Thermodynamic and experimental investigations on the growth of thick aluminum nitride layers by high temperature CVD

Abstract To achieve AlN bulk growth, high temperature CVD process using chlorine chemistry was investigated. High growth rate and high crystalline quality are targeted for both polycrystalline and epitaxial AlN films grown on (0 0 0 1) α - Al 2 O 3 Sapphire and (0 0 0 1) off axis 4H SiC or on axis 6H SiC single crystal substrates. Thermodynamic calculations were carried out to select the more appropriate inert materials for the reactor and to understand the chemistries of Al chlorination and AlN deposition steps. The reactants were ammonia ( NH 3 ) and aluminum chlorides ( AlCl x ) species formed in situ using chlorine gas ( Cl 2 ) reaction with high purity Al wires. Deposition temperature was varying from 1100 to 1800 ∘ C . Influences of temperature, total pressure, Cl 2 flow rate and carrier gas (Ar or H 2 ) on growth rate, surface morphology and crystalline state are presented. As results, films morphology is related to a variation of the thermodynamic supersaturation. As-grown AlN layers surface morphologies were studied by SEM, FEG-SEM and AFM. Crystalline state, crystallographic orientations and epitaxial relationships with substrates were obtained from θ / 2 θ X-ray diffraction and X-ray pole figure, respectively. Growth rates up to 200 μ m h − 1 have been reached for polycrystalline AlN layers.


Introduction
Aluminum nitride AlN is a wide bandgap III-V semiconductor material. AlN single crystal is expected to be a promising substrate material because of its wide direct bandgap (Eg ¼ 6:2 eV), its high electrical resistivity (between 10 9 and 10 13 O cm at 300 K) and high thermal conductivity (3:3WK À1 cm À1 ), its small difference in thermal expansion coefficient and small lattice mismatch with SiC and GaN. The availability of AlN single crystal substrates is expected for applications such as group III nitride optoelectronics blue and UV LEDs and LDs, high frequency and high power devices (high electron mobility transistors: HEMT) or surface acoustic waves (SAW) emitters and detectors.

Thermodynamic analysis
Complex equilibria thermodynamic analysis of AlN HTCVD process was carried out using a procedure based on minimization of the whole Gibbs energy [38] with Factsage TMr using SGTE [39] and FACT [40] thermodynamic databases. Let us note that thermodynamic data for the condensed phase hAlNi are extrapolated when T4930 C [41].
Our AlN HTCVD process consists in two steps. The first step is the synthesis of aluminum chlorides via the in situ reaction between Cl 2 and Al (chlorination reaction). The second step is the deposition step which correspond to the chemical reaction between AlCl x species and NH 3 with either H 2 or Ar as carrier gas. Fig. 1 shows the effect of temperature at P ¼ 10 À2 atm (a) and pressure at T ¼ 650 C (b) on the formation of different AlCl x species by reaction between Al and Cl 2 . It appears that equilibrium partial pressure of AlCl 3 is larger than that of AlCl below 790 C and that AlCl and AlCl 2 only exist at high temperature above 500 C( Fig. 1(a)). Concerning the effect of total pressure ( Fig. 1(b)), the increasing of AlCl 3 and AlCl 2 partial pressures is similar while the Al 2 Cl 6 =AlCl ratio exhibits a strong increase with a crossover at about 0.2 atm. This thermodynamic study of Al chlorination by Cl 2 is in agreement with previous thermodynamic calculations of reaction between Al and HCl [17,18].
