**Enhanced Silver Nanowire Composite Window Electrode Protected by Large Size Graphene Oxide Sheets for Perovskite Solar Cells**

#### **Hongye Chen 1, Min Li 1, Xiaoyan Wen 1, Yingping Yang 1, Daping He 1, Wallace C. H. Choy <sup>2</sup> and Haifei Lu 1,\***


Received: 12 January 2019; Accepted: 31 January 2019; Published: 2 February 2019

**Abstract:** Despite the outstanding features of high transmittance and low sheet resistance from silver nanowire (Ag NW) based transparent electrodes, their applications in perovskite solar cells (PVSCs) as window electrodes encounter significant obstacles due to the stability issue brought by the corrosion of halogen species from perovskite layer. In this study, we used large size graphene oxide (LGO) sheets as the protective barrier for bottom Ag NW nano-network. Contributed by the LGO with average size of 60 μm, less GO sheet was necessary for forming the fully covered protective barrier with fewer cracks, which consequently improved the optical transparency and anticorrosive ability of the composite electrode compared to the one from relatively small size GO. Our experiments demonstrated the composite electrode of Ag NW/LGO. The glass substrate exhibited transmittance of 83.8% and 81.8% at 550 nm before and after partial reduction, which maintained 98.4% and 95.1% average transmittance (AVT) of the pristine Ag NW electrode. Meanwhile, we utilized the steady hot airflow to assist the fast solvent evaporation and the uniform GO film formation on Ag NW electrode. Before the application of composite electrode in organic-inorganic hybrid perovskite solar cells, the operational stability of composite electrodes from different sizes of GO with perovskite film fabricated on top were characterized under continuing external bias and light irradiation. Experimental results indicate that the Ag NW electrode protected by LGO could maintain original resistance for more than 45 h. Finally, the PVSC fabricated on Ag NW/LGO based composite electrode yielded a power conversion efficiency (PCE) of 9.62%, i.e., nearly 85% of that of the reference device fabricated on the commercial indium-tin oxide (ITO) glass. Our proposed low temperature and solution processed bottom electrode with improved optical transparency and operational stability can serve as the very beginning layer of optoelectronic devices, to promote the development of low cost and large area fabrication perovskite solar cells.

**Keywords:** Ag nanowires; graphene oxide; perovskite solar cell; transparent bottom electrode

#### **1. Introduction**

Organic-inorganic hybrid perovskite has been demonstrated as a promising candidate of light absorber because of its excellent properties for photovoltaic devices. The power conversion efficiency (PCE) has achieved over 23% since it was first introduced in 2009 with PCE of 3.8% [1–7]. With the increase in efficiency and potential commercial value, researchers have paid more attention to controlling the cost of preparation, by avoiding the use of high energy consuming instruments and

adopting low temperature and solution processed methods for each layer. For example, the low temperature and solution processed interfacial layers, including electron [8–14] and hole [15–19] transporting layers, have been intensively studied to improve the efficiency of PVSCs. Carbon based materials, such as carbon paste [20,21], carbon nanotubes [22–25] and graphene [26], have also been used as top electrodes in PVSCs to replace conventional top metal electrodes formed by thermal evaporation, to reduce the cost of device fabrication and improve the stability of device simultaneously. Besides, metal nanowire electrode through spray-coating or spin-coating method is another favored candidate of top electrode for semi-transparent PVSCs [27,28].

For the bottom electrode of PVSCs, the most commonly used transparent conductive electrodes at present are indium-tin oxide (ITO) and fluorine-doped tin oxide (FTO) films fabricated by evaporation or sputtering in high vacuum environment, which however have the following problems. First, due to the large amount of electrical energy consumption and machine loss during the preparation process, the fabrication cost of ITO and FTO is relatively high. Second, the indium element required for ITO is a rare element. The rising of its price could be predicted in the long-term. Third, the metal oxide electrode has natural brittleness. The optoelectronic device based on this type of electrode will be greatly limited in flexibility and folding characteristics. As a result, several candidate materials have been investigated as bottom electrode to replace ITO or FTO, such as carbon nanotubes [29], graphene [30,31], conductive polymers [32] and metal nanowires [33–36]. Among these materials, silver nanowire (Ag NW) network is regarded as the most promising candidate because of the low sheet resistance and high transmittance. Moreover, silver nanowires have excellent flexibility, making them suitable for flexible optoelectronic device. However, the application of silver nanowire as bottom electrode to perovskite solar cells has encountered various obstacles, especially the chemical stability upon the perovskite materials because the halide species released from the perovskite layer will easily penetrate through the conventional interfacial layer, corrode the bottom silver nanowires, and thus lead to degradation of the electrical conduction [37–39].

To address the chemical stability issue, various attempts have been made to protect the bottom silver nanowire electrode. For instance, Han et al. [38] proposed a dense fluorine-doped ZnO (FZO) layer deposited on the surface of Ag NWs by pulsed laser deposition (PLD) to enhance the corrosion resistance of the electrode and ensure the conductivity of the electrode, achieving a PCE of 3.29% from perovskite device. Kim et al. [39] demonstrated a sandwich protection structure of ITO/Ag NW/ITO by successive-spin-coating method and annealing at 250 ◦C to enhance the conductivity of the composite electrode, resulting in a PCE of 8.44% from PVSC. Based on the sandwich protection structure, Lee et al. [37] proposed using amorphous Al-doped zinc oxide (AZO) instead of ITO as a protective layer for the purpose of reducing the use of indium and achieved a PCE of 13.93% from PVSC. In addition, our previous research [40] reported an anti-corrosive film composed of self-assembled graphene oxide (GO) flakes on a silver nano-network under ambient conditions, which can effectively prevent halide corrosion against the Ag NWs, and finally revealed a PCE of 9.23% by optimizing the amount of graphene oxide and reducing agent.

We demonstrated the formation of fully covered film composed of large size graphene oxide sheets (LGO) as the protective barrier for bottom Ag NW nano-network through utilizing a time-saving strategy of steady hot airflow, which is beneficial for the film formation with good uniformity. We found that the anti-corrosive abilities of the composite Ag NW electrodes from large size GO sheets were significantly enhanced compared to those composited from small size GO sheets. They were experimentally demonstrated by characterizing the resistance of composite electrode under continuing external bias and light irradiation. Importantly, the composite electrode of Ag NW/LGO after partially reduction showed an excellent optical transmittance, reaching 81.8% at 550 nm and maintaining 95.1% average transmittance (AVT) of the pristine Ag NW electrode. Finally, the PVSC fabricated with Ag NW/LGO based composite electrode demonstrated a power conversion efficiency (PCE) of 9.62%, corresponding to nearly 85% of the efficiency from reference device on commercial ITO glass.

#### **2. Materials and Methods**

#### *2.1. Materials*

Graphene oxide powder was purchased from Anhui Lianruan Education Technology Co., Ltd. (Hefei, China). ITO glasses were purchased from YINGKOU OPV TECH NEW ENERGY Co., Ltd. (Yingkou, China). Poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate) PEDOT:PSS (Heraeus Clevios PVP AI 4083), Lead (II) iodide(PbI2, >99.99%) and 2,9-dimethyl-4,7-diphenyl-1,10-Phenanthroline (BCP, >99%) were purchased from Xi'an Polymer Light Technology Corp. (Xian, China). Methylammonium iodide (MAI, 99%) was purchased from Greatcell Solar PTY Ltd. (Queanbeyan, Australia). [6,6]-Phenyl-C61 butyric acid methyl ester (PCBM, >99.5%) was purchased from Luminescence Technology Corp. (Shanghai, China). Dimethyl sulfoxide (DMSO, anhydrous, ≥99.9%), chlorobenzene (anhydrous, 99.8%) and 2-Propanol (anhydrous, 99.5%) were purchased from Sigma-Aldrich Co. LLC. (Shanghai, China). γ-Butyrolactone (GBL, anhydrous, 99.9%) was purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). Ethanol, acetone and sodium borohydride were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). All chemicals were used without further purification.

#### *2.2. Preparation of Large Size GO*

Large size GO sheets were prepared using a centrifugal classification method according to previous report [41]. Briefly, appropriate amount of GO powder was taken for preparing an aqueous dispersion with a concentration of 2 mg/mL, which was stirred for 24 h. After centrifugation at 5000 rpm for 30 min, the bottom solution (about 30% in volume) was kept and diluted to 2 mg/mL for the next cycle of centrifugation. After repeating the above process seven times, the gel at bottom was collected and used for investigations in this study.

#### *2.3. Preparation of Ag NW/RLGO Composite Electrodes*

The glass substrates were cleaned ultrasonically with acetone, deionized water and ethanol for 15 min in sequence, and then dried with N2, followed by UV/ozone treatment for 15 min. The pristine Ag NW dispersion was diluted to 1.3 mg/mL before use. Highly conductive Ag NW transparent electrodes were fabricated by spin-coating the Ag NW dispersion at 2500 rpm for 30 s and annealing at 150 ◦C for 10 min in air. According to our previous research [40], 32 μmol NaBH4 were added to 2.5 mL LGO aqueous dispersion (0.25 mg/mL). Then, the dispersion was kept for 12 h for the partial reduction of LGO. Fifty microliters of partially-reduced LGO dispersion (RLGO) were dropped onto the Ag NW transparent electrode with area of 2.89 cm2, allowing the liquid to disperse onto the entire substrate uniformly. Finally, the substrate was dried under a steady hot airflow, which could greatly speed up the evaporation of the solution. The "steady hot air flow" was achieved by a commercial adjustable electric hair dryer (FH6618, FLYCO, Shanghai, China) through carefully controlling wind and power.

#### *2.4. Preparation of Perovskite Solar Cells on Ag NW/RLGO Composite Electrodes*

For the fabrication of perovskite solar cells, the transparent electrodes of Ag NW/RLGO and commercial ITO glass were used. The Ag NW/RLGO composite transparent electrodes were prepared as described above. The control ITO substrates were ultrasonically cleaned with detergent, deionized water, acetone, and ethanol for 15 min in sequence. To prepare hole transport layer, the as-received PEDOT:PSS solution was spin-coated on the transparent electrodes at a speed of 2000 rpm for 30 s and annealed on a hotplate at 125 ◦C for 10 min. The samples were then transferred to a glove-box after cooling to room temperature. The perovskite MAPbI3 films were fabricated on the substrates according to a previously reported method [42]. Generally, PbI2 and MAI with 1:1 ratio were mixed in the mixing solvent of GBL and DMSO (volume ratio = 7:3) to form the precursor with a concentration of 1.2 M. The mixture solution was stirred at 60 ◦C overnight before use. The perovskite films were fabricated by a successive two step spin-coating process of 1000 rpm for 10 s and 4000 rpm for 30 s. During the second step of spin-coating, 150 μL of toluene were quickly dropped on the substrate, which was then annealed on a hotplate at 100 ◦C for 10 min. The PCBM layer was deposited on the as-formed perovskite film by spin-coating its chlorobenzene solution (17 mg/mL) at 2000 rpm for 30 s, followed by annealing at 100 ◦C for 10 min. Subsequently, a thin BCP film, used as buffer layer, was deposited by spin-coating its saturated solution at 3000 rpm for 60 s. Finally, the device was completed by thermal evaporation of 150 nm thick Ag cathode under 3.0 × <sup>10</sup>−<sup>4</sup> Pa.

#### *2.5. Characterizations*

The surface morphology and cross section of the as-prepared samples were examined by field-emission scanning electron microscope (FE-SEM, Zeiss Ultra Plus, Oberkochen, German). GO sheets were measured by optical microscope (XJP-107JX, Pudan Optical Instrument Co., Ltd., Shanghai, China). The diffused transmission spectra of the transparent electrodes were obtained from a UV-vis spectrophotometer (UV2600, Shimadzu, Tokyo, Japan) with an integrating sphere. The sheet resistances of the electrodes were measured using a four-point probe system with a current source-meter (Keithley 2400, Tektronix, Beaverton, OR, USA). To characterize the resistance variation of composite electrode before and after device fabrication, two electrical contacts were formed on the two sides of the electrode using silver paste and a digital multimeter (DT9206, FLUKE Corporation, Elite, WA USA) was used for resistance measurement. The photocurrent density–voltage (J–V) characteristics of all devices were analyzed using an AM1.5G solar simulator (Oriel Sol3A, Newport Corporation, Irvine, CA, USA) and Keithley 2400 sourcemeter (Beaverton, OR, USA), scanning from −0.1 to 1.2 V at a scan rate of 0.1 V s<sup>−</sup>1. The IPCE (Newport Corporation, Irvine, CA, USA) system was employed to study the quantum efficiency of the solar cells.

#### **3. Results and Discussion**

In our previous research [40], we demonstrated that Ag NWs underneath perovskite layer was severely corroded after the formation of perovskite film, which is the most critical issue that hinders their application in PVSCs as bottom electrode. To solve the chemical instability and improve the optical transparency, we used a fully-covered film composed of LGO sheets on silver nanowire electrode as an anti-corrosive barrier. As depicted in Figure 1, three different GO sheets with average sizes of 60 μm, 30 μm and 3 μm were successfully separated by a centrifugal classification method. The obtained GO sheets were diluted to an appropriate concentration, dropped on clean silicon wafer with a 300 nm SiO2 layer, and then characterized by optical microscope. In Figure 1, we can clearly observe that most GO sheets were thin and well classified, which was beneficial to the transmittance of Ag NW/GO composite electrodes and our later research. Figure 1b,d,f shows the size distributions of the three different GO sheets.

**Figure 1.** (**a**,**c**,**e**) The optical microscope images of 60 μm, 30 μm and 3 μm GO sheets, respectively;and (**b**,**d**,**f**) the size distributions of 60 μm, 30 μm and 3 μm GO sheets, respectively.

It is also worth mentioning our strategy of forming fully-covered GO film on Ag NW substrate. Our previous method of forming the GO film was obtained through self-assembly and the solvent was dried naturally under ambient condition, which was very time consuming and not suitable for large area film fabrication. Here, a rapid GO film formation approach was tested by using the steady hot airflow for the drying of the solvent. We believe this strategy is compatible with roll-to-roll technology. The preparation procedure of the composite electrode is shown in Figure 2a. To illuminate the difference from two approaches for the formation of composite electrode, a relatively high concentration of GO solution (0.5 mg/mL) was used. Figure 2b shows the photo of Ag NW/LGO composite electrodes that were prepared by the two methods with the large area of 6.25 cm2. Sample 1 was prepared by a conventional method, in which water was evaporated naturally. Sample 2 was prepared with the assistance of a steady hot airflow, as indicated in Figure 2a. The preparation of Sample 1 was time

consuming and resulted in non-uniform film formation, as indicated in Figure 2b. However, with the assistance of steady hot airflow, a flat and uniform GO film could be rapidly fabricated, and the preparation process of the Ag NW/LGO composite was efficient and suitable for large-scale fabrication.

**Figure 2.** (**a**) Schematic representation of the Ag NW/LGO composite electrode fabrication process; and (**b**) photos of Ag NW/LGO composite electrodes fabricated from relatively high concentration of LGO solution (0.5 mg/mL) on the area of 6.25 cm2 using two different methods.

To evaluate the anticorrosive effect of the film made of different size of GO sheets, the same volumes of GO aqueous solution with a constant concentration of 0.25 mg/mL were dropped on three identical Ag NW electrodes and allowed to dry under a steady hot airflow, as discussed above. Two electrical contacts using silver paste were formed on the two sides of the electrode to measure its resistance variation before and after device fabrication. PEDOT:PSS layer, perovskite (MAPbI3) layer and PCBM layer were fabricated on the composite electrodes in sequence. The control sample of Ag NW electrode without GO was also prepared following the same procedure. The resistance variations of the Ag NW/GO composite electrodes were measured in a glove-box filled with N2 (O2 <0.1 ppm, H2O <0.1 ppm) after device fabrication. As illustrated in Figure 3a, the resistances of all composite electrodes remained stable for the first 10 h, including the control sample. This was because of the glove-box, in which water and oxygen content were low. The process of decomposition of perovskite layer was relatively slow. Meanwhile, the corrosion of halide species against silver nanowires was also a gradual process in such condition. Over time, the resistance of the control electrode changed dramatically, by exhibiting more than 100 times the original resistance during 8 h. This result directly reflected the severe corrosion of the silver nano-network underneath the perovskite film. In contrast, the resistances of the electrodes protected by GO sheets increased slowly, and the resistance increasing rates decreased as the sizes of GO sheets increased. The inset in Figure 3a clearly indicates that the Ag NW electrode protected by 60 μm GO sheets showed an excellent chemical stability, which can keep stable with negligible variation for 24 h. In addition, we also measured the resistance of the Ag NW electrode protected by 60 μm GO sheets under continuous 0.8 V bias and light irradiation (0.5 sun), as shown in Figure 3b. The result clearly demonstrates that, even under continuous electricity and irradiation, the composite electrode was almost unaffected and maintained a stable resistance, proving the outstanding protective effect of large size GO sheets.

**Figure 3.** (**a**) The resistance variation of the Ag NW electrodes protected by different sizes of GO sheets (40 μL, 0.25 mg/mL) before and after the fabrication of perovskite devices: and (**b**) the resistance of the Ag NW electrode protected by 60 μm GO sheets under continuous 0.8 V bias and light irradiation (0.5 sun).

To illuminate the enhanced stability of the composite electrode of Ag NW/LGO against perovskite layer, the electrode was characterized by SEM. As evidenced by the SEM images from the silver nano-network coated with 60 μm GO sheets (Figure 4), the silver nanowires were widely covered by single graphene oxide sheet, which acted as anti-corrosive barrier against halide ions from perovskite. Since the average length of the silver nanowires was about 30 μm, it is easy to understand that a larger area of GO sheet can protect more silver nanowires from corrosion. In other words, less GO would be necessary to fully cover the same area of Ag NW electrode than one composed of small-sized GO sheets, as schematically depicted in Figure 5. Due to the physical stacking of the GO sheets, it is also reasonable to believe that the penetration of halide species through the cracks indicated by the dotted lines formed between GO sheets could still happen and caused the deterioration of the electrode. Therefore, the gradual increase of the resistance of electrode composited by 30 μm and 3 μm GO sheets after the fabrication of perovskite layer could be attributed to the formation of more cracks or defects in the anticorrosive film. Stacking more GO sheets on the electrode to block the cracks with additional GO layers might be an efficient method to prevent such penetration; however, this would be harmful to the optical transparency and electrical conductivity of transparent electrode. Solar cells fabricated on the composite electrode based on that strategy would be difficult to optimize to achieve good performance. It should also be noted that, even though the combination of metal nanowire electrode and single- or double-layer graphene in CVD method being able to obtain the best optical, electrical and stability properties, the hydrophobic feature of the film would add difficulty to the fabrication of the charge transporting layer and perovskite film. Therefore, relatively large GO sheets with average size of 60 μm were used for demonstration.

**Figure 4.** (**a**) Large-scale; and (**b**) enlarged SEM images of silver nano-network under the protection of 60 μm GO sheets. The red arrows inside indicate the crinkles of GO.

**Figure 5.** Schematic of Ag NW electrodes protected by: (**a**) large GO sheets; and (**b**) small GO sheets.

Even though the reduced stacking layer from large size GO sheets helped decrease the vertical resistance between bottom Ag NW electrode and charge transporting layer fabricated above, the lateral resistance of GO was poor due to the high degree of oxidation. Partially-reduced LGO was still essential to improve the lateral conductivity, which could help the efficient collection of photo-induced carriers generated at the position away from the metal nanowires. Hence, an appropriate amount of NaBH4 was added into the pristine LGO solution as a reducing agent to enhance the electrical conductivity of LGO film. More details are shown in Experimental Section 2.3. To characterize the optical property of the composite transparent electrodes, diffused transmission spectra of the silver nano-network electrodes coated with 40 μL of LGO (average size of 60 μm) solution with and without partial reduction by optimal amounts of NaBH4 are shown in Figure 6. The spectra indicated the composite electrodes of Ag NW/LGO and Ag NW/RLGO including the glass substrate exhibited transmittances of 83.8% and 81.8%, respectively, at the wavelength of 550 nm. Contributed by the size effect of GO sheet, as discussed above, the quantity of GO sheets forming the anticorrosive film on the electrode surface could be reduced. The optical loss of bare silver nanowire electrode brought by the addition of 60 μm LGO and 60 μm RLGO were minimized to 1.6% and 4.6% of average transmittance (AVT), respectively, at the spectral range of 400–800 nm.

**Figure 6.** Diffused transmission spectra of silver nano-network electrodes and the network electrode coated by the same amount of 60 μm LGO sheets with and without chemical reduction, and the commercial ITO substrate. The inset is the performance parameters of these samples.

