Low-Temperature Processed TiOx/Zn1−xCdxS Nanocomposite for Efficient MAPbIxCl1−x Perovskite and PCDTBT:PC70BM Polymer Solar Cells

The majority of high-performance perovskite and polymer solar cells consist of a TiO2 electron transport layer (ETL) processed at a high temperature (>450 °C). Here, we demonstrate that low-temperature (80 °C) ETL thin film of TiOx:Zn1−xCdxS can be used as an effective ETL and its band energy can be tuned by varying the TiOx:Zn1−xCdxS ratio. At the optimal ratio of 50:50 (vol%), the MAPbIxCl1−x perovskite and PCBTBT:PC70BM polymer solar cells achieved 9.79% and 4.95%, respectively. Morphological and optoelectronic analyses showed that tailoring band edges and homogeneous distribution of the local surface charges could improve the solar cells efficiency by more than 2%. We proposed a plausible mechanism to rationalize the variation in morphology and band energy of the ETL.


Introduction
The ever-growing energy demand relies mainly on the combustion of fossil fuel, which continues to cause serious resource depletion and environmental pollution. Solar energy is a proven renewable energy that is environmentally friendly and free from regional restriction. Inorganic solar cells (i.e., Si-based, gallium arsenide, copper indium gallium selenide, cadmium telluride, etc.) offer high efficiency (more than 25%) and stability, however, the technology development is hindered by a sophisticated manufacturing process in addition to the handling of toxic materials [1]. While dye-sensitized, polymer and perovskite solar cells are the emerging photovoltaic devices that are both lightweight and low cost, they however have their own set of problems that inhibit their large-scale production as was highlighted in recent review papers [1][2][3]. To realize their practical applications, device stability is one of the critical factors that needs to be understood and overcome. Different approaches have been proposed for a more stable polymer [4][5][6], dye-sensitized and perovskite solar cells [7][8][9][10][11].
Interface engineering can be considered as the most crucial factor in order to achieve high performance and stable solar cells, because it shapes the pathways of photogenerated charge carriers from an active layer to metal electrodes. Both polymer solar cells and perovskite solar cells can be easily fabricated in n-i-p configuration where electrons are collected at the bottom electrode, or p-i-n configuration where electrons are collected at the top electrode, via simple solution processes. In terms of device performances, p-i-n devices often exhibit more reliable performances and better environmental stability. However, their efficiencies are slightly lower than that of n-i-p structures due to the barrier in Fermi levels at the interface between the electron transport layer and the metal electrode [12]. Significant efforts have been devoted to developing interfacial materials for better band alignment, defect passivation and charge separation. While ZnO [13,14], SnO 2 [15], Al 2 O 3 [16], MoO x [17], core-shell nanoparticles/structures [18], and organic compounds [3,19,20] have all been reported as materials for electron transport layer (ETL), TiO 2 remains the most widely used and essential ETL material for both polymer solar cells and perovskite solar cells owing to the ease of tuning its optical and electrical band gaps. Comprehensive reviews on ETL materials, properties and interface engineering have been reported in recent papers [12,21,22]. Different concepts of tailoring TiO 2 have been proposed and showed promising results including metal-doped TiO 2 [23], amino acid-modified TiO 2 [24], plasmon-mediated TiO 2 [25], or functionalized organic/polymer on TiO 2 [26,27].
Instead of doping TiO 2 , synthesizing ETL composite of two metal oxides with appropriate band edges is another approach in facilitating electron transport, offering better efficiency and minimizing recombination rate. ETL composites of SrTiO 3 /TiO 2 [28], BaTiO 3 /TiO 2 [29], and In 2 O 3 /TiO 2 [30] have been shown to improve the overall device efficiency. Nevertheless, it is critical to further optimize the suitable mixing ratio (molar ratio, volume ratio) in order to achieve high power conversion efficiency. Controlling film thickness and morphology are the main challenges in using nanocomposites as ETL. It is also unclear how work function would vary at nanometer level when two semiconducting metal oxides are mixed together and the influence of lateral inhomogeneity of work function on the device performance.
