Epitaxial GaN using Ga(NMe2)3 and NH3 Plasma by Atomic Layer Deposition

Low temperature deposition of high-quality epitaxial GaN is crucial for its integration inelectronic applications. Chemical vapor deposition at approximately 800 °C using SiC with anAlN buffer layer or nitridized sapphire as substrates is used to facilitate the GaN growth. Here,we present a low temperature atomic layer deposition (ALD) process usingtris(dimethylamido)gallium(III) with NH3 plasma. The ALD process shows self-limitingbehaviour between 130-250 °C with a growth rate of 1.4 Å/cycle. The GaN films produced werecrystalline on Si(100) at all deposition temperatures with a near stochiometric Ga/N ratio withlow carbon and oxygen impurities. When GaN was deposited on 4H-SiC, the films grewepitaxially without the need for an AlN buffer layer, which has never been reported before. The bandgap of the GaN films was measured to be ~3.42 eV and the fermi level showed that the GaN was unintentionally n-type doped. This study shows the potential of ALD for GaN-basedelectronic devices. Abstract Low temperature deposition of high-quality epitaxial GaN is crucial for its integration in electronic applications. Chemical vapor deposition at approximately 800 °C using SiC with an AlN buffer layer or nitridized sapphire as substrates is used to facilitate the GaN growth. Here, we present a low temperature atomic layer deposition (ALD) process using tris(dimethylamido)gallium(III) with NH 3 plasma. The ALD process shows self-limiting behaviour between 130-250 ° C with a growth rate of 1.4 Å/cycle. The GaN films produced were crystalline on Si(100) at all deposition temperatures with a near stochiometric Ga/N ratio with low carbon and oxygen impurities. When GaN was deposited on 4H-SiC, the films grew epitaxially without the need for an AlN buffer layer, which has never been reported before. The bandgap of the GaN films was measured to be ~3.42 eV and the fermi level showed that the GaN was unintentionally n-type doped. This study shows the potential of ALD for GaN-based electronic devices. 2 q measurement and X-ray Reflectivity (XRR) mode to measure film. The software PANalytical X’Pert reflectivity was used to fit the XRR data using a two-layer model of the GaN substrate. A LEO 1550 Scanning electron microscope (SEM) with an acceleration energy of 10-20 kV was used to characterise the morphology of the film. The film composition and chemical bonding environments of the atoms in the film were characterized with a Kratos AXIS Ultra DLD X-ray photoelectron spectroscopy (XPS) equipped with Ar (0.5 keV) sputtering source. CasaXPS software was used to evaluate the data. Gaussian-Laurentius functions and Shirley background were used to fit the experimental XPS data. The absorption spectroscopy measurements were conducted using a custom fibre optical setup consisting of a light source (Ocean Optics DH-2000-BAL), a detector (Avantes AvaSpec-Dual) and a bifurcated optical on the GaN film. The RBS/ERDA measurement showed that the GaN film contained 45.7 at.% Ga, 47.2 at.% N, 2.8 at.% C, 3.1 at.% O and 1.2 at.% H giving a Ga/N ratio of 0.97, which indicates a slightly under-stochiometric film composition. This shows that even when subtracting the Ga Auger peak from the N 1s XPS region, the film composition and Ga/N ratio is not reliable and other techniques are required to determine an accurate atomic content. In addition, C impurities were not detected by XPS after clean sputtering whilst RBS/ERDA showed C present in the film, showing the latter is preferred for detection of impurities. In addition, sputtering was used in the XPS measurement, which could cause preferential sputtering and change the composition and impurity levels in the film. Nonetheless, XPS is highly important for accessing the chemical environment of each element.