Based on the excellent chemical stability and transmittance of the Ag NW/LGO composite electrodes, we further evaluated the transparent electrodes for perovskite solar cell applications. To fabricate PVSCs based on the Ag NW nano-network transparent electrodes protected by anti-corrosive LGO film, PEDOT:PSS, MAPBI3, PCBM and BCP were deposited as hole transport layer (HTL), absorber layer, electron transport layer (ETL) and buffer layer, respectively. The devices were finally finished after thermal evaporation of the silver counter electrode. The schematic diagram and cross section SEM image are shown in Figure 7a,b. The control devices of perovskite solar cells on commercial ITO electrode were also fabricated for reference. More details on device fabrication can be found in Experiment Section 2.4.

**Figure 7.** (**a**) Schematic illustration of the perovskite solar cell structure on an Ag NW/LGO composite electrode; and (**b**) cross-sectional SEM image of the perovskite solar cell fabricated on Ag NW/LGO composite electrode.

The current density–voltage (J–V) curves of PVSCs fabricated on ITO glass, the pristine Ag NW electrode without GO and the composite electrode of Ag NW nano-network with 60 μm LGO sheets are shown in Figure 8a, and the performance parameters are summarized in Table 1. The performance of PVSCs was obtained under 1 sun illumination (AM 1.5G, 100 mW/cm2) irradiated from the bottom electrode side covered with a black shadow mask. The perovskite solar cell using the Ag NW/RLGO electrode showed an open circuit voltage (VOC) of 0.87 V, short-circuit current (JSC) of 15.43 mA/cm2, fill factor (FF) of 70.9%, and PCE of 9.62%. The incident photon-to-electron conversion efficiency (IPCE) of the device based on the composite electrode is shown in Figure 8b. The calculated JSC from the IPCE spectrum is 15.23 mA/cm2, which is consistent with the value of 15.43 mA/cm2 from the J–V curve. Figure 8c exhibits the J–V hysteresis characteristics of the cells based on the optimized Ag NW/RLGO composite electrode using a dwell time of 10 ms for both forward (JSC to VOC) and backward (VOC to JSC) scan directions. The detailed hysteresis performance parameters are summarized in Table 2. As shown in Table 2, a PCE of 9.62% was achieved in the backward scan, whereas the PCE value was 9.08% in the forward scan. The little hysteresis phenomenon might be due to the coating of the PCBM layer [43,44]. The stabilized photocurrent and PCE at the maximum power output point (MPP) under AM1.5 sun illumination are shown in Figure 8d, which exhibited stabilized photocurrent of 12.72 mA/cm2 and PCE of 9.35% at 0.76 V under 400 s continuing light irradiation, agreeing well with the value measured from J–V curves (Figure 8c). As illustrated in Table 1, the JSC value from Ag NW/RLGO composite electrode accounted for 95% of the value of reference device, which was consistent with the optical transparency ratio of the composite electrode and ITO glass, indicating their qualification of optical transparency and electrical conductivity for transparent bottom electrode. The slight decrease of the FF value from the Ag NW/RLGO composite electrode-based device could be attributed to relative larger surface roughness of the composite electrode. Even though the coating of GO and PEDOT:PSS layers could improve the smoothness of bare silver nanowire electrode, the surface of composite electrode in large scale was still worse than the ITO film possessing nanometer surface roughness. The additional GO layer and thick PEDOT:PSS layer slightly affected the charge transferring between the composite electrode and perovskite layer, as evidenced by the slight increase of the serial resistance in Table 1. Nevertheless, a PCE of 9.62% was achieved, which corresponded to

nearly 85% of that of the reference device with an ITO electrode, indicating that it is suitable to use our Ag NW/RLGO based electrode for PVSCs. More importantly, our Ag NW composite electrode protected by large size GO sheets exhibited superior chemical stability over those that are pristine or protected by small size GO sheets, and maintained 95.1% average transmittance of pristine Ag NW electrode. Furthermore, the fast GO film formation method, which was assisted by steady hot airflow, was beneficial to large scale fabrication technology. Future optimization on the surface of composite electrode and perovskite film quality on the composite electrode would undoubtedly boost the efficiency of the devices.

**Figure 8.** (**a**) J–V curves of the PVSC fabricated on the Ag NW nano-network electrode protected by 60 μm LGO sheets. Data from commercial ITO glass and Ag NW electrode without GO are also presented. (**b**) IPCE performances of the devices fabricated on the nano-composite electrode and ITO substrate. The blue and red solid lines indicate the integrated current of the ITO-based and composite electrode based devices, respectively. (**c**) J–V curves of the PVSC fabricated on the Ag NW nano-network electrode protected by 60 μm LGO sheets measured under forward and backward bias scanning. (**d**) Steady photocurrent (blue) and PCE (red) under 1 sun illumination of the Ag NW/RLGO based PVSC.


**Table 1.** Summarized photovoltaic parameters of PVSC devices based on Ag NW/RLGO composite electrodes, control ITO substrates and pristine Ag NW electrodes without GO.



#### **4. Conclusions**

In summary, different sizes of GO sheets from a centrifugal classification method were used and compared for the formation of fully-covered barrier on silver nanowire electrode with the assistance of steady hot airflow. Experimental results show that the size of GO sheet played an important role in preventing the corrosion of halogen species. By employing large size GO sheets, the anti-corrosive ability of Ag NW/LGO composite electrodes were significantly improved compared to those with small size GO sheets, as proven by monitoring the resistance variation under external bias and light illumination. Moreover, the Ag NW/RLGO composite electrode exhibited excellent transmittance of 81.8% at 550 nm and maintained 95.1% average transmittance (AVT) of the pristine Ag NW electrode. Consequently, the PVSCs fabricated on the Ag NW/RLGO based composite electrodes achieved a power conversion efficiency (PCE) of 9.62%, i.e., nearly 85% of the PCE from the solar cell on commercial ITO. These results clearly demonstrate the potential of Ag NW/RLGO composite as window electrode for PVSCs, which would be beneficial to large-scale low-temperature fabrication and future commercialization of PVSCs.

**Author Contributions:** H.C. and H.L. conceived the idea. H.C. and H.L. designed the experiment and guided the experiment. M.L., X.W. and W.C.H.C. helped the data analysis. Y.Y. and D.H. helped the device fabrication. H.C. conducted data collection and wrote the manuscript. H.C. and H.L. revised the manuscript.

**Acknowledgments:** This work was supported by the National Natural Science Foundation of China (Grant No. 11704293), Hubei Provincial Natural Science Foundation of China (Grant No. 2017CFB286) and the Fundamental Research Funds for the Central Universities (WUT: 2018IB008, 2018IVB016).

**Conflicts of Interest:** There are no conflicts to declare.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **High-Dose Electron Radiation and Unexpected Room-Temperature Self-Healing of Epitaxial SiC Schottky Barrier Diodes**

**Guixia Yang 1, Yuanlong Pang 1, Yuqing Yang 1, Jianyong Liu 1, Shuming Peng 1, Gang Chen 2, Ming Jiang 3, Xiaotao Zu 3, Xuan Fang 4, Hongbin Zhao 5, Liang Qiao 3,\* and Haiyan Xiao 3,\***


Received: 31 December 2018; Accepted: 28 January 2019; Published: 2 February 2019

**Abstract:** Silicon carbide (SiC) has been widely used for electronic radiation detectors and atomic battery sensors. However, the physical properties of SiC exposure to high-dose irradiation as well as its related electrical responses are not yet well understood. Meanwhile, the current research in this field are generally focused on electrical properties and defects formation, which are not suitable to explain the intrinsic response of irradiation effect since defect itself is not easy to characterize, and it is complex to determine whether it comes from the raw material or exists only upon irradiation. Therefore, a more straightforward quantification of irradiation effect is needed to establish the direct correlation between irradiation-induced current and the radiation fluence. This work reports the on-line electrical properties of 4H-SiC Schottky barrier diodes (SBDs) under high-dose electron irradiation and employs in situ noise diagnostic analysis to demonstrate the correlation of irradiation-induced defects and microscopic electronic properties. It is found that the electron beam has a strong radiation destructive effect on 4H-SiC SBDs. The on-line electron-induced current and noise information reveal a self-healing like procedure, in which the internal defects of the devices are likely to be annealed at room temperature and devices' performance is restored to some extent.

**Keywords:** electron irradiation; room temperature self-healing; noise; electron-induced current; I–V curve

#### **1. Introduction**

Silicon carbide (SiC) is an ideal material for high-frequency, high-power, and high-temperature devices due to its strong hardness, excellent thermal/mechanical stability, high-thermal conductivity, and controlled electrical properties [1,2]. With the advance of state-of-the-art growth and epitaxy techniques in the past decade, a variety of high-quality SiC-based structures and devices have been fabricated [1], making them very competitive toward third-generation wide-gap semiconductors. Thus, SiC-based structures have recently regained significant attention in advanced device applications. Furthermore, owing to its large intrinsic fundamental band gap (up to 3.26 eV [3]), SiC is not suitable to visible and infrared light stimulation or irradiation, and their radiation resistance is stronger than that of the conventional semiconductors (such as Si-based materials); therefore, SiC-based diodes are very promising candidates to replace silicon devices for atomic battery light sensors, nuclear cell photosensors [4,5], or particle detectors [6,7].

Required by long-term exposure applications, the radiation resistance and durability of SiC diodes (as radiation detectors and atomic battery light sensors) are very critical and have been subjected to extensive research. For particle detector applications, the detectivity and radiation resistance of SiC diodes were studied under gamma ray [7,8], proton [7,8], neutron [9], electron [8,10], and X-ray [6] irradiations. The results indicated that the electrical properties of radiated SiC diodes and their radiation resistance depended on types, energies, flux, fluence, and the absorbed dose of irradiated particles. It was found that when the total dose of 32-MeV proton was 8.5 × <sup>10</sup><sup>12</sup> cm−<sup>2</sup> and the total absorbed dose of 1.25-MeV γ-ray was 1000 kGy (Air), the electrical properties of 6H-SiC diodes were basically unchanged [7]. Within the range of 1–10 Gy absorbed dose, the radiation damage to 4H-SiC diodes caused by 22-MeV electron and 6-MeV X-ray photon beams hardly affected the radiation-induced current [6]. When the neutron fluence exceeded 5.7 × <sup>10</sup><sup>16</sup> cm−2, the charge collection efficiency of the 4H-SiC diode started to drop [9]. For atomic battery light sensor applications, electrons were an important radiation source [4]; thus, many studies had focused on the ability of 4H-SiC diodes' resistance to electron irradiation. For example, Eiting et al. studied the radiation resistance of 4H-SiC p-i-n junction betavoltaic irradiated with 8.5 GBq 33P source with an average beta decay energy of 77 keV, and found that the devices did not experience degradation under the irradiation of 33P source for four half-lives of 33P source, thus demonstrating a good electron radiation resistance [4]. Chandrashekhar et al. experimentally studied the devices' radiation resistance performance under the 1 mCi 63Ni β-radiation source for ten days, and the results suggested such irradiation conditions did not cause the degradation of 4H-SiC diodes [5].

In practical applications for both radiation detectors and atomic battery light sensors in many extreme conditions, such as nuclear, aerospace, military, and astrophysics, the total expected irradiation dose is normally enormous. However, the most common research of experimental or theoretical study on SiC electron radiation resistance is still relatively low; for example, the largest total 8.2-MeV electron fluence ever reported is 9.48 × 1014 cm−<sup>2</sup> [8,10]. Large fluence of electron irradiation is not only common in the radiation detection experiments, but also critical in the integrated design of the next-generation atomic battery, and thus deserves further study.

Moreover, the general principle of particle detector and atomic battery light sensor is based on induced-current generation (by radiation of electrons, neutrons, photons, and heavy ions) and detection (of the induced-voltage signals in diodes). The previous research were mostly focused on the correlation between the internal defects of SiC materials (or devices) and their electrical parameters under electron irradiation, such as relationship between defect levels and I–V as well as C–V characteristic curves [5,8]. However, this apparent relationship is not suitable to explain the intrinsic response of irradiation effect, since defects are not easy to characterize and it is complex to distinguish whether the defects come from the raw material or exists only upon irradiation. Therefore, a more straightforward quantification of irradiation effect is greatly needed to establish the direct correlation between irradiation-induced current and the radiation fluence.

For this purpose, low-frequency noise information, especially 1/f noise, can be used as a characterization means for the sensitive detection of internal defects in electronic materials [11] and devices [12,13]. This method in principle can also be utilized to characterize radiation damage to electronic devices [12]. Under the beam radiation, the internal defects and the structural damage will lead to the increase of low-frequency noise power spectral density *SV* [14,15]. For example, Babcock et al. [16] studied the radiation resistance of Ultra High Vacuum/Chemical Vapor Deposition SiGe heterojunction bipolar transistor (HBT) irradiated by 60Co γ-ray and found that the increase of the total absorbed dose resulted in the performance degradation of the SiGe HBT. The current gain *β* was decreased with the increase of absorbed dose, and the internal defects were increased, while the noise information *SV* increased in the low-frequency range, as compared with that before irradiation. Therefore, the 1/f noise can be used as a tool to characterize the defects in materials and devices, and correlate the defects with devices performance.

Here, we reported the study of electron-induced current and radiation resistance of SiC Schottky barrier diodes (SBDs) under high-fluence electron irradiation as well as an in situ noise diagnostics for defect–electrical property analysis. Through on-line electron-induced current, I–V curve, noise information, and SBDs' radiation resistance to the environment of electron irradiation had been analyzed, and a self-driven healing process was observed at room temperature, which led to some extent of electrical performance recovery.

#### **2. Materials and Methods**

#### *2.1. 4H-SiC SBDs Samples and Irradiation Experimental Conditions*

The epitaxial 4H-SiC SBDs used in this experiment were provided by State Key Laboratory of Wide-Band Gap Semiconductor Power Electronics located at Nanjing, China. Its structure is shown in Figure 1a, where the photosensitive area of SBDs is 3 mm × 3 mm and the voltage-withstand range is −100 to 100 V. The 4H-SiC SBDs were 4H-SiC epilayers grown by chemical vapor deposition on SiC substrates of 360 mm thickness, with nitrogen doped with a net doping density of 1 × 1018 cm−<sup>3</sup> and micropipe density of 1 micropipe cm−2. The buffer was n-type, 1 μm thick, with a net free carrier concentration of 1 × <sup>10</sup><sup>18</sup> cm−3. The epilayer was n-type, 12 <sup>μ</sup>m thick, with a net free carrier concentration of 3 × 1015 cm<sup>−</sup>3. The ohmic contact was obtained by deposition of a 1000-Å-thick layer of Ni and a 3-μm-thick layer of Au. The Schottky contact was obtained by radio frequency magneton sputtering of a 1000-Å-thick layer of Ni at room temperature. In order to reduce the influence of environment on the device, SiO2 of 1000 Å thickness and Si3N4 of 1000 Å thickness were grown on Ni by sputtering method.

**Figure 1.** (**a**) The schematic diagram of Silicon carbide (SiC) Schottky barrier diodes (SBDs) and (**b**) the principle of electron-induced current generation under irradiation.

Electron beam irradiation experiments were performed on the electron accelerator of Sichuan Forever Holding Co., Ltd, located at Mianyang, China. The electron energy was 1.8 MeV, the electron flux was 9.62 × <sup>10</sup><sup>12</sup> cm−<sup>2</sup> <sup>s</sup>−1, and the total electron fluence was 9.05 × <sup>10</sup><sup>17</sup> cm−2. In order to avoid overheating of SiC SBDs during electron beam irradiation, the bottom of the irradiated metal platform was continuously cooled with water, and meanwhile, the temperature of the SiC SBDs was monitored by a thermocouple and controlled at a constant temperature of 25 ◦C. The home-made real-time on-line current test system was used to monitor and record the changes of the electron-induced current of SiC SBDs during irradiation and 30 min after electron irradiation. Then the devices were kept at room temperature (25 ◦C) for 72 h without heating. The devices performance tended to decrease with the increasing electron fluence, but it was not certain whether the devices were disabled or not. In the case of reverse breakdown, the devices would be disabled completely, and the reverse breakdown state can be regarded as the worst-case state of devices performance. The devices were reverse biased to breakdown at 200 V voltage by Keithley 6517B high impedance/electrometer located at Mianyang, China toward the end of the experiment, and the damage of the device after irradiation was evaluated when the devices performance at reverse breakdown state was considered as the worst-case state. The I–V curves and noise information of SiC SBDs before and after electron beam irradiation, room-temperature self-healing, and reverse breakdown were measured by Keithley 2635 sourcemeter and the self-developed noise parameter test system located at Mianyang, China.

#### *2.2. The Electron-Induced Current Test*

When subjected to radiations of electron beam, γ-ray, and neutron, current can be induced in SBDs due to the ionization effect within Schottky barrier junctions, which is shown in Figure 1b. The principle of the electron-induced current can be understood by taking the electrons as an example—1.8 MeV electrons penetrate the SBDs completely. The n-type 4H-SiC material of the Schottky barrier junction was irradiated by incident electrons with an energy larger than the material's fundamental electronic band gap; thus electron–hole pairs would be generated in n-type 4H-SiC material. The Fermi lever in the Ni Schottky contact was less than the Fermi lever in the n-type 4H-SiC material; therefore, the average energy of electrons in the n-type 4H-SiC material was greater than the average energy of those in the Ni Schottky contact. The difference in the average electron energy can be expected to transfer electrons from the n-type 4H-SiC material to the Ni Schottky contact until the average electron energies were equal. While the holes can only stay in n-type 4H-SiC material, the electrons were pulled toward the Ni Schottky contact. As a result, the width of the depletion region became narrower, and the contact potential difference decreased. The incident electron energy was converted into electrical energy. Once the external circuit was short-circuited, current could flow through the Schottky barrier junction. In this way, the separation of the electron–hole pairs can be achieved, and the induced current *IP*, whose direction was from the Ni Schottky contact to the n-type 4H-SiC material through the Schottky barrier junction, can be produced. The current equation of Schottky barrier junction under electron beam irradiation is

$$I = I\_0(e^{cII/kT} - 1) - I\_P \tag{1}$$

where *I* and *I*<sup>0</sup> are the current flowing through the Schottky barrier junction and the reverse saturation current, respectively; e, *U*, and *T* stand for the electron charge, the applied voltage, and temperature, respectively; and *<sup>k</sup>* is Boltzmann constant (*<sup>k</sup>* = 1.38 × <sup>10</sup>−<sup>23</sup> J K<sup>−</sup>1).

When the applied voltage *U* is 0 V, the Formula (1) becomes

$$I = -I\_P.\tag{2}$$

In this study, the open-circuit induced current of SiC SBDs has been tested by a real-time on-line current test system, as shown in Figure 2a. This test system consisted of the fixtures, low-loss cables, 10-channel scanning card, data interface, Keithley 6517B high impedance/electrometer, and self-compiled control software (installed in the control computer). The fixtures were used to fix SiC SBDs and transmit current signals to low-loss cables. The 10-channel scanning card was installed at the rear panel of Keithley 6517B high impedance/electrometer and they were used to capture multiple electron-induced currents. The self-compiled control software was used to control and record current data detected by Keithley 6517B high impedance/electrometer in real time. The system can achieve a 10-channel real-time signal acquisition with the current and voltage accuracies of each signal acquisition reaching way up to 1 fA and 1 nV, respectively.

**Figure 2.** The illustration for (**a**) real-time on-line current test system and (**b**) test system for noise parameters.

#### *2.3. Noise Information Test*

In this study, the changes of noise information before and after irradiation of SiC SBDs have been tested by using the noise parameter test system. The noise parameter test system consists of low-noise bias, adapter, Stanford Research Systems SR560 voltage amplifier, acquisition card, and XD3020 [17] noise analysis software with the system bandwidth being 0–10<sup>6</sup> Hz, the background noise being <4 nV/√Hz (@1 kHz), and the signal magnification capability being 100–105, as shown in Figure 2b.

The noise of the diode was typically composed of two or three components of shot noise, 1/f noise, and generation-recombination noise (G-R noise). The noise power spectral density (*S*) can be written as:

$$S(f) = A + \frac{B}{f^{\gamma}} \tag{3}$$

where *A* is the shot noise amplitude, *B*, the 1/f noise amplitude, and the exponents of *γ* is usually taken to be unity.

In this work, based on the noise spectrum information of SiC SBDs, the electron irradiation effects of SiC SBDs were studied by using the parameter information in Equation (3). The damage degree of the device had been determined according to the change of noise power spectral density before and after irradiation and room-temperature self-healing.

#### *2.4. I–V Curve Test*

The current–voltage (I–V) curve of a semiconductor device contained the electrical information of dark current as well as break-over voltage. The real-time on-line electron-induced current and I–V characteristics can characterize the device to some extent. The I–V curve with voltage range of −100 to 2.5 V before and after irradiation of SiC SBDs had been measured with Keithley 2635 sourcemeter in this work. The damage degree of the device had been determined according to the change of I–V curve before and after irradiation and room-temperature self-healing.

