Superlattices and Microstructures 40 (2006) 496–500www.elsevier.com/locate/superlattices

Thermodynamic analysis of Si doping in GaN

I. Halidou∗, Z. Benzarti, T. Boufaden, B. El Jani

Unite de Recherche sur les HeteroEpitaxies et Applications (URHEA), Faculte des Sciences 5000 Monastir, Tunisia

Received 20 September 2006; accepted 25 September 2006Available online 13 November 2006

Abstract

Equilibrium calculations of Si-doping in GaN are investigated using the Gemini code. The method of thecalculation is based on the minimisation of the Gibbs free energy. Experimental growth conditions are usedfor the calculation. The variables are the amount of the dopant and the temperature. The results show theformation of a solid Si3N4 compound with a certain quantity of the input SiH4, that is the silicon precursorin our MOVPE system. Si3N4 formation can explain the limitation of Si incorporation and the surfaceroughening as revealed by MOVPE Si doped layers.c© 2006 Elsevier Ltd. All rights reserved.

Keywords: Thermodynamic calculations; MOVPE; GaN; Si doping

1. Introduction

Fabrication of nitride devices requires performing a controlled and reproducible n- and p-typedoping. Many works have been devoted to studying n-type doping of GaN using silicon (Si),germanium (Ge) and selenium (Se) as dopants [1,2]. Satisfactory electrical properties of Sidoped GaN have been reported but still suffer from compensation effects and surface degradation.However, for good ohmic contacts, high electron concentrations as well as smooth surfaces areneeded. Therefore, understanding the growth mechanism of Si doping is still of interest. Mostof the studies were focused on the effect of the growth parameter (growth temperature, dopantinput concentration, growth rate, III/V ratio. . . ) and thus were done empirically. Equilibriumcalculations can provide useful information necessary to understand the growth mechanism.Thermodynamic analysis of InGaAlN alloy growth was reported [3]. However, little has beendone about Si-doping in GaN growth.

∗ Corresponding author. Tel.: +216 73 500 274; fax: +216 73 500 278.E-mail address: ibrahim.halidou@fsm.rnu.tn (I. Halidou).

0749-6036/$ - see front matter c© 2006 Elsevier Ltd. All rights reserved.doi:10.1016/j.spmi.2006.09.023

I. Halidou et al. / Superlattices and Microstructures 40 (2006) 496–500 497

Fig. 1. Temperature dependence of equilibrium pressures of species arising from the reaction of the system:40 µmol TMG, 0.18 mol NH3, 0.09 mol N2, and 0.09 mol H2. The total pressure is 1 atm. Only the most importantspecies are represented for clarity. It should be noted that solid GaN compound disappears above 1200 K.

In this paper, we perform equilibrium calculations of MOVPE Si-doping in GaN growth usingSiH4 as Si precursor. The results provide guidelines for understanding the growth mechanismsuch as the Si incorporation limit and surface roughening.

2. Calculation procedure

For the thermodynamic calculations, we have used the GEMINI code provided byThermodata [4]. We have already successfully used this computer programme to study carbonand vanadium doping of GaAs, etching of GaAs and AlAs in CCl4 + H2 ambient and GaNdecomposition under H2 ambient [5–7]. The results showed good agreement with experimenteven though MOVPE is a non-equilibrium process. But, a near equilibrium can be reached at thevapour–solid interface. The principle of the calculations is based on the standard minimisationof the Gibbs free energy of a system of compounds. This allows the determination of the systemcomposition when equilibrium is reached.

For MOVPE Si-doping in GaN layer growth, the reactants are trimethylgallium (TMG),ammonia (NH3), the carrier gas (N2 + H2) and SiH4. The products are all possible compoundsthat may be formed from a mixture of these reactants for a given temperature. For eachcompound, the thermodynamic data such as standard enthalpy of formation (1H◦

f (298 K)),standard entropy formation (1S◦ (298 K)), and the heat capacity as a function of the temperature(Cp(T )) are required. Some of the data are provided by Thermodata and the others are takenfrom Refs. [4,8].

3. Results and discussion

We start from a vapour composition used in our MOVPE growth experiment for undoped GaNlayer deposition. The mixture contains: 40 µmol TMG, 0.18 mol NH3, 0.18 mol H2 and/or N2.For the calculation, the total pressure is 1 atm. In Fig. 1, the species issuing from this mixtureas a function of temperature are shown. It should be noted that the mole concentration of solid

498 I. Halidou et al. / Superlattices and Microstructures 40 (2006) 496–500

Fig. 2. Partial pressure of species containing Si as a function of input mole amount of SiH4. Saturation occurs above 50nmol SiH4 due to the formation of solid Si3N4 indicating thermodynamic limit of Si incorporation.

compounds is expressed in partial pressure units (atm) used for gas species. In this manner,we can compare the amount of solid and gas species in equilibrium at a given temperature.The amount of solid species represents the mole numbers condensed to reach the vapour–solidequilibrium per unit volume. It appears from Fig. 1 that the partial pressure of H2 and N2 isconstant for the whole temperature range (900–1500 K). The dominant species for gallium andcarbon are GaCH3 and CH4. NH3 has a high equilibrium partial pressure in opposition to NHxspecies (NH2 and NH). From this calculation, it also appears that solid GaN cannot be depositedabove 1200 K and that the GaN gas specie is stable above this temperature.

