GROWTH, STRUCTURAL, ELECTRONIC AND OPTICAL CHARACTERIZATION OF NITRIDE SEMICONDUCTORS

Transcription

1 GROWTH, STRUCTURAL, ELECTRONIC AND OPTICAL CHARACTERIZATION OF NITRIDE SEMICONDUCTORS A dissertation presented to the faculty of the College of Arts and Sciences of Ohio University In partial fulfillment of the requirements for the degree Doctor of Philosophy Costel Constantin November 2005

2 Constantin, Costel. Ph.D. November Physics & Astronomy Growth, Structural, Electronic and Optical Characterization of Nitride Semiconductors (98pp.) Director of Dissertation: Arthur R. Smith This project investigates the growth, optical, electronic, surface, magnetic and bulk properties of scandium gallium nitride on Sapphire(0001), manganese scandium nitride on MgO(001), heterostructures of cubic gallium nitride and scandium nitride on Mg(001), and chromium nitride on MgO(001) grown by radio frequency molecular beam epitaxy. The growth of Sc x Ga 1 x N films has been performed at a substrate temperature of 650 C. The diffraction and optical experiments confirm the existence of two main regimes of growth; for high Sc concentration (x 0.54), a rocksalt crystal structure is obtained. For low x (x 0.17), a wurtzite-like crystal structure is observed with local lattice distortions at the sites where the Sc atoms incorporate substitutionally into the Ga sites. The growth of Mn x Sc 1 x N films, with x = , has been performed at a substrate temperature of 500 C. A rocksalt structure is observed for the Mn x Sc 1 x N films. Magnetic measurements preformed on the Mn 0.03 Sc 0.97 N film show ferromagnetic with a T C 50 K. As the manganese concentration is increased to x = 0.05, the ferromagnetism is reduced. The growth of heterostructures c-gan(001)/scn(001)/mgo(001) and ScN(001)/c- GaN(001)/MgO(001) adopt a cubical symmetry of the MgO(001) substrate.

3 Constantin, Costel. Ph.D. November Department of Physics & Astronomy Growth, Structural, Electronic and Optical Characterization of Nitride Semiconductors Grown by rf-plasma Molecular Beam Epitaxy (98pp.) Director of Dissertation: Arthur R. Smith This project investigates the growth, optical, electronic, surface, magnetic and bulk properties of scandium gallium nitride on Sapphire(0001), manganese scandium nitride on MgO(001), heterostructures of cubic gallium nitride and scandium nitride on Mg(001), and chromium nitride on MgO(001) grown by radio frequency molecular beam epitaxy. The growth of Sc x Ga 1 x N films has been performed at a substrate temperature of 650 C. The diffraction and optical experiments confirm the existence of two main regimes of growth; for high Sc concentration (x 0.54), a rocksalt crystal structure is obtained. For low x (x 0.17), a wurtzite-like crystal structure is observed with local lattice distortions at the sites where the Sc atoms incorporate substitutionally into the Ga sites. The growth of Mn x Sc 1 x N films, with x = , has been performed at a substrate temperature of 500 C. A rocksalt structure is observed for the Mn x Sc 1 x N films. Magnetic measurements preformed on the Mn 0.03 Sc 0.97 N film show ferromagnetic with a T C 50 K. As the manganese concentration is increased to x = 0.05, the ferromagnetism is reduced. The growth of heterostructures c-gan(001)/scn(001)/mgo(001) and ScN(001)/c- GaN(001)/MgO(001) adopt a cubical symmetry of the MgO(001) substrate.

4 The zincblend c-gan grown atop of ScN(001) shows a smoother surface (predominantly 2D growth) as compared to the rocksalt ScN(001) grown on atop of c-gan(001). The growth of stoichiometric CrN(001) films is performed at a substrate temperature of 450 C. A novel growth method of highly crystalline stoichiometric CrN(001) films has been proposed. The room temperature scanning tunneling microscopy together with resistivity versus temperature experiments reveal the electronic behavior of CrN(001) films to be metallic below T N 270 K, and semiconductor above T N. Approved: Arthur R. Smith Associate Professor of Physics & Astronomy

5 To Mia and Anca.

6 Acknowledgments I want to thank my parents Maria and Niculae Constantin who worked very hard to bring the food home, and who helped me to achieve the pre-doctoral education. I am deeply indebted to my supervisor Dr. Arthur Reed Smith whose help, stimulating suggestions and encouragement helped me in all the time of research for and writing of this thesis. I want to thank my colleagues Muhammad Haider, Rong Yang, Hamad Al-Brithen, Dr.Haiqiang Yang, and Dr.Erdong Lu, for the nice collaborations and conversations we had. I would like to thank Steven Diehl for helping me with the thesis class file. I would like to acknowledge Randy, Doug, and Roger, for their patience to turn our scribbles into nice vacuum parts we needed for our lab. Finally, I would like to thank all the people from the Physics Department for their genuine kindness and professionalism.

7 7 Table of Contents Abstract Dedication Acknowledgments List of Tables List of Figures Introduction Instruments Growth Methods Reflection High Energy Electron Diffraction (RHEED) Scanning Tunneling Microscopy (STM) Atomic Force Microscopy (AFM) X-ray Diffraction (XRD) Resistivity versus Temperature (ρ-t) Molecular Beam Epitaxial Growth of Scandium Gallium Nitride Films Introduction Experimental Procedure Results and Discussions Summary Composition-Dependent Structural Properties in ScGaN Alloy Films: A Combined Experimental and Theoretical Study Introduction Experimental Methods Experimental Results Summary Mixing Rocksalt and Wurtzite Structure Binary Nitrides to Form Novel Ternary Alloys: ScGaN and MnGaN Introduction Experimental Details Results and Discussion Summary Investigation of MnScN(001)/MgO(001) Grown by Molecular Beam Epitaxy: A Possible Dilute Magnetic Semiconductor

8 6.1 Introduction Experimental Details Results and Discussions Summary Growth of Heterostructures c-gan(001)/scn(001)/mgo(001) and ScN(001)/c- GaN(001)/MgO(001) by Molecular Beam Epitaxy Introduction Experimental Procedure Results and Discussion Summary Metal/Semiconductor Phase Transition in Chromium Nitride(001) grown by rf-plasma-assisted Molecular-Beam Epitaxy Introduction Experimental Details Results and Discussions Summary Conclusions Bibliography Appendices A List of Publications B List of Contributed Talks in Conferences B.1 I was the presenter B.2 I was a co-author C List of Contributed Posters in Conferences C.1 I was the presenter

9 9 List of Tables 3.1 Flux ratio r = J Sc /(J Sc + J Ga ), Sc composition x = N Sc /(N Sc + N Ga ), XRD 0002 peak amplitude, XRD 0002 FWHM, RHEED FWHM, measured c/a ratio, expected c/a ratio (based on Ref. [27]), and c/a ratio for ideal wurtzite structure (based on Ref. [24] for different Sc x Ga 1 x N samples). Sample 1-4 correspond to regime I, samples 5-7 to regime III XRD peak amplitudes and FWHM for ScGaN and peak amplitudes for MnGaN films for 0002, 0004, and 0006 peaks CrN 002 FWHM, lattice constants a, a, and a 0, for 3 different film thicknesses of CrN(001)/MgO(001)

10 10 List of Figures 2.1 (a) Schematic diagram of reflection energy electron diffraction (RHEED); (b) RHEED pattern of a MgO(001) flat surface; (c) RHEED pattern of a CrN(001) rough surface Ewald sphere in 2D A schematic diagram showing the Ewald sphere construction intersecting a series of lattice rods lying perpendicular to the plane of the sample. Where L is the distance between the sample and phosphorous screen; s is the distance between the streaks seen on the screen; and, h = 2π / a is the distance between reciprocal lattice rods Simplified electronics schematic model of a Scanning Tunneling Microscope (STM) operated in constant current mode A schematic diagram of the AFM Schematic diagram of x-ray diffraction Schematic setup for Resistivity versus Temperature measurement (a) RHEED pattern of wurtzite GaN (000 1) along [11 20]; (b)-(d) RHEED patterns of Sc x Ga 1 x N for low Sc concentration along same azimuth as (a); (e) RHEED pattern of rocksalt ScN (111) along [1 10]; (f)-(h) RHEED pattern of Sc x Ga 1 x N for high Sc concentration along same azimuth as (e) (a) Side view diagram of GaN [000 1] along [11 20]; (b) reciprocal space map corresponding to (a), 3-index notation corresponds to ν 1 ν 2 ν 3 which label the reciprocal-lattice points; (c) side view diagram of ScN [111] along [1 10]; (d) reciprocal space map corresponding to (c) using similar 3-index notation. Size of dots is proportional to their intensity; (e) schematic model of ScGaN for low x regime showing local distortions of the bond angle θ b (a) XRD of a representative spectrum with x = 0.05 showing the two sapphire peaks [0006 & 00012] and three ScGaN peaks [0002, 0004, & 0006]; (b) sapphire 0006 peaks for the films with x = 0.05; 0.14; 0.17; (c) ScGaN 0002 peaks at 2θ 34.5 for the films with x = 0.05; 0.14; (a) The lattice spacing a and c vs. x for low Sc concentration

11 3.5 (a) Optical absorption measurements of all the films; numbers are the Sc/(Sc + Ga) flux ratios r, and tangent line for r = 0.89 exemplifies method of obtaining E t. Small peaks at 3.4 ev for large x are instrumental in origin and of no importance here. (b) deduced E t values versus r and x for all the films showing the 3 different regions. Most error bars are too small to be seen µm 2µm AFM images of Sc x Ga 1 x N/GaN/Sapphire(0001) with (a) x = 5%, and (b) x = 17%] HRTEM images of Sapphire, GaN, and Sc x Ga 1 x N; (a) 30nm 30nm HRTEM image showing well ordered GaN and Sapphire crystals; (b) 30nm 30nm HRTEM image showing GaN and Sc x Ga 1 x N with x = 5% ; (c) 15nm 15nm HRTEM image showing GaN and Sc x Ga 1 x N with x = 17% [the black arrows show the interface between GaN and Sc x Ga 1 x N layers for figures (b) and (c)] HRTEM and FFT images of Sc x Ga 1 x N; (a) 35nm 35nm HRTEM image of Sc x Ga 1 x N with x = 5%; (b) FFT image of (a); (c) 9nm 9nm HRTEM image of Sc x Ga 1 x N with x = 17%; (d) FFT image of (c) RBS images of Sc x Ga 1 x N/Sapphire(0001) with x = 5% [figure (a)], and x = 17% [figure (b)] Images of Sc x Ga 1 x N crystal without stacking faults [figure (a)], and with stacking faults [figure (b)] RHEED pattern of ScGaN and MnGaN along [1120. The Sc incorporation measured with the RBS in (a) is found to be x 5%, and also for the case of MnGaN (b), Mn incorporation is x 5 % RBS measurements. (a) MnGaN; and (b) ScGaN each with x 5 % (a) & (b) XRD spectra of ScGaN and MnGaN with x 5% showing the two sapphire peaks [0006 & 00012] and the three ScGaN / MnGaN peaks [0002, 0004, & 0006] ScGaN lattice spacing c vs. a for low Sc concentration Schematic model of ScGaN for low Sc composition showing local distortions of the bond angle θ N Sc N and θ N Ga N XRD spectrum of a MnScN(001)/MgO(001) sample with 5% Mn concentration SQUID measurement of MnScN sample with 3% Mn concentration SQUID measurements of two control samples of ScN [# 71 and # 73], and two samples of MnScN with Mn concentration of 3% and 5%, respectively XRD θ-2θ scan of c-gan/scn/mgo(001) heterostructure XRD θ-2θ scan of ScN/c-GaN/MgO(001) heterostructure

