Optical Optimization of the TiO2 Mesoporous Layer in Perovskite Solar Cells by the Addition of SiO2 Nanoparticles.

In this work, SiO2 nanoparticles (NPs) were integrated into the mesoporous TiO2 layer of a perovskite solar cell to investigate their effect on cell performance. Different concentrations of SiO2/ethanol have been combined in TiO2/ethanol to prepare pastes for the fabrication of the mesoporous layer with which perovskite solar cells have been fabricated. Addition of SiO2 NPs of 50 and 100 nm sizes produces an enhancement of cell performance mainly because of an improvement of the photocurrent. This increment is in good agreement with the theoretical predictions based on light scattering induced by dielectric SiO2 NPs. The samples using modified scaffolds with NPs also present a significant lower current-potential hysteresis indicating that NP incorporation also affects the ion accumulation at the perovskite interface, providing an additional beneficial effect. The results stress the importance of the appropriated management of the optical properties on further optimization of perovskite solar cell technology.

Introduction

Ever since the seminal reports in sensitized[1,2] and in all-solid solar cell configurations,[3,4] halide perovskites have been widely investigated as light harvesters in solar cells because of their optimal properties, including direct band gap, large absorption coefficients, and high carrier mobility.[5] Indeed, the significant work carried out in the research resulted in a power conversion efficiency (PCE) of perovskite solar cells (PSCs) close to 23%.[6] The main strategy to enhance PCE has been the optimization of the deposition process and of the halide perovskite harvester by the use of multication/anion perovskites.[7−11] However, other strategies have been employed including the incorporation of noble metal nanoparticles (NPs), increase of the thickness of the absorbing material, adding scattering layers, or the use of back reflectors and other light-trapping mechanisms.[12−14] In these cases, the improvement on solar cell performance is attained by the management of the optical properties of the device. For example, deposition of gold and silver NPs is a common approach because of their high (and size tuned) scattering and absorption cross-sections.[15] They have also been shelled with SiO2[16−18] to protect them from the corrosive effect of iodine and incorporated in the architecture of PSCs, observing an enhancement of the device performance.[17] Theoretical studies predict that core–shell Au/SiO2 NPs provide optical absorption enhancement in PSCs in spectral regions where the perovskite material has relatively poor absorption.[19−21] It is suggested that this enhancement is mainly based on a plasmonic effect, but it is not the only way in which Au/SiO2 NPs can help to enhance the PSC performance. Recently, we have used Au/SiO2 NPs with a smaller Au core than required to produce significantly enhanced light absorption due to the near-field plasmonic effect.[8] However, we have observed an increase of cell performance by modification of the interfacial properties.[22] Light scattering is also a way in which light harvesting can be improved because of the light-trapping enhancement. Nanopatterned mp-TiO2 layers enhance the light-harvesting efficiency of PSCs.[23] In addition, the high scattering efficiency observed for big Au/SiO2 NPs demonstrates to be an effective way of improving light harvesting because of the redirection of the incident light into solar cells with increased optical path lengths (light-trapping effect).[24−27] In terms of material preparation process and cost, the total elimination of Au can be highly interesting if SiO2 dielectric NPs could also induce a beneficial role on PSCs. The use of a mesoporous layer of SiO2 instead of the conventional TiO2 has been investigated.[28] However, the record performance for PSCs has always been reported with a TiO2 mesoporous layer.[7−11] Here, we propose further enhancement of the TiO2 mesoporous scaffold by the introduction of SiO2 NPs on it.

In this study, we synthesized SiO2 NPs, incorporating them within the mesoporous TiO2 layer to fabricate PSCs. The cell structure was fluorine-doped tin oxide (FTO)/compact TiO2/TiO2 mesoporous containing SiO2/CH3NH3PbI3/spiro-OMeTAD/Au (Figure ). Two different sizes of SiO2 NPs, 50 and 100 nm, and different concentrations within the mesoporous layer were tested to exploit the effect of these NPs in the solar cell performance.