#### **3. Results and Discussion**

#### *3.1. High-Dose Electron Irradiation*

Figure 3 shows the real-time curves of the open-circuit electron-induced current in SiC SBDs. It can be seen that when the energy of electrons was 1.8 MeV, and the electron flux was 9.62 × <sup>10</sup><sup>12</sup> cm−<sup>2</sup> <sup>s</sup><sup>−</sup>1, a current of 1.27 × <sup>10</sup>−<sup>5</sup> A was induced due to the ionization effect. However, as the electron fluence

further increased, the electron-induced current of the device decreased continuously. When the electron fluence reached 1 × <sup>10</sup><sup>15</sup> cm<sup>−</sup>2, the current decreased to 1.24 × <sup>10</sup>−<sup>5</sup> A, accounting for only a decrease of 2.36%, which was consistent with the results reported by Nava et al. with the total electron fluence being 9.48 × 1014 cm−<sup>2</sup> [8,10]. SiC SBDs can be considered as sufficient to resist the irradiation of electrons with a fluence under 1 × <sup>10</sup><sup>15</sup> cm−2. However, with a further increase of electron fluence, the electron-induced current of SiC SBDs began to decrease sharply, particularly when the electron fluence was ≤<sup>3</sup> × 1017 cm−2. When the electron fluence was 1 × 1016 cm−2, the current decreased to 1.03 × <sup>10</sup>−<sup>5</sup> A, showing a decrease of 18.90%. When the electron fluence was 1 × 1017 cm−2, the current decreased to 5.00 × <sup>10</sup>−<sup>6</sup> A, accounting for a decrease of 60.63%. At the end of irradiation, the total electron fluence reached 9.05 × 1017 cm−<sup>2</sup> and the current became 1.82 × <sup>10</sup>−<sup>6</sup> A, showing a decrease of 85.70%. It was noteworthy that when the electron fluence was >3.87 × 1017 cm−2, the electron-induced current fluctuation of SiC SBDs increased sharply and the current fluctuation could reach up to 86%.

**Figure 3.** The real-time changing profiles of the induced current with electron fluence.

When SiC SBDs were irradiated by high dose of electrons, a large number of defects would be generated inside the silica layer [18], silicon carbid layer [19–22], and the interfaces [23,24] between the metal and silicon carbide. These defects would produce a removal effect [25,26] on charge carriers at 1.8 MeV electron irradiation, leading to the decrease of current.

Hemmingsson et al. studied the deep level defects in electron-irradiated 4H-SiC epitaxial layers grown by chemical vapor deposition using deep level transient spectroscopy (DLTS). The measurements performed on electron-irradiated p-n junctions in the temperature range 100–750 K revealed both electron and hole traps with thermal ionization energies ranging from 0.35 to 1.65 eV [27], which led to the deterioration of device performance. Iwamoto et al. investigated the formation and evolution of defects in 4H-SiC Schottky barrier diodes and correlated with the SBDs' performances [28]. The SBDs were irradiated with 1 MeV electrons to a fluence of 1.00 × <sup>10</sup><sup>15</sup> cm−2. Current–voltage, capacitance–voltage, and DLTS measurements were used to study the effect of defects on the SBDs performance. It was found that the DLTS defect levels (EH1, EH3, and Z1/2) were very likely to be partly responsible for the charge collection efficiency reduction after electron irradiation. EH1 and EH3 were related to carbon interstitials and Z1/2 was related to carbon vacancies. DLTS study on n-type 4H-SiC (0001) epilayers also showed that the carrier lifetime in an n-type 4H-SiC epilayer, measured by differential microwave photoconductance decay, had been significantly improved from 0.73 μs (as-grown) to 1.62 μs (after oxidation, 1300 ◦C) as the Z1/2 and EH6/7 centers had been reduced from (0.3–2) × <sup>10</sup><sup>13</sup> cm−<sup>3</sup> to below the detection limit (1 × <sup>10</sup><sup>11</sup> cm−3) by thermal oxidation of epilayers at

1150–1300 ◦C [29]. Based on the above results, we suggested that the current fluctuation might due to the increasing defects in the devices, such as interstitials and vacancies. These defects led to a slower and less stable carrier motion, which resulted in current fluctuation.

Figure 4 shows the I–V curves (including reverse breakdown region) of SiC SBDs before and after irradiation. From Figure 4b, it can be seen that the reserve current of SiC SBDs before irradiation increases with the increase of reserve bias, showing typical diode characteristics. For the 2-h irradiation sample, the forward current of the device dropped sharply from 2.39 × <sup>10</sup>−<sup>5</sup> A (before irradiation) to 2.06 × <sup>10</sup>−<sup>8</sup> A at a voltage of 2.5 V, while the absolute value of the reverse dark current further increased (Figure 2b). When the voltage became −100 V, the current was −8.76 × <sup>10</sup>−<sup>7</sup> A, showing a serious degradation of the device characteristics. At the reverse breakdown of the device, the I–V curve of the device had been tested. It was found that the forward current was further reduced compared to the case of 2 h after irradiation. When the voltage became 2.5 V, the current dropped to 1.42 × <sup>10</sup>−<sup>9</sup> A, while the voltage became −100 V, the current was −1.3 × <sup>10</sup>−<sup>9</sup> A, which was similar to the I–V characteristics of the resistance. The electron capture levels had been proved to be induced by the electron irradiation, and it would drastically influence the resistance in the bulk crystal and the Schottky barrier of SiC SBDs [30]. Under 2 MeV electron irradiation, the Schottky barrier was found to decrease and the resistance of the bulk crystal was found to increase with the electron fluence, leading to the decrease of the forward current. The Schottky barrier of SiC SBDs decreased from 1.25 eV (pre-irradiation) to 1.17 eV (post-irradiation, 1.00 × <sup>10</sup><sup>17</sup> cm−2). The decrease of the barrier height was thus responsible for the increase of the reverse current. On the other hand, the increase of electron-induced defects in passivation layers was expected to result in the increase of the reverse leakage [31,32]. In this work, the I–V curves (2 h after electron irradiation) exhibited similar behavior. The SiC SBDs could be considered as the resistance since the barrier height disappeared and the bulk crystal resistance increased further after the SiC SBDs were reverse biased to breakdown. The I–V curves (reverse breakdown) of SiC SBDs were similar to the I–V characteristics of the resistance.

**Figure 4.** The I–V curves for SiC SBDs before and after electron beam irradiation. (**a**) The foward bias region and (**b**) the reverse bias region.

Besides the electrical properties (electron-induced current and I–V curves), the noise information of SiC SBDs also reflects the internal defect states before and after irradiation and the reverse breakdown, as shown in Figure 5a. It can be seen that in the frequency range of 10<sup>2</sup> Hz to 10<sup>5</sup> Hz, the noise power spectral density, *SV*, exhibited a clear dependence on the electron irridation, with *SV* (reverse breakdown) > *SV* (2 h after irradiation) > *SV* (pre-irradiation). Ziel et al. [33] proposed that under small fluence irradiation, induced current was mainly contributed by two components: (1) the current *I*0*eqU*/*kT* due to the injection of electrons from the semiconductor into the metal and (2) –*I*<sup>0</sup> due

to the injection of electrons from the metal into the semiconductor. Both currents contained carriers that pass independently and randomly through the junction barrier, thereby showing as pure intermediate frequency shot noise. In this work, the shot noise of current SiC SBDs was in the frequency range of 10<sup>2</sup> to 10<sup>5</sup> Hz.

**Figure 5.** The noise power spectral densities of SiC SBDs. (**a**) Before and after electron beam irradiation and in a reverse breakdown state; (**b**) the low-frequency noise curve within 72 h after electron irradiation; and (**c**) the changing profiles of the low-frequency noise amplitude within 72 h after electron irradiation.

The current noise power spectral density (*SI*) of SBDs' shot noise is [34]

$$S\_I = 2e(I + 2I\_0) \tag{4}$$

The SBDs resistance can be obtained by Equation (1)

$$R = \frac{d\mathcal{U}}{dI} \approx \frac{kT}{\mathcal{c}(I + I\_0)}\tag{5}$$

The voltage noise power spectral density *SV* is

$$A = S\_V = S\_I \times R^2 = \frac{2(kT)^2(I + 2I\_0)}{\varepsilon(I + I\_0)^2} \tag{6}$$

The current *I* is

$$I = \frac{2ne}{t} \tag{7}$$

where *R* stands for the diode resistance, *n* and *t* are the carrier concentration and time for drifting through the barrier, respectively.

When *I I*0, Equation (6) can be simplified as

$$S\_V \approx \frac{2(kT)^2}{eI} = \frac{t(kT)^2}{e^2n} \tag{8}$$

As the current decreased, *n* decreased, which made *SV* to increase according to Equation (8), thus leading to *SV* (2 h after irradiation) > *SV* (pre-irradiation). When SiC SBDs first experienced the electron beam irradiation and then subjected to reverse breakdown, a large number of defects would be produced inside SiC SBDs, which made the noise power spectral density of SiC SBDs to further increase compared with SBDs that only irradiated by electrons, therefore showing *SV* (reverse breakdown) > *SV* (2 h after irradiation).

#### *3.2. Room-Temperature Self-Healing*

As shown in Figure 5a, the noise spectra of SiC SBDs before and after irradiation as well as in reverse breakdown state are 1/f noise in the frequency range of 100 to 10<sup>2</sup> Hz. 1/f noise had an extrinsic origin that arose from defects [35]. The defects of electronic origin lying in the gap acted as electron traps leading to both mobility and carrier density fluctuations. In the frequency range of 10<sup>0</sup> to 10<sup>2</sup> Hz, *SV* reached the minimum before irradiation, while reached the maximum at 2 h after electron irradiation, and *SV* even was greater than that of the reverse breakdown. However, if the SiC SBDs were left alone for 72 h after electron irradiation at room temperature, their *SV* could become lower than that of *SV* at reverse breakdown state, showing a significant decrease. The observed *SV* (2 h after irradiation) > *SV* (72 h after irradiation) was partially due to the fact that part of the defects were rapidly annealed at room temperature; thus, the effect of carrier removal was weakened, leading to the increase of carrier concentration. According to Hooge's equation [36], the relationship between *SV* and carrier concentration *n* are given by Equation (9).

$$S\_V(f) = \frac{B}{f^\gamma} \propto \frac{\alpha\_H}{n} \tag{9}$$

where *α<sup>H</sup>* stands for the Hooge parameter. The *SV* corresponding to 1/f noise was inversely proportional to the carrier concentration inside the device. As carrier concentration increased, *SV* decreased.

The noise power spectra at 2 h and 72 h after irradiation (shown in Figure 5a) cannot clearly reflect the annealing effect of the internal defects in SiC SBDs at room temperature. Therefore, at 72 h after irradiation exposure, we have carried out the tracking measurement on the low-frequency noise of SiC SBDs irradiated by electron beam at room temperature, which is shown in Figure 5b. It can be clearly seen that in the frequency range of 100~102 Hz, the noise power spectral density of SiC SBDs decreased with time, and the observable frequency range of 1/f noise gradually decreased. At 2 h after irradiation, 1/f noise can be observed in the frequency range of 1–100 Hz; at 4 h after irradiation exposure, 1/f noise can be observed in the frequency range of 1–30 Hz; and at 72 h after irradiation exposure, 1/f noise frequency further decreased into 1–6 Hz, while 1/f noise will be annihilated by shot noise above these frequency ranges.

The temporal variation of 1/f noise amplitude of the SiC SBDs irradiated by electrons can be calculated by Equation (9), and the results are shown in Figure 5c, where the 1/f noise amplitude *B* decreases exponentially with time *t*, and this relationship is further well fitted by Equation (10).

$$B = 7.81 \times 10^{-13} e^{-t/0.81} + 2.50 \times 10^{-15} \tag{10}$$

where the unit of *<sup>B</sup>* is V2·Hz<sup>−</sup>1, the unit of *<sup>t</sup>* is s.

Then the declining rate is

$$\frac{dB}{dt} = -9.64 \times 10^{-13} e^{-t/0.81} \tag{11}$$

At 4.5 h after irradiation, *B* declined sharply, while the declining rate decreased with the increase of time, and the declining rate followed Equation (11). At 72 h after irradiation, the declining rate tended to be constant. The value of *<sup>B</sup>* decreased from 6.75 × <sup>10</sup>−<sup>14</sup> <sup>V</sup>2·Hz−<sup>1</sup> (2 h after irradiation) to 2.05 × <sup>10</sup>−<sup>15</sup> <sup>V</sup>2·Hz−<sup>1</sup> (72 h after irradiation), accounting for 96.96% decrease.

Figure 5b,c shows that the internal defects of the SiC SBDs continuously decrease at room temperature, weakening the removal effects of the defects on carriers. Back in 1966, Fischerrr et al. [37] had systematically investigated the temperature annealing of traps produced by 6–88 MeV electron irradiation in n-type Ge. In their experiment, the electron irradiation temperature was 85 K, and it was found that some traps disappeared when the temperature increased to near 200 K, thereby, leading to an increase of the carrier concentration. In 2009, Messina et al. [38] found that a significant portion of *E <sup>γ</sup>* centers, which induced in amorphous silica at room temperature by γ-irradiation up to 79 kGy, spontaneously decayed after the end of irradiation. In 1984, Yamaguchi et al. [39] found that effective room-temperature self-healing of radiation-induced defects in both p-type and n-type InP after electron irradiation leading to the recovery of InP solar cell properties. Besides Ge, amorphous silica, and InP solar cell, room-temperature self-healing of radiation-induced defects were also shown in Si-based devices [25,40–43], as reported by Pease et al. Pease et al. [25] found the normalized change in reciprocal resistivity in the drain to source (Δ1/*RDS*) of Si-based power Metal-Oxide-Semiconductor Field-Effect Transistors degraded between 2 and 10% (most less than 5%) over a period of 24 h after room-temperature proton and neutron irradiation. These phenomena suggested that the internal defects generated inside the semiconductor material under irradiation can be annealed and annihilated below room temperature. However, as far as the room-temperature annealing of SiC materials is concerned, currently there is still not enough experimental evidences to demonstrate that some defects of SiC materials can be annealed after electron irradiation at room temperature. In this work, we used low-frequency noise to show that electron irradiation can produce the defects in SiC devices and these induced defects can be further annealed at room temperature. Assuming only electron irradiation and *α<sup>H</sup>* remained constant, regardless of the amount of electron irradiation, part of internal defects in SiC material at 72 h after irradiation can be calculated to decrease by 3.04% as compared with the same internal defects at 2 h after irradiation. Since 1/f noise cannot be directly used to differentiate the types of defects, it is impossible to know what kinds of defects are annealed at room temperature. However, this property can still be used to characterize the declining trend of internal defects with time after electron irradiation via the noise information, and this method is also applicable for the defects characterization in other semiconductor materials and devices.

Furthermore, noise characterization were used to demonstrate that the SiC SBDs after electron irradiation have undergone an unexpected self-annealing or relaxation process at room temperature, thereby, reducing the induced defects while increasing the carrier concentration. Such room-temperature annealing was consistent with the on-line electron-induced current curve after irradiation and vice versa. As shown in Figure 6, the data collected by the on-line current–voltage test system shows that the electron-induced current of the device does not decrease to the background level within 19 min after the exposure to electron irradiation, but instead presents a linear increasing profile with the current rising from 10−<sup>10</sup> to 10−<sup>7</sup> A. This phenomenon suggested that within 19 min after the irradiation, the device was subjected to a room-temperature annealing or relaxation, curing some of the internal defects, and recovering the electrical performance to a certain level. This phenomenon was also consistent with the I–V characteristics of the SiC SBDs, demonstrating that although the performance of the SiC SBDs was greatly lowered after the exposure to 9.05 × 1017 cm−<sup>2</sup> electron irradiation, this type of semiconductor device can still recover its electrical properties to some extent under this unexpected self-annealing mechanism.

**Figure 6.** The real-time electron-induced current curve of SiC SBDs after the end of exposure to electron beam irradiation.

#### **4. Conclusions**

The electrical performance of SiC SBDs irradiated by high-dose, high-energy (1.8 MeV) electron beam had been investigated in this work. It was found that electron beam had a strong radiation destructive effect on 4H-SiC SBDs. Their electrical performance was greatly reduced when exposed to high-energy electron beam with a fluence as high as 9.05 × <sup>10</sup><sup>17</sup> cm−2, the electron-induced current reduced by 85.70%, and the device characteristics degraded seriously. The on-line electron-induced current and noise information revealed a self-healing like procedure, in which the internal defects of the devices were likely to be annealed at room temperature and devices performance was restored to some extent. Although the mechanism of this self-healing is under investigation, the high-dose irradiation and noise diagnostics study reported here can provide useful information for designing next-generation atomic battery and detectors for extreme environment applications.

#### **5. Patents**

Guixia Yang, Yuanlong Pang, Fansong Zeng, et al. Low-noise bias device, the Chinese patent of invention, patent number: ZL 201510729381.1.

**Author Contributions:** All the authors contributed to the conception, design, and performance of the experiment, the analysis of the data, and the writing of the paper. G.Y., H.X., and L.Q. initiated and discussed the research problem; Y.P., Y.Y., and M.J. performed the irradiation and on-line current experiments; G.Y., S.P., and J.L. performed the I–V and noise experiments; X.F. and H.Z. performed defect analysis; L.Q., X.Z., and G.Y. made the figures and analyzed the data; and G.Y., H.X., and L.Q. wrote the paper.

**Acknowledgments:** This research was funded by NSAF Joint Foundation of China (Grant No. U1530129), the Institution of Nuclear Physics and Chemistry (INPC), China. Project (Grant No. 2015CX03), National key laboratory of Materials Behavior and Evaluation Technology in Space Environment (Grant No. 6142910010403).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Improving Two-Step Prepared CH3NH3PbI3 Perovskite Solar Cells by Co-Doping Potassium Halide and Water in PbI2 Layer**

#### **Hsuan-Ta Wu 1, Yu-Ting Cheng 1, Ching-Chich Leu 2,\*, Shih-Hsiung Wu <sup>3</sup> and Chuan-Feng Shih 1,4,\***


Received: 25 March 2019; Accepted: 23 April 2019; Published: 27 April 2019

**Abstract:** Incorporating additives into organic halide perovskite solar cells is the typical approach to improve power conversion efficiency. In this paper, a methyl-ammonium lead iodide (CH3NH3PbI3, MAPbI3) organic perovskite film was fabricated using a two-step sequential process on top of the poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) hole-transporting layer. Experimentally, water and potassium halides (KCl, KBr, and KI) were incorporated into the PbI2 precursor solution. With only 2 vol% water, the cell efficiency was effectively improved. Without water, the addition of all of the three potassium halides unanimously degraded the performance of the solar cells, although the crystallinity was improved. Co-doping with KI and water showed a pronounced improvement in crystallinity and the elimination of carrier traps, yielding a power conversion efficiency (PCE) of 13.9%, which was approximately 60% higher than the pristine reference cell. The effect of metal halide and water co-doping in the PbI2 layer on the performance of organic perovskite solar cells was studied. Raman and Fourier transform infrared spectroscopies indicated that a PbI2-dimethylformamide-water related adduct was formed upon co-doping. Photoluminescence enhancement was observed due to the co-doping of KI and water, indicating the defect density was reduced. Finally, the co-doping process was recommended for developing high-performance organic halide perovskite solar cells.

**Keywords:** perovskite solar cells; water doping; potassium halide doping

#### **1. Introduction**

The first report on lead halide organic perovskites for photovoltaic applications was published in 2009 [1]. Kojima et al. used methylammonium lead iodide (CH3NH3PbI3, MAPbI3) to replace organic dyes in dye-sensitized solar cells (DSSCs), where mesoporous titanium oxide (TiO2) and a liquid electrolyte were used, achieving a power conversion efficiency (PCE) of 3.8%. Recently, the best solar cell efficiency achieved was 22%, giving perovskites a reasonable chance to reach commercial competiveness [2]. PCE of the MAPbI3-based solar cell has been close to 20% in both mesoporous structure devices [3] as well as in planar heterojunction architectures [4]. High temperature annealing (>400 ◦C) is required to crystallize the TiO2 layers used in mesoporous-type solar cells. Compared to the high-temperature processing of mesoporous solar cells, the planar heterojunction perovskite photovoltaics has the advantage of a low-temperature (100 ◦C) solution process, and, therefore, can be adopted in the roll-to-roll production of flexible devices [5].