For the Si doping calculation, we add SiH4 with different amounts in the above mixture. Theinput SiH4 amount lies in the range used in experiment. For clarity, Fig. 2 displays only the partialpressure of the most important compounds with the dopant element (Si) as a function of the inputSiH4 mole fraction. The temperature is fixed at 1400 K, a typical value for GaN MOVPE growthand the total pressure is 1 atm. It should be noted that at this temperature, solid GaN formationis excluded as can be seen in Fig. 1. Therefore, the amount of SixNy species represented in Fig. 2is exaggerated. Indeed, the absence of solid GaN compound enhances the partial pressure ofGa and N containing species. Keizer et al. [9] proceeded in the same manner for equilibriumcalculations of Si-doping in GaAs by MOVPE. However, their results showed a good agreementwith the experiment. As can be seen from Fig. 2, SiNH appears to be the dominant specie. Thepartial pressure of all the species depends linearly on the amount of input SiH4 until a criticalvalue. Indeed, at an input SiH4 of 50 nmol, the partial pressure of the species becomes constantdue to the formation of solid Si3N4 compound. This results in an upper limit for Si incorporation.Another feature that can be observed from Fig. 2 is the absence of gas species such as SiN,Si2N. . . Sassi et al. [10] reported that the formation of these species requires the presence of NH2and NH radicals. Since the dissociation of NH3 in the gas phase produces a small amount of theseradicals, the probability of the formation of SiN, Si2N is very low. These authors also showedthat the formation of solid Si3N4 compound is the result of the reaction between NH3 and Si. Wehave already seen that NH3 has a high equilibrium partial value (Fig. 1). We therefore attributethe formation of SiNH and solid Si3N4 to a reaction between NH3 and SiH4. The SiHx species

I. Halidou et al. / Superlattices and Microstructures 40 (2006) 496–500 499

Fig. 3. Real time-dependent He–Ne laser reflectometry signal during Si-doping in GaN. The amplitude of the signal isconstant during the deposition of the undoped layer. The sudden drop of the signal when the SiH4 flow is turned on isattributed to the formation of solid Si3N4 as predicted by the thermodynamic calculation.

is the result of the dissociation of SiH4. Finally, it is also seen that from our calculation, no solidSi is formed. This often appears in the Si-doping of GaAs [9].

4. Comparison with the experiment

We have already reported on Si-doping of GaN by MOVPE growth using SiH4 as Siprecursor [11,12]. Si was shown to affect significantly the GaN layer morphology. Indeed, upondoping we have observed a decrease of the amplitude of the in situ He–Ne laser reflectometrysignal and the oscillations disappear above a critical value of SiH4 flow rate. AFM images ofthe layers reveal an increase of the surface roughening. Thus the damping of the reflectivitysignal is due to surface roughening against SiH4 flow. Our studies also showed a saturationtendency of electron concentration for SiH4 flow of about 45 nmol/min. Similar results werereported in Refs. [13–15]. Liu et al. [15] have reported an electron concentration saturation valueof 1.7×1020 cm−3 for SiH4 flow of about 700 nmol/min. The surface roughening upon doping isattributed to the anti-surfactant effect of Si [16,17]. Considering the thermodynamic calculationresults, it can be seen that the surface roughening as well as the electron concentration saturationappears for SiH4 amounts in the range where solid Si3N4 is formed. Therefore, the growthchanges upon the Si doping can be attributed to the formation of solid Si3N4.

To clarify this point of view, we deposited a Si doped GaN layer on a 0.3 µm thick undopedlayer. The amount of TMG, NH3 and N2 are the same as that used for the equilibrium calculation.The SiH4 flow was 100 nmol/min, a value which lies in the saturation zone predicted both byexperimental and theoretical results. Fig. 3 shows the in situ reflectivity signal record versus thegrowth time. During the deposition of the undoped GaN layer, the amplitude of the oscillationsis constant indicating a 2D smooth surface. When the SiH4 flow was turned on into the MOVPEgrowth chamber for the growth of the Si-doping GaN layer, the signal drops suddenly and reachesa value below that of the bare sapphire substrate indicating a 3D growth mode (maximumroughness). The formation of an in situ SiN mask by introducing simultaneously NH3 andSiH4 into the MOVPE growth chamber was reported by several groups in the so called “Si/N

500 I. Halidou et al. / Superlattices and Microstructures 40 (2006) 496–500

treatment growth method” [18–22]. GaN can not grow on a SiN mask resulting in a 3D growthmode. We thus think that the damping of the reflectivity can be related to a formation ofsolid Si3N4 as in MOVPE Si/N treatment growth. Neugebauer [23] reported that under N-rich growth conditions (MOVPE growth), Si segregates to the surface where it can form Si3N4islands whereas under Ga-rich conditions (MBE growth) surface segregation does not occur. Thisexplains the morphology differences between the Si-doping in GaN layers grown by MOVPE andMBE.

