Growth of gallium oxide thin films from gallium acetylacetonate by atomic layer epitaxy Minna Nieminen," Lauri Niinisto"" and Eero Rauhalab "Helsinki University of Technology, Laboratory of Inorganic and Analytical Chemistry, FIN-021 50 Espoo, Finland bUniversity of Helsinki, Department of Physics, Accelerator Laboratory, FIN-0001 4 Helsinki, Finland Gallium oxide thin films have been deposited by atomic layer epitaxy (ALE) using Ga(acac), (acac =pentane-2,4-dionate) and either water or ozone as precursors. Films were grown on silicon (loo), soda lime and Corning glass substrates. The influence of the deposition parameters (e.g. pulse duration, growth and source temperatures) on film growth were studied and by a proper choice of the parameters a self-controlled growth was demonstrated around 370 "C.Spectrophotometry, X-ray diffraction (XRD), Rutherford back-scattering spectroscopy (RBS) and X-ray photoelectron spectroscopy (XPS) were used to determine the refractive index, thickness, crystallinity and stoichiometry of the films. All the films were amorphous and highly uniform with only small thickness variations. The films deposited with water contained a considerable amount of carbon as an impurity whereas ozone as an oxidizer gave stoichiometric Ga203 films. Gallium oxide is a thermally and chemically stable material which is insulating at room temperature but semiconducting at higher temperatures. The n-type semiconducting property of gallium oxide is due to a slight oxygen deficit in the crystal lattice. Ga203 is the single stable oxidation state of gallium under normal conditions.It exists in several crystalline forms, of which the monoclinic, high temperature P-Ga203 modifi- cation is the most stable one. Because of its optical and electrical properties, gallium oxide has found applications in metal-insulator structures on GaAs' and facet coatings for GaAs-based lasers. Recently, the use of gallium oxide thin films as gas sensors has attracted increasing intere~t.~-~ At high temperatures (800-1000°C) the films can be used as oxygen sensors while at lower temperatures (<7OO"C) they can be used to make sensors for reducing gases. A small number of articles concerning the growth of gallium oxide thin films has been published so far. Fleischer and co- workers 2,6 have produced stoichiometric Ga203 thin films with a high-frequency sputtering process using an ultrapure Ga203 ceramics target.Macri et uL5 also used the same target when they studied the growth of films by reactive radio- frequency magnetron sputtering. Homogeneous Ga203 films have been deposited by electron-beam evaporation using a single-crystal high-purity Gd,Ga5012 source7 which, however, introduces a few per cent of gadolinium into the films, especially on the surface region.8 A spray pyrolysis process was used to grow stable P-Ga203 and a-Ga20,:Co9 thin films on glass substrates. Hariu et al." deposited gallium oxide by reactive vapour evaporation of gallium in an oxygen atmosphere. To our knowledge, the use of chemical methods [chemical vapour deposition (CVD) and atomic layer epitaxy (ALE)] in the growth of gallium oxide thin films has not been reported. The ALE process has certain unique features when compared to conventional thin film deposition techniques." The growth rate is not dependent on the rate of the reactions provided that the dose in each reaction step is high enough to give a monolayer coverage of the surface.Accordingly, growth is uniform over large areas and no thickness monitoring is needed because it is determined by the number of reaction cycles. The relatively high substrate temperature eliminates any weak bonds from the surface by re-evaporation. Together with the lack of vapour-phase reactions, this results in highly stable and stoichiometric films.Nucleation phenomena are appar- ently modified to allow predominantly two-dimensional (2D) nucleation ensuring very uniform layers even in ultra-thin structures or conformal coatings. ALE has been mainly applied to the growth of 111-V and 11-VI semiconducting thin films and oxide layers.12 This work is part of our studies of