One key aspect that affects the performance of planar heterojunction perovskite solar cells is the quality of the organic perovskite film, which is determined by the thermodynamics and the growth kinetics of the film [6–8]. Solution-processed perovskite films usually have abundant defects when compared to single-crystal samples. The introduction of additives into the perovskite precursor solution was reported to be an effective way to prepare a high quality perovskite film with fewer defects, leading to enhanced device performance [4,9–11]. Potassium halides were wildly used as additives in the perovskite precursor solutions adopted for solar cell research [12–14]. Potassium halide-doped perovskite solar cells that have a record PCE of more than 20%, without I–V hysteresis, have been constructed [13,15]. Potassium halides were found to significantly facilitate the crystal growth of perovskite films and ameliorate the perovskite morphology, resulting in a reduced density in trap states and enhanced device performance [13,16]. However, it is not easy to form a homogeneous organic precursor by adding considerable amounts of potassium halide salts due to its restricted solubility in some organic solvent used for organic perovskite processing, which limits the application of potassium halides as additives. Water is a good solvent for potassium halide salts. Additionally, water additives have been reported to enhance the property of a two-step processed MAPbI3 [17,18]. Water additives changed the characteristics of dimethylformamide (DMF), which is a general solvent for PbI2, thus helping to make a homogeneous PbI2 solution. A smooth and dense PbI2 film was fabricated by adding a small amount of water into the PbI2 precursor with DMF, and then a high-quality MAPbI3 was obtained after the methylammonium iodide (MAI) conversion. Therefore, water was considered to be a suitable candidate for a co-doping additive for potassium halides during the MAPbI3 solution process.

However, it is challenging to deposit high-quality organic perovskite films on PEDOT:PSS, which is wildly used as the hole-transporting material in planar heterojunction perovskite solar cells. To obtain a homogeneous film structure on top of an organic surface, such as the PEDOT:PSS film, through a simple-solution process is not easy for an ionic material (such as MAPbI3) [17]. Compared to MAPbI3, PbI2 is less polar and, therefore, can easily form a continuous film on the PEDOT:PSS surface. Therefore, the two-step sequential process (PbI2 layer + MAI conversion) is an appropriate way to fabricate a perovskite film when PEDOT:PSS is used as the hole-transporting layer [17]. However, very few studies have reported the effects of water or potassium halide additives on two-step processed perovskite solar cells [19]. According to this report, the alkali metal halides (KCl, NaCl, and LiCl) were incorporated with the PbI2 layer and chelated with Pb2<sup>+</sup> ions, enhancing the crystal growth of PbI2 films that, in turn, improved the crystallinity of the perovskite films and their photovoltaic properties. However, to the best of our knowledge, no report has investigated the performance of a planar heterojunction perovskite device by considering their interactive effects by using both water and potassium halides.

In this work, we proposed an effective way to enhance the efficiency of the MAPbI3-based perovskite device through the co-doping of water and potassium halides (KI, KBr, and KCl) during the PbI2 deposition process. Systematic studies of the effects of water and potassium halides co-doping on the thin film and device were investigated and discussed. To construct a planar heterojunction perovskite solar cell, a two-step process was employed to fabricate the MAPbI3 film. PbI2 was first deposited on PEDOT:PSS film, then the MAI was spin-coated on the PbI2 layer, followed by thermal annealing. Water and potassium halides were added into the PbI2 precursor solution to elucidate their influence on the performance of perovskite solar cells. As a result, the PCE of devices made from these additive-enhanced perovskites increased from 8.8% (based on pristine perovskite) to 13.9%.

#### **2. Materials and Methods**

#### *2.1. Chemicals*

Lead (II) iodide (PbI2, 99.9985%), potassium iodide (KI, 99.995%), and potassium chloride (KCl, 99.997%) were purchased from Alfa Aesar. Anhydrous N,N-dimethylformide (DMF, 99.8%), 2-propanol (IPA, 99.5%), and chlorobenzene (CB, 99.8%) were purchased from Sigma-Aldrich (Saint Louis, MO, USA). Potassium bromide (KBr, IR spectroscopic) was purchased from Honeywell (Morristown, NJ, USA). Ultra-pure wa ter (Resistivity > 18.2 MΩ·cm at 25 ◦C, Baker Analyzed LC/MS Reagent) was purchased from J.T. Baker (Radnor, PA, USA). Poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS), methylammonium iodide (MAI, >98%), bathocuproine (BCP, >99.5%), and phenyl-C61-butyric acid methyl ester (PC61BM, >99.5%) were purchased from Uni-onward (New Taipei City, Taiwan). All materials were used as received.

#### *2.2. Device Fabrication*

The fabrication process of the referenced standard perovskite solar cell (Ref) was as follows: patterned ITO-coated (indium tin oxide) glass substrates were cleaned ultrasonically in acetone and isopropanol for 10 min, and then dried in an oven for 15 min at 110 ◦C. After a UV-ozone treatment of 30 min, the substrates were transferred into a N2-filled glovebox. Filtered PEDOT:PSS was spin-coated at 2000 rpm onto the ITO substrate and baked on a hot plate at 110 ◦C for 10 min. The PbI2 precursor was prepared by dissolving PbI2 in the DMF solvent (370 mg/mL) and heated to 70 ◦C on a hot plate. Following this, the hot PbI2-DMF precursor was spin-coated on top of the PEDOT:PSS layer and baked on a hot plate at 90 ◦C for 10 min. The MAI (45 mg/mL in IPA) precursor was spin-coated on the crystalline PbI2 layer and annealed at 110 ◦C for 1 h to form the MAPbI3 perovskite. After annealing for 20 min, the pure IPA solution was spun on the MAPbI3 layer to wash out redundant MAI, and then the annealing process continued for 40 min. The PC61BM (20 mg/mL in CB) was spin-coated on the perovskite. BCP of 10 nm and Al of 150 nm were sequentially evaporated on the top with a deposition rate of 0.4 Å/s and 3–5 Å/s, respectively. For the co-doping samples, H2O with a 1, 2, or 3 volume percent, and potassium halide of various concentrations were added into the PbI2- DMF solution. The water added into the precursor was ultra-pure water.

#### *2.3. Characterizations*

The crystallinity of the MAPbI3 film was examined by X-ray Diffraction (XRD) with ultraX 18 (Tokyo, Japan). The microstructure of the films was characterized by high resolution scanning electron microscope (HR-SEM) SU8000 (HITACHI, Tokyo, Japan). The electrical analysis was performed by E5270B Precision IV Analyzer (Keysight, Santa Rosa, CA, USA) under AM 1.5 G illumination. The absorption analysis was performed by UV-vis-NIR spectrophotometer U-4100 (HITACHI, Tokyo, Japan). The photoluminescence (PL) was performed by Jobin Yvon LabRAM HR micro-Raman system (HORIBA, Kyoto, Japan) and the excitation laser wavelength was set to 532 nm. The Fourier transform infrared spectroscopies (FTIRs) were performed by Nicolet FTIR instrument (Thermo Fisher Scientific, Waltham, MA, USA). The Raman spectroscopy was performed by Raman microscope (Renishaw, Wotton-under-Edge, UK). The depth profiles were performed by Auger Electron Spectroscopy MICROLAB 350 (Thermo Fisher Scientific, Waltham, MA, USA).

#### **3. Results and Discussion**

The photovoltaic properties of the devices (structure: glass/ITO/PEDOT:PSS/MAPbI3/PC61BM/ BCP/Al) were investigated. Figure 1 displays the photocurrent density–voltage (J–V) curves of the perovskite solar cells prepared with and without the incorporation of water and potassium halide additives in the PbI2-DMF precursor. The concentration of each cell was at the optimal conditions for cell properties. Table 1 lists the corresponding cell parameters. The cell performance was measured using an aperture metal mask of 2.2 mm × 3.2 mm to a device active area of 2 mm × 3 mm. The PCE of

device that was prepared using the 2 vol% water-doped PbI2 precursor increased from 8.8% (without water) to 12.0% with a short-circuit current density (JSC) of 22.5 mA·cm<sup>−</sup>2, an open-circuit voltage (VOC) of 0.9 V, and a fill factor (FF) of 59.2%. The photovoltaic performance, as a function of the water content, is illustrated in Figure S1, and the corresponding photovoltaic parameters are listed in Table S1. It was found that the PCE of devices that were prepared using a water content below 3 vol% were improved, and that the best performance was achieved by adding 2 vol% water in PbI2. The VOC and FF increased slightly, and the JSC increased markedly by increasing the water content to 2 vol%, resulting in the best PCE. Further increasing the water content to 3 vol% degraded the cell property. It has been reported that a planar heterojunction solar cell based on a high-quality perovskite film has a power conversion efficiency of 18% with a remarkably high FF value of 0.85 [17]. The study indicated that adding small amounts of water into the PbI2-DMF precursor made the solution more uniform, forming a smooth PbI2 film on top of the PEDOT:PSS, with high crystallinity and large crystalline domains. The perovskite film fabricated from the high-quality PbI2 film was highly pure and dense, without any pinhole. Our research showed similar observations. Water addition improved the morphology and crystallinity of the PbI2 films, as revealed in the SEM images (Figure S2). Moreover, the thickness of the PbI2 films increased with the content of water. The thickness was 156, 192, 189, and 194 nm for 0, 1, 2, and 3 vol% water incorporated PbI2 films, respectively. Figure S3 shows the MAPbI3 films fabricated from the water-doped PbI2 films. The grain size of all of the three water-incorporated films, Figure S3b–d, was larger than the film without water (Figure S3a). It has been observed that a rapid crystallization of organo-halide perovskites into the expected tetragonal cell for MAPbI3 occurred on exposure to small amounts of moisture [20]. You et al. have proposed a growth mode via thermal annealing of the perovskite precursor film in a humid environment (e.g., ambient air) to greatly improve the film quality, grain size, carrier mobility, and lifetime [21]. They indicated that, due to the strong hydroscopic nature of MAI, exposing the perovskite precursor to moisture during film formation could result in the accumulation of moisture within grain boundaries, inducing grain boundary creep and, subsequently, merging adjacent grains together. In addition, moisture could also provide an aqueous environment to enhance the diffusion length of the precursor ions, further promoting perovskite grain growth. The PbI2 films fabricated by the water-containing precursor possibly retained residual water or a water-related function group, which helped perovskite grain growth, until the deposition of MAI.

**Figure 1.** Photocurrent density–voltage (J–V) curves of perovskite solar cells prepared without and with water and/or potassium halide additives under their optimized concentrations.


**Table 1.** The corresponding photovoltaic parameters of devices in Figure 1.

Conversely, the PCE of photovoltaic devices fabricated by the PbI2-DMF precursor with potassium halides (KI, KBr, and KCl) as the only additives became worse. As displayed in Figure 1, the PCE deteriorated from 8.8% for pristine MAPbI3 to 7.5%, 8.2%, and 8.0%, by adding 1 mg/mL KI, KBr, and KCl, respectively. For KCl- and KI-doped samples, the degradation of PCE was caused by the decrease in the JSC and the fill factor, while for KBr-doped samples, only the fill factor decreased. Figure S4 illustrates the changes in photovoltaic performance of the potassium halide-doped devices as a function of the amount of potassium halide added. Notably, the devices were further improved by incorporating potassium halide into the water-PbI2-DMF precursor. By co-doping with 2 vol% water and potassium halides, the PCE was improved from 8.8% (pristine perovskite) to 13.9% (4 mg/mL KI), 11.9% (1 mg/mL KBr) and 12.8% (2 mg/mL KCl). The changes in the photovoltaic performance of the water and potassium halide co-doped devices as a function of the potassium halide additives is also illustrated in Figure 2. The performance of devices was enhanced by adding KI or KCl into the 2 vol% water-doped PbI2-DMF precursor, while KBr-doped devices had inferior properties than those without KBr. Using 4 mg/mL KI as the additive provided the best performance of the device. Figure S4 further plots the cell parameters as functions of the content of the potassium halides for devices prepared by PbI2 without water; all of the cell parameters degraded and Table S2 lists the corresponding photovoltaic parameters of the devices.

**Figure 2.** The changes in (**a**) *V*OC; (**b**) *J*SC; (**c**) FF (fill factor); (**d**) power conversion efficiency (PCE) of the potassium halide-doped devices with 2 vol% water co-doping as a function of doping amounts of KI, KBr, and KCl.

Figure 3 shows comprehensive XRD analyses of the potassium halides-doped PbI2 and MAPbI3 films with and without water incorporated with PbI2. The PbI2 and MAPbI3 films were deposited on ITO/PEDOT:PSS. Without water, the XRD intensity increased markedly for all the KCl-, KBr-, and KI-doped PbI2 films (Figure 3a). However, the XRD of the MAPbI3 films prepared on the potassium-doped PbI2 were similar to that without doping (Figure 3c). Water incorporation increased the XRD intensity of PbI2, but did not affect the XRD intensity of the perovskite film (Figure 3a,c). Co-doping of water and potassium halide increased the XRD intensity of both the PbI2 and MAPbI3 significantly (Figure 3b,d). The increase in the XRD intensity was also accompanied by an increase in the grain size.

*Nanomaterials* **2019**, *9*, 666

**Figure 3.** X-ray Diffraction (XRD) of PbI2 prepared by incorporating (**a**) KCl, KBr, and KI (without water) and (**b**) KCl, KBr, and KI with water additives. XRD of MAPbI3 on PbI2 prepared using the same conditions as (**a**) and (**b**) are shown in (**c**) and (**d**), respectively.

Figure 4 shows the top-view and cross-section SEM images of the PbI2 films that were co-doped with 2 vol% water and potassium halides (KI: 4 mg/mL, KBr: 1 mg/mL, and KCl: 2 mg/mL). The film thickness was unchanged by the potassium halides, but the surface coverage was improved and the grain size was enlarged, which agreed with the XRD observation. Figure 5 displays the SEM images of the co-doped films. Adding potassium halides (especially KI and KBr) to the 2 vol% water-incorporated PbI2 precursor enhanced MAPbI3 grain growth, which was denser with higher continuity, allowing for effective charge generation and dissociation in perovskite films. Such a highly continuous and large grain structure was also beneficial for carriers to transport through the film.

**Figure 4.** Plane and cross-section scanning electron microscope (SEM) images, respectively, of PbI2 prepared with (**a**,**e**) 2% water, (**b**,**f**) 2% water + KI (4 mg/mL) (**c**,**g**) 2% water + KBr (1 mg/mL), and (**d**,**h**) 2% water + KCl (2 mg/mL). (Scale bar = 500 nm).

**Figure 5.** Plane and cross-section SEM images, respectively, of MAPbI3 on PbI2 with (**a**,**e**) 2% water, (**b**,**f**) 2% water + KI (4 mg/mL) (**c**,**g**) 2% water + KBr (1 mg/mL), and (**d**,**h**) 2% water + KCl (2 mg/mL). (Scale bar = 500 nm).

To exploit the effect of co-doping water and potassium halides in the PbI2 layer, the Fourier transform infrared spectroscopy and Raman spectroscopy were adopted, as shown in Figure 6. The PbI2 films were prepared on Si/PEDOT:PSS because glass absorbs the IR light. In Figure 6a, characteristic peaks related to the C=O stretching were found around 1715 cm−<sup>1</sup> [22]. According to the literature, the C=O vibration shifts to lower frequency around 1650 cm−<sup>1</sup> for DMF, which forms Lewis adducts due to the reaction of the PbI2 layer [23,24]. In our sample, the C=O stretching was found to be blue-shifted, which might be caused by the formation of a PbI2-DMF-H2O adduct by incorporating water through a Lewis acid–base reaction. The other possibility was that the PbI2-DMF precursor reacted with the PEDOT:PSS under-layer because of the incorporation of water by partially dissolving the PEDOT:PSS near the PbI2/PEDOT:PSS interface. A similar observation has been reported by Winther et al. [25]. Figure 6b shows the Raman spectra of the PbI2 layer with or without potassium halides and water. Si-related signals were found at 521 cm−<sup>1</sup> (Si–Si LO), 940, and 987 cm−<sup>1</sup> (Si–OH) [26]. Notably, the O–H and C–H stretching of PEDOT split into two peaks at 2865 and 2945 cm<sup>−</sup>1, owing to the incorporation of metal halides and water, particularly for KBr- and KCl-doping [27]. Evidence from FTIR and Raman spectroscopies indicated that the doping of water and potassium halides could form some new adducts and change the interfacial chemistry near the PbI2/PEDOT:PSS.

**Figure 6.** (**a**) Fourier transform infrared (FTIR) and (**b**) Raman spectroscopies of PbI2 films doped using potassium halides with and without water.

Figure 7 displays the absorption spectra (Figure 7a) and PL spectra (Figure 7b,c) of the perovskiteglass structures prepared on the PbI2 layers that were doped with water, potassium halides, and potassium halides and water co-dopants. In the visible region (470–800 nm), the absorbance of the MAPbI3 on PbI2, prepared with the co-doped potassium halides and water, increased, indicating improved crystallinity. With potassium-only additives, the absorbance of MAPbI3 decreased. The PL-quenching effect of KBr-doping was observed (Figure 7b), while PL enhancement was found in the KI- and KCl-doped films. The steady-state photoluminescence (PL) was an effective way to detect the trap states within the perovskite layer. The higher PL intensity indicated fewer traps or defects within the films and improved crystallinity. Figure 7c shows the PL spectra of the co-doped water and potassium halides for MAPbI3 films. The PL intensity of the perovskite films obviously increased by co-doping, except for the KBr-doped film. In particular, the PL intensity of the co-doped KI and water increased by five times. Therefore, it was concluded that the co-doping of water and potassium halides increased the grain size of MAPbI3 and eliminated radiative defects that contributed to the strong PL response.

**Figure 7.** (**a**) Absorption spectra and (**b**,**c**) photoluminescence (PL) spectra of perovskite films prepared using PbI2 with different potassium halide and water additives. The samples were measured with this structure: glass/ITO/PEDOT:PSS/MAPbI3/PC61BM.

Auger electron spectroscopy (AES) was used to investigate the elementary depth distribution of the full device structure using KI and 2 vol% water co-doping, as shown in Figure 8. The potassium signal was found to be uniformly distributed across the perovskite and penetrated into the PEDOT:PSS. It has been reported that KI can provide an extra I-ion source that affects the coordination with Pb2<sup>+</sup> and compensates for the I vacancy [15]. Therefore, in addition to the positive effect due to water- and potassium halides-induced grain growth, the enhancement of PL intensity can be associated with the passivation of the K<sup>+</sup> cation and halide anions on the grain boundaries of perovskite film [15,16] and the perovskite interface, thus, reducing the trapped states. In other words, the addition of potassium halides and water improved the quality of the perovskite by reducing the traps and interfacial radiative recombination centers, allowing for effective charge generation and collection. Such a highly continuous and large grain structure was also beneficial for carriers to transport through the film. This improvement reduced series resistance (*R*S) and increased shunt resistance (*R*SH), improving the *J*SC, FF, and *V*OC as observed in Table 1. Some studies partially attributed good photovoltaic performance to a significant absorption improvement because the use of additives led to a denser perovskite film with less pinholes [14,28]. However, our research results showed water and/or potassium halide additives only slightly enhanced the absorption of perovskite. The pristine MAPbI3, prepared by our process, exhibited a compact microstructure with pure tetragonal structure phases, and the absorption effect was considered to be minor. It is worth noting that the KBr additive was harmful to the device, where even its physical properties, shown in Figures 3–5, seemed as good as KI. We attributed this to the possible formation of a small amount of MAPbBr3, through the incorporation of Br even though no secondary phase was found in the XRD spectra. The existence of MAPbBr3 within the MAPbI3 film may cause an energy barrier and inhibit charge transport from the perovskite layer to the ITO [12].

**Figure 8.** Auger electron spectroscopy (AES) depth profile of MAPbI3 prepared on PEDOT:PSS coated ITO (indium tin oxide) glass.

However, incorporation of potassium halides had little influence on the microstructure of the perovskite without any water additive. As shown in Figure S5, the surface morphology and the cross-section images of the films were similar, regardless of the types of potassium halides. The grain size was unchanged by the potassium halide additives, which is in agreement with the unchanged XRD FWHM (full width at half maximum) observation shown in Figure 3. The additive effect of each potassium halide can be clearly observed in Figures S6–S8. To conclude, the potassium halides had positive influences on the film quality and the device, only if they were co-doped with the water additive. Without water, the potassium halide additive made the solution more inhomogeneous, worsening the film properties. Because the solubility of potassium halides in PbI2-DMF was low, they were unevenly dispersed within the solution without water, forming an inhomogeneous film. The different optimal concentration of each potassium halide for the photovoltaic properties was influenced by their solubility.

