Perovskite solar cells with a hybrid electrode structure

Organic-inorganic hybrid perovskites are one of the most promising low cost and solution-processable photovoltaic (PV) materials.^1,2 To date, perovskite solar cells (PSCs) with a simple n-i-p sandwich structure [Fig. 1(a)] have yielded power conversion efficiencies (PCEs) above 25%, surpassing the PCEs of polycrystalline Si devices (∼22%).³ Similar to their Si-based analogs, it is predicted that perovskites can achieve even higher PCEs using a back-contact device architecture.^4,5 The first operational back-contact perovskite solar cell (BC-PSC) incorporated a quasi-interdigitated electrode (QIDE) in which a microarray of anodes is placed above the planar cathode [Fig. 1(b)].^6–8 This robust architecture is more suitable for perovskite photoabsorbers, as BC-PSCs using a conventional interdigitated back contact require a reduction in electrode spacing to accommodate the shorter carrier diffusion lengths compared to silicon, likely leading to a prohibitive increase in electrode defects.^9,10 Despite this advantage, the reported performances of BC-PSCs incorporating QIDEs have plateaued at 5%, suggesting that a greater understanding of the device physics is required.⁷ 

In principle, the PCEs of BC-PSCs and conventional sandwich PSCs should be comparable,^4,9,10 but the difference in record efficiencies between the two architectures indicates that there may be inherent issues with the former structure. However, even if sandwich and back-contact PSCs are fabricated using the same materials, comparing the underlying physics of the architectures remains complicated, as the two structures will use active layers with different properties, i.e., BC-PSCs will give rougher perovskite layers due to the electrode structure or the sandwich PSC will suffer transmission losses due to the substrate and its active layers. Overcoming these issues requires the creation of a hybrid-structured device in which the back-contact and sandwich modes (SWMs) are incorporated into a single solar cell sharing an identical active area. This would allow for the direct comparison of the PSC performance in either the back-contact or sandwich mode using exactly the same perovskite photoabsorber layer.

Recently, DeLuca et al. reported bifacial BC-PSCs incorporating transparent back-contacts with illumination from either side giving similar performance [Fig. 1(b)].¹¹ For our purposes, such a device is an excellent candidate to realize the aforementioned PSC with a hybrid structure with the deposition of an additional electrode system on top of the perovskite layer of a BC-PSC, yielding a device with one common cathode and two anodes (or vice versa) [Fig. 1(c)]. Herein, we demonstrate an example of a PSC with such a hybrid electrode structure using transparent QIDEs and perform a comparative analysis between the sandwich and back-contact operation modes.

The first stage of building a hybrid-structured PSC was to fabricate BC-PSCs using transparent QIDEs, as described by DeLuca et al.¹¹ A cross-sectional scanning electron microscope (SEM) image of a hybrid device shows that the ∼350 nm Cs_0.05FA_0.79MA_0.16PbI_2.49Br_0.51 perovskite layer uniformly covers the transparent QIDE, which consists of a microstructured anode [∼1.5 µm wide indium tin oxide “fingers” with a CuSCN coating (ITO|CuSCN) spaced every ∼2.5 µm] on a continuous cathode [fluorine-doped tin oxide with a TiO₂ layer (FTO|TiO₂) on glass] (Fig. 2). The anode and cathode are separated by an ∼100 nm Al₂O₃ insulating layer to prevent shorting. The device was completed by spin coating an ∼60 nm CuSCN hole transporting layer (HTL) to conformally cover the exposed perovskite surface, followed by evaporating ∼80 nm of gold through a mask to form the top electrode for the sandwich part (see Fig. S1 of the supplementary material). A more detailed description of the fabrication process is given in the supplementary material. A photograph of the device is shown in Fig. S2 of the supplementary material.

