
v, P. Heuell, V.F. Tsvetkov, and B. Bullemer, Surface Science, 315 (1994) 248.
. Martensson, Surface Science, 330 (1995) L639.
T. Ozawa, T. Nagamura, and T. Nakata, Jpn. J. Appl. Phys., 34 (1995) 3351.
6 (1996) 49.
ience, 350 (1996) 247.
ce Science, 221 (1989) 244.
awa, S. Kono, Surface Science, 356 (1996) 28.
Surface Science, 82/83 (1994) 434.
hramtsova, S.V. Ryzhkov, A.A. Saranin, A.B. Chub, V.G. Lifshits, Surface Science, 316 (1994) L1034.
(1995) 45.
R17001.
14. B.S. Meyerson, E. Ganin, D.A. Smith, and T.N. Nguyen, J. Electrochem 15. G.R. Srinivasa 16. W. Kern, J. Electrochem. Soc., 137 (1990) 1887. 17. C.S. Chang, I.S.T. Tsong, Y.C. Wang, and R.F. D 18. M.A. Kulako 19. F. Owman, P 20. Y. Marumoto, T. Tsukamoto, M. Hirai, M. Kusaka, M. Iwami, 21. L. Li, and I.S.T. Tsong, Surface Science, 351 (1996) 141. 22. M.A. Kulakov, G. Henn, B. Bullemer, Surface Science, 34 23. S. Tanaka, R.S. Kern, R.F. Davis, J.F. Wendelken, and J. Xu, Surface Sc 24. F. Owman, P. Martensson, J. Vac. Sci. & Technol., 14 (1996) 933. 25. H. Daimon, S. Nagano, T. Hanada, S. Ino, S.Suga, Y. Murata, Surfa 26. X. Chen, T. Abuk 27. Y. Taguchi, M. Date, N. Takagi, T. Aruga, M. Nishijima, Applied 28. A.V. Zotov, E.A. K 29. R. Kaplan, Surface Science, 215 (1989) 111. 30. V.M. Bermudez, Applied Surface Science, 84 31. J.E. Northrup and J. Neugebauer, Phys. Rev. B, 52 (1995)
139
32. J.E. Northrup, Phys. Rev. Lett., 57 (1986) 154. 33. P. Badziag, Surface Science, 337 (1995) 1. 34. P. Badziag, Surface Science, 352 (1996) 396.
nce 201 (1988) L513.
hys. Rev. B, 42 (1990) 7230.
cience, 165 (1986) 83.
, Surface Science, 237 (1990) 35.
.J. Nemanich, and R.F. Davis, submitted to J. Appl. Phys.
SU (1993).
. Electrochem. Soc.
) 313.
winner, U. Kafader, P. Wetzel, and C. Pirri, J. Electron. Spect. and Rel. Phenoma.
ff, L. Kubler, F. Lutz, M. Diani, and D. Bolmont, Solid State Comm., 83 (1992) 823.
rbone, R. Rochow, Phys. Rev. B, 46 (1992) 13215.
51/252 (1991) 305.
llaguet, L. Kubler, M. Diani, J.L. Bischoff, G. Gewinner, P. Wetzel, N. Becourt, Surface Science, 339 (1995) 363.
olmont, Applied Surface Science, 68 (1993) 575.
R.G. Agostino, R. Fasel, J. Osterwalder, and L. Schlapbach, Surface
35. T. Abukawa, C.Y. Park, S. Kono, Surface Scie 36. P. Martensson, G. Meyer, N.M. Amer, E. Kaxiras, and K.C. Pandey, P 37. B.B. Pate, Surface S 38. A.V. Hamza, G.D. Kubiak, and R.H. Stulen 39. M.D. Allendorf, D.A. Outka, Surface Science, 258 (1991) 177. 40. S.W. King, M.C. Benjamin, R.S. Kern, J.P. Barnak, D. Hanser, R 41. Jacob van der Weide, Ph.D. dissertation, NC 42. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to J 43. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to J. Electrochem. Soc. 44. C. Pirri, U. Kafader, G. Gewinner, and P. Wetzel, Solid State Comm., 89 (1994 45. G. Ge 46. J.L. Bischo 47. E. Puppin, C. Ca 48. L. Kubler, F. Lutz, J.L. Bischoff, and D. Bolmont, Surface Science, 2 49. S. Jui 50. M. Diani, J.L. Bischoff, L. Kubler, and D. B 51. O.M. Kuttel,
140
Science, 312 (1994) 131.
141
6. Interaction of Atomic Hydrogen with (3x3) 6H-SiC (0001)Si Surfaces.
To be Submitted for Consideration for Publication
to
Surface Science
by
Sean W. King, *Mark C. Benjamin, John P. Barnak,
*Robert J. Nemanich, and Robert F. Davis
Department of Materials Science and Engineering
*Department of Physics
North Carolina State University
Raleigh, NC 27695
142
6.1. Abstract
X-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES), low
energy electron diffraction (LEED), and temperature programmed desorption (TPD) were
used to examine the interaction of atomic hydrogen with (3x3) 6H-SiC (0001)Si surfaces. It
was found that atomic hydrogen exposure selectively removes silicon from the SiC surface
converting the (3x3) surface to a (1x1) surface. Selective removal of silicon was witnessed
by the reduction and removal of the Si-Si bonding Si 2p XPS peak from (3x3) (0001)Si 6H-
SiC surfaces exposed to atomic hydrogen. Additional etching of the SiC surface was
indicated by the reduction in the Si LVV/C KLL ratio in AES from 1.3 to 0.4 following
exposure of (3x3) surfaces to a remote rf H plasma. TPD of atomic H treated (3x3) SiC
surfaces showed weak hydrogen desorption in the range of 400-600°C where desorption from
silicon atoms would be expected by analogy to (111) Si. However, the hydrogen desorption
signal increased at higher temperatures where hydrogen desorption from carbon sites would
be expected based on analogy to (111) diamond surfaces. C-H termination of the SiC surface
was supported by the observation of some C-C bonding after thermal desorption of rf plasma
treated SiC surfaces at T > 1000°C.
143
6.2. Introduction
Hydrogen is a common constituent in many semiconductor processes including
chemical vapor deposition (CVD), reactive ion etching (RIE), wet chemical cleaning, gas
source molecular beam epitaxy (GSMBE), and rapid thermal annealing (RTA) [1-12].
Knowledge of the interaction and chemistry of hydrogen at semiconductor surfaces is
therefore of great importance in order to understand the fundamentals of the above processes.
For these reasons, many surface analytical studies concerned with the interaction of hydrogen
with semiconductor surfaces such as silicon, diamond, and gallium arsenide [13-46] have
been conducted. However, there have been relatively few surface analytical studies
concerned with the interaction of hydrogen and silicon carbide (SiC) [47-51]. SiC is a wide
bandgap semiconductor (Eg (6H-SiC) = 3.0 eV) which is of interest for high power, high
frequency, and high temperature electronic devices due to its excellent oxidation resistance,
high saturation electron drift velocity (vsat = 2x107 cm/s), high breakdown voltage (EB = 2.5
MV/cm) , high thermal conductivity (κ = 4.9 W/cm K), and high melting point (Tmelt ≈
3000°C) [52,53]. Due to moderately close lattice matching (?a/ao AlN/SiC = 0.8%, GaN/SiC
= 3.5%), SiC is also of interest as a heteroepitaxial substrate for growth of III-V nitride
compounds which in turn are of interest for blue/UV optoelectronic applications as well as
high power and high frequency devices [54]. Unfortunately though, many of the same
properties which make SiC of interest in these demanding conditions also makes it a
challenging material to work with from a processing point view. Therefore in order for SiC
to succeed in many of these applications, advances must be made in SiC processing such as
144
growth, etching, contact formation, and surface cleaning [52]. As many of these processes
are currently based on using hydrogen or hydrogenated species, an increased understanding
of the interaction of hydrogen with SiC surfaces should assist in the further development of
these processes.
Some of the first investigations of the interaction of hydrogen with SiC revealed that
etching of SiC by hydrogen occurred at high temperatures (>1500°C) [55-58]. The work of
Chu and Campbell [55] in particular found hydrogen to be a non-preferential etchant for
single crystal hexagonal SiC yielding useful etch rates of 2-4 microns/min. at temperatures of
1600-1700°C. Further, Bartlett and Mueller [57], found an H2 etch prior to SiC CVD to be
instrumental in obtaining good homoepitaxy. Subsequent hydrogen/SiC studies focused on
the interaction of high energy H+ and D+ ions (i.e. sputtering and implantation) with
polycrystalline 3C-SiC which in these studies was being considered as a first wall material
for thermonuclear fusion reactors [59-65]. However, more recent studies have investigated
the interaction of low energy (i.e. thermally generated) atomic hydrogen with polycrystalline
3C-SiC surfaces [47-48]. The first such study by Allendorf et al [47] used temperature
programmed desorption (TPD) and Auger electron spectroscopy (AES) to investigate the
adsorption and desorption of thermally generated hydrogen on sputter cleaned polycrystalline
3C-SiC surfaces. They observed > 1 ML adsorption of hydrogen which was observed via
TPD to desorb in a broad temperature range from 400-1000°C. Analysis of the broad
desorption feature indicated two desorption peaks at 700 and 850°C characterized by first
order desorption with activation energies of 63 and 72 kcal/mole respectively. Due to the
polycrystalline nature of these surfaces, Allendorf et al [47] unfortunately were not able to
assign these two desorption features to desorption from specific sites. However, they did
145
show a reduction in the AES Si/C ratio from 1.31 to 0.46 after atomic H exposure suggesting
etching or selective etching of the SiC surface. A similar effect was observed by Lannon et
al [49] for polycrystalline 3C-SiC films exposed to (H/H2)+ ions of various energies (10-
2000 eV). In this case, the 3C-SiC films were grown in situ via carbonization of (001)
silicon wafers with C2H4 and were not exposed to any sputtering. Finally, the reaction of
thermally generated atomic hydrogen with polycrystalline 3C-SiC films was also studied by
Kim and Olander [48] using modulated molecular beam mass spectrometry. In this case,
they were able to observe etching of SiC by atomic hydrogen at temperatures ranging from
300-1100K via the detection of SiH4, CH4, and C2H2 reaction products. Based on their
study and those of Allendorf et al [47], Kim and Olander proposed a precursor model for
SiH4 and CH4 formation based on a surface composed of adsorbed atomic hydrogen
overlaying mono and dihydrides of silicon and carbon on the SiC surface. In their model, the
dihydrides act as the precursors for SiH4 and CH4 generation and the production of these
species follows a first order reaction between dihydrides and the overlayer of adsorbed
atomic hydrogen. However, the results of Kim and Olander and those of Kim and Choi [58]
showed pronounced and enhanced etching of polycrystalline 3C-SiC surfaces at grain
boundaries suggesting that their results may be more indicative of processes occurring at
grain boundaries rather than at crystalline surfaces. Therefore in this study, we have chosen
to examine the reaction of atomic hydrogen with single crystal (3x3) reconstructed (0001)Si
6H-SiC surfaces.
The (3x3) reconstructed (0001)Si 6H-SiC surface is rapidly becoming a well
characterized semiconductor surface [66-73] and common starting point for most MBE
growth of SiC, AlN, and GaN on (0001)Si 6H-SiC substrates [73-75]. We and others have
146
shown this surface to consist of an incomplete bilayer of silicon overlaying the SiC surface
[66-68,76]. STM investigations have additionally shown this bilayer to be arranged in a
structure similar to the (7x7) Si (111) DAS model [70-72]. In our case, this surface was
prepared by chemical vapor cleaning processes and was not exposed to any sputtering
processes which are know to create a number of surface defects which can control surface
chemistry. Surface analytical techniques such as UV and x-ray photoelectron spectroscopy
(UPS and XPS), Auger electron spectroscopy (AES), low energy electron diffraction
(LEED), and temperature programmed desorption (TPD) were used to study not only the
effects of atomic hydrogen on the chemistry of SiC surfaces but on the electronic structure of
(3x3) 6H-SiC (0001)Si surfaces as well. The details of the effect on the electronic structure,
however, well be presented in a separate paper [77].
In this paper, we show that exposure of (3x3) 6H-SiC (0001)Si surfaces to atomic
hydrogen from a remote rf plasma source results in complete removal of all Si-Si bonded
silicon from the surface leaving a (1x1) surface. In turn, TPD showed weak H2 desorption in
the range of 400-600°C where desorption from silicon atoms would be expected based on
analogy to (111) Si surfaces [13-19]. A corresponding rise in H2 desorption was observed at
higher temperatures where H2 desorption from carbon sites would be expected based on
analogy to (111) diamond surfaces [32,34] and which in turn suggests C-H termination of the
SiC surface. This was supported by the observation of some C-C bonding after thermal
desorption of the rf plasma treated SiC surface at T > 1000°C. Similar to results previously
reported by other researchers, we also observed removal of silicon from single crystal (3x3)
6H-SiC (0001)Si SiC surfaces using smaller fluxes of atomic hydrogen generated from a hot
rhenium filament. In this case, complete removal of all Si-Si bonded silicon was not
147
observed, but TPD was able to detect H2 desorption in the range of 300-500°C indicative of
Si-H desorption. However, the hydrogen desorption in this temperature range was broad and
irregular resembling desorption from (2x1) B/Si (001) surfaces [78].
6.3. Experimental
All experiments described below were conducted using a unique ultra high vacuum
(UHV) configuration which integrates several completely independent UHV surface
preparation, thin film growth and surface analysis systems via a 36 ft. long transfer line
having a base pressure of 9x10-10 Torr (see Refs. 76 and 79 for details of the transfer line,
and many of the associated systems). The experiments described in this paper employed the
SiC atomic layer epitaxy (ALE)/temperature programmed desorption (TPD), Auger electron
spectroscopy (AES), low energy electron diffraction (LEED), x-ray photoelectron
spectroscopy (XPS), and remote H plasma systems. A brief description of these systems is
provided below.
The SiC ALE system consisted of a UHV chamber with a base pressure of 3x10-10
Torr and was equipped with a residual gas analyzer (RGA) and a variety of gas dosers. For
TPD experiments, the RGA (a 0-200 amu quadrapole gas analyzer from Hiden Analytical
Ltd.) was housed in a separate differentially pumped cylindrical chamber (similar in design
to that of Smentkowski and Yates [80] ). The RGA chamber had a 0.5 cm diameter orifice at
the head of the RGA for TPD experiments and an approximately 50 cm2 "sunroof" which
could be opened to allow monitoring of residual gases in the system. The sample heating
stage for the TPD experiments consisted of a wound tungsten heating filament positioned
148
close to the back of the sample and mounted on a boron nitride disk [79]. A W/6%Re-
W/26%Re thermocouple was employed to measure the temperature of the backside of the
wafer. Heating profiles to 1100°C were easily obtained using a programmable
microprocessor and 20 amp SCR power supply. Actual surface/sample temperatures (i.e.
those reported herein) were measured using an infra-red pyrometer with a spectral response
of 0.8 to 1.1 µm and a emissivity setting of 0.5. The estimated experimental accuracy for the
latter temperatures was estimated to be ± 25°C..
Low fluxes of atomic hydrogen were generated in the ALE system using a hot
filament fabricated from 0.25 mm diameter rhenium wire. Temperatures >1700°C as
measured by the previously mentioned optical pyrometer were used to generate the atomic
hydrogen. The rhenium filament was positioned approximately 3 1/2" away from the SiC
wafer. No attempts were made to try and accurately measure the atomic H flux at the SiC
surface, and exposures were quoted in units of Langmuirs (10-6 Torr sec.). All hot filament
atomic H exposures were conducted without heating of the SiC by the sample heater (i.e.
room temperature). Any heating of the SiC surface by the hot filament is felt to be minimal
and at most could have raised the surface temperature by 100°C.
The XPS experiments were performed in a stainless steel UHV chamber (base
pressure = 2x10-10 Torr) equipped with a dual anode (Mg/Al) x-ray source and a 100 mm
hemispherical electron energy analyzer (VG CLAM II). All XPS spectra reported herein
were obtained using Al Kα radiation (hν = 1486.6 eV) at 12 kV and 20 mA emission
current. XPS analysis typically required less than 1 hour during which time the pressure
never increased above 9x10-10 Torr. Calibration of the binding energy scale for all scans
was achieved by periodically taking scans of the Au 4f7/2 and Cu 2p3/2 peaks from
149
standards and correcting for the discrepancies in the measured and known values of these two
peaks (83.98 and 932.67 eV, respectively) [81]. Curve fitting of most data was performed
using the software package GRAMS 386. A combination Gaussian-Lorentzian curve shape
with a linear background was found to best represent the data. The Auger electron
spectrometer and the low energy electron diffraction optics were mounted on a six way cross
off the transfer line and pumped through the transfer line. In the AES analysis, a 3 keV, 1mA
beam was used. Each Auger electron spectrum was collected in the undifferentiated mode
and numerically differentiated. In LEED an 80 eV, 1mA beam was used.
The plasma system consisted of an all metal seal stainless steel vacuum chamber
pumped by a 330 l/s turbomolecular pump. The base pressure of this system was 4x10-9
Torr and was limited by the double o-ring sealed quartz tube attached to the top of the system
where the rf discharge was produced. The process gases flowed through this quartz tube and
an inductively coupled plasma was generated using an rf power supply (13.56 MHz) and rf
matching network attached to a copper coil wrapped around the quartz tube. The sample was
located 40 cm down from the center of the rf coil. An inline Nanochem purifier and filter
was used for point of use purification of hydrogen. Sample heating in the plasma system
was conducted using a sample heater similar in design to the one previously described in the
ALE system. Depending on the chamber pressure and rf power, the plasma could be
maintained in the quartz tube or extended down toward the sample region. For the
experiments described in this study, an rf power of 20 W and a chamber pressure of 10-15
mTorr was used which confines the plasma to the quartz tube (i.e. a remote plasma). The
plasma system was also equipped with a differentially pumped 0-100 amu RGA which
allowed direct analysis of the purity of the process gases. RGA analysis of the hydrogen
150
(99.999% purity) used in these experiments after in situ purification revealed that the
impurity level of these gases were below the baseline of the system (<1ppm).
The n-type (ND=1018/cm3), 1" diameter, off axis (4° toward {11-20}) (0001)Si 6H-
SiC wafers used in this research were supplied by Cree Research with an ≈ 1 µm n-type
(ND=1017/cm3) 6H epilayer and a 1000Å thermally grown oxide. The thermal oxide was
removed by a 10 min. dip in 10:1 HF. The unpolished back side of each wafer was
subsequently coated via RF sputtering with tungsten to increase the heating efficiency of the
SiC, as the latter is partially transparent to the infrared radiation emitted from the tungsten
filament heater. After coating the backside of the SiC wafer with tungsten, the SiC wafers
were ultrasonically rinsed in trichloroethylene, acetone and methanol each for 5-10 min. and
then exposed to the vapor from a 10:1 buffered HF solution for 10 min. The wafers were
then mounted to a 1" diameter ring shaped Mo sample holder using Ta wire and inserted into
the transfer line load lock for experimentation. The SiC wafer was then loaded into the SiC
ALE system and annealed in 10-6 Torr SiH4 for 15 min. at 1050°C. This produced an
oxygen free (3x3) reconstructed surface [76]. For comparison purposes, Si (111) wafers
were also examined in this study. In this case, n-type (0.8-1.2 ? cm) Si (111) wafers were
cleaned by dipping in 10:1 HF to remove the thermal oxide and then annealed in the ALE
system at 950°C. This produced an oxygen free (7x7) reconstructed surface.
6.4. Results
6.4.1. Interaction of rf Atomic H with (3x3) vicinal (0001)Si 6H-SiC surfaces
151
Figure 6.1(a) shows an AES spectrum obtained from a (0001)Si 6H-SiC wafer after
annealing in SiH4 to produce the (3x3) reconstructed SiC surface. As illustrated, the
treatment removes oxygen from the SiC surface to levels below the detection limit of AES
and produces a silicon rich surface. Accounting for the 2:1 difference in sensitivity between
Si and C in AES [47], the Si/C peak to peak height (pph) ratio for the 3x3 surface shown here
was 1.35. Figure 6.2(a) shows an XPS spectrum of the Si 2p core level obtained from the
same surface. As displayed, two peaks at 99.5 and 101.5 eV were detected by XPS and
which were indicative of Si-Si and Si-C bonding respectively. This figure indicates the > 1
Si/C pph ratio in AES for the (3x3) 6H-SiC surface is primarily due to excess silicon
deposited on the surface. In previous studies, it has been shown that the (3x3) reconstruction
corresponds to an incomplete bilayer coverage of silicon on top of the SiC surface [67,68,76].
Figure 6.1(b) shows an AES spectrum obtained from the (3x3) 6H-SiC (0001)Si
wafer after a 1 min., 450°C remote rf plasma treatment (20W, 15 mTorr). As can be seen,
the intensity of the Si LVV transition is greatly reduced by the rf plasma exposure and the
corresponding Si/C pph ratio is reduced to 0.4. Correspondingly, a reduction/elimination of
the Si-Si bonding peak in the Si 2p XPS spectrum was also observed (see Figure 6.2(b)).
(Note: the shift in energy of the Si 2p anc C 1s core levels with H plasma processing is
related to band bending effects discussed in a seperate paper [77]). Additionally, the LEED
pattern from this surface was observed to switch from (3x3) to (1x1). The (1x1) pattern was
composed of broad dots suggestive of increased surface disorder. This etching phenomena
was observed to occur throughout the temperature range investigated (25-800°C). At this
point, we also note that the pph Si/C ratio obtained in AES here was very similar to the value
152
of 0.47 obtained by Allendorf et al [47] from polycrystalline 3C-SiC films exposed to
thermally generated atomic H.
As hydrogen can not be detected by either AES or XPS, TPD was used to confirm
that hydrogen adsorbed or desorbed from the SiC surface. Figure 6.3 shows a TPD spectrum
obtained from the 450°C rf plasma treated SiC surface. In our case, we were only able to
observe a weak desorption feature centered around 600°C. Beyond this feature, the intensity
of the H2 signal was observed to gradually increase to the endpoint of our temperature data
acquisition hardware which is 950°C. This gradual hydrogen desorption at higher
temperatures could be related to the hydrogen desorption observed from 450-950°C by
Allendorf et al [47]. However, it could also be due to desorption of subsurface hydrogen or
outgassing from our heater. Allendorf et al [47] have previously noted desorption/outgassing
of subsurface hydrogen at 1000°C from their polycrystalline 3C-SiC films due to residual
hydrogen trapped in the SiC during CVD growth of the SiC film. However in their case, the
outgassing feature was observed to be quite abrupt and intense whereas we observed a
gradual rise in the H2 signal. Additionally, we typically observe higher levels of outgassing
of subsurface hydrogen from rf plasma treated silicon wafers at much lower temperatures
(data not shown). This could however be related to the observation of Keroak and Terreault
[65] that implanted deuterium desorbs from silicon at relatively lower temperatures than SiC
(1200 vs. 2700°C).
To test outgassing of our heater as a source of hydrogen in TPD experiments on
plasma treated SiC, we acquired a TPD spectrum of the heater just prior to the SiC TPD
experiments (see Figure 6.4(a)). As can be seen, an essentially featureless spectra was
obtained with the detected H2 signal being two orders of magnitude lower than that from the
153
plasma treated SiC. To test outgassing of the moly sample holder, we additionally performed
TPD on a 1" diameter molybdenum plate exposed to the same plasma conditions as the SiC
wafer (see Figure 6.4(b)). In this case, we observed a large and broad desorption spectrum
centered at ≈ 550°C which is typical for surface and subsurface desorption.
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
SiC
Figure 6.1. AES of (3x3) reconstructed (0001)Si 6H-SiC (a) before remote H plasma, and (b) after remote H plasma (1 min., 20 W, 15 mTorr, and 450°C)
154
96 98 100 102 104 106
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Si-Si
Si-C
(a)
(b)
Figure 6.2. XPS of the Si 2p core level from (3x3) reconstructed (0001)Si 6H-SiC (a) before remote H plasma, and (b) after remote H plasma (1 min., 20 W, 15 mTorr, and 450°C)
300 400 500 600 700 800 900
a.m
.u #
2 (a
rb. u
nits
)
Temperature (ÞC)
Figure 6.3. TPD of (1x1) 6H-SiC (0001)Si after remote H plasma exposure (1 min., 20 W, 15 mTorr, and 450°C), (ß = 1°C/sec.)
155
200 300 400 500 600 700 800 900
a.m
.u. #
2 (a
rb. u
nits
)
Temperature (ÞC)
(b)
(a)
Figure 6.4. TPD of (a) sample heating stage after outgassing, and (b) molybdenum plate after remote H plasma exposure (1 min., 20 W, 15 mTorr, and 450°C), (ß=1°C/sec.)
Finally, XPS of the C 1s core level before and after TPD revealed the presence of a
second C 1s peak after TPD (see Figure 6.5). However after TPD, the SiC surface still
displayed a (1x1) LEED pattern. Before TPD, the XPS showed only one C 1s peak. After
TPD, a second peak at higher binding energy appeared and which is indicative of some C-C
bonding at the surface. In the TPD experiments, the SiC wafer dwells at the maximum
temperature (≈1000°C) for less than a minute. In our experience, the time at this temperature
is not sufficiently long enough to result in the volatilization of a enough silicon to produce
this much C-C bonding at the surface alone. Therefore, the appearance of some C-C bonding
must be related to the loss of hydrogen from the SiC surface.
156
280 282 284 286 288
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
Figure 6.5. XPS of the C1s core level from (0001)Si 6H-SiC (a) before remote H plasma, (b) after remote H plasma (1 min., 20 W, 15 mTorr, and 450°C), and (c) after annealing at 1000°C.
6.4.2. Interaction of Thermal Atomic H with (3x3) vicinal (0001)Si 6H-SiC surfaces
In order to separate out subsurface hydrogen outgassing and other plasma related
induced effects, atomic H generated via cracking H2 over a hot rhenium filament was also
used as a source of atomic hydrogen. For comparison purposes, TPD spectra were first
acquired from (7x7) Si (111) surfaces exposed to atomic hydrogen generated by the hot
filament. In these experiments, it was observed that the surface switched to (1x1) after the
atomic H exposure. Figure 6.6 shows a TPD spectrum acquired from the hydrogen
terminated (1x1) Si (111) surface. As can be seen, a sharp desorption peak centered at ≈
475°C typical of monohydride (ß1) desorption from silicon was observed [13,15-18]. The
157
shoulder at lower energies was assigned to dihydride and trihydride (ß2,3) desorption from
silicon [13,15-18]. The ability to detect these features demonstrates the ability of the hot
filament to produce atomic hydrogen as well as validates our TPD apparatus.
200 300 400 500 600
a.m
.u #
2 (a
rb. u
nits
)
Temperature (ÞC)
Figure 6.6. TPD of Si (111) after room temperature exposure to 2000£ H2 with rhenium filament at > 1700°C (ß=1°C/sec.)
Figure 6.7 shows an AES spectrum of a (3x3) 6H-SiC (0001)Si surface before and
after exposure to atomic H (2000 Langmiur H2) from the hot rhenium filament. In this case,
the hot filament H exposure was observed to still maintain a Si/C ratio of > 1 but the ratio did
decrease from 1.35 to 0.9 and the LEED pattern was observed to switch from (3x3) to (1x1)
(however, in this case the LEED pattern displayed sharp dots). This observation was
supported by a similar reduction in the intensity of the Si-Si Si 2p peak in XPS. Unlike the 158
(111) Si surface however, a sharp desorption feature was not observed from the SiC surface
exposed to atomic H (see Figure 6.8) with only broad H2 desorption in the range of 350-
650°C being detected. In this case, the SiC TPD spectra more closely resembled the TPD
spectra of Kim at al [78] obtained from (2x1) B/Si (001) surfaces. Additionally, XPS did not
detect any C-C bonding at the SiC surface after TPD of the SiC surfaces treated with atomic
H from the hot filament.
30 130 230 330 430 530 630 730
(a)
(b)
dN(E
)/dE
Electron Energy (eV)
Si
C
Figure 6.7. AES of (3x3) 6H-SiC (0001)Si (a) before atomic H exposure and (b) after room temperature exposure to 2000£ H2 with hot filament at > 1700°C.
159
100 200 300 400 500 600 700 800
a.m
.u. #
2 (a
rb. u
nits
)
Temperature (ÞC)
Figure 6.8. TPD of (0001)Si 6H-SiC after room temperature exposure to 2000£ H2 with rhenium filament at > 1700°C (ß=1°C/sec.)
As a final check, TPD spectra were also acquired from chemically vapor/silane
cleaned (3x3) SiC surfaces which had been cooled in the silane to ≈ 300°C (see Figure 6.9).
The samples were cooled in SiH4 to 300°C to hopefully maintain some hydrogen termination
of the SiC surface without preferentially losing silicon from the surface. In this case, more
pronounced hydrogen desorption at 475-525°C was observed and which was more
comparable to hydrogen desorption from the Si (111) TPD spectra. However, the intensity of
this hydrogen desorption feature was not nearly intense or sharp as that observed from the Si
(111) wafer. This could be related, though, to the inability of the silane flux to maintain a
hydrogen terminated surface at low temperatures.
160
300 400 500 600 700
a.m
.u. #
2 (a
rb. u
nits
)
Temperature (ÞC)
Figure 6.9. TPD of (3x3) 6H-SiC (0001)Si after cooling to 300°C in 10-6 Torr SiH4 (ß=1°C/sec.)
6.5. Discussion
In order to gain a better understanding of the observed etching of SiC surfaces
presented above, the authors feel that it is necessary to first separately consider the etching of
the two elemental components of SiC (i.e. silicon and diamond). Previous studies have
reported etching of silicon [14,24,28], diamond [32,34,37,39], and amorphous silicon carbon
alloys [82] by atomic hydrogen produced either via plasma excitation or thermal
decomposition. In the case of silicon, atomic hydrogen etching has been associated with the
formation of di and trihydrides (SiH2(a) and SiH3(a)) on the silicon surface. These species
have been shown to be the precursors for the final etch product, silane (SiH4(g)) [14,24].
However, SiH2(a) and SiH3(a) species were only observed to form at low temperatures (0-
161
200°C) where hydrogen desorption is negligible. At higher temperatures (>350°C) were
hydrogen desorption from di and trihydride sites is more appreciable, the silicon surface was
observed to be terminated by mostly monohydrides (SiH(a)) and the yield of etch products
such as SiH4(g) and Si2H2(g) decreased. Correspondingly, studies on H plasma cleaning of
silicon have observed etching/roughening of silicon surfaces at temperatures < 450°C and no
surface roughening at temperatures > 450°C [28].
In the case of diamond, etch products such as CHx and C2Hx have not been observed
from (001) or (111) single crystal diamond surfaces exposed to atomic H. TPD experiments
on single crystal (001) [32,34] and (111) [38] diamond surfaces have only shown
recombinative desorption of hydrogen at ≈ 1000°C. Etch precursors such as CHx or C2Hx at
≈ 425°C have only been observed via TPD from (001) and (111) oriented polycrystalline
CVD diamond films exposed to atomic H [39]. However, Kuttel et al [37] clearly have
observed etching of (001) and (111) single crystal diamond surfaces at temperatures up to
870°C in a microwave H plasma via a measured decrease in RMS surface roughness from 7
nm to 1 nm after plasma processing. The long etch time (17 hr.) required by Kuttel to
produce these results indicates that the formation of CHx species on diamond surfaces by
atomic H is perhaps slow or inefficient.
The simple observation that H plasma etching of diamond surfaces occurs at higher
temperatures than that observed in the case of silicon suggests that perhaps the atomic H etch
rates for silicon and carbon in SiC will be different. In fact, this has already been observed
for amorphous SiC films. Using XPS and glancing incidence XRD, Kalomiros et al [82]
have shown preferential etching of silicon from an a-SiC surface and the formation of a
polycrystalline hydrogenated carbon layer by a rf plasma process at 230°C. This behavior is
162
exactly what should be predicted based on analogy to the observed etching characteristics of
diamond and silicon. This behavior is also consistent with our observation of the selective
removal of excess silicon from (3x3) (0001)Si 6H-SiC surfaces by atomic H generated both
by plasma excitation and thermal decomposition.
A second observation to be made based on analogy to silicon and diamond is that
atomic H etching of surfaces generally occurs at low temperatures where hydrogen
desorption is low and a fully hydrogenated surface can be maintained. Therefore in order to
better understand the etching characteristics of SiC surfaces in atomic H, we have attempted
to estimate the hydrogen surface coverage of silicon carbide surfaces in an atomic H flux. To
do this, we have used a simple model in which we consider hydrogen adsorption and
desorption from silicon and carbon sites separately. The adsorption/desorption processes at
silicon and carbon sites are modeled using published data for the kinetics of these processes
on (001) silicon [23,31] and (001) diamond surfaces respectively [32,34]. Although the
experimental data presented in this study is for (111)/(0001) oriented surfaces, the choice to
use kinetic parameters from (001) orientations is primarily for sake of consistency as kinetic
parameters for hydrogen desorption from (111) diamond surfaces are not available. The
values for Edes and ν used in these calculations are presented in Table 6.1.
163
Table 6.1. Kinetic Parameters used to model hydrogen adsorption/desorption from Si and C sites on SiC surfaces. (Note: all data the data shown below are for first order processes).
Si C ß1 ß2,3 Hamza [32] Thomas [34] Edes (kcal/mol) 47 [23] 25 [31] 37 72.7
ν (#/sec) 8x1011 107 3x105 1013
For the carbon sites, hydrogen desorption was modeled using the kinetic parameters
determined for hydrogen desorption from (001) diamond surfaces by both Hamza et al [32]
and Thomas et al [34]. This is primarily a result of the extremely low value of 37 kcal/mol
for Edes reported by Hamza et al [32] which is lower than that reported for silicon. This
value for Edes is extremely surprising given the simple fact that hydrogen desorption from
diamond surfaces occurs at temperatures 500°C higher than on silicon. Therefore, the data of
Thomas et al [34] was used as well since in this case Edes = 72.7 kcal/mol which is higher
than most values reported for H2 Edes from silicon surfaces. Additionally, Allendorf et al
[47] have also reported Edes = 72 kcal/mol and ν = 1013/sec for hydrogen desorption from
polycrystalline SiC surfaces.
To estimate the SiC hydrogen surface coverage, we employ the method of Schulberg
et al [94] in which we assume a steady state equilibrium between the incoming flux of atomic
H and surface desorption of H2. Desorption kinetics are typically described by the general
Polyani-Wigner rate expression [31]:
desorption rate = -dØ/dt = νn exp(-Edes/RT) (1)
164
where: n = the reaction order Ø = the adsorbate surface coverage ν = the pre-exponential factor, νo = 1028/cm2 sec,
ν1 = 1013/sec, ν2 = 10-2 cm2/sec Edes = the activation energy for desorption
In steady conditions, the flux of adsorbates leaving the surface via desorption will be
equal to the incoming flux of adsorbate times the adsorbate sticking coefficient. The sticking
coefficient is described by:
S = So(1 - Ø/Ømax)n (2) where S = Sticking Probability So = Initial Sticking Probability (i.e. S at Ø = 0) Ømax = Maximum Surface Coverage
The combination of equations 1 and 2 allows the determination of the steady state surface
coverage of the adsorbate [94]. Based on this model and the desorption kinetic data tabulated
above, we have therefore estimated the surface coverage of hydrogen on silicon and carbon
sites both as a function of temperature and flux. The results are presented in Figures 6.10-
6.13 and are based on the assumption of a unity initial sticking coefficient (i.e. So = 1)
Figures 6.10 and 6.11 show the estimated monohydride and di/trihydride surface
coverages for silicon sites as a function of temperature and flux (ML/sec). Figure 6.10
indicates that in fluxes typical of low pressure hydrogen plasma processes, all silicon sites
should be saturated with hydrogen up to temperatures of ≈ 800°C. However, Figure 6.11
indicates that the concentration of di/trihydrides should start to decrease around 500-600°C.
165
Although the temperature range is slightly higher, this is clearly consistent with the
previously reported etching behavior of silicon in a remote H plasma. Accordingly, etching
of silicon in SiC should stop or decrease at temperatures above ≈ 600°C. However, as shown
in Figures 6.12 and 6.13, carbon sites on SiC surfaces should remain saturated with hydrogen
up to temperatures of 1000-1100°C indicating that etching of carbon in SiC could continue
up to temperatures these very same temperatures. We also note the significant difference in
predicted concentration of occupied carbon sites based on the data of Hamza et al [32] and
Thomas et al [34]. Based on the data of Hamza et al [32], carbon sites would be expected to
be saturated with hydrogen up to 1200°C in a 100 ML/sec H flux. However, the data
Thomas et al would predict that at least of half of these sites would be empty under the same
conditions.
