Efficiency enhancement of perovskite solar cells using structural and morphological improvement of CH₃NH₃PbI₃ absorber layers

In the recent years, perovskite solar cells have been one of the most interesting scopes for researchers, due to their low cost and easy process ability, as evidenced by their unprecedented cell efficiency progress during the past several years. The organolead halide perovskite was first used in 2009 by Miyasaka in dye-sensitized solar cells as a dye [1]. The organometal halides have the general formula ABX₃, where A is an organic cation (e.g. ${{\rm{CH}}}_{3}{{{\rm{NH}}}_{3}}^{+},{\rm{HC}}{({{\rm{NH}}}_{2})}_{2}{}^{+}$), B is a metallic ion (e.g. Pb²⁺, Sn²⁺), and X is a site occupied by a halogen ion (e.g. I⁻, Cl⁻, Br⁻) [2]. Perovskite CH₃NH₃PbI₃ is a direct band gap (E_g) semi-conductor (E_g ∼ 1.55 eV at room temperature), which together with its unique properties such as a large diffusion length (∼0.1–1 μm), high mobility of the charge carriers (∼5 cm² V⁻¹ s⁻¹), high optical absorption (∼10⁵ cm⁻¹), low cost, and simple processing deposition make it a suitable candidate for use as an absorber layer in the solar cells [3–5]. In order to use perovskite structures as the absorber layers in solar cells, an ideal morphology is a uniform and pinhole free layer, which prevents electron–hole recombination and amplifies photo-current. Thus the morphology, crystallinity, compactness, and composition of layers are required to be optimized. Several groups have shown that the perovskite morphologies such as the grain size and crystallinity highly affect the performance of solar cells [6, 7]. Highly efficient perovskite solar cells have been obtained using the perovskite grains as large as possible [8].

Moreover, the optical and electrical properties of the perovskite structures and, therefore, solar cell functionality, significantly depend upon the deposition conditions (e.g. molar ratio of precursors). Up to the present time, several researchers have tried to overcome these challenges using various chemical and physical deposition methods. The CH₃NH₃PbI₃ layers can be prepared by one-step or two-step solution methods [9–12]. The two-step sequential solution method is a simple and more controllable deposition technique for fabricating highly efficient CH₃NH₃PbI₃ solar cells [13]. In this procedure, first, a PbI₂ solution is deposited on a substrate, and then by introducing methylammonium iodine (MAI) to isopropanol (IPA) and reaction with MAI, CH₃NH₃PbI₃ is formed. Although in the two-step method sometimes some unconverted PbI₂ remains in the system, the perovskite uniformity and grain size are more controllable [11]. Park et al have investigated the effect of MAI concentration on the morphology of perovskite grains to obtain highly efficient perovskite solar cells [14].

Further suitable electron selective materials such as TiO₂, ZnO, and CdSe quantum dot/phenyl-C61 butyric acid methyl ester (PCBM) composite and appropriate hole selective materials such as Spiro-MeOTAD and Poly3hexylthiophene (P3HT) are required in order to have a good extraction of hole and electron generated carriers in the perovskite (PSK) layer which can lead to improvement of PSK solar cells efficiency [15].

In this work, a two-step spin coating method was used for the preparation of the CH₃NH₃PbI₃ perovskite layers. Then the effect of MAI concentration on the structural and optical properties of the perovskite film and, therefore, power conversion efficiency (PCE) of the fabricated solar cell are reported.

For solar cell fabrication, fluorine-doped tin oxide (FTO) substrates were patterned by an HCl solution (2 M in deionized water) and Zn powder. Then the etched substrates were ultra-sonicated in a detergent, deionized water, acetone, ethanol, and isopropanol, each for 10 min, followed by drying at 100 °C, and subsequently treated by UV-ozone for 15 min. The cleaned FTO substrates were coated with a compact layer of TiO₂ by spin coating of a mild acidic solution of tetraisopropyl orthotitanate in ethanol (2000 RPM for 30 s), followed by annealing at 500 °C for 30 min, treating with a 0.02 M TiCl₄ solution in deionized water at 70 °C for 30 min, washing with deionized water, and drying at 100 °C. Then the substrates were annealed at 500 °C for 30 min. The mesoporous TiO₂ layer was deposited on the TiO₂ compact layer by spin coating (5000 RPM for 30 s) of the TiO₂ paste containing 20 nm sized TiO₂ nanoparticles diluted in ethanol (2:7 weight ratio), and immediately dried at 70 °C for 30 min. According to the literature, the efficient thickness of the mesoporous TiO₂ layer is in the range of 400–600 nm [16]. By controlling the spin coating conditions and TiO₂ nanoparticle precursor concentration, an efficient film thickness of 400 nm was obtained.

