1. Introduction
In solar cells, heterojunction and sandwich architectures are two kinds of promising structure configurations that have been widely applied for designing efficient devices [1-3]. Specifically, the heterojunction structure is fabricated by stacking up two different semiconductor materials together with different doping to control the conduction type, i.e. the n-type and p-type. Therefore, the typical heterojunction device demonstrates the p-n characteristic, and the built-in electric field formed at the two materials' interface contact is beneficial for charge separation. As for sandwich structure, it is usually composed of multiple solid material layers such as light absorption layer, electron transporting material (ETM) and hole transporting material (HTM), or by configuration of combination of solid layer and liquid electrolyte, which is considered effective for charge transport [4-6].
Although these two solar cell structures possess their own advantages for good performance, there still remain challenges in practical working. As a successful example of heterojunction architecture, copper indium gallium selenide(CIGS)-based solar cells offer inherent advantages of efficient carrier separation and collection, and even yield comparable efficiency to Si wafer-based cells [7]. Nevertheless, the interface of the CIGS heterojunction device could also bring about several problems. First, lots of defect states will be introduced into the heterojunction interface, which shall arouse great trap recombination. Second, since there exist band offsets at the CIGS/CdS interface, barrier spikes could possibly form at both the conduction band and valence band site, which will greatly reduce the collection efficiency of the free carriers and lead to relatively low open voltage [8, 9]. As a typical representative of the sandwich architecture, perovskite solar cells, with mixed organic-inorganic halide perovskites (MAPbX3, MA = CH3NH3, X = I, Cl, Br) as absorbers, inorganic metal oxides (TiO2, ZnO, Al2O3, etc.) as the ETMs and organic molecules (Spiro-OMeTAD, etc.) as the HTMs, also offer tempting prospects on energy conversion efficiency from less than 5% to over 20% since 2009 [10, 11]. But the utilization of the HTM will greatly increase the total cost due to the complex device fabrication process and the difficulty in HTM synthesis. Additionally, the light-induced degradation of the HTM could be a threat to the device stability [12]. To overcome these shortcomings, designing CIGS and perovskite devices with novel structures may be an effective way to enhance the stability and acquire better solar cell performance.
Here, we have designed two novel devices of sandwich CIGS and heterojunction perovskite, referring to the advantages of the architectures of sandwich perovskite (standard) and heterojunction CIGS (standard) solar cells respectively. The similar property of the polycrystalline characteristic and large absorption coefficient of CIGS and perovskite materials makes it possible for their architecture reference. The key features of the modulated devices were analyzed by a one-dimensional simulator wxAMPS, which is a new version of the AMPS-1D that is well adapted to modeling various homo-junction, heterojunction, multi-junction solar cells [13-15]. By further varying of the thickness and doping density of absorber layer, we explored their effects on the sandwich CIGS and heterojunction perovskite solar cells' performances.
2. Device simulation parameters
In the wxAMPS simulation, the standard heterojunction CIGS solar cell was in configuration of ZnO/CdS/CIGS, where CIGS is utilized as the p-type absorber layer and ZnO/CdS as the n-type layers, as shown in Fig. 1(b). The standard sandwich perovskite solar cell was in configuration of TiO2/MAPbI3/Spiro-OMeTAD, as shown in Fig. 1(a). The modulated novel sandwich CIGS and heterojunction perovskite devices were depicted in Figs. 1(d) and 1(c), which are referring to the structures of sandwich perovskite solar cell (Fig. 1(a)) and heterojunction CIGS solar cell (Fig. 1(b)), respectively. Tables 1 and 2 summarize the input modeling parameters for the standard heterojunction CIGS and sandwich perovskite devices, respectively, most of which were selected from recent reported experimental works. For sandwich CIGS and heterojunction perovskite devices, it simply replaced the absorber of the sandwich perovskite device with the CIGS layer and the absorber of the heterojunction CIGS with perovskite device. In our simulation, the AM1.5 solar radiation spectrum was adopted as the light source. Surface recombination rates of both the front and the back were set to 1 × 107 cm/s. Energy levels of defects in the simulated thin film materials were located at the center of their bandgap with the Gaussian-type energetic distribution (characteristic energy 0.1 eV). The absorption coefficients of the CIGS and perovskite layer were derived from Refs. [16, 17], respectively.