0
0.2
0.4
0.6
0.8
1
200 400 600 800 1000 1200 1400 1600
Ø/Ø
max
Temperature (ÞC)
1 100 10 4
10 6
Figure 6.10. Mono-hydride surface coverage on silicon sites of SiC.
166
0
0.2
0.4
0.6
0.8
1
200 400 600 800 1000 1200 1400 1600
Ø/Ø
max
Temperature (ÞC)
1100
10 4
Figure 6.11. Di-hydride surface coverage on silicon sites of SiC.
0
0.2
0.4
0.6
0.8
1
200 400 600 800 1000 1200 1400 1600
Ø/Ø
max
Temperature (ÞC)
1
10
100
Figure 6.12. Hydrogen surface coverage on carbon sites of SiC based on kinetic data of Hamza [32].
167
0
0.2
0.4
0.6
0.8
1
200 400 600 800 1000 1200 1400 1600
Ø/Ø
max
Temperature (ÞC)
1
100
10 4
10 6
Figure 6.13. Hydrogen surface coverage on carbon sites of SiC based on kinetic data of Thomas et al [34].
Based on analogy to silicon and Figures 6.10-6.13, one would expect etching of SiC
in atomic H at temperatures < 1000°C and no etching at temperatures ≈ > 1000°C. In reality,
SiC is probably etched by atomic H at all temperatures. However, the etch rate probably
exhibits some temperature dependence. At low temperatures (i.e. RT-500°C), the etch rate is
probably low due to limited thermal activation and perhaps surface mobility. As the
temperature is increased into the range of 500-1000°C, the etch rate should decrease due to
significant desorption of hydrogen through silicon sites and the reduction of etch precursors
such as SiH3(a). However, the etch rate in this temperature range will probably still be
measurable due to limited desorption from carbon sites and hence the ability to form CHx
etch products. At temperatures > 1000°C, the etch rate should decrease due to increased
desorption of hydrogen from both carbon and silicon sites but may eventually increases due
to increased thermal activation and possible volatilization of silicon at temperatures >1500°C.
168
This type of etch rate dependence has been partially observed by Kim and Olander [48] in
their modulated molecular beam mass spectrometry studies. In these studies, they observed
SiH4, CH4, and C2H2 etch products from polycrystalline 3C-SiC films during atomic H
exposure over the temperature range of RT-800°C. However, the yield of SiH4 from SiC
was observed to initially increase with temperature and then start to decrease at ≈ 500°C.
The yield of CH4, however, was observed to gradually increase over the entire temperature
range investigated (0-800°C). Though Kim and Olander [48] explain the decrease in SiH4
production to depletion of silicon from the surface, their results are also clearly consistent
with our explanation based on enhanced hydrogen desorption from silicon sites. However,
our inability to avoid selective removal of silicon from the (3x3) (0001)Si 6H-SiC surface in
the rf H plasma over the temperature range of 0-800°C is also consistent with the results of
Kim and Olander [48].
In the higher temperature range (1000-1700°C), it has been previously noted that the
etch rate of SiC in molecular hydrogen (i.e. H2) increases with increasing temperature. In
this case, we feel that the actual etching of SiC is due to atomic H produced by thermal
decomposition of molecular H2. Figure 6.14 shows the predicted percent dissociation of
molecular hydrogen into atomic hydrogen based on thermodynamic calculations (note figure
produced by HSC program). As can be seen, significant production of atomic H is only
predicted to occur at temperatures of 1500-1700°C. So these higher temperatures for
molecular H2 etching are at least to some extent probably necessary to produce a significant
concentration of atomic hydrogen. The higher temperatures probably also assist in the
volatilization of silicon from the surface as well as increasing the surface mobility.
169
0.0
5.0
10
15
1000 1500 2000 2500 3000
% H
2 Dis
soci
atio
n
Temperature (ÞC)
Figure 6.14. Percent dissociation of H2 into H as function of temperature.
At this point, it is worth considering some of the potential errors in this simplified
model. First for this model to be valid, diffusion of hydrogen from carbon to silicon sites
(and vice versa) must be minimal. For Si-Ge alloys this subject has sparked much debate. In
this system, a lowering of Tmax for monohydride (ß1) desorption from silicon was observed
with the addition of germanium to the surface [83,84]. This lowering of Tmax has been
argued to be due to weakening of the Si-H bond due to electronic matrix effects [83]. Others,
however, have argued that the lowering of ß1 Tmax can be described by considering
hydrogen diffusion from silicon sites to germanium sites where the activation energy for
desorption is lower (Edes Si-H = 55 kcal/mol vs. Ge-H = 35 kcal/mol) [84,85]. This is
reasonable for silicon surfaces where the activation energies for hydrogen diffusion on (100)
and (111) surfaces have been determined to be ≈ 30-41 kcal/mol [86,87] and 35 kcal/mole
170
[88] respectively. Unfortunately for the case of SiC, activation energies for hydrogen surface
diffusion have not been determined. However, activation energies for surface diffusion of Si
and C atoms on SiC surfaces have been reported. Based on surface diffusion length
measurements in CVD, Kimoto et al [89,90] determined an activation energy of 82 kcal/mol
for surface diffusion of Si or C atoms on (0001)Si 6H-SiC surfaces. These measured values
are in agreement with the calculations of Takai et al [91], which found activation barriers of
106-126 kcal/mol for carbon diffusion on (111) and (-111) 3C-SiC surfaces. As much lower
activation energies for self diffusion of Si on Si (111) [92,93] and (001) [93] surfaces have
been determined (18-35 and 6-25 kcal/mole respectively) it seems reasonable to expect the
mobility of H atoms on SiC surfaces to be lower than on Si surfaces. Accordingly, H surface
diffusion should be negligible in the temperature range studied here and should not affect our
model.
Another possible source of error in our model is that Tmax for hydrogen desorption
from Si and C sites on SiC could be significantly different from that observed from silicon
and diamond respectively. As previously mentioned, it has been suggested that the addition
of Ge to Si lowers Tmax for ß1 H2 desorption from Si sites due to a weakening of the Si-H
bond by germanium [83]. It has additionally been observed that boron doping (1019/cm3) of
Si also lowers Tmax ß1 whereas dosing Si (001) surfaces with diborane (B2H6) produces
two ß1 desorption features [78]. These effects have been explained based on varying
electronic effects in which the Si-H bond is weakened by the more electronegative dopant or
due to changes in the work function by the dopant or substitutional atom. Clearly, the
addition of carbon to the silicon lattice (and vice versa) could have similar effects. The
addition of carbon changes both the work function (and band gap) as well as creating a more
171
polar bond with silicon to weaken any Si-H bonds at the surface. This line of reasoning
could explain our inability to observe sharp desorption features from atomic H treated (3x3)
(0001)Si 6H-SiC surfaces.
In the case of (3x3) SiC surfaces exposed and cooled in SiH4 to 300°C, H2
desorption at 475°C was observed and which was consistent with the observed H2 desorption
from Si (111) surfaces. This can be explained by the fact that for this surface all of the
silicon atoms terminating the SiC surface have Si-Si backbonds as the (3x3) surface consists
of a bilayer of silicon atoms. However, in the case of the hot filament atomic H treated (3x3)
SiC surfaces, a broad range of H2 desorption in the temperature range of 200-600°C was
observed and which was similar in appearance to that observed from B2H6 treated Si (001)
surfaces [78]. In this case, some of the Si bilayer has been removed via etching by atomic H
and hence desorption occurs from Si atoms with Si-Si and Si-C backbonds (i.e. C-Si-H or Si-
Si-H). For the silicon atoms with carbon backbonds, the Si-H bond is weakened as the
underlying carbon atom is more electronegative and withdraws charge which the silicon atom
would share with hydrogen. Accordingly, the weaker Si-H bonds translates into a lower
Edes. This range of Edes is what gives Figure 6.8 its broad nature. However, we note that in
the studies of Ascherl et al [95] dosing silicon surfaces with trimethylsilane (SiH(CH3)3) did
not result in any change in Tmax for ß1.
In the case of the H plasma treated (0001)Si 6H-SiC surfaces, TPD showed a weak
H2 desorption feature at ≈ 625°C with a gradual increase in H2 signal at higher temperatures.
This feature is intermediate to what would be expected from carbon and silicon sites based on
analogy to silicon or diamond surfaces. This feature could be related to H2 desorption from
172
CHx sites, but is higher in temperature than previous reports of CHx desorption from
polycrystalline CVD diamond films [39]. This TPD spectrum, however, could be greatly
effected surface roughening induced by the atomic H etching of the SiC surface. This could
produce greatly affect Edes.
Setting aside our simple desorption model, etching of SiC by atomic hydrogen can
also be predicted based on simple comparison of the Si-C, Si-H, and C-H bond energies
which are 76.5, 70.8, and 98.8 kcal/mol respectively [97]. As can be seen, the C-H bond is
actually stronger than the Si-C bond and energy can be gained by breaking a Si-C bond and
forming a C-H bond. Although, the reaction between molecular hydrogen (H2) and SiC to
form Si-H and C-H bonds is not energetically favorable, i.e.
Si-C + H-H <=> Si-H + C-H (i) 76.5+104 <===> 70.4 + 98.8 179.7 <====> 169.2 kcal/mol
The reaction between pre dissociated atomic H and SiC is, i.e.
Si-C + 2H <=====> Si-H + C-H (ii)
76.5 <====> 70.4 + 98.8
Thus for the (0001)Si 6H-SiC surface, it is more energetically favorable for atomic
hydrogen to insert itself into a Si-C bond forming C-H bonds rather than simply terminating a
Si dangling bond and forming a relatively weak Si-H bond. Thus a C-H terminated SiC
surface would be expected after atomic hydrogen exposure. This is in agreement with the < 1
Si/C pph ratio we and other have observed from SiC surfaces exposed to atomic H. This also
agrees with our observation of the formation of some C-C bonding in XPS after TPD at
1000°C. At temperatures of 1000°C it would be expected that any hydrogen adsorbed on
173
carbon would desorb leaving behind some C-C bonding. Additionally, we have
previously noted that in HF wet chemical processing H termination of silicon atoms at
(0001)Si 6H-SiC surfaces is highly unstable due to the polarity of the underlying Si-C bonds
[96]. OH- termination is instead favored due to the ability of the Si-OH bond to cancel the
dipole produced by the Si-C bond below. For these same reasons, Si-H termination of
(0001)Si 6H-SiC surfaces in vacuum are equally unstable and hence there is an additional
driving force for the H atom to insert itself into the Si-C bond. It is also worth mentioning
that in the (3x3) structure there are actually a significant number of Si-Si bonds. We note
that in this case, these bonds are compressed and distorted to values far from the their
equilibrium value due to the differences in the lattice constants between Si and SiC (≈ 20 %
[10,70-72]). As such, these bonds are probably more reactive with atomic hydrogen and it
may be more energetically favorable to form SiHx species rather than to remain in the
distorted (3x3) structure.
Finally, the above results and discussion indicate that in H plasma cleaning of SiC,
silicon should be added to the plasma chemistry in order to compensate for the selective
removal of silicon from the SiC surface due to etching. The addition of silicon to the plasma
(via SiH4 or Si2H6) should also assist in the reduction of silicon oxides which are
particularly difficult to remove in H plasmas. Lin et al [51] have previously investigated
cleaning of HF dipped SiC surfaces using a 1:1 H2:He mixture in an ECR plasma source
(650°C, 90 min., 5x10-4 Torr). In their case, they reported the removal of C-C, C-O, and C-
F species from the SiC surface but were not successful in completely removing silicon oxides
(Si-O) from the surface. The addition of silicon to the plasma chemistry therefore could
assist in the removal of silicon oxides through chemical reduction and formation of more
174
volatile sub oxides. Initial investigations in our lab using H2/1%SiH4 mixtures in plasma
cleaning of HF dipped SiC surfaces have shown an enhanced removal of silicon oxides.
However, deposition of silicon was a particular problem indicating that such a cleaning
process requires delicately balancing the silicon deposition rate with its etching rate. This
will require tight process control and perhaps only a very narrow processing window will be
available.
6.6. Conclusion
In conclusion, we have shown that atomic hydrogen exposure selectively removes
silicon from (3x3) 6H-SiC (0001)Si surfaces. Atomic hydrogen exposures reduces and
removes the Si-Si bonding Si 2p XPS peak and converts the (3x3) LEED pattern to (1x1).
Additional etching of the SiC surface was indicated by the reduction in the Si LVV/C KLL
ratio in AES from 1.3 to 0.4 following exposure of (3x3) surfaces to a remote rf H plasma.
TPD of atomic H treated (3x3) SiC surfaces showed weak hydrogen desorption in the range
of 400-600°C where desorption from silicon atoms would be expected by analogy to (111)
Si. However, the hydrogen desorption signal increased at higher temperatures where
hydrogen desorption from carbon sites would be expected based on analogy to (111)
diamond surfaces. C-H termination of the SiC surface was supported by the observation of
some C-C bonding after thermal desorption of rf plasma treated SiC surface at T > 1000°C.
Based on these observations, we conclude that atomic H processing of SiC surfaces
selectively removes silicon from the surface and favors C-H termination.
175
6.7 Acknowledgments
The authors would like to thank Cree Research, Inc. for supplying the 6H-SiC wafers.
The research was supported by the Office of Naval Research under contract and through the
Department of Education through an Electronic Materials/GAANN Fellowship.
6.8 References: Examples of H2 in Si Semiconductor Processing 1. S.S. Iyer, M. Arienzo, and E. de Fresart, Appl. Phys. Lett, 57, 893 (1990). 2. H.J. Stein, Appl. Phys. Lett., 32, 379 (1978). 3. K.G. Drujif, J.M.M. de Nijs, E. van der Drift, E.H.A. Granneman, and P. Balk, Appl. Phys. Lett., 67, 3162 (1995). 4. G.R. Srinivasan, J. Cryst. Growth, 70, 201 (1984). 5. H. Habuka, J. Tsunoda, M. Mayusumi, N. Tate, and M. Katayama, J. Electrochem. Soc., 142, 3092 (1995). H2 in SiC Processing 6. N. Nordell, A. Schoner, and S.G. Andersson, J. Electrochem. Soc., 143, 2910 (1996). 7. T. Kimoto, H. Nishino, W.S. Yoo, and H. Matsunami, J. Appl. Phys., 73, 726 (1993). 8. T. Kimoto and H. Matsunami, J. Appl. Phys., 76, 7322 (1994). 9. D.J. Larkin, P.G. Neudeck, J.A. Powell, and L.G. Matus, Appl. Phys. Lett., 65, 1659 (1994). 10. H.S. Kong, J.T. Glass, and R.F. Davis, J. Appl. Phys., 64, 2672 (1988). 11. P. Liaw and R.F. Davis, J. Electrochem. Soc., 132, 642 (1985). 12. J.A. Powell, L. G. Matus, and M.A. Kuczmarski, J. Electrochem. Soc., 134, 1558 (1987). H2 and Silicon
176
13. G. Schulze and M. Henzler, Surface Science, 124, 336 (1983). 14. L.H. Chua, R.B. Jackman, and J.S. Foord, Surface Science, 315, 69 (1994). 15. S.M. Gates and S.K. Kulkarni, Appl. Phys. Lett., 60, 53 (1992). 16. B.G. Koehler, C.H. Mak, D.A. Arthur, P.A. Coon, and S.M. George, J. Chem. Phys., 89, 1709 (1988). 17. C.C. Cheng, S.R. Lucas, H. Gutleben, W.J. Choyke, and J.T. Yates, J. Am. Chem. Soc., 114, 1249 (1992). 18. S.M. Gates, R.R. Kunz, and C.M. Greenlief, Surface Science, 207, 364 (1989). 19. M. Liehr, C.M. Greenlief, M. Offenberg, and S.R. Kasi, J. Vac. Sci. Technol. A, 8, 2960 (1990). 20. J.J. Boland, Phys. Rev. Lett., 65, 3325 (1990). 21. R.M. Wallace, P.A. Taylor, W.J. Choyke, and J.T. Yates, Surface Science, 239, 1 (1990). 22. H. Kobayashi, K. Edamoto, M. Onchi, and M. Nishijima, J. Chem. Phys., 78, 7429 (1983). 23. K. Sinniah, M.G. Sherman, L.B. Lewis, W.H. Weinberg, J.T. Yates, and K.C. Janda, J. Chem. Phys., 92, 5700 (1990). 24. C.M. Greenlief, S.M. Gates, and P.A. Holbert, Chem. Phys. Lett., 159, 202 (1989). 25. M.K. Farnaam and D.R. Olander, Surface Science, 145, 390 (1984). 26. D. Ludden, R. Tsu, T.R. Bramblett, and J.E. Greene, J. Vac. Sci. Technol. A, 9, 3003 (1991). 27. S.M. Gates, C.M. Greenlief, and D.B. Beach, J. Chem. Phys., 93, 7493 (1990). 28. J.S. Montgomery, T.P. Schneider, R.J. Carter, J.P. Barnak, Y.L. Chen, J.R. Hauser, and R.J. Nemanich, Appl. Phys. Lett., 67, 2194 (1995). 29. K. Nakashima, M. Ishii, I. Tajima, and M. Yamamoto, Appl. Phys. Lett., 58, 2663 (1991). 30. Y. Morita and H. Tokumoto, Appl. Phys. Lett., 67, 2654 (1995). 31. C. Kleint and K.D. Brzoska, Surface Science, 231, 177 (1990). Hydrogen and Diamond
177
32. A.V. Hamza, G.D. Kubiak, and R.H. Stulen, Surface Science, 237, 35 (1990). 33. V.S. Smentkowski, H. Jansch, M.A. Henderson, and J.T. Yates, Surface Science, 330, 207 (1995). 34. R.E. Thomas, R.A. Rudder, and R.J. Markunas, J. Vac. Sci. Technol. A, 10, 2451 (1992). 35. B. Pate, Surface Science, 165, 83 (1986). 36. S. Matsumoto, Y. Sato, and N. Setaka, Carbon, 19, 234, (1981). 37. O.M. Kuttel, L. Diederich, E. Schaller, O. Carnal, and L. Schlapbach, Surface Science, 337, L812 (1995). 38. Y. Mitsuda, T. Yamada, T.J. Chuang, H. Seki, R.P. Chin, J.Y. Huang, and Y.R. Shen, Surface Science Letters, 257, L633 (1991). 39. L.H. Chua, R.B. Jackman, J.S. Foord, P.R. Chalker, C. Johnston, and S. Romani, J. Vac. Sci. Technol. A, 12, 3033 (1994). 40. T. Ando, M. Ishii, M. Kamo, and Y. Sato, J. Chem. Soc. Faraday Trans., 89, 1783 (1993). 41. J. van der Weide and R.J. Nemanich, Appl. Phys. Lett., 62, 1878 (1993). Hydrogen and GaAs and InP 42. G.V. Jagannathan, M.L. Andrews, and A.T. Habig, Appl. Phys. Lett., 56, 2019 (1990). 43. I. Suemune, Y. Kunitsugu, Y. Tanaka, Y. Kan, and M. Yamanishi, Appl. Phys. Lett., 53, 2173 (1988). 44. M. Yamada and Y. Ide, Jpn. J. Appl. Phys., 33, L671 (1994). 45. C.M. Rouleau and R.M. Park, J. Appl. Phys., 73, 4610 (1993). 46. Y. Sakamoto, T. Sugino, H. Ninomiya, K. Matsuda, and J. Shirafuji, Jpn. J. Appl. Phys., 34, 1417 (1995). Hydrogen and SiC 47. M.D. Allendorf and D.A. Outka, Surface Science, 258, 177 (1991). 48. Y. Kim and D.R. Olander, Surface Science, 313, 399 (1994). 49. J.M. Lannon, J.S. Gold, and C.D. Stinespring, 77, 3823 (1995).
178
50. A.O. Konstantinov, N.S. Konstantinova, O.I. Kon'kov, E.I. Terukov, and P.A. Ivanov, Semiconductors, 28, 209 (1994). 51. M.E. Lin, S. Strite, A. Agarwal, A. Salvador, G.L. Zhou, N. Teraguchi, A. Rockett, and H. Morkoc, Appl. Phys. Lett., 62, 702 (1993). SiC/III-N Devices 52. R.F. Davis, Advances in Ceramics, 23, 477 (1987). 53. R.F. Davis, G. Kelner, M. Shur, J. Palmour, J.A. Edmond, Proc. of the IEEE 79, 677 (1991). 54. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10, 1237 (1992). H2 Thermal Etching 55. T.L. Chu and R.B. Campbell, J. Electrochem. Soc., 112, 955 (1965). 56. J.M. Harris, H.C. Gatos, and A.F. Witt, J. Electrochem. Soc., 116, 380 (1969). 57. R.W. Bartlett and R.A. Mueller, Mat. Res. Bull., 4, S341 (1969). 58. D. Kim and D. Choi, J. Am. Ceram. Soc., 79, 503 (1996). H and H2 Ion Implantation/Sputtering 59. J. Bohdansky and J. Roth, J. Nucl. Mater., 122/123, 1417 (1984). 60. U.R. Ajerk, J. Irr. Results, 1, 23 (1997). 61. K. Sone, M. Saidoh, K. Nakamura, R. Yamada, Y. Muragami, T. Shikama, M. Fukutomi, M. Kitajima, and M. Okada, J. Nucl. Mater., 98, 270 (1981). 62. M. Mohri, K. Watanabe, and T. Yamashina, J. Nucl. Mater., 75, 7 (1978). 63. T. Yamashina, M. Mohri, K. Watanabe, H. Doi, and K. Hayakawa, J. Nucl. Mater., 76/77, 202 (1978). 64. J. Bohdansky, H.L. Bay and W. Ottenberg, J. Nucl. Mater., 76/77, 163 (1978). 65. D. Keroack and B. Terreault, J. Vac. Sci. Technol. A, 14, 3130 (1996). (3x3) 6H-SiC (0001)Si Surfaces and Characterization 66. R. Kaplan and T.M Parrill, Surface Science, 165, L45 (1986). 67. R. Kaplan, Surface Science, 215, 111 (1989). 68. V.M. Bermudez, Appl. Surf. Sci., 84, 45 (1995).
179
69. S. Nakanishi, H. Tokutaka, K. Nishimori, S. Kishida, and N. Ishihara, Appl. Surf. Sci., 41/42, 44 (1989). 70. V. van Elsbergen, T.U. Kampen, and W. Monch, Surface Science, 365, 443 (1996). 71. L. Li and I.S.T. Tsong, Surface Science, 351, 141 (1996). 72. M.A. Kulakov, G. Henn, and B. Bullemer, Surface Science, 346, 49 (1996). 73. R.S. Kern, Ph.D. dissertation, NCSU (1996). 74. A. Fissel, U. Kaiser, E. Ducke, B. Schroter, and W. Richter, J. Cryst. Growth, 154, 72 (1995). 75. A. Fissel, B. Schroter, and W. Richter, Appl. Phys. Lett., 66, 3182 (1995). 76. S.W. King, M.C. Benjamin, R.S. Kern, D. Hanser, J.P. Barnak, R.J. Nemanich, and R.F. Davis, submitted to J. Appl. Phys.. 77. M.C. Benjamin, S.W. King, J.P. Barnak, R.S. Kern, R.F. Davis, and R.J. Nemanich, submitted to J. Appl. Phys.. 78. H. Kim, G. Glass, S.Y. Park, T. Spila, N. Taylor, J.R. Abelson, and J.E. Greene, Appl. Phys. Lett., 69, 3869 (1996). Experimental Stuff 79. J. van der Weide, Ph.D. Dissertation, NCSU. 80. V.S. Smentkowski and J.T. Yates, Jr., J. Vac. Sci. Technol. A, 7 3325 (1989). 81. Perkin Elmer XPS Handbook. 82. J.A. Kalomiros, E.C. Paloura, A. Ginoudi, S. Kennou, S. Ladas, Ch. Lioutas, N. Vouroutzis, G. Voutsas, D. Girginoudi, N. Georgoulas, and A. Thanailakis, Solid State Comm., 96, 735 (1995). 83. B.M.H. Ning and J.E. Crowell, Surface Science, 295, 79 (1993). 84. N.M. Russell and J.G. Ekerdt, Surface Science, 369, 51 (1996). 85. L. Surnev and M. Tikhov, Surface Science, 138, 40 (1984). Surface Diffusion 86. A. Vittadini, A. Selloni, and M. Casarin, Surface Science, 289, L625 (1993). 87. G.A. Reider, U. Hofer, and T.F. Heinz, Phys. Rev. Lett., 66, 1994 (1991).
180
88. G. Li, Y.C. Chang, R. Tsu, and J.E. Greene, Surface Science, 330, 20 (1995). 89. B. Voigtlander and A. Zinner, Surface Science, 292, L775 (1993). 90. K.N. Tu, J.W. Mayer, L.C. Feldman, Electronic Thin Film Science for Electrical Engineers and Materials Scientists, pg 132 (1987). Macmillan Pub. Co., New York. 91. T. Kimoto and H. Matsunami, J. Appl. Phys., 78, 3132 (1995). 92. T. Kimoto and H. Matsunami, J. Appl. Phys., 75, 850 (1994). 93. T. Takai, T. Halicioglu, and W.A. Tiller, Surface Science, 164, 327 (1985). 94. M.T. Schulberg, M.D. Allendorf, and D.A. Outka, Surface Science, 341, 262 (1995). 95. M.V. Ascherl, J.H. Campbell, J. Lozano, and J.H. Craig, Jr., J. Vac. Sci. Technol. A, 13, 2721 (1995). 96. S.W. King, R.J. Nemanich, and R.F. Davis, submitted to the J. Electrochem. Soc. 97. B. Douglas, D.H. McDaniel, and J.J. Alexander, Concepts and Models of Inorganic Chemistry, p. 74, John Wiley & Sons, New York, (1983).
181
7. EX SITU AND IN SITU SURFACE CLEANING PROCESSES FOR
(0001) AlN, GaN, AND AlxGa1-xN SURFACES
To be Submitted for Consideration for Publication
to the
Journal of Applied Physics
by
Sean W. King, Laura L. Smith, John P. Barnak, *James A. Christman,
Michael D. Bremser, *Robert J. Nemanich, and Robert F. Davis
Department of Materials Science and Engineering
*Department of Physics
North Carolina State University
Raleigh, NC 27695
182
7.1 Abstract
Exposure to numerous acids and bases and UV/O3 oxidation were used to determine
the best ex situ cleaning techniques for the (0001) surfaces of AlN and GaN. HF and HCl
were found to produce surfaces with the lowest coverage of oxygen after wet chemical
cleaning of AlN and GaN, respectively. However, AES and XPS revealed the surfaces to be
terminated with F and Cl which inhibited re-oxidation prior to loading into vacuum. TPD
showed that temperatures of 650 and 850°C were necessary to thermally desorb the Cl and F,
respectively. UV/O3 oxidation was observed to reduce the surface carbon coverage for both
AlN and GaN though incompletely. This was related to the relative inert chemical nature and
resistance to oxidation of GaN and AlN surfaces. In situ remote hydrogen plasma exposure
at 450°C removed halogens and hydrocarbons remaining after ex situ cleaning of both AlN
and GaN surfaces; however, oxide free surfaces were not be achieved. Complete thermal
desorption of the surface oxide from (0001) GaN in UHV was only achieved at temperatures
> 800°C where some GaN decomposition occurred. Annealing GaN in NH3 at 800°C
reduced the surface oxide without loss of surface stoichiometry.
183
7.2. Introduction
GaN and AlN are completely miscible semiconductors with wide direct band gaps of
3.40 and 6.2 eV, respectively [1-3]. Combined with the 1.9 eV direct band gap of InN [1],
the III-V nitride materials system is accordingly of considerable interest for many visible-UV
optoelectronics device applications [4]. The recent demonstration of a blue laser diode based
on a InGaN quantum well (QW) structure [5] highlights many of the recent advances which
have been made in this field. However, GaN and AlN are also of interest for high power,
high frequency, and high temperature device applications due to the other extreme materials
properties that they exhibit including: high temperature stability (Tmelt AlN ≅ 2850°C), high
saturation electron drift velocity (vsat GaN = 2.7x107 cm/sec), high thermal conductivity (κ
AlN = 3.0 W/cm K), and high breakdown voltage (GaN = 2.0 MV/cm) [2-3]. The recent
observation of a negative electron affinity for AlN [6] and AlxGa1-xN [7] alloys also makes
these materials candidates for field emitters in cold cathode electron devices. However, the
advancement of GaN and AlN toward use in these applications will demand the continued
development of new and improved processes to fabricate these materials and devices. As
surface cleaning processes are the foundations on which most semiconductor fabrication
steps are built [1-3], it is likely that improvements in GaN and AlN processing can be further
assisted by optimization of surface cleaning processes for GaN and AlN.
Experience gained in silicon and gallium arsenide technology has shown that the
criteria for surface cleanliness must include removal of not only native oxides and organic 184
contaminants but also metallic impurities, particulate contaminants, adsorbed molecules, and
residual species left by previous processes [8-10]. The deleterious effects of incomplete
removal of all these various contaminants are numerous. In Si homoepitaxy, for example,
improper removal of surface oxides and organic contaminants have been shown to result in
an increased density of line and planar defects in epitaxial films from <104/cm2 to >
1010/cm2 [11-16]. In fact, recent studies on the heteroepitaxial growth of SixGe1-x alloys on
Si (100) have shown that surface defects produced in the Si substrate by residual
organic/carbon contamination act as the preferred sites for misfit dislocation generation [17].
Most importantly however, these increased epitaxial defect densities have been shown to
decrease device yield and performance [18-19]. Surface cleaning is also important in other
semiconductor processes including metal contact formation and gate dielectric formation. In
the case of metal contact formation, improper removal of surface oxides prior to metal
deposition has been demonstrated to result in higher contact resistances and lack of contact
parameter uniformity [20-26]. Control of the metal contact Schottky barrier height has also
been shown to be dependent on surface preparation [27-33]. In this case, control of the
Schottky barrier height is dependent on controlling the existence and density of
surface/interface states which determine the surface Fermi level position. Surface and
interface states can frequently be related to surface dangling bonds which can be passivated
by various surface cleaning processes. Control of interface states is also of importance in
MISFET devices where interface states can produce changes in threshold voltage and must
be controlled to ensure uniformity in device operation. Drift in MIS device parameters has
also been linked to the presence of alkali ions at the interface which can, in part, be
controlled by surface cleaning prior to insulator deposition [34].
185
Due to the above observations, numerous studies have been conducted on Si [8-10]
and GaAs [35-71] which have been concerned with obtaining clean, structurally well ordered
surfaces. For GaAs, these studies have covered a wide range of ex situ and in situ processes
including wet chemical [35-53], UV/O3 oxidation [54-62], thermal desorption 63,64],
chemical beam [65], and atomic H cleaning [66-71]. In the case of AlN and GaN, however,
there have been comparatively fewer investigations of processes to obtain clean surfaces of
these materials [72-88]. In fact to the authors knowledge, there have been no such studies for
AlN. Most studies of the surface properties of AlN in vacuum have relied either on ion
bombardment/sputtering [72-74] or in situ preparation of polycrystalline AlN films on Al via
nitriding with hydrazine [75] or N2+ ions [76,77].
For GaN, perhaps the first surface cleaning study was that of Hedmann and
Martensson [78] who used XPS to examine single crystal HVPE GaN/Al2O3 (0001) films
etched in H3PO4 at 100°C and then annealed in situ at 300°C. In this case, they reported that
the in situ anneal removed some oxygen and carbon contaminants but were unsuccessful in
completely removing these contaminants at this temperature. Since the early reports by
Hedmann and Martensson, there have only recently been more investigations of surface
cleaning processes for GaN. These stuides have been conducted on high quality MOCVD
and OMVPE GaN films grown on (0001) Al2O3 and 6H-SiC substrates [79-88].
The first such investigation by Khan et al [79], examined in situ cleaning of MOCVD
GaN (0001) films in MBE via annealing in an evaporated flux of Ga (2-5x1015/cm2 min) at
600 or 900°C. In situ Auger electron spectroscopy (AES) analysis of GaN films cleaned in
this fashion showed the CKLL peak intensity to be 2% of the NKLL signal while the O
contamination was close to the AES sensitivity limit. However, low energy electron
diffraction (LEED) of these surfaces displayed only unreconstructed (1x1) diffraction
186
patterns. This technique has since been used by Bermudez et al to prepare clean GaN
surfaces to study the interfaces and interaction of GaN with Ni [80], Al [81], and O2 [82]. In
the latter study, Bermudez concluded that surfaces prepared via annealing in a Ga flux show
upward bending of 0.9 eV due to surface Fermi level pinning by surface states of unidentified
character. These states were removed by re-exposure of the clean GaN surfaces to molecular
oxygen (O2) which decreased the band bending by ˜ 0.15 eV. In contrast, Bermudez also
found that surfaces prepared by only wet chemical cleaning in 1:10 NH4OH:H2O showed
only 0.4±0.2 eV upward band bending. In addition, Bermudez also investigated N2+
sputtering for cleaning GaN surfaces in situ and found the technique to be equivalent to
annealing in a beam of Ga [82]. Nitrogen ion sputtering has also been used by Sung et al
[83] to examine the polarity of GaN films grown on (0001) Al2O3. Their time-of-flight
scattering and recoiling spectrometry (TOF-SARS) and classical ion trajectory simulations
indicate that (0001) GaN surfaces prepared in this fashion are nitrogen terminated with Ga
atoms comprising the second layer. These findings, however, are contrary to those of Daudin
et al [84] who using ion channeling and convergent beam electron diffraction techniques
concluded that smooth MOCVD GaN films grown on (0001) Al2O3 are Ga terminated
whereas rough/pyramidal GaN films are N terminated.
In a previous study [85], we have investigated cleaning of GaN surfaces using HF
and HCl wet chemical processes followed by in situ thermal desorption. In this study, it was
found that HCl:DI wet chemical processes produced the lowest coverages of oxygen and
carbon contaminants but that HF wet chemistries resulted in GaN surfaces for which thermal
desorption of carbon contaminants in situ was more efficient [85]. More recent studies by
others have investigated cleaning of GaN using wet chemical treatments based on warm
187
NH4OH [86] and HF solutions buffered with (H2FNH4) [87]. In situ cleaning of GaN
surfaces using a H2:He plasma has also been demonstrated by Hughes et al [88].
In this study, we have chosen to investigate both ex situ and in situ cleaning of AlN
and GaN surfaces using numerous surface analytical techniques including Auger electron
spectroscopy (AES), x-ray photoelectron spectroscopy (XPS), ultra-violet photoelectron
spectroscopy (UPS), low energy electron diffraction (LEED), and temperature programmed
desorption (TPD). A broad range of chemistries and techniques for cleaning both GaN and
AlN surfaces was selected in order to better facilitate a direct comparison of the various
different cleaning processes available. In this study, we have chosen to investigate UV/O3
oxidation for ex situ carbon contamination removal and a variety of standard wet chemistries
and ex situ chemical vapor exposures for oxide removal. Wet chemistries based on H2SO4
and H3PO4 solutions common in GaAs technology for chemical oxide growth were also
investigated [41-44]. The in situ cleaning processes examined included thermal desorption,
exposure to hydrogen plasmas, and annealing in fluxes of Al, Ga, NH3 and SiH4.
Additionally, the use of GaN and In as a passivating/protective layer for AlN surfaces was
additionally investigated.
7.3. Experimental
7.3.1. Integrated Surface Preparation and Analysis System.
All experiments described below were conducted using a unique ultra high vacuum
(UHV) configuration which integrates several completely independent UHV surface
preparation, thin film growth and surface analysis systems via a 36 ft. long transfer line
188
having a base pressure of 9x10-10 Torr (additional details of the transfer line, and many of
the associated systems are provided in Refs. 89,90). The experiments described in this paper
employed the III-V nitride gas source molecular beam epitaxy (GSMBE), Auger electron
spectroscopy (AES), low energy electron diffraction (LEED), x-ray/UV photoelectron
spectroscopy (XPS/UPS), and remote H2 plasma CVD systems. A brief description of these
systems is provided below.
The III-N GSMBE system consisted of a UHV chamber with a base pressure of
3x10-10 Torr and was equipped with a residual gas analyzer (RGA) and a variety of gas
dosers, and Knudsen cells. The RGA (a 0-200 amu quadrapole gas analyzer from Hiden
Analytical Ltd.) was housed in a separate differentially pumped cylindrical chamber (similar
in design to that of Smentkowski and Yates [91]) which had a 0.5 cm diameter orifice at the
head of the RGA for TPD experiments and an approximately 50 cm2 "sunroof" for
monitoring residual gases in the system. The sample heating stage in the ALE system
consisted of a wound tungsten heating filament positioned close to the back of the sample
and mounted on a boron nitride disk [89]. A W/6%Re-W/26%Re thermocouple was
employed to measure the temperature of the backside of the wafer. Surface temperatures and
heating profiles to 1100°C were easily achieved using a programmable microprocessor and
20 amp SCR power supply. Actual surface/sample temperatures (i.e. those reported herein)
were recorded using an infra-red pyrometer with a spectral response of 0.8 to 1.1 µm and a
emissivity setting of 0.5. The estimated experimental accuracy for the latter temperatures
was estimated to be ± 25°C.
