- Email: [email protected]
Effects of hydrogen on GaN metalorganic vapor-phase epitaxy using tertiarybutylhydrazine as nitrogen source
ARTICLE IN PRESS
Journal of Crystal Growth 266 (2004) 347–353
Effects of hydrogen on GaN metalorganic vapor-phase epitaxy using tertiarybutylhydrazine as nitrogen source Yu Jen Hsu, Lu Sheng Hong*, Jyh Chiang Jiang, Jing Chong Chang Department of Chemical Engineering, National Taiwan University of Science and Technology, 43, Keelung Road, Section 4, Taipei 106, Taiwan
Abstract Gallium nitride (GaN) layers have been synthesized from mixtures of trimethylgallium, tertiarybutylhydrazine (TBHy), and hydrogen (H2) at 958 K using the metalorganic vapor-phase epitaxy technique. The role of H2 in the reaction system is explored by investigating the film property changes at various H2 concentrations. Epitaxial growth of GaN with carbon incorporation less than 3 mol% is accomplished at a low H2 concentration of 20 mol%, while it turns gradually into polycrystalline growth with high amount of carbon incorporation when the H2 concentration is increased up to 90 mol%. Density functional theory calculations are performed to examine the thermochemistry of the H2 elimination reaction of gaseous TBHy on Ga-terminated GaN surface. The result suggests that at high H2 concentrations, a C–N containing species derived from TBHy due to incomplete b-hydride elimination may be ascribable to the appearance of polycrystalline phase and high amount of carbon incorporation. r 2004 Elsevier B.V. All rights reserved. PACS: 81.15.Kk; 77.84.Bw; 03.67.Lx Keywords: A1. Density functional theory; A3. Metalorganic chemical vapor deposition; B1. Gallium nitride; B1. Tertiarybutylhydrazine
1. Introduction Gallium nitride (GaN) is a promising material for applications in photonic devices working under drastic circumstance because it shows unequaled properties, such as large direct band-gap energy, higher thermal conductivity, large free exciton, and optical phonon energies. Both GaN and its related alloys have become the most favorable III–V semiconductor for manufacturing short*Corresponding author. Tel.: +886-2-2737-6650; fax: +8862-2737-6644. E-mail address: [email protected] (L.S. Hong).
wavelength devices such as blue–ultraviolet light-emitting diodes (LEDs) and laser diodes. The successful development of GaN-based bluelight LED has made the full color LED display come true. Up to now, epitaxial growth of GaN is accomplished on a sapphire substrate by the metalorganic chemical vapor deposition (MOCVD) technique using trimethylgallium (TMG) and ammonia (NH3) as the reactants [1–4]. However, high deposition temperature (usually larger than 1273 K) and large V/III ratio (over several thousand times) are necessary to acquire stoichiometric GaN layers. This crucial requirement can be attributed to the high N–H
0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.02.065
ARTICLE IN PRESS 348
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
bond strength (93.3 kcal/mol [5]) of the NH3 molecule, which makes it difficult to dissociate enough amounts of active nitrogen species to react with TMG molecules. Meanwhile, the imperative conditions create three problems in the established process. First, the heating elements easily lose function due to the corrosive nature of NH3 at such high operating temperatures. Second, the incorporation of indium to fabricate InGaN/GaN heterostructures turns difficult because of the low dissociation temperature of InN (B773 K) [6]. Third, high process temperature prohibits the choice of substrate other than sapphire. Therefore, finding a nitrogen source alternative to NH3 that can decompose at lower temperatures has turned into a focus of recent research interests. Through many nitrogen sources that can decompose at low temperature, undoubtedly, hydrazine (N2H4) is the first candidate to replace NH3 as a new nitrogen precursor for its quite low decomposition temperature of about 700 K [7]. Unfortunately, the high toxicity and explosive nature restrict the choice of N2H4 as a CVD precursor. In contrast, hydrazine-based compounds, such as monomethylhydrazine and dimethylhydrazine (DMHy), which show moderate chemical properties, have been proposed as nitrogen sources for epitaxial growth of GaN [8–10]. Many reports do show quite positive results in lowering the synthesis temperature and the V/III precursor feed ratio. However, carbon contamination seems a negative problem for these compounds. It is suggested that the stronger C–N bond energy (72.9 kcal/mol) than N–N (38.3 kcal/ mol) [5] in DMHy is responsible for the carbon incorporation into GaN layers due to the formation of carbon containing fragments such as N(CH3) and N(CH3)2 during dissociation process. Another newly developed compound named tertiarybutylhydrazine (TBHy) [11–13] has also been applied for epitaxial growth of GaN. In general, the C–N bond strength between t-butyl ligand and nitrogen is less than that between the methyl or ethyl group and nitrogen. Therefore, the C–N bond in TBHy is expected to dissociate at lower decomposition temperatures than DMHy and should have less possibility of carbon incor-
poration into GaN layers when TBHy is used as the nitrogen precursor. We have recently studied the growth kinetics of GaN epitaxy from both the TMG/DMHy and the TMG/TBHy reaction systems [14,15]. Our results show that stoichiometric GaN layers could be obtained at a V/III precursor feed ratio less than 20 for both the reaction systems. In addition, the GaN layers formed using TBHy show much less carbon incorporation than those formed using DMHy. A further endeavor to suppress the carbon incorporation shows that adding a large amount of hydrogen (H2) to the TMG/DMHy reaction system could reduce the carbon content down to several molar percentages. A plausible explanation is that the possible carbon contaminants N(CH3)x would be stabilized by H2 addition. Here in this article, the role of H2 in the TMG/ TBHy system which shows completely reverse character to that in the TMG/DMHy system is reported. The properties of the deposited layers at various H2 concentrations in the TMG/TBHy system are investigated. Moreover, model reactions are devised to explore the H2 effects using a computational chemistry prediction based on the hybride Hartee–Fock/density functional level of theory (DFT).
