Alternatives to silicon for solar cells

Michael P.C. Watts

Last time I started to talk about the solar cell business, and whether there was any realistic opportunity for new technologies given the state of the industry. This is the second of a series of blogs to try and answer this question. If you hate serials, the complete analysis is available on my web site www.impattern.com. Let’s start by looking at the potential for other semiconductors.

What limits the conversion efficiency of different materials ?

The sun is a hot black body radiation source that is being converted to electricity by a room temperature band gap semiconductor. Any photon with energy greater than the band gap creates a single electron hole pair, so low energy photons below the bandgap do not create current, and high energy photons do not generate any more current than band gap photons. This limits the maximum conversion to around 30%, as shown below. The optimal band gap for our sun is around 1 eV, and silicon is an optimal material.

Convolution of solar spectrum and a single junction band gap semiconductor from www.pveducation.org

The alternatives to silicon such as GaAs, CdTe, and CIGS all have band gaps around 1 eV, all offering the same maximum possible efficiency.

The alternate materials, CdTe and CIGS, are direct bandgap semiconductors, and as a result have much high absorbtion for a given thickness than indirect bandgap silicon. Higher absorbtion means that the cells can be 20x thinner, enabling “Thin-film solar cells”. Thin cells use less raw materials and can be fabricated on large area substrates. The challenge in thin film PV solar is how to create thin sheets of low defect crystals on either glass, plastic or metal films; the most difficult challenge in hetero-epitaxy. CdTe cells are made by sublimation. CIGS are more exotic in that they are a binary solid solution of CuInSe and CuGaSe, with a number of different deposition strategies; the most effective are sputtering or co-evapaoration. There is a particular wide range in efficiencies reported for CIGS, lower cost production techniques produce much lower efficiencies.

Single crystal GaAs has the best efficiency that is close to the theoretical maximum with polycrystalline silicon at 20%. There are additional losses when the cells are assembled in to modules. Average production module performance summarized by IRENA are; thick polycrystalline silicon 14%, CdTe (9.5%) and CIGS (9%). Seeing as all these material have the same theoretical maximum the differences are in the internal efficiency, which is a typically dependent on the quality of the epitaxy and the semiconductor growth process. The challenge is for the direct bandgap have to match growth quality of silicon. Silicon growth has been researched for as long as transistor have been made, so there is a huge learning advantage to silicon.

The lower efficiency of the non silicon material means that there must be a proportionally larger area of cells to generate the same energy. This has proved to be a major barrier to home installations, but not to large generation plants. The same applies to thin film silicon cells which much less efficient because they absorb less of the light, as a result about twice the area is needed to generate the same power. The market place has shown that even with significantly lower prices, thin film silicon has not been able to compete.

To date, none of the single junction alternatives have created cells with better internal efficiency than thick polySi cells, and certainly do not suggest any immediate prospect of exceeding mature poly-silicon technology.

What levels of cell manufacturing productivity are being achieved ?

Looking at published data for facility capacities and substrate size I estimate that ;

Poly-crystalline Si process line moving 0.15×0.15 m substrates at an impressive 1200 substrates an hour = 27 m2 per hour.

First Solar process CdTe on 1.2×0.6m substrates in 60 MW facility which works out as at 100 substrates an hour = 30 m2 per hour

A German CIGS plant operates at 30 MW facility = 15 m2 per hour.

Amorphous Si thin film plant operating at 75MW , processing 2.2×2.2m glass at 30 an hour = 154 m2 per hour.

My first take is that with an efficiency disadvantage, the cost structure and manufacturing scale of non-Si thin film solutions today are behind polySi, and which explains their competitive challengies.

What is the module cost per watt opportunity ?

To understand the opportunity in module cost, let’s consider a test case of a new factory that Panasonic is building in Malaysia for $540 M, with a capacity of 300 MWatts per year, and will require 1,500 staff. It is easy to calculate the contribution of 5 year deprecation per watt ($0.40) and labor ($0.10). The total material costs in 2012 for a polycrystalline silicon module were estimated as $0.69, with a total module cost of $1.2.

A thin film module factory with the same productivity, capital costs, labor and cell efficiency would have a negligible silicon cost, and higher substrate cost netting out at $0.95 /W (- 20% of polySi) for the module. A thin film factory running at maximum productivity would have 6x higher throughput, making the depreciation and labor components negligible, netting a module cost of $0.55 /W (-54% of polySi).

The manufacturing opportunity for thin film with the same cell efficiency and raw material cost is to halve the total module cost. The latest polySi factories are requiring a capital investment of $2 /W to generate panels at $1 /W. It looks like a thin film factory supporting the highest throughputs could support a capital investment of $4-5/W.

My take is that the alternatives to silicon have a significant leaning deficit to overcome to get closer to efficiency parity, then they need to invest in the most productive manufacturing capacity. These 2 steps would lead to a real 2x cost advantage.

I have combined all my solar blogs in a paper at www.impattern.com