Development of SnO₂ Composites as Electron Transport Layer in Unencapsulated CH₃NH₃PbI₃ Solar Cells

Abstract

Improving morphological and electronic properties of the electron transport layer (ETL) is a critical issue to fabricate highly efficient perovskite solar cells. Tin dioxide is used as an ETL for its peculiarities such as low-temperature solution-process and high electron mobility and several handlings have been tested to increase its performances. Herein, SnO₂:ZnO and SnO₂:In₂O₃ composites are studied as ETL in planar n-i-p CH₃NH₃PbI₃ solar cells fabricated in ambient air, starting from glass/ITO substrates. Morphological, electrical and optical properties of zinc- and indium-oxide nanoparticles (NPs) are investigated. First-principle calculations are also reported and help to further explain the experimental evidences. Photovoltaic performances of full devices show an improvement in efficiency for SnO₂:In₂O₃–based solar cells with respect to pristine SnO₂, probably due to a suppression of interfacial charge recombination between ITO/ETL and ETL/perovskite. Moreover, a better homogeneity of SnO₂:In₂O₃ deposition with respect to SnO₂:ZnO composites, conducts an increase in perovskite grain size and, consequently, the device performances.

1. Introduction

Perovskite solar cells (PSCs) represent a fast-developing technology and have been attracting enormous interest for the last decade thanks to their high efficiencies and low production costs that are both fundamental requirements for scalability. The recent record of 25.5% [1], established by UNIST, demonstrates the great potential of PSCs.

The PSC is a multi-component device and there are many aspects of the production process that can be addressed in order to promote the scalability. The perovskite is sandwiched between an electron transport layer (ETL) and a hole transport layer (HTL), which extract and transport the photo-generated electrons and the holes in the perovskite to the external circuit, respectively.

In the conventional architecture, namely n-i-p, the ETL and the transparent conductive oxide (TCO), such as indium tin oxide (ITO), form the front electrode providing a good transparency to visible light. The perovskite is deposited on top of the ETL and the full device is completed with a back electrode composed by a metal-coated HTL.

The ETL plays a central role in obtaining PSCs with high performances and sustainable costs. In addition to the electron extraction, this component hinders the hole transfer, thus preventing undesirable charge recombination processes and maximizing the current generated by the solar cell. Suitable ETL should provide a proper alignment of the valence and conduction bands (VB and CB) with the perovskite ones.

Titanium dioxide has been the most used ETL material, providing so far the best results. However, TiO₂ has several drawbacks. It has a relatively low film electron mobility (~10⁻⁵ cm² V⁻¹ s⁻¹) [2] and a high-temperature deposition process (about 500 °C), which increases the energy pay-back time and hinders its use in tandem or flexible solar cells [3]. Moreover, TiO₂-based PSCs degrade rapidly under UV illumination due desorption of oxygen adatoms [4], and such instability limits further applications. Hence, the design of alternative materials to substitute TiO₂ as ETL is a current challenge in the field of PSCs. In this context, tin(IV) oxide (SnO₂) is considered the most promising candidate as ETL. In particular, SnO₂ nanoparticles (NPs) provide good anti-reflective properties and thermal stability [5], and are less hygroscopic [6] and more stable under UV radiation [7] than TiO₂. The latter two properties avoid the degradation of the perovskite layer and ensure the long term stability of the device [8]. Moreover, SnO₂ has a wider bandgap (~3.8 eV [9]), which avoids absorption of high energy photons. Its higher electron mobility (~10⁻³ cm² V⁻¹ s⁻¹ [10]) reduces charges accumulation at the ETL/perovskite interface, improving the electron transport efficiency and decreasing undesired carrier recombination processes [11]. Only a few years ago, several groups achieved good results using SnO₂ as ETL [12,13,14], but the performances were still much lower than TiO₂-based devices. The reason was attributed to the high temperature (450 °C) used during the annealing that produces a large amount of charge traps and recombination centers in the layer such as oxygen vacancies [15]. To avoid these defects, deposition processes with low temperature have been developed. Jiang et al. [2] used SnO₂ colloidal precursor to procure dense and pinhole-free film. The SnO₂ nanoparticles precursor was spin coated on ITO substrate and post-annealed at 150 °C. Actually, PSCs based on SnO₂ are widely studied in many works [16,17,18,19]. The low temperature process suggested the potential of SnO₂ as ETL also in tandem with perovskite-silicon solar cells [20,21]. Several strategies have been adopted to further increase its performance. The addition of dopants [22,23,24], surface passivation treatments [25,26], and bilayers composed by different ETL materials [27,28,29,30] or the use of nanocomposites [31] are typical material handlings used to modify the electronic properties, or improve the ETL/perovskite interfacial contact and the band alignment.

