Impact of dynamic co-evaporation schemes on the growth of methylammonium lead iodide absorbers for inverted solar cells

A variety of different synthesis methods for the fabrication of solar cell absorbers based on the lead halide perovskite methylammonium lead iodide (MAPbI3, MAPI) have been successfully developed in the past. In this work, we elaborate upon vacuum-based dual source co-evaporation as an industrially attractive processing technology. We present non-stationary processing schemes and concentrate on details of co-evaporation schemes where we intentionally delay the start/end points of one of the two evaporated components (MAI and PbI2). Previously, it was found for solar cells based on a regular n-i-p structure, that the pre-evaporation of PbI\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$_2$$\end{document}2 is highly beneficial for absorber growth and solar cell performance. Here, we apply similar non-stationary processing schemes with pre/post-deposition sequences for the growth of MAPI absorbers in an inverted p-i-n solar cell architecture. Solar cell parameters as well as details of the absorber growth are compared for a set of different evaporation schemes. Contrary to our preliminary assumptions, we find the pre-evaporation of PbI2 to be detrimental in the inverted configuration, indicating that the beneficial effect of the seed layers originates from interface properties related to improved charge carrier transport and extraction across this interface rather than being related to an improved absorber growth. This is further evidenced by a performance improvement of inverted solar cell devices with pre-evaporated MAI and post-deposited PbI2 layers. Finally, we provide two hypothetical electronic models that might cause the observed effects.


Experimental details -Characterisation methods j-V characterisation
The solar cell j-V characterisations were done in a standardized setup using a sun simulator with halogen lamp in combination with a Keithley 2400 source measure unit. To contact the substrate, a substrate holder with lead-outs enables a 2-wire measurement. The active area was limited to 0.096 cm 2 for every solar cell by applying a shadow mask. For measuring the current-voltage characteristics, the setup is calibrated with a silicon cell with known parameters to an illumination of one sun (100 mW/cm 2 , 25°C, AM 1,5). All measurement settings except lamp intensity and position are controlled by a home-made LabVIEW program. After selecting the desired adjustments, the program independently measures the values and saves them into a .txt file.

SEM and EDX measurements
The used scanning electron microscope (SEM), was a Zeiss Supra 40 VP, which is based on the GEMINI technology. Mainly the Inlens detector was used with working distances from 1.5 mm to 3 mm. The applied voltage was fixed at 3 kV with the 20 µm aperture. Magnification values between 5000 and 100000 were used.
For the energy dispersive X-ray spectroscopy (EDX) measurements a Bruker XFlash 630M was used. This detector was mounted on the Zeiss Supra 40 VP allowing EDX measurements directly after taking SEM images of the morphology. For this purpose, the magnification was fixed at 500 with 8 mm working distance. The used voltage was fixed at 10 kV with the 120 µm aperture. By using these parameters, the interaction depth reached values around 500 nm to 600 nm which secures a detection of all used layers. The data was evaluated with the help of the software ESPRIT 2.

In situ XRD measurements
The in situ X-ray diffraction (XRD) measurements were carried out through exchangeable Kapton windows in the chamber walls. Cu-K α radiation at 35 kV and 40 mA with 1.54 Å wavelength was used for the measurements, and the K β radiation was damped by a Ni filter to 5 % of the K α intensity. For the XRD detection, three Dectris Mythen 1K detectors are connected to each other in a circular arc in front of the exit Kapton window. The used incidence angle of the X-Ray source was set at 11°a llowing an in situ 2Θ range from 8°to 36°. Due to the connection points of the three Dectris Mythen 1k detectors, blind spots at 17.3°and 26.7°occur in the corresponding color maps.

TRPL analysis
Time resolved photo luminescence (TRPL) measurements took place in a standardized home-made single photon counting setup. A pulsed diode laser with a wavelength of 638 nm was used for creating charge carriers. The laser intensity was damped between 0.001 % and 100 % intensity with the help of a neutral density filter wheel. While the frequency of the laser pulse is in principle adjustable, for these experiments a frequency of 1 MHz was used. The width of every laser pulse is 88 ps with 12.6 pJ energy and a focus area of 4.5 × 10 −5 cm 2 creating an energy density of 267 nJ/cm 2 in the full intensity modality. A "Pico Quant" detector was used for the photon detection with wavelengths between 270 nm and 850 nm. The time interval for every channel of the multi-channel analyser was set to 56 ps which is therefore the time increment of the data points. The measurement of higher excitation laser intensities is coupled to a higher recombination rate. To avoid immediately reaching the counting limit of the detector, damping filters in front of the detector were placed for several measurements. These filters can attenuate the incoming photon rate from the laser between 1 % and 100 % before reaching the detector. All measurements were saved into a .txt file for evaluations in python.