2. Experimental and computational details A vertical-type CVD reactor was used to perform GaN deposition experiments. The precursors used were TBHy (Japan Pionics) and TMG (US. Epichem) for nitrogen and gallium sources, respectively. A Pd-membrane purifier (Johnson Matthey) was used to purify H2 gas, while an adsorption-type column was used for N2 purification. The ultra-pure H2, after regulated by a mass flow controller, was passed through a vaporizer to carry out the precursor vapor. The partial pressure of TMG was 2 103 Torr and the V/III precursor feed ratio was kept constant at 20. To explore the effects of H2 on film properties, the precursor streams were premixed with a dilute gas stream composed by H2 and/or N2 gases. The flow rate of the dilute H2 was varied by suitably adjusting the N2 flow rate keep a constant total
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
flow rate of 1000 sccm. The total pressure during deposition was kept constant at 5 Torr. Deposition experiments were performed on GaN-grown on sapphire substrates (GaN (2 mm)/ a-Al2O3) which were fabricated by a commercialized Aixtron MOCVD apparatus using the TMG/ NH3 reaction system at 1323 K. Therefore, all the film depositions in this work were basically a homogeneous growth process. A substrate was placed on a SiC-coated graphite susceptor and indirectly heated by a resist heater beneath the susceptor. A proportional integral derivative controller was used to manipulate the temperature of the susceptor with a thermocouple embedded inside the susceptor holder. An estimation of the heat conduction between the susceptor holder and the substrate was made which showed little difference between them. Therefore, the susceptor temperature was taken as an index of the deposition temperature. All the deposition experiments were performed at 958 K. After deposition experiments, the properties of the deposited films were evaluated by the following techniques. The crystalline structure was examined using a low-angle X-ray diffraction (XRD) technique to exclude the signal from the substrate. The crystalline bonding orders of the deposited layers were analyzed using Raman scattering spectroscopy. The bonding state of the deposited film was analyzed using the Fourier transform infrared spectrometer (FT-IR). A scanning electron microscope (SEM) was applied to observe the surface morphology of the deposited films. The chemical composition of the deposited film was investigated using X-ray photoelectron spectroscopy (XPS) technique. All computations were performed using Gaussian 98 program package [16]. The geometries were fully optimized using the three-parameter exchange functional of Becke [17] with the gradient-corrected correlation functional of Lee et al. [18] (B3LYP) and the standard 6-31G basis set (denoted as B3LYP/6-31G) for the DFT calculations. The calculated electronic energies were corrected to include zero-point energies (ZPEs) and thermal corrections obtained from vibrational frequency calculations. To obtain higher correlation energies, a series of single point calculations
349
on the B3LYP/6-31G geometries were carried out using a larger basis set (aug-cc-pVTZ) [19] (denoted as B3LYP/aug-cc-pVTZ//B3LYP/ 6-31G).
3. Results and discussion Fig. 1 shows two examples of SEM photographs, showing the surface of the deposited layers grown at various H2 concentrations. The film thicknesses, determined from SEM cross-sectional images, are all around 200 nm in spite of different H2 concentrations. Despite the same V/III precursor feed ratio, the surface at the lower H2 concentration (20 mol%) is smooth in morphology (Fig. 1a), while a rough surface with granular structure is observed at the higher H2 concentration (100 mol%) (Fig. 1b). This result suggests that two-dimensional growth prefers at low H2 concentrations. Fig. 2 shows the XRD patterns of the deposited layers as a function of H2 concentration. The major diffraction peak located at about 34.5 is the hexagonal (0 0 0 2) plane for GaN. As the H2 concentration increases, another diffraction peak located at 36 which is assigned to the (1 0 1) plane for GaN appears gradually. In addition, the Raman spectra showing the scattering peaks for
Fig. 1. SEM images of GaN layers deposited on GaN-grown on sapphire substrates at H2 concentrations of (a) 20 and (b) 100 mol%. Deposition conditions: partial pressure of TMG=2 103 Torr, V/III=20, substrate temperature= 958 K, total pressure=5 Torr, total carrier gas flow rate=1000 sccm.
ARTICLE IN PRESS 350
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
Fig. 2. XRD patterns of the GaN layers deposited at H2 concentrations of (a) 20, (b) 50, (c) 80, and (d) 100 mol%.
Fig. 4. FT-IR spectra of the GaN layers deposited at H2 concentrations of (a) 20 and (b) 100 mol%.
Table 1 Composition of the deposited layers grown at various H2 ambient concentrations Sample no.
H2 ambient concentration (%)
Ga (%)
N (%)
C (%)
1 2 3 4
20 50 80 100
48.7 51.1 49.6 45.4
48.5 46.0 46.4 48.2
2.8 3.0 4.0 6.4
Note: Data are determined by ESCA narrow scan analyses in molar percentage.