In this paper, we explore the effects of combining tin dioxide with other metal oxides to form composites deposited on two different ITO-based substrates for realizing methylammonium lead triiodide perovskite (MAPbI₃) solar cells (device configuration is shown in Figure S1). In particular, we built unencapsulated PSCs with a n-i-p planar structure using SnO₂:In₂O₃ or SnO₂:ZnO composites as ETL. Both ETL and MAPbI₃ were spin-coated in ambient air. For MAPbI₃ deposition, a “one-step” spinning procedure is used, with opportune anti-solvent, which proved to be one of possible strategies to counteract moisture during the fabrication process [32,33]. Indeed, many studies focus on the manufacturing of perovskite to make the whole process compatible with standard environmental conditions [34,35,36,37,38]. Doped Spiro-OMeTAD [39] was employed as HTL and Au and ITO as back and front electrodes, respectively. Photovoltaic performances, also in this case acquired in air, were compared. We also performed a state-of-the-art density functional theory (DFT) [40,41] study in order to provide a microscopic understanding of the ETL electronic structure.

We found an improvement in fill factor (FF) both for SnO₂:In₂O₃- and SnO₂:ZnO-based devices, leading to an increment in efficiency only if In₂O₃ NPs were used. Both experimental and theoretical results suggest that the introduction of indium oxide nanoparticles can refine the interfacial contact with ITO and perovskite inducing a uniform and pinhole-free perovskite film deposition and enhancing the electron transfer. SnO₂:In₂O₃ composites, deposited as ETL in PSCs without control of environmental conditions, allow to reach satisfactory efficiencies.

2. Experimental Section

2.1. Materials

Indium oxide (In₂O₃) nanopowder, zinc oxide (ZnO) nanopowder, lead iodide (PbI₂) (99.999%), spiro-OMeTAD, 4-tert-butylpyridine (t-BPY), FK 209 Co(III) TFSI salt, dimethyl formamide (DMF), dimethyl sulfoxide (DMSO), lithium-bis(tri-fluoromethylsulfonyl)-imide (Li-TFSI), diethyl ether, acetonitrile (ACN) and chlorobenzene (CBZ) were purchased from Sigma-Aldrich/Merck, St. Louis, MO, USA. The tin(IV) oxide 15% w/w in H₂O colloidal dispersion was obtained from Alfa-Aesar, by Thermo Fisher GmbH, Kandel, Germany. Methylammonium iodide (MAI) was prepared following the procedure in the reference [42]. An excess of methylamine (50 mL) was reacted with hydroiodic acid (20 mL) in 200 mL of ethanol under N₂ atmosphere at room temperature for 2 h. Crystallization of MAI was achieved using a rotary evaporator. Methylamine (33 wt% in absolute ethanol) and hydroiodic acid (57 wt% in water) were purchased from ACROS Organics, Geel, Belgium.