Introduction
Gallium nitride (GaN) is a key material in microelectronics based on the group 13-nitride materials. This is due to its favourable properties of wide and direct bandgap, high thermal stability, high breakdown field and good electron mobility. 1,2,3,4 In high frequency electronics, GaN films are preferably deposited on SiC due to its high thermal conductivity. 5 However, it is difficult to deposit GaN directly on SiC by CVD due to poor wetting of GaN on SiC. 6 Therefore, an AlN buffer layer is used to facilitate GaN growth. 7,8,9 Currently, thin films of electronic grade epitaxial GaN are deposited by chemical vapor deposition (CVD) processes using trimethylgallium (TMG) and ammonia (NH3) at temperatures between 800-1000 °C. 10 The high deposition temperatures are required to obtain highly crystalline films but also to overcome the poorly suited precursor combination of TMG and NH3, which leads to high N/Ga precursor ratios of 10 3 . They also hinder the deposition of highly conformal GaN films on topographically complex surfaces and on temperature sensitive materials. To meet these challenges, the low temperature time-resolved CVD route known as atomic layer deposition (ALD) has been studied for GaN. Deposition of GaN by ALD has been achieved using TMG or triethylgallium (TEG) with N2/H2 plasma 11 , NH3 plasma 11,12 or thermally with NH3 13 . However, these processes suffer non-stoichiometric Ga/N ratios and high carbon and oxygen impurities.
Additionally, thermal ALD with TMG and NH3 requires deposition temperatures of 400 °C, which is relatively high for an ALD process. 13 Similar issues has been reported for ALD of GaN using TEG, were impurities of C and O was the primary issue. 14,15 . ALD routes using GaCl and GaCl3 together with NH3 deposited at high temperature (>400 °C) have also been investigated, but render films with Cl impurities. 16,17 The disadvantage of TMG and TEG for ALD of GaN is their strong M-C bonds, which makes it difficult to remove all of their ligands from the deposited precursor at low temperatures. 18 Amides (M-NR2) have more reactive and desirable M-N bonds, which would improve the surface chemistry during the NH3 precursor pulse. Herein, we report low temperature deposition of GaN using tris(dimethylamido)gallium(III), (Ga(N(CH3)2)3 1, with NH3 plasma by ALD. Precursor 1 has been previously used for ALD of Ga2O3 19 and Ga2S3 20 and CVD of GaN 21 , but not for ALD of GaN. Notably, all previous reports using 1 has resulted in amorphous films at all deposition temperatures. 19,20,21 We show that 1 displayed self-limiting behaviour in a temperature range from 130-250 °C on Si(100). Deposition of GaN on 4H-SiC(0001) rendered epitaxial films with near stochiometric composition and very low impurity levels of carbon and oxygen. To the best of our knowledge, this is the first report of GaN grown epitaxially directly on SiC substrate.

Experimental details Precursor synthesis
The reaction and manipulations were carried out under a nitrogen atmosphere on a Schlenk line using Schlenk air-free techniques and in a Glovebox-Systemtechnik dry box. All anhydrous solvents were purchased from Sigma-Aldrich TM and further dried with 4Å molecular sieves.
All NMR spectra were measured with an Oxford Varian 300 and 500 MHz spectrometers.
Solvents peaks were used as an internal standard for the 1 H NMR (300 MHz) and 13 C{ 1 H} NMR (125 MHz) spectra. The melting point was determined in a capillary sealed under N2 with a Stuart ® SMP10 melting point apparatus and is uncorrected. GaCl3 (99.99%) was purchased from ACROS-Organics TM and lithium dimethylamide (95%) from Sigma-Aldrich™, and both were used without further purification.

Synthesis of Tris(dimethylamido)gallium(III) dimer (1)
Tris(dimethylamido)gallium(III) dimer 1 was synthesised using a modified literature procedure ( Figure 1). 22 A room temperature suspension of lithium dimethylamide (5.2 g, 102 mmol) in nhexanes (150 mL) was slowly added to a -20 °C solution of GaCl3 (6.0 g, 34.1 mmol) in nhexanes (100 mL) via cannula and the mixture was stirred at this temperature for 1 h. The reaction was slowly warmed to room temperature and stirred for a further 16 h. The reaction mixture was filtered through a pad of Celite and concentrated under reduced pressure to give a solid. The solid was purified by recrystallisation from Et2O at -35 °C to give the compound 1 as a solid (4.61 g, 63%).

Thermogravimetric Analysis and Differential Scanning Calorimetry
The volatilization curve was collected using a TA Instruments thermogravimetric analysis Q500 tool. The ramp experiment of 1 was undertaken inside a nitrogen glovebox in a tared aluminum pan loaded with 18 mg of 1. The furnace was heated at a rate of 10 °C/min to 400 °C with a maintained nitrogen flow rate of 60 sccm. The differential scanning calorimetry (DSC) measurement for 1 was performed on a TA Instruments DSC Q10 tool. The sample was prepared in a sealed aluminum pans in a N2 filled glovebox and weighed approximately 0.2 mg.
The experiments heated the sample of 1 and a blank reference pan at a rate of 10 °C/min to 400 °C.