Fig. 3. Raman spectra of the GaN layers deposited at H2 concentrations of (a) 20, (b) 50, and (c) 100 mol%.
the deposited layers at various H2 concentrations are shown in Fig. 3. The result is characterized by two important features. First, the appearance of the characteristic modes at 567 and 735 cm1 indicates the epitaxial growth of GaN [20]. Second, the intensity of the scattering peaks decreases as the H2 concentration is raised, indicating a comparatively poor epitaxial property for GaN layers deposited at high H2 concentrations. Fig. 4 shows two examples of FT-IR spectra for the samples deposited at various H2 concentra-
tions. Comparing the vibration mode around 1380–1240 cm1 which belongs to the C–N bond stretching [21], the adsorption intensity of the C–N bond stretching for the layer deposited at the higher H2 concentration (Fig. 4b) is stronger than that deposited at the lower H2 concentration (Fig. 4a), which indicates that C–N bond stretching is more pronounced for the samples deposited at high deposition H2 concentrations. This result is quite different from the case when DMHy is used as the nitrogen source [14]. To ascertain the novel behavior of H2 in the TMG/TBHy reaction system, ESCA narrow scan for C 1 s signal is performed to quantitatively analyze the total carbon contamination content in the layers deposited at various H2 concentrations. Table 1 tabulates the compositions of all constituent
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
atoms in the layers deposited at various H2 concentrations. Though the atomic ratio of Ga to N for all the samples are near stoichiometric in spite of different H2 concentrations, the total carbon content which is the evidence of carbon incorporation increases with increasing H2 concentration. From Table 1, it clearly shows that the carbon incorporation amount increases from 2.8 to 6.4 mol% when the H2 concentration is raised from 20 to 100 mol%. Since the carbon content obtained from ESCA measurement shows the same trend with the FTIR measurement result shown in Fig. 4, i.e., C–N bond stretching in the deposited layers increases with H2 concentration, the carbon contamination is mainly originated from a C–N containing species which is most plausibly a derivative from TBHy. The concentration of this C–N containing species is proportional to H2 ambient concentration. In other words, introducing N2 instead of H2 to this reaction system could relieve the appearance of the C–N containing species. This C–N containing species results in not only carbon incorporation but also poor epitaxial property and film morphology. The detailed chemistry remains unclear. A possible reaction mechanism based on a quantum chemistry calculation is proposed as follows. First of all, an adduct composed of TBHy and TMG molecules which is optimized by the B3LYP/6-31G is used as a starting material. That is, the adduct is simulated as a simplified model for a TBHy molecule adsorbed on a Gaterminated GaN surface. Then, the adduct, denoted as TMGTBHy, is used to execute a series of theoretical predictions of some reaction potential energy surfaces. A potential energy surface for C–N bond breaking is performed to compare the experimental results. Fig. 5a shows the result of the potential energy diagram for TMGTBHy. Both the activation energy and the heat of reaction (in kcal/mol) including the ZPE correction and higher level correlation calculations are also shown. A transition state structure with an energy barrier of 72.73 kcal/mol is found since it is characterized by one imaginary frequency of 907.5837 cm1. From this imaginary frequency vibration, two strong displacement vectors origi-
351
Fig. 5. (a) Potential energy diagram for the TMG/TBHy reaction system. Both the activation energy and the heat of reaction calculated in kcal/mol are carried out using the B3LYP/6-31G level and the values in parentheses are achieved using a larger basis set (aug-cc-pVTZ). (b) Potential energy diagram for the TMG/DMHy reaction system. Both the activation energy and the heat of reaction calculated in kcal/ mol are carried out using the B3LYP/6-31G level and the values in parentheses are achieved using a larger basis set (augcc-pVTZ).
nated from two hydrogen atoms: one from t-butyl ligand and the other from hydrazine ligand (NHNH2) are observed. The final predicted products include a hydrazine ligand (NHNH) which is a C–N free species adsorbed on a Gaterminated GaN surface, a gaseous isobutene, and
ARTICLE IN PRESS 352
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
a gaseous H2 molecule as shown as ðCH3 Þ3 Ga : NH2 NHCðCH3 Þ3 -ðCH3 Þ3 Ga : NHNH þ C4 H8 þH2 :
ð1Þ
The predicted result strongly suggests that bhydride elimination reaction does occur in the tbutyl ligand and a H2 molecule is formed eventually. According to this calculation result, the presence of excess H2 will tend to inhibit the forward reaction path of TMGTBHy based on the Principle of LeChatelier. By contrast, low H2 concentrations favor the reaction path to form the C–N free species. This explains well the experimental fact why the carbon contamination amount is lower for the GaN layers grown at low H2 concentrations. Furthermore, the C–N containing species, possibly TBHy itself, due to its complicated molecular structure may show steric hindrance effect during film formation process. This may be the reason for the three-dimensional film growth and the poor crystalline property at high H2 concentrations. An analogous adduct composed of DMHy and TMG denoted as TMGDMHy is also used for the same potential energy surface for C–N bond breaking. Fig. 5b shows the result of the potential energy diagram for the TMGDMHy. The overall reaction equation is given by ðCH3 Þ3 Ga : NH2 NðCH3 Þ2 -ðCH3 Þ3 Ga : NHNCH3 þCH4 :
ð2Þ
Comparing the calculation results in Fig. 5, two important differences are concluded as follows. First, the fragment of the final products adsorbed on the Ga-terminated GaN surface for TBHy is a carbon-free nitrogen species (NHNH), while a methyl-containing nitrogen species (NHNCH3) for DMHy is predicted. This difference may explain why the GaN layers grown using TBHy have much less carbon contamination than those formed using DMHy [15]. Also, the ambiguity of the role of ambient H2 played in each reaction system can be clarified. That is, compared with the retarding effect of H2 for TBHy to form a carbonfree species, adding excess H2 into the TMG/ DMHy reaction system may decrease carbon incorporation through a further reaction of H2 and NHNCH3 which takes away the carbon
contaminant (methyl ligand). Second, the larger endothermic heat of reaction of TMGTBHy (39.39 kcal/mol) over TMGDMHy (4.36 kcal/ mol) indicates that higher reaction temperature may be required for the GaN film growth using TBHy. This result is coincident with the necessary synthesis temperature range reported for the TMG/TBHy reaction system (958–973 K) versus that for the TMG/DMHy reaction system (853– 903 K) [14,15].