Figure 1

Schematic of the PSC structure where the TiO2 mesoporous layer is modified with SiO2 NPs. Compact TiO2 and spiro-OMeTAD as electron and hole selective contacts, respectively.

Experimental Procedure

Synthesis of SiO2 NPs

The sol–gel method was used for the synthesis of SiO2 NPs with diameters of 50 and 100 nm by hydrolyzing tetraethylorthosilicate (TEOS) in a mixture of ethanol, water, ammonia, and a surfactant. TEOS, ethanol, and ammonia were used as the silica precursor, the common solvent, and the catalyst, respectively. Typically, 5.5 mL of ethanol, 11 mL of deionized water, and 0.8 mL of ammonia were mixed and stirred at 30 °C (for 100 nm SiO2 NPs) or 40 °C (for 50 nm SiO2 NPs). At the same time, a solution containing 5.5 mL of ethanol and 1.7 mL of TEOS was prepared and stirred at 30 °C (for 100 nm SiO2 NPs) or 40 °C (for 50 nm SiO2 NPs). Then, the latter was added to the first solution, and the reaction mixture was maintained at 30 or 40 °C (depending on the desired particle size) for 2 h. During this time, the reaction turned turbid slowly because of the formation of silica NPs. Once the reaction was completed, 3-amino-propyltriethoxysilane was added dropwise to stabilize the particles. The resulting solution was heated at 80–90 °C to remove the excess of ammonia. The precipitate was collected by centrifugation and washed several times with water and ethanol; then, it was dried overnight at 100 °C. Figures S1 and S2 show the transmission electron microscopy (TEM) images of the SiO2 NPs with 50 and 100 nm sizes, respectively.

Cell Fabrication

PSCs were fabricated in a structure based on FTO glass substrates, which were partially etched using HCl and zinc powder, cleaned with soap and deionized water, sonicated in a mixture of acetone/ethanol and then in acetone/isopropyl alcohol, and dried with compressed air. The substrates were then treated by an ultraviolet–ozone lamp for 15 min. To prepare the TiO2 blocking layer (bl-TiO2), titanium diisopropoxide dis(acetylacetonate) in ethanol (1:9, v/v) was deposited by spray pyrolysis at 450 °C, and then the film was annealed at 450 °C for 30 min in air. SiO2 (0.4 mg) powder in 10 mL of ethanol was stirred for 1 night. Different concentrations of SiO2 solution (0, 10, 20, and 50 μL) were combined with a dilute commercial TiO2 paste (30NRT) in ethanol (1:5 weight ratio). Then, the SiO2/TiO2 (mp-TiO2) layer was deposited on the bl-TiO2 layer by spin-coating at 2000 rpm for 10 s. After drying at 100 lC for 10 min, it was annealed in air up to 500 aC for 30 min. Lithium bis-(strifluoromethanesulfonamide) (Li-TFSI, 35 mM) solution in acetonitrile was deposited on the substrates by spin-coating at 3000 rpm for 10 s. This Li-TFSI deposition has been performed for samples using SiO2 with 50 nm, but not for 100 nm, to rule out the influence of treatment instead of NP incorporation. Then, the substrates were annealed at 450 PC for 30 min.[29] PbI2 (622 mg) and 105 mg of dimethyl sulfoxide were mixed in 944 mg of dimethylformamide (DMF) solution inside a glovebox. This solution was heated at 65 oC and then mixed into 214 mg of CH3NH3I. The completely dissolved solution was spin-coated on the substrate layer at 4000 rpm for 50 s. Also, on the rotating substrate, diethyl ether was dripped, which caused rapid vaporization of DMF. Then, the deposited film was heated at 100 wC to obtain a dense CH3NH3PbI3 film. Then, 72.3 mg of spiro-OMeTAD, 28.8 μL of 4-tert-butyl pyridine, and 17.5 μL of Li-TFSI solution (520 mg/mL of Li-TFSI in acetonitrile) were dissolved in 1 mL of chlorobenzene and deposited on the perovskite layer as a hole-transfer layer by spin-coating at 4000 rpm for 30 s. Finally, the electrode contacts (60 nm Au) were deposited on the cells by thermal evaporation in a high vacuum chamber.