using modified aluminium oxide thin films as catalyst supports.13 One possible way to modify alumina is to replace some of the cations by other trivalent cations. Because the properties of gallium are similar to those of aluminium, we have chosen gallium to replace some of aluminium in the A120, structure. In this paper we have studied the growth of gallium oxide thin films in a flow- type atomic layer epitaxy (ALE) reactor using Ga(acac), and water, oxygen or ozone as precursors.In the ALE method the source materials are alternately pulsed into the reactor chamber.14-16 Between the reactant pulses the excess of reactant and gaseous side-products are purged out with an inert gas pulse. In an ideal case, one monolayer of the first reactant is chemisorbed on the substrate and this monolayer reacts with the second precursor pulsed onto the substrate resulting in the formation of a solid film. A controlled layer-by-layer growth is achieved by repeating this reaction cycle. The influence of the deposition parameters such as pulse duration, pressure, substrate and source temperature on the film growth were studied in detail.Experimental Gallium oxide thin films were deposited in a flow-type ALE reactor as described el~ewhere.'~,~~ The reactants were alter- nately introduced into the reactor and nitrogen with a purity of 99.999% was used as a carrier and purging gas. The source material for gallium was Ga(acac), (acac =pentane-2,4-dionate) which was synthesized from 99.999% GaC1, (Strem Chemicals, Inc.) using the method described by Belcher et al." Water, oxygen or ozone were used as an oxygen source. Ga(acac), was evaporated from an open aluminium crucible held at 130°C in the source furnace. Water was contained in thermo- statted glass reservoir held at 20 "C and it was introduced into the reactor through a capillary by means of its own vapour pressure.Ozone was produced by feeding oxygen gas (99.998%) to the reactor through an ozone generator (Fischer model 502). The concentration of ozone was cu. 10% (60 g mP3) and the gas flow rate during the pulse was about 60cm3 min-' (measured for the oxygen gas). The film deposition took place at a reduced pressure (approx. 1.5 mbar) in the temperature J. Muter. Chem., 1996, 6(l),27-31 range 350-400°C. The effect of the source temperature of Ga(acac), (120-150 "C)as well as the duration of the precursor pulses C2000-2750 ms for Ga(acac),, 2000-3500 ms for H20 and 600-1200 ms for O,] and the purge pulses (2000-4500 ms) on the film growth were studied. Soda lime glass (5 x5 cm2), Corning 7059 glass (5 x5 cm2) and silicon( 100) were used as substrates.Thermal analysis was used to study the thermal behaviour of Ga(acac),. Simultaneous thermogravimetry (TG) and differential thermal analysis (DTA) curves were recorded in a Seiko TG-DTA instrument of the SSC 5200 series. A pressure of 1-2 mbar and nitrogen atmosphere were chosen to simulate the growth conditions in the ALE reactor. The heating rate was 10 "C min-' and sample weight was ca. 10 mg. Thicknesses and refractive indices of the films were evaluated by fitting the transmittance and reflection spectra1* measured with a Hitachi U-2000 double-beam spectrophotometer in the region of 370-1100 nm. The crystallinity of the films was determined by XRD measurements with a Philips MPD1880 powder diffractometer using Cu-Ka radiation.Film composition and stoichiometry were determined by Rutherford back-scattering spectroscopy (RBS) using ions from the 2.5 MV Van de Graaff accelerator of the University of Helsinki. Both 4He and 'H ions with a scattering angle of 170" were used. By varying the 'H beam energy, both oxygen and carbon could be detected with good sensitivity due to nuclear potential and nuclear resonance scattering-enhanced cross- sections. Film thicknesses and possible heavier impurities, on the other hand, are readily revealed by the 4He spectra. X-Ray photoelectron (XP) spectra were recorded with a Kratos Analytical XSAM 165 spectrometer using a monochromatic aluminium X-ray source. The intention of the analysis was to determine whether the films were stoichiometric throughout