In addition, the formation of crystalline TiO 2 requires high temperature annealing (500 • C) which is not suitable for fabricating devices on flexible substrates, as well as for many other practical applications. Several groups have addressed this issue by developing low-temperature processes. Kim et al. found that the devices made of ultra-thin amorphous TiO x by atomic layer deposition are more stable than crystalline TiO 2 [31,32]. Recently, Deng et al. reported that high efficiency perovskite solar cells could be achieved using TiO x processed at room-temperature [33]. Another study by Wang et al. demonstrated that efficient perovskite devices made of ETL nanocomposites of graphene/TiO 2 could be processed at 150 • C [34].
Here, we show that low-temperature (80 • C) processed TiO x :Zn 1−x Cd x S (T:Z) can be used as an effective ETL in both p-i-n perovskite and polymer solar cells. Zn 1−x Cd x S was chosen due to the ease of synthesis and highly tunable bandgap by varying Zn/Cd molar ratios [35]. We examined performances of the p-i-n devices using different loadings of Zn 1−x Cd x S in TiO x and carried out electrochemical impedance spectroscopy, ultraviolet photoelectron spectroscopy (UPS) and scanning Kelvin probe microscopy (SKPM) measurements. The local morphology with corresponding work function obtained from SKPM can shed more light into the fundamental understanding of electron transport layer in solar cells. at room temperature for 30 min. The mixture was heated with 700 W microwave (Sharp Cooperation, Osaka, Japan) radiation for 3 min. Zn 1−x Cd x S nanopowder was filtered and rinsed with 2-butanol [36]. The nanopowder was dried at 250 • C for 2 h and calcined at 500 • C for 1 h. Suspension of 0.4 mg/mL Zn 1−x Cd x S in 2-butanol was used in all experiments.

Device Fabrication
Polymer and perovskite solar cells were fabricated on pre-patterned ITO glass (Luminescence Technology Corporation, Taipei, Taiwan, 25 × 25 mm 2 , 5 Ω/sq). Prior to the fabrication, the substrates were cleaned by ultrasonication under detergent, DI water, acetone, and isopropyl alcohol sequentially. The substrates were further exposed under 395 UV light for 20 min.
Convective deposition [39] or spin coating method was used for depositing of hole transporting, photoactive, and electron transporting layers. A cleaned glass microscope slide (75 × 25 × 1 mm 3 , Fisher PA, Charlotte, NC, USA) was used as a deposition blade and placed at 45 • with respect to the patterned ITO substrate. PEDOT:PSS was used as the hole transporting layer for both solar cells. 20 µL of PEDOT:PSS (100 nm) were coated twice by convective deposition at the speed of 3000 µm/s. The coated substrate was annealed on a hot plate at 120 • C for 30 min in the ambient environment (%RH 50-60). The polymer solar cell fabrication was continued under ambient conditions while the PEDOT:PSS coated substrates were transferred to a glove box for perovskite solar cell fabrication.

Characterization
Surface morphologies were characterized by atomic force microscope (NanoWizard 3 NanoScience, JPK Instruments, Berlin, Germany).). Bruker OSCM-PT-R3 AFM probes (Bruker Nano, CA, USA) were used. The solar cell performance was measured using 0.1 cm 2 mask under 100 mW/cm 2 and 1.5 AM filter (Abet Technologies, Milford, CT, USA). The light intensity was calibrated by 4 cm 2 solar cell reference (VLSI Standards, Mountain View, CA, USA). I-V measurements were conducted using source measurement unit (PXI-4130, National Instrument, Austin, TX, USA with 10 mV scan step from 1 V to −1 V. Incident photon-to-current efficiency IPCE was measured using QEPVSI-b (Newport Corporation, Irvine, CA, USA). Impedance spectroscopy and Mott-Shockey (M-S) measurements were carried out using CHI 660E electrochemical work station (CH Instruments, Bee Cave, TX, USA). For M-S measurement, a small AC voltage of 5 mV was applied under constant illumination to measure device impedance as a function of frequency (100 Hz to 1 MHz). UPS was carried out at the Synchrotron Light Research Institute.
