<%BANNER%>

Growth and Characterization of Gallium Nitride on Lattice-Matched Magnesium Calcium Oxide

Permanent Link: http://ufdc.ufl.edu/UFE0024993/00001

Material Information

Title: Growth and Characterization of Gallium Nitride on Lattice-Matched Magnesium Calcium Oxide
Physical Description: 1 online resource (311 p.)
Language: english
Creator: Gerger, Andrew
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: calcium, camgo, cao, gan, latticematched, liftoff, magnesium, mgcao, mgo, oxide
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GROWTH AND CHARACTERIZATION OF GALLIUM NITRIDE ON LATTICE-MATCHED MAGNESIUM CALCIUM OXIDE By Andrew Phillip Gerger August 2009 Chair: Brent Gila Cochair: Cammy Abernathy Major: Materials Science and Engineering As the interest for Gallium Nitride based materials and devices expand, focus on controlling the quality of material is becoming a priority. In this study, the focus of GaN growth upon a lattice tunable oxide is investigated. The oxide utilized is MgCaO. This oxide has the unique feature of maintaining single phase as the metal ratios are varied. Lattice parameter of the oxide is therefore able to be changed and matched to that of GaN. With a matched substrate, the subsequent GaN growth has less strain which results in decreased level of defects in the film, which would normally lead to degraded performance and early device failure. Higher quality material lends itself to further applications. Applications such as nanobar fabrication and oxide etching resulting in GaN film transfer to other substrates. Another issue encountered in the expansion of GaN devices is heat generation. Heat buildup in high-power RF devices has been the crutch for most material systems. Having the ability to dissipate heat from the device permits greater performance at higher operating loads and longer device lifetime. Therefore, to transfer GaN onto high thermal conductivity substrates would warrant a greater range for device operation, as the heat is more efficiently dissipated through the substrate. The MgCaO substrate can be easily etched away while leaving overgrowth nitride intact, lending itself to the concept of lift-off or pick-and-place. Transfer of GaN onto diamond substrates or other comparable materials would enable the high thermal conductivity and would facilitate the GaN devices to work at higher temperatures and power loads. The substrate transfer could be accomplished after the material growth or after the device fabrication was complete. In this work, the growth and characterization of single crystal MgCaO epitaxial films have been demonstrated. This has been demonstrated on single crystal GaN epifilms grown on sapphire, but SiC can be employed as well, eliminating the underlying GaN epifilm. Oxide thicknesses over 100nm with surface RMS roughness of less than 0.5nm have been achieved and employed as a successful substrate for nitride based overgrowth.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Andrew Gerger.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Gila, Brent P.
Local: Co-adviser: Abernathy, Cammy R.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024993:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024993/00001

Material Information

Title: Growth and Characterization of Gallium Nitride on Lattice-Matched Magnesium Calcium Oxide
Physical Description: 1 online resource (311 p.)
Language: english
Creator: Gerger, Andrew
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: calcium, camgo, cao, gan, latticematched, liftoff, magnesium, mgcao, mgo, oxide
Materials Science and Engineering -- Dissertations, Academic -- UF
Genre: Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GROWTH AND CHARACTERIZATION OF GALLIUM NITRIDE ON LATTICE-MATCHED MAGNESIUM CALCIUM OXIDE By Andrew Phillip Gerger August 2009 Chair: Brent Gila Cochair: Cammy Abernathy Major: Materials Science and Engineering As the interest for Gallium Nitride based materials and devices expand, focus on controlling the quality of material is becoming a priority. In this study, the focus of GaN growth upon a lattice tunable oxide is investigated. The oxide utilized is MgCaO. This oxide has the unique feature of maintaining single phase as the metal ratios are varied. Lattice parameter of the oxide is therefore able to be changed and matched to that of GaN. With a matched substrate, the subsequent GaN growth has less strain which results in decreased level of defects in the film, which would normally lead to degraded performance and early device failure. Higher quality material lends itself to further applications. Applications such as nanobar fabrication and oxide etching resulting in GaN film transfer to other substrates. Another issue encountered in the expansion of GaN devices is heat generation. Heat buildup in high-power RF devices has been the crutch for most material systems. Having the ability to dissipate heat from the device permits greater performance at higher operating loads and longer device lifetime. Therefore, to transfer GaN onto high thermal conductivity substrates would warrant a greater range for device operation, as the heat is more efficiently dissipated through the substrate. The MgCaO substrate can be easily etched away while leaving overgrowth nitride intact, lending itself to the concept of lift-off or pick-and-place. Transfer of GaN onto diamond substrates or other comparable materials would enable the high thermal conductivity and would facilitate the GaN devices to work at higher temperatures and power loads. The substrate transfer could be accomplished after the material growth or after the device fabrication was complete. In this work, the growth and characterization of single crystal MgCaO epitaxial films have been demonstrated. This has been demonstrated on single crystal GaN epifilms grown on sapphire, but SiC can be employed as well, eliminating the underlying GaN epifilm. Oxide thicknesses over 100nm with surface RMS roughness of less than 0.5nm have been achieved and employed as a successful substrate for nitride based overgrowth.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Andrew Gerger.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Gila, Brent P.
Local: Co-adviser: Abernathy, Cammy R.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024993:00001


This item has the following downloads:


Full Text

PAGE 1

GROWTH AND CHARACTERIZATION OF GALLIUM NITRIDE ON LATTICEMATCHED MAGNESIUM CALCIUM OXIDE By ANDREW PHILLIP GERGER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Andrew Phillip Gerger 2

PAGE 3

To my family, friends, and all who have the desire to learn 3

PAGE 4

ACKNOWLEDGMENTS First off, Id like to thank my parents Stephen and Melody for all the love, support, and guidance they have given me throughout my life. They have bestowed upon me a desire of knowledge and hard work. In addition, my acad emic friends and mentors have taught me immensely in the ways of research and academia I have been fortunate to work at MAIC throughout most of my stint at the University of Florida, and through that position, I have learned many useful skills both intellectuall y and physically with the repair s of much equipment. Also, the university setting has allowed me to do what Ive enjoyed since a young boy, which is taking things apart and trying to fix them. Ill always remember the countless high-precision repairs of equipment, to the duct-tape fixes. The best of which allowed my plasma head to operate more stable by the use of a five dollar desk fan that I borrowed from my Lovely! I want to thank my committee members, Dr Stephen Pearton, Dr. David Norton, and Dr. Fan Ren for the classroom knowledge, and research guidance they have all shared. Immense thanks goes to my advisors, Dr. Brent Gila and Dr. Cammy Abernathy, for the knowledge and skills they have bestowed upon me throughout my years at UF. Their insight and experience with research and lab work has been crucial to my success. I will take away the notion that from motorcycles to molecular beam epitaxy, learning happe ns everywhere. Also to the University of Florida, thank you for the Nationa l Titles and the parking tickets Keep growing and allowing other bright scientists and engineers to accomplish their goals. Finally, with the support of my wonderful fiance, soon to be wife Kara, this work has come to fruition. She kept lighting the fire under me to get things done, and I thank her greatly. The support and love from my darling w ill always be appreciated and cherished. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .........................................................................................................................9 ABSTRACT ...................................................................................................................................23 CHAPTER 1 INTRODUCTION................................................................................................................. .25 Motivation...............................................................................................................................25 Dissertation Outline........................................................................................................... .....26 2 LITERATURE REVIEW AND BACKGROUND................................................................28 Substrate Issues.......................................................................................................................28 III-Nitride Crysta l Structures..................................................................................................28 Current Alternate Substrates...................................................................................................29 Lithium Gallate................................................................................................................ 29 Lithium Aluminate..........................................................................................................30 Gadolinium Oxide...........................................................................................................31 Zinc Oxide.......................................................................................................................31 3 EXPERIMENTAL APPROACH AND INSTRUMENTATION..........................................34 Materials Characterization......................................................................................................34 Atomic Force Microscopy (AFM)...................................................................................34 Scanning Electron Microscopy (SEM)............................................................................35 Cathodoluminescence (CL).............................................................................................37 Reflection High-Energy Elect ron Diffraction (RHEED)................................................38 Auger Electron Spectroscopy (AES)...............................................................................38 X-Ray Photoelectron Spectroscopy (XPS)......................................................................39 X-Ray Diffraction (XRD)................................................................................................39 Photoluminescence (PL)..................................................................................................41 Contact Angle Goniometer..............................................................................................42 Hall-effect measurements (Hall).....................................................................................42 Current-Voltage Measurements (I-V).............................................................................42 Ellipsometry....................................................................................................................43 Transmission Electron Microscopy (TEM).....................................................................43 Material Growth Techniques..................................................................................................43 Molecular Beam Epitaxy (MBE).....................................................................................44 Metal Organic Chemical Vapor Deposition (MOCVD).................................................46 5

PAGE 6

4 MBE GROWN MGCAO........................................................................................................61 Epitaxy Challenges.................................................................................................................61 Alloying Calculations For MgO and CaO..............................................................................61 Gallium Nitride Substrate Preparation....................................................................................63 Standard MBE Oxide Growth Procedures..............................................................................66 Binary Oxide Growth of MgO and CaO.................................................................................70 Growth of MgCaO Via Digital Growth Mode.......................................................................71 Growth of MgCaO Via Continuous Growth Mode................................................................78 MgCaO TEM Characterization...............................................................................................79 MgCaO and MgO Thermal Expansion Measurements..........................................................80 Mounting Methods for MOCVD Growth...............................................................................83 5 MBE GAN ON OXIDE........................................................................................................124 Sample Preparation...............................................................................................................124 GaN calibration growths on GaN.........................................................................................128 MBE GaN on MgCaO..........................................................................................................133 XPS Study of MgCaO Annealing.........................................................................................141 6 MOCVD GAN GROWTH...................................................................................................180 MOCVD Overview...............................................................................................................180 MOCVD GaN Substrates.....................................................................................................183 Back-Side Sample Coating...................................................................................................187 SILICON BACKSIDE COATED SAMPLES.....................................................................190 Annealing of MgCaO in Ammonia......................................................................................193 MOCVD GaN on MgO........................................................................................................195 MOCVD GaN on Sc2O3.......................................................................................................196 Contact Angle Measurements...............................................................................................199 MOCVD GaN on Sc2o3 Capped MgO..................................................................................202 MOCVD GaN on MgCaO Capped with MBE GaN............................................................205 7 CHEMICAL STABILTY AN D ETCHING OF MGCAO...................................................256 Wet Etching of MgCaO........................................................................................................256 Dry Etching of MgCaO and GaN.........................................................................................258 8 GAN FEATURE FABRICATION AND LIFTOFF............................................................271 Nano Dimension GaN Features............................................................................................271 GaN on Oxide Nanobars.......................................................................................................272 Lift-Off Solution...................................................................................................................274 Lateral Etching for Liftoff....................................................................................................275 9 SUMMARY AND CONCLUSIONS...................................................................................300 Summary of MgCaO on GaN...............................................................................................300 6

PAGE 7

Summary of MBE GaN on MgCaO.....................................................................................301 Summary of MOCVD GaN on Oxide..................................................................................302 Summary of MgCaO Ch emical Stability..............................................................................303 Summary of GaN Feature Fa brication and Liftoff...............................................................303 Future Work Directions........................................................................................................3 04 Future Work in Progress with ZnO......................................................................................305 LIST OF REFERENCES.............................................................................................................307 BIOGRAPHICAL SKETCH.......................................................................................................311 7

PAGE 8

LIST OF TABLES Table page 2-1 Properties of alternative substrates....................................................................................33 4-1 Material properties for GaN, MgO, and CaO....................................................................88 4-2 Auger peak height rati os showing various cleaning processes and the resulting surface contamination........................................................................................................91 4-3 XPS signal intensities of the contamination elements compared to the nitrogen peak of the samples that were pretreated with the standard cleaning process............................93 4-4 XPS signal intensities of the contamination elements compared to the nitrogen peak of the samples that were not pretreated with the standard cleaning process.....................93 5-1 Sample growth conditions and characte rization data of silicon doped MBE GaN on MgCaO.............................................................................................................................160 5-2 Sample HALL data from silicon doping..........................................................................160 6-1 Contact angle measurements of the various surfaces employed in the growth of GaN with both 1l and 5l water drops...................................................................................235 7-1 Possible reaction of MgO with water...............................................................................261 7-2 Results from the wet etch study.......................................................................................264 7-3 Properties of the various chemicals used on the MgCaO................................................265 8

PAGE 9

LIST OF FIGURES Figure page 3-1 Veeco Dimension 3100 AFM with Nanoscope V controller.............................................48 3-2 Veeco Multimode AFM with Nanoscope IIIa controller...................................................48 3-3 JEOL 6400 SEM.............................................................................................................. ..49 3-4 FEI XL40 Field-Emission SEM.........................................................................................49 3-5 Possible RHEED patterns. A) Amorphous diffraction pattern. B) Polycrystalline diffraction pattern. C) Single crystal diffraction pattern. [Reprinted with permission from B.P. Gila, 2000. Growth and Characteri zation of Dielectric Materials for Wide Bandgap Semiconductors. PhD dissertation (pg. 47, Figure 3-11). University of Florida, Gainesville, Florida.]............................................................................................50 3-6 AES Perkin-Elmer PHI 660 Scanning Auger Multiprobe.................................................51 3-7 Perkin-Elmer XPS/ESCA PHI 5100 ESCA.......................................................................51 3-8 Custom XPS system attached to Varian MBE...................................................................52 3-9 Phillips APD 3720 XRD....................................................................................................53 3-10 Phillips MRD XPert system with a 5-crystal analyzer.....................................................53 3-11 Phillips 3100 high-temperature XRD inst rument with a helium atmosphere....................54 3-12 Omnichrome helium-cadmium ultraviolet laser at 325nm along with a SpectraPhysics argon ion laser.......................................................................................................55 3-13 PL setup showing optics, monochromat or, and PMT on the left of the image..................55 3-14 Rudolph V-530044 Auto EL IV ellipsometer....................................................................56 3-15 JEOL 200CX TEM............................................................................................................57 3-16 Modified Riber model 2300 MBE used for oxide growths...............................................58 3-17 Modified Varian Gen II MBE syst em utilized for the GaN growths.................................58 3-18 Veeco/Emcore P75 TurboDisk vertical react or utilized for GaN growths. A) Image of the MOCVD growth chamber (left) and load-lock (right). B) View of the full tool.....................................................................................................................................59 3-19 Image of the wafer susceptor being heated by the resistance coil heater..........................60 9

PAGE 10

4-1 Graphical representation of the lattices of GaN and MgO or CaO. Also included is the outline of the matching (0001) and (111) planes.........................................................86 4-2 Vegard's Law results in a la ttice matched ternary to GaN.................................................87 4-3 Lattice mismatch of oxide to GaN, show ing the change with Mg content within a ternary alloy of MgCaO.....................................................................................................87 4-4 Plot of various ternary ratios, with TEC included in calculations.....................................89 4-5 Removal of carbonous photoresist from a sa mple exposed to UV-ozone in the small box ozoner. The removal rate was calculated to be 72 /min, showing effective removal of carbon from surfaces for this process..............................................................90 4-6 XPS study of GaN surfaces before and af ter vacuum annealing at 700C. The figures on the left (A,B,C) did not have the st andard pretreatment of HCl/UV-O3/BOE, while the spectrums on the right (D,E,F) did have the pretreatment process....................92 4-7 AFM images of MBE binary oxides Mg O and CaO. A) MgO growth, Ra=1.23nm vertical scale of 15nm. B) CaO growt h, Ra=0.815nm vertical scale is 15nm................94 4-8 XRD scans of the binary shown in Figur e 4-4, the scans are also in Log scale to emphasize the oxide peaks in relation to the large GaN signal since the oxides are only a few hundred thick. A) CaO substrate showing the (222) CaO peak, the GaN peak, along with small peaks from the sapphire substrate. B) MgO substrate shown the (222) MgO peak, the GaN peak, and additional stray peaks from the substrate...................................................................................................................... .......95 4-9 XPS scans of the binary oxides MgO and CaO, which show the corresponding elements in addition with an indium peak from the mounting technique and a carbon peak which is from atmospheric carbon from the transfer out of the oxide MBE and into the XPS chamber. A) MgO. B) CaO........................................................................96 4-10 Omega rocking curves of the (222) and (002) peaks of MgO. The FWHM of the peaks indicate the degree of crystallin ity. A) (222) MgO omega RC, FWHM of 1.882 degrees. B) (002) MgO omega RC, FWHM of 4.198 degrees...............................97 4-11 AES scans of continuously grown sample (at left), and digita lly grown sample (at right).22...............................................................................................................................98 4-12 XRD scan showing change in atomic sp acing of the MgCaO ternary around the large GaN (004) peak. This shift is accomplis hed through varying the time sequencing of the shutters: (Mg time(s) / Ca time(s)). It can be seen that adding more calcium, shifts the peak position to smaller angl es which is a larger lattice spacing.......................98 4-13 XRD scan of standard UOE GaN used as substrates for the oxide calibrations. The GaN 004 peak can be seen which includes the alpha and beta signals. Along with the GaN peak can be seen some peaks from the sapphire substrate........................................99 10

PAGE 11

4-14 AFM images of the starting UOE GaN subs trate showing pits in the GaN film. A) Cross-section scan of a pit. B) 1m scan of an area without a pit, showing a smooth surface, yet lacking growth te rraces, average roughness (Ra)=3 ..................................100 4-15 AFM images of our in-house grown MO CVD GaN substrate showing nice growth terraces and roughness values of a few angs trom. A) 1mX1m scan area with roughness of 1 B) Larger scan area of same sample...................................................101 4-16 Omron Zen computer programmed digital relay. A) Self-installed digital relay that interfaced with the shutter controllers to allow for automated switching of the source shutters. B) View showing th e relay to the right and the sw itches that were used to manually growth the digital samples...............................................................................102 4-17 XRD scan of the initial set of digital MgCaO MBE growths. The spectrum indicates that increasing the amount of time that the magnesium shutter was open, caused a shift of the oxide lattice closer to the Ga N peak near 73 degrees. The timing is in seconds, therefore 10/10 is 10seconds of the respective metal followed by 10 seconds of the other................................................................................................................... ....103 4-18 Sample set with varying timing of the shutter sequencing. The magnesium flux was 3.7E-8 Torr BEP and calcium flux was 3.5E-8 Torr BEP. The spectrum shows that with these fluxes, an approximately latt ice-matched oxide can be grown with a shutter sequence composed mainly of calcium exposure................................................104 4-19 XRD scan of two samples with the same 10 second timing of shutter exposures, with varying calcium flux. As can be seen the sample with calcium flux of 8.2E-8 Torr BEP, and corresponding magnesium flux of 4.4E-8 Torr BEP.......................................105 4-20 XRD scan showing The GaN peaks under a 50nm layer of MgCaO, in which the oxide peaks are aligned to the GaN peak s and therefore can not be seen. The samples were grown with the installed digi tal time delay relay at sequencing of 10 seconds and then alternating sequenci ng of 15 seconds magnesium and then 5 seconds calcium...............................................................................................................106 4-21 Auger scan showing the concentrations of the elements at the surface before depth profiling. This top spectrum is for the c ontinuous growth indicating a Ca/O ratio of 63.2%, and a Mg/O ratio of 36.8%. The lowe r scan is the sample grown in the digital method, indicating a Ca/O ratio of 65% and a Mg/O ratio of 35%. The higher calcium content at the surface is expected since the last shutter open for the digital growth was the calcium. Since th e 10 second exposure would lay down approximately one of two angstrom of ma terial, the slight increase in calcium concentration is expected.................................................................................................107 4-22 Auger depth profiling with argon spu ttering, showing the comparison between ternary oxide growths with the continuous a nd digital growth modes. As seen, the large segregation of oxygen is not present as earlier samples, since the growth rate has been reduced..............................................................................................................108 11

PAGE 12

4-23 AFM images of continuously grown MgCa O at 500C. A) Five micron scans area of the sample with average roughness of five angs troms. B) Closer view of the sample, showing average roughness of just over three angstrom.................................................109 4-24 RHEED image of MgCaO during growth in the MBE system. The spots in columns are indicative of a single crystall ine surface with some roughness.................................110 4-25 XRD scan of lattice-matched MgCaO showi ng no indication of the (222) oxide peak..111 4-26 XRD spectrum comparing the two growth modes of digital and continuous. Both samples are lattice matched to the GaN substrate. A) Full 2-Theta scan of the samples. B) Tight view of the GaN (004) peak showing no additional oxide peak.......112 4-27 TEM images of lattice-matched MgCaO on GaN. The images are of the same sample, simply two different areas. The important note is that the large threading dislocations in the GaN s ubstrate are not propagating in to the oxide. Also, the images are inverted to th at the oxide is now above the GaN, for clarity.........................113 4-28 XRD stress measurements performed on CaO. The expansion of the lattices of both GaN and CaO can be seen, as the CaO peak shifts more than the GaN..........................114 4-29 High temperature XRD measurements of MgO on GaN. As seen, the MgO lattice expand at a slightly larger am ount than the GaN substrate.............................................114 4-30 Ternary alloy rich with MgO during HT -XRD measurements. The phase separation of the ternary oxide is shown as th e sample is heated above 500C.................................115 4-31 Repeat of the HT-XRD from Figure4-30. The sample was heated to 800C, back to room temperature, then heated a se cond time to confirm no phase change....................115 4-32 Full XRD scan of the ternary oxide from Figure 4-31, after multiple heatings. The scan shows that the sample did pha se separate into the binaries.....................................116 4-33 HT-XRD measurement of a calcium oxi de rich ternary alloy. Again, phase separation occurred..........................................................................................................116 4-34 Room temperature, 300C, and 700C XRD of the sample in Figure 4-33, showing the evolution of the binaries...................................................................................................117 4-35 Room Temperature scan of the sample in Figure 4-34, indicating the binary oxide peaks, the GaN (004) peak, and the GaN (003) forbidden peak......................................118 4-36 HT-XRD measurement of a digital 10: 10 oxide sample. The sample was latticematched with GaN at room temperature, then as the measurement progresses, so did the evolution of the binary oxides....................................................................................118 12

PAGE 13

4-37 XRD scan of a lattice matched MgCaO samp le that was annealed in the MBE growth chamber at 800C for one hour. No phase separation was seen, indicating that the phase change is pressure dependant.................................................................................119 4-38 AFM scans of the lattice-matched MgCaO sample that was annealed during the HTXRD measurements. A) Before HT-XRD, 5m scan area, z-scale of 10nm. B) Before HT-XRD, 1m scan area, z-scale of 5nm. C) After HT-XRD, 5m scan area, z-scale of 30nm. D) After HT-XRD 1m scan area, z-scale of 20nm..........................120 4-39 AFM scans of the CaO sample that was annealed during the HT-XRD measurements. A) Before HT-XRD, 5m s can area, z-scale of 15nm. B) Before HT-XRD, 1m scan area, z-scale of 15nm. C) After HT-XRD, 5m scan area, zscale of 100nm. D) After HT-XRD, 1m scan area, z-scale of 100nm..........................121 4-40 Phase diagram for MgO and CaO, above 1600C.............................................................122 4-41 Isothermal section of the ternary diag ram of magnesium, oxygen, and calcium at 1227C...............................................................................................................................122 4-42 Mounting procedure for oxide samp les grown for MOCVD GaN growths. Molybdenum blocks with quarter wafer cutout s. A) Back of holder showing clip and wire mounting. B) Front of the holder............................................................................123 5-1 Molybdenum sample holder block on hotpl ate used to perform indium mounting of samples for MBE growths...............................................................................................144 5-2 RHEED image of GaN on GaN growth ha ving a (3X2) reconstr uction, that emerges shortly after the start of growth........................................................................................144 5-3 Tapping mode AFM of initial results of GaN on GaN MBE growth. A) 1m area of the growth at 630C. B) Larger area view showing the pits which originate from the Uniroyal GaN substrate....................................................................................................145 5-4 AFM tapping images of MBE GaN on GaN at different gallium fluxes, 1m2 areas A) Ga flux 1.77E-7 Torr BEP, Ra: 1.01nm, ver tical scale: 10nm. B) Ga flux 2.05E-7 Torr BEP, Ra: 1.02nm, vertical scale: 10n m. C) Ga flux 2.53E-7 Torr BEP, Ra: 2.03nm, vertical scale: 20nm...........................................................................................146 5-5 AFM tapping images of MBE GaN on GaN at varying substrate temperatures, 1m2areas. A) Ts=725C, Ra: 2.97nm, vertic al scale: 20nm. B) Ts=700C, Ra: 1.57nm, vertical scale: 10nm. C) Ts=630C, Ra: 2.89nm, vertical scale: 20nm.............147 5-6 AFM tapping images of MBE GaN on GaN at varying nitrogen flows, 1m2. A) Flow=0.5sccm, Ra: 3.22nm, ver tical scale: 40nm. B) Flow=1.0 sccm, Ra: 3.89nm, vertical scale: 40nm. C) Flow=1.5sccm Ra: 1.48nm, vertic al scale: 20nm...................148 5-7 Optimized MBE GaN growth conditions on UOE GaN..................................................149 13

PAGE 14

5-8 In-house grown MOCVD GaN showing growth terraces. A) 1mX1m area. B) 5mX5m area................................................................................................................150 5-9 Optimized MBE GaN conditions on in -house grown MOCVD Ga N. A) 1mX1m area. B) 5mX5m area.................................................................................................151 5-10 SEM images of initial results of MB E GaN on MgCaO, indicating a rough film. Also included is the EDS spectrum of the film, showing the calcium and oxygen from the underlying layer. The magnesium peak is barely visible ju st to the right of the low energy gallium peak. A) Low ma gnification SEM. B) Higher magnification SEM. C) EDS spectrum..................................................................................................152 5-11 RHEED images of first MBE GaN on MgCa O. A) Pattern after 5 minutes of GaN growth, taking on a slight poly-crystalline pattern. B) Patte rn at the end of the one hour growth showing single crystalline 3D growth.........................................................153 5-12 First MBE GaN on MgCaO growth. A)AFM image showing starting MgCaO surface morphology with smooth terraces. B) MBE GaN growth on MgCaO, with Ra=6nm. C) Closer view of the GaN growth showing bumpy morphology..................154 5-13 XRD scan of the first MBE GaN on Mg CaO which is shown in Figure 5-12. The GaN peak (alpha and beta) are shown with no additional peaks. This indicates that the oxide stayed single phase, and is lattice matched to the GaN....................................155 5-14 RHEED images of GaN growth at 700C on MgCaO in MBE. A) Pattern after 10 minutes of GaN growth. B) After 1 hour growth. C) At the end of the two hour growth..............................................................................................................................156 5-15 AFM images showing MBE GaN growth on MgCaO and the different morphology produced from substrate temperatures at 700C versus 800C, vertical scales of 80nm and 30nm for images A and B, respectively. A) 700C substrate temperature, Ra=8nm, 1mX1m area. B) 700C, 5 mX5m area. C) 800C substrate temperature, Ra=3nm, 1mX1m area. D) 800C, 5mX5m area...............................157 5-16 AFM images showing MBE GaN growth on MgCaO with AlN exposure and the different morphology produced from substrat e temperatures at 700C versus 800C. A) 700C substrate temperature, Ra=5 nm, 1mX1m area. B) 700C, 5mX5m area. C) 800C substrate temperature, Ra=4nm, 1mX1m area. D) 800C, 5mX5m area................................................................................................................158 5-17 RHEED images showing MBE GaN grow th on MgCaO after a short AlN exposure prior GaN growth. A) After 15 minutes. B) After 1 hour. C) End of growth of 2 hours.................................................................................................................................159 5-18 AFM images of MBE GaN on binary oxides of MgO and CaO. A) MBE GaN on CaO (B), Ra=7.03nm vertical scale of 70nm. B) MgO, substrate for (A), Ra=1.23nm vertical scale of 15nm. C)MBE GaN on CaO (D), Ra=8.54nm 14

PAGE 15

vertical scale of 70nm. D) CaO substrate for (C), Ra=0.815nm vertical scale is 15nm................................................................................................................................161 5-19 XRD scans of the binary oxides used in the growths of Figure 5-18, the scans are also in Log scale to emphasize the oxide peaks in relation to the large GaN signal since the oxides are only a few hundred thick. A) CaO substrate showing the (222) CaO peak, the GaN peak, along with sm all peaks from the sapphire substrate. B) MgO substrate shown the (222) MgO peak, the GaN p eak, and additional stray peaks from the substrate...................................................................................................162 5-20 X-ray diffraction spectrum of the lattic e-matched MgCaO film used for the GaN films grown to larger thicknesses....................................................................................163 5-21 RHEED images of MgCaO at different thicknesses. A) Approximately 200nm thick film, showing signs of surface charging. B) Approximately 100nm, showing rotation of the film.........................................................................................................................164 5-22 AFM images of MgCaO different thicknesses. A) Approximately 200nm. B) Approximately 100nm.....................................................................................................165 5-23 3D AFM images showing the morphological differences between the two growths, grown on the same block, with differing st arting oxide substrate thicknesses. A) GaN with starting substrate oxide of 100nm (Figure 5-22b). B) GaN with substrate oxide of 200nm (Figure 5-22a)........................................................................................166 5-24 Optical profilometer scan of the clippe d area from the sample in Figure 5-23, grown for 500minutes. The Y-profile indicated a step height of around 400nm, which corresponds to the cross-section SEM images.................................................................167 5-25 Field-Emisson SEM images showing a cr oss-section of the GaN on MgCaO sample grown for 500 minutes. The thicknesses of the oxide and GaN are approximately 100nm and 400nm, respectively. A) 20,000X secondary electron image of the sample, showing the sapphire (bottom) then the MOCVD GaN, followed by the MgCaO, and finally the MBE GaN on the top. B) 150,000X image of the sample showing a more detailed view of the MBE GaN growth.................................................168 5-26 AFM images showing the rough mor phology of the 20 hour MBE GaN growth on MgCaO. A) 252m area. B) 12m area..........................................................................169 5-27 AFM images showing the progression of growths from the starting MOCVD GaN to the final MBE GaN. A) MOCVD GaN, s howing nice terraced growth. B) MgCaO growth showing replication of the MOCVD GaN substrate. C) MBE GaN growth on the MgCaO showing a rough morphology.......................................................................170 5-28 FE-SEM images of the 20 hour MBE GaN growth shown in Figure 5-26. Crosssection images depict a columnar structur e of the MBE GaN. A) Image showing the columnar structure of MBE GaN on the MgCaO on MOCVD GaN on Sapphire 15

PAGE 16

(bottom). B) Higher magnification. C) Showing delamination of both MBE GaN and oxide from the oxide and MOCVD GaN, respectively.............................................171 5-29 XRD scan of thick MBE GaN on MgCaO, from Figure 5-28, showing an oxide with slightly smaller lattice than the GaN. Th ere is no sign of phase separation of the oxide, again confirming ear lier measurements................................................................172 5-30 Secondary electron and cathodoluminesence (CL) imagines of the thick 20 hour MBE GaN growth with a comparison with standard in-house grown MOCVD GaN. A) Secondary electron image of the MBE GaN surface at 1,000X. B) Secondary electron image of the MBE GaN surface at 5,000X. C) CL image of image A. D) CL image of image B. E) CL image of MOCVD GaN 1,000X. F) CL image of MOCVD GaN 5,000X showing dark areas corr esponding to threading dislocations.....173 5-31 Photoluminesence (PL) spectra of GaN samples from a standard MOCVD in-house run, and the thick 1.5m MBE sample. A) PL spectra from MOCVD GAN showing band-edge at 360nm, and some blue and ye llow emission from defects. B) PL spectra from thick MBE GaN on MgCaO s howing band-edge, no blue emission, but increased yellow emission from nitrogen vacancies........................................................174 5-32 XPS spectrum of MgCaO before the 700C vacuum anneal to examine hydroxide decomposition.................................................................................................................. 175 5-33 Magnesium 2s peak before and after h eating. A) Before heating, the broadness of the peak to higher energies signifies a hydroxide layer. B) After heating, only one peak indicating no magnesium hydroxide.......................................................................176 5-34 Oxygen 1s for the pre-heated and post-heat ed sample. The peak changes are due to the surface hydroxide layers being decomposed during the vacuum anneal. A) Before annealing. B) After annealing.............................................................................177 5-35 Calcium 2p before and after heating. The far left peak at higher energies corresponds to calcium hydroxide. A) Be fore heating, showing a strong hydroxide peak. B) Post heating, showing a la rge reduction in the hydroxide peak.......................178 5-36 XPS full spectrum scan of the MgCaO sa mple after heating to 700C in the MBE chamber. The carbon peak seen befo re heating was largely reduced.............................179 6-1 AFM images showing the starting sapphire surface used for GaN MOCVD growths, having roughness of around 1 A) 300nm scan area. B) 1 micron scan area. C) 5 micron scan area..............................................................................................................2 07 6-2 Standard data collected during MOCVD gr owth run on the TurboDisk reactor. The emissivity of the surface allows temperat ure measurement while the reflectivity off the sample surface indicates surface roughne ss and layer thickness. The numbers indicate the various la yers of the recipe...........................................................................208 16

PAGE 17

6-3 Image taken in MBE of 180nm Ti/W back side coating. Thermocouple temperature at 700C, yet, surface is cooler since coati ng was not very efficient at absorbing the IR radiation from the heater.............................................................................................209 6-4 AFM of sample grown with 180nm Ti/W b ackside coating. Lack of strong terraces indicate a growth temperature lower th an the thermocouple reading of 500C................210 6-5 SEM images of the Ti/W coating. A) Middle of wafer at 500X magnification. B) Middle of wafer at 5,000X magnification. C) Edge of the wafer at 500X magnification. D) Edge of th e wafer at 5,000X magnification.......................................211 6-6 Wafer image in MBE with 500nm tungsten on the back of the wafer. The sample had more blackbody radiation, indicating a higher temperature than the thinner coating. The thermocouple reading of 700C is a more accurate measure of the sample temperature with the thicker coating...................................................................212 6-7 AFM images of the MgCaO grown on the wafer with 500nm of tungsten coating on the back. A) Five micron scan. B) One micron scan.....................................................213 6-8 Optical images of the metal backside co atings for samples after MOCVD exposure. A) Molybdenum. B) Tungsten. C) Tantalum. D) Silicon.............................................214 6-9 Optical image showing the silicon contamination around the e dge of the sapphire wafers used for GaN growth............................................................................................215 6-10 EDS line scan across the edge of the wafer shown in Figure 6-9, confirming the silicon contamination.......................................................................................................215 6-11 Ring of bad growth after MOCVD GaN gr owth on the silicon wafer which had a ring of silicon on the front side...............................................................................................216 6-12 Some etching of the silicon backside co ating during initial atte mpts are removing the silicon on the front of the sapphire wafers from Union Carbide.....................................217 6-13 Image showing complete backside coati ng removal after 12 minutes in the silicon etchant..............................................................................................................................217 6-14 AFM image showing contamination on the epi-surface of the sapphire wafer from the silicon etching............................................................................................................ 218 6-15 SEM images showing the ring left after etching the silicon fro m the front of the wafer. A) BSE image. B) SE image...............................................................................219 6-16 EDS spectrum of the ring area shown in Figure 6-15......................................................219 6-17 AFM images of the MOCVD GaN grow th on the sapphire wafer after silicon etching. A) Five micron scan area. B) One micron scan area showing sharp terraces, some threading dislocations, and having roughness of 1 ..............................................220 17

