Multi-junction solar cells paving the way for super high-efficiency

The development of high-performance solar cells offers a promising pathway toward achieving high power per unit cost for many applications. Various single-junction solar cells have been developed and efficiencies of 29.1%, 26.7%, 23.4%, 22.1%, and 21.6% (a small area efficiency of 25.2%) have been demonstrated¹ with GaAs, Si, CIGSe, CdTe, and perovskite solar cells, respectively. However, single-junction solar cells may be capable of attaining AM1.5 efficiencies of up to 30%–32%, as shown in Fig. 1.² That is, state-of-the-art of single-junction solar cells are approaching the Shockley–Queisser limit.³ An important strategy to raise the efficiency of solar cells is stacking solar cell materials with different bandgaps to absorb different colors of the solar spectrum. This so-called multi-junction (MJ)^4,5 approach can reduce thermalization loss due to a high-energy photon absorbed by a small-bandgap material and below-bandgap loss due to a low-energy photon of insufficient energy to excite an electron in a high-bandgap material, as shown in Fig. 2.^6, Figure 3 shows the principle of wide photoresponse using MJ solar cells for the case of a triple-junction solar cell.⁷ Solar cells with different bandgaps are stacked one on top of the other so that the solar cell facing the sun has the largest bandgap (in this example, this is the GaInP top solar cell with a bandgap energy Eg of 1.8–1.9 eV). This top solar cell absorbs all the photons at and above its bandgap energy and transmits the less energetic photons to the solar cells below. The next solar cell in the stack (here the InGaAs middle solar cell with an Eg of 1.41 eV) absorbs all the transmitted photons with energies equal to or greater than its bandgap energy and transmits the rest downward in the stack (in this example, to the Ge bottom solar cell with an Eg of 0.67 eV). Of all the so-called third generation solar cell strategies,⁸ only MJ designs have been successful in surpassing the detailed-balance limit of single-junction solar cells. Such successful achievements are thought to be due to longtime R&D since the late 1970s, bandgap engineering including lattice matching, high-quality epitaxial growth, and so forth.

The operating principles of MJ solar cells were suggested by Jackson⁹ as long ago as 1955, and they have been investigated since 1960.¹⁰ This concept was most successfully implemented in III–V compound semiconductor solar cells, since a compound semiconductor has a good range of lattice parameters and bandgaps to choose from. High efficiencies of 32.8%¹ under one-sun and 35.5%¹¹ under concentration with two-junction solar cells, 37.9%¹² under one-sun and 44.4%¹² under concentration with three-junction solar cells, 46.1%¹³ under concentration with four-junction solar cells, 38.8%¹⁴ under one-sun with five-junction solar cells, and 39.2%¹⁵ under one-sun and 47.1%¹⁵ under concentration with six-junction solar cells have been demonstrated, as shown in Fig. 4 and Fig. 6. Figure 4 shows the chronological improvement in conversion efficiency^7,16 of concentrator MJ and one-sun MJ solar cells in comparison with those of crystalline Si, GaAs, CIGS, and perovskite single-junction solar cells.

In the days to come, Si-based tandem solar cells¹⁶ such as III–V/Si,^17,18 II–VI/Si,¹⁹ chalcopyrite/Si,²⁰ CZTS/Si,²¹ and perovskite/Si²² tandem solar cells are expected to play a more important role as high-efficiency, low-cost solar cells move closer to industrial manufacturing. In addition, there are other approaches such as perovskite/perovskite,²³ III–V/CIGSe,²⁴ and perovskite/CIGSe²⁵ MJ solar cells that are still at a lower technology readiness level but may become very attractive candidates for photovoltaic energy conversion in the future.

Here, we discuss the perspectives of MJ solar cells from the viewpoint of efficiency and low-cost potential based on scientific and technological arguments and possible market applications. In addition, this article provides a brief overview of recent developments with respect to III–V MJ solar cells, III–V/Si, II–VI/Si, perovskite/Si tandem solar cells, and some new ideas including so-called 3rd generation concepts.⁸ 

The fundamental processes in photovoltaic power conversion are shown in Fig. 5, where incident sunlight of energy above the semiconductor bandgap can be absorbed (1) and excess energy dissipated as a thermalization loss (3); photons below the bandgap energy can pass through the solar cell unabsorbed (2). Excess radiative recombination proceeds due to the presence of photogenerated carriers (4). At forward operating voltages, the free energy of the carriers is determined by the quasi-Fermi level separation that defines the solar cell voltage at the electrical contacts to the solar cell, V = μe–μh.