EIS and Mott-Schotkky measurements were carried out to understand the recombination losses and charge transport properties upon mixing Zn 1−x Cd x S into TiO x . Figure 1c shows Nyquist plots for perovskite solar cells and polymer solar cells at different loading of Zn 1−x Cd x S. The high frequency intercept of the Nyquist plot attributed to the series resistance and the low-frequency intercept corresponds to the recombination resistance (R rec ). As R rec gets larger, lower leakage current and recombination occur at the surface. It was clearly observed that R rec increased for devices containing Zn 1−x Cd x S, which had lower recombination losses. In addition, the lifetime of free carriers (τ) could be calculated from the maximum frequency (f m ) obtained from Bode plot as shown in Figure 1d by The carrier lifetime increased from 46 µs to 68 µs and to 82 µs as Zn 1−x Cd x S content increased from 0 to 100%. It is also well known that the interfacial charge density is inversely proportional to the slope of the Mott-Schottky plots (Figure 1e) as follows: where C is capacitance, ε o is vacuum permittivity, ε r is dielectric constant of material, N is charge carrier density, A is electrode area, V is applied potential, V fb is flat band potential, k B is Boltzmann constant, T is absolute room temperature and e is elementary charge of electron. As the charge carrier density is smaller than 5 × 10 20 cm −3 [43], the difference in flat band potential is negligible. constant, T is absolute room temperature and e is elementary charge of electron. As the charge carrier density is smaller than 5 × 10 20 cm −3 [43], the difference in flat band potential is negligible. Assuming an equivalent dielectric constant for all devices under illumination, the slopes of T100, T75:Z25, T50:Z50, T25:Z75 and Z100 are −2.2 × 10 10 , −2.3 × 10 10 , −2.4 × 10 10 , −1.8 × 10 10 , −1.0 × 10 10 respectively, indicating the highest interfacial charge density and resulting in the highest current density.       To better understand the nature of the starting materials, we performed detailed characterizations of TiO x , Zn 1−x Cd x S and their nanocomposites. SEM images and EDS spectra ( Figure S3) to confirm that the prepared nanocomposites form uniform, smooth thin films and addition of Zn 1−x Cd x S in TiO x does not effect on the morphology of the films. TEM images ( Figure S4) reveal that the TiO x is in amorphous form with small amount of anatase crystals of TiO 2 while Zn 1−x Cd x S is fully crystalline with crystallite size of about 5 nm. UV-Vis spectra and Tauc plots are shown in Figure S5. The band gaps calculated from Tauc plot of T100 and Z100 thin films are 3.7 and 2.5 eV, respectively, which are in good agreement with the band gap of TiO x and Zn 1−x Cd x S reported in the literatures [33,35]. The band gap values obtained from highest slopes in the Tauc plot of T75:Z25, T50:Z50, T25:Z75 films share the same value of 3.47 eV, which is lower than that of the pure T100 film. The nanocomposite ETL films also exhibit minor slopes, indicating indirect transitions or existence of gap states [44]. The indirect band gap is shifted from 3.0 eV for T75:Z25 to 3.1 eV for T50:Z50 to 3.25 eV for T25:Z75.  To better understand the nature of the starting materials, we performed detailed characterizations of TiOx, Zn1−xCdxS and their nanocomposites. SEM images and EDS spectra ( Figure  S3) to confirm that the prepared nanocomposites form uniform, smooth thin films and addition of Zn1−xCdxS in TiOx does not effect on the morphology of the films. TEM images ( Figure S4) reveal that the TiOx is in amorphous form with small amount of anatase crystals of TiO2 while Zn1−xCdxS is fully crystalline with crystallite size of about 5 nm. UV-Vis spectra and Tauc plots are shown in Figure S5. The band gaps calculated from Tauc plot of T100 and Z100 thin films are 3.7 and 2.5 eV, respectively, which are in good agreement with the band gap of TiOx and Zn1−xCdxS reported in the literatures [33,35]. The band gap values obtained from highest slopes in the Tauc plot of T75:Z25, T50:Z50, T25:Z75 films share the same value of 3.47 eV, which is lower than that of the pure T100 film. The nanocomposite ETL films also exhibit minor slopes, indicating indirect transitions or existence of gap states [44]. The indirect band gap is shifted from 3.0 eV for T75:Z25 to 3.1 eV for T50:Z50 to 3.25 eV for T25:Z75.   