Impurities in epitaxial silicon wafers

5.1.5 Impurities in epitaxial silicon wafers

The electrical quality of a silicon crystal depends on the amount of impurities and defects in the crystal lattice. In this subchapter we are therefore first discussing a possible metal contamination of our material and in the following approach the task to reduce the doping concentration in EpiWafers. In [73] it has been reported that a metal contamination occurs during processing in the RTCVD160 depending on temperature and duration. Interstitial iron concentrations above 4x10¹¹ cm^-3 were measured for wafers after 4 min annealing in the RTCVD160 at 1100°C. This value should be higher for our wafers as temperature and process times are higher. Mainly transition metals, such as Fe, Cr, Cu, Mo and Co, lead to a significant decrease in carrier lifetime because they form deep levels in the band-gap of silicon, which lead to an increased recombination [79-81]. In Figure 5.7 the SRH point defect lifetimes at an injection level of 1x10¹⁵ cm^-3 were calculated for varying p-type and n-type doping concentrations and for four different iron concentrations. The equation used for the calculations can be found in appendix C. An iron contamination of more than 5x10¹⁰ cm^-3 could explain a decreased SRH lifetime of below 200 µs for a p-type sample with a doping concentration of 8x10¹⁶ cm^-3. For an n-type sample an iron contamination is not that critical

(see Figure 5.7 right) and SRH lifetimes above several milliseconds can still be reached. If another metal impurity is assumed, i.e. chromium, the SRH lifetime for n-type silicon changes also stronger with doping concentration. This depends mainly on the capture cross section of the metal impurity, which is in case of interstitial iron much lower for holes than for electrons. A summary of capture cross sections for interstitial iron and other transition metal impurities can be for example found in [81].

10¹⁴ 10¹⁵ 10¹⁶ 10¹⁷

phosphorous doping concentration [cm^-3] Figure 5.7 SRH point defect lifetimes at a fixed injection level for different doping concentrations of boron and phosphorus. Four different iron concentrations were modelled.

Impurities can form precipitates in the crystal or they can occur at interstitial and substitutional sites, respectively. Because the unintentional iron contamination level in wafers is usually higher than that of other metal impurities (see [82]) the iron concentration is a good indicator for the overall metal contamination in a silicon wafer. In p-type wafers the interstitial iron concentration can be determined by measuring the lifetime before and after the dissociation of FeB pairs [83]. For the previous subchapter epitaxial p-type wafers have been processed with different reorganization steps prior to epitaxial growth. Those p-type samples have been used for interstitial iron imaging using photoluminescence according to [84]. All samples showed interstitial iron concentrations between 1x10¹⁰ cm^-3 and 5x10¹⁰ cm^-3 which is close to the detection limit of this method. Such iron concentrations could lead, however, for a highly doped p-type wafer to a reduced SRH lifetime in the range of 200 µs – 1000 µs (see Figure 5.7). To be sure, that no other contaminations degrade the bulk lifetime in our EpiWafers we are going to investigate the influence of a phosphorus gettering step on the effective carrier lifetime in the next subchapter.

In contrast to the results from [73] that were presented at the beginning of this paragraph, an interstitial iron concentration of 1x10¹⁰ cm^-3 in detached EpiWafers is one order of magnitude lower than expected. A possible explanation for that could be the proven gettering effect of porous silicon [85]. To confirm this, the interstitial iron concentration of the grinded reference wafers was also measured. The grinded area showed a mean interstitial iron concentration of 5x10¹⁰ cm^-3, which means that all epitaxially grown layers processed in this work show a similar amount of interstitial iron. It is not clear what the contamination source

in [73] has been and how this contamination could be reduced for the samples processed in this work. But these findings underline that the state of the reactor changes constantly and that reactor cleanliness and maintenance is of huge importance.

The high doping concentrations of 8x10¹⁶ cm^-3 limits the overall bulk lifetime in our samples not only because of SRH lifetime but also because of Auger recombination. This value was chosen at the beginning of this work because several processes with high doping concentrations, both with phosphorus and boron, changed the background in the RTCVD160 continuously. Because of this background a minimum doping concentration was mandated and comparable values are only guaranteed for doping concentrations that are above that background doping. For determination of the background doping an epitaxial layer was grown on a silicon sample without adding of any dopant gas. The incorporated doping concentration was then determined using spreading resistance profiling (SRP) measurements and resulted in a quite high boron background of 2x10¹⁶ cm^-3.

This high background doping originates from the numerous different processes performed in our lab-type RTCVD160 reactor. Especially epitaxial emitter depositions need doping concentrations above 1x10²⁰ cm^-3 [86]. This memory effect was already discussed in previous works [46, 51], but the source of the background doping was not clear. It was proposed to deposit a thin undoped layer prior to each process to avoid cross-doping from the dummy wafers and the reactor walls, respectively. This was tested but was unfortunately not sufficient to lower the background. The lowest measured background concentration of 1x10¹⁴ cm^-3 in a similar RTCVD reactor was reported by Bau [27]. This value was achieved in a quartz carrier after wet chemically cleaning and baking under hydrogen prior to processing. In the course of this work a new quartz carrier was introduced for the sole purpose of depositing lowly doped layers. The first measurement of the background showed a doping concentration of only 3x10¹³ cm^-3 which is to our knowledge lower than all previously published values for similar reactors. The background doping can therefore be clearly attributed to the quartz carrier in our lab-type reactor. The shown memory effect requires to focus on either n-type or p-type deposition in one quartz carrier to avoid compensation. Compensating dopants can lower the desired doping concentration and make a comparability of the achieved results difficult. Additionally, compensated silicon suffers from lower carrier mobilities and therefore reduced minority carrier diffusion lengths [87].

However, measurements on compensated EpiWafers will show in the following chapter that it can also lead to higher lifetimes depending on the compensation level as stated in [88].

Because the old quartz carrier showed mainly a boron background and n-type material is more tolerant to metal contamination, it was decided that the new quartz carrier will be used for n-type wafers only. Even after seven processes with a phosphine doping of 3x10¹⁶ cm^-3 the background was stable at 3x10¹³ cm^-3.

The results on p-type EpiWafers presented in this subchapter show that the bulk lifetime in epitaxially grown material could be limited by metal impurities, but mainly because of

the high doping concentration. The high background doping in the RTCVD160 reactor could be eliminated to enable EpiWafer growth with doping concentrations as low as 3x10¹³ cm^-3. Wafers with different n-type doping concentrations are therefore processed in the following subchapter and have to reveal if the SRH recombination in epitaxially grown wafers can be lowered by reducing the doping concentration.