12 7.3 RHEED images of c-gan(001)/scn(001)/mgo(001); (a) MgO(001) RHEED pattern before the growth of ScN; (b) ScN(001) RHEED pattern at a sample temperature of 850 C; (c) and (d) GaN(001) RHEED patterns at the sample temperature of 750 C, and 27 C, respectively RHEED images of ScN(001)/c-GaN(001)/MgO(001); (a) c-gan(001) RHEED pattern at the sample temperature of 750 C; (b) ScN(001) RHEED pattern at a sample temperature of 850 C AFM images of c-gan(001)/scn(001)/mgo(001). The gray scale and the root mean square roughness of a, b, and c are 332 Å and 9.87 Å, 344 Å and 12.0 Å, and 284 Å and 9.75 Å, respectively AFM images of ScN(001)/c-GaN(001)/MgO(001). The gray scale and the root mean square roughness of (a), (b), and (c) are 206 Å and 5.91 Å, 218 Å and 4.52 Å, and 220 Å and 14.9 Å, respectively Sequence of RHEED patterns during the CrN growth on MgO(001) substrate. (a) MgO substrate after the heating; (b) CrN buffer layer; (c) CrN layer grown at 450 C; (d) After the temperature of the substrate goes up to 650 C with the plasma on for about 10 minutes; (e) The final layer of CrN XRD results for CrN(001)/MgO(001). Lorentzian fittings of K α1 and K α2 of MgO 002 and CrN 002 peaks; (inset) Rocking curve of CrN 002 peak RT STM of CrN(001)/MgO(001). (a) 250Å 250Å STM image of CrN showing smooth terraces. V S =+1 V, I T =0.2 na; (b) 25Å 25Å atomic resolution image showing the 1 1 square lattice periodicity and the fcc conventional unit cell of the CrN lattice. V S =+0.7 V, I T =0.2 na; (c) The features at A, B, and C are LTD s. The features 1-5 are CrN vacancy islands. V S =+0.6V, I T =0.6 na STS of CrN(001)/MgO(001). I-V S curve (right axis) taken on similar image area as in Fig. 3(c); together with NC-V S (left axis) derived from I-V S curve ρ vs. T measurement showing metallic (I) and semiconductor (II) regions. There is evident thermal hysteresis with a width of 20 K. The inset shows the linear fit of ln (ρ/ρ 0 ) vs. 1/T for region (II), the straight line represents the linear fit to the data having slope = (E g /2k B )

13 13 Chapter 1 Introduction When I think myself about science and research... I think of scuba diving dip in the ocean that nobody else has ever wonder to go to. And the chances are... that you will get to discover new creators which might puzzle the common sense, but they would be totally adapted to their environment. With this in mind, in this thesis, I explored the combination between wurtzite (hexagonal) and rocksalt (cubic) binaries nitrides to find exotic novel ternary alloys. I explored the magnetic doping of rocksalt semiconductors to form new dilute magnetic semiconductors. I explored new heterostructures between zincblend semiconductors and rocksalt semiconductors. I explored the interesting phase transitions of hexagonal and stoichiometric rocksalt materials. In the last couple of years, binary III-nitrides semiconductors such as wurtzite GaN, AlN, and InN, have received a lot of attention due to their high thermal conductivity, high breakdown field, and electron mobility properties[1, 2]. In addition, the wide range of bandgaps which they have [e.g. 6.2 ev (AlN), 3.4 ev (GaN), and 1.9 ev (InN)], made their isocrystalline ternary alloys, such as Al x Ga 1 x N, and In x Ga 1 x N, suitable for optoelectronic applications ranging from ultraviolet to visible spectrum range. Although, the combination of isocrystalline binaries to form isocrystalline ternary alloys have obvious advantages, would be interesting to see what happens in the case of combining non-isocrystalline binaries to form novel ternary alloys. Recently, Ranjan and Bellaiche predicted that the combination of wurtzite GaN with

14 14 rocksalt ScN would result in a novel ternary Sc x Ga 1 x N alloys, which would have a hexagonal crystal structure[3]. They also predicted that the Sc x Ga 1 x N alloys would exhibit: i) a wide range of bandgap; and ii) large electromechanical responses with the increase of the Sc concentration. Experimentally, not too much work has been done on this material. Little and Kordesch grew ScGaN by sputtering and reported amorphous or microcrystalline films and a linearly decreasing optical band gap (from 3.5 down to 2.0 ev) with increasing Sc concentration[4]. In the the chapter three, we report the growth of crystalline Sc x Ga 1 x N alloys on Sapphire(0001) substrate by molecular beam epitaxy, and in-depth bandgap engineering and crystal structure identification with the increase of Sc concentration between x = 0 to x = 1. In chapter four, we present more experimental evidences towards the support of the metastable hexagonal phase found at Sc x Ga 1 x N alloys with Sc concentration up to Here, we also present experimental observations at the boundary (with Sc concentration, x = 0.17) between the hexagonal/rocksalt crystal phase transition. Chapter five is dedicated to a structural comparison between two GaN alloys, mainly, Sc x Ga 1 x N and Mn x Ga 1 x N (both with x = 0.05). The motivation for the chapter six of this thesis came from the work of Al-Brithen et al. who grew ScMnN alloys by molecular beam epitaxy, the Sc concentration was varied between 0% and 100%[6]. Alloy type behavior is observed since the Vegard s law is obeyed. The structure is observed to be face-centered tetragonal rocksalttype crystal structure with Sc and Mn cations and N anions. Al-Brithen et al. also suggested the possibility of forming a dilute magnetic semiconductor for low Sc concentration ( 5). Hence, in chapter six, we present preliminary magnetic results for Mn x Sc 1 x N alloys with x = , grown on MgO(001) substrates. Chapter seven contains a structural comparison between the c-gan(001)/scn(001) and the ScN(001)/c-GaN(001) heterostructures grown of MgO(001) substrates. This zincblend c-gan to rocksalt ScN, and viceversa heterostructures, are important because, in order to create the spin injection mechanism, one needs heterostructure system between a semiconductor (e.g. c-gan) and a DMS (e.g. MnScN) system. Since the Mn concentration is only few percents into the ScN host lattice, the structure that the MnScN adopts is the same fcc rocksalt type as ScN. Another advantage

15 15 of combining the c-gan and ScN structures is the lattice mismatching of 0.64 which is very small. Therefore, there will be less defects propagating into the heterostructure. Within the frame of the this thesis purpose, in chapter eight, we present structural, surface and electronic properties of stoichiometric rocksalt CrN(001)/MgO(001) films. The motivation that made us start this project was; first, the challenge to obtain very high crystalline stoichiometric films of CrN(001)/MgO(001), since, so far, all the CrN films were grown by sputtering, which is more aggressive technique as compared to the method we used, radio-frequency plasma molecular beam epitaxy[7, 10]; and, second, the controversial electronic properties reported in the literature up to now, some groups reported that CrN at room temperature is a weak metal or metal[8, 7]. While other groups reported that the CrN at room temperature is a semiconductor[9, 10, 11]. Chapter nine contains the conclusion of the work which was presented in this thesis. Three appendices A, B, and C are included at the end, one for each, paper contribution, contributed talks in conferences, and contributed posters.

16 16 Chapter 2 Instruments 2.1 Growth Methods All our films are grown by radio-frequency plasma assisted molecular beam epitaxy with N 2 as a plasma source. The home built growth chamber is kept under ultra high vacuum (usually with a ultra high vacuum of torr) which is provided by a cryostat that uses 4 He. The chamber pressure is monitored by an ion gauge which has a measuring range of torr. The growth chamber has four ports as sources for the metal elements, such as Ga, Cr, Mn, and Sc. We used either effusion cells or e-beam cells for the evaporation of Cr and Sc, whereas for the Mn and Ga we used only effusion cells. The evaporated flux is determined by a crystal thickness monitor which is mounted in-situ of the growth chamber. The nitrogen pressure of the radio-frequency plasma source is controlled through a valve, and we usually use a nitrogen pressure of to torr for most of our growths. We usually used MgO substrates with crystal orientations of (001) and (110) for growth of materials with cubic symmetry. In the case of growth of materials with hexagonal symmetry, we used MOCVD GaN/Sapphire(0001) Ga-polar and GaN/Sapphire(000 1) N-polar. The size of the substrates is 10mm 10mm.5mm. The substrates are cleaned with acetone and isopropanol, after which they are mounted on a sample holder made of molybdenum. Following this, the samples are introduced into the chamber through a load-lock chamber which has a base pressure of 10 7 torr,

17 17 then the sample is transferred to the UHV analysis chamber through a gate valve. The analysis chamber is pumped with a ion pump and a titanium sublimation pump. The ion pump can reach a pressure up to torr. To transfer the sample in-situ from the analysis chamber to the growth chamber, a transporting rod is used which passes through a gate valve between the two chambers. The sample is then attached to the growth stage in the growth chamber. The growth stage can be manipulated in X, Y, and Z directions as well as both rotational angles, θ and φ. The desired growth temperature for the sample is achieved by using a graphite filament which is mounted behind the sample onto the growth stage. Finally, the sample growth temperature is measured in two ways, i) by using a thermocouple of type C which is in contact with one molybdenum temperature shield that is placed under the graphite filament, and ii) by an ex-situ optical pyrometer. 2.2 Reflection High Energy Electron Diffraction (RHEED) The surface condition of the film during the growth is monitored in-situ with reflection high energy electron diffraction (RHEED) experiment. The geometry of RHEED is shown in Fig. 2.1(a). The STAIB electron gun sends an accelerated electron beam (usually 20 kev) which is incident on the surface with a glancing angle (β 3 ). The electrons are diffracted by the crystal structure of the sample and then impinge on a phosphor screen mounted opposite to the electron gun into the growth chamber. Because of the fact that the energetic electron beam has an incident angle β of only few degrees on the sample, the strong interaction between the electron beam and the electronic system of the crystal limits the penetration depth to only a few Å. Therefore we can consider that the electrons are diffracted by a 2D structure, rather than a 3D structure. Figure 2.2 shows the Ewald sphere (in 2D) intersecting reciprocal lattice rods where the Laue diffraction criteria is satisfied ( k = G), with k being the difference between the incident wavevector k I and the reflected wavevector k R, and with G being the reciprocal lattice vector. The RHEED diffraction phenomenon can even be seen better in figure 2.3, where L is the distance between the sample

18 18 and the phosphorous screen; h is the distance between reciprocal lattice rods; and, s is the distance between diffraction streaks on the screen. The wavelength of the incident beam is λ = Å (for a kinetic energy of electrons of 20 KeV), using this to calculate the incident wavevector we obtain k I = 100 Å 1, which is the radius of the Ewald sphere. If we consider a ScN system with a fcc crystal structure and a lattice constant a = Å, one can calculate a distance between the reciprocal lattice rods to be h = 2.8 Å 1. Now, one can see that the ratio of k I to h is 36 which means that the ewald sphere is very large compared to the distance between the reciprocal lattice rods. Therefore, the Ewald sphere will only intersect only a couple of rods [ e.g. (-10); (00); (10)]. The resulting RHEED pattern is a series of streaks and spots. Qualitatively, if a surface is atomically flat, then a series of spots extended vertically is observed by RHEED pattern which can be seen in Fig. 2.1(b). The streakiness effect in the vertical direction of the diffraction spots observed in the figure 2.1 is due to the fact that we have a small range of wavelengths incident on the screen and that the beam is not and infinitesimally thin (which means that the Ewald sphere will have a finite thickness)[12]. If the surface is rough, then the RHEED pattern shows a spot-like behavior like in figure 2.1(c). The spot-like behavior of the RHEED pattern is due to the transmission of the electrons through 3-dimensional islands formed on the surface. One can easily find information about the crystal unit cell by looking at the distance between the streaks and dots in Fig. 2.1(c). By measuring the FWHM of the diffraction streaks one can quantify the in-plane grain size or correlation length - defined as the length scale over which the atomic positions are quantitatively correlated. Most of the softwares existent on the market today can automatically measure coherence length during growth, provided that there is a prior calibration to a known sample.

19 19 (a) Electron Diffraction Patterns Electron Gun Electron Beam CCD Camera Sample Diffracted Beam Screen (b) MgO 100 (c) CrN 100 d in-plane d out-of-plane Figure 2.1: (a) Schematic diagram of reflection energy electron diffraction (RHEED); (b) RHEED pattern of a MgO(001) flat surface; (c) RHEED pattern of a CrN(001) rough surface.

20 20 k I = 2 - incident vector - reflected vector k R G - reciprocal lattice vector Ewald sphere in 2D k R 2 G k I Figure 2.2: Ewald sphere in 2D

21 21 Ewald sphere k I k I = 2 - incident vector - reflected vector k R G - reciprocal lattice vector (b) Sample Side view Reciprocal lattice rods k R e - s L h h [00] streak (a) Top view Phosphorous screen Figure 2.3: A schematic diagram showing the Ewald sphere construction intersecting a series of lattice rods lying perpendicular to the plane of the sample. Where L is the distance between the sample and phosphorous screen; s is the distance between the streaks seen on the screen; and, h = 2π / a is the distance between reciprocal lattice rods.