Characterization

The current–voltage (J–V) curves were obtained with a scan rate of 50 mV/s in an Abet Technologies Sun 2000 Class A AM 1.5G solar simulator with a Keithley 2612 Source Meter, where the light intensity was adjusted with an NREL-calibrated Si solar cell with a KG-5 filter to 1 sun of intensity (100 mW/cm2) in the Institute of Advanced Materials (INAM) at Universitat Jaume I. The measurements were performed using a shadow mask with an area of 0.101 cm2. Ultraviolet–visible absorption spectra were recorded by using a Cary 500 Scan VARIAN spectrophotometer in the 250–800 nm wavelength range. Impedance spectroscopy measurements were performed using an Autolab PGSTAT-30 instrument equipped with a frequency analyzer module at 0.1 sun light illumination. The dc bias was selected at 0 V, and the ac perturbation was 60 mV covering the frequency range of 0.1 Hz to 1 MHz. The integration time was 0.125 s, and the number of cycles was equal to one.

Results and Discussion

To demonstrate the effect of SiO2 NPs on the PSC performance, the fabricated cells were tested (with different concentrations and sizes of SiO2) under 1 sun condition in air. Figures and 3 show the dependence of the average device performance parameters when the concentration of SiO2 NPs, with 100 and 50 nm sizes, respectively, varies from 0 to 10 and 50 μL. To ensure reliability, at least 15 devices were made for each concentration, but usually more devices are prepared in a single condition. We have analyzed several batches at each condition, and two operators prepared the devices to avoid dispersion from batch to batch, independent of the skill of a single operator. No cell was ruled out of the analysis despite some batches presenting a significant lower performance than the average. This procedure implies the preparation of a large amount of samples, as we have prepared in this work, to have a confidence interval as high as 90%, where no overlap is observed between the intervals for the samples having the highest PCE and the reference, as we show in Figures S3 and S4, making the study statistically meaningful. The boxplot featuring the short-circuit current density (Jsc), open circuit potential (Voc), fill factor (FF), and PCE of all of the cells with different SiO2 NP concentrations are compared in Figures and 3. Clearly, the addition of SiO2 NPs demonstrates an increase in the Jsc and hence in the PCE. In particular, the highest PCE for samples containing 50 and 100 nm size SiO2 NPs embedded in the TiO2 mesoporous scaffold was achieved by using 20 and 10 μL solutions, respectively. No significant variation in light absorbance was observed after the addition of SiO2 NPs, see Figure S5.

Figure 2

Boxplot showing the minimum, quartile 1, mean, median, quartile 3, and maximum for the photovoltaic performance of cells using different concentrations of SiO2 NPs of 100 nm size. (a) Jsc, (b) Voc, (c) FF, and (d) PCE. Concentration is indicated in μL of a solution of 0.4 mg of SiO2 powder in 10 mL of ethanol.

Figure 3

Boxplot showing the minimum, quartile 1, mean, median, quartile 3, and maximum for the photovoltaic performance of cells using different concentrations of SiO2 NPs of 50 nm size. (a) Jsc, (b) Voc, (c) FF, and (d) PCE. Concentration is indicated in μL of a solution of 0.4 mg of SiO2 powder in 10 mL of ethanol.

The values of the photovoltaic parameters for the cells with the highest PCE with different SiO2 NP concentrations and sizes are summarized in Table . Again, the cells with NPs exhibit higher photovoltaic parameters. Clearly, the observed PCE enhancement for the device incorporating SiO2 NPs mainly comes from the great improvement of Jsc because the Voc and FF exhibit only a slight change. The maximum PCE of the devices without NPs reaches 15.04%, whereas the maximum PCE of those incorporating the NPs reaches 16.51 and 16.71% for sizes of 50 and 100 nm, respectively. The J–V curves of these champion cells are plotted in Figure S6.

