Since the organometal halide perovskite solar cells (PSCs) were first invented in 2009,^[1] their power conversion efficiencies (PCEs) have been increased rapidly,^[2–4] and they exhibited promising application prospects. Various methods have been developed to improve the photovoltaic performance of PSCs, such as two-step sequential deposition method,^[5] vacuum flash-assisted method,^[6] and solvent engineering method.^[7] Such progress has partially removed obstacles for PSCs on the commercial road. However, most PSCs with high PCEs are usually hole-transport-materials-based (HTM-based), for which expensive hole transport materials (such as spiro-OMeTAD) and noble metal electrodes are usually indispensable.^[8–11] Moreover, most fabrication techniques by now request a nitrogen protected atmosphere or a high vacuum environment, which increases the fabrication cost a lot and hinders the further industrialization.^[12–14] In contrast, paintable carbon-based (PC-based) PSCs are comparatively low-cost both in material and the required equipment. They are hole-transport-layer-free and work efficiently because the perovskite material can simultaneously transport both electrons and holes.^[15,16]

The carbon electrode is very cheap, stable, environment-friendly, and can be easily prepared by directly painting carbon paste on the perovskite layer with a low-temperature annealing process.^[17] Though the PCEs of PC-based PSCs remain relatively low compared with those of HTM-based PSCs, their intrinsic advantages continue to inspire efforts to improve their performance. Wei and co-workers developed a new type of carbon film with good flexibility and conductivity that could be simply hot pressed onto the perovskite layer. A PCE up to 13.5% has been achieved.^[18] Chen and co-workers utilized a solvent strategy based on the two-step sequential method to prepare the perovskite layer. A shiny and smooth perovskite surface was obtained and an outstanding PCE of 14.4% has been achieved.^[19] It is widely recognized that the leading factor that influences the performance of HTM-based PSCs is the quality of the perovskite film, with aspects such as grain sizes, crystal boundaries, and surface coverage.^[20] However, the working principle of PC-based PSCs is significantly different from that of HTM-based PSCs,^[21] and the investigation on influencing factors of PC-based PSCs performance is still insufficient.

The antisolvent treatment method is usually directly applied to the perovskite layer^[22] or on the PbI₂ layer in HTM-based PSCs.^[20,23–25] Few pieces of literature about antisolvent treatments on PbI₂ layers in PC-based PSCs have been reported. In this work, we used three types of antisolvents (chlorobenzene (CB), toluene (TL), and diethyl ether (DE)) in the antisolvent extraction treatments on PbI₂ layers in PC-based PSCs to control the device qualities. At the same time, PSCs without (W/O) antisolvent treatment were also prepared as a reference. Comprehensive research on the influencing factors including PbI₂ residual, the surface morphology of the perovskite film, the surface roughness of the perovskite film, and the contact status of the perovskite/carbon electrode interface in PC-based PSCs was performed. Remarkably, all the preparation procedures were completed in the ambient air with a relative humidity about 40%. Humidity is one of the causes of variations in device performance.^[26,27] We found that the primary factor is the contact status of the perovskite/carbon electrode interface, and an even, smooth perovskite surface could promote this interface to integrate entirely. The poor contact could seriously suppress the device performance, especially the short-circuit photocurrent density (J_SC) and the fill factor (FF).^[28–30] After optimization, a champion device with an open-circuit voltage (V_OC) of 921 mV, a J_SC of , an FF of 60.6%, and a PCE of 10.02% was achieved.

