Phosphorus Diffusion Gettering

The wordgetteringdescribes the removal of and/or lowering the recombination activity of impurities in a material during high temperature treatments. Hence, it is commonly accompanied by an improvement of the material quality indicated by a higher minority carrier lifetime. A typical temperature range of 600^◦C to 900^◦C for PDG is used in the present work. Gettering mechanisms are classified into two main terms:

• External getteringdescribes the gettering of impurities by an externally applied layer, e.g., the PSG layer. Ideally, it is possible to remove most of the gettered impurities after PDG by etching the phosphosilicate glass off. However, the fact that impurities are gettered within the glass layer is not confirmed experimentally for⁵⁷Co in [124]. As described extensively in [16], there is experimental evidence for the metal accumulation at the PSG/Si interface resulting from the formation of SiP precipitates in the highly P-doped kink region of the emitter profile. This has been emphasized by

2.3 POCl3Emitter Formation and Diffusion Gettering 37

Ourmazd and Schröter using high resolution electron microscopy (HREM) to image NiSi₂particles close to SiP precipitates [11] and by Gilles et al.’s observation of the 3d transition elements Mn, Fe and Co occupying substitutional lattice sites in the kink region [125].

• Internal getteringterms the gettering at internally present lattice sites like, e.g., dislocations and grain boundaries. It should be noted that this gettering type does not lead to a reduction of the total impurity concentration. Depending on the impurity’s configuration at the respective gettering site this gettering does not necessarily result in its lower recombination activity, i.e., higher lifetime.

Thus, the word gettering is not entirely appropriate since it is commonly associated with a ma-terial improvement. However, the most of the gettered impurities form precipitates or complexes at nucleation sites which are less detrimental to the electronic material quality than more homo-geneously dispensed interstitial impurities. A prominent example is interstitial iron Fei inducing a deep SRH level within the Si band gap. The correlation between recombination activity and concentration, distribution, stoichiometry as well as shape of precipitated impurities is studied by Donolato [126], Istratov et al. [10], Chen et al. [8], Buonassisi et al. [9] and Schön et al. [70]. In particular, Kveder et al. presented a model being capable of fitting contrast values of dislocations versus temperature measured by electron beam induced current (EBIC) [7]. Note that the EBIC contrast of a defect is directly connected to its recombination activity. According to their model, it turns out that the shape of the contrast versus temperature curve of the investigated dislocation gives evidence of its metal contamination. Hence, fitting these curves before and after PDG allows a qualitative description of reducing the impurity concentration at dislocations by gettering.

Rinio et al. showed that external gettering by the phosphorus emitter is the dominant mechanism during a low temperature annealing (LT anneal) [127]. They annealed comparable mc-Si samples with and without the emitter present at low temperatures below 600^◦C. After the emitter removal (of samples with emitter), all samples are processed into heterojunction solar cells. Light beam induced current (LBIC) measurements will reveal a beneficial effect only if the emitter is present during the LT anneal.

The authors attributed the benefit to a gettering mechanism dominated by the external gettering sink, i.e., the emitter and the PSG. Impurities seem to be gettered mainly by the emitter region and not by internal gettering sites. This might be expected since the distance that impurities have to overcome in order to be gettered at the surface is only the sample thickness of 260 µm in this experiment. In contrast to that lateral distances from intra-grain regions to the nearest grain boundary acting as an internal gettering sink are in the order of millimeters. Note that their experiment cannot distinguish whether impurities are mainly gettered by the PSG layer, by the emitter itself or by the interface between these two regions.

This issue is particularly addressed in Chapter5.

Segregation getteringis one of the two main mechanisms occurring in PDG. Similar to the segrega-tion during crystallizasegrega-tion of silicon, impurities have a different solubility limit in the crystallized silicon than in the gettering region which is the melt in this case. The process is therefore also termed as segre-gation into a second phase. Depending on the temperature it can involve segresegre-gation between the solid solution of an impurity in silicon and its solid second phase, e.g., metal-silicide formation. A frequently found metal-silicide stoichiometry is MSi₂ [72]. At higher temperature the second phase can also be a liquid alloy. In addition, it is possible that the phosphosilicate glass (PSG) forms at least partly a liquid phase with higher solubility for impurities [128]. This aggregation state is likely to occur in highly P doped SiO₂leading to a reduced melting point. The segregation coefficient is the ratio between the equi-librium impurity concentration in silicon and the one in the respective second phase. On the occasion that the silicon solution is supersaturated with impurities the gettering is termedrelaxation gettering. It would completely stop if the impurity concentration is below the solubility, whereas segregation getter-ing continues due to the difference in solubility limits of the gettergetter-ing layer and the Si bulk [129]. A supersaturated solution is typically accompanied by the formation of precipitates [69].

A consequence of the already mentioned kick-out mechanism is the higher amount of mobile inter-stitial impurities, which can easily diffuse through the Si bulk. This is an important fact when trying to explain phosphorus gettering phenomena. If impurities can easily diffuse they are more likely to be trapped at gettering sites located some distance away from the original impurity spot. Gettering sites in mc-Si are for example grain boundaries and dislocations offering more space for the gettered impurities than the undistorted crystal lattice. The reaction equation describes the kick-out of a substitutional metal impurity M_sby a Si interstitial:

Ms+SiiSi+Mi (2.14)

Since it is based on the injection of point defects (Si interstitials) that type of gettering is known as injection gettering. This mechanism typically enhances PDG.

