Wafer Bonding Creates Record-breaking Four-junction Cell
To prevent the formation of efficiency sapping defects, conventional multi-junction cells are built with lattice-matched materials. But this restriction can be lifted with wafer-bonding, which enables the fabrication of a four-junction cell with record-breaking efficiency, say Rainer Krause and Bruno Ghyselen from Soitec and Frank Dimroth from the Fraunhofer Institute for Solar Energy Systems.
Concentrating photovoltaics(CPV) involves the generation of electricity through the focusing of sunlight onto solar cells.
At Soitec of Bernin, France, we are one of the world’s leading manufacturers of this class of solar system. We manufacture modules that use Fresnel lenses to focus sunlight onto cells at a concentration of over 500 suns and we are analysing optical concentrators that would allow us to increase further the concentration level with the same degree of reliability (see Figure 1). To ensure that these modules generate as much energy as possible, they are mounted on tracking systems that follow the sun’s position in the sky from dawn to dusk. Such systems are deployed on solar farms, and deliver their best returns in locations with bright sunshine (regions said to have a high value for Direct Normal Irradiation, or DNI).
Figure 1. Soitec’s concentrating photovoltaic modules employ high efficiency III-V multi-junction solar cells, a Fresnel lens array with a relatively small aperture, and glass for the front cover and bottom plate
We are currently making our systems more competitive by working on different elements of the system. On the solar cell side, by teaming up with researchers from Fraunhofer ISE and CEA-Leti, we are running a programme to take cell efficiency to a new level. Our new devices feature four junctions – one more than the traditional multi-junction cell – bringing the achievable maximum efficiency to around 50 percent. Onthe way to hitting this level of performance, in 2013 we raised the world record for efficiency to 44.7 percent by uniting a top tandem cell of GaInP and GaAs with a bottom tandem cell made from GaInAsP and GaInAs.
We are by no means the only multi-junction cell manufacturer or developer that has devoted a great deal of effort to improving device efficiency. That’s because gains in efficiency can lead to significant reductions in the levelised cost of energy.
Before we unveiled our ground-breaking device, increases in the record for efficiency – which have recently increased by about one percent a year – resulted from refinements to the conventional triple-junction cell. In its standard form (see Figure 2), it comprises a germanium (0.7 eV) bottom junction and middle and top junctions of GaAs (1.4 eV) and GaInP (1.9 eV).
Figure 2. A state-of-the-art triple junction solar cell features a bottom junction made from germanium, and additional junctions based on GaInAs and GaInP
Several firms, including AZUR-Space in Germany, and Boeing-Spectrolab and Emcore in the United States, manufacture this incumbent design using mature, high-yield industrial production processes. Such devices are deployed in commercial systems, where they can reach efficiencies of 41 percent [1, 2].
Features of the standard device include two terminals – a front contact and a backside contact – and the connection of the three cells in series, so that a single voltage is delivered at the cell level. The epitaxial structure of the multi-junction cell is usually formed by MOCVD. There are limits to what is possible with this approach, due to epitaxy and lattice matching, and if the material quality is not high, charge carrier recombination can impair device performance. So, to optimise the efficiency of the cell, it is crucial to carefully balance the characteristics of every junction, so that they work well together.
Two well-known routes exist for improving the efficiency of multi-junction cells. One is to fine-tune its absorption profile, so that the contribution from every junction combines to propel the overall efficiency to a new high. And there is also a more ambitious approach: To add additional junctions, starting with a move from three to four.
We are not alone in targeting a four-junction device with cells operating at the optimum bandgap energies of 1.9 eV, 1.4 eV, 1.1 eV and 0.7 eV. It has been demonstrated that it is possible to reach these energies by introducing dilute nitrides, such as GaInNAs, into the conventional GaInP/GaAs/Ge stack . Meanwhile, researchers at NREL followed by other industrial groups, have proposed the use of an inverted metamorphic four-junction solar cell .
Our approach is different. While the number of junctions and the choice for their energies remain basically the same, a totally different and key technological step is added to the epitaxy tool box: wafer-bonding. The concept of applying wafer bonding in this arena is not completely new [5, 6]. However, we are the first to use this to form, at full wafer level, cells with an efficiency that exceeds every triple-junction device. Introducing wafer bonding has enabled the marriage of lattice-mismatched materials without the creation of dislocations, so GaAs and InP can be united.
To join materials such as these, two crystal structures must be brought together to form covalent bonds at the interface. Success demands that materials are carefully prepared, and their surface roughness is low. Once these conditions are met, it is possible to yield multi-junction solar cells based on defect-free material.
We have developed a specific wafer-bonding process for uniting InP- and GaAs-based materials to form a transparent, electrically conductive design. In our case, bonding is not restricted to bulk wafers, but includes ternary and quaternary epilayers, thereby enabling an incredibly wide choice of optimum materials.