Figure 2 shows the detailed preparation procedures of the perovskite layers in our experiments. The entire experiment can be divided into two stages. In the first stage, the pre-deposited PbI₂ films were divided into three groups and treated with CB, TL, and DE, respectively. Drops of antisolvents were directly loaded onto the PbI₂ substrates with different waiting time (5 s, 15 s, and 25 s). The extraction time was set up according to the previous report and adjusted for the experimental conditions in this work.^[20] Then the PbI₂ films with antisolvent extraction treatment were immersed in an MAI/IPA solution for 30 min. In the first stage, the optimal extraction time was determined for the next stage. In the second stage, the antisolvent treated PbI₂ films with optimal extraction time and the PbI₂ films without treatment were immersed in the MAI/IPA solution for different lengths of time. The soaking time for films with antisolvent treatment was 1 min, 5 min, 10 min, and 30 min; and for films without treatment was 20 min, 30 min, 40 min, and 60 min.^[19,31] In this stage, the optimal soaking time was determined. The photovoltaic parameters of the PSCs in the first stage are listed in table 1. Samples were recorded as **-5, **-15, and **-25 PSCs (** represents that antisolvents were used, and numbers indicate the extraction time). Both the CB and TL treated PSCs showed that the PCE increases with longer extraction time. The DE-treated samples did not show this trend mostly because of its fast evaporation rate. An overlong extraction time ( s) for DE would lead to random and irregular damage to the morphologies of the PbI₂ films. Similar phenomena were observed in CB and TL treated samples with extraction time . Thus the optimal extraction time was determined, which was 25 s for CB and TL and 5 s for DE.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) Schematic of the preparation procedures of perovskite layers with or without antisolvent extraction treatment on PbI2 films.

Table 1.

Table 1.

Table 1. Photovoltaic parameters of PSCs treated with different antisolvents and extraction time. The soaking time was set at 30 min. All the values were averaged from at least 5 devices. . Experimental Extraction VOC/mV JSC/ FF/% PCE/% groups time/s CB 5 884 11.25 48.0 4.77 15 928 9.10 57.7 4.87 25 909 10.28 60.0 5.61 TL 5 948 4.71 43.7 1.95 15 927 9.79 51.0 4.63 25 897 10.75 55.0 5.30 DE 5 936 10.05 50.4 4.74 15 937 6.87 36.3 2.34 25 918 9.30 43.2 3.69

Table 1. Photovoltaic parameters of PSCs treated with different antisolvents and extraction time. The soaking time was set at 30 min. All the values were averaged from at least 5 devices. .

In the second stage, the soaking time of the antisolvent treated samples with the optimal extraction time was studied. The optimal soaking time for samples without the treatment was also included. Figure 3 shows the XRD patterns of the perovskite films prepared by different antisolvent extraction treatments or no treatment. The diffraction peaks at around 14.1°, 24.4°, 28.4°, 31.8°, and 37.6° can be assigned to the (110), (202), (220), (310), and (312) planes of the tetragonal perovskite structure.^[32,33] The diffraction peak at around 12.6° can be assigned to the (001) plane of the hexagonal PbI₂.^[20] All the perovskite films prepared by antisolvent extraction treatments showed PbI₂ residual when the soaking time was less than 10 min. In Fig. 3(c), the TL treated perovskite films showed an obvious diffraction peak at 12.6 even when the soaking time was prolonged to 30 min, which indicates that the transformation of PbI₂ to perovskite was quite slow and considerable amount of PbI₂ was left. It is believed that when MAI inserts into the PbI₂ crystal lattice, the grain volume would expand and result in a compact perovskite layer capping on the surface of the PbI₂ film which blocks the further diffusion and insertion of MAI.^[34] This would lead to an incomplete transformation of PbI₂ to perovskite and degrade the device performance since the residual PbI₂ would slow down the electron separation and transportation.^[35] By comparing the diffraction peaks at 12.6° from W/O-30 PSCs in Fig. 3(a) and CB-30 PSCs in Fig. 3(b), the CB samples were found to show a faster transformation of PbI₂ to perovskite. This is because the CB extraction treatment can change the PbI₂ film to a porous status, which accelerates the transformation process.^[36] In Fig. 3(d), the DE-treated samples showed much better XRD spectra when the extraction time was prolonged to 30 min, which indicates that perovskite films with no PbI₂ residual and high crystallinity were obtained. All experimental groups showed stronger perovskite diffraction peaks with longer soaking time, especially the (110) and (220) planes, demonstrating that higher crystallinity films were obtained. The photovoltaic parameters of the PSCs in the soaking time experiments are summarized in table 2. The J–V curves of the best performance PSCs from each experimental group are shown in Fig. 4.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) XRD patterns of perovskite films prepared by (a) non-antisolvent extraction treatment, (b) CB extraction treatment with an extraction time of 25 s, (c) TL extraction treatment with an extraction time of 25 s, and (d) DE extraction treatment with an extraction time of 5 s.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (color online) J–V curves of the best performance PSCs from each experimental group in the soaking time experiments.