Similar to the compact layer, this layer was also annealed at 500 °C for 30 min. After mesoporous TiO₂ deposition, it was treated with TiCl₄ solution, as described for the compact layer. Then the CH₃NH₃PbI₃ perovskite layer was deposited on the mesoporous TiO₂ layer. A poly3-hexylthiophene)P3HT) layer (10 mg ml⁻¹ in anhydrous chlorobenzene) was spin-coated on the perovskite layer at 1500 RPM for 40 s. Finally, a 100 nm thick Au cathode was deposited on it by a thermal evaporator with a shadow mask (Nanostructured Coatings Co, Iran).

In order to make the CH₃NH₃PbI₃ absorber layer, the PbI₂ precursor solution (1 M in anhydrous dimethylformamide (DMF)) was deposited on the mesoporous TiO₂ substrate and also separately on a glass substrate for the perovskite layer characterization. Based on the efficient conditions introduced in the literature, this process was performed by a two-step spin coating at 3000 RPM for 5 s and 6500 RPM for 5 s, followed by heating at 40 °C for 3 min, and 100 °C for 5 min [17]. After thermal treatment of the layers, the CH₃NH₃I solution in anhydrous isopropanol alcohol (IPA) with various concentrations (7, 10, 20 and 40 mg ml⁻¹) was dropped on the samples and rested for 20 s, then spun at 4000 RPM for 20 s, and subsequently heated at 100 °C for 5 min. The CH₃NH₃PbI₃ layers with different MAI concentrations of 7, 10, 20 and 40 mg ml⁻¹ were named as S₁, S₂, S₃, and S₄.

The structure of the layers was investigated using x-ray diffraction (XRD, Cu kα radiation, λ = 1.5406 Å, Bruker, AXS). The morphology and surface roughness of the samples were examined by a scanning electron microscope (SEM, VEGA//HITACHI S-4160 AND TESCAN) and an atomic force microscope (AFM, Arapasohesh, Iran). The optical properties of the samples were studied by UV–visible spectroscopy (1800 UV, Shimadzu).

The density current–voltage (J–V) characteristic of the fabricated solar cells based on the CH₃NH₃PbI₃ absorber layer was measured using an Ivium Stat potentiostat (model: XRE, Netherlands) under a calibrated AM 1.5 solar simulator with a light intensity of 100 mW cm⁻² (Sharif Solar 10-2, Iran). The thickness of the layers measured from the cross-sectional FESEM images and listed in table 1. To study the charge carrier recombination dynamics, the open circuit voltage decay analysis technique was used.

The XRD patterns for the CH₃NH₃PbI₃ perovskite samples (S₁, S₂, S₃, and S₄) are shown in figure 1. The observed diffraction peaks at 14.09°, 19.99°, 24.51°, 28.47°, 31.87°, 35.04°, 38.5°, 40.76° and 43.41° were assigned to the (110), (112), (202), (220), (310), (312), (322), (224) and (314) planes of the tetragonal perovskite structure [18]. The unconverted PbI₂ peak can also be observed at 12.6° in the patterns; its intensity decreased as the concentration of MAI solution was increased. For the sample with a high concentration of MAI solution (S₄), the observed PbI₂ peak was very weak, which indicated that almost all the PbI₂ were converted to the CH₃NH₃PbI₃ perovskite. From the most intense (110) and (220) diffraction peaks, the average crystallite size of the layers (D) was estimated using the Sherrer equation (equation (1)):

$$\begin{eqnarray}&&D=\displaystyle \frac{0.89\,\lambda \,}{\beta \,\cos \,\theta }.\end{eqnarray} \tag{ 1 }$$

In this equation, λ is the x-ray wavelength, β is the full width at half-maximum of the diffraction peak in radians, and $\theta$ is the Bragg's angle.

Zoom In Zoom Out Reset image size

Based on equations (2)–(4), the lattice parameters and unit cell volume of the perovskite layers were calculated using the (110) and (220) diffraction peaks:

$$\begin{eqnarray}&&n\,\lambda =2d\,\sin \,\theta \end{eqnarray} \tag{ 2 }$$

in which n is the order of diffraction, λ is the x-ray wavelength, d is the lattice inter-planar spacing of the crystal, and $\theta$ is the Bragg's angle

$$\begin{eqnarray}&&\displaystyle \frac{1}{{{d}_{hkl}}^{2}}=\displaystyle \frac{({h}^{2}+{k}^{2})}{{a}^{2}}+\displaystyle \frac{{l}^{2}}{{c}^{2}},\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&V={a}^{2}\times c.\end{eqnarray} \tag{ 4 }$$

In equation (3), h, k, and l are the miller indices of the peaks, and a and c are the lattice parameters of the unit cell. The calculated results were summarized in table 1.

