Longbin
Qiu‡
,
Sisi
He‡
,
Zonghao
Liu
,
Luis K.
Ono
,
Dae-Yong
Son
,
Yuqiang
Liu
,
Guoqing
Tong
and
Yabing
Qi
*
Energy Materials and Surface Sciences Unit (EMSSU), Okinawa Institute of Science and Technology Graduate University (OIST), 1919-1 Tancha, Kunigami-gun, Onna-son, Okinawa 904-0495, Japan. E-mail: Yabing.Qi@OIST.jp
First published on 20th October 2020
Hybrid chemical vapor deposition (HCVD) has been employed in the fabrication of perovskite solar cells (PSCs) and modules (PSMs), and it shows strong promise for upscalable fabrication. The conventional HCVD process needs a relatively long processing time (e.g., several hours) and the fabricated PSCs often exhibit salient hysteresis, which impedes utilization of this technology for mass production. Herein, we demonstrate a rapid HCVD (RHCVD) fabrication process for PSCs using a rapid thermal process, which not only significantly reduces the deposition time to less than 10 min, but also effectively suppresses hysteresis. This markedly reduced deposition time is comparable to that of solution-coating processes. Furthermore, the shorter processing time inside the furnace reduces the exposure time of the glass/ITO/SnO2 substrates under vacuum, which helps maintain the high quality of the SnO2 electron-transport layer and results in a lower density of gap states. Finally, PSMs with a designated area of 22.4 cm2 fabricated via RHCVD achieved an efficiency of 12.3%, and maintained 90% of the initial value after operation under continuous light illumination for over 800 h.
HCVD is a promising method as compared to the solution-based ones because of its advantages, such as uniform deposition across large areas, low cost, solvent-free, and readiness for integration with other thin film solar technologies (e.g., thin film silicon solar cells) to form tandem solar cells.13 Currently, the decay rate between small area solar cells and large area solar modules upon upscaling is 1.3% per decade of the area increase,15 which is approaching other mature photovoltaic technologies. HCVD is a two-step deposition process. In the first step, inorganic precursor materials (e.g., PbI2, PbCl2, CsI, etc.) are deposited by thermal evaporation,14 spray coating13 or spin coating.16 In the second step, organic precursor materials (e.g., FAI, MAI, MABr, etc., where FA is formamidinium and MA is methylammonium) are sublimed in the first heating zone of a CVD tube furnace, and subsequently driven by a gas flow (e.g., dry N2, Ar, or dry air) towards the second heating zone, where the organic precursor vapor reacts with the inorganic precursor that is pre-deposited on the substrate, leading to perovskite film growth.14,15 Based on the pressure and heating zone temperatures, a variety of HCVD techniques can be developed to fabricate perovskite films, including atmospheric pressure HCVD,16–18 low-pressure HCVD,14 single-zone HCVD19,20 and double-zone HCVD.14 However, all of the HCVD processes usually need a relatively long processing time (2–3 hours), which severely limits their mass-production capabilities for solar cell fabrication. Reducing the deposition time is still a challenge that needs to be fully addressed for the further development of the module-scale HCVD technique. Furthermore, it has been found that the long deposition time has a detrimental effect on the ETL, such as SnO2 or TiO2, which deteriorates the solar module performance.15 Also, hysteresis was observed due to the un-optimized interface between this ETL layer and the perovskite layer. Incorporation of an additional buffer layer such as C60 improves the HCVD processed solar cell performance by reducing the negative impact of vacuum annealing on the ETL.15 However, this additional layer increases the cost and complexity of the deposition process.
Herein, we report a rapid HCVD (RHCVD) process to fabricate PSCs and PSMs (Fig. 1). The RHCVD process greatly reduces the deposition time from several hours to within 10 min, which is comparable with typical solution coating processes. As compared with the regular HCVD process, the RHCVD processed PSCs exhibits only slight hysteresis even without any additional interfacial buffer layer. Using RHCVD, we obtained an efficiency of 15.5% for the small-area PSC (active area = 0.1 cm2). Furthermore, we show that the RHCVD process is readily upscalable for fabricating PSMs, the area of which is only limited by the size of the CVD tube furnace. As a demonstration, we fabricated PSMs on 5 cm by 5 cm substrates with a designated area of 22.4 cm2 and a geometric fill factor of approximately 90%. The performance of the PSMs was up to 12.3% with almost no hysteresis. The 224 times increase in the device area with an absolute PCE reduction of 3.2% (i.e., a decay rate of 1.36% per decade of the area increase), further confirms the upscalability of this RHCVD method. The operational lifetime of the encapsulated PSMs was tested under continuous light illumination with a steady voltage output (the initial maximum power point (MPP) voltage) and the PSM maintained 90% of its initial performance after operation for more than 800 h.