Ultraviolet photoelectron spectroscopy (UPS) was used to characterize the work function and valence band energy of TiO x :Zn 1−x Cd x S electron transport layer. UPS spectra of 10 nm T:Z thin films coated on silicon substrates are shown in Figure 2a. The effective work functions of the films were obtained by subtracting the source energy (40.8 eV) by the energy difference (∆KE) between the low kinetic energy edge (LKE) and the spectrometer/sample Fermi level: Ultraviolet photoelectron spectroscopy (UPS) was used to characterize the work function and valence band energy of TiOx:Zn1−xCdxS electron transport layer. UPS spectra of 10 nm T:Z thin films coated on silicon substrates are shown in Figure 2a. The effective work functions of the films were obtained by subtracting the source energy (40.8 eV) by the energy difference (ΔKE) between the low kinetic energy edge (LKE) and the spectrometer/sample Fermi level: The valence band energy (EVB) was estimated by subtracting the source energy (40.8 eV) by the energy difference (ΔKE') between the LKE and the high kinetic energy (HKE) edge for photoemission as below: (4) The work functions of T100, T75:Z25, T50:Z50, T25:Z75 and Z100 thin films with respect to the vacuum level are 4.22, 4.00, 3.70, 4.08 and 3.60 eV, respectively (Table S1). The valence band energy of T100, T75:Z25, T50:Z50, T25:Z75 and Z100 thin films with respect to the vacuum level are −7.42, −7.20, −6.90, −7.28 and −7.80 eV, respectively. Figure 2b shows a proposed general band energy diagram of TiOx:Zn1−xCdxS nanocomposite electron transport layer. The band energy levels of ITO, PEDOT, perovskite/PCDTBT, PC70BM and Al were taken from other studies [12,32].
Scanning Kelvin probe microscopy (SKPM) is an effective technique to monitor simultaneously morphology/topography and contact potential difference (VCPD) between the atomic force microscopy tip and sample surface. When the AFM tip and sample is far apart and not connected electrically, their Fermi levels are different. As the AFM tip approaches the sample surface, an electrical force is induced between the tip and the sample surface, brought their Fermi levels to line-up and formed contact potential difference (CPD). To measure this CPD, an external bias is applied at the contact area until the surface potential difference between the tip and the sample becomes zero. The magnitude of the applied bias equals to the work function difference between the tip and the sample, which means the work function of the sample can be calculated once the work function of the tip is known. The measured CPD is defined as the difference in work function between a tip and a sample as follows: The valence band energy (E VB ) was estimated by subtracting the source energy (40.8 eV) by the energy difference (∆KE') between the LKE and the high kinetic energy (HKE) edge for photoemission as below: The work functions of T100, T75:Z25, T50:Z50, T25:Z75 and Z100 thin films with respect to the vacuum level are 4.22, 4.00, 3.70, 4.08 and 3.60 eV, respectively (Table S1). The valence band energy of T100, T75:Z25, T50:Z50, T25:Z75 and Z100 thin films with respect to the vacuum level are −7.42, −7.20, −6.90, −7.28 and −7.80 eV, respectively. Figure 2b shows a proposed general band energy diagram of TiO x :Zn 1−x Cd x S nanocomposite electron transport layer. The band energy levels of ITO, PEDOT, perovskite/PCDTBT, PC 70 BM and Al were taken from other studies [12,32].