The breakdown between power generated by the solar cell and these losses is illustrated in Fig. 2.⁶ For a single-junction solar cell, the two largest losses are the thermalization and below-Eg losses, both of which are significantly mitigated with the addition of semiconductor junctions with different bandgap energies in an MJ device. This is because a larger portion of the solar spectrum is then absorbed close to the bandgap of one of the semiconductors and, therefore, experiences less thermalization of carriers. All other fundamental losses (note that resistance losses are not regarded as fundamental losses, though they are unavoidable in practical devices and, in fact, benefit from lower currents that are typically achieved with more junctions) increase with an increased number of semiconductor junctions and are discussed in detail elsewhere.³ We note that all three remaining losses fundamentally depend on solar cell temperature and can, therefore, be reduced by operating the solar cell at a lower temperature.

The so-called “Boltzmann loss” is an entropic loss associated with the increase in the occupancy of optical modes on re-emission of light that results in a voltage loss. Practically, it can be recovered in two ways, conventionally by increasing the solar concentration on the solar cell, or equivalently, by restricting the radiative emission from the cell. Dividing up the solar spectrum between an increasing number of junctions results in emission at multiple wavelengths and hence a larger Boltzmann loss. The loss can be mitigated by paying attention to the geometrical optical arrangement, for example, through solar concentration or an angular restriction of radiative recombination.

The Carnot loss arises from establishing carriers at finite temperature in a band and, hence, rises as further junctions establish additional bands occupied by photogenerated carriers. The loss cannot be recovered except in very unusual circumstances where the PV cell is operated at low temperatures.

The emission loss is unavoidable in a conventional solar cell owing to the reciprocity²⁶ between absorption and emission encapsulated by Kirchoff's law of radiation. The detailed-balance limit³ considers only this unavoidable radiative recombination loss through thermodynamic arguments. The external radiative efficiency (ERE) describes how closely a sub-cell comes to this thermodynamic limit as all other non-radiative recombinations are considered potentially avoidable. In an optimally configured MJ solar cell, the emissive loss is small, but in a series-connected tandem solar cell where current flow is constrained by one sub-cell, there can be a significant transfer of energy down the tandem absorber stack. This is known as radiative coupling and discussed in more detail below. An extreme case of radiative coupling arises if the reciprocity between absorption and emission is lifted, potentially using magneto-optical materials,²⁷ which allows the efficiency of an infinite tandem stack (asymptotes to 86.8%) to be raised to the Landsberg efficiency of 93.5%. At the Landsberg limit, each of the component junctions operates arbitrarily close to Voc and electrical power delivered from the solar cell infinitesimally slowly.²⁸ 

To achieve efficient operation, the photogeneration rates in a series-connected solar cell should be closely matched. While the choice of the bandgap and, hence, absorption threshold for the component junctions plays a primary role, the sub-cell photogeneration can be optimized by adjusting the thickness of the junctions such that an overperforming sub-cell can allow some light to pass unabsorbed into an underlying junction.²⁹ This approach works well for static solar spectra (such as AM0), but for the terrestrial spectrum, the spectral irradiance varies throughout the day and between seasons with noticeable effects on system performance.³⁰ If the semiconductor material is radiatively efficient, sub-cells with excessive photogeneration will radiate the excess into lower lying junctions with a small fraction of this escaping from the top of the solar cell. Radiative dominated behavior has been observed in III–V solar cells³¹ and some perovskite materials,³² even appearing as a measurement artifact in MJ devices.^33,34 Since radiative coupling transfers energy down the MJ stack, the effect can mitigate the effects of spectral mismatch under blue-rich spectral conditions³⁵ as well as offering some freedom in tolerable absorber bandgap configuration,³⁶ in particular enabling efficient operation of higher-gap top solar cells in a silicon-based tandem.³⁷ 

III–V semiconductor materials have many advantages for high-efficiency solar cells in general and the MJ solar cell in particular. III–V semiconductors consist of elements from the group III (Al, Ga, and In) and V (N, P, As, and Sb) columns in the periodic table arranged in a zinc- blend (or wurtzite) crystal structure. The highest single-junction efficiency has consistently remained a GaAs solar cell due to its bandgap matching the solar spectrum and its high ERE, as shown in Fig. 1. III–V alloys (hereafter III–Vs) cover a wide-bandgap range from 2.4 eV down to almost 0.0 eV, as shown in Fig. 6, with III–N alloys covering much higher bandgaps³⁸ (not shown).

