Introduction
How to dramatically reduce the cost-per-Watt of solar-generated electricity is the primary challenge in present photovoltaic industry. The solution is none but (a) increasing the solar-electric power conversion efficiency and (b) reducing the manufacturing cost. In contrast to conventional Si or GaAs bulk solar cells, low-cost materials and higher conversion efficiency are more favorable, which have been driving the development of current thin film (2nd generation) and multiple-junction/multiple-band (3rd generation) solar cells assisted with concentrating photovoltaic (CPV) systems [1,2].
Semiconductors are the backbone materials used in current solar cells. The band-gap property of semiconductors plays a critical role in photovoltaic effect - the conversion of solar radiation into electrical energy. Semiconductors can only efficiently absorb photons with energy higher than the band gap, where the excess energy of photons is wasted and converted into heat through electron–phonon scattering and subsequent phonon emission. Considering the loss of the absorbed photon energy above the semiconductor bandgap, Shockley and Queisser theoretically analyzed an upper limit to the conversion efficiency of 31% for a single-bandgap solar cell [3]. This is called the maximum thermodynamic efficiency or detailed balance limit. If the hot electrons and holes can convert their excess energy into electrical or chemical free energy instead of heat, then the limited efficiency can be increased to 66% [4]. The 3rd generation photovoltaic cells are dedicated to harvesting more solar energy by reducing or eliminating the energy loss effects as mentioned above. The approaches include the hot carrier extraction, multi-excitation, multi-junctions and intermediate bands (IBs). The hot carrier extraction and multi-excitation are the two fundamental ways converting hot carriers or excitons to useful work to enhance the efficiency of solar cells. The former produces an enhanced photovoltage by extracting hot carriers before they relax via phonon scattering, the latter an enhanced photocurrent with one high energy photon generating more than one electron-hole pairs. These processes are more favorable in quantum confined heterostructures such as quantum dots (QDs) due to quantized energy levels leading to a slow relaxation for hot-carriers [5]. However, the quantum efficiency of these processes is too low to be applicable to practical devices. On the other hand, the design of multi-junction and intermediate band (IB) aims to possess multiple bandgaps covering the whole solar spectrum and thereby harvesting more solar energy. In such “full spectrum” solar cells, the thermalization loss is much less than that of a single junction cell. In the following, we will give a brief review to the development of multifunction solar cells and followed by a detailed description of the progress on IB solar cells based on ZnTe:O dilute alloy, III-V self-organized QDs and colloidal QDs.
The prevailing multi-junction approach in practice employs series-connected materials with different bandgaps to form double-junction and triple-junction tandem schemes, etc. With optimal bandgaps, each junction covers a select range of the solar spectrum. For double junction solar cells, InGaP/GaAs has been extensively investigated and demonstrated with the reliable efficiency up to 25% at 1× Sun, and 30%–32% at 300–500× Suns [6]. Especially, at ultra-high concentration up to 3000× Suns, InGaP/GaAs solar cells are still capable of keeping the efficiency greater than 30% and have no significant degradation [6]. It makes InGaP/GaAs more attractive at ultra-high CPV applications. For triple junction solar cells, InGaP/GaAs/Ge is the representative structure with demonstrated efficiency about 30%–33% at 1× Sun and 40%–42% at 300–500× Suns [7]. However, the band-gap combination of 1.9, 1.4, and 0.65 eV in the current triple-junction structure is not optimal. One of the optimizations is using 1.0 eV bandgap material as the bottom cell. However, the direct growth of device-quality of either lattice-matched InGaAsN or lattice-mismatched In0.37Ga0.63As 1.0 eV as the bottom cell on Ge is challenging. A recent approach called inverted metamorphic multi-junction (IMM) solar cells, with a metamorphic growth of In0.37Ga0.63As cell on the top of the epitaxial heterostructure and then inverted mount of the whole structure on a foreign substrate followed by the removal of GaAs substrate, can deliver the efficiency about 33.8% at 1× Sun [8]. An additional advantage of IMM solar cells is the potential of reusing expensive GaAs substrates and thereby lowering down the cost. The manufacturing of the above GaAs-based tandem solar cells is involved in state-of-the-art metal-organic chemical vapor disposition (MOCVD) epitaxy process and the use of non-cheap substrates; thereby resulting in a high cost. They could not be widely employed in terrestrial photovoltaic systems without CPV. CPV can much reduce the area size of expensive cells substituted by much cheaper optical elements, therefore reduce the overall system cost and also offer other benefits such as increased efficiency [9]. To lower cost, other materials on cheaper substrates are also exploited. InGaN/Si double-junctions cell has been acquired considerable attention due to the widely-tunable bandgap of InGaN and an intrinsic low-resistance ohmic junction between In0.46Ga0.54N and Si allowing carrier tunneling without heavy doping [10]. However, the poor quality of metamorphic epitaxial InGaN on Si makes practical device still elusive. Also, CdTe/Si double-junction cell has been proposed [11]. In addition, such tandem scheme is applied to a:Si-H-based thin film solar cells, with Si:H/a-SiGe:H/nc-Si:H triple junctions, can achieve efficiency up to 15.1% [12].