In the following a description of the microscopic processes in PDG is presented based on the model suggested by Seibt and Kveder [72]. Fig.2.11is divided into six steps which does not necessarily mean that gettering proceeds in these fixed sequences. Some processes might also occur simultaneously.

1. The silicon surface is exposed to an atmosphere containing POCl3-N2, N2and O2gas.

2. The highly P doped SiO2 (PSG) is formed by consuming a considerable amount of silicon as it gives rise to a volume expansion of ≈130 % [3,130]. It is also accompanied by the injection of silicon interstitials.

If the phosphorus content exceeds its solid solubility in silicon precipitation occurs, namely SiP precipitates are formed. As stated in the introduction the total phosphorus concentration depends on the gas mixture (gas flow rates and partial pressures), temperature and time. Precipitates might be a second source for silicon interstitials due to their volume expansion. The main gettering mechanism is sketched for an interstitial impurity which is transported towards the PSG layer by segregation.

3. The gettered impurity forms a second phase at the interface between PSG and the Si bulk. As already mentioned this phase can be liquid or solid. Here the frequently observed compound is sketched: metal-silicide MSi₂.

The third source for interstitials is the pairing of P with a Si interstitial PSi_i. Schröter et al. claim this pair to be the main source of Si self-interstitials [16].

4. This pair is a fast diffusing complex which is one of the mechanisms leading to the specific kink-and-tail shape of P profiles. Note that phosphorus also diffuses via vacancies in the kink region which is, for the sake of clarity, not sketched here.

5. PSiidissociates at a certain depth of the Si bulk and injects the Si interstitial.

6. Due to the different processes of injecting Si self-interstitials a supersaturation of these point de-fects is established. This leads to the already described injection gettering. A substitutional impu-rity is kicked out by the Si atom and subsequently free to be gettered. In contrast to segregation gettering this mechanism is a non-equilibrium process due to the supersaturation of Si interstitials which means a concentration beyond its equilibrium value.

In addition to segregation gettering, relaxation gettering will take place in regions where supersaturation with dissolved impurities is reached [69]. The mechanisms described above emphasize the importance of Si self-interstitials in PDG of substitutional impurities which are kicked into mobile interstitial sites by the Si interstitials.

2.3 POCl3Emitter Formation and Diffusion Gettering 39

PSG Si PSG Si

Injection of Si interstitials

e.g MSi₂ formation

PSi_i pair formation Fast diffusing PSi_i

PSG Si

PSi_i dissociation

PSG Si

Kick-out and gettering

3 4

5 6

SiP precipitate at PSG/Si interface

Highly P doped SiO₂ formation

PSG Si

Segregation

Si 2 1

Silicon ( interstitial) Phosphorus Impurity Oxygen

Kick-out Predeposition

in POCl₃/N₂ + N₂ + O₂ atmosphere

Figure 2.11: Six steps depicting PDG by segregation and injection gettering. Note that all steps can occur simultaneously but are depicted separately for the sake of clarity.

Note that iron is preferably located at an interstitial lattice site or bound to boron in p-type Si [59].

Hence, injection gettering should not be relevant in the particular case of iron gettering. Schröter et al., however, describe a gettering mechanism at the PSG/Si interface involving substitutional impurities such as iron in the kink region of the phosphorus emitter [16]. It should be mentioned still that iron is not always the only lifetime-limiting impurity in mc-Si as will be discussed in Chapter 6. Iron gettering pro-ceeds as relaxation and segregation gettering. The latter is explained by the pairing between negatively charged substitutional iron and the positively charged phosphorus. Experimental evidence of such a pair-ing between the active phosphorus and the substitutional iron was first given by Gilles et al.. It explains the strongly increased solubility of 3d transition elements within the emitter region by four orders of magnitude and hence the strong segregation gettering of iron [125]. It also stresses the strong impact of the active phosphorus concentration [P⁺] on the emitter’s PDG behavior.

Nevertheless, precipitation is of high importance particularly for mc-Si which in turn correlates with the density of structural defects acting as nucleation centers as pointed out by Schön et al. [70]. Further-more, Syre et al. emphasize the role of vacancies located in the near-surface emitter region (kink) during PDG [118]. They observed a correlation between the oxygen content within the emitter and the amount of gettered iron. In their publication, vacancy-oxygen complexes are assumed to form which act as nu-cleation centers for iron precipitates. In contrast to Gilles et al., Syre et al. claim an indirect influence of phosphorus on the gettering of iron originating from the formation of vacancies during the P in-diffusion. They suggest an iron-oxygen-vacancy complex Fe-O-V. Chen et al. identify the phosphorus-vacancy complex P₄V as the dominant complex being formed in high concentration phosphorus emitters and being the most effective gettering sink for transition metals [131].

The various processes, involving either interstitials or vacancies, underline the fact that the exact atomistic mechanisms behind PDG are still unclear and that PDG is subject of current research.