22 Scanning Tunneling Microscopy (STM) At the end of growth, after assuring that the surface smoothness is suitable for the STM experiment, the sample is transferred in-situ into the analysis chamber. The STM is host into the analysis chamber. Figure 2.4 shows the electronic setup of the STM used in constant current mode. The principle of STM is rather simple. It can be compared with that of a gramophone (an old-fashion record player). Just as a gramophone, the STM uses a sharp Tip to map the surface topography in real space. But the difference between a gramophone and an STM is that the STM Tip does not touch the surface. The Tip is brought to within 1 nanometer distance of the conducting surface which needs to be scanned. If a positive voltage is applied between the Tip and the surface then the electrons would start to quantum tunnel from the Tip to the sample. The tunneling current is very low, on the order of pico and nanoamperes, which implies that the tunneling process is very difficult. Also the tunneling current is exponentially proportional with the distance [I T k exp( K d), where k, and K are two different constants]. As a rule of thumb, for every 2.5 Å increase in the distance between the Tip and the sample the current becomes a factor of 1000 lower. The feedback electronics in figure 2.4 is used to keep the tunneling current constant. Going into even more details, the preamplifier [Fig. 2.4] takes the tunneling current I t and converts it into a voltage of typically V t = 1 V. In the next step, this voltage V t is amplified logaritmically. Because the tunneling current and therefore the voltage V t is an exponential function of the distance d between the tip and the surface. The logarithm of the voltage V t is a measure for the distance between the Tip and the surface. The next step is to subtract a chosen reference voltage V R from the voltage V t. If the difference between the 2 voltages is zero, then the Tip is at the right distance. If the difference between the two voltages is nonzero, then voltage is applied to the piezo tube in z-direction in order to approach or to retract the Tip from the surface. The STM image is obtained by plotting XY-scan versus Z-distance. Another useful information about the electronic structure of the surface can be obtained by doing Scanning Tunneling Spectroscopy (STS). To do this, the electronic

23 23 XY-Ramp X high-voltage amplifier Y high-voltage amplifier Computer Feedback Reference voltage high-voltage amplifier Z Inch-worms log preamplifier I Tip X,Y,Z Piezo Tube Path followed by the Tip Figure 2.4: Simplified electronics schematic model of a Scanning Tunneling Microscope (STM) operated in constant current mode. feedback box has to be turned off, and a set of Voltages is applied between the Tip and the surface. The Tip is set at a specific X, and Y coordinates, while the tunneling current is measured at every applied Voltage. The I-V curve gives information about the filled and empty states at the surface of the material.

24 Atomic Force Microscopy (AFM) We usually use Atomic Force Microscopy (AFM) to look at larger surface morphology. The instrument we use is made by Park Scientific Instruments, and it utilises a sharp Tip which moves over the surface of a sample in a raster scan. The Tip is at the end of a cantilever which bends in response to the Van der Waals force between the tip and the sample. In Figure 4.1 is shown how the incident laser beam is deflected off the cantilever and then by measuring the difference signal (A-B), changes in the bending of the cantilever can be measured. The movement of the Tip or sample is performed by an extremely precise positioning device made from piezo-electric ceramics, most often in the form of a tube scanner [Fig. 4.1]. The plotting of distances in x-, y-, and z- directions create the AFM image. Laser Source Reflected Laser Beam B A Incident Laser Beam Photo-Detectors Tip Cantilever Sample x, y, z Piezo Tube Scanner Path followed by the Tip Figure 2.5: A schematic diagram of the AFM.

25 X-ray Diffraction (XRD) We use x-ray diffraction to find bulk properties of our mbe-grown semiconductors and metals. We use a Rigaku machine which has a Cu source. From the Cu source both emissions can be seen, K α1 and K α2 with the wavelengths of λ 1 = 1.54 Å and λ 1 = Å, respectively. Figure 2.6 shows how the incoming x-ray beam K I is reflected from two planes of atoms. From the difference of path between K I and K R the atomic spacing d can be calculated from the formula 2.1. Where n is an integer number for constructive interference, θ is the diffraction angle where the constructive interference occurred, and for λ we usually use a average wavelength value of Å. d = nλ 2sin(θ) (2.1) K I K R d d sin( d sin( Figure 2.6: Schematic diagram of x-ray diffraction 2.6 Resistivity versus Temperature (ρ-t) We used resistivity versus temperature (ρ-t) measurements in order to look at the electronic properties of our CrN films. We used van der Pauw [13] method

26 which is basically a 4-point probe technique. The home-made setup is presented in a schematic diagram in figure 2.7. Four contacts of Indium are soldered at the corner of a 3mm 3mm sample, and I have designed a labview program which controls the Keithley and the lock-in amplifier. In order to obtain resistivity versus temperature in every point, we need to determine 2 different resistances across different contacts [R A = V 43 /I 12, and R B = V 14 /I 23 ]. The lock-in amplifier generates an ac current of 1 µamp across contacts 1 and 2 (I 12 ), and then it measures the output voltage across contacts 4 and 3 (V 43 ), therefore obtaining resistance R A. Similarly is done to obtain resistance R B. The sheet resistance can be calculated by using formula 2.2, where t is a geometrical factor, which for the case of a square sample it can be assumed to be unity. The final resistivity can be calculated using formula 2.3, where d is the thickness of the material. R S = π ln 2 26 R A + R B t (2.2) 2 ρ = R S d (2.3)

27 27 Computer Keithley Thermocouple Lock-in Amplifier R = 1 M Indium contacts 3 1 Figure 2.7: Schematic setup for Resistivity versus Temperature measurement

28 28 Chapter 3 Molecular Beam Epitaxial Growth of Scandium Gallium Nitride Films 3.1 Introduction In this chapter, we explore alloy formation in ScGaN using as a method of growth rf molecular beam epitaxy over the Sc fraction range x = 0-100%. The two binaries that are combined to form ScGaN alloys is ScN and GaN. ScN is known to be a rocksalt (cubic) semiconductor with an indirect bandgap from Γ X of 1 ev and a direct transition E t at the X point of ev [14, 15, 16, 17, 18, 19]. GaN is a wurtzite (hexagonal) semiconductor with an direct bandgap of 3.4 ev and lattice constants a = Å, and c = Å. Given the fact that the Sc-Sc spacing in ScN (3.183 Å) is very close to the c-plane Ga-Ga spacing in GaN (3.189 Å), it is interesting to investigate the possibility of creating novel lattice-matched crystalline ScGaN ternary alloys [20]. It is thus important to consider what kind of bonding Sc will have in GaN. Takeuchi et al. predicted a metastable wurtzite phase (w-scn) for ScN [21]. However, recently Farrer and Bellaiche have found that w-scn should be unstable and that instead a layered hexagonal phase having nearly 5-fold coordination, denoted h-scn, which can be arrived at by flattening the bilayer of the wurtzite structure, should be metastable [22, 3]. In fact, such a structure has also been predicted to exist for MgO (h-mgo) [23]. Yet, aside from the report of Little & Kordesch, who grew ScGaN by

29 29 sputtering and reported amorphous or microcrystalline films and a linearly decreasing optical band gap (from 3.5 down to 2.0 ev) with increasing Sc concentration [4], little is known experimentally regarding the actual crystal structures of ScGaN alloys. For example, one might expect more than one growth regime for crystalline films. If so, then novel transitional properties, as suggested by Farrer and Bellaiche, might be observed at the boundary between the two different regimes [22]. 3.2 Experimental Procedure The ScGaN film growth is done using a custom MBE system that employs Ga and Sc effusion cells and a rf plasma N source with N 2 as the source gas. The sapphire(0001) substrates are first heated to 900 C. The N plasma source is applied using 500 W with a N 2 flow rate of 1.1 sccm (chamber background pressure = Torr). The r = J Sc /(J Sc +J Ga ) flux ratio is set to specific values with constant N flux while keeping J Sc + J Ga < J N (N-rich conditions). Growth of ScGaN begins following a 30 minute substrate nitridation and continues to a thickness of nm. The substrate temperature is 650 C. The growth is monitored in-situ by RHEED using a 20 kev e -beam, and the films are studied ex-situ by X-ray diffraction (XRD) with Cu Kα X-rays, optical absorption (OA), and Rutherford backscattering (RBS). A series of ScGaN films on sapphire(0001) with r in the range between 0 and 1 have been grown and analyzed. Presented in Fig. 3.1 are the RHEED patterns of 8 such ScGaN films, including the 2 endpoints at r=0 [GaN(000 1)] and r=1 [ScN(111)]. The direction of the RHEED beam is along [11 20] for GaN and correspondingly [1 10] for ScN. 3.3 Results and Discussions Calculation of the RHEED patterns is accomplished by considering the structural models. Figure 3.2(a) shows a side-view model of the [000 1]-oriented wurtzite GaN bi-layers as viewed along [11 20], and Fig. 3.2(b) shows the associated reciprocal space map with the spots labeled by the reciprocal space indices ν 1 ν 2 ν 3. The intensity of

30 30 the diffraction spots with ν 1 odd, calculated from the structure factor for wurtzite, alternates along the k z direction, and the spots with ν 1 = 0 and ν 3 odd have zero intensity. Clearly the RHEED pattern of GaN(000 1) shown in Fig. 3.1(a) is in good agreement with the calculated reciprocal space map, and the measured ratio of S in,gan to S out,gan for GaN is 1.92, which agrees well with the expected value c/( 3a/2) = Å/2.762 Å = Fig. 3.2(c) shows the side-view model of ScN(111) as viewed along [1 10] together with the corresponding reciprocal space map shown in Fig. 3.2(d). The spot intensities are calculated from a superposition of the structure factors of two inequivalent 111- oriented fcc grains (note that fcc is not 2-fold symmetric about 111, and for a single grain the diffraction pattern is asymmetrical about the 00 rod). Clearly the RHEED pattern of ScN(111) shown in Fig. 3.1(e) is in excellent agreement with the calculated reciprocal space map in Fig. 3.2(d), and the ratio of the lateral spot spacing S in,scn to S out1,scn is measured to be 2.86, in good agreement with the expected value of Comparing the rocksalt and wurtzite patterns, we note that the measured spacing S out2,scn [see Fig. 3.1(e)] is very close to 2 the measured spacing S out,gan [see Fig. 3.1(a)], due to the fact that the Sc layer spacing a/ 3 = Å in ScN is very close to the Ga layer spacing c/2 = Å in GaN. Yet clearly the RHEED patterns of wurtzite and rocksalt are easily distinguishable. Shown in Figs. 3.1(b-d) are the RHEED patterns for ScGaN alloy films having r in the range 0<r Over this range of r, the RHEED patterns are in good agreement with the wurtzite reciprocal space map. However, we note that the structure factor for the layered hexagonal phase is identical to that of wurtzite. To determine if the ScGaN sample with low Sc composition is consistent with wurtzite or not, it is necessary to consider the measured lattice constants a and c (discussed below). As can be seen in Figs. 3.1(f-h), the RHEED patterns for ScGaN films having r in the range 0.54 r<1 are in good agreement with the rocksalt pattern. At r = 0.54, a very diffuse and ring-like rocksalt pattern is seen, suggesting a polycrystalline rocksalt film [compare with Fig. 3.2(d)]. A less polycrystalline rocksalt pattern is evident for r = Also for r = 0.89, the RHEED pattern, although ring-like, clearly shows the

31 31 wurzite / hexagonal rocksalt / cubic a) GaN (0001) e) ScN (111) S out, GaN S in, GaN S out-2, ScN S in, ScN S out-1, ScN b) J Sc/(J Sc+J Ga) = f) J Sc/(J Sc+J Ga) = 0.54 c) J Sc/(J Sc+J Ga) = 0.21 g) J /(J +J ) = 0.78 Sc Sc Ga d) J Sc/(J Sc+J Ga) = 0.29 h) J Sc/(J Sc+J Ga) = 0.89 Figure 3.1: (a) RHEED pattern of wurtzite GaN (000 1) along [11 20]; (b)-(d) RHEED patterns of Sc x Ga 1 x N for low Sc concentration along same azimuth as (a); (e) RHEED pattern of rocksalt ScN (111) along [1 10]; (f)-(h) RHEED pattern of Sc x Ga 1 x N for high Sc concentration along same azimuth as (e).

32 32 uc (a) Wurzite 3 2 a c (c) Rocksalt a N Ga Sc a 3 3a (b) S out, GaN S in, GaN (d) 104 S out-1, ScN 105 S in, ScN S out-2, ScN (e) Sc Ga b N Figure 3.2: (a) Side view diagram of GaN [000 1] along [11 20]; (b) reciprocal space map corresponding to (a), 3-index notation corresponds to ν 1 ν 2 ν 3 which label the reciprocal-lattice points; (c) side view diagram of ScN [111] along [1 10]; (d) reciprocal space map corresponding to (c) using similar 3-index notation. Size of dots is proportional to their intensity; (e) schematic model of ScGaN for low x regime showing local distortions of the bond angle θ b.