In addition to the tandem series of different bandgap materials, an alternative approach to pursue “full spectrum” solar cells is to directly modify the band structures by introducing additional electron/hole levels into the forbidden band gap of the single-bandgap host material. With respect to the conduction and valence bands, these additional trap levels as IBs provide increased absorption due to multi-photon processes, as depicted in Fig. 1. In addition to the direct transition from the valence band (VB) to conduction band (CB), the transition via the IB states occurs by absorbing two sub-gap photons in sequence. Therefore, the short-circuit current (Isc) is increased without sacrificing the open circuit voltage (Voc). The concept was initially proposed in 1960 [13], followed by the argument that trap levels prefer to serving as non-radiative recombination centers which reduce the conversion efficiency [3]. Further theoretical analysis predicts the efficiency of an IB cell implemented by a homo-junction is able to surpass the Shockley-Queisser limitation with an ultimate limit of 63.1%, comparable to triple junction cells [14]. This analysis assumes that (a) negligible non-radiative recombination; (b) ideal carrier transportation (all photo-carriers can transport out to terminals before recombination); (c) optimal bandgap energy of the host material (1.9 eV) and position of IB states (0.7 eV from the band edges); (d) non-overlapping absorption (higher energy photons are not involved in lower energy transitions). Additionally, because of the parallel process inside, an IB cell performs more stable than a triple junction cell when the input solar spectrum is distorted by weather. Therefore, an IB cell, potentially achieving a better figure of merit than a multi-junction cell, has become an exciting research topic. Instead of early-proposed deep-level impurity, recent efforts were focused on more promising approaches with the use of dilute semiconductor alloys and quantum-confined nanostructures such as QDs. Dilute semiconductor alloys mainly refer to dilute nitride III-V semiconductors and dilute oxygen II-VI semiconductors. Because of the big electro-negativity difference between nitrogen atoms with other Group V atoms in III-V host materials, such as As in GaAs, deep level states may be formed when nitrogen atoms replace the Group V atoms. So does oxygen in II-VI semiconductors. How to minimize the induced defects in dilute alloys to avoid non-radiative recombination is the primary task in current stage of the developments. The further challenge is how to increase the efficiency of extracting the photo-excited carriers. On the other hand, quantum confined heterostructures such as quantum dot (QD) heterostructures are extensively studied in order to achieve IBs with applications in solar cells. The main challenge left here is not the minimization of defects but more focused on how to efficiently extract the photo-excited carriers without degradation on the open circuit voltage.
Dilute alloys
In dilute semiconductor alloys, the deep-level states can act as localized isoelectronic defects at low concentration [15, 16] or form an energy band when their wave-functions are coupled at high concentration. This behavior is described by the band anti-crossing (BAC) model, as expressed in Eq. (1) [17].where EC(k) is the energy level of the conduction band, EL is the energy level of localized isoelectronic states, and V is the adjusted parameter describing the coupling effect between EC(k) and EL. x is the mole composition in GaAs1-xNx or ZnTe1-xOx.
The band diagram of ZnTe1-xOx with an IB is predicted based on BAC model, as shown in Fig. 2 [18]. It should be noticed that the band diagrams of ZnTe1-xOx and GaAs1-xNx are not optimal for operating IB absorption as the aforementioned. Further band diagram optimization by tuning the band gap may be achieved by forming ternary alloys, such as AlGaAs, GaInP, GaAsP, ZnCdSe, MgCdSe, ZnSeTe, etc. Although the idea of an IB solar cell based on the dilute semiconductor alloys is extremely attractive, the experimental research is rarely reported and only ZnTe1-xOx material system was studied with some details so far.