Source materials in the GSMBE included Al, Ga, NH3, and SiH4. Al (99.9999%)
was evaporated from a 25 cc "cold lip" Knudsen cell and Ga (99.99999%) was evaporated
189
from a 25 cc dual filament Knudsen cell. The NH3 (99.9995%) was further purified via an in
line purifier connected directly to a leak valve mounted on the GSMBE chamber. The SiH4
(99.995%) was used as received with out any additional purification. Sample exposure to the
NH3 and SiH4 was obtained using "molecular beam" dosers similar to the design of Bozack
et al [92]. Collimation of the ammonia and silane into a molecular beam focused onto the
sample was achieved with this doser using a 13 mm diameter x 2 mm thick glass capillary
array with a ten micrometer pore size (Galileo Electro Optics Inc.). Though, the doser to
sample distance was fixed at 2", no attempts were made to accurately measure the flux of
NH3 or SiH4 and hence all exposures are quoted as Langmuirs (£ = 10-6 Torr sec).
The XPS and UPS experiments were performed in a stainless steel UHV chamber
(base pressure = 2x10-10 Torr) equipped with a dual anode (Mg/Al) x-ray source, He I UV
lamp, and a 100 mm hemispherical electron energy analyzer (VG CLAM II). All XPS
spectra of AlN reported herein were obtained using Al Kα radiation (hν = 1486.6 eV) at 12
kV and 20 mA emission current. For GaN and AlGaN all XPS spectra were acquired using
Mg Kα radiation (hν = 1253.6 eV) at 10 kV and 20 mA emission. XPS analysis typically
required less than 1 hour during which time the pressure never increased above 9x10-10 Torr.
Calibration of the binding energy scale for all scans was achieved by periodically taking
scans of the Au 4f7/2 and Cu 2p3/2 peaks from standards and correcting for the discrepancies
in the measured and known values of these two peaks (83.98 and 932.67 eV, respectively
[74]). Significant sample charging, however, was observed during XPS of bulk
polycrystalline AlN wafers. To correct for these charging effects, the C 1s peak from these
AlN surfaces was assigned a value of 285.7 eV and all the other core levels (O1s, Al 2p, N
1s, F 1s) shifted accordingly. The value of 285.7 eV for the C 1s core level was based on the
190
observation that adventitious carbon on thin AlN surfaces (˜30Å) was observed to occur at
this energy. Curve fitting of most data was performed using the software package GRAMS
386. A combination Gaussian-Lorentzian curve shape with a linear background was found to
best represent the data.. The Auger electron spectrometer and the low energy electron
diffraction optics were mounted on a six way cross off the transfer line and pumped through
the transfer line. In the AES analysis, a 3 keV, 1 mA beam was used. Each Auger electron
spectrum was collected in the undifferentiated mode and numerically differentiated. In
LEED an 80 eV, 1 mA beam was used.
The remote plasma CVD system consisted of a metal seal stainless steel vacuum
chamber pumped by a 330 l/s turbomolecular pump. The base pressure of this system was
4x10-9 Torr. The process gases flowed through a quartz tube mounted at the top of the
chamber, and the plasma is excited by rf (13.56 MHz) applied through a copper coil wrapped
around the quartz tube. The sample was located 40 cm below the center of the rf coil. An in
line Nanochem purifier and filter was used for point of use purification of hydrogen and
silane. Sample heating in the plasma system was conducted using a sample heater similar in
design to the one previously described in the ALE system. The plasma system was also
equipped with a differentially pumped 0-100 amu RGA which allowed direct analysis of the
purity of the process gases. RGA analysis of the hydrogen and silane (both 99.999% purity)
used in these experiments after in situ purification revealed that the impurity level of these
gases were below the baseline of the system (<1 ppm)
7.3.2. Samples and Ex Situ Preparation:
191
The surfaces of a variety of AlN and GaN samples were investigated. The AlN
samples were derived from films (1) epitaxially grown on 6H-SiC (0001)Si by
organometallic vapor phase epitaxy (OMVPE) [93] or gas source-molecular beam epitaxy
(GSMBE) [94], (2) deposited via reactive ion sputtering on Si (111) and (3) hot pressed
polycrystalline AlN wafers. The GaN surfaces investigated were those of GaN films
epitaxially deposited on epitaxial AlN buffer layers grown on 6H-SiC (0001) by OMVPE
[93] and GSMBE [94]. For GaN and AlN films on SiC substrates, the back sides of the
substrates were RF sputter coated with tungsten. This was done to improve the heating
efficiency of these samples as SiC, GaN, and AlN are all partially transparent to the infrared
radiation emitted by our tungsten filament heaters.
The experimental system employed for the ex situ UV/O3 exposures described in this
study employed a high intensity Hg lamp positioned in close proximity (˜ 1 cm) too the AlN
and GaN samples. In order to increase the concentration of generated O3 (i.e. to increase the
oxidation rate), the UV/O3 box was purged with 1 L/sec O2 during some UV exposures.
Further details of this process have been previously described [54-62]. CMOS grade acids
and bases and high resistivity (18.4 M?) de-ionized water were used in all the ex situ wet
chemical cleaning processes examined. The wet chemical cleans investigated included
various mixtures of the following acids and bases: HCl, HF, NH4F, HNO3, H2SO4, H3PO4,
H2O2, NH4OH, NaOH, KOH, acetic acid, and lactic acid. The selection of these chemicals
was based on their extensive usage in standard microelectronic processes. Unless otherwise
noted, all AlN and GaN samples were rinsed in DI water and blown dry with N2 after any wet
chemical processes. After each ex situ cleaning process, the AlN or GaN sample was
mounted to a molybdenum sample holder and loaded into the transfer line load lock for
subsequent surface analysis or in situ cleaning 192
7.4. Results
7.4.1. Ex situ cleaning of AlN
Figure 7.1(a) shows an AES spectrum of the surface of an as-received OMVPE AlN
sample on which a thin oxide layer has formed during ambient exposure. Additional
oxidation via a UV/O3 treatment was investigated initially as a possible ex situ method for C
removal from the nitride surfaces. Both AES and XPS were used to examine an OMVPE
AlN film which had been previously cleaned in trichloroethylene, acetone, and methanol for
5 min. in each solvent and then exposed to UV/O3 for 10 min. at room temperature. As
shown in Figure 7.1(b), the combination of solvent cleaning and UV/O3 exposure reduced
the intensity of the CKLL peak due to surface C by ˜ 50%. A similar decrease in the intensity
of the C 1s core level was also observed in XPS.
Longer UV/O3 exposures (30 min. - 1 hr.) with or without a solvent preclean did not
further appreciably decrease the surface C coverage. However, Figure 7.1(b) does show that
the AlN surface was further oxidized by the UV/O3 treatment. This was confirmed by XPS
analysis of the O 1s core level from bulk polycrystalline AlN wafers (see Figure 7.2). XPS
of the O 1s core level from the AlN wafer after solvent cleaning showed a broad peak which
was deconvoluted into two peaks at 531.3 eV and 533.0 eV. The intensity of both O 1s
peaks was observed to increase after the UV/O3 exposure (see Figure 7.2(a,b)). The reported
binding energies of the O 1s core level from various forms of aluminum oxide (Al2O3) have
ranged from 530.7-532.5 eV [95-100], while the reported binding energy positions for the O
193
1s core level from various nitrate compounds (i.e. MNO3) have ranged from 532.7-533.6 eV
[100]. Therefore, it is tempting to assign the O 1s peak at 531.3 eV to Al-O bonding and the
O 1s peak at 533.0 eV to N-O bonding. However, the second O 1s peak at 533.0 eV could
alternatively be due to aluminum hydroxides [95-99]. Previous XPS examinations of various
aluminum oxides and hydroxides (sapphire, gibbsite, bayerite, bauxite, boehmite, and
diaspore) by Tsuchida et al [97], have shown that the binding energy of the O 1s core level of
OH- species (hydroxides) is typically 532.0-532.3 eV while for O2- species it is typically
530.7-531.5 eV. Additionally, the binding energy of the O 1s core level for H2O has been
reported to be 533.3 eV [100]. Further, Tsuchida et al [97] were successful in deconvoluting
the broad O 1s spectrum from boehmite (AlO(OH)) and diaspore (AlO(OH)) into two
separate peaks located at 530.7 and 532.2. eV which they attributed to O2- and OH- species.
By analogy to these previous observations, it therefore seems likely that the native oxide and
UV/O3 generated oxides on AlN surfaces may alternatively be composed primarily of oxygen
bonded to aluminum in both Al-O and AlO-OH states.
The issue of N-O bonding versus AlO-OH bonding could be resolved simply by the
detection of chemically shifted Al 2p and N 1s core levels which would indicate the presence
of either Al-O or N-O bonding. Interestingly though, no chemically shifted peaks were
observed in XPS spectra of the Al 2p and N 1s core levels from these surfaces (see Figure 7.3
and 7.4). Closer inspection of the literature, however, reveals that relative to pure Al, the
chemical shift of the Al 2p core level for both Al-O and Al-N bonding is approximately the
same = 1.4-1.5 eV [100]. However, the reported binding energy for the N 1s core level for
N-O bonding is 401.4-402.8 eV [100] which is significantly different from the reported
binding energy of 397.3 eV for N-Al bonding [76]. Since in our case only a single N 1s peak
at 398.0 eV indicative of N-Al bonding was observed (Figure 7.4), it seems likely that very
194
little N-O bonding is present at the surface and that the native and UV/O3 oxides are
composed mostly of aluminum oxides and hydroxides. Comparing the relative intensities of
the two O 1s peaks, it was also observed that the higher binding energy peak increased in
intensity with UV/O3 exposure whereas the N 1s intensity decreased (data not shown). This
suggests also that the O 1s peak at 533.0 eV is not related to N-O bonding but to AlO-OH
bonding. Therefore it appears that these surface oxides are predominantly composed of
hydroxide (OH-) species (see Figure 7.2 (a,b)). However, it should be pointed out that in this
case the energy of the photoemitted Al 2p and N 1s core levels are ˜ 1410 and 1000 eV
respectively and accordingly are not very sensitive to the outermost surface layer of AlN (λ ˜
20Å). Hence, the presence of some N-O bonding at the surface can not be absolutely ruled
out. In fact, more surface sensitive photoemission is needed to completely resolve this issue.
30 130 230 330 430 530 630 730
(a)
(b)
(c)
Al C
N
O
dN(E
)/dE
F
Electron Energy (eV)
Figure 7.1. AES survey spectra of OMVPE AlN: (a) as received, (b) solvent cleaned and 20 min. UV/O3 exposure, and (c) 3 min. dip in 10:1 buffered HF (BHF).
195
528 530 532 534 536 538
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
O2- OH -
Figure 7.2. XPS of O 1s core level from bulk AlN wafer (a) as received, (b) UV/O3 exposure, and (c) 10:1 BHF.
70 72 74 76 78
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Al-N
Figure 7.3. XPS of Al 2p core level from bulk AlN wafer after a 10:1 BHF dip.
196
392 396 400 404 408 412 416 420 424
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
N-Al
AlNPlasmon
Figure 7.4. XPS of N 1s core level from bulk AlN wafer after a 10:1 BHF dip.
As AlN is reasonably inert, oxidation of an AlN surface in a typical laboratory
ambient was not observed to proceed rapidly. Therefore, UV/O3 exposure was used to
repeatedly grow a thin oxide on an AlN surface to assess the efficacy of wet chemical
removal of this oxide. Among the wet chemicals investigated, 1:1 HCl:DI, 1:1
NH4OH:H2O2, RCA SC1 and SC2 solutions were observed to significantly reduce the surface
oxide on UV/O3 treated AlN surfaces. However as shown in Figure 7.1(c), 10:1 buffered HF
(7:1 NH4F:HF) processes were observed to most dramatically reduce/remove the UV/O3
oxide on AlN surfaces. A comparison of these various wet chemical cleans is provided in
Table 7.1 and Figure 7.5. For comparison, the CKLL/NKLL and OKLL/NKLL AES peak to peak
height (pph) ratios for UV/O3 treated AlN surfaces etched in NH4OH:H2O2 was ˜ 0.32 and
0.58 respectively, while for UV/O3 treated AlN surfaces etched in 10:1 buffered HF (BHF),
CKLL/NKLL and OKLL/NKLL ratios of 0.22 and 0.12 respectively were obtained (see Table 7.1).
197
In terms of carbon coverage, the RCA SC1 clean was observed to be equivalent to the BHF
clean, however, the RCA SC2 clean was observed to leave a slightly larger carbon coverage.
As displayed in Figure 7.1, it is also interesting to note that a change in the AlLVV line shape
from that typical of aluminum oxide to that of aluminum nitride was also observed after the
10:1 BHF clean (see Figure 7.6) [85]. Further examination of Figure 7.2(c) also reveals that
the 10:1 BHF treatment reduces the intensity of the OH- O 1s core level to that of the O2- O
1s core level. This suggests that HF primarily attacks hydroxide (OH-) species on AlN
surfaces. Although, a detailed comparison between solutions was not made, similar results
were also obtained with both 10:1 HF, 10:1 BHF, and 40% NH4F.
Table 7.1. OKLL/NKLL and CKLL/NKLL AES pph ratios from OMVPE AlN surfaces given various wet chemical treatments following a UV/O3 oxidation (uncorrected for differences in sensitivity).
C/N O/N Al/N UV/O3 0.27 2.57 0.66 10:1 BHF 0.22 0.12 0.24 1:1 HCl:DI 0.29 0.36 0.27 1:1 NH4OH:H2O2 0.32 0.58 0.27 RCA SC1 0.20 0.21 0.30 RCA SC2 0.33 0.21 0.30
198
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
(c)
(d)
(e)
AlC
NO
F
Figure 7.5. AES of (0001) OMVPE AlN after UV/O3 oxidation and oxide removal with (a) 1:1 NH4OH:H2O2, (b) 1:1 HCl:DI, (c) 10:1 BHF, (d) RCA SC1, and (e) RCA SC2. (spectra normalized to NKLL).
30 50 70 90 110 130
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
Figure 7.6. Close up of AlLVV from AES of OMVPE AlN after (a) UV/O3 and (b) 10:1 BHF. 199
It is also important to note that the presence of a small concentration of fluorine was
detected in the AES spectra shown in Figure 7.1(c) and 7.5(c). Relative to XPS, AES is
particularly insensitive to detecting the presence of fluorine and electron stimulated
desorption effects are also a problem [101,102]. Therefore, XPS was employed to further
investigate AlN surfaces cleaned using HF based solutions. For this purpose, a thin (˜ 30Å)
AlN film was intentionally grown on 6H-SiC (0001) by GSMBE [94] in order to avoid any
possible charging effects. Figure 7.7(a) shows an XPS spectrum of the F 1s region from a
GSMBE AlN film after dipping in 10:1 buffered HF (BHF) for 10 min. A broad peak was
detected and which was deconvoluted using a Gaussian-Lorentzian distribution into two
peaks having peak positions of 686.8 and 688.5 eV (see Table 7.2). These peaks were
assigned to Al-F and N-F bonding based on analogy to previous reports of XPS from
AlF3.H2O and NF3 [100,103,104]. The carbon and oxygen contamination on this surface was
studied as well. Figure 7.8(a) shows that after HF dipping most of the carbon is located at
285.7 eV which is typical of adventitious carbon and is indicative of a mixture of C-O and C-
H bonding [100]. The O 1s core level from the HF treated surface was observed to be quite
broad (Figure 7.9(a)) and in this case fitted to a single Gaussian-Lorentzian line shape
centered at 533.1 eV. Thermal desorption of the F, C, and O contamination from this surface
was also studied and is described in the following section.
200
Table 7.2. XPS core level positions and full width half maxima (Γ, FWHM) from a GSMBE AlN surface after dipping in 10:1 BHF and annealing at various temperatures.
Al 2p (Γ) N 1s (Γ) O 1s (Γ) C 1s (Γ) F 1s (Γ)____ ) ) ) BHF Dip 75.4 (1.9) 398.6 (1.8 533.1 (3.2 285.8 (2.3 686.8 (2.2)
688.5 (2.2) 400°C 75.3 (1.9) 398.6 (1.8) 533.1 (2.9) 285.7 (2.1) 686.7 (2.1) 530.7 (1.5) 688.7 (2.3) 600°C 75.2 (1.9) 398.5 (1.8) 533.1 (2.4) 285.5 (1.8) 686.5 (2.0) 531.4 (1.9) 688.5 (2.6) 800°C 75.2 (1.9) 398.4 (1.8) 532.9 (2.6) 285.5 (1.8) 686.5 (1.9) 531.1 (1.4) 688.3 (3.0) 950°C 75.2 (1.9) 398.4 (1.8) 532.8 (2.9)
682 684 686 688 690 692 694
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
Al-F N-F
Figure 7.7. XPS of the F 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C.
201
280 282 284 286 288 290
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
C-Si
Adventitious Carbon
Figure 7.8. XPS of the C 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C.
526 528 530 532 534 536
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
Figure 7.9. XPS of the O 1s core level from a 30Å AlN GSMBE film on (0001) 6H-SiC after (a) dipping in 10:1 BHF, and annealing for 15 min. at: (b) 400°C, (c) 600°C, (d) 800°C, and (e) 950°C.
202
Other wet chemistries based on H2SO4, H3PO4, and NaOH etc. were also investigated
for both oxide and carbon removal from AlN surfaces as these chemicals are commonly used
in GaAs processing [35-53] as well as for wet chemical etching of III-V nitride compounds
[105-108]. Treatments in concentrated H2SO4 and H3PO4 were observed to leave residual
sulfate and phosphate on the surface which was probably related to difficulties in rinsing
these chemicals off the AlN surface due to their viscous nature. H2O2:H2SO4 (Piranha etch)
was observed to be good for removing gross carbon contamination from AlN surfaces
(photoresist, etc.). NaOH was observed to leave traces of Na on the surface, which however,
were successfully removed below the detection limits of XPS with an RCA clean (see Figure
7.10). It should be noted, though, that the detection limits of XPS is ˜ 0.1 atomic % and Na
contamination levels below this level have been historically noted to wreak havoc with Si
MOSFET devices [34]. More dilute levels of H3PO4 were moderately successful for oxide
removal, but it was observed that when etching AlN in H3PO4 at higher temperatures (100-
150°C) the, AFM RMS surface roughness increased from as low as 20Å to as high as 200Å.
203
1060 1065 1070 1075 1080 1085 1090
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 7.10. XPS of Na 2p from polycrystalline AlN surface after (a) etching in NaOH and (b) an RCA clean.
7.4.2. In Situ Processing of AlN
The chemistry and thermal of desorption of F, C, and O contaminants on AlN
surfaces after HF processing was further investigated using AES, XPS, and temperature
programmed desorption (TPD). Fig. 7.7(b) shows that after annealing an HF dipped AlN
surface at 400°C for 15 min., the two F 1s peaks become more distinguishable and positioned
at 686.7 and 688.7 eV. Annealing at 600°C resulted in a reduction in intensity of the higher
binding energy peak, as shown in Figure 7.7(c); this peak was further reduced after annealing
at 800°C. Complete elimination of the low binding energy peak at 686.5 eV was not
achieved until annealing at 950°C for 15 min. (Fig. 7.7(e)). During this series of thermal
desorptions, the C 1s and O 1s core levels were also monitored in XPS. The intensity of the
204
C 1s core level was observed to decrease appreciably after the 600°C anneal, but complete
removal of carbon from the AlN surface was not observed until after the 950°C anneal
similar to fluorine (see Figure 7.8(b-e)). The intensity of the O 1s peak initially slightly
decreased after the 400 and 600°C anneals presumably due to water and CO desorption.
However. the O1s intensity almost doubled in intensity after the 950°C anneal (see Figure
7.9(b-e)). This was attributed to reaction of the AlN surface with water desorbing from the
chamber wall due to heating of the walls during the anneal. It is also important to note that
after the 400°C anneal, two O 1s peaks can again be resolved. However after the 950°C
anneal, only one O 1s peak could be resolved suggesting that the hydroxides dehydrogenate
forming O2- species. Similar desorption effects were also seen for carbon and oxygen from
solvent cleaned and UV/O3 processed AlN surfaces.
In a separate study, TPD was performed on an amorphous AlN film sputtered on Si
(111) which had been subsequently dipped in 10:1 BHF (see Figures 7.11). For this surface,
a strong desorption peak for m/e- 16 and 18 (H2O) was observed to occur at temperatures of <
200°C (see Figure 7.4(a)) which is agreement with the observed decrease in O 1s intensity in
XPS for AlN surfaces annealed at 400°C. A large TPD peak was detected for desorption of
m/e- 19 and 20 (F and HF) at approximately 400°C while a small peak for m/e- 38 (F2) was
detected at 500°C (see Figure 7.11(b,c)). This is also in agreement with the observed decease
in intensity of the 688.7 eV F 1s peak after annealing fluorinated AlN surfaces at this
temperature. Desorption features at 400-500°C for m/e- 2, 12, 16, 18, and 28 were also
detected and are probably related to desorption of H2O, CO, and various organic and
hydrocarbon contaminants. A general increase in m/e- 2 (H2), m/e- 19 (F), and m/e- 28 (N2 or
CO) occurred after 600°C and continued until 1000°C after which annealing was stopped.
205
100 200 300 400 500 600
Cou
nts (
arb.
uni
ts)
Temperature (ÞC)
(a)
(b)
(c)
Figure 7.11. TPD of m/e- (a) 18, (b) 20 and (c) 38 from 10:1 BHF dipped AlN (ß = 20°C/min.).
As complete thermal desorption of fluorine from AlN occurred only at elevated
temperatures, the lower temperature process of remote H plasma cleaning was investigated
for this purpose and for the removal of other contaminants. The details regarding the RF
plasma system and its operation are described elsewhere [97]. Atomic H has been previously
shown to be extremely efficient for the removal of halogens from Si (001) surfaces
[109,110]. Figure 7.12 shows an XPS spectrum of the F 1s region taken from a
polycrystalline AlN wafer before and after exposure to a remote H plasma. Figure 7.12(a)
shows the presence of a large quantity of F on the AlN surface after dipping in 10:1 BHF.
Figure 7.12(b) reveals that almost complete removal of fluorine is achieved after exposure of
the AlN wafer to a 15 mTorr, 20W, remote H plasma at 450°C for 5 min. This technique was
206
also efficient for removal of C from AlN surfaces, as shown by the AES spectra in Figure.
7.13(a) and (b). However, it was not found efficient for removing oxygen from AlN
surfaces.
Annealing AlN in fluxes of Al (˜.1 ML/sec), Ga (˜.1 ML/sec), and NH3 (˜ 1 sccm)
was investigated. Though effective for removing fluorine and carbon, none of these
processes was found particularly affective in further removing oxygen from the AlN surface
(see Figure 7.14). No attempts were made to evaporate a thin layer of Al or Ga onto the AlN
surface at room temperature and then thermally desorb this layer at higher temperatures via
the method of Khan [79]. Exposure to silane at high temperatures (1000°C) was the only in
situ clean found capable of appreciably removing oxygen from an AlN surface (see Figure
7.15). Unfortunately, the loss of oxygen produced by the silane exposure was at the expense
of some deposition of silicon presumably due to the formation of Si-N bonding at the surface
(data not shown).
207
685 687 689 691 693 695 697
(a)
(b)
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
Figure 7.12. XPS of F 1s core level from polycrystalline AlN wafer cleaned in 10:1 BHF (a) before and (b) after remote H plasma exposure at 450°C (15 mTorr, 20W). Note F 1s shifted due to sample charging.
30 130 230 330 430 530 630 730
(a)
(b)
dN(E
)/dE
Electron Energy (eV)
Al CN O
Figure 7.13. AES survey spectra of OMVPE AlN after: (a) UV/O3 and 10:1 BHF dip and (b) remote H plasma at 450°C. 208
526 528 530 532 534 536
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 7.14. XPS of O 1s from (0001) GSMBE AlN after (a) annealing at 1000°C and (b) annealing in a 0.1 ML/sec flux of Al at 1000°C.
526 528 530 532 534 536 538
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 7.15. The XPS O 1s core level from an AlN surface (a) before and (b) after annealing in a SiH4 flux.
209
As previous studies on GaAs have demonstrated that a thin (<1Å) In layer can be
used as a passivation layer to protect GaAs surfaces in air and then be later thermally
desorbed in vacuum [111], In was investigated as a passivation layer for AlN. These films
were deposited on OMVPE and GSMBE AlN films in situ immediately after growth (i.e. no
exposure to atmosphere). Unfortunately, In was not found to be a suitable passivation layer
for AlN as the In films were observed to ball up on the AlN surface instead of wetting the
AlN surface. This effect left part of the AlN surface exposed to ambient and therefore some
of the AlN surface was allowed to oxidize. As illustrated in Figure 7.16, after thermal
desorption of the In layer at 750°C, a significant concentration of oxygen was present on the
AlN surface. The level of this oxygen contamination was equivalent to that left after wet
chemical processing in HF. Therefore, 200Å of GaN was also investigated as an alternative
passivation/protecting layer for AlN. In this case, complete surface coverage was less of a
problem and thermal desorption at ˜ 950°C resulted in an essentially oxygen and carbon free
surface (see Figures 7.17 and 7.18). However, complete desorption of the GaN film was not
observed to occur at this temperature. In order to more completely desorb the GaN
passivation layer, extended annealing at temperatures of > 1000°C were required. However
even after extended annealing at >1000°C, AES and XPS still detected a persistent trace of
Ga on the surface (see Figure 7.19, and Tables 7.3 and 7.4). Additionally, the extended high
temperature annealing was observed to result in increased oxidation of the AlN surface due to
heating of the chamber walls causing water to desorb and react with the AlN surface (see
Figure 7.17(e)).
210
Table 7.3. XPS core level positions and FWHM (Γ) from a 200Å OMVPE GaN capping layer on OMVPE AlN after annealing at various temperatures. Al 2p (Γ) Ga 3d (Γ) Ga 2p3/2 (Γ) O 1s (Γ) C 1s (Γ)____ As Received 20.5 (1.6) 1118.4 (1.8) 532.8 (2.4) 285.7 (2.1) 531.1 (1.5) 500°C 20.2 (1.6) 1118.0 (1.9) 533.2 (2.0) 284.8 (2.0) 531.2 (2.0) 750 75.2 (1.8) 20.1 (1.5) 1117.9 (1.8) 531.8 (2.8) 284.5 (1.9) 950 75.3 (1.7) 20.3 (1.3) 1118.2 (1.9) 532.4 (3.8) 21.5 (1.9) 1119.6 (2.2) >1000°C 75.3 (1.6) 20.2 (1.5) 1118.8 (2.7) 532.6 (3.5)
Table 7.4. Ratio of integrated intensity of Al 2p to Ga 3d and Ga 2p3/2. Al2p/Ga3d Al2p/Ga2p3/2 As Received 0.0 0.0 500°C 0.0 0.0 750°C 0.5 0.06 950°C 3.33 0.11 >1000°C 7.2 0.31
526 528 530 532 534 536 538
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
211
Figure 7.16. XPS of O 1s from In capping layer on OMVPE AlN (a) as received, (b) after annealing at 600°C, and (c) after annealing at 750°C.
526 528 530 532 534 536 538
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
Figure 7.17. XPS of O 1s core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as received, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) > 1000°C.
280 282 284 286 288
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
Figure 7.18. XPS of C 1s core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as received, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) >1000°C. 212
1114 1116 1118 1120 1122 1124
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
(d)
(e)
Figure 7.19. XPS of Ga 2p3/2 core level from 200Å GaN capping layer on (0001) AlN buffer layer, (a) as received, (b) after annealing at 500°C, (c) 750°C, (d) 950°C, and (e) >1000°C.
7.4.3. Ex situ Cleaning of GaN
Figure 7.20 shows an AES spectrum from an as received GaN film grown via
GSMBE. As this sample was kept on a laminar flow bench for the period of time that it was
removed from vacuum (1 day), the level of C contamination was quite small. To investigate
the ability of UV/O3 to remove carbon from GaN surfaces, the GaN film was exposed to
UV/O3 for 10-20 min. and investigated with AES. Figure 7.20(b) shows that this exposure
resulted in additional oxidation of the GaN surface without significantly affecting the level of
C contamination. In fact, the C level actually increased, probably as a result of the increased
handling of the sample during unmounting and remounting for the UV/O3 treatment.
213
Examination of the UV/O3 treated GaN surfaces with LEED did not detect any diffraction
patterns, indicating that the O3 generated oxide is likely amorphous. To test that our UV/O3
apparatus was operating correctly, a GaAs substrate was examined simultaneously with an
OMVPE GaN film. As shown in Figure 7.21, AES showed complete removal of carbon
from the GaAs surface after a UV/O3 treatment in agreement with the results of others
[54,58]. Further examination of more contaminated (0001) OMVPE GaN surfaces using
XPS, however, did show a reduction in surface carbon coverage on these surfaces after
UV/O3 exposure (see Figure 7.22). As in the case of AlN, complete removal of carbon was
not detected but a reduction in carbon was observed in XPS and AES with the C 1s peak in
XPS shifting to a higher binding energy indicative of oxidation of the surface carbon (see
Figure 7.22 and Table 7.5).
30 230 430 630 830 1030 1230
(a)
(b)
(c)
dN(E
)/dE
Electron Energy (eV)
GaON
C
Cl
Ga
Figure 7.20. AES survey spectra from GSMBE GaN after (a) 1 day in air on laminar flow bench, (b) UV/O3 oxidation, and (c) 5 min. etch in 1:1 HCl:DI.
214
Table 7.5. XPS core level positions for OMVPE GaN surfaces after various ex situ treatments (Γ= FHWM). Treatment Ga 3d (Γ) N 1s (Γ) Ga2p3/2 (Γ) O 1s (Γ) C 1s (Γ)____ As received 20.1 (1.4) 397.7 (1.3) 1118.2 (1.8) 532.4 (3.1) 285.3 (2.0) Solvents 20.1 (1.4) 397.7 (1.3) 1118.2 (1.8) 532.4 (3.1) 285.3 (2.0) UV/O3 20.3 (1.4) 397.9 (1.3) 1118.5 (1.9) 532.7 (3.1) 285.8 (2.3) UV/O3 24hr. 20.8 (1.7) 398.2 (1.4) 1119.0 (2.1) 533.0 (3.2) 285.8 (3.9) 531.5 (1.9) HCl:DI 20.2 (1.4) 397.9 (1.4) 1118.3 (1.7) 532.7 (2.9) 285.5 (2.1) BHF Vapor 20.8 (1.4) 398.2 (1.2) 1118.8 (1.8) 533.2 (3.6) 286.1 (2.3) 403.8 (2.0) 11120.5 (1.9)
30 230 430 630 830 1030 1230
(a)
(b)
dN(E
)/dE
Electron Energy (eV)
GaC O
GaAs
Figure 7.21. AES survey spectra from (001) GaAs (a) before and (b) after UV/O3.
215
280 282 284 286 288 290
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Figure 7.22. XPS of C 1s core level from (0001) OMVPE GaN after (a) ultrasonification in trichloroethylene, acetone, and methanol, and (b) UV/O3 exposure.
To further assist in the removal of carbon from the GaN surface in later experiments,
the UV/O3 box was purged with 1 L/sec. of oxygen to increase the concentration of ozone
generated within the box. The O2 purge did result in a further decrease in the surface carbon
coverage by the UV/O3 treatment, but complete carbon removal was still not achieved (see
Figure 7.23). However, the oxygen purge also enhanced the oxidation rate of the GaN
surfaces during the UV/O3 exposures. This was observed by an almost complete
disappearance of the NKLL and N 1s peak in AES and XPS respectively (see Figure 7.23). In
addition, the Ga and N core levels were observed to broaden and shift to higher binding
energies by ˜ 0.7-0.8 eV and 0.5 eV respectively (see Table 7.5). However despite the high
degree of oxidation of this GaN surface, chemically shifted N or Ga core levels could not be
216
detected/resolved to indicate the existence of either Ga-O or N-O bonding. As with AlN, the
observed chemical shifts for Ga core levels in XPS for Ga-N and Ga-O bonding are both
approximately the same. For the Ga 3d core level, binding energies of 19.6-21.0 eV [100]
have been reported for Ga2O3 and for GaN the reported Ga 3d core levels are 19.2-20.3
[81,100]. However as mentioned previously, the reported chemical shifts for the N 1s core
level for N-Ga (BE=397.2 [100]) and N-Ox (BE=400-405 [100]) bonding are much larger. In
fact, for oxidized InN a large chemically shifted N 1s core level at ˜ 405 eV has been
observed and attributed to NO and NO2 species [112]. As a large chemically shifted N 1s
core level was not observed in XPS of the UV/O3 treated GaN surfaces and the N 1s core
level was observed to broaden and decrease in intensity relative to the Ga core levels, it
seems likely that the oxide formed on GaN by the UV/O3 treatment is composed mostly of
Ga-O bonded oxygen.
30 230 430 630 830 1030 1230
dN(E
)/dE
Electron Energy (eV)
O
GaNCGa
217
Figure 7.23. AES survey scan of OMVPE GaN after a 24 hr UV/O3 exposure with 1L/sec. flowing O2. XPS of the O 1s core level from an OMVPE GaN surface undergoing a 24 hr. UV/O3
exposure with 1 L/sec flowing O2 showed the development of a second O 1s peak at 531.5
eV possibly due to the formation of stoichiometric Ga2O3 (O1s = 530.8 eV [100]). Prior to
UV/O3 exposure or after a short UV/O3 exposure, only a single broad O 1s core level
(FWHM = 3.1 eV) centered at 532.4-532.7 eV could be detected from these GaN surfaces
(see Figure 7.24 and Table 7.5). Therefore by analogy to AlN surfaces, it seems likely that
the oxides formed on air exposed and UV/O3 treated GaN surfaces are also composed mostly
of O2- and OH- species (see Figure 7.24).
526 528 530 532 534 536 538
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
(c)
O2-
OH -
Figure 7.24. XPS of O 1s core level from (0001) OMVPE GaN (a) after solvent cleaning, (b) after UV/O3 oxidation for 25 min., and (c) after UV/O3 oxidation for 24 hr. with 1 L/sec flowing O2.
218
The combination of UV/O3 oxidation and exposure to the same acids and bases
investigated above for oxide removal from AlN was also used to determine the best wet
chemical method for removing oxides from GaN surfaces. In this case, solutions of HCl,
NH4OH, and HF were found to be very effective for oxide removal. Table 7.6 summarizes
the AES OKLL/NKLL and CKLL/NKLL pph ratios obtained from GaN surfaces undergoing
UV/O3 oxidation followed by numerous wet chemical treatments. As illustrated in Table 7.6,
1:1 HCl:DI was found to produce the lowest OKLL/NKLL ratio of the wet chemistries
investigated and as illustrated in Fig. 7.20(c) and Fig. 7.25(a) significant coverages of Cl
were observed on HCl treated GaN surfaces. Slightly higher OKLL/NKLL ratios were obtained
with 10:1 HF and 10:1 BHF solutions, but F was not detected by either AES or XPS.
However, the lowest CKLL/NKLL ratios were obtained with these HF based solutions. It
should be noted, however, that the values presented in Table 7.6 represent the very best
values obtained with each clean and as illustrated in Figure 7.25, on average similar results
were obtained for both 1:1 HCl:DI, 10:1 BHF, and 1:1 NH4OH:H2O2 wet chemical
treatments. In fact we have previously reported slightly higher values for OKLL/NKLL and
CKLL/NKLL ratios for HCl and HF treated GaN surfaces. These higher values could be related
to longer times required to mount samples, laboratory ambient and cleanliness (i.e. pollen
count, smoke, air circulation) and the fact that Teflon beakers were used for HF processes
and Pyrex/glass beakers were used for all other wet chemical processes. It should be noted,
however, that for HCl treatments, the oxygen surface coverage was observed to be inversely
related to the amount of Cl detected on the surface (i.e. higher Cl coverage equals lower
oxygen coverage). A similar relation was also observed between carbon and oxygen (i.e.
higher carbon coverage equals lower oxygen coverage). All GaN surfaces treated in either 219
HCl:DI, HF, BHF, or NH4OH:H2O2 displayed (1x1) LEED diffraction patterns indicating
removal of the UV/O3 oxide. Other wet chemistries including , RCA SC1 and RCA SC2,
1:1:7 H2SO4:H2O2:DI, H3PO4, and Acetic were also investigated. RCA SC1 and SC2 were
found to reduce the UV/O3 oxide on GaN surfaces but (see Figure 7.26) the SC2 clean was
found to leave more carbon on the surface relative to the SC1 clean which is similar to what
was observed for the AlN surfaces. The H2SO4 and H3PO4 cleans were found to leave
residual sulfates and phosphates on the GaN surfaces with accordingly higher oxygen levels.