Many of the III–Vs can be grown as single-crystal layers on single-crystal substrates (e.g., GaAs, InP, Ge, etc.) using liquid phase epitaxy (LPE), molecular beam epitaxy (MBE), organometallic vapor phase epitaxy (OMVPE), and hydride vapor phase epitaxy (HVPE) techniques. Many III–Vs can be doped both n-type and p-type over a wide density range and are direct gap for nearly complete thin-film absorption. They can have high mobilities, relatively long lifetimes, and low interface recombination when higher bandgap alloys are used for heterojunction passivation (i.e., window and back-surface field layers) that result in diffusion lengths longer³⁹ than the required single-pass absorption thicknesses. Thus, light-trapping techniques often used in silicon photovoltaics have not been much required for III–V materials. The primary disadvantage of III–V solar cells is their sensitivity to defects that act as deep non-radiative recombination sites, such as dislocations, impurities, and phase boundaries, but they also have a relatively low density if there are intrinsic native defects. While III–Vs can be grown with a high degree of crystal perfection that avoids the problems of defects, it can presently be done only at a relatively high cost that results in the other primary disadvantage of III–V solar cells. Due to the high cost of III–V solar cell fabrication, they have been used in applications that leverage the efficiency advantage on the balance of system costs, such as space, concentrator, and other area or weight constrained applications. Concentrator applications have also been intimately tied to III–V MJ development because of the logarithmic increase in voltage with concentration for each junction.

From the beginning of MJ solar cell research, III–V alloys have been the material of choice due to the advantages listed above. In addition, the wide range of bandgaps appropriate for solar collection at the same lattice constant, as shown in Fig. 6, has provided many avenues for creating defect-free monolithic MJ configurations. Several of the III–V MJ strategies with their achieved efficiencies are summarized in Fig. 6 and mentioned briefly in the following paragraph.

Initially, multi-terminal structures were considered,^40,41 but since low-resistance tunnel junction interconnects consisting of heavily doped wide-bandgap materials have been demonstrated in III–V materials,^42,43 series-connected MJ solar cells have dominated. Such two-terminal devices facilitate the integration of the solar cells into modules and the connection to inverters. Target bandgap combinations for series-connected MJ solar cells have been calculated for specific spectra assuming detailed-balance^44,45 and more realistic ERE limitations.⁴⁶ Recently, wafer bonding,¹³ transparent conductive oxide layer insertion,⁴⁷ and mechanical stacking¹⁷ for MJ solar cell formation have been demonstrated. The first III–V MJ strategy considered was the easily lattice-matched AlGaAs/GaAs structure,^48–50 but difficulties with defects associated with oxygen incorporation into AlGaAs restricted progress.⁵¹ The surprising stability⁵² and quality of epitaxially grown GaInP lattice-matched to GaAs finally resulted in the GaInP/GaAs solar cell with an efficiency of 29.5%⁵³ to exceed the theoretical potential of any single-junction solar cell. Spontaneous ordering in GaInP material also resulted in adjustable bandgaps for the same alloy composition.⁵⁴ The use of Ge substrates was primarily introduced to GaInP/GaAs solar cells for the mechanical and cost advantages over GaAs substrates, but the serendipitous introduction of a diffuse Ge junction also resulted in a slight voltage increase without much added cost or complexity. These GaInP/GaAs/Ge three-junction (3J) solar cells have remained the standard for space and concentrator applications to this day. Improvements on this 3J solar cell revolved around replacing the low-Eg Ge junction with a higher-Eg III–V junction. Dilute nitride GaInNAs(Sb) that is lattice-matched to GaAs was a very promising 1.0 eV candidate⁵⁵ but sufficient quality was not obtained over many years with industry standard OMVPE growth. Using (arguably) more expensive MBE growth, excellent concentrator 3J solar cells with a dilute nitride bottom junction⁵⁶ have been demonstrated.⁵⁷ While high densities of threading dislocations (TDD) are detrimental to III–V quality,⁵⁸ metamorphic growth (which allows the mismatched strain to be relaxed slowly in compositionally step graded buffer layers) has allowed high-quality lattice-mismatched InGaAs junctions to be grown with TDD∼1 × 10⁶ cm⁻² on GaAs substrates. The inverted metamorphic multi-junction (IMM) strategy⁵⁹ that grows the low-bandgap InGaAs junctions last has been demonstrated with three,^60,61 four,^62,63 and six junction¹⁵ III–V solar cells with extremely high efficiencies. A natural consequence of the IMM strategy is the removal of the substrate, which has advantages for cost with potential substrate reuse, light-weight, and flexible solar cells,^64,65 and back-surface reflectors for photon recycling.^31,66 The dislocations in mismatched InGaAs could alternatively be isolated from the high-quality top junctions by bifacial growth⁶⁷ or through the use of strain-balanced GaAsP/InGaAs bi-layers.⁶⁸ The upright, metamorphic growth of lattice-mismatched top junctions on an active Ge junction (UMM) has also been demonstrated with high efficiency.^69,70 Interestingly, very different 3J bandgap combinations used in the 3J IMM⁶⁰ and 3J UMM⁶⁹ resulted in very similar efficiencies as a result of the absorption gaps in the terrestrial spectra.⁴⁶ More recently also, five-junction UMM solar cells were developed and used in concentrator photovoltaics.⁷¹ Even higher quality low-bandgap junctions have been obtained by separately growing lattice-matched junctions on InP and wafer bonding with junctions grown on GaAs substrates for a 4J concentrator ¹³ and 5J one-sun solar cell.¹⁴ Space solar cells with AM0 efficiencies >32% are available as 4J UMM and 5J IMM structures.