Such thermodynamic calculations indicate that AlCl x species formed by chlorination remain stable between 500 and 1500 1C. Above 1500 1C, AlCl 3 partial pressure decreases rapidly while AlCl (and AlCl 2 ) partial pressure increases. Fig. 2 shows the thermodynamic study of homogeneous reactions in the gas phase (starting from AlCl x synthesis at T ¼ 650 C, NH 3 ,H 2 ,A r and without formation of any condensed phase). This indicates that AlCl plays an important role at high temperature above 110 0 1C. Above 1200 1C, AlCl 3 partial pressure decreases while those of AlCl and AlCl 2 increase. AlCl partial pressure becomes larger than that of AlCl 3 above 1450 1C. Al 2 Cl 6 and AlCl 2 H partial pressures decrease rapidly whereas those of AlCl 2 and Al (gas) increase strongly at high temperature. But these other Alreactants are negligible (less than 1%). So, AlCl 3 and AlCl seem to be the most important Al sources in HTCVD process at low temperature (above 500 1C) and high temperature (above 1500 1C), respectively. H 2 remains stable but decreases slowly at high temperature to the advantage of H, HCl and also Cl which increase strongly with increasing temperature. H and Cl appear only above 1100 1C. NH 3 partial pressure is very low and does not appear in Fig. 2. Thermodynamically, NH 3 is rapidly dissociated in N 2 and H 2 when the temperature increases (P NH3 % 10 À9 atm at 300 1C). In these calculations, NH 3 thermal dissociation is not inhibited so N 2 seems to be the only one N-reactant of this process which is not in good agreement with N 2 actual high thermal stability. But it is well known that NH 3 dissociation is catalyzed by metals and hindered by ceramicslike quartz (SiO 2 ) [42]. Furthermore, in previous thermodynamic studies, NH 3 dissociation was sometimes not considered because of this kinetic aspect [42] or of the formation of an adduct compound in gas phase such as AlCl 3 .NH 3 [12,[41][42][43] or Cl 2 AlNH 2 [44,45].T o  conclude on this point, kinetically, NH 3 (and perhaps NH 2 or NH at high temperature) seems to be the most important N source of AlN HTCVD.
The contribution of AlCl and AlCl 3 on AlN deposition step between 600 and 2000 1C was also studied in N 2 or NH 3 atmosphere starting from the four following equations: Thermodynamic AlN deposition yield Z was calculated with the following formula: Using N 2 as N source, AlN synthesis cannot be realized with only AlCl 3 (Eq. (2)) but is thermodynamically possible with AlCl (Eq. (1)). In the second case, both AlCl and AlCl 3 can react with NH 3 to form AlN (Eqs. (3) and (4)) which shows the great importance of H in this CVD process. Results of our thermodynamic calculations indicate that in both cases the use of a AlCl-AlCl 3 mixture improves AlN deposition yield than AlCl 3 alone. AlN deposition from AlCl 3 and NH 3 decreases near 1200 1C and seems to be impossible above 1300 1C without AlCl. With AlCl species, AlN deposition is thermodynamically possible until 1600 1C. The decrease of AlN deposition at high temperature is due to the formation of the AlCl þ 2HCl þ 1 2 N 2 þ 1 2 H 2 gas mixture, which is the most stable combination for gaseous species at high temperature [44,46]. However, in our experiments, AlN deposition was realized until 1800 1C which contradicts the thermodynamic impossibility on AlN growth above 1600 1C. An explanation could be that experimental thermodynamic data on AlN above 930 1C [41] and other compounds such as AlCl 3 .NH 3 [12,42,43] are not available. Furthermore, the existence of a Cl 2 AlNH 2 species which is also not included in these calculations is supposed by ab-initio calculations above 600-700 1C [44] and by analogy with Cl 2 BNH 2 [45,[47][48][49] (coming from direct reaction in gas phase between BCl 3 and NH 3 ) which was measured by HT mass spectrometry measurements in CVD conditions and seems to be a key species in BN deposition process [50,51]. Fig. 3(a) shows AlN deposition yield as a function of AlCl 3 and NH 3 partial pressures at 1100 1C and 10 À2 atm. The maximum AlN deposition yield is found for a low AlCl 3 partial pressure and a high NH 3 partial pressure. Fig. 3(b) shows AlN deposition yield as a function of AlCl 3 and H 2 partial pressures at 1100 1C and 10 À2 atm. AlN yield increases with increasing H 2 partial pressure at constant AlCl 3 and NH 3 partial pressures. Thermodynamically, the use of H 2 carrier gas promotes the formation of AlN compared to an inert gas such as Ar.
