Stacking Sidesteps The Strain In Multi-junction Cells
The solar industry is changing. Although silicon is still the dominant material for solar cell manufacture, alternative technologies like thin-film and concentrator photovoltaics (CPV) are emerging. The latterof these promises to excel in sunnier climes and involves the focusing of incident sunlight onto cells with a typical area of just 10-100 mm2. By focusing the light by a factor of several hundred, it is possible to minimize the total expenditure on these relatively expensive cells, and ultimately realize an acceptable cost-per-Watt at the system level.
Generating costs for CPV systems are also influenced by the efficiency of the cell. Single-junction solar cells are inappropriate, because they are limited to a theoretical maximum efficiency of about 30 percent, due to thermalization and transmission losses. In comparison, multi-junction cells can produce far higher efficiencies, because they cut thermalization losses by using several cells to absorb different parts of the solar spectrum.
Today the dominant multi-junction solar cell technology is the monolithic triple-junction (InGaP/(In)GaAs/Ge) solar cell that was originally developed and commercialized for space applications. These devices feature In0.5Ga0.5P, In0.01Ga0.99As and germanium cells that are lattice-matched to a germanium substrate to ensure excellent material quality and photovoltaic performance.
The conversion efficiency record for a multi-junction cell under concentrated irradiation has been broken on several occasions over the last few years. Spectrolab has made the most recent claim for the record, and in August 2009 it announced an efficiency of 41.6 percent, which was achieved under a concentration of 364 suns. But this record may not last for long because further improvements in monolithic multi-junction cells are expected. These could come through either adding more junctions to the stack, or using lattice-mismatched solar cells. In the latter case, cell compositions are modified to yield a superior combination of bandgap energies. The downside of this approach is that each cell differs in its crystal lattice spacing, so additional buffer layers are needed to pin the crystallographic defects and prevent them from degrading the performance of the active layers.
Regardless of the form of monolithic multi-junction solar cell, current matching between different cells is essential, due to the inherent series connection of these integrated devices. In addition, there is a need for tunnel junctions, which are applied to electrically connect the different cells in the stack. These junctions can handle the high peak tunneling currents.
Fig.1. Multi-junction cells deliver higher efficiencies than single-cell equivalents by using materials with differing bandgaps to absorb different parts of the solar spectrum. This approach reduces thermalization losses
In real-life CPV applications there is an additional complication too - non-uniform illumination levels on the solar cell. This may lead to local current densities exceeding the tunnel junction design value, and ultimately higher resistances and voltage losses. Deviations in the spectral distribution of the incident sunlight occur all the time, because they depend on changes in geographical, seasonal, daily and climatic conditions. This makes it very tough to current match cells for optimal energy yield, and imposes very stringent requirements on cell design for specific operating conditions and locations. The latter of these also needs to be very well documented.
Fig. 2.Mechanically stacked junctions employ an electrical contact for every junction. Thanks to this approach, there is no need to current-match the junctions, or use tunnel diodes
Stacking cells: pros and cons
A promising alternative to these monolithic cells is a mechanically stacked multi-junction architecture. With this approach, different single-junction solar cells are integrated by mechanically placing them into a stack, such that each cell absorbs a different part of the incident spectrum (figure 2). Each of these cells has a separate electrical contact. This means that they do not have to be connected in series, which is a massive benefit because it removes the need for tunnel junctions and current matching.
Eliminating the need for current matching also produces additional, important advantages – it allows full exploitation of the power generated by every cell within the device, and it creates an inherent robustness against variations in the spectral distribution of the incident light. What’s more, this approach offers the freedom to realize any combination of cells with different energy bandgaps, without the need to worry about lattice-matching issues.
However, all of these advantages have to be weighed against three specific challenges associated with mechanically-stacked solar cells that have hampered progress by their developers, and prevented commercialization: bulkiness; a complex electrical architecture; and optical coupling requirements.
Bulkiness is to a certain extent inevitable, due to the use of different solar cells and their associated substrates in a mechanical stack. In addition, multiple substrates push up costs. Specifically for CPV applications, the thermal mass of the full mechanical stack also offers a major challenge concerning heat dissipation.
The second issue, the complex electrical architecture, stems from the need to provide individual electrical interconnections to every cell. This can be addressed at the system level with an intelligent string and inverter design (a string is a number of individual cells that are series connected by external circuits to obtain a larger DC output voltage, which can then be converted to an AC source for the grid with an inverter). However, this does not eliminate the need to integrate electrical leads on every cell in the stack, which has a major impact on cell and stack development and technology. This is especially a concern for cell interconnects that need to be placed on one of the intermediate cell surfaces within the stack.
Compared to its monolithic cousin, a mechanically stacked multi-junction solar cell has additional surfaces and interfaces, and it requires additional adhesive layers in between the different cells. These increase the number of sources of optical loss, which might have an important impact on the final performance of middle and bottom cells in such a stack. Moreover, in order to fully transmit the non-absorbed light from an upper cell to the ones beneath it in the stack, the higher cell must be optically transparent to its sub-bandgap radiation.
