单晶硅提拉法生长工艺，熔体碳杂质的下降

Czochralski (CZ) crystal growth of single crystallinesilicon (Si) is invariably accompanied by transport of impurities such ascarbon (C), oxygen (O), and related products from chemical reactions in thehigh temperature range. C contamination in the growing crystal is detrimentalfor the minority carrier lifetime, which is one of the critical parameters ofwafer for power device [1]. Reduction of C contamination in the CZ-Si crystalis required for the production of high performance Si wafer.

Contamination of C in Si crystal mainly originates fromcarbon monoxide (CO) generation on the graphite components. And, incorporationof C occurs prior to the growth stage [2]. The reason is the CO generationstarts from the preheating stage, and reaches the maximum in the melting stageof the CZ-Si crystal growth. The generated CO incorporated into the Sifeedstock and accumulated in the melt. Thus, C prediction with quasi-staticassumption is not sufficient to account for C accumulation in CZ-Si crystalgrowth [3]. Therefore, transient global simulations of heat and mass transportwere performed for the melting process of CZ–Si crystal growth. A virtual PIDcontroller for the temperature was introduced to realize the power control ofthe heater. Accumulation of C in Si feedstock was predicted for the entiremelting process, which consists of preheating, melting and stabilizationstages.

To reduce the C contamination effectively, it isessential to control the C transport from its generation, incorporation andaccumulation in the growth. Besides the CO generation by the reaction betweengraphite and SiO, the reaction between quartz crucible and graphite susceptorwas also taken into account as another CO source [4]. Since softening cruciblecontacts with susceptor, this reaction is actively involved at the hightemperature. The contact reaction process is shown schematically in Fig. 1. Thegenerated SiO and CO could be transported from the narrow gap to the gasdomain. Back diffusion of CO may affect C contamination in Si feedstock.

To suppress the CO generation, silicon carbide(SiC)
coating could be applied conventionally to the gas-guide above the meltsurface. Three cases, including no coating, SiC coating and no reaction, werecompared for CO concentrations above the melt surface, as shown in Fig. 2(a).Without the consideration of coating reaction, hardly any CO could beincorporated into the Si melt as C contamination. This is inconsistent with thepractical process. Even the reactivity of SiC is lower than graphite, reactionbetween SiC and gaseous SiO must be taken into account due to the shortdistance between gas-guide and Si feedstock. Actually, SiC coating with low COgeneration could only reduce the C incorporation partially in real growth, asshown in Fig. 2(b). Therefore, these two extra CO sources were introduced inthe present study, and implemented as boundary conditions of the transientglobal simulation of melting process of CZ-Si growth.

According to the physical properties of
argon (Ar) gas,the Péclet number of
CO transport in the typical gas flow of
CZ-Si growth isvery high, around
2000 [5]. This indicates the
enhancement of the Ar flow abovethe
melt could reduce the incorporation
rate of C efficiently. Parameterstudies
of furnace pressure and gas flow rate
were conducted on theaccumulation of
C during the melting process.
Convective transport of SiO andCO
dominated at the higher flow rate. Thus,
generation and back diffusion ofCO
were both suppressed by the intense gas
flow. Because of the combination ofthese two positive effects, the increase in flow rate could reduce the Ccontamination by logarithm, as shown in Fig. 3. According to the accumulationof C, the final C content depends on the growth duration and contamination fluxat the gas/melt interface.

References


[1] Y. Nagai, S. Nakagawa and K. Kashima, J. Cryst. Growth, 401, 737 (2014).


[2] R. W. Series and K. G. Barraclough, J. Cryst. Growth, 63, 219 (1983).


[3] X. Liu, B. Gao and K. Kakimoto, J. Cryst. Growth, 417, 58 (2015).


[4] B. Gao, S. Nakano and K. Kakimoto, J. Cryst. Growth, 314, 239 (2011).

[5] D. E. Bornside, R. A. Brown, T. Fujiwara, H. Fujiwaraand T. Kubo, J. Electrochem. Soc., 142, 2790 (1995).