High performance perovskite solar cells by hybrid chemical vapor deposition

Organometal halide based perovskites are promising materials for solar cell applications and are rapidly developing with current devices reaching 19% efficiency. In this work we introduce a new method of perovskite synthesis by hybrid chemical vapor deposition (HCVD), and demonstrate efficiencies as high as 11.8%. These cells were found to be stable with time, and retained almost the same efficiency after approximately 1100 h storage in dry N2 gas. This method is particularly attractive because of its ability to scale up to industrial levels and the ability to precisely control gas flow rate, temperature, and pressure with high reproducibility. This is the first demonstration of a perovskite solar cell using chemical vapor deposition and there is likely still room for significant optimization in efficiency.


Introduction
Organometal halide perovskite based solar cells have been rapidly improving in recent years.3][4][5][6] Due to perovskite's broad spectral absorbance, low temperature processing, and low cost of materials it is a promising material for low cost solar cells.To achieve easier fabrication and better reproducibility, we used a two-step method of perovskite synthesis where a layer of metal halide (e.g.PbCl 2 and PbI 2 ) is deposited, followed by the addition of an ammonium halide salt (e.g.methyl ammonium iodide, formamidinium bromide).In this method, a solid powder is converted to the gas phase in a dedicated zone of the tube furnace and then deposited onto substrates downstream in another zone of the tube furnace.Because of the use of solid precursors, this method can be viewed as a form of hybrid chemical vapor deposition (HCVD).][9][10][11] Perovskite has been previously synthesized by CVD, 12,13 but this work is the rst to use CVD to demonstrate any solar cell efficiency.The reported stability of perovskite lms can vary signicantly.In one case, perovskite is reported to decay substantially to 20% of its original performance over the course of 6 days in a N 2 environment, 6 and in others it is reported to be stable for 500 h in air without encapsulation. 14erovskite solar cells made by HCVD were shown to be stable up to 1100 h in a N 2 environment, which warrants further study of stability under ambient conditions.