The results obtained show that although there is a PbI₂ peak in sample S₁, this sample has a better crystallinity than the other ones.

In order to evaluate the morphological structure of the samples, FESEM images were recorded for all samples and shown in figure 2. As it is obvious in these figures, particles are agglomerated with rather uniform distribution in all samples. Samples S₂, S₃, and S₄ have nearly the same particle shapes and a uniform morphology. On the other hand, the S₁ sample has a higher particle size and roughness. By comparison of the figures, we can see that with increase in the MAI concentration to more than 7 mg ml⁻¹, the grain sizes significantly decrease from the average value of around 600 nm for the S₁ sample to 150, 250 and 280 nm for the S₂, S₃ and S₄, samples, respectively.

Zoom In Zoom Out Reset image size

These results have demonstrated that proper concentration of MAI solution leads to the larger grain sizes and therefore, better crystal quality. Such a crystal grain size synthesized with the MAI concentration solution of 7 mg ml^–1 may explained with a typical Ostwald ripening phenomenon. The MAI solution with low concentration (S₁) could likely induce the Ostwald ripening mechanism for island coalescence during crystal growth [19]. In fact, due to atoms leave small islands more readily than larger islands, the newly germinated nuclei, which have radius less than the critical radius, are not much stable and may dissolve in the solution and accumulate over larger nuclei, which lead to the formation of larger grains. With increase in the MAI concentration, the Gibbs free energy could decrease, and the driving force for the nucleation increases. This phenomenon may lead to the formation of a critical nucleus with a small size. As a result, with a higher concentration of MAI solution (S₂–S₄ samples), a film with smaller grain size and lower porosity will form. In fact, due to for the S₁ sample the growth was run by the low MAI concentration, when MAI solution dropped on PbI₂ film and rested for 20 s (before spin), it was diffused in the PbI₂ layer and the nuclei was sparsely distributed (localized nucleation) and then the larger grain of CH₃NH₃PbI₃ is formed. But for the samples prepared by higher MAI concentration (S₂, S₃ and S₄ samples), further crystal growth is inhibited, because nucleation and crystal growth are terminated immediately after adding MAI solution to PbI₂ layer. By increasing MAI concentration from S₁ to S₄ samples, the nuclei density raise and then the growth of grain sizes decrease [6]. In figure 2, the insets show the cross-sectional image of the samples for estimating the average thickness of the layers.

However, it seems that an increase in the grain size for an MAI concentration higher than 10 mg ml⁻¹ (S₄ sample) is related to the other dominant mechanisms. On the other hand, we propose that with increase in the MAI concentration (by more than 10 mg ml⁻¹), diffusion paths through the grain boundaries can form inside the PbI₂ layer, which can cause more MAI molecules to reach the underlying PbI₂ layer, and a more portion of PbI₂ is converted to the CH₃NH₃PbI₃ perovskite structure. The XRD patterns for the S₄ sample (figure 1) is an evidence for this discussion that there is a very weak PbI₂ peak indicating that almost the whole PbI₂ is converted to the CH₃NH₃PbI₃ perovskite structure.

The 2D and 3D topographic AFM images of the S₁ and S₄ samples are shown in figure 3. These images confirm the presence of pinholes in the perovskite films with a low MAI concentration (S₁ sample). The surface roughness profile of the samples shows that the sample with a higher MAI concentration (S₄) has a lower roughness.

Zoom In Zoom Out Reset image size

In order to investigate the optical properties of the perovskite thin films, the transmission spectra were measured in the range of 400–1100 nm and illustrated in figure 4(a). As shown in this figure, the S₂ and S₄ samples show the maximum and minimum transmission in the measured wavelength range, respectively. The higher/lower transmission of the S₂/S₄ sample could be due to the lower/higher thickness of this sample in comparison with the S₃ and S₄ samples. Although the S₁ sample has the lowest thickness compared with the other samples, values of its transmission are lower than the S₂ and S₃ samples and higher than the S₄ sample (table 1). This means that for this sample, other reasons such as grain size (d), morphology and surface roughness (higher porosity and roughness, see figures 2 and 3) should also be considered for this effect [20]. According to Rayleigh scattering (S) equation (5) that established for such low grain sizes in a few hundred nanometers ranges (compared to the visible light wavelength, $\lambda$):

$$\begin{eqnarray}&&S=c\left(\displaystyle \frac{{d}^{3}}{{\lambda }^{4}}\right)\,(c\,{\rm{i}}{\rm{s}}\,{\rm{a}}\,{\rm{c}}{\rm{o}}{\rm{n}}{\rm{s}}{\rm{t}}{\rm{a}}{\rm{n}}{\rm{t}}).\end{eqnarray} \tag{ 5 }$$

Zoom In Zoom Out Reset image size

We expected the S₁ sample (higher grain size than other samples) has lower transmittance. But transmittance of the S₁ sample is higher than S₄ sample which is likely because of much lower thickness of this sample than the S₄ sample that may overcome the Rayleigh scattering phenomena [20–22].

