Elsevier

Journal of Crystal Growth

Volume 463, 1 April 2017, Pages 1-9
Journal of Crystal Growth

Investigation of dislocation cluster evolution during directional solidification of multicrystalline silicon

https://doi.org/10.1016/j.jcrysgro.2017.01.027Get rights and content

Highlights

  • The dislocation distances within dislocation clusters can reach nm-scale.

  • It is impossible to measure the dislocation density by means of EPD counting.

  • Dislocation clusters can be divided into light and dense clusters.

  • Which type occurs depends on crystallographic orientation of the grains.

  • An annihilation mechanism of dislocation clusters has been found.

Abstract

Dislocation clusters are the main crystal defects in multicrystalline silicon and are detrimental for solar cell efficiency. They were formed during the silicon ingot casting due to the relaxation of strain energy. The evolution of the dislocation clusters was studied by means of automated analysing tools of the standard wafer and cell production giving information about the cluster development as a function of the ingot height. Due to the observation of the whole wafer surface the point of view is of macroscopic nature. It was found that the dislocations tend to build clusters of high density which usually expand in diameter as a function of ingot height. According to their structure the dislocation clusters can be divided into light and dense clusters. The appearance of both types shows a clear dependence on the orientation of the grain growth direction. Additionally, a process of annihilation of dislocation clusters during the crystallization has been observed. To complement the macroscopic description, the dislocation clusters were also investigates by TEM. It is shown that the dislocations within the subgrain boundaries are closely arranged. Distances of 40–30 nm were found. These results lead to the conclusion that the dislocation density within the cluster structure is impossible to quantify by means of etch pit counting.

Introduction

Multicrystalline Silicon (mc-Si) is actually the main raw material for industrial production of solar cells. This material is characterized by crystal defects such as grain boundaries, precipitations and inclusions as well as other impurities. In particular dislocations can be detrimental for the solar cell efficiency [1], [2]. They tend to build networks of sub-grain boundaries which are commonly mentioned as dislocation clusters [3], [4], [5], [6], [7], [8]. In particular the sub-grain boundaries act as recombination active centres [9], [10]. Sub-grain boundaries can also be mentioned as small angle grain boundaries which can end within a grain or single crystal, respectively.

In [5] the sub-grain structure was studied by means of white-beam X-ray topography (WB-XRT). The sub-grain boundaries were formed due to the rearrangement of dislocations during the crystal growth process. They are mainly aligned along growth direction, and the sub-grains have a slightly tilted crystallographic orientation according to the primary grain orientation. The tilt is mainly a rotation around an axis parallel to the growth direction. Further it is concluded in [4], [5], [8] that dislocations are mainly generated during the growth process in the vicinity of the liquid-solid phase boundary. This phenomenon was also observed by means of in-situ X-ray topography for other materials like aluminium [11], [12], [13].

The origin of a dislocation cluster needs an initial event of dislocation generation. These starting points – or seeds – are mainly located at grain boundaries [7], [14], [15]. Once the dislocation clusters are generated, they tend to successively increase their diameter with the proceeding solid-liquid interface. Thus the amount of dislocations becomes inevitably higher with increasing ingot height.

In this work, dislocation clusters have been studied from a macroscopical viewpoint in dependence of the mc-Si grain structure. The investigations have been carried out by automatized methods normally used for process controlling. They are optimized for short measurement times accepting a lower accuracy. The benefit is a high number of measurements in a short time. The results are supposed to be typical for multicrystalline silicon.

For an insight into the structure on a microscopic scale, the investigations have been additionally supported by transmission electron microscopy (TEM). These measurements have been carried out on a few samples which show typical defect structures. These microscopic investigations usefully complement the information obtained by WB-XRT [5].

Section snippets

Sample preparation

Multicrystalline G5 ingots with dimensions of about 80×80×30 cm3 were produced by the block casting technique (directional solidification). To ensure the growth of dislocations, no modern high performance multi (HPM) process has been applied. Instead, an old standard process was used without supporting the seeding behaviour at the beginning of the crystallization. Therefore, the seeding of the grains has taken place statistically at the crucible bottom.

The ingots have been sliced into bricks,

Texture and defect structure

Fig. 3 shows the typical grain structure and the dislocation distribution of the mc-Si wafer from the top region of a G5 ingot. The grain size ranges from about 5–30 mm. The straight grain boundaries are supposed to be twin boundaries. The dislocation distribution – visible by a dark contrast – is very inhomogeneous, as already known from literature [3], [4], [6]. Regions of high dislocation density are site by site to regions with low dislocation density.

The dislocation clusters are made of

Description of the dislocation structure

The dislocation clusters are made of a network of sub-grain boundaries emanating from dislocation arrays [5]. Within the mc-Si structure they have to interact with grain boundaries, which may stop the expansion of the dislocations clusters or which could been overcome, as shown in Fig. 6. This behaviour may correspond with the type of grain boundary [19]. If the grains have at least one common glide system, dislocations may pass the boundary (Fig. 6d). It is also possible that arrays at grain

Conclusion

By automated analysing tools of wafer and solar cell production, the evolution of the dislocations structure has been described as a function of the ingot height from a macroscopical point of view. Additionally, TEM investigations have provided information about the structural characteristics of dislocations clusters. The TEM results also confirm previous observations made by WB-XRT [5] and other groups [6], [8]. The dislocations clusters are made of a network of sub-grain boundaries (small

Acknowledgements

The authors are grateful to P. Werner and I. Ratschinski for supporting the TEM investigations.

References (31)

  • X.J. Chen et al.

    Study on thermal stress in a silicon ingot during a unidirectional solidification process

    J. Cryst. Growth

    (2008)
  • D. Franke et al.

    Silicon ingot casting: process development by numerical simulations

    Sol. Energy Mater. Sol. Cells

    (2002)
  • B. Sopori et al.

    Performance limitations of mc-Si solar cells caused by defect clusters

    ECS Trans.

    (2009)
  • S. Würzner et al.

    The relationship between microstructure and dislocation density distribution in multicrystalline silicon

    J. Appl. Phys.

    (2010)
  • M. Kivambe et al.

    The microstructure of dislocation clusters in industrial directionally solidified multicrystalline silicon

    J. Appl. Phys.

    (2011)
  • Cited by (0)

    View full text