Parameter and unit | Compact TiO2 | CH3NH3PbI3 | Spiro-OMeTAD hline |
εr | 100 [18] | 30 [19] | 3 [20] |
Eg (eV) | 3.2 | 1.55 [21] | 3.17 [20] |
χ (eV) | 4 | 3.9 [22] | 2.05 [20] |
Thickness (μm) | 0.03 | 0.35 | 0.15 |
Na (cm-3) | 0 | 0 | 3 × 1018 |
Nd (cm-3) | 5 × 1019 [18] | 0 | 0 |
Nc (cm-3) | 1 × 1021 | 2.5 × 1020 [23] | 2.8 × 1019 [24] |
Nv (cm-3) | 1 × 1021 | 2.5 × 1020 [22] | 1 × 1019 [24] |
Parameter and unit | ZnO | CdS | CIGS |
εr | 9 | 10 | 13.6 |
Eg (eV) | 3.3 | 2.4 | 1.15 |
χ (eV) | 4.4 | 4.2 | 4.5 |
Thickness (μm) | 0.2 | 0.05 | 3 |
Na (cm-3) | 0 | 0 | 2 × 1016 |
Nd (cm-3) | 1 × 1018 | 1.1 × 1018 | 0 |
Nc (cm-3) | 2.2 × 1018 | 2.2 × 1018 | 2.2 × 1018 |
Nv (cm-3) | 1.8 × 1019 | 1.8 × 1019 | 1 × 1019 |
3. Results and discussions
Table 3 summarizes the modeling performance parameters of the standard/preliminary modulated CIGS and perovskite solar cells. The preliminary modulated solar cell means the active layer in the modulated structure exhibits the same property as in the original standard structure. For both the standard and sandwich CIGS solar cells, the CIGS layer is set to be 3 μm thick, with doping density equal to 1 × 1016 cm-3; and for both the standard and heterojunction perovskite solar cell, the perovskite layer is set to be 0.35 μm thick, with doping density equal to 1 × 1014 cm-3 (intrinsic). However, in terms of the performance values such as the Voc, Jsc, FF and PCE values of preliminary modulated devices, they are inferior to the standard devices. Here, we adjusted the thickness and doping density of the active layer of the modulated solar cells to explore their effects on the device characteristics.
Solar cells | Active layer property | JSC (mA cm-2) | VOC(V) | FF(%) | PCE (%) | |
CIGS | Heterojunction | 3 μm, doping (1 × 1016) | 37.42 | 0.63 | 79.02 | 18.48 |
Sandwich | 3 μm, doping (1 × 1016) | 34.34 | 0.62 | 65.74 | 14.09 | |
Perovskite | Sandwich | 0.35 μm, doping (1 × 1014) | 24.08 | 1.08 | 76.13 | 19.71 |
Heterojunction | 0.35 μm, doping (1 × 1014) | 18.24 | 0.71 | 73.79 | 9.6 |
When decreasing the thickness of the CIGS layer in the sandwiched one from 3 to 0.35 μm, yet the doping concentration fixed to 1 × 1016 cm-3, there is little change of Jsc but Voc is greatly increased (Fig. 2(a)). The value of the efficiency also gets enhanced as a result of the decreased CIGS thickness. Based on this, the doping effect of the absorber layer is further explored with the thickness fixed to 0.35 μm. It can be seen in Fig. 2(b) that when decreasing the doping density of CIGS layer from 1 × 1016 cm-3 to 1 × 1014 cm-3, the device performance improves a lot, especially in the FF and Voc value. An optimized sandwich CIGS device with efficiency of 20.7% is obtained. The reasons for these variations can be explained as follows. Instead of improving the photoelectric property, the addition of p-type HTM even promotes a negative effect on the sandwiched one. This is because the CIGS/HTM interface can provide a weak electric field which is beneficial for hole transporting but also a heterojunction contact which would intensify recombination. Since the span of the interface electric field is much shorter than the width of CIGS layer (3 μm) and there are few carriers generating in the CIGS layer adjacent to CIGS/HTM interface, the electric field originally exploited to separate photo-induced carriers is weak for lack of enough carriers, hence the recombination effect takes the lead. When decreasing the thickness of the CIGS layer in the sandwiched one, although Jsc reduces as a result of less light absorption, the built-in electric field could cross the whole active layer and the recombination rate is greatly reduced. Note that Voc depends mainly on the built-in electric field, depletion width and recombination rate, therefore, sandwiched ultra-thin CIGS film leads to rather high Voc and PCE. As doping density decreases, the original n-p-p+ junction gradually turns into n-i-p junction, and CIGS absorption material acts as an i layer. In the i layer, trap-assisted recombination centers are reduced significantly and minority carrier lifetime and diffusion length are longer than in the doped layer, which suppresses the recombination of photon-generated carriers in the CIGS layer. Hence, it is revealed that the sandwich CIGS with thin/intrinsic absorber demonstrates reduced recombination rate (Fig. 2(d)) and broader span of the built-in electric field (Fig. 2(c)) than in standard structure with thick/doped absorber.