the bulk and also to detect any carbon remaining in the films.Therefore, the films were sputtered until there was no change in the intensities of theo C Is, 0 1s and Ga 3d peaks. This depth was cu. 100-150 A. The charge-up shifts, depending on the insulating sample were corrected using the C 1s signal (284.6 eV) from the unsputtered sample. Results and Discussion Optimization of growth parameters Deposition with water. In the preliminary studies the subli- mation behaviour of Ga(acac), was studied by varying the source temperature. The experiments were performed using a growth temperature of 385°C and a reduced pressure of cu. 1.5 mbar. The pulse durations of Ga(acac), and H,O were 600 and 2000 ms, respectively.When the source temperature was above 120°C a constant growth rate was observed. The sublimation behaviour of Ga(acac), was verified by TG-DTA studies which indicated a sharp and complete volatilisation at around 130 "C under a reduced pressure of 1-2 mbar (Fig.1). The dependence of the growth rate on the growth tempera- ture is shown in Fig. 2. As a result of preliminary studies the pulse durations of Ga(acac), and ,H,O were chosen to be 2250 and 2500 ms, respectively. The growth ?ate increased with increasiq temperature, namely from 0.25 A per cycle at 350 "C to 0.55 A per cyclc at 400°C. A temperature-independent growth rate of 0.33 A per cycle was obtained between 365 and 380 "C, indicating a very narrow plateau of self-controlled ALE growth.When the growth temperature was 380°C or lower, the films were transparent and had a green colour of inter- ference, but at temperatures over 380°C the films became less transparent and also the interference colour was changed to brown-green. In order to verify the self-controlled growth of the films, the effect of source and purge pulse durations on the film growth rate at 370°C were studied. The dependence of the growth 28 J. Muter. Chew., 1996,6( l), 27-31 I"'" .-F-5fL 3 E 1' I -10 55 110 163 220 275 330 385 440 495 5SO TIT Fig. 1 Thermoanalytical curves for Ga(acac), recorded with a heating rate of 10°C min-' in flowing nitrogen at a pressure of 1-2mbar. The sample mass was 13 mg. 0.8 0.7-I-a, 0.6 02 0.53 0.4 0.3 sg 0.2 L UJ0.1 04 340 350 360 370 380 390 400 410 depositiontemperaturePC Fig.2 Dependence of the growth rate on the deposition temperature. Pulse durations were 2250 ms for Ga(acac), and 2500 ms for H20. rate on H20 pulse [Ga(acac), 2250 ms] is presented in Fig. 3(u). The saturation of the growth rate at a constant level is seen when the water pulse duration is longer than 2500 ms. This indicates that a certain minimum pulse duration is needed to obtain a full coverage of the substrate. The dependence of the growth rate on the Ga(acac), pulse (H20 pulse 3000 ms) is shown in Fig.3(b). The growth rate was found to be indepen- dent of the Ga(acac), pulse dutation between 2000 and 2800 ms.This confirms that the growth is self-controlled. The purge gas pulse durations had no effect on the growth rate, indicating that the constant surface state was achieved. The effect of the total pressure on the growth rate was also studied between 0.5 and 2.5 mbar. The consumption of the precursor was the same at all pressures and no effect on the film growth rate was observed. Deposition with oxygen. The optimized parameters obtained for gallium oxide film growth using water as an oxygen source were then used in experiments where oxygen was studied as an oxidizer. However, when oxygen was used no film growth was detected. Varying the substrate temperature (350-390 "C) also had no effect on film growth. This indicates that oxygen is not reactive enough to facilitate film growth at the tempera- tures studied.Deposition with ozone. The dependence of the growth rate on the growth temperature is presented in Fig. 4. The source temperature for the gallium precursor was 130 "C and the pulse durations of Ga(acac), and 0, were 2250 and 600 ms, respectively. Molsa and co-workers studied the growth of Ce021g and Y20320using ozone as an oxygen source, and according to their results the growth rate saturated to a c 0.1o'2 1"/;::;I