Increasing cell efficiencies, trimming material costs and replicating manufacturing processes employed in the LED industry could cut CPV's electrical generation costs to below those of silicon

RICHARD STEVENSON REPORTS



Some makers of CPV modules are adopting a similar approach to automakers by employing robots for many manufacturing steps. A switch to a manufacturing process that is similar to that used in the LED industry could enable a reduction in capital expenditure by a factor of up to ten.

The concentrating photovoltaic (CPV) industry is in poor shape. A sharp decline in the price of silicon cells and the failure the thin-film pioneer Solyndra have scared away investors in all emerging solar technologies, and led to the death of many makers of CPV systems and their components. Big names to fall by the wayside include the US firms GreenVolts and SolFocus, and many smaller outfits have fared no better, including UK start-up Circadian Solar that has just shut its doors.

Given this sorry state of affairs, and the combination of a rail strike and gloomy skies, it would not have been surprising to see many glum faces at Advances in Photovoltaics. But that was certainly not the case: optimism permeated throughout this meeting.

During this conference on 5 February at Imperial College London, several speakers began their talks by highlighting the encouraging report from market analyst GTM Research, which suggests that CPV should fall from $3/W in 2011 to below $1.2/W by the end of this decade, a decline that is large enough to make this technology competitive with other forms of photovoltaics.

Bringing down the cost of CPV will hinge on the introduction of cheaper, more efficient multi-junction cells and lower-cost modules and tracker systems. Several options are available for succeeding in all these endeavours, and some were outlined at the meeting: Andy Johnson, director of CPV technology at the global epiwafer provider IQE discussed approaches to bringing down chip costs; Geoff Duggan, CTO at module developer Fullsun Photovoltaics outlined the cost-savings that could be realised by switching from robotic-heavy processes to those used in the LED sector; and several speakers from academia and industry detailed modifications to cell architecture that could catapult efficiencies to 50 percent and beyond.

Today, IQE, a company with around 90 multi-wafer MBE and MOCVD tools, nets four-fifths of its sales from wireless products. So why is it interested in CPV? Because, according to Johnson, the potential sales are huge. He pointed out that installations of PV are rising, should hit 50 GW by 2015, and if CPV captures just a tiny fraction of that market, it will equate to many, many wafers – just 1 GW of power from CPV will require the production of approximately 300,000 6-inch wafers. “In terms of wafer volumes, a small proportion of the global solar market will be bigger than the entire III-V market of today.” IQE has two keys roles to play in the CPV industry: bringing down costs and increasing efficiency.

To cut costs, the company is promoting growth on larger substrates. Johnson argues that if an epiwafer foundry performs deposition of multi-junction cells on state-of-the-art, multi-wafer tools, savings result from a combination of lower cost-of-ownership, a longer campaign length between maintenance breaks and higher yields.

Today, the conventional multi-junction cell is formed by depositing (In)GaAs and GaInP on germanium substrates. This foundation is available in 6-inch and 8-inch formats, but the latter, which is not an established product, is very expensive. Note that it is possible to accommodate both sizes of substrate in the latest multi-wafer tools, such as the Aixtron 2800 G4: this MOCVD reactor is capable of housing eight 6-inch wafers or five 8-inch wafers.

Cost savings can also result from depositing layers at faster growth rates, because this boosts the throughput of the reactor. IQE’s process engineers have succeeded in increasing MOCVD growth rates while maintaining high crystal quality, and can now deposit InGaAs and GaInP epilayers on germanium substrates at 15 m/hr and 5 m/hr, respectively. The company is targeting a growth cycle time of 90 minutes, and has long-term ambitions to eventually shortening this to just an hour.

Commitment to providing the CPV industry with state-of-the-art chips is highlighted by IQE’s seven-year exclusive manufacturing agreement with US multi-junction cell provider Solar Junction. IQE is to provide all the epitaxy for the US firm, and is also supporting efforts to get the cells qualified for use in space.

Growing the epilayers for Solar Junction’s cells presents new challenges for IQE, due to the insertion of a lattice-matched dilute nitride rather than germanium for the bottom junction. This switch to a quaternary cell ups the output voltage from around 3.2 V to 3.5 V, and is the key to the company’s record-breaking triple-junction cells. What’s more, this device architecture offers a promising path to four, five and six junction cells with ever higher efficiencies.

Development of these novel devices at Solar Junction involved growth by MBE. This is the starting point for IQE. However, the company is considering whether it is better to deposit the dilute nitride by MBE before growing the remaining layers by MOCVD – and ultimately, whether it is possible for MOCVD to form the entire structure.

 

By 2020, CPV systems could deliver lower generating costs than those based on silicon, according to GTM Research

Taking the strain

Introducing dilute nitrides is by no means the only option for modifying the conventional triple-junction cell so that it can reach higher efficiencies. Another way forward, detailed by Ned Ekins-Daukes from Imperial College London, is to turn to strain-balanced layers for each cell. These can all be lattice-matched to the substrate.