PAGE 18

6-18 AFM images of 100nm MgCaO before be ing annealed in the MOCVD. A) One micron scan length, showing average roughness of half a nanometer. B) Five micron scan length.......................................................................................................................221 6-19 AFM images of the MgCaO after high te mperature annealing in ammonia in the MOCVD. A) Five micron scan area. B) One micron scan area having roughness near 20nm. C) Cross-section scan show ing feature heights of nearly 30nm..................222 6-20 3D animation of the MgCaO after MOCVD anneal........................................................223 6-21 RealTemp data of the MgCaO anneal..............................................................................223 6-22 XRD scan of the MgCaO sample annealed in the MOCVD. A) Before annealing. B) After annealing, showing a s light increase in the MgO p eak, and a slight broadening around the GaN peak.......................................................................................................224 6-23 EDS results of the front of the a nnealed MgCaO to verify no additional contamination.................................................................................................................. .225 6-24 AFM line scan showing how the sapphire will form pits under ammonia annealing at high temperatures.............................................................................................................2 26 6-25 RealTemp data of growth of GaN on MgO.....................................................................226 6-26 SEM images of MOCVD GaN on MgO, s howing areas of good growth and pitting. A) Region of somewhat smooth growth. B) Area with high density of pits..................227 6-27 TEM image of the MOCVD growth of GaN on MgO. Image indicates nucleation layer induced stress, and oxide bloc king of threading dislocations.................................228 6-28 RHEED image of MBE grow th of scandium oxide.........................................................229 6-29 XRD scan of scandium oxide s howing oxide peak around 65 degrees 2 ......................229 6-30 AFM images of scandium oxide for MOCV D GaN substrate. A) Large area scan. B) Smaller area scan having roughness of 0.6nm............................................................230 6-31 MOCVD quarter wafer sample holder showing degradation..........................................231 6-32 RealTemp data from first attemp t at MOCVD GaN on scandium oxide........................231 6-33 SEM images of the aborted GaN on scandium oxide sample, showing uniform coverage with GaN island nuclei. A) SE image at 2,000X. B) SE image at 20,000X..232 6-34 SEM images of the aborted GaN on scandium oxide sample after growth continued, showing some coalescence of the GaN island nuclei. A) SE image at 2,000X. B) SE image at 20,000X.............................................................................................................23 3 18

PAGE 19

6-35 Contact angle measurements taken for various substrates for GaN growth. The measurement was conducted with a rame-h art goniometry with DI water. The change in contact angle indicates a differe nce in surface energy resulting in varying surface mobility of GaN on the surface under high temperature growth regimes. A) Standard sapphire substrate epi-polished w ith no treatment. B) Sapphire surface of wafer with silicon coated backside after etching to remove silicon on epi-side of wafer. C) Lattice matched MgCaO. D) MgO. E) Sc2O3. F) Standard MOCVD GaN. G) GaN sample interrupted right af ter island formation at 1090C, of the low temperature nucleation layer............................................................................................234 6-36 Oxide samples after contact angle measur ement showing affected areas. A) MgO. B) MgCaO........................................................................................................................236 6-37 AFM images showing the MgCaO surface before and after contact angle measurement with DI water. A) Before measurement. B) After measurement showing etching effect of the water.................................................................................237 6-38 Contact angle measurements of MgO and MgCaO showing the change of angle over a two minute exposure to the water drop. Red and blue lines indicate the location of the drop right after initial contact. The bl ack and white image shows the actual water drop after two minutes of exposure. A) MgO. B) MgCaO............................................238 6-39 XRD scan of the scandium oxide capped MgO...............................................................239 6-40 AFM scans of the Sc2O3 capped MgO. A) Cross-section showing growth features around 10nm tall. B) Five micron scan length having roughness of 1.8nm...................239 6-41 RealTemp data showing the growth of GaN on the Sc2O3 capped MgO. Good oscillations are indica tive of nice growth........................................................................240 6-42 Optical micrographs showing the GaN growth on Sc2O3 capped MgO. Hexagonal growth pits can be seen across most of the sample surface. A) 20X. B) 100X.............241 6-43 SEM images of the MOCVD GaN on Sc2O3 capped MgO. A) 1,000X magnification. B) 10,000X magnifica tion of hexagonal surface pit...............................242 6-44 AFM images of MOCVD GaN on Sc2O3 capped MgO showing terraces growth. A) 5m by 5m scan. B) 1m by 1m s can, having average roughness of 1nm................243 6-45 Cross-section FE-SEM of thick MOCVD GaN on oxide. The figure is labeled to show the stack structure of growths.................................................................................244 6-46 Additional cross-secti on FE-SEM images of the GaN grown on oxide in the MOCVD...........................................................................................................................245 6-47 TEM image of the MOCVD GaN on Sc2O3 capped MgO..............................................246 19

PAGE 20

6-48 PL spectrum comparisons during the multip le stages of growth of MOCVD GaN on oxide. The reduction in yellow emission with a strong band-edge emission is promising of high quality GaN........................................................................................247 6-49 PL spectrum comparison between the growth on MBE oxide, versus a standard GaN sample grown on sapphire...............................................................................................248 6-50 CL images of the GaN on MBE oxide compared to standard MOCVD GaN on sapphire, with gray levels equal. The growth on MBE oxide shows much reduced threading dislocations, which can be seen in the GaN on sapphire images which are dark areas in the images. A) GaN on sa pphire at 1,000X. B) GaN on sapphire at 5,000X. C) GaN on MBE oxide at 1,000X. D) GaN on MBE oxide at 5,000X............249 6-51 RealTemp data of MOCVD GaN grown on MgCaO capped with MBE GaN................250 6-52 SEM images of MOCVD GaN on MgCaO capped with MBE GaN. Growth shows terraces with some incomplete coalescence.....................................................................251 6-53 AFM images of the MOCVD GaN on MgCaO capped with MBE GaN. A) 20m scan area showing large growth platelet s. B) 1m scan showing roughness around 1nm..................................................................................................................................252 6-54 Cross-section FE-SEM of the MOCV D GaN on MgCaO capped with MBE GaN. The image indicates incomplete coalescence of the GaN top film..................................253 6-55 PL spectrum of the MOCVD GaN on MgCaO capped with MBE GaN. The signal was very weak from the sample surface being quite uneven...........................................254 6-56 CL images of the MOCVD GaN on MgCa O capped with MBE GaN, showing some emission from the growth edges of the incomplete coalescence. A) 5,000X. B) 1,000X..............................................................................................................................255 7-1 MgCaO starting surface for etching experi ment showing terraced growth. A) 1m scan having roughness less than 1nm. B) Larger area view with roughness around 1.5nm...............................................................................................................................262 7-2 AFM images of MgCaO exposed to vari ous solutions for five minutes. The AFM images on the left are the height images, while those on the right show the phase image. A) Acetone exposure. B) Metha nol exposure. C) DI water exposure..............263 7-3 AFM tapping image showing the 5 micron scan area that was imaged just before this larger view. The scan outline is from th e movement of the hydroxide layer on the surface of this sample which was exposed to DI water for 5 minutes.............................264 7-4 Etch rate results of MgCaO and GaN in a chlorine plasma, showing over an order of magnitude less etching for the oxide compared to GaN..................................................266 20

PAGE 21

7-5 Etch rate results of MgCaO and GaN in methane plasma, showing almost an order of magnitude less etching for the oxide compared to the GaN............................................267 7-6 Optical profilometer results from dry et ching of GaN. Similar results from the MgCaO dry etch resulted.................................................................................................268 7-7 SEM images showing the dry etching of GaN and the resulting features set forth through the lithography process. A) Highe r magnification of GaN etch. B) Low magnification of GaN etch features.................................................................................269 7-8 ICP etching of GaN and MgCaO in chlori ne and methane chemistry. A) Chlorine chemistry. B) Methane chemistry...................................................................................270 8-1 In the above SEM images, a BSE image indicates the compositional difference between the GaN bar and the MgCaO field. Al so, an EDS line scan verifies a rise in calcium and oxygen in the location of the MgCaO.........................................................281 8-2 Above are AFM images of the GaN bars on MgCaO showing surface morphology between the materials. A section analysis is included, indicating a vertical height of just over 200nm for the bar. This height in dicates how the oxide acts as an etch stop, since the GaN growth thickness was approximately 200nm...........................................282 8-3 SEM images of metal contacts on GaN bars on a field of MgCaO.................................283 8-4 Auger analysis of the sample in Figure 83. A) Spot analysis of the oxide field. B) Spot analysis of the GaN bar...........................................................................................284 8-5 Shown above is the process of transferring the grown film or features onto a foreign substrate. The features can be from the millimeter to micrometer scale........................285 8-6 Optical profilometer scan of the circ ular PR features produced onto the MgCaO surface, to simulate GaN. Width of the f eature is 350 microns. A) Cross-section of the sample showing height and width. B) 3D rendering of the feature..........................286 8-7 Optical profilometer scan of the recta ngular PR features produced onto the MgCaO surface, to simulate GaN. Feature dimensions are approximately 100 by 500 microns. A) Cross-section of the sample showing height and width. B) 3D rendering of the feature....................................................................................................287 8-8 Optical microscope images of the PR features on MgCaO. A) 5X overview indicating the features of interest. B) Corner of a rect angular feature. C) Edge of a circular feature.................................................................................................................288 8-9 Optical images of the undercutting of PR, showing the differences with acid concentration. A) 5X image with 2% acid. B) 100X image with 2% acid. C) 5X image with 20% acid. D) 100X image with 20% acid....................................................289 8-10 Oxide thickness versus lateral etch rate...........................................................................289 21

PAGE 22

8-11 Optical images showing etching differenc es with oxide composition. A) MgO, 2% acid. B) MgO, 20% acid. C) Mg0.25Ca0.75O, 2% acid. D) Mg0.25Ca0.75O, 20% acid.....290 8-12 Etch rate versus oxide composition.................................................................................290 8-13 Images of sample submerged in acid for second time, showing minimal additional etching. A) Circular feature 1st etching. B) Circular feature 2nd etching. C) Rectangular feature 1st etching. D) Rectangular feature 2nd etching..............................291 8-14 Optical profiler images of the PR foldi ng after first round out etching. A) Circular feature, 2% acid. B) Rectangular feature, 2% acid. C) Ci rcular feature, 20% acid. D) Rectangular feature, 20% acid....................................................................................292 8-15 Optical images of the 1.5 micron thick MBE GaN film that had been successfully lifted off the substrate and transferred onto another substrate. A) Low magnification showing a scale bar for the size of the circul ar GaN film. B) Image showing areas of GaN film that had not been mounted adequately to the mounting wax. C) Region that had a particle on the oxide and the MBE GaN grow around it, showing terraced growth. D) Crack in film from poor mounting to the wax..............................................293 8-16 AFM images of the underside of the rele ased GaN film at various scan areas. Roughness around 3-4nm................................................................................................294 8-17 AFM images of the oxide upon which the GaN sample was grown...............................295 8-18 Cross-section FE-SEM image of the th ick MBE GaN on oxide sample. This shows the columnar type growth which results in the granular textur e seen by the AFM of the interface with the oxide..............................................................................................296 8-19 Optical images of a sample of 5 micron thick MOCVD GaN removed from the GaN substrate after etching the intermediate oxide substrate. A) Image showing the entire 3X4mm sample. B) 5X magnification im age showing the GaN surface that was removed from the oxide...................................................................................................297 8-20 AFM images showing the surface morphology of the 5 micron GaN film that was separated from the substrate through etchi ng the oxide. This was the surface that was in contact with the oxide during gr owth. A) 20 micron scan length having roughness of 5.44nm. B) smaller area scan showing morphology.................................298 8-21 Released 1.5 micron GaN film from substr ate and transferred to a glass slide. The sample lifted-off the substrate was a half inch circle.......................................................299 9-1 PL data of ZnO deposited upon MgCaO and Sapphire using PLD. Growth upon MgCaO shows a four times increase in band-edge emission for ZnO............................306 22

PAGE 23

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GROWTH AND CHARACTERIZATION OF GALLIUM NITRIDE ON LATTICEMATCHED MAGNESIUM CALCIUM OXIDE By Andrew Phillip Gerger August 2009 Chair: Brent Gila Cochair: Cammy Abernathy Major: Materials Science and Engineering As the interest for Gallium Nitride ba sed materials and devices expand, focus on controlling the quality of material is becoming a priority. In th is study, the focus of GaN growth upon a lattice tunable oxide is inve stigated. The oxide utilized is MgCaO. This oxide has the unique feature of maintaining singl e phase as the metal ratios are varied. Lattice parameter of the oxide is therefore able to be changed and matched to that of GaN. With a matched substrate, the subsequent GaN growth has less strain which re sults in decreased level of defects in the film, which would normally lead to degraded performance and early device failure. Higher quality material lends itself to further applications. Applications such as nanobar fabrication and oxide etching resulting in GaN film tr ansfer to other substrates. Another issue encountered in the expansion of GaN devices is heat generation. Heat buildup in high-power RF devices has been the cr utch for most material systems. Having the ability to dissipate heat from the device perm its greater performance at higher operating loads and longer device lifetime. Ther efore, to transfer GaN onto hi gh thermal conductivity substrates would warrant a greater range for device operatio n, as the heat is more efficiently dissipated through the substrate. The MgCaO substrate can be easily etched away while leaving 23

PAGE 24

overgrowth nitride intact, lending itself to the concep t of lift-off or pick-and-place. Transfer of GaN onto diamond substrates or other comparab le materials would enable the high thermal conductivity and would faci litate the GaN devices to work at higher temperatures and power loads. The substrate transfer could be accomplished after the material growth or after the device fabrication was complete. In this work, the growth and characteriza tion of single crystal MgCaO epitaxial films have been demonstrated. This has been dem onstrated on single crystal GaN epifilms grown on sapphire, but SiC can be employed as well, el iminating the underlying GaN epifilm. Oxide thicknesses over 100nm with surface RMS roughness of less than 0.5nm have been achieved and employed as a successful substrate for nitride based overgrowth. 24

PAGE 25

CHAPTER 1 INTRODUCTION Motivation With todays need for high speed and high power electronics, the role of material quality is of utmost importance. The presence of defects in the semi-conducting material leads to degraded performance and early device failure. The III-nitride semiconductors have advantages over standard silicon devices in this high power field. Advantages which include operation at higher temperatures for longer durations, larger power density, and many other benefits for the electronics field. The III-nitride of focus for this research is Gallium Nitride (GaN.) With a large direct bandgap, GaN is suitable for both power applicat ions and optical applications. Since the demonstration of GaN-based viol et LEDs in 1970s, gallium nitride and other III-nitrides have been researched for many devices including HE MTs and Laser Diodes. Unfortunately, the potential of the III-nitride semiconductors has ye t to be reached. The speed and power capability of devices have been hindered by the defects associated with the growth of the material. At first glance, growing films on foreign subs trates may seem quite straightforward, but heteroepitaxy has its complications. The larges t issue encountered with epitaxial growth is minimizing the strain between the starti ng substrate and the growing material.1,2 This strain is greatly based on the lattice of th e two material systems. If the lattice parameters are too different, strain develops at the interface which re sults in misfit. This strain results in the possibility of film cracking or buckling under the stress. Finally, the strain may hinder epitaxy and promote 3D growth resulting in a ve ry rough surface morphology. Until heteroepitaxy growth can result in low-defect material, the fu ll potential of Gallium Nitride will be hindered. 25

PAGE 26

Another issue with growth of different mate rials is the thermal expansion differences. When the materials have large expansion variances, strain will develop at different temperatures. This strain can result in defect formation in the growing material.3,4 Large GaN substrates for homoepitaxial growth are not refined and therefore result in a very poor quality template for further growth. The cost of free-standi ng GaN substrates also limits their use, as a single 2 wafer costs nearly $10K.5 At the present, heteroepitaxy is the method used for most all of toda ys commercial and rese arch applications. With the ability for the starting substrate to be selectively removed, the GaN can be transferred onto a substrate with more desirable properties. To remove this subs trate without damage to the top GaN material and to make this process rapid and technically simp le would be very beneficial to multiple high power applications. This is the ma in application for this research. Dissertation Outline The objective of this study wa s to grow a lattice matching oxi de to gallium nitride that could in turn be used as a template for furthe r gallium nitride growths. The study also included complete characterization of the oxide for use in for MOSFET technology. The background for this research which includes a literature review of pertinent research is contained within Chapter 2. It also contains general information on the growth of gallium nitride and material characteristics. Along w ith the properties, information on the defects normally present in the material is examined. Chapter 3 discusses the experimental approach and instrumentation utilized throughout the cour se of the research. Moving to Chapter 4, the growth of MgCaO by molecular beam epitaxy is examined. Within Chapter 4, the lattice matching criteria and the methods for the successf ul lattice matching of oxide to gallium nitride are presented. Within Chapter 5, the lattice parameter of MgCaO is adjusted to match that of the GaN substrates. The reason for this shifting ba sed on the ratio of the binary oxides, MgO and 26

PAGE 27

CaO, is explained with supporting characterization. Chap ter 6 includes the stab ility of the oxide in the sense of wet and dry etching and the re sultant morphology of the surface after various etchings. Next, Chapter 7 covers the growth of gallium nitride on Mg CaO by molecular beam epitaxy. The MBE growth of GaN is supported by references to other MBE GaN experiments. The MOCVD growth of gallium nitride is presented in Chapter 8. The multiple variations of substrates treatments and growth methods are described and thoroughly examined. The results of bad growths are examined and some interesting results were i nvestigated. Moving to Chapter 9, the experiments involved in th e transfer of the top GaN onto a new substrate via the removal of the oxide is presented. This culmination of freely released GaN reveals the ultimate application of this research which is the transf er of the GaN onto a more desirable substrate. Finally Chapter 10 concludes the research with some final remarks and thoughts for future works and applications. 27

PAGE 28

CHAPTER 2 LITERATURE REVIEW AND BACKGROUND Substrate Issues At first glance, growing films on foreign s ubstrates may seem quite straightforward, but heteroepitaxy has its complications. The larges t issue encountered with epitaxial growth is minimizing the strain between the starting substrat e and the growing material. This strain is greatly based on the lattice of th e two material systems. If the lattice parameters are too different, strain develops at the interface which results in misfit. These misfit dislocations propagate through the material in the growth dire ction to minimize the strain energy associated with the atom spacing. This strain also lends to the possibility of film cracking or buckling under the stress. Finally, the strain may hinder epita xy and promote 3D growth resulting in a very rough surface morphology. Until heteroepitaxy growth can result in low-de fect material, the full potential of Gallium Nitride will be hindered. Large interest has been focused on this strain relaxation arena. The concept of growing buffer layers of various ma terials before the main GaN growth has been utilized for some time, the most prominent be ing Aluminum Nitride (A lN). Yet, these buffer layers are at times, unwanted variables in the fi nal device. Therefore, to grow the GaN on a lattice-matched, abrupt interface would result in a l eap in the direction of low-defect material. III-Nitride Crystal Structures The III-nitrides crystallize in tw o phases, hexagonal wurtzite ( -phase) and cubic zincblende ( -phase) polytypes. The wurtzite phases of AlN, GaN and InN have direct bandgaps of 6.2, 3.4, and 0.9eV respectively. The cubic zi ncblendes have smaller bandgaps of 4.9, 3.2, 0.9eV respectively. Hexagonal wurt zite phase is the most ther modynamically stable structure and therefore, almost all scientific research as well as commercial devices, are based on this 28

PAGE 29

polytype.6 The wurtzite structure co nsists of two interpenetra ting hexagonal close packed sublattices. One lattice consists of nitrogen a nd the other contains the group III atom. The two sublattices have an offset along the c axis by 5/8 the cell height. Fi nding a matching substrate for growth has been investigated, lead ing the research to various oxides. Previous published oxide substrates for GaN growth include Gadolinium Oxide (Gd2O3), Zinc Oxide (ZnO), Lithium Gallate (LiGaO ), and2 Lithium Aluminate (LiAlO ). Each substrate has specific advantages and disadvant ages, with some similarities. 2Current Alternate Substrates Throughout the years of gallium nitride advancements, research has replaced the normal sapphire substrate for many others. The constant desire for less st rain between the substrate and the growing film has lead researchers to try more other crystalline substrates. Many of the experimental substrates have ha d marginal success but all still ha ve drawbacks that the next try to overcome. The alternative substrate propert ies are tabulated in Table 2-1, along with the standard substrates of sa pphire and silicon carbide. Lithium Gallate Lithium Gallate (LGO or LiGaO2), has been studied for pi ezoelectric devices since 1965.7 LGO is an orthorhombic crystal with a, b, and c lattice constants of 5.407, 6.379, and 5.014 respectively. GaN growth on LGO is reporte d to have better qual ity than when grown upon sapphire or SiC. This is resultant of the fact that the average la ttice mismatch between LGO and GaN is around 0.9%.7 This low mismatch is attributed to the hexagonal symmetry of the crystal which is slightly distorted from the wurtzite structure. One issue that plagues LGO is the large interface roughness. Ev en with the roughness, the critical thickness of GaN on this substrate is quite large. Enough material for de vices is able to be grown before relaxation occurs. Lithium diffusion into the film has been re ported to be an issue with the film growth, but 29

PAGE 30

using a surface nitridation prio r to growth limits the diffusion of both the Lithium and Oxygen. LGO is a polar material, therefor e, resulting in the ability to grow single polarity material, allowing for good carrier confinement in heterointerfaces. A main drawback for LGO is its reactivity. It is highly volatile above 850C, which may limit its uses for certain applica tions, like post-growth annealing. Also, the as-grown wafers of LGO contain antiphase domains where the growth kinetics are vastly different. These antiphase domains are small regions of anion terminated material on the cation terminated face. The importance of these antiphase domains is that GaN will not adhere to these anion, oxygen, terminated regions.8 As far as electrical property results, they have been limited. What is known is that growth upon LGO substrates results in high carrier concentrations and low mobility.9 LGO is also an insulator, thus allowing for device insu lation with the normally re quired implant isolation step in processing. Also, consistency in vendor manufacturing of the substrates is unreliable, even from the same vendor.7Film-substrate removal has been shown to work with the LGO substrates through chemical means. The film is then bonded to a host substrat e. The size of the transferred sample reported is 1X3mm.7Lithium Aluminate Lithium Aluminum Oxide (LAO) has a tetragonal crystal structure with lattice parameters; a=b=5.1687, and c=6.2679. For GaN grown by vapor-phase epitaxy, the lattice mismatch with LAO is -1.4% in the c-direc tion and -0.1% in the a-direction.10 LAO substrates are grown by the Crochralski method, which re sults in the substrate being po lished instead of an epilayer surface. 30

PAGE 31

The GaN films can be separated from the L AO substrate by two met hods. If the film is quite thick, the film will delaminate spontaneously. Yet, for thin films, or if sidewall growth occurs, etching is required in heated HCl. As can be inferred, there is measurable stress in the films grown upon LAO, as compared to the stress-free films grown upon LGO. It is reported that no surface morphology benef it accrues to the use of LAO as a template over sapphire substrates.10 Also, the films grown on LAO dem onstrated impurity concentrations identical to the best films grown on sapphire. The GaN films did exhibit lower oxygen, carbon, and chlorine contamination compared to films on LGO substrates under same growth conditions. Lithium impurities were not detected in the Ga N film, indicating that LAO is chemically and thermally very stable.10Gadolinium Oxide Gd2O3 has a lattice constant of 3.86, which is larger than the GaN overgrowth resulting in a 20% mismatch. Unlike GaN growth on sapphi re, where a 30 in-plane rotation occurs to reduce the mismatch, GaN growth on rare-earth oxides are fully rela xed without rotation.11 An intermediate buffer layer is usually employed to minimize the defects caused by the large lattice mismatch. This layer helps to reduc e the density of interface traps. Zinc Oxide Zinc Oxide (ZnO) has a wurtzite crystal structure with lattic e parameter of 3.24950.12 This results in a mismatch to GaN of around 1.9%. ZnO is also transparent to visible light, and is able to achieve high electri cal conductivity from doping. This conductivity can be a drawback if an electrically insulating substrate is desired for isolation. Its reported that the GaN/ZnO st ructure would have to be grow n at temperatures less than 800C, or the film would peel o ff the ZnO single crystal substrate.13 A buffer layer of (In, Ga)N would be required to prevent the delamination of the film.13 Also, the ZnO substrate orientation 31

PAGE 32

has a large effect on the crysta l quality, therefore only certain crystallographic orientations allows for epitaxial growth of GaN. 32

PAGE 33

Table 2-1. Properties of alternative substrates GaN ZnO LiGaO2LiAlO2Gd2O3Al2O3SiC Structure Wurtz Wurtz Ortho Tetrag Bixbyite Rhomb Wurtz Lattice Parameter ( ) 3.186 3.325 5.407 5.1687 4.758 3.076 Atom spacing in (111) plane 3.828 Mismatch to GaN (%) 2.4 0.9 -1.4 20.1 16 -3.5 Melting Temp ( C ) 2527 1975 2668 2040 2700 Bandgap (eV) 3.4 5.3 9.5 3.26 Dielectric constant 9.5 11.4 9.4 10 Thermal Expansion Coeff. (10^-6/K) 6.20E06 8.30E06 7.50E06 1.03E05 33

PAGE 34

CHAPTER 3 EXPERIMENTAL APPROACH AND INSTRUMENTATION Materials Characterization The advantage of the characterization techni ques required for this research is the availability of all the instruments. The Su rface, Chemical, and Elect rical characterization techniques are readily availa ble in house and access is ma de accessible through the Major Analytical Instrumentation Center and multiple research groups. Atomic Force Microscopy (AFM) Characterization of the growth surface is pertinent to the optimi zation of the growth conditions. The ability to visually inspect a surface reveals key as pects that lead the experimental approach. AFM data allows the surface roughness and morphology to be determined, which is unmatched by any other imag ing technique. This data tells the crystal grower of the growth regime, whether by 2D or 3D growth, since an atomically flat surface results from 2D growth. The method of data collection comes from a probe which senses the Van Der Waal forces of the surface atoms. Th e probe deflection is interpreted as a height through the instrument and a resulting image is produced. Roughness is then calculated from the height image. The AFMs that will be utilized are the V eeco Dimension 3100 SPM (Figure 3-1) with the newest Nanoscope V controller and the Veeco Multimode (Figure 3-2) with a Nanoscope IIIa controller. Tapping mode will be the primary im aging technique. In this technique, a silicon probe oscillates above the surface, normally with a chosen amplitude between 20-50nm. This variation can be controlled depending on the am ount of power supplied to the piezo beneath the tip to allow for different types of samples to be scanned. A surface that is atomically flat with very little surface contamination might require only 20nm of oscillation, while samples that have 34

PAGE 35

a larger amount of contamination or surface water might need larger oscillations to avoid the tip from becoming stuck to the surface. The amplitude of the oscillations then varies as the probe interacts with the sample surface forces. This change in amplitude is detected and processed into a height signal. The vertical resolution of the instrument is approximately 0.01, while the lateral resolution is dependant on the probe. Th e probes normally utilized are Veecos standard TESP silicon tips having a tip radius between 8-10nm. The cantilever resonant frequency is around 300kHz. When samples with rough topographies ar e needed scanned, contact mode can be employed which can give a better image than tapp ing mode. The reason behi nd this result is that the contact mode scans a silicon nitride pyramid tip over the surface which is attached to a gold V-shaped cantilever. The gold has a very small spring constant which results in the tip being able to track large topography changes at a faster scan rate than tapping mode could give. The draw-back to current contact tips are the larger tip radii of betw een 10-20nm. This larger radii limits the lateral resolution of the scan. Scanning Electron Microscopy (SEM) Electron microscopy has been a staple for ma terials characterization for decades. The ability to gain topological data al ong with chemical analysis is inherent to the technique. Surface imaging is utilized to see features like isla nds and defects of the material surface. The instrument uses a beam of electrons emitted from a tungsten hair-pin filament, which produces electrons when a current is passe d through the wire which heats the filament. At the same time an anode applies a field at the tip of the filament which begins to emit electrons when they have higher energy than the work-function of the tungste n filament. When this beam interacts with the sample surface, the interacti on produces electrons of various energies, auger, secondary, and backscattered electrons. 35

PAGE 36

This instrument does not employ an a uger detector so only the secondary and backscattered electrons are detect ed. When the impinging electrons experience an almost elastic collision and escape the sample in approximately the same orientation that they entered, a backscattered electron is emitted and detected by a solid state detector positioned directly above the sample. When inelastic coll isions occur in the sample, s econdary electrons are produces which have energies less than the source beam. These electrons can be collected easily by an Everhart-Thornley detector which is a positivel y-biased faraday cage around a scintilator and then to a photomultiplier tube. Backscattered electrons show compositional differences in the sample due to areas of different atomic numbe rs, while secondary electrons are utilized for topographic imaging of the sample surface. An electron beam is rastered across the sample surface and the resultant secondary electrons are collected and processed into an image. The ease of use and access to these microscopes allow for rapid characterization of the growth material. A JEOL JSM 6400 SEM will be utilized in the characterization (Figure 3-3). The instrument is capable of secondary and back scattered imaging under accelerating voltages between 0.4KV and 40KV. Normal operation is between 7-15KV. Magnification between 10X to 300,000X are capable. Resolution for this instrume nt is stated to be 3.5nm, while actual visual inspection results in clearly defi ning features of approximately 50nm. This limitation is due to the system vacuum and chamber isolation. An Oxford energy dispersive x-ray detector or EDS detector allows for qualitative analysis of the elements present in the sample. If required, sample coating is performed in house by either a carbon evaporation coater or a Au-Pd sputter coater. The coatings are normall y deposited to a thickness of 10nm to allow for adequate heat dissipation and electron grounding. Most of th e oxide samples are thin enough 36

PAGE 37

that a coating is not required when scanning at 7KV accelerating voltage. This is a result of the electron beam interaction volume can be up to a micron deep into the sample, thus most of the electrons are grounded through the GaN substrat e which is grounded through the use of carbon paint on the side of the sample. This method allows for an undisturbed surface for the most accurate imaging. The JEOL 6400 also contains an Oxford en ergy dispersive x-ray (EDX) detector, also called (EDS). This x-ray detector uses a liquid n itrogen cooled silicon-lithium (SiLi) detector to create electron-hole pairs which create a current pulse. This ma gnitude of this current pulse depends of the energy of the incoming x-ray which can then be used to create a spectrum of energies versus intensity of the sample. These xray energies are characteri stic to the sample and can be used to qualify and quantify the elements present in the sample. The imagine software of the instrument can be linked to this detector to create an x-ray map of th e sample which indicate the exact location of specified elements. When the features of interest become too small for the JSM 6400 to resolve, a FEI XL-40 field-emission SEM is employed (Figure 3-4). With a LaB6 filament, the field-emission gun allows for a tighter electron b eam with a higher current density than the standard thermionic emission gun. The instrument also uses a turbo pump resulting in a higher vacuum than the diffusion pump of the JSM 6400 SEM. The higher vacuum allows for a larger mean free path for the electrons with less scattering and theref ore a smaller probe on the surface of the sample. Magnifications exceeding 500,000X are capable from the increased spatial resolution of the more coherent electron beam. Cathodoluminescence (CL) Cathodoluminescence is a technique that utilizes an electron beam to excite emission of photons from a sample. As the electrons impinge the sample, they excite transitions that emit 37

PAGE 38

light from the interaction volume of the electr on beam. For this work, the electron beam is supplied by the JEOL 6400 SEM used for the imaging and EDS analysis of the samples. The CL capabilities are supplied by a Ga tan MonoCL with wavelength range from 200nm to 800nm. The detector is linked with the instruments imagi ng software to allow for images of the sample during light emission. This allows for defect de tection and quantificati on within the top volume of the sample surface. Reflection High-Energy Electron Diffraction (RHEED) While under the ultra-high vacuum of the grow th chamber, the sample surface quality can be determined. RHEED utilizes an electron gun operating between 6-10KV and a view screen coated in phosphor. The electron beam is directed onto the sample surface at a glancing angle, normally less than 5 degrees. The electrons ar e diffracted off the first few layers of surface atoms, onto the phosphor screen. The patterns whic h are seen on the screen can indicate whether the sample is single crystalline, polycrystall ine, or amorphous. The surface morphology can be determined also, whether the surface is smooth or rough, and whether the gr owth is 2D or 3D. While an atomically smooth, single crystal surfac e results in a streaky pa ttern, while a slightly rougher surface shows a spotty reco nstruction. Polycrystalline samples are depicted as rings or arcs, and amorphous samples show almost no patter n at all shown in Figur e 3-5. RHEED is used to characterize the starting substrate and subs equent growths. RHEE D can also give an indication of the thickness of an oxide when th e pattern becomes fuzzy. This is because the surface will begin to charge and distort the electr on beam when the film becomes thicker than the electrons ability to conduct into the normally conductive substrate. Auger Electron Spectroscopy (AES) AES allows elemental analysis of samples w ith a higher degree of quantitative reliability than standard SEM. The analysis depth is ap proximately the top few mo nolayers of material. 38

PAGE 39

This is a result of the Auger process where a main electron beam interacts with the sample atoms. The beam ejects an inner core electron from the atoms which then gets filled by an electron from a more outer shell. This electr on transition releases an x-ray or a low energy Auger electron. Since these elect rons have low energy, they have a very shallow escape depth, resulting in a very surface sens itive analysis. Along with surf ace analysis, AES can be coupled with sputtering to perform depth profile elementa l analysis through the use of a sputter gun. This technique allows the verification of homogeneit y of the growth film, showing any compositional changes throughout the film thickne ss. Interfaces can also be characterized to locate any contamination or native oxide on the starting substrate. An AES Perkin-Elmer PHI 660 Scanning Auger Multiprobe will perform the Auger characterization (Figure 3-6). X-Ray Photoelectron Spectroscopy (XPS) Based on the Photoelectric principle, explained by Al bert Einstein in 1905, a monochromatic x-ray sour ce impinges a sample, thereby creati ng energy transitions by releasing a photoelectron. This electron is detected by a hemispherical analyzer and the energy is related to the binding energy of the atoms in the samp le. This binding energy is related to the stoichiometry of the material and can reveal wh at phases and oxidation stat es are present in the materials. The analysis depth is approximately the top 1-10nm of material, since the kinetic energy of the photoelectrons is not large enough to escape from deeper in the sample. Depth profiling is available with the use of a sputter gun to probe chemical bonding at interfaces in the sample. Analysis of samples is performed by either a Perkin-Elmer XPS/ESCA PHI 5100 ESCA system (Figure 3-7) or a XPS tool that was built directly onto the Varian MBE (Figure 3-8). X-Ray Diffraction (XRD) The crystalline quality of the films will be characterized through XRD. In this method, x-rays are diffracted off the sample to produce characteristic peaks of atomic planes. This 39

PAGE 40

technique can give indication of the degree of crystallinity of the film while also indicating the lattice spacing of the material. Therefore, for interests of lattice matching film to substrate, XRD is an easy method that requires little sample-prep. The peaks of the substrate and film should overlap in a lattice-matched system. For the GaN growth upon MgCaO, the ratio of Magnesium to Calcium can be adjusted to result in the lattice spacing required for GaN lattice matching. This technique is based on Braggs La w which is stated as Equation 3-1. n = 2dsin (3-1) Where n is multiple of whole x-ray wavelengths, is the incident x-ray wavelength, d is the atomic spacing of the mate rial being analyzed, and is the Bragg angle in degrees. This equation states that constructive interference of x-rays occurs only at certain angles from the sample surface. At these angles large intensity of x-rays emerge from the sample and are indicated by a peak in the resultant spectrum. X-ray diffraction reveals the st ate of crystallinity of the sa mple, where a very sharp peak in the spectrum indicates a single crystalline film, whereas a very broad peak is a indication of a polycrystalline film, and finally amorphous materi als do not result in any peaks. A measure of crystal quality is th rough the calculation of the full width at half maximum (FWHM) of the peak. For initial verification and tuning of the metal ratios of the oxide, a Phillips APD 3720 XRD (Figure 3-9) will be ut ilized scanning at angle incr ements of 0.005 degrees. Higher resolution rocking-curve analysis will utilize a Phillips MRD XPert system with a 5-crystal analyzer (Figure 3-10). Finall y, exact thermal expansion measurements can be taken with a Phillips 3100 high-temperature XRD instrument with a helium atmosphere (Figure 3-11). The sample is heated as a spectrum is taken show ing expansion of the lattice parameters. These stress measurements of the films are used to determine the exact regime for optimum growth. 40