Figure 4(b) shows the variation in the absorption coefficient (α) with photon energy for the perovskite thin films. The absorption coefficients were obtained from equation (6):

$$\begin{eqnarray}&&\alpha =2.303\displaystyle \frac{A}{t},\end{eqnarray} \tag{ 6 }$$

where A is the absorbance and t is the thickness of the layer. As expected, the values for the absorption coefficients of the samples were high. The absorption coefficient exceeded 2 × 10⁵ cm⁻¹ at higher photon energies. Accordingly, consistent with the transmittance spectra, S₁ sample had the highest absorption coefficient, which is likely to be due to the larger grain sizes (lower carrier recombination) among these samples [20, 23]. Since stronger light absorption, and subsequently, more photo-generated carriers can be achieved from films contain larger grains [20]. To calculate the direct bandgap of the samples, the Tauc's equation was used:

$$\begin{eqnarray}&&\alpha h\upsilon =A{(h\upsilon -{E}_{{\rm{g}}})}^{\displaystyle \frac{1}{2}},\end{eqnarray} \tag{ 7 }$$

where h is the Planck's constant, hυ is the incident photon energy, and A is an energy-independent constant. Then the direct band gaps of the layers were determined based on the extrapolated liner regression of ${(\alpha h\nu )}^{2}-h\nu$ curves (figure 4(c)). The values obtained for the band gaps are listed in table 2. Although the absorption edge of the samples shows a little shift, the band gaps of all samples are nearly the same (about 1.550 eV), which is consistent with the reported values [24].

In order to investigate the effects of different perovskite thin films synthesized by various MAI concentrations on the solar cell performance, four solar cells including the CH₃NH₃PbI₃ absorber layers were fabricated. Figure 5 shows the schematic mesoporous device structure, cross-sectional image, and its energy band diagram. The TiO₂ blocking layer (bl-TiO₂) covers the etched FTO substrate. A uniform mesoporous TiO₂ layer with a thickness of about 400 nm is deposited on bl-TiO₂, and it is smoothly covered by a perovskite layer with a thickness of about 170–480 nm and then coated with P3HT as the hole transport layer (HTL). Then the Au cathode with a thickness of about 100 nm is evaporated on HTL.

Zoom In Zoom Out Reset image size

Figure 5(c) exhibits the energy levels of each device component and extraction of free charge carriers generated in the perovskite absorber layer under illumination by the electron and hole selective material.

The current density–voltage (J–V) characteristics curves of fabricated solar cells are shown in figure 6. The parameters extracted from the J–V curves such as PCE, fill factor (FF), short circuit-current density (J_sc), series resistance (R_s), shunt resistance (R_sh), and open-circuit voltage (V_oc) were summarized in table 2. PCE and FF were obtained from equations (8) and (9):

$$\begin{eqnarray}&&{\rm{P}}{\rm{C}}{\rm{E}}( \% )=\displaystyle \frac{{V}_{{\rm{\max }}}\times {J}_{{\rm{\max }}}}{{P}_{{\rm{i}}{\rm{n}}}}\times 100,\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&{\rm{F}}{\rm{F}}=\displaystyle \frac{{V}_{{\rm{\max }}}\times {J}_{{\rm{\max }}}}{{V}_{{\rm{o}}{\rm{c}}}\times {J}_{{\rm{s}}{\rm{c}}}}.\end{eqnarray} \tag{ 9 }$$

Zoom In Zoom Out Reset image size

R_s and R_sh were measured from the slope of the J–V curves of solar cells at the J_sc and V_oc points of the curves. The best photoelectrical parameters were recorded for the solar cell fabricated to the S₁ absorber layer; they were consistent with the absorption coefficient (figure 4(b)). The results obtained show that the solar cell prepared from the absorber layer with a low MAI concentration (S₁) has the highest efficiency of 11.6%. This may be attributed to the large grain size and appropriate cover of the mesoporous TiO₂ layer, which prevent shunt path and efficient recombination of charge carriers in the layer, and also the beneficial effect of some residual PbI₂ in the absorber layer of the device S₁ [25]. As reported, the presence of excess PbI₂ can improve crystallization of the perovskite layer by reducing the lattice strain [26]. As explained, the S₁ sample has a significantly larger grain size, which causes the reduction of grain boundaries, leading to a lower carrier recombination rate and a higher PCE.