go 50 lsoo 2000 2500 3ooo 3500 4ooo water pulse durationjms 0 -0.3 O+ I 180 2OOO 2200 2400 2600 2800 3000 Ga(a~ac)~pulse duration/ms Fig. 3 Dependence of the growth rate at 370°C on the H20 (a) and Ga(acac), (b)pulse durations constant value when the pulse duration of ozone was longe? than 400 ms.A temperature-independent growth rate of 0.22 A per cycle was obtained between 350 and 375°C. When the growth temperature was raised above 375°C the growth rate increased as a consequence of decomposition of the precursor, i.e. the growth took place via CVD. In order to compare better the results obtained using water as an oxidizer, a substrate temperature of 370 "C was chosen for further experiments. The dependence of the growth rate on the ozone pulse duration [Ga(acac),, 2250 ms] is shown in Fig. 5(u). When the ozone pulse is longer than 800 ms saturation of the growth rate at a constant level of 0.28 A per cycle is indicated. Similarly, the growth rate was found to be independent of the Ga(acac), pulse duration between 2000 and 2500 ms [Fig.5(b)]. These results verify that the thin film growth is self-controlled ALE growth. Dependence of the film thickness on reaction cycles The dependence of the film thickness on the reaction cycles at 370 "C is shown in Fig. 6. When water was used as .an oxygen source a linear relation with a growth rate of 0.33 A per cycle was observed. Similarly, using ozone a! an oxidizer gave a linear relation with a growth rate of 0.28 A per cycle. However, these growth rates are much less than one monolayer per cycle, which is probably caused by steric hindrance due to the 0.8 0.7 + II 340 350 380 370 380 390 400 410 deposition temperaturePC Fig.4 Growth rate as a function of temperature. Ga(acac), and 0, pulse durations were 2250 and 600 ms, respectively. $-600 800 la00 1200 1400 53 0.6 ozone pulse duration/ms I LI I 0 0 0.2 0-3 1 0 is00 1750 2OOO 2250 2500 2750 3OOO Ga(a~ac)~pulse duration/rns Fig.5 Dependence of the growth rate at 370°C on the O3 (a) and Ga(acac), (b)pulse durations bulky Ga(acac), molecule. The higher growth rates of films grown with water may be due to the formation of films which contain some GaO(0H) in analogy with A10(OH),21 since more hydroxy residues are probably left in the films grown with water than in those films grown with ozone. Other properties XRD measurements made on films grown on all substrates and with both oxidizers revealed that films were amorphous. The thickness measurements indicated that films were uniform, showing only small thickness variations (typically < 1%) in the gas flow direction, within the substrate length of 5 cm.The films deposited on soda lime glass and Corning 7059 glass had the same growth rate, whereas the growth rate ?f the films deposited on silicon decreased to yalues of 0.23 A per cycle (water as an oxidizer) and 0.21 A per cycle (ozone). The refractive index at a wavelength of 580nm had a constant value of 1.8 when water was used as the oxygen source. The films grown using ozone as the oxygen source were highly transparent in the visible region and the refractive index of the films at a wavelength of 580 nm was 1.9.Neither film thickness nor deposition temperature was observed to have an effect on the refractive indices. The increase in refractive index for films deposited with ozone as compared with films grown with water 400 350 1 300 250 \2 200 5 150 .-5 100 so 0 0 2oM) 4000 6OOO 8000 10000 12000 number of cycles Fig.6 Dependence of the film thickness on the number of cycles at 370 "C:0,water; A,ozone. The pulse durations for Ga(acac),, H20 and O3were 2250, 3000 and 1000 ms, respectively. J. Mater. Chem., 1996, 6( l), 27-31 29 may be due to higher film densities in the former case. Passlack et uL7 reported refractive indices between 1.84 and 1.89 for thin gallium oxide films grown with electron-beam evaporation, which agrees with our results.The refractive index of bulk Ga203 is between 1.92 and 1.95.22 Stoichiometry and carbon residues were quantitatively deter- mined by RBS. The back-scattering spectra of films grown on silicon( 100) using water and ozone as the oxygen