PAGE 41

Photoluminescence (PL) Photoluminescene (PL) is an optical characteri zation technique that utilizes the principle of optical pumping. Employing a laser as the ligh t source allows for accurate luminescence data since the extremely narrow wavelength of laser li ght permits characterization of the light from the sample and not interference from the light source. Utilizing a Omnichrome helium-cadmium ultraviolet laser (Figure 3-12) the sample is frontside illuminated with en ergy higher than that of interest. The UV light of 325nm excites the various el ectronic states in the sample, including defects and the bang-edge. Using this technique samples are probed for the various types of defects and impurity levels. This information is then utilized to adjust further grows. A monochromator is used to capture lumine scence as any desired wavelength, normally between 350nm to 800nm. The monochromator di rects the light into a PMT (photomultiplier tube) that amplifies the signal and outputs th e intensity versus wavelength (Figure 3-13). GaN samples normally have multiple emission regions. The first is the band edge luminescence around 361nm which is desired to be the most intense emission from the sample. The second region is in the yellow/ green region of the spectrum. This yellow emission is due to nitrogen vacancies in the material and for high qua lity materials this emission is suppressed. The final peak that is normally seen in spectrums from our PL setup is a duplicate peak of the band edge that appears at 720nm. Again, this red peak is actually just an artifact of the scan. As a standard, a piece of n-GaN from Un iroyal is scanned from 350-800nm. There is also a chance of a blue emissi on around 440nm which is indicative of oxygen doping. This blue emission is seen in our MOCVD grown samples due to an overdue change on the gas purifiers. The oxygen enters the system from small leaks in the nitrogen source manifold and over time degrades the purifiers which nor mally keep the oxygen out of the system. The oxygen can then become trapped in the galliu m bubblers and permanently contaminates the 41

PAGE 42

gallium metalorganic. Any other oxygen doping from either autodoping or other sources of contamination in the system would s how as an increase to this emission. Contact Angle Goniometer A rame-hart contact angle goniometer is used to measure the contact angle of droplets of deionized with a sample surface. The water drop was dispensed with an automated dispensing unit, capable of dispensing down to 0.1L of fluid per step. The measurements were taken with a 1L and 5L drop for comparison. The tangent a ngle to the contact point of the fluid and the sample surface is taken as the contact angle. Th is angle gives relative in formation of the surface energies of the samples with a larger contact angle having a highe r surface energy than one with a small angle. Hall-effect measurements (Hall) Used to analyze the electrical characteris tics of the grown GaN film, Hall measurements will be performed. Discovered in 1879, the Hall measurements are based on the principle that charge carriers are affected in the presence of a magnetic field. Th e materials carrier concentration, type, and mobility are all determined from this technique. Current-Voltage Measurements (I-V) To determine the quality of the MgCaO layer, an IV measurement is used to determine the breakdown field strength. In th is measurement, a limit to the current is set, and the voltage on the sample is swept from negative to positive. The current limit is usually set quite low, on the order of 100nA. The voltage is increased until the current reaches the set limit, where forward and reverse breakdowns are reached. Th is voltage can then be divided by the film thickness to calculate the breakdown field strength. A Hewlett Packard Model 4155C Semiconductor Parameter Analyzer will be used for the IV measurements. 42

PAGE 43

Ellipsometry Ellipsometry is an optical te chnique that is non-destructive which is used to measure the thicknesses of thin films. The principle operation is to reflect plane-polarized light off the sample into an analyzer. When the light impinge s the sample, a change in polarization occurs at which time the phase change is detected by the an alyzer. Using silicon as the substrate, the reflected light can measure the thickness of the transparent film. A Rudolph V-530044 Auto EL IV ellipsometer was used for this work (Figure 3-14). Transmission Electron Microscopy (TEM) Transmission electron microscopy (TEM) us es a high energy beam of electrons to penetrate and pass through a thin sl ice of material, result ing in an image of the atoms and defects of that material. Used for the analysis of th e microstructure and def ects within a sample, TEM has extremely high lateral resolu tion allowing for indivi dual atoms to be seen. Samples are prepared for TEM analysis through the use of a focused ion beam or FIB. A FEI Strata dualbeam 235 FIB was used in the sample preparation. The samples are first coated with a thick layer of protecting carbon, then the FIB uses a beam of gallium ions to sputter the sample and mill away material until an extrem ely thin slice of material is le ft. This thin slice is electron transparent which allows for the imaging electron b eam to either transmit or forward scatter onto a detector, forming an image. The TEM that wa s used almost entirely for the samples of GaN and oxide was a JEOL 200CX (Figure 3-15). Material Growth Techniques Many tools have been developed for the grow th of thin films throughout the decades. The growth techniques utilized in this research include molecular beam epitaxy and metalorganic chemical vapor deposition. These tools will be used to grow oxides and nitrides for the research. 43

PAGE 44

Molecular Beam Epitaxy (MBE) Molecular Beam Epitaxy (MBE) is a technique that provides extremely precise control of a growing film. MBE is a great research tool be cause of the high level of control of the growing surface. Another advantage of MB E over other growth techniques is that the growth chamber is under ultra-high vacuum, 10-10 Torr, which keeps the background contamination level to a minimum. Source material comes in solid, liquid, or ga s depending on the type of film to be grown. For this research, solid sources of magnesium, calcium, and scandium, and liquid gallium, are heated in Knudsen effusion cells. The temperat ure of these cells will determine the vapor pressure of the material. This vapor pressure causes a pressure difference in the cell and the chamber, creating a beam of atoms which impinge s the substrate. The flux of atoms from the cell is approximated by the reduced Equation (3-2)14 2 1 2)( )22118.1( MTP PAe F (3-2) Where P is the pressure in the source crucible, F is the flux, A is the area of the aperture, D is the distance from the sample to the aperture, M is th e molecular weight of the species, and T is the temperature of the species. The substrate is held at a temperature to which supplies the impinging atoms enough energy to be mobile of the surface. This mobilit y allows the atoms to mi grate on the surface to a high energy site on the crystalline substrate. This migration allows for the growing film to imitate the starting terraced GaN substrate. At low temperatures, the atoms have less energy for migration and therefore, chances of the atoms resting where they impinge the sample is quite high, resulting in islands or 2D nucleation. With this knowledge of terrace controlling growth, it is reasonable to conclude that the more terraces available for atom attachment, the faster the 44

PAGE 45

growth with take place, while st ill in layer-by-layer or epitaxial controlled growth. The starting MOCVD GaN substrates have a large amount of terr aces available of oxide attachment. This is a method to verify if the sample is at an ade quately high temperature, by noticing that the oxide replicates the GaN terraces. The higher temperatures used for large atom mi gration can also be de trimental to growth, since the desorption rate of the impinging atom s increase. Therefore for growth of multiple atomic species, this can result in a non-stoichiome tric growth. The species with the higher vapor pressure will be the species that desorbs at the faster rate, ther efore, a balance needs to be found which will equate absorption / desorption rate of the different species to result in a suitable growth. With that in mind, controlling the fluxes of the metals to create a lattice-matched oxide is not only based on temperature of the metal effusion oven, but also on the substrate growth temperature. The oxide material was grown on MOCVD GaN in a modified Riber model 2300 MBE (Figure 3-16). An overpressure of elemental oxygen is supplied by an Oxford radio frequency plasma source operating at 13.56 MHz, keepi ng the chamber at pressures between 10-6 torr and 10-5 torr. The oxygen source is molecular oxygen of research grade. The GaN grown on top of the oxide was deposited in a modified Varian Gen II MBE system (Figure 3-17). The Varian system utilizes a hot-lip gallium cell to prevent gallium deposition at the end of the crucible which mitigates the flux. Elemental nitrogen is supplied by an Oxford radio frequency plasma head in which research grade molecular nitrogen is passed. The Varian MBE tool is equipt with other elements to faci litate doping experiments. The chamber contains magnesium for p-type doping of the GaN, along with silicon for n-type doping. Both systems employ RHEED analysis to verify and observe changes in the growing film. 45

PAGE 46

Metal Organic Chemical Va por Deposition (MOCVD) Metalorganic chemical vapor deposition (MOCVD) is a growth technique that has seen much improvement over that last few decades. Th e switch from horizontal to vertical reactors has allowed industrial tools to gr ow many wafers at a time and with great uniformity. This work utilizes a Veeco/Emcore P75 TurboD isk vertical reactor handling two inch wafers (Figure 3-18). Being a commercial tool, the computer control and automation makes a growth run very nice. Yet, the principles that govern MOCVD growths are very sensitive to slight changes in growth conditions and can therefore require mu ch iteration to change one variable. The TurboDisk reactor has a showerhead de sign which allows source material to be injected separately onto the wafer surface. Th is separation minimizes pre-reactions of the metalorganics before reaching the substrate. Th e substrate is rotated up to 1500rpm to allow for good uniformity and to create a good flow profile of the gasses. The susceptor in which the substrate rests, is heated thr ough a resistance heater directly below the susceptor (Figure 3-19). Temperatures reached in the tool can exceed 1100C. The nitrogen source is created through the decomposition of ammonia on the substrate. Carrier gasses of nitrogen and hydr ogen are used to supply the samp le with the metalorganics. Besides the gallium source, aluminum is availabl e for aluminum nitride growth and also indium for alloying is necessary. The tool has two sources of gallium on the MOCVD tool, TMG (Trimethylgallium) and TEG (Triethylgallium). The choice of source for Ga N growth depends on the structure that is to be created. TEG is mainly used for HEMT struct ures and other devices that need rapid changes to the structure. TMG is the standard for most al l other GaN material. This is the choice source since TMG is less expensive than TEG and for the unintentional benef it of additional carbon incorporation for more insulati ng properties. TMG has inherently more carbon by-products than 46

PAGE 47

TEG from the methyl group compared to the et hyl group. Carbon has been shown (Impact of carbon on trap states in n-type GaN grown by metal organic chemi cal vapor deposition) to help mitigate the doping effects from residual oxygen and ot her contaminates in the growth chamber. The carbon seems to sit in vacancies and allows for less conductive GaN material. Sample growth can be monitored by emissivity compensated pyrometry. Utilizing an infra-red light source of 930nm, 10nm bandwidth, reflectance off the sample surface can indicated the degree of roughness of the surf ace along with layer thickness because of interference with the light as the layer grows. In addition, the same detector can measure the black-body radiation off the sample and give an accurate measure of the surface temperature. This method is superior to simple measuremen ts by the thermo-couple as the thermo-couple is mounted below the heater, and based on the gas fl ows in the tool, the sample surface can have a completely different temperature th an what the thermo-couple measures. 47

PAGE 48

Figure 3-1. Veeco Dimension 3100 AF M with Nanoscope V controller. 1 Figure 3-2. Veeco Multimode AFM with Nanoscope IIIa controller. 48

PAGE 49

Figure 3-3. JEOL 6400 SEM. Figure 3-4. FEI XL40 Field-Emission SEM. 49

PAGE 50

A B C Figure 3-5. Possible RHEED patterns. A) Amor phous diffraction pattern. B) Polycrystalline diffraction pattern. C) Single crystal diffraction pattern. [Reprinted with permission from B.P. Gila, 2000. Growth and Characteri zation of Dielectric Materials for Wide Bandgap Semiconductors. PhD dissertation (pg. 47, Figure 3-11). University of Florida, Gainesville, Florida.] 50

PAGE 51

Figure 3-6. AES Perkin-Elmer PHI 660 Scanning Auger Multiprobe. Figure 3-7. Perkin-Elmer XPS/ESCA PHI 5100 ESCA. 51

PAGE 52

Figure 3-8. Custom XPS system attached to Varian MBE. 52

PAGE 53

Figure 3-9. Phillips APD 3720 XRD. Figure 3-10. Phillips MRD XPert sy stem with a 5-crystal analyzer. 53

PAGE 54

Figure 3-11. Phillips 3100 high-temperature XR D instrument with a helium atmosphere. 54

PAGE 55

Figure 3-12. Omnichrome helium-cadmium ultr aviolet laser at 325nm along with a SpectraPhysics argon ion laser. Figure 3-13. PL setup showing optics, monochr omator, and PMT on the left of the image. 55

PAGE 56

Figure 3-14. Rudolph V-530044 Auto EL IV ellipsometer. 56

PAGE 57

Figure 3-15. JEOL 200CX TEM. 57

PAGE 58

Figure 3-16. Modified Riber mode l 2300 MBE used for oxide growths. Figure 3-17. Modified Varian Gen II MB E system utilized for the GaN growths. 58

PAGE 59

A) B) Figure 3-18. Veeco/Emcore P75 TurboDisk vertical reactor utilized for Ga N growths. A) Image of the MOCVD growth chamber (left) and load -lock (right). B) View of the full tool. 59

PAGE 60

Figure 3-19. Image of the wafer susceptor being heated by the re sistance coil heater. 60

PAGE 61

CHAPTER 4 MBE GROWN MGCAO Epitaxy Challenges Quality epitaxy is not a trivial matter to accomplish. For the growth of high quality material, the number one factor is the starting substrate. Substrat es can either composed of the same exact same material as the epitaxial film to be grown, or they can be of a different material. Growing on a foreign substrate is termed hete roepitaxy, and is the only economical method of growing gallium nitride. Currently the most wide ly used substrate for growing gallium nitride is sapphire, Al2O3, which is an extremely stable substr ate which can withstand the conditions normal to GaN growth. Yet, sapphi re is not a great substrate since the lattice para meter is quite different than GaN and it is not a good heat conductor, which is needed for GaN devices operating at high power loads. Si nce no bulk process has been devel oped to date that can create large boules of GaN material, very expensive wafers of HVPE GaN thick films are the only method of true homoexpitaxial growth. Again this possesses an expense issue that has not proven economical for further device manufacturing. Therefore, growing a gallium nitride film on a non-GaN substrate poses challenges. These challenges begin with the fact that the atoms being deposited normally will not come to rest exactly in a non-straining position of the host substrate. That is, until now. Alloying Calculations For MgO and CaO One oxide that has a fairly good lattice sp acing match to GaN is magnesium oxide (MgO). This oxide has a mismatch to GaN of -6.5% on the (111) crystal face.14 This is better than the mismatch of Al2O3 to GaN, which is 16%.2 Yet, there is still a large amount of strain in the lattice, which can result in defects at the interface, which inevitably propagate through the growing film. Examination of other oxides whic h are closer matched to GaN resulted in the 61

PAGE 62

binary calcium oxide (CaO). This oxide has a mismatch that is 6.6%14 to GaN along the (111) face. Considering Vergards Law (Equation 4-1)15, which states that a good approximation of a materials properties can be assumed from th e properties of the compounds contained in the solution, a mixture of CaO and MgO could result in an exactly lattice matched system with GaN. Therefore, the oxide for this research is a tern ary alloy of magnesium, calcium, and oxygen. The oxide will grow on the GaN basal plane (0001) by growing with the oxide (111) plane. Through growing on the (111) plane, the oxide minimizes the energy of the exposed surface while having the lowest energy configuration with the GaN (0001) face. Advantages of this oxide compared to standard substrates used for GaN growth, is th e ability to tune the la ttice parameters without resulting in phase separation. Through varying the calcium con centration in the ternary, the oxide is latticed matched to the GaN film as seen in the graphical representation of the material lattice models in Figure 4-1. This lattice matc hing allows for reduced defects in the film, and improved morphology of the resulting surface. Shown in Figure 4-2 is the calculation for a lattice matched system, and also a graphical representation of the mismatch for varying compositions in Figure 4-3. aMgCaO = xaMgO + (1 x)aCaO Equation 4-1 Having a good interface between substrate a nd film not only creates good quality growth material, but also has the advantage of decreasi ng defects at the interf ace. These defects are traps that hinder device performa nce. HEMT passivation is a good example of how the electrical characteristics can change by simply having a go od dielectric interface for the device. Being lattice matched with GaN, MgCaO can reduce th e interface traps that pl ague the electrical characteristics of most GaN de vices. Starting with a good interface results in lowering of 62

PAGE 63

interface traps and reduces leakage of a device. For a good interface, the thermal expansion of the materials should also be considered. The thermal expansion of the binaries, MgO and CaO, have both been found for relatively high temperatures.3,12,16 Experiments for GaN have also resulted in values for thermal expansion around the same temperatures.17,18 Listed in Table 4-1 are the thermal coefficients reported along with other parameters of GaN and the two binaries of interest. As reported, the MgO and CaO thermal expansions seem to trend linear with temperature. This is unlike GaN which is repo rted to have a slightly increasing expansion at elevated temperatures. Since this trend varies through reports17,18,19, a linear approximation is utilized for calculations. Thermal expansion m easurements can be explored to determine the exact expansion. This non-linear trend might cause strain even wh en the lattice parameters are matching. Therefore, if this becomes and issue, a solution would be to in duce a lattice mismatch to intentionally strain the GaN film to compen sate. Figure 4-4 is a pl ot of various ternary compositions versus temperature with th ermal expansion taken into account. Adjusting the composition of the oxide to facilitate the ther mal expansion might also be an advantageous device property. Strain induced mobility changes occur in a variety of materials as the lattice is stra ined. With the lattic e response of the AlGaN/GaN system, devices could be fabricated to take a dvantage of a strained film or etched cantilever for macroscopic pressure or strain measurements.20 Gallium Nitride Substrate Preparation The substrates used for the oxide growths we re standard gallium nitride on sapphire. The samples were either in-house grown GaN with the TurboDisk tool, or UOE GaN which were calibration or engineering wafers from UOE. Eith er way, the pretreatment of the substrates were the same for the epitxial growth process. 63

PAGE 64

While gallium nitride is a very stable material at room temperature, the atmosphere in the lab will inevitably deposit carbonous residues and oxidize the surface of a sample. Unless samples of MOCVD GaN are loaded into the vacuum environment of the MBE tool within a few days, all samples must go through a process to clean and de-oxidize the starting GaN surface. The cleaning procedure begins with a hydrochloride acid wash to remove the carbon residues from the surface. A th ree minute soak in a 1:1 solution of HCl and deionized water will clean the surface of most of th e bulk carbon on the sample. The sample is then rinsed with deionized (DI) water to neutraliz e the acid and rinse the acid pr oducts from the sample followed by a blow dry with dry nitrogen gas, before the next step. Following the acid wash, the sample is then placed in a partially sealed container whic h has a bank of ultraviolet lights above the sample surface. When the UV lights are turned on, th e high energy light creates ozone from the atmospheric oxygen. The ozone (O3) is highly reactive and therefore scav enges carbon from the sample surface. This process was tested with a high concentration of carbon to verify the cleaning or carbon removing capabil ities of the UV-ozone cleaner. For a large sample of carbon, a single crysta l silicon two inch wa fer was coated with approximately one micron of photoresist. Photoresist is a derivative of PMMA and therefore is on the order of pure carbon. The sample was anal yzed by an ellipsometer to allow for accurate thickness measurements. The ellipsometer is accu rate to within a few angstrom. The coated sample was then placed in the box UV-O3 cleaner and allowed to be irradiated for some time. The sample was noticed to increase in temperatur e up to 50-70C from the light absorption of the UV lamps. The sample was periodically removed and analyzed with the ellipsometer. Shown in Figure 4-5, the photoresist wa s very efficiently etched away at a rate of 72 /min. This result 64

PAGE 65

shows that the small amount of residual carbon on the surface of the starting GaN can be very effectively removed by this UV-ozone process. The normal process of cleaning the GaN sa mple after the HCl wash, is a 25 minute exposure in the UV-ozone cleaner. Once the samp le was removed from the cleaner, it went through a wash in buffered oxide etchant (BOE). This exposure to ozone not only caused a reaction with the surface carbon, but unfortunate ly would also oxidize the GaN surface causing GaO formation. The buffered oxide etchant is a mixture of ammonium fl uoride and hydrofluoric acid. This solution is very effective at rem oving oxides of various so rts, including GaO. Therefore after a five minute wash in the BO E solution, the sample goes through one final DI water rinse followed by a dry nitrogen blow dry. Samples were examined by auger electron spectroscopy to determine the carbon and oxygen content. The results are tabulated in Table 4-2, which shows the various treatments performed and the resultant car bon to nitrogen and oxygen to n itrogen ratio for the sample surfaces. As it can be seen the combination of the HCL/DI wash for three minutes, followed by a 25 minute exposure in the UV-ozone system, endi ng with a five minute wash in BOE resulted in the best contamination removal, and cleanest surface. After the pretreatment procedure, the sample s are then mounted and loaded into the MBE system. Before growth, the samples are anne aled under vacuum to 700C and held at that temperature for ten minutes. This 700C anneal is found to produce a cleaner surface than without. Since the Varian MBE chamber has an XPS system attached to the buffer extension, the samples can now be annealed and scanned in the XPS without ever breaking the vacuum. The results are seen in Figure 4-6 which depicts the nitrogen, oxygen, and carbon binding energy peaks. The study involved samples that had been processed with the above mentioned 65

PAGE 66

pretreatment of HCl, UV-O3, and BOE, along with samples that did not have the pretreatment. The samples were scanned before and after a 700C anneal for ten minutes. The differences in peak intensities before and after are compar ed in Table 4-3, and Table 4-4. Again the contamination elements of carbon and oxygen are co mpared against the nitr ogen peak which is part of the sample. It can be seen that the sa mples that were annealed showed improvements of surface cleanliness since the ratio s decreased. This study also verified the previous study that used AES to quantify the contamination, showing that the pretreatment process resulted in a cleaner surface than without. As a final note on the subject of surface preparation, it is noticed that during the annealing the RHEED does improve. When firs t loaded into the chamber, the RHEED image from the sample surface is hazy with no distingu ishable pattern. As the sample begins to heat up, the first pattern begins to form when the adsorbed water on the surface is liberated as can be determined by a chamber pressure increase. The patte rn seen at this time is slightly streaky but faint. As the sample temperature continues to increase, the streaky (1X3) reconstruction begins to form and is vibrant once reaching and annealin g at 700C. This strong pattern indicates a clean surface ready for epitaxial growth. Standard MBE Oxide Growth Procedures The substrates used for these calibration runs were UOE MOCVD GaN on sapphire. This material is our standard material for calib rations, since it is in plenty supply and is well characterized. Pieces on UOE GaN, normally ar ound 1cmX1cm, were cleaved out of a full two inch wafer by the normal method. This method involves using a diamond scribe to scratch the sample dimensions on the back of the sapphire followed by a compressive stress on the front GaN surface which causes the scratch to cleave. The pieces were prepped in the normal manner 66

PAGE 67

with a 3 minute bath of 1:1 solution of HCL: DI water, followed by a 25 minute exposure to ozone, and finally a soak in buffered oxide etch for 5 minutes. The samples would then be mounted upon standard MBE molybdenum sample holders, which are also termed blocks. Utilizing a hot plate, a standard 4 molybdenum sample holder, was allowed to reach temperat ures of approximately 200C wh ile resting upon a aluminum foil coated hot plate. The aluminum foil was for c ontamination issues to reduce particulate matter entering the MBE system. At this temperature, indium had enough energy to melt but not enough to oxidize very rapidly. While in the liquid state, the indium was spread thin across the entire molybdenum surface with a stainless steel razor blade. Once the entire surface had been coated with indium, a small chunk of indium was placed in the center of the holder and allowed to melt and pool. At this point the actual sa mple to be grown upon was rested via tweezers on the pool. The sample was then gently moved around the block while the excess indium was scrapped off the block with the razor blade. As the indium under the sample became thinner and thinner, capillary forces be gan to immobilize the sample. Once a good adhesion had been established and the sample was in the center of the block, the entire block was removed from the hot plate and allowed to cool. No rmally a strip of thin tantalum foil was used to clip the sample to the block with a screw. This clip gave extr a support to keep the sample adhered to the block even though all of the support of the sample was from the capillary forces of the liquid indium. Yet, at the normal growth temperature of 500C the evaporation rate of the indium was not that significant, and therefore most of the samples did not require clips. Oxide samples were grown in a modified Riber model 2300 MBE. The substrate thermocouple was calibrated by using the melting points of gallium antimonide, GaSb, (707C) and indium antimonide, InSb, (525C). Pieces of GaSb and InSb were heated in the growth 67

PAGE 68

position with nitrogen plasma impinging on the su rface. This reduced the chance for loss of the group V, Sb, during the heating, which would re sult in an incorrect melting temperature. The metal sources were heated in sta ndard Knudsen Cells. The Magnesium cell temperature was about 360C while the Calcium was in the vicinity of 412C. The temperatures of the cells were less important than the emer ging flux and the growth rate. There was an overpressure of elemental oxygen, supplied by an Oxford Radio Frequency plasma source operating at 13.56 MHz, which kept the chamber between the high 10E-6 Torr and low 10E-5 Torr. This was the pressure regime for optimum growth of the oxides. The growth rate was determined by mounti ng a piece of silicon wafer next to the GaN substrate on the same block during growth. Afte r growth, the silicon sample was examined by an ellipsometer. The ellipsometer would measur e the index of refraction of the film and an internal calculation would result in the film th ickness with a variability of plus or minus five angstroms. Silicon was used as a substrate for th e thickness measurements since the ellipsometer works off the principle of light reflection off interfaces. The film on the GaN on sapphire is too transparent to effectively be measured with the instrument available. The variation between a single crystal of the oxide film grown on GaN versus the polycrystall ine or amorphous film grown on the silicon sample is not significantly different to be an issue for determining growth rate. If monolayer control was of interest, exact thickness measurements would be taken by growing a very thick oxide layer and doing TEM or cross-section SEM analysis to get an exact thickness and resulting growth rate. After indium mounted was complete, the samp les were loaded into the MBE load-lock and a vacuum was drawn on the chamber by a molecular drag pump backed by a roughing pump. The pressure was lowered into the 10E-5 Torr ra nge before the chamber was opened up to an ion 68

PAGE 69

pump which would pull the pressure down to 10E-8 Torr. Once accomplished, the samples, now on a trolley carrier, would be m oved into the buffer chamber on th e system. Since this chamber rarely sees atmosphere, the pressure within th is buffer chamber can reach into the 10E-10 Torr regime. Once the system was ready to go, which in clude heating of the effusion cells and filling the cryo-panels with liquid nitr ogen, a magnetic transfer arm would move one of the sample blocks into the growth chamber. The block w ould hook to the substrate heater by a lock and key mechanism. Once in the growth chamber, the sample was rotated to face the source panel to be cleaned by an in-situ anneal. Without any gas fl owing and all source shutters closed, the sample would be heated to 700C in 10-15 minutes. The pressure in the chamber would begin in the mid 10E-9 Torr, and as the sample temperature is in creased the pressure in the chamber would rise into the 10E-8s as surface water from the atmosphere evaporates. The 700C annealing step was determined to re sult in maximum surface cleaning while having a minimum surface stoichiometric change. It is shown that below 750C in a MBE ultra-high vacuum environment, that the d ecomposition rate is almost zero. Since the activation energy for decomposition in a vacuum environment for GaN is 3.6eV21, the decomposition rate rapidly increases above 800C After this cleaning step, the sample would be cooled to the growth temperature and rotated away so the sample surface is 180 degree s away from the source materials. The oxygen plasma would be lit at this time. Depending on the flow of research gr ade molecular oxygen, the plasma would take around ten minutes to come to operating conditions. Operating conditions were determined by a stable chamber pressure and a stable photon output as detected by a 69

PAGE 70

photodiode attached to the view-port which is on the opposite end of the di scharge tube of the plasma head crucible. Binary Oxide Growth of MgO and CaO The first step in creating a ternary alloy is to grow the binaries separately and examine the resultant film. The first samples grown were MgO and CaO on UOE GaN. The substrate temperature was set to 200C as this was the gr owth condition of the previous growths of MgO for dielectric testing. The flux of magnesium was set to 4E-8 Torr BEP (beam equivalent pressure) and the calcium was se t to 5E-8 Torr BEP. The o xygen flow was set to 0.25sccm which would result in a chamber pressure of 5.5E-6 Torr and a photodiode reading of around 130mV. To begin growth the oxygen and correspondi ng metal source shutters were opened, then the sample would be rotated to the growth posi tion which directly faces the oxygen plasma. The sample would be rotated at 15 rpm (revolutions per minute) to allow for the best uniformity across the sample surface. The samples were grown for 30 minutes, resu lting in thicknesses around 300 since the conditions allowed a growth rate of 10 per minute. The binary oxide samples were then characterize d fully to gain insight for the next step of alloying. In figure 4-7, tapping mode AFM imag es show surfaces that have average roughness of around 1nm, as the MgO surface is somewhat rougher than the CaO, since the CaO surface has raised bumps which are believed to be regi ons of increased growth rate or too low of substrate temperature. When the surface temperat ure is too low, the mobility of the impinging atoms is quite low resulting in either poor crys tallinity or clustering of the atoms. X-ray diffraction was performed in Figure 4-8 to find the exact lattice para meter of the binary films. The XRD spectrum shows the oxide peaks, the Ga N substrate peak, along with small additional peaks from the sapphire substrate. As can be s een the vertical scale is logarithmic, since the oxide film is quite thin and therefore is barely visible in linear plots. To further characterize the 70

PAGE 71

films, they were scanned by XPS to determine if any other elements were present. The XPS scans (Figure 4-9) show the photoelectric peaks of the binary elements, along with a carbon peak which is always present on sample surfaces when they have seen atmosphere. This atmospheric carbon can be removed by a short argon sputter in the XPS chamber. In addition to the XRD scans of the Mg O film, high-resolution X-ray analysis was performed on a thick MgO film on GaN. Shown in Figure 4-10 is the omega rocking curve for the sample, showing the MgO (222) and the MgO (002) peaks. This high resolution scan allows for accurate measure of the peaks. As shown in the figure, the (222) peak is the surface parallel plane and is s figure of merit fo r crystallinity of the film. The full-width of half-maximum (FWHM) for this peak is 1.882 degrees. The (002) MgO omega rocking curve has a FWHM of 4.198 degrees. The (002) plane is not surface pa rallel, and gives additional insight into the quality of the film. Since the (222) plane is much smaller than the (002) plane, this indicates that the film might have a low angle grain boundary that could be due to some dislocations in the film. Dislocations are to be expected since there is a 6.5% mi smatch between the MgO and the GaN substrate, which could cause stress in the film. Growth of MgCaO Via Digital Growth Mode Now that the binary oxides have been characterized and a growth rate with flux determined, the next step is to combine the bina ries and grow a ternary oxide. For this task, there were two thoughts on how to properly grow th is material. The first was to simply allow both metal fluxes to hit the substrate and gr ow normally. This might pose issues with segregation, even though the metals have the sa me valence and should replace each other in the growing lattice. Therefore th e initial experiments were done by alternating the metal fluxes which impinge the sample. This type of growth had been performed and published earlier.22 71

PAGE 72

The elemental sources were shuttered at sp ecific time intervals to create a homogenous composition throughout the film. This type of grow th has been termed a digital alloy. This method allows for alternating layers of MgO an d CaO, resulting in a homogenous film (Figure 411). Along with results showing continuous un iformity throughout the film, by varying the timing in which the shutters are open, allows for different amounts of meta l atoms to incorporate into the film. By changing th e amount of time the individual shutters are open, the resulting alloy shifts from one binary alloy to the other. This allows for composition changes, shifting the oxide lattice parameter as seen in Figure 4-12. As a comparison, Figure 4-13 shows the XRD scan of a standard UOE GaN surface. Also to note is that these UOE GaN substrat es were calibration runs from Uniroyal. Being the case, each wafer can have a slightly different diffraction spectrum which results from varying thicknesses of the GaN layer, and def ects present in both the GaN and the sapphire. Also to note is that these substrates do have pi ts in the GaN film which could possibly extend to the sapphire surface, as seen in Figure 4-14. It is worthy to also note that the UOE GaN does not show strong growth terraces normally seen in MOCVD GaN. As a comparison, Figure 4-15 shows the surface of the GaN we grow in-house on out MOCVD tool. Our in-house GaN shows nice terraced growth and sharp edges with low roughness. Even though UOE GaN substrates are used for calibration runs, once good recipes are created, the growths then commence on the high quality in-house grown, MOCVD GaN. The sequenced or digital grow ths for calibration were grow n 30 minutes. This not only gave a film that would be thick enough for thickness measurements, but to allow for enough material for depth profiling Auge r to confirm uniformity. Just as a more thorough detailing of the procedure, these initial digital growths were done manua lly by hand, flipping the shutter 72

PAGE 73

switches for the entire growth. As can be inferred, this does not lend itself to enjoyable times during these growths. Also, this short sequenc ing did not allow much time to perform other duties involved with MBE growth, like adjusting the plasma h ead or looking at the RHEED pattern. Therefore, after these initial calibrati on growths, a computer-controlled digital switching relay was installed onto the shutte r control manifold. The digita l relay was an Omron Zen which couple allows programming of any recipe to cont rol the open/closed positi ons of the shutters, shown in Figure 4-16. This device allowed for the ability to progr am a recipe to open and close the different shutters at any timing needed for growth. This allowed for the freedom to check conditions of the MBE system and verify stability of the plasma head which had a tendency to drift over time resulting in an increase in the re flective power. Building this interface with the shutter controllers was a great assistance for these growths, and also allowed exact timing of the switching since manual switching woul d always be off now and again. The growths of the ternary began with the standard pretreatment of 3 minutes HCL:H2O (1:1), 25 minutes UV-O3, and then 5 minutes of BOE. After the sample was loaded into the MBE, and transferred to the grow th chamber, the sample was heated to the standard 700C for a ten minute anneal as an in-situ clean. These first growths were performed at 300C at oxygen flow of 0.3sccm with plasma pow er at 300W, resulting in a chamber pressure of around 2E-6 Torr. For this initial set of digital growths, the shutters were opened and cl osed at sequences that allowed for three cycles per minute. The cycles tried were ten seconds of magnesium followed by ten seconds of calcium, then fifteen seconds of magnesium and five seconds calcium, and finally five seconds of magnesium followed by fifteen seconds calci um. Therefore, with initial growth rates of around 30 /min, this would result in a frac tion of a monolayer up to a couple monolayers of the various binari es grown during the cycles. The flux of the magnesium cell was 73

PAGE 74

around 5E-8 Torr BEP (beam equivalent pressure) while the calcium was around 3E-7 Torr BEP. The growths would always begin wi th magnesium since this would result in finishing the growth with CaO. CaO seemed to be a bit more stable in air than MgO, so this was assumed to help the stability if further processing was needed. After the three different sequencing sample s were grown, XRD was performed to locate the oxide peak in relation to the GaN peak. This would indicate the direction for the next samples. Seem in Figure 4-17, the first set of samples were calcium rich since the oxide peaks were to the left of the GaN peak, indicating a la rger lattice parameter than GaN. The figure does show how nicely the lattice spacing changes in accordance with the varying amount of time that each metal source was open to the sample. By th e addition of more magnesium, the peak shifted closer to the GaN spacing. Also to note is that since this set of samples were calcium rich to begin with, as the amount of calcium exposure in creased, so did the film thickness. Therefore the sample with the longest calcium exposure resu lted in the thickest growth, which resulted in the largest peak height in the XRD spectrum. Also to note, while sitting on a stool flippi ng the shutter switches, an oddity was noticed on the mass spectrometer. The pressure read by the mass spectrometer showed a drop while the calium cell was open, compared to when the ma gnesium cell was opened. The pressure would drop from 2E-6 Torr to around 4E-7 Torr. This pressure drop indi cated that there was insufficient oxygen overpressure from the 0.3sccm flow. Therefore the next sample set was grown under a slightly higher flow. After the results of the first digital growths, a few changes were made to result in improved growth. The first adjustment was to increase the oxygen flow to 0.4sccm to allow for more active oxygen species to impi nge on the sample surface. The increased flow brought the 74