Table 1

Values of Jsc, Voc, FF, and PCE for Champion PSCs with Different Concentrations and Sizes of SiO2 NPs under the Irradiation of 1 sun Intensity (100 mW/cm2; AM 1.5G)

size (nm)	SiO2 (μL)	Jsc (mA/cm2)	Voc (V)	FF (%)	PCE (%)
50	0	19.35	1.055	73	15.04
	10	20.03	1.065	74	15.99
	20	20.55	1.061	75	16.51
	50	19.72	1.052	72	15.15
100	0	20.06	1.006	72.47	14.63
	10	22.59	1.023	72.32	16.71
	20	20.17	0.998	69.94	14.09
	50	20.38	1.033	74.29	15.64

To highlight the cause of the observed enhancement of PCE performance, the optical properties of the mesoporous scaffold have been theoretically analyzed to calculate the effect of incorporation of SiO2 NPs of different sizes. The obtained results are consistent with the scattering of SiO2 NPs inside the TiO2 scaffold calculated by the Mie theory.[30] First, the scattering cross-section (σscattering) of a SiO2 NP immersed in the TiO2 bulk matrix induces a clear scattering enhancement by increasing the size (radius) of the NP, as illustrated in Figure a. This behavior agrees with previous calculations carried out with metal NPs in polymer media[15] and can qualitatively explain the enhancement of the solar cell efficiency when 100 nm size SiO2 NPs are used inside the TiO2 scaffold. Moreover, the dependence of the scattering as a function of the filling factor (FF, volume of SiO2 inside the TiO2 matrix) will exhibit an optimum concentration, as presented in Figure b for a size of 100 nm. Here, it was important to take into account the mesoporous nature of TiO2: it seems that for low concentrations, SiO2 is totally covered by TiO2, whereas for a high FF, the NPs “can see” the influence of the air present in the porous layer, and an effective index medium has to be considered.[15] Under these conditions, σscattering decreases for large concentrations, where the effective refractive index of the matrix will approach that of SiO2. Nevertheless, because the total scattering is proportional to the number of NPs (and hence, FF), there is an optimum concentration that maximizes the scattering of light and, with it, the efficiency of the solar cell (see Figure b). In addition, other effects such as the improvement of the contact between the perovskite and the mesoporous layer by the presence of large size SiO2 NPs can also contribute for enhancing the PSC performance.

Figure 4

(a) Integrated scattering cross-section as a function of the size of the SiO2 NP inside the TiO2 matrix. (b) Integrated scattering cross-section per FF as a function of FF. Size was fixed to 100 nm.

On the other hand, the enhancement on cell performance observed when 100 nm NPs are used instead of 50 nm NPs is significantly more moderate than that expected from the theoretical analysis, as highlighted in Figure a. Here, it is worthy to note that the thickness of the scaffold, ∼200 nm, is of the same order as the SiO2 NP size, and consequently an excessive increase of the NP concentration would affect seriously the morphology of this layer and even light transmission, effects that were not considered in Figure .

The PCE enhancement achieved, according to Table , was 9 and 14% for NPs 50 and 100 nm in diameter, respectively, which can be explained as a consequence of the light-trapping effect of the NPs. These NPs act like mirrors inside of the mesoporous layer; the sunlight is scattered on it, enabling a longer optical path length, and thereby a higher amount of light is absorbed by the perovskite. One of the most interesting characteristics of the method proposed here to enhance the PCE is the fact that this TiO2–SiO2 nanocomposite is obtained through a simple mixing step, indicating the feasibility of the proposed method easily upscalable. There are other light-trapping mechanisms, as nanotextures,[31] that theoretically can achieve a higher enhancement, but require a complicated fabrication method that is difficult to transfer to industrial applications.