The J–V characteristics of the hybrid-structured PSC devices were recorded for both operation modes with rear (glass) side illumination. The sandwich mode (SWM) operates when the back-contact mode (BCM) is open circuited (J_BCM = 0), and vice versa, allowing the SWM and BCM to operate independently of each other. Figure 3 shows J–V curves under 1 sun AM1.5G illumination and in the dark. The relevant photovoltaic parameters [short-circuit current density (J_SC), open-circuit voltage (V_OC), fill factor (FF), and PCE] for both operation modes are listed in Table I. The disparity between forward and reverse scans is due to the hysteresis phenomenon, which is commonly ascribed to the existence of mobile ions within the perovskite material.^1,6–9,11 Both operation modes show similar levels of hysteresis (taking the ratio of reverse to forward scan). The stabilized power output for PSC devices operated in sandwich (black) and back-contact (red) modes under 1 sun illumination was obtained from maximum power point tracking experiments (see Fig. S3). The statistical data for photovoltaic parameters for both operation modes are listed in Table S1 of the supplementary material.

The device has an identical active area (0.08 cm²) for both SWM and BCM configurations that, therefore, differ only in the geometry of their electrical contacts (Figs. 2 and S2). This suggests that the SWM and BCM configurations should have the same attainable optical power at zero voltage (J_sc). However, the experiment shows that the J_sc of the SWM and BCM are different and far below the theoretical maximum potential J_sc calculated based on wave-optic simulations (accounting for all interference, parasitic absorption, transmission, and reflection losses).¹¹ This indicates that there may be a collection loss associated with the device electronic configuration, which has not been considered earlier.^4,10,11

To understand this discrepancy, the J-V curve characteristics were first modeled using a simple triple-diode equation to capture the characteristics of the p-i-n layers without prior knowledge of the device contact geometry. The goal of the model is to provide an initial estimate of the effective collection length (L_c) by evaluating the J-V curve characteristics between the SWM and BCM configurations,

This model assumes that the collection efficiency is 100% and is independent of the applied voltage; however, this assumption is subsequently corrected (see the discussion below). The n_0p, n_0i, and n_0n are the diode-ideality factors (see Table S2 of the supplementary material) at the p-, i-, and n-type layers, respectively (obtained numerically using global optimization to minimize errors between experimental and fitting). The (kT/q) is the thermal voltage at 300 K, R_S is the series resistance, R_SH is the shunt resistance, and J_sc is the theoretical maximum short-circuit current density (obtained from the optical simulation). The J_0p, J_0i, and J_0n are dark saturation current densities at the p-, i-, and n-type active layers, respectively. J_0p and J_0n contain information on the depletion regions from which the initial guess of L_c can be interpolated (details in the supplementary material). This model alone is insufficient to fit the whole experimental data (even if global iterations of R_S and R_SH are taken into account) but is sufficient to obtain an initial estimation of the L_c for each configuration.

Next, a model developed within a reciprocal relation between carrier injection and collection was used to determine in greater detail the source of the discrepancy and to explain the difference in collection efficiencies between sandwich and back-contact geometries.^12–15 The collection efficiency is therefore not only limited by the device geometry but also by the voltage applied to the respective structures. The collection efficiency (η_c) is formulated as (details in the supplementary material)^12–15 

where η_i is the injection efficiency, η_out is the light out-coupling efficiency, A(λ) is the absorption of the active layer, ϕ_sun(λ) is the 1 sun irradiation, and ϕ_bb(λ) is the black-body radiation, all of which are the same for both configurations. In contrast, the voltage-dependent photoluminescence quantum yield [PLQY(V)] and voltage dependent dark-current [J_D(V)] containing the triple diode equation vary and are therefore dependent on the two contact geometries. This highlights the importance of Eq. (1) in obtaining the initial estimate of L_c to be used in Eq. (2), from which after subsequent iterations, we can extract the effective collection length (L_c). Figure 4(a) depicts the collection efficiency as a function of applied voltage [Eq. (2)] and shows that the collection efficiency is indeed not constant across an applied voltage (i.e., the quasi-Fermi level splitting is not constant) and is dependent on the collection length (L_c) as interpolated from the model. Based on this model, L_c is at least an order of magnitude shorter in the BCM geometry than in the standard SWM configuration. Open circuit photovoltage decay (OCVD) measurements reveal a shorter electron lifetime for the BCM than for the SWM by a factor of ∼4 (Fig. S4), which supports the findings of this model.