PAGE 75

chamber pressure to 7E-6 Torr. This also resulted in a stable pressure reading of the mass spectrometer during growth. The next change wa s to decrease both metal fluxes, with a greater reduction for the calcium. The growth rates for the samples then became on the order of 10 /min, which had the positive benefit that only a fraction of a monolayer of the individual binaries would grow before the other would star t. Unfortunately, this resulted in having to manually flip the shutter switches for a longer tim e if a thicker sample was needed. Again the pretreatment and in-situ cleaning procedures were the same as th e previous set. The resulting XRD spectrum showed that at a magnesium fl ux of 3.7E-8 Torr BEP, and a calcium flux of 3.5E-8 Torr BEP, that the material could now be fa irly lattice matched to GaN. As can be seen in Figure 4-18, the ternary oxide peak is now ju st at a smaller spacing than the GaN spacing. The sample with 15 seconds magnesium and 5 seconds calcium still has a defined peak at higher angles, while the equal timed sample has a barely noticeable bump to the right of the GaN peak. The ultimate goal is to allow for equal timing of shutter sequences to make flux comparisons simpler. Therefore the next sample set locked the timing of the shutters to equal ten seconds exposures. The differences came with th e calcium flux which was varied from just over 8E-8 Torr BEP, to 1E-7 Torr BEP, while the ma gnesium flux remained constant at 4.4E-8 Torr BEP. The oxygen flow remained the same at 0.4sccm. The XRD scan in Figure 4-19 shows that both fluxes result in decent matching with the GaN substrate, yet the sample with slightly lower calcium flux showed no noticeable signs of a peak in the spectrum. This lack of peak indicates that the oxide peak is hidden behind the GaN p eak and is therefore una ble to be resolved. Fortunately, this is exactly what was hope d for and looks like the goal was reached. It was at this time that the time delay relay was installed on the syst em that was shown in Figure 4-16. Again, this assisted immensely in the growth process not only to the increased 75

PAGE 76

automation but for the increased precision of shutter timing. The following samples were all grown with this new addition. Now that the gr owth rate was slowed down, with magnesium flux of 3.5E-8 Torr BEP and calcium flux of 4.9E-8 Torr BEP, the variation of sequence timing is not as much of a factor. Seen in Figure 4-20, iden tical condition growths with timing of 10 seconds alternating fluxes, or with 15 seconds of magnesium and 5 sec onds of calcium, do not show any peaks around the GaN peaks. This indicates that the slight variation of metal atoms is not substantial enough to shift the latti ce enough to be noticeable in the XRD spectrum. Therefore, it is determined that the optimal method of as certaining the exact amount of each metal to incorporate would involve growths having a su bstantial difference in lattice and then extrapolating the difference. After slowing the growth rate down, it was thought that perhaps the oxygen flow should be slowed also. This notion is from a previous study performed on the straight magnesium oxide in which the oxygen flow was decreased and the br eakdown field of the sample was determined. The XRD results showed that a change from 0.5 sccm to 0.2 sccm did not show any oxide peak, indicating that the lower overpre ssure of oxygen does not play a s ubstantial role in the lattice parameter at growth rates around 1nm per minute. It was predicted that the hi gh growth rate that had previ ously been used, could be the reason that the composition of the film would change throughout the la yer thickness. This segregation of increased oxygen further from the GaN interface could possibly be reduced or eliminated by the slowing of the growth. It is a common practice to slow everything down during growth to fix one problem and then figure out the best method of growing faster. One of the great advantages to molecular beam epitaxy over other growth methods is to be able to control atom by atom growth of the sample. Ther efore, samples of the ternary oxide were grown 76

PAGE 77

at 200C and oxygen conditions of 0.25 sccm at 300W forward power on the plasma head. The flux of metals were 6.8E-8 Torr BEP for the magnesium and 9.4E-8 Torr BEP for the calcium. The sample was grown at 200C since this was th e condition of the earlier depth profile sample which showed compositional changes. The sample s were grown for thirty minutes, one with the digital approach timed to alternating ten sec onds metal exposures, since the metal flux ratios were calibrated to give approximately lattice ma tched growth, an the other in a non-digital or continuous mode with both metal shutters open at the same time. Initial surface scans of the growths (Figure 4-21) indicate a slightly higher calcium to oxygen ratio for the digital approach, indicating a Ca/O ratio of 65% and a Mg/O rati o of 35%, compared to the continuous growth with a Ca/O ratio of 63.2%, and a Mg/O ratio of 36.8%. The higher calcium content at the surface is expected since the last shutter open fo r the digital growth was the calcium. Since the 10 second exposure would lay down approximately one of two angstrom of material, the slight increase in calcium concentration is expecte d. The higher calcium calculated percentage is attributed to the sensitivity fact ors of the elements from the auger software. The ratio differences indicate the change for the samples. Next a de pth profile was conducted to verify or counter the earlier results. The depth prof iles are show in Figure 4-22. The results show that there is no significant segregation of the el ements within the thickness of the oxide film. This result supports the thoughts earlier that a slower growth would increa se the uniformity and atom incorporation into the film. As a note, the star t of the depth profile scan does show a drop in calcium content for both growths. This drop is attributed to surface roughness from the argon sputtering and surface residual carbon. The importance of this experiment is that it shows that with a slow growth rate and optimal conditions the digital growth me thod of alternating the metal shutters is not required to grow a uniform ternary film of MgCaO. 77

PAGE 78

Growth of MgCaO Via Continuous Growth Mode Once it was known that the MgCaO can be grown uniformly in a continuous mode while all source shutters are open to the sample, the next step is to find the optimum conditions for growth. Samples of oxide were tried at various temperatures, from 100C up to 600C. Temperatures between 300C to 500C all resu lted in nice terraces growth up to 100nm thicknesses. Thicker samples lost the terraced morphology as they grew thicker increasingly as the substrate temperature was lowered. Therefore it was concluded that the best temperature to grow MgCaO was at 500C. This temperature is above the metal source temperatures, and therefore once the metal atom hits the surface it wants to leave the surface, but by supplying higher oxygen flows, some blocking occurs and allows the sample to grow with a growth rate of around 1nm/min. The results are shown in Figure 4-23 which shows the AFM of the surface. Also Figure 4-24 shows the RHEED image during gr owth, showing a spotty pattern in columns which indicate a single crystal films with some surface roughness. Finally in Figure 4-25, the XRD pattern shows a lattice-matched film to the Ga N since the oxide (222) peak is not visible. Figure 4-26 shows XRD data supporting that bot h growth methods, digital and continuous, produce single phase, lattice-matched oxides. Since the oxygen flows were increased to s upply necessary overpressure on the sample, it was found that by increasing the plasma wattage, th e active species could be increased. Using an OceanOptics spectrometer, it was seen that by increasing the wattage from 300W to 350W forward power, the atomic oxygen emission increas ed, and indicating higher concentrations of oxygen atoms which can incorporate into the samp le. This result was supported by literature which indicates a larger amount of dissociation with both higher flow rates and higher powers.23, 24 78

PAGE 79

Therefore it seems that the optimum c onditions for MgCaO growth are by having a substrate temperature of 500C, and an oxygen fl ow of 1.0sccm with forward power of 350W through the RF plasma head. To maintain lattic e matching with the GaN, the flux of magnesium atoms is 4.9E-8 Torr BEP, while the calcium flux is 5.4E-8 Torr BEP. This set of conditions results in high quality smooth ternary oxide grow th with a matched lattice parameter to gallium nitride. MgCaO TEM Characterization In depth characterization of a 400nm thick MgCaO film was performed by TEM to characterize the interface properties of the oxide and GaN substrate. Shown in Figure 4-27, the oxide film shows good crystallinity while the substrate GaN shows regions of threading dislocations. These dislocations are created at the interface of the sapphire substrate for GaN growth. Threading dislocations are the main reas on that this oxide is be ing investigated. Since the oxide is still a cubic system, and the GaN is hexagonal, the slip systems are difference for the different materials. The oxide grows on the (111) plane on top of the GaN (001) basal plane. Since the threading dislocations of the GaN propagate parallel to the growth direction, the drive is to continue the dislocation propagation vertical th rough the top oxide layer. Fortunately for the research, the oxide has a slip plane which is perpendicular to the dislocation propagation direction. The MgCaO slip plane is the (111) plane. This plane is the face that is directly contacting the GaN basal plane. Therefore, as the thread ing dislocations from the GaN material impinge upon the oxide (111) face, they are halted. The only directions the defects can propagate are horizontal in the material, which is perpendicular to the growth direction. Being the case, the MgCaO therefore acts as a dislocation barrier. This barrier action prevents the threading 79

PAGE 80

dislocations from the GaN substrate to propagate into the top GaN layer which will be grown on the oxide, shown in later chapters. Therefore, since the oxide can be grown with the same lattice parameter as the GaN template, the film can be grown with minimal st rain in the film. When the subsequent GaN growth is grown upon the lattice-matched oxide s ubstrate, the film should have minimal stress during growth. Minimizing stress in the film then allows for a reduction in defect associated with the stress. In turn, less threading dislocatio ns will be present in the material, resulting in a high quality material which can facilitate better device properties. Therefore by utilizing the MgCaO as a substrate, defect propagation from the GaN under the oxide is minimized, and the top grown film will have less stress because of growth is commencing upon a lattice-matched substrate. MgCaO and MgO Thermal Expansion Measurements Since the oxide samples were grown on GaN at 500C, the lattice parameters are effectively matched at this temperature if inte nded through the metal fluxes. When the oxide continues for further top GaN gr owth, the lattice expansion at those temperatures might induce some strain into the top film, which is why determ ining the expansion coefficient of this oxide is important. Since a lattice-matched oxide peak ca n not be seen in XRD, because the GaN peaks overlap, then for high temperature XRD measurem ents, it is required that the oxides not be lattice matched to verify peak movement. Th erefore samples of calcium and magnesium rich ternary oxides were grown and annealed at elevat ed temperatures for a stress measurement. The samples were examined in a Phillips HT-XRD on a tantalum heater with an enclosed atmosphere of helium. Initially the binary expansion measurements were conducted. The CaO results are shown in Figure 4-28, which shows the peak degrees sh ift to lower values, indicative of increasing 80

PAGE 81

lattice spacing from the thermal ener gy. The shift in lattice shows a 2 angle decrease of 0.45 for the 400C change. The same is true of the MgO shown in Figure 4-29, having a 2 angle decrease of 0.65 for the 400C change. The GaN in both samples had the same shift in peak angle with a 2 decrease of 0.15 for the 100C to 500C m easurement. Therefore, supporting the literature results for the thermal e xpansion of the binaries, they are larger than that of GaN, with MgO having a higher TCE than CaO. With th is knowledge a good extr apolation would give approximately a fair estimate of a lattice-matched MgCaO TCE. Next, two samples of MgCaO were grown using the digital approach to allow for accurate alloying. The first sample grown had al ternating magnesium and calcium exposures of approximately 15 seconds magnesium, 5 seconds calcium, while the next had 5 seconds magnesium and 15 seconds calcium. The results were a bit confusing at first, since the ternary oxide peak would shift as expected then would start to disappear. The reason behind this, was that after scanning the full range it was noticed that the ternary was actually segregating into the binaries. Once the annealing temperature had reach ed approximately the growth temperatures of around 300C for these particular samples, the cha nge began to occur. Shown in Figure 4-30, the sample which was approximately 75% magnesium oxide, began to expand normally then shift to higher angles. The measurement was repeated twice in Figure 4-31 after the initial measurement, just to confirm no phase change during another heating. The expansion was then exactly the same as the binary MgO oxide expans ion, confirming that this peak was from MgO. At the end, a full scan was taken at room temper ature (RT) to verify that now two peaks are shown for the oxides in Figure 4-32. Next a sample of 75% CaO was measured in the same manner. The sample showed a ternary peak to the left of the GaN (004) peak, and to the right of the GaN (003) forbidden peak. 81

PAGE 82

For some samples of high quality GaN, the forbi dden peaks or duplication peaks appear. This is rare but is shown to occur in many material systems.25 The same result occurred as with the previous sample. Shown in Figure 4-33, the te rnary peak began to fade as the CaO peak emerged. Figure 4-34 shows only the room temperature, 300C, and 700C temperature for a more clear view of the binary peak evolutions of CaO and MgO. Finally Figure 4-35 shows the room temperature scan after the measurement. Finally, a sample with digital sequencing of 10:10, was stressed. The room temperature scan showed no oxide peaks and wa s therefore considered latticematched to GaN. The results of the HT-XRD are shown in Figure 4-36 which ve ry nicely shows the evolution of the binary oxide peaks. The stress measurements performed on the oxi de samples give insigh t into the workings of the materials at different temperatures that the sample might encounter during growths. The binary oxides show similar re sults to published data, while the ternary measurements gave interesting results. The fact that the ternary oxide would separate in to the binaries during annealing temperatures below 800C, show a st rong pressure dependence on diffusion within these samples. It was verified through additiona l experimentation that no segregation of phase separation occurred for a nneals in the vacuum of the MBE cham ber. After growth a sample was scanned in the XRD, then re-mounted for an anneal at 800C for one hour in the MBE chamber. The scan after anneal showed no oxide peaks, Fi gure 4-37, indicating a stab le film for annealing in vacuum. This result was also confirmed for the MBE GaN on oxide growths in Chapter 5. In addition to the XRD results, AFM was performed on the samples that were annealed at atmospheric pressure in the HT-XRD. Figur e 4-38 shows the surface of the MgCaO sample which was lattice-matched, before and after anneali ng. As it can be seen, the surface did become 82

PAGE 83

rougher after annealing. The terraces became more like trough and hills. The roughness increased from 0.46nm and 0.89nm for the 1m and 5m scans respectively, to 1.4nm and 1.8nm respectively. Also to note is the sample which was pure CaO. Shown in Figure 4-39, the surface before and after anneali ng are vastly different in term s of morphology and roughness. The roughness values for the scan area of one mi cron went from just unde r 1nm to over 8nm. The change is surface morphology supports the HT-XRD results which trend towards shortrange diffusion of the oxide. Since this material is unique in that no other research has been performed on the MgCaO system for thin films, the information available is quite sparse at the time of research. A phase diagram was found for full alloying between MgO and CaO from 1963, Figure 4-40.26 This diagram was for bulk material, and for temper atures between 1600C to 2800C, which is well above the temperature for this research. Also found was an isothermal section at 1227C for the ternary system of magnesium, calci um, and oxygen, shown in Figure 4-41.27 Even these are the only evidence of research done for the MgO:CaO system, the data is for bulk materials not grown with MBE. The kinetics of growth in MBE environments are vastly difference than growth at higher pressures and rates. The atomistic control of growth allowed for by MBE gives the ability to create materials that some phase diagrams say are not likely. Even though the phase diagram shows a two-phase region for the compositions that are investigated in this research, the x-ray analysis, RHEED analysis, and TEM analysis all show a single phase ternary oxide at growth conditions and in the room ambient environment. Mounting Methods for MOCVD Growth Since normal mounting of samples involve us ing the capillary forces of molten indium on a molybdenum block, a new method of mounting was developed for the oxide growths for subsequent MOCVD growth. I ndium would contaminate the gr owth chamber of the MOCVD 83

PAGE 84

during the high temperature high pr essure conditions for growth. Therefore, an indium-free mounting solution was determined through many trials. The first step was to use a molybdenum block with a recessed cut-out in the shape of either a full 2 inch wafer or a quarter of a two inch wafer. The samples would then be placed in from the back of this holder and clipped or wire d in with tantalum wire (Figure 4-42), which shows the back and front of a sa mple after mounting. Initial trials with this method resulted in oxide growths which were smooth, but lacked terr aces. From earlier trials with temperature calibrations, it has been shown that while the te rnary oxide can be grown with a low surface roughness at 100C, the morphology lacks the terraces desired for further nitride growth. The terraces then supply preferential nu cleation sites for the GaN growths. Next trial was to add a piece of silicon wafer to the back of the GaN substrate during mounting. Since silicon will absorb the IR ligh t from the heater, the thought is that perhaps enough phonon conduction will occur from the silicon wa fer and the rough polish of the back of the sapphire substrate. Since the vacuum cond itions of the MBE minimize heat transfer, the slight contact between samples did allow the sample to reach slightly higher temperatures based on the growth morphology, but still the subs trate was below desired temperatures. To allow direct absorption of heat to the subs trate, the next trials were sets of backside coatings of refractory metals. The first trial wa s with a coating of titanium and tungsten in an alternative layer configuration of (20nm/40nm) to a total thic kness of 180nm. The alternating layers were used to compensate for thermal expa nsion differences in the metals with the sapphire and GaN material. This coating allowed for a ho tter surface but was still s lightly transparent to the IR light, based on growth and the results of pyrometer. 84

PAGE 85

To compare substrate temperatures in th e MBE for oxide growth, a V-series Ircon Modline Plus infrared thermometer pyrometer was used. Since emissivity values are adjustable for calibration, the same emissivity was used for all measurements based off of the first sample run. The emissivity would be set to the thermocouple temperature based off of an indium mounted sample, whether this was the actual temperature of the substrate or not. Only the comparative values are of importance. Since all samples had a GaN surface, comparisons could be made along with the resulting oxide morphology, since samples less than 400C do not replicate the terraces of the starting GaN, while samples around 500C show nice terraces morphology. The next sample had a coating of 500nm of tungsten. This sample showed normal type growth terraces seen by indium mounting. Therefore the thicker co ating allowed for adequate IR absorption and heat build-up. The pyrometer in dicated that from all the mounting techniques and coating thicknesses, the thicker coating ha d the highest surface temperature, while the thinner coating was next, then the silicon back ed sample, and finally the lone-mounted GaN on sapphire wafer had the lowest temperature. Ot her samples of tantalum, and molybdenum were tested for heat absorption and to test in the MOCVD reactor for la ter growths. Those results are shown in the MOCVD GaN growth chapter. 85

PAGE 86

Figure 4-1. Graphical representati on of the lattices of GaN and MgO or CaO. Also included is the outline of the matching (0001) and (111) planes. 86

PAGE 87

OCaMg Therefore xyields x x aOCaMgxX5.05.0: %9.49499.0: 393.3)1()978.2(186.31 Figure 4-2. Vegard's Law results in a lattice matched ternary to GaN. 0.00.20.40.60.81.0 -8 -6 -4 -2 0 2 4 6 8 Percent lattice mismatchPercent Mg Figure 4-3. Lattice mismatch of oxide to GaN, showing the change with Mg content within a ternary alloy of MgCaO. 87

PAGE 88

Table 4-1. Material properti es for GaN, MgO, and CaO. GaN MgO CaO Structure Wurtzite Rock Salt Rock Salt Lattice Parameter (a) 3.186 4.2112 4.799 Atom spacing in (111) plane --2.978 3.393 Mismatch to GaN (%) ---6.5 6.6 Melting Temperature ( C ) 2527 3073 2860 Bandgap (eV) 3.4 8 7 Dielectric constant 9.5 9.8 11.8 Thermal Expansion Coeff. (K^-1) 6.20E-06 1.35E-05 1.28E-05 88

PAGE 89

3004005006007008009001000 3.175 3.180 3.185 3.190 3.195 3.200 3.205 3.210 3.215 Lattice Parameter vs. TemperatureLattice (A)Temperature (K) 54% CaO 52% CaO 50% CaO 48% CaO 46% CaO GaNFigure 4-4. Plot of variou s ternary ratios, with TEC included in calculations. 89

PAGE 90

Figure 4-5. Removal of carbonous photoresist from a sample exposed to UV-ozone in the small box ozoner. The removal rate was calculated to be 72 /min, showing effective removal of carbon from surfaces for this process. 90

PAGE 91

Table 4-2. Auger peak height ratios showing various clean ing processes and the resulting surface contamination. Carbon/Nitrogen Oxygen/Nitrogen As received 0.45 0.97 Acetone/Methanol 0.55 0.80 HCl/DI (1:1) (3min) 0.40 0.24 HCl (3min) / O3 (25min) ND 1.47 O3 (25min) / HCl (3min) 0.13 0.26 HCl (3min) / O3 (25min) / BOE (5min) 0.06 0.45 91

PAGE 92

A) (D B) (E C) (F Figure 4-6. XPS study of GaN surfaces before a nd after vacuum annealing at 700C. The figures on the left (A,B,C) did not have the standa rd pretreatment of HCl/UV-O3/BOE, while the spectrums on the right (D,E,F) di d have the pretreatment process. 92

PAGE 93

Table 4-3. XPS signal intensities of the contamination elements compared to the nitrogen peak of the samples that were pretreated with the standard cleaning process. No anneal 700C anneal Carbon/Nitrogen 0.74 0.60 Oxygen/Nitrogen 0.60 0.58 Table 4-4. XPS signal intensities of the contamination elements compared to the nitrogen peak of the samples that were not pretreated with the standard cleaning process. No anneal 700C anneal Carbon/Nitrogen 0.89 0.62 Oxygen/Nitrogen 0.81 0.39 93

PAGE 94

A) B) Figure 4-7. AFM images of MBE binary oxide s MgO and CaO. A) MgO growth, Ra=1.23nm vertical scale of 15nm. B) CaO grow th, Ra=0.815nm vertical scale is 15nm. 94

PAGE 95

A) B) Figure 4-8. XRD scans of the binary shown in Figure 4-4, the scans are also in Log scale to emphasize the oxide peaks in relation to the large GaN signal since the oxides are only a few hundred thick. A) CaO substrate showing the (222) CaO peak, the GaN peak, along with small peaks from the sapphi re substrate. B) MgO substrate shown the (222) MgO peak, the GaN peak, and addi tional stray peaks from the substrate. 95

PAGE 96

A) Binding Energy (eV) N(E) Min: 7Max: 18447 1000900 800 700 600 500 400 300 200 100 0 In 3d5 C 1s O 1s Mg 2p Mg 2s B) Binding Energy (eV) N(E) Min: 13Max: 27427 1000900 800 700 600 500 400 300 200 100 0 C 1s Ca 2p3 In 3d5 O 1s Figure 4-9. XPS scans of the binary oxides MgO and CaO, which show the corresponding elements in addition with an indium peak from the mounting technique and a carbon peak which is from atmospheric carbon from the transfer out of the oxide MBE and into the XPS chamber. A) MgO. B) CaO. 96

PAGE 97

A) B) Figure 4-10. Omega rocking curves of the (222) and (002) peaks of MgO. The FWHM of the peaks indicate the degree of crystallinity. A) (222) MgO omega RC, FWHM of 1.882 degrees. B) (002) MgO omeg a RC, FWHM of 4.198 degrees. 97

PAGE 98

0501001502000 10k 20k 30k 40k 50k 60k 70k 80k Ca Mg N Ga Ocycles 0204060801000 10k 20k 30k 40k 50k 60k Mg Ca O N Gacycle Figure 4-11. AES scans of continuously grown samp le (at left), and dig itally grown sample (at right).22 6870727476 5/15 70.175 8/12 71.79 10/10 73.9counts (arb)2(degrees) Figure 4-12. XRD scan showing change in at omic spacing of the MgCaO ternary around the large GaN (004) peak. This shift is accomplished through varying the time sequencing of the shutters: (Mg time(s) / Ca tim e(s)). It can be se en that adding more calcium, shifts the peak position to smaller angles which is a larger lattice spacing. 98

PAGE 99

Figure 4-13. XRD scan of standard UOE GaN used as substrates for the oxide calibrations. The GaN 004 peak can be seen which includes the alpha and beta signals. Along with the GaN peak can be seen some peaks from the sapphire substrate. 99

PAGE 100

A) B) Figure 4-14. AFM images of the starting UOE GaN substrate showing pits in the GaN film. A) Cross-section scan of a pit. B) 1m scan of an area without a pit, showing a smooth surface, yet lacking growth te rraces, average roughness (Ra)=3 100

PAGE 101

A) B) Figure 4-15. AFM images of our in-house grow n MOCVD GaN substrate showing nice growth terraces and roughness values of a few angs trom. A) 1mX1m scan area with roughness of 1 B) Larger scan area of same sample. 101

PAGE 102

A) B) Figure 4-16. Omron Zen computer programmed digital relay. A) Self-installed digital relay that interfaced with the shutter controllers to allow for automated switching of the source shutters. B) View showing th e relay to the right and the sw itches that were used to manually growth the digital samples. 102

PAGE 103

Figure 4-17. XRD scan of the initial set of digital MgCaO MBE growths. The spectrum indicates that increasing the amount of time that the magnesium shutter was open, caused a shift of the oxide lattice closer to the GaN peak near 73 degrees. The timing is in seconds, therefore 10/10 is 10sec onds of the respective metal followed by 10 seconds of the other. 103

PAGE 104

Figure 4-18. Sample set with varying timing of the shutter sequencing. The magnesium flux was 3.7E-8 Torr BEP and calcium flux was 3.5E-8 Torr BEP. The spectrum shows that with these fluxes, an approximately lattice-matched oxide can be grown with a shutter sequence composed mainly of calcium exposure. 104

PAGE 105

Figure 4-19. XRD scan of two samples with the same 10 second timing of shutter exposures, with varying calcium flux. As can be seen the sample with calcium flux of 8.2E-8 Torr BEP, and corresponding magnesium flux of 4.4E-8 Torr BEP. 105

PAGE 106

Figure 4-20. XRD scan showing The GaN peak s under a 50nm layer of MgCaO, in which the oxide peaks are aligned to the GaN peaks a nd therefore can not be seen. The samples were grown with the insta lled digital time delay relay at sequencing of 10 seconds and then alternating sequencing of 15 seconds magnesium and then 5 seconds calcium. 106

PAGE 107

Figure 4-21. Auger scan showing the concentrati ons of the elements at the surface before depth profiling. This top spectrum is for the c ontinuous growth indicating a Ca/O ratio of 63.2%, and a Mg/O ratio of 36.8%. The lower s can is the sample grown in the digital method, indicating a Ca/O ratio of 65% and a Mg/O ratio of 35%. The higher calcium content at the surface is expected since the last shutter open for the digital growth was the calcium. Since th e 10 second exposure would lay down approximately one of two angstrom of ma terial, the slight increase in calcium concentration is expected. 107

PAGE 108

Figure 4-22. Auger depth pr ofiling with argon sputtering, sh owing the comparison between ternary oxide growths with the continuous a nd digital growth modes. As seen, the large segregation of oxygen is not present as earlier samples, since the growth rate has been reduced. 108

PAGE 109

A) B) Figure 4-23. AFM images of continuously grow n MgCaO at 500C. A) Five micron scans area of the sample with average roughness of fi ve angstroms. B) Closer view of the sample, showing average roughness of just over three angstrom. 109

PAGE 110

Figure 4-24. RHEED image of MgCaO during gr owth in the MBE system. The spots in columns are indicative of a single cr ystalline surface with some roughness. 110

PAGE 111

Figure 4-25. XRD scan of lattice-matched Mg CaO showing no indication of the (222) oxide peak. 111

PAGE 112

A) B) Figure 4-26. XRD spectrum comparing the two growth modes of digital and continuous. Both samples are lattice matched to the GaN substrate. A) Full 2-Theta scan of the samples. B) Tight view of the GaN ( 004) peak showing no additional oxide peak. 112

PAGE 113

Figure 4-27. TEM images of lattice-matched MgCaO on GaN. The images are of the same sample, simply two different areas. The important note is that the large threading dislocations in the GaN subs trate are not propagating into the oxide. Also, the images are inverted to that the oxide is now above the GaN, for clarity. 113

PAGE 114

66 68 70 72 74 1 10 100 1000 10000 GaN CaOCountsAngle (2 Theta) 500C 300C 100C Figure 4-28. XRD stress measurements performe d on CaO. The expansion of the lattices of both GaN and CaO can be seen, as the CaO peak shifts more than the GaN. 70 72 74 76 78 80 1 10 100 1000 10000 100000 GaN MgOCountsAngle (2 Theta) 500C 300C 100C Figure 4-29. High temperature XRD measurements of MgO on GaN. As seen, the MgO lattice expand at a slightly larger amount than the GaN substrate. 114

PAGE 115

70 72 74 76 78 80 0.1 1 10 100 1000 10000 100000 1000000 RT scan 300C 400C 500C 600C 700C 800CCountsAngle (2 Theta) Figure 4-30. Ternary alloy rich with MgO during HT-XRD measurements. The phase separation of the ternary oxide is shown as the sample is heated above 500C. 707274767880 10 100 1000 10000 100000 CountsAngle (2 Theta) RT 300C 1st ramp 800C 1st ramp 500C 2nd ramp 800C 2nd ramp RT after 2nd Figure 4-31. Repeat of the HT-XRD from Figure4 -30. The sample was heated to 800C, back to room temperature, then heated a se cond time to confirm no phase change. 115

PAGE 116

65 70 75 80 85 1 10 100 1000 10000 100000 GaN (004) CaO (222) MgO (222)CountsAngle (2 Theta) Figure 4-32. Full XRD scan of the ternary oxide from Figure 4-31, after multiple heatings. The scan shows that the sample did pha se separate into the binaries. 66 68 70 72 74 0.1 1 10 100 1000 10000 100000 1000000 CountsAngle (2 Theta) RT 300C 500C 700C 800C Figure 4-33. HT-XRD measurement of a calci um oxide rich ternary alloy. Again, phase separation occurred. 116

PAGE 117

606264666870727476788082 0.1 1 10 100 1000 10000 100000 1000000 CountsAngle (2 Theta) RT 300C 700C65 70 75 10 100 CountsAngle (2 Theta) RT 300C 700C Figure 4-34. Room temperature, 300C, and 700C XRD of the sample in Figure 4-33, showing the evolution of the binaries. 117

PAGE 118

65 70 75 80 85 10 100 1000 10000 100000 1000000 GaN (003) GaN (004) CaO (222) MgO (222)CountsAngle (2 Theta) Figure 4-35. Room Temperature scan of the samp le in Figure 4-34, indicating the binary oxide peaks, the GaN (004) peak, and the GaN (003) forbidden peak. 65 70 75 80 1 10 100 1000 10000 GaN (004) CaO (222) MgO (222)CountsAngle (2 Theta) 100C 300C 500C 700C 800C Figure 4-36. HT-XRD measuremen t of a digital 10:10 oxide samp le. The sample was latticematched with GaN at room temperature, then as the measurement progresses, so did the evolution of the binary oxides. 118

PAGE 119

60 65 70 75 80 85 10 100 1000 10000 100000 CountsAngle (2 Theta) Figure 4-37. XRD scan of a lattice matched MgCaO sample that was annealed in the MBE growth chamber at 800C for one hour. No phase separation was seen, indicating that the phase change is pressure dependant. 119

PAGE 120

A) (B C) (D Figure 4-38. AFM scans of the lattice-matche d MgCaO sample that was annealed during the HT-XRD measurements. A) Before HT-XRD 5m scan area, z-scale of 10nm. B) Before HT-XRD, 1m scan area, z-scale of 5nm. C) After HT-XRD, 5m scan area, z-scale of 30nm. D) After HT-XRD 1m scan area, z-scale of 20nm. 120

PAGE 121

A) (B C) (D Figure 4-39. AFM scans of the CaO sample that was annealed during the HT-XRD measurements. A) Before HT-XRD, 5m scan area, z-scale of 15nm. B) Before HTXRD, 1m scan area, z-scale of 15nm. C) After HT-XRD, 5m scan area, z-scale of 100nm. D) After HT-XRD, 1m scan area, z-scale of 100nm. 121

PAGE 122

Figure 4-40. Phase diagram for MgO and CaO, above 1600C. Figure 4-41. Isothermal secti on of the ternary diagram of ma gnesium, oxygen, and calcium at 1227C. 122

PAGE 123

A) B) Figure 4-42. Mounting procedure for oxide samples grown for MOCVD GaN growths. Molybdenum blocks with quarter wafer cutout s. A) Back of holder showing clip and wire mounting. B) Front of the holder. 123

PAGE 124

CHAPTER 5 MBE GAN ON OXIDE Sample Preparation Once grown in the Riber MBE, samples of Mg CaO were then transferred into the Varian Gen II nitride MBE. The transfer process does subject the samples to atmosphere for a brief period of time. The short time at atmosphere was either used to transf er the oxide sample onto the Varian sample holder or for oxide characteri zation. If the sample was to be characterized with AFM or XRD, the samples were transferred to the instruments in a self contained vacuum container which contained desiccant to re duce the exposure to atmospheric humidity. Samples were mounted in one of two ways which included either indium mounting on a molybdenum block, or by wire moun ting full or quarter wafers havi ng a backside coating. When growths were on the order of a few hours, th e preferred method of mounting was with the indium. Utilizing a hot plate, a standard 4 mo lybdenum sample holder, also called a block, was allowed to reach temperatures of approximately 200C while resting upon a aluminum foil coated hot plate, Figure 5-1. The aluminum foil was fo r contamination issues to reduce particulate matter entering the MBE system. At this temper ature, indium had enough energy to melt but not enough to oxidize very rapidly. While in the liquid state, the indium was spread thin across the entire molybdenum surface with a stainless steel razor blade. Once the entire surface had been coated with indium, a small chunk of indium was placed in the center of the holder and allowed to melt and pool. At this point the actual sa mple to be grown upon was rested via tweezers on the pool. The sample was then gently moved around the block while the excess indium was scrapped off the block with the razor blade. As the indium under the sample became thinner and thinner, capillary forces began to immob ilize the sample. Once a good adhesion had been established and the sample was in the center of the block, the entire block was removed from the 124

PAGE 125

hot plate and allowed to cool. No rmally a strip of thin tantalum foil was used to clip the sample to the block with a screw. This clip gave extr a support to keep the sample adhered to the block. Even when the sample is up to growth temperatur es, this thin layer of molten indium would keep the sample from falling off the block, but the clip s gave extra support if the indium evaporation rate exceeded the time for growth. In the hi ghand ultra-high vacuum of the MBE growth chamber, the indium would begin to evaporate off the block when up to temperature. Since most of the time, the growth temperature of GaN was near 700C, the eva poration rate can be significant and cause loss of sample adhesion after many hours. The secondary method of mounting samples for GaN growth in the MBE was by the use of a molybdenum block that had a cut-out recess for the wafer itself. Therefore a full two inch wafer could be mounted and grown without the need for indium. This method had advantages as well as disadvantages. The main advantage was that without indium, there would be no evaporation of the mounting media, allowing fo r unlimited growth time. Unfortunately, since the wafer is recessed in the block, the surface of the sample is slightly below that of the molybdenum block itself. This height differenc e causes the electron beam from the RHEED gun to be blocked and unable to diffract off the sample surface. Therefore without the ability to watch the RHEED pattern, no in-situ characteriza tion of the growing surface was available. Ultimately, samples mounted within the recesse d blocks were grown with known recipes and were used for long duration growths. In addition these samples had to have a backside coating to allow for sufficient infra-red absorption from the sample heater. A coating of refractory metal such as tungsten, molybdenum, or tantalum, would be deposited on the backside of the wafer, which was discussed in Chapter 4. In addition to the refractory metal coatings, silicon could also 125

PAGE 126

be used as an IR absorbing coating, as the sili con has a very low vapor pressure at the growth temperatures normally used, even unde r the ultra-high vacuum atmosphere. Once the sample was mounted, the block wa s transferred into the Varian load-lock chamber. This chamber was initially pumped down by a roughing pump, then by a molecular drag pump. The pressure would drop to the neighborhood of 10E-5 Torr before the molecular drag pump was closed and the chamber would be opened to an ion pump pulling the pressure into the 10E-8s. If necessary, th e samples could be heated while in the load lock to remove the atmospheric moisture that is always present on samples after they have been exposed to atmosphere. The heated would be accomplished by heating lamps on the inside walls of the load lock. Since most of the growths require temper atures higher than what could be accomplished within the load lock, so this procedure would only be needed if a faster pump-down time was desired. The samples would then be transferred in to the buffer chamber via a magnetic trolley system. Once ready for a growth, the sample bloc ks were transferred in to the growth chamber with a magnetic transfer arm. The block would be attached to the substrate heater by the standard pin and spring mechanis m employed by most MBE tools. Solid sources available on the nitride MBE include silicon, magnesium, gadolinium, scandium, aluminum, in addition to gallium. The gallium is containe d in a pyrolytic-boronnitride crucible in a hot lip effusion cell. The hot lip cell helps reduce th e build-up of gallium at the end of the crucible which c ould eventually block the gallium from exiting the cell. The gallium cell temperature would normally be se t to approximately 890C, which would produce a flux with a beam equivalent pressure, or BEP, in the range of 2E-7 Torr. The gallium was 126