Formation of IB in ZnTe1-xOx
The methods for synthesizing ZnTe1-xOx alloy or oxygen isoelectronic centers in ZnTe are very diverse, which include crystallizing ZnTe powder at high temperature with adding oxygen [16] or ZnO [19], pulsed-laser deposition (PLD) [20], ion-implantation with pulsed laser melting [18], molecular beam epitaxy (MBE) [20,21], and so on. When oxygen atoms are introduced into ZnTe crystal, only OTe states around 0.4 eV below the CB edge are confirmed to be radiative, and therefore suitable for operating IB conversion [22,23]. The electrical and optical properties of other type oxygen impurities including O interstitials, OZn, have not been fully understood. Nevertheless, as shown in Fig. 3, the absorption coefficiency of oxygen related IBs can be as high as 104 cm-1, and it also depends on the oxygen ambient pressure in growth.
Demonstration of IB operation
IB energy conversion based on ZnTe:O was demonstrated by the use of the device structure shown in Fig. 4(a) [24]. The calculated band diagram is depicted in Fig. 4(b).
The sub-bandgap (1.5-2.2 eV) response as strong as ZnTe bandgap (>2.2 eV) was clearly observed in the ZnTe:O sample as shown in Fig. 5, which was attributed to the intermediate states formed by oxygen doping in ZnTe. The mechanism for generating sub-bandgap photocurrent is illustrated in Fig. 6. An electron is excited the first photon from the VB to IB, then from the IB to CB by the second photon since the monochromic light beams used in this experiment provides multiple photons in the same energy. Once the photon energy is less than the height of the IB, no electron can be delivered from the VB to IB. Therefore, monochromatic response the ZnTe:O diode cuts off the at 1.5 eV, the edge of oxygen states. Theoretically, the total photocurrent including sub-bandgap and bandgap response can be determined from three-level rate equations coupling with current continuous equations in the diode [25-28]. Furthermore, a two-beam excitation experiment was introduced to confirm a much lower energy photon response which transits electron from the IB to CB in ZnTe:O as shown in Fig. 7. The one exciting source was a 1.55 μm (0.8 eV) laser beam only capable to deliver electron from the IB to CB, and the other was a 0.65 μm (1.9 eV) laser beam that can excite electron from the VB to IB directly but not the VB to CB. As shown in Fig. 8 (a), there was no detectable response (zero Voc) when only 1.55 μm laser was turned on, but significant Voc and Isc were observed under 0.65 μm laser excitation due to second-photon absorption described previously. And more importantly as shown in Fig. 8(b), both Voc and Isc monotonically increased with the power of the 1.55 μm laser when the 0.65 μm laser was turned on, indicating that 1.55 μm photons assisted to generate more electrons in the CB only from the IB, which is the most direct evidence of photovoltaic conversion through ZnTe:O IBs.
Finally, comparing the ZnTe:O diode to the controlled ZnTe diode, it showed a 100% enhancement in Isc but 15% drop in Voc and overall 50% improvement in power conversion efficiency as shown in Fig. 9. Regardless of demonstration of IB energy conversion, either Isc, Voc or conversion efficiency for ZnTe and ZnTe:O diodes are less than 1%, still far away from the theoretical value [25] under AM1.5 as shown in the following Table 1. In order to realize a high efficiency of ZnTe:O based IB solar cells, two main issues need to be further addressed. One is the photo-carrier recombination and transportation in ZnTe:O absorber, the other is the diode structure.
Tab.1 Theoretical values for ideal ZnTe and ZnTe:O diodes |
| Isc/(mA·cm-2) | Voc/V | efficiency | |
|---|---|---|---|
| ZnTe | 7.4 | 1.81 | 12% |
| ZnTe:O | 17.1 | 1.59 | 21% |
Carrier recombination and transportation
Time resolved photoluminescence (TRPL) as shown in Fig. 10 revealed that electron lifetime in the IB of ZnTe:O was about 4 µs, about several orders longer than those in type I alignment quantum well or QD structures of direct band semiconductors such as GaAs, InP, ZnTe etc, which are usually less than nanosecond. The long electron lifetime in the IB means a slow recombination rate from the IB to the CB, which is beneficial to increasing electron occupation in IB under a steady-state light excitation, and therefore favors to realize IB conversion. However, the electron lifetime in the CB of ZnTe:O are less than 100 ps, in the same order of a time of electron transporting out from the depletion region (assuming 100 cm2/(V·s) mobility, and 1 µm thickness, and 2 V self-built bias ). In other words, the carrier extraction efficiency is limited by the short carrier lifetime in the CB.