33 33 rocksalt-type spot positions. The results suggest that each monocrystal has rocksalt structure for growth at large Sc composition, consistent with predictions[22, 24]. Incorporation x N Sc /(N Sc + N Ga ), where N Sc and N Ga are the number of Sc and Ga atoms within a given volume respectively, was obtained using RBS; the results are given in Table 3.1. Whereas r and x are about the same for the cubic regime (r 0.54), for the hexagonal regime (r 0.29), x is consistently smaller than r, suggesting that the sticking coefficients S Ga and S Sc are different at low r values but similar at high r values. This behavior is reasonable given that the N-polar (and N-terminated) wurtzite surface would have a single dangling bond per N atom [note that the more stable surface structure (in Ga-rich conditions) for GaN(000 1) would have 1 layer of Ga atoms atop the N atoms (Ga adlayer) with an even lower surface diffusion barrier)][25, 26]. In comparison, from the model of Fig. 3.2(c), the (111)-oriented, N-terminated rocksalt surface would have 3 dangling bonds per N atom, resulting in larger surface diffusion barrier of both Sc and Ga atoms, rougher growth mode, and larger sticking coefficients for both Sc and Ga. This agrees with the RHEED patterns which are more spotty for the cubic regime compared to the hexagonal regime. The lattice constant a is directly obtained from the RHEED pattern using a peak fitting program. The RHEED calibration is performed using a GaN substrate grown by metal-organic chemical vapor deposition. The resulting a values are plotted vs. r and x in Fig. 3.4(a) for the small x values, where it is seen that a increases with both r and x. Crystallinity and out-of-plane lattice constant information was determined using XRD, and shown in Fig. 3.3 are results for ScGaN layers with low x. Figure 3.3(a) shows the entire XRD spectrum ( ) for the film with x = 0.05 where 5 peaks are seen - 2 sapphire peaks (0006 and 00012) and 3 ScGaN peaks (0002, 0004, and 0006). The spectra are corrected so that both sapphire peaks give the same lattice constant (c = Å); the alignment of the sapphire 0006 peaks for films with x = 0.05, 0.14, and 0.17 is presented in Figure 3(b). The observed ScGaN peak shift for the same 3 samples is shown in Fig. 3.3(c); compared to x = 0, the 0002 peak shifts to the left for x = 0.05 and 0.14 but interestingly shifts slightly back to the right for x = 0.17, from which the c-spacings are calculated using Gaussian peak fitting and

34 34 Bragg s law to be Å, Å, and Å, respectively, compared to Å for x = 0. The c values are plotted in Fig. 3.4(b) vs. r and x. Over the range 0<x<0.17, a increases by a net 0.08 Å while c increases by only a net Å. Thus for the small x range, the Sc x Ga 1 x N lattice expands predominantly within the c-plane. Such anisotropic expansion as well as the decrease of c between x = 0.14 and 0.17 would not be expected in the case of alloying of iso-crystalline binary compounds. The anisotropic expansion suggests the following picture, as illustrated by the schematic model shown in Fig. 3.2(e): at the location where a Sc atom substitutes for a Ga atom, a local lattice distortion occurs in which the N-Sc-N bond angle θ b (having one leg in the c-direction) reduces compared to the 108 for a wurtzite bond (in ideal h-scn, θ b = 90 ). In other words, the internal parameter u [see Fig. 3.2(a)] varies within the alloy away from the w-gan value [22]. For h-scn (x = 1), Farrer et al. have predicted a = 3.66 Å, c = 4.42 Å, and c/a = [22]. Thus a can potentially locally increase by up to 14.77% of a GaN = Å, depending on x; at the same time, c would locally decrease. Further experimental evidence for the low x regime points to the same model. First, we note substantial broadening of the RHEED diffraction lines with increasing x (see FWHM values vs. x in Table 3.1). Second, we note the substantial intensity decrease and broadening of the 0002 ScGaN XRD peak with increasing x [see Fig. 3.3(c) and also Table 3.1]. Such behavior is indicative of an increased spread of lattice constants with increasing x, resulting in reduction of the long range order or of the maximum correlation length of the crystal. Very recently, Ranjan and Bellaiche have calculated the c/a ratio for an ideally ordered Sc x Ga 1 x N having x = 0.5 to be 1.55[27]. Considering the c/a ratio for GaN (x = 0) to be = Å/3.189 Å, linear interpolation gives decreasing values of c/a for increasing x values which can be compared with the measured c/a values of our low x samples. This data is presented in Table 3.1 which also shows values for the ideal wurtzite c/a ratios vs. x which are based on the recent report by Moreno- Armenta et al. [24] The data show that the change in c/a over the range x = 0 to 0.17 is for the wurtzite case versus for the model of Ranjan and Bellaiche.

35 a) Intensity (counts) (Ga,Sc)N 0002 Sapphire 0006 Sapphire x 6.78%Sc = 0.05 Whole Spectrum (Ga,Sc)N 0004 (Ga,Sc)N (degrees) b) c) %Sc x = 0.05 K %Sc x = 0.14 K %Sc x = %Sc x = %Sc x = %Sc x = 0.17 Intensity (counts) Intensity (counts) ideal 10 sapphire ScGaN c = (degrees) 2 (degrees) degrees 2 (degrees) Figure 3.3: (a) XRD of a representative spectrum with x = 0.05 showing the two sapphire peaks [0006 & 00012] and three ScGaN peaks [0002, 0004, & 0006]; (b) sapphire 0006 peaks for the films with x = 0.05; 0.14; 0.17; (c) ScGaN 0002 peaks at 2θ 34.5 for the films with x = 0.05; 0.14; 0.17.

36 lattice constant c (Angstroms) lattice constant a lattice constant c lattice constant a (Angstroms) x Figure 3.4: (a) The lattice spacing a and c vs. x for low Sc concentration. The latter is thus in better agreement with the experiment which finds a change in c/a over this range to be Extrapolating straight line fits of the measured a and c vs. x data to x = 1 results in a(x=1) = 3.60 Å and c/a = 1.45 for a hypothetical ScN in hexagonal phase. Thus we find that the extrapolated c/a ratio is significantly smaller than that predicted for w-scn (1.6); and, the extrapolated a is significantly larger than that predicted for w-scn (3.49 Å)[21], closer in fact to the value predicted for h-scn (3.66 Å)[22]. In fact, we note that a increases faster, and that c decreases, with x between 0.14 and Optical absorption measurements have been obtained for all the films grown with various r values. To obtain a value for E t, the quantity (hνα) 2 is plotted vs. hν, as shown in Fig. 3.5(a). Then E t is estimated from the energy intercept of a straight

37 37 Table 3.1: Flux ratio r = J Sc /(J Sc + J Ga ), Sc composition x = N Sc /(N Sc + N Ga ), XRD 0002 peak amplitude, XRD 0002 FWHM, RHEED FWHM, measured c/a ratio, expected c/a ratio (based on Ref. [27]), and c/a ratio for ideal wurtzite structure (based on Ref. [24] for different Sc x Ga 1 x N samples). Sample 1-4 correspond to regime I, samples 5-7 to regime III. ScGaN Sample flux ratio r composition x XRD peak Ampl XRD FWHM RHEED FWHM measured c/a c/a from Ref. [27] c/a from Ref. [24] line tangent to the curve near its inflection point. The resulting E t values are plotted vs. r and x in Fig. 3.5(b). The E t = 3.37 ev of wurtzite GaN is obtained at x = 0. The E t = 2.15 ev is obtained at r = 1 for ScN grown on MgO(001)[18]. As can be seen, three different regions, consistent with the RHEED results, can be distinguished : I) low Sc fraction (0< r <0.30); II) intermediate Sc fraction - transitional regime (0.30< r <0.54); and III) high Sc fraction (0.54< r <1). In both regions I and III, E t decreases monotonically with increasing x. For region I, extrapolating a straight line fit of the E t values to x = 1 (pure ScN) obtains an E t = 2.3 ev - significantly smaller than the value of 3.0 ev predicted for the E t of w-scn[21]. The extrapolation from low x however is probably not a good estimate of the E t of h-scn since layered hexagonal is not iso-crystalline with the low x regime (wurtzite with local N-Sc-N bond distortions). As RHEED indicates, rocksalt structure is observed for larger x (region III), and within region III, E t decreases linearly towards the rocksalt value of 2.15 ev at x = 1. Using a straight line fit to the E t values of region III and extrapolating to x = 0 yields

38 38 a) (h ) (10 ev cm ) (a*e)^2/10^ h Eg(eV) ( ev) b) Eg (ev) E (ev) t (I) (II) Sc / (Sc + Ga) flux ratio r RBS ratio x JSc/(JGa+JSc) Sc / (Sc + Ga) (III) Figure 3.5: (a) Optical absorption measurements of all the films; numbers are the Sc/(Sc + Ga) flux ratios r, and tangent line for r = 0.89 exemplifies method of obtaining E t. Small peaks at 3.4 ev for large x are instrumental in origin and of no importance here. (b) deduced E t values versus r and x for all the films showing the 3 different regions. Most error bars are too small to be seen.

39 39 an E t = 2.72 ev for a hypothetical GaN in rocksalt structure. Finally, the E t values in region II (encircled points) with r in the range show comparatively large variations, which are consistent with the expected instability of the crystal structure near the transition between the stable hexagonal and cubic regimes. 3.4 Summary To summarize, we have shown that the disparate ground state crystal structures of ScN and GaN lead to two distinct regimes of structural and optical properties. For both low x and high x, alloy-type behavior is observed. For x 0.54, rocksalt structure is found, in agreement with predictions[22, 24]. For small x up to 0.17, an anisotropic expansion of the ScGaN lattice is observed which is interpreted in terms of local lattice distortions of the wurtzite structure in the vicinity of Sc Ga substitutional sites in which there is a decrease of the N-Sc-N bond angle. This tendency toward flattening of the wurtzite bilayer is consistent with a predicted h-scn phase[22].

40 40 Chapter 4 Composition-Dependent Structural Properties in ScGaN Alloy Films: A Combined Experimental and Theoretical Study 4.1 Introduction In chapter 3 we observed experimentally that ScGaN alloys exhibit 3 regimes of growth [e.g. wurtzite (hexagonal), transitional regime, and rocksalt (cubic)] as the Sc concentration is varied between 0% to 100%. In this chapter, we quantitatively measure the variation of the a and c lattice constants with Sc fraction in the low Sc concentration regime (0-17%) which are found that can be well explained by the predictions of first-principles theory. In chapter 3, it was also found that lattice constant a increases 13 times as much as c over this range 0-17%, thus resulting in a decrease in the c:a ratio. This decrease compared the best with predictions based on a hexagonal ScN phase as the high x endpoint. We also explained that this anisotropic lattice constant expansion can be understood in terms of a model in which there occur local lattice distortions of the wurtzite structure in the vicinity of Sc Ga substitutional sites. In particular, as the out-of-plane N-Sc-N bond angle

41 41 decreases, the lattice constant a increases. And by geometry, the in-plane N-Sc-N bond angle should increase. For Sc-N and Ga-N bond lengths roughly equal, then this distortion would also result in a decrease of c. The slight experimentally observed increase in c for small x is attributed to the larger bond length of Sc-N (e.g Å in rocksalt ScN) in comparison to Ga-N (e.g Å in wurtzite GaN). However, in the chapter 3, neither the bond angles nor the bond lengths were directly determined (measured or predicted), which is necessary to prove the lattice distortion model. In this chapter, we report only the experimental part obtained for ScGaN with small Sc concentration only (Sc fraction up to 17%). I will also briefly describe the theoretical results performed by our theory collaborators, Nancy Sandler (Condensed Matter and Surface Science Program, Department of Physics and Astronomy, Ohio University, Athens, OH & Institut de Ciència de Materials de Barcelona CSIC, Campus UAB, Barcelona, Spain) and Pablo Ordejón (Institut de Ciència de Materials de Barcelona CSIC, Campus UAB, Barcelona, Spain). We present new experimental data resolving the ScGaN lattice along the growth direction and supporting the conclusion that the ScGaN lattice is a crystalline alloy containing substitutional Sc atoms. As well, we observe that, above a certain Sc fraction, there is evidence of defects, most likely in the form of stacking faults. 4.2 Experimental Methods The details about the ScGaN film growth is presented in chapter 3, the only difference in this study is that we also grow a buffer layer of GaN at a substrate temperature of 650 C, under Ga-rich conditions, and to a thickness of 50 nm. Similarly as in chapter 3, the growth is monitored in-situ by RHEED using a 20 kev e -beam. The films are measured ex-situ by XRD, atomic force microscopy (AFM), high resolution transmission electron microscopy (HRTEM), and Rutherford backscattering (RBS).