4. Conclusion Effects of hydrogen on the metalorganic vaporphase epitaxy of GaN using the TMG/TBHy reaction system are investigated. Deposition experimental results show that adding excess H2 into the reaction system not only increases carbon incorporation but also deteriorates the crystalline property and surface morphology of the GaN layers. The role of H2 during reaction is explored by an approach of computational chemistry prediction to understand the possible carbon incorporation path and surface reaction mechanism. The predicted result shows that the b-hydride elimination reaction occurs in the t-butyl ligand of the adsorbed TBHy to produce a carbon-free nitrogen species (NHNH), a gaseous isobutene and a H2 molecule. As a comparison, a similar simulation when DMHy is used as the nitrogen source is also performed. The result shows that a methyl-containing nitrogen species (NHNCH3) dominates the carbon incorporation behavior. The simulation results explain quite well the completely different role of H2 played in these two reaction systems. An advanced quantum chemical calculation technique that considers much well defined Ga-terminated GaN surface to simulate all possible reaction pathways for these reaction systems is now underway.
Acknowledgements The authors thank National Science Council of Taiwan for financial support under Contract NSC 91-2214-E-011-019.
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
References [1] Y. Ohba, A. Hatano, J. Crystal Growth 145 (1994) 214. [2] C.J. Sun, M. Razeghi, Appl. Phys. Lett. 63 (1993) 973. [3] G.P. Yablonskii, A.L. Gurskii, E.V. Lutsenko, I.P. Marko, B. Schineller, A. Guttzeit, O. Schon, M. Heuken, K. Heime, R. Beccard, D. Schmitz, H.H. Juergensen, J. Electron. Mater. 27 (1998) 222. [4] S. Nakamura, M. Senoh, T. Mukai, Jpn. J. Appl. Phys. 30 (1991) L1708. [5] J. Emsley, The Elements, Oxford University Press, New York, 1991. [6] D.A. Neumayer, J.G. Ekerdt, Chem. Mater. 8 (1996) 9. [7] D.K. Gaskill, N. Bottka, M.C. Lin, J. Crystal Growth 77 (1986) 418. [8] R.T. Lee, G.B. Stringfellow, J. Crystal Growth 204 (1999) 247. [9] H. Saito, T. Makimoto, N. Kobayashi, Jpn. J. Appl. Phys. 35 (1996) L1644. [10] H. Sato, H. Takahashi, A. Watanabe, H. Ota, Appl. Phys. Lett. 68 (1996) 3617. [11] S. Nishide, T. Yoshimura, Y. Takamatsu, A. Ichige, K. Pak, N. Ohshima, H. Yonezu, J. Crystal Growth 189/ 190 (1998) 325. + [12] U.W. Pohl, C. Moller, K. Knorr, W. Richter, J. Gottfriedsen, H. Schumann, K. Rademann, A. Fielicke, Mater. Sci. Eng. B59 (1999) 20. + [13] U.W. Pohl, K. Knorr, C. Moller, U. Gernert, W. Richter, J. Bl.asing, J. Christen, J. Gottfriedsen, H. Schumann, Jpn. J. Appl. Phys. 38 (1999) L105.
353
[14] Y.J. Hsu, L.S. Hong, K.F. Huang, J.E. Tsay, Thin Solid Films 419 (2002) 33. [15] Y.J. Hsu, L.S. Hong, J.E. Tsay, J. Crystal Growth 252 (2003) 144. [16] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, V.G. Zakrzewski, J.A. Montgomery Jr., R.E. Stratmann, J.C. Burant, S. Dapprich, J.M. Millam, A.D. Daniels, K.N. Kudin, M.C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G.A. Petersson, P.Y. Ayala, Q. Cui, K. Morokuma, P. Salvador, J.J. Dannenberg, D.K. Malick, A.D. Rabuck, K. Raghavachari, J.B. Foresman, J. Cioslowski, J.V. Ortiz, A.G. Baboul, B.B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R.L. Martin, D.J. Fox, T. Keith, M.A. Al-Laham, C.Y. Peng, A. Nanayakkara, M. Challacombe, P.M.W. Gill, B. Johnson, W. Chen, M.W. Wong, J.L. Andres, C. Gonzalez, M. Head-Gordon, E.S. Replogle, J.A. Pople, Gaussian 98, Revision A.11.1, Gaussian, Inc., Pittsburgh, PA, 2001. [17] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [18] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785. [19] D.E. Woon, T.H. Dunning Jr., J. Chem. Phys. 98 (1993) 1358. [20] V.Y. Davydov, R.E. Dunin-Borkovski, V.G. Golubev, J.L. Hutchison, N.F. Kartenko, D.A. Kurdyukov, A.B. Pevtsov, N.V. Sharenkova, J. Sloan, L.M. Sorokin, Semicond. Sci. Technol. 16 (2001) L5. [21] Q. Fu, C.B. Cao, H.S. Zhu, Mater. Lett. 42 (2000) 166.
