1.

INTRODUCTION

Figue 1 shows that using high quality, inexpensive silicon wafers as templates for the growth of overlying high performance cells offers limiting efficiency close to the maximum possible. If the costs of doing this can be kept small compared to relatively fixed module costs such as those of module encapsulation, frames and junction boxes, this offers the potential for a higher efficiency product at lower cost per unit power or energy output. A range of techniques for producing such tandem cell stacks currently under investigation both within the authors’ groups and elsewhere are outlined and recent progress is reviewed. Key contenders include both GaP/GaAs and Si/Ge metamorphic approaches, “monolayer” interfacial strain relaxation approaches, lattice matched approaches using III-V alloys, chalcogenides or strain-balanced, short-period superlattices, “all-silicon” tandems based on manipulating silicon’s bandgap using quantum confinement, new materials such as perovskites; or finally, mechanically stacking exfoliated cells on a silicon wafer.

Figure 1.

Tandem cell stack and limiting efficiency, with (green) and without (red) the bottom cell constrained to be silicon.

2.

WAFER PRICES/DEPOSITION COSTS

Average selling prices (ASP) of monocrystalline 156 mm quasi-square silicon wafers for photovoltaic use have reduced from a high of $10/wafer in 2008 to just over $1/wafer at the end of 2012 (Fig. 2a). Similarly low prices prevailed during 2013.

Figure 2.

(a) Quarterly ASP of mono 156 nm quasi-square Si wafers 2008-2013 (various sources); (b) International Technology Road Map (ITRPV) market share projections of mono and mono-like wafers (March 2012).

Ongoing price reductions are anticipated from reduced crystallisation costs, including the production of “quasi-mono” material, and from improved sawing processes as well as from thinner wafers. The 2012 ITRPV Roadmap (Fig. 2b) shows what might be possible, particularly given an additional incentive such as provided by the present work, with 90% of all wafers used by the industry projected to be mono or mono-like by 2020. These inexpensive, high-quality wafers would provide ideal templates for the overgrowth of thin, crystalline, wider-bandgap cells, greatly boosting the efficiency potential of Si wafer-based approaches. Even if multicrystalline materials remain the norm, some approaches, such as those involving perovskites would work well.

MOCVD deposition costs for III-V materials are presently high but large reductions are expected due to the extrensive growth of the LED lighting market. Costing studies for photovoltaic use suggests competitive costs could ultimately be attained [1]. The closest step in present silicon cell processing is PECVD deposition of a circa 70 nm nitride antireflection coating, costing less than 5c/W to deposit [1].

3.

EFFICIENCY GAINS

Figure 1 shows the limiting efficiency of tandem cell stacks on Si compared to that of stacks with a free choice of substrate. All cells are assumed to be radiatively limited, apart from the Si cell where performance is limited by intrinsic Auger recombination. For a 2-cell stack, the optimum bandgap of the top cell is 1.7 eV, although a lower bandgap cell can be used if partly transmissive of wavelengths above its bandgap. For a 3-cell stack, the optimal bandgaps for the upper cells are 2.0 eV and 1.5 eV, although again lower values can be used if these cells are partly transmissive. Such partly transmissive designs may be preferred in practice since they will reduce the thickness of deposited material required. Increasing the number of cells in the stack continues to increase Si-based cell efficiency, but at a slower rate than for the unconstrained case.

The highest efficiency tandem cells to date use GaInP/Ga(In)As as the two uppermost cells and either GaInAs or Ge as the bottommost cell, with efficiencies in the 35-40% range for 3-cell stacks for non-concentrated sunlight. Using these two uppermost cells on a Si substrate is likely to give the quickest path to high efficiency both in a monolithic or a mechanical stack, although a range of other options may be able to be developed to a similar level of sophistication in the future.

4.

OPTIONS

Key options include both GaP/GaAs and Si/Ge metamorphic approaches, “monolayer” interfacial strain relaxation approaches, lattice matched approaches using III-V alloys, chalcogenides or strain-balanced, short-period superlattices, “all-silicon” tandems based on manipulating silicon’s bandgap using quantum confinement, new materials such as perovskites; or finally, mechanically stacking exfoliated cells on a silicon wafer.