The as-deposited Cs0.1FA0.9PbI3 film shows a smooth surface and grain sizes in the range of 300–500 nm (Fig. 2c). SEM micrographs confirm the good uniformity of the surface morphology of the RHCVD deposited perovskite films across a relatively large area (20 μm × 13 μm) (Fig. S3†). The XRD result confirms the full conversion of PbI2 to perovskite, as there is a strong peak at 14.0° and no observable peak at 12.7° (Fig. 2d). The smooth baseline and the absence of the peak at around 11.6° indicate that the formed perovskite film is mainly pure α-phase. High resolution Cs 3d core-level X-ray photoelectron spectroscopy (XPS) data verify the incorporation of Cs cations into the perovskite (Fig. S4†), which agrees well with our previous study on the Cs0.1FA0.9PbI2.9Br0.1 perovskite film prepared by the regular HCVD method.15 The composition ratio between Cs and Pb is determined by XPS to be approximately 1:
10, which agrees well with the desired ratio. Our previous studies suggest that iodide (I−) can easily desorb from iodide-based perovskite films leading to generation of iodine (I2) gas under vacuum conditions and subsequently being pumped out.25,26 During the HCVD or RHCVD processes, the large amount of the organic precursor FAI in the crucible (0.1 g) provides sufficient supply of I−, minimizing the deterioration of the perovskite films during the RHCVD process. Furthermore, the fact that perovskite films prepared by RHCVD show high quality and uniformity over a large area (even without additional annealing as compared to most solution processed perovskite films) indicates that RHCVD is an efficient way to form perovskite films, possibly thanks to the dual function of IR heating in promoting perovskite formation, as well as uniformly heating the converted perovskite films to enhance the film crystallinity. The electronic structures of the deposited perovskite Cs0.1FA0.9PbI3 film has been further investigated by UPS (Fig. S5†). The deposited perovskite showed an n-type nature with a work function of 3.92 eV and valence band maximum of 1.50 eV below the Fermi level. The absorbance of the perovskite film has been characterized by UV-vis absorption measurements and the optical band gap of 1.56 eV is determined on the basis of the Tauc-plot (Fig. S6†).
Our solar cells are based on the perovskite films deposited by RHCVD and the SnO2 ETL. In our previous work, the SnO2 ETL was fabricated by sputtering at room temperature.27 The crystallinity of the sputtered SnO2 films was studied as a function of the post-annealing temperature in the range of 100–500 °C. The room temperature deposited SnO2 film showed an amorphous structure, and the crystallinity was only observed after post-annealing at temperatures over 300 °C.27 However, the post-annealing process also deteriorated the quality of the SnO2 films.27 Similar to the reports of Park and coworkers, annealing of the amorphous SnO2 ETL at 500 °C crystallizes SnO2. However, the enhanced crystallinity reduces the device performance and leads to large hysteresis, due to the increased interface capacitance.28 Furthermore, the vacuum annealing process during the regular HCVD process not only damages the deposited SnO2 films, but also increases the density of gap states located between the valence band and Fermi level, which might further decrease the performance of the as-prepared solar cell devices.15 On the other hand, the solution processed films using SnO2 nanocrystal solutions exhibit a higher quality compared with sputtered amorphous SnO2 films after a similar vacuum annealing process in a CVD tube furnace.15 In this work, we performed a detailed study to investigate the influence of the sputtered amorphous SnO2 (i.e., without post-annealing to avoid damage) and solution-processed SnO2 films on the PSC device performance.29,30 First, a nanocrystal SnO2 layer was prepared by spin coating, or alternatively, an amorphous SnO2 layer was prepared by sputter coating on the ITO/glass substrates. The surface morphology of the SnO2 film was first characterized by atomic force microscopy (AFM). As can be seen in Fig. S7,† the solution processed nanocrystal SnO2 film shows a smoother morphology compared with the sputtered SnO2 film. The surface roughness (RMS) for the sputtered SnO2 and solution coated nanocrystal SnO2 films is 2.79 and 0.88 nm, respectively (Fig. S7†). The RMS of the sputtered amorphous SnO2 layer is similar to that of the ITO substrate (2.45 nm) due to a conformal coating, while the SnO2 nanocrystals help reduce the surface roughness of the substrate.27
The PCE for the PSC device based on the spin coated SnO2 nanocrystal ETL is 15.0% in the reverse scan and 14.1% in the forward scan (Fig. S8†). Comparing the PCEs from the reverse scans, the PSC devices have negligible differences based on solution coated SnO2 and sputter deposited SnO2. However, the PSC devices with a sputter-deposited SnO2 ETL showed 7 times higher hysteresis index (0.06 versus 0.41; hysteresis index = (PCEreverse − PCEforward)/PCEreverse) and 57.4% lower forward scan efficiency. The severe hysteresis issue in the PSC devices can be attributed to the sputtered SnO2 ETL that has a combination of an amorphous structure, high surface roughness and a large defect density.27–29 Hence, we selected solution coated SnO2 nanocrystal films as the ETL to fabricate subsequent PSC devices.