5. Conclusion

Equilibrium calculations, based on standard minimisation of Gibbs free energy, of Si-dopingGaN by MOVPE was presented. The results show that Si incorporation is limited by theformation of solid Si3N4 compound. This compound can explain the Si anti-surfactant effecton the GaN layer during MOVPE processing. A good agreement is obtained between thecalculation and the experimental results such as electron concentration saturation and surfaceroughening upon doping. Equilibrium calculations can therefore be used to predict MOVPEgrowth mechanisms of GaN doping even though kinetics can take place as MOVPE is a non-equilibrium process.

References

[1] S. Nakamura, S. Pearton, G. Fasol, The Blue Laser Diode, Springer, Berlin, 2000.[2] G.-C. Yi, B.W. Wessels, Appl. Phys. Lett. 69 (1996) 3028.[3] A. Koukitu, Y. Kumagai, H. Seki, J. Cryst. Growth 221 (2000) 743.[4] B. Cheynet, J.D. Dubois, A. Rivet, Thermodata/INPG/CNRS, 6, rue du Tour de l’Eau, 38400 Saint Martin d’Heres

Cedex, France.[5] A. Bchetnia, A. Rebey, T. Boufaden, B. El Jani, J. Cryst. Growth 207 (1999) 15.[6] A. Rebey, L. Beji, B. El Jani, P. Gibart, J. Cryst. Growth 191 (1998) 734.[7] A. Rebey, T. Boufaden, B. El Jani, J. Cryst. Growth 203 (1999) 12.[8] JANAF Thermochemical Tables, 2nd ed. NSRDS-NBS 37, Natl. Bur. Std. US, Washington DC, 1971.[9] L.C. Keizer, X. Tang, R.J.C. Van Meerten, L.J. Giling, J. Cryst. Growth 102 (1990) 667.

[10] S. Sassi, J.C. Bureau, K. Chafik, Eur. Phys. J. AP 7 (1999) 207.[11] I. Halidou, Z. Benzarti, Z. Chine, T. Boufaden, B. El Jani, Microelectron. J. 32 (2001) 137.[12] Z. Benzarti, I. Halidou, O. Tottereau, T. Boufaden, B. El Jani, Microelectron. J. 33 (2002) 995.[13] H. Murakami, T. Asaki, H. Amano, K. Hiramatsu, N. Sawaki, I. Akasaki, J. Cryst. Growth 115 (1991) 648.[14] N. Koide, H. Kato, M. Sassa, S. Yamasaki, K. Manabe, M. Hashimoto, H. Amano, K. Hiramatsu, I. Akasaki,

J. Cryst. Growth 115 (1991) 639.[15] X. Liu, L. Wang, D.-C. Lu, D. Wang, X. Wang, L. Lin, J. Cryst. Growth 189–190 (1998) 287.[16] S. Tanaka, S. Iwai, Y. Aoyagi, Appl. Phys. Lett. 69 (1996) 4096.[17] K.S. Son, D.G. Kim, H.K. Cho, K. Lee, S. Kim, K. Park, J. Cryst. Growth 261 (2004) 50.[18] E. Frayssinet, B. Beaumont, J.P. Faurie, P. Gibart, Zs. Makkai, B. Pecz, P. Lefebre, P. Valvin, MRS Internet J.

Nitride Semicond. Res. 7 (2002) 8.[19] S. Sakai, T. Wang, Y. Morishima, N. Naoi, J. Cryst. Growth 221 (2000) 334.[20] T. Kunisato, Y. Nomura, H. Ohbo, T. Kano, N. Hayashi, M. Hata, T. Yamaguchi, M. Shono, M. Sawada, A. Ibaraki,

Phys. Status Solidi c 0 (2003) 2066.[21] S.E. Park, S.M. Lim, C.R. Lee, C.S. Kim, O. Byungsung, J. Cryst. Growth 249 (2003) 487.[22] T. Bottcher, J. Dennemarck, R. Kroger, S. Figge, D. Hommel, Phys. Status Solidi c 0 (2003) 2039.[23] J. Neugebauer, Phys. Status Solidi c 0 (2003) 1651.