To provide a more intuitive picture, a fully coupled optical-electrical simulation was performed to map the spatial dependence of collection efficiency on the BCM geometry [Fig. 4(b)]. This map is constructed by “pumping” a uniform generation rate on top of the spatially dependent generation rate in every spatial point of the map and “probing” a change in total collected majority carriers. The maximum collection efficiency takes place near the adjacent contacts where an intense space-charge region is expected. This is consistent with experimental findings reported previously.⁶ As applied bias, the collection efficiency is reduced where the range of its distance is much shorter than the half-pitch size demonstrated here (36 nm collection length vs 1.25 μm pitch). In other words, the electric field required for charge separation is reduced with the distance from the closest adjacent two contacts. Note that in the experimental data, the collection efficiency is effectively zero at voltages close to V_oc, meaning that there is no charge extraction at V_oc, and is equivalent to the maximum injection efficiency in the dark condition (maximum quasi-Fermi level splitting).

BC-PSCs must overcome the limitations of their collection efficiencies to achieve their maximum potential efficiencies. To solve this problem, a large depletion region can be created to improve the charge collection efficiency, for example, not only near the contacts but also at the grain boundaries.¹⁶ Recently, Tainter et al. showed that effective charge segregation (probably at the grain boundaries) may improve long-range charge extraction in a similar device geometry.¹⁷ Grain boundaries may form an array of p-i-n junctions through the interfaces between the two electrodes and that is conceptually equivalent to the effective depletion region discussed here. However, in our case, it seems that only the interfaces between active layers and contacts serve as p-i-n junctions. Engineering the back-contact geometry by utilizing a cross-grating back-contact may break this limitation by localizing larger carrier generation around the contact geometry.⁵ In addition to this, the importance of the voltage dependent collection efficiency should be considered in interpreting any data obtained from operando characterizations of thin-film semiconductors when using a back-contact geometry as a platform.

For the purpose of simplifying the analysis and modeling of the devices presented here, we considered the influence of the back-contact electrode on the operation of the devices in SWM configuration to be negligible. However, comparing the performances for our device in SWM configuration with standard sandwich-type planar PSC devices reported in the literature requires caution, as the influence of the embedded back-contact electrodes on device performance in SWM configuration is yet to be explored, and the low efficiencies obtained suggest that the presence of the back-contact electrodes may have a detrimental effect. Nevertheless, the objective of utilizing this hybrid electrode structure is to reveal the inherent issues with the back-contact architecture, in particular, and to understand the mechanisms through which the electrodes are limiting device performance. Both the experimental and theoretical results in this study suggest that the issue with charge collection needs to be addressed for the back-contact architecture to fulfill its potential.

In summary, the fabrication of a PSC device with a hybrid structure was achieved by depositing a top electrode onto a back-contact device with t-QIDEs. The performance of the device in both operation modes was tested, with the PCE of the SWM being three times higher than that of the BCM. The discrepancy between the PCEs of the SWM and BCM operation modes was rationalized using a coupled optical-electrical simulation method and a semianalytical model. The BCM suffers from a charge collection efficiency at least an order of magnitude lower than the SWM configuration because of its intrinsically narrow depletion region across the contact interfaces. Future work on the back-contact perovskite solar cell should focus on understanding and overcoming this fundamental limitation, such as through interfacial doping.

See the supplementary material for experimental details including materials, fabrication of transparent quasi-interdigitated electrodes, deposition of the CuSCN layer, PV device fabrication, characterization details, and details of the solar simulation setup and device structures, J–V curves, photovoltaic parameters, open circuit photovoltage decay curves, electron lifetime estimations, and physical parameters used in simulations of collection efficiency.