ITO layer and cleaning process
For all processes ITO coated glass substrate were used. These samples were purchased from Kintec with the pattern KT18274 and 25.0 mm × 25.0 mm × 1.1 mm size. The ITO surface showed a polished grade with a conductivity of 10 Ω/sq. Every cleaning and NiO preparation run was consisting of 24 substrates allowing different process variations for the same substrate batch. After removing the substrates from the package, every sample was engraved with a sequence number starting from R000. When the engraving was finished, the substrates were placed inside probe carriers filled with isopropanol. These probe carriers were put into an ultrasonic bath for 25 min at 50°C. After the ultra sonic bath, every substrate was removed from the probe carrier, washed up with distilled water and dried with a nitrogen stream.

Synthesis of the NiO layer
The NiO layer was deposited via electron beam evaporation in an Alliance Concept Eva450 chamber. The substrate carrier fits all 24 prepared samples and was rotating while the deposition was running. Furthermore, the substrate carrier design allows to place all samples at approximately the same distance from the substrate carrier centre. Due to these facts, there should only be minor difference of the NiO layer deposition between the different samples depending on geometric factors. All homemade NiO layers were fabricated at 150°C and 8 sccm O 2 for 4 min which led to a layer thickness of approximately 20 nm to 30 nm.

Synthesis of MAPI solar cells
When the absorber was finished, the buffer layer PCBM had to be deposited. This film was created with a 20 mg ml solution in chlorobenzene. The solution was dropped onto the substrate, which was rotated at 2000 rpm for 50 s. After the buffer layer, the ETL, consisting of ZnO nano particles in a dispersion of 2.5 wt% in isopropyl alcohol (IPA), was spin-coated at 4000 rpm for 30 s with a subsequent heating at 90°C for 5 min. The last step is the evaporation of the metal electrodes which was realized by evaporating 100 nm thick Ag-films with the help of a shadow mask.

Evaporation schemes
The subsequent passage visualizes the temperature and pressure development of all evaporation schemes in Figure 1. Blue curves indicate the PbI 2 temperature, green curves the MAI temperature, red curves the substrate temperature and black points 2/6 the measured pressure. Dashed lines inside the plots mark the opening and closing times of the corresponding shutter. The following acronyms were used in the submitted work:

j-V parameter statistics
During the experiments a large quantity of solar cells were prepared. Figure 3 and Figure 4 list the key characterization parameters of the j-V analysis for Eva 1-4. The box-plot visualizes the statistic distribution while the orange line marks the mean values. In addition, all measured values are presented as circles next to the box plot. Non-functional cells did not contribute to the statistics but were mentioned in the corresponding figure caption.

Morphology absorber images
The absorber morphology in dependence of the corresponding evaporation scheme is displayed in Figure 5. With respect to the EDX measurements, larger grain structures seem to require a I Pb stoichiometry of 3 or above. However, the cross section images of EVA 4 also suggest large grains with a top layer of smaller grains probably caused by the PbI 2 post depostion. Therefore, it can be concluded that the pre-deposition of MAI on NiO is beneficial for the crystallization behavior of MAPI in the inverted structure. Eva 4 (bottom right). The magnification was fixed at 80000.

5/6
Statistics of the wet chemical reference cells Wet chemical reference cells were fabricated from the same substrate batch as the coe-evaporated samples presented in the main manuscript. The base substrate (glass and ITO) and the NiO deposition are equal to the preparation scheme for the co-evaporated solar cells. The absorber was prepared as follows: The source solution for the wet chemical MAPI deposition consists of 658 mg PbI 2 and 228 mg MAI, which were dissolved in 1 ml co-solution containing 100 µl dimethyl sulfoxide (DMSO) and 900 µl dimethyl formamide (DMF). The absorber layer was spin-coated by dropping 100 µl of the source solution on an ITO/HTL substrate and spinning at 4000 rpm for 30 s. While the substrate was rotating, 1 ml diethyl ether was given onto the substrate after 10 s. Afterwards, the substrate was heated at 100°C for 10 min. All following steps are equal to the manufacturing of the co-evaporated solar cells. Figure 6 and Table 1 are listing the key parameters of the wet chemical reference cells. The summary of all wet chemical reference samples shows a large variation of solar cell parameters. While efficiencies over 15 % have been achieved, the average efficiency is only 9.57 % due to the large distribution of measurement values. Especially the comparison with the co-evaporated absorbers is of great interest. Also, the wet chemical reference samples show not only a large distribution of key parameters, but 2 of 6 cells are non-functional, indicating a problem which is not necessarily caused by the absorber. While the efficiency of the wet chemical reference process is in general higher due to higher current densities, the fill factor is heavily increased by EVA 4. However, the statistical distribution shows that the base process for both preparation methods is not optimized at this state of work. all reference cells 0.99 ± 0.03 18.74 ± 3.03 9.57 ± 4.13 50.09 ± 16.84 reference cells of the discussed batch 1.02 ± 0.03 17.75 ± 0.72 9.03 ± 2.83 48.50 ± 12.46 Table 1. Average and standard deviation (SD) of solar cell parameters from wet chemical reference cells. The first line refers to the values of reference cells from 5 different batches of substrates (23 functional cells, 4 non-functional cells, number of batches: 5). The second line refers to wet-chemically prepared solar cells from the same substrate batch as the co-evaporated solar cells discussed in the main manuscript (4 functional cells, 2 non-functional cells: number of batches: 1)