The influence of a variation to thermodynamic equilibrium defined as gas phase supersaturation was also studied. Experimentally, a variation of supersaturation in gas phase is known to have a great effect on surface morphology and crystallographic orientation of deposits. Usually, the epitaxial growth is favored with a low supersaturation [52,53]. In the case of AlN a low supersaturation is obtained at high temperature and for partial pressures of reactants near the thermodynamic equilibrium. In the present investigations, supersaturation a was calculated with the following formula: a ¼ P Al species Â P N species P eq Al species Â P eq N species (6) where P is the partial pressure of reactant species and P eq is the calculated thermodynamic equilibrium partial pressure of Al and N species over solid AlN at given temperature and total pressure. Fig. 4 shows calculated gas phase supersaturation a between 1000 and 2000 1Ca t1 0 À2 atm with different H 2 =AlCl 3 ratio at a constant NH 3 partial pressure. Supersaturation decreases rapidly with increasing temperature and slowly with increasing H 2 =AlCl 3 ratio.
Our thermodynamic calculations indicate that AlCl is a necessary precursor to grow AlN at high temperature with high growth rate, it is also known that AlCl is very reactive with quartz (SiO 2 ) reactor and this reaction is accelerated in the presence of H 2 [11]. For this reason, AlCl 3 is generally favored and generated by decreasing the temperature of the chloride source at 500 C [17,18,20,21,23,24,27,28]. In our case, a water-cooled quartz tube (and a reactants dilution) prevent the reaction of AlCl with quartz. Based on these thermodynamic calculations of AlCl x synthesis and AlN deposition step, it is possible to grow AlN by HTCVD at low pressure in a water-cooled quartz reactor.

Experimental procedure
A schematic diagram of the HTCVD apparatus is shown in Fig. 5. The HTCVD set-up consists in a home-built vertical coldwall reactor composed of two reaction zones [35,54]. The first one is the chlorination area where aluminum chlorides are generated via in situ reaction at 640 1C( T melt: Al ¼ 660 C) between Cl 2 and high purity Al wire contained in an inner tube heated by a lamp furnace. Then, AlCl x species are mixed with NH 3 and either H 2 or Ar as carrier gas. The second zone whose walls are water-cooled is the AlN deposition area. The substrate is located on top of a graphite susceptor heated by induction and the deposition temperature is measured on the graphite surface using InfraRed pyrometry. The NH 3 flow is in a (20-100 sccm) range and diluted in 200-1000 sccm of Ar or H 2 . The Cl 2 flow rate is fixed between 2.5 and 100 sccm and diluted with Ar.
AlN films were deposited at low pressure on (0 0 0 1) a-Al 2 O 3 Sapphire and (0 0 0 1) off axis 4H SiC or on axis 6H SiC single crystals between 1100 and 1800 1C. Let us note that a reaction occurs between Al 2 O 3 and C above 1500 1C which induces the formation of CO, Al 2 O and Al gases in agreement with our thermodynamic calculations and a previous study of the Al-C-O ternary system [55]. Before AlN deposition, Sapphire and SiC substrates were chemically etched in a HF 5% solution for 10 min. A thermal cleaning under H 2 atmosphere was also carried out above 1000 1C for 15 min. As-grown AlN layers surface morphologies were studied by SEM, FEG-SEM and AFM. Crystalline state, crystallographic orientations and epitaxial relationships with substrates were obtained from y=2y X-ray diffraction and X-ray pole figure, respectively.