#### **4. Conclusions**

In conclusion, a way to improve the perovskite solar cells (structure: glass/ITO/PEDOT:PSS/ MAPbI3/PC61BM/BCP/Al) was proposed by co-doping water and potassium halides in the PbI2 layer, which was coated on the PEDOT:PSS layer based in a two-step sequential process. When only potassium halides were added to PbI2, the PCE of the devices became worse, while the PCE of the devices prepared using the 2 vol% water-doped PbI2 precursor increased from 8.8% (without water) to 12.0%. By co-doping with the 2 vol% water and a potassium halide, the PCE was improved from 8.8% (pristine perovskite) to 13.9% (4 mg/mL KI), 11.9% (1 mg/mL KBr), and 12.8% (2 mg/mL KCl). XRD results showed that the incorporation of water and a potassium halide improved the crystallinity and enlarged the grain size. SEM images showed that the grain of PbI2 became coarse and continuous upon co-doping. FTIR spectra showed characteristic peaks, related to C=O stretching, around 1715 cm−1, which was probably caused by the formation of a PbI2–DMF–H2O adduct or an interfacial reaction near the PbI2/PEDOT:PSS interface. Raman spectra revealed that O–H and C–H stretching of PEDOT split into two peaks at 2865 and 2945 cm−1, owing to metal halides and water incorporation. Obvious PL enhancement was caused by the co-doping of water and KI, reducing the defect density. Together with the AES observations, it was found that KI distributed uniformly within the perovskite layer and penetrated into the PEDOT:PSS layer, which suggested that the elimination of the defects in the film and interface upon KI doping was one of the reasons to improve the KI–water co-doped solar cell.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/5/666/s1: Figure S1: The tendency of photovoltaic performance of the water-doped devices as a function of water amount; Table S1: The corresponding photovoltaic parameters of devices in Figure S1; Figure S2: SEM morphology and cross-section of PbI2 films with (a,e) 0%, (b,f) 1%, (c,g) 2%, and (d,h) 3% water; Figure S3: SEM morphology of perovskite films prepared by PbI2 with (a) 0%, (b) 1%, (c) 2%, and (d) 3% water; Figure S4: The (a) *V*OC; (b) *J*SC; (c) FF; (d) PCE of the potassium halide-doped devices as a function of doping amounts of KI, KBr and KCl; Table S2: The corresponding photovoltaic parameters of devices in Figure S4 and Figure 2; Figure S5: The SEM cross-sectional images of the potassium halide-doped MAPbI3 films (without water additive), (a) ref, (b) 1 mg/mL KI, (c) 1 mg/mL KBr, and (d) 1 mg/mL KCl; Figure S6: SEM surface morphology of the perovskite films with different KI and water additives in the PbI2 layer; Figure S7: SEM surface morphology of the perovskite films with different KBr and water additives in the PbI2 layer; and Figure S8: SEM surface morphology of the perovskite films with different KCl and water additives in the PbI2 layer.

**Author Contributions:** Conceptualization, Y.-T.C. and C.-F.S.; Data curation, H.-T.W. and C.-C.L.; Formal analysis, H.-T.W. and Y.-T.C.; Investigation, H.-T.W. and S.-H.W.; Methodology, Y.-T.C.; Project administration, H.-T.W. and C.-F.S.; Resources, C.-C.L., S.-H.W. and C.-F.S.; Supervision, C.-C.L. and C.-F.S.; Validation, H.-T.W. and Y.-T.C.; Writing—original draft, H.-T.W. and C.-C.L.; Writing—review & editing, H.-T.W., S.-H.W. and C.-F.S.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors are grateful for the support of the Ministry of Science and Technology of the Republic of China under Contract No. MOST 106-2221-E-006-225- and MOST 107-2221-E-006-160-, Taiwan. This work was also financially supported by the Hierarchical Green-Energy Materials (Hi-GEM) Research Center, from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) and the Ministry of Science and Technology (MOST 107-3017-F-006-003) in Tainan. The authors are also grateful for the support of the Department of Industrial Technology, Ministry of Economic Affairs, Taiwan.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **E**ffi**cient Ni**/**Au Mesh Transparent Electrodes for ITO-Free Planar Perovskite Solar Cells**

**Dazheng Chen 1,2,\*, Gang Fan 1, Hongxiao Zhang 2, Long Zhou 2, Weidong Zhu 1, He Xi 1,3, Hang Dong 1, Shangzheng Pang 1, Xiaoning He 1, Zhenhua Lin 1, Jincheng Zhang 1,2, Chunfu Zhang 1,2,\* and Yue Hao <sup>1</sup>**


Received: 4 May 2019; Accepted: 17 June 2019; Published: 28 June 2019

**Abstract:** Indium thin oxide (ITO)-free planar perovskite solar cells (PSCs) were fabricated at a low temperature (150 ◦C) in this work based on the transparent electrode of photolithography processed nickel/gold (Ni/Au) mesh and the high conductivity polymer, PH1000. Ultrathin Au was introduced to increase the conductivity of metal mesh, and the optimal hexagonal Ni (30 nm)/Au (10 nm) mesh (line width of 5 μm) shows a transmittance close to 80% in the visible light region and a sheet resistance lower than 16.9 Ω/sq. The conductive polymer PH1000 not only smooths the raised surface of the metal mesh but also enhances the charge collection ability of metal mesh. The fabricated PSCs have the typical planar structure (glass/Ni-Au mesh/PH1000/PEDOT:PSS/MAyFA1−yPbIxCl3−x/PCBM/BCP/Ag) and the champion PSC (0.09 cm2) obtains a power conversion efficiency (PCE) of 13.88%, negligible current hysteresis, steady current density and PCE outputs, and good process repeatability. Its photovoltaic performance and stability are comparable to the reference PSC based on the ITO electrodes (PCE = 15.70%), which demonstrates that the Ni/Au mesh transparent electrodes are a promising ITO alternative to fabricate efficient PSCs. The relatively lower performance of Ni/Au based PSC results from the relatively slower charge extraction and stronger charge recombination than the ITO based PSC. Further, we tried to fabricate the large area (1 cm2) device and achieve a PCE over 6% with negligible hysteresis and steady current density and PCE outputs. The improvements of perovskite film quality and interface modification should be an effective approach to further enhance the device performance of Ni/Au based PSCs, and the Ni/Au mesh electrode may find wider applications in PSCs and flexible devices.

**Keywords:** metal mesh; transparent electrode; photolithography; perovskite solar cell; large-area solar cell

#### **1. Introduction**

Organic-inorganic hybrid perovskites solar cells (PSCs) have attracted more and more attention due to their advantages of low fabrication cost, light weight, solution processability, tunable light absorption range, bipolar transport properties, large-area manufacturing, and compatibility to both rigid and flexible substrates. Their rapid progress of increased power conversion efficiency (PCEs) from 3.8% to 23.32% and improved stability indicate many potential applications, including in photovoltaic plants, photovoltaic curtains, building integrated photovoltaic materials, wearable electronics devices, and even space power systems [1–10]. And the efficient bottom transparent electrode is primary and crucial to meeting these versatile applications. Nowadays, transparent conductive oxides, such as the indium thin oxide (ITO) is the most commonly used transparent electrode in PSCs. However, ITO

prices are becoming more and more expensive because of the rising cost of indium. On the other hand, the ITO shows a relatively large sheet resistance, brittleness, and poor mechanical robustness [11,12]. There will be cracks on ITO surface at a bending radius (<10 mm) against repeated bending, which will further decrease its conductivity and degrade the device performance, making it incompatible with low-cost technology, such as roll-to-roll printing [13].

To overcome the drawbacks of ITO, many alternatives have been developed with good electrical and optical properties, such as graphene [14,15], carbon nanotubes [16], metal nanowires [17,18], transparent conducting oxide nanocrystal [19], conducting polymer [20,21] ultrathin metal films [21–24], and metal-mesh transparent conductive electrodes (TCEs) [25,26]. Among the alternatives, metal meshes have excellent ductility and can be easily fabricated by mature evaporation process, which is more suitable for large-area device production. Metal-mesh TCEs are highly bendable as well, and silver metal-mesh TCEs have been successfully used to fabricate organic solar cells, PSCs, and flexible devices [5,27]. More importantly, the conductivity and transmittance of metal mesh can be adjusted by structure parameters such as the metal-grid pitch, line width, and film thickness. And there are many ways to achieve size-tunable metal-mesh TCEs, such as laser sintering of nanoparticle ink, lithographic patterning, grain boundary lithography, nanoimprint lithography, and photolithography [28–31]. Photolithography is a standard fabrication process in semiconductor devices and integrated circuits, which determines their feature size (≤7 nm) and operation performance. It is believed that the photolithography process provides more precise control of the shape and size of the objects it creates and can create patterns over an entire surface cost-effectively [32]. In addition, the photolithography related processes are low-temperature technology and compatible to large area substrate (18 inches) or flexible substrates. Therefore, photolithography is a promising process to prepare highly consistent metal mesh for transparent electrode and solar cell applications [33–35].

We have proved in our previous work [36] that the adhesion of Ag on a glass substrate is relatively poorer than that of Ni, and Ni mesh processed by photolithography has been successfully used to fabricate PSCs, but the device performance (PCE = 5.74%) should be further improved. In this work, we optimize the thickness and shape of Ni meshes, Au is used to increase the conductivity of metal mesh, the high conductivity polymer PH1000 is employed to smooth the raised surface of metal mesh and enhance the charge collection ability of metal mesh, the optimized two-step MAyFA1-yPbIxCl3-x perovskite is used as the light absorber, and the fabricated ITO free PSCs have a structure of Glass/hexagonal Ni (30 nm)-Au (10 nm) mesh/PH1000/PEDOT:PSS/MAyFA1−yPbIxCl3−x/PCBM/BCP/Ag. The champion PSC (0.09 cm2) obtains a PCE of 13.88%, negligible current hysteresis, steady current density and PCE outputs, good repeatability and storing stability. Further, we tried to fabricate the large area (1 cm2) devices under the same process, and achieve a PCE over 6% with negligible hysteresis and stable steady-state outputs. The comparable performance to the ITO based reference PSC demonstrates that the Ni/Au mesh transparent electrodes are a promising ITO alternative to fabricate efficient PSCs.

#### **2. Materials and Methods**

#### *2.1. Materials*

All materials and reagents, Methylammonium iodide (MAI, 99.8%, Dyesol, Queanbeyan, Australia), Formamidinium iodide (FAI, 99.8%, Dyesol, Queanbeyan, Australia), Lead iodide (PbI2, 99.999%, Sigma-Aldrich, Saint Louis, MI, USA), Lead chloride (PbCl2, 99.999%, Sigma-Aldrich, Saint Louis, MI, USA), Phenyl-C61-butyric acid methyl ester (PCBM, 98%, Nano-C, Westwood, MA, USA), Poly(3,4-ethy-lenedioxythiophene) poly(styrenesulfonate) (PEDOT:PSS, Clevios PVP Al 4083, Heraeus, Hanau, Germany), high conductivity polymer PH1000 (Clevios PH1000, Heraeus, Hanau, Germany), N,N'-Dimethylformamide (DMF, 99.8%, Aladdin, Beijing, China), Chlorobenzene (CB, 99.8%, Sigma-Aldrich, Saint Louis, MI, USA), Bathocuproine (BCP, 96%, Sigma-Aldrich, Saint Louis, MI, USA) Isopropanol (IPA, 99.5%, Sigma-Aldrich, Saint Louis, MI, USA), photoresist (AZ6112, Ruicai, Suzhou, China), and developing solution (2.38%, NMD-3, Ruicai, Suzhou, China), were used as received without further purification.

#### *2.2. Metal Mesh Preparation*

Firstly, glass substrates (2 <sup>×</sup> 2.5 cm2) were ultrasonically cleaned in deionized (DI) water, acetone, ethyl alcohol, and DI water for 20 min, respectively. Secondly, the nitrogen gun was used to dry the substrates, and spin-coat the positive photoresist on substrates by two-steps of 500 rpm for 5s and 4000 rpm for 30 s with the acceleration of 4000 rpm/s, followed by baking on a hot plate at 100 ◦C for 90 s to cure the photoresist. Thirdly, the substrates were naturally cooled and exposed for 2.3 s under a shadow mask and then developed for 60 s in the developing solution. Then, the samples were rinsed in flowing DI water and dried by nitrogen gun. Fourthly, an optical microscope was employed to check the defined patterns and the samples were transferred into the E-beam evaporation system and nickel/gold (Ni/Au) mesh of x nm/10 nm were deposited under a pressure below 5 <sup>×</sup> 10−<sup>4</sup> Pa. After the metal mesh deposition, the samples were immersed in the acetone solution and lifted off by a low power ultrasonic bath for several ten seconds. The completed metal-mesh show a line width of 5 μm, and the active areas of 0.09 mm<sup>2</sup> and 1 cm2.

#### *2.3. PSCs Preparation and Characterization*

At first, the prepared metal-mesh substrates were UV-ozone treated by 15 min, and the high conductive polymer PH 1000 were spin-coated on the metal mesh at 1000 rpm for 60 s and annealed on a hotplate at 150 ◦C for 15 min. Secondly, the PEDOT:PSS was spin-coated at 6000 rpm for 45 s and 150 ◦C annealing for 15 min. Thirdly, the MAyFA1−yPbIxCl3−<sup>x</sup> precursor solution was prepared by mixing PbI2, PbCl2 that was dissolved in DMF solution, and stirred for 2 h at 75 ◦C, the solution was then spin coated onto the PEDOT/PSS layer at 3000 rpm for 45 s; Mixing FAI, MAI in the solvent of DMF, stirring at room temperature until completely dissolved, then spin coated it onto the PbI2/PbCl2 layer at 3000 rpm for 45 s; after thermally annealing at 100 ◦C for 10 min the perovskite layer was formed. Fourthly, the PCBM (20 mg/mL in CB) was spin-coated on the perovskite layer at 2000 rpm for 40s, and the BCP (0.5 mg/mL in IPA) was spin-coated on the PCBM at 6000 rpm for 45s. Finally, the Ag (100 nm) electrode was thermally evaporated under a shadow mask and the fabricated devices (seeing Figures S1 and S2 in Supporting Information) have an active area of 0.09 cm<sup>2</sup> and 1 cm2. As a reference, the ITO based PSCs, ITO/PEDOT:PSS/Perovskite/PCBM/BCP/Ag, were also fabricated under the same process conditions.

The current density-voltage (J-V) characteristics were measured with a source measurement unit of Keithley 2400 and simulated AM1.5 G sun light (100 mW/cm2, SEN-EI Electric. Co. Ltd, XES-300T1, Osaka, Japan). The incident photo-to-current conversion efficiency (IPCE) was measured by the quantum efficiency measurement system (SCS10-X150, Zolix instrument. Co. Ltd, Beijing, ChinaZolix Instrument. Co. Ltd). The four-point-probe system was utilized to measure the sheet resistance of electrodes. The UV–visible spectrophotometer (Perkin-Elmer Lambda 950, Waltham, MA, USA) was used to characterize the transmittance spectra of different samples. The film morphology was characterized by a JSM-7800F extreme-resolution analytical field emission scanning electron microscope (SEM) (JEOL Ltd., Tokyo, Japan) and atomic force microscopy (AFM) (Agilent 5500, Santa Clara, CA, USA). Electrochemical impedance spectroscopy (EIS) measurements were performed on an electrochemical workstation (CHI600E, Shanghai Chenhua, Shanghai, China) with a 10 mV amplitude perturbation and frequencies between 100 Hz and 1 MHz. M-S plots were recorded on the same system under AC excitation amplitude of 30 mV at a frequency of 5 kHz. Transient photocurrent (TPC) measurement was performed with a system excited by a 532 nm (1000 Hz, 3.2 ns) pulse laser. Transient photovoltage (TPV) measurement was performed with the same system excited by a 405 nm (50 Hz, 20 ms) pulse laser. A digital oscilloscope (Tektronix, D4105, Beaverton, OR, USA) was used to record the photocurrent or photovoltage decay process with a sampling resistor of 50 Ω or 1 MΩ, respectively. All the measurements were performed under ambient atmosphere at room temperature.

#### **3. Results and Discussion**

The geometry of ITO-free planar PSCs with an inverted structure based on a metal mesh transparent electrode is shown in Figure 1. Here, the two-step solution processed MAyFA1−yPbIxCl3−<sup>x</sup> perovskite layer is chosen as the light absorber; the PEDOT:PSS and PCBM act as the hole and electron transport layers (HTL and ETL), respectively; the BCP is further used to modify the electron collection at the PCBM/Ag interface. In particular, to obtain optimal metal mesh with high transmittance and low resistance, the photolithography process was chosen to precisely define the line width and space between metal lines. Two curial interfaces related to the metal mesh were designed to improve the properties of Ni/Au transparent electrode, then the corresponding performance of ITO-free PSCs with small and large active areas are discussed as follows.

**Figure 1.** Process of indium thin oxide (ITO)-free perovskite solar cells (PSCs) based on a metal mesh transparent electrode.

#### *3.1. Optimization of Ni*/*Au Mesh Transparent Electrode*

There are two curial interfaces introduced by the metal mesh, including the metal/substrate and metal/HTL interfaces. First, the good adhesion of metal mesh on substrate is necessary to fabricate the PSCs. Although the silver material possesses the lowest resistivity, it has been found in our previous work [36] that the adhesion of Ag mesh on glass substrates is relatively weak, and the mesh is easily damaged in the lift-off process, thus the Ni mesh is chosen to fabricate PSCs. However, the relatively low conductivity of Ni limits the improvement of device performance. In this work, an ultrathin Au is introduced to enhance the conductivity of Ni, and the shape and thickness of Ni/Au mesh has been further optimized. On the other hand, as the PEDOT:PSS HTLs (about 10 nm) is prepared by solution coating, the film quality is highly related to the smoothness of the underlying layer. Considering the thickness of the Ni/Au metal mesh has exceeded 50 nm, the high conductivity PHI000 (about 100 nm) is deposited to smooth the raised surface of the metal mesh, and simultaneously enhance the hole collection ability of Ni/Au electrode. It is noted that the PH1000 only causes a little loss of transmittance in the visible region [5], therefore the Ni/Au/PH1000 electrode is selected in the ITO-free PSCs.

Figure 2a provides the transmittance of spectra of Ni/Au mesh (square, hexagon) and ITO on glass, and the glass substrate. It can be seen that the glass shows the highest transmittance at over 90% in 300 nm to 850 nm, the commercial ITO coated glass substrate possesses an average transmittance over

80%, with the oscillation behavior related to the optical interference at the ITO/glass interface. For the Ni/Au meshes, although the highest transmittance is lower than the ITO sample, a relatively higher transmittance is obtained in the 300–370 nm and 420–500 nm regions. Their average transmittance has also exceeded 80% and the hexagon mesh shows a slightly low transmittance compared to the square mesh. Consequently, the optical performance of the Ni/Au mesh can meet the requirement of the transparent electrode. For the electrical property, the four-point-probe method is used to measure their sheet resistances. The corresponding values are also in Figure 2, showing that the hexagon mesh (22.6 Ω/sq) has lower resistance than the square mesh (30.7 Ω/sq), but larger resistance than that of the ITO (10 Ω/sq). From Figure S3, it is clear that the PSC with the Ni/Au square mesh electrode shows better performance compared to the device with a pure Ni electrode (58 Ω/sq); and the PSC with a hexagon Ni/Au electrode performs better than the device with a square Ni/Au electrode. Their improved PCEs mainly come from the increased JSC values, which is in line with the results of sheet resistance. Thus, the hexagon Ni/Au mesh should be more suitable for fabricating ITO-free PSCs, and its thickness is further optimized. Here, the thickness of Au is fixed on 10 nm, and the thicknesses of Ni vary by 10 nm, 20 nm, 30 nm, and 40 nm. Table 1 shows the corresponding sheet resistances of the Ni/Au meshes. When the thickness of Ni increases from 10 to 40 nm, the sheet resistance values are 33.8 Ω/sq, 22.6 Ω/sq,16.9 Ω/sq and 13.6 Ω/sq, respectively. It is clear that the thicker the metal is, the lower the sheet resistance is. However, too thick a metal will lead to a worse lift-off effect, such as uneven edges of the metal wire and partial fracturing of the grid, which will induce a weak conductivity in the electrode. Meanwhile, if the thickness is less than 10 nm, the adhesion between Ni and glass substrate becomes very poor. Also, with metal deposited by electron beam evaporation it is difficult to form a homogeneous film at a thickness lower than 10 nm, as that will lead to a weaker conductivity [37]. Therefore, the thickness of Ni is a trade-off, which should be further determined by the device performance. As shown in Figure 2b and Table 1, all the devices show similar VOC and FF values, and the JSC dominants the overall PCEs. When the thickness of Ni is 10 nm or 40 nm, a lower PCE of about 10% is limited by the relatively poor JSC values of 15.43 mA/cm<sup>2</sup> and 16.16 mA/cm2. While if the 20 nm or 30 nm Ni is used, a JSC exceeding 20 mA/cm<sup>2</sup> can be achieved for PSCs, and the PSCs with30 nm Ni obtain superior PCE of 13.72%. As a result, the optimal metal mesh transparent electrode should be the hexagon Ni (30 nm)/Au (10 nm) mesh electrode, and the corresponding device performance is further investigated.

**Figure 2.** (**a**) Transmittance spectra of Ni (20 nm)/Au (10 nm) mesh (square, hexagon) and ITO on glass, and the glass substrate; (**b**) density-voltage (J-V) curves for PSCs with hexagon Ni(x nm)/Au (10 nm) (x = 10, 20, 30, 40 nm) metal mesh electrodes. The sheet resistances of Ni/Au and ITO electrode are also displayed in Figure 2.