Ga_1-xIn_xN alloys (that have been successful for light emitting devices using low-In content) have also been suggested for MJ solar cell materials because the alloy theoretically covers the full bandgap range.³⁸ Some high-bandgap GaInN solar cells have been demonstrated,⁷² but the fabrication of the high-In GaInN alloys to capture the infrared portion of the solar spectrum has remained challenging. High-In GaInN alloys suffer from polarization charge,⁷³ contact inversion layers,⁷⁴ phase decomposition,⁷⁵ and large lattice mismatch.⁷⁶ 

Silicon is a material that combines multiple benefits. It is earth abundant, almost 30% of the Earth's crust is formed from silicon. It can be purified to extremely high levels (typically less than 0.001% impurities in solar cell material) and grown into mono- or multi-crystalline ingots, which are then diced and processed to solar cell devices absorbing sunlight between 300 and 1200 nm. The bandgap is close to ideal for a single-junction solar cell and has resulted in hero devices with up to 26.7% conversion efficiency.⁷⁷ Mechanical strength and stability are further advantages and the cost of silicon solar cells has come down 82% (solar costs have fallen 82% since 2010, https://www.pv-magazine.com/2020/06/03/solar-costs-have-fallen−82-since-2010/) just between 2010 and 2020 mainly due to increased mass manufacturing and economies of scale. Fully processed devices are sold at approximately 0.12 US$/W (prices according to http://pvinsights.com/(27.4.2021)) or 27 US$/m² at an average solar-electric conversion efficiency of 22.5% AM1.5g. This is less than the price for many building materials like floor tiles. One could argue that the perfect solar cell material has been already found, but some limitations may still need to be overcome. Being an indirect semiconductor, silicon requires a certain thickness (typically 150 μm) to absorb sunlight, its manufacturing processes are energy intensive, and the conversion efficiency of silicon single-junction solar cells is fundamentally limited to 29.5% by Auger recombination.^78,79 Auger recombination describes the three carrier energy transfer from a photogenerated carrier to an electron in the conduction band and, therefore, becomes even more important at high concentrations. This fundamental intrinsic recombination process determines the charge carrier lifetimes of ultra-pure silicon, different from direct semiconductors like GaAs or GaInP, which are limited by radiative recombination.

The fundamental efficiency limit of silicon single-junction solar cells can be overcome by MJ devices as described above and such solar cells may still benefit from using silicon as the bottom junction and substrate material. This is attractive because most thin-film absorbers need a support as they are too thin to be self-sustained. Also, silicon with a 1.1 eV bandgap is close to the optimum for dual-junction and triple-junction devices. The ideal bandgap energy⁸⁰ for one additional absorber above Si is 1.7 and 2.0/1.5 eV in the case of two absorbers. The exact bandgaps depend on several factors such as transparency of the upper layers and the long-wavelength response of the silicon bottom solar cell. In fact, this has turned out to be one of the challenges in manufacturing tandem solar cells on silicon. The light-trapping features have to move from the front to the rear of the wafer to allow the deposition of planar thin-film absorbers at the front.^81–83 This can be done by implementing a pyramid texture,⁸⁴ spheres,⁸⁵ or nanostructure gratings⁸⁶ on the back of the silicon wafer to increase the light path through the silicon and, therefore, enhancing absorption close to the indirect bandgap.

The most successful examples for Si-based tandem solar cells in terms of conversion efficiency are combinations of III–V compounds with Si (see example in Fig. 7 right). GaAsP/Si tandem solar cells have reached AM1.5g conversion efficiencies of 23.4% (monolithic, two-terminals),^87,88 GaAs/Si up to 32.8% (four-terminal),¹⁷ and GaInP/Ga(In)As(P)/Si up to 35.9% (two-terminal^89,90 and four-terminal).¹⁷ Some groups are involved in growing the III–V layers directly on silicon, which is a challenge due to the large difference in the lattice constant of 3%–4% and thermal expansion coefficient. Other groups have used wafer bonding or gluing to make the connection, followed by a removal of the growth substrate. Of course, the latter can be economically attractive only if the growth substrate is removed with high yield and reused for the further growth of III–V layers.⁹¹ Direct growth of the III–V absorbers onto silicon using methods like metal-organic vapor phase epitaxy is challenging in terms of achieving low enough defect densities for highest efficiency devices, but continuous progress is being made in the field of metamorphic III–V growth on silicon.^92,93 With further research, this problem may be solved in the near future, making III–V direct growth the method of choice for realizing tandem solar cells on silicon that combine high performance and reliability and that can be manufactured at competitive costs.

