Silicon for solar applications: the path to perfection

Introduce a slight imperfection to make an ultra-pure material even more perfect – this daring and seemingly counterintuitive approach is the fruit of decades of scientific research. Hans Queisser, a founding father of photovoltaics, retraces the history of this research in his aptly titled review ‘Defects in Semiconductors: Some Fatal, Some Vital’ (Science, 1998). It tells the story of “a solid with irreproducible properties” to the technological workhorse silicon is today. The enabler for this success story was the control and use of imperfections in a smart way nowadays termed defect engineering.

Silicon-based photovoltaics stands firmly on the understanding of the vital and fatal impurities Queisser describes. In every solar cell the vital impurities, mainly dopants and passivating agents, are precisely dosed for optimum cell performance. The fatal impurities, primarily transition metals, must be understood in their deleteriousness and kept below critical levels. At the same time, costly and unnecessary over-purification should be avoided. Defect engineering and control is therefore critical to reach maximum performance and reliability at minimal cost.

Summarising, purity is no end in itself: wafers destined for photovoltaic applications must include well-dosed impurities to reach ideal electronic properties. Indeed, if we extracted a wafer from the Avogardo Sphere and converted it to a solar cell, the result would be nothing but a very expensive mock-up – and a group of hopping mad crystallographers.

12N, 9N, 6N purity feedstock for a 5N purity product

Based on the experience described above, silicon feedstock is engineered to reach an optimal conductivity in the wafer by adding just the right amount of friendly impurities. This step, known as doping, is typically done with boron or phosphorous. As a result, the purity of even the cleanest feedstock is reduced from a purity of 99.9999999999% (12 Nines) to a purity of about 99.9999% (6 Nines). In addition, as the feedstock is molten and solidified to incorporate the dopant atoms, other impurity elements will also enter the melt from the crucible walls or from the furnace itself. As a result, typical impurity concentrations in mono or multi-crystalline wafers are high enough to reduce the material to a purity of 5N. The N-purity is thus not necessarily a relevant quality criterion.

Why then use the expensive high-purity feedstock? In the European installation frenzy between 2007 and 2009, skyrocketing margins for both feedstock and cell producers led to a tumultuous market in which almost any silicon metal was good enough to make a sale. In some cases, improper casting control or unsuitable feedstock introduced intolerable amount of fatal impurities into the wafer. This resulted in cells with limited performance or maybe worse – reliability issues in the long term.

The solar community has since learned two lessons. First, it was recognised that controlling relevant defects in the wafers is a prerequisite. Second, using high “N” purities is an expensive over-simplification. An indication of this learning is reflected in the market place where high N-purity is no longer a sales pitch: the price premium mandated by 9N+ over so called second-grade solar silicon (6N-8N) has all but vanished over the last three years (data from www.pvinsights.com).

Defects in Solar Cells: Some Fatal, Some Equal, Some Vital

The solar engineer has the choice between two approaches to reach high-performance wafers: the simpler one is based on a tabula rasa approach and requires high-purity feedstock. Recycling of cuts from earlier castings and/or blending with lower grades of silicon is naturally forbidden in this scenario. While this is a technologically viable path, it comes at the price of high feedstock expense and low yield.

The experienced solar engineer will use his scientific understanding and his own empirical experience to use price-competitive feedstock and recycling to the best economical advantage without compromising cell performance and reliability. This entails an understanding of the concentration limits of fatal “enemies” and the less relevant secondary impurities as well as an exact addition of vital dopants. The solidification of the ingot casting may also be understood and used as a purification step allowing further softening of purity limits.

Producers of solar-grade silicon have recognised this business opportunity and have established simplified purification processes based on gas-phase deposition such as flow bed reactors or metallurgical purification methods. A stringent control of fatal impurities in the feedstock such as molybdenum, zirconium, tungsten and other so-called deathniums may be necessary to a level below 10 part per billion (8N) in some cases. Other transition metals such as iron, aluminum or copper or main group elements such as oxygen, carbon or calcium will be acceptable at a much higher level without compromising cell performance and reliability. Ultimately, it is the solar cell itself that is the best empirical sensor for the optimal feedstock.

The challenge of the supply chain is to understand the impact of impurities on the solar cell and to provide the best possible product based on this feedback. It is therefore to be expected that the feedstock producers will explorer further potential for cost reduction in their purification processes – moving away from absolute purity levels towards a fine-tuning of impurity concentrations. The narrowing price gap between 6N-8N and >9N material is an indicator that the market approves of this evolution.