Journal of Crystal Growth 266 (2004) 347–353
Effects of hydrogen on GaN metalorganic vapor-phase epitaxy using tertiarybutylhydrazine as nitrogen source Yu Jen Hsu, Lu Sheng Hong*, Jyh Chiang Jiang, Jing Chong Chang Department of Chemical Engineering, National Taiwan University of Science and Technology, 43, Keelung Road, Section 4, Taipei 106, Taiwan
Abstract Gallium nitride (GaN) layers have been synthesized from mixtures of trimethylgallium, tertiarybutylhydrazine (TBHy), and hydrogen (H2) at 958 K using the metalorganic vapor-phase epitaxy technique. The role of H2 in the reaction system is explored by investigating the film property changes at various H2 concentrations. Epitaxial growth of GaN with carbon incorporation less than 3 mol% is accomplished at a low H2 concentration of 20 mol%, while it turns gradually into polycrystalline growth with high amount of carbon incorporation when the H2 concentration is increased up to 90 mol%. Density functional theory calculations are performed to examine the thermochemistry of the H2 elimination reaction of gaseous TBHy on Ga-terminated GaN surface. The result suggests that at high H2 concentrations, a C–N containing species derived from TBHy due to incomplete b-hydride elimination may be ascribable to the appearance of polycrystalline phase and high amount of carbon incorporation. r 2004 Elsevier B.V. All rights reserved. PACS: 81.15.Kk; 77.84.Bw; 03.67.Lx Keywords: A1. Density functional theory; A3. Metalorganic chemical vapor deposition; B1. Gallium nitride; B1. Tertiarybutylhydrazine
1. Introduction Gallium nitride (GaN) is a promising material for applications in photonic devices working under drastic circumstance because it shows unequaled properties, such as large direct band-gap energy, higher thermal conductivity, large free exciton, and optical phonon energies. Both GaN and its related alloys have become the most favorable III–V semiconductor for manufacturing short*Corresponding author. Tel.: +886-2-2737-6650; fax: +8862-2737-6644. E-mail address: [email protected] (L.S. Hong).
wavelength devices such as blue–ultraviolet light-emitting diodes (LEDs) and laser diodes. The successful development of GaN-based bluelight LED has made the full color LED display come true. Up to now, epitaxial growth of GaN is accomplished on a sapphire substrate by the metalorganic chemical vapor deposition (MOCVD) technique using trimethylgallium (TMG) and ammonia (NH3) as the reactants [1–4]. However, high deposition temperature (usually larger than 1273 K) and large V/III ratio (over several thousand times) are necessary to acquire stoichiometric GaN layers. This crucial requirement can be attributed to the high N–H
0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.02.065
ARTICLE IN PRESS 348
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
bond strength (93.3 kcal/mol [5]) of the NH3 molecule, which makes it difficult to dissociate enough amounts of active nitrogen species to react with TMG molecules. Meanwhile, the imperative conditions create three problems in the established process. First, the heating elements easily lose function due to the corrosive nature of NH3 at such high operating temperatures. Second, the incorporation of indium to fabricate InGaN/GaN heterostructures turns difficult because of the low dissociation temperature of InN (B773 K) [6]. Third, high process temperature prohibits the choice of substrate other than sapphire. Therefore, finding a nitrogen source alternative to NH3 that can decompose at lower temperatures has turned into a focus of recent research interests. Through many nitrogen sources that can decompose at low temperature, undoubtedly, hydrazine (N2H4) is the first candidate to replace NH3 as a new nitrogen precursor for its quite low decomposition temperature of about 700 K [7]. Unfortunately, the high toxicity and explosive nature restrict the choice of N2H4 as a CVD precursor. In contrast, hydrazine-based compounds, such as monomethylhydrazine and dimethylhydrazine (DMHy), which show moderate chemical properties, have been proposed as nitrogen sources for epitaxial growth of GaN [8–10]. Many reports do show quite positive results in lowering the synthesis temperature and the V/III precursor feed ratio. However, carbon contamination seems a negative problem for these compounds. It is suggested that the stronger C–N bond energy (72.9 kcal/mol) than N–N (38.3 kcal/ mol) [5] in DMHy is responsible for the carbon incorporation into GaN layers due to the formation of carbon containing fragments such as N(CH3) and N(CH3)2 during dissociation process. Another newly developed compound named tertiarybutylhydrazine (TBHy) [11–13] has also been applied for epitaxial growth of GaN. In general, the C–N bond strength between t-butyl ligand and nitrogen is less than that between the methyl or ethyl group and nitrogen. Therefore, the C–N bond in TBHy is expected to dissociate at lower decomposition temperatures than DMHy and should have less possibility of carbon incor-
poration into GaN layers when TBHy is used as the nitrogen precursor. We have recently studied the growth kinetics of GaN epitaxy from both the TMG/DMHy and the TMG/TBHy reaction systems [14,15]. Our results show that stoichiometric GaN layers could be obtained at a V/III precursor feed ratio less than 20 for both the reaction systems. In addition, the GaN layers formed using TBHy show much less carbon incorporation than those formed using DMHy. A further endeavor to suppress the carbon incorporation shows that adding a large amount of hydrogen (H2) to the TMG/DMHy reaction system could reduce the carbon content down to several molar percentages. A plausible explanation is that the possible carbon contaminants N(CH3)x would be stabilized by H2 addition. Here in this article, the role of H2 in the TMG/ TBHy system which shows completely reverse character to that in the TMG/DMHy system is reported. The properties of the deposited layers at various H2 concentrations in the TMG/TBHy system are investigated. Moreover, model reactions are devised to explore the H2 effects using a computational chemistry prediction based on the hybride Hartee–Fock/density functional level of theory (DFT).