Results and discussions
In this experimental investigation, AlN films were deposited at different temperatures, in a range of 1100-1800 1C on different single crystal substrates: (0 0 0 1) Al 2 O 3 , (0 0 0 1) off axis 4H SiC and on axis 6H SiC [35]. Different conditions were tested in order to reach epitaxial growth. The influence of growth temperature, total pressure and chlorine flow rate were studied. Films morphology is also related to a variation of the thermodynamic supersaturation. Crack-free polycrystalline AlN layers were obtained with a thickness up to 150 mm and growth rates up to 200 mmh À1 . Their color varied from white to yellow and to brown with increasing Cl 2 flow rate. Yellow and brown colors of deposits seem to be related to an enrichment by aluminum in AlN interstitial sites [3,11,12]. Mirror-like AlN layers were grown up to a thickness of 30 mm but exhibited cracks. Any chlorine has not been found from X-EDS in all fabricated AlN films. Fig. 6(a) is the typical kinetic curve which shows the variation of AlN growth rate on 4H SiC substrates as a function of the inverse of temperature in the 1100-1800 1C range at P ¼ 10 Torr.
Growth rate varied from 10 to 30 mmh À1 . At low temperature, this curve represents the kinetic regime of AlN deposition. High growth rates are obtained at high temperature which correspond to the diffusion regime [52]. A slight drop of deposition rate is observed at very high temperature and probably corresponds to the decrease of thermodynamic supersaturation (Fig. 4) and/or the beginning of powder formation by homogeneous nucleation in the gas phase [52,[56][57][58]. Contrary to thermodynamic prediction, we definitely observed the formation of condensed AlN above 1600 1C. A new further assessment of thermodynamic data on the whole Al-N-H-Cl system at high temperature should be achieved as for the Si-C-H-Cl system [53,[59][60][61]. Fig. 6(b) shows the variation of AlN growth rate as a function of total pressure at 1750 1C. AlN deposition rate decreases with increasing pressure between 10 and 100 Torr as previously reported by Suzuki and Tanji [58] but for a temperature of 1000 1C. Our thermodynamic analysis shows that AlN growth rate increases as a logarithmic function with increasing total pressure. The experimental trend cannot be explained by a thermodynamic approach but probably by chemical reactions, kinetic, gas dynamics and/or reactor's design effects. Fig. 6(c) shows the variation of AlN growth rate as a function of Cl 2 flow rate at 1750 1C and 10 Torr. Growth rate increases linearly with increasing Cl 2 flow rate which is in agreement with our thermodynamic calculations. So, in these conditions, the deposition of AlN is probably limited by the formation of AlCl x species, in agreement with previous investigations based on CVD using AlCl 3 [9,46,58] and on hydride vapor phase epitaxy using the reaction between HCl and Al [20,21,24,28].
The effect of carrier gas (Ar and/or H 2 ) was also studied. The use of H 2 as carrier gas allows to obtain higher growth rate as expected from thermodynamic calculations, more dense deposits and smoothest surface morphology than Ar. But it is still unclear if it comes from a chemical or a thermal effect. Fig. 7 shows an X-ray diffraction pattern of preferred oriented (0 0 0 1) 2H AlN film deposited on (0 0 0 1) Sapphire substrate.
From y=2y scans related to diffracting vectors normal to the (0 0 0 1) surface of a-Al 2 O 3 single crystal substrate, it has been deduced that each AlN sample exhibits a simple crystallographic orientational relationship with the substrate; i.e. ð0001Þ AlN ==ð0001Þ a-Al 2 O 3 (Fig. 7(a)). Let us recall that both the structure AlN and a-Al 2 O 3 are hexagonal P6 3 mc (a ¼ 3:1114Å, c ¼ 4:9792Å) and trigonal R À 3c (a ¼ 4:7617Å, c ¼ 12:9947Åo r a ¼ 4:7617Å and a ¼ 55:3004 for rhombohedral axes), respectively. However, from systematic plots of logðIÞ¼f ð2yÞ,i t could also be deduced that part of AlN layers were polycrystalline ( Fig. 7(b)). Same conclusions has been found for films deposited on 4H and 6H SiC single crystal substrates.