The ideal attribute for a MJ sub-cell material is one whose absorption threshold can be tuned over the solar spectrum, is stable, non-toxic, and efficient, and can be integrated into a tandem stack at low cost. The integration of sub-cells poses a particularly awkward challenge, since sub-cells that can perform well in isolation can become impaired when integrated with additional sub-cells, either via impurity diffusion or via an excessive thermal budget. Mechanically stacking separate sub-cells in a multi-terminal device is one means by which incompatibilities can be overcome and serves as a useful proof of concept. For brevity, we survey here only two-terminal tandem solar cells fabricated from at least one novel material.

Metal halide perovskite materials are strong, direct-gap semiconductors with optical absorption, typically extinguishing sunlight in a layer of 200–400 nm thickness. Their absorption threshold can be tuned over a wide wavelength range owing to the wide range of alloy combinations with a ABX3 structure, where A is an organic amine cation, B is the metal cation, and X is the halide anion. Typical choices for A are methylammonium iodide or formamidinium iodide Cs, B is commonly Pb and/or Sn, and X is a halide, typically I and/or Br, resulting in compounds such as (FA_0.75Cs_0.25)Pb(I_0.8Br_0.2)₃ a popular material for the top solar cell of a tandem. These materials offer a tantalizing and unprecedented range of photovoltaic solar cell absorber material combinations that can be prepared both via solution processing or via vapor deposition.⁹⁴ Several permutations of perovskite tandem have been attempted⁹⁵ and, in particular, the material system offers a range of opportunities for achieving wide-gap absorbers that are well suited for many tandem solar cell applications.⁹⁶ 

The versatility of the perovskite material has made it a popular choice for a silicon tandem architecture as a top solar cell with a potentially facile preparation, as illustrated in Fig. 7 (left). In this popular and fast-moving field, the aim is to establish a stable, wide-gap perovskite material^96,97 that is compatible with a suitable silicon bottom solar cell.⁹⁸ One of the outstanding challenges is to improve the voltage of the wide-gap perovskite top cell.⁹⁵ 

A 29.2% perovskite/silicon tandem solar cell was achieved by spin-coating a 1.68 eV perovskite material onto a n-type heterojunction silicon solar cell with a textured rear only. Interconnection between the silicon and perovskite material was achieved using a transparent, conducting metal-oxide ITO layer rather than a tunnel junction.⁴⁷ The device was stable under testing, retaining 95% of its initial efficiency after 300 h of operation. An announcement of a 29.5% tandem solar cell was also recently made, but no technical details are available at the time of writing (https://www.pv-tech.org/news/oxford-pv-pushes-tandem-shj-perovskite-cell-conversion-efficiency-to-record-29.52). A double textured perovskite/ n-type heterojunction silicon tandem solar cell achieved a 25.2% power conversion efficiency where perovskite precursors were co-evaporated to form a conformal film over the textured silicon surface and interconnected using a nanocrystalline silicon tunnel junction.⁹⁹ A similar result, 25.7%, has also been achieved using a solution processed perovskite absorber and metal-oxide interconnection layer.¹⁰⁰ 

Copper Indium Gallium Selenide (CIGS) solar cells can also provide a convenient and commercially mature low-gap solar cell with strong absorption that, with certain alloy fractions, delivers lower energy band-edge than silicon. The thin absorber enables thin, flexible solar cells to be made¹⁰¹ and wide-gap perovskite materials offer an opportunity to augment the efficiency in a tandem configuration. Generally speaking, the film roughness of CIGS has complicated the fabrication of efficient tandem devices. The first reported perovskite/CIGS tandem solar cell used a thick PEDOT:PSS layer, achieving an efficiency of 10.9%,¹⁰² and later, the CIGS was polished to yield a smooth surface and a much higher efficiency of 22.4%.¹⁰³ More recently, self-assembled monolayers have been shown to form an effective interfacial layer between the perovskite and the rough CIGS material, resulting in a tandem efficiency of 24.2%.^104,105

Absorption thresholds as low as 1.2 eV can be achieved using mixed Pb–Sn perovskite materials¹⁰⁶ offering the opportunity for a perovskite/perovskite tandem solar cell. A 24.8% 1.77/1.22 eV perovskite tandem solar cell has been achieved by paying particular attention to Sn oxidation and a low-optical loss tunnel junction that employed an ALD deposited SnO2 layer interlayer between solar cells.¹⁰⁷ 

The principal challenge for fabricating organic solar cells has been to find molecular absorber materials that operate efficiently in the infrared wavelength range.¹⁰⁸ A 1.72 eV PBDB-T:F-M/1.26 eV PTB7-Th:O6T-4F:PCBM device achieved an efficiency of 17.3%.¹⁰⁹ In comparison with the perovskite/perovskite tandem solar cell above, the principal loss in this organic tandem device is the low solar cell voltages obtained for each sub-cell, in addition to a marginally impaired External Quantum Efficiency (EQE) and solar cell fill factor.