Two strategies can be used to increase the carrier extraction efficiency. One is to decrease the electron transport time dependent on the carrier mobility and the self-built field, which are the intrinsic parameters of semiconductors. The other is to increase the carrier lifetime in the CB of ZnTe:O.
As shown in the Fig. 11, high density of unoccupied IB states results in quick decay from electron from the CB to the IB based on the three level rate equations [29]. Therefore, it is straight-forward to increase electron occupation in the IB in order to increase the lifetime in the CB. Several approaches have been reported to increase the electron occupation in the IB.
First of all, the higher light intensity can boost electron occupation as shown in the Fig. 12. The carrier lifetime at the CB can be as long as 1 ns when the inject photon flux is about 1020 cm-2·s-1, roughly equal to 1000 suns of AM1.5 spectrum. On the contrary, the carrier lifetime in the IB is dropped in one order under 1020 cm-2·s-1 photon flux. Alternatively, the occupation in the IB could be enhanced by the self-built field regarding the carrier recombination and transportation in the depletion region of an ideal diode structure, as shown in Fig. 13 [25]. Once electron-hole pairs are generated in the depletion region with electrons in the IB, holes will be pulled out from the depletion region very quickly. Therefore, the recombination rates of electron from the IB to VB are expected to be much smaller than that without bias field in the case of TRPL measurement. However, net electron accumulation in the depletion region will reform the band diagram, electrical field distribution in the depletion region and width of the depletion region. Finally, a more practical idea of using electrically doped IB has been studied experimentally and theoretically [26,30,31], where the electron in the IB are thermally generated from doped impurities whose energy levels are slightly below the IB edge.
Junction structure
The initial demonstration based on p-ZnTe/n-GaAs introduced high density interface defects, inter-diffusion impurities, and dislocations in the active region of the junction, which reduce the photocurrent and output voltage of the photo-diode. Although it is very difficult to achieve n-ZnTe, ZnTe based p-n homojuction has been reported using Al diffusion in ZnTe, which shows the Voc is as large as 0.9 V [32]. On the other hand, n-ZnO/n-ZnSe is a good candidate for forming a heterojunction for p-ZnTe with minimizing the optical loss and electrical recombination loss, although the large lattice mismatch between ZnSe and ZnTe is still an issue to be addressed carefully [33].
QDs
QDs are semiconducting crystal islands in nanometer dimensions providing strong quantum confinements on the carriers. Figure 14(a) shows an atomic force microscope (AFM) image of InAs QDs. Due to the resulting discrete energy levels and delta-function density of states, QDs promise supreme optical and electrical properties compared to the conventional semiconductor heterostructures. Specially, coherently-strained self-organized QDs, formed on a two-dimensional wetting layer in the Stranski-Krastanow growth mode [34,35] with dislocation free, have been well demonstrated with unparallel performance in many optoelectronic devices such as lasers, light emitting diodes, and photodetectors. For example, self-organized In(Ga)As/GaAs QD lasers exhibit near-ideal device characteristics [36], including ultralow threshold current (Jth<20 A/cm2), nearly temperature invariant operation, high output power (14 W), large modulation bandwidth (f-3dB = 24.5 GHz), near-zero chirp and α-parameter, and tunable emission wavelength in the range of 0.9~1.5 µm. The large strain fields inside the dots and surrounding materials can significantly inhibit the propagating dislocations, leading to low defect densities in device active regions [37]. It promises the direct epitaxy of QD lasers with significantly improved performance compared to conventional III-V quantum well heterostructures on Si [37,38]. For the application in photovoltaics, regarding at least two aspects such as IBs and multi-excitation, QDs can offer the possibilities for enhancing the efficiency of single-junction solar cells>63% [5].
The big advantage of QDs severing as IBs is their easy tunable bandgap. The emission/or absorption wavelengths of QD heterostructures are largely determined by the dot size and composition as well as the bandgap of the surrounding barrier layers. In addition to the increase of indium composition, the larger the size, the longer the wavelength of light absorbed and emitted. For example, the emission wavelength of self-assembled In(Al, Ga)As QDs can be continuously tuned from visible to ~1.5 µm by controlling the dot size and composition and/or using different barrier layers [36,39,40]. Considering the photoluminescence (PL) peak is close to the absorption band edge in direct band-gap semiconductors, the PL characterization can help to understand the optical absorption property. Shown in Fig. 14(b) are the room-temperature PL emission spectra of InAs QDs grown on GaAs substrates, wherein different emission wavelengths are achieved by altering the indium composition or layer thickness of the InGaAs capping layers. Thus, it is much easier to achieve an optimum bandgap that corresponds to the highest possible solar-electric energy conversion. Also, it is expected the mixture of QDs of different sizes or composition are capable of harvesting the maximum proportion of sunlight.