Table 2.

Table 2.

Table 2. Photovoltaic parameters of PSCs in the soaking time experiments. All the values were averaged from at least 5 devices. . Experimental groups Soaking time/min VOC/mV JSC/ FF/% PCE/% W/O 20 923 17.28 50.5 8.05 30 944 15.90 55.4 8.32 40 925 16.88 60.3 9.42 60 890 16.27 55.1 7.98 CB-25 1 962 11.29 40.8 4.43 5 952 10.04 47.4 4.53 10 943 10.86 50.9 5.21 30 909 10.28 60.0 5.61 TL-25 1 968 4.77 44.3 2.05 5 983 3.85 40.3 1.53 10 963 9.30 43.0 3.85 30 897 10.75 55.0 5.30 DE-5 1 969 5.80 34.5 1.94 5 959 9.21 39.9 3.52 10 948 8.94 43.7 3.70 30 936 10.05 50.4 4.74

Table 2. Photovoltaic parameters of PSCs in the soaking time experiments. All the values were averaged from at least 5 devices. .

We found the optimal soaking time was 40 min for W/O samples and 30 min for samples treated with CB, TL, and DE. The best samples in different groups were recorded as W/O-40 PSCs, CB-30 PSCs, TL-30 PSCs, and DE-30 PSCs. The W/O-40 PSCs showed the best performance with a V_OC of 925 mV, a J_SC of , an FF of 60.3%, and a PCE of 9.42%. The corresponding XRD spectra indicated that almost no redundant PbI₂ remained. However, an overlong soaking time (60 min) was not suitable for W/O samples despite the fact that the perovskite diffraction peaks became stronger (see table 2). With the ion migration process during film growth, a dissolution-recrystallization process would occur in the perovskite layer and large sparsely distributed perovskite crystals would form on the surface.^[19] This demonstrated that an overlong soaking time was improper for high-quality perovskite films and thus led to lower PCE.

The surface SEM images of the perovskite films from W/O-40, CB-30, TL-30, and DE-30 PSCs are shown in Fig. 5. Except for the DE-30 PSCs, the surface morphologies of these perovskite films were similar with a full perovskite coverage and very few pinholes. Their grains were uniform with sizes of about 200–300 nm. By contrast, the DE-30 PSCs showed inferior perovskite surface morphologies. A poor perovskite coverage and a considerable quantity of pinholes could be observed in Fig. 5(d). The crystal sizes were relatively smaller ( ) and thus brought more grain boundaries. These pinholes and grain boundaries caused more charge recombination and bigger series resistance due to the large density of charge traps.^[37] Thus the photocurrent density and fill factor were reduced severely. AFM was performed to further explore the topographies of these perovskite films. As shown in Fig. 6, the average roughnesses of the perovskite films were 32.5 nm for W/O-40 PSCs, 66.3 nm for CB-30 PSCs, 66.7 nm for TL-30 PSCs, and 68.6 nm for DE-30 PSCs. All the perovskite films prepared by antisolvent extraction treatment were rougher than the films without treatment. A reasonable explanation is that when antisolvents were introduced to PbI₂ films, the organic solvent in the precursor solution was removed too fast for the PbI₂ crystal particles to crystallize slowly and uniformly. Large grain particles or even bulks gathered on the film surface and brought undesired surface fluctuations. Thus the perovskite films prepared subsequently showed uneven surfaces. Lots of research has demonstrated that many kinds of solvent treatment methods could turn PbI₂ films into a porous status as the CB extraction treatment did.^[36,38,39] Though a porous structure made the transformation from PbI₂ to perovskite faster and completed, increased surface roughness was also observed from the perovskite films.^[34]

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. Surface SEM images of the perovskite films from (a) W/O-40, (b) CB-30, (c) TL-30, and (d) DE-30 PSCs. The scale bars represent .

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. (color online) AFM images of the perovskite films from (a) W/O-40, (b) CB-30, (c) TL-30, and (d) DE-30 PSCs.