The solar cell fabricated with the S₂ layer has the worst photo-electrical parameters, which may be due to a high recombination carrier in this layer. Since this sample has the lowest grains and light absorption and the highest grain boundaries that lead to high recombination of charge carriers.

In order to study the charge carrier recombination dynamics in the fabricated devices with various MAI concentrations, the open-circuit voltage decay analysis was carried out. Figure 7 shows the normalized V_oc decay curves for each solar cell. As shown in this figure, a fast drop in the photo-voltage occurs in devices based on the S₂ and S₄ layers after turning the light off. Among these devices, photo-voltage of the S₁ device decay is slower than the other samples. The observed trend can be attributed to the presence of grain boundaries and defects. The defects in the crystal structure are generally expected to decrease the lifetime of the photo-generated charge carriers in the absorber as additional recombination pathways speed-up the loss of excess carriers [27]. Also the interfaces could introduce trap states for charge transport that influence the recombination dynamics. If there are a significant number of shallow trap states for both electrons and holes, long living charge carriers may be expected due to a slow de-trapping process. Therefore, the fast and slow voltage decay regime would correspond to the recombination of free charges and the recombination of the slowly de-trapped carriers, respectively. On the other hand, in contrast to the shallow trap states, if the number of deep trap states that act as recombination centers increases, the carrier lifetime decreases. Therefore, by comparing the results, for the solar cell based on the S₁ absorber layer, due to higher grain sizes, may shallow trap states are more active than in other devices. On the other hand, a fast drop in the photo-voltage occurs in devices based on the S₂ and S₄ layers (lower grain size) after turning the light off. This may be due to higher accumulation of charge carriers at the interface which may lead to rapid carrier recombination. Moreover, in devices based on the S₁ and S₃ layers, the electrostatic (ES) potential due to accumulation of slow mobile ion charges at both sides of the interface may cause dramatically long decay times. This additional ES potential for devices based on the S₂ and S₄ layers is significantly lower than that of devices based on the S₁ and S₃ samples as a result of the higher recombination rate. This ES goes beyond the natural built-in voltage of the interface under illumination and it somewhat compensates for the higher recombination during open-circuit conditions [28]. As a result of this phenomenon, we can see an additional photovoltage in photo-voltage decay curves which still has not reached zero during 14 s (figure 7). The higher open circuit voltage of devices based on the S₁ and S₃ samples (∼0.93 V) compared to that for other devices (0.56 and 0.86 V) could be due to the much higher additional ES in devices based on the S₁ and S₃ layers. However, one may note that from the given set of experiments, it is difficult to identify the real origin of the increased charge carrier lifetime for the fabricated solar cells.

Zoom In Zoom Out Reset image size

In this work, the morphological, structural, optical, and photo-voltaic properties of the perovskite layers and fabricated solar cell based on the perovskite layers [(FTO(400 nm)/C-TiO₂ (30 nm)/Meso-TiO₂ (400 nm)/CH₃NH₃PbI₃ (170–480 nm)/P3HT (30 nm)/Au (100 nm)] were studied. A two-step spin coating method was used for the preparation of the CH₃NH₃PbI₃ perovskite layers with various MAI concentrations (7, 10, 20 and 40 mg ml⁻¹). The band gaps of all samples were about 1.550 eV. Moreover, the results obtained showed that the sample prepared by a high concentration of precursor solutions (40 mg ml⁻¹) had a pure CH₃NH₃PbI₃ perovskite structure and a lower surface roughness. However, in application to the absorber layer in solar cell devices, the sample prepared with a low concentration of precursor solutions (7 mg ml⁻¹) had better photo-electrical properties. Because this sample has higher grain sizes and absorption that have significant effects on the photo-electrical properties of the solar cell device. The results obtained show that by controlling the deposition conditions, due to the reduction in charge recombination, PCE enhancement of the perovskite solar cell (up to 11.6%) is possible. In this work, a device with an absorber layer prepared from a low concentration of MAI solution (7 mg ml⁻¹) was introduced as an efficient structure for the mesoporous perovskite solar cell.

The authors gratefully acknowledge the financial support of the Shahrood University of Technology and the equipment services from the Nano Optoelectronic Laboratory (NOPL) of the Tarbiat Modares University.