PAGE 127

research grade of six nines pure. Without th is high purity, contamin ation would result in significant doping of the gallium nitride film. The nitrogen source was supplied by passing molecular nitrogen through a radiofrequency plasma head. The Oxford plasma head would operate at 300W forward power, at 13.56Mhz. Cracking of the molecular nitrogen resu lted in atomic nitrogen, along with charged species and electron to be emitted from the PBN cr ucible of the plasma head. A diffuser plate was utilized to keep a backing pressure in the crucible and to spread the exiting species across the entire substrate holder. This allows for approximately a uniform grow across the full two inch wafer if needed. Gas flow was varied for the different experiments, and varied from 0.5sccm up to 2sccm, resulting in a chamber pressure between high 10E-6 Torr to low 10E-5 Torr. Normal procedure for initiati ng nitride growth, first involves heating the substrate facing the sources while the shu tters are closed. The nitrogen is not flowing, and theref ore the pressure in the chamber is 10E-9 Torr. For GaN on GaN calibration samples the nitrogen plasma is lit and shuttered open, exposing the sample to the n itrogen plasma, when the sample reaches300C. For the growth on oxides, the sample is not exposed to the nitrogen plasma until the sample temperature reaches the growth temperature. Ther e is normally a pressure spike to the low 10E8 Torr when the surface water as the surface water is heated until desorbing from the surface of the block. The pressure then again falls one d ecade as the water is pum ped out of the chamber by the cryo pump or the liquid ni trogen cryo panels. The samples were heated to 700C and allowed to rest at this temperature for ten minutes to allow a final cleaning of volatiles from the surface. Then the sample temperature is lowere d to the growth temperature. Once the desired growth temperature was reached the source sh utters were opened and growth would begin. 127

PAGE 128

The 700C annealing step was determined to result in maximum surface cleaning while having a minimum surface stoichiometric change. It is shown that below 750C in a MBE ultrahigh vacuum environment, that the decompositi on rate is almost zero. Since the activation energy for decomposition in a vacuum environment for GaN is 3.6eV21, that the decomposition rate rapidly increases above 800C GaN calibration growths on GaN The Varian MBE utilized for this research was initially not set up for smooth gallium nitride growths. The system was utilized for the research of spintronic materials, which require different conditions for growth. Spintronic ma terials also do not requ ire the smooth, terraced surfaces that this research requires, and they usually have fairly low surface roughness but a very bumpy surface texture. Since this was the case, the first challenge was to figure out a set of conditions that would give rela tively smooth GaN material. To begin the process of growing smooth Ga N, homoepitaxy was attempted. Substrates of GaN on sapphire were to be used as the starting templates. The reasoning behind this method is that the first issue is to be able to grow good quality smooth GaN. By growing homoepitaxially, the issues encountered during nucleation can be avoided by gr owing on a good template and the work can be focused on the actual recipe fo r good GaN. MOCVD samples from Uniroyal and our in-house Emcore/Veeco Turbodisc reactor we re prepared for growth by our standard cleaning recipe of HCl, UV-oz one, and buffered oxide etch. Samples were grown between temperatures of 630C up to 750C. This temperature spread was to determine the correct surface energy needed to allow for migration of the atomic species once they impinged on the sample. If the temper ature is not sufficiently high, then the atoms dont have enough energy to migrat e on the surface to find a site to rest. When the sample surface is too hot, then the atom s will either land of the surf ace and then not stick, which 128

PAGE 129

decreases growth rate significantly, or they wi ll have too high mobility and begin to cluster together since the top of a cluste r is the highest energy site to re st. Both of these scenarios are undesirable and lead to poor growth morphology. Therefore, finding the correct growth temperature along with the correct amount of atomic fluxes is cruc ial to good crystalline growth. All samples were also grown under relative ly similar gallium conditions. The flux of gallium atoms from the Knudsen cell had a beam e quivalent pressure, or BE P, of 1.5e-7 to 2.5e-7 Torr. This rate of metal atoms woul d result in a GaN growth rate around 10 /min. Now, while the gallium flux was maintained within a small ra nge, the nitrogen flux was varied considerably. The nitrogen flow into the RF plasma h ead was altered from 0.4sccm (standard cubic centimeters) up to 2sccm. The change was to determine a V/III ratio that would result in smooth GaN growth. Determining the correct III/V ratio is crucia l to good growth of gallium nitride. Too high a ratio will result in a N-stable regime that results in pit formation in the growing surface, while too low a ratio will allow excess ga llium to pool together in droplet s. The correct ratio allows a smooth surface to grow and become atomically flat.28 Knowing this scenar io is only have the battle, since the ratio will change not only by the amount of fluxing atoms, but also by the substrate temperature since higher growth temperatures can preferentially limit the incorporation of the gallium atoms. Initial samples were grown at a substrate te mperature of 630C, and gallium flux of 2e-7 Torr BEP. Nitrogen flow was kept low, 0.7s ccm, to establish a low V/III ratio growth morphology. RHEED indicated a smooth surface before growth, showing streaky (1X1) reconstruction of the starting GaN surface. Afte r growth was initiated, a faint (3X2) pattern emerged Figure 5-2, but faded 20-30 minutes into growth, after approximately 20-30nm of GaN 129

PAGE 130

had been deposited. The resulting RHEED has a streaky/spotty appear ance indicating a rough 3D surface, which was confirmed by tapping mode AFM in Figure 5-3. The surface roughness is not terrible, as it resulted in about a 1nm roughness across 1m2. The AFM roughness comparisons were performed on a 1m2 area, and a 25m2 area was scanned just to confirm uniformity. Also, in Figure 5-3 large circular pits can be seen in the AFM image. These pits are pits in the starting Uniroyal GaN surface from po or CVD growth. While the pits in the Uniroyal substrates would be detrimental to devices mate rial, as a calibration s ubstrate, they have no effects on the overall growth. At that point, the gallium flux was adjusted from 1.77e-7 Torr BEP, to 2.53e-7 Torr BEP. This resulted showed that above 2e-7 Torr BEP, the surface began to roughen to just under 2nm roughness, while the samples grown at 2 and 1.77e-7 Torr BEP had equivalent roughness of around 1nm average roughness (Ra) as shown in Figure 5-4. The next trial was to adjust the substrate te mperature from 630C up to 725C. This is the temperature range that most MBE GaN is grown, so should be the range to give the best results in this system. As can be seen through Figure 5-5, the change in substrate temperature (Ts) from 725C to 700C reduced the averag e roughness but did not change the morphology of the film. The grainy texture is still pr esent, and the higher roughness ca lculated was probably from the depression in the scan area. When the samp le temperature was reduced to 630C, the film morphology changed into a more platelet, smooth surface. The roughness value is still high on this sample, but this is presumed to be from the round features on the surface. These features seem to be gallium droplets that have accumulate d on the surface. Droplets form when either the growth is too high in gallium flux, or the substrate is too cold to incorpor ate the gallium atoms. The colder surface is the reason behind this result Understanding that 630C is too cold for the 130

PAGE 131

growth, and 725C seems too high, the next experiments were conducted at substrate temperatures of either 675C or 700C. After the previous experiments, the gall ium effusion cell was changed during a chamber venting. The vent was due to the substrate heat er not functioning properly. When inspected, one of the power leads to the heater had separated. The power lead was replaced and reinstalled. While the chamber was up to atmosphere, the ga llium cell was changed for a new dual-filament cell. This dual filament cell would give better control of the gallium flux and by having the lip of the cell hotter than the base, there was a reduc ed chance for gallium droplets to form at the end of the cell and in turn leak out shorting the cell. This change in cell br ought a change in cell operating temperatures, but the gallium flux was adjusted for temperature and therefore flux conditions would be repeatable from previous experiments. Finally, the third variable in sample conditions was varied. The nitrogen flux was controlled by the flow of molecu lar nitrogen through the RF plasma head, which was operated at 300W. Gas flow was varied between 0.5sccm up to 1.5sccm. This change in flow produced a chamber pressure difference between 1E-5 Torr to 3E-5 Torr. This change in gas flow could have two effects on the surface. The higher gas flows would be positively supplying more atomic nitrogen species to the sample surface fo r incorporation, since only atomic nitrogen would be used for GaN growth because molecula r nitrogen will not crack on the 700C surface. The disadvantage for the higher flow could be site blocking and crowding on the sample surface. Site blocking would have the eff ect of creating pits and changing the growth mode to 3D. Since smooth GaN still had not been proven in this MBE instrument, the higher flows were not a concern as these runs were calibrations to find a regime that would produce smooth GaN. 131

PAGE 132

The effect of increasing the ni trogen flow was quite dramatic. Seen in Figure 5-6, as the gas flow was changed in steps from 0.5sccm, to 1.osccm, and finally to 1.5sccm, the roughness of the films was reduced. Subs trate temperature was 700C and gallium flux was 2.2E-7 Torr for all growths. Through increasing the nitrogen impinging on the sample surface, the surface pits began to be reduced. The change from 0.5 to 1.0sccm had a large effect on the sample morphology as the 1.0sccm sample began to have areas of large, flat growth. The average roughness increased slightly for th e 1.0sccm over the 0.5sccm, but this is presumably due to the large open pit areas. When the flow was further increased to 1.5sccm, the larger flat areas began to be replaced by long connections of brain-like growth. This flowing type of growth was much improved on the roughness from around 3.5nm to 1.5nm, compared to the lower flow conditions. The low roughness value is the desire d outcome and therefore 1.5sccm seemed to be the best condition for nitrogen flow. The gas flow of 1.5sccm produced a background pressure in the chamber of 3E-5 Torr, which is low enoug h not to cause beam scattering from the sources. Combining the best of the experimental va riables, the sample was grown under the best conditions possible for this MBE system. The cond itions of growth were a substrate temperature of 675C, a gallium flux of 2.2E-7 Torr BEP, and nitrogen plasma of 300W at 1.5sccm. The result was a further reduction in surface roughness down to below approximately 1nm, as seen in Figure 5-7. Finally, the conditions that resulted in the lowest roughness and best surface morphology were performed on high quality GaN substrat es. Using our in-hous e grown MOCVD GaN, which shows very clean well defined terraces (F igure 5-8), the growth commenced. Figure 5-9 shows the starting GaN substrate along with the final MBE GaN grown on top. The roughness of the sample is just under 7 which is very suitable device material. 132

PAGE 133

MBE GaN on MgCaO Once a suitable recipe for growing MBE GaN was determined from the previous section, the ultimate step is to grow this material on the lattice matched magnesium calcium oxide. The MgCaO was grown in the Riber oxide MBE which is in the same laboratory as the Varian nitride MBE. The ternary oxide was grown upon a starting GaN substrate, and the binary components were adjusted so that the ternary would have the same lattice spacing as the GaN. Being this the case, it would be assumed that the growth would be similar to the GaN on GaN growths. After growth in the Riber MBE the star ting MgCaO substrates were remounted for growth in the Varian system. The reason behind the remounting is that for the beginning oxide samples, the pieces were indium mounted to the molybdenum block. Under the oxygen plasma, the indium on the block absorbs quite a bit of oxygen during the growth run. If the oxide sample block were to be placed directly into the ni tride chamber and heated, excess oxygen from this indium would be emitted into the chamber. This oxygen is undesirable for nitride growth, since the oxygen would contaminate the source materials, and could be incorporated into the growing GaN film. Oxygen doping in GaN acts as an n-t ype dopant. The is caused by either a gallium vacancy, or more likely a gallium vacancy next to a oxygen sitting on a nitrogen site.29 Therefore, the samples of ternary oxide were remounted on fresh indium on a molybdenum block that is normally used for growth in the Varian nitride MBE. When the samples are remove d from the oxide growth blocks, they are packaged into sealed containers and placed in a portable v acuum jar for transport to the Major Analytical Instrumentation Center for characterization. Th e samples were scanned by AFM to determine surface roughness and morphology, in addition to X-ra y diffraction to verify lattice spacing and matching to the GaN substrate. In all, the samples would be exposed to atmosphere for 133

PAGE 134

approximately 2 hours between emergence from th e oxide MBE and installation into the nitride MBE. This minimizes exposure to atmospheric humidity, which can be quite high in Florida. Once loaded into the Varian system, a sample is transferred into the growth chamber and heated up facing opposite the sources. This practice helps to mitigate any stray gallium or nitrogen from hitting the sample surface before desired. While facing in the opposite direction, the source atoms would have to bounce off the back of the chamber wall and hit the surface, which is highly unlikely since the walls of the ch amber are cooled by liquid nitrogen. Since the samples were exposed to atmosphere during the transfer, they were heated to 700C for a ten minute annealing clean. This is the same pro cedure used in all our GaN cleanings, and works equally effectively for the stable oxides. The oxides are very stable in the vacuum environment of the MBE chambers. This was verified through not noticing any change in the chamber pressure during the anneal to 700C, and in addition, the samples had been tested up to 900C in the chamber and no pressure change was detected. In addition, the ternary was scanned by XRD directly after growth and after annealing in the MBE chamber up to 700C and no change was noticed. This data was shown in Chapter 4. The first MBE GaN growth on MgCaO, turned out to be slightly rough, on the order of 7nm average roughness, which was also visible in secondary electron SEM images, Figure 5-10, along with EDS analysis of the film. Since th is sample was only grown for two hours, which would have grown a film on the order of 100n m, the underlying oxide film shows up in the spectrum. The sample was grown at 630C, since this was the normal temperature used to grow the dilute magnetic materials commonly grown in the Varian MBE. To compensate for this low temperature, the gallium flux was approximately 2E-7 Torr BEP, and the nitrogen flow was increased to 1.6sccm. The RHEED indicated that after a few minutes in to the growth, there 134

PAGE 135

seemed to be a poly-crystalline pattern beginning to form, indicated in Figure 5-11. This pattern disappeared and a standard strea ky/spotty pattern emer ged indicating a 3D crystalline surface. This trend of RHEED pattern from a slightly poly to a rough crystalline surface seemed to be a common theme in most all the MBE GaN on oxide th at followed. A possibl e explanation for this event is that the surface coverage is not comp lete until a short time of growth has past. Therefore localized regions of the sample ar e not fully coalesced and therefore the RHEED pattern depicts a poly-crystalline surface while it is actually not a complete coverage until a short time has past. Also the heights of the plateau s from the starting GaN, replicated with the oxidized, would only allow the electron beam from the RHEED gun to hit some of the high plateaus and therefore look to be poly-crystalline. The surface of the starting oxide was terraced from the replica tion of the substrate GaN. This oxide was grown in a digita l approach that was explained in Chapter 4. Shown in Figure 512 was the starting MgCaO surface, along with AFM images of the MBE GaN. Again the surface did not replicate the terraces, but none of the calibration samples followed the substrate morphology. Also for additional verification of growth, a powder XRD scan was taken on the sample. Shown in Figure 5-13 the (004) GaN peak s, alpha and beta, are seen and no additional peaks. This indicates that the MgCaO is lat tice matched and remained as a single phase during the two hour growth of GaN at 630C. This was in addition to the 10 minute 700C annealing that was performed on the oxide post growth in the Ri ber MBE chamber. The results indicate that the oxide can be successfully used as a stable template for subsequent GaN growth. Subsequent GaN on oxide growths were pe rformed at higher substrate temperatures, following the better results seem at 675C and above Therefore the next growth set was grown at substrate temperatures of 700C and 800C to exam ine if the higher temper ature results in higher 135

PAGE 136

surface mobility and therefore be tter growth. During growth of the 700C sample the RHEED showed approximately the same pattern as th e previous growth at 630C, Figure5-14, where a slightly poly-crystalline pattern was seen for a short time dur ing the beginning of growth, followed by a 3D crystalline pattern after one a nd two hours of growth. The AFM images of the surfaces showed that as the temperature was increased from the 630C to the 700C, the film actually became slightly rougher with average roughness just over 8nm. The next sample was grown at 800C, and the roughness decreased below 3nm, but the morphology took on larger grain-like features, al l oriented in the same direction. This is likely the cause of too high growth temperature that resulted in the gallium atoms migrating to clusters and forming larger nuclei from the beginning of growth. The tapping mode AFM images are shown in Figure 5-15. After the initial results were examined, it was proposed that the start of growth was the issue with the rough morphology of th e growths. Therefore, since AlN has been used as a buffer layer for growth of GaN on sapphire it could possibly help the growth on MgCaO as well. 30, 31, 32 The aluminum nitride has an innate ability to abso rb strain and allow for less defects to be grown into the GaN layer. Being the case, the same conditions from the previous growths at 700C and 800C were repeated, now using a short alumin um nitride capping laye r on the oxide. The samples were exposed to aluminum at a flux of 2.7E-7 Torr BEP, while the nitrogen plasma shutter was opened and allowed to grow aluminum nitride for one minute. This should create only a few monolayer of AlN, but possibly assist in the GaN growth. The results of the exposure are shown in the AFM images in Figure 5-16. As can be seen, the AlN exposure did allow a smoother growth at the 700C substrate temperatur e, but did not show an improvement during the growth at 800C. In addition, the aluminum e xposure did not show any noticeable effects on the RHEED pattern, from Figure 5-17. These sa mples are all tabulated in Table 5-1. 136

PAGE 137

During the growths at 700C and 800C fo r the AlN exposure and those without, the growths were doped n-type with silicon from the solid silicon source on the Varian MBE. The silicon source had a flux of 3.5E-10 Torr BEP, which was barely detectable with the flux monitor. This incorporation of silicon into the growing films was intentional to determine electrical properties of the growths. As can be seen in Table 5-2, the incorporation of silicon into the growing films was effected quite a bit by having the aluminum nitride exposure. It was at this point that the substrate heater in the system malfunctioned. While the heater was being repaired, the gallium effu sion cell was exchanged for a brand new dualfilament cell. Again, this is a very nice oven to us e since the ability to heat the lip of the crucible hotter than the base, allows for no built-up of gallium at the end of the cell. Following the gallium cell exchange, the samples were all grown at 675C in accordance with the calibration results. Also the nitrog en flux was set at 1.5sccm, resulting in a chamber pressure of just over 3E-5 Torr. The gallium was set to a flux of 1.8E-7 Torr BEP. The first two samples were 60 minute growths set on substrates of the individual binary oxide. A sample of MgO and CaO were mounted on the same block and grew for 60 minutes. The resulting GaN film was very granular shown in Figure 5-18, wh ich also includes the AFM images of the oxides before GaN growth. The XRD scans of these oxides (Figure 5-19) show the exact lattice positions of the oxides relative to the GaN peak. Therefore, this result shows that the conditions for GaN growth in the Varian MBE system ar e not set for nucleati ng on a non-lattice matched substrate. The next results were based on the fact that for device material, the GaN layer must be thicker than a few hundred angstroms. Theref ore, the following growth was attempted for 500 minutes. This would explore the film growth properties after a longe r time at the growth 137

PAGE 138

temperature and examine if any cracking from st ress would be introduced from a thicker film. The oxide substrates used in this run were la ttice-tuned MgCaO, verified by XRD in Figure 5-20, mounted on the same block. During growth of these substrates, the RHEED images looked similar (Figure 5-21) with the only differences be ing that that sample B was grown thicker to get an idea of how the oxide morphology would change on longer growths. The differences in the RHEED images come from the fact that the thicker oxide exhibited charging effects on the surface, making the RHEED spots look larger and blu rry. This is an expected result as these films have high dielectric constants and theref ore hold a charge from the impinging electron beam. The oxide surface through AFM showed that the thicker oxide did begin to roughen, increasing to almost 1nm, compared to the thinner film at just over 0.3nm, shown in Figure 5-22. The GaN growth on these substrates was accomplishe d with the same conditions as the previous growths on the binary oxides which include galli um flux of 1.7E-7 Torr BEP, 1.5 sccm nitrogen flow, and a substrate temperature of 675C. The resulting films were quite different in morphology yet similar roughness values, as show n in Figure 5-23. The film grown on the thicker, rougher oxide, had a more bumpy l ook to the surface, while the GaN grown on the thinner oxide had an almost leafy look as there were regions of flat smoot h growth. A scan from a Veeco Wyko optical profilometer was utilized to scan the area near where the tantalum clip was placed to secure the sample to the block. Since the clip blocks the impinging atoms from reaching the surface, area under the clip has exposed oxide. The step height measured is indicative of the thickness of the MBE grown GaN, which is around 0.4m, shown in Figure 524. For additional insight into the GaN growth, a cross-sectio n was performed on the sample with the leafy morphology. As can be seen in Figure 5-25, the thickness of the oxide and GaN correspond well to the predicted and measured valu es. Also in the SEM images, it can be seen 138

PAGE 139

that the MBE GaN structure does follow a slight ly 3D morphology as th e growth progresses, with better material the furthe r from the interface. This corresponds to the RHEED pattern, which showed that the streaky/sp otty 3D pattern was beginning to become less spotty and more 2D growth. It can then be in ferred that perhaps an even longe r growth could result in a smooth film. After all the calibrations and experiments with the GaN on MgCaO within the MBE, it was time to grow a sample over 1m that could then be used to further processing, since this would be about the minimum thickness needed fo r device material. The final growth was performed on a 130nm thick lattice-matched Mg CaO. The oxide surface was quite smooth and had terraced growth. This sample was a full 2 wafer, with a backside silicon coating of 10m which was for efficient heating of the substr ate through infra-red light absorption from the heater. The wafer was mounted in a recessed bl ock with tantalum wire holding the sample in place. After oxide growth, the wafer was transfer red, via a portable vacuum container, for AFM imaging. After imaging, the sample was loaded in to the load-lock of the Varian MBE system. For this sample, the gallium flux was increased to 2.2E-7 Torr BEP, to facilitate a faster growth rate, since previously the growth rate was on the order of 8 /min. All other conditions remained the same. The growth was conducted for 1180 minutes or almost 20 hours. After growth, the sample was characterized through AFM, and cro ss-section FESEM. The AFM of the surface indicated a very rough morphology on the orde r of 20nm, with a ve ry hill and trough morphology, as can be seen in Figure 5-26,. As a comparison of the growth surfaces, Figure 527 depicts the starting MOCVD GaN surface, th e MgCaO growth, and finally the MBE GaN surface. The sample was cross-sectioned and im aged in a field-emission SEM. The images, shown in Figure 5-28, indicate a MBE GaN growth of columnar structures on the MgCaO. The 139

PAGE 140

structure of this growth is probably a result of the increased gallium flux and growth rate, which was increased from 8 /min to 12 /min. Also seen is an area were the MBE GaN film had delaminated from the oxide, and also areas we re the MBE GaN held onto the oxide, and the oxide separated from the MOCVD GaN. This indicates equivalent bonding between all the layers of the structure. In addition to the morphology characterizati ons, the sample was then analyzed through XRD. Shown in Figure 5-29, the sample shows some sapphire peaks, the (004) GaN peak, and the oxide peak which is slightly to the right of the GaN peak. Th ere are no additional peaks from MgO or CaO, indicating that even a 20 hour growth at 675C in the MBE caused no phase segregation of the oxide. Finally, optical characterization of this thick sample was performed. Cathodoluminesence (CL) was performed on the sample to see the areas of defects present in the sample. Cathodoluminesence gives a good indication of quantity of threading dislocations, since these defects show up as dark areas in the imag e. As a comparison, a sample of standard inhouse grown MOCVD GaN on sapphire is shown al ong with the thick sample (Figure 5-30). These CL images of the MBE GaN and the MO CVD GaN were gathered with the same photon counts per second therefore dir ect comparisons can be made. It is obvious that the concentrations of threading dislocations are much less in the MBE GaN than the MOCVD GaN. To verify if all the luminescence from the samp le is coming from the band-edge, or if point defects are playing a role in the image, photoluminescence was pe rformed. Shown in Figure 531, photoluminescence was performed to get ad deep er insight into the CL images from Figure 5-30. What results is showing how the standa rd in-house MOCVD GaN has a strong band-edge, but multiple defects. The yellow emission aroun d 550nm is explained as mainly being from 140

PAGE 141

nitrogen defects.33 Along with the yellow emission, a blue emission around 450nm was shown in the spectra. This blue emission can be attributed to oxygen contamination that is from a contaminated gallium source of the system. This blue emission is not seen in the MBE grown GaN, which is a really good sign that autodopin g is not occurring, and the MgCaO substrate is very stable during growth. The MBE GaN also had a strong band-edge emission, along with a much larger yellow emission than the MOCVD Ga N. Based on the CL image, which showed a decrease in threading dislocati ons, this yellow emission is mo st probably from a very large population of nitrogen vacancies. This increased nitrogen vacancy is quite probable since the growth was not well tuned and therefore is a ma in reason why there is so many point defects causing the large emission. XPS Study of MgCaO Annealing It has been found that by exposing samp les of MgCaO to atmospheric humidity for extended periods of time, results in the sample becoming roughened and visibly changing colors. Also, the oxide rapidly etches in water. This supports the idea that a hydroxide layer forms readily on the sample surface and over time begins to etch away at the material. Since samples are transferred in air between the oxide growth to ol and the MBE nitride system, it is only logical that a surface hydroxide will form. Therefore, to examine this surface layer and the effect of heating the sample, the sample will be analyzed by XPS. The XPS analysis chamber is connected to the nitrid e MBE via a buffer extension, thereby maintaining vacuum for the sample during transferring. A study was found that worked to char acterize hydroxides of both MgO and CaO.34 The research used XPS to determine the binding energies of the hydroxides and the temperatures under vacuum at which point the hy droxide is removed from polycry stalline oxides. Therefore, 141

PAGE 142

by analyzing a sample of MgCaO, similar results should be obtained since the surface of the ternary oxide should have both magnesium and calcium oxide bonds exposed. A sample of lattice-matched MgCaO with thickness of approximately 130nm had been grown on a full two inch GaN wafer with silic on backside coating. The sample was scanned with XPS before any heat treatments, which show the elements of interest with some carbon from exposure to the atmosphere in Figure 5-32. Both the photoelectri c lines and some auger lines are present in the spectrum, yet only th e photoelectric lines are analyzed for energy changes. A full spectrum of the samples was taken, along with multiplex scans of the magnesium, calcium, and oxygen peaks. The peak s of interest were the highest intensity photoelectron lines corresponding to the Mg2s, Ca2p, and the O1s. The multiplex scans will give a higher resolution with steps of 0.025eV with pass energy of 17.9eV and 50ms per step. The x-rays were produced from a Mg anode which gave higher counts than an Al anode. Neither anode would have overlapping peaks with any of th e three elements of inte rest. A curve fit was performed on the magnesium peak since the peak did have a slight hump at higher energies, indicative of a hydroxide. The curv e fit was for clarity of the loca tion of the peak in Figure 5-33. The sample showed strong peaks correspondi ng to the magnesium, calcium, and oxygen, along with a peak for carbon. This carbon is atmosphe ric carbon that settled on the surface during the exposure to atmosphere during the AFM characteri zation. Multiplex scans of the elements show a single peak for the magnesium, while the calc ium revealed three separate peaks, and the oxygen showed a large peak and a much smaller peak at lower energy. This sample was then transferred through the buffer extension into th e Varian MBE. The sample was heated to 700C in 10 minutes at a pre ssure of 4E-9 Torr. After annealing at 700C for 10 minutes, the sample was allow to cool in 20 mi nutes in the chamber. Again the sample was 142

PAGE 143

moved to the XPS chamber and a duplicate scan was performed, both full spectrum and multiplex. After the anneal, the carbon peak was no ticeably reduced, which is to be expected. The multiplex of the magnesium showed a strengthening of the peak, which is again expected since the surface carbon was reduced, allowing fo r a stronger magnesium signal. The same effect occurred with the main calcium peak, yet the oxygen signal was reduced. This reduction in oxygen signal is a good indicati on that a surface hydroxide layer had been removed. Also to note is that the side peak from the magnesi um signal, which is i ndicative of magnesium hydroxide, disappeared. The oxygen peaks are th e peaks bound to the calcium and magnesium, and after anneal there was a shif ting of the peak energies, indi cating a change in surface oxygen bonding. This is a result shown in the published wo rk sited earlier indicati ng hydroxide removal. For the calcium spectrum, the high energy peak corresponding to calcium hydroxide was almost eliminated after the 700C anneal. The multiplex data for before and after annealing for oxygen and calcium is shown in Figures 5-34 and 5-35, respectively. Finally, a full scan was taken to show a reduction in surface carbon from the heating in Figure 5-36. This data corresponds to the literature, which states that the hydroxide of MgO is removed in vacuum around 240C, while the hydroxide from CaO is a bit more robust and does not decompose until temperatures around 500C. Therefore, this shows that for growth of GaN on the ternary oxide, annealing th e sample at a high temperature is needed to decompose both hydroxides of the MgO and CaO complex to obtain a clean surface. 143

PAGE 144

Figure 5-1. Molybdenum sample holder block on hotplate used to perform indium mounting of samples for MBE growths. Figure 5-2. RHEED image of Ga N on GaN growth having a (3X2 ) reconstruction, that emerges shortly after the start of growth. 144

PAGE 145

A) B) Figure 5-3. Tapping mode AFM of initial results of GaN on GaN MB E growth. A) 1m area of the growth at 630C. B) Larger area view showing the pits which originate from the Uniroyal GaN substrate. 145

PAGE 146

A) B) C) Figure 5-4. AFM tapping images of MBE GaN on GaN at different gallium fluxes, 1m2 areas A) Ga flux 1.77E-7 Torr BEP, Ra: 1.01nm, ver tical scale: 10nm. B) Ga flux 2.05E-7 Torr BEP, Ra: 1.02nm, vertical scale: 10n m. C) Ga flux 2.53E-7 Torr BEP, Ra: 2.03nm, vertical scale: 20nm. 146

PAGE 147

A) B) C) Figure 5-5. AFM tapping images of MBE GaN on GaN at varying substrate temperatures, 1m2areas. A) Ts=725C, Ra: 2.97nm, vertic al scale: 20nm. B) Ts=700C, Ra: 1.57nm, vertical scale: 10nm. C) Ts= 630C, Ra: 2.89nm, vertical scale: 20nm. 147

PAGE 148

A) B) C) Figure 5-6. AFM tapping images of MBE GaN on GaN at varying nitrogen flows, 1m2. A) Flow=0.5sccm, Ra: 3.22nm, ver tical scale: 40nm. B) Flow=1.0 sccm, Ra: 3.89nm, vertical scale: 40nm. C) Flow=1.5s ccm, Ra: 1.48nm, vertical scale: 20nm. 148

PAGE 149

Figure 5-7. Optimized MBE GaN growth conditions on UOE GaN. 149

PAGE 150

A) B) Figure 5-8. In-house grown MOCVD GaN showing growth terraces. A) 1mX1m area. B) 5mX5m area. 150

PAGE 151

A) B) Figure 5-9. Optimized MBE GaN conditions on in-house grown MOCVD GaN. A) 1mX1m area. B) 5mX5m area. 151

PAGE 152

A) B) C) Figure 5-10. SEM images of initial results of MBE GaN on MgCaO, indicating a rough film. Also included is the EDS spectrum of th e film, showing the calcium and oxygen from the underlying layer. The magnesium peak is barely visible just to the right of the low energy gallium peak. A) Low magnifi cation SEM. B) Higher magnification SEM. C) EDS spectrum. 152

PAGE 153

A) B) Figure 5-11. RHEED images of first MBE GaN on MgCaO. A) Pattern after 5 minutes of GaN growth, taking on a slight poly-crystalline pattern. B) Patte rn at the end of the one hour growth showing single crystalline 3D growth. 153

PAGE 154

A) B) C) Figure 5-12. First MBE GaN on MgCaO growth. A)AFM image showing starting MgCaO surface morphology with smooth terraces. B) MBE GaN growth on MgCaO, with Ra=6nm. C) Closer view of the GaN growth showing bumpy morphology. 154

PAGE 155

Figure 5-13. XRD scan of the first MBE GaN on MgCaO which is shown in Figure 5-12. The GaN peak (alpha and beta) are shown with no additional peaks. This indicates that the oxide stayed single phase, and is lattice matched to the GaN. 155

PAGE 156

A) B) C) Figure 5-14. RHEED images of GaN growth at 700C on MgCaO in MBE. A) Pattern after 10 minutes of GaN growth. B) After 1 hour growth. C) At the end of the two hour growth. 156

PAGE 157

A) (B C) (D Figure 5-15. AFM images showing MBE GaN growth on MgCaO and the different morphology produced from substrate temperatures at 700C versus 800C, vertical scales of 80nm and 30nm for images A and B, respectively. A) 700C substrate temperature, Ra=8nm, 1mX1m area. B) 700C, 5 mX5m area. C) 800C substrate temperature, Ra=3nm, 1mX1m area. D) 800C, 5mX5m area. 157

PAGE 158

A) (B C) (D Figure 5-16. AFM images showing MBE GaN growth on MgCaO with AlN exposure and the different morphology produced from substrate temperatures at 700C versus 800C. A) 700C substrate temperature, Ra=5nm, 1 mX1m area. B) 700C 5mX5m area. C) 800C substrate temperature, Ra=4nm, 1mX1m area. D) 800C, 5mX5m area. 158

PAGE 159

A) B) C) Figure 5-17. RHEED images showing MBE GaN growth on MgCaO after a short AlN exposure prior GaN growth. A) After 15 minutes. B) After 1 hour. C) End of growth of 2 hours. 159

PAGE 160

Table 5-1. Sample growth conditions and char acterization data of silicon doped MBE GaN on MgCaO. Sample Substrate Temp N Flow (sccm) Growth Time AlN Exposure 1 700 1.3 2 No 2 800 1.1 1 No 3 700 1.3 2 Yes 4 800 1.1 1 Yes Table 5-2. Sample HALL data from silicon doping. Sample Ra (nm) Rs (Ohms) Mobility (cm2/Vs) Carrier Density (cm-3) 1 8.32 2 3.45 4514K 6.15 1.12E+16 3 5.45 3.48K 44.86 1.99E+18 4 4.80 2.07K 38.27 3.94E+18 160

PAGE 161

A) (B C) (D Figure 5-18. AFM images of MBE GaN on bina ry oxides of MgO and CaO. A) MBE GaN on CaO (B), Ra=7.03nm vertical scale of 70nm. B) MgO, substrate for (A), Ra=1.23nm vertical scale of 15nm. C)MB E GaN on CaO (D), Ra=8.54nm vertical scale of 70nm. D) CaO substrate for (C ), Ra=0.815nm vertical scale is 15nm. 161

PAGE 162

A) B) Figure 5-19. XRD scans of the binary oxides used in the growths of Figure 5-18, the scans are also in Log scale to emphasize the oxide peaks in relation to the large GaN signal since the oxides are only a few hundred thick. A) CaO substrate showing the (222) CaO peak, the GaN peak, along with small peaks from the sapphire substrate. B) MgO substrate shown the (222) MgO peak, the GaN peak, and additional stray peaks from the substrate. 162

PAGE 163

Figure 5-20. X-ray diffraction sp ectrum of the lattice-matched MgCaO film used for the GaN films grown to larger thicknesses. 163

PAGE 164

A) B) Figure 5-21. RHEED images of MgCaO at diff erent thicknesses. A) Approximately 200nm thick film, showing signs of surface charging. B) Approximately 100nm, showing rotation of the film. 164

PAGE 165

A) B) Figure 5-22. AFM images of MgCaO different thicknesses. A) Approximately 200nm. B) Approximately 100nm. 165

PAGE 166

A) B) Figure 5-23. 3D AFM images showing the morphological differences between the two growths, grown on the same block, with differing star ting oxide substrate thicknesses. A) GaN with starting substrate oxide of 100nm (Fi gure 5-22b). B) GaN with substrate oxide of 200nm (Figure 5-22a). 166

PAGE 167

Figure 5-24. Optical profilometer scan of th e clipped area from the sample in Figure 5-23, grown for 500minutes. The Yprofile indicated a step he ight of around 400nm, which corresponds to the cross-section SEM images. 167

PAGE 168

A) B) Figure 5-25. Field-Emisson SEM images showin g a cross-section of the GaN on MgCaO sample grown for 500 minutes. The thicknesses of the oxide and GaN are approximately 100nm and 400nm, respectively. A) 20,000X secondary electron image of the sample, showing the sapphire (bottom) then the MOCVD GaN, followed by the MgCaO, and finally the MBE GaN on the top. B) 150,000X image of the sample showing a more detailed view of the MBE GaN growth. 168

PAGE 169

A) B) Figure 5-26. AFM images showing the rough morphology of the 20 hour MBE GaN growth on MgCaO. A) 252m area. B) 12m area. 169