We have further studied the advantages of RHCVD compared with the regular HCVD process. The most significant advantage is the much shorter deposition time thanks to the rapid thermal process. RHCVD can lead to increased throughput and hence lower cost. It also helps reduce the vacuum annealing effect on the SnO2 ETL.15Fig. 3a compares the representative J–V curves for the PSC devices fabricated based on RHCVD and regular HCVD. Similar to our previous study, the longer vacuum annealing process in the regular HCVD process deteriorates the quality of the SnO2 ETL, and the PSC device performance is much lower. The photovoltaic parameters are summarized in Table 1. For RHCVD, the champion PSC device shows a PCE of 15.5%. In contrast, for regular HCVD, the PCE is only approximately 7.6%. Although the addition of a thin layer of C60 and formation of a SnO2/C60 double-layer ETL could help improve the PSC performance as reported in our previous work,15 this additional step of vacuum deposition of C60 will increase the complexity of the fabrication and increase the LCOE. For RHCVD, because the high-quality of the SnO2 ETL is maintained during the much shorter vacuum annealing process, the PSC device performance is much higher. PSC devices fabricated via a regular HCVD process show a lower current density, and the incident-photon-to-current-efficiency (IPCE) spectra (Fig. 3b) confirm the results obtained from the J–V curves (Table 1). The current density of the PSC fabricated by the RHCVD process was determined to be 22.3 mA cm−2 from the J–V curve and 21.2 mA cm−2 from the EQE curve (4.93% difference), and the current density of the PSC fabricated by regular HCVD was determined to be 18.8 mA cm−2 from the J–V curve and 18.0 mA cm−2 from the EQE curve (4.26% difference). Although these differences are not negligible, they are within a relatively low level (i.e., below 5% of the current density) (see Table S1†), which suggests that our measurements were reasonably accurate.31 The optical properties of the SnO2 films after the RHCVD process and the regular HCVD process have been characterized (Fig. S9†). As shown in Fig. S9,† the transmittances of both SnO2 films on a quartz substrate are similar to each other, and the optical band gap of both films determined based on the Tauc-plot is 3.85 eV. This large band gap is expected to show a deep valence band and suitable hole blocking properties.
Scan direction | V OC (V) | J SC (mA cm−2) | FF (%) | PCE (%) | ||
---|---|---|---|---|---|---|
RHCVD | Champion | Reverse | 0.99 | 22.3 | 70.2 | 15.5 |
Forward | 0.97 | 22.6 | 64.5 | 14.1 | ||
Average | Reverse | 0.98 ± 0.01 | 22.2 + 0.4 | 65.8 ± 3.5 | 14.3 ± 0.8 | |
Forward | 0.96 ± 0.02 | 22.3 + 0.4 | 63.9 ± 1.9 | 13.9 ± 0.6 | ||
Regular HCVD | Champion | Reverse | 0.96 | 19.0 | 42.1 | 7.6 |
Forward | 0.93 | 18.8 | 40.0 | 7.0 | ||
Average | Reverse | 0.91 ± 0.05 | 17.5 + 1.0 | 43.9 ± 1.4 | 7.0 ± 0.3 | |
Forward | 0.93 ± 0.04 | 17.5 + 1.2 | 42.4 ± 1.7 | 6.8 ± 0.3 |
We further studied the defect density of the SnO2 films that experienced different vacuum annealing processes to understand their influence on the PSC device performance. Ultraviolet photoemission spectroscopy (UPS) and XPS measurements were carried out to characterize the surface properties of the SnO2 films after the vacuum annealing process. To understand the surface properties, SnO2 films coated on ITO substrates were placed into the CVD tube furnace to simulate the vacuum annealing process during RHCVD and regular HCVD. No organic precursors were loaded into the CVD furnace for the study on the effects of temperature and vacuum conditions. As shown in Fig. 3c, the SnO2 film that experienced the RHCVD vacuum annealing process shows a lower work function (WF) of 3.66 eV, while for the SnO2 film that experienced the regular HCVD vacuum annealing process the WF is much higher (3.90 eV). As for the ETL layer, a lower WF is expected to better facilitate electron extraction.32,33 This explains why the PSC devices based on the SnO2 ETL that experienced long vacuum annealing showed poorer performance. The gap state density between the valence band minimum and the Fermi level also increases for the longer vacuum annealing process (Fig. 3c).15 The