Film deposition
The films were deposited in a Picosun R-200 ALD system equipped with a Litmas Remote Inductively Coupled Plasma Source. The system used a base pressure of 400 Pa with continuous N2 (99,999%, further purified with a getter filter to remove moisture) flow through the deposition chamber. The Si (100) and 4H-SiC (0002) substrates were cut into 1.5×1.5 cm pieces and Si (100) was used without further cleaning whilst 4H-SiC was cleaned with RCA-1 (solution of 1 part H2O2 (30%), 1 part NH3 (25%) and 5 parts H2O) and RCA-2 (1 part H2O2 (30%), 1 part HCl(37%) and 6 parts H2O) 23 solutions to remove organic and inorganic contaminants on the surface prior to deposition. The substrates were loaded into the reactor onto a heated substrate holder and the system was heated to 450 °C for 120 minutes before each run with a continuous N2 flow (300 sccm) to minimize the oxygen content in the chamber.
Approximately 500 mg of 1 was placed in a glass vial in a stainless-steel bubbler without a diptube for incoming carrier gas. The temperature of the bubbler was set at 120 °C with a N2 flow of 100 sccm to aid transporting the precursor vapor into the deposition chamber. A 10 s N2 purge was used after the pulse of 1. The NH3 (AGA/Linde, 99.999 %) plasma used as the nitrogen source was an Ar (99.999 %, further purified with a getter filter to remove moisture)/NH3 (100/75 sccm) mixture, ignited using a plasma power of 2800 W. A plasma pulse of 9 s followed by a 10 s purge was used with the above parameters unless otherwise stated.

Characterisation
The crystallinity of deposited films was studied using a PANalytical EMPYREAN MRD XRD with a Cu-anode x-ray tube and 5-axis (x-y-z-v-u) sample stage both in q-2q and grazing incidence (GIXRD) configuration. Strain in the films was estimated from the GIXRD measurement for each diffraction peak values of sin 2 y and e using the software X'Pert Stress. PANalytical X'Pert PRO with a Cu-anode tube and Bragg-Brentano HD optics was used for q-2q measurement and X-ray Reflectivity (XRR) mode to measure film. The software PANalytical X'Pert reflectivity was used to fit the XRR data using a two-layer model of the Elemental composition of the films was obtained using Rutherford backscattering spectrometry (RBS) and time-of-flight elastic recoil detection analyses (ToF-ERDA). The measurements were carried out in a 5-MV NEC-5SDH-2 pelletron tandem accelerator. 2 MeV 4 He + ions were employed for RBS and detected in a scattering angle of 170°. Two different geometries, azimuth angle 5°+tilt angle 2° and azimuth angle 40°+tilt angle 2°, were chosen in order to minimize channelling effects. In addition, more suppression of the probable channelling effects was undertaken by multiple small random-angular movements around the equilibrium angles within a range of 2°. RBS spectra were fitted by SIMNRA 7.02 code 24 with an ~1% statistic uncertainty to determine elemental compositions. Recoils, in ToF-ERDA, were detected at 45° angle between the primary beam and a ToF-E detector telescope in a gas ionization chamber (GIC) using a 36-MeV 127 I 8+ beam incident at 67.5° with respect to the sample surface normal.
The ToF-E detector telescope consisted of two circular carbon foils with 8 and 5 µg/cm 2 thicknesses, 6 mm radius, a 0.05-msr solid angle (ΔΩ), and a flight distance of 425 mm between the foils. Utilizing a ToF-GIC setup provides the system with a good energy resolution and enhanced ion species separation in terms of mass and energy. 25 Average elemental compositions were also obtained from ToF-ERDA time-energy coincidence spectra using two different software packages, CONTES 26 and Potku 27 . Systematic uncertainties of the experiment, discussed in more detail elsewhere 28 in particular for light elements, were estimated to be a maximum of 5-10%, whereas statistic uncertainties arisen from the number of experimental counts were ˂ 2.3%. However, the relative elemental concentrations were obtained with higher accuracy 29,30 . The stopping power data required for both RBS and ERDA simulations was retrieved from SRIM2013 code. 31 Cross-sectional transmission electron microscope (TEM) samples were prepared by the traditional sandwich approach, which includes sample cutting, gluing, polishing and ion milling. A Gatan Precision Ion Polishing System Model 691 operated at 5 kV and 40 mA, with an Ar ion source, was used to make the samples electron transparent. Scanning transmission electron microscopy (STEM) and selective area electron diffraction (SAED) characterization were performed using the Linköping double Cs corrected FEI Titan 3 60-300 operated at 300 kV.