PAGE 170

A) B) C) Figure 5-27. AFM images showing the progres sion of growths from the starting MOCVD GaN to the final MBE GaN. A) MOCVD GaN, showing nice terraced growth. B) MgCaO growth showing replication of the MOCVD GaN substrate. C) MBE GaN growth on the MgCaO showing a rough morphology. 170

PAGE 171

A) B) C) Figure 5-28. FE-SEM images of the 20 hour MB E GaN growth shown in Figure 5-26. Crosssection images depict a columnar structur e of the MBE GaN. A) Image showing the columnar structure of MBE GaN on the MgCaO on MOCVD GaN on Sapphire (bottom). B) Higher magnification. C) Showing delamination of both MBE GaN and oxide from the oxide and MOCVD GaN, respectively. 171

PAGE 172

606570758085 0.1 1 10 100 1000 10000 100000 1000000 MgCaO (222)CountsAngle (2 Theta) Figure 5-29. XRD scan of thick MBE GaN on MgCaO, from Figure 5-28, showing an oxide with slightly smaller lattice than the GaN. There is no sign of phase separation of the oxide, again confirming earlier measurements. 172

PAGE 173

A) (B C) (D E) (F Figure 5-30. Secondary electron and cathodolumin esence (CL) imagines of the thick 20 hour MBE GaN growth with a comparison with standard in-house grown MOCVD GaN. A) Secondary electron image of the MBE GaN surface at 1,000X. B) Secondary electron image of the MBE GaN surface at 5,000X. C) CL image of image A. D) CL image of image B. E) CL image of MOCVD GaN 1,000X. F) CL image of MOCVD GaN 5,000X showing dark areas corres ponding to threading dislocations. 173

PAGE 174

A) B) Figure 5-31. Photoluminesence (PL) spectra of GaN samples from a standard MOCVD in-house run, and the thick 1.5m MBE sample. A) PL spectra from MOCVD GAN showing band-edge at 360nm, and some blue and ye llow emission from defects. B) PL spectra from thick MBE GaN on MgCaO s howing band-edge, no blue emission, but increased yellow emission from nitrogen vacancies. 174

PAGE 175

Binding Energy (eV) N(E) Min: 207Max: 118687 10009008007006005004003002001000 O 2s Mg 2p Mg 2s O 1s Ca 2s C 1s Ca 2p3 Mg KVV O KVV Ca LVV Figure 5-32. XPS spectrum of MgCaO before the 700C vacuum anneal to examine hydroxide decomposition. 175

PAGE 176

A) B) Figure 5-33. Magnesium 2s peak before and after heating. A) Before heating, the broadness of the peak to higher energies signifies a hydroxide layer. B) After heating, only one peak indicating no magnesium hydroxide. 176

PAGE 177

A) B) Figure 5-34. Oxygen 1s for the pre-heated and po st-heated sample. The peak changes are due to the surface hydroxide layers being decomposed during the vacuum anneal. A) Before annealing. B) After annealing. 177

PAGE 178

A) B) Figure 5-35. Calcium 2p before and after heating. The far left peak at higher energies corresponds to calcium hydroxide. A) Be fore heating, showing a strong hydroxide peak. B) Post heating, showing a la rge reduction in the hydroxide peak. 178

PAGE 179

Binding Energy (eV) N(E) Min: 187Max: 91427 10009008007006005004003002001000 Ca LVV O KVV Ca 3s Mg KVV O 2s Mg 2p Ca 2s Ca 2p Mg 2s O 1s Figure 5-36. XPS full spectrum scan of the Mg CaO sample after heating to 700C in the MBE chamber. The carbon peak seen be fore heating was largely reduced. 179

PAGE 180

CHAPTER 6 MOCVD GAN GROWTH MOCVD Overview In todays marketplace of semiconductor materi als, keeping up with demand is crucial. Once a new material is developed, the supply must keep up with the demand. This is an essential component of business which in a round about way dr ives the engine of science. The need for high throughput manufacturing of devices requires that the raw materials are plentiful. While MBE growth allows for very precise and controll ed materials, the growth rate and throughput is low. In comparison, metal-organic chemical va por deposition or MOCVD, is a technique which can have a substantially higher material grow th rate than MBE which results in faster manufacturing. While the notion of higher throughput is well received, the MOCVD process is not without issues. These issues in clude a drastically different gr owth environment and different requirements for substrate compatibility. Compared to the ultra-high vacuum environment of the MBE chamber, the MOCVD reactor operates at pressures in the millitorr range. Also, the extremely pure elemental sources in the MBE ar e replaced with metal-organic sources and a variety of other gases. Even with these vast differences, the capability to improve productivity drives the innovation and research of MOCVD semiconductor growths. The challenges of the MOCVD tool will be discussed individually. Starting with the source materials for growth of gallium nitride, wh ich is the focus of this work, there are many factors involved with the incorporation of the el ements. The gallium source is from either the metal-organics of TMG (Trimethylgallium) or TEG (Triethylgallium). The choice of source for GaN growth depends on the structure that is to be created. TEG is mainly used for HEMT structures and other devices that need rapid changes to the struct ure. TMG is the standard for 180

PAGE 181

most all other GaN material. This is the choi ce source since TMG is less expensive than TEG and for the unintentional benefit of additional carbon incorporation for more insulating properties. TMG has inherently more carbon by-products than TEG from the methyl group compared to the ethyl group. Carbon has been shown to help mitigate the doping effects from residual oxygen and other contamin ates in the growth chamber.35 The carbon seems to sit in vacancies and allows for less conductive GaN material. Therefore all grows for this research utilizes TMG since the main focus is growing GaN on this new ternary oxide substrate and less concerned with the carbon incorporation. The reaction of the TMG on the surface involve s the absorption of the molecule onto the surface of the substrate. The substrate must then supply the molecule with enough energy to separate the gallium at om from the methyl group. This thermal energy is adequate when the substrate temperature is above 350K.36 At this point the methyl group must desorb from the surface to avoid scattering and allow for other metal-organic molecules to reach the surface. This removal of by-products is assisted by the high flow rates of th e other gasses across the sample surface. Since the TMG molecule is quite large, a carrier gas is needed to transport the molecule from the source to the substrate. H ydrogen is normally used as a carrier gas which percolates through the TMG bubbler in which th e TMG is contained. The hydrogen carrier gas picks up TMG molecules from the bubbler and flows the material to the reactor chamber. Since the ability of TMG to be captured by the fl ow of hydrogen gas is largely controlled by temperature, the bubbler is submerged in a water bath in which the temper ature is regulated to within 0.1C of the normal set point of zero degree s. Even with the low temperatures, there is also a carrier gas dilution to lower the concentration of TMG in the gas flow. 181

PAGE 182

The other necessary element, nitrogen, is supplied by gaseous ammonia. The ammonia supply is the vapor from liquid ammonia and regulated by mass flow controllers. Unlike the TMG which can react on the substrate when the temperature only warm, ammonia requires higher thermal energy to crack the molecule and re lease the nitrogen atom. Temperatures greater than 1000C are required to crack half the ammonia molecules, and is therefore quite inefficient, which then limits some of the substrate use discussed in chapter two.37 The low temperature nucleation step involving the am monia is attributed to a reac tion between the ammonia and the TMG in which hydrogen atoms are successively st ripped from the absorbed metal-organic ammonia compound. This reaction is very involved and has many pathways to the release of the nitrogen atom. The ammonia also adds additional flow into the gas stream. With the addition of the ammonia, the issues encountered with fluidyna mics come into play. Having smooth laminar flow of gasses through the reacto r and across the sample surface is crucial in the ability to grow epitaxial films. When the gas flow become s turbulent, recirculation can occur creating undesirable results for the growth. Therefore, a high flow of gasses are needed to maintain the laminar flows, which adds additional changes si nce each gas has different heat absorption. This results in substrate temperature changes every time a gas is changed or flows adjusted. Having the higher substrate temperature for the ammonia decompositi on, is not entire a bad condition. Increased temperatures result in higher surface mobility for the impinging atoms. This increased mobility allows for the atoms to migrate along the surface to find a favorable site for incorporation. The substrate temperature is monitored by both a thermocouple a few millimeters from the underside of the sample and also by a pyromet rer. This emissivity compensated pyrometer 182

PAGE 183

uses an infra-red light source of 930nm and 10nm bandwidth. Refl ectance off the sample surface can indicated the degree of roughness of the surface along with layer thickness because of interference with the light as the layer grows. In addition, the same detector can measure the black-body radiation off the sample and give an accurate measure of the surface temperature. This method is superior to simple measuremen ts by the thermo-couple as the thermo-couple is mounted below the heater, and based on the gas fl ows in the tool, the sample surface can have a completely different temperature th an what the thermo-couple measures. To assist in uniform growth across the enti re sample the substrate holder is rotated. Normally a rotation of 1500 revolutions per minute is used for this tool. The rotation is also mandatory in providing laminar flow across the wa fer for the gasses. The gasses are supplied to the sample by a showerhead type in jector that is capable of adju sting the flows for the inner and outer portions of the wafer. Again, by adjusting flows creates temperature variations across the wafer from the gas heat capaci ties, which can alter growth. Therefore, it can be seen that every little ch ange in any of the cond itions of temperatures, flows, or concentrations can change a growth qui te drastically. This is the reason it takes an immense amount of samples to perfect a recipe for good GaN growth in a MOCVD reactor. Even with these challenges, the ultimate reward for the many years of trial and error can result in a recipe that can allow for the material to grow in the manner desired. MOCVD GaN Substrates All oxides were grown on GaN substrates. The two substrates employed during the experiments were either MO CVD GaN on sapphire by Uniroyal or in-house grown MOCVD GaN on sapphire. Initially all grow ths occurred on the Uniroyal s ubstrates, since this material was readily available. This material was useful for initial calibrations, but ultimately useless for further device fabrication. The dow nfall of this material is that most of the wafers were 183

PAGE 184

engineering wafers from Uniroyal and therefore had imperfections. Most all the wafers had pits in the GaN surface, in which some extended all the way to the sapphire substrate, which was shown in Figure 4-14. Therefore this surface did not allow for large areas of good material for device fabrication, and therefore was ultim ately only good for oxide growth morphology characterization. Most all growths which were to be used for further processing and for templates for further GaN top growth, employed substrates manufactured our in-house MOCVD. The tool is a Veeco/Emcore P75 Turbodisk vertical reactor. Th is system is setup for growth on a single two inch wafer. The tool uses ammonia as the nitrogen source with molecular nitrogen as the carrier gas. For the metalorganics sources, the instru ment has triethylgallium (TEG), trimethylgallium (TMG), trimethylindium (TMI), and trimethylaluminum (TMA.) The variety of metalorganic sources allows for any group III nitride material to be grown. In addition to the metalorganics, the system uses silane (SiH4) for n-type doing, and bis-cyclop entadienyl magnesium (Cp2Mg) for a p-type dopant. Since this system was intended for commercial use, most all controls are automated through a computer interface. This automation makes it quite nice when many changes take place at the same time as for growth regimes lik e recovery in GaN where temperature, pressure, and gas flows are all changing at the same time. Unfortunately, being a commercial tool, the down side is that slight changes in recipes caus e large changes in growths. With this in mind, even getting the recipe for the initial GaN growth on sapphire took many years to fine tune to get the high quality and smooth surfaces that can be grown today. All GaN growths take place on single crystal sa pphire (Al2O3) substrates oriented in the (0001) direction which is normal to the wafer surf ace. The sapphire is epi-polished one the front 184

PAGE 185

and fine ground on the reverse. The use of sa pphire as the starting su bstrate was a decision based on a wide knowledge base. Even with the undesirable propertie s of sapphire for GaN growth, it leads the field in subs trates since it is inexpensive co mpared to other substrates and extremely chemically and thermally stable. Ma ny methods have been devised to get around the issue of lattice mismatch between the sapphire and GaN that have lead to surprisingly good quality material. Still, the quality nor speed of MOCVD GaN can match that of HVPE GaN in which high quality material of hundred of microns have been fabricated. Most starting sapphire was purchased through Crystal Systems Inc. The starting substrates of (0001) sapphire are of extrem ely good quality in respect to surface morphology. AFM characterization of the sa pphire surface Figure 6-1 reveal s uniformly spaced terraces indicating the miscut of plus or minus one degree from the manufactur ers dicing and polishing procedure. These terraces are very good to have since they will be pref erential nucleation sites for the GaN growth. The sapphire wafers are then loaded onto grap hite susceptors with a proprietary coating which prevents degradation under the MOCVD atmosphere. The sample is loaded via a loadlock chamber which is flushed with nitrogen to remove any oxygen in the chamber, since oxygen will cause unwanted doping of the film. The ho lders are then transferred into the growth chamber onto a spindle attached to a computer controlled smart motor. After loading the susceptor, and loading a recipe into the computer, the actual growth can commence Since the MOCVD environment entails high temperatures and relatively high pressures compared to other growth methods, surface kineti cs and fluid dynamics play a large role in growth quality. Since gas flows are in the multip le liters per minute regime, it can be assumed that the material will encounter some type of boundary layer at the surface of the wafer. This is 185

PAGE 186

the reason that most MOCVD processes have a diffusion limited rate limiting step. Since the arena of fluid dynamics can be a bit complicated and lengthy, the equatio ns governing the flow of the gasses in the MOCVD reactor are presente d in the Appendix. To minimize the effect of this boundary layer at the surface of the wafer, the wafer is rotated at 1500 rpm in this particular chamber. This rotation allows for source ma terial to diffuse through the boundary layer and react on the surface of the sample rather than react before the su rface and cause particulates to settle on the surface. The rotati on also allows for good uniformity in all directions radial from center. Growth of GaN on sapphire follows a few ba sic principles: growing a low temperature nucleation layer; annealing of that low temperat ure nucleation layer which turn into small nuclei or islands of high quality GaN; coalescence of the islands into a smooth continuous layer; vertical growth from this continuous layer to the desired thickness. The steps and conditions involved in this process can be visualized through Figure 6-2 and the following explanation. Area 1 is the high temperature hydrogen clean of the sapphire up to 1100C. Area 2 is the low temperature nucleation step when approximately 50nm of GaN is deposited. Area 3 is the annealing of this low temperature nucleation layer, at which time the GaN decomposes and reforms at various locations, producing islands or nuc lei. Area 4 is the 3D or coalescence stage at which time there is a lower TMG flow than the main growth layer, and the temperature is hotter producing lateral grow th to coalesce the islands. Finally area 5 is the main growth layer at which point the vertical growth in creases. To note is the oscill ations seen by the reflectivity. These oscillations show minimu m and maximum to the reflectivity from the thickening of the GaN layer. These layers are related to the wa velength of the IR light used to gather the reflectivity data and are used to sh ow layer smoothness and growth rate. 186

PAGE 187

Back-Side Sample Coating In ultra-high vacuum environments, h eat transfer is limited to phonon conduction and infra-red radiation absorption. In conventiona l MBE operations, the samp le of interest is mounted to a metal block, molybdenum, by the us e of indium. The indium, which is molten when mounting samples, acts as a conduction pathway for phonons between the molybdenum block and the sample. Since most samples have a rough polish on the backside, there are very few contact points between the sample and moly bdenum block without the indium. Therefore the sample either needs to be absorbent to infra-re d radiation to increase te mperature, or an infrared transparent substrate, like sapphire, needs to have materials that absorb the infra-red deposited on the rough back-side. Unfortunately, samples grown for use as substr ates in the MOCVD are restricted from the use of indium. Indium would contaminate the growth chamber and sample holders during the high temperature and high pressure s of growth. Therefore, a ne w mounting and coating system was to be developed, one th at did not involve indium. Requirements of a coating are the ability to absorb sufficient infrared radiation to heat the substrate, to be compatible with the highvacuum and high-temperature environment of the MBE, and to be an effective coat ing that does not delaminate or wa rp the substrate. In the ultrahigh vacuum of the MBE, vapor pressure of any coating needs to be considered. With temperatures reaching 800C and above at times, the coating needs to have a sufficiently low vapor pressure to not a ffect the heating of the substrate. Next the coating must not warp the substrate during heating or cool ing and be robust enough to adhe re to the substrate without delaminating or peeling. Th erefore, thermal expansion di fferences between the sapphire substrate and the metal back-sid e coating would need to be si milar enough to avoid problems. 187

PAGE 188

It was found that the MOCVD environment pl aced a whole other set of conditions on the coating. With temperatures reaching 1200C and a very harsh atmosphere of hydrogen and ammonia, the coating would have to withstand another round of requirements. With reactive ammonia at high temperatures, th e reacting of the metals with this gas would have to be considered. When ammonia reacts on a hot substrate it not only supplie s a lone nitrogen atom which wants to react with most all elements, but it releases three atomic hydrogen atoms that can cause havoc with the sample and coating. With the strict requirements of this coati ng, only a few materials co uld be used with the properties needed and the availabi lity of the material. First wa s a series of experiments with multiple refractory metals including tungsten, titanium, molybdenum, and tantalum. Other researchers have used a combination of titanium and tungsten for backside coatings in MBE. These researchers found out that slow temperat ure ramps were needed to prevent substrate cracking due to the ther mal stresses introduced from the coa ting. During current experiments it was found that an alternati ng layered structure of titanium and tungsten (20nm/40nm, respectively) worked fine for the coating in th e MBE since it was calculated based on thermal expansion data on the sapphire and the metals. The first coating experiment involved alternating layers of titanium and tungsten at thicknesses of 20nm and 40nm, re spectively. All metal coati ngs were deposited in a argon sputter deposition tool. The metal thicknesses were based off thermal expansion data from the metals and the sapphire substrate to minimize the effect of the coating. In addition to these alternating layers, the total thickness was limited to 180nm, to minimize induced stress. During oxide growth, it was noticed that during the 700C cleaning stage that the sample did not has as much blackbody radiation as a normal block as seen in Figure 6-3. Therefore it was inferred that 188

PAGE 189

the coating thickness was not sufficient for adequa te IR absorption. After the sample was grown, it was noticed through the AFM scan (Figure 64), that the surface was not very terraced, indicating a low growth temperature, even t hough the roughness was low. This supported the notion that the coating was an inefficient absorber of IR radiation from the sample holder. After the sample was used during an MOCVD growth run, the backside coating was examined to determine if any changes had occurre d. Visually, the outside edge around the entire quarter wafer had a different color than the middle. This indicated that some etching or nitridation had occurred. SEM analysis showed a different morphology of the edge versus the middle of the sample, as seen in Figure 6-5. EDS did not indicate any differences in the elements present. It was then decided to try a coating with pure tungsten at a thickness of 500nm. The thicker coating should allow more absorption of heat for growth. Visually, at 700C in the MBE chamber, the sample emitted more blackbody radiation, and therefore was determined to be at a higher temperature (Figure 6-6). The final growth seemed to be smoother as the terraces were more pronounced as seen in Figure 6-7. Samples of the other refractory metals were tried, including Molybde num, and Tantalum. The oxide growths were all identi cal, indicating that the thicker 500nm coatings are sufficient for heating the sample. After MOCVD exposure, the coatings all had some form of discoloration around the edge of the wafer, shown in Figure 68. Therefore all the samples were scanned in the SEM with EDS spectra to analyze the discolor ations. The results were the same as for the tungsten sample, in which the EDS simply shows the metal and the sapphire elements, while the images show a roughened surface around the edges. It is suspected that the edges of the samples are exposed to some ammonia during the MOCVD runs, and reacts on the surface. Through 189

PAGE 190

literature, most of the metals do react with am monia to form volatile by -products or nitrides.38, 39, 40SILICON BACKSIDE COATED SAMPLES After the successful trials of the refractory metals for backside coatings, the next trials involved in-stock sapphire wafers with silicon coatings. To avoid the process of growing MOCVD GaN on sapphire then sputtering metal on the back, it was logical to try growing GaN on sapphire which already had silicon on the ba ck from the vendor. There was many wafers from Union Carbide that had CVD silicon on the b ack of polished sapphire wafers. The coating was 10m thick, which would allow for plenty of absorption of the IR radiation for MBE oxide growth. Also beneficial is the fact that the thermal expansions of silicon and sapphire are quite similar, since silicon has a TEC of around 4.68 while single crystal sapphire is around 4.5. Unfortunately, after the first growth attempt in the MOCVD reactor, it was found that during the manufacturers CVD process, some silicon had been deposited on the front of the wafer as seen in Figure 6-9 and confirmed with EDS in Figure 6-10. Since silicon oxide is used for a masking material to prevent GaN MOCVD growth, it was noti ced that the growth with in a half inch from the outside of the two inch wafe r was bad seen in Figure 6-11. Through extensive literature review on etchan ts and chemical compatibilities, a method of removing the silicon from the epi-side of th e wafer was constructed. While the CVD silicon on the backside of the wafer is desirable, any silicon on the front of the wafer will cause unwanted effects during MOCVD growth of GaN. This silicon on the front of the wafer must be removed without damaging the epi-polished sapphire Therefore, to remove the silicon on the front of the wafer, the backside coating must be protected. Initially, photoresist was spun onto the back of the wafer, as Shipley 1045 positive p hotoresist is resistant to degradation in acidic chemistries. Therefore, the etchant used to re move the front side sili con will not affect the 190

PAGE 191

backside coating. After PR coati ng the wafer was immersed in a bath of standard silicon etchant of nitric acid, ammonium fluor ide, hydrofluoric acid, and wa ter. The solution ratio is 120NHO3:60H20:6NH4F:1HF. This solution is reported to have an etch rate 100-200nm/min.41 The nitric acid is a strong oxidizer, and th erefore will oxidize any si licon exposed while the ammonium fluoride and hydrofluoric acid works to remove the silicon oxide that is formed. This solution is not reported to have a detrimental e ffect on sapphire. Even if there would be an etching effect on the sapphire, re ports have indicated that pitti ng of the sapphire before MOCVD GaN growth has an advantageous effect of a higher quality growth compared to untreated sapphire.42 The wafer was submerged in the acid soluti on and sonicated while not icing any changes. Instantly the solution started a ttaching the photoresist protecting the back of the wafer. The solution started turning yellow as the silicon began to etch. At the five minute mark, the silicon along the edge of the wafer was becoming noticeably etched, unfortunately, so was the silicon on the back of the wafer (Figure 6-12). At this point there was no saving the wafer, so the acid bath was continued to see how long it would take to etch off all the silicon, which took 12 minutes (Figure 6-13). Once the wafer has been treated in the aci d solution while sonicating the bath for 12 minutes, it is rinsed in DI water then rinsed in a solution of IPA, methanol, and then DI water to remove any residual contamination from the silic on etching. As seen in the AFM images, the surface still contained large amounts of etch products that were not rinsed from the surface of the sapphire, in Figure 6-14. The second trial to remove the silicon ring on the front of the sapphire was more of a success. To protect the backside silicon and the photoresist, once the ba ckside of the sapphire 191

PAGE 192

wafer was spun coated with the photoresist, it was pressed onto the polished side of a silicon wafer. This sandwich was then hard backed at 150C for 10 minutes to bo nd the wafers together. At this point, a cotton swab was used to scr ub the edge with the acid solution. This was not resulting in a noticeable removal, so the wafer sandwich was then submerged in the bath. The solution began turning yellow and the sacrificia l silicon wafer began to etch along with the silicon ring on the sapphire. Afte r 5 minutes the ring on the front began to disappear. Using as an indicator, when rinsed with DI water, if a ny water would bead up at th e edge where the silicon was, then the sample was returned to the acid solution until this water beading did not occur. After the acid etch, the photores ist bonding the wafers together was removed by an acetone bath while sonicating. Then an acetone rinse, followed by DI water, followed by a final rinse in IPA, METH, and DI water. This method seemed to remove the front silicon well, but still resulted in some etching of the backing silicon since both the PR and sacrificial silicon wafer were being attacked. Finally, a scrap sapphire wafer was mounted to the back of the silicon coated wafer with photoresist and heated on a hotplat e to bond the wafers together. Then the ring of silicon on the front of the wafer was scrubbed with a cotton swap which was soaked in the acid solution for an additional eight minutes. This removed most of the silicon, so a final complete submersion in the acid resulted in a front surface that seemed to be free of silicon. The cleaned wafer was analyzed with the SEM and the images are show n in Figure 6-15, which show a slight ring where the silicon was. EDS (Figure 6-16) did not indicate any residual silic on of the surface, even though there is a visible marking. The residual ring on the outside of the wafer still cause d poor GaN growth within a quarter inch from the edge of the wafer. Fort unately, the inner part of the wafer grew normally 192

PAGE 193

with terraced growth and roughness of only a few angstroms as seem in Figure 6-17. Therefore, the intermediate step of depositing a refractor y metal on the back of wafers between MOCVD GaN growth on sapphire, and the MBE oxide gr owth, can be eliminated, which can reduce surface contamination from the sputter tool. Annealing of MgCaO in Ammonia To verify stability of the MgCaO in the MOCVD environment, a sample of oxide was annealed in ammonia. Ammonia is a very harsh chemical which attacks many materials, especially at high temperatures. Latticematched MgCaO was grown on n-type MOCVD GaN with a backside coating of titanium and tungs ten with alternating la yers of 20nm and 40nm respectively to a total thickness of 180nm. The oxide was grown to a thickness of 100nm, and characterized with AFM to allow for a baseline before annealing (Figure 6-18). The sample was then annealed in the MOCVD under ammonia for 25 minutes at 1090C. This temperature is the normal temperature used to clean the starting sa pphire wafers, and is hotter than any normal growth which occurs around 1050C. The long expos ure time of 25 minutes was to see changes on a long time basis and extrapolate data for sh orter exposures. The oxide surface would be exposed to the ammonia atmosphere for a only a short time while the low temperature nucleation GaN layer is annealed up to the roughening temperature of 1090C Once this temperature is reached, the TMG flow continue s and the oxide surface would be protected at that point. The results of the anneal are shown in Fi gure 6-19 which indicates that the surface did roughen under the reducing effects of the ammonia. The surface roughness started near a half of a nanometer roughness and after etching was around 20nm roughness. This roughening is depicted in a 3D fashion in Figure 6-20. The RealTemp data is shown in Figure 6-21, which indicates the temperature and reflectivity of the sample surface. As can be seen, the temperature was increased in the normal fashion for a st andard growth up to 530C, at which point the 193

PAGE 194

ammonia is introduced. After the ammonia begins to flow the temperature is ramped up to 1090C, and allowed to rest at this temperature for the 25 minute anneal. After the 25 minute anneal, the sample is cooled down to room te mperature before removing from the chamber. The results of the reflectivity are quite interesting. In the beginning, the sample shows a slight dip in the reflectivity, which is normal to this system during the beginning of sample rotation and gas flow. Then the re flectivity increases as the temper ature begins to rise, and this is attributed to the su rface hydroxide layer beginning to be thermally removed. At the point of high temperature annealing, the refl ectivity drops, as this is attr ibuted to the beginning of the surface roughening effects of the heat and ammonia. The reflectivity then begins to rise, as this was initially thought to be a resu lt of surface reconstruc tion as seen in sapphire high temperature annealing.43 Yet, after the post AFM analysis, the su rface was actually getting rougher, but from the fact that the surface features were getting larger as indicated from the Z-range of the image. This shows that the surface at high magnificat ions was becoming smoother, but the annealing caused these watermelon patch features on the surface that individually are very smooth, but macroscopically very rough. Therefore, as the rise in reflectivity was concerned, it was from these larger flat areas that caused an increase in signal. To further characterize the MgCaO afte r high temperature ammonia annealing, the sample was scanned by XRD before and after an nealing. The spectrums are shown in Figure 622. Shown in the spectrums are the GaN peak and some substrate defect peaks. The interesting to note is the slight peak co rresponding to MgO which pops up af ter annealing. The peak at 78 degrees is a minute amount of MgO which has formed, perhaps from excess magnesium in the sample after MBE growth. As a reminder that the scales of the XRD spectra are in Log scale so the peak is just slightly higher than the backgrou nd signal. To note is that there was no increase 194

PAGE 195

in a peak associated with CaO, and therefore in dicates that there was not excess calcium in the sample. There is also a slight broadening of the GaN peak to lo wered angles, indicating that the ternary oxide did shift to sli ghtly richer calcium concentra tion, supporting the MgO formation. EDS analysis was performed on the sample to verify if any of the backside coating of titanium and tungsten had been etched and deposit ed on the front surface. The results are shown in Figure 6-23. As can be seen, no additional elem ents were found from the back of the sample. The analysis did show the GaN substrate since the interaction volume of the electron beam is deeper than the thickness of the oxide surface. In addition to the GaN, the calcium and oxygen peaks are present, while the magnesium peak ove rlaps the gallium L-line and is therefore not visible. Even with the surface roughened to around 20nm GaN growth should be viable on this surface. It has been showed with sapphire substr ate, that even samples up to tens of nanometers roughness, the grown GaN quality is the same as when grown on atomically flat substrates.44 This shows the robustness of GaN growth by MOCVD, and that the growth is very forgiving of the starting surface roughness. A dditionally, it was found that sapphire itself will fo rm etch pits after annealing in ammonia for extended periods of time. Shown in Figure 6-24, etch pits are seen in the surface of sapphire, which was found during a backside metal annealing run. This shows that ammonia has quite the ability to etch many surfaces, ev en those as stable as sapphire. MOCVD GaN on MgO The first trial, was to use straight MgO ar e the substrate. The MgO was grown in the MBE to a thickness of approximately 150nm. The oxide had a roughness of approximately 0.8nm, but without a terraced morphology. Growth in the MOCVD began with a 600C clean step since the sample was freshly grown in th e MBE, and no surface reconstruction is desired. The rest of growth follows the standard re cipe used to grow GaN on sapphire. The low 195

PAGE 196

temperature nucleation was performe d at 530C, and the reflectivity began to rise. As the sample temperature was raised to the 1090C anneal temperature, the reflectivity was steady until near 950C when the reflectivity began dropping. The standard sequence of 3D layer and recovery layers were used, and the refletivity signal seemed quite nice as there were large oscillations, seen in Figure 6-25 showing the RealTemp data. Post growth analysis showed a surface with regions of smooth growth and others that were quite pitted, Figure 6-26. The sample was then milled in a focused ion beam (FIB) and imaged with a TEM. The resulting image is shown in Figure 6-27. The TEM indicates a top GaN film with a large amount of stress and defect s. The one thing to notice is that the large threading dislocations in the substrate GaN do not propagate through the oxide into the top GaN film. Therefore the MgO did have some improve ment in blocking the threading dislocations, and the defects seen in the top film ar e most likely caused by improper nucleation. MOCVD GaN on Sc2O3Utilizing a similar recipe used for our standa rd GaN growth on a 2 sapphire wafer, GaN is grown on a Sc2O3 substrate. The Sc2O3 is grown in our Riber Oxide MBE. The starting substrate is MOCVD GaN on sapphire. The subs trate had the metallized backside-coating to avoid indium mounting. The indium would cr eate unwanted effects on the MOCVD tool. Once the sample was mounted and transferred to the grow th chamber, it was heated to 700C as an insitu thermal clean. This process is found to im prove the interface between the starting GaN and the grown oxide. After a 10minutes anneal at the 700C, the substrate is cooled to the growth temperature to await deposition. The Sc source is a Knudsen Cell with an alumina crucible, heated to 1190C. The oxygen is supplied by a Oxfo rd RF plasma head operating at 300W. The Sc2O3 is grown at 10 /min. The RHEED and XRD images c onfirm a crystalline film (Figure 6196

PAGE 197

28 and Figure 6-29, respectively), while AFM (Figure 6-30) shows an extrem ely smooth surface. The surface oxide replicates the starting substrate GaN with the terraces. The sample is then transferred to the MOCVD reactor to grow the subsequent GaN. The sample is inserted through the load lock onto a substrate holder that is designed specifically for a quarter of a two inch wafer (Figure 6-31). One thi ng to note about this partic ular holder is that it has a small chip where the corner of the wafer sits. These holders are uniquely designed to handle the extreme conditions of a MOCVD react or. These conditions involve temperatures over 1200C, in an atmosphere of hydrogen and amm onia. The holder itself is graphite with a proprietary coating to prevent de gradation. Once this coating is scraped or chipped, then the graphite itself is exposed to the environment th at can cause degradation. As seen in Figure 6-31 after growth runs, some amounts of powder are seen near the chi p, and under the grown sample. This powder was analyzed and is simply carbon from the degradation of the inside of the holder. Growth begins with a thermal anneal to clea n the surface and prepar e it for nucleation. This high temperature anneal normally reaches nearly 1200C when growing GaN on sapphire. The anneal occurs under a hydrogen ambient to facilitate cleaning a nd surface reconstruction (maybe, find paper and ref). For the Sc2O3 sample, since the starting surface is nearly atomically flat, a lower temperature anneal was performed. By lowering the temperature of the anneal to simply facilitate cleaning of the surf ace from atmospheric carbon and residue since being out of the MBE, it s hould allow for minimal impact to the starting surface. After the elevated temperature anneal, the sa mple temperature is lowered to 530C for the low temperature nucleation. This temperature allows for little migr ation of the source materials on the surface. This is beneficial for producing a uniform layer of GaN. The V\III ratio is very high in this part of the growth, again to furthe r the uniform nucleation la yer. By utilizing the 197

PAGE 198

reflectometer or RealTemp, the layer progress can be monitored. As the reflection of the laser off the sample increases, it is indicative of a smooth er surface. Letting th e reflectometer increase to around 11% during this nucleati on, it indicates a more or less complete GaN seed layer has been deposited. The normal starting reflection of th e sapphire wafer is around 7%. At this point, the TMG is stopped while still flowing ammonia. The sample is then ramped to 1090C to allow for surface migration of the thin GaN seed layer. The high temperature is needed to allow the GaN to actually decompose and move around the surface where it starts to coalesce into islands. These islands, depending on the am ount of TMG flown during the nuc leation layer, can result in heights of up to 1um. These islands are the real seeds for the subsequent growth. The reflectometer drops during this migration, as the GaN forms islands, the surface becomes rougher due to an almost 50% coverage of these islands on the surface. Once these islands are formed, the Ga source is reintroduced to the system. The V\III ratio is now lower, allowing a la rger amount of Ga to interact with the surface. Now at these high temperatures, the Ga has an extremely high mobility on the surface resulting in high migration and a growth regime which is conducive to lateral growth. This lateral growth is exactly what is needed to allow coalescence of th ese islands. Without this step the islands never fully coalesce and a very textured surface result s after further growth (show the rest of the growth of the aborted Sc2o3 on GaN). After the is lands begin to grow to gether, the reflectivity begins to increase as a complete surface is starting to form. It is at this time when we lower the V\II ratio even further for faster growth during the main growth layer. Lowering the V\II ratio facilitates a growth rate which is nearly 2 microns per hour. Since the islands have coalesced and the surface is fairly flat, this results in uniform vertical growth. 198

PAGE 199

During the first attempt at growth on Sc2O 3, it was found that the surface migration is vastly different from that of sapphire. After th e low temperature nucleation step and the ramp up to 1090C for island formation, the reflectivity of the sample plummeted from a value of 3.5% to almost zero within a time of 30 seconds, Figure 6-32. After aborti ng the run at that point and characterizing the sample, Figure 6-33, it was determ ined that the growth had gone just fine and the reason behind the rapid decrease in reflection was that the surface ki netics of the scandium oxide is largely different than the aluminum oxi de surface. The mobility on the Sc2O3 surface at 1090C allowed for incredibly high mi gration rates. This is indicat ive of a surface that has a high surface energy, allowing for the thin nucleation la yer to rapidly migrate to island formations. The sample was then reinserted into the MOCVD reactor and the growth was continued where it left off. To avoid decomposition of th e GaN islands, the sample was heated up to the 3D layer under ammonia. The reflectivity di d not show any improvement after a 30 minute growth. The surface was then characterized a nd found that the islands did coalesce to some degree as seen in Figure 6-34. Therefore MO CVD GaN growth on scandium oxide can be achieved. The next step is to use scandium oxi de as a capping layer for MgO or MgCaO which can then still be used as a lift-off layer. Contact Angle Measurements During MOCVD GaN growth, it was noticed th at after the low temperature nucleation layer was grown and then annealed, the rate of 3D roughening was vastly different among samples. It was first noticed that the rate of roughening of the Sc2O 3 was very rapid when substrate temperatures reached the standard te mperature of around 1090C. At first glance this was thought to be from possibly the sample bei ng ejected from the holder as the reflectance dropped extremely fast. Once the run was aborted, the sample was found to still be in the holder but had a very rough looking appearance. Afte r SEM analysis, it was found to have very 199