2.2. Device Fabrication and Characterization

We built PSCs with the following layout ITO/ETL/MAPbI₃/doped-Spiro-OMeTAD/ Au using SnO₂:ZnO or SnO₂:In₂O₃ composites as ETL and compare their performances with the SnO₂–based device. Two different ITO glass substrates were used: the former was purchased by Kintec (named comITO) and the latter was fabricated in our laboratories (named labITO) using a RF magnetron sputtering system MRC-643 equipped with a loadlock. Sputtering processes were carried out by using an ITO ceramic target (10 wt% SnO₂) in argon atmosphere at RF-power of 400 W. ITO films were deposited on non-intentionally heated Corning Eagle XG glass substrate. All the ITO substrates were treated under UV-ozone for 30 min before the ETL film deposition to enhance the substrate wettability. Three ETL solutions were prepared in air starting from diluting an equal volume of Alfa-Aesar SnO₂ colloidal dispersion with deionized water (SnO₂:H₂O = 1:1), or 1.5% w/w dispersions of In₂O₃ or ZnO NPs in deionized water (SnO₂:In₂O₃ = 1:1 and SnO₂:ZnO = 1:1). The ETL solutions were centrifuged and supernatant was spin-coated in air at 6000 rpm for 50 s and the substrates were dried on a hot plate at 150 °C for 30 min. MAPbI₃ layer was obtained with the “one-step” solvent engineering approach [43,44]. The MAPbI₃ precursor solution (1.4 M) was obtained in a N₂-filled glove box by dissolving PbI₂ and MAI in a mixture of DMF:DMSO (9:1, volume ratio). ETL-based substrates were treated again under 7 min of UV-ozone. Finally, the precursor solution was spin-coated onto substrates in air at 1000 rpm (200 rpm/s) for 10 s and then at 5000 rpm (1000 rpm/s) for 45 s, dripping the diethyl ether anti-solvent in the proper time during the second step spinning. Then, the substrates were dried on a hot plate beforehand at 50 °C for 2 min and after at 100 °C for 10 min. Environmental conditions were monitored (45–70% of relative humidity and 20–25 °C). A 0.060 M solution of Spiro-OMeTAD, as HTL, was prepared in a N₂-filled glove box dissolving the Spiro-OMeTAD powder in CBZ. After, considering 1 mL of solution, it was sequentially doped as in the reference [39] with 28 μL of t-BPY, 17 μL of the lithium solution (1.8 M Li-TFSI in ACN), and 7.5 μL of the cobalt solution (4.0 M FK 209 Co(III) TFSI salt in ACN) few hours before the HTL spin coating deposition. Finally, the doped Spiro-OMeTAD solution was spin-coated in a N₂-filled glove box at 4000 rpm for 30 s and 80 nm gold electrodes were deposited using a thermal evaporator (by Morfield).

Scanning electron microsopy (SEM) images were collected by Thermo Fisher Scientific Phenom pro X SEM. Dynamic light scattering (DLS) measurements were performed with a Zetasizer Nano ZS He-Ne laser 633 nm, 4 mW. Steady-state photoluminescence (PL) was measured with a Renishaw Raman inVia Re-flex system with excitation at 514 nm by an Argon ion laser. Transmission spectra were measured using the PerkinElmer λ-900 spectrophotometer. KLA-Tencor P7 profilometer was used to measure the thickness of the ETLs. Work functions of two different ITO substrates were carried on through Kelvin Probe Technique using an SKP5050 Scanning Kelvin Probe (KP Technology, Wick, Caithness, Scotland). Solar cells were characterized under the AM1.5G spectrum generated by a class AAA dual-lamp solar simulator (WACOM Electric Co. LTD, Tanaka, Fukaya-shi, Saitama) in uncontrolled moisture and temperature conditions. The current–voltage characteristics of the devices were obtained by applying external voltage while recording the generated photocurrent using a Keithley (Model 2651A) high power system source meter. JV data acquisition was performed with scan rate of 2 V/s for all the electrical measurements and the scan range was 1.2 V to −0.1 V (reverse scan direction). The irradiated area was 0.11 cm², defined by applying a shadow mask. JV characteristics were recorded immediately after the gold electrode evaporation.