As Many other reflections of very weak intensity are also identified as corresponding to these of substrate a-Al 2 O 3 . Then it is straightforward to deduce from these pole figures the following epitaxial relationships:  With increasing the temperature of HTCVD process, we have observed that the 1 2 3 3 reflections of AlN become less broad and that the proportion of powder-like AlN diminishes. The crystallinity of AlN layers in epitaxy on (0 0 0 1) off axis 4H SiC substrates was characterized using the o scan method (rocking curves, RCs) [62,63]. X-ray rocking curves were measured on both 0 0 0 2 and 1 01 2 AlN reflections in order to study out of plane (tilt) and in plane (twist) distortions, respectively. The lowest full-width at half-maximum (FWHM) values were 2100 arcsec (0:58 ) for (0 0 0 2) and 1400 arcsec (0:39 ) for (1 01 2) in the case of a 10 mm AlN layer deposited in the following conditions: T ¼ 1750 C, P ¼ 10 Torr, low reactant flow rates and high carrier gas dilution. These values are important compared to those currently obtained on (0 0 0 1) on axis 6H SiC [25] or (0 0 0 1) Sapphire [31,33] and it is interesting to note that using this HTCVD process, tilt is more important than twist contrary to HVPE process [20,25,31,33]. Fig. 9 shows SEM images of AlN layers grown on 4H SiC substrates at various deposition temperatures with constant partial pressure of reactants and a total pressure of 10 Torr. With decreasing the supersaturation by increasing the deposition temperature, the surface morphologies obtained are sphere-like and become facetted and then smooth. So, rougher deposits are obtained at low temperature and smooth films at high deposition temperature. These smooth AlN layers obtained at high temperature were related to the AlN epitaxial growth from the previous y=2y X-ray diffraction and X-ray pole figure experiments.
However, several cracks were observed on such smooth AlN layers probably because of the difference of thermal expansion  coefficient between AlN and the substrate at high temperature, lattice mismatch substrate-deposit and temperature ununiformity on the substrate surface. Fig. 10 shows a SEM and AFM images of an AlN layer deposited at 1750 1C and 10 Torr on 6H SiC (on axis). This AlN layer exhibits hexagonal based hillocks ( Fig. 10(a)). Those observed by AFM consist in polygonized spiral which very probably come from a preferential epitaxial growth around screw dislocations ( Fig. 10(b) and (c)). Defect formation analysis, as done for SiC growth [64,65], should be very useful to understand the growth mechanisms. This growth mode is related to low supersaturation conditions and epitaxial growth. RMS roughness between 20 and 35 nm was obtained for such AlN layers.

Conclusions
Thermodynamic calculations on AlCl x synthesis and AlN deposition step indicate that it is possible to grow AlN by HTCVD at low pressure in a water-cooled quartz reactor. AlCl 3 and AlCl seem to be the most important Al-reactant in HTCVD process at low temperature and high temperature, respectively. However, several compounds formations by homogeneous reactions in gas phase and associated chemical mechanisms are still unclear. Experimentally, the growth of thick AlN layers on Sapphire and SiC substrates was achieved between 1100 and 1800 1C by HTCVD. Experiments prove that the use of H 2 as carrier gas instead of Ar allows to obtain higher growth rates, more dense deposits and smoothest surface morphologies. AlN epitaxial layers were grown when the deposition process was performed with low supersaturation conditions, i.e. high temperature and with partial pressures of reactants near the thermodynamic equilibrium. This epitaxial growth of AlN was also related to smooth surface morphology and exhibits hexagonal based hillocks related to a spiral growth around screw dislocations on axis 6H SiC. With increasing Cl 2 flow rate, AlN growth rate increases linearly and the epitaxial layer becomes polycrystalline with increasing deposition rate. The maximum growth rate of polycrystalline layers obtained at high temperature was about 200 mmh À1 . To conclude, HTCVD seems to be very promising for the development of a further bulk growth process to produce AlN single crystal substrates.