enlarged UPS valence features (Fig. 3d) and high resolution XPS valence features (Fig. S10†) further confirm the increased gap states between the valence band and the Fermi level. These gap states lower the hole blocking barrier and increase recombination, hence lowering the VOC and the PCE of the PSCs.27 We used the space-charge limited-current (SCLC) technique to characterize the electronic properties of the devices based on the symmetric sandwich structure (see the inset in Fig. 3e). The mobility was calculated using the Mott–Gurney law.34 The mobility for the freshly coated SnO2 film was 5.0 × 10−4 cm2 V−1 s−1. After the shorter vacuum annealing process (10 min) to simulate the RHCVD process, the mobility of the SnO2 film became 5.1 × 10−4 cm2 V−1 s−1, which was almost the same as before. However, the mobility for the SnO2 film that experienced a longer vacuum annealing process of 230 min showed a much lower mobility of 1.7 × 10−4 cm2 V−1 s−1, which is likely the result of the increased gap states acting as scattering/trapping centers. These observations strongly suggest that RHCVD not only significantly reduces the deposition time, but also maintains the high quality of the SnO2 nanocrystal ETL, which is the main reason for the improvement of PCE up to 15.5%.
To help understand the main reason responsible for the current density reduction, we studied the energy level alignment based on perovskite and SnO2 for the case of RHCVD versus regular HCVD (Fig. S11†). Similar to the previous reports on planar PSCs based on the SnO2 and TiO2 ETL, the energy level mismatch at the perovskite/TiO2 interface prevents efficient charge extraction and therefore leads to reduced current density.35–39 Here in this work, the conduction band of the SnO2 film that experienced the regular HCVD process shows an upward shift, and the work function also increases by 0.24 eV. There is an energy barrier for electron extraction from the perovskite film to SnO2. For the SnO2 film that experienced the RHCVD process, the energy level alignment between the perovskite and SnO2 is better (Fig. S11†). The low work function helps facilitate electron extraction. Furthermore, with a longer vacuum annealing during the regular HCVD process, more gap states are formed above the valence band of the SnO2 film, which lowers the hole blocking properties of this interface and increases charge carrier recombination (Fig. S11†).27
To further verify that vacuum annealing deteriorates charge carrier extraction and increases charge carrier recombination, we fabricated PSCs based on solution-coated perovskite films deposited on a SnO2 ETL with and without vacuum annealing, and the results are shown in Fig. S12.† As we can see, the PSC device based on the SnO2 ETL film after a long time of vacuum annealing shows lower performance with reduced open circuit voltage, short circuit current density and fill factor, which agrees with our previous results.15 On the other hand, the PSC device based on the SnO2 ETL film after a short time of vacuum annealing as in the RHCVD process does not show much difference between the control sample case and the RHCVD-treated SnO2 case. The faster carrier extraction from the perovskite film to the SnO2 ETL after the RHCVD process is further verified by the time resolved photoluminescence (TRPL) spectra (Fig. S13 and Table S2†). The resulting curves are fitted by a double exponential model with fast and slow decay time constants (τ1 and τ2).36 The fast decay time constant τ1 corresponds to the quenching of charge carriers by electron extraction from CH3NH3PbI3 to SnO2.27 With the reduced vacuum annealing process time, the fast time constant τ1 decreased from 17.8 ns for CH3NH3PbI3 on SnO2-HCVD to 10.9 ns for CH3NH3PbI3 on SnO2-RHCVD. This observation also implies that a longer vacuum annealing process may increase the gap states and deteriorate the quality of the SnO2 layer.
In the case of regular HCVD, the combination of mismatched energy levels between the SnO2 ETL and the perovskite film, increased gap states in the SnO2 ETL, and reduced mobility of the SnO2 film contributed to the decreased solar cell performance. The major differences between the regular HCVD and RHCVD are summarized in Table 2.