Fig.14 (a) AFM image of uncapped InAs metamorphic quantum dot layer; (b) room-temperature photoluminescence spectra of self-assembled InAs quantum dots incorporating two different InGaAs capping layers ( I—less indium composition with/or thinner thickness, II—more indium composition with/or thicker thickness) |
A heterostructure of multilayer stacks of two-type QDs on silicon substrates, including a top-stack of In0.5Ga0.5As QDs with a GaAs capping layer and a bottom-stack of InAs QDs with an In0.15Ga0.85As capping layer, has been proposed to harvest more solar light. This heterostructure, for the first time, has been implemented in QD lasers on Si substrates in which the bottom-stack of InAs QDs act as an efficient dislocation filter [37]. It could further be implemented onto a solar cell aiming to absorb more of infrared light with the top-stack of In0.5Ga0.5As QDs firstly and then with the bottom stack of InAs QDs. The epitaxy of the structure on Si is started with an initial GaAs buffer layer grown by MOCVD and followed by MBE growth of the rest structure. Figure 15 depicts the PL spectra emitted from both short-wavelength In0.5Ga0.5As QDs and long-wavelength InAs QDs on Si substrates.
Another advantage is that QDs can be fabricated into a variety of different forms such as colloidal QDs [41], which are synthesized in chemical solvent with low cost and also can easily be combined with organic polymers, dyes, or integrated into porous films. Thus, it has capability of integration of these materials on inexpensive substrates such as Si, glass, metal sheets and flexible plastics. Colloidal QDs are tunable over a broad spectral range by varying their size during the synthesis. Figure 16 is the PL and absorption spectral characterization of mixed PbSe colloidal QDs with different sizes. A promising approach employs hybrid organic semiconductor solar cells embedded with PbSe QDs to capture more of the infrared wavelength of sunlight [42].
Though many efforts have been made, so far all investigations of QD solar cells have not demonstrated significantly improved conversion efficiency, compared to bulk homo-junction devices. The fundamental puzzle here is the low extraction efficiency of photo-carriers from quantum-confined islands and their subsequent trapping and recombination in surrounding wet-layers. It looks a nature association with QDs as a high-efficiency heterostructure for carrier injection, but may not in a “reverse” process. Recently, a theoretical study shows an enhanced carrier extraction via introducing a thin high-bandgap barrier and tunneling process [43]. Another practical problem is the absorption volume of QDs is limited by the size and sparse density of QDs. A simple approach of using a multi-stack of QDs would generate dislocation and defects, which act as carrier traps and degrade the efficiency of solar cells. Recent studies demonstrated an improved Voc and Isc with the use of a strain compensation layer to reduce the overall strain accumulation and hence defect generation [44,45].
Conclusion
Overall, the IBSC based on dilute alloys and QDs exhibits strong potential for the high conversion efficiency and low-cost manufacturing capability. The IB conversion of both ZnTe:O dilute alloy and InAs/GaAs-based QDs have been experimentally demonstrated. In the case of ZnTe:O alloy, ZnTe:O cells showed an enhanced conversion efficiency with respect to ZnTe controlled cells. However, the absolute value was less than 1% mainly due to high level structural defects in the p-ZnTe/n-GaAs junction, which may be solved by the use of a p-ZnTe/n-ZnSe junction with a stress-released buffer or ZnTe homo-junction. Also, more fundamental limitation for realizing IB conversion lies in that non-radiative states, possible O interstitials and OZn, may be simultaneously and significantly introduced when achieving radiative OTe. In the case of InAs/GaAs-based QDs, the junction technology is well developed and the direct growth of QD heterostructures on Si substrates has also been demonstrated. However, so far there is no example that the GaAs cell with QDs is better than the GaAs homo-junction bulk cell. In addition to the material quality control in growth, how to efficiently extract the photo-carriers before their recombination, without significant degradation of the open circuit voltage, is the main challenge.