Table 7.6. AES pph ratios of UV/O3 and wet chemical processed OMVPE GaN surfaces Treatment OKLL/NKLL CKLL/NKLL GaLMM/NKLL 3 hr. in air 0.22 0.07 As Received 0.15 0.21 0.59 Solvents 0.20 0.24 0.61 UV/O3 0.85 0.19 0.81 HCl:DI 0.12 0.23 0.62 NH4OH:H2O2 0.29 0.22 0.73 10:1 HF 0.15 0.10 0.69 40% NH4F 0.23 0.14 0.77 10:1 BHF 0.23 0.20 1:1:7 H2SO4:H2O2:DI 0.33 0.22 0.63 85% H3PO4 @125°C 0.27 0.15 0.76 RCA SC1 0.25 0.24 0.60 RCA SC2 0.16 0.35 0.62 BHF Vapor 0.43 0.35 0.77
220
30 130 230 330 430 530 630 730
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
(c)
Cl CONGa
Figure 7.25. AES of (0001) OMVPE GaN after UV/O3 oxidation and oxide removal with (a) 1:1 HCl:DI, (b) 10:1 BHF, and (c) 1:1 NH4OH:H2O2 (spectra normalized to NKLL).
30 230 430 630 830 1030 1230
dN(E
)/dE
Electron Energy (eV)
Ga C
NO
Ga
(a)
(b)
(c)
Figure 7.26. AES of (0001) OMVPE GaN after UV/O3 oxidation followed by (a) RCA SC1 and (b) RCA SC2 (spectra normalized to NKLL). 221
As we have previously found an ex situ exposure to the equilibrium vapor from a
BHF solution to be useful for UV/O3 oxide removal from (0001)Si 6H-SiC surfaces, this
approach was also investigated for oxide removal from (0001) GaN surfaces. As displayed
in Table 7.6 and Figure 7.27, the oxygen surface coverages observed in AES from UV/O3
and BHF vapor treated GaN surfaces were not found to be lower than those for typical wet
chemical approaches. However, XPS analysis seemed to indicate a lower oxygen coverage
for the vapor cleaned GaN surfaces compared to the wet chemical cleaned surfaces. The
reasons for this are currently unclear but may be related to enhanced electron beam oxidation
of the GaN surface due to physisorbed water left on the surface by the vapor treatment. One
other significant difference between the BHF vapor treatment and the wet chemical cleans
was the detection of large amounts of fluorine on the BHF vapor treated surfaces (see Figure
7.27) compared to the BHF wet chemical cleans for which no fluorine was detected. XPS of
the F 1s core level from the BHF vapor treated surfaces showed a broad F 1s peak which
could be deconvoluted into two separate peaks at 686.6 and 688.2 eV similar to BHF wet
chemical processed AlN surfaces (see Figure 7.28). In addition, chemically shifted Ga 2p
and N 1s core level peaks at 1120.5 and 403.8 eV presumably due to Ga-F and N-F bonding
were also detected indicating the possible formation of both GaF3 and NF3 on the surface (see
Figure 7.29, 7.30 and Table 7.5). It is interesting to note that the intensity of the Ga 2p3/2 Ga-
F peak was observed to decrease in intensity along with the 686.6 eV F1s peak after rinsing
in DI water whereas the N 1s peak at 403.8 eV was not effected. Both Ga-F and N-F Ga2p3/2
and N1s peaks were completely removed by cleaning in 1:1 NH4OH:H2O2. As a final note,
preliminary AFM examinations indicated that all of the above wet chemical treatments (with
the exception of H3PO4) did not increase the GaN RMS surface roughness above that of the
as grown surface ˜ 10Å.
222
30 230 430 630 830 1030 1230
C
N
OGa
Cl
F
Ga
dN(E
)/dE
(a)
Electron Energy (eV)
(b)
(c)
Figure 7.27. AES of (0001) OMVPE GaN (a) HCl:DI dip, (b) UV/O3 oxidation, and (c) HF vapor cleaning.
684 686 688 690 692 694
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Ga-F
N-F
Figure 7.28. XPS of F1s core level from (0001) GaN after UV/O3 oxidation followed by (a) 10:1 BHF vapor clean, and (b) a DI rinse. 223
1114 1116 1118 1120 1122
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(b)
(a)
Ga-F
Ga-N
Figure 7.29. XPS of Ga2p core level from (0001) OMVPE GaN after UV/O3 oxidation and (a) BHF vapor clean and (b) DI rinse and NH4OH:H2O2 clean.
393 396 399 402 405
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
N-Ga
N-F
Figure 7.30. XPS of N 1s core level from (0001) OMVPE GaN after UV/O3 oxidation and (a) BHF vapor clean and (b) DI rinse and NH4OH:H2O2 clean. 224
7.4.4. In Situ Processing of GaN
In a previous study, we used AES to examine the thermal desorption of O, C, and Cl
contaminants left on GaN surfaces after various DI:HCl:HF:MeOH wet chemical processes
[85]. In this study, it was observed that annealing these GaN surfaces at 800°C resulted in
only incomplete desorption of carbon and oxygen contaminants [85]. In the GaN/AlN
capping layer study, however, complete desorption of carbon from the GaN surface was
achieved after annealing at 950°C (see Figure 7.18). In the case of oxygen, almost complete
thermal desorption of oxygen was also achieved in this temperature range (see Figure 7.17).
However in separate TPD experiments, it was observed that at Tsub >˜ 800°C an increase in
m/e- 69 (Ga) could be detected and which exponentially increased with temperature (see
Figure 7.31). This suggests that the oxygen and carbon contaminants were removed from the
GaN surface through sublimation of the GaN film. Unfortunately due to experimental
difficulties we were not successful in observing Cl desorption from GaN surfaces using TPD.
However, the concentration of the former was observed to be decreased below the detection
limit of AES by annealing at 600°C or greater (see Figure 7.32). This is contrast to GaAs
surfaces where GaCl was observed to desorb at ˜ 220°C [113].
Further investigation of the removal of native and UV/O3 oxides from GaN via
annealing in fluxes of Ga and NH3 was also conducted. Using AES, both procedures were
observed to result in a reduction in the amount of oxygen on GaN surfaces (see Fig. 7.33(a-
d)). However, oxygen could not be removed below the detection limit (˜ 0.1%) of our AES
system. This however, was later found to be related to electron beam oxidation of the GaN
surface by the Auger e-beam during AES analysis. Figure 7.34 shows a series of AES
spectra taken sequentially for differing periods of time from an as grown GSMBE GaN 225
surface. As this figure illustrates, the intensity of the OKLL peak was observed to increase
with the number of scans. When scanning very quickly or performing only one scan,
essentially no oxygen could be detected from the as grown GaN surface in AES. A detailed
examination of Ga vapor flux cleaning was not investigated here, as it has been previously
thoroughly examined by Kahn [79] and Bermudez [80-82]. However, as shown in Figure
7.33(d) essentially atomically clean GaN surfaces were obtained by annealing in NH3 at
800°C. Further, (2x2) reconstructions were observed in LEED from OMVPE and GSMBE
GaN films annealed in ammonia which has not been observed from GaN surfaces cleaned via
annealing in a Ga flux.
Table 7.7. XPS core levels from GSMBE GaN surfaces after various treatments. Treatment Ga 3d (Γ) Ga 2p3/2 (Γ) N 1s (Γ) O 1s (Γ) C1s (Γ) UV/O3 20.3 (1.6) 1118.1 (1.9) 387.7 (1.3) 532.0 (2.7) 285.0 (1.8) 650°C UHV 20.1 (1.6) 1117.8 (1.9) 397.5 (1.2) 531.0 (2.2) 283.9 (2.3) NH3-650°C 20.1 (1.6) 1117.8 (1.9) 397.5 (1.3) 531.2 (2.7) 283.9 (2.3) Ga-650°C 20.0 (1.8) 1117.4 (2.2) 397.6 (1.6) 532.0 (2.5) NH3-800°C 20.0 (1.6) 1117.5 (1.8) 397.5 (1.4) 531.0 (2.6)
226
0 10 0
1 10 -11
2 10 -11
3 10 -11
4 10 -11
5 10 -11
200 300 400 500 600 700 800 900 1000
m/e
- 69
(arb
. uni
ts)
Temperature (ÞC)
Figure 7.31. m/e- 69 (Ga) signal from GSMBE GaN as a function of surface temperature.
30 130 230 330 430 530 630 730
dN(E
)/dE
Electron Energy (eV)
Ga C
N
O
Cl(a)
(b)
(c)
(d)
Figure 7.32. AES of (0001) OMVPE GaN after (a) HCl vapor clean, and annealing at (b) 300°C, (c) 450°C, and (d) 600°C. (spectra normalized to NKLL).
227
30 230 430 630 830 1030 1230
dN(E
)/dE
Electron Energy (eV)
(a)
(b)
(c)
(d)
Ga CN
O Ga
Figure 7.33. AES survey spectra from GSMBE GaN (a) exposed to UV/O3, and after annealing in: (b) UHV at 650°C, 20 min., (c) NH3 (5x10-6 Torr) at 650°C, 20 min, and (d) NH3 (5x10-6 Torr) at 800°C, 25 min. (spectra normalized to NKLL).
30 230 430 630 830 1030 1230
dN(E
)/dE
Electron Energy (eV)
O
Ga
N
Ga
(a)
(b)
(c)
(d)
Figure 7.34. AES survey spectra from GSMBE GaN after various sequential scans (a) 1, (b) 3, (c) 8, and (d) 9 scans. (spectra normalized to NKLL). 228
The surface electronic structure of (0001) OMVPE GaN surfaces after wet chemical
processing and in situ processing was also investigated using UPS. UPS spectra of a (0001)
OMVPE GaN surface after cleaning in methanol, 1:1 HCl:DI, and annealing in NH3 are
displayed in Figure 7.35 along with a UPS spectrum of an as grown (2x2) GSMBE GaN film.
It was observed that after wet chemical processing the VBM of the GaN surface was located
at -3.8 eV below the system Fermi level perhaps suggesting that the GaN surface Fermi level
is inverted as the bandgap of GaN is only 3.4 eV. After annealing in NH3, however, the
VBM of the GaN surface was observed to move closer to the Fermi level to ˜ 3.0 eV below
EF. In comparison, the VBM of an as grown (2x2) GSMBE GaN film was observed to be
located ˜ 2.7 eV below the Fermi level. However the position of the GaN core levels relative
to the GaN VBM were observed to be the same for the as grown GSMBE GaN surface and
the OMVPE GaN film annealed in NH3 at 800°C (see Table 7.8).
-10 -8 -6 -4 -2 0
Cou
nts (
arb.
uni
ts)
Electron Energy (eV)
-2.7
-3.0
-3.8
(d)
(c)
(b)
(a)
Figure 7.35. UPS of OMVPE GaN after (a) rinsing in methanol, (b) 1:1 HCl:DI, and (c) annealing in NH3 at 800°C, (d) as grown (2x2) GSMBE GaN.
229
Table 7.8. Ga and N core levels from GaN relative to the GaN VBM after various processes. Treatment Ga 3d-VBM Ga2p3/2-VBM Ga3p3/2-VBM N1s-VBM As Received 17.5 eV 1115.5 eV 102.8 eV 395.0 eV 1:1 HCl:DI 17.7 eV 1115.8 eV 103.1 eV 395.2 eV 800°C-NH3 18.4 eV 1116.3 eV 103.6 eV 395.8 eV (2x2) 18.4 eV 1116.3 eV 103.6 eV 395.8 eV
The use of a remote hydrogen plasma for cleaning GaN was also investigated. Figure
7.36(a) shows an AES spectrum from a HCl:DI cleaned OMVPE GaN film before plasma
exposure. Exposure to a 15 mTorr, 20W remote H plasma for 5 min. at 100°C resulted in the
complete removal of both Cl and C from the surface within the detection limits of AES.
However, the level of oxygen was found to increase dramatically, due to the low temperature
plasma exposure (see Figure 7.34(b)). A second exposure of the same sample to a H plasma
at 450°C for 10 min. resulted in a reduction of the oxygen level close to that prior to the first
H plasma exposure. Higher sample temperatures during plasma processing were investigated
(600-800°C), but further reduction of the surface oxygen levels was not realized.
230
200 400 600 800 1000 1200
dN(E
)/dE
Electron Energy (eV)
Cl CN
O Ga
(a)
(b)
(c)
Figure 7.36. AES survey spectra from OMVPE GaN (a) after oxide removal with 1:1 HCl:DI, (b) 100°C H plasma exposure, and (c) 450°C H plasma exposure.
7.4.5. Ex Situ Cleaning of AlGaN Surfaces
Many of the wet chemical processes found effective for oxide and carbon removal for
GaN and AlN surfaces were also investigated for surfaces of OMVPE AlxGa1-xN (x˜0.5)
films. As shown in Figure 7.37, AlGaN surfaces were observed to retain both Cl and F on
the surface for both HCl and HF wet chemical processes. HF vapor cleaning was also
observed to result in the formation of a large amount of F on the surface leading to AlF3
formation (see Figure 7.36 and 7.37).
231
30 230 430 630 830 1030 1230 1430 1630
(a)
(b)
(c)
(d)
(e)
dN(E
)/dE
Electron Energy (eV)
CN
O
Cl
F
Ga Al
Figure 7.37. AES of AlGaN (a) as received, (b) after solvent cleaning in trichloroethylene, acetone, methanol, and isopropanol, (c) 1:1 NH4OH:H2O2, (d) 1:1 HCl:DI, and (e) 10:1 BHF.
30 230 430 630 830 1030 1230 1430 1630
(a)
(b)
Al C
N
O
dN(E
)/dE
Electron Energy (eV)
F Ga Al
Figure 7.38. AES of AlGaN after (a) UV/O3 exposure and (b) HF vapor oxide removal.
232
70 72 74 76 78 80 82
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
(a)
(b)
Al-N
Al-F
Figure 7.39. XPS of Al 2p core level after UV/O3 oxidation and (a) HF vapor oxide removal, and (b) DI rinse and NH4OH:H2O2.
7.5. Discussion
7.5.1. As Received and UV/O3 surfaces
Prabhakaran et al [86] have previously used XPS to examine the native oxide formed
on GaN and observed two O 1s core levels located at 531.3 and 532.7 eV which is in
agreement with our results from native and UV/O3 oxides on both AlN and GaN surfaces.
Prabhakaran et al [86] assigned the lower binding energy peak to Ga2O3 based on the
observation that this peak was removed via etching in NH4OH which is known to dissolve
Ga2O3. However, Prabhakaran et al [86] did not offer an explanation as to the origin of the
233
higher binding energy O 1s core level. Although, Prabhakaran et al [86] did observe a
decrease in intensity on the lower binding energy side of the Ga 2p and N 1s core levels after
etching in NH4OH which they speculated may be due removal of an oxynitride phase at the
surface. Using Zr Mξ radiation (hν = 151.6 eV), Bermudez [82] has recently studied the
interaction of O2 with UHV prepared atomically clean (0001) GaN surfaces. In this case, he
was not able to identify whether oxygen bonded with either N or Ga on the surface but was
able to identify the appearance of a satellite on the Ga 3d core level due to oxygen exposure
which may indicate the formation of Ga-O bonds.
In our XPS analysis of the O 1s core level from as received and UV/O3 treated AlN
and GaN surfaces described here, the presence of oxygen in two different chemical states was
attributed to oxygen bonded to Al (AlN) or Ga (GaN) in both O2- and OH- chemical states.
This was primarily based on the lack of observation of a chemically shifted N 1s core level at
˜ 401-402 eV. The formation of aluminum hydroxides on both aluminum and aluminum
nitride has been previously observed using XPS and FTIR [95-99]. However in both cases,
the hydroxides were observed to form only after suspension in water (H2O) for extended
periods of time (days-weeks). Additionally, Nylund and Olefjord [95,96] observed the
spontaneous decomposition of hydroxides on aluminum into aluminum oxides when inserted
into UHV. Obviously, in our case the "hydroxide" species seem to be much more stable and
were easily detected in UHV by XPS. For these reasons, the higher binding energy O 1s core
level may alternatively be due to oxygen bonded to Al or Ga in a chemical state somewhere
between O2- and OH-. However, we do not feel that the higher binding energy O 1s peak is
due to N-O bonding. In our case the NKLL and N 1s signal in AES and XPS respectively
were observed to decrease in intensity with increased UV/O3 exposure and accordingly the
intensity of both O 1s core levels was observed to increase and the Ga core levels 234
correspondingly were observed to broaden. This argues against the formation or inclusion of
oxynitride phases in the UV/O3 oxide and perhaps the native oxide. In fact one would expect
the more volatile N-O species to be removed from the GaN surface during the UV/O3
treatment which may be the limiting factor in growing an oxide on GaN by this technique.
Finally in our case, we did not observe a decrease in intensity on the low binding energy side
of the N 1s core levels after etching UV/O3 or native oxides in HCl, HF, or NH4OH. Perhaps
during the initial oxidation of GaN surfaces, NOx species form and then are gradually
removed from the surface allowing the formation of a more stoichiometric gallium oxide on
the surface. More detailed analysis of oxidized GaN and AlN surfaces using FTIR could
help resolve this issue.
As for the nature of the carbon contaminants accumulated on both AlN and GaN
surfaces in ambient and wet chemical processing, the XPS C 1s core level was generally
observed to be located at 285.2-285.8 eV with a broad FWHM of 2.2-3.0 eV. Prabhakaran et
al [86] has made similar observations for GaN surfaces and concluded that the adventitious
carbon is composed mostly of C-H bonding. However, in our case we feel that the surface
carbon is really a mixture of C-H and C-O bonding as Miyauchi et al [115] has previously
shown that the C 1s core level for CH2, C-O, and O-C=O bonded carbon contaminants on
silicon surfaces are located at 284.6, 286.3 and 288.4 eV respectively. The exact energy
position of the C 1s core level for these contaminants could be affected by the differences in
band bending at the surfaces of Si, GaN, and AlN. However, we feel that the observation of
a decrease in intensity of the OKLL transition after annealing at > 500°C (see Figure 7.32)
simultaneous with the decrease in intensity and shift to lower binding energy for the C 1s
core level (see Figure 7.18) supports the contention that significant concentrations of C-O
bonded carbon are present on the GaN and AlN surfaces.
235
The ability to use UV/O3 oxidation to completely remove carbon below the detection
limits of AES from GaAs surfaces and not from GaN and AlN surfaces was initially
troublesome. Except for freshly grown surfaces, the UV/O3 exposure was always observed
to reduce the level of carbon detected by both AES and XPS from both GaN and AlN
surfaces. Additionally, the surface carbon C 1s peak was always observed to be shifted to
higher binding energies after the UV/O3 exposure which is indicative of oxidation of the
surface carbon and the formation of more C-O bonds (see Figure 7.22 and Table 7.5).
However as mentioned, complete carbon removal was never achieved and the authors have
accordingly postulated several ideas as to why this is so. It should be first pointed out,
though, that Baunack et al [62] and Fominski et al [57] have previously reported incomplete
removal of carbon from silicon surfaces using UV/O3 oxidation, and we have also observed
this for (0001) 6H-SiC surfaces (see Chapter 3). In the case of GaN, AlN, and SiC, the
inability to completely remove carbon by UV/O3 oxidation may be related to several factors.
The first factor is the extreme chemical inertness of these materials and their resistance to
oxidation which prevents the formation of a complete passivating oxide to prevent carbon
accumulation during sample mounting and transfer. A similarly related factor is the wide
band gap of these materials (especially compared to GaAs). It may be that in addition to the
generation O3, the UV light from the Hg lamp may also create electron hole pairs at the
surface of GaAs which assists in the oxidation of surface carbon via transfer of electrons/hole
pairs from the GaAs to the surface carbon. As the nitrides have much larger bandgaps, fewer
electron/hole pairs may be generated at the surface by the Hg lamp or transfer of electrons
from the Nitride and SiC surface may be not as efficient. Finally in the case of GaN and
AlN, the bonding in these materials is much more ionic than in the case of GaAs and
therefore adventitious carbon may be more strongly bonded or attracted to the nitride
236
surfaces. In summary, the UV/O3 oxidation treatment was found most useful when trying to
remove gross carbon contamination left by photoresist, silver paste, extensive storage in
Fluoroware containers, or (in the case of SiC) polishing treatments.
One further point regarding the oxidation of GaN surfaces that the authors would like
to address, is that as grown GSMBE GaN surfaces were not observed to oxidize rapidly in air
as detailed in Table 7.6. It was generally observed that freshly grown GSMBE GaN surfaces
acquired very few contaminants during overnight storage on a laminar flow bench, and a
similar resistance to contamination was also observed for freshly grown OMVPE GaN films
stored in Flouroware containers (note: Ma et al. [114] have also observed this for fresh
HVPE films). It was also generally observed that the surface carbon initially accumulated on
the nitride surfaces was more closely related to CHx species which was indicated by the
lower binding energy of 284.8-285.0 eV for the XPS C 1s peak from these surfaces.
However, with further aging and processing more C-O bonded carbon was observed to
accumulate on the nitride surfaces shifting the C 1s peak to higher binding energies of 285.6-
286 eV. As previously mentioned, it was sometimes observed that the UV/O3 and wet
chemical cleaning processes described above deposited/accumulated more oxygen and
carbon contaminants on the surface than were previously detected from these surfaces prior
to cleaning. As can be imagined, this greatly complicated our cleaning experiments.
Therefore, most of the experiments described above were conducted on GaN and AlN
surfaces which had been allowed to "age" for several weeks in their Fluoroware containers.
One exception, is the data presented in Table 7.7 for which the GSMBE film was "aged" on a
laminar flow bench for only one day.
237
7.5.2. Wet Chemical and HF Vapor Processing
Ex situ wet chemical cleaning and oxide removal from (0001) GaN and AlN surfaces
with HCl and HF based solutions were observed to produce surfaces of these materials with
the lowest levels of oxygen but with significant concentrations of Cl and F, respectively. It is
interesting to note that in HF and HCl wet chemical processing, F was exclusively detected
on AlN surfaces whereas Cl was exclusively detected on GaN surfaces. This suggests that F
and Cl exclusively bond with Al and Ga atoms at AlN and GaN surfaces respectively. In
fact, this may be expected simply based on comparison of the bond strengths of halogens
with Ga, Al, and N [116]. As Table 7.9 illustrates, the bond strengths of Al-F and Ga-Cl are
much larger than those for N-Cl and N-F. Thus preferential bonding of F and Cl with Al and
Ga over N should be expected based simply on bond strengths. However, XPS of HF
processed AlN surfaces showed two F 1s core levels at 686.5-686.8 eV and 688.3-688.7 eV
which were attributed to both Al-F bonding and N-F bonding. The lower binding energy F
1s peak at ˜ 686.6 eV is in excellent agreement with the value of 686.3 eV previously
reported for fluorine in AlF3.H2O [100]. As the authors are currently not aware of any
reported values for the F 1s core level from NFx species, the assignment of the F 1s peak at
688.5 eV to N-F bonding is based primarily on the observation of a chemically shifted N 1s
core level at ˜ 404 eV from BHF vapor treated OMVPE GaN and AlGaN surfaces (see Figure
7.30).
238
As displayed in Figures 7.28-30 and 7.39, chemically shifted N 1s, Ga 2p3/2, and Al
2p core levels were observed from GaN and AlGaN surfaces treated with a combination
UV/O3-BHF vapor treatment. These chemically shifted core levels were attributed to the
formation of NF3, GaF3 and AlF3 species respectively [117-119]. As only two F 1s peaks
positioned at ˜ 686.5 and 688.5 eV were detected from these surfaces it seems natural to
assign these two peaks to N-F bonding and either Al-F or Ga-F bonding. To further support
our assignment of the higher binding energy F 1s peak at ˜ 688.5 eV to N-F bonding, we note
that for the BHF vapor treated GaN surface the F 1s peak at 686.5 eV and the chemically
shifted Ga 2p3/2 core level were reduced in intensity by rinsing in DI water while the F ls
peak at 688.5 eV and the chemically shifted N 1s peak were not (see Figure 7.28 and 29).
Table 7.9. Bond energies of Cl, F, and H with Al, Ga, and N [116]. Bond Bond Energy (kJ/mol) Al-F 664 Ga-F 577 N-F 238 H-F 566 Al-Cl 511 Ga-Cl 481 N-Cl 381 H-Cl 428 H-H 432 Al-H 285 Ga-H 275 N-H 339
It is important to note, however, that the F 1s peak at ˜ 688.5 eV may also be due to
physisorbed HF. TPD analysis of BHF treated AlN surfaces showed a desorption peak for
both m/e- 20 and 38 at 400°C and 500°C respectively which corresponds will with the
observed decrease in intensity of the 688.5 eV F 1s peak in XPS of fluorinated AlN surfaces.
239
However, the 688.5 eV F 1s was not observed to completely disappear until after
temperatures of 950°C were reached which is were the 686.5 eV was also observed to
disappear. As the authors are not currently aware of any reported values of the binding
energy of the F 1s core level from HF, the authors therefore must leave open the possibility
that N-F and H-F bonded fluorine is present on BHF processed AlN surfaces.
The observation of N-F bonding on HF wet chemically processed AlN surfaces raises
the question of whether some N-Cl bonding is also present on HCl wet chemically treated
GaN surfaces. The observation of both Ga-F bonding and N-F bonding on BHF vapor
treated GaN surfaces also raises the question as to why F is exclusively detected from HF wet
chemically processed AlN surfaces and not from GaN surfaces as well. Unfortunately, we
were not able to detect any chemically shifted Ga or N peaks for HCl wet chemical or vapor
treated GaN surfaces. Further, Cl was extremely difficult to detect with XPS due to week
sensitivity and hence we can't say a lot about what Cl is bonded to on GaN except based on
our previous bond strength arguments. However, we do note that the bond strength of N-Cl
is ˜ 150 kJ/mol larger than the N-F bond strength so N-Cl bonding on GaN should perhaps be
expected based on the observation of N-F bonding on AlN. As for the specificity for F
adsorption on AlN surfaces in HF and Cl adsorption on GaN in HCl, the authors speculate
that this behavior may be actually related to the differences in bandgap between these two
materials. Ohmi [120] has previously explained the ability of HF to hydrogen terminate
silicon surfaces based on the excellent alignment of the H+ ion with the valence maximum of
silicon in HF solutions. This close alignment allows for efficient transfer of electrons and the
formation of a covalent bond. Ohmi also states that the energy position of an ion in an
aqueous solution is a direct function of the electronegativity of the ion with more
electronegative ions having a lower (more negative) energy. As fluorine is more
240
electronegative than chlorine (Pauling electronegativity for F=4 and Cl=3.2), the F- ion
should lie at a lower energy than the Cl- ion. As the bandgap of AlN is much larger than
GaN, it should follow that the VBM of AlN lies below that of GaN. Accordingly, interaction
of F- with the VBM of AlN would be expected whereas the Cl- ion would be expected to
better interact with the VBM of GaN (see Figure 7.40).
GaN AlN1:1 HCl 10:1 HF
H+
OH-
F-SO42-
OH-
F-
K+
O3-
Cl- Cl-
SO42-
EF
Figure 7.40. Schematic illustrating alignment of Cl- and F- ions with VBM of GaN and AlN in 1:1 HCl:DI and 10:1 HF respectively.
One important aspect about the scenario illustrated in Figure 7.40 is that the
termination of GaN and AlN in aqueous solutions may be dependent on the doping level and
type (i.e. the position of the Fermi level). Accordingly, termination of GaN surfaces by H or
OH ions may be more favored for p-type material leading to less chlorine termination. In
+
-
241
contrast however, increased Cl and F termination of GaN and AlN surfaces could perhaps be
alternatively achieved by adjustment and optimization of the pH of HCl and HF solutions
respectively which would control the energy position of Cl and F ions in these solutions. - -
Complete Cl and F termination of GaN and AlN surfaces is desirable as these
contaminants have been observed to desorb from these surfaces at much lower temperatures
than both carbon and oxygen contaminants. In fact for GaN surfaces, complete desorption of
oxides and carbon contaminants was only observed at temperatures where some GaN surface
decomposition was observed to occur (see Figure 7.17-19 and 7.31). In the case of AlN, a
completely fluorine terminated AlN surface is even more desirable as thermal desorption of
oxides from this surface was not even observed to be possible at temperatures of 1000-
1100°C (see Figure 7.9). Additionally, as the bond strengths of N, Ga, and Al with Cl and F
are so strong, tying up dangling bonds at these nitride surface with Cl and F should stabilize
and inhibit re-oxidation of the surface in air. It is also important to note that we have
empirically observed a direct correlation between the halogen surface coverage and the
carbon and oxygen surface coverage on these surfaces. Typically, the larger the halogen
coverage the lower the oxygen and carbon coverage. We have also empirically observed that
larger oxygen coverages correspond to lower carbon coverages and vice versa. This is in
agreement with the previous observations of Ingrey [35] for III-V arsenide and phosphide
surfaces. Ingrey [35], has previously noted that there are a finite number of adsorption sites
on semiconductor surfaces for which oxygen, carbon, and halogen contaminants compete.
Saturation of these sites with one particular specie hinders the adsorption or contamination of
the surface by other species.
Putting the scenario in Figure 7.40 aside, it is interesting to note that HF processes
were generally observed to leave surfaces with fewer carbon contaminants relative to HCl
242
processes. This may be related to the fact that HCl and other processes were conducted in
glass (Pyrex) beakers whereas HF processes were conducted in Teflon beakers (note: HF
etches glass). However, the RCA SC1 clean which was also conducted in a Pyrex beaker
was observed to leave a similar level of carbon contaminants to HF processed GaN and AlN
surfaces. In contrast, the RCA SC2 clean which is an HCl based solution was observed to
always leave a higher carbon coverage. This perhaps suggests that the chemical state of
carbon species in HF are different from those in HCl. In fact, it was observed that the C 1s
peak for surface carbon on HF treated GaN surfaces was typically 0.2-0.4 eV higher in
binding energy than that for HCl treated GaN surfaces. This in turn implies that the HF
processes may leave more C-O bonded carbon whereas HCl processes leave more C-H
bonded carbon. The importance of C-H bonded carbon vs. C-O bonded carbon will be more
fully discussed in the subsequent section on thermal desorption. However, it is worth
mentioning here that combinations of HF and HCl may produce cleaner GaN surfaces due to
enhanced replacement of carbon for chlorine on the GaN surface.
Removal of the UV/O oxide from AlN surfaces using HF was initially surprising
given the known chemical inertness of sapphire (Al O ) [121]. However, there have been
previous reports of etching amorphous aluminum oxide films by HF [121,122] H PO [123],
and CF RIE [124]. Additionally surfaces of Al [95,96], Al O [97], and AlN [98,99] are also
all known to be composed to some extent of aluminum hydroxides (AlO-OH) which may be
more chemically reactive. In fact, our XPS results show that HF attacks primarily the oxide
associated with the higher binding energy O 1s core level which we have previously
attributed to oxygen in a OH chemical state (see Figure 7.2). On the other hand, oxide
removal from GaN surfaces using HCl, HF, and NH OH chemistries was not surprising given
that they already have been previously reported and are commonly used to remove oxides
3
2 3
3 4
x 2 3
-
4
243
from GaAs surfaces prior to metal contact deposition [35-53]. Solutions of 1:1:7
H SO :H O :DI are also used to clean GaAs surfaces, but do so by oxidizing the surface and
contaminants, forming a thin passivating oxide layer on the GaAs surface which is then
thermally desorbed in situ typically prior to epitaxy [40-44]. In our case, the H SO :H O
cleans were observed to remove carbon from the surface but were not observed to form an
oxide layer which again is not suprising given the relative chemical inertness of GaN
compared to GaAs.
2 4 2 2
2 4 2 2
As the AES spectra displayed in Figures 7.27 and 7.38 illustrate, the BHF vapor
treatment does not appear to be very effective for removal of oxides from AlN and GaN
surfaces relative to other wet chemical processes. However as previously mentioned, the
oxide coverage for the BHF vapor treated surfaces appeared to be much lower in XPS.
Although we have no clear explanation for this anomalous behavior, we do note that both
AlF and GaCl are extremely hygroscopic and like to form AlF H O and GaCl H O [117-
119]. Accordingly, these BHF vapor treated surfaces may have retained several monolayers
of physisorbed water on the surface which when excited by the electron beam during AES
analysis may have oxidized the GaN and AlGaN surfaces. In the case of XPS analysis, these
ebeam induced oxidation effects would not be expected to occur. Accordingly, better results
with BHF vapor cleaning may be achieved under parameters where GaF or AlF formation
is prohibited (i.e. by heating substrate or not using a UV/O exposure prior to the vapor
treatment).
. .3 3 3 2 3 2
3 3
3
7.5.3. Thermal Desorption and Capping Layers
244
In a previous paper, we examined the thermal desorption of oxygen and carbon
contaminants on GaN surfaces after wet chemical cleaning in HCl:MeOH, HCl:DI, HF:DI,
and HF:MeOH solutions [85]. It was observed that for all surfaces incomplete removal of
these contaminants was not achieved even after annealing at 800°C [85]. However, thermal
desorption of carbon contaminants at 800°C was observed to be greater for HF treated
surfaces relative to HCl treated surfaces [85]. In particular, the carbon desorption from the
HF:MeOH (MeOH = methanol) treated GaN surfaces was observed to be the most complete
of the wet chemical treatments investigated [85]. These observations suggest perhaps that the
chemical state of the carbon contaminants left on GaN surfaces after HCl and HF wet
chemical processing are somehow different. In fact as mentioned above, we have previously
observed that the binding energy of the C 1s peak for surface carbon on GaN surfaces is
typically located at ˜ 0.4 eV higher binding energy for HF processed GaN surfaces compared
to HCl processed GaN surfaces. In turn, this difference in binding energy for the surface
carbon C 1s suggests that HF wet chemical processes leave more C-O bonded carbon
contaminants on surfaces whereas HCl processes leave more C-H bonded carbon
contaminants. This is important as in XPS examinations of the thermal desorption of carbon
contaminants from both GaN and AlN surfaces, we have observed that the intensity of the C
1s core level decreases and shifts to lower binding energy with higher annealing
temperatures. Generally, after wet chemical processing, the C 1s core level is located at ˜
285-286 eV and shifts to 284-285 eV after annealing at 500-600°C (see Table 7.2 and 7.7 and
Figure 7.18). Ignoring the possibility of band bending effects, this indicates that most of the
C-O bonded carbon desorbs at temperatures = 500-600°C leaving behind only C-H bonded
carbon which apparently desorbs at much higher temperatures. As HF processes tend to
x
245
leave more C-O bonded carbon contaminants, it should therefore not be surprising that
thermal desorption of carbon contaminants at 800°C from these surfaces should be more
complete. Additionally as carbon contaminants left by MeOH should also be expected to be
composed predominantly of C-O bonding, it should not be surprising that the thermal
desorption of carbon contaminants from MeOH:HF treated GaN surfaces is also more
complete at 800°C. These observations and conclusions further emphasize the need for the
investigation of HF:HCl and even HF:HCl:MeOH wet chemical processes for cleaning GaN
surfaces.
Despite all of the above observations, no wet chemical treatments were observed to
have a significant effect on the desorption of oxides from GaN (or AlN) surfaces. Unlike
GaAs, complete thermal desorption of the surface oxide for GaN surfaces was not observed
to occur at 650°C [63,64]. In our case, complete thermal desorption of the surface oxide
from GaN in UHV was not observed to occur until temperatures of 900-950°C where
achieved at which point significant decomposition of the GaN film was observed to occur
with our RGA (see Figure 7.17, and 7.31). Additionally, the XPS spectra of the Ga 2p
core level from a GaN surface annealed in UHV at 950°C (see Figure 7.19) displayed two
core levels perhaps indicating the preferential loss of nitrogen from the surface and hence the
stoichiometry of the GaN film/surface (note: this in contrast to the results of Munir and
Searcy [125] which concluded that GaN decomposes/sublimes congruently).
3/2
The discrepancy between oxide desorption from GaAs and GaN surfaces can be
explained, however, by considering the fact that the oxides from both GaAs and GaN
probably leave the surface as either Ga-O, As-O, or N-O species instead of O . In order for
this to happen, Ga-As or Ga-N bonds must be broken. As the Ga-N bond is much stronger
than the Ga-As bond, higher temperatures will be required to break the Ga-N bond and hence
2
246
desorb the oxide. Therefore it should not be surprising that thermal desorption of oxygen
from GaN surfaces occurs at slightly higher temperatures than for GaAs and that the
temperature at which this occurs is also the temperature at which some decomposition of the
GaN film can be observed. In fact in the case of GaAs, As incongruently sublimes from the
GaAs in UHV at 650°C and GaAs surfaces accordingly have to be annealed in a flux of As in
order to maintain a stoichiometric surface during oxide desorption. Accordingly, annealing
GaN in fluxes of N or Ga may also be necessary to maintain a stoichometric surface during
thermal desorption of the surface oxide (annealing GaN in fluxes of Ga and NH will be
discussed further in the following section). The above line of reasoning also explains the
inability to thermally desorb the oxide from AlN surfaces. The Al-N bond is significantly
stronger than the Ga-N bond and accordingly higher temperatures than were investigated
here would be necessary in order to break the Al-N bond and allow desorption of the oxide as
either Al-O or N-O.