PAGE 200

uniformly spaced GaN islands. This result supports the notion that the thin layer of low temperature GaN had very high mobility on th e Sc2O3 surface, allowing the material to reorganize into these islands with very high mobility. To gain support for this theory, contact angle measurements were taken with a variety of samples to see how the surface energies changed with the different substrate materials. Contact angle measuremen ts were taken with a goniometer with digital capture abilities usi ng DI water with the sessile drop method. The images were then analyzed with ImageJ with a plug-in called Drop Analysis or DropSnake.45 To calculate the exact angle of th e water droplet, user inputted poi nts along the drop circumference was used to formulate a curve fitting the drop. The analysis was performed with both a 1l drop and a 5l for comparison. With the smallest drop let of water, the force from the weight of the drop is minimized and gives the most accurate measurement. Shown in Figure 6-35 a variety of samples were measured to compare the different substrates for GaN growth. As can be seen, the Sc2O3 has the largest contact angle and therefore the highest surface energy of all the subs trates. This result supports the experimental data and verifies that the GaN mobility on the Sc2O 3 substrate is much higher than it would be at the same temperature on a sapphire substrate. Al so to note is that the other oxides of MgO and MgCaO also have higher surface energies than th e sapphire. Since surface energy is related not only to the morphology of the surface, but also th e charge and bonds on at that surface, it is realistic to think that the surface energy is a reliable measure of GaN wetting during growth. Since the surface roughness of all the samples inve stigated are nearly identical, a few angstroms different, then the surface charge will be the main factor in the contact angle differences. These measurements give a fair approximation as to th e action that the GaN molecule will have on the 200

PAGE 201

substrate surface, since the GaN bond has a fractiona l ionic character of 0.51 and water is quite polar. The substrates by Union Carbide, with a back side silicon coating, were also analyzed. After the wafers had been bonded to a sacrificia l wafer by photoresist and etched in the silicon etch solution, the surface needed to be characterized for any sapphire damage. What was noticed was that the surface showed signs of chemical m echanical polishing with scratches visible. The contact angle was found to be similar to the MgCaO sample, with an angle of 45 degrees. As an interesting side note, a normal GaN growth on sapphire that was aborted post high temperature annealing just befo re introduction of TMG at 1090C, was also analyzed. In SEM and AFM, tall islands of almost 1um can be s een on the surface on a rougher substrate. The contact angle on this sample had the highest of all samples scanned, at 89 degrees. All measurements are tabulated in Table 6-1. Since the MgCaO and MgO samples are shown to be instable in water one can assume that there would be some etch ing occurring during the contact a ngle measurement. This was verified visually by a discoloration under th e water drop once the water had been removed, Figure 6-36. When scanning in AFM, the area under the water had a vastly different morphology than the rest of the surface (Figure 637). Also, since this etching was expected, the contact angle measurement was performed twice, th e first as soon as the water was released from the needle injector, and the second performed two mi nutes after the first. The angles of both the MgCaO and the MgO decreased indicating a lowering of the surface energy under the drop (Figure 6-38). Since the AFM showed a rougheni ng of the oxide after th e water contact, it can be concluded that the underlying GaN was being exposed which has a lower surface energy than the oxide. 201

PAGE 202

Overall, the effect of surface forces plays an important role in the nucleation and growth of GaN by MOCVD. Every new substrate whic h is grown upon in the MOCVD will have a different response to the conditi on of growth. Therefore, by supporting the data seen by the reflectivity during growths, the surface energies s eem to play a large role in the kinetics of growth. MOCVD GaN on Sc2o3 Capped MgO After determining that GaN growth will occur on scandium oxide, the next step is to apply that principle to an oxide that can be wet etched. Ultimately, the main focus of this work is to have an intermediate oxide layer that can be wet etched awa y, releasing the top GaN material. Therefore the next experiment was to grow magnesium oxide on GaN and cap it with scandium oxide. The capping is to protect that MgO surface from the effects of the room air humidity which forms a hydroxide layer. It was shown in the previ ous chapter that the hydroxide requires elevated temperatures to be removed from the surfa ce. Therefore by a capping method, a high temperature clean stage ca n be avoided and reduce surface degradation. The magnesium oxide sample is grown upon a quarter of a two inch wafer of GaN, on which 500nm of tungsten had bee deposited on the b ackside and mounted in the standard fashion in within a molybdenum holder with a quarter cutout and wired in with tantalum wire. The GaN substrate is heated to the standa rd 700C anneal for ten minutes to obtain a clean surface. The sample is then cooled to 400C, for MgO depos ition. The flux of magnesium was 5.5E-8 Torr BEP and the oxygen source was 1.0sccm at 350W. The MgO was grown for two hours before being capped with 20 nm of scandium oxide. Figure 6-39, and 6-40, show the XRD and AFM of this oxide sample. The oxide sample is then transferred into the MOCVD reactor at which time, the standard growth recipe commenced. As seen in the Real Temp data in Figure 6-41, the sample began with 202

PAGE 203

a low temperature nucleation which causes the reflec tivity to decrease, for the fact that the MBE oxide surface was already very re flective, so the GaN material ac tually causes some roughening. Next the reflectivity began to climb to a high value, which indicates a full layer of low temperature GaN. During the high temperature anneal, the reflectivity begins to decrease before the anneal temperature of 1090C is reached. Ag ain, this is the same trend noticed during the growth on scandium oxide. For that reason, once the sample reaches the high temperature, the TMG is turned back on with minimal flow to al low for the GaN islands to coalesce. After ten minutes had passed, the recovery layer begins and the V/III ratio is decreased to increase the vertical growth. As it can be seen, oscillations were strong and did not dampen very fast, indicating a good growth layer of GaN. After growth the sample had a large region of very smooth GaN. This was a great result, and was supported through the following analysis. The surface was first imaged with an optical microscope as shown in Figure 6-42, showi ng a smooth film, with some pits. A higher concentration of pits indicates that the 3D layer needed to be held a bit longer, or the main layer V/III layer needed to be increased. The surf ace was then analyzed with SEM and EDS which showed a smooth film with hexa gonal pits in Figure 6-43, and only gallium and nitrogen through EDS. The AFM of the surface show terraced gr owth is roughness around 1nm in Figure 6-44. To get a more intuitive idea of the film through the bulk, the sample was cross-sectioned and imaged with the field-emission SEM. The re sults show a continuous 4m thick GaN layer on top of the oxide (Figures 6-45 and 6-46). In addition, Figure 6-47 shows the TEM of this sample, indicating stress in the t op GaN layer near the oxide surface. This stress can be from the fact that the scandium oxide cap is not the sa me crystal system or la ttice spacing of the GaN. 203

PAGE 204

Next some measurements of the quality of the top GaN layer were performed. Photoluminescence or PL was performed on this sample at room temperature. The results are extremely exciting since there is a strong bandedge emission with almost no yellow emission commonly seen in MOCVD GaN growths. The yellow emission is normally attributed to nitrogen vacancies which are quite prevalent in MOCVD GaN on sapphire si nce they are stacked within the threading dislocations normally seen in high amounts. In Figure 6-48, the PL spectra is compared to the spectrum taken on the substr ate GaN, and the spectrum taken after the oxide was grown but before the top GaN growth. As can be seen, the starting substrate GaN had yellow emission, along with some blue emissi on thought to be from oxygen doping from a contaminated TMG bubbler. Therefore, since ther e is no blue emission and very minimal yellow emission seen in this sample, it supports the notio n of low defects in the material. Also included in Figure 6-49 is the sample compared to a st andard UOE GaN sample which shows a high level of defects. The next level of GaN quality verificati on is done through cathodoluminescense imaging. CL is a technique that uses the electron beam from an SEM to excite transitions in the sample that emit light. Seen in Figure 6-50, the sample is compared to a standard MOCVD GaN sample at various magnifications. Also shown is the se condary electron images to get an idea of any features and the results they produce. The standa rd sample shows a much modeled look of dark and light areas. The dark areas are non-recombin ative centers that are from the high density of defects along threading di slocations. Since threading disloca tions produce these dark areas, and non are really seen in the GaN on oxide sample, this adds to the support of an increased GaN film quality. To have equal comparisons, the co unts per second on the photomultiplier of the CL detector are equal for both samples. Also, the gray level of the bright areas was set to the same 204

PAGE 205

level of 170 out of the 255 gray levels. These images therefore support the PL data which show a reduction in defect emissions, and that the ma in component of CL signal is from band-edge emission. MOCVD GaN on MgCaO Capped with MBE GaN Theoretically, if the substrate has a smoot h terraced morphology with similar lattice spacing as the epi-layer, then nuc leation should not be required a nd growth will occur as if is homoepitaxy. The impinging atoms will sit prefer entially sit along the te rrace ledges and grow laterally. This growth should minimize dislocati on formation since there ar e no nuclei to interact with each other, which normally form defects at their boundaries. The main concern is that the conditions fo r normal high-temperature growth can possibly have an effect on the substrate. The atmosphe re is composed of ammonia gas which at high temperatures has an etching effect for most mate rials. By supplying a small amount of gallium to the high temperature ammonia during heat-up, then the ammonia will act to form gallium nitride on the surface which is continuously deco mposing by the high-temperature. Balancing the supply of gallium from TMG to form GaN, with the decomposition rate of GaN, can adequately protect the substrate surface until normal growth is started. Once the substrate has reached growth temperature, the flows of gasse s are adjusted to increase the rate of GaN deposition and begin normal high-temperature grow th. Without the slight TMG flow to the surface, the ammonia degrades the starting subs trate surface at elevat ed temperatures. To allow for a starting template, a growth of MgCaO was loaded into the Varian MBE system and capped with 20nm of MBE GaN. This sample was then loaded into the MOCVD for further growth. The growth bega n by heating the sample under ammonia. Once the sample reached 800C, TMG was introduced into the chamber at a very small amount. As the sample was increased to the growth temperature of 1050C, the V/III ratio was decreased from 205

PAGE 206

30,000 to the normal growth layer ratio of 3,000. The idea behind this process was that the sample already had nucleated GaN on the surface of the oxide, and therefore is set for continued growth without nucleation. The RealTemp data for the run is shown in Figure 6-51 which does s how oscillations of growth, but they began to dampen quite rapi dly. Therefore growth only continued for 30 minutes. The surface of the sample through SEM and AFM are shown in Figures 6-52 and 6-53, respectively, which show a somewhat smooth grow th that indicates not complete coalescence of islands or a porous growth. The cross-section imaging in Figure 6-54 shows large continuous grain-like growths that relate to the plane-view of the surface showing separa te nucleation sites. The PL intensity was extremely low for this sample. The cause is most likely from the uneven surface that is shown thr ough the cross-section images. When a sample has a very rough surface, the emitted light is difficult to gather and results in low intensity. The data was integrated for three times as long as a normal sample in addition to full open slits on the monochromator. The spectrum is shown in Figur e 6-55. To get more insight into GaN quality, CL images show some emission at the tips of the nuclei, but no mode ling seen in standard MOCVD GaN growths, Figure 6-56. 206

PAGE 207

A) B) C) Figure 6-1. AFM images showing the starting sapphire surface used for GaN MOCVD growths, having roughness of around 1 A) 300nm scan area. B) 1 micron scan area. C) 5 micron scan area. 207

PAGE 208

Figure 6-2. Standard data collected during MOC VD growth run on the TurboDisk reactor. The emissivity of the surface allows temperat ure measurement while the reflectivity off the sample surface indicates surface roughne ss and layer thickness. The numbers indicate the various la yers of the recipe. 208

PAGE 209

Figure 6-3. Image taken in MBE of 180nm Ti/W backside coating. Thermocouple temperature at 700C, yet, surface is cooler since coating was not very efficient at absorbing the IR radiation from the heater. 209

PAGE 210

Figure 6-4. AFM of sample grown with 180nm Ti/W backside coating. Lack of strong terraces indicate a growth temperature lower than the thermocouple reading of 500C. 210

PAGE 211

A) (B C) (D Figure 6-5. SEM images of the Ti/W coating. A) Middle of wafer at 500X magnification. B) Middle of wafer at 5,000X magnification. C) Edge of the wafer at 500X magnification. D) Edge of th e wafer at 5,000X magnification. 211

PAGE 212

Figure 6-6. Wafer image in MBE with 500nm tungs ten on the back of the wafer. The sample had more blackbody radiation, indicating a higher temperature than the thinner coating. The thermocouple reading of 700C is a more accurate measure of the sample temperature with the thicker coating. 212

PAGE 213

A) B) Figure 6-7. AFM images of the MgCaO grown on the wafer with 500nm of tungsten coating on the back. A) Five micron scan. B) One micron scan. 213

PAGE 214

A) (B C) (D Figure 6-8. Optical images of th e metal backside coatings for samples after MOCVD exposure. A) Molybdenum. B) Tungsten. C) Tantalum. D) Silicon 214

PAGE 215

Figure 6-9. Optical image show ing the silicon contamination around the edge of the sapphire wafers used for GaN growth. Figure 6-10. EDS line scan across the edge of the wafer shown in Figure 6-9, confirming the silicon contamination. 215

PAGE 216

Figure 6-11. Ring of bad growth after MOC VD GaN growth on the silicon wafer which had a ring of silicon on the front side. 216

PAGE 217

Figure 6-12. Some etching of the silicon backsi de coating during initia l attempts are removing the silicon on the front of the sapphire wafers from Union Carbide. Figure 6-13. Image showing complete backside co ating removal after 12 minutes in the silicon etchant. 217

PAGE 218

Figure 6-14. AFM image showing contamination on the epi-surface of the sapphire wafer from the silicon etching. 218

PAGE 219

A) (B Figure 6-15. SEM images showing the ring left after etching the silicon from the front of the wafer. A) BSE image. B) SE image. Figure 6-16. EDS spectrum of the ring area shown in Figure 6-15. 219

PAGE 220

A) B) Figure 6-17. AFM images of the MOCVD GaN growth on the sapphire wafer after silicon etching. A) Five micron scan area. B) One micron scan area showing sharp terraces, some threading dislocations, and having roughness of 1 220

PAGE 221

A) B) Figure 6-18. AFM images of 100nm MgCaO before being annealed in the MOCVD. A) One micron scan length, showing average roughness of half a nanometer. B) Five micron scan length. 221

PAGE 222

A) B) C) Figure 6-19. AFM images of the MgCaO after high temperature annealing in ammonia in the MOCVD. A) Five micron scan area. B) One micron scan area having roughness near 20nm. C) Cross-section scan show ing feature heights of nearly 30nm. 222

PAGE 223

Figure 6-20. 3D animation of the MgCaO after MOCVD anneal. 0500100015002000250030003500 0 2 4 Time (s)Reflectivity %400 500 600 700 800 900 1000 1100 Temperature (C) Figure 6-21. RealTemp data of the MgCaO anneal. 223

PAGE 224

A)606570758085 0.1 1 10 100 1000 10000 100000 1000000 Before AnnealCountsAngle (2 Theta) B)606570758085 0.1 1 10 100 1000 10000 100000 1000000 After AnnealCountsAngle (2 Theta) Figure 6-22. XRD scan of the MgCaO sample ann ealed in the MOCVD. A) Before annealing. B) After annealing, showi ng a slight increase in th e MgO peak, and a slight broadening around the GaN peak. 224

PAGE 225

Figure 6-23. EDS results of the front of the annealed MgCaO to verify no additional contamination. 225

PAGE 226

Figure 6-24. AFM line scan showing how the sa pphire will form pits under ammonia annealing at high temperatures. 01000200030004000500060007000 0 2 4 6 8 Time (s)Reflectivity %400 500 600 700 800 900 1000 1100 Temperature (C) Figure 6-25. RealTemp data of growth of GaN on MgO. 226

PAGE 227

A) B) Figure 6-26. SEM images of MOCVD GaN on MgO, showing areas of good growth and pitting. A) Region of somewhat smooth growth. B) Area with high density of pits. 227

PAGE 228

Figure 6-27. TEM image of the MOCVD growth of GaN on MgO. Image indicates nucleation layer induced stress, and oxide bl ocking of threading dislocations. 228

PAGE 229

Figure 6-28. RHEED image of MBE growth of scandium oxide. 60 65 70 75 80 0.1 1 10 100 1000 10000 100000 1000000 CountsAngle (2 Theta) Figure 6-29. XRD scan of scandium oxi de showing oxide peak around 65 degrees 2 229

PAGE 230

A) B) Figure 6-30. AFM images of scandium oxide for MOCVD GaN substrate. A) Large area scan. B) Smaller area scan having roughness of 0.6nm. 230

PAGE 231

Figure 6-31. MOCVD quarter wafer sa mple holder showing degradation. Figure 6-32. RealTemp data from first attempt at MOCVD GaN on scandium oxide. 231

PAGE 232

A) B) Figure 6-33. SEM images of the aborted GaN on scandium oxide sample, showing uniform coverage with GaN island nuclei. A) SE image at 2,000X. B) SE image at 20,000X 232

PAGE 233

A) B) Figure 6-34. SEM images of the aborted GaN on scandium oxide sample after growth continued, showing some coalescence of th e GaN island nuclei. A) SE image at 2,000X. B) SE image at 20,000X 233

PAGE 234

A) (B C) (D E) (F G) Figure 6-35. Contact angle measurements taken for various substrates for GaN growth. The measurement was conducted with a rame-hart goniometry with DI water. The change in contact angle indicates a difference in surface energy resulting in varying surface mobility of GaN on the surface under high temp erature growth regimes. A) Standard sapphire substrate epi-polished with no treatment. B) Sapphi re surface of wafer with silicon coated backside after etching to remove silicon on epi-side of wafer. C) Lattice matched MgCaO. D) MgO. E) Sc 2O3. F) Standard MOCVD GaN. G) GaN sample interrupted right after island fo rmation at 1090C, of the low temperature nucleation layer. 234

PAGE 235

Table 6-1. Contact angle measurements of the various surfaces employed in the growth of GaN with both 1l and 5l water drops. Sample Description Contact Angle (deg rees) 1l Contact Angle (degrees) 5l Standard Sapphire 20 17 Sapphire with Backside Silicon after Etch 45 46 MgCaO 47 38 MgCaO after 2 minutes --34 MgO 57 61 MgO after 2 minutes --47 MOCVD GaN 71 69 Sc2O3 79 78 GaN sample after island formation 89 89 235

PAGE 236

A) B) Figure 6-36. Oxide samples afte r contact angle measurement showi ng affected areas. A) MgO. B) MgCaO. 236

PAGE 237

A) B) Figure 6-37. AFM images showing the MgCa O surface before and after contact angle measurement with DI water. A) Before measurement. B) After measurement showing etching effect of the water. 237

PAGE 238

A) B) Figure 6-38. Contact angle measurements of MgO and MgCaO showing the change of angle over a two minute exposure to the water dr op. Red and blue lines indicate the location of the drop right afte r initial contact. The black and white image shows the actual water drop after two minutes of exposure. A) MgO. B) MgCaO 238

PAGE 239

606570758085 0.1 1 10 100 1000 10000 100000 1000000 CountsAngle (2 Theta) Figure 6-39. XRD scan of th e scandium oxide capped MgO. A) B) Figure 6-40. AFM scans of the Sc2O3 capped MgO. A) Cross-section showing growth features around 10nm tall. B) Five micron s can length having roughness of 1.8nm. 239

PAGE 240

Figure 6-41. RealTemp data show ing the growth of GaN on the Sc2O3 capped MgO. Good oscillations are indicative of nice growth. 240

PAGE 241

A) B) Figure 6-42. Optical micrographs showing the GaN growth on Sc2O3 capped MgO. Hexagonal growth pits can be seen across most of the sample surface. A) 20X. B) 100X. 241

PAGE 242

A) B) Figure 6-43. SEM images of the MOCVD GaN on Sc2O3 capped MgO. A) 1,000X magnification. B) 10,000X magnifica tion of hexagonal surface pit. 242

PAGE 243

A) B) Figure 6-44. AFM images of MOCVD GaN on Sc2O3 capped MgO showing terraces growth. A) 5m by 5m scan. B) 1m by 1m scan, having average roughness of 1nm. 243

PAGE 244

Figure 6-45. Cross-section FE-SEM of thick MO CVD GaN on oxide. The figure is labeled to show the stack structure of growths. 244

PAGE 245

Figure 6-46. Additional crosssection FE-SEM images of th e GaN grown on oxide in the MOCVD. 245

PAGE 246

Figure 6-47. TEM image of the MOCVD GaN on Sc2O3 capped MgO. 246

PAGE 247

300 400 500 600 700 0.0 0.2 0.4 0.6 0.8 1.0 1.2 GaN on Oxide Oxide on GaN Starting GaNIntensity (arb units)Wavelength (nm) Figure 6-48. PL spectrum comparisons during th e multiple stages of growth of MOCVD GaN on oxide. The reduction in yellow emissi on with a strong band-edge emission is promising of high quality GaN. 247

PAGE 248

300 400 500 600 700 0.0 0.2 0.4 0.6 0.8 1.0 1.2 GaN on Oxide GaN standardIntensity (arb units)Wavelength (nm) Figure 6-49. PL spectrum comparison between th e growth on MBE oxide, versus a standard GaN sample grown on sapphire. 248

PAGE 249

A) (B C) (D Figure 6-50. CL images of the GaN on MBE oxide compared to standard MOCVD GaN on sapphire, with gray levels equal. The growth on MBE oxide shows much reduced threading dislocations, which can be seen in the GaN on sapphire images which are dark areas in the images. A) GaN on sa pphire at 1,000X. B) GaN on sapphire at 5,000X. C) GaN on MBE oxide at 1,000X. D) GaN on MBE oxide at 5,000X. 249

PAGE 250

010002000300040005000 0 2 4 6 Time (s)Reflectivity %400 500 600 700 800 900 1000 1100 Temperature (C) Figure 6-51. RealTemp data of MOCVD GaN grown on MgCaO capped with MBE GaN. 250

PAGE 251

A) B) Figure 6-52. SEM images of MOCVD GaN on MgCaO capped with MBE GaN. Growth shows terraces with some incomplete coalescence. 251

PAGE 252

A) B) Figure 6-53. AFM images of the MOCVD Ga N on MgCaO capped with MBE GaN. A) 20m scan area showing large growth platelet s. B) 1m scan showing roughness around 1nm. 252

PAGE 253

A) B) Figure 6-54. Cross-section FE-SEM of the MOCVD GaN on MgCaO capped with MBE GaN. The image indicates incomplete coalescence of the GaN top film. 253

PAGE 254

300 400 500 600 700 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 Intensity (arb. units)Wavelength (nm) Figure 6-55. PL spectrum of the MOCVD GaN on MgCaO capped with MBE GaN. The signal was very weak from the sample surface being quite uneven. 254

PAGE 255

A) B) Figure 6-56. CL images of the MOCVD Ga N on MgCaO capped with MBE GaN, showing some emission from the growth edges of th e incomplete coalescence. A) 5,000X. B) 1,000X. 255

PAGE 256

CHAPTER 7 CHEMICAL STABILTY AND ETCHING OF MGCAO Along with the ability to grow and characte rized materials, the ab ility to control the processing of materials is essentia l. With the knowledge of etchi ng and stability of the materials used, a logical sequence of processing steps can be ascertained. Samples with and without the top GaN are examined for the ability to be etched for further device fabrication. Wet Etching of MgCaO Having the ability to etch material fast a nd controllably is one of the most desired components of device fabrication. The ability to we t etch a material can me more valuable than dry etching for the fact that dry etching can introduce damage from the high energy plasma. Also, wet etching can be much faster at etching some materials over others. One of the most useful properties of MgCaO is the ability to be rapidly wet etched. It has been shown that the (100) surface of MgO dissolves in liquid water at a rate faster than the (111) face.46 The possible dissolution of the oxide is summari zed in Table 7-1. It is also stated that the rate of the MgO dissolution increases as the acidity of the solution increases. Since the ternary oxide is composed of MgO it is reasonable that this assumption will hold with the ternary along with the binary oxide. Since one of the concerns with using a new material for device fabrication is compatibility with the chemicals used for proce ssing, the ternary oxide wa s exposed to a variety of normally used chemicals. The experiment set forth was to test a few chemicals used in the lithography and processing of normal materials and examine the effects that occurred. Therefore, a sample of lattice-matched MgCaO was diced into many pieces and exposed to deionized water, methanol, and acetone. Thes e chemicals are essent ial components to many processing steps. The samples were placed in be akers of the respective chemicals and allowed to 256

PAGE 257

soak for five minutes while gently swirling th e solution. After the five minute exposure, the samples were removed and rapidly rinsed in DI water, then blown dry with nitrogen gas. The samples were then analyzed with tapping mode AFM. The star ting surface is shown in Figure 7-1, while the samples exposed to the various chemicals are shown in Figure 7-2. The AFM results are given with height images whic h give the values for the roughness, while phase imaging was performed which indicate areas on th e sample which are harder or softer. The phase image allows for a qualitative view of diff erences along the surface. One other interesting result shown in Figure 7-3, which shows a 20 micron by 20 micron scan area, is that after scanning at a higher magnification the scan area is visible. This visi ble scan area is attributed to the compression of the oxide surface which is still fully hydrolyzed from the DI water soak. This result has been seen in other samples that have been exposed to atmospheric humidity and then scanned. The AFM is capable of showing a vari ation of the surface after scanning resulting from the movement of the hydroxide layer on the surf ace. All the samples showed some form of etching from all the chemicals. The roughness va lue of the starting surface was on the order of 9 while the acetone roughened the surface to just over 12 the methanol increased to just over 2 nm, and the DI water had the most etching with a result of just over 13 nm roughness. These results are tabulated in Table 7-2. Since the literature indicates that the oxide dissolution increases w ith increasing acidity, some properties are tabulated in Table 7-3. Sinc e the sample with the least etching is the one with acetone, it was still unclear of the mechanis m, since the pH of acet one is slightly higher than both methanol and water. Therefore, the other thought was that possi bly the polarity of the solution has some effect on the etching. Also included in Table 7-3 are the dipole moments of the solutions. Since the differences of the solutions are not that strong, it was concluded that this 257

PAGE 258

mechanism was probably minute. Therefore, the fact that the acetone had the smallest effect does support the notion and literature that more ac idic solutions show hi gher dissolution of the oxide. In addition to the various chemicals tried, two other pieces of the oxide were placed in phosphoric acid at various concentr ations. Past experience show s strong etching of the oxide with phosphoric acid. The samples were exposed to acid at concentrations of 2% and 20% for 5 minutes. After removal from the acid and a DI ri nse, the samples were imaged with AFM. The oxide had been completely removed in both conc entrations, and the substrate GaN was exposed. The surface looked just like the starting GaN surface with terraces and no oxide present. Therefore, phosphoric acid was determined to be an effective method of removing the MgCaO for processing and etching. Dry Etching of MgCaO and GaN Once a GaN film has been grown on the MgCaO surface, the next step is the creation of devices. With this goal in mind, th e first step is to identify the conditions needed to dry etch the GaN film. Since it is well known that GaN is qui te stable with most we t etches, dry etching is the preferred method of creating GaN features. In addition, the oxide upon which the GaN film is grown rapidly etches in wet chemistries, so a wet etch for the GaN is not the preferred approach. In addition to the GaN film, determini ng the etch rate of the MgCaO at the same time allows for determining the selectivity of the etch. With standard optical lithography techniques, photoresist (PR) wa s patterned on the UOE GaN samples and MBE grown MgCaO in the shape of narrow rectangular bars. Samples were then loaded into a Unaxis 790 reactor with pressures around 5mTorr. The ion density was controlled by a 2 MHz, 3-turn inductively coupled plasma (ICP) source, while the ion energy was controlled by separate rf-biasing (13.56 MHz) of the sample holder. After etching the 258

PAGE 259

samples were rinse in acetone to remove the resi dual PR. The etch depth was determined with an optical profilometer. Various plasma chemistries were examined to see differences in conditions. Plasmas of Cl2/Ar and CH4/H2/Ar were analyzed. CH4/H2/Ar produced etch rates around 20-70 /min for the oxide, while etch rates around 100 /min was obtained with Cl2/Ar plasma under the same conditions. The GaN on the other hand, showed ra tes of almost an order of magnitude higher. The slow MgCaO etch rates are limited by the low volatilities of the oxide etch products. The volatilities of both the MgCl2 and CaCl2 etch products is very low as the melting point of CaCl2 is 782C, and MgCl2 is 714C and therefore a layer of thes e products forms and inhibits further etching.47 In addition the potential me tal hydride etch products of CaH2 and MgH2 also have low volatility. Shown in Figure 7-4, it can be seen that the MgCaO etch rate is much less than the GaN etch rate at all powers examined within the Cl2/Ar plasma. Also shown for reference in the graphs is the d.c. self-bias created on the samp le holder during the etches. This self-bias increases with increasing rf chuck power indicati ng a rise in incident ion energy. The slower etching of the oxide indicates that this chemistry is well suited to etch the a top layer of GaN, while not rapidly removing the underl ying oxide. This indicates th at the MgCaO can be used as an efficient etch stop during de vice fabrication. With being quite robust to the dry etch conditions, very accurate etchings are not necess ary with this system and the etching will not punch through the oxide which can be used as a device to device isolation field. Figure 7-5 shows the data for dry etching with CH4/H2/Ar chemistry. Shown in the graphs, the MgCaO etch rates are very slow co mpared to the GaN. Also to note is that roughening or degradation of the GaN surface was not detected during these etch runs. These 259

PAGE 260

results are shown in Figure 7-6 which shows the results of the op tical profilometer data of the GaN etchings and some SEM images in Figure 7-7. Finally in Figures 7-8, the eff ect of the ICP power on the etch rates of the materials is shown. When increased, the pow er creates a more conductive plasma as the ion density is increased, and in turn reduces the self-bias. In both chemistries, the MgCaO still had a lower etch rate than the GaN. 260

PAGE 261

Table 7-1. Possible reacti on of MgO with water. 1) [MgO] + OH[MgO]OH(2) [MgO] + H+ [MgO]H+ (3) [MgO] + H2O [MgO]H+OH(4) [MgO] + OH [MgO]OH (5) [MgO] + H2O [MgO]H2O (6) [MgO] + H2O [MgO]H+OH261

PAGE 262

A) B) Figure 7-1. MgCaO starting surface for etching experiment showi ng terraced growth. A) 1m scan having roughness less than 1nm. B) Larger area view with roughness around 1.5nm. 262

PAGE 263

Figure 7-2. AFM images of MgCaO exposed to various solutions for five minutes. The AFM images on the left are the height images, while those on the right show the phase image. A) Acetone exposure. B) Meth anol exposure. C) DI water exposure. 263

PAGE 264

Figure 7-3. AFM tapping image showing the 5 micron scan area that was imaged just before this larger view. The scan outline is from th e movement of the hydroxide layer on the surface of this sample which was exposed to DI water for 5 minutes. Table 7-2. Results from the wet etch study. Sample Average Roughness Z Range Before etch 0.86nm 14.3nm Acetone 5 min 1.22nm 13.6nm Methanol 5 min 2.03nm 23.8nm DI water 5 min 13.2nm 78.0nm 264

PAGE 265

Table 7-3. Properties of the vari ous chemicals used on the MgCaO. Name Structure Dipole Moment (D) pH values Acetone 2.91 pKa= 19.3 Methanol 1.69-2.87 pKa= 15.5 Water 1.85-2.90 pKa= 15.74 Phosphoric Acid 2.78-3.02 pKa= 2.12 265

PAGE 266

8090100110120130140 100 1000 MgCaO GaN20 Cl2/ 10 Ar Etch Rate (A/minute)RF Power (W)170 180 190 200 210 220 230 240 250 DC Bias DC Bias (V) Figure 7-4. Etch rate results of MgCaO and GaN in a chlorine plasma, showing over an order of magnitude less etching for the oxide compared to GaN. 266

PAGE 267

8090100110120130140 100 MgCaO GaN5 CH4/ 10 H2/ 5 Ar Etch Rate (A/minute)RF Power (W)140 160 180 200 220 240 DC Bias DC Bias (V) Figure 7-5. Etch rate results of MgCaO and GaN in methane plasma, showing almost an order of magnitude less etching for the oxide compared to the GaN. 267

PAGE 268

Figure 7-6. Optical profilometer results from dr y etching of GaN. Similar results from the MgCaO dry etch resulted 268

PAGE 269

A) B) Figure 7-7. SEM images showing the dry etching of GaN and the resulting features set forth through the lithography process. A) Highe r magnification of GaN etch. B) Low magnification of GaN etch features. 269

PAGE 270

A) 050100150200250300 10 100 1000 MgCaO GaN Etch Rate (A/minute)ICP (W)200 220 240 260 280 300 320 340 360 380 400 20 Cl2/ 10 Ar 120 W RF DC Bias (V) DC Bias B) 050100150200250300 10 100 MgCaO GaN5 CH4/ 10 H2/ 5 Ar 120 W RF Etch Rate (A/minute)ICP (W)200 220 240 260 280 300 320 340 360 380 DC Bias DC Bias (V) Figure 7-8. ICP etching of GaN and MgCaO in chlorine and methane chemistry. A) Chlorine chemistry. B) Methane chemistry. 270

PAGE 271

CHAPTER 8 GAN FEATURE FABRICATION AND LIFTOFF Another issue encountered in the expansion of GaN devices is heat generation. Heat buildup in high-power radio frequency (RF) devi ces has been the crutch for most material systems. Having the ability to dissipate heat from the device permits greater performance at higher operating loads and longer device lifetime. Unfortunately, the substrates currently used for epi-growth of GaN, are poor thermal conducto rs. Therefore, having the ability to then transfer these features onto high thermal conductivity substrates would allow a greater range for device operation. As the heat is more efficiently dissipated through the su bstrate, thicker films of this high-quality material can be grown. Th icker and larger devices will allow for larger power loads used in high-performance applicat ions. Therefore, to transfer GaN onto high thermal conductivity substrates would warrant a gr eater range for device operation, as the heat is more efficiently dissipated through the substrate. The benefit of the MgO based substrate to be etched away while leaving the nitride intact, lends itself to the concept of lift-off or pick-andplace. Using the procedures of flip-chip-bondin g, the GaN can be bonded to a separate substrate where it can be further processed into devices. Transfer of GaN onto diamond substrates or other comparable materials would enable the hi gh thermal conductivity a nd would facilitate the GaN devices to work at higher te mperatures and power loads. Th e substrate transfer could be accomplished after the material growth or after the device fabrication was complete. Nano Dimension GaN Features In todays expanding field of nano-based devices science more than practicality rules the laboratory. With the popularity of nanowires an d their application to optical and electronic fields, there has been one issue to overcome, organization. What is meant by organization, is the arrangements of these nanowires on the growth s ubstrate. Seeding or nucleating these nanowires 271

PAGE 272

usually results in a random arrangement on th e growth substrate which resemble hair.48, 49 Along with this randomness, the dimensions of th e nanowires vary, which results in variable concentrations of dopants throughout the wire. Instead of growing these wires directly on th e substrate to the dimensions desired, the wires will be patterned and etched from a con tinuous film, in which the nano-dimension is the thickness. Therefore, not only can the nitride be patterned and fabricat ed into devices on the insulating oxide below, the GaN can be patterned in to features, such as bars or cylinders. These features can then be utilized as devices in th e power-electronic or opto electronic fields. There has been some success in growing vertical GaN features in MOCVD using a masking technique (Ref. 4-5.) In this technique a mask of usually silicon dioxide or silicon nitride is deposited then thr ough the use of optical lithograp hy, openings are made. These openings are where the GaN grows. These wire s are on the order of a few hundred nanometers in diameter and only a slightly taller. The vertical dimension is limited by the height of the mask. Above the mask, lateral growth occurs. GaN on Oxide Nanobars To overcome the obstacles of nano-feature growth, (Ref. 25), this work utilized a different approach for fabricati ng the nanowires. Instead of grow ing these wires directly on the substrate to the dimensions desired, the wires are be patterned and et ched from a continuous film, in which the nano-dimension is the thickness. Therefore, not only can the nitride be patte rned and fabricated into devices on the insulating oxide below, the GaN can be patterned in to features, such as bars or cylinders. These features can then be utilized as devices in the power-electronic or optoelec tronic fields. Left on the oxide field, the isolation betw een features is well suited for close proximity device operation. Analysis through XPS to determine the band offsets of the MgCaO with GaN verifies this 272