When increasing the thickness of the perovskite layer in heterojunction structure from 0.35 μm to 3 μm, yet the doping concentration fixed to 1 × 1014 cm-3, both the Jsc and Voc are increased, thus the device efficiency is enhanced accordingly (Fig. 3(b)). This can be ascribed to the more captured sunlight as a result of the thicker absorber thickness. However, the Voc of the heterojunction perovskite device is inferior than in the sandwich structure, which is related to the Fermi levels of the respective layers. In the sandwich device, Voc depends on the difference between the Efn of ETM and the Efp of HTM; in the heterojunction device, Voc depends on the difference between the Efn of ETM and the Efp of perovskite, the smaller difference of the latter leads to low Voc, which is negative for high-efficiency solar cell (Fig. 3(a)). Consequently, to promote the performance of heterojunction structure as much as possible, improving the Voc is of great importance. Here we fix the absorber thickness as 3 μm and further explore the effect of the doping density. As Fig. 3(c) shows, when the doping is less than 1 × 1016 cm-3, there is no evident change of the Voc value, but with the doping density increasing from 1 × 1016 to 1 × 1017 cm-3, the Voc as well as the efficiency gets significantly enhanced. This can be ascribed to the much more efficient charge separation effect induced by the intensified built-in electric field through heavy doping. As Fig. 3(d) shows, when increasing the doping from 1 × 1016 to 1 × 1017 cm-3, the built-in electric field is greatly intensified and even stronger than in the standard sandwich perovskite solar cell, which will enhance the efficiency of the device.

Therefore, for the sandwich CIGS solar cell, the thin/intrinsic CIGS film is more favorable for charge transfer and will lead to better device performance; and for the heterojunction perovskite solar cell, a relatively thick and heavily doped perovskite layer will benefit in promoting its photoelectric property. But since it is hard to fabricate intrinsic CIGS film and there is an interface contact problem in sandwich CIGS, and the thick/doped heterojunction perovskite device could introduce defects induced by pin-holes, it is still a challenge to fabricate the CIGS and perovskite solar cells with novel structures and many works remain to be done to promote their practical application.
4. Conclusion
In this paper, we designed two thin-film polycrystalline solar cells with novel structures: sandwich CIGS and heterojunction perovskite, referring to the advantages of the architectures of sandwich perovskite (standard) and heterojunction CIGS (standard) solar cells, respectively. The characteristics of the devices were investigated by wxAMPS simulation platform with variation of the thickness and doping density of the absorber layers. It is revealed that the sandwich CIGS solar cell with intrinsic/thin CIGS film could lead to better device performance since there is less interface recombination and more efficient charge transfer. The optimized efficiency of sandwich CIGS is 20.7%, much higher than the standard heterojunction CIGS structure (18.48%). The heterojunction perovskite solar cells can be more efficient at employing thick/doped perovskite absorber (16.9%) than these typically utilizing thin and weak-doping/intrinsic perovskite films (9.6%) as a result of the intensified built-in electric field. This structure modulation proves to be useful and can be further applied for other solar cells' design to acquire better performance.