3
As displayed in Figures 7.9 and 7.17, it is interesting to note that the two O 1s core
levels attributed to oxygen in O and OH chemical states merge into one peak of smaller
FWHM at ˜ 800°C. Previous analysis of the thermal decomposition of aluminum hydroxides
(Al(OH) ) using thermogravimetric analysis (TGA), has shown that these materials
decompose into aluminum oxides at temperatures of ˜ 500°C [126]. This is 300°C lower than
the temperature at which we observe the two O 1s chemical states to merge into one and
suggests that perhaps the higher binding energy O 1s peak is due oxygen bonded to
aluminum in some other chemical state than OH . In fact, this higher binding energy O 1s
state could be related to the non-bridging vs. bridging oxygen state seen in silicon oxide films
[127].
2- -
3
-
247
Based on the large bond strengths and the observation that both AlF and GaCl
sublime at 1300°C and 800°C [116-119], the rather high temperature stability of F and Cl on
AlN and GaN surfaces should not be surprising. For F on AlN, complete thermal desorption
of fluorine was not observed until temperatures of 950°C were achieved. However, TPD and
XPS experiments revealed the loss of fluorine from AlN surfaces at much lower temperatures
of T <500°C and which was correlated with desorption of either physisorbed HF or
desorption from nitrogen sites. Desorption of fluorine from AlN at T > 500°C was
correlated mainly with desorption from Al sites, however, desorption of fluorine from N sites
may also occur at T > 500°C. Complete desorption of Cl from GaN was observed to occur
at a slightly lower temperature of ˜ 700-800°C. Due to the inability to detect Cl with XPS
and experimental difficulties with the TPD system, we were not able to correlate Cl
desorption to specific sites (i.e. N or Ga). However as illustrated in Figure 7.32, a significant
decrease in intensity of Cl detected by AES was observed after annealing an HCl vapor
treated GaN film at 450°C. Similar to AlN, Cl desorption from GaN at T < 450°C may be
related to desorption of physisorbed HCl or desorption of Cl from N sites. Based on the Ga-
Cl bond strength, however, it should be expected that Cl desorption from GaN surfaces at
T > 500°C should occur predominantly from Ga sites.
3 3
sub
sub
sub
sub
sub
As for the capping layers on AlN, our results indicate that the success of this idea is
dependent on finding a capping layer which (1) uniformly covers the AlN surface (i.e. no
exposed AlN), (2) can be desorbed at T < 800°C, and (3) can be deposited in situ via either
MBE or OMVPE. Clearly, GaN is too thermally stable to be used for this purpose and
complete coverage with In and other group III (Ga, Th), IV (Sn, Pb), and V (As, Sb, Bi)
elements is likely to always be a problem. Group III-As, P compounds, however, are likely
to exhibit both the lower temperature stability and coverage needed. Unfortunately, most
248
OMVPE and MBE systems do not currently have the capability of depositing both nitride,
phosphide, and arsenide compounds. Therefore, InN may be the best alternative. InN has
been reported to decompose in UHV at ˜ 600°C [128] and should wet/cover the AlN surface.
Unfortunately, growth of pure InN is extremely difficult and has not been frequently
reported. However, In Ga N alloys may work just as well as pure InN capping layers. x 1-x
As for the difficulty in completely desorbing OMPVE GaN capping layers from AlN,
the authors speculate that Ga may be perhaps trapped at the steps of the AlN surface.
Alternatively, residual TMA or TEA may have been left in the OMVPE reactor at the start of
the GaN capping layer growth and which lead to the formation of a thin Al Ga N layer at
the AlN/GaN interface and which is stable at > 1000°C. In fact, the authors note that the
problem of complete GaN desorption was not observed for GaN films grown on AlN buffer
layers by GSMBE. In this case, there is no chance of intermixing of Al and Ga during
growth. Interdiffusion of the AlN and GaN during growth is another alternative but unlikely
for the growth temperatures used in this study (˜1000°C).
x 1-x
7.5.4. Chemical Vapor Cleaning and H Plasma Processes
As mentioned above, annealing GaN surfaces in a flux of Ga or N in order to
maintain a stoichiometric surface may be a more appropriate method for removal of surface
oxides as opposed to simple thermal desorption where some decomposition and or
sublimation of the GaN film is required in order to completely remove the oxide. Annealing
GaN surfaces in fluxes of reactive gas/vapor species such as Ga, N , and NH , may also be +2 3
249
beneficial in that these species may assist in the removal of carbon and oxygen via reacting
with the contaminants and forming more volatile species which desorb at lower temperatures.
In fact, it is now common to anneal silicon and silicon carbide surfaces in reactive fluxes of
Ga, Ge, Si, SiH , Si H , and GeH to assist both in maintaining a stoichiometric surface as
well in removing the surface oxide [129-135]. Further, Kahn et al [79] and Bermudez [80-
82] have demonstrated cleaning of GaN via annealing of GaN surfaces in a flux of Ga.
However, our results have shown that annealing in a NH flux can also be used for obtaining
an atomically clean GaN surface. This approach is more closely analogous to flux cleaning
of other III-V compounds where it has been found necessary to anneal in a flux of the group
V component during oxide desorption in order to counteract incongruent sublimation of the
group V component [65].
4 2 6 4
3
Figure 7.33 illustrates that NH was found to be effective for removal of surface
carbon at temperatures < 600°C which is especially impressive in comparison to thermal
desorption in which carbon removal was incomplete even at temperatures of 800°C.
Ammonia has been previously reported to be an excellent scavenger of hydrocarbons [137].
In Figure 7.34, it is also shown that annealing GaN in 5x10 Torr NH3 at 800°C completely
removes carbon from the GaN surface and leaves only a submonlayer coverage of oxygen.
Although the submononlayer coverage of O in AES can be attributed to electron beam
induced oxidation of the GaN surface during the AES analysis as indicated in Figure 7.34
[136], small traces of oxygen were almost always detected by XPS from GaN surfaces
annealed in NH at 800-900°C. However, NH cleaned GaN surfaces were observed to
display (2x2) reconstructions in LEED. The (2x2) reconstruction has only been previously
observed from as grown RF and ECR MBE samples [138,139].
3
-6
3 3
250
UPS analysis of the electronic structure of NH cleaned GaN surfaces also indicates
no band bending at the GaN surface. For NH cleaned, undoped, OMVPE GaN samples the
VBM was observed to be located at ˜ 3.0 eV below the system Fermi level (see Figure 7.35).
Given that one would expect the Fermi level of undoped GaN to lie 0.1-0.2 eV below the
GaN CBM (E = 3.4 eV), this result indicates an essentially flat band condition for NH
prepared GaN surfaces. This is in contrast to the UPS results obtained by Bermudez [82] for
GaN surfaces prepared via annealing in a Ga flux. In this case, Bermudez observed the GaN
VBM to be located at ˜ 2.4 eV below E indicating significant band bending at the surface
[82]. This result is in much better agreement with our observation of E -VBM = 2.7 eV for
as grown (2x2) GSMBE GaN films examined in situ (see Figure 7.35) It is important to note,
however, that in the case of GaN surfaces prepared by Ga flux cleaning, Bermudez observed
the presence of surface states in his UPS spectra [82] which had previously lead him to
conclude that the GaN VBM was 0.5 eV closer to E [81]. In our case, surface states were
not clearly visible in our UPS spectra and assuming even that we are overestimating the
VBM by 0.5 eV as Bermudez previously did indicates that the (2x2) as grown surface would
be flat band and the NH cleaned OMVPE surfaces are inverted. At this point it is important
to note photovoltage effects could be strongly influencing our results and leading to our
discrepancies with the results of Bermudez. In fact, Bermudez has previously reported the
observation of a photovoltaic shift of ˜ 0.2 eV for the GaN VBM and Ga 3d core level for
GaN surfaces irradiated with a high pressure Hg arc lamp [81]. Finally, we note that
Bermudez observed that the GaN VBM was ˜ 3.0 eV below E for GaN surfaces prepared by
only wet chemical cleaning in 1:10 NH OH:DI.
3
3
g 3
F
F
F
3
F
4
Our plasma cleaning results indicate that H plasmas are very efficient for in situ
removal of halogen and carbon species at temperatures of 450°C which is 300-400°C lower
251
than the temperature required to desorb these contaminants in UHV or annealing in a flux of
NH or Ga. Halogen removal from silicon surfaces using atomic H has been shown to be via
an Ely-Rideal mechanism [110]. In this mechanism, atomic H is able to extract halogens
from silicon without being thermally accommodated at the surface. The H atom can
accomplish this due to the 220 kJ/mol (relative to 1/2H2(g)) of excess potential energy
residing in the H atom. In the case of GaN and AlN, a detailed study has not yet been made
to ascertain whether Ely-Rideal is operable in halogen extraction from these surfaces.
However in Figure 7.12, it was shown that a small amount of F was still present on the AlN
surface after H plasma exposure at 450°C. Given the large flux of atomic H (1016/cm2s
[109]) produced in the plasma, complete removal of F from AlN would be expected if Ely-
Rideal were operable.
3
In one of the ground breaking papers by Nakamura et al [140] it was shown that
annealing p-type GaN in ammonia led to high resistivity material due to compensation of the
p-type dopants with hydrogen. Therefore, one natural concern with regard to cleaning GaN
surfaces via annealing in NH or a H plasma is whether these processes will lead to
compensation of p-type dopants in GaN films. H plasma processes have been previously
shown to lead to compensation of other III-V compounds (InGaAlP and InGaP [141]).
Additionally, Pearton et al [142-143] have intentionally used an ECR H plasma at 250-
400°C to implant H into GaN, AlN, and InN to study the effect of compensation of these
materials. For ECR plasma exposure at 250°C they have observed significant compensation
of both n and p type material. As such H plasma cleaning processes should perhaps operate
at temperatures > 500°C in order to avoid incorporation of H in the subsurface of GaN films.
In the case of NH , Nakamura et al [141] has previously shown that annealing p-type GaN in
3
3
252
1 atm. NH will lead to compensation of the p-type dopants. However for the NH cleaning
processes we have described here, a much lower flux of ammonia was used and hence
hydrogen incorporation and p-type dopant compensation should be much less. In fact, we
note that Kim et al [144] have recently achieved the growth of p-type GaN via NH -GSMBE
with out post growth annealing. In this case the NH flux and temperature used to grow the
GaN is equivalent to the NH flux and temperature we have used to clean the GaN surface.
3 3
3
3
3
Finally, it is also interesting to examine a few cases of homoepitaxial growth on GaN.
In the first such report by Gassmann et al [145], GaN was grown on bulk GaN single crystals
by MBE. In this case the GaN crystal was annealed/cleaned at only 675°C where MBE GaN
growth occurred. Correspondingly, cross sectional TEM showed the presence of a 50 nm
highly defective area at the interface between the MBE GaN film and the GaN single crystal.
As we have shown, thermal desorption at 675°C still leaves a significant amount of carbon
and oxygen on the GaN surface. In contrast, Ponce et al [146] did not observe such a
defective layer in cross section TEM of the interface between a GaN epilayer grown by
MOCVD on a GaN single crystal. In this case, Ponce et al [146] did not specify their surface
cleaning/preparation except that growth occurred at 1050°C. In our thermal desorption
studies, we have shown that thermal desorption at 950°C in UHV is sufficient for complete
desorption of carbon and oxygen from the GaN surface. However, Ponce et al [146] still
note the presence of dislocations originating at the interface between the GaN film/wafer
interface which they attribute to non-optimized surface prep conditions.
7.6. Conclusions
253
Of the numerous acids and bases examined, HF and HCl solutions produced AlN and
GaN surfaces with the lowest coverages of oxygen respectively. However using AES and
XPS, significant amounts of F and Cl were detected on these surfaces after dipping in HF
and HCl, respectively. It is hypothesized that these halogens tie up dangling bonds at these
nitride surfaces hindering re-oxidation of the surface. Fluorine was very thermally stable
requiring temperatures of > 850°C for desorbtion. Remote H plasma exposure was effective
for removing halogens and hydrocarbons from both AlN and GaN surfaces at temperatures of
450°C, but was not particularly efficient for oxide removal. Annealing GaN in NH3 at 700-
800°C produced clean as well as stoichiometric GaN surfaces.
7.7 Acknowledgments
The work described herein was supported by the ONR under contract N00014-91-J-
1416. Appreciation is expressed to Cree Research, Inc. for the 6H-SiC wafers.
7.8 References III-V Nitride Properties 1. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10 1237 (1992). 2. J.H. Edgar, J. Mater. Res., 7 235 (1992). 3. R.F. Davis, Proc. of IEEE, 79 702 (1991). 4. H. Morkoc and S.N. Mohammad, Science, 267 51 (1995). 5. S. Nakamura, S. Masayuki, and S. Yasanobo, Appl. Phys. Lett., 68 3269 (1996). 254
6. M.C. Benjamin, M.D. Bremser, T.W. Weeks, Jr., S.W. King, R.F. Davis, and R.J. Nemanich, Applied Surface Science, 104/105 455 (1996). 7. M.C. Benjamin, C. Wang, R.F. Davis, and R.J. Nemanich, Appl. Phys. Lett., 64 (1994). Surface Cleaning Reviews 8. W. Kern, RCA Review, 39 278 (1978) 9. W. Kern, J. Electrochem. Soc., 137 1887 (1990). 10. T. Ohmi, J. Electrochem. Soc., 143 1957 (1996). Structural Defects and Improper Surface Cleaning 11. B.A. Joyce, J.H. Neave, and B.E. Watts, Surface Science, 15 1 (1969). 12. B. Meyerson, E. Ganin, D. Smith, and T. Nguyen, J. Electrochem. Soc., 133 1232 (1986). 13. J.H. McFee, R.G. Swartz, V.D. Archer, S.N. Finegan, and L.C. Feldman, J. Electrochem. Soc., 130 3083 (1989). 14. A. Miyauchi, Y. Inoue, M. Ohue, N. Momma, T. Suzuki, and M. Akiyama, J. Electrochem. Soc., 137 3257 (1990). 15. C. Galewski, J. Lou, and W.G. Goldham, IEEE Trans. Semicond. Manfact., 3 93 (1990). 16. M.K. Sanganeria, M.C. Ozturk, G. Harris, K.E. Violette, I. Ban, C.A. Lee, and D.M. Maher, J. Electrochem. Soc., 142 3961 (1995). 17. F.K. LeGoues, MRS Bulletin, 21 38 (1996). 18. G.R. Srinivasan and B.S. Meyerson, J. Electrochem. Soc., 134 1518 (1987). 19. G.R. Srinivasan, J. Cryst. Growth, 70 201 (1984). Metal Contacts and Surface Cleaning 20. K. Prasad, Vacuum, 46 127 (1995). 21. L.M. Yeh, Y. Xie, and P.H. Holloway, J. Appl. Phys., 65 3568 (1989). 22. Z. Lilienal-Weber, N. Newman, J. Washburn, and E.R. Weber, Appl. Phys. Lett., 54 356 (1989). 23. F. Ren, A.B. Emerson, S.J. Pearton, T.R. Fullowan, and J.M. Brown, Appl. Phys. Lett., 58 1030 (1991). 255
24 . M. Kodama, Electronics Letters, 30 (1) 89 (1994). 25. R. van de Walle, R.L. Van Meirhaeghe, W.H. Laflere, and F. Cardon, J. Appl. Phys., 74 1885 (1993). 26. G. Stareev and H. Kunzel, J. Appl. Phys., 74 7592 (1993). Schottky Barrier and Surface Cleaning 27. S.D. Offsey, J.M. Woodall, A.C. Warren, P.D. Kirchner, T.I. Chappell, and G.D. Pettit, Appl. Phys. Lett., 48 475 (1986). 29. W. Monch, J. Vac. Sci. Technol. B, 4 1085 (1986). 30. L.J. Brillson, M.L. Slade, R.W. Viturro, M.K. Kelly, N. Tache, G. Margaritondo, J.M. Woodall, P.D. Kirchner, G.D. Pettit, and S.L. Wright, J. Vac. Sci. Technol. B, 4 919 (1986). 31. K. Smit, L. Koenders, and W. Monch, J. Vac. Sci. Technol. B, 7 888 (1989). 32. A. Zangwill, , (Cambridge Univ. Press, New York, 1988) pp. Physics at Surfaces 221-231. 33. E.H. Rhoderick, and R.H. Williams, Metal-Semiconductor Contacts, 2nd ed. (Oxford University Press, New York, 1988), pp. 5-17. Alkali Ions and MISFETS 34. E. Yon, W.H. Ko, and A.B. Kuper, IEEE Trans. Electron Devices, 13 276 (1966). Wet Chemical GaAs Surface Cleaning 35. S. Ingrey, J. Vac. Sci. Technol. A, 10 829 (1992). 36. C.C. Chang, P.H. Citrin, and B. Schwartz, J. Vac. Sci. Technol., 14 943 (1977). 37. P.A. Bertrand, J. Vac. Sci. Technol., 18 28 (1981). 38. A. Munoz-Yague, J. Piqueraz, and N. Fabre, J. Electrochem. Soc., 128 149 (1981). 39. R.P. Vasquez, B.F. Lewis, and F.J. Grunthaner, J. Vac. Sci. Technol. B, 3 791 (1983). 40. J.M. Woodall, P. Oelhafen, T.N. Jackson, J.L. Freeouf, and G.D. Pettit, J. Vac. Sci. Technol. B, 1 795 (1983). 41. J. Massies and J.P. Contour, Appl. Phys. Lett., 46 1150 (1985). 256
42. J. Massies and J.P. Contour, J. Appl. Phys., 58 806 (1985). 43. G. Hughes and R. Ludeke, J. Vac. Sci. Technol. B, 4 1109 (1986). 44. J.P. Contour, J. Massies, H. Fronius, and K. Ploog, Jpn. J. Appl. Phys., 27 L167 (1988). 45. Z.H. Lu, C. Lagarde, E. Sacher, J.F. Currie, and A. Yelon, J Vac. Sci. Technol. A, 7 646 (1989). 46. Y. Hirota, K. Sugii, and Y. Homma, J. Electrochem. Soc., 138 799 (1991). 47. S.M. Mokler, P.R. Watson, L. Ungier, and J.R. Arthur, J. Vac. Sci. Technol. B, 10 2371 (1992). 48. Z. Song, S. Shogen, M. Kawasaki, and I. Suemune, Appl. Surf. Sci., 82/83 250 (1994) 49. Z. Song, S. Shogen, M. Kawasaki, and I. Suemune, J. Vac. Sci. Techol. B, 13 77 (1995). 50. H. Yao, S. Yau, and K. Itaya, Appl. Phys. Lett., 68 1473 (1996). 51. K. Matsushita, N. Suzuki, S. Okuyama, and Y. Kumagai, Jpn. J. Appl. Phys., 35 5293 (1996). 52. L.J. Gomez Zazo, M.T. Montojo, J.L. Castano, and J. Piqueras, J. Electrochem. Soc., 136 1480 (1989). 53. D.E. Aspnes, J. Vac. Sci. Technol., 17 1057 (1980). UV/O Cleaning 354. S. Ingrey, W.M. Lau, and N.S. NcIntyre, J. Vac. Sci. Technol. A, 4 984 (1986). 55. J.R. Vig, J. Vac. Sci. Techonl. A, 3 1027 (1985). 56. M. Tabe, Appl. Phys. Lett., 45 1073 (1984). 57. V.Y. Fominski, O.I. Naoumenko, V.N. Nevolin, A.P. Alekhin, A.M. Markeev, and L.A. Vyukov, Appl. Phys. Lett., 68 2243 (1996). 58. J.A. McClintock, R.A. Wilson, and N.E. Byer, J. Vac. Sci. and Technol., 20 241 (1982). 59. R.F. Kopf, A.P. Kinsella, and C.W. Ebert, J. Vac. Sci. Technol. B, 9 132 (1991).
25760. M. Suemitsu, T. Kaneko, and M. Miyamoto, Jap. J. Appl. Phys., 28 2421 (1989).
61. S.J. Pearton, F. Ren, C.R. Abernathy, W.S. Hobson, and H.S. Luftman, Appl. Phys. Lett., 58 1416 (1991). 62. S. Baunack and A. Zehe, Phys. Stat. Solid A, 115 223 (1989). Thermal Desorption/ Chemical Beam Cleaning 63. R.P. Vasquez, B.F. Lewis, and F.J. Grunthaner, Appl. Phys. Lett., 42 293 (1983). 64. A.J. SpringThorpe, S.J. Ingrey, B. Emmerstorfer, P. Mandeville, and W.T. Moore, Appl. Phys. Lett., 50 77 (1987). 65. T.H. Chiu, W.T. Tsang, M.D. Williams, C.A.C. Mendonca, K. Dreyer, and F.G. Storz, Appl. Phys. Lett., 65 3368 (1994). H Plasma Cleaning of GaAs/InP 66. M. Yamada and Y. Ide, Jpn. J. Appl. Phys., 33 L671 (1994). 67. T. Kikawa, I. Ochiai, and S. Takatani, Surface Science, 316 238 (1994). 68. E.J. Pettit and F. Houzay, J. Vac. Sci. Technol. B, 12 547 (1994). 69. Y. Sakamoto, T. Sugino, H. Ninomiya, K. Matsuda, and J. Shirafuji, Jpn. J. Appl. Phys., 34 1417 (1995). 70. R.W. Bernstein and J.K. Grepstad, J. Vac. Sci. Technol. A, 7 581 (1989). 71. C.M. Rouleau and R.M. Park, J. Appl. Phys., 73 4610 (1993). AlN Surface Studies 72. V. Grafe, H. Reinhardt, D. Schalch, and A. Scharmann, Phys. Stat. Sol. A, 136 K105 (1993). 73. C.G. Olson, J.H. Sexton, D.W. Lynch, A.J. Bevolo, H.R. Shanks, B.N. Harmon, W.Y. Ching, and D.M Wieliczka, Solid State Comm., 56 35 (1985). 74. R.V. Kasowski and F.S. Ohuchi, Phys. Rev. B, 35 9311 (1987). 75. H.H. Madden and D.W. Goodman, Surface Science, 150 39 (1985). 76. J.A. Taylor and J.W. Rabalais, J. Chem. Phys., 75 1735 (1981). 77. Y. Kido, M. Kakeno, K. Yamada, T. Hioki, J. Kawamoto, and M. Tada, J. Phys. D, 15 2067 (1982). GaN Surface Studies 78. J. Hedman and N. Martensson, Physica Scripta, 22 176 (1980). 258
79. M.A. Khan, J.N. Kuznia, D.T. Olson, and R. Kaplan, J. Appl. Phys., 73 3108 (1993). 80. V.M. Bermudez, R. Kaplan, M.A. Khan, and J.N. Kuznia, Phys. Rev. B, 48 2436 (1993). 81. V.M. Bermudez, T.M. Jung, K. Doverspike, A.E. Wickenden, J. Appl. Phys., 79 110 (1996). 82. V.M. Bermudez, J. Appl. Phys., 80 1190 (1996). 83. M.M. Sung, J. Ahn, V. Bykov, D.D. Koleske, A.E. Wickenden, J.W. Rabalais, Phys. Rev. B, 54 14652 (1996). 84. B. Daudin, J.L. Rouviere, and M. Arlery, Appl. Phys. Lett., 69 2480 (1996). 85. L.L. Smith, S.W. King, R.J. Nemanich, and R.F. Davis, J. Elect. Mater., 25, 805 (1996). 86. K. Prabhakaran, T.G. Andersson, and K. Nozawa, Appl. Phys. Lett., 69 3212 (1996). 87. H. Ishikawa, S. Kobayashi, Y. Koide, S. Yamasaki, S. Nagai, J. Umezaki, M. Koike, and M. Murakami, J. Appl. Phys., 81 1315 (1997). 88. W.C. Hughes, W.H. Rowland, Jr., M.A.L. Johnson, S. Fujita, J.W. Cook, Jr., J.F. Schetzina, J. Ren, and J.A. Edmond, J. Vac. Sci. Technol. B, 13 1571 (1995). Experimental Refs. 89. J. van der Weide, Ph.D. disseration, NCSU (1994). 90. S.W. King, R.S. Kern, M.C. Benjamin, J.P. Barnak, R.J. Nemanich, and R.F. Davis, submitted to J. Appl. Phys. 91. V.S. Smentkowski and J.T. Yates Jr., J. Vac. Sci. Technol. A, 7 3325 (1989). 92. M.J. Bozack, L. Muehlhoff, J.N Russel Jr., W.J. Choyke, and J.T. Yates, Jr., J. Vac. Sci. Technol. A, 5 1 (1987) 93. T.W. Weeks, Jr., M.D. Bremser, K.S. Ailey, E. Carlson, W.G. Perry, and R.F. Davis, Appl. Phys. Lett. 67 401 (1995). 94. S.W. King, W.G. Perry, E.P. Carlson, R.J. Therrien, K.M. Tracy, R.J. Nemanich, and R.F. Davis, submitted to J. Appl. Phys.. 259
Aluminum Hydroxides 95. A. Nylund and I. Olefjord, Surf. Intf. Analysis, 21 283 (1994). 96. A. Nylund and I. Olefjord, Surf. Intf. Analysis, 21 290 (1994). 97. T. Tsuchida and H. Takahashi, J. Mater. Res., 9 2919 (1994). 98. P. Bowen, J.G. Highfield, A. Mocellin, and T.A. Ring, J. Am. Ceram. Soc., 73 724 (1990). 99. J.G. Highfield, P. Bowen, Anal. Chem., 61 2399 (1989). Nitrogen-Oxygen 100. XPS Handbook, Jill Chastain, Ed., Perkin Elmer Eden Prairie, MN (1992). Fluorine ESD 101. T.J. Chuang, H.F. Winters, and J.W. Coburn, Surface Science, 2, 514 (1978). 102. J.W. Coburn, H.F. Winters, and T.J. Chuang, J. Appl. Phys., 48, 3532 (1977). Group III Halogens 103. D.R. Lide, CRC Handbook of Chemistry and Physics, 71st ed, (CRC Press, New York 1991) p. 4-41, 9-87. 104. V.I. Nefedov, Y.A. Buslaev, Y.V. Kokunov, Zh. Neorg. Khim., 19 1166 (1974). Wet Chemical Etching of AlN 105. T.J. Chu and R.W. Kelm, Jr., J. Electrochem. Soc., 122 995 (1975). 106. X.D. Wang, U. Mazur, K.W. Hipps, and J.T. Dickson, Thin Solid Films, 240 45 (1994). Wet Chemical Etching of GaN Surfaces 107. S.S. Kocha, M.W. Peterson, D.J. Arent, J.M. Redwing, M.A. Tischler, and J.A. Turner, J. Electrochem. Soc., 142 L238 (1995). 108. T. Kozawa, T. Kachi, T. Ohwaki, Y. Taga, N. Koide, and M. Koike, J. Electrochem. Soc., 143 L17 (1996). Halogen Removal with Atomic H 109. J.P. Barnak, S.W. King, J.S. Montgomery, Ja-Hum Ku, and R.J. Nemanich in Ultraclean Semiconductor Processing Technology and Surface Chemical Cleaning , M. Liehr, M. Heyns, M. Hirose, and H. Parks eds., (Mater. Res. and Passivation Soc. Proc. 386, Pittsburgh, PA, 1995) pp. 357-362. 110. C.C. Cheng, S.R. Lucas, H. Gutleben, W.J. Choyke, and J.T. Yates, Jr., J. Am. 260
Chem. Soc., 114 1249 (1992). In Passivation Layer 111. C.K. Peng, S.L. Tu, S.S. Chen, and C.C. Lin, Appl. Phys. Lett., 66 2549 (1995). InN Oxidation 112. C.P. Foley and J. Lyngdal, J. Vac. Sci. Technol. A, 5 1708 (1987). Cl Desorption from GaAs 113. C. Sasaoka, Y. Kato, and A. Usui, J. Cryst. Growth, 115 94 (1991). Air Exposure of GaN 114. J. Ma, B. Garni, N. Perkins, W.L. O'Brien, T.F. Kuech, and M.G. Lagally, Appl. Phys. Lett., 69 3351 (1996). C-O and C-H Bonded Carbon 115. A. Miyauchi, Y. Inoue, M. Ohue, N. Momma, T. Suzuki, and M. Akiyama, J. Electrochem. Soc., 137 3257 (1990). AlF and GaF 3 3116. CRC Handbook, D.R. Lide, Ed., CRC Press, Boston (1990). 117. N.A. Solntseva, A.M. Zagudaev, E.A. Burakov, and T.V. Stratonova, Tsvetnye Metally, 13 51 (1986) 118. P.F.M. van Gaans and J.C. van Miltenburg, J. Solution Chem., 20 335 (1991). 119. W.C. Simpson, T.D. Durbin, P.R. Varekamp, and J.A. Yarmoff, J. Appl. Phys., 77 2751 (1995). Energy of Ions in Aqueous Solutions 120. T. Ohmi, J. Electrochem. Soc., 143 1957 (1996). Etching of Aluminum Oxide 121. R.G. Frieser, J. Electrochem. Soc., 113 357 (1966). 122. R.S. Nowicki, J. Vac. Sci. Technol., 14 127 (1977). 123. T. Kawabe, M. Fuyama, and S. Narishige, J. Electrochem. Soc., 138 2744 (1991). 124. B. Zhou and W.F. Ramirez, J. Electrochem. Soc., 143 619 (1996). GaN Decomposition/Sublimation 125. Z.A. Munir and A.W. Searcy, J. Chem. Phys., 42 4223 (1965). 261
Hydroxide Decomposition 126. I. Chen, S. Hwang, and S. Chen, Ind. Eng. Chem. Res., 28 738 (1989). 127. G. Hollinger and F.J. Himpsel, Appl. Phys. Lett., 44, 93 (1984). InN Decomposition 128. R.D. Jones and K. Rose, J. Phys. Chem. Solids, 48 587 (1987). CVC Cleaning 129. S. Wright and H. Kroemer, Appl. Phys. Lett, 36, 210 (1980). 130. J.F. Morar, B.S. Meyerson, U.O. Karlsson, F.J. Himpsel, F.R. McFeely, D. Rieger, A. Taleb-Ibrahimi, and J.A. Yarmoff, Appl. Phys. Lett., 50, 463 (1987). 131. M. Racanelli, D.W. Greve, M.K. Hatalis, and L.J. van Yzendoorn, J. Electrochem. Soc., 138, 3783 (1991). 132. H. Hirayama, R. Tatsumi, A. Ogura, and N. Aizaki, Appl. Phys. Lett., 51, 2213 (1987). 133. H. Hirayama and T. Tatsumi, J. Appl. Phys., 66, 629 (1989). 134. K. Saito, T. Amazawa, and Y. Arita, J. Electrochem. Soc., 140, 513 (1993). 135. A.Fissel, B. Schroter, and W. Richter, Appl. Phys. Lett., 66, 3182 (1995). Electron beam induced oxidation 136. J.L. Melendez and C.R. Helms, J. Electrochem. Soc., 141 1973 (1994). NH Scavenging carbon 3137. F.C. Sauls, W.J. Hurley, L.V. Interrante, P.S. Marchetti, and G.E. Maciel, Chem. Mater., 7 1361 (1995). (2x2) Reconstruction 138. P. Hacke, G. Feuillet, H. Okumura, and S. Yoshida, Appl. Phys. Lett., 69 2507 (1996). 139. K. Iwata, J. Asahi, S. Yu, K. Asami, H. Fujita, M. Fushida, and S. Gonda, Jpn. J. Appl. Phys., 35 L289 (1996). Hydrogen Passivation of Acceptors 140. S. Nakamura, N. Iwasa, M. Senoh, and T. Mukai, Jpn. J. Appl. Phys., 31 107 (1992). 141. V.A. Gorbylev, A.A. Chelniy, A.Y. Polyakov, S.J. Pearton, N.B. Smirnov, R.G. Wilson, A.G. Milnes, A.A. Cnekalin, A.V. Govorkov, B.M. Leiferov, O.M. Borodina, and A.A. Balmashnov, J. Appl. Phys., 76 7390 (1994). 262
142. J.M. Zavada, R.G. Wilson, C.R. Abernathy, and S.J. Pearton, Appl. Phys. Lett., 64 2724 (1994). 143. S.J. Pearton, C.R. Abernathy, P.W. Wisk, W.S. Hobson, and F. Ren, Appl. Phys. Lett., 63 1143 (1993). 144. W. Kim, A. Salvador, A.E. Botchkarev, O. Aktas, S.N. Mohammad, and H. Morkoc, Appl. Phys. Lett., 69 559 (1996). Homepitaxial Growth on GaN 145. A. Gassmann, T. Suski, N. Newman, C. Kisielowski, E. Jones, E.R. Weber, Z. Liliental-Weber, M.D. Rubin, H.I. Helava, I. Grzegory, M. Bockowski, J. Jun, and S. Porowski, J. Appl. Phys., 80 2195 (1996). 146. F.A. Ponce, D.P. Bour, W. Gotz, N.M. Johnson, H.I. Helava, I. Grzegory, J. Jun, and S. Porowski, Appl. Phys. Lett., 68 917 (1996).
263
8. X-ray Photoelectron Spectroscopy analysis of GaN/(0001)AlN and
AlN/(0001)GaN
Growth Mechanisms
To be Submitted for Consideration for Publication
to the
Journal of Applied Physics
by
Sean W. King, William G. Perry, Eric P. Carlson, Robert J. Therrien, Robert J. Nemanich,
and Robert F. Davis
Department of Materials Science and Engineering
North Carolina State University
Raleigh, NC 27695
264
8.1. Abstract
The growth mechanisms of GaN on (0001) AlN and AlN on (0001) GaN have been
investigated using x-ray photoelectron spectroscopy (XPS), low energy electron diffraction
(LEED), and Auger electron spectroscopy (AES). It has been found that GaN growth on
(0001) AlN at low temperatures (650-780°C) occurs via a Stranski-Krastanov 3D type
growth mechanism. The GaN on (0001) AlN growth mechanism, however, switches to a
Frank van der Merwe/layer by layer type growth mechanism at higher temperatures
(>800°C). AlN growth on (0001) GaN was observed to occur via a FM/layer by layer
growth mechanism in the temperature range investigated (750-900°C). We propose a model
based on the interaction of atomic hydrogen with GaN/AlN surfaces which shows that the
surface kinetics of hydrogen desorption/ammonia decomposition is the determining factor for
the GaN growth mechanism.
265
8.2. Introduction
GaN and AlN are completely miscible semiconductors with wide band gaps of 3.40
and 6.2 eV, respectively. Many potential applications for these materials including UV-
visible/optoelectronics, high-power, high-frequency, and high-temperature electronic devices
have been recently realized [1-3]. To date, the two most popular techniques for growth of
these materials has been organometallic vapor phase epitaxy (OMVPE) and electron
cyclotron resonance-molecular beam epitaxy (ECR-MBE). The success and improved
understanding of the OMVPE and ECR-MBE techniques has been greatly aided by several
studies of the growth mechanisms of GaN and AlN on various substrates by these techniques
[4-10]. An alternative to OMVPE and ECR-MBE growth of GaN, is reactive molecular
beam epitaxy (RMBE) which essentially replaces an ECR N2 source with NH3 in an MBE
system (i.e. NH3-GSMBE). NH3-GSMBE is currently gaining increased attention due to its
inherent simplicity relative to ECR-MBE and the improved electrical and optical properties
of GaN films grown by this technique [11-14].
The first reports of growth of AlN, GaN, and AlxGa1-xN alloys by NH3-GSMBE
were those by Yoshida et al [15-19]. They reported successful growth of highly resistive
single crystals of AlN (as determined by reflection high energy electron diffraction
(RHEED)) on (0001) and (11-20) Al2O3 substrates at temperatures of 1000 and 1100°C
respectively [15,16]. The GaN films grown at 700°C on (0001) Al2O3 were also single
crystalline though conductive with high carrier concentrations of 1019-1020/cm3 and
mobilities of 30 cm2/Vsec [17]. Use of an AlN buffer layer for growth of GaN on Al2O3,
266
did not significantly improve the electrical properties of their films. However, Yoshida et al
did make the key observation that the band edge cathodoluminescence (CL) intensity from
their GaN films grown on AlN/Al2O3 were 25 times more intense than those films grown
directly on Al2O3 [18,19].