Results and Discussion
Tris(dimethylamido)gallium(III) dimer 1 has been previously used in ALD studies, however its thermal properties have not been reported as far as we know. Thermogravimetric analysis (TGA) showed 1 evaporated in a single step from 140-230 °C, with only 5% of residual mass ( Figure 2). The 1 torr vapour pressure of 1 was shown to occur at 109 ± 5 ºC its ∆H of vaporization was 59.5 kJ mol -1 ( Figure S1). Differential scanning calorimetry (DSC) showed an endothermic peak at 103 °C and a small broad exothermic event with a peak at 193 °C. The endothermic event peak corresponds to the melting point of 1 (103-106 °C), which matches our observed DSC value, whilst the exothermic event indicates high temperature decomposition.
These results show that 1 is a thermally stable precursor and has favourable properties for use in ALD of GaN. To study if 1 could be used in self-limiting deposition of GaN, ALD experiments on Si (100) substrates were undertaken using varied pulse times for 1 and NH3 plasma whilst the temperature was maintained at 200 °C. The GaN growth per cycle (GPC) saturates at 1.4 Å/cycle when the pulse time of 1 is 4 s or longer (Figure 3a). The GPC saturates at the same value for NH3 plasma when pulse times of 6 s or longer were used. This is indicative of a selflimiting growth and surface chemistry between 1 and NH3 plasma. The growth per cycle was constant between 130-250 °C when using a 4 s pulse time for 1 and 9 s NH3 plasma pulse time ( Figure 3b). As the bubbler temperature was set to 120 °C , lower deposition temperature was not conducted due to the risk of condensation of 1 on the substrate. Reactor temperatures of ³250 °C resulted in a decreased GPC, indicating desorption of the precursor from the substrate.

(c) (d)
The chemical bonding environment and atomic composition in the GaN films deposited at 250 °C were analysed by XPS ( Figure 5). High resolution XPS spectra were used to analyse the chemical environment of the Ga 2p3/2 and N 1s regions. Two peaks were used to obtain a good fit for the Ga 2p3/2 at 1117.5 eV and 1119.4 eV, which were attributed to Ga-N and Ga-O bonds, respectively (Figure 5a). Three peaks were used to fit the N 1s at 397.1 eV, 394.8 eV and 392.7 eV, which were assigned to N-Ga and two Ga Auger peaks, respectively (Figure 5b). This is in accordance with previous XPS measurements on GaN thin films. 33 showing the latter is preferred for detection of impurities. In addition, sputtering was used in the XPS measurement, which could cause preferential sputtering and change the composition and impurity levels in the film. Nonetheless, XPS is highly important for accessing the chemical environment of each element.  (Figure 6d). This shift of the Fermi level indicates that the deposited GaN is n-type doped. This is expected as GaN normally shows n-type behaviour due to impurities and defects in the film such as oxygen and nitrogen vacancies. 38,39,40 Intensity (arb. units) Binding energy (eV)  By replacing the Si (100) substrate with 4H-SiC (0001), which is a more suitable substrate for GaN with respect to the lattice mismatch, XRD in q-2q geometry showed only the GaN (0002) and (0004) Figure S2. This is on par with previous high temperature CVD study of GaN films were an AlN seed layer was deposited on the SiC substrate. 8 . These findings could revolutionize the transistor industry, especially for the production of high electron mobility transistors (HEMTs). Being able to deposit very thin GaN directly on SiC without the need for an AlN seed layer allows for increased design freedom when producing HEMTs, which could in turn improve their performance. Finally, a thermal ALD route to GaN was investigated using 1 and NH3 without plasma discharge. The same pulse parameters were used as the plasma process, but a 16 s of NH3 pulse was used in a temperature range of 150-400 °C . GaN films could be obtained also via this thermal process, however, the films were X-ray (q-2q and GIXRD) amorphous. The XPS analysis showed the films contained high amounts of C and O impurities, 4 at.% and 15 at.%, respectively ( Figure S3). The Ga/N ratio could not be accurately be measured due to the Ga Auger overlap with N 1s. The high level of impurities could explain the amorphous nature of these films, which are likely to distort the long-range order of the atoms in the lattice. This also highlights the efficiency of the NH3 plasma to remove carbon and oxygen impurities from the growing film surface.

Conclusions
We have successfully been able to deposit GaN on Si (100) and 4H-SiC (0001) using tris(dimethylamido)gallium(III) together with NH3 by both thermal and plasma ALD. The NH3 plasma process showed surface saturation could be obtained with a GPC of 1.4 Å/cycle after 4 s of 1 and 6 s of NH3 plasma. The growth per cycle was independent of temperature between 130-250 °C and the growth rate was 3-5 times higher than previously reported ALD of GaN.
The deposited film at 250 °C showed a Ga/N ratio of 0.97 with 2.8 at.% of C and 3 at. By being able to deposit high quality GaN films at these low temperatures also present the opportunity to deposit on temperature sensitive materials which has previously been hindered due to the high deposition temperatures. The ability to deposit GaN directly on SiC also reduces the number of material interfaces which could potentially improve the transistor performance.   Intensity (arb. units)