4.1

Metamorphic approaches

Collaborators at Ohio State University have pioneered the growth of III-V cells on silicon for solar applications [2-4] through metamorphic structures based on graded GaAs_yP_l-y or Ge_ySi_l-y buffer layers (Fig. 3a). Of these approaches, the former is preferred on device grounds, since the GaAs_yP _l-y bandgap is higher than that of Si allowing the Si substrate to be used as an active cell in he stack. The buffer layer can serve as a heteroface window for the Si cell as well as accommodating the tunnelling/defect junction required to connect overlying cells in series with the Si cell.

Figure 3.

(a) Cross-sectional dark-field STEM image of an MOCVD GaAs0.7 P0.3/step graded GaAsyPl-y/Si structure [2]; (b) EQE for CZ Si sub-cells with and without a GaP epitaxial front window. Also shown is the improved infrared response of a FZ Si bottom layer cell developed for this work [2].

Good preliminary results have been obtained in interfacing the underlying Si cell with overlying layers (Fig. 3b). This is the most likely route to demonstration of a near-term efficiency above 30% in a monolithic tandem device on silicon. One disadvantage is the deposition time required for the buffer layer, which may be as thick as or thicker than the overlying cells (Fig. 3a).

4.3

Lattice matched approaches

4.3.1

III-V cells

From Fig. 5a, GaN and AIN have smaller lattice spacing than Si, allowing alloys with phosphides and arsenides to be lattice matched. The properties of the alloy system GaN_xP_l-x-yAs_y has been investigated in some detail [7]. Alloys with bandgaps in the range 1.5-2.0 have been synthesized [7] with operational monolithic cell on Si demonstrated [7]. Performance has been modest due to the less than ideal properties of layers incorporating substantial N. If this issue could be solved, a triple junction stack within this material system would be possible.

Figure 4.

(a) “Lattice matching in a monolayer” – lowest energy configuration for Ge on Si, once more than 12 monolayers deposited (after [5]); (b) “Out-of-sequence” tandem. Photons are shared between the Si and Ge, with rear light-trapping features increasing Ge absorption of photons below the Si bandgap.

Figure 5.

(a) Bandgap versus lattice parameter of a range of semiconductor compounds [6]; (b) Lattice match to kesterite and stannite.

4.3.2

Chalcogenide cells

As apparent from Fig. 5, some chalcogenide compounds are close to being lattice-matched to Si. Of particular interest is CZTS (kesterite), with both it and alloys with stannite (Cu₂FeSnS₄) being in this class (Fig. 5b). The advantage of this material system is that it is likely to attract significant research attention over the coming decade as a non-toxic, earth-abundant alternative to both CdTe and CIGS thin-films. Thin-film tandem cells in this system are also likely to be of increasing interest, given the likely need for any alternative to Si to at least match its energy conversion efficiency. The present work could build on such developments. The first step is to demonstrate CZTS epitaxy on Si, with this being recently achieved at NREL and on sapphire at UNSW.

4.3.3

Strain balanced layers

Another approach to lattice matching is to use strain-balanced layers as has proved effective in III-V cell technology [10]. Strain balancing of GaN_xP_l-x-yAs_y layers has also been explored [11], given the superior properties of the dilute nitrides. More radical combinations may be possible, as suggested by Fig. 5a, if thicknesses are restrained to a few monolayers. Layer by layer deposition approaches may be particularly attractive to allow the required control and may be economically feasible for semitransparent layers with wavelength selective light-trapping features, as developed for “micromorph” cells.

5.

DISCUSSION AND CONCLUSION

The above approaches can be compared using a matrix of key indicators, including probability of successful medium term implementation, likely performance, thermal expansion issues and likely cost.

The III-V approaches offer the ultimate in efficiency from the Si tandem approach and realistic opprotunities for nearterm implementation. The III-V based tandems may be ideal for applications such as concentrator cells where cost constraints can be relaxed.

CZTS-based cells offer the opportunity for lattice matching to Si with potentially low costs. There is considerably more uncertainty, however, regarding the timescale for implementation of high performance devices.

Perovskite-silicon tandem cells look particularly interesting for near-term implementatin at low cost. Issues in this case relate to the poorer stability of the perovskite cells compared to silicon and the high percentages of toxic Pb in the most successful implementations to date.