Scanning Kelvin probe microscopy (SKPM) is an effective technique to monitor simultaneously morphology/topography and contact potential difference (V CPD ) between the atomic force microscopy tip and sample surface. When the AFM tip and sample is far apart and not connected electrically, their Fermi levels are different. As the AFM tip approaches the sample surface, an electrical force is induced between the tip and the sample surface, brought their Fermi levels to line-up and formed contact potential difference (CPD). To measure this CPD, an external bias is applied at the contact area until the surface potential difference between the tip and the sample becomes zero. The magnitude of the applied bias equals to the work function difference between the tip and the sample, which means the work function of the sample can be calculated once the work function of the tip is known. The measured CPD is defined as the difference in work function between a tip and a sample as follows: where V CPD is the CPD between the tip and the sample, ϕ tip and ϕ sample are the work functions of the tip and the sample, respectively and eis the elementary charge. To obtain the work function of the AFM tip, freshly cleaved highly oriented pyrolytic graphite (HOPG, ϕ = 4.6 eV) and thermally evaporated gold thin film (Au, ϕ = 5.1 eV) were used. For all measurements, the substrate was grounded. Topographical and potential maps of the pure TiO x , Zn 1−x Cd x S and mixed TiO x :Zn 1−x Cd x S thin films are shown in Figure 3. Three-dimensional maps combining topographic and V CPD data clearly show that the T50:Z50 film possesses the least variation in surface potential compared to other ETL films. Potential histograms of all samples are shown in Figure 4a. The work function of T50:Z50 is 3.75 ± 0.28 eV compared to 3.47 ± 0.44 eV of T100 and 3.77 ± 0.29V eV of Z100, which are comparable to UPS data. FWHM value of T50Z50 has minimum value, indicating the least variation in work function at nanoscale level. It is expected that the work function values of ETL composite films obtained from SKPM technique are slightly lower than those obtained from UPS method due to water adsorption and hydrocarbon contamination (Tables S1 and S2). Although the work function value of T100 film estimated from SKPM method is about 0.4 eV lower than that of UPS value, it is in good agreement with plasma enhanced atomic layer deposited TiO x films reported in Kim's study [32]. Figure 4b-f shows profiles at three different locations from SKPM image of T25:Z75 film, which indicates that the surface potential depends on the film roughness. As the tip scanned over a particle, the height abruptly increased about 5 nm, and the local V CPD decreased about 200 mV, leading to an increase at the local work function. The distribution of charges at the surface and potential step can be attributed to crystallographic orientation and atomic relaxation [45]. It is possible that electronegative character of Zn 1−x Cd x S results in a partial electron transfer with TiO x that leads to an increase in local work function [45]. When the height decreased about 3 nm, the local V CPD increased about 100 mV. Within 1-2 nm variation in height, the local charge distribution could be fluctuated about 50 mV. The overall results indicate that an effective ETL composites of mixed metal oxides should have root mean square roughness of about 0.5 nm and scanning probe microscopy can be used as an effective tool to predict the homogeneity of local charges.
where VCPD is the CPD between the tip and the sample, φtip and φsample are the work functions of the tip and the sample, respectively and eis the elementary charge. To obtain the work function of the AFM tip, freshly cleaved highly oriented pyrolytic graphite (HOPG, φ = 4.6 eV) and thermally evaporated gold thin film (Au, φ = 5.1 eV) were used. For all measurements, the substrate was grounded. Topographical and potential maps of the pure TiOx, Zn1−xCdxS and mixed TiOx:Zn1−xCdxS thin films are shown in Figure 3. Three-dimensional maps combining topographic and VCPD data clearly show that the T50:Z50 film possesses the least variation in surface potential compared to other ETL films. Potential histograms of all samples are shown in Figure 4a. The work function of T50:Z50 is 3.75 ± 0.28 eV compared to 3.47 ± 0.44 eV of T100 and 3.77 ± 0.29V eV of Z100, which are comparable to UPS data. FWHM value of T50Z50 has minimum value, indicating the least variation in work function at nanoscale level. It is expected that the work function values of ETL composite films obtained from SKPM technique are slightly lower than those obtained from UPS method due to water adsorption and hydrocarbon contamination (Tables S1 and S2). Although the work function value of T100 film estimated from SKPM method is about 0.4 eV lower than that of UPS