**Table 1.** Photovoltaic parameters of PSCs with various Ni(x nm)/Au (10 nm) meshes.

#### *3.2. Performance of PSCs with Optimal Ni*/*Au Metal Mesh*

Figure 3a presents the J-V curves of the champion PSC with the electrode and reference PSC with an ITO electrode. It can be seen that the PSC based on Ni/Au mesh obtains a champion PCE of 13.88%, VOC of 0.94 V, JSC of 21.14 mA/cm2, and FF of 69.75% at reverse voltage scan direction; while the forward scanned PCE is 13.39% with VOC = 0.93 V, JSC = 21.04 mA/cm2, and FF = 68.42%(Figure 3c). This negligible current hysteresis in the PSC based on Ni/Au mesh electrode is strongly related to the effects of PCBM passivation and BCP interface medication, which has been proved in our previous works [3,4]. Further, the steady-state outputs of current density and PCE at the maximum power point voltage of PSC based on Ni/Au mesh are shown in Figure 3d, the nearly unchanged current density and PCE outputs during 160 s illustrate the good operation stability of PSC. At the same time, the reference ITO based PSC shows a PCE of 15.70%, VOC of 0.96 V, JSC of 21.60 mA/cm2, and FF of 75.80%. Meanwhile, as shown in Figure S4, the lower leakage current and better rectification characteristics are in line with the relatively higher performance of PSC based on the ITO electrode. It is clear that the PSC based on Ni/Au mesh can obtain comparable PCE to that of the ITO based PSC, which demonstrates that the hexagon Ni/Au mesh is a promising ITO alternative. On the other hand, compared to ITO based PSC, the low VOC and FF of Ni/Au based PSC may be explained by the relatively weak charge transport and collections ability. As J-V curves under AM 1.5 G illumination reveal the photovoltaic performance of PSC in the entire absorption range, to further study the photovoltaic conversion at single incident light wavelengths, the IPCE spectra are measured and the results are shown in Figure 3b. The integrated JSC values of 19.29 mA/cm2 (Ni/Au-PSC) and 19.95 mA/cm<sup>2</sup> (ITO-PSC) agree well with the measured JSC in J-V measurements, which manifest the dependability of the J-V curve measurement. However, it should be noted that the IPCEs of ITO based PSC are higher than that of Ni/Au based PSC at each single wavelength from 300 nm to 800 nm. As the overall optical transmittance of Ni/Au and ITO are close, the perovskite film quality and the charge transport properties may account for the relatively low performance of the Ni/Au based PSC.

First, as the same preparation method of perovskite film has been evaluated in our previous work [3], in the present work the primary concern is the morphology of perovskite films. Figure 4a,c and Figure 4b,d show the SEM and AFM images of perovskite film based on Ni/Au mesh and ITO electrodes. It has been observed that both the perovskite films are compact and smooth, there is no significant difference in the grain sizes and root-mean-square roughness (RMS) values. From this it can be understood that the PH1000 buffer layer has smoothened the raised surface of Ni/Au mesh, which ensures a similar film morphology of perovskite on PEDOT:PSS/PH1000/Ni/Au/Glass and PEDOT:PSS/ITO/Glass samples. Here, we extracted the series resistance (Rs) and shunt resistance (Rsh) from the J-V curves by the parameter extraction method in our previous work [38]. As show in Figure S5, the experimental data are well reproduced by the fitting curves and the corresponding Rs, Rsh, ideality factor, saturation current are listed in Table S1. It is clear that the PSC based on Ni/Au mesh show a RS of 1.2 <sup>Ω</sup>·cm2 and RSH of 5.548 kΩ·cm2, while that of the ITO based PSC are 1.0 <sup>Ω</sup>·cm2 and 6.398 kΩ·cm2, combining the larger ideality factor and saturation current values, thus the more efficient carrier transport contributes to the better performance of ITO based PSC. To further study the charge transport properties, transient photocurrent (TPC) and transient photovoltage (TPV) measurements are carried out. The TPC can reflect extraction and transport property of carriers, and the TPV provides insight to carrier recombination property in a solar cell. As shown in Figure 5a, the ITO based PSC has a relatively faster photocurrent decay (0.46 μs) compared with the Ni/Au based PSC (1.96 μs), suggesting that the extraction and transport of carriers are more efficient in ITO based PSC. Meanwhile, the photovoltage decay processes are displayed in Figure 5b, the fitted charge recombination lifetimes are 1389.04 μs and 1183.34 μs for PSC based on ITO and the Ni/Au mesh, respectively. What is more, EIS measurements were carried out to evaluate the carriers' recombination in the corresponding PSCs. The corresponding Nyquist plots, the equivalent circuit diagram, and fitted results are shown in Figure S6 and Table S2. It can be found that, the PSC based on Ni/Au mesh achieved a relatively smaller Rrec (carrier recombination resistance) and shorter electron lifetime. Simultaneously, the sheet resistance of Ni (30 nm)/Au (10 nm) mesh (16.9 Ω/sq) is higher than that of ITO. Combining the results of AFM, SEM, TPC, TPV, and conductivity, the relatively low performance of PSC based on Ni/Au mesh can be understood.

**Figure 3.** (**a**) J-V curves and (**b**) incident photo-to-current conversion efficiency (IPCE) spectra of PSCs based on Ni/Au mesh and PSCs based ITO measured under 100 mW/cm<sup>2</sup> AM 1.5 G illumination; (**c**) J-V curves in forward and reverse scans and (**d**) steady output current density and PCE of PSCs based on Ni/Au mesh and the maximum power point (Vmax = 0.67 V).

In addition, the statistical results of photovoltaic parameters and the store stability of Ni/Au based PSCs are shown in Figures 6 and 7, the device number in the statistical analysis is 15. From Figure 6, it can be seen that most of the VOC values concentrate in the range from 0.84 V to 0.98 V and eight of them exceed 0.92 V; the JSC ranges from 14 mA/cm<sup>2</sup> to 24 mA/cm2, with eight of them higher than 19 mA/cm2; the FF from 62% to 76%, with nine of them larger than 69%; the PCE from 9% to 14% with eight of them exceeding 12%. All the photovoltaic parameters exhibit Gaussian distribution, which suggests the good reproducibility of Ni/Au based PSCs. Here, their FF and VOC values should be further improved to obtain high performance PSCs. Besides the optimization of Ni/Au mesh electrode, more efficient charge transport layers [3,39], and interface passivation [1,40] are feasible approaches to achieve this goal. Furthermore, from Figure 7, the PSCs based on Ni/Au mesh and ITO present a similar trend of PCE degradation. After being stored in N2 for 240 h, the Ni/Au based PSCs keep 58% of their initial PCE values, and the ITO based devices maintain 60% of their initial PCE values. Therefore, the PSCs based on Ni/Au mesh possess comparable photovoltaic performance and stability

to the reference ITO based PSCs, and the Ni/Au mesh is one of the promising ITO alternatives to fabricate PSCs. Nowadays, owing to the improvement of perovskite quality and interface modification, the PCE of planar ITO based PSCs has exceeded 23% [1]. It is believed that there is plenty of room for performance improvement of PSCs based on metal mesh transparent electrodes, which will be further investigated in our future works.

**Figure 4.** SEM and AFM images of perovskite film based on (**a**,**c**) Ni/Au mesh and (**b**,**d**) ITO electrodes. The unit of RMS values are nm.

**Figure 5.** (**a**) Transient photocurrent (TPC) and (**b**) transient photovoltage (TPV) decay curves of the PSCs based on Ni/Au mesh and ITO electrodes.

**Figure 6.** Histograms of photovoltaic parameters. (**a**) JSC, (**b**) VOC, (**c**) FF, and (**d**) PCE for PSC based on Ni/Au mesh electrode. The histograms shown are the photovoltaic parameters for 15 devices.

**Figure 7.** Storing stability of Ni/Au mesh and ITO based PSCs without encapsulation in N2 for 240 h.

What is more, the large-area (1 cm2) Ni/Au mesh-based PSCs are fabricated under the same process conditions as the small-area (0.09 cm2) PSCs. Figure 8a,b shows the J-V curve, photovoltaic parameters, and steady outputs of large-area Ni/Au based PSCs and ITO based reference PSCs. The Ni/Au based PSC obtains a PCE of 6.01%, with VOC = 1.04 V, JSC = 11.28 mA/cm2, and FF = 51.28% at the reverse

voltage scan direction, as well the steady PCE and current density outputs under continuous AM 1.5 G illumination during 160 s. The forward scanned PCE = 5.93%, VOC = 1.05 V, JSC = 10.81 mA/cm2, and FF = 52.27%, thus the current hysteresis effect is negligible. While the ITO based PSC shows a higher PCE of 9.01% with VOC = 1.02 V, JSC = 16.24 mA/cm2, and FF = 55.59%, it should be noted that the JSC and FF of large area PSCs are obviously lower than the small area PSCs (Figure 3), which is due to relatively poor charge extraction and strong charge recombination related to the more defects and traps in the larger-area perovskite layer. This can be partly proved by the TPC and TPV results. As shown in Figure 8c,d, the large-area PSC displays a slower photocurrent decay (3.59 μs) compared to the small-area cell (1.96 μs), suggesting inefficient charge transport and extraction in the large-area cell. Meanwhile, a faster photovoltage decay (644.49 μs) than the small-area one (1183.34 μs) reveals the relatively strong charge recombination in the large-area cell. Therefore, the perovskite film quality and interface modification are crucial to further improving the performance of large-area devices. Considering the drawbacks of ITO, the meal mesh based large-area PSCs may find wider applications in the near future.

**Figure 8.** (**a**) J-V curves in forward and reverse scans for large-area PSCs based metal mesh, (**b**) steady output characteristics curve, (**c**) TPC and (**d**) TPV decay curves.

#### **4. Conclusions**

In summary, we carefully designed and deposited hexagon Ni/Au metal mesh transparent electrodes with high transmittance and low resistance by photolithography and the e-beam evaporation process. To be an efficient electrode for PSCs, the Ni was used to improve the adhesion of metal mesh to glass substrate and the Au was used to increase the conductivity of the metal mesh. The conductive polymer PH1000 not only smooths the raised surface of metal mesh but also enhances the charge collection ability of metal mesh. The optimal hexagonal Ni (30 nm)/Au (10 nm)/PH1000 electrodes were employed to fabricate ITO-free PSCs. The champion PSC (0.09 cm2), with a typical planar structure, obtained a PCE of 13.88%, negligible current hysteresis, steady current density and PCE outputs, good repeatability and storing stability. The comparable performance to the ITO based reference PSC demonstrates that the Ni/Au mesh transparent electrodes are a promising ITO alternative to fabricate efficient PSCs. Further, we tried to fabricate the large area (1 cm2) devices under the same low-temperature process, and achieved a PCE over 6% with negligible hysteresis and stable steady outputs. And the perovskite film quality and interface modification are crucial to further improving the performance of large-area devices. Considering the drawbacks of ITO, the meal mesh-based PSCs may find wider applications in the near future.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/7/932/s1, Figure S1: Photos of ITO-free PSCs (0.09 cm2) based on Ni/Au mesh electrode (a) light incident surface (b) backlight surface, Figure S2: Photos of large-area ITO-free PSCs (1 cm2) based on Ni/Au mesh electrode (a) light incident surface (b) backlight surface., Figure S3: J-V curves of ITO-free PSCs (0.09cm2) based on Ni (30 nm) square, Ni(20 nm)/Au(10 nm) square, and Ni(20 nm)/Au(10 nm) hexagon mesh electrodes. The sheet resistance of pure Ni and Ni/Au meshes are about 58 Ω/sq, 30.7 Ω/sq, and 22.6 Ω/sq, thus the PSC with Ni/Au hexagon electrode obtains higher Jsc, FF, and PCE values, Figure S4: Semi-log plots of dark JV curves for PSC based on Ni/Au mesh and ITO electrodes, Figure S5: JV curves under AM 1.5G illumination for PSC based on Ni/Au mesh and ITO electrodes. Here the symbols represent the experimental data and the solid lines indicate the fitting curves, Figure S6: Nyquist curves of PSCs based on Ni/Au mesh and ITO electrodes. The insert is the equivalent circuit for the fittings, where Rs represents the series resistive elements related to connections and devices, Rrec the carrier recombination resistance and, CPE the constant phase element.; Table S1: Rs, Rsh, saturation current, and ideality factor extracted from JV curves under AM 1.5G illumination, Table S2: EIS parameters extracted from Nyquist curves. The electron lifetime is the reciprocal of the frequency of the maximum point of the semi-circular response.

**Author Contributions:** D.C. and C.Z. conceived the idea, designed the experiment and guided the experiment. G.F., H.Z. and L.Z. conducted most of the electrode preparation, device fabrication, data collection and wrote the manuscript; D.C. and C.Z. revised the manuscript; W.Z., H.X. helped the device measurement, X.H. and Z.L. helped the film characterization, H.D. and S.P. helped the data analysis. J.Z. and Y.H. supervised the group. All authors read and approved the manuscript.

**Funding:** This work is partly supported by the Fundamental Research Funds for the National 111 Center (Grant No.B12026), Fundamental Research Funds for the Central Universities (XJS191107), Shaanxi Postdoctoral Research Grant Program (30102180038), Class General Financial Grant from the China Postdoctoral Science Foundation (2016M602771), and Natural Science Foundation of China (61804113, 61704128, 61704131, and 61604119).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Dopant-Free Hole Transport Materials with a Long Alkyl Chain for Stable Perovskite Solar Cells**

#### **Kai Wang 1,**†**, Haoran Chen 1,**†**, Tingting Niu 1, Shan Wang 1, Xiao Guo <sup>1</sup> and Hong Wang 2,\***


Received: 22 May 2019; Accepted: 24 June 2019; Published: 28 June 2019

**Abstract:** Hole transport materials are indispensable to high efficiency perovskite solar cells. Two new hole transporting materials (HTMs), named 4,4- -(9-nonyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA-1) and 4,4- -(9-methyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline)(CZTPA-2), were developed by different alkyl substitution methods. The two compounds, containing a carbazole core and triphenylamine (TPA) groups with different lengths of the alkyl chain, were designed and synthesized through a two-step synthesis approach. The power conversion efficiency (PCE) was found to be affected by the length of the alkyl chain, reaching 7% for CZTPA-1 and 11% for CZTPA-2. Furthermore, the CZTPA-2 still maintained 89.7% of its original performance after 400 h. The proposed results demonstrate the effect of carbon chain substituents on the efficiency of perovskite solar cells (PSCs).

**Keywords:** perovskite solar cell; alkyl chain; hole transporting materials; stable

#### **1. Introduction**

Perovskite solar cells (PSCs) have attracted much attention in recent years owing to their excellent optoelectronic properties, high absorption coefficients, easy solution processability, high carrier mobility, and so on [1–5]. Many efforts have been made to increase the efficiency of PSCs. Surprisingly, it has increased dramatically from 3.8% [1] to 24.2% [6] in a short period of time.

As is well-known, 2,2- ,7,7- -Tetrakis[*N*,*N*-di(4-methoxyphenyl)amino]-9,9- -spirobifluorene (Spiro-OMeTAD) [7,8] is a common hole transporting material (HTM) with good performance. However, it is still a formidable challenge to develop this material due to various drawbacks, such as a complicated synthesis process and high cost. Therefore, new HTMs with a simple synthesis process and low cost need urgently to be developed.

HTMs construct the hole transport layer (HTL) that is needed to block electron transport, enhance hole transport, and prevent direct contact between the perovskite layer and the electrode, which causes annihilation. An ideal HTM should have such characteristics as good hole mobility, good hydrophobicity, suitable energy levels, and the ability to be prepared in solution [9–11]. Currently, common HTMs can be classified as inorganic substances, organic polymers, or organic small molecules depending on the type of material. Low cost, high stability, and hole mobility are among the many advantages that HTMs based on inorganic materials have. Kamat and co-workers first introduced copper iodide as an HTM and achieved an average PCE of 6% [12]. Then, inorganic semiconductors (CuSCN, NiO, CuI) as HTMs were developed [13,14]. In addition to inorganic HTMs, polymer-based HTMs have been explored in PSCs. Poly-[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA) was

the first conjugated polymer to be used as an HTM in PSCs [15] and maintained the highest PCE of any reported polymeric HTM [16]. Accompanied by the development of conjugated polymer HTMs, Poly(3-hexylthiophene-2,5-diyl (P3HT) [17] was originally used in organic solar cells (OSCs) as the active layer was introduced into the hole transport layer. Compared to polymeric HTMs, small molecule HTMs have the advantages of a determined molecular weight and simple purification for PSCs. As a group with good stability and solubility, triphenylamine (TPA) is widely used in organic small molecule HTMs [18–20]. The TPA group in particular can influence the optoelectrical properties of HTMs due to its non-planar geometry. Furthermore, the carbazole group has a high carrier mobility in OSCs and as good a performance as HTMs; thus, it is currently regarded as a core in PSCs [21–27].

In 2014, Sun and co-workers [27] designed a series of HTMs based on carbazole, named X19 and X51. The X51 realized a PCE of 9.8% due to a higher charge-carrier mobility and conductivity than X19. Subsequently, Nazeeruddin and co-workers [28] studied bridged carbazole with biphenyl, by using silolothiophene as the bridge, which obtained a PCE of 13.1%. They found that compared to the spirofluorene-linked triphenylamine HTMs, novel silolothiophene-linked methoxy triphenylamines (Si-OMeTPAs) enable more stable PSCs. In the same year, Tang and co-workers [29] used carbazole as a core with four TPAs as the side groups to achieve a PCE of up to 18.32%.

Recently, Khaja Nazeeruddin and co-workers [30] used anthra[1,2-b:4,3-b- :5,6-b- :8,7 b---]tetrathiophene as the core, by changing the alkyl chain length of the methoxy groups on the triarylamine sites, to develop a series of materials. The device based on a methyl substitute named ATT-OMe was found to have the best PCE of 18.13%, which was better than that of the device based on other longer alkyl chain HTMs. They believed that the presence of alkyl chains decreased the hole-transport properties. However, the performance of PSCs based on carbazole is far from that of the classical Spiro-OMeTAD. Therefore, more efforts are needed to develop new small molecule HTMs that match with perovskite instead of the Spiro-OMeTAD.

In this work, we developed two HTMs—4,4- -(9-nonyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA-1) and 4,4- -(9-methyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA-2)—based on TPA as the end group. They both have the advantage of simple synthesis steps and low cost. Both HTMs are synthesized by one-step Suzuki coupling. The cost of the raw material 3,6-Dibromocarbazole (\$0.3/g) plus 4-methoxy-*N*-(4-methoxyphenyl)-*N*-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)aniline (\$15/g) is clearly lower than that of Spiro-OMeTAD (\$220/g). Notably, the CZTPA-2 with the longer alkyl chain achieved a better PCE of 11.79% with a short current density (*J*sc) of 21.80 mA/cm2, an open circuit voltage (*V*oc) of 0.99 V, and a fill factor (FF) of 54.59%. This is attributed to the significantly improved hole mobility of CZTPA-2, resulting in a significant increase in device efficiency.

#### **2. Materials and Methods**

#### *2.1. Materials*

Unless otherwise noted, all reagents used in the experiments were purchased from commercial sources and used without further purification. 3,6-Dibromo-9H-carbazole, *N*,*N*-bis(4-Methoxyphenyl)-4-(4,4,5,5-tetraMethyl-1,3,2-dioxaborolan-2-yl)-BenzenaMine, lead iodide (PbI2), methylammonium iodide (MAI), acetonitrile (99.8%), chlorobenzene (99.9%), and dimethylformamide (DMF) (99%) were purchased from Sigma-Aldrich. 4-tert-butylpyridine (TBP) and Li-bis-(trifluoromethanesulfonyl) imide (Li-TFSI) were purchased from TCI. 2,2- ,7,7- -tetrakis-(*N*,*N*-di-p-methoxyphenylamine)-9,9'-spirobifluorene (Spiro-OMeTAD) (99.0%) was purchased from Xi'an Polymer Light Technology Co., Ltd.

Perovskite precursor: The perovskite precursor was obtained by mixing PbI2 and MACl (in a molar ratio of 1:1) in DMF with a concentration of 350 mg/mL, and was then stirred at 60 ◦C overnight in a glovebox.

Spiro-OMeTAD: The 2,2- ,7,7- -Tetrakis(*N*,*N*'-di-p-methoxyphenylamine)-9,9'-spirobifluorene (Spiro-OMeTAD) was doped with TBP and Li-TFSI. A total of 73.2 mg of Spiro-OMeTAD (Xi'an Polymer Light Technology Co., Ltd., Xi'an, China) was dissolved in 1 mL of chlorobenzene (CB) with 28.8 μL of 4-tert-butylpyridine (TBP) and 17.6 μL of Li-bis-(trifluoromethanesulfonyl) imide (Li-TFSI).