2. Experimental and computational details A vertical-type CVD reactor was used to perform GaN deposition experiments. The precursors used were TBHy (Japan Pionics) and TMG (US. Epichem) for nitrogen and gallium sources, respectively. A Pd-membrane purifier (Johnson Matthey) was used to purify H2 gas, while an adsorption-type column was used for N2 purification. The ultra-pure H2, after regulated by a mass flow controller, was passed through a vaporizer to carry out the precursor vapor. The partial pressure of TMG was 2 103 Torr and the V/III precursor feed ratio was kept constant at 20. To explore the effects of H2 on film properties, the precursor streams were premixed with a dilute gas stream composed by H2 and/or N2 gases. The flow rate of the dilute H2 was varied by suitably adjusting the N2 flow rate keep a constant total
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
flow rate of 1000 sccm. The total pressure during deposition was kept constant at 5 Torr. Deposition experiments were performed on GaN-grown on sapphire substrates (GaN (2 mm)/ a-Al2O3) which were fabricated by a commercialized Aixtron MOCVD apparatus using the TMG/ NH3 reaction system at 1323 K. Therefore, all the film depositions in this work were basically a homogeneous growth process. A substrate was placed on a SiC-coated graphite susceptor and indirectly heated by a resist heater beneath the susceptor. A proportional integral derivative controller was used to manipulate the temperature of the susceptor with a thermocouple embedded inside the susceptor holder. An estimation of the heat conduction between the susceptor holder and the substrate was made which showed little difference between them. Therefore, the susceptor temperature was taken as an index of the deposition temperature. All the deposition experiments were performed at 958 K. After deposition experiments, the properties of the deposited films were evaluated by the following techniques. The crystalline structure was examined using a low-angle X-ray diffraction (XRD) technique to exclude the signal from the substrate. The crystalline bonding orders of the deposited layers were analyzed using Raman scattering spectroscopy. The bonding state of the deposited film was analyzed using the Fourier transform infrared spectrometer (FT-IR). A scanning electron microscope (SEM) was applied to observe the surface morphology of the deposited films. The chemical composition of the deposited film was investigated using X-ray photoelectron spectroscopy (XPS) technique. All computations were performed using Gaussian 98 program package [16]. The geometries were fully optimized using the three-parameter exchange functional of Becke [17] with the gradient-corrected correlation functional of Lee et al. [18] (B3LYP) and the standard 6-31G basis set (denoted as B3LYP/6-31G) for the DFT calculations. The calculated electronic energies were corrected to include zero-point energies (ZPEs) and thermal corrections obtained from vibrational frequency calculations. To obtain higher correlation energies, a series of single point calculations
349
on the B3LYP/6-31G geometries were carried out using a larger basis set (aug-cc-pVTZ) [19] (denoted as B3LYP/aug-cc-pVTZ//B3LYP/ 6-31G).
3. Results and discussion Fig. 1 shows two examples of SEM photographs, showing the surface of the deposited layers grown at various H2 concentrations. The film thicknesses, determined from SEM cross-sectional images, are all around 200 nm in spite of different H2 concentrations. Despite the same V/III precursor feed ratio, the surface at the lower H2 concentration (20 mol%) is smooth in morphology (Fig. 1a), while a rough surface with granular structure is observed at the higher H2 concentration (100 mol%) (Fig. 1b). This result suggests that two-dimensional growth prefers at low H2 concentrations. Fig. 2 shows the XRD patterns of the deposited layers as a function of H2 concentration. The major diffraction peak located at about 34.5 is the hexagonal (0 0 0 2) plane for GaN. As the H2 concentration increases, another diffraction peak located at 36 which is assigned to the (1 0 1) plane for GaN appears gradually. In addition, the Raman spectra showing the scattering peaks for
Fig. 1. SEM images of GaN layers deposited on GaN-grown on sapphire substrates at H2 concentrations of (a) 20 and (b) 100 mol%. Deposition conditions: partial pressure of TMG=2 103 Torr, V/III=20, substrate temperature= 958 K, total pressure=5 Torr, total carrier gas flow rate=1000 sccm.
ARTICLE IN PRESS 350
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
Fig. 2. XRD patterns of the GaN layers deposited at H2 concentrations of (a) 20, (b) 50, (c) 80, and (d) 100 mol%.
Fig. 4. FT-IR spectra of the GaN layers deposited at H2 concentrations of (a) 20 and (b) 100 mol%.
Table 1 Composition of the deposited layers grown at various H2 ambient concentrations Sample no.
H2 ambient concentration (%)
Ga (%)
N (%)
C (%)
1 2 3 4
20 50 80 100
48.7 51.1 49.6 45.4
48.5 46.0 46.4 48.2
2.8 3.0 4.0 6.4
Note: Data are determined by ESCA narrow scan analyses in molar percentage.