Efforts at IMEC
At IMEC, which is based in Belgium, I am working with several other researchers to develop mechanically-stacked solar cells that address the above-mentioned problems. We hope that our efforts will ultimately enable these devices to fulfill their high-efficiency potential and deliver the benefits associated with their inherent robustness to spectral variations.
The technologies used to produce these cells are compatible with high-throughput manufacturing. Specifically, a mechanically stacked triple-junction InGaP/GaAs/Ge cell is under development, exploiting the same sub-cells employed in current state-of-the-art monolithic cells.
In the proposed configuration, the InGaP and GaAs subcells are processed such that the germanium substrate, onto which the epitaxial layers are deposited, is removed. In this way, an InGaP cell with a typical thickness of about 1 μm and a 3-4 μm thick GaAs cell can be stacked on a separately realized germanium bottom cell. This architecture will allow the extraction of the full current generated by the germanium cell, which is not the case in a monolithically stacked triple-junction cell.
Removing the substrates from the top and middle cells should lead to easier transport of excessive heat towards the heat sink. A further benefit is that it allows interconnection of the contact grids of the different cells from the stack’s front or rear side. This makes the electrical design less complex, opening the door to stacked cell processing on the wafer scale, which is a key element in upscaling this technology for the production of high-efficiency concentrator cells.
Final integration of the individual cells into a mechanical stack also requires know-how and tools from the semiconductor manufacturing industry, specifically 3Dstacking expertise, an area of technology where IMEC has considerable strength. Adopting this approach will produce high-quality bonding, integration and interconnection processes, while also offering the possibility to perform high-accuracy alignment of the different cells in the stack. This should result in a good alignment of the different contact grids applied to the front and backside of the individual cells, such that optical coupling between the different cells in the stack is not hampered by unwanted additional shadowing losses due to the contact structures.
One significant step towards the fabrication of this triplejunction stacked cell has been the fabrication of a dualjunction version based on GaAs and germanium. This has been made by combining 4 μm, one-side-contacted GaAs solar cells and separately connected germanium solar cells. The thinned-down GaAs solar cells were bonded on top of the germanium solar cells using silicone sealant. Transmission measurements on a layer of this sealant with relevant thickness (~20 μm) revealed that the transmission loss through this layer is limited to 3.5 percent in the 900-1800 nm wavelength range.
Fig.3. IMEC has already built a mechanically stacked dual-junction cell based on GaAs and germanium. The next goal is the addition of an InGaP top cell
The presently realized mechanical stack (figure 3) employs individual GaAs and germanium cells that have a nonmatched cell area and contact grid design. This means that the longer-wavelength portion of incident light that passes through the GaAs cell is not optimally coupled to the underlying germanium cell. Performance is also compromised by the absence of an anti-reflective coating at the rear of the GaAs cell that limits the transmittance of infrared radiation to the bottom cell. But the good news is that the applied technologies for thinning down the GaAs cell and bonding it on top of an active germanium solar cell are readily applicable in an optimized, stacked-cell design.
We have found that thinned-down, one-side-contacted GaAs cells integrated in the mechanical stack exhibit identical results to those obtained previously on similar stand-alone cells. They can produce conversion efficiencies exceeding 23 percent (1 sun, AM1.5), close to our institution’s best results for regular GaAs solar cells on a germanium substrate of 24.7 percent. The main limiting factor for the one-side contacted GaAs cells is the relatively low fill factor. This is caused by the use of the two-point measurement method during ‘I-V’ characterization, which is imposed by the small available area for contacting the rear-side grid in the present configuration.
The separately contacted germanium bottom cell in our stack exhibits a conversion efficiency of almost 2 percent (1 sun, AM1.5). The low efficiency can be attributed to the absence of an anti-reflective coating at the rear of the GaAs cell, and the use of a non-optimized coating on the front side of the germanium cell. Applying an optimized coating to both these surfaces should lift the efficiency of the germanium bottom cell to approximately 3-3.5 percent, which is well above the contribution of the germanium cell in conventional triple-junction cells.
Efforts will now be directed at increasing solar cell performance through the reduction of reflective losses at different cell surfaces and matching of the area and contact grid of the III-V and germanium cells. The bonding process demands high-accuracy alignment of the different cells’ contact grids, which can be realized with flip-chip bonding, according to scanning acoustic microscopy measurements. When an ultra-thin, one-side contacted InGaP top cell is added to the structure, this should yield mechanically stacked triple-junction solar cells for CPV applications that feature efficiencies of around 40 percent and enhanced spectral robustness.
Finally, it should be noted that the application of a contacting layout that makes the contacts of all cells in the stack accessible from the stack’s front and rear surfaces - as demonstrated in the present work – is instrumental in allowing wafer-scale processing of the full mechanical stack right up to the final step. This way, dicing to individual solar cells can be performed as a last step, right before cell laydown and interconnection.