Since the preliminary work by Yoshida et al, a steady improvement in the quality of
GaN films grown by NH3-MBE has been achieved. GaN films grown directly on (0001)
Al2O3 by Powell et al [20] at 760-780°C were found to exhibit carrier concentrations as low
as 1-4x1018/cm3 and mobilities as high as 100-110cm2/Vsec. A further reduction in carrier
concentrations to 2x1017/cm3 was additionally demonstrated by Yang et al [11] and Kamp
[14] et al by the use on an AlN buffer layer on (0001) Al2O3 [11,14]. Finally, through
optimization of the NH3 flux, Kim et al [13] have been able to grow at 850°C highly resistive
GaN films with carrier concentrations < 1014/cm3 and mobilities as high as 200 cm2/Vsec
[13]. The reduction of the background carrier concentrations to these levels, has additionally
allowed Kim et al and Yang et al to recently achieve Mg p-type doping of GaN without post
growth annealing [11,13].
As improvements and increased understanding of OMVPE and ECR-MBE
techniques were aided by several studies on the growth mechanisms of GaN and AlN on
various substrates [4-10], it should be expected that further improvement in NH3-GSMBE
growth of GaN and AlN should also be aided by such studies. In this paper, we have used
surface analytical techniques such as x-ray photoelectron spectroscopy (XPS), Auger electron
spectroscopy (AES), and low energy electron diffraction (LEED) to study the initial growth
mechanisms of GaN on AlN and AlN on GaN by the NH3-GSMBE technique. We show
267
that at low growth temperatures (Tsub<800°C) GaN growth on AlN proceeds via a Stranski-
Krastanov (SK) growth mechanism (2D-3D) and at higher temperatures (Tsub>800°C) GaN
growth on AlN proceeds via a Frank van der Merwe (FM)/layer by layer mechanism. The
change from a SK to a FM growth mechanism is attributed to a transition from a fully
hydrogen terminated GaN surface to a partially hydrogen terminated surface with increasing
temperature. AlN was observed to grow on GaN in a FM/layer by layer mechanism
throughout the temperature range investigated.
8.3. Experimental
8.3.1. Thin Film Growth and Analysis System
All experiments described below were conducted using a unique ultra high vacuum
(UHV) configuration which integrates several completely independent UHV thin film growth
and analysis systems via a 36 ft. long transfer line having a base pressure of 9x10-10 Torr
(see Ref. 21 for details of the transfer line, and many of the associated systems). The
experiments described in this paper employed the III-V nitride gas source molecular beam
epitaxy (GSMBE), Auger electron spectroscopy (AES), low energy electron diffraction
(LEED), and x-ray photoelectron spectroscopy (XPS) systems.
The GSMBE system with a base pressure of 3x10-10 Torr was designed and
constructed specifically for the growth of III-V nitride thin films. The sample heating stage
consisted of a wound tungsten heating filament positioned close to the back of the sample
268
and mounted on a boron nitride disk [21]. A W/6%Re-W/26%Re thermocouple was
employed to measure the temperature of the backside of the wafer. Heating profiles to
1100°C were easily achieved using a programmable microprocessor and 20 amp SCR power
supply. Actual surface/sample temperatures (i.e. those reported herein) were measured using
an infra-red thermometer with a spectral response of 0.8 to 1.1 µm and a emissivity setting of
0.5. The experimental accuracy for the substrate temperatures was estimated to be ± 25°C.
Source materials in the GSMBE included Al, Ga, and NH3. Al (99.9999%) was
evaporated from a 25 cc "cold lip" Knudsen cell and Ga (99.99999%) was evaporated from a
25 cc dual filament Knudsen cell. The NH3 (99.9995%) was further purified via an inline
purifier connected directly to a leak valve mounted on the GSMBE chamber. Sample
exposure to the NH3 was obtained using "molecular beam" dosers similar to design of Yates
et al [22]. Collimation of the ammonia into a molecular beam focused onto the sample was
achieved with this doser using a 13 mm diameter x 2 mm thick glass capillary array with a
ten micrometer pore size (Galileo Electro Optics Inc.). The doser to sample distance was
fixed at 2". This doser arrangement enhanced the ammonia flux to the sample by a factor of
10-100 relative to the background ammonia flux. Analysis of the ammonia using a
quadrapole residual gas analyzer (RGA) revealed extraneous peaks at 28 and 44 indicating
that CO, N2, and CO2 were the principal contaminants in the gas.
The XPS experiments were performed in a stainless steel UHV chamber (base
pressure = 2x10-10 Torr) equipped with a dual anode (Mg/Al) x-ray and a 100 mm
hemispherical electron energy analyzer (VG CLAM II). All XPS spectra reported herein
were obtained using Mg Kα radiation (hν = 1253.6 eV) using 12 kV and 20 mA emission
current. XPS analysis typically required less than 1 hour during which time the pressure
269
never increased above 9x10-10 Torr. Calibration of the binding energy scale for all scans
was achieved by periodically recording scans of the Au 4f7/2 and Cu 2p3/2 peaks from
standards and correcting for the discrepancies in the measured and known values of these two
peaks (83.98 and 932.67 eV, respectively) [23]. Curve fitting of most data was performed
using the software package GRAMS 386. A combination Gaussian-Lorentzian curve shape
with a linear background was found to best represent the.
The Auger electron spectrometer and the low energy electron diffraction optics were
mounted on a six way cross off the transfer line and pumped through the transfer line. In the
AES analysis, a 3 keV, 1mA beam was used. Each Auger electron spectrum was collected
and numerically differentiated. In LEED an 80 eV, 1mA beam was used.
8.3.2. Substrate and Thin Film Preparation and Analysis
The substrates used in this research were ≈ 1.5x1.5 cm2 pieces cut from 1 3/16"
diameter on and off-axis (4° toward (11-20)), n-type (Nd=1018/cm3) 6H-SiC (0001)Si
wafers obtained from Cree Research, Inc. All wafers were received with an ≈ 1 µm n-type
epitaxial layer (Nd=5x1017) on which was grown ≈ 500-1000Å of thermal oxide. After
removal of the thermal oxide with 10:1 HF, the unpolished back side of each wafer was
subsequently coated via RF sputtering with tungsten to increase the heating efficiency of the
SiC, as the latter is partially transparent to the infrared radiation emitted from the tungsten
filament heater. All wafers were then ultrasonically rinsed in acetone and methanol, exposed
to the vapor from a 10:1 buffered HF solution for 10 min, and mounted using Ta wire to a 1"
270
diameter Mo disk with an approximately 1.5 cm2 square hole cut in the center. Each
wafer/Mo assembly was then fastened to a ring shaped Mo sample holder using Ta wire and
inserted into the transfer line load lock. The in situ procedure used for the final cleaning step
of the 6H-SiC substrates was similar to that described by Kaplan and Kern et al [24,25] and
is described in detail elsewhere [26]. Briefly, each SiC wafer was annealed in the GSMBE
system in a flux of 10-6-10-5 Torr SiH4 for ≈ 15-20 min at 950-1050°C. Analysis via AES
and XPS revealed oxygen-free, silicon terminated SiC surfaces which displayed either (1x1)
or (3x3) LEED patterns. If a (3x3) LEED pattern was obtained (indicative of the formation
of a ˜ bilayer of free silicon on the surface [24,26]), the sample was annealed in UHV at
1050°C for 5-10 min. to desorb the excess silicon. This procedure resulted in a (1x1) LEED
diffraction pattern. Growth of the AlN buffer layer was always initiated on an oxygen free
(1x1) 6H-SiC (0001) surface.
To initiate the deposition of AlN, the 6H-SiC wafer was raised to a temperature of
1050°C at which point the shutter to the Al Kcell (at 1150°C) was opened. A few seconds
later, ammonia was admitted into the system which created a total pressure of ≈ 10-5 Torr.
Growth proceeded at a rate of 1000 Å/hr. for approximately 15 min. after which the Al cell
was shuttered. The sample was allowed to cool in ammonia until approximately 800-900°C
at which point the ammonia valve was closed. The AlN films displayed (2x2) reconstructed
surfaces in LEED immediately after growth. This reconstruction was sensitive to either
contamination or temperature, as a (1x1) LEED pattern was observed several hours after
growth. SEM analysis showed the films to be free of surface topography at 10 kX. The
films were too resistive for electrical measurements. Additional details regarding this
research have been published elsewhere [27,28].
271
To achieve the growth of the GaN films on the AlN buffer layer, the latter was heated
to 650-800°C in 10-4 Torr (˜ 50 sccm) ammonia for 10 min., after which the Ga cell (at
1020°C) was opened and growth allowed to proceed. The growth rate for these conditions
was determined via cross sectional SEM to be ≈ 2000 Å/hr. After the desired GaN thickness
had been achieved, the Ga cell was closed and the GaN film allowed to cool in ammonia to
approximately 600°C after which the ammonia valve was closed. Films of AlN were also
deposited at 800°C and 10-5 Torr NH3 on undoped (0001) GaN films previously grown by
both GSMBE and organometallic vapor phase epitaxy (OMVPE) [29]. Other parameters
were similar to those used in the growth of the AlN buffer layer. The OMVPE GaN films
were cleaned by annealing in 10-4 Torr NH3 at 800°C for 15 min. Auger and x-ray
photoelectron spectroscopies did not detect the presence of any oxygen or carbon [30].
8.3.3. Growth Mode Analysis
The experimental procedure and analysis used to study the growth modes of GaN
(AlN) on AlN (GaN) was similar to that used by Sitar et al. [31] to study the growth modes
of AlN and GaN on (0001)Si 6H-SiC and (0001) Al2O3. Therefore, only a brief description
of the procedure will be given here.
A series of sequential depositions each having a thickness of ≈ 0.5-5Å of GaN (AlN)
on AlN (GaN) were conducted until a GaN (AlN) film thickness of ≈ 35Å was achieved.
Following each deposition, XPS, LEED, and AES analysis were performed. Each series of
depositions and analysis was completed within 12-14 hours to ensure the cleanliness of the 272
surfaces and interfaces. To determine the growth mode of the GaN (AlN) film, the ratio of
the initial integrated intensity of the Al 2p (Ga 3d) core level from the AlN buffer layer
(OMVPE GaN film) (Io) was measured against the integrated intensity of the Al 2p (Ga 3d)
core level from the GaN/AlN interface (Is). Is/Io was plotted against the calculated GaN
thickness (= growth rate x growth time). Theoretical curves for the expected Al 2p (Ga 3d)
attenuation from layer-by-layer (Frank van der Merwe (FM)), layer-by-layer plus island
(Stranski-Krastanov (SK)), and island (Volmer-Weber (VW)) growth modes were
simultaneously plotted and compared with the experimentally determined attenuation to
elucidate the growth mode(s). The following relations were used for FM, SK, and VW
growth modes:
FM: Is/Io = exp (-t/λ) (1)
where:
t = thickness of the growing film
λ = mean free path of the photoelectron being measured
Io = Initial intensity of the substrate core level
Is = Intensity of the substrate core level with a overlying film of thickness t
SK: Is/Io = (1-θ) + exp(-t/λ) (2)
where θ is the surface coverage of the film/islands.
VW: Is/Io = (1-θ)exp(-q/λ) + exp(-t/λ) (3)
where q = thickness of the film before onset of 3D growth 273
The following relation from Briggs and Seah [32] was used to calculate the mean free
paths of the core levels of interest:
λ = 0.41(aE)1/2 + 538E-2 monolayers (4)
E = Kinetic energy of the photoelectron
a = (ρMw/NA)1/3 where ρ = density of overlying film
Mw = Molecular weight of film
NA = Avogadro's Number
From the above relation, the mean free path of Al 2p photoelectrons in GaN was determined
to be 19Å, and the mean free paths of Ga 3d, 3p, and 2p photoelectrons in AlN were
determined to be 19, 18 and 6Å respectively.
8.4. Results
8.4.1. Growth of GaN films.
Figure 8.1 shows a plot of the attenuation of the Al 2p core level as a function of the
overlying GaN film thickness for growth at 650°C. Curves for the expected attenuation for
Frank van der Merwe (FM), Stranski-Krastanov (SK), and Volmer-Weber (VW) growth
modes are also included. The Al 2p core level intensity actually increases after the first few
274
GaN depositions. This effect is believed to be due to forward scattering effects where the
trajectory of a photoelectron emitted by an underlying atom is scattered toward the direction
of the overlying atom due to the positive charge of the nucleus of the latter [33,34]. This can
cause an increase in the intensity of the core levels along certain crystallographic directions
[33]. However, the forward scattering effect diminishes with further depositions, and the
experimental data starts to follow the curve expected for FM growth. However, at a GaN
thickness of 10-12 Å the Al 2p attenuation starts to exhibit a positive deviation again from
the curve for FM growth and approaches the curve expected for SK growth.
Unreconstructed (1x1) LEED patterns were displayed throughout the entire sequence
of GaN depositions at 650°C (see Fig. 8.2(a)). The oxygen levels measured by AES and
XPS throughout the series of 650°C GaN depositions were found to be either undetectable or
less than 1% of a monolayer. The microstructure of the surface of the GaN films grown at
650°C is shown in Figure 8.3(a). These films were n-type and extremely conductive,
exhibiting four point probe sheet resistances of 10-2 ?/sq which is indicative of free carrier
concentrations of 1019/cm3 or greater [35]. Photoluminescence spectra of these films
showed very broad donor bound exciton emission of weak intensity (see Figure 8.4(a)).
275
0.20
0.40
0.60
0.80
1.00
1.20
0 5 10 15 20 25 30
I/Io
I/Io FM
I/Io VW
I/Io SK
I / I
o
GaN Thickness (Å)
Figure 8.1. Attenuation of Al 2p core level from AlN buffer layer as a function of overlying GaN film thickness for T =650°C. Filled circle = experimental I/I , filled diamond = theoretical I/I for Frank van der Merwe layer by layer growth, filled triangles = theoretical I/I for Volmer Weber and Stranski-Krastanov growth.
sub o
o
o
(a) (b) Figure 8.2. LEED diffraction patterns from (a) (1x1) (0001) GaN, and (b) (2x2) (0001) GaN.
276
(a) (b)
(c)
Figure 8.3. SEM micrographs at 10 kX from GaN films grown in GSMBE at (a) 650°C, (b) 750°C, and (c) 800°C.
277
(a)
(b)
Figure 8.4. Photoluminesence (PL) at 4K of NH -GSMBE GaN grown at (a) 650°C and (b) 800°C.
3
278
Figure 8.5 shows a plot of the Al 2p attenuation as a function of GaN thickness for
films grown on AlN at 800°C. The SEM micrograph in Figure 8.3(b) shows a very smooth
surface containing occasional "pits" similar in appearance to those observed in thin OMVPE
GaN surfaces which have been speculated to be due to the incomplete coalescence of flat
GaN islands [29]. The RMS surface roughness of this film as determined by AFM was ˜ 40Å
which is comparable to the < 20Å typically measured from OMVPE GaN surfaces.
0
0.2
0.4
0.6
0.8
1
1.2
0 5 10 15 20 25 30 35 40
I/Io (Al 2p)
I/Io (FM Theory)
o
GaN Thickness (Å)
Forward ScatteringEffects
Figure 8.5. Attenuation of Al 2p core level from AlN buffer layer as a function of overlying GaN film thickness for T =800°C. Filled circle = experimental I/I , filled triangle = theoretical I/I for Frank van der Merwe layer by layer growth.
sub o
o
279
Reconstructed (2x2) LEED patterns were displayed throughout the series of GaN
depositions on AlN at 800°C (see Fig. 8.2(b)). These surfaces were similarly sensitive to
surface contamination and temperature as with the (2x2) reconstructed AlN surface.
Typically, (1x1) LEED patterns developed after 3-4 hours in a vacuum of 10-9 to 10-8 Torr.
The (2x2) reconstructed surfaces could be restored by annealing in 10-5 Torr NH3 for 5 min.
No attempts were made to determine the actual structure of the (2x2) reconstruction.
However, the work of Bernholc et al [36] indicate that a (2x2) N adatom reconstruction is
energetically most favorable for Ga terminated (0001) GaN and a (2x2) N vacancy
reconstruction is energetically favorable for the nitrogen terminated (0001) GaN surface. For
the experiments described heree, gallium termination. For AlN growth on the silicon face of
6H-SiC (0001), one expects the formation of Si-N bonds at the AlN/SiC interface which
inturn implies Al termination of the AlN film. Accordingly, Ga termination should be
expected for GaN growth Al terminated AlN. The oxygen levels detected by both AES and
XPS from GaN films grown at 800°C were similarly found to be undetectable or less than
1%.
The GaN films grown at 800°C were found to be more resistive (0.2-1M?) in contrast
to those grown at 650°C. Sheet resistances for the former were typically too high for four
point probe measurement. The carrier concentrations (ND-NA) of the 800°C films
determined by CV measurements were in the range of 1-5x1017/cm3. GaN films grown at
825°C were more resistive with a carrier concentration of 2-5x1016/cm3. However, Hall
measurements found n = 2x1017/cm3 and µ = 60cm2/Vsec. PL for these films displayed
sharp (4 meV) donor-bound exciton emission and very little D-A emission. (see Figure
8.4(b)). 280
8.4.2. Growth of AlN films on GaN.
A plot of the attenuation of the Ga 3p, 2p, and 3d core levels from an OMVPE GaN
as a function of overlying AlN thickness is shown in Figure 8.6. As can be seen, FM growth
of AlN on OMVPE GaN was observed to occur (similar results were obtained for AlN
growth on GSMBE GaN). No surface topography was observed on the AlN surface via
SEM, but the AlN/GaN was observed to have cracked on occassion. The cause of the
cracking is currently not known. The resistivity of these AlN films were beyond the range of
our experimental capabilities and CV did not indicate any charge.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 5 10 15 20 25 30 35 40
I/Io Ga 3d
3d Theory
I/Io Ga3p
3p Theory
I/I o
AlN Thickness (Å)
Figure 8.6. Attenuation of Ga 3d and 3p core levels from OMVPE GaN as a function of overlying GaN film thickness for Tsub = 800°C. Filled circle = experimental I/I for Ga 3d, filled square = experimental I/I for Ga 3p, empty circle and square = experimental I/I for Frank van der Merwe layer by layer growth.
o
o o
281
8.5. Discussion
8.5.1. GaN on AlN Growth Mechanisms
8.5.1.1 Strain Effects:
The above XPS CL attenuation analysis shown in Figures 1 & 5 clearly illustrates
that the growth of GaN on monocrystalline AlN by NH3-GSMBE initially proceeds in a
FM/layer by layer mechanism for the temperature range investigated. However using XPS
and SEM, we have observed that growth can either continue in a FM/layer by layer fashion
or switch into a three dimensional SK type growth regime. Further inspection of Figure 1
reveals that the deviation from the expected Al 2p attenuation for FM growth and the
experimentally observed Al 2p attenuation occurs at ≈ 10-12Å which is also reported
Matthews-Blakeslee critical thickness for GaN on AlN [37]. This observation is clearly in
line with the classical interpretation of SK type growth where interfacial strain forces 3D
growth.
The differences between the different types of growth modes (FM, SK, and VW) are
typically explained in terms of macroscopic surface and interfacial energies. Classically,
Frank van der Merwe (FM) growth is predicted when the sum of the film surface energy (σf)
and the substrate-film interfacial energy (σi) is less than the substrate surface energy (σs) (i.e.
σf +σi < σs). Volmer-Weber (VW) growth is predicted for the opposite case, i.e. when σf +
σi > σs. Prediction of Stranski-Krastanov (SK) growth requires a more detailed analysis of
the substrate-film interfacial energy, σi. Several different factors are lumped together in σi.
282
The two most important factors contributing to the interfacial energy are: σb, the energy due
to the chemical bonds formed between the two materials at the interface, and σst, the strain
formed at the interface due to imperfect lattice matching. The interfacial energy, σi, can then
be described as σi = σst + σb. Thus, Stranski-Krastanov growth is described by the situation
where the expression σf + σb + σst < σs is satisfied initially at the start of film growth.
However, as growth continues the interfacial strain energy term increases dramatically and
switches the equality of the expression to σf + σb + σst > σs. At this point, the interfacial
stain energy can be lowered by either generation of misfit dislocations (as envisioned by
Matthews and Blakeslee [38-40]) or switching into three dimensional growth (i.e. surface
roughening [41]). Clearly, stress relaxation via switching to 3D growth is clearly in line with
our observations of SK type growth of GaN at 650°C.
In contrast to GaN growth at 650°C, no deviations between the Al 2p attenuation and
theory for FM growth was observed for GaN growth at 800°C. Correspondingly, SEM and
AFM analysis of thicker films (2000-5000Å) revealed smoother surfaces with RMS
roughness ˜ 40 Å consistent with FM/layer by layer growth. Clearly for strain relaxation via
surface roughening to be the cause of SK type GaN growth at 650°C, a different strain
relaxation mechanism must be in operation during GaN growth at 800°C. A different strain
relaxation mechanism at 800°C could occur as a result of a lowering of strain due to better
lattice matching between GaN and AlN at 800°C. However, the thermal expansion
coefficients (c and a) for GaN are both larger than those of AlN, and hence the lattice
matching and strain is actually poorer at 800°C [42].
283
Clearly, an alternative explanation is that the interfacial strain at the GaN/AlN surface
is relieved via misfit dislocation generation. The switch from strain relaxation via surface
roughening to strain relaxation via misfit dislocation would imply that some sort of thermal
activation is required to nucleate a misfit dislocation. Owing to the large moduli of the III-V
nitrides, this is reasonable as the energy to create a dislocation is quite high (i.e. dislocation
energy ≅ Gb2, where G=E/2(1+ν) an EGaN=300 GPa [43]). Therefore, one could
alternatively argue that at 800°C the GaN film grows psuedomorphically on the AlN and that
at the MB critical thickness strain relaxation occurs via misfit dislocation generation rather
than via surface roughening. However, the authors note that the underlying AlN buffer layer
although thin (≈250Å) and monocrystalline is thicker than the critical thickness for AlN on
6H-SiC (≈40Å) and is highly defective with a large density of misfit dislocations which
certainly propagate into the GaN film. Accordingly, there are already plenty of defects at the
GaN/AlN interface to act as low energy nucleation sites for misfits in the GaN film. (In fact
in low strain SixGe1-x alloy growth on Si, misfits are found to preferentially nucleate at
dislocations and other defects at the Si surface [44]).
It is worth noting that a switch from 3D to 2D GaN growth with substrate
temperature for NH3-GSMBE growth of GaN directly on (0001) Al2O3 has additionally
been observed using RHEED by Powell et al [20]. Their RHEED studies showed that films
grown at Tsub < 760°C exhibited well defined transmission spots indicative of three
dimensional island growth and at Tsub > 780°C, RHEED exhibited sharp Kikuchi lines and a
streaky (1x1) patterns indicative of two dimensional growth [20]. As the lattice matching
and interfacial bonding between GaN and Al2O3 are completely different to the case of GaN
on AlN, one would not expect to see a switch from 3D to 2D growth in the same temperature 284
range if strain relaxation mechanisms were responsible for the changes in growth mechanism.
This in turn suggests that surface processes such as adsorption, desorption, and diffusion are
playing an important role in determining the growth mechanism of GaN on AlN in NH3-
GSMBE. In fact, in Si and Ge GSMBE growth using SiH4, Si2H6, and GeH4 a transition
from 3D to 2D growth with increasing substrate temperature has also been observed [45-49].
In this homoepitaxial case, the transition from 3D to 2D growth was explained in terms of
sluggish hydrogen desorption kinetics/sight blocking at low growth temperatures [50]. At
this point, the authors provide a simple model for the interaction of NH3 with GaN surfaces
in MBE and which shows that hydrogen desorption/ammonia decomposition may also be the
determining factor for the NH3-GSMBE GaN growth mechanism on all substrates including
Al2O3, AlN, Si, and GaN.
8.5.1.2. Hydrogen desorption
Based on RGA studies of the surface cracking of ammonia, Kamp et al [14] proposed
the following sequences of reactions for describing the reaction of ammonia with GaN
surfaces:
NH3g ---- NH3ad (i) NH3ad ---- NH2ad + Had (ii) NH2ad +Had ----- NHad + 2Had (iii) NHad + 2Had ------ Nad + 3Had (iv) 2Had --- H2g (v) 2Nad --- N2ad (vi) - No growth
285
Gaad + Nad --- GaNs (vii) - GaN growth
For simplicity, we first start by modeling reaction (v), i.e. adsorption and desorption
of hydrogen from GaN surfaces. Our approach is to estimate the steady state surface
coverage of hydrogen on GaN and AlN surfaces in the presence of an atomic hydrogen flux.
This will allow us to estimate the hydrogen surface coverage during NH3-GSMBE growth in
the case where ammonia decomposition on GaN and AlN surfaces is rapid and not the
limiting reaction. The authors also note that this model is a reasonably accurate description
for the interaction of GaN/AlN surfaces with hydrogen plasmas in which there is known to be
a large concentration of atomic hydrogen [51].
Desorption kinetics are typically described by the general Polyani-Wigner rate
expression [52]:
rate = -dθ/dt =νnnexp(-Edes/RT) (5) where: n = the reaction order θ = the adsorbate surface coverage ν = the pre-exponential factor,νo=1028/cm2 sec,ν1=1013/sec,
ν2=10-2 cm2/sec Edes = the activation energy for desorption
In steady state conditions, the flux of adsorbates leaving the surface via desorption will be
equal to the incoming flux of adsorbate times the adsorbate sticking coefficient. The sticking
coefficient is described by:
S = So(1 - θ/θmax)n (6)
286
The combination of equations 5 and 6 allows the determination of the steady state surface
coverage of the adsorbate [53].
For hydrogen on GaN, desorption of hydrogen can be expected from both Ga and N
sites. Chiang et al [54] using isothermal analysis and direct recoil time of flight (TOF) mass
spectroscopy have recently determined that deuterium desorption from Ga sites on Ar+
sputter cleaned polycrystalline GaN surfaces is second order with an activation energy of 9
kcal/mol. The findings of Chiang et al for H desorption from Ga are similar to those found
for hydrogen desorption from Ga in GaAs where a second order desorption activation energy
of 13 kcal/mol was found [55]. Second order desorption is consistent with the following
mechanism:
2Dads ---- GaD2 -----D2(g) + Ga. (viii)
Chiang et al speculate that the rate limiting reaction in this expression is actually diffusion of
the deuterium to Ga sites and that D2 desorption is actually rapid. Unfortunately, Chiang et
al did not provide any information regarding hydrogen desorption from nitrogen sites except
that hydrogen desorption from nitrogen sites is complete by 600°C. Despite this we have
estimated that hydrogen desorption from nitrogen in GaN is second order with an activation
energy of 55 kcal/mol. This estimation is based on the similar bond energies of hydrogen to
nitrogen and silicon in ammonia and silane (431.8 and 398.3 kJ/mol respectively [56]) and
the determination of second order desorption kinetics of hydrogen from Si (111) with Edes =
57.5 kcal/mol by Schulze and Henzler [57]. (Note: desorption of hydrogen from nitrogen
sites using 1st order desorption kinetics with Edes=45 kcal/mol were also calculated). To the
287
authors knowledge, there have been no reports on Edes for hydrogen desorption specifically
from a nitrogen site to support this assumption. However, hydrogen desorption sites from
polycrystalline MoN with Edes ≅ 50 kcal/mol have been reported by Choi et al [58-60]. For
hydrogen desorption from AlN we considered hydrogen desorption from nitrogen sites in
AlN to be similar to hydrogen desorption from nitrogen sites on GaN surfaces and hence
assumed Edes ≅ 50 kcal/mol. For hydrogen desorption from Al sites on AlN we used the
experimentally determined activation energies of hydrogen desorption from metallic (111) Al
surfaces which were 17 kcal/mol for 0th order desorption kinetics[61,62]. In our
calculations, we assumed an activation energy of 17 kcal/mol and 2nd order desorption
kinetics.
Using Edes(Ga) = 9 kcal/mol, Edes(Al) = 17 kcal/mol, Edes(N) = 55 kcal/mol,ν2=
10-2cm2/sec, So=1 and Ømax = 2.3x1015/cm2, we were able to generate plots of hydrogen
surface coverage on Ga, Al, and N sites of GaN and AlN surfaces as a function of
temperature for various atomic H fluxes (see Figures 8.7, 8.8, 8.9). As can be seen in Figure
8.7, very little hydrogen remains adsorbed to Ga at any temperature except for extremely
large fluxes of atomic H (1024/cm2sec.). However, for H fluxes of 1017-1018/cm2sec (i.e.
typical NH3 MBE conditions) hydrogen is stable on Al sites on AlN to temperatures of
100°C (see Figure 8.9). In contrast, saturation of nitrogen sites with hydrogen is observed to
occur for all fluxes up to temperatures of 400-550°C (see Figure 8.8). These figures also
show that desorption of hydrogen from nitrogen sites is nearly complete at 600°C for low H
fluxes. This is consistent with the observations of Chiang et al. [55] and lends support to our
assumptions for the activation energy for desorption of hydrogen from nitrogen. Further
288
inspection of Figure 8.8, reveals that for atomic H fluxes of 1017-1018/cm2sec. hydrogen
desorption does not become significant until temperatures of 700-800°C. This is of
importance in that H fluxes of this magnitude are comparable to the NH3 fluxes used in MBE
growth of GaN. Therefore, it is not unreasonable to expect similar hydrogen surface
coverages for GaN surfaces exposed to an ammonia flux. In fact, the authors argue that the
simple model we have presented here for the interaction of atomic hydrogen with a nitrogen
site is exactly analogous to the case of H desorption/decomposition of chemisorbed NH3 (i.e.
reactions iii-v).
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
10e1510e1710e1810e2010e2210e24
Ø/Ø
max
Temperature (ÞC)
Edes
= 7 kcal/mol, 2 nd Order
v = 0.02 cm 2/sec
Figure 8.7. Hydrogen surface coverage of Ga sites as a function of temperature and flux.
289
0
0.2
0.4
0.6
0.8
1
300 400 500 600 700 800 900 1000
10e1510e1710e1810e2010e22
Ø/Ø
max
Temperature (ÞC)
Edes
= 50 kcal/mol, 1 st Order
v = 10 13 /sec
Figure 8.8. Hydrogen surface coverage of N sites as a function of temperature and flux.
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
10e1510e1710e1810e2010e2210e24
Ø/Ø
max
Temperature (ÞC)
Edes
= 17 kcal/mol, 2 nd Order
v = 0.02 cm 2/sec
Figure 8.9. Hydrogen surface coverage of Al sites as a function of temperature and flux.
290
Based on the above findings, it now seems reasonable to ascribe the observed change
in GaN growth mechanism from SK to FM with increasing growth temperature as being due
to reactive site blocking by incomplete desorption (decomposition) of hydrogen (ammonia)
from GaN surfaces. At temperatures below 700°C, the GaN surface is so saturated with
hydrogen that the hydrogen effectively blocks reactive sites at steps, kinks, etc. preventing
incorporation of Ga into the GaN lattice. As unincorporated Ga may not be able to desorb
from the GaN surface at these temperatures Ga clusters may form leading to 3D growth
(alternatively, the unincorporated Ga clusters could grow to form Ga droplets which have
also been observed [13,20]). At higher temperatures (Tsub>780°C), hydrogen
desorption/ammonia decomposition is more complete leaving more open reactive sites with
which to incorporate the incoming Ga and thus preventing Ga cluster formation.
Additionally, it is intriguing to note that figure 8.8 shows that for higher atomic H/NH3
fluxes of 1020-1022/cm2sec characteristic of CVD growth, hydrogen saturated surfaces are
maintained up to temperatures of 950-1050°C. Correspondingly, this is the temperature
range in which OMVPE growth of high quality GaN is observed to occur. This suggests that
the need for high temperatures in CVD growth of GaN is not to decompose GaN per se but to
speed the kinetics up to the point that the surface can handle the large fluxes of species
impinging on it.
Figures 8.7 and 8.8 may additionally explain the extremely low activation energies
for Ga desorption from GaN observed by Jones et al [63] during NH3-MBE growth. For
growth in the temperature range of 625-740°C, Jones et al using desorption mass
spectroscopy (DMS) observed the activation energy for Ga desorption was 1.4 eV (32.2
291
kcal/mol). At higher temperatures (740-875°C), a lower activation energy for Ga desorption
of 0.4 eV (9.2 kcal/mol) was observed. Both of these values are extremely low in
comparison to those observed E for Ga desorption from GaAs which are observed to range
from 2.8 eV (64.5 kcal/mol) to 4.9 eV (112 kcal/mol) [64-67] These values correspond to the
activation energies for evaporation of Ga from liquid Ga and sublimation of Ga from GaAs
respectively [64]. Jones et al explained their low values as being to due to the presence of a
"hydrogen terminated liquid Ga pool which is microscopic in size" at low temperatures and a
hydrogen terminated nitrogen site at high temperatures. From Figure 8.7, it obvious that in
typical NH3-MBE growth conditions, hydrogen termination of Ga sites is not likely. Further
at higher temperatures where they observe a change in the activation energy, Figure 8.8
shows that there is very little hydrogen termination of either Ga or N sites. Instead, we
suggest that at high temperatures Ga desorption occurs as GaHx species which would be
expected to have the very low activation energies observed (i.e. Edes for AlHx on Al (111) =
27 kcal/mol). At lower temperatures, Ga desorption is an average of desorption as GaHx and
desorption from Ga droplets (i.e. (2.8 + 0.4)/2 = 1.6 ≈ 1.4 eV).
act
Finally, the authors recognize that several other detailed studies of the growth
modes/mechanisms of GaN and AlN have been previously performed [4-10] Unfortunately,
most of these studies were concerned with growth of AlN and GaN on other substrates such
as Si, Al2O3, and MgO, or low temperature/amorphous OMVPE/CVD GaN/AlN buffer
layers deposited on Al2O3. Further, most of these studies were conducted using other
growth methods such as OMVPE/CVD and ECR-MBE both of which are sufficiently
different from our growth method. With this is consideration, the authors do not feel that
these studies merit further discussion.
292
In summary, it has been shown that the growth mechanism of GaN on AlN is
dependent on growth temperature switching from SK type growth to FM type growth with
increasing temperature. Based on previous observations in Si/Ge GSMBE [45-49], we
provide a model which indicates that the surface kinetics of hydrogen desorption/ammonia
decomposition may be the determining factor in the change of growth kinetics. These
findings indicate that in order for 2D growth of high quality GaN to be achieved at
temperatures lower than Tsub ≈ 780°C, time must be allowed for hydrogen to desorb from
the surface. Hence, migration enhanced or ALE type growth schemes may be necessary for
growth of GaN in this regime.
8.5.2. AlN growth mechanisms on (0001) GaN
As shown in Figure 8.6 AlN growth on GaN is observed to occur in a layer by layer
fashion. This complements the observations by Sitar et al [30] in which it was observed that
ECR-MBE AlN grows in a layer by layer fashion on both Al2O3 and SiC. By the same
method, we have additionally observed 2D FM/layer by layer growth of AlN on (0001) 6H-
SiC by NH3-GSMBE .
8.5.3. Surface Reconstructions
Typically reports of most RHEED analysis of GaN films has consisted of the
observation of either "spotty" or "streaky" unreconstructed (1x1) GaN surfaces. Only, a few
reports of the observation of (2x2) reconstructed GaN surfaces have been made [68-71], and
293
until recently, all such reports of (2x2) reconstructions came from GaN films grown on
(0001) 6H-SiC substrates by either ECR or RF MBE (perhaps indicating the better quality of
epitaxy achieved on these substrates). However, Iwata et al have recently reported the
observation of (2x2) and (4x4) reconstructed GaN surfaces grown on (0001) sapphire using
an improved ECR source [71]. Our observation of (2x2) reconstructed GaN surfaces is the
first such case for GaN films grown via NH3-MBE. In our case, the observation of (2x2)
reconstructions coincided with a corresponding decrease in surface roughness for our GaN
films. Based on this, the authors are skeptical of the claims of layer by layer growth and
"smooth" GaN surfaces obtained by other RF, ECR, and NH3-GSMBE in which (1x1)
RHEED patterns were observed.
8.6 Conclusions
The following conclusions can be drawn:
1.) Growth of GaN at 650°C on AlN occurs via a Stranski-Krastanov (SK) growth
mechanism resulting in a relaxed GaN film.
2.) Growth of GaN at 800°C on AlN occurs via a Frank van der Merwe growth
mechanism resulting in a strained GaN film.
3.) Growth of AlN at 800°C on GaN occurs via a Frank van der Merwe growth
mechanism (but apparently does not result in a strained film).
294
4.) The surface kinetics of hydrogen desorption/ammonia decomposition is the
determining factor in which type of GaN growth mechanism occurs.