Results and discussion
Multi-zone HCVD has independent precise control of pressure, gas ow rate, methyl ammonium iodide (MAI) temperature, and substrate temperature, which helps improve reproducibility and performance.For example, by controlling the pressure (1 Pa to 1 atm), the diffusion rate (D g f T 3/2 /P, where D g is the gas diffusion constant, T is the temperature, and P is the pressure) 15 can be varied over a wide range, which is expected to provide more exibility in controlling the lm growth and possibly will lead to higher performance.The easy but rigorous process control and capacity to fabricate large uniform lms is also expected to facilitate scale-up.In this work, perovskite is fabricated by a two-step method where lead chloride is rst deposited onto appropriate substrates by thermal evaporation in high vacuum, followed by vapor phase deposition of MAI using HCVD.Lead chloride is used because it was observed to produce a more uniform lm by evaporation than PbI 2 , and is reported to increase the charge diffusion length in perovskite lms. 16,17It is expected that other metal halide precursors prepared by vacuum evaporation or solution processing are also compatible with our HCVD method.For example, a solution based method of PbI 2 deposition was used with HCVD and it demonstrated some efficiency.Fig. 1a shows an example of the HCVD deposition process, in which PbCl 2 and MAI are used.The substrates (uorine doped tin oxide (FTO) glass substrate/TiO 2 compact layer pre-deposited with PbCl 2 ) and MAI are loaded into two separate temperature control zones of the furnace.The furnace is then sealed, pumped down to a certain pressure and purged with inert gas.The diffusion of MAI takes place in two phases.First the MAI must diffuse to the substrate in the gas phase (D g f T 3/2 /P).Once the MAI has landed onto the substrate it must diffuse through the lm (i.e., solid diffusion) to react with the metal halide present at the bottom surface.In the case of solid diffusion, the diffusion constant is determined by the Arrhenius equation (D s f e ÀC/kT , where D s is the gas diffusion constant, C is a constant, k is Boltzmann's constant). 18Higher temperatures will increase both gas and solid diffusion rates, and increase the speed of perovskite conversion.
A schematic diagram of a complete solar cell and the layers within is shown in Fig. 1b.The electron transport layer (ETL) of the cell is FTO coated with a thin lm of titanium oxide made by spray pyrolysis.This is followed by the deposition of a metal halide layer and HCVD perovskite synthesis.The hole transport layer (HTL) is a lm of spiro-MeOTAD spun onto the perovskite followed by gold top electrode deposition via evaporation.A photograph of a complete, functioning device is shown in Fig. 2a.From this image, we can see that the lm is partially transparent and homogeneous.The absorbance spectrum was taken on a complete device (Fig. 2b) and the absorption edge was measured to be at 781 nm (ref.14) making it suitable for applications like photovoltaic windows and multi-junction solar cells. 19he furnace zone with MAI was heated to 185 C, and the zone with the substrate was heated to a nominal temperature of 130 C. Due to the ow of hot gas, the temperature inside the growth region is a gradient between the two set temperatures.Specically, substrates were usually loaded in the region that reaches a maximum temperature between 160 C and 170 C. Samples that were placed in a region hotter than 170 C were not observed to correctly form perovskite.The initial PbCl 2 layer was converted into a yellow lm of PbI 2 in the high temperature regions of the furnace (ESI, Fig. S1 †).This suggests that the substrate temperature was too high, and that there is a maximum temperature of perovskite formation around 170 C. Substrates that were loaded in zones with a maximum temperature of 145 C were found to be unstable.Perovskite formed at this temperature rapidly turned transparent when exposed to air.Independent control of temperature zones and control of pressure provide signicant advantages over existing vapor deposition methods. 9The additional degree of control may open the door for fabrication of a wider range of perovskite materials.
Thermal evaporation was used to deposit the metal halide layer because it offers good uniformity and thickness control.The resulting perovskite lms show a relatively smooth morphology with complete coverage (RMS roughness ¼ 28 AE 10 nm, on   This journal is © The Royal Society of Chemistry 2014 silicon).Fig. 3a and b show AFM images of perovskite lms in true proportions (i.e., the lengths along the lateral and vertical directions have the same scale).This surface roughness is similar to the roughness of the FTO/TiO 2 substrates used for solar cells ($27 nm).A perovskite lm on the FTO/TiO 2 substrate was measured to have a roughness of $50 nm.In contrast, solar cells that are fabricated by a solution process oen leads to incomplete coverage, and consequently a much higher roughness. 20aving lower surface roughness allows for thinner lms on top of the perovskite, such as the hole transport layer and metal contact layer, which likely will lead to better optimized devices.Also in this work, the two top lms correspond to the most expensive materials used, and having thinner lms may help reduce manufacturing cost.The grains of perovskite lms prepared by HCVD are typically between 0.2 and 2 mm in size, with an average size of approximately 0.6 AE 0.4 mm.
It appears that PbCl 2 is mostly converted to perovskite during the CVD process.Film thicknesses are measured before and aer the CVD deposition process and it is found to increase by a factor of 3.0 AE 0.2.It is possible to make an estimate of the thickness increase of fully converted perovskite assuming densities of lead chloride and perovskite to be 5.85 g cm À3 and 4.17 g cm À3 , respectively. 21With these assumed densities, the expected volume would increase by a factor of 3.1, which is in good agreement with measured values and indicates that the perovskite is fully converted.Fig. 3c shows the X-ray diffraction (XRD) spectrum of the perovskite lm and the observed characteristic diffraction peaks at 14 and 28.4 correspond to the (110) and ( 220) planes.The spectrum also reveals the presence of other orientations: (112) at 20.0 , (211) at 23.5 , (202) at 24.5 , (312) at 32.0 , (224) at 40.6, and (314) at 43.2 . 22There are also peaks associated with the substrate, which are predominately due to SnO and are found at 26.5 , 33.7 , 37.9 , and 51.5 .The relatively high intensity of peaks at orientations other than (110) and (220) suggests that perovskite lms grown by HCVD are not highly oriented with respect to the substrate, which appears consistent with the AFM image in Fig. 3a.For some perovskite samples, X-ray diffraction spectra reveal a peak at 12.6 associated with the presence of PbI 2 .If present, these peaks are generally smaller compared to the main (110) perovskite peak, indicating that the resulting lm mainly consists of perovskite.
One of the most important parameters observed is the thickness of the pre-deposited PbCl 2 layer and subsequent perovskite.The average performance of solar cells with different PbCl 2 layer thicknesses is shown in Fig. 4a.There appears to be a peak in efficiency at PbCl 2 layer thickness approximately 100 nm, or approximately 300 nm of perovskite.It has recently been proposed that lower open-circuit voltage (V oc ) and ll factor (FF) observed for solar cell devices with a perovskite lm thickness in excess of 300 nm can be due to the low hole mobilities of the electron blocking layer. 23Alternatively, poor performance with thick lms may be due to perovskite grains smaller than the thickness of the lm. 24This trend is most clearly observed when considering the ll factor as a function of thickness (ESI, Fig. S2 †).Furthermore, it was found that post annealing substrates in air rather than in N 2 signicantly improves V oc , short-circuit current density (J sc ), and FF.The surface morphology aer air annealing was observed to be different than surfaces annealed in a N 2 environment.It is very likely that the effect we observe in this work is similar to the solvent annealing effects as reported recently by Xiao et al. 24 A set of air annealed solar cells, including the best performing cell demonstrated, is shown in Fig. 4b.This batch of solar cells had an average measured PbCl 2 thickness of 103 AE 3 nm and an average measured perovskite thickness of 296 AE 4 nm.The devices from this batch annealed in N 2 had an average efficiency of 6.1% and a maximum of 8.8%, while the devices annealed in air had an average efficiency of 10.8%, and a maximum of 11.8%.
To monitor the stability of perovskite solar cells made by HCVD, some samples were measured aer 450-1100 h of storage in a N 2 glove box.On average, the cells maintained their efficiency (ESI, Fig. S3 †).For example, the highest performing sample was measured about 1100 h aer the initial measurement and was found to have similar efficiency to the original measurement.Aer 1100 h, the cell's V oc decreased slightly (À1%) but J sc (+2%) and FF (+5%) increased.The solar cell performance measurements were performed in ambient air with a temperature of $25 C and a relative humidity of $50%.Between the measurements, the samples were stored in the dark, inside a N 2 glove box with H 2 O and O 2 levels less than 0.1 ppm.The cell has spent a total of approximately 8 h outside of the N 2 glove box for measurements.The stability observed in our solar cells is likely related to the fact that perovskite in our HCVD method reaches temperatures above 160 C. At a temperature above 160 C, excess MAI and incorporated water are expected to evaporate completely.The HCVD process forms perovskite with a constant ow of MAI and may be prone to absorb excess MAI.A lm of MAI is hygroscopic and will likely cause instability issues.Thermogravimetric analysis shows that a MAI powder sample starts to lose mass at 145 C (ESI, Fig. S4 †).This is consistent with the observation that the minimum temperature of stable perovskite formation is higher than 145 C.
Similar to perovskite solar cells that are fabricated with other methods, devices fabricated by HCVD suffer from hysteresis.The top performing cell showed substantially lower (7.0%)efficiency when measured in the reverse bias (ESI, Fig. S5 †).Further optimization of parameters such as layer thickness, MAI deposition temperature and time, and annealing conditions could yield performance similar to those of the highest performing solar cells, but is beyond the scope of this study.