2.3. First-Principle Calculations

We performed periodic spin-polarized DFT calculations with the GGA/PBE density functional [45,46] and plane wave basis set with the Vienna Ab-initio Simulation Package program (VASP, version 5.4.1) [47,48,49,50]. We computed the CB energy position of all the investigated ETL materials following the approach proposed by Toroker et al. [51]. We compared the results with the CB edge of the MAPbI₃ perovskite (−4.13 eV) that has been calculated at the same level of theory [52]. The ionic cores were represented with projector-augmented wave (PAW) potentials [53,54] and, in particular, 2s²2p⁴ electrons for O, 5s²5p² electrons for Sn, 3d¹⁰4p² electrons for Zn, and 5s²5p¹ electrons for In were treated as valence electrons. The long-range interactions were taken into account and treated with Grimme’s DFT-D3(BJ) correction method [55,56,57]. The break condition of the SC-loop was set to 10⁻⁵ eV. All atoms were allowed to relax freely to equilibrium in bulks and slab models until the force on each of them was less than 0.03 eV/Å. Bulk structure optimizations were performed considering the most thermodynamically stable phase of each metal oxide. These are the rutile structure (space group P4₂/mnm) for SnO₂, the bixbyte structure (space group Ia$\overline{3}$) for In₂O₃ and the wurtzite structure (space group P6₃mc) for ZnO. Surface slab models were constructed starting from PBE-D3BJ optimized bulk geometries. The surfaces considered in our study are the most stable non-polar surfaces, i.e., the (110) surface for SnO₂ [58], the (111) surface for In₂O₃ [59] and the (10$\overline{1}$0) surface for ZnO [60]. Slabs dimensions are chosen to reach an accurate description of the bulk properties in the inner layers. Thus, we built 5-layer slab models of SnO₂ (110) and ZnO (10$\overline{1}$0) containing 30 and 20 atoms, respectively, and a 2-layer slab (160 atoms) for In₂O₃ (111). In all the models, a 10 Å vacuum slab was inserted. Plane waves energy cut-off has been set to 800 eV in all calculations. Γ-centered Monkhorst-Pack k-point meshes were used for SnO₂ (6 × 6 × 9), ZnO (8 × 8 × 4) and In₂O₃ (4 × 4 × 4) bulks, and these have been scaled accordingly in surface models. Experimental band gaps were used in the approach implemented to calculate CB edges [51] (SnO₂ = 3.6 eV [61]; In₂O₃ = 3.7 eV [62]; ZnO = 3.2 eV [63]).

3. Results and Discussion

We carried out experimental characterizations in order to understand the effect of nanoparticles on the solar cell parameters. DLS measurements were performed on the ETL precursor solutions and different NPs size distribution patterns were obtained (Figure 1). The solution with In₂O₃ NPs shows a unique predominant population at ~100 nm. Such a narrow distribution of hydrodynamic radii is expected to generate smooth, compact and homogeneous ETLs.

In SnO₂:ZnO solution, the highest peak is at ~5 nm, highlighting a predominant NPs population with smaller dimensions than the solution containing only SnO₂ NPs. Composites and SnO₂ precursor solutions were deposited on ITO as described in experimental section. This procedure provides layers with almost the same thickness (~30 nm) for all the ETLs. SEM images of the perovskite layer were acquired on each ETL (Figure 2). When MAPbI₃ is deposited on SnO₂:In₂O₃ we observe larger perovskite grains with respect to the SnO₂:ZnO composite. This is ascribed to uniform distribution of NPs into SnO₂:In₂O₃, which improves the perovskite crystal growth, generating large grains. Instead, small aggregates of SnO₂:ZnO composite generate pinholes onto ITO substrates and form rougher surfaces that increase the density of nucleation sites for the perovskite crystal growth [64]. In Table S1 and Figure S2 data and images of grain analysis are shown. These results are in agreement with the observations deduced with DLS measurements.

Transmission UV-Vis spectra (Figure S3) reveal that both indium and zinc oxide NPs do not affect the optical transparency and anti-reflective properties of the tin dioxide/ITO unit. Transparency is about 80% in the visible spectrum, increasing up to 90% in the range 550–600 nm. The band gap extracted from Tauc plot is ~3.8 eV (inset of Figure S3), in agreement with other works [65,66,67].