42 Experimental Results Shown in Fig. 4.1(a) is a 2µm 2µm AFM image of Sc x Ga 1 x N/GaN/Sapphire(0001) with x = 5%. The surface consists of closely packed hillocks. The root-mean-square (rms) surface roughness is 115 Å with a peak-to-valley height of 875 Å. These results suggests a 3D growth mode, which is consistent with growth under N-rich conditions. As we increased the Sc concentration to x = 17% [Fig. 4.1(b)], the root-mean-square (rms) surface roughness and the peak-to-valley height are 38.4 Å and 419 Å, respectively - less than that obtained for x = 5%. The growth mode, however, is still clearly 3D, and the reduction in roughness is attributed to an increase in surface diffusion barrier, resulting in smaller, more densely-packed grains. The roughness indicated in these AFM images is consistent with the results of RHEED studies in the same composition range (0-17%), which also indicated 3D growth mode for the ScGaN grown under these conditions[39]. Nonetheless the RHEED patterns indicated a crystalline surface. Figure 4.2 shows high-resolution TEM (HRTEM) images for Sc x Ga 1 x N/GaN/sapphire films with x = 5% and 17%. Figure 4.2(a) shows the interface between the well ordered GaN and sapphire layers. These layers show a high degree of crystalline perfection. The GaN layer in particular shows the good crystallinity consistent with the Ga-rich buffer layer growth conditions. The interface between the Sc x Ga 1 x N with x = 5% and GaN layers is shown in Fig. 4.2(b), with the interface indicated by arrows. As seen, at 5% Sc composition, there is no strong difference seen in the image between the GaN and Sc 0.05 Ga 0.95 N layers, showing that the 5% layer has fairly high crystalline quality with Sc substituting for Ga. However, a slight increase in the contrast variations within the Sc 0.05 Ga 0.95 N layer suggests not only the presence of the alloy but also the possibility of a slightly non-uniform distribution of Sc. As seen in Fig. 4.2(c), the interface between the 17% Sc-containing layer and the GaN layer is much more obvious. This is due to 1) increased contrast variations within the Sc 0.17 Ga 0.83 N layer; and 2) apparent random variations in the stacking sequence of the c-planes. The contrast variations are interpreted as fluctuations in the Sc alloy concentration, whereas the apparent variations in the stacking sequence

43 43 (a) Sc0.05Ga0.95 N (b) Sc0.17Ga0.83N 0 2 m 2 m Figure 4.1: 2µm 2µm AFM images of Sc x Ga 1 x N/GaN/Sapphire(0001) with (a) x = 5%, and (b) x = 17%] are attributed to the incorporation of a number of stacking faults. Since the stacking sequence of hcp GaN and fcc ScN are ABABAB... and ABCABC...; if a layer like C is added to the hcp sequence, a fcc structure is obtained. Therefore, near the transitional regime between the wurtzite GaN and rocksalt ScN more precisely at Sc concentration 17% the appearance of stacking faults is natural since the crystal starts to develop toward the most stable rocksalt ScN structure. Nonetheless, the continuation of the c-planes along the c-direction clearly remains as the layer grows, which implies that Sc is incorporated into the Ga sublattice even during 3D growth. The Sc does not seem to incorporate into interstitial sites since this would likely lead to greater degradation of the crystal structure or even to an amorphous film. In Figure 4.3 are shown expanded views of the X-TEM images of Sc x Ga 1 x N for x = 5% and 17% together with their corresponding fast fourier transforms (FFT s).

44 44 (a) GaN 5 nm (b) Sapphire Sc Ga N nm (c) GaN Sc Ga N nm GaN Figure 4.2: HRTEM images of Sapphire, GaN, and Sc x Ga 1 x N; (a) 30nm 30nm HRTEM image showing well ordered GaN and Sapphire crystals; (b) 30nm 30nm HRTEM image showing GaN and Sc x Ga 1 x N with x = 5% ; (c) 15nm 15nm HRTEM image showing GaN and Sc x Ga 1 x N with x = 17% [the black arrows show the interface between GaN and Sc x Ga 1 x N layers for figures (b) and (c)].

45 45 The FFT image of Sc x Ga 1 x N with x = 5% [Fig. 4.3(b)] shows a sharp spot pattern, consistent with the real-space image showing a highly crystalline layer. The out-ofplane (y-direction) spacing of the spots is the reciprocal of the c-lattice constant, while the in-plane (x-direction) spacing of the spots is the reciprocal of the a-lattice constant. Therefore, for the ideal wurtzite structure, the x:y spot spacing ratio should equal c/a 5.19:3.20 = 1.62, and the experimental data is in good agreement with this value. The FFT image of Sc x Ga 1 x N with x = 17% [Fig. 4.3(d)] shows that the diffraction spots are elongated into horizontal diffraction streaks. This spot-elongation indicates only 1-dimensional periodicity which is along the c-direction; the long-range periodicity in the in-plane direction is lost although the short-range periodicity within a single in-plane layer is still seen [Fig. 4.3(c)]. This loss of long-range in-plane periodicity is consistent with the loss of stacking order seen in the X-TEM image [Fig. 4.3(c)] which is attributed to the existence of stacking faults. Shown in Figure 4.4 are Rutherford backscattering (RBS) measurements for the Sc x Ga 1 x N films with x = 5% [Fig. 4.4(a)] and x = 17% [Fig. 4.4(b)]. The Sc and Ga signals are each clearly seen in the RBS spectra beginning at the onset energies of 2.2 MeV and 2.45 MeV, respectively. The measured random spectra (black circles) are compared with RUMP simulations (black lines) for both Sc x Ga 1 x N/Sapphire(0001) films where Sc compositions of x = 5% and 17% were used in the simulations. As can be seen from the RBS spectra of Fig. 4.4(a), the RUMP fit is quite good, confirming the composition values indicated. Aligned spectra (open triangles), in which the incident alpha particle beam is aligned with the crystalline c-axis, indicate the degree of ion crystal channelling; the more elastic backscattering present in the aligned spectrum, the less channelling there is, indicating more backscattering centers in the film. In the case of Sc x Ga 1 x N with x = 5% [Fig. 4.4(a)], the amount of backscattering from Ga and Sc is 54% and 71%, respectively. This significant amount of backscattering in comparison with a pure GaN film (only few % backscattering at most) suggests a large number of scattering centers within the crystalline channels. In the case of Sc x Ga 1 x N with x = 17%

46 46 (a) Sc Ga N (b) FFT Sc Ga N nm (c) Sc Ga N (d) FFT Sc Ga N nm Figure 4.3: HRTEM and FFT images of Sc x Ga 1 x N; (a) 35nm 35nm HRTEM image of Sc x Ga 1 x N with x = 5%; (b) FFT image of (a); (c) 9nm 9nm HRTEM image of Sc x Ga 1 x N with x = 17%; (d) FFT image of (c)

47 47 (a) Normalized Yield random aligned RUMP simulation x = 0.05 Sc Ga 3.05 MeV He 168 o backscattering (b) Normalized Yield random aligned Energy (MeV) x = 0.17 backscattering Sc Ga 0 Energy (MeV) Figure 4.4: RBS images of Sc x Ga 1 x N/Sapphire(0001) with x = 5% [figure (a)], and x = 17% [figure (b)]

48 48 [Fig. 4.4(b)], the backscattering from Ga and Sc sites is increased to 94% and 100%, respectively. This indicates an even greater concentration of backscattering centers in the film. In a previous chapter [39], it was supposed that the backscattering of the aligned spectra was attributed to the lattice distortions arising from the incorporation of Sc onto Ga lattice sites. As shown below in our recent theoretical calculations, the lattice distortions are positively found to occur as expected. However, the amount of lattice distortion for x 6% in ScGaN is found to be only a few degrees at most around any given Sc substitutional site. While the lattice distortions could affect the channelling, it is uncertain whether the cumulative effect of such small distortions would be sufficient to cause the observed backscattering. In the case of the ScGaN film with x = 17%, the X-TEM image clearly suggests the presence of stacking faults in the film; stacking faults would increase the number of backscattering sites as a random variation in stacking sequence results in cations blocking otherwise-open channels. This can be understood clearly from Fig. 4.5(b) where cations incorporating at the faulted sites block the otherwise-open channels. In the case of x = 5%, the X-TEM image did not indicate the presence of many stacking faults; it therefore suggests that either (a) local lattice distortions due to Sc incorporation are surprisingly sufficient to cause significant backscattering; or, (b) stacking faults, despite not being observed for 5% Sc concentration in X-TEM, are in fact present in significant quantities in the 5% film. A possible origin of increasing stacking faults with increasing Sc concentration is a reduction in surface diffusion which kinetically limits site sampling during growth, leading to growth of faulted domains. The reduced surface diffusion can be related to the bonding of Sc atoms in comparison with Ga atoms. If one considers that in the ground state, Sc forms 6 bonds with N (rocksalt structure) whereas in the ground state, Ga forms only 4 bonds with N (wurtzite structure), then on a surface, it could be that Sc atoms have more bonds to share with N compared with Ga atoms. Then the N atom will be more likely to bond to the Sc atom site compared to the Ga atom site. This effectively increases the sticking coefficient but decreases the overall surface diffusion and makes

49 49 (a) Sc Sc xga1-xn (0001) no faults (b) Sc xga1-xn (0001) faulted Ga N A [ 1120] [0001] [ 1210] C B [ 211 0] Figure 4.5: Images of Sc x Ga 1 x N crystal without stacking faults [figure (a)], and with stacking faults [figure (b)] it more likely for N atoms to bond at faulted sites. Therefore, as Sc concentration on the surface increases, the probability of forming stacking faults also increases. To summarize these experimental findings, we observe that ScGaN with small Sc composition in the range 0-17% can be grown using MBE under N-rich conditions with Sc atoms substituting for Ga atoms. We find that the alloy remains crystalline even at 17% Sc concentration; however, stacking disorder is observed in X-TEM images and confirmed in FFT images. And the backscattering seen in RBS also confirms defects, which are assumed to therefore be stacking faults. Despite these defects, the effect of Sc incorporation on the lattice is to locally distort the lattice, and this effect is not altered despite the presence of the evident stacking faults. In fact, we find that the lattice constants measured are evidently not affected by the stacking faults due to their agreement with the theoretical calculations. Briefly, our theory collaborators, Dr. Nancy Sandler, and Dr. Pablo Ordejón, performed firstprinciple Density Functional Theory calculations, and they generated a wurtzite GaN

50 50 in super-cells containing 32 atoms. The Ga atoms were substitutionally replace by Sc for concentrations of Sc x Ga 1 x N with x = 0, 6.25, 12.5, 18.7, and The subsequent measurement of the bond-lengths and angles around Sc and Ga/N atoms shows that the angles are distorted more around Sc atoms, as compared to the angles around Ga or N atoms. This is an indication of the tendency towards flattening the wurtzite bi-layer, which is very consistent with the experimental evidences presented in chapters 3, and Summary In summary, we have presented a combined experimental and theoretical study (attached in appendix A) of ScGaN with Sc concentration up to about 17% (experiment) and 25% (theory). The calculations are in overall good agreement with the experiment, and a good match is found between the variation of lattice constants with Sc fraction from experiment vs. that from theory. Furthermore, the variation of bond-lengths and in-plane and out-of-plane angles as functions of Sc fraction has been studied, and the results agree with a picture in which the addition of Sc to the GaN lattice induces local lattice distortions which tend toward a flattening of the wurtzite bi-layer. The experimental evidence shown here also suggests the presence of stacking disorder for the higher Sc fraction (17%). It is likely that kinetic limitations during growth or mechanisms of stress release in Sc-rich regions may be cause of the stacking disorder. Despite this disorder, the lattice constants of the 17% sample appear to be entirely determined by the local lattice distortions (and not by the stacking defects), as also predicted by the theory.

51 51 Chapter 5 Mixing Rocksalt and Wurtzite Structure Binary Nitrides to Form Novel Ternary Alloys: ScGaN and MnGaN 5.1 Introduction In the chapters 3, and 4 we focused our attention to structural characterization of ScGaN alloys. In this chapter, we make a structural comparison between MnGaN and ScGaN systems for Mn(Sc) concentration of 5%. Only a select few special material systems present the opportunity for ideal lattice-matched epitaxy. It is therefore a great challenge to study materials, for example alloys, in which the separate binary materials are not isocrystalline. In this chapter, we compare two such cases: 1) MnGaN; and 2) ScGaN. While GaN (bandgap 3.37 ev) has wurtzite structure and tetrahedral bonding, both MnN and ScN are face-centered tetragonal (fct) with octahedral bonding. Combining these disparate structures would appear to be quite difficult; yet, it is of significant interest to do so in order to form new materials having novel properties, including semiconducting, magnetic, and even ferroelectric. The two cases, MnGaN and ScGaN, are quite different in terms of both their interest and

52 52 their growth. ScN is in fact a rocksalt semiconductor with a in the range Å, an indirect bandgap from Γ- X of 1 ev, and a direct transition Et at the X point of ev [28, 14, 15, 16, 17, 18, 19]. The properties of ScGaN have been thoroughly explained in the previous chapters (3, and 4). On the other hand, MnN is a metal and has the fct structure with a = b = 4.22 Å and c = 4.12 Å [29]. One reason to combine GaN with MnN is for the possibility to form a room-temperature ferromagnetic MnGaN semiconductor [30]. Though challenging, growth of wurtzite MnGaN alloy by molecular beam epitaxy using radio frequency plasma has been reported [31, 32, 33]. In addition, cubic MnGaN has also recently been grown on GaAs (001) and found to have p-type conductivity [34]. The purpose of this chapter is to make a structural comparison between ScGaN and MnGaN alloys mainly focusing on the low Sc (Mn) composition (x < 20%). 5.2 Experimental Details In this chapter, we present results from ScGaN and MnGaN alloys using a custom MBE system that employs Mn, Ga, and Sc effusion cells and a rf plasma N source with N 2 as the source gas. Details about the ScGaN alloys growth is presented in chapters 3 and 4. The MnGaN films are grown using substrates of wurtzite Ga-polar GaN (0001) grown by metalorganic chemical vapor deposition (MOCVD) on sapphire. The substrates are loaded into the MBE chamber and heated up to 550 C, which is maintained throughout the entire growth. The r = J Mn / (J Mn + J Ga ) flux ratio is set to specific values with constant N flux while keeping J Mn + J Ga < J N (N-rich conditions). The growth begins with a 30 nm thick GaN buffer layer, and then the Mn shutter is opened to grow the (Ga,Mn)N layer with thickness in the range mm. Finally, a 30 nm thick GaN cap layer is grown. The growth is monitored in-situ by reflection high energy electron diffraction (RHEED) using a 20 kev e - beam, and the films are studied ex-situ by X-ray diffraction (XRD) with Cu Kα X-rays, optical absorption (OA) (ScGaN), and Rutherford backscattering (RBS).