Experimental
Details involving synthesis, fabrication and measurement are included in the ESI.†

Conclusions
To summarize, we developed a new method of perovskite synthesis by hybrid chemical vapor deposition (HCVD).The solar cells fabricated by this method not only achieved efficiencies as high as 11.8%, but also exhibited decent stability showing almost the same efficiency aer approximately 1100 h storage in dry N 2 gas.HCVD is of particular advantage due to its ability to scale up to industrial levels and the ability to precisely control gas ow rate, temperature, and pressure with high reproducibility.It is expected that process optimization of HCVD can further improve performance.

Fig. 1
Fig. 1 Hybrid chemical vapor deposition based perovskite synthesis (a) diagram of the HCVD furnace and MAI deposition onto metal halide seeded substrates.(b) Layered structure of a complete solar cell fabricated by a HCVD process.The complete solar cell is a glass substrate followed by a FTO layer, an electron transport layer, a perovskite layer, a hole transport layer, and a top metal contact.

Fig. 2
Fig. 2 Transmission through the complete perovskite solar cell.(a) Back view of the complete HCVD solar cell with front illumination from a fluorescent light bulb; (b) UV-vis spectrum of a complete cell showing an absorption edge at 781 nm.

Fig. 3
Fig. 3 Analysis of perovskite films.(a and b) Atomic force microscopy 3D images of the perovskite thin film on the silicon substrate in true proportion.The silicon substrate was prepared under the same conditions and at the same time as those used for the highest performing device.(a) 40 Â 40 mm 2 scan of a perovskite film with 18 nm RMS roughness.(b) 2 Â 2 mm 2 image showing the laminate structure of the perovskite grains.(c) X-ray diffraction spectrum of the working perovskite solar cell.The major perovskite peaks (110) at 14.1 , and (220) at 28.4 are clearly shown.

Fig. 4
Fig. 4 Perovskite solar cell performance.(a) Efficiency as a function of film thickness.Scatter plot of average efficiencies of solar cells at different seed layer thicknesses.Black dots represent cells that were annealed in N 2 and the red diamonds represent cells that were annealed in air.(b) I-V measurements on the top performing batch of solar cells.The cells have $300 nm of perovskite grown by HCVD.The curves shown were measured $500 h after device fabrication.