Fig. 3. Raman spectra of the GaN layers deposited at H2 concentrations of (a) 20, (b) 50, and (c) 100 mol%.
the deposited layers at various H2 concentrations are shown in Fig. 3. The result is characterized by two important features. First, the appearance of the characteristic modes at 567 and 735 cm1 indicates the epitaxial growth of GaN [20]. Second, the intensity of the scattering peaks decreases as the H2 concentration is raised, indicating a comparatively poor epitaxial property for GaN layers deposited at high H2 concentrations. Fig. 4 shows two examples of FT-IR spectra for the samples deposited at various H2 concentra-
tions. Comparing the vibration mode around 1380–1240 cm1 which belongs to the C–N bond stretching [21], the adsorption intensity of the C–N bond stretching for the layer deposited at the higher H2 concentration (Fig. 4b) is stronger than that deposited at the lower H2 concentration (Fig. 4a), which indicates that C–N bond stretching is more pronounced for the samples deposited at high deposition H2 concentrations. This result is quite different from the case when DMHy is used as the nitrogen source [14]. To ascertain the novel behavior of H2 in the TMG/TBHy reaction system, ESCA narrow scan for C 1 s signal is performed to quantitatively analyze the total carbon contamination content in the layers deposited at various H2 concentrations. Table 1 tabulates the compositions of all constituent
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
atoms in the layers deposited at various H2 concentrations. Though the atomic ratio of Ga to N for all the samples are near stoichiometric in spite of different H2 concentrations, the total carbon content which is the evidence of carbon incorporation increases with increasing H2 concentration. From Table 1, it clearly shows that the carbon incorporation amount increases from 2.8 to 6.4 mol% when the H2 concentration is raised from 20 to 100 mol%. Since the carbon content obtained from ESCA measurement shows the same trend with the FTIR measurement result shown in Fig. 4, i.e., C–N bond stretching in the deposited layers increases with H2 concentration, the carbon contamination is mainly originated from a C–N containing species which is most plausibly a derivative from TBHy. The concentration of this C–N containing species is proportional to H2 ambient concentration. In other words, introducing N2 instead of H2 to this reaction system could relieve the appearance of the C–N containing species. This C–N containing species results in not only carbon incorporation but also poor epitaxial property and film morphology. The detailed chemistry remains unclear. A possible reaction mechanism based on a quantum chemistry calculation is proposed as follows. First of all, an adduct composed of TBHy and TMG molecules which is optimized by the B3LYP/6-31G is used as a starting material. That is, the adduct is simulated as a simplified model for a TBHy molecule adsorbed on a Gaterminated GaN surface. Then, the adduct, denoted as TMGTBHy, is used to execute a series of theoretical predictions of some reaction potential energy surfaces. A potential energy surface for C–N bond breaking is performed to compare the experimental results. Fig. 5a shows the result of the potential energy diagram for TMGTBHy. Both the activation energy and the heat of reaction (in kcal/mol) including the ZPE correction and higher level correlation calculations are also shown. A transition state structure with an energy barrier of 72.73 kcal/mol is found since it is characterized by one imaginary frequency of 907.5837 cm1. From this imaginary frequency vibration, two strong displacement vectors origi-
351
Fig. 5. (a) Potential energy diagram for the TMG/TBHy reaction system. Both the activation energy and the heat of reaction calculated in kcal/mol are carried out using the B3LYP/6-31G level and the values in parentheses are achieved using a larger basis set (aug-cc-pVTZ). (b) Potential energy diagram for the TMG/DMHy reaction system. Both the activation energy and the heat of reaction calculated in kcal/ mol are carried out using the B3LYP/6-31G level and the values in parentheses are achieved using a larger basis set (augcc-pVTZ).
nated from two hydrogen atoms: one from t-butyl ligand and the other from hydrazine ligand (NHNH2) are observed. The final predicted products include a hydrazine ligand (NHNH) which is a C–N free species adsorbed on a Gaterminated GaN surface, a gaseous isobutene, and
ARTICLE IN PRESS 352
Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
a gaseous H2 molecule as shown as ðCH3 Þ3 Ga : NH2 NHCðCH3 Þ3 -ðCH3 Þ3 Ga : NHNH þ C4 H8 þH2 :
ð1Þ
The predicted result strongly suggests that bhydride elimination reaction does occur in the tbutyl ligand and a H2 molecule is formed eventually. According to this calculation result, the presence of excess H2 will tend to inhibit the forward reaction path of TMGTBHy based on the Principle of LeChatelier. By contrast, low H2 concentrations favor the reaction path to form the C–N free species. This explains well the experimental fact why the carbon contamination amount is lower for the GaN layers grown at low H2 concentrations. Furthermore, the C–N containing species, possibly TBHy itself, due to its complicated molecular structure may show steric hindrance effect during film formation process. This may be the reason for the three-dimensional film growth and the poor crystalline property at high H2 concentrations. An analogous adduct composed of DMHy and TMG denoted as TMGDMHy is also used for the same potential energy surface for C–N bond breaking. Fig. 5b shows the result of the potential energy diagram for the TMGDMHy. The overall reaction equation is given by ðCH3 Þ3 Ga : NH2 NðCH3 Þ2 -ðCH3 Þ3 Ga : NHNCH3 þCH4 :
ð2Þ
Comparing the calculation results in Fig. 5, two important differences are concluded as follows. First, the fragment of the final products adsorbed on the Ga-terminated GaN surface for TBHy is a carbon-free nitrogen species (NHNH), while a methyl-containing nitrogen species (NHNCH3) for DMHy is predicted. This difference may explain why the GaN layers grown using TBHy have much less carbon contamination than those formed using DMHy [15]. Also, the ambiguity of the role of ambient H2 played in each reaction system can be clarified. That is, compared with the retarding effect of H2 for TBHy to form a carbonfree species, adding excess H2 into the TMG/ DMHy reaction system may decrease carbon incorporation through a further reaction of H2 and NHNCH3 which takes away the carbon
contaminant (methyl ligand). Second, the larger endothermic heat of reaction of TMGTBHy (39.39 kcal/mol) over TMGDMHy (4.36 kcal/ mol) indicates that higher reaction temperature may be required for the GaN film growth using TBHy. This result is coincident with the necessary synthesis temperature range reported for the TMG/TBHy reaction system (958–973 K) versus that for the TMG/DMHy reaction system (853– 903 K) [14,15].