InP bulk material is more expensive than GaAs or germanium wafers, but this is not a stumbling block for us, because we can leverage our Smart Cut technology, which we have used in volume manufacturing for more than a decade. Armed with this, we use only a very thin layer of initial material, and can consequently re-use the InP substrate many times. Once a thin layer of InP has been extracted from the substrate, it can be transferred to many different types of carrier. This led us to introduce the acronym ‘InPOX’ as a generic name for ‘InP-on-X’, where X can include silicon, GaAs, germanium and sapphire. An example of this is the transfer of a 0.5 µm-thick layer of InP to 100 mm sapphire. Working with partners at Fraunhofer ISE and CEA-Leti, we have used our Smart Cut technology in conjunction with direct wafer bonding and III-V epitaxial growth to produce a record-breaking four-junction cell. Our collaborators contribute expertise associated with III-V material growth, fabrication of engineered substrates and epitaxial lift-off and bonding techniques.
The device that resulted features a bottom GaInAs junction with a bandgap of 0.7 eV, overlaid with GaInAsP, GaAs and GaInP cells with bandgaps of 1.0 eV, 1.4 eV and 1.9 eV, respectively (see Figure 3). Features of this photovoltaic include low shading losses, which results from front metal contacts with a finger width of 5 µm, and a double layer anti-reflection coating that ensures minimal reflection.
Figure 3. Fabrication of Soitec’s four-junction cell involves wafer-bonding and substrate removal
Using 100 mm substrates, cells with an area of 5.2 mm2 have been processed and characterised to form four active junctions. Development of the fabrication process revealed approaches for dramatically increasing the yield of good solar cells, and this helped us to produce wafers with a yield of functional devices in excess of 95 percent (see Figure 4 for images of finished cells, at wafer level and after die separation).
Figure 4: Soitec’s cell at wafer level (a) and after die separation (b)
Our partners at the Fraunhofer ISE CalLab measured our cell’s quantum efficiencies and current-voltage characteristics under one-sun AM1.5d (1000 W/m2, 25 °C) standard test conditions using a spectrally adjustable solar simulator. To determine performance under concentration, a Xenon Flash simulator with adjustable distance between the flash bulb and the measurement plane illuminated the devices.
A three-dimensional laser microscope enabled an accurate measurement for mesa edge, and thus an accurate value for the efficiency of these small area concentrator solar cells .
Measurements revealed that the peak quantum efficiency for all four cells is well above 85 percent, and that device efficiency hits 44.7 percent at 297 suns. Even at 1000 suns there is no indication of a breakdown in peak tunnel current density, while the high fill-factor of 86.5 percent at this concentration indicates a moderate resistance of the bond interface, which operates well up to current densities of several A cm-2.
We know that it is possible to deliver even higher efficiencies from our four-junction cell. Our plan is to realise this, and efforts at improving device structures are already underway, directed at the optimisation of sub-cell characteristics and further material quality improvement. In the longer term, this will help to trim the generating costs associated with concentrating photovoltaics and empower this technology to displace other types of cell.
The authors acknowledge the contributing work from the CEA Leti team in Grenoble France and the laboratory work of the Fraunhofer ISE team in Freiburg. Also acknowledged is the contributing work from Helmotltz-Zentrum in Berlin. This program is supported by the French Environment and Energy Management Agency (ADEME) through the “Investissements d’Avenir", pending European Commission notification agreement.
 W. Guter et. al.Proceedings of the CPV-7th International Conference on Concentrating Photovoltaic Systems, 2011, pp. 5-8.
 J. H. Ermer et. al.IEEE Journal of Photovoltaics 2209 (2012)
 V. Sabnis et. al.8th International Conference on Concentrater Photovoltaic Systems, Toledo, Spain 14-19 2012
 M. Stan et. al. Journal of Crystal Growth 3121370 (2010)
 K. Tanabe et. al.Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, USA,
2006, pp. 768-71.
 D. Bhusari et. al.“Direct semiconductor bonding technology (SBT) for high efficiency III-V multi-junction solar cells," in
Proceedings of the 37th IEEE Photovoltaic Specialists Conference, Seattle, Washington, USA, 2011, pp. 1-4.
 M. Meusel et. al.Progress in Photovoltaics: Research and Applications 11499 (2003).
The strengths of CPV
Concentrating photovoltaic systems are already cost-competitive in very sunny climes, and in these locations there are several reasons for turning to this form of energy generation. For example, aside from the tiny solar cells located at the focal points of the lenses, the production of a CPV system generally employs low-cost materials, such as glass or silicone, and involves fully automated mass-production conditions.
Another attribute of CPV is that the record-breaking cells not only outperform their rivals at room temperature, but exhibit a far slower decline in efficiency with temperature, as well as much less aging degradation. What’s more, mounting of these cells in tracker systems leads to a more constant power output curve throughout the day (see figure), resulting in not only more energy harvested, but also a higher energy production during peak hours – when it is most valuable. On top of all these arguments, which are associated with the levelised cost-of-energy, there are environmental considerations, such as the small physical footprint of CPV, its absence of water consumption, high levels of recycling, and the opportunity for dual land use with agriculture or animals.
By mounting solar modules on a system that tracks the position of the sun in the sky from dawn until dusk, CPV systems generate significant levels of electricity threoughout the day.