Fully inorganic thin-film tandems can be made using chalcopyrite materials, and the well-established CIGS solar cell material is one alloy combination from a large array of the penternary (In_1-xGa_x)(S_ySe_1-y)₂ material system that can span 1–2.43 eV.¹¹⁰ A mechanical stack composed of a 1.48 eV CdTe/0.95 eV CuInS₂ double junction achieved an efficiency of 15.3%.¹¹¹ A 1.68 eV CuGaSe₂/1.1eV CuInGaSe₂ mechanically stacked device achieved an efficiency of 8.5%¹¹² while a 1.89 eV GaInP/1.42 eV GaAs/1.20 eV CIGSe mechanically stacked tandem has achieved 24.2%.¹¹³ Combining CIGS films into a monolithic tandem structure has proven difficult, owing to the complexity of forming the second junction without impairing the performance of the first. Alloying with Ag has provided a new dimension to tackle this problem, since Ag alloys not only have marginally higher bandgaps, their lower melting point temperature also helps reduce the thermal budget for forming the tandem solar cell and reduces compositional disorder.¹¹⁴ Monolithic tandem efficiencies remain low (∼3%)¹¹⁵ but the potential to exceed 25% with this approach exists if the difficulties associated with sub-cell integration can be overcome.¹¹⁶ Monolithic chalcopyrite tandem devices on silicon have been attempted, and a 1.65 eV Cu₂ZnSnS₄ (CZTS)/1.1 eV Si tandem achieved an efficiency of only 3.5%, likely limited by incomplete sulfurization and inadvertent silicon solar cell degradation.²¹ 

Higher efficiencies have been obtained for a 1.8 eV CdZnTe/1.1 eV Si tandem device, which achieved an efficiency of 17%.¹⁹ 

Antimony selenosulfide Sb₂(S,Se)₃ forms 1D ribbons¹¹⁷ and by varying the Se/S atomic ratio, offers an adjustable absorption threshold from 1.7 to 1.1 eV. To date, a 10% efficient single-junction solar cell has been demonstrated¹¹⁸ and a proof-of-concept 1.74 eV Sb₂S/1.22 eV SbSe achieved an efficiency of 7.9%.¹¹⁹ 

Organic absorber materials are well suited for absorbing visible wavelengths and can, therefore, form the high bandgap junction in a hybrid organic-silicon device. A dye sensitized solar cell was partnered with a silicon solar cell to form a 1.8 eV dye/1.1 eV Si mechanical stack tandem cell with an efficiency of 14.7%.¹²⁰ The convention for the interconnection of a tandem solar cell is a series connected stack, but this is only one means by which multiple absorbers can be arranged, and several other permutations are possible in a combination of series and/or parallel connection.¹²¹ Specifically, the combination of a wide-gap solar cell in parallel with two lower gap solar cells has the merit of lower sensitivity to variation in the incident solar spectrum¹²² and this has been demonstrated as a so-called “voltage matched” tandem whereby a pentacene layer absorbs photons at energies above 1.8 eV that undergo singlet fission to produce two electron hole pairs.¹²³ A fully parallel singlet fission device has also been demonstrated¹²⁴ in addition to a conventional series-connected tandem.¹²⁵ 

All the results reported in Sec. III.C are derived from small area cells, cells that are much smaller than 1 cm². For all these approaches to achieve their practical potential, high efficiency will need to be maintained over large areas. III–V multi-junction solar cells are manufactured on 6-in. wafers and subsequently interconnected in series to form a module. The promise of thin-film tandem cells to which all but the silicon-based tandems aspire, is to expand the substrate size significantly, ideally coating an entire sheet of module glass. One of the few tandem technologies that achieved this on a manufacturable scale was the micromorph tandem from Oerlikon, where an amorphous silicon a-Si/microcrystalline silicon μc-Si tandem configuration was manufactured over an area of 1.4 m² delivering an i initial 11% module efficiency. The micromorph technology was rendered obsolete when the costs of the more efficient c-Si modules dropped significantly below that of the micromorph tandem.

Other MJ solar cells composed of II–VI, chalcopyrite, kesterite compound, and perovskite solar cells are thought to have similar potential as III–V compound MJ solar cells. In order to realize high-efficiency MJ solar cells using these materials, reducing non-radiative recombination and resistance losses by learning from the progress achieved in III–V compound MJ solar cells is necessary.