53 Results and Discussion In Figure 5.1 are presented the RHEED patterns of ScGaN and MnGaN taken along [1120] direction. Qualitatively, both RHEED patterns in Fig. 5.1 are spotty suggesting roughness at the surface of the films. The fact that the spots in Fig. 5.1(b) do not connect with each other implies that the MnGaN film exhibits a rougher surface as compared to the ScGaN film from Fig. 5.1(a). The streak spacings for both films are very close to the wurtzite GaN values, and only careful measurement can reveal the differences (shown below). In Figure 5.2 is presented RBS results for the MnGaN (a) and ScGaN (b) films. The onsets of Ga, Mn, and Sc are at energies of 2.45, 2.3, and 2.17 MeV, respectively. For both MnGaN and ScGaN films, the filled circles represent the random spectra, the empty triangles represent the aligned spectra, and the black lines represent the RUMP simulations for films having 5% Mn or Sc incorporation. The random spectrum and the 5% Mn RUMP simulation for MnGaN [Fig. 5.2(a)] are in reasonable agreement (Note that the substrate is GaN/sapphire so that the backscattering from GaN continues past the end of the film). For the case of ScGaN [Fig. 5.2(b)], good agreement is also found between the random spectrum and the 5% Sc RUMP simulation (Note since the substrate was sapphire, the peaks drop off at the low energy side). Thus we find that both Mn and Sc are incorporated into GaN under N-rich conditions. The errors of the RUMP simulation concentrations are considered to be ± 1%. The aligned MnGaN spectrum in Fig. 5.2(a) shows no signal from the Mn, indicating that the Mn is substitutional on the Ga sites, with little or no bond angle distortion. For example, the He ion does not see much Mn when the beam is aligned along the c-axis. We note that N-doped Mn clusters have recently been proposed to play a key role in the ferromagnetism of Mn-doped GaN [35]. Our RBS result cannot rule out this possibility if the fraction of Mn existing as clusters is only a small percentage (i.e. 10%) of the total Mn incorporated, since such a concentration will fall below the noise level. Further work is required to explore this possibility. In the case of ScGaN [Fig. 5.2(b)], the aligned spectrum (beam aligned along substrate

54 54 ScGaN a) J /(J +J ) = Sc Sc Ga MnGaN b) J /(J + J ) = Mn Mn ga Figure 5.1: RHEED pattern of ScGaN and MnGaN along [1120. The Sc incorporation measured with the RBS in (a) is found to be x 5%, and also for the case of MnGaN (b), Mn incorporation is x 5 %.

55 55 (a) Normalized Yield GaN random spectrum aligned spectrum RUMP simulation 3.05 MeV He 168 o Mn end film Ga end Mn Ga (Ga,Mn)N 5% Mn 5 0 (b) Energy (MeV) random spectrum aligned spectrum RUMP simulation ScGaN 5% Sc Normalized Yield MeV He 168 o channeling 5 Sc Ga 0 Energy (MeV) Figure 5.2: RBS measurements. (a) MnGaN; and (b) ScGaN each with x 5 %

56 56 c-axis) contains components from both Ga and Sc (although that for Sc is hard to see in the picture). Quantitatively, the channelling in the ScGaN film is 76% for Ga and 54% for Sc. The significant amount of backscattering in the aligned ScGaN spectrum shows that the He ions are scattered by both Ga and Sc. Clearly, the case is very different compared to MnGaN. The backscattering could imply an amorphous film; however, as was shown the RHEED pattern has significant in-plane atomic ordering. Moreover, it is known that the RBS channelling can be affected by only tiny (e.g. 0.1 Å) shift of an atom site[36]. Therefore, a more intriguing possibility is that the ScGaN crystal is in fact well-ordered locally but has bond angle distortions, discussed further below. Out-of-plane lattice information was determined using XRD. Shown in Fig. 5.3(a) and 5.3(b) are the entire XRD spectra ( ) for ScGaN and MnGaN with x 5%. For both samples, there are five main peaks seen: two sapphire substrate peaks (0006 & 00012) and three alloy peaks (0002, 0004, and 0006). In previous work, we found that the c lattice constant for MnGaN with 5% Mn was essentially unchanged with respect to wurtzite GaN [37]. For the case of ScGaN, the c-values are found to increase slightly (as shown below) with increasing Sc fraction (as presented in chapter 3). We note in the case of the MnGaN a few very low-intensity peaks at 44.79, 53.14, 58.49, 60.04, 64.89, and The peak at agrees well with η-mn 3 N (44.81 ) [29]; that at is in good agreement with MnGaN The peak at agrees well with Mn 4 N 112 (58.53 ) or Mn 044 (58.62 ); that at is somewhat close to MnGa 012 (60.40 ) or Mn 334 (60.61 ). The peak at agrees well with either MnGa 020 (64.89 ) or Mn 235 (64.48 ); finally, the peak at agrees well with Mn 037 (82.46 ) or Mn 4 N 113 (82.88 ). In the case of ScGaN, there are no observable secondary peaks. A comparison of peak intensities from Figure 5.3(a) & (b) can give more information about the structure of the films. The XRD peak amplitudes of ScGaN and MnGaN are presented in Table 5.1. The magnitude of the 0002, 0004, and 0006 peaks of MnGaN stay within the same order of magnitude, indicating good crystallinity of the film. This agrees well with the RBS results. On the other hand, the amplitude of the ScGaN peaks decreases by two orders of magnitude with increasing peak index (from 0002 to 0006), and in addition their width increases with increasing peak index,

57 a) Intensity (counts) (Ga,Sc)N 0002 Sapphire 0006 Sapphire %Sc x = 0.05 ScGaN Whole Spectrum (Ga,Sc)N 0004 (Ga,Sc)N (degrees) (degrees) 1.E+07 b) 10 7 (Ga,Mn)N 1.E E Intensity (counts) 1.E E E E (Ga,Mn)N 0004 Sapphire 0006 x ~ 0.05 MnGaN Whole Spectrum Sapphire (Ga,Mn)N E (degrees) Figure 5.3: (a) & (b) XRD spectra of ScGaN and MnGaN with x 5% showing the two sapphire peaks [0006 & 00012] and the three ScGaN / MnGaN peaks [0002, 0004, & 0006]

58 58 both of which indicate a decrease in the long range ordering for the ScGaN crystal structure (compared to GaN) [38]. However, the decrease in the long range ordering need not imply an amorphous or polycrystalline structure. Indeed, amorphous or polycrystalline behavior would be inconsistent with the ScGaN RHEED pattern [Fig. 5.1(a)]. In the case of an amorphous film, we would not expect the diffraction lines, and in the case of a polycrystalline film, we would expect a ring-like RHEED pattern. The data therefore suggests that the ScGaN for low Sc concentration (x 17 %) can be described as a distorted wurtzite lattice in which the Sc-N-Sc bond angles θ b with one bond in the c-direction get slightly smaller compared to the wurtzite value of 108. Table 5.1: XRD peak amplitudes and FWHM for ScGaN and peak amplitudes for MnGaN films for 0002, 0004, and 0006 peaks. ScGaN peak Amp. (cts.) ScGaN FWHM ( of 2θ MnGaN peak Amp. (cts) A reason for bond angle distortion is based on the recent prediction of Farrer and Bellaiche who have found that ScN has a metastable-layered hexagonal phase, which can be arrived at by flattening the bilayer of the wurtzite structure [22, 3]. While the ScGaN alloy does not have this layered hexagonal structure, the same tendency of Sc to form more octahedral-type bond angles with N results in a local distortion of the crystal in the vicinity of the Sc atoms. This distortion can also explain the large He backscattering observed in RBS as well as the decreasing XRD intensity with the peak order. More details about the experimental work for this ScGaN alloy is explored in depth in chapter 3. In Figure 5.4 is plotted ScGaN lattice spacing c vs. lattice spacing a for low Sc concentration (x 0.17). The lattice constant a is obtained directly from the RHEED pattern using a peak fitting program; the lattice spacing

59 59 c was determined from the XRD spectra. Over the range 0 < x < 0.17, a increases by a net 0.08 Å while c increases by only a net Å. Such anisotropic expansion would not be expected in the case of alloying of iso-crystalline binary compounds. c (A) a (A) Figure 5.4: ScGaN lattice spacing c vs. a for low Sc concentration. The anisotropic expansion of the ScGaN lattice suggests the picture that is illustrated in Fig Further experimental evidence for low x regime points to the same model presented in chapter 3. First, we note substantial broadening of the RHEED diffraction lines with increasing x. Second, we note the substantial intensity decrease and broadening of the 0002 ScGaN XRD peak with increasing x. Such behavior is indicative of an increased spread of (local) lattice parameter with increasing x, resulting in reduction of the long-range order or of the maximum correlation length of the crystal.

60 60 Sc Ga N N-Sc-N N-Ga-N Figure 5.5: Schematic model of ScGaN for low Sc composition showing local distortions of the bond angle θ N Sc N and θ N Ga N. 5.4 Summary In summary, we have studied MnGaN and ScGaN alloys with small Mn and Sc compositions. We find that the two cases exhibit completely different incorporation and crystal structure phenomena. For MnGaN, the Mn atoms are incorporated into the Ga lattice sites with little effect on the wurtzite crystal structure. On the other hand, for the ScGaN, the Sc also incorporate into the Ga sites; however, the bond angles are distorted, resulting in marked effects on the XRD peak intensities as well as the RBS channelling. The conclusion is that the ScGaN lattice is nominally wurtzitelike but with local lattice distortions.

61 61 Chapter 6 Investigation of MnScN(001)/MgO(001) Grown by Molecular Beam Epitaxy: A Possible Dilute Magnetic Semiconductor 6.1 Introduction In the last couple of years, much interest has been focused on the doping the III-V semiconductors (e.g. GaAs and GaN) with the transition metal ions (e.g. Mn, Cr) to obtain dilute magnetic semiconductor (DMS). Moreover, wurtzite GaN system has attracted a lot of attention lately due to the prediction by Dietl et al. of ferromagnetism above room temperature for GaMnN DMS system [49]. One of the ultimate goals for DMS systems is to be able to use and control the spin of the electrons, besides the classical electron charge. But, as of this moment, there are still problems that need to be solved in the DMS systems, such as, growth, incorporation, and precipitations. It is known that transition metals such as manganese prefers octahedral bonding configuration with nitrogen rather than tetrahedral bonding (found in the

62 62 wurtzite GaN structure) [29]. In order to solve some of the problems mentioned before regarding DMS systems, we want to incorporate the manganese into a semiconductor that has the same octahedral bonding configuration between the anions and the cations. Hence, in this chapter of the thesis, we present preliminary magnetic results of isocrystalline system of MnScN (a possible DMS system). MnScN films were grown on MgO(001) substrates by molecular beam epitaxy. The Mn concentration of the films was varied between 3% and 5%. Magnetic measurements were performed on four films with superconducting quantum interference device (SQUID): two films of ScN and two films of MnScN with 3% and 5% Mn concentration. 6.2 Experimental Details We grew MnScN films on top of ScN(001)/MgO(001) substrates by radio frequency plasma assisted molecular beam epitaxy (rf-mbe). We used e-beam source for Sc cell, and effusion cell for Mn. We used N 2 as a plasma source, and the base pressure of nitrogen and the plasma source power were set at torr and 500 Watt respectively. The MgO(001) substrates are cleaned ultrasonically with acetone and isopropanol, after which they were introduced into the growth chamber. Prior to the buffer layer growth of ScN(001), the MgO(001) substrates were heated up to 1000 C for 30 minutes. The buffer layer growth is grown subsequently on the MgO(001) substrate at a temperature of 800 C, and with a thickness of 50 nm. After the growth of ScN(001)/MgO(001), the substrate temperature is reduced to 520 C, and then the MnScN films were grown at Mn concentrations of 3-5% and with a thickness of 290 nm. During the entire growth, the Sc flux was maintained constant to cm 2 s 1, while the Mn flux was varied in the range of cm 2 s 1. The total flux ratio was chosen to be given by the ratio of manganese to the total metal in the film [R = J Mn /(J Mn + J Sc )]. The growth is monitored in-situ by reflection high energy electron diffraction (RHEED). After the samples were taken out of the growth chamber, ex-situ measurements by x-ray diffraction (XRD), and by SQUID were performed.