We performed steady-state PL measurements on solar cell precursors without HTL, i.e., nominally identical MAPbI₃ films deposited on ITO/SnO₂, ITO/SnO₂:ZnO, ITO/SnO₂:In₂O₃ and ITO alone for comparison, in order to test the proposed ETLs for their ability to maintain a high photoluminescence [68]. A reduction in PL intensity is indicative of enhanced non-radiative recombination and thus reduced quasi-Fermi level splitting. Therefore, luminescence should be maximized to avoid V_OC losses. The measured PL spectra are shown in Figure 3. No major PL quenching is found. Similar emission is observed for all the samples, with peak intensity by only a factor ~2 lower than the reference ITO/MAPbI₃ stack. This suggests only slightly increased non-radiative recombination vs. the reference stack (which could be at the ETL/MAPbI₃ interface and/or in the MAPbI₃ film itself when deposited on the ETLs vs. ITO). In addition, the tested ETLs should, in principle, all allow for similar V_OC.

Full devices, starting from commercial ITO (comITO) substrates, were fabricated, introducing these composites as ETL and compared with SnO₂ layer. All the devices with SnO₂ or SnO₂:In₂O₃ as ETL were working, whilst some SnO₂:ZnO-based devices did not work. In Figure 4 box plot charts of normal distribution of PSCs are reported for all PV parameters. Short circuit current density decreases in solar cells with ZnO NPs inserted in the SnO₂ ETL, probably due to growth of smaller perovskite grains (Figure 2c) with more grain boundaries, which act as recombination centers. Conversely, no drop in V_OC is recorded, as expected by PL spectra. Generally, this parameter is less dependent on the undesired recombination processes at interfaces, with respect to Jsc.

We note that Jsc and V_OC for tin dioxide and SnO₂:In₂O₃ composite-based PSCs are quite similar, whereas the FF is lower for the reference. Ideally, FF depends on V_OC [69], but in the real devices many losses are present. This parameter improves when composites are used, probably due to higher electron mobility and transport ability of SnO₂:In₂O₃ or SnO₂:ZnO ETL leading to improved charge transport balance and the reduction in the interface charge accumulation [70,71]. Furthermore, the homogeneous population of the SnO₂:In₂O₃ precursor solution ensures a better interfacial contact of ITO and the perovskite with the composite film, which further enhances the charge transport. In

Table 1. Average data and standard deviation from J-V measurements of SnO₂, SnO₂:ZnO and SnO₂:In₂O₃-based PSCs starting from comITO substrates.

ETL	Jsc [mA/cm2]	VOC [V]	FF [%]	PCE [%]
SnO2	18 $\pm$ 1	1.01 $\pm$ 0.02	61 $\pm$ 3	11.2 $\pm$ 1.2
SnO2:ZnO SnO2:In2O3	12 $\pm$ 1 17 $\pm$ 1	0.99 $\pm$ 0.04 1.02 $\pm$ 0.02	79 $\pm$ 1 73 $\pm$ 3	9.5 $\pm$ 1.1 13.4 $\pm$ 0.6

average data and standard deviation of PSCs photovoltaic parameters, realized starting from commercial ITO substrates, are reported. The enhancement was confirmed by the fabrication of different batches (in Figures S4 and S5 and Table S2, JV data and EQE measurements extracted from another batch are reported as evidence).

This effect is also evident if we use ITO fabricated in our laboratories (labITO). Results are reported in Figure 5 and in

Table 2. Photovoltaic parameters of devices with different ETLs starting on labITO.

ETL	Jsc [mA/cm2]	VOC [V]	FF [%]	PCE [%]
SnO2	19.5	0.98	65.8	12.5
SnO2:ZnO	16.7	1.02	70.2	12.0
SnO2:In2O3	23.3	1.00	70.3	16.4

, where J-V and PV parameters of devices are shown, respectively.