4. Conclusion Effects of hydrogen on the metalorganic vaporphase epitaxy of GaN using the TMG/TBHy reaction system are investigated. Deposition experimental results show that adding excess H2 into the reaction system not only increases carbon incorporation but also deteriorates the crystalline property and surface morphology of the GaN layers. The role of H2 during reaction is explored by an approach of computational chemistry prediction to understand the possible carbon incorporation path and surface reaction mechanism. The predicted result shows that the b-hydride elimination reaction occurs in the t-butyl ligand of the adsorbed TBHy to produce a carbon-free nitrogen species (NHNH), a gaseous isobutene and a H2 molecule. As a comparison, a similar simulation when DMHy is used as the nitrogen source is also performed. The result shows that a methyl-containing nitrogen species (NHNCH3) dominates the carbon incorporation behavior. The simulation results explain quite well the completely different role of H2 played in these two reaction systems. An advanced quantum chemical calculation technique that considers much well defined Ga-terminated GaN surface to simulate all possible reaction pathways for these reaction systems is now underway.
Acknowledgements The authors thank National Science Council of Taiwan for financial support under Contract NSC 91-2214-E-011-019.
ARTICLE IN PRESS Y.J. Hsu et al. / Journal of Crystal Growth 266 (2004) 347–353
References [1] Y. Ohba, A. Hatano, J. Crystal Growth 145 (1994) 214. [2] C.J. Sun, M. Razeghi, Appl. Phys. Lett. 63 (1993) 973. [3] G.P. Yablonskii, A.L. Gurskii, E.V. Lutsenko, I.P. Marko, B. Schineller, A. Guttzeit, O. Schon, M. Heuken, K. Heime, R. Beccard, D. Schmitz, H.H. Juergensen, J. Electron. Mater. 27 (1998) 222. [4] S. Nakamura, M. Senoh, T. Mukai, Jpn. J. Appl. Phys. 30 (1991) L1708. [5] J. Emsley, The Elements, Oxford University Press, New York, 1991. [6] D.A. Neumayer, J.G. Ekerdt, Chem. Mater. 8 (1996) 9. [7] D.K. Gaskill, N. Bottka, M.C. Lin, J. Crystal Growth 77 (1986) 418. [8] R.T. Lee, G.B. Stringfellow, J. Crystal Growth 204 (1999) 247. [9] H. Saito, T. Makimoto, N. Kobayashi, Jpn. J. Appl. Phys. 35 (1996) L1644. [10] H. Sato, H. Takahashi, A. Watanabe, H. Ota, Appl. Phys. Lett. 68 (1996) 3617. [11] S. Nishide, T. Yoshimura, Y. Takamatsu, A. Ichige, K. Pak, N. Ohshima, H. Yonezu, J. Crystal Growth 189/ 190 (1998) 325. + [12] U.W. Pohl, C. Moller, K. Knorr, W. Richter, J. Gottfriedsen, H. Schumann, K. Rademann, A. Fielicke, Mater. Sci. Eng. B59 (1999) 20. + [13] U.W. Pohl, K. Knorr, C. Moller, U. Gernert, W. Richter, J. Bl.asing, J. Christen, J. Gottfriedsen, H. Schumann, Jpn. J. Appl. Phys. 38 (1999) L105.
353
[14] Y.J. Hsu, L.S. Hong, K.F. Huang, J.E. Tsay, Thin Solid Films 419 (2002) 33. [15] Y.J. Hsu, L.S. Hong, J.E. Tsay, J. Crystal Growth 252 (2003) 144. [16] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, V.G. Zakrzewski, J.A. Montgomery Jr., R.E. Stratmann, J.C. Burant, S. Dapprich, J.M. Millam, A.D. Daniels, K.N. Kudin, M.C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G.A. Petersson, P.Y. Ayala, Q. Cui, K. Morokuma, P. Salvador, J.J. Dannenberg, D.K. Malick, A.D. Rabuck, K. Raghavachari, J.B. Foresman, J. Cioslowski, J.V. Ortiz, A.G. Baboul, B.B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R.L. Martin, D.J. Fox, T. Keith, M.A. Al-Laham, C.Y. Peng, A. Nanayakkara, M. Challacombe, P.M.W. Gill, B. Johnson, W. Chen, M.W. Wong, J.L. Andres, C. Gonzalez, M. Head-Gordon, E.S. Replogle, J.A. Pople, Gaussian 98, Revision A.11.1, Gaussian, Inc., Pittsburgh, PA, 2001. [17] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [18] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785. [19] D.E. Woon, T.H. Dunning Jr., J. Chem. Phys. 98 (1993) 1358. [20] V.Y. Davydov, R.E. Dunin-Borkovski, V.G. Golubev, J.L. Hutchison, N.F. Kartenko, D.A. Kurdyukov, A.B. Pevtsov, N.V. Sharenkova, J. Sloan, L.M. Sorokin, Semicond. Sci. Technol. 16 (2001) L5. [21] Q. Fu, C.B. Cao, H.S. Zhu, Mater. Lett. 42 (2000) 166.