The allowable cost per unit area of solar cell modules largely depends on module efficiency.^137,138 For example, a 30%-efficient solar cell costing 3.5 times as much as a 15%-efficient solar cell of the same area will yield equivalent overall photovoltaic system costs¹³⁷ due to the balance of system costs. Therefore, high-efficiency solar cells will have a substantial economic advantage over low-efficiency solar cells, as the cost of fabricating the former is low enough. Additionally, efficiency improves the environmental impact of photovoltaic modules as less materials are needed to produce them. For space applications, high-efficiency solar cells have significant payload advantages. Although the III–V MJ solar cells have demonstrated an extremely high conversion efficiency with up to 39.2%,¹⁵ further cost reduction is still necessary to access terrestrial photovoltaic markets.

Figure 10 shows a comparison of expected module costs as a function of module production volume for III–V tandem cells with/without high-speed deposition, III–V/Si tandem devices, and concentrators reported by the authors¹³⁸ and cost analytical results for rapid deposition (HVPE; Hydride Vapor Phase Epitaxy)¹³⁹ and Si tandem¹⁷ reported by NREL. Therefore, ways for module cost reduction are reduction in film thickness, a high growth rate of the III–V layers, reuse of substrates, concentration of light, use of Si as substrate material and bottom cells, and an increase in module production volume, as shown in Fig. 10. The results suggest that there are ways to realize costs of less than $1/W for III–V compound MJ solar cell modules by scaling up the production volume to 100 MW/year with a high-speed growth method or with Si-based tandem solar cells. Many of these technologies are current fields of research.

Cost analysis of perovskite and perovskite/Si tandem solar cell modules has been reported.^140,141 About $0.5/W, comparable with crystalline Si and CdTe solar cell modules, was estimated as a manufacturing cost of perovskite and perovskite/Si tandem solar cell modules. One highly uncertain aspect of the module cost for tandem solar cell modules is the level of encapsulation that will be required to maintain tolerable module performance with a module efficiency of 24% over the working lifetime, 30 years.¹⁴⁰ For a silicon tandem, the underlying silicon cell degradation is generally low (less than 0.5%/year) with standard encapsulation methods, but it is not known what level of encapsulation will be required to deliver long-term efficient operation from perovskite on silicon tandem solar cells. What has been established is that the degradation rate of the top cell should be below 0.9%/year for a silicon tandem to remain financially viable.¹⁴² 

Of all the so-called third generation solar cell strategies,⁸ only MJ designs have been successful to surpass the detailed-balance limit of single-junction solar cells. Table II shows potential efficiencies of 3rd generation solar cells.^8,143–145 Demonstrated efficiencies of III–V MJ solar cells at one-sun are nearing levels of 40% and those under concentration are approaching levels of 50%. Most of the underlying physics of these cells have been understood, but sophisticated engineering and high-quality materials are an essential prerequisite for achieving higher levels of efficiency. While low-cost solar cell materials are desirable for tandem solar cells, only high-voltage junctions, as quantified by the ERE,^26,146 with well-chosen bandgaps matched to the application spectra will be helpful for surpassing the efficiency of single-junction silicon. Quantification of spectral efficiency¹⁴⁷ is a convenient metric to judge how to choose tandem partners, and more comprehensive multi-junction models are also available.¹⁴⁸ The challenge for low-cost tandem materials is to bring the best together in a way that preserves the high-quality junctions. This has already been achieved in high-efficiency III–V multi-junction devices, but here, lower processing costs are needed without compromising on the required quality for flat-plate areas. Alternatively, renewed development of robust and economical terrestrial concentrator systems could result in high demand for the most efficient multi-junction solar cells possible. The continued development of multi-junction solar strategies through multiple pathways and a sufficiently large market is likely to bring the technology closer to the economics of single-junction silicon and to provide clean, economical, and efficient energy, especially for area constrained applications.

It can be expected that silicon-based tandem solar cells will receive further growing attention by the photovoltaic industry as efficiencies for single-junction solar cells reach a plateau and many of the opportunities for cost reduction have already been implemented. Innovation will most likely come from more efficient devices, and it has already been shown that Si-based tandem solar cells can achieve nearly 36% conversion efficiency in a two- or four-terminal configuration. This proves that Si-based tandem cells can get close to the best triple-junction solar cells ever reported in the literature and radiative efficiency limits are even as high as 45.2%, 49.6%, and 52.2% for two-junction, three-junction, and four-junction cells,⁸⁰ leaving sufficient room for further development. Using combinations of silicon and III–V materials allows the PV industry to keep many established processes that have already been scaled to large volumes. But at the same time new processes must be implemented and scientists will continue to have different opinions on which technologies are favorable for silicon-based tandem solar cells. The market will accept all those solutions that provide sufficiently high efficiencies combined with economically attractive production processes. We believe that III–V/silicon tandem solar cells will have a significant efficiency advantage compared with conventional silicon single-junction devices, and although they will be more expensive when they enter the market, their costs will reduce over a period of time. The exact efficiency number for market entry may be debatable, but probably it is on the order of 30% (AM1.5g) or more. The reduction in costs will depend on the market size, as indicated in Fig. 10, but for this to materialize, entry markets must be found. An example is electric cars where the high performance of the solar cells directly translates into longer driving distances before re-charging of the battery. Such conveniences often convince companies and customers to pay a premium price.