We note an enhancement of performances caused by higher Jsc values. Then, we characterized labITO and comITO substrates. Transmittance and absorbance spectra are reported in Figure 6 and main optical parameters Eg and A_av (specifically, band gap and average optical absorbance weighted on the solar spectrum between 350 and 800 nm), sheet resistance, resistivity and charge carrier concentrations are shown in

Table 3. Thickness, optical and electrical parameters of comITO and labITO.

TCO	Thickness [nm]	Rsh [ohm/sq]	ρ [Ω∗cm] × 10−4	Ne [cm−3] × 1021	Aav [%]	Eg [eV]
comITO	180	10	1.8	1.2	8	3.83
labITO	285	11	3.2	0.55	9	3.65

. In Figure 6 it is evident that the commercial TCO presents a higher free carrier absorption in the red-NIR spectral region, due to a doping level two times higher than ITO film produced in our laboratory, as evident in Table 3. The higher Eg value observed for comITO with respect to labITO can be predominantly attributed to a higher charge carrier concentration, according to the Burstein–Moss effect for n-type degenerate semiconductors. This is the reason for the observed blue shift of the absorption edge. However, ITO films have a quite similar average optical absorbance value. Furthermore, another analysis was 2D work function maps, acquired over an area 5 × 10 mm² (shown in Figure 7), indicating mean absolute values of 4.82 eV and 4.62 eV for labITO and comITO, respectively. These values highlight that ITO involves different interfacial behavior when it is used as substrate. Then, the enhancement of Jsc parameter could be due to combined effects between lower charge carrier concentration and a better interfacial contact of ITO/ETL when labITO is used. It is also important to evidence that if SnO₂:In₂O₃ is used as ETL, electrical characteristics recorded immediately after full device fabrication are improved with respect to PSCs based on pristine SnO₂ which instead require a storage time to increase its PV performance [72,73].

Spin-polarized DFT calculations were also performed to gain insight on the CB positions of the materials under investigation. The proper alignment between the CB edges of the ETL and the absorber material is critical for a proper solar cell operation. The CB edge of the perovskite should lie at a higher energy than the ETL one in order to ensure the electron transfer between the two materials. Their energy difference represents a first estimation of the efficiency of the electron injection process. The obtained results elucidated that tin dioxide CB plays the major role in settling the V_OC of the solar cell also when combined with other oxides. In fact, as shown in Figure 8, the CB edge position of the ZnO and In₂O₃ compounds lies higher in energy than SnO₂ CB. Thus, SnO₂ CB position defines in both composites the V_OC of the device.

4. Conclusions

In this study, CH₃NH₃PbI₃ solar cells have been fabricated in ambient air for testing the performances of the SnO₂:In₂O₃ and SnO₂:ZnO composites as ETL. Insertion of In₂O₃ NPs inside the SnO₂ solution effectively improved the device performances. In particular, a higher efficiency has been independently measured from starting ITO substrates. The enhancement could be due to a better interfacial contact of ITO and perovskite with SnO₂:In₂O₃. Moreover, we addressed this behavior to the distribution of homogenous aggregates in precursor solution, refining the interface with the perovskite and making it more uniform with respect to ZnO NPs, which promotes the growth of small perovskite grains, which are detrimental for PSCs performances. Composites have provided quite the same V_OC value of the SnO₂-based device, confirming PL and theoretical results that suggest the SnO₂ CB also in composites define the driving force of the electron transfer process. Therefore, improvements and worsening obtained using composites depend essentially on NPs size distribution, which influences the uniformity and interfacial ETL/perovskite. Our work points out the importance of morphological effects caused by use of composites and benefits offered by ETL precursor solutions with a homogeneous population in order to obtain a high quality of interfaces. In future, different characterization techniques (such as dark J-V) are planned to be used to deeply explore the observed phenomena. Further computational studies are needed to understand at atomic level how the indium- and zinc-oxide structures influence the quality of the interface. At the same time, the investigation of different SnO₂:In₂O₃ ratio in ETL precursor solutions could lead to further improvements of the interfacial contact with the perovskite, so to enhance the electron transfer process.