PAGE 273

application. The determination of the valence band offset of the MgCaO/GaN heterojunction results in a value of 0.65 eV. As the bandgap di fference of the materials is 4.01 eV, the nested interface band alignment results in a conduction band offset of 3.36eV.50 This offset is well within the parameters required for device to device isolation. Once the GaN film has been grown upon the Mg CaO, the film may be patterned through optical lithography methods to obtain the nano-scal e features. As a proof of concept to nanoscale optical lithography, Intel has just converted over ha lf of their production line of microprocessors to the 65nm technology from the pr evious 90nm (Ref. 23.) As state of the art technology continues to become production t echnology, Intel is looking at their 45nm lithography. Realistically for our capabilities at UF, our smalle st line width through optical lithography will be on the order of 400nm. Yet, the technology is available and proven. Our control of the nano-dimension is through growth thickness. Processing of the GaN on oxide is facilitate d by the fact that the nitride and oxide have quite different etch rates. The etch rates fo r the MgCaO films are lower than for GaN by more than an order of magnitude, and standard plasma chemistry is not capable of a high removal rate of the MgCaO layers.47 This results in an etch barrier if wanting to completely etch the GaN into features, and expose the underlying oxide. Feature isolation can then be realized, as the chance of etching through the oxid e into the underlying substrate is minimized. Next, GaN bars were fabricated from the materi al. As talked about earlier, the MgCaO is quite resilient to the dr y etch conditions needed for GaN etchi ng. Therefore, the oxide acts as an efficient etch stop during the GaN etch step. Fo r proof of concept, larger horizontal dimension features have been fabricated to make processing more straightforward. Shown in the Figure 81, are GaN bars on a MgCaO field, with EDS line scan across the bar and field to show the 273

PAGE 274

elemental changes. AFM scans show the dimens ions and morphology of the features in Figure 8-2. Following the growth and etching of the GaN into these bar features, lithography was utilized to pattern and deposit metal contacts to the featur es. The idea behind this process was that by placing ohmic contacts on the GaN bar, a resulting current/voltage or IV curve can give a deeper insight into the properties of the GaN film and the dimension effects. As seen in Figure 8-3, the metal contacts were spaced evenly across a GaN bar, with Auger analysis of the bar and the field in Figures 8-4. Once contacted and IV teste d, no conduction through the GaN feature could be gained. A current was finally measured once the voltage was increased to nearly thirty volts. This conduction was from the tunneling of current through the oxide under the contact pads and into the underlying GaN substrate. The high voltage was indicative of the high breakdown voltage of this highly insulating film. To verify the oxide properties, the breakdow n voltage was determined for the latticematched MgCaO. A sample of 70nm thick oxide on highly conductive n-GaN was patterned with contact pads for a Schottky contact. The bottom contact was the n-GaN substrate in which indium was wiped onto the side. The average vo ltage at a compliance of 100nA, was 14V. This equates to a breakdown field of approximately 2MV/cm. This s upports the fact this oxide can not only be used as a template fo r GaN growth, it can also be used as an insulating field oxide for device fabrication. Lift-Off Solution Finally, the notion that the oxi de can be readily etched whil e leaving the nitride intact, lends itself to the concept of lift-off. Using the procedures of flip-chip-bonding, the GaN or GaN features can be bonded to a separate substrate where they can be further processed into devices. 274

PAGE 275

Even though the MgCaO is quite robust to dr y etching, an advantage for the MgCaO is its ability to be easily etched through wet chemistry. A dilute etch of Phosphoric Acid (2%) rapidly etches the oxide but leaves the GaN unaffected (R ef. 25.) The ability for the MgCaO to be wet etched at a very rapid rate allows for large und ercut of the GaN, releasing the film from the starting substrate. This method is simpler than the method of using an excimer laser required for removal from sapphire substrates. There is commercial interest for the selective transfer of films and small features of GaN onto various substrates. Raytheon is one such company that is inte rested in the transfer of GaN onto diamond substrates. Since diamond not only is electrically insulati ng, it has a very high thermal conductivity. This high co nductivity would facil itate the GaN devices to work at higher temperatures and power loads. Show in Figure 8-5, is a represen tation of the transfer process. Lateral Etching for Liftoff Samples of MBE oxide were patterned with photoresist (PR) through standard lithography techniques so that feat ures of various dimensions were created on the oxide surface. These features were created to simulate final device structures of GaN that would be grown on the oxide surface. After device creation transferring the devices to other substrates is a highly desirable trait; therefore, to test this ability the PR features will simulate the testing. Shown in Figures 8-6 and 8-7, features of rectangular and circul ar dimensions have been fabricated of approximately 2 microns tall on the oxide. Figure 8-8 shows an optical micrograph of the features. The dimensions of the rectangular features are approximate ly 100m by 550m, and the circle is approximately 350m in diameter. For the lift-off process, the samples were submerged in a weak phosphoric acid solution. The phosphoric acid has been shown to readily etch the oxide without harming the PR features. Each sample had exposure to a 2% and a 20% acid solution, to compare lateral etch rates, which 275

PAGE 276

undercut the PR features, resul ting in liftoff (Figure 8-9). Th e oxide samples are of various thicknesses and compositions, which can then relate lateral etch rate to these properties. The thickness result will show the necessary oxide th ickness to allow for rapid removal of future GaN devices, while the composition results will indicate the etch rate if an oxide of different lattice parameter is desired. The samples were submerged in a beaker containing the various acids for 60 seconds while gently swirling. Swirling th e solution allows for removal of etch products within the ledge created by the undercutting. Sonication also works, but is too aggressive for the photoresist features. The samples were then removed and quickly rinsed with DI water to neutralize the acid solution. Once removed from the water, samples were blown dry with compressed room air. At this point, the samples were examined by an optical microscope, with digital imaging capabilities. To note, all the samples exposed to the acid soluti ons did have undercutting of the PR features. Also, all samples in the 20% solu tion exhibited a greater amount of etching. The samples were divided into series and some results where extrapolated. Two of the samples were grown in the same manner, with th e only difference being growth time, resulting in 50nm and 230nm. These oxides are lattice-matched to GaN and therefore give a good idea of the etch rates for device material. The idea behind th is experiment is that possibly a thicker oxide will etch faster since the undercut aspect ratio wi ll allow for a larger opening. A larger opening should allow easier flow of acid solution to reac h fresh oxide and remove the etch products. Shown in Figure 8-10, the etch rates of the diffe rent thicknesses seem quite similar. The measurements show that effectively the late ral etch rates are the same for the various thicknesses. Since this is the case, the thickn ess of the oxide has ne gligible effect on the undercutting ability to the extent of the dimensions measured. 276

PAGE 277

The next set of samples investigated was co mprised of oxides with varying compositions. One sample was pure MgO, while another wa s approximately 30% MgO, and a third was directly matched to GaN with around 50% MgO. The though for this set was that the Mg might have a different etching ability than the Ca. Si nce it is shown that pure CaO reacts more readily with water and acids than MgO, which is said to be fairly insoluble with water.12 The results of this experiment confirms the notion that a ternar y composition with more Mg is more resistant to the acid etch. Both concentrations of acids s howed the same trend towards a higher etch rate with increasing calcium content in the ternary MgCaO, as illustrated in Figures 8-11 and the graphical representation in Figure 8-12. For testing, one of the samples was re-submerge d into the acid solutions for an additional 60 second exposure for a total of 120 seconds. The thought was the undercut etching would continue, possibly at a slower ra te due to the aspect ratio of the height to lateral width not allowing for adequate removal of etch products. After examining the samples, it was found that there was very little additional etching, as seen in Figure 8-13. It was then thought that the reason for these results was twofold. After the initial 60 seco nd exposure to the acid and subsequent blow drying the PR ledge that was produced from the undercut either broke and rested upon the underlying clean GaN surface or prohi bited further etching due to trapped air, thus restricting the acid from reaching the remaining oxide. Since the initial acid exposure etched the oxide from the open field into the undercut features, no air became trapped, allowing for adequate removal of oxide etch products. It was therefore determined that for any industrial processing, etching steps would nece ssarily have to be performed in a single pass, in order to yield the desired results. 277

PAGE 278

To verify if the PR ledge had broken after etching, the sample was analyzed using a Wyko optical profilometer. Show in Figure 814, there was evidence of a height difference between the area of PR containi ng oxide underneath and areas where oxide had been etched. This confirms that the ledge had released and fallen onto the sample. Also this result was confirmed by the measurement of the PR drop he ight. All samples showed approximately the same PR drop distance as the th ickness of the oxide as veri fied through field-emission SEM analysis. In addition to the fabrication and liftoff of small features in the micron size, the need might rise to transfer a full wafe r film to another substrate. To test this app lication, a large millimeter size piece of MBE GaN on MgCaO was cleaved and mounted to a glass slide with black mounting wax. The glass slide is heated on a hotplate and a small chip of the wax is placed on the slide and allowed to melt. A piece of the sample is then placed very gently on the wax so that the wax is either smaller than the sample or gently placed where no wax will coat the side of the sample. If the wax coats the side s of the sample, no amount of etching will release the layer since its the ed ge with the exposed oxide that allows the acid etchant to do the work of liftoff. A piece of mounted GaN on oxide was placed in a bottle of 2% and one in a bottle of 20% phosphoric acid solution. Allowed to sit, the sample within the 20% acid solution totally released after approximately 24 hours soak with an occasional agitation. Optical images show the released GaN layer in Figure 8-15 which had an approximate diameter of 3mm. Some areas indicate where the mounting wax did not contact the GaN film during mounting by the circular dark areas. Also shown is an area where there had been a small particulate on the oxide surface before growth of the top GaN. This area s hows how the top GaN nucleated around the particle and then coalesced together by th e terraced look of the film around the particle. Shown in Figure 278

PAGE 279

8-16 the released 1.5 micron GaN layer was then imaged with AFM. The AFM image shows the surface that was in at the inte rface with the MgCaO. The su rface roughness of this layer is around 3-4nm, indicating that the oxide surface was not directly replicated during the growth of the MBE GaN, as the starting oxide surface had roughness around 0.8nm (Figure 8-17). The surface does have the same textured look as the oxide, and along with the columnar growth seen in cross section imaging (Figure 8-18) the r ougher surface is understandable. The morphology could possibly be improved by different MBE growth conditions of the gallium nitride. The reason behind the scanning of this surface is th e fact that depending on the type of device wanted, nitrogen-polar or gallium-polar, this surface could be used for devices rather than the top of the last grown GaN layer. Following the successful lift-off of the 1.5 micr on MBE GaN, the next experiment was to remove the 5 micron thick MOCVD GaN from the scandium oxide capped magnesium oxide which was shown in Chapter 6. The sample was mounted on black wax and submerged into 20% phosphoric acid for 2 days before the sample was separated (Figure 8-19). The sample size was approximately 3mm by 5mm. The top GaN la yer which was in contact with the oxide was imaged through AFM (Figure 8-20), indicated a slightly rough surface most likely from the islands of the buffer layer. Also to note, was th at this sample was extremely resistive, as HALL was not able to be performed, and the IV curv e from the HALL sample showed resistance on the order of six mega-ohms. In addition to the 3mm circular GaN lift-off, our corporate partners hip with Raytheon had similar success. After growth material was shippe d to their laboratories, half inch circles were cut out of the samples by a laser. These large diameter samples represent the ultimate challenge for lift-off as the undercutting of the film with only 100nm of oxide posed a huge aspect ratio 279

PAGE 280

between lateral distance and vert ical oxide height. Yet, with the challenges came huge successes with the successful release of full 0.5 diameter circles of GaN. Shown in Figure 8-21, the samples had been mounted to glass slides with Cr ystalBond. The samples were then left to soak for five days in a 2% phosphoric acid solution before the substrate was fully removed. These results prove that extremely large areas of GaN epi-film can be transferre d to foreign substrates with the use of MgCaO. Overall, the successful release of the top GaN layer from the underlying GaN on sapphire substrate by the etching of the intermediate oxi de opens the door for great possibilities. The realization of a lattice matched oxide that can be easily wet etched as a release layer for lift-off of a GaN epi-layer has been proven. Wi th this capability the transfer of the GaN film to a variety of substrates can be accomplished and create a ne w field that can use high thermally conductive substrate for device platforms. These types of substrates can then allow for removal of heat generated by the extreme operating levels capable of gallium nitride. Then finally the limits of gallium nitride devices can be explored and improved. 280

PAGE 281

Figure 8-1. In the above SEM images, a BSE image indicates the compositional difference between the GaN bar and the MgCaO field. Al so, an EDS line scan verifies a rise in calcium and oxygen in the location of the MgCaO. 281

PAGE 282

Figure 8-2. Above are AFM images of the GaN bars on MgCaO showing surface morphology between the materials. A section analysis is included, indicating a vertical height of just over 200nm for the bar. This height in dicates how the oxide acts as an etch stop, since the GaN growth thickness was approximately 200nm. 282

PAGE 283

Figure 8-3. SEM images of metal cont acts on GaN bars on a field of MgCaO. 283

PAGE 284

A) B) Figure 8-4. Auger analysis of the sample in Figure 8-3. A) Spot analysis of the oxide field. B) Spot analysis of the GaN bar. 284

PAGE 285

Figure 8-5. Shown above is the pr ocess of transferring the grown fi lm or features onto a foreign substrate. The features can be from the millimeter to micrometer scale. 285

PAGE 286

A) B) Figure 8-6. Optical profilometer scan of the circular PR features produced onto the MgCaO surface, to simulate GaN. Width of the f eature is 350 microns. A) Cross-section of the sample showing height and width. B) 3D rendering of the feature. 286

PAGE 287

A) B) Figure 8-7. Optical profilometer scan of the rectangular PR features produced onto the MgCaO surface, to simulate GaN. Feature dime nsions are approximately 100 by 500 microns. A) Cross-section of the sample showing height and width. B) 3D rendering of the feature. 287

PAGE 288

A) B) C) Figure 8-8. Optical microscope images of th e PR features on MgCaO. A) 5X overview indicating the features of interest. B) Corner of a rect angular feature. C) Edge of a circular feature. 288

PAGE 289

A) (B C) (D Figure 8-9. Optical images of the undercutti ng of PR, showing the differences with acid concentration. A) 5X image with 2% acid. B) 100X image with 2% acid. C) 5X image with 20% acid. D) 100X image with 20% acid. 50 230 0 10 20 30 40 50 60 70 Lattice Matched MgCaO to GaNEtch Rate (microns/min)MgCaO Thickness 2% acid 20% acid Figure 8-10. Oxide thickness versus lateral etch rate. 289

PAGE 290

A) (B C) (D Figure 8-11. Optical images s howing etching differences with oxide composition. A) MgO, 2% acid. B) MgO, 20% acid. C) Mg0.25Ca0.75O, 2% acid. D) Mg0.25Ca0.75O, 20% acid. 0102030405060708090100 15 20 25 30 35 40 45 50 55 60 Etch Rate (microns/min)Percent Mg 2% acid 20% acid Figure 8-12. Etch rate versus oxide composition. 290

PAGE 291

A) (B C) (D Figure 8-13. Images of sample submerged in acid for second time, showing minimal additional etching. A) Circular feature 1st etching. B) Circular feature 2nd etching. C) Rectangular feature 1st etching. D) Rectangular feature 2nd etching 291

PAGE 292

A) (B C) (D Figure 8-14. Optical profiler images of the PR folding after first round out etching. A) Circular feature, 2% acid. B) Rectangular feature, 2% acid. C) Ci rcular feature, 20% acid. D) Rectangular feature, 20% acid. 292

PAGE 293

A) (B C) (D Figure 8-15. Optical images of the 1.5 micron thick MBE GaN fi lm that had been successfully lifted off the substrate and transferred onto another substrate. A) Low magnification showing a scale bar for the size of the circul ar GaN film. B) Image showing areas of GaN film that had not been mounted adequately to the mounting wax. C) Region that had a particle on the oxide and the MB E GaN grow around it, showing terraced growth. D) Crack in film from poor mounting to the wax. 293

PAGE 294

Figure 8-16. AFM images of the underside of th e released GaN film at various scan areas. Roughness around 3-4nm. 294

PAGE 295

Figure 8-17. AFM images of the oxide upon which the GaN sample was grown. 295

PAGE 296

Figure 8-18. Cross-section FE-SEM image of th e thick MBE GaN on oxide sample. This shows the columnar type growth which results in the granular textur e seen by the AFM of the interface with the oxide. 296

PAGE 297

A) B) Figure 8-19. Optical images of a sample of 5 micron thick MOCVD GaN removed from the GaN substrate after etching the intermediate oxide substrate. A) Image showing the entire 3X4mm sample. B) 5X magnifi cation image showing the GaN surface that was removed from the oxide. 297

PAGE 298

A) B) Figure 8-20. AFM images showing the surface morphology of the 5 micron GaN film that was separated from the substrate through etching the oxide. This was the surface that was in contact with the oxide during growth. A) 20 micron scan length having roughness of 5.44nm. B) smaller area scan showing morphology. 298

PAGE 299

Figure 8-21. Released 1.5 micron GaN film from substrate and tran sferred to a glass slide. The sample lifted-off the substrate was a half inch circle. 299

PAGE 300

CHAPTER 9 SUMMARY AND CONCLUSIONS From the initial thoughts of this work, to the successful results, the hard work and long hours have been fruitful. The id ea of a lattice-tunable oxide subs trate for the growth of gallium nitride was novel and had issues to overcome. Bei ng able to create a substrate that has the same lattice-spacing as GaN opens the experimental door for the growth of very high-quality material. The lack of strain at th e interface of these materials, allows for reduced stress in the growing film which has had a large reduction in threading di slocations. Since the MgCaO slip-plane is directed perpendicular to the GaN, the oxide ac ts as a blocking layer for defect propagation. In addition to the highly suited substrate fo r GaN growth, the oxide has the ability to be wet etched extremely rapidly resulting in a succes sful lift-off mechanism for the top GaN layer. Having the ability to transfer a thin film of GaN to a new substrate opens the possibilities for higher power devices. When placed on a substr ate with high thermal conductance, the GaN devices can operate at higher pow ers, since the increased heat generation is removed from the device by the substrate acting as a heat sink. The successful transference of the film has been proven without bowing or cracking in the layer afte r transfer. This demonstration solidifies the idea of a lattice-matched oxide suited for film transfer. Summary of MgCaO on GaN Samples of MgCaO have been grown on MOCVD GaN in a modified Riber model 2300 MBE. The oxides were normally grown to a thickness of 100nm, usually at a substrate temperature of 500C. There was an overpressu re of elemental oxygen, supplied by an Oxford Radio Frequency plasma source operating at 13.56 MHz. The metal sources were heated in standard Knudsen Cells. The Magnesium cell temperature was around 360C while the Calcium was set in the vicinity of 412C. These temperatures were set based on the stoichiometry of the 300

PAGE 301

film, not on elemental fluxes ratios. Equal metal ratios, resulting in 1: 1 stoichimetry of the lattice matched film, were verified through XRD. Two methods were examined that controlled the metal flux ratios. In the first method the el emental sources were shuttered at specific time intervals to create a homogenous composition throughout the film. This type of growth has been termed a digital alloy. This method allows for alternating layers of Mg O and CaO, resulting in a homogenous film. Changing the amount of time the individual shutters are open, allows for composition changes, shifting th e oxide lattice parameter. The other method, and preferred method, was to simply adjust the temperature of the effusion cell to adjust the emerging flux, thereby adjusting the incorporation of elements into the growing film. Adjusting the elemental fluxes resulted in a shift of lattice parameter. This shift allows for lattice matching with the lattice of GaN. The resulting film morphologies of the lattice-matched oxid e successfully replicated the starting GaN terraces, indicating high mobility of the impinging atoms on the substrate surface. Average roughness of the latticematched films was on the same order as the GaN substrate, which were a few angstroms. When the oxide was not lattice-matched, the roughness increased as the mismatch in lattice para meters caused stress in the growing film. Also, the morphologies of non-matched films had difficulty replicati ng the starting GaN subs trate. At extreme mismatch, or when the binaries were grown, no terraces were replicate d. The MgO films would occasionally grow large crysta llites which would decorate th e terrace ledges of the underlying GaN. The conditions causing this phenomenon we re normally from the substrate being at too high of a temperature or when the growth rate was too high. Summary of MBE GaN on MgCaO The GaN grown on top of the oxide was depos ited in a modified Varian Gen II MBE system. Various samples were grown at substr ate temperatures between 600C and 800C. The 301

PAGE 302

nitride was grown to thicknesses between 10nm to 1.5m to give a realisti c film that could be used for device construction. The Surface morp hology was characterized by AFM and SEM. Since the MBE system was setup for the growth of dilute magnetic materials and spintronic materials, extremely smooth GaN growth was never accomplished. The growths did show a reduction in threading disloc ations through cathodoluminescen ce and a strong band-edge in photoluminescence with an increased level of ni trogen vacancies compared to in-house grown MOCVD GaN. Doping of the GaN was accomplishe d through a silicon flux during some of the nitride growths. HALL measurements were performed to verify doping concentration and mobility. Non-doped samples were also grown and measurement to confirm that auto-doping from the oxide substrate was not occurring. No auto-doping was detected and the HALL results did show successful doping ability. In addi tion, XRD examination after the extremely long growth times did not show any phase separation of the MgCaO, indicating a stable substrate for MBE GaN growth. Summary of MOCVD GaN on Oxide The results obtained from the MOCVD GaN gr owths are promising for high quality GaN material. Initial work with GaN growth on MgO has showed that the oxide is stable in the MOCVD environment and growth will occur. The localized smooth GaN material produced does have stress at the oxide interface as s een through TEM. After capping the MgO with scandium oxide, the growth became better and allowed for a thick uniform layer of MOCVD GaN. The surface did have pitting, possibly from incorrect V/III ratios, but again, the nucleation layer is key to the subsequent growth. It was shown that the MgCaO is affected at temperatures around 1100C under ammonia, but not to a vast degree. Th rough the deposition of a thin capping layer of MBE GaN, the subsequent growth produced terraced growths, yet with incomplete coalescence. 302

PAGE 303

The optical characterization methods utilized for these growths have revealed an improvement in GaN quality. PL spectra have indicated strong band-edge emission and almost no emission elsewhere. The PL data was suppor ted through the CL data, which indicated very minute amounts of non-recombinative centers, which emerge as dark areas in the CL image. Overall, the success of MOCVD GaN growth on MgCa O or any material to speak of, will rest in the key step of nucleation. With a successful nucl eation step, the main grow th material will have improved qualities and good morphology. Summary of MgCaO Chemical Stability One of the most interesting aspects of this ternary system is the stability in certain chemistries and the instability in others. The oxide is quite robust to similar dry etch chemistry used for the etching of GaN. The etching is ove r an order of magnitude less. Therefore this oxide can act as an etch stop during dry et ch processing of GaN into devices. Within wet chemistry, MgCaO has incredibly high etch rates. Since the oxide forms hydroxides at a high rate, the disso lution of the oxide in aqueous environments is rapid. With minimal effects from acetone, which increase as the oxide is exposed to methanol, and maximum with phosphoric acid, the material is suited for processes which require lateral undercutting of features. Summary of GaN Feature Fabrication and Liftoff Through lithography and dry etching, GaN film s on MgCaO have been dry etched into nanobars. These features have vertical dimensi ons with the thickness of the GaN growth. By stopping GaN growth less than 100nm, the features can then be cl assified as true nano-sized features. With proven lithogr aphy, arrays of these nano-featur es can be created on the highly insulative MgCaO. Since the band offsets of the GaN and oxide are adequate for device isolation, large areas of devices can be fabricated on the same wafer. 303

PAGE 304

Lateral etching experiments have shown th at etch rates up to 50m per minute are possible with MgCaO. This result opens the door for liftoff of devices after fabrication on the wafer, or complete film transference to other su bstrates. It has been shown that MBE GaN films 1.5m thick and 5m MOCVD GaN films have b een successfully transferred to foreign substrates with areas larger than 125mm2. These transferred films can then be mounted to substrates with high thermal conductivities, like diamond, for use in high power devices. Future Work Directions Future work required involves work on the Ga N growth initiation on the oxide surface. Even with the lattice-matched substrate, surface en ergies play a crucial ro le in the nucleation of gallium nitride on the MgCaO. Once a good nucl eation can be perfected, the realization of a stress-free, extremely low defect GaN material will be produced. Work on the nucleation would involve probing the surface kinetic s and reactions taking place at th at oxide/nitride interface. Additional research into the work that was c onducted with an alumin um nitride buffer layer would give solidification on the results seen th rough this work. The aluminum nitride buffer layer might aid in producing a smooth gallium nitride film, but would also introduce additional stress into the growth. Since AlN and GaN have difference lattice spacing, the different might introduce threading dislocations into the grown GaN. It is the threading dislocations which have been shown through cathodoluminescence to have b een reduced through the use of the MgCaO. Additionally, since th e lattice parameter of the ternary MgCaO is tunable from 2.978 to 3.393 this system could be applied to other ma terials. Materials with hexagonal symmetry parallel to the growing face, and having lattice spacing with this ra nge could be used in the same manner that GaN was used in this research. Ot her material systems incl ude, but are not limited to AlN, SiC, and ZnO. Also, since MgCaO is quite etchable, a system based on the other semiconductors could benefit from the lift-off capab ilities and be transferre d to other substrates. 304

PAGE 305

One other thought that could be a great a dvantage for devices, would be to grade the ternary from one lattice to a different spacing. As an example, suppose you start with a GaN substrate and want to grow ZnO on top for a mu ltiple wavelength light source. The oxide would start lattice-matched to GaN and then as the film grows, the flux ratio would be changed to incorporate additional calcium thereby sh ifting the lattice pa rameter to the 3.325 required for ZnO. This would then allow the GaN substrat e and the ZnO layer to have an LED or LASER structure and be operated at the same time. Since the oxide has a high break-down field as it is very insulative, the different materials could be biased differently allowing for simultaneous operation. The device structure possibilities are endless. Future Work in Progress with ZnO At the time of publication, some results have been collected with the use of MgCaO as a substrate for zinc oxide grow th. The ZnO was deposited by PLD on MgCaO films that were lattice matched to GaN, but can be tuned to th e lattice of ZnO. The substantial improvement only after the first set of samples is very exc iting. Shown in Figure 91, the PL spectrum of a variety of ZnO samples are compared. The resu lts show that a large enhancement of the bandedge emission is seen on samples of silv er doped ZnO on MgCaO compared to growth on sapphire. Results like this ZnO show promise for the MgCaO system. A pplications to other materials system could result in improvements in different arenas. Overall it can be said that MgCaO has a vast set of applications yet to be explored, with the possibi lity of great outcomes. 305

PAGE 306

350400450500550600650 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 Intensity (arb. units)Wavelength (nm) Ag-doped ZnO on Sapphire Ag-doped ZnO on ZnO Buffer on Saphhire Ag-doped ZnO on MgCaO Ag-doped Zno on ZnO Buffer on MgCaO MgCaO300 K Figure 9-1. PL data of ZnO deposited upon Mg CaO and Sapphire using PLD. Growth upon MgCaO shows a four times increase in band-edge emission for ZnO. 306

PAGE 307

LIST OF REFERENCES 1. Pearton, Gallium Nitride Processing for Electro nics, Sensors and Spintronics, (SpringerVerlag). 2006. 2. L. Liu, J.H. Edgar, Materials Science and Engineering R 37 (2002) 61. 3. S.K. Srivastava, Solid State Sciences, 8 (2006) 573. 4. K. Barghout, J. Chaudhuro, Journal of Materials Science 39 (2004) 5817 5823. 5. Parrish, M, Wafer Price Q uote, Cree, Inc. (2007). 6. Liu, Y, Basic Issues for High Quality GaN-Growth by MBE Microtechnology and Nanoscience. Goteborg, Sweden, Chalmers University of Technology. (2005). 7. William A. Doolittle, Tom Kropewnicki, C. Carter-Coman, S. Stock, Paul Kohl, Nan Marie Jokerst, Robert A. Metzger, Sangbeom Kang, Kyeong Kyun Lee, Gary May, April S. Brown, J. Vac. Sci. Technol. B 16.3., May/Jun 1998. 8. William A. Doolittle, Sa ngbeom Kang, April Brown, Solid-State Electronics 44 (2000) 229-238. 9. P. Kung, A. Saxler, X. Zhang, D. Walker, R. Lavado, M. Razeghib, Appl. Phys. Lett., Vol. 69, No. 14, 30 September 1996. 10. M.D. Reed, O.M. Kryliouk, M.A. Mastro, T.J. Anderson, Journal of Crystal Growth 274 (2005) 14. 11. M. Hong, J. Kwo, S. N. G. Chu, J. P. Mannaerts, J. Vac. Sci. Technol. B Vol. 20, No. 3, May-Jun 2002. 12. CRC Press CRC Handbook of Chemistry and Physics, (81st Ed). 13. Takeshi Ohgaki, Shigeaki Sugimuraa, NaokiOhashi, Isao Sakaguchia, TakashiSeki guchia, Hajime Hanedaa, Journal of Crystal Growth 275 (2005) e1143e1148. 307

PAGE 308

14. Foxon, C.T., Handbook of Crystal Growth Vol.3. Elsevier Science, 1994. 15. Razeghi, M, Compound Semiconductors and Crystal Growth Techniques, Fundamentals of Solid State Engineering, 2nd Ed, Springer, pg. 469, (2006). 16. S.S. Kushwah, J. Shanker, Physica B, 225 (1996) 283-287. 17. M Leszczynskit, T Suskit, P Perlint, H Tei sseyret, I Grzegoryt, M Bockowskit, J Junt, S Porowskit, J Major, J. Phys. D: Appl. Phys 28 (1995) A149-A153. 18. V. Kirchner, H. Heinke, D. Hommel, J. Z. Domagala, M. Leszczynski, Appl. Phys. Lett ., Vol. 77, No. 10, 4 September 2000. 19. C. Roder, S. Einfeldt S. Figge, D. Hommel, Physical Review B 72, 085218, 2005. 20. B. S. Kang, S. Kim, J. Kim, F. Ren, Appl. Phys. Lett., Vol. 83, No. 23, 8 December 2003. 21. N. Grandjean, J. Massies, F. Semond, Appl. Phys. Lett. Vol. 74, No. 13, 29 March 1999. 22. A.H.Onstine, A. Herrero, B.P. Gila, D. Stod ilka, C.R. Abernathy, J. LaRoche, F. Ren, S.J. Pearton, Mat. Res. Soc. Symp. Proc Vol. 811, 2004 Materials Research Society. 23. C. S. Corr, P. G. Steen, W. G. Graham, Plasma Sources Sci. Technol 12 (2003) 265. 24. D. M. Kearns, D. R. Gillen, D. Voulot, R. W. McCullough, W. R. Thompson, J. Vac. Sci. Technol A 19.3., May-Jun 2001. 25. R. D. Vispute, S. Choopun, Y. H. Li, D. M. Chalk, S. B. Ogale, R. P. Sharma, T. Venkatesan A. Iliadis., Mat. Res. Soc. Symp Proc. Vol. 583. 26. R.C. Doman, J.B. Bar, R.N. McNally, A.M. Alper, J. Am. Ceram. Soc., 314 (1963). 27. B. Halstedt, J. Phase Equilib ., 662-675 (1993). 308

PAGE 309

28. B. Heying, R. Averbeck, L. F. Chen, E. Haus, H. Riechert, J. S. Speck, J. Appl. Phys. Vol. 88, No. 4, 15 August 2000. 29. M. Rummukainen, J. Oila, A. Laakso, K. Saarinen, A. J. Ptak, T. H. Myers, Appl. Phys. Lett. Vol. 84, No. 24, 14 June 2004. 30. K. Kaneko, N. Iizuka, N. Suzuki, Phys. Stat. Sol ., 2, No. 7 (2005). 31. A.M. Sanchez, F.J. Pacheco, S.I. Molina, J. Stemmer, J. Aderhold, J. Graul, Materials Science and Engineering B 80 (2001) 299. 32. A. Ponce, A. M. Sanchez, S. I. Molina, F. Fedler, J. Stemmer, J. Graul, Phys. Stat. Sol ., 192, No. 2 (2002). 33. Michael A. Reshchikova, Hadis Morko, J. Appl. Phys. 97, 061301, 2005. 34. H. S. Craft, R. Collazo, M. D. Losego, Z. Sitar, J.-P. Maria, J. Vac. Sci. Technol Nov/Dec 2008. 35. A. Armstrong, A. R. Arehart, B. Moran, S. P. DenBaars, U. K. Mishra, J. S. Speck, S. A. Ringela, Appl. Phys. Lett ., Vol. 84, No. 3, 19 January 2004. 36. R. Lin, R.I. Masel, Surface Science 258 (1991) 225-234. 37. D.D. Koleske, A.E. Wickenden, R.L. Henry, Code 6861, Electronic Science and Technology Division Naval Research Laborat ory, Washington, D.C. 20375. 38. C. H. Kunsman, J. Am. Chem. Soc., 1928, 50 (8), 2100-2113. 39. Hitoshi Shindo, Chikashi Egawa, Takaharu Onishi, Kenzi Tamaru, J.C.S. Faraday I, 1980,76,280-290. 40. S.H. Mohamed, Surface & Coatings Technology 202 (2008) 2169. 309

PAGE 310

41. Kirt R. Williams, J. Microelectromechanical Systems Vol. 5, No. 4, 1996. 42. D S Peng, Y C Feng, W X Wang, X F Liu, W Shi, H B Niu, 2007 J. Phys. D: Appl. Phys. 40 1108-1112. 43. F. Dwikusuma, D. Saulys, T. F. Kuecha, Journal of The Electrochemical Society 149 11, G603-G608-2002. 44. Y. Golan, P. Fini, D. Dahan, F. Wu, S. Zamir, J. Salzman, J.S. Speck, Eur. Phys. J. AP 22, 11 (2003). 45. A.F. Stalder, G. Kulik, D. Sage, L. Barbieri, P. Hoffmann, Colloids And Surfaces A: Physicochemical And Engineering Aspects vol. 286, no. 1-3, pp. 92-103, September 2006. 46. Truong, Thanh N., Goodman, D.W ., Johnson, Michael A., Gunster, Jens, Stefanovich, Eugene V, J. Phys. Chem. B 1999, 103 (17), pp 3391. 47. M. Hlad, L. Voss, B.P. Gila, C.R. Abernathy, S.J. Pearton, F. Ren, Applied Surface Science 252 (2006) 8010. 48. Liu, J, Appl. Phys. Lett. 83(20), 2003. 49. Calarco R., Appl. Phys A(87): 499-503, 2007. 50. J.-J. Chen, M. Hlad, A.P. Gerger, B.P. G ila, F. Ren, C.R. Abernathy, S.J. Pearton, Journal of Electronic Materials Vol. 36, No. 4, 2007. 310

PAGE 311

BIOGRAPHICAL SKETCH Andrew Phillip Gerger grew up in Lake County, Florida. He is son to Stephen and Melody Gerger, grandson to Elizabeth Kinz er, and brother to Amy Gerger. During high school he had a fascination with scienc e and would occasionally get in trouble for some homemade experiments. After gr aduating from Leesburg High School in 2001, Andy began his academic career at the University of Florida. He majored in Materials Science and Engineering, graduating with a Bachelors in Scienc e degree in 2005 with a minor in Sales Engineering. Andy then became Dr. Gilas first graduate student within the Abernathy research group. Next came his Master of Science degree in 2007. Continued research in the field of semiconduc tor thin film growth and characterization earned Andrew his Doctor of Philo sophy degree in the summer of 2009. 311