The simplest solution to realize a III–V/silicon tandem solar cell is a two-terminal device where the GaAsP top cell is grown directly onto a silicon bottom junction. The silicon junction may be formed by diffusion or implantation of P, GaP serves as a front surface field, and the rear may be formed by a combination of a SiO_x passivation with a nanostructured grating for light diffraction. Aluminum can be sputtered and point contacts formed by laser firing. Such a device would fulfill the requirements of manufacturability and low cost, but their performance is today still falling behind a good silicon single- junction solar cell. This may certainly change in the next coming years as defects in the III–V epitaxial layers are better controlled and issues with the GaP passivation to silicon solved. But it may also be necessary to develop more complex device architectures implementing III–V layer transfer from a GaAs substrate, tunnel oxide passivation for silicon, four-terminal architectures, or additional junctions. Some of these approaches have already helped achieve up to 36% efficiency, but they suffer from disadvantages in terms of manufacturability and cost. Finding a path toward realizing a III–V/silicon tandem product will remain the subject of continued discussion in the coming years. But we can be confident that once performance and cost are shown to be economically attractive, nothing will stop the large-scale growth of this technology. The reliability of this technology should be comparable to silicon solar cells and presently, there seems to be no restriction to scale manufacturing to the GW level. Also, presently, there seems to be no restriction to scale manufacturing at the GW level.

In order to realize a society fully based on renewable energy, solar cells with the highest efficiency are attractive because they reduce the required system area and the need for using materials. As single-junction solar cells are limited to 30%–32% conversion efficiency under one-sun, MJ or tandem solar cells are expected to contribute to higher performances. The concept of MJ solar cells was first and most successfully implemented using III–V compound semiconductors and such products have already become the standard technology in space. There is a need to further improve their conversion efficiency of III–V MJ solar cells and reduce their cost to achieve widespread terrestrial deployment. At the same time, perovskite materials have emerged as an alternative solution to form MJ devices, but reliability, module integration, and large volume manufacturing are still subjects of current research and development.

In this paper, we provide perspectives for MJ solar cells from the viewpoints of efficiency and low-cost potential based on scientific and technological arguments and possible market applications. Two, three, four, five, and six-junction solar cells have potential efficiencies of 36.6%, 44.0%, 48.8%, 50.4%, and 51.4% under one-sun, respectively. For realizing higher efficiency MJ solar cells, we highlighted the importance of improving the external radiative efficiency of solar cell materials, or in other words, improving material quality and decreasing defect density in the bulk and at interfaces. Further decreasing resistance losses and applying light management for better absorption or photon recycling are important objectives to accomplish. There is a wide range of technological options under development that will lead to further efficiency improvement; most of these options revolve around target III–V MJ solar cells, III–V/Si, II–VI/Si, and perovskite/Si tandem solar cells. The potential for <1 $/W MJ solar cell modules exists for III–V based devices if new technologies such as high-speed deposition, Si-based tandem solar cells, or the use of concentration are employed with high efficiency and manufacturability. Once the technology is exposed to some terrestrial markets, cost reduction will happen, driven by an increase in production volumes. The history of the silicon photovoltaic industry bears testimony to this. So, the main question now is how to make the technology enter the terrestrial markets with volumes on the order of several hundred MW/year. Besides the III–Vs, other materials may have advantages in terms of production cost and they may enter the field of multi-junction technology quickly once the materials show high external radiative efficiency and reliability. II–VI/Si, chalcopyrite/Si , and especially perovskite/Si tandem solar cells, are being developed rapidly and are expected to play an important role as high-efficiency and low-cost solar cells in the future. As ideal bandgap combinations with the highest efficiency, MJ solar cells are often found in lattice-mismatched systems, and effecting efficiency improvements by making reductions in bulk recombination based on a further understanding of non-radiative recombination is necessary. Reduction in surface and interface recombination, efficient optical coupling and low loss electrical interconnection of sub-cells, and effective photon recycling of bottom solar cells are also key elements for high-efficiency MJ solar cells. At this point, nobody can predict which concept will be the most successful, but we firmly believe that at least one multi-junction solar cell technology with efficiency beyond the limits of silicon will emerge as a major player in the photovoltaic market. The only questions at this point of time are which materials will take the lead in this direction and when will this happen.

M.Y., F.D., J.G. and N.J. wish to express thanks to the NEDO, EC, DOE and ARENA, respectively for their support.

The data that support the findings of this study are available from the corresponding author upon reasonable request.