63 Results and Discussions Figure 6.1 shows a typical XRD (θ - 2θ) spectrum of the MnScN film with 5% Mn concentration. There are only two peaks seen, one MnScN 002 peak which overlap with the ScN 002 peak (coming from the buffer layer) which has the 2θ position at 39.81, and one MgO 002 with the corresponding 2θ of 42.95, respectively. If we take a wavelength of 1.54 Å for the K α of the Cu XRD source, then we found a out-ofplane lattice constants of c = 4.52 Å for MnScN, and a c = Å lattice constant for the MgO substrate (note that the spectrum was corrected to the MgO 002 peak). Brithen et al. reported that manganese incorporates into the host ScN lattice and it actually forms Mn x Sc 1 x N alloy[6]. The alloy obeys the Vegards law, and by using equation 6.1, deduced by Al-Brithen et al. [6], we can calculated the expected lattice constant c alloy = 4.48 Å. c alloy = (1 x c ) c ScN + x c c MnN (6.1) Where x c = 0.05, c S cn = Å, and c MnN = Å (note that the lattice constants of ScN and MnN are given in Al-Brithen s paper [6]). According to Al- Brithen et al., the Mn x Sc 1 x N alloy shows a noticeable shift in the XRD peak starting with x c = 0.19, hence, it is conceivable that for our case we would not expect any shift in the XRD peak, since our x c is only 0.05 Mn concentration. Therefore, our deduced lattice constant for MnScN of 4.52 Å agrees much better with the lattice constant for ScN of Å reported by Al-Brithen et al. [6]. Hysteresis magnetic measurements are presented in Figure 6.2 for a MnScN film with a manganese concentration of 3%. The hysteresis magnetic measurements of the Mn 0.03 Sc 0.97 N were taken at a temperatures of 5, 12, and 50 K, respectively [Fig.6.2]. After subtracting the background, a strong magnetic signature is evident for Mn 0.03 Sc 0.97 N at temperatures of 5 and 12 K, indicating ferromagnetic behavior. However, as the temperature is increased to 50 K, it looks like the intensity of the magnetic moment is decreasing indicating a reduction of ferromagnetism for this film. Therefore, the Curié temperature for this film is 50 K.

64 64 Intensity (counts) MnScN 002 MgO ( o ) Figure 6.1: XRD spectrum of a MnScN(001)/MgO(001) sample with 5% Mn concentration. In Figure 6.3 is presented the magnetic measurements for all four films, two control samples of ScN (# 71 and # 73) and two films of MnScN with Mn concentration of 3% and 5%, respectively. The temperature for the magnetic measurements in Fig.6.3 for the samples #71, #339, and #340 is 5 K, whereas for the sample #73 is 8 K. The sample #339 is the Mn 0.05 Sc 0.95 N and at 5 K there is no hysteretic behavior indicating that the film is nonmagnetic. It might be possible that as we increase the Mn concentration, the magnetism that was detected in the sample Mn 0.03 Sc 0.97 N disappears. This could be due to the fact that as we increased the Mn concentration to 5% in to ScN lattice, the average distance between Mn atoms is reduced and these atoms align antiferromagnetically with each other, hence reducing the total ferromagnetism of the film. One of the control sample [#73] shows ferromagnetic behavior in Fig.6.3 whereas a similar sample #71 which was grown under very similar

65 65 Figure 6.2: SQUID measurement of MnScN sample with 3% Mn concentration. conditions as sample #73 does not show any magnetic behavior. We don t understand at the moment were is the magnetism coming from in the case of sample #73, but we believe that the ScN is nonmagnetic, and probably the magnetism seen in sample #73 could be due to some measurement errors from our collaborators who performed the SQUID measurements. We also performed energy dispersive x-ray to see if the magnetism could come from magnetic clusters of iron, cobalt, or other magnetic materials. No other magnetic and nonmagnetic materials were present in the film except the expected ones.

66 66 Figure 6.3: SQUID measurements of two control samples of ScN [# 71 and # 73], and two samples of MnScN with Mn concentration of 3% and 5%, respectively. 6.4 Summary In summary, MnScN films with 3% and 5% Mn concentration were grown by radiofrequency molecular beam epitaxy. Our XRD results agree well with the XRD results obtained by Al-Brithen et al. for the Mn x Sc 1 x N alloy growth[6]. The hysteresis magnetic measurements show ferromagnetic behavior for Mn 0.03 Sc 0.97 N with a Curié temperature of 50 K. As the Mn concentration is increased, the ferromagnetism is reduced. One possible explanation for this phenomenon would be that, as we put more manganese into the ScN host lattice, the average distance between manganese atoms is reduced and these atoms align antiferromagnetically with each other, therefore reducing the total magnetic moment of the film.

67 67 Chapter 7 Growth of Heterostructures c-gan(001)/scn(001)/mgo(001) and ScN(001)/c-GaN(001)/MgO(001) by Molecular Beam Epitaxy 7.1 Introduction In the chapter 6, we presented the magnetic properties of MnScN alloys with 3%, and 5% Sc incorporation. As we showed in chapter 6, MnScN could be a possible dilute magnetic semiconductor. But in order to produce spin injection, MnScN should be grown on top of a semiconductor such as cubic GaN. Lately, our group focused on the growth, structural and surface reconstructions of the cubic GaN(001) grown on MgO(001) substrates. We found with cathodoluminescence measurements that c- GaN is a semiconductor with direct optical bandgap of 3.2 ev. From XRD we found a out-of-plane lattice constant c = 4.52 Å. The film/substrate epitaxial orientation relationship is found to be 001 c GaN 001 MgO and 110 c GaN 110 MgO. In addition, we found that there is only 1% lattice mismatch between c-gan(001) and

68 68 ScN(001). Since both systems (c-gan and ScN) have octahedral bonding with the N-atoms, we want to go a step further and ask the question: how would the c-gan and ScN semiconductors behave in heterostructure system. Therefore, in this chapter, we investigate the structural behavior of heterostructures such as ScN(001)/c- GaN(001)/MgO(001) and c-gan(001)/scn(001)/mgo(001) grown by MBE. 7.2 Experimental Procedure c-gan(001)/scn(001)/mgo(001) and ScN/c-GaN(001)/MgO(001) heterostructures were grown by molecular beam epitaxy (MBE) using a radio frequency (RF) plasma source for nitrogen and an effusion cell for gallium and scandium. The base pressure of nitrogen and the plasma source power are set at torr and 500 Watt respectively. The substrate temperature for c-gan layer is 750 C, and 850 C for the ScN layer. Usually the thickness of the layer that is grown directly onto MgO(001) substrate is between 220 nm and 500 nm. In order to study the heteroepitaxial layer growth under stress conditions, we choose to deposit the topmost layers of c-gan(001) and ScN(001) only with thicknesses varying from 8 nm to 13 nm. The Sc flux is chosen to be /cm 2 sec, whereas the flux for Ga is varied between /cm 2 sec and /cm 2 sec. The growth is monitored in real time using RHEED. Finally samples are characterized ex situ by AFM and XRD using Cu Kα for more surface and structural studies. 7.3 Results and Discussion Figure 7.1 shows a typical XRD θ-2θ scan of c-gan(001)/scn(001) heterostructure, obtained over the 2θ range of 25 to 130. The spectrum shows mainly two orders of c-scn [002 and 004] and two orders of MgO [002 and 004] peaks. The XRD 002 and 004 peaks of the topmost layer of c-gan are convoluted together with the ScN 002 and 004 peaks because the lattice constant of c-gan of c = 4.52 Å is very close to the lattice constant of ScN [51] of c = 4.51 Å. In addition, the intensity of

69 ScN002 MgO002 MgO004 Intensity (cts) ScN ( o ) Figure 7.1: XRD θ-2θ scan of c-gan/scn/mgo(001) heterostructure. the c-gan 002 peak is 2,110 counts, which is about 17.5 times smaller than that of the ScN 002 peak. This difference in the peak intensity is due to the difference in the thickness of the grown layers, as c-gan has a thickness of 13.1 nm, whereas the ScN layer has a thickness of nm. Looking at the case when a ScN layer with a thickness of 8.3 nm is grown on top of a nm thick layer of c-gan on MgO(001). The XRD from the figure 7.2 shows that, besides, the convoluted ScN into the c-gan [002 and 004] peaks, there are also the ScN 111 and 222 peaks present at 2θ angles of and 72.34, respectively. The full width at half maximum (FWHM) intensity in the θ-2θ scan of ScN 111 peak is By using the Scherrer formula 7.1, we calculated a coherence length of 36 nm

70 GaN002 MgO002 MgO004 Intensity (cts) ScN111 ScN222 GaN ( o ) Figure 7.2: XRD θ-2θ scan of ScN/c-GaN/MgO(001) heterostructure. for the ScN 111 peak, along the c-axis of growth. For the general case, we took K = 0.9, and the wavelength of the x-ray to be 1.54 Å. L = Kλ cos(θ)f W HM(2θ) (7.1) In Figure 7.3 is shown the RHEED patterns of c-gan(001)/scn(001) heterostructure. Figure 7.3(a) shows the RHEED pattern of the MgO(001) substrate prior to the growth. Figure 7.3(b) shows the RHEED pattern which was acquired at the end of growth of the ScN(001) layer. The fact that the spots are elongated vertically indicate that there is a 2-dimensional combined with a 3-dimensional growth. In addition, the spots position tells us that the grains are mainly 001 oriented as compared to the substrate. Moreover, by calibrating the previous MgO(001) RHEED pattern [Fig. 7.3(a)] with a lattice constant of Å [the corresponding lattice constant of

71 71 (a) (c) S28_ S32_ (b) MgO(001) 750 o C (d) S32_ S32_ o C 27 o C Figure 7.3: RHEED images of c-gan(001)/scn(001)/mgo(001); (a) MgO(001) RHEED pattern before the growth of ScN; (b) ScN(001) RHEED pattern at a sample temperature of 850 C; (c) and (d) GaN(001) RHEED patterns at the sample temperature of 750 C, and 27 C, respectively. MgO(001)], we found a in-plane lattice constant of the ScN(001) layer to be a ScN(001) = 4.47 Å ± Å. After growing c-gan(001) on top of ScN(001), we recorded the RHEED pattern from Figure 7.3(c) and we observed that the surface become more 3-dimensional since the spots have more circular shape and they are not so much elongated in the vertical direction as compared to the 7.3(b). We obtained a in-plane lattice constant of c-gan(001) layer to be a GaN(001) = 4.49 Å ± Å. The final RHEED pattern has been acquired after the film was cooled down to 27 C, and we obtained a lattice constant of a GaN(001) = 4.48 Å ± Å.

72 72 (a) S28_ a (b) 750 o C S28_ e 850 o C Figure 7.4: RHEED images of ScN(001)/c-GaN(001)/MgO(001); (a) c-gan(001) RHEED pattern at the sample temperature of 750 C; (b) ScN(001) RHEED pattern at a sample temperature of 850 C. In Figure 7.4 is presented the RHEED pattern of the ScN(001)/c-GaN(001). The RHEED pattern from Figure 7.4(a) shows the surface condition of the first layer of c- GaN(001) on top of the MgO(001) substrate. The RHEED pattern looks very streaky indicating a smooth surface [mostly 2-dimensional growth]. It looks like the growth of c-gan(001)/mgo(001) is smoother than the case when ScN(001) is grown on top of MgO(001) substrate. The atomic force microscopy of heterostructure c-gan(001)/scn(001) presented in Figs. 7.5(a-c) show that the c-gan, which is grown on top of ScN(001), has a very similar surface morphology with the c-gan(001) grown only on MgO(001) substrates. It is interesting to notice that the root mean square (rms) of the c-gan grown on the

73 73 (a) (b) 10 mx10 m (c) mx5 m [100] mx2 m [010] (Substrate orientation) Figure 7.5: AFM images of c-gan(001)/scn(001)/mgo(001). The gray scale and the root mean square roughness of a, b, and c are 332 Å and 9.87 Å, 344 Å and 12.0 Å, and 284 Å and 9.75 Å, respectively.