Figure 7 shows the cross-sectional SEM images of the best performance PSCs in these four experimental groups. Combined with the AFM results, it could be inferred that a rough surface aggravated the poor contact between the perovskite layer and carbon electrode. The surface fluctuations in the perovskite films may lead to an empty hole in the perovskite/carbon electrode interface because of a bump on the perovskite surface, as circled in Fig. 7(b); an uncovered area because of a hollow, as in Fig. 7(c); or even a capping layer levitating above the film surface, as in Fig. 7(d). Due to an uneven perovskite surface, every situation is possible at the contact interface, resulting in a poor contact status. Moreover, the carbon paste was viscous and abundant of organic solvents (such as epoxy) and a direct painting method was utilized to prepare the carbon electrode. Unlike HTM-based PSCs wherein the hole-transport-layers or metal electrodes are deposited onto the perovskite surface in a molecular or atomic scale by specialized techniques (e.g., thermal evaporation method), the painted carbon electrode may be deposited in a micron scale. The inherent nature of carbon paste and the painting technique used in PC-based PSCs made it hard to fulfill the complete contact between the perovskite layer and carbon electrode. Figure 7(a) shows a relatively better contact status of W/O-40 PSCs. According to the topographies from the AFM images and the contact problems from the cross-sectional SEM images, a simple diagram is displayed in Fig. 7(e) to illustrate the contact status of the perovskite/carbon electrode interface. The rightmost circled area shows a more compact contact owing to a flat perovskite surface compared with the other three cases.

Fig. 7.

	Figure Option ViewDownloadNew Window
	Fig. 7. (color online) Cross-sectional SEM images of (a) W/O-40, (b) CB-30, (c) TL-30, and (d) DE-30 PSCs. The scale bars represent . (e) A simulated diagram of the perovskite/carbon electrode interface.

Table 3 summarizes the influencing factors and performance of W/O-40, CB-30, TL-30, and DE-30 PSCs. The CB-30 PSCs were used as the reference. With a little PbI₂ residual, the PCEs of the TL-30 PSCs decreased by 6% compared to that of the reference cells. The DE-30 PSCs lost 16% in PCEs as a result of poor surface morphology. The compact perovskite/carbon electrode contact of W/O-40 PSCs helped to enhance the PCEs by 68%. Clearly, the contact status between the perovskite layer and the carbon electrode dominated the device performance in PC-based PSCs. A flat and smooth perovskite surface would also greatly assist PCEs improvement. No PbI₂ residual and good surface morphologies could help to improve the device performance but their influence was not as significant as that of the contact status. W/O-40 PSCs showed the best performance in all the experiments. After optimization, the highest PCE up to 10.02% was achieved. The J–V curve and photovoltaic parameters are shown in Fig. 8.

Fig. 8.

	Figure Option ViewDownloadNew Window
	Fig. 8. (color online) The J–V curve and photovoltaic parameters of the champion PSC from W/O-40 PSCs.

Table 3.

Table 3.

Table 3. Comparison between influencing factors and performance of PSCs in the soaking time experiments. . Influencing W/O-40 CB-30 TL-30 DE-30 factors PbI2 residual none none a little none Surface good good good poor morphology Surface 32.5 66.3 66.7 68.6 roughness/nm Perovskite/carbon compact poor poor poor electrode contact PCE/% 9.42 5.61 5.30 4.74

Table 3. Comparison between influencing factors and performance of PSCs in the soaking time experiments. .

By comparing the device performance of PC-based PSCs prepared with or without the antisolvent extraction treatment, we demonstrated that the dominating factor of PC-based PSCs performance is the contact status of the perovskite/carbon electrode interface. An even perovskite surface can be a great help to improve the contact status. The effects of the other two factors (PbI₂ residual and the surface morphology) were limited in the case of a poor contact, even though they are crucial in HTM-based PSCs. CB-30 PSCs exhibited the highest PCEs among all the antisolvent-treated samples, however, their PCEs improvement was limited compared with the PSCs treated with TL and DE. The poor contact between the perovskite layer and carbon electrode resulting from a rough perovskite surface seriously influenced their PCEs. To conclude, a good perovskite/carbon electrode contact should be the prime requisite in high-performance PC-based PSCs.