8.7. References
1. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B 10 1237 (1992). 2. J.H. Edgar, J. Mater. Res. 7 235 (1992). 3. R.F. Davis, Proc. of IEEE 79 702 (1991). 4. H. Amano, I. Akasaki, K. Hiramatsu, N. Koide, and N. Sawaki, Thin. Solid Films 163 415 (1988). 5. I. Akasaki, H. Amano, Y. Koide, K. Hiramatsu, and N. Sawaki, J. Cryst. Growth 98 209 (1989). 6. W. Qian, M. Skowronski, M. De Graef, K. Doverspike, L.B. Rowland, and D.K. Gaskill, Appl. Phys. Lett. 66 1252 (1995). 7. K. Hiramatsu, S. Itoh, H. Amano, I. Akasaki, N. Kuwano, T. Shiraishi, and K. Oki, J. Cryst. Growth 115 628 (1991). 8. N. Kuwano, T. Shiraishi, A. Koga, K. Oki, K. Hiramatsu, H. Amano, K. Itoh, and I. Akasaki, J. Cryst. Growth 115 281 (1991). 9. K. Wang, D. Pavlidis, J. Singh, J. Appl. Phys. 80 1823 (1996). 10. K. Uchida, A. Watanabe, F. Yano, M. Kouguchi, T. Tanaka, and S. Minagawa, J. Appl. Phys. 79 3487 (1996). 11. Z. Yang, L.K. Li, and W.I. Wang, Appl. Phys. Lett. 67 1686 (1995). 12. L.K. Li, Z. Yang, W.I. Wang, Electronic Letters 31 2127 (1995). 13. W. Kim, O. Aktas, A.E. Botchkarev, A. Salvador, S.N. Mohammad, and H. Morkoc, J. Appl. Phys. 79 7657 (1996). 14. M. Kamp, M. Mayer, A. Pelzmann, A. Thies, H.Y. Chung, H. Sternschulte, O. Marti, and J. Ebeling, 1995 Fall MRS. 15. S. Yoshida, S. Misawa, A. Itoh, Appl. Phys. Lett. 26 461 (1975).
295
16. S. Yoshida, S. Misawa, Y. Fujii, S. Takada, S. Gonda, and A. Itoh, J. Vac. Sci. Technol., 16 990 (1979). 17. S. Yoshida, S. Misawa, and S. Gonda, J. Appl. Phys. 53 6844 (1982). 18. S. Yoshida, S. Misawa, and S. Gonda, Appl. Phys. Lett. 42 (1983). 19. S. Yoshida, S. Misawa, and S. Gonda, J. Vac. Sci. Technol. B 1 250 (1983). 20. R.C. Powell, N.E. Lee, J.E. Greene, Appl. Phys. Lett. 60 2505 (1992). 21. J. van der Weide, Ph.D. Dissertation, NCSU (1994). 22. See for example - M.J. Bozack, L. Muehlhoff, J.N. Russel Jr., W.J. Choyke, and J.T. Yates, Jr., J. Vac. Sci. Technol. A 5 1 (1987). - C.C. Cheng, R.M. Wallace, P.A. Taylor, W.J. Choyke, and J.T. Yates, Jr., J. Appl. Phys. 67 3693 (1990). - A. Winkler and J.T. Yates, Jr., J. Vac. Sci. and Technol. A 6 2929 (1988). - P.L. Hagans, B.M. DeKoven, J.L. Womack, J. Vac. Sci. Technol. A 7 3375 (1989). - C.T. Campbell and S.M. Valone, J. Vac. Sci. Technol. A 3 408 (1985). 23. XPS Handbook, Perkin Elmer. 24. R. Kaplan, Surface Science, 215 111 (1989). 25. R.S. Kern, Ph.D. Dissertation NCSU (1996). 26. S.W. King, M.C. Benjamin, R.J. Nemanich, and R.F. Davis, submitted to J. Electrochem. Soc., 27. S. Tanaka, Ph.D. Dissertation NCSU (1996). 28. A. Aboelfotoh, R.S. Kern, C.I. Harris, and R.F. Davis, Appl. Phys. Lett., 69 2873 (1996). 29. . T.W. Weeks, Jr., M.D. Bremser, K.S. Ailey, E.Carlson, W.G. Perry, E.L. Piner, N.A. El-Masry, and R.F. Davis, J. Mater. Res. 11 1011 (1996). 30. S.W. King, L.L. Smith, J.P. Barnak, Ja-Hum Ku, J.A. Christman, M.C. Benjamin, M.D. Bremser, R.J. Nemanich, and R.F. Davis, Fall MRS, 1996. 31. Z. Sitar, L.L. Smith, and R.F. Davis, J. Cryst. Growth 141 11 (1994). 32. D. Briggs and M.P. Seah, Practical Surface Analysis, 2nd Edition, Vol. 1, Wiley & Sons, New York (1990). 296
33. W.F. Egelhoff, Crit. Rev. in Solid State and Material Sciences 16 213 (1990). 34. S.A. Chambers, Surface Science Reports 16 261 (1992). 35. A. Skfo, R. Thegra, P. Epeap, E. At, J. Irrep. Results, 57 530 (1996). 36. J. Bernholc, Spring MRS Symp. 1996. 37. S. Krishnankutty, R.M. Kolbas, M.A. Khan, J.N. Kuznia, J.M. Van Hove, and D.T. Olson, J. Electron. Mater. 21 437 (1992). 38. J.W. Matthews and A.E. Blakeslee, J. Cryst. Growth 27 118 (1975). 39. J.W. Matthews and A.E. Blakeslee, J. Cryst. Growth 29 273 (1975). 40. J.W. Matthews and A.E. Blakeslee, J. Cryst. Growth 32 265 (1975). 41. A.G. Cullis, MRS Bulletin 21 21 (1996). 42. H.P. Maruska and J.J. Tietjen, Appl. Phys. Lett., 15 327 (1969). 43. I. Akasaki and H. Amano, in Properties of Group III Nitrides, J.H. Edgar, Ed., Inspec., London (1994). 44. F.K. LeGoues, MRS Bulletin 21 38 (1996). 45. S.M. Mokler, W.K. Liu, N. Ohtani, and B.A. Joyce, Appl. Phys. Lett. 59 3419 (1991). 46. S.M. Mokler, W.K. Liu, N. Ohtani, and B.A. Joyce, Appl. Surf. Sci. 60/61 92 (1992). 47. S.M. Mokler, W.K. Liu, N. Ohtani, and B.A. Joyce, J. Cryst. Growth 120 290 (1992). 48. K. Sakamoto, H. Matsuhata, K. Miki, T. Sakamoto, J. Cryst. Growth 157 295 (1995). 49. L. Chen, T. Chou, W. Tsai, G. Huang, H. Tseng, H. Lin, and C. Chang, Jpn. J. Appl. Phys. 34 L869 (1995). 50. S. Gates, and S. Kulkarni, Appl. Phys. Lett. 60 53 (1992). 51. J. Barnack, Foo Foo, and G. Rabbit, Ph.D. Dissertation, NCSU (1997). 52. D.H. Parker, M.E. Jones, and B.E. Koel, Surface Science 233 65 (1990). 53. M. Schulberg, M. Allendorf, and D. Outka, Surface Science 341 262 (1995).
297
54. C.M. Chiang, S.M. Gates, A. Bensaoula, J.A. Schultz, Chem. Phys. Letters 246 275 (1995). 55. W. Mokwa, D. Kohl, and G. Heiland, Phys. Rev. B 29 6709 (1984). 56. Concepts and Models of Inorganic Chemistry, B. Douglas, D. H. McDaniel, and J.J. Alexander, eds. John Wiley & Sons, New York pg. 78 (1983). 57. G. Schulze and M. Henzler, Surface Science 124 336 (1983). 58. J. Choi, H. Lee, L. Thompson, Appl. Surf. Sci., 78 299 (1994). 59. H. Lee, J. Choi, C. Colling, M. Mudholkar, and L. Thompson, Appl. Surf. Sci., 89 121 (1995). 60. C. Colling, J. Choi, and L. Thompson, J. Catalylsis, 160 35 (1996). 61. A. Winkler, Ch. Resch, and K.D. Rendulic, J. Chem. Phys. 95 7682 (1991). 62. V. Zhukov, A. Ferstl, A. Winkler, K.D. Rendulic, Chem. Phys. Lett. 222 481 (1994). 63. C.R. Jones, Ting Lei, R. Kaspi, and K.R. Evans, 1995 Fall MRS. 64. J.P. Reithmaier, R.F. Broom, and H.P. Meier, Appl. Phys. Lett. 61 1222 (1992). 65. A.H. Kean, C.R. Stanley, M.C. Holland, J.L. Martin, and J.N. Chapman, J. Cryst. Growth 111 189 (1991). 66. N. Sugiyama, T. Isu, and Y. Katayama, Jpn. J. Appl. Phys. 28 L287 (1989). 67. R.P. Burns, K.A. Gabriel, and D.E. Pierce, J. Am. Ceramic Soc. 76 273 (1993). 68. W. Hughes, W. Rowland, Jr., M. Johnson, S. Fujita, J. Cook, J. Schetzina, J. Ren, and J. Edmond, J. Vac. Sci. Technol. B 13 1571 (1995). 69. K. Iwata, J. Asahi, S.J. Yu, K. Asami, H. Fujita, M. Fushida, and S. Gonda, Jpn. J. Appl. Phys. 35 L289 (1996). 70. M.E. Lin, S. Strite, A. Agarwal, A. Salvador, G.L. Zhou, N. Teraguchi, A. Rockett, and H. Morkoc, Appl. Phys. Lett. 62 (1993) 702. 71. H. Liu, A.C. Frenkel, J.G. Kim, and R.M. Park, J. Appl. Phys. 74 6124 (1993).
298
9. XPS MEASUREMENT OF THE SiC/AlN BAND-OFFSET AT THE
(0001) INTERFACE
Presented at the 1995 Fall MRS Conference
III-V Nitride Symposium
by
Sean W. King , Mark C. Benjamin , Robert J. Nemanich , Robert F. Davis , * ** ** *
and Walter R.L. Lambrecht†
*Department of Materials Science and Engineering
**Department of Physics
North Carolina State University
Raleigh NC 27695
†Department of Physics
Case Western Reserve University
Cleveland, OH 44106
299
9.1 Abstract
X-ray photoelectron spectroscopy has been used to determine the band-offset at the
SiC/AlN (0001) interface. First, the valence band spectra were determined for the bulk
materials and analyzed with the help of calculated densities of states. Core levels were then
measured across the interface for a thin film of 2H-AlN on 6H-SiC which allowed the
determination of a band offset of 1.4 ± 0.3 eV. The analysis of the discrepancies between
measured peak positions and densities of states obtained within the local density
approximation provides information on self energy corrections in good agreement with
independent agreement with independent calculations of the latter.
9.2. Introduction
Silicon carbide wafers are being used increasingly as substrates for the growth of III-
V nitride thin films. In particular, SiC is rather closely lattice matched to AlN (?a/a = 0.9%)
which is often used as a buffer layer for GaN growth. The availability of bulk 6H-SiC
substrate wafers of high quality is instrumental for this purpose. Since SiC can also be grown
on AlN layers on SiC [1], one may also consider the use of SiC as an active quantum well
layer in a AlN/SiC/AlN heterostructure device. From both points of view, the band-offset at
the SiC/AlN interface is of obvious interest. To date, only two previous values are available:
a theoretical value by Lambrecht and Segall [2] which was for the (110) interface between
zincblende SiC and AlN; and an experimental value obtained indirectly from measurements
o
300
of the Fermi level of 2H-AlN grown on 6H-SiC (0001) by Benjamin et al. [3]. The
investigations described here provide a more direct experimental determination of the band
offset at the basal plane interface between 6H-SiC and wurtzite AlN. The procedure consists
of measuring the core levels at the interface between a thin film of AlN (0001) grown on a
6H-SiC (0001) substrate and separately determining the energy of the valence band edges
with respect to the core levels for thick films. Calculated densities of states were used to aid
in the determination of the valence band edge and allowed us to obtain additional information
on the electronic structure of the materials. In particular, we obtained results for the
difference in the quasi-particle self energy shifts of the N 2s and C 2s bands with respect to
those of the upper N 2p and C 2p like valence bands.
9.3. Experiment
A unique and integrated ultra high vacuum (UHV) system consisting of a 36 ft. long
UHV transfer line to which several thin film deposition and surface analysis units were
connected was employed in this research. The details of this integrated system have been
previously described [4]. The as-received, n-type (N ˜ 10 /cm ), vicinal 6H-SiC (0001)
substrate wafers containing a one micron thick, n-type (N ˜ 10 /cm ) 6H-SiC (0001)
epitaxial layer were sequentially dipped in 10% HF for 5-10 min. to remove the thermally
grown 750Å silicon oxide surface layer, rinsed in 18.4 M? de-ionized water, blown dry with
N , mounted to a molybdenum sample holder, loaded into ultra high vacuum (UHV) and
degassed at 250, 500, 700, and 900°C for 30 min. each and annealed in a 10 -10 Torr flux
of silane at 950°C for ˜ 20 min. X-ray photoelectron spectroscopy (XPS) and Auger electron
spectroscopy (AES) analyses of the SiC surface revealed that oxygen and non-carbidic
17-18 3d Si
16-17 3
-7
d Si
2
-6
301
carbon had been removed below the detection limits of these instruments. LEED displayed a
sharp (1x1) pattern. The SiC bulk core levels and valence band of the SiC were measured via
XPS.
Each AlN film was subsequently grown via gas-source molecular beam epitaxy
(GSMBE), at 700°C and 10-6 Torr total pressure using ULSI (99.9995%) NH and a flux of
high purity Al (99.999%) evaporated from a Knudsen cell at 1050°C as sources. The
temperature of 700°C was chosen to minimize any reaction between the SiC substrate and the
AlN. In order to prevent the formation of Si N at the SiC/AlN interface, the SiC wafer was
exposed to the Al flux for 5 min. at 700°C prior to the introduction of NH into the system.
Very thin films (10-20Å) were deposited to investigate the AlN/SiC heterojunction/interface.
Thicker films (200Å) were then deposited to measure the bulk AlN core levels and valence
band. The films were then subsequently transferred within a UHV environment to the
chambers containing the XPS, AES, and LEED units for analyses of the surface chemistry
and structure. Analysis via AES and XPS indicated that the films were stoichiometric and
contained < 5% ML of surface oxygen. LEED displayed a sharp (1x1) pattern. Further
details of the growth and cleaning procedures are described elsewhere [5]. All XPS analysis
was performed using the Al anode (hν = 1486.6 eV) at 20 mA, and 12 kV (240W). Due to
the inherently poor signal/noise ratio in XPS valence band spectra, 50 or more scans of this
region were acquired and summed together. All AES spectra were taken using a beam
voltage of 3 keV and an emission current of 1 mA. LEED was performed using rear view
optics, a beam voltage of ˜ 100 eV, and emission current of 1 mA. Calibration of the XPS
binding energy scale was performed by measuring the position of the Au 4f and shifting the
spectrums such that the peak position occurred at 83.98 eV.
3
3 4
3
7/2
302
9.4. Theory
The densities of states (DOS) used for the analysis of the valence band spectra were
calculated using the linear muffin-tin orbital [6] and density functional methods in the local
density approximation (LDA) [7]. It is important to realize that strictly speaking the band
structures obtained in this theory are not corresponding to the energies for extracting an
electron from the material as measured in photoemission. The latter are quasiparticle
energies and differ from the LDA Kohn-Sham eigenvalues by a self energy correction [8].
This is, among other things, responsible for the well known underestimate of the bandgaps by
the LDA. While these corrections have been found to be rather insensitive to the structure
[9], they are expected to depend on the amount of localization of the states involved. We will
show below that these corrections shift the C 2s and N 2s bands from the LDA calculated
positions with respect to the valence band edge. The available calculations of these
corrections using the GW method (i.e. using the leading term in Hedin's many body theory
[10] with G the one-electron Green's function and W the screened Coulomb interaction) for
SiC [11-13] and AlN [14] show that they are about constant (but not quite, see below) over
the upper valence band but are discontinuous across the ionicity gap. The present
comparison between LDA calculated DOS and measured valence band spectra confirms this
picture, and can be used to obtain an experimental value for these self-energy corrections.
9.5. Results
A comparison of the valence band spectra for 6H-SiC (0001) measured from the
substrate and the calculated DOS is presented in Figure 9.1. The latter is shown unbroadened
Si
303
as obtained from the highly accurate tetrahedron method and with a Gaussian broadening by
0.5 eV. The calculated and measured DOS are aligned to each other in the upper valence
band region. The major atomic orbital character of these peaks is indicated. The reference
level in these spectra is based on the Au 4f standard and is thus not directly related to any
intrinsic materials property of SiC. Thus only relative energy differences are meaningful.
The major reason for using the comparison to calculated DOS is that this allows for a more
precise determination of the actual valence band edge. First, we note that good agreement is
obtained between theoretical and calculated peak positions to about 10 eV binding energy.
The peak intensities of the spectra are influenced by matrix elements and details of the
experimental set-up such as collection solid angle and emission angle with respect to the
surface normal, not accounted for by the DOS. These intensities also depend slightly on the
background subtraction procedure. Here a linear background subtraction was used. The
recent discovery of the presence of surface states in the SiC band gap may also confuse the
direct location of the valence band maximum [15].
7/2
304
Figure 9.1. XPS spectra (arbitrary units) and theoretical densities of states (in states per unit cell per eV) of 6H-SiC.
The major requirement for an accurate band-offset determination is E - E where E
is the energy of the valence band maximum (VBM) and E is any core level of SiC.
However, the broadening near the valence band edge hinders an unambiguous determination
of this edge. In this case, it is preferable to determine the energy separation of the core levels
from the well defined C 2p-Si 3s peak and take the position of the VBM with respect to that
peak from the calculation. If we make use of the calculated GW corrections, the alignment
can be done even more accurately as explained below.
v c v
c
On closer inspection, it has been observed that the experimental C 2s peak is shifted
by about 1.0 eV to lower energy from its theoretical position and the C 2p-Si 3s peak is
shifted by about 0.4 eV. The reason for aligning the spectra in this manner is precisely the
self energy effects mentioned above. Indeed, the GW calculations by Rohlfing et al. [11] and
305
Backes et al. [12] predict a 0.4 eV shift from the valence band for the X3v eigenvalue in 3C-
SiC, which in that case, is closely associated with the C 2p-Si 3s peak. Wenzien et al. [13]
obtain only a 0.2 eV for this shift, so that this must be considered the approximate
uncertainty for this alignment procedure. The former calculations reveal that the C 2s band
self-energy correction is approximately 1.0 eV larger than that of the valence band maximum
while the latter shows this correction to be 1.4-1.6 eV. The present measurement indicates
that the former two are in better agreement with experiment. In summary then, we find that
the valence band maximum on the energy scale of Figure 9.1 lies at 2.2 ± 0.2 eV. We also
see that this agrees well with a simple straight line extrapolation from the half height point of
the experimental valence band edge.
Figure 9.2 shows a similar analysis for the valence-band spectrum of AlN. In this
case, we note that the upper valence band width is only 6 eV wide (in the theoretical DOS).
Rubio et al's [14] GW calculations predict a 2.0 eV shift for the N 2s peak and a 0.6 eV shift
for the N 2p-Al 3s like peak. We see that if we align one of these including the above
correction, the shift for the other is well reproduced. Due to the insulating nature of AlN,
some shifting of the AlN core levels and valence band spectra may expected due to charging.
However, the AlN was deposited thin enough (200Å) such that electrons tunneling from the
conducting SiC substrate into the AlN should minimize and make negligible any charging
affects. For these reasons, our indirect approach of comparing experimental valence band
data with the theoretical DOS is justified. With respect to the same Au 4f based reference
level, we then find the valence-maximum of AlN lies at 4.1 eV and is again in good
agreement with a direct straight line extrapolation of the edge. The core levels for the bulk
and interface system shown in Table 9.1 have also been measured on the same energy scale
as employed for the studies described above.
7/2
306
Figure 9.2. XPS spectra (arbitrary units) and densities of states (in states per unit cell per eV) of 2H-AlN
These data are then substituted into the expression
-?E =(E - E ) - (E - E ) + (E - E ) (1) SiC v v c
SiC AlN AlN SiC b v c b c c
AlNi
where the subscripts b and i indicate bulk and interface respectively. While in the above, E
and E are all positive electron binding energies, it is customary to give ?E in terms of the
energy levels which are the negative of the binding energies. Hence the minus sign on the
left of Eq.1. Using different core levels and remembering the uncertainties in the alignment
of each valence band spectrum we finally arrive at a value of ?E = 1.4 ± 0.3 eV.
v
c v
v
307
Table 9.1. Valence-band maxima and core levels measured on the same Au 4f7/2 based reference scale.
bulk SiC Si 2p 101.3 C 1s 283.5 VBM 2.2 bulk AlN Al 2p 75.3 N 1s 398.5 VBM 4.1 AlN/SiC heterojunction Si 2p 101.5 C 1s 283.6 Al 2p 74.9 N 1s 398.2
9.6. Discussion
The value obtained for the band-offset is in quite good agreement with the previously
calculated offset of 1.5 eV for the (110) zincblende interface [2]. This is perhaps somewhat
surprising since the latter is a non-polar interface while here we deal with a polar
heterovalent interface. In fact, from simple electron counting rules, one expects that a purely
N terminated surface would have an excess of 1/4 electron and thus must reconstruct its
surface for example by having one N vacancy every 4 N atoms in order to maintain charge
neutrality. In reality one may have a missing dimer every 4 instead of a simple vacancy or
any other arrangement which is equivalent in net charge. At present, it is not known on an
308
atomic scale what the interface structure is like, but we may note that if 1/4 of the N are
missing at the interface, this is for electron counting purposes equivalent to mixing the
nitrogen layer with C (group IV) anions. One expects that this would lower the dipole from
that of a non-polar interface [16] by a few 0.1 eV. However, a slightly larger degree of
intermixing may completely wipe out this interface dependence. This is probably indirectly
indicative that there are an equal number of Al-C and N-Si bonds at the interface.
The value of 1.4 eV obtained for the SiC/AlN band offset is larger than the
previously reported experimental value of 0.8 eV [3]. The discrepancies between these two
values may be related to the experimental techniques or to the preparation of the SiC/AlN
interface. The experiments described in this report employed recently developed surface
preparation processes that result in atomically clean and ordered SiC prior to AlN deposition.
In contrast, the SiC surface preparation of the prior study would typically exhibit a small
amount of oxygen at the interface ˜ 25-50% ML. In addition, the gas source MBE employed
in the present study results in a higher quality interface as opposed to the ECR technique
employed previously. The ECR technique has been shown to result in more damage and an
excess of Si-N bonding at the interface [17]. In the study presented in this paper, no oxygen
was detected at the SiC/AlN interface. Furthermore, the use of an initial Al flux prior to
ammonia exposure avoids the formation of a large amount of Si-N bonding. These factors
can strongly influence the band alignment between two semiconductors. In addition, the
initial study also assumed flat bands in the SiC near the interface while upward band bending
was noted as a distinct possibility. Since that band-offset value was based on the assumption
of alignment of the measured Fermi level of AlN and the bulk n-type doped SiC, it indeed
probes the macroscopic band alignment (affected by band bending) rather than the offset in
309
the immediate vicinity of the interface. This differs from the present XPS investigation
because of the limited escape depth of the photoelectrons.
9.7 Conclusion
By combining XPS studies of the valence band spectra with calculated DOS and a
careful analysis of the alignment between the two, taking into account known self-energy
corrections to the LDA band structures, the position of the valence band maxima of the 6H-
SiC (0001) and 2H-AlN (0001) with respect to their core levels has been determined. A
subsequent measurement of core levels at the heterojunction between a thin film of 2H-AlN
grown on top of 6H-SiC then allowed us to extract a band offset of 1.4 ± 0.3 eV. The latter is
in good agreement with the calculated value of the (110) zincblende SiC/AlN which indicates
that polar interface specific effects were averaged out by atomic level reconstructions of the
interface leading to an equal amount of Si-N and C-Al interface bonds.
9.8 Acknowledgments
The work at CWRU was supported by NSF (DMR-92-22387): the research at NCSU
was supported by ONR (NOOO14-92-J-1477). Appreciation is expressed to Cree Research
Inc. for the 6H-SiC wafers used in this study.
9.9 Addendum
310
Since the presentation of this paper, several additional experiments have been
conducted to examine interface effects on the band alignment between 2H-AlN and 6H-SiC.
As we have previously determined the position of the SiC core levels relative to the 6H-SiC
VBM and the AlN core levels relative to the AlN VBM, measurement of the valence band
discontinuity for various different interface conditions requires only measuring the difference
between the AlN core levels and the SiC core levels at the SiC/AlN interface. Table 9.2
summarizes the Si2p-Al2p results obtained from thin AlN films grown on 6H-SiC substrates
under various conditions. In these experiments, both on and off axis 6H-SiC substrates were
used, and both Al and NH3 pre-exposures were used to initiate growth of AlN on the SiC
substrates. For consistency, all the measurements reported here were taken from 20Å
AlN/6H-SiC interfaces. As this table displays, values for Si 2p - Al 2p were observed to
range from ˜ 26.5 - 27.0 eV. This in turn implies a variation in the 2H-AlN/6H-SiC valence
band discontinuity of ˜ 0.5 eV. This is somewhat consistent with the recent first principle
calculations by Majewski et al [18] which showed a variation in the valence band
discontinuity for the (001) SiC/AlN interface of 0.9 eV (?Ev = 1.5 - 2.4 eV). This variation
was found to correspond to differences in the bonding at the SiC/AlN interface with the
purely mixed C/N interfaces resulting in a discontinuity of 1.5 eV and a mixed Al/Si interface
resulting in a discontinuity of 2.4 eV. Unfortunately, Table 8.2 does not display a clear trend
between the substrate orientation, surface reconstruction, or pre-growth orientation.
Additionally, our calculations indicate the valence band discontinuity to range from 0.9-1.4
eV.
311
Table 9.2. Various measurements of Si2p-Al2p from a AlN/6H-SiC interface for different substrates, surface reconstructions, and pre-growth treatments. Orientation Reconstr. Al/NH3 Tsub(TC) Si2p-Al2p (0001), off (1x1) Al 950°C 26.6 (0001), off (1x1) Al 1250°C 26.6 (0001), on, with epi (1x1) Al 950°C 26.6 (0001), off (1x1) Al 950°C 26.8 (0001), off (1x1) Al 950°C 26.95 (0001), off (1x1) Al 1350°C 26.96 (0001), on, wit epi (1x1) Al 1400°C 26.95 (0001), off (1x1) NH3 1250°C 26.45 (0001), on (3x3) NH3 1250°C 27.3 (0001), on (3x3) NH3 1250°C 26.5 (0001), off (vx3v3)R30° NH3 1250°C 26.7 (0001), off (1x1) NH3 1250°C 26.7 (0001), off (3x3) NH3 1250°C 26.9 (0001), on, with epi (v3xv3)R30° NH3 1250°C 26.7 (000-1), on (1x1), Si NH3 1250°C 26.5 (000-1), on (1x1), graphiteNH3 1250°C 26.8 (1-100) (1x1) NH3 1250°C 27.0 (1-100) (1x1) Al 1250°C 27.0
9.10 References
1. L.B. Rowland, R.S. Kern, S. Tanaka, and R.F. Davis, Appl. Phys. Lett. 62, 3333 (1993). 2. W.R.L. Lambrecht and B. Segall, Phys. Rev. B 43, 7070 (1991). 3. M.C. Benjamin, C. Wang, R.F. Davis, and R.J. Nemanich, Appl. Phys. Lett., 64, 3288 (1994). 4. Jacob van der Weide, Ph.D. dissertation, North Carolina State University, 1993. 5. S.W. King, R.J. Nemanich, and R.F. Davis, unpublished. 6. O.K. Andersen, O. Jepsen, M. Sob, in Electronic Band Structure and its
312
Applications, ed. M. Yussouf (Springer, Heidelberg 1987), p. 1. 7. W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965). 8. L.J. Sham and W. Kohn, Phys. Rev. 145, A561 (1966). 9. W.R.L. Lambrecht, B. Segall, M. Yoganathan, W. Suttrop, R.P. Devaty, W.J. Choyke, J.A. Edmond, J.A. Powell, and M. Alouani, Phys. Rev. B 50 10722 (1994). 10. L. Hedin, Phys. Rev. 139, A796 (1965). 11. M. Rohlfing, P. Kruger, J. Pollmann Phys. Rev. B 48, 1791 (1993). 12. W.H. Backes, P.A. Bobbert, W. van Haeringen Phys. Rev. B 51, 4950 (1995). 13. B. Wenzien, P. Kackell, F. Bechstedt Phys. Rev. B 52, (1995). 14. A. Rubio, J.L. Corkill, M.L. Cohen, E.L. Shirley, and S.G. Louie, Phys. Rev. B 48, 11810 (1993). 15. M.C. Benjamin, S.W. King, R.F. Davis, and R.J. Nemanich, unpublished. 16. W.R.L. Lambrecht and B. Segall, Phys. Rev. B 41, 2832 (1990). 17. Z. Sitar, L.L. Smith, and R.F. Davis, J. Cryst. Growth 141, 11 (1994). 18. J.A. Majewski, M. Stadele, and P. Vogl, to be published in the proceedings of the Fall 1996 Materials Research Society.
313
10. Dependence of (0001) GaN/AlN Valence Band Discontinuity on
Surface Reconstruction
and Growth Temperature
Presented at the Fall 1996 MRS Conference
in the III-V Nitride Symposium
by
Sean W. King, Mark C. Benjamin*, Robert J. Nemanich*, and Robert F. Davis
Department of Materials Science and Engineering
*Department of Physics
North Carolina State University
Raleigh NC 27695
314
10.1 Abstract
X-ray and ultraviolet photoelectron spectroscopies have been used to determine the
heterojunction valence band discontinuity at the (0001) GaN/AlN interface. Discontinuity
values of 0.6±0.2 eV were determined for GaN grown on AlN at 650°C, and 0.9±0.2 eV for
GaN grown on AlN at 800°C.
10.2. Introduction
The semiconductor compounds of GaN and AlN are completely miscible with band
gaps of 3.40 and 6.2 eV, respectively. Many potential applications based on heterostructures
and bandgap engineering of these two materials and their alloys have been recently realized
including UV-visible/ optoelectronics and high-frequency devices [1-3]. Charge transport
and quantum confinement are a few of the many important parameters which can affect these
devices [4]. Therefore, reliable knowledge of the valence band discontinuity, ?Ev for the
GaN/AlN interface is extremely important to the advancement of III-V nitride technology.
Accordingly, several authors using a variety of characterization techniques have reported ?Ev
values for GaN/AlN heterojunctions fabricated by different growth techniques including
ECR-MBE [5-7], NH3-Gas Source-MBE [8], and OMVPE [9,10] on Al2O3 [9,10] and 6H-
SiC [5-8] substrates. Unfortunately, the electrical, optical, and microstructural characteristics
315
of these materials and the associated interfaces prepared/deposited by the various techniques
are quite different, and the corresponding values reported for ?Ev range from 0.5 eV [9] to
1.4 eV [7,8] (see Table 10.1). In a separate paper [11], the authors have shown that for GaN
grown via NH3-gas source MBE on a high temperature monocrystalline AlN buffer layer, a
3D to 2D growth transition can be observed with a corresponding change in the electrical,
optical, and microstructural properties of the resulting GaN films. In this paper, we show that
the valence band discontinuity for the (0001) GaN/AlN interface also changes with this
transition.
Table 10.1. Published data for GaN/AlN ?Ev.
?Ev GaN/AlN Orientation Authors Technique 0.8±0.3 eV (0001) - G. Martin et al, [5,6] - XPS 0.5 eV (0001) - J. Baur et al, [9,10] - Metal Impurities 1.36±0.07 eV (0001) - Waldrop & Grant, [8] - XPS 1.4 eV (0001) - Z. Sitar et al, [7] - CL/PL 0.65 eV (0001) - M. Wang, et a; [13] - Au SBH/Theory 0.85 eV (110) - Alabensi et al, [14] - Theory 0.44-0.75 eV (110) - Bernholc, et al [15] - Theory 0.48-1.12 eV (0001) - Ke, et al [16] - Theory 0.75 eV (001) - " " - Theory 0.81 eV (110) - " " - Theory 0.77 eV (111) - " " - Theory
10.3. Experimental
10.3.1. Thin Film Growth and Analysis
316
All experiments described below utilized a UHV system which integrates several
completely independent UHV thin film growth and analysis units via a 36 ft. long transfer
line having a base pressure of 9x10-10 Torr [12]. The experiments described in this paper
employed only the III-N GSMBE, XPS/UPS, and LEED units [11]. The details of our
sample preparation and growth have been described elsewhere [11] and, therefore, only a
brief description of these components of the research will be given here.
The GSMBE system (base pressure = 3x10-10) Torr was designed and constructed
specifically for the growth of III-V nitride thin films. Source materials in the GSMBE
included Al (99.9999%), Ga (99.99999%), and NH3 (99.9995%). Al and Ga were
evaporated from 25 cc "cold lip" and dual filament Knudsen cells respectively. The XPS and
UPS experiments were performed in a UHV chamber (base pressure = 2x10-10 Torr) and
equipped with a dual anode (Mg/Al) x-ray source, a differentially pumped helium resonance
UV lamp, and a 100 mm hemispherical electron energy analyzer (VG CLAM II). All XPS
spectra were obtained using Mg Kα radiation (hν = 1253.6 eV) at 12 kV and 20 mA
emission current. A combination Gaussian-Lorentzian curve shape with a linear background
best represented the XPS data. All UPS spectra were acquired using the unmonochromated
He I line (hν = 21.2 eV) from the UV lamp. The LEED patterns were obtained using an 80
eV, 1 mA beam.
317
The substrates used in this research were ≈ 1.5x1.5 cm2 pieces cut from 1 3/16 in.
diameter off-axis (4° toward (11-20)) n-type (Nd=1018/cm3) 6H-SiC (0001)Si wafers
obtained from Cree Research, Inc. All wafers were received with an ≈ 1 µm n-type epitaxial
layer (Nd=5x1017cm3) on which was grown ≈ 500-1000Å of thermal oxide. They were
ultrasonically and sequentially rinsed in trichloroethylene, acetone, and methanol, dipped in
10:1 buffered HF for 10 min., and mounted to a Mo sample holder using Ta wire. An
opaque W coating was deposited via RF sputtering on the unpolished side of each wafer
which allowed substrate temperatures of 1100°C to be easily achieved with a tungsten
filament heater. The in situ procedure used for the final cleaning step of the 6H-SiC
substrates was similar to that described by Kaplan and Kern [17,18]. Briefly, each SiC wafer
was annealed in the GSMBE system in a flux of 10-6-10-5 Torr SiH4 for ≈ 15-20 min at 950-
1050°C. Analysis via AES and XPS revealed oxygen-free, silicon terminated SiC surfaces
which displayed (1x1) LEED patterns.
An approximately 250Å monocrystalline AlN film grown at 1050°C in ≈ 10-5 Torr
NH3 on each (0001)Si 6H-SiC wafer was used as the buffer layer for growth of GaN. The
AlN films/buffer layers displayed (2x2) reconstructed surfaces in LEED immediately after
growth. This reconstruction was sensitive to either contamination or temperature, as a (1x1)
LEED pattern was observed several hours after growth. To achieve the growth of the GaN
films on the AlN buffer layer, the latter was heated to 650-800°C in 10-4 Torr (˜ 50 sccm)
ammonia, after which the Ga cell was opened and growth allowed to proceed. After the
desired GaN thickness had been achieved, the Ga cell was closed and the GaN film allowed
to cool in ammonia to approximately 600°C after which the ammonia valve was closed.
10.3.2. GaN/AlN ?Ev analysis
The method used in this research for calculating the GaN/AlN valence band discontinuity
was similar to that of Grant and Waldrop [8,20]. The basic scheme of this approach is to
318
reference the valence band maximum energy to a core level energy from each semiconductor
and then use the measured difference between the two core level energies from a junction
between the two semiconductors to indirectly determine the discontinuity. Specifically, the
position of one core level (CL) from the substrate (AlN) is measured with respect to the
substrate valence band maximum (VBM) i.e. (VBM-CL)AlNbulk. Subsequently, a thin layer
(≈15-20Å) of the second semiconductor (GaN) is deposited on the substrate and the
difference between the substrate and film core levels are measured, i.e. (CLAlN -
CLGaN)interface. Finally, the thickness of the overlying film is increased beyond the
sampling depth of XPS (≈ 250Å) and the CL-to-VBM energy is measured for the film, i.e.
(VBM-CL)GaNbulk. The valence band discontinuity between the two semiconductors is
then given as:
-?Ev(GaN/AlN)= (VBM-CL)AlNbulk - (VBM-CL)GaNbulk + (CLAlN-CLGaN)int. (1)
In the measurements by Grant & Waldrop [8] and Martin et al [5,6], XPS was used to
determine both the core level and valence band maxima energies. In the measurements
described herein, UPS was used to determine the VBM of AlN and GaN due to the increased
signal to noise ratio
(S/N) in the UPS VB spectra. Core level energies were measured via XPS.