#### *2.2. Device Fabrication*

The SnO2 layer was spin-coated on an ITO substrate at 3000 rpm for 30 s, which was cleaned in UV–ozone and then annealed at 150 ◦C for 30 min. The perovskite layer was spin-coated at 4000 rpm for 30 s by an anti-solvent method. The details of the operation are as follows: 100 μL CB was rapidly added after 5 s of spin-coating with perovskite solution; and the perovskite films were annealed at 100 ◦C for 5 min. All of the above processes were performed in the nitrogen glovebox. Then, Spiro-OMeTAD and two HTMs were dissolved in the CB (10 mg/mL), and then spin-coated upon the perovskite layer at 3000 rpm for 30 s. Then, molybdenum trioxide (MoO3) and gold (Au) were thermally evaporated on the hole transporting layer. The effective area of the cell is 0.05 cm2.

#### *2.3. Device Characterization*

The cross-sectional images of PSC were taken by scanning electron microscopy (SEM) (ZEISS Merlin, Carl Zeiss Microscopy, Jena, Germany). The UV–visible absorption spectra were measured using a UV Spectrophotometer (SHIMADZU UV-1750, East Test Technology Co., Ltd, Shenzhen, China). Photoluminescence (PL) spectra were obtained using a spectrofluorometer (HitachiF-7000, Hitachi High-Technologies Corporation, Shenzhen, China). Thermogravimetric (TGA) analysis was performed on a Mettler Toledo TGA2.

The device was measured under AM 1.5 G solar irradiation with an intensity of 100 mW/cm2 through an Enlitech SS-F5-3A solar simulator. The instrument was calibrated on standard solar cells. *J–V* properties were measured by the method of Enlitech Ltd. (Kaohsiung, Taiwan) and a Keithley (Cleveland, OH, USA) 2400 source meter under dark conditions. The external quantum efficiency (EQE) spectra were measured using a solar cell IPCE test system (CROWNTECH Inc., model QTEST HIFINITY (Macungie, PA, USA)).

The synthesis method and details of the experimental procedure are shown in the supporting information (SI).

#### **3. Results**

The details of the experimental procedure are shown in the Supplementary Information (SI). The design principle was to improve planarity as well as increase solubility. The carbazole group with simple structures has good hole transporting ability. By inserting N atoms with different alkyl chain lengths, the optoelectronic properties and solubility can be adjusted. Herein, we chose 4-methoxy-*N*-(4-methoxyphenyl)-*N*-(4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)phenyl)aniline units as the raw material, used Suzuki coupling, and substituted the carbazole core with different lengths of alkyl chains. The aimed-for materials were obtained by simple column chromatography separation and recrystallization.

Figure 1a shows the absorption spectra of CZTPA-1 and CZTPA-2 in chloroform and as a coating on quartz substrates. Absorption peaks at 335 nm for CZTPA-1 and 327 nm for CZTPA-2 in the solution were observed. Relative to the solution, the CZTPA-1 film exhibits a redshift of 2 nm with an onset of 409 nm, corresponding to an optical bandgap of 3.03 eV, whereas CZTPA-2 aligns to a narrower bandgap of 2.98 eV (onset of 416 nm). Both HTMs have a slight redshift in the film compared with the solution, suggesting that aggregation exists in the films.

**Figure 1.** (**a**) Optical absorption spectra of 4,4- -(9-nonyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA-1) and 4,4- -(9-methyl-9H-carbazole-3,6-diyl)bis (*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA-2) in chloroform and thin films spin-coated from chloroform. (**b**) Cyclic voltammetry of CZTPA-1 and CZTPA-2.

From Figure 1b, the highest occupied molecular orbital (HOMO) level of CZTPA-1 is −4.89 eV, while in the CZTPA-2 there is a higher HOMO of −4.83 eV (the energy level of Ferroceneis is −0.54 eV by the cyclic voltammetry method). Compared to the two HTMs, the Spiro-OMeTAD has the higher energy level as the HOMO is −4.64 eV. Based on the relationship of *E*LUMO = *E*HOMO + *E*g, the lowest unoccupied molecular orbital (LUMO) level value was calculated to be −1.86 eV for CZTPA-1, −1.85 eV for CZTPA-2, and −1.74 eV for Spiro-OMeTAD (Table 1). CZTPA-2 has similar electrochemical properties compared with CZTPA-1, which indicates that the electrochemical properties of the molecule remain stable with a change in the alkyl chain length.


**Table 1.** Optical and electrochemical properties of CZTPA-1 and CZTPA-2.

(a) Maximum absorption peak in CH2Cl3 solution; (b) Maximum absorption peak of films on quartz glass; (c) Optical bandgap calculated from the absorption onset of films: *E*<sup>g</sup> opt = 1240/λonset eV; (d) *E*LUMO = *E*HOMO + *E*<sup>g</sup> opt.

We then found that CZTPA-1 and CZTPA-2 match well with the energy level of the perovskite from Figure 2a. Furthermore, the thermal stability of the two HTMs was measured by thermogravimetric analysis (TGA) (Figure 2b). Both CZTPA-1 and CZTPA-2 exhibit good thermal stability with decomposition temperatures (Td, 5% weight loss) at 391.8 ◦C and 384.6 ◦C, respectively. As we expected, increasing the length of the alkyl chains reduces the thermal stability.

The photoluminescence (PL) spectra in Figure 2c show the maximum emission peak at 431 nm for CZTPA-1 and at 425 nm for CZTPA-2 in film. CZTPA-2 was slightly more blue-shifted than CZTPA-1, caused by the increase in the length of the alkyl chains.

Figure 2d shows the PL spectra of the perovskite, perovskite with Spiro-OMeTAD, perovskite with CZTPA-1, and perovskite with CZTPA-2. Strong PL quenching was observed after the HTMs were coated on perovskite films. Respectively, compared with the original perovskite film, the PL intensity was reduced to 7%, 22%, and 18% after coating with Spiro-OMeTAD, CZTPA-1, and CZTPA-2. Thus, we think that CZTPA-2 has better charge separation than CZTPA-1 and a smaller *Jsc* and FF in the PSCs compared to Spiro-OMeTAD devices. In short, this means that the hole transfer capabilities of CZTPA-2 are superior because of their better charge transfer capability.

**Figure 2.** (**a**) Energy level diagram of a perovskite solar cell (PSC) with CZTPA-1, CZTPA-2, and 2,2- ,7,7- -Tetrakis[*N*,*N*-di(4-methoxyphenyl)amino]-9,9- -spirobifluorene (Spiro-OMeTAD). (**b**) Thermogravimetric analysis (TGA) diagrams of CZTPA-1 and CZTPA-2. (**c**) Photoluminescence (PL) spectra of CZTPA-1 and CZTPA-2 thin film, excitation at 350 nm. (**d**) Photoluminescence spectra of perovskite, perovskite with Spiro-OMeTAD, perovskite with CZTPA-1, and perovskite with CZTPA-2, excitation at 500 nm.

Figure 3 shows two cross-sectional scanning electron microscopy (SEM) images of the PSC with the structure of ITO/SnO2/perovskite/HTMs/Au. The PSC includes an ≈460 nm perovskite capping layer and an ≈35 nm HTM layer (CZTPA-1 or CZTPA-2). From the SEM image, we found that the HTM layer deposited on the perovskite layer and the boundaries of each layer are clear. Atomic force microscopy (AFM) images of the two materials spin-coated onto perovskite films are shown in Figure S7, which represent the perovskite/CZTPA-2 (a) and the perovskite/CZTPA-1 (b) films, respectively. The roughness of both samples is slightly high, which may be due to the lower thickness of the HTMs. After a comparison, it was found that the roughness of the film in which the HTM is CZTPA-2 (RMS = 20.968 nm) is significantly lower than that of CZTPA-1 (RMS = 28.662 nm); this is attributed to the solubility of CZTPA-2 being higher such that it could better cover the film and further improve the carrier transport of the device.

**Figure 3.** (**a**) A cross-sectional SEM image of the CZTPA-1 PSC. The scale bar is 200 nm. (**b**) A cross-sectional SEM image of the CZTPA-2 PSC. The scale bar is 200 nm.

In order to compare the properties of the two HTMs, we used them as the hole transport layer of perovskite solar cells to compare device performance. Between them, the effective area of the cells is 0.05 cm2, and the scanning rate is 0.02 V/s. As shown in Figure 4a,b, the PSCs with CZTPA-2 achieve a PCE of 11.79% with an open-circuit voltage (*V*oc) of 0.99 V, a short-circuit current density (*J*sc) of 21.8 mA cm<sup>−</sup>2, and a fill factor (FF) of 54.59%, while the PSCs with CZTPA-1 achieve a lower PCE of 6.05% under the condition of no doping. In contrast, the best cell is based on Spiro-OMeTAD, as the HTM achieves the PCE of 16.77%. We also compared the Spiro-OMeTAD PSC without any doping, which achieved a PCE of 11.74%. The photovoltaic performance of the PSCs based on the two HTMs (dopant-free) and the Spiro-OMeTAD were investigated under AM 1.5 illumination (100 mW cm<sup>−</sup>2). The champion device performance plots of CZTPA-1 and CZTPA-2 are shown in Table 2. The device fabricated with CZTPA-1 as the HTM yielded a promising PCE of 6.05% with a *J*sc of 20.58 mA cm<sup>−</sup>2, a *V*oc of 0.77 V, and an FF of 38.01%. We calculated the average PCE of the CZTPA-2 devices to be 10.15% ± 0.90. The CZTPA-1 devices own an average PCE of 5.27% ± 0.57. The average PCE of the Spiro-OMeTAD (dopant-free) is 10.02% ± 0.98, and the corresponding PCE of Spiro-OMeTAD is 15.65% ± 0.71. All data are based on the values obtained from 20 devices. From these, we can speculate that CZTPA-2 exhibits better performance compared with CZTPA-1. The FF and *V*oc of CZTPA-1 are obviously lower than those of CZTPA-2 in the PSC devices. Furthermore, CZTPA-2 obtains a slightly higher PCE compared with the dopant-free Spiro-OMeTAD, which is attributed to the higher *J*sc.

**Figure 4.** (**a**) *J–V* characteristics of PSCs based on the two hole transport materials (HTMs), the Spiro-OMeTAD, and the Spiro-OMeTAD (dopant-free). (**b**) External quantum efficiency (EQE) and *J*sc spectra of PSCs with CZTPA-2. (**c**) Histograms of PCEs measured in 20 cells of CZTPA-1. (**d**) Histograms of PCEs measured in 20 cells of CZTPA-2.


**Table 2.** Photovoltaic data of PSCs based on the two HTMs, the Spiro-OMeTAD, and the Spiro-OMeTAD (dopant-free).

When preparing the solution, we found the CZTPA-1 dissolution rate to be slower than the CZTPA-2 dissolution rate. What is more, the solubility of CZTPA-2 was found to be better than that of CZTPA-1 with an increasing alkyl chain length and CZTPA-2 deposits in perovskite with better crystallization. These results also demonstrate that CZTPA-2 achieves better performance.

To test the reproducibility of the devices, we fabricated 20 devices in several different batches. The devices are shown in Figure 4c,d. As shown in the PCE histogram of the corresponding device data, the average PCE of CZTPA-2 and CZTPA-1 is 10.15% and 5.27%, respectively.

Moreover, the EQE spectrum of PSCs with CZTPA-2 is also shown in Figure 4b. The integral of the current densities calculated from the EQE spectra is 21.32 mA cm−<sup>2</sup> for CZTPA-2 according predominantly to the experimental data.

To calculate the hole mobility of the two HTMs, we constructed a device with a configuration of ITO/PEDOT:PSS/HTM/Au using the space-charge-limited current (SCLC) method, and the *J–V* characteristics of this device were studied in the dark. Hole mobility is calculated by the Mott–Gurney equation of *J* = 9εrε0μVa2/8L3, so we change the form of the formula to obtain μ = 8d3/9εrε0(*J* <sup>1</sup>/2/Va)2, where ε<sup>r</sup> is the relative dielectric constant of the transport medium (ε<sup>r</sup> = 3 for organic materials), ε<sup>0</sup> is the permittivity of free space (8.85 <sup>×</sup> 10−<sup>12</sup> C V−<sup>1</sup> m−1), *J* is the dark current density (mA cm−2), and d is the thickness of the active layer [31]. d is 48 nm for CZTPA-1 and 56 nm for CZTPA-2. Figure 5 shows that the hole mobility of Spiro-OMeTAD (doped) is 1.01 <sup>×</sup> <sup>10</sup>−<sup>3</sup> cm<sup>2</sup> <sup>V</sup>−<sup>1</sup> <sup>s</sup><sup>−</sup>1. The hole mobility of CZTPA-1 is 4.68 <sup>×</sup> 10−<sup>5</sup> cm2 V−<sup>1</sup> s<sup>−</sup>1, and CZTPA-2 has the higher hole mobility of 8.06 <sup>×</sup> 10−<sup>5</sup> cm2 V−<sup>1</sup> s<sup>−</sup>1. However, they are all lower than the hole mobility of Spiro-OMeTAD (doped). Compared to CZTPA-1, CZTPA-2 has a higher hole mobility, which leads to good hole transport and enhances the charge transport in a planar PSC. CZTPA-2's higher hole mobility can be attributed to its high hole transport capability. These results suggest that both a fast charge transfer and high hole transport capability contribute to a high PCE.

**Figure 5.** (**a**) Space-charge-limited current (SCLC) *J* <sup>1</sup>/2–*V* characteristics of CZTPA-1-, CZTPA-2-, and Spiro-OMeTAD-based hole-only devices measured in the dark. (**b**) Stability of CZTPA-1, CZTPA-2, and Spiro-oMeTAD (dopant-free) without encapsulation.

We tested the stability of the three HTM devices without encapsulation by storing them in the dark under air conditions for at least 400 h. The PCE over time curve was plotted and is shown in Figure 5b. The PCE of the CZTPA-2 device was still over 10%. In comparison, the PCE of Spiro-OMeTAD (dopant-free) device dropped below 10%. CZTPA-1 exhibited low performance, with 68.2% of the original PCE. CZTPA-2 and Spiro-OMeTAD (dopant-free) maintained 89.7% and 81.6% of their initial PCE, respectively. This test verified that the device based on CZTPA-2 has the best stability of the three HTMs. The long alkyl chain, which has good morphology, may influence the stability of a PSC device.

Finally, we also tested the capacitance of the three HTMs. The capacitance versus frequency was plotted and is shown in Figure 6. The capacitance is mainly caused by charge or ion accumulation at the perovskite interface, which leads to interfacial recombination. We can see the capacitance of the device based on CZTPA-2 is obviously smaller than that of CZTPA-1 and Spiro-oMeTAD (dopant-free), which confirms that the CZTPA-2 device has less interfacial recombination and a higher PCE.

**Figure 6.** Capacitance versus frequency based on the devices of ITO/SnO2/perovskite/HTMs/Au.

#### **4. Conclusions**

Two new and low-cost hole transporting materials based on a carbazole core were designed and synthesized using a simple synthesis process. CZTPA-2 (dopant-free) achieved the best performance, with a PCE of 11.79%, a *J*sc of 21.80 mA/cm2, a *V*oc of 0.99 V, and a FF of 54.59%, which was slightly higher than that of Spiro-OMeTAD (dopant-free) and CZTPA-1 (dopant-free) and attributed to its higher hole transport mobility. The PL spectra, scanning electron microscopy images, and photoelectric properties indicate that CZTPA-2 with a longer alkyl chain has better optoelectrical properties. The CZTPA-2 (dopant-free) device also had the best stability, which remained at 89.7% of its original PCE after 400 h compared to CZTPA-1 and Spiro-OMeTAD (dopant-free). Besides this, its hole transport layer thickness is 35 nm. Therefore, we think that CZTPA-2 can also be used to modify an interface when compared to traditional HTMs (above 100 nm). The device based on CZTPA-2 exhibited good stability. We conclude that a longer alkyl chain may promote solubility and enhance the perovskite layer's crystallinity.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/7/935/s1, Figure S1: 1H NMR spectrum for compound **1.** Figure S2: 1H NMR spectrum for compound 2. Figure S3: 1H NMR spectrum for CZTPA-1. Figure S4: 1H NMR spectrum for CZTPA-2. Figure S5: 13C NMR spectrum for CZTPA-1. Figure S6: 13C NMR spectrum for CZTPA-2. Figure S7: AFM images (5 μmx5 μm) of CZTPA-2 (a) and CZTPA-1 (b) films.

**Author Contributions:** K.W. conceived the idea; K.W. designed the experiment; H.C. conducted device fabrication; X.G. and T.N. helped in device measurement; K.W. conducted device fabrication and data collection and wrote the manuscript; K.W. and S.W. revised the manuscript; Writing—review & editing, H.W.; Supervision, H.W.

**Funding:** This work was supported by the Key Technologies R&D Plan Projects of Guangdong Province (Nos. 2015B010127013, 2016B010124004, 2017B010112003), by the Science and Technologies Plan Projects of Guangzhou City (Nos.201604046021, 201905010001), and by the Science and Technology Development Special Fund Projects of Zhongshan City (Nos. 2017F2FC0002, 2017A1009).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Solution-Processed PEDOT:PSS**/**MoS2 Nanocomposites as E**ffi**cient Hole-Transporting Layers for Organic Solar Cells**

#### **Madeshwaran Sekkarapatti Ramasamy, Ka Yeon Ryu, Ju Won Lim, Asia Bibi, Hannah Kwon, Ji-Eun Lee, Dong Ha Kim \* and Kyungkon Kim \***

Deprtment of Chemistry and Nanoscience, Ewha Womans University, 52 Ewhayeodae-gil, Seodaemun-gu, Seoul 03760, Korea

**\*** Correspondence: dhkim@ewha.ac.kr (D.H.K.); kimkk@ewha.ac.kr (K.K.)

Received: 6 July 2019; Accepted: 11 September 2019; Published: 16 September 2019

**Abstract:** An efficient hole-transporting layer (HTL) based on functionalized two-dimensional (2D) MoS2-poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) composites has been developed for use in organic solar cells (OSCs). Few-layer, oleylamine-functionalized MoS2 (FMoS2) nanosheets were prepared via a simple and cost-effective solution-phase exfoliation method; then, they were blended into PEDOT:PSS, a conducting conjugated polymer, and the resulting hybrid film (PEDOT:PSS/FMoS2) was tested as an HTL for poly(3-hexylthiophene):[6,6]-phenyl-C61-butyric acid methyl ester (P3HT:PCBM) OSCs. The devices using this hybrid film HTL showed power conversion efficiencies up to 3.74%, which is 15.08% higher than that of the reference ones having PEDOT:PSS as HTL. Atomic force microscopy and contact angle measurements confirmed the compatibility of the PEDOT:PSS/FMoS2 surface for active layer deposition on it. The electrical impedance spectroscopy analysis revealed that their use minimized the charge-transfer resistance of the OSCs, consequently improving their performance compared with the reference cells. Thus, the proposed fabrication of such HTLs incorporating 2D nanomaterials could be further expanded as a universal protocol for various high-performance optoelectronic devices.

**Keywords:** organic solar cells; MoS2; hole-transporting layer; oleylamine

#### **1. Introduction**

Organic solar cells (OSCs) have many striking properties such as flexibility, solution processability, light weight, and simple manufacturing, especially if compared with their inorganic counterparts. To enhance their performance, numerous strategies have been proposed, including novel photoactive materials, morphology control, interfacial engineering, plasmonic nanoparticles incorporation, and alternative buffer layers and electrodes [1–6]. Their power conversion efficiency (PCE) has been recently improved up to >13% with rapid advances in new photovoltaic materials [7]. In the typical bulk heterojunction (BHJ) OSCs configuration, a photoactive blend layer consisting of acceptor/donor pairs is sandwiched between a bottom transparent anode and a top low-work-function cathode, combined with the corresponding interlayers. Such interlayers are crucial for determining the overall PCE and stability of OSCs because they reduce the potential energy barrier between photoactive layer and electrodes, enhancing the extraction of holes and electrons at the anode and cathode, respectively.

Until now, many hole-transporting layer (HTL) materials, such as conducting conjugated polymers [8–10], conjugated polyelectrolytes [11,12] metal oxides/sulfides [13–17], and graphene oxide and its hybrid films [18–21], have been explored for use in OSCs. Among them, the conjugated polymer poly(3,4 ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) has been the most widely used due to its adequate work function for creating a good ohmic contact between active layer and anode, solution processability, and high conductivity. However, its hygroscopic and acidic nature often induces chemical instability between active layers and indium tin oxide (ITO) anodes, affecting the device stability and efficiency [22,23]. Moreover, there is a clear surface energy mismatch between PEDOT:PSS (hydrophilic nature) and the active layer (hydrophobic and made of, e.g., poly(3-hexylthiophene) (P3HT)) [24,25]. To overcome such drawbacks, various PEDOT:PSS modification strategies, such as incorporating metal nanoparticles [26–28], modification by metal salts [29,30], polymer doping [31,32], and hybridization with graphene [33,34], have been developed. Interfacial engineering with long alkyl chains is an alternative but attractive method to reduce the surface energy mismatch between HTL and active layer and also to accomplish desirable molecular orientation in the active layer for enhancing the charge transport in OSCs [35].