source are illustrated in Fig. 7 and 8, respectively. The solid lines represent the theoretical simulations by a computer program.23 A small linear background has been added to the simulated silicon signal in Fig. 8 to account for the low energy tail. Helium ion scattering at 2000 keV was mainly used for determining the layer thickness and possible impurities. In addition, the 12C('H, 'H)12C resonance at 1735 keV24 and proton nuclear potential scattering at 2500 keV25 were used to determine the carbon and oxygen contents, respectively. The same sample structure was, however, indicated by all energies and by both He and H ions.Fig. 7 shows the proton scattering yields from C, 0 and Ga at 2000 keV. At 1735 keV the carbon signal is enhanced by about 10 times relative to that shown in Fig. 7. The films grown using water as the oxidizer had a carbon impurity content of ca. 30 atom% (25 atom% Ga, 45 atom% 0 with a total of 2.26 x 10'' atom cm-2) as determined by RBS. The XPS analysis discussed below gave a lower value 'H ion energy/keV Fig.7 Back-scattering spectrum for 2.0MeV 'H ions incident on a 320 nm thick Ga,,,O,.,C, sample on a silicon substrate (-, experimen-tal data; -, theoretical spectrum). The precursors used were Ga(acac), and water.The theoretical Ga, 0,Si and C signals are shown. 4He ion energylkev Fig. 8 Rutherford back-scattering spectrum for 2.0 MeV ,He ions incident on a 197 nm thick Ga203 film on a silicon substrate (-a, experimental data; -, theoretical spectrum). The precursors used were Ga(acac), and ozone. The theoretical Ga, 0 and Si signals are shown. 30 J. Muter. Chem., 1996, 6(l), 27-31 (< 9 atom%) but nevertheless indicated a significant amount of carbon in the film. The effect of preferential sputtering of carbon at the surface is believed to account for a lower XPS quantification of carbon compared with RBS results. The gallium precursor is responsible for the carbon impurity of the films. Since the growth of films with water was found to be very controllable at 370°C it is not likely that carbon is due to the thermal decomposition of the Ga(acac), ligand in the gas phase.Obviously the adsorbed Ga(acac), molecule reacts with the H20 pulse with a different mechanism than with 03, resulting in carbon residues in the former case. The reactor temperature is too low (370 "C) and water as the precursor is not reactive enough to drive off the carbon residues. Recently, Yoshida et al. found in CVD-deposited SrTiO, a carbon content of 7 atom% even at 6OO0C,but this could be reduced by increasing the oxygen partial pressure.26 The films deposited with ozone were stoichiometric, almost pure Ga203 films with ca. 1 atom% carbon impurity (total area density 1.78 x lo1* atom cm-2) as determined by proton back-scattering at 2500 and 1735 keV.There was no difference in the atomic composition between films grown on soda lime glass and silicon. No impurities other than carbon were detected. Chlorine, for example, with the possible signal located between 4He ion energies 1100 and 1300 keV, (Fig. 8) had an impurity content below the detection level of 0.2 atom%. The amount of atoms (expressed in atom cm-2) were used with the thicknesses obtained by the optical transmission or reflection spectroscopy to estimate the densities of the films. The calcu- lated densities for films grown on silicon and soda lime glass were 5.6 and 5.3 g ~m-~, respectively. These values are lower than that of bulk P-Ga203 (5.88 g ~m-~).~' The XP spectrum of a 170nm thick Ga,03 film deposited using ozone as the oxidizer on a silicon substrate is presented in Fig.9(a) (before sputtering) and (b)(after sputtering). Since the ratio of gallium :oxygen atoms in the film was 3959, the film was shown to be close to stoichiometry using the standard sensitivity factors. The carbon concentration in films deposited with ozone was, after sputtering, close to zero (around 1 atom% but dominated by noise in the spectrum). No other binding energylev Fig.9 XP spectrum of a 170nm thick Ga,O, film on a silicon substrate. (a) XPS survey (before sputtering), (b) Ga 3d peak (after sputtering). impurities were