value, it is in good agreement with plasma enhanced atomic layer deposited TiOx films reported in Kim's study [32]. Figure 4b-f shows profiles at three different locations from SKPM image of T25:Z75 film, which indicates that the surface potential depends on the film roughness. As the tip scanned over a particle, the height abruptly increased about 5 nm, and the local VCPD decreased about 200 mV, leading to an increase at the local work function. The distribution of charges at the surface and potential step can be attributed to crystallographic orientation and atomic relaxation [45]. It is possible that electronegative character of Zn1−xCdxS results in a partial electron transfer with TiOx that leads to an increase in local work function [45]. When the height decreased about 3 nm, the local VCPD increased about 100 mV. Within 1-2 nm variation in height, the local charge distribution could be fluctuated about 50 mV. The overall results indicate that an effective ETL composites of mixed metal oxides should have root mean square roughness of about 0.5 nm and scanning probe microscopy can be used as an effective tool to predict the homogeneity of local charges. We suggest the following mechanism to rationalize for morphological and electronic variation of TiOx:Zn1−xCdxS nanocomposite films. In pure TiOx film, nanocrystals can be formed from Ti-O monomers via hydrolysis and condensation reactions which are strongly dependent on reaction temperature. Solvent evaporates, supersaturates the polymeric precursor solution and leads to the We suggest the following mechanism to rationalize for morphological and electronic variation of TiO x :Zn 1−x Cd x S nanocomposite films. In pure TiO x film, nanocrystals can be formed from Ti-O monomers via hydrolysis and condensation reactions which are strongly dependent on reaction temperature. Solvent evaporates, supersaturates the polymeric precursor solution and leads to the formation of a large number of Ti-O species. At low temperature (80 • C), the nuclei impinge on each other and impede further grain growth via diffusion, resulting in a relatively smooth film of small grains (RMS: 0.7 nm), but the mixture of amorphous and crystalline phases (dominantly amorphous phase) leads to broad distribution of work function (3.47 ± 0.44 eV). Pure Zn 1−x Cd x S film follows similar mechanism as TiO x film, however, there is no polymeric precursor. The pure Zn 1−x Cd x S film is in smooth, crystalline phase (RMS: 0.9 nm), resulting in narrow distribution of work function (3.77 ± 0.29 V). When preexisting Zn 1−x Cd x S seeds are present in the initial stage of the synthesis, the Ti-O species are immobilized onto the seed crystals to overcome the activation energy barrier for nucleation, leading to the formation of bigger grains compared to that of pure films and creating an ordered electronic network of TiO x surrounding the grains. For T75Z25 film, the density of Zn 1−x Cd x S in the composite film is too low, resulting in rougher film due to scattering of the large crystalline grains. For T25Z75 film, the density of Zn 1−x Cd x S in the composite film is too high, the grains are interconnected by the polymer chains similar to the mechanism proposed by Yang's group [11], however, the volume of TiO x is not enough to bridge the gaps between grains, resulting in a rough surface. For T50Z50 film, the density of Zn 1−x Cd x S is optimal in the TiO x matrix, the distribution of seed crystals and its proximity interaction with surrounding polymer chains is the most effective, resulting in the smoothest film among all. As the morphology and internal structures of ETL films are tuned by addition of Zn 1−x Cd x S, the band energy of ETL is simultaneously modified due to presence of Zn 1−x Cd x S energy states and surrounding TiO x electronic network, which are supported by UV-Vis, UPS and SKPM results. Ultimately, appropriate addition of Zn 1−x Cd x S can be used to maximize the optoelectronic output and stability of ETL film and the corresponding devices.

Conclusions
This study has demonstrated that efficient perovskite and polymer solar cells were successfully fabricated using low-temperature processed TiO x :Zn 1−x Cd x S thin film as an electron transport layer. The ETL band energy can be simply tuned by varying TiO x :Zn 1−x Cd x S ratio. The addition of Zn 1−x Cd x S in TiO x lowers band gap, alters valence band positions and work function levels, reduces recombination at interfaces and increases charge carrier lifetime. The impact of local surface potential homogeneity of TiO x :Zn 1−x Cd x S nanocomposite electron transport layer on the performance of p-i-n perovskite and polymer solar cells has been revealed. Together with tailoring band edges, a homogeneous distribution of the local surface charges has been shown to improve solar cells' efficiency by more than 2%.