Single and few-layer molybdenum disulfide, a two-dimensional (2D) transition metal dichalcogenide (TMDC), has recently received much interest in electronics and optoelectronics research due to its excellent optical (bandgap: 1.8 eV), electrical (device mobility: 10–130 cm2 V−<sup>1</sup> S<sup>−</sup>1), and mechanical (Young modulus: 270 GPa) properties [36,37]. Among the key preparation/exfoliation methods for TMDCs, namely, micromechanical cleavage [38], chemical vapor deposition [39], and liquid-phase exfoliation (LPE) [40], the latter is more attractive because it is scalable and cost-effective. MoS2 has been tested as HTL for OSCs [41–43] to exploit its extraordinary optical and electrical properties in photovoltaics; nevertheless, the results have revealed that neat MoS2 is not sufficient to replace PEDOT:PSS as OSC HTL, possibly because of its work function mismatch and unexpected phase transition. Hence, Xing et al. fabricated PEDOT:PSS/WS2 hybrid films and demonstrated their applicability as effective OSC HTLs [44]. However, the long-time (48 h) sonication they adopted for TMDC exfoliation in the PEDOT:PSS aqueous dispersion may affect the structure of both PEDOT:PSS and MoS2 in the final product; therefore, innovative strategies for effectively integrating these materials in OSCs are still highly demanded.

Here, we report the fabrication of oleylamine-functionalized MoS2 (FMoS2) combined with PEDOT:PSS as an effective hybrid HTL (PEDOT:PSS/FMoS2) for use in conventional P3HT:[6,6]-phenyl-C61-butyric acid methyl ester (PCBM)-based OSCs. The so-obtained OSCs exhibited better PCE and short-circuit current density (Jsc) values compared with the reference cell having simple PEDOT:PSS as HTL. FMoS2 was characterized by various spectroscopic techniques including Raman spectroscopy, ultraviolet–visible (UV-Vis) absorption and transmittance, photoluminescence (PL), and transmission electron microscopy (TEM); the active layer microstructure and the surface properties of the hybrid HTL were analyzed by grazing-incidence wide-angle X-ray scattering (GIWAXS), atomic force microscopy (AFM), and contact angle measurements. Electrochemical impedance spectroscopy (EIS) measurements were carried out using an electrochemical analyzer (IVIUMSTAT.XR, IVIUM Technologies) under illumination at 0.1 V.

#### **2. Experimental**

#### *2.1. Materials and Methods*

The following chemicals were used in our experiment: molybdenum (IV) sulfide (<2 μm, 99%) and oleylamine (Sigma-Aldrich, Gyeonggi-do, Korea), P3HT (1-Material, Gyeonggi-do, Korea), PEDOT:PSS (Heraeus Deutschland GmbH & Co., Leverkusen, Germany), isopropyl alcohol (IPA) (Dae-Jung Chemicals & Metals Co., Ltd., Gyeonggi-do, Korea), and methanol (Samchun Chemicals, Seoul, Korea).

#### *2.2. Synthesis of FMoS2 Nanosheets and PEDOT:PSS*/*FMoS2 Hybrids*

FMoS2 nanosheets were synthesized according to the liquid-phase exfoliation method reported in literature [45], with small modifications. Briefly, bulk MoS2 powder (200 mg) was bath-sonicated in oleylamine (2 mL) by using a Branson ultrasonic bath for 20 min and successively stirred at 60 ◦C for 12 h in an N2-filled glove box. Then, 1,2-dichlorobenzene (DCB) (18 mL) was added, and the dispersion was further bath-sonicated for 5 h. The resulting suspension was centrifuged at 4000 rpm, and the top 80% dark-green color supernatant, which contains excess oleylamine, DCB, and FMOS2 was collected. Then, the FMoS2 nanosheets were separated by adding excess acetone, followed by sonication for 2 min and high-speed centrifugation (10000 rpm). The separated FMoS2 nanosheets were settled at the bottom of the centrifuge tube, which was re-dispersed in a small amount of IPA by mild sonication, and different concentrations (5, 20, and 50 μL) of the resulting dispersion were added into PEDOT:PSS:methanol (1:1 V%) aqueous solutions, which were successively ultrasonicated for 30 min to obtain PEDOT:PSS/FMoS2 hybrid solutions.

#### *2.3. Fabrication of OSCs*

The OSCs having device architectures of ITO/PEDOT:PSS/P3HT:PCBM/LiF/Al and ITO/(PEDOT:PSS/FMoS2)/P3HT:PCBM/LiF/Al were fabricated as follows. ITO-coated glass substrates were cleaned via sequential ultrasonication in acetone, IPA, and distilled water, followed by oxygen plasma treatment for 10 min; then, they were spin-coated with a PEDOT:PSS (Clevios P VP Al 4083) or PEDOT:PSS/FMoS2 solution at 4000 rpm for 40 s and dried at 130 ◦C for 30 min to complete the HTL deposition. Next, an active layer consisting of a P3HT:PCBM (1:0.6 wt%) binary blend solution was spin-coated on the resulting HTL layer at 2500 rpm for 40 s inside an N2-filled glove box and annealed at 150 ◦C. Finally, LiF and Al layers were deposited by thermal evaporation. The active area of the fabricated OSCs was 0.06 cm2.

#### *2.4. Characterization*

The absorption properties of the samples were analyzed using a UV-Vis absorption spectrometer (Cary 5000, Varian, Inc.). Raman spectra were recorded on a Horiba Jobin-Yvon spectrometer. The emission properties were investigated with a luminescence spectrometer (LS55 Perkin Elmer). The TEM measurements were carried out on a JEOL JSM-2100-F system. The surface morphologies were investigated using a tapping-mode atomic force microscope (Veeco D3100). The water contact angles of the samples were measured with a KSV CAM 101 instrument. The GIWAXS analysis was conducted at the PLS-II 9A U-SAXs beamline of the Pohang Accelerator Laboratory (Korea) at the following operating conditions: incidence angle of ~0.12◦, wavelength of 1.12 Å, and sample-to-detector distance of 224 nm. The GIWAXS patterns were recorded using a 2D charge-coupled device camera (Rayonix, SX-165, USA) with an exposure time of 10–30 s. The JV properties of the solar cells were measured with a Keithley 2400 solar cell IV measurement system under AM 1.5 G illumination at 100 mW cm<sup>−</sup>2.

#### **3. Results and Discussion**

OSCs having two different device architectures, ITO/(PEDOT:PSS/FMoS2)/P3HT:PCBM/LiF/Al and ITO/PEDOT:PSS/P3HT:PCBM/LiF/Al (for comparison), were fabricated as schematized in Figure 1.

First, we synthesized FMoS2 nanosheets via the solution-phase ultrasonic exfoliation of bulk MoS2 in the presence of oleylamine and 1,2-dichlorobenzene as a solvent; then, they were incorporated in different concentrations (5, 20, and 50 μL) into PEDOT:PSS, and the resulting PEDOT:PSS/FMoS2 (denoted as PEDOT:PSS/FMoS2(5), PEDOT:PSS/FMoS2(20), and PEDOT:PSS/FMoS2(50) according to the FMoS2 loading) was used as HTL for conventional OSCs.

Raman spectroscopy is a powerful nondestructive technique for monitoring structural changes in 2D materials [46]. The Raman spectrum of bulk MoS2 showed two characteristic peaks at 374.83 and 402.05 cm−<sup>1</sup> corresponding, respectively, to the E1 2g and A1g vibrational modes (Figure 2a); the first arose from the in-plane vibration of Mo and S atoms, while the second resulted from the out-of-plane vibrations of sulfur [47,48]. As regards FMoS2, the peaks for both the E1 2g and A1g vibrational modes were blue-shifted toward higher wavenumbers (respectively, 382.66 and 405.66 cm−1), suggesting interactions between oleylamine and MoS2. Moreover, the wavenumber difference between these two vibrational modes is closely related to the layer number present in the MoS2 nanosheets [49], and in our case, this difference decreased from 27.2 cm−<sup>1</sup> for bulk MoS2 to 23 cm−<sup>1</sup> for FMoS2 nanosheets, demonstrating the successful exfoliation of MoS2 nanosheets during the oleylamine treatment.

**Figure 1.** Fabrication process for poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS)/oleylamine-functionalized MoS2 (FMoS2) hybrid hole-transporting layer (HTL) for organic solar cells.

The absorption properties of the FMoS2 nanosheets were further investigated via UV-Vis absorption spectroscopy; their spectrum (Figure 2b) clearly showed two characteristic absorption peaks of MoS2 at 618 and 677 cm−<sup>1</sup> corresponding, respectively, to the A1 and B1 direct excitonic transitions with the energy split from valence band spin–orbital coupling [50]. Furthermore, unlike bulk MoS2, FMoS2 yielded dark-greenish dispersion in 1,2-dichlorobenzene. These results clearly indicate some alteration in the surface properties of MoS2 due to the oleylamine treatment [51].

Bulk MoS2 is an indirect bandgap semiconductor that does not exhibit any photoluminescence; however, upon exfoliation, its luminescence increases with decreasing its layer thickness, so that single-layer MoS2 shows the highest photoluminescence due to its transition into a direct bandgap semiconductor [52,53]. As expected, FMoS2 exhibited significant photoluminescence (see the PL spectra in Figure S1, Electronic Supporting Information (ESI)), which clearly proves the successful layer thinning of MoS2 during the functionalization process.

**Figure 2.** (**a**) Raman spectra of bulk and oleylamine-functionalized MoS2 (FMoS2). (**b**) Ultraviolet– visible light absorption spectrum of FMoS2. (**c**,**d**) Transmission electron microscopy images of FMoS2 nanosheets.

The TEM images of the FMoS2 nanosheets are displayed in Figure 2c,d, showing a thin nanosheet morphology with sizes of several hundred nanometers. A careful observation of the nanosheet edges reveals the presence of few-layer nanosheets, confirming the effectiveness of the liquid-based exfoliation with oleylamine. AFM measurements were carried out (Figure S2, ESI) to further evaluate the layer thickness; that of FMoS2 was ~6.7 nm, suggesting the existence of few-layer nanosheets, while the reported thickness of monolayer MoS2 ranges between 0.9 and 1.2 nm [54].

To improve the performance of conventional PEDOT:PSS-based HTL for OSCs, we incorporated it with FMoS2 via a simple solution-blending method because we believed that the introduction of 2D sheet-like MoS2 functionalized with a long-chain primary alkyl amine (oleylamine) would have made the PEDOT:PSS surface more hydrophobic, facilitating the following deposition of the hydrophobic active layer. In addition, the amine group of oleylamine tends to be located near Mo atoms in MoS2 due to metal–amine interactions, while its long alkyl chain with –CH3 groups is oriented toward the active layer, and this kind of configuration should enforce the active layer with a desirable molecular orientation for efficient charge transport in OSCs; P3HT thin films deposited on insulator substrates modified with –CH3 groups formed face-on orientation because of π–H interactions [55,56].

The contact angles of ITO with PEDOT:PSS and PEDOT:PSS/FMoS2 containing 5, 20, and 50 μL of FMoS2 were 30◦, 47◦, 54◦, and 56◦, respectively (Figure S3, ESI), which indicates that the hydrophobicity of PEDOT:PSS was slightly increased by the FMoS2 addition and, hence, the hydrophobic active layer solution was more compatible on hybrid HTL than that of the hydrophilic PEDOT:PSS one.

The surface morphology of the various samples was compared via tapping-mode AFM analysis (Figure S4, ESI); the root-mean-square (rms) roughness value of PEDOT:PSS was 1 nm and decreased down to 0.69 nm for PEDOT:PSS/FMoS2(5), suggesting a smooth surface morphology in the hybrid HTL. However, PEDOT:PSS/FMoS2(50) exhibited an rms roughness value of 0.97 nm, indicating that the addition of higher FMoS2 concentrations would decrease the film smoothness.

All the synthesized PEDOT:PSS and PEDOT:PSS/FMoS2 hybrid films exhibited similar UV-Vis transmittance values (Figure S5a, ESI), showing that the FMoS2 addition did not affect any absorption property of the PEDOT:PSS matrix. As regards the P3HT:PCBM (active layer) films spin-coated on glass substrates predeposited with PEDOT:PSS or PEDOT:PSS/FMoS2 HTLs (Figure S5b), for all the samples, their absorbance ranged from 400 to 650 nm, with a maximum at 512 nm, and two shoulders around 550 and 600 nm. The existence of vibronic feature at 600 nm suggests that the P3HT film existed in a high degree of ordered crystalline lamella due to strong interchain interactions [57].

The current–voltage (JV) characteristics of the fabricated P3HT:PCBM OSCs having PEDOT:PSS/FMoS2 as HTL are shown in Figure 3a. Their performance is compared with that of reference devices having PEDOT:PSS as HTL in Table 1. The reference cells showed PCE = 3.25%, Jsc = 7.92 mA cm<sup>−</sup>2, Voc = 0.671 V, and FF = 0.61. The FMoS2 incorporation led to significant PCE and Jsc improvements; in particular, the device based on PEDOT:PSS/FMoS2(5) exhibited the highest PCE, Jsc, and FF.

The external quantum efficiency (EQE) measurements (Figure 3b) showed improved EQE for the hybrid HTL-based OSCs compared with the reference cells and confirmed also their increased Jsc, demonstrating the enhanced charge extraction at the HTL/active layer interface and the charge collection at the electrodes [58,59]. The photovoltaic parameters such as PCE, Jsc, FF and Voc as a function of FMoS2 in PEDOT:PSS HTLs are plotted in Figure 3c, d, e and f respectively.

**Figure 3.** (**a**) Current density–voltage curves, (**b**) external quantum efficiency (EQE) profiles, (**c**) power conversion efficiencies (PCE), (**d**) short-circuit current density (Jsc), (**e**) fill factor, and (**f**) open-circuit voltage (Voc) values of organic solar cells based on poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) and PEDOT:PSS/oleylamine-functionalized MoS2 (FMoS2) as hole-transporting layers. The reported average PCE values are extracted from nine identical cells for each sample.

**Table 1.** Photovoltaic performance of poly(3-hexylthiophene):[6,6]-phenyl-C61-butyric acid methyl ester-based organic solar cells having poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSSS) and PEDOT:PSS/oleylamine-functionalized MoS2 (FMoS2) as hole-transportation layers.


To understand the charge transport, we analyzed the microstructure (chain-orientation and crystallinity) of the active layer (P3HT:PCBM) on both the PEDOT:PSS and PEDOT:PSS/FMoS2 samples by GIWAXS (Figures 4 and 5). Charge transport in conjugated polymers occurs either in the π–π staking direction or the chain backbone one, which is the fastest but its vertical alignment of chains backbones along the z direction is rarely observed [60,61]. In general, P3HT crystallizes into two main configurations, namely, edge-on and face-on orientations; in the former, both chain backbone and π–π staking directions lie parallel to the substrate; in the latter, π–π staking occurs perpendicular to the substrate, which is a desirable orientation in OSCs for vertical charge transport [62]. Figure 4 shows the GIWAXS diffraction patterns of P3HT:PCBM thin films deposited on PEDOT:PSS and PEDOT:PSS/FMoS2 HTLs. In both cases, the thin films exhibited strong (100), (200), and (300) diffractions along the z axis, confirming the existence of the strong edge-on lamellae configuration of P3HT [63]. In addition, the absence of π–π staking peak (010), corresponding to the face-on orientation near the z axis, indicates that P3HT preferentially adopted the edge-on configuration in both PEDOT:PSS and PEDOT:PSS/FMoS2 HTLs. Since the use of –CH3 group-functionalized substrates tends to promote the face-on orientation of P3HT [55,56], we aimed to improve such configuration of the active layer by incorporating the described oleylamine (having –CH3 groups)-functionalized

MoS2 into PEDOT:PSS, but we did not observe any significant difference in its molecular orientation, maybe because the low FMoS2 concentrations used were not sufficient for such change. Thus, we can conclude that the PCE and Jsc enhancement in the OCSs having PEDOT:PSS/FMoS2 as HTL may be due to its surface compatibility for the active layer deposition, as observed in the AFM and contact angle measurements.

**Figure 4.** Grazing-incidence wide-angle X-ray scattering diffraction patterns of poly(3-hexylthiophene):[6,6]-phenyl-C61-butyric acid methyl ester thin films deposited on (**a**) poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) and PEDOT:PSS combined with (**b**) 5, (**c**) 20, and (**d**) 50 μL of oleylamine-functionalized MoS2.

**Figure 5.** (**a**) In-plane and (**b**) out-of-plane spectra of poly(3-hexylthiophene):[6,6]-phenyl-C61-butyric acid methyl ester thin films deposited on poly (3,4-ethylendioxythiophene): poly(styrenesulfonate) (PEDOT:PSS) and PEDOT:PSS combined with 5, 20, and 50 μL of oleylamine-functionalized MoS2 samples obtained from grazing-incidence wide-angle X-ray scattering.

Electrical impedance spectroscopy (EIS) was performed to investigate the charge transport dynamics of the OSCs fabricated with PEDOT:PSS and PEDOT:PSS/FMoS2(5) as HTL (Figure 6). This analysis allowed us to observe the current response by applying alternating current voltage as a function of frequency; the OSCs with PEDOT:PSS/FMoS2(5) demonstrated slightly lower charge transfer resistance, revealing that the holes were effectively transported from the active layer to the anode (ITO). In order to elucidate the origin of the improvement in the photovoltaic performance, especially both FF and Jsc for PEDOT:PSS/FMoS2(5), we further calculated the resistance of the devices. In general, it is well known that lower series resistance (RS) and higher shunt resistance (RSH) are required to achieve higher FF in the solar cell device [64]. Based on the J–V curves obtained from the

devices, it is clearly revealed that the device with PEDOT:PSS/FMoS2(5) as HTL showed the lowest RS while maintaining higher RSH, leading to enhancement in charge extraction. The corresponding RS value of cells employing PEDOT:PSS/FMoS2(5) as HTL was 134.8 <sup>Ω</sup>·cm2, while the reference showed 180.0 <sup>Ω</sup>·cm2. Lower RS indicates that better interfacial contact and charge collection efficiency were obtained due to the addition of the conducting FMoS2 layer. In the case of the RSH, no significant changes in the shunt resistance were observed for the devices. In the point of view of the identical RSH, barrier resistance at the interface and the leakage current level flowing across the photoactive layer is similar. Therefore, the addition of FMoS2 might contribute to extract photoexcited charges efficiently by lowering the RS.

**Figure 6.** Electrical impedance spectra of organic solar cells based on poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) and PEDOT:PSS/oleylamine-functionalized MoS2 (5 μL) (PEDOT:PSS/FMoS2(5)) as hole-transportation layer.

#### **4. Conclusions**

The application of solution-processed PEDOT:PSS/FMoS2 hybrids as effective HTLs for OSCs has been successfully demonstrated. Raman, UV-Vis, PL, TEM, and AFM analyses confirmed the successful exfoliation of bulk MoS2 into few-layer nanosheets in the presence of oleylamine via a simple and cost-effective solution-based method. The OSCs fabricated with the synthesized PEDOT:PSS/FMoS2 hybrids as HTL exhibited PCE values up to 3.74%, which is 15.08% higher than that of the reference cells having simple PEDOT:PSS as HTL. The hybrid HTL films showed better surface properties for the deposition of the hydrophobic active layer, consequently, the charge-transfer resistance was minimized for OSCs fabricated with hybrid HTL compared with reference cells, improving the OSC performance. Due to their simple preparation method, 2D FMoS2-incorporated PEDOT:PSS-based HTL provides valuable alternative HTL for OSCs.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/9/1328/s1, Figures S1 and S2: PL spectra and AFM image of oleylamine-functionalized MoS2 (FMoS2) respectively, Figures S3 and S4: Contact angles and AFM images of PEDOT:PSS and PEDOT:PSS combined with FMoS2 respectively, Figure S5: (a) UV-Vis transmittance spectra of PEDOT:PSS and PEDOT:PSS combined with FMoS2, (b) UV-Vis absorbance spectra of P3HT:PCBM thin film spin-coated on PEDOT:PSS and PEDOT:PSS FMoS2.

**Author Contributions:** M.S.R., prepared the content of this research, carried out device fabrication, analysis of results and wrote the manuscript. K.Y.R. performed GIWAXS measurements and interpreted the data. J.W.L. performed IPCE measurements and guided for manuscript revisions. A.B., H.K., J.-E.L. performed the design of device configuration, performed AFM and PL measurements. D.H.K. and K.K. guided the overall scope of this research and wrote the manuscript.

**Funding:** This work was supported by National Research Foundation of Korea Grant funded by the Korean Government (2017R1A2A1A05022387 and 2016M1A2A2940914) and by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20173010013340 and 20163030013900).

**Conflicts of Interest:** There are no conflicts to declare.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