detected. On the other hand, the films grown with water contained a higher concentration of carbon (around 9 atom%) which was not removed by sputtering. The Ga 3d peak was found at 20.5 eV and the full width at half-maximum (FWHM) was 1.6 eV [Fig.9(b)]. The binding energy of Ga 3d 3 4 5 6 M. Fleischer and H. Meixner, Sensors Actuators B, 1991,4,437. W. Hanrieder and H. Meixner, Sensors Actuators B, 1991,4,401. P. P. Macri, S. Enzo, G. Sberveglieri, S. Groppelli and C. Perego, Appl. Surf. Sci., 1993,65166,277. M. Fleischer, W. Hanrieder and H. Meixner, Thin Solid Films, 1990, 190,93. photoelectrons corresponded to that given in ref. 27 for Ga203 powder, indicating that gallium is present as Ga3+ in the films. 7 M. Passlack, E. F. Schubert, W. S. Hobson, M. Hong, N. Moriya, S. N. G. Chu, K. Konstadinidis, J.P. Mannaerts, M. L. Schnoes and G. J. Zydzik, J. Appl. Phys., 1995,77,686. 8 M. Passlack, M. Hong, E. F. Schubert, J. R. Kwo, J. P. Mannaerts, Conclusions S. N. G. Chu, N. Moriya and F. A. Thiel, Appl. Phys. Lett., 1995, 66, 625. The present study demonstrates that by using a chemical deposition method (ALE), Ga203 thin films can be grown on glass and silicon substrates. The precursors used were Ga(acac), and either water or ozone. No film growth was detected when oxygen was used as the oxidizer. The growth rate of the films was dependent on the pulse durations of the 9 10 11 12 13 H-G. Kim and W-T. Kim, J. Appl. Phys., 1987,62,2000. T. Hariu, S. Sasaki, H. Adachi and Y. Shibata, Jpn. J. Appl. Phys., 1977, 16, 841. T. Suntola and J. Hyvarinen, Ann.Rev. Muter. Sci., 1985, 15, 177. L. Niinisto and M. Leskela, Thin Solid Films, 1993,225, 130. M. Nieminen, L. Niinisto and R. Lappalainen, Mikrochim. Acta, 1995, 119, 13. precursors and on the growth temperature, but a narrow plateau of self-controlled ALE growth was observed around 370°C. The growth rate was much less than one monolayer per cycle and this is probably caused by steric hindrance due to the bulky Ga(acac), molecule. The films were amorphous and highly uniform with only small thickness variations over 14 15 16 17 T. Suntola, Muter. Sci. Rep., 1989,4,261. M. Leskela and L. Niinisto, in Atomic Layer Epituxy, ed. T. Suntola and M. Simpson, Blackie and Son, Ltd., Glasgow, 1990,p.1. T. Suntola, in Handbook of Crystal Growth, vol. 3, ed. D. T.J. Hurle, Elsevier, Amsterdam, 1994, p. 601. R. Belcher, C. R. Jenkins, W. I. Stephen and P. C. Uden, Tulunta, the substrate area of 5 x 5 cm2. RBS and XPS measurements 1970, 17,455. indicated that the films deposited with water contained a considerable amount of carbon as an impurity, whereas the films grown with ozone were stoichiometric and had a carbon content of ca. 1 atom%. No other impurities were detected. 18 19 20 M. Ylilammi and T. Ranta-aho, Thin Solid Films, 1993,232, 56. H. Molsa and L. Niinisto, Muter. Res. SOC. Syrnp. Proc., 1994, 335, 341. H. Molsa, L. Niinisto and M. Utriainen, Adv. Mater. Opt. Electr., 1994,4,389. 21 L. Hiltunen, H. Kattelus, M. Leskela, M. Makela, L. Niinisto, The authors wish to thank Dr. Kevin S. Robinson (Kratos Ltd., Manchester, UK) for XPS measurements. The assistance of Ms. Heini Molsa and Mr. Pekka Soininen is gratefully acknowledged. This work was suppmkd in part by the 22 23 E. Nykanen, P. Soininen and M. Tiitta, Muter. Chem. Phys., 1991, 28, 379. CRC Handbook of Chemistry and Physics, CRC Press, Boca Raton, FL, 69th edn., 1988, p. B-92. J. Saarilahti and E. Rauhala, Nucl. Instrum. Methods Phys. Res. B., Academy of Finland. 24 1992,64,734. E. Rauhala, Nucl. Instrum. Methods Phys. Res. B., 1985, 12,447. 25 M. Luomajarvi, E. Rauhala and M. Hautala, Nucl. Instrum. References 1 A. Callegari, P. D. Hoh, D. A. Buchanan and D. Lacey, Appl. Phys. Lett., 1989,54, 332. 26 27 Methods Phys. Res. B., 1985,9,255. M. Yoshida, H. Yamaguchi, T. Sakuma, Y. Miyasaka, P-Y. Lesaicherre and A. Ishitani, J. Electrochem. Soc., 1995,142,244. R. Carli and C. L. Bianchi, Appl. Surf. Sci., 1994,74,99. 2 M. Fleischer, L. Hollbauer and H. Meixner, Sensors Actuators B, 1994,18-19, 119. Paper 5/03276H; Received 22nd May, 1995 J. Muter. Chem., 1996, 6(1), 27-31