This approach has been pioneered by researchers at Imperial College, and led to the launch of spin-off company Quantasol, which was acquired by JDSU in the summer of 2011. Success by the start-up included a world record efficiency of

 28.3 percent for a single-junction cell, which laid the foundations for JDSU’s construction of a triple-junction cell operating at 42.5 percent. This featured optimisation of absorption and current matching.

In addition to highlighting these record-breaking cells, Ekins-Daukes outlined various opportunities to reach efficiencies of 50 percent or more. He claimed that it is possible to realise a 50.1 percent efficiency with a four-junction cell based on bandgaps of 1.99 eV, 1.55 eV, 1.15 eV and 0.66 eV.

According to Ekins-Daukes, fabricating a cell with a bandgap close to 1 eV presents the biggest challenge. If InGaAs were employed, this layer would be relaxed and riddled with threading dislocations that drag down the output voltage and impair efficiency. However, it is possible to overcome this issue by growing the structure upside down, before turning to wafer bonding and substrate removal processes.

Alternatives for a 1 eV layer are dilute nitrides and dilute bismides. Echoing the thoughts of Johnson, Ekins-Daukes said that MBE is the superior growth method for depositing dilute nitrides, but dilute bismides, such as GaAsBi – which are also lattice-matched to GaAs – can be deposited by both MBE and MOCVD. Another option for realising absorption at around 1 eV was discussed by Tony Krier from Lancaster University: Add GaSb quantum dots to a GaAs solar cell. He and his co-workers have pursued this, fabricating cells with a ten-layer stack of GaSb nanostructures. They form rings rather than dots, with outer and inner diameters of around 23 nm and 10 nm.

Photoluminescence measurements reveal a peak associated with the GaSb rings at 1.4 m. However, the associated gains in increased spectral coverage come at the expense of a decline in the open circuit voltage at one sun from 1 eV to 0.6 eV. Krier blames this on the trapping of holes, which then act as recombination centres. Encouragingly, increasing concentration from 1 sun to 2500 suns boosts the open circuit voltage to around 0.9 eV. What’s more, the addition of the GaSb rings produces a 6 percent gain in short-circuit current to 27.9 mA cm-2, indicating that these nanostructures may be able to deliver an increase in cell efficiency.

Switching to InAlAs/InGaAs multi-junction cells with a lattice constant close to that of InP offers yet another opportunity for increasing device efficiency. Donagh O’Mahony from University College Cork is pursuing this. He explained that the efficiency sweet spot is at 5.80-5.81 Å, so it is nearer to the lattice constant of InP (5.87 Å) than that of GaAs (5.65 Å). Manufacturing devices on InP is not viable, however, argued O’Mahony, because chip costs would be prohibitively expensive. Instead, he believes that engineers must start with a GaAs substrate and employ multi-stage grading layers.

The promise of these novel devices is yet to be fulfilled. Initially the single-junction cells that had an alternative lattice constant and were grown on GaAs had an efficiency of 6-7 percent, compared with 12.8 percent for those formed on InP.  This difference is partly due to a fall in output voltage by up to 0.3 eV. It is also worth noting that even the cell on InP falls short of what should be possible – O’Mahony claims that greater than 20 percent efficiency is to be expected.

Mirroring LED production

Another area where there is room for cutting costs is the manufacturing approach for making the modules. According to Geoffrey Duggan, CTO of Fullsun Photovoltaics, today several firms within the industry are using many robots for module manufacture. The capital expense required to build a module production plant is then very high, and this prevents many start-ups from entering the CPV sector, due to the challengers of obtaining finance.

Duggan argued that those firms that have built production lines with robotic processes, such as market leader Soitec, should continue on that path. However, he advises newcomers to try and adopt the assembly methods used in the microelectronic and optoelectronic industries. This means using surface mount components and turning to smaller cells – dimensions of less than 2 mm by 2 mm are recommended.

Several companies, including Fullsun, Panasonic and Semprius, are pursuing this alternative approach that can slash capital expenditure by a factor of five to ten. The latter has developed a module with an output of more than 90 W, which features hundreds of microcells, weights just 6.5 kg, employs a silicone-on-glass lens array and operates at a concentration of 1100 suns.

Given the difficulties of launching a CPV firm today – the collapse of a high proportion of the 50 or more companies formed between 2005 and 2008 has led investors to shy away from this industry – newcomers will need far lower capital expenditure than before to enter this sector.

Winning CPV deployment contracts would be aided by more efficient cells produced at lower costs, but success on all fronts is possible, thanks in part to the efforts of those that attended Advances in Concentrator Photovoltaics