It is well-known that the mesoporous scaffold also plays an important role in the recombination and decrease of hysteresis.[8,32] In addition to the increase of PCE with the incorporation of SiO2, a significant decrease in the J–V hysteresis is observed, see Figure  and Table . The forward and reverse sweeps of the J–V curve for reference devices with no SiO2 and for devices with SiO2 NPs of 50 nm are summarized in Table S1. The hysteresis index (HI) has been quantified by calculating as follows[33]where JRev(Voc/2) and JFor(Voc/2) are the currents when half of the Voc voltage is applied for reverse (from Voc to zero) and forward (from zero to Voc) voltage scans, respectively. A HI of 0 corresponds to a cell without significant hysteresis, whereas a HI of 1 corresponds to a system in which the hysteresis is as high as the photocurrent.

Figure 5

J–V hysteresis of champion PSCs with SiO2 NPs (50 nm) and without NPs (ref.). RS, reverse scan from Voc to zero. FS, forward scan from zero to Voc.

Table 2

Hysteresis Factor Extracted from the Data Using the Definition Provided in Eq

SiO2 (μL)	0	10	20	30
HI	0.102	0.032	0.018	0.024

Different aspects contribute to hysteresis in PSCs, among them, ion migration and accumulation at perovskite interfaces is one of the most influencing ones.[34] One easy way to weigh the effect of ion accumulation is by open circuit voltage decay (OCVD) measurements.[35] Light soaking produces an accumulation of ions at the interface. This charge accumulation causes an electrostatic potential adding its value to the built-in potential from the Voc. After switching off the illumination in OCVD measurements, two regimes are observed. First, a fast decrease is observed as the built-in potential induced by carrier photogeneration is quickly removed. For longer times, a significantly slower decay is observed as the elimination of the electrostatic potential is related with the slow ion migration that removes the ion accumulation. OCVD measurements for PSCs with and without SiO2 NPs are plotted in Figure . Although both kinds of samples present the same behavior in the fast decay regime, a clear difference is observed in the slow decay regime. Samples with SiO2 NPs are able to remove the electrostatic potential induced by ion accumulation faster than reference cells, pointing to a lower ion accumulation and/or a faster dynamic in samples with SiO2 NPs. In this sense, the addition of SiO2 NPs is not just affecting the optical properties of the mesoporous scaffold but also affecting the ion migration properties of the perovskite deposited in it. Alternatively, ion accumulation was also monitored by the low-frequency capacitance.[32,36−38] Samples with SiO2 NPs present lower low-frequency capacitance, see Figure S7, pointing to a lower ion accumulation in good agreement with the OCVD measurements.

Figure 6

OCVD for cells with and without NPs (50 nm SiO2).

Conclusions

SiO2 NPs with 50 and 100 nm sizes have been integrated into a conventional mesoporous TiO2 layer of the PSCs, observing an enhancement of cell performance up to 14%, in comparison with reference samples, mainly because of an increase in the photocurrent. These results are in good agreement with the theoretical predictions based on the increase of light scattering induced by the presence of SiO2. Light scattering is favored by the increase of the NP size. However, as the size that is able to produce significant scattering is comparable with the optimum mesoporous layer thickness, other aspects such as the layer morphology, the loss in transparency, and the reduction of the effective refractive index have to be properly balanced to optimize the SiO2 concentration in the mesoporous scaffold. The addition of SiO2 NPs also affects the ion accumulation at the perovskite interfaces, causing an additional beneficial effect, reducing the cell hysteresis significantly. Here, we show a method, easy to implement, to increase the PSC performance by the modification of the mesoporous TiO2 scaffold with the addition of dielectric SiO2 NPs, pointing concretely to the development of new mesoporous layers as a way to further increase the performance of PSCs. Optimization of optical properties has allowed the latest improvement in performance in very high-efficient GaAs solar cells and has to be necessarily considered as the reported efficiency of the photovoltaic technology is approaching the Shockley–Queisser limit, as it is the case for PSCs.