319
0 2 4 6 8 10 12Binding Energy (eV)
He I VBM = 3.5 eVHe Iß
UPSx 10
Figure 10.1. UPS Spectra of (2x2) (0001) AlN surface.
10.4. Results
Determination of the energy of the AlN VBM was the most difficult aspect in the
measurement of the GaN/AlN valence band discontinuity. The AlN VBM position in the
UPS data was complicated by the presence of artifacts created by significant amounts of
emission from He Iß radiation (see Figure 10.1). Location of the VBM in the UPS VB
spectra was determined by extrapolating a straight line through the leading edge of the
spectra to the energy axis. A value of 71.5±0.1 eV was determined for Al2p-VBMAlN after
320
analysis of several AlN films. Due to complications with He Iß emission, Al2p-VBMAlN
was also determined using XPS VB spectra. In this case, a lower value of 71.3±0.3 eV was
found for Al2p-VBMAlN (data not shown). Therefore, in these studies, a value for Al2p-
VBMAlN = 71.4 ±0.2 eV was used (see Table 10.2). This number is slightly higher than our
previously reported value of 71.2±0.3 eV determined by aligning the theoretical valence band
density of states (VBDOS) of Lambrecht et al. [21] to the AlN VB XPS spectra.
Determination of the VBM for GaN was relatively straightforward. As shown in
Figure 10.2, extrapolation of the leading edge of the UPS VB spectra for GaN grown at
800°C to the energy axis yields a value of 2.4 ± 0.1 eV which is in excellent agreement with
the value of 2.4 ± 0.2 eV additionally obtained from XPS VB spectra (data not shown).
However, a slight change in the Ga and N core levels relative to GaN VBM occurred as a
function of GaN growth temperature. Table 10.2 lists the Ga and N core levels with respect
to the GaN VBM. Comparison of UPS VB spectra from 650 and 800°C GaN indicates that
the position of the GaN VBM moves only 0.2 eV closer to the Fermi level when increasing
the growth temperature from 650°C to 800°C. This indicates that the 0.6-0.7 eV change in
the CL-VBMGaN values with growth temperature resulted primarily from changes in the
positions of the GaN core levels.
321
Table 10.2. Al 2p and N 1s Core levels referenced to AlN VBM.
1050°C AlN Al2p-VBM 71.4±0.2 eV 650°C GaN Ga 3d-VBM 17.8±0.1 eV Al2p-Ga3d 54.1 800°C GaN Ga 3d-VBM 18.4±0.1 eV Al2p-Ga3d 53.9
0 2 4 6 8 10
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
He IHe I VBM = 2.4 eV
β
UPS x 10
Figure 10.2. UPS spectra of (2x2) (0001) GaN surface.
The positions of the Al 2p, N 1s, and Ga 3p, 3d, and 2p core levels were recorded as a
function of thickness of GaN. Examination of the data revealed that the CLGaN-CLAlN
were similar to within ±0.2 eV regardless of film thickness. However, for purposes of
322
comparison, values of ?Ev were calculated using AlNCL-GaNCL data taken at film
thicknesses in the range of 18-22Å. As noted by Grant and Waldrop, this thickness is beyond
the reported critical thickness for GaN on AlN [22] and should minimize strain effects.
Values for the (0001) GaN/AlN valence band discontinuity for GaN on AlN were
calculated using the data in Table 10.2,. For GaN grown at 800°C on AlN,
?Ev (GaN/AlN) = (VBM-CL)AlNbulk - (VBM-CL)GaNbulk + (CLAlN-CLGaN)int
= -71.4 + 18.4 + 53.9
= 0.9 ± 0.2 eV
Similarly, for GaN grown at 650°C, a ?Ev of 0.6±0.2 eV was calculated.
10.5. Discussion
Two reports of the (0001) GaN/AlN band alignment based on XPS measurements
have been published by Martin et al. [5,6] and Grant and Waldrop [8]. Both groups of
investigators arrived at a type I band alignment between GaN and AlN but the with
significantly different valence band discontinuities of 0.8±0.3 eV [5,6] and 1.36±0.07 eV [8].
As similar values for CLAlN-CLGaN
323
Table 10.3. CL-VBM data for AlN and GaN reported by various investigators. Al 2p-VBMAlN Ga 3d-VBMGaN King et al [21] 71.2±0.3 eV Bermudez et al [23] 71.9±0.2 eV 18.4±0.2 eV This Paper 71.4±0.2 eV 17.8±0.1 eV (650°C) 18.4±0.1 eV (800°C) Waldrop & Grant [8] 70.6±0.07 eV 17.76±0.07 eV Martin et al [5,6] 70.6±0.3 eV 17.1±0.3 eV
were measured in this paper and by Martin et al. and Grant and Waldrop, most of the
discrepancy in ?Ev arises from the values determined for CL-VBMAlN and CL-VBMGaN.
Table 10.3 summarizes the Al2p-VBMAlN and GaNCL-VBMGaN data reported in the ?Ev
measurements of this paper and those by Martin et al and Grant & Waldrop. Values for
Al2p-VBMAlN and Ga3d-VBMGaN from studies by Bermudez et al. [23] concerned with
the UHV reaction between Al and GaN are also included.
A comparison of the data in Table 10.3 shows a large distribution in the reported
Al2p-VBMAlN and Ga3d-VBMGaN values. As these values are extremely sensitive to the
location of the AlN/GaN VBM, it is possible that the observed discrepancy is a result of the
different methods used to locate the VBM. In the study reported here and in the research of
Bermudez et al. [23], the energy position of the VBM for AlN and GaN was determined by a
straight-line extrapolation of the leading edge of the XPS/UPS valence band spectra to the
energy axis. If a significant number of occupied states in the bandgap of AlN or GaN exist at
the surface, photoemission could occur from these states and would cause the valence band to
falsely appear closer to the Fermi level. This would result in a larger Al2p-VBMAlN or
324
Ga3d-VBMGaN value which would agree with the observation that the Al2p-VBM and
Ga3d-VBM values reported here and by Bermudez et al. are much larger than those by
Martin et al. and Grant & Waldrop. Theoretical calculations have yet to indicate the presence
of any types of states in the gap (i.e. surface states), and experimental work has yet to provide
direct evidence of surface states. We also note that for GaN films grown at 650°C (i.e.,
conditions identical to Grant and Waldrop) we obtained a value of 17.8±0.1 eV for Ga3d-
VBMGaN which is identical to the value reported by Grant and Waldrop. This indicates that
photoemission from surface states is probably not the source of discrepancy among the
various measurements.
However, the presence of surface states in sufficiently small densities as to be
undetected in photoemission can still pin the Fermi level in the gap and create band bending
at the surface. The presence of band bending at the surfaces of wide band gap
semiconductors could result in photo-voltage effects which could also account for the
observed discrepancies in the Cl-VBM data. This may explain why our value of 17.8±0.1 eV
for Ga3d-VBMGaN from (1x1) unreconstructed GaN films grown at 650°C and our value of
18.4±0.1 eV from (2x2) reconstructed GaN films grown at 800°C. However, we observed no
evidence of photovoltage effects. Without knowledge or direct evidence of surface states, the
observed discrepancies in ?Ev cannot be ascribed to photovoltage effects.
One final and possible source of the discrepancies in the reported ?Ev measurements
could result from differences in the types and densities of defect, strain and stoichiometry in
the GaN and AlN films used in the various studies. The recent theoretical calculations by Ke
et al. [16] have shown a strong variation in ?Ev (0001) GaN/AlN from 0.48 to 1.12 eV based
on the internal relaxation parameter, u (defined as d/c where d is the bond length along the
325
stacking direction and c is the lattice constant for the c axis). Their results show a very small
?Ev=0.48 eV when the GaN bond length along the c axis is forced into that of the ideal
hexagonal structure and a large ?Ev=1.12 eV when GaN is allowed to adopt the c axis bond
length of its bulk structure. For the more realistic intermediate case of the two extremes, they
calculated a ?Ev of 0.83 eV. The results of Ke et al. clearly illustrate the magnitude of the
effect that strain and defects at the GaN/AlN interface can play on the band alignment
between these materials. In a prior report [11], we have noted that the electrical, optical, and
microstructural properties of GaN films grown by the NH3-GSMBE technique at
temperatures of 650 and 800°C are different. In particular we have observed that GaN
growth at 650°C occurs via a 2D/Stranski-Krastanov type growth mechanism with the
corresponding films exhibiting rough surfaces, extremely high carrier concentrations
(N=1019-20/cm3), and broad, weak PL spectra. In contrast, GaN growth at 800°C was
observed to occur via a two dimensional/Frank van der Merwe growth mechanism with the
corresponding films exhibiting smooth surfaces, lower carrier concentrations (n=1016/cm3)
and sharp PL. Clearly, the defect levels in these films and at the interfaces are significantly
different and could be the cause for the increase in ?Ev from 0.6 to 0.9 eV with the increase
in growth temperature. Unfortunately, without detailed knowledge of the microstructure of
the GaN/AlN interfaces examined in this and other studies, it is not currently possible to link
these variations.
326
10.6. Conclusions
In conclusion, XPS/UPS has been used to determine the heterojunction valence band
discontinuity for the (0001) GaN/AlN interface. For GaN grown on AlN at 650°C, a valence
band discontinuity of 0.6±0.2 eV was determined, while a discontinuity of 0.9±0.2 eV was
determined for GaN grown on AlN at 800°C.
10.7. Addendum
Since the completion of this work, there have been several new experimental and
theoretical findings published which point to the existence of surface states on GaN and AlN
surfaces. First, Bermudez [24] has recently corrected his reported value for Ga 3d - VBMGaN
of 18.4 ± 0.2 eV to 18.0 ± 0.2 eV. He attributed the higher value of Ga 3d- VBMGaN to Fermi
level pinning by the existence of undetected surface states at the edge of the GaN VBM in the
prior study [23]. These surface states are apparently extinguished by O2 exposure which
unpins the GaN surface Fermi level and allowed for the more accurate measurement of Ga 3d
- VBMGaN.
327
The above findings by Bermudez [24] are additionally complemented by recent local
density approximation (LDA) supercell calculations by Northrup and Neugebauer [25] for
nonpolar (10-10) and (11-20) 2H-GaN surfaces. Their calculations show the presence of N
derived states (SN) just above the GaN VBM and Ga dangling bond states (SGa) just below
the GaN conduction band minimum (CBM) for the unrelaxed "ideal" (11-20) and (10-10)
surfaces. However when these surface are allowed to relax, both states move out of the gap
with the SN state dropping below the VBM and the SGa state moving above the CBM.
Similar results have also been obtained for the (11-20) surface of 2H-AlN by Kadas et al [26]
in their ab initio Hartree-Fock total energy calculations. Their results showed a sharp
localized state in the AlN band gap well below the AlN CBM associated with Al surface
atoms and states just above the AlN VBM associated with N surface atoms. Both of these
theoretical calculations therefore suggest that the surface state observed by Bermudez on
(0001) GaN could be associated with a nitrogen derived state. This suggestion is supported
by the recent time of flight scattering and recoiling spectrometry (TOF-SARS) studies of
Sung et al [27] which have found that GaN surfaces prepared similarly to Bermudez are N
terminated with Ga comprising the second layer of atoms. Unfortunately, these results are in
contradiction with the ion channeling and convergent beam electron diffraction studies of
Daudin et al [28] which have found smooth OMVPE GaN surfaces on (0001) Al2O3 to be Ga
terminated and rough GaN surfaces to be N terminated.
Although the films examined in this study are expected to be either Al or Ga
terminated (based on the fact that growth occurred on the Si face of (0001) 6H-SiC), the
theoretical calculations of Bernholc [15] have shown that for (2x2) Ga terminated GaN a N
adatom reconstruction is the most energetically favorable. Therefore, the above findings
reemphasize that both our values for Al2p - VBM for (2x2) (0001) AlN and Ga 3d - VBM
for (2x2) (0001) GaN may be both over estimated due to the unknowing inclusions of
nitrogen derived surface states in the location of the AlN and GaN VBM. This would bring
the calculated ?Ev GaN/AlN into closer agreement with the results of Grant and Waldrop [8]
328
and further away from theory. However if the overestimation of the values for both the
Al2p-VBMAlN and Ga3d-VBMGaN are the approximately the same, then the values for ?Ev
GaN/AlN reported above will not be affected. Clearly, all of the above indicates the need for
theoretical investigations of the electronic structure of both (0001) AlN and GaN surfaces.
More experimental investigations are also needed to determine the polarity of (0001) III-V
nitride films grown on both 6H-SiC and Al2O3 as well as to determine what effect various
processes may have the surface termination of these films.
10.8. Acknowledgments
The authors would like to thank Cree Research, Inc. for supplying the SiC substrates in this
research. This work was supported by the ONR under Contract N00014-92-J-1477.
10.9. References
1. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B 10 p. 1237 (1992). 2. J.H. Edgar, J. Mater. Res. 7 p. 235 (1992). 3. R.F. Davis, Proc. of IEEE 79 p. 702 (1991). 4. A. Morgan and J. Williams, Physics and Technology of Heterojunction Devices, Wiley & Sons, New York 1985, pp. 241-244. 5. G. Martin, S. Strite, A. Botchkarev, A. Agarwal, A. Rockett, H. Morkoc, W.R.L. Lambrecht, and B. Segall, Appl. Phys. Lett. 65 p. 610 (1994). 6. G. Martin, S. Strite, A. Botchkarev, A. Agarwal, A. Rockett, W.R.L. Lambrecht, B. Segall, and H. Morkoc, J. Electronic Materials 24 p. 225 (1995). 7. Z. Sitar, M.J. Paisley, B. Yan, R.F.Davis, J. Ruan, and J.W. Choyke, Thin Solid Films, 200 p. 311 (1991). 329
8. J.R. Waldrop and R.W. Grant, Appl. Phys. Lett. 68 p. 2879 (1996). 9. J. Baur, K. Maier, M. Kunzer, U. Kaufmann, and J. Schneider, Appl. Phys. Lett. 65 p. 2211 (1994). 10. J. Baur, K. Maier, M. Kunzer, U. Kaufmann, Materials Science and Engr. B 29 p. 61 (1995). 11. S.W. King, M.C. Benjamin, R.J. Nemanich, and R.F. Davis, to be published. 12. J. van der Weide, Ph.D. Dissertation, NCSU (1994). 13. M.W. Wang, J.O. McCaldin, J.F. Swenberg, T.C. McGill, R.J. Hauenstein, Appl. Phys. Lett. 66 p. 1974 (1995). 14. E.A. Albanesi, W.R.L. Lambrecht, and B. Segall, J. Vac. Sci. Technol. B 12 2470 (1994). 15. Bernholc, Spring MRS Symposium 1996. 16. S. Ke, K. Zhang, and X. Xie, J. Appl. Phys. 80 p. 2918 (1996). 17. R. Kaplan, Surface Science, 215 p. 111 (1989). 18. R.S. Kern, Ph.D. dissertation, NCSU (1996). 19. S. Tanaka, Ph.D. dissertation, NCSU (1995). 20. E.A. Kraut, R.W. Grant, J.R. Waldrop, and S.P. Kowalczyk, Phys. Rev. Lett. 44 p. 1620 (1980). 21. S.W. King, M.C. Benjamin, R.J. Nemanich, R.F. Davis, MRS Proceedings 22. S. Krishnankutty, R.M. Kolbas, M.A. Khan, J.N. Kuznia, J.M. Van Hove, and D.T. Olson, J. Electron. Mater. 21 437 (1992). 23. V.M. Bermudez, T.M. Jung, K. Doverspike, and A.E. Wickenden, J. Appl. Phys. 79 p. 110 (1996). 24. V.M. Bermudez, J. Appl. Phys., 80 p. 1190 (1996). 25. J.E. Northrup and J. Neugebauer, Phys. Rev. B., 53 p. R10477 (1996). 26. K. Kadas, S. Alvarez, E. Ruiz, P. Alemany, Phys. Rev. B., 53 4933 (1996). 27. M.M. Sung, J. Ahn, V. Bykow, D.D. Koleske, A.E. Wickenden, and J.W. 330
Rabalais, Phys. Rev. B, 54 p. 14652 (1996). 28. B. Daudin, J.L. Rouviere, and M. Arlery, Appl. Phys. Lett., 69 p. 2480 (1996).
331
11. Gas Source Molecular Beam Epitaxy Growth of
Scandium Nitride
on (111) 3C and (0001) 6H-SiC
To be Submitted for Consideration for Publication
to the
Journal of Crystal Growth
by
Sean W. King, Kieran M. Tracy, David W. Bray, Eric P. Carlson, Robert J. Therrien,
William G. Perry, Robert J. Nemanich, and Robert F. Davis
Department of Materials Science & Engineering
North Carolina State University
Raleigh NC 27695.
332
11.1 Abstract
Films of ScN were grown via gas source molecular beam epitaxy (10-5 Torr NH3,
800-1050°C) on (111) 3C-SiC epilayers previously deposited on vicinal and on axis (0001)Si
6H-SiC substrates. Analysis via SEM on the former revealed featureless surfaces at a
magnification of 10 kX, and a slight step morphology on the latter which presumably
mimicked the steps on the vicinal surface. All films exhibited hexagonal (1x1) LEED
patterns indicating growth of (111) oriented ScN. However, TEM showed the films to be
polycrystalline with columnar grains oriented at ˜ 15° to the (0001) direction of the SiC
substrate. Omega-2θ XRD scans additionally showed a FWHM of 1047 arc sec. The
conductivity of the films decreased with decreasing growth temperature. The sheet resistance
of films grown at 1050°C were = 5 ? cm. Hot probe measurements showed all ScN films to
be n-type. Ultra-violet photoelectron spectroscopy of the ScN films grown at 1050°C and
800°C showed the valence band maximum of ScN to be positioned 1.6 and 1.2 eV
respectively below the system Fermi level, indicating a minimum band gap of 1.6 eV.
11.2. Introduction
Scandium nitride (ScN) possesses the NaCl crystal structure and a lattice constant,
ao, of ≈ 4.503Å [1,2]. Early reports [3-5], of the growth of single crystals of ScN reported
333
large deviations (1-5%) from stoichiometry (i.e. ScN1-x), presumably due to incomplete
nitridation of the scandium charge. Possibly for this reason, it is still the subject of much
debate as to whether ScN is a semiconductor or a semi-metal [1,2,4,6-11]. Theoretical
calculations indicate that the fundamental band gap of ScN is 0.0 eV or very small (< 0.1 eV)
[8-10]. However, optical transmission experiments of ScN fabricated by a variety of
techniques have measured bandgaps ranging from 1.5-2.1 eV [1,2,4,7]. Many other
properties of ScN are not known or in dispute. Table 11.1 summarizes some of the reported
values for the physical properties of ScN.
Despite the lack of detailed knowledge, ScN is of interest in III-V nitride technology
due to its reasonably close lattice matching with cubic and hexagonal GaN (ao (0001) GaN =
3.189Å, a (111) ScN = 3.139Å [17]) and its possible band gap of 2.1 eV [2]. The reported
low resistivity [16] of ScN makes it a viable alternative to TiN for use as a low resistivity
ohmic contact to n-type GaN. Further, the achievement of p-type ScN could lead to the
development of a low resistivity contact to p-type GaN. Finally, the moderately close lattice
matching of ScN to SiC could also allow ScN to be used as a conducting buffer layer to
replace the insulating AlN or AlGaN buffer layers currently employed in the growth of GaN
on SiC substrates. Successful growth of GaN films on a Sc/GaN/Al2O3 structure has already
been recently demonstrated [18]. Finally, the high temperature/thermodynamic stability and
possible bandgap of 1.5-2.1 eV makes ScN an attractive replacement for InN in the
fabrication of blue LEDS and Lasers. However, a serious impediment to the use of ScN in
both of the last two applications is the difference in crystal structure between ScN (NaCl) and
GaN (ZB or wurtzite). In order to determine whether ScN can be used in any of these
applications, the authors have begun an investigation regarding the growth of ScN and
334
GaScN alloys on (0001) 6H-SiC and (0001) GaN. The results of initial research regarding
the growth of ScN on (111) 3C-SiC and (0001) 6H-SiC via gas source-molecular beam
epitaxy (GSMBE) are described in the following sections.
Table 11.1. Properties of ScN
Crystal Structure NaCl [1] Lattice Constant 4.503Å [1] Bandgap, Eg 0.0-2.6 eV [1,4,5,10,11] Conductivity type. n or p? n-type [1,2,11] Carrier Concentration 9x1019-8.3x1020/cm3 [1,2,11] Electron Mobility 28-150 cm2/Vsec [1,2] Doping Si, C, Zn, and Mg, no p-type ScN [2] Resistivity 25.4 µ? cm [16] Electron Effective Mass 0.1-0.2mo [14] Dielectric Constant ε = 5.2 [9], ε = 10.8 [14] Hardness 1170±150 kg/mm2 for 50 gm load [16] Thermal Expansion Coefficient 8.68x10-6/°C [9], 8.1x10-6/°C [2] Tmelt 2550±50°C [16] Oxidation Resistance Inert in air to temperatures of 600°C [16] Etchants Completely dissolves in boiling HCl and HNO3 [16]
11.3. Experimental
A gas source-molecular beam epitaxy (GSMBE) system, designed and constructed
specifically for the growth of ScN, AlN, and GaN thin films, was used in these studies. The
GSMBE was a component of an ultrahigh vacuum transfer line which allowed for in situ
analysis of the ScN films by low energy electron diffraction (LEED), Auger electron
spectroscopy (AES), x-ray photoelectron spectroscopy (XPS), and ultra-violet photoelectron
spectroscopy (UPS). Source materials in this system consisted of NH3 (99.9995%), Sc
335
(99.99%), Ga (99.99999%), and Al (99.99999%). The metals were evaporated from standard
Knudsen cells. Impurities in the scandium charge including, Al, Ca, Cr, Cu, Mn, Si & Y,
were each = 30 ppm. However, significant amounts of fluorine were also detected by XPS
from Sc films evaporated onto 6H-SiC (0001). The source of the fluorine is currently
undetermined but is believed to be the scandium charge. A base pressure of 10-10 Torr was
achieved in the GSMBE system via a 400 l/s turbo pump and a 500 l/s ion pump. During
nitride growth, the GSMBE system was pumped by the turbo pump only in order to prevent
irreversible damage to the ion pump. Substrate temperatures of 1100°C were easily achieved
via a hot tungsten filament heater.
Vicinal and on axis, 6H-SiC wafers (0001) (Nd ≈ 1018/cm3) with a 1 micron 3C-SiC
epitaxial layer (Nd≈1017/cm3) were provided by Cree Research Inc. The unpolished side of
these wafers were coated with an opaque tungsten film via RF sputtering to increase the
thermal heating efficiency of the SiC, as the latter is transparent to the infra-red radiation
emitted by the tungsten filament heater. After sputter coating, the wafers were ultrasonicated
in trichloroethylene, acetone, and methanol for 10 min. each and then dipped in 10:1 buffered
HF for 10 min. to remove the surface oxide. The wafers were subsequently
degassed/annealed at 1050°C in the GSMBE system for 10-15 min. to further desorb the
surface oxide. This surface displayed a (√3x√3)R30° reconstructed surface in LEED.
Growth of ScN was initiated on this surface in 10-5-10-4 Torr NH3 at temperatures ranging
from 800-1050°C.
336
11.4. Results
11.4.1. Thermodynamics
In an initial attempt to understand the growth of ScN and deduce a set of optimized
growth parameters, an MBE/CVD equilibrium diagram for ScN was computed using the
software package HSC. The computed equilibrium diagram shown in Figure 11.1, shows
that under GSMBE conditions (i.e. 10-5-10-4 Torr NH3), ScN is the equilibrium phase to
900°C. This phase is stable at higher temperatures, but is in equilibrium with a significant
amount of Sc(v). At still higher temperatures ScN decomposes completely into Sc(v) and
N2(g). For comparison purposes, equilibrium diagrams for AlN, GaN, and InN were also
calculated using HSC (see Figures 11.2,11.3, and 11.5). The equilibrium diagram of ScN
closely resembles the equilibrium diagram determined for AlN (see Figure 11.2) indicating
the possibility that the optimized growth conditions for ScN and AlN should be
approximately the same. In contrast, the equilibrium diagrams computed for GaN and InN
(see Figures 11.3 and 11.5) showed these materials to become thermodynamically unstable at
much lower temperatures than ScN and AlN. In fact in order to stabilize these materials
under typical experimental conditions, it was found necessary to exclude N2 formation from
the HSC equilibrium calculations. In a sense, this is a crude way of imposing/including
kinetic limitations in thermodynamic calculations. However, this approach was shown to be
337
valid as the computed equilibrium for GaN and InN with N2 removed from the equilibrium
products was found to be in close agreement with the experimentally determined equilibrium
of GaN and InN in vacuum (see Figure 11.4 and 11.5).
10 -9
10 -7
10 -5
0.001
0.1
10
1000
10 12
10 14
10 16
10 18
10 20
10 22
10 24
600 800 1000 1200 1400 1600
NH
3 Pre
ssur
e (T
orr)
NH
3 Flux (#/cm2sec)
Temperature (ÞC)
ScN
Sc(v) + N2
ScN + Sc(v)
OMVPE/CVD
GSMBE
Flux = P/¦2šmkT
Figure 11.1. ScN equilibrium diagram computed using HSC.
338
10 -7
10 -5
0.001
0.1
10
1000
10 14
10 16
10 18
10 20
10 22
10 24
600 800 1000 1200 1400 1600 1800
NH
3 Pre
ssur
e (T
orr)
NH
3 Flux (#/cm2sec)
Temperature (ÞC)
AlN
Al + N2
Al + AlNOMVPE
GSMBE
Figure 11.2. AlN equilibrium diagram computed using HSC.
10 -3210 -2910 -2610 -2310 -2010 -1710 -1410 -1110 -810 -5
0.0110 110 410 7
10 1010 13
10 -1110 -810 -50.011010 410 710 1010 1310 1610 1910 2210 2510 2810 3110 34
200 400 600 800 1000 1200 1400 1600
NH
3 Pre
ssur
e (T
orr)
NH
3 Flux (#/cm2sec)
Temperature (ÞC)
Ga + N2
GaN HSC: Ga-NH3 Equilibrium with N
2
HSC: Ga-NH3 Equilibrium without N
2
GSMBE
OMVPE
Figure 11.3. GaN equilibrium diagram computed using HSC.
339
10 -10
10 -8
10 -6
10 -4
10 -2
10 0
10 2
10 4
10 6
10 8
10 10
10 12
10 14
10 16
10 18
10 20
10 22
10 24
10 26
10 28
10 30
600 700 800 900 1000 1100 1200
NH
3 Pre
ssur
e (T
orr)
NH
3 Flux (#/cm2sec)
Temperature (ÞC)
GaN Decomp. Rate(Munir & Searcy)
Equil. NH3 Flux
(Thurmond & Logan) OMVPE
GSMBE
GaN(s)
Ga + N2
Figure 11.4. GaN equilibrium diagram computed using experimental data of Thurmond and Logan [19] and Munir and Searcy [20].
10 -18
10 -16
10 -14
10 -12
10 -10
10 -8
10 -6
10 -4
10 -2
10 0
10 2
10 4
10 4
10 6
10 8
10 10
10 12
10 14
10 16
10 18
10 20
10 22
10 24
200 400 600 800 1000
3 Pre
ssur
e (T
orr)
NH
3 Flux (#/cm2sec)
Temperature (ÞC)
InN
In(l) +N2
GSMBE
OMVPE
HSC: In-NH3 Equilibrium with N
2
HSC: In-NH3 Equilibrium without N
2
InN Decomposition rate/N2
Vapor, Jones and Rose
Figure 11.5. InN equilibrium diagram computed using HSC and experimental data of Jones and Rose [21].
340
11.4.2. Growth
In situ AES and XPS analysis of ScN films grown in 10-5 Torr NH3 at 800-1050°C
detected the presence of only Sc and N (see Figure 11.6 and 11.7). Small traces of oxygen
were detected in AES (see Figure 11.6) but were attributed to electron beam induced
oxidation during AES. Other contaminants (most notably fluorine) were not detected. A
more detailed XPS analysis of the Sc2p1/2,3/2 and N 1s core levels showed the ScN films to
be stoichiometric within the detection limit of this analytical technique ( ≈ 0.1 at.%) (see
Figure 11.7). No unreacted Sc was detected on the surface. The XPS spectrum shown in
Figure 11.7 is extremely similar to that obtained by L. Porte [13] from ScN films prepared by
argon sputtering Sc onto Ta in 3x10-7 Torr N2. In situ UPS analysis showed a shift in the
ScN valence band maximum (VBM) from 1.6 eV below the Fermi level for the 1050°C films
to 1.2 eV below EF for the 800°C films (see Figure 11.8). Correspondingly, an increase in
the resistance of the ScN films with decreasing growth temperature was observed. For ScN
films grown at 1050°C, typical sheet resistances measured by four point probe were ˜ 5 ? cm.
Films grown at lower temperatures < 950°C were too resistive for four point probe
measurements. Hot probe measurements revealed that all ScN films were n-type. Doping of
ScN with Si and C from SiH4 and C2H4 respectively was investigated. The introduction of
Si did not appreciably decrease the resistivity of the ScN films. Capacitance-Voltage
measurements indicated ND-NA ≈ 5x1017/cm3. In contrast, the resistivity of C doped films
showed a dramatic decrease to 1.2 ? cm and ND-NA = 1019-1020/cm3 (CV measurements).
341
Unfortunately, hot probe measurements indicated these films were still n-type. Subsequent
SIMS analysis showed no incorporation of carbon into the ScN films but an increase in the
oxygen contamination by 2-3 order of magnitudes. The source of the oxygen is presumably
H2O or O2 impurities in the C2H4. However, these results indicate that oxygen may be an
excellent n-type dopant for ScN.
Hexagonal (1x1) LEED patterns were obtained at Ep ≈ 50 eV from all ScN films
indicating growth of (111) oriented ScN. Examination of these films via SEM revealed
featureless surfaces to magnifications of 10,000X. However, some films grown on off-axis
(0001) 6H-SiC exhibited a step like structure apparently mimicking the steps on the SiC
substrate. Transmission electron microscopy of 800°C ScN showed the films to be
polycrystalline with the grains oriented at an angle of ˜ 15° to the (0001) direction of the SiC
substrate (see Figure 11.9). This is in agreement with the wide FWHM of 1047 arc sec
observed in XRD Omega-2θ scans of similar films. A similar microstructure was observed
for ScN films grown on a 5000Å GaN/AlN/6H-SiC structure (see Figure 11.10).
342
100 200 300 400 500 600 700
dN(E
)/dE
Electron Energy (eV)
O KLL
Sc LMM
Sc MNN
N KLL
Figure 11.6. AES survey scan of ScN film grown at 800°C on (111) 3C-SiC.
392 396 400 404 408
Cou
nts (
arb.
uni
ts)
Binding Energy (eV)
N 1s
Sc 2p3/2
Sc 2p1/2
Figure 11.7. XPS spectrum of N 1s and Sc 2p3/2,1/2 2p3/2,1/2 core levels from a ScN film grown at 800°C.
343
-14 -12 -10 -8 -6 -4 -2 0 2
Cou
nts (
arb.
uni
ts)
Energy Below Fermi Level (EF = 0) (eV)
He Iß
He I VBM = -1.6 eV
Figure 11.8. UPS spectrum of ScN film grown at 800°C on (111) 3C-SiC.
344
Figure 11.9. TEM of ScN film grown at 800°C on (111) 3C-SiC.
Figure 11.10. TEM of 800°C ScN on 5000Å (0001) GaN/AlN/6H-SiC.
11.5. Discussion
The above thermodynamic calculations and growth studies show that ScN films can
be easily grown/deposited via NH3-GSMBE in the temperature range of 800-1050°C.
Dismukes et al [1,2] have previously demonstrated growth of single crystal films of ScN on
(11-20) Al2O3 via hydride vapor phase epitaxy in the temperature range of 850-930°C. As
800 and 1050°C have been determined to be our optimized growth temperatures for GaN and
AlN respectively, ScN should be much easier to incorporate into these compounds than InN
which is highly unstable at these temperatures in MBE. Previous attempts in this research to
grow InN via NH3-GSMBE at these temperatures have been completely unsuccessful.
Unfortunately, the difference in crystal structure between ScN and AlN/GaN will limit the
345
range over which these compounds can be successfully combined to form equilibrium alloys.
The results of S. Lee et al. [22], however, have shown that single phase, NaCl structure
AlxTi1-xN with x as large as 0.8 can be fabricated via plasma enhanced chemical vapor
deposition. The observation by these investigators of metastable AlN particles with the NaCl
structure is also extremely encouraging.
Our UPS measurements on the 1050°C ScN films which show the ScN VBM to be
1.6 eV below the system Fermi level are in agreement with the UPS results of Porte [13]
which also found the VBM of ScN to be 1.5-2.0 eV below EF. These measurements indicate
that ScN has a minimum bandgap of 1.5 eV. The observation that the ScN VBM shifted 0.4
eV closer to EF with decreasing growth temperature indicates a decrease in donors
presumably due to nitrogen vacancies.
11.6. Conclusions
Scandium nitride films have been successfully deposited on (111) 3C/(0001)6H-SiC
substrates by NH3-GSMBE in the temperature range of 800-1050°C. All films were n-type
with a maximum conductivity of 5 ? cm for films grown at 1050°C. The conductivity of
these films decreased with decreasing temperature. Analysis via SEM showed featureless
surfaces at a magnification of 10 kX. Analysis via TEM indicated the films were
polycrystalline with grains oriented at ˜ 15° to the (0001) direction. A minimum bandgap of
1.6 eV for ScN was deduced via UPS measurements.
11.7 Acknowledgments
346
The authors would like to thank Cree Research, Inc. for supplying the SiC substrates
used in this research. This work was supported by the ONR under contracts N00014-91-J-
1410 and N00014-92-J-1477.
11.8 References
1. J.P. Dismukes, W.M. Yim, J.J. Tietjen, and R.E. Novak, RCA Review, 31 p. 681 (1970). 2. J.P. Dismukes, W.M. Yim, and V.S. Ban, J. Cryst. Growth, 13/14 p. 365 (1972). 3. R. Didchenko and F.P. Gortsema, J. Phys. Chem. Solids, 24 p. 863 (1963). 4. G. Busch, E. Kaldis, E. Schaufelberger-Teker, and P. Wachter, in Les Elements des Terres Rares, Edition du CNRS, Colloque Internationale No. 180, Tome I (1970), p. 359. 5. M.D. Lyutaya, A.B. Goncharuk, and I.I. Timofeeva, Z. Prik. Khimii, 48 p. 721 (1975). 6. N. Sclar, J. Appl. Phys., 33 p. 2999 (1962). 7. N. Sclar, J. Appl. Phys., 35 p. 1534 (1964). 8. K. Schwarz, P. Weinberger, and A. Neckel, Theor. Chim. Acta, 15 p. 159 (1969). 9. A. Neckel, P. Rastl, R. Eibler, P. Weinberger, and K. Schwarz, J. Phys. C, 9 p. 579 (1976). 10. R. Monnier, J. Rhyner, T.M. Rice, and D.D. Koelling, Phys. Rev. B, 31 p. 5554 (1985). 11. G. Travaglini, F. Marabelli, R. Monnier, E. Kaldis, and P. Wachter, Phys. Rev. B, 34 p. 2876 (1986). 12. W. Lengauer, J. Solid State Chemistry, 76 p. 412 (1988). 13. L. Porte, J. Phys. C: Solid State Phys., 18 p. 6701 (1985). 14. G. Harbeke, E. Meier, and J.P. Dismukes, Optics Communications, 4 p. 335 (1972). 347
15. M.D. Lyutaya and V.F. Bukhanevich, Russian J. Inorganic Chem., 7 p. 1290 (1962). 16. G.V. Samsonov, M.D. Lyutaya, and V.S Neshpor, Z. Prik. Khimii, 36 p. 2108 (1963). 17. S. Strite and H. Morkoc, J. Vac. Sci. Technol. B, 10 p. 1237 (1992). 18. D. Koleske, A. Wickenden, J. Freitas, R. Kaplan, S. Prokes, this symposium. 19. C.D. Thurmond and R.A. Logan, J. Electrochem. Soc., 119 p. 622 (1972). 20. Z.A. Munir and A.W. Searcy, J. Chem. Phys., 42 p. 4223 (1965). 21. R.D. Jones and K. Rose, J. Phys. Chem. Solids, 48 p. 587 (1987). 22. S. Lee, B. Kim, H. Kim, J. Lee, J. Appl. Phys. 80 p. 1469 (1996).
348