74 74 (a) (b) 10 mx10 m (c) mx5 m [100] mx2 m [010] (Substrate orientation) Figure 7.6: AFM images of ScN(001)/c-GaN(001)/MgO(001). The gray scale and the root mean square roughness of (a), (b), and (c) are 206 Å and 5.91 Å, 218 Å and 4.52 Å, and 220 Å and 14.9 Å, respectively.

75 75 ScN [Fig. 7.5(c)] is about 5 times smaller than the rms of c-gan(001) grown directly on MgO(001) substrate. The AFM images of the heterostructure ScN(001)/c-GaN(001) is presented in figure 7.6(a-c). In figure 7.6(a) is shown that ScN grown on c-gan forms rectangularlike shaped plateau that are interconnected with each other; clearly different from plateau-pyramid surface morphology found by Al-Brithen and Smith of ScN(001) grown only on MgO(001) substrate [51]. Comparing the rms = 14.9 Å of ScN(001)/c- GaN(001) [Fig. 7.6(c)] with rms = 9.75 Å of c-gan(001)/scn(001) [Fig. 7.5(c)] it seems that c-gan(001) grows smoother on ScN(001) rather than vice-versa. This implies that the zincblend of c-gan structure relaxes faster and is more stable than the rocksalt structure of ScN in a heterostructured situation.

76 Summary In summary, we grew c-gan(001)/scn(001) and ScN(001)/c-GaN(001) heterostructures by molecular beam epitaxy (MBE). Based on our XRD, AFM, and RHEED results, we found that the c-gan(001)/scn(001) has a very similar surface morphology with the c-gan(001) grown only on MgO(001) substrates. The RHEED, AFM and XRD measurements indicate that the surface of c-gan(001)/scn(001) is smoother and shows no other extra crystalline phases as compared to the surface of ScN(001)/c- GaN(001).

77 77 Chapter 8 Metal/Semiconductor Phase Transition in Chromium Nitride(001) grown by rf-plasma-assisted Molecular-Beam Epitaxy 8.1 Introduction In this chapter, we investigate the structural and electronic properties of stoichiometric single-phase CrN(001) thin films grown on MgO(001) substrates by radiofrequency N plasma-assisted molecular-beam epitaxy. For many years, CrN has received a lot of attention due to its high hardness and corrosion resistance [52, 53]. In addition, CrN is also interesting due to its magnetic, optical, and electronic properties. It is known that CrN is paramagnetic (PM) with a B1 NaCl crystal structure at room temperature; at T Néel, reported in the range K, the material undergoes a phase transition to antiferromagnetic (afm) with an orthorhombic P nma crystal structure [54, 55, 56]. Filippetti et al. found theoretically that the magnetic stress is the driving force for the lattice distortions, and thus linked the magnetic and

78 78 structural transitions [8]. So far, however, the electronic properties have been less understood. Filippetti et al. reported, based on theoretical calculations, that CrN is a metal in its PM state but a weak metal in the afm state [56]. This agrees well with an earlier experimental work, based on CrN polycrystalline powders, which found metallic behavior with, however, an abrupt increase of the resistivity with increasing temperature at 286 K [7]. On the other hand, Herle et al., based on synthesized CrN powders, reported that CrN is a semiconductor with a bandgap of 90 mev as measured by resistivity [9]. Recently, Gall et al. grew crystalline thin films of CrN by sputtering and reported that CrN behaves as a semiconductor (Mott-type insulator) with an optical gap of 0.7 ev [10]. Although the electronic behavior at room temperature has become more clear, the value of the bandgap as well as its detailed nature, remain to be determined; moreover, the discrepant results reported for low temperatures, remain to be resolved. In this chapter, we present results concerning the unique structural and electronic properties of CrN(001). We show that high-quality epitaxial layers are grown by molecular beam epitaxy (MBE), having crystalline (001) orientation and atomicallysmooth surface [as measured by scanning tunneling microscopy (STM)] and semiconducting behavior at 300 K. Bulk ex situ measurements confirm the semiconducting behavior at 300 K, but furthermore a transition to the metallic state is observed at 285 K. This transition coincides with the known magnetic/structural transition. 8.2 Experimental Details The experiments are performed in a custom-designed hybrid molecular beam epitaxy/scanning tunneling microscopy (MBE/STM) system. The MBE system employs a Cr effusion cell and a radio frequency N plasma source with N 2 as the source gas. During the entire growth, the N plasma source is set to a power of 500 W with the nitrogen flow rate about 1.1 sccm (growth chamber pressure is Torr). The entire growth process is monitored by reflection high energy electron diffraction (RHEED), after which the samples are transferred to the in situ STM. Finally, x-ray

79 diffraction (XRD) and resistivity vs. temperature (R-T) measurements are performed ex situ. Resistivity is measured using the 4-point Van der Pauw geometry Results and Discussions The MgO(001) substrates are first heated to 900 C for 30 minutes under the presence of N-plasma, after which the RHEED pattern looks streaky [Fig. 8.1(a)], indicating a suitable smooth substrate surface. A buffer layer of CrN is grown at a sample temperature T s <200 C with thickness t B 45 nm. The RHEED pattern of the buffer layer [Fig. 8.1(b)] looks spot-like, suggesting rough growth. Next, the temperature is increased to 450 C and a CrN layer is grown with a thickness of t CrN 90 nm, after which the RHEED pattern [Fig. 8.1(c)] shows a streaky behavior but also some extra spots. At this point, the Cr shutter is closed, followed by a second and rapid increase of T s to 650 C. The RHEED pattern becomes more streaky [Fig. 8.1(d)] and the extra spots fade away. After 14 minutes at 650 C a final 4 nm CrN layer is grown. The ultimate RHEED pattern [Fig. 8.1(e)] is very streaky suggesting a high quality, atomically smooth surface with (001) orientation. Shown in Fig. 8.2, are the main peaks seen in XRD over the range 30 < 2θ < 140 the two orders (002 and 004) of MgO and CrN. The Cu K α1 and K α2 002 peaks of MgO (CrN) are seen at 2θ=42.99 and (2θ=43.92 and ), respectively. The inset of Fig. 8.2 shows the Lorentzian fit for the rocking curve of the CrN 002 peak, giving a full width half maximum (FWHM) of the K α1 CrN 002 peak of Γ ω =0.09, which is smaller than the corresponding Γ ω =0.10 of the MgO 002 substrate peak. This further indicates the high quality crystallinity of the CrN film. Lattice parameters measured for several CrN(001) films are summarized in Table 8.1. The measured out-of-plane lattice constant a 4.14 Å is consistent with values of Å reported in earlier work[7, 11]. The measured in-plane lattice constant a obtained from RHEED varies from 4.04 Å for the thinnest film (t CrN =335 Å) to 4.13 Å for the thickest film (t CrN =1490 Å), increasing with thickness. The thinnest films are evidently in-plane compressively strained. Note that a a for thicker films and that the agreement between a and a improves with thickness.

80 80 (a) [100] MgO MgO [110] MgO (b) CrN (c) CrN (d) CrN (e) CrN Figure 8.1: Sequence of RHEED patterns during the CrN growth on MgO(001) substrate. (a) MgO substrate after the heating; (b) CrN buffer layer; (c) CrN layer grown at 450 C; (d) After the temperature of the substrate goes up to 650 C with the plasma on for about 10 minutes; (e) The final layer of CrN.

81 81 Intensity (counts) CrN(001)/MgO(001) K K Intensity (counts) K K MgO 002 CrN (degrees) Rocking Curve of CrN002 K (degrees) K Figure 8.2: XRD results for CrN(001)/MgO(001). Lorentzian fittings of K α1 and K α2 of MgO 002 and CrN 002 peaks; (inset) Rocking curve of CrN 002 peak. Moreover, the relaxed lattice constant a 0 is obtained for each film using the following equation (1) with ν being the Poisson ratio for CrN [10, 52]. 1 ν a 0 = a 1 + ν + a 2ν 1 + ν (8.1) With ν=0.29, the relaxed lattice constant a 0 obtained is 4.10 Å for the two thinnest (<1000 Å) films but for the thickest film a 0 =4.14 Å which we conclude to be the most probable relaxed bulk value for CrN and consistent with previously reported values of Å[10, 57, 58].

82 Regarding the stoichiometry, Rutherford backscattering has been done for our thickest film (t CrN =1490 Å) and it was found that Cr:N=1:1 with 5% uncertainty. 82 Table 8.1: CrN 002 FWHM, lattice constants a, a, and a 0, for 3 different film thicknesses of CrN(001)/MgO(001). CrN Sample thickness of CrN layer t CrN (Å) CrN 002 FWHM Fω ( ) lattice constant a (Å) lattice constant a (Å) relaxed bulk lattice constant a 0 (Å) Figure 8.3 shows room temperature STM images of CrN(001) acquired at sample bias V s =+0.7V. The surface of CrN, on a scale of 300Å 300Å [Fig. 8.3(a)], consists of smooth terraces separated by steps of height 2.07 Å (a 0 /2) which indicates that the growth is epitaxial. The atomic resolution image of CrN(001) in Fig. 8.3(b) clearly shows the cubic 1 1 structure. The atoms viewed in this image are most likely Cr, although a detailed surface calculation remain to be performed. The conventional surface cell, and the primitive surface unit cell rotated by 45 are shown. Features 1 and 2 in Fig. 8.3(b) are most likely CrN vacancy islands which formed during heating up to 650 C. In the image of Fig. 8.3(c), besides CrN vacancy islands (3, 4, and 5), there are some brighter areas labelled A, B, and C. Similar features, referred to as long-range-topographic distortions (LTD s) [59], have been observed recently by Al- Brithen et al. [19] on the non-polar (001) surface of semiconducting ScN as well as for various non-polar (110) surfaces of III-V semiconductors like GaAs [59]. They are only observed on semiconductor surfaces. For example, they are not seen on MnN(001) or Mn 3 N 2 (010) surfaces in our system, which are metallic [60]. The explanation of the mechanism of the LTD features has been explained elsewhere [59, 61], but briefly (for this case) the LTD s can be attributed to downward band-bending in the vicinity of

83 83 positively charged ionized donors, which locally enhances the tunneling current (for either bias polarity). Room temperature scanning tunneling spectroscopy (STS) measurements averaged over an area as in Fig. 8.4(c) have been performed. The current-voltage (I V S ) curve is presented in Fig. 8.4(right axis), showing a non-linear shape, typical of a semiconductor. The corresponding normalized conductance (NC) (di/dv)/(i/v ), which is related to the local density of states (LDOS), is plotted vs. V S in Fig. 8.4(left axis) and was computed according to the method described by Feenstra [62], using a broadening of 0.05 V. The NC-V S spectrum shows that a dip in the LDOS is observed near the Fermi level [the zero of the voltage in Fig. 8.4]. Such a dip is consistent with a small bandgap; however, the 300 K thermal broadening ( 50 mev) precludes a direct measure of its size. Qualitatively, the Fermi level E F is on the conduction band side of the minimum, implying the sample is n-type, which agrees with a separate hot point probe measurement, which showed that the majority carriers are electrons. To confirm the semiconductor behavior seen in STM at 300 K, ex situ resistivity vs. temperature (ρ-t) measurements were taken between 77 and 450 K. As shown in Fig. 8.5, the resistivity varies linearly (slightly increasing) with increasing T between 77 and 260 K [region (I)], showing metallic behavior. At 260 K, ρ begins to increase more, and above 280 K ρ increases steeply, consistent with a first-order phase transition. Above 280 K [region (II)], ρ decreases exponentially with temperature, indicating semiconductor behavior. Measuring ρ-t with decreasing temperature, thermal hysteresis is observed with a width of 20 K. As mentioned before, a steep increase in ρ has been observed by Tsuchiya et al. and Browne et al. [7, 63] but not a transition from metallic to the semiconductor state, as we observe. Plotted as an inset of Fig. 8.5 is ln(ρ/ρ 0 ) versus 1/T for the data of region (II). Fitting the data using the linear equation ln(ρ/ρ 0 )=(E g /2k B )(1/T), we derive the bandgap E g =71.0 ± 0.3 mev with ρ 0 = Ω-cm.

84 84 (a) 100 Å (c) A 5 B C 10 Å 3 4 (b) 1 [110] 2.87 Å [1 1 0] 2 10 Å Figure 8.3: RT STM of CrN(001)/MgO(001). (a) 250Å 250Å STM image of CrN showing smooth terraces. V S =+1 V, I T =0.2 na; (b) 25Å 25Å atomic resolution image showing the 1 1 square lattice periodicity and the fcc conventional unit cell of the CrN lattice. V S =+0.7 V, I T =0.2 na; (c) The features at A, B, and C are LTD s. The features 1-5 are CrN vacancy islands. V S =+0.6V, I T =0.6 na.