J. Mater. Chem. A, 2019, 7, 6920–6929 | This work | |
---|---|---|
Deposition time | A total of 230 min (including ramping, deposition, and cooling) | Less than 10 min with a rapid ramping and deposition process |
The heating component | In the regular HCVD process, the tube furnace is heated by resistive heating (i.e., using filaments); the ramping of the temperature is slow and temperature fluctuation is large | For the RHCVD process, we use a rapid thermal annealing tube furnace, which uses an infrared (IR) heating component with stable control of temperature. The IR heating component is mechanically movable along the furnace tube, which enables fast heating and cooling. This is the first report to apply this rapid thermal annealing process for the fabrication of PSCs and PSMs |
ETL structure | The SnO2 ETL was deposited by a sputtering process at room temperature and it is amorphous in structure. Due to the prolonged vacuum annealing process, the SnO2 quality has been reduced with increased gap states. A thin layer of C60 has been incorporated to maintain the electron extraction and transport properties. However, this complicated the deposition process with one more vacuum deposition step | In this work, the SnO2 ETL was prepared using a solution of SnO2 nanocrystals |
Hysteresis and device performance | Due to the prolonged vacuum annealing process and induced gap states in the SnO2 ETL, the solar cells and modules showed hysteresis. And a module with a 22.4 cm2 designated area showed a PCE of 10%, and that with a 91.8 cm2 designated area showed a PCE of 9.34% | In the case of RHCVD, the solar cells and modules showed substantially reduced hysteresis with high-quality SnO2. PSMs with a designated area of 22.4 cm2 showed a PCE of 12.3% |
Operational stability | The T80 lifetime is around 500 h | The T90 lifetime is over 800 h. The much better operation stability is most likely due to the higher quality of the SnO2 ETL, i.e., the new RHCVD method leads to a lower density of gap states in the SnO2 ETL |
The balance between the temperature ramping time, deposition time and perovskite layer thickness has been studied to further optimize the RHCVD process. Higher temperatures can shorten the deposition time. However, the deposition rate of FAI and the reaction between deposited FAI and PbI2 need to be balanced. If FAI is deposited faster than its reaction with PbI2, there will be excess FAI on top of the perovskite film. When FAI is deposited substantially slower than its reaction with PbI2, the feeding of FAI will be the time limiting factor and a longer time is needed for the complete conversion of PbI2 to perovskite. The optimized condition for RHCVD corresponds to a balanced condition required for vaporization of FAI, diffusion into PbI2 and reaction with PbI2.15,40 Upon temperature increase, both the FAI deposition rate and the reaction rate with PbI2 increase. As the temperature increased to 170 °C (RHCVD-2 in Fig. 4a), the device performance started to decrease (Fig. 4b), which indicates a faster increase of the FAI deposition rate compared with its reaction rate with PbI2, and a pale surface on the top of the perovskite film was observed after being taken out from the tube. On the other hand, with a slower ramping rate (RHCVD-3 in Fig. 4a), the total vacuum annealing time will increase, and the WF of the SnO2 ETL prepared from the SnO2 nanocrystal rises slightly (Fig. 4c). Based on these observation, a shorter reaction time and higher ramping rate benefit the formation of high-quality perovskite films and SnO2 film. The thickness of the PbI2 film also influences the deposition time because a thicker PbI2 film would need more FAI for the conversion to perovskite. To ensure sufficient light absorbance, a film thickness of 370 nm for the perovskite was chosen in this work (Fig. S14†).5
HCVD is promising for upscalable fabrication of PSCs. 10 cm × 10 cm PSMs have been demonstrated for the regular HCVD process and the upscaling of the size from 0.1 cm2 to 91.8 cm2 shows a small absolute PCE decay rate of 1.3% per decade of the area increase.15 Similarly, in this work we also tested upscalable capability of RHCVD. Here the CVD tube furnace used for the RHCVD process has an inner diameter of 96 mm, which is the only limitation for the size of the substrates to be used. We fabricated a PSM on a 5 cm × 5 cm substrate using the RHCVD process. The PSM has a designated area of 22.4 cm2 with 7 cells connected in series. For each subcell the ITO stripe width is 6.6 mm, and there is a 0.1 mm P1 patterning line between each sub stripe. The length of each sub-cell is 4.8 cm, with 1 mm space at the edges that are perpendicular to the patterning lines. The edges that are parallel to the patterning line have a space of 1.5 mm and are used for wiring of the electrodes (Fig. S15†). The typical geometric fill factor for the module is approximately 90%.15,27 The champion PSM has a PCE of 12.3% and almost no hysteresis (Fig. 5a and b). Furthermore, the continuous operation of the PSM under continuous light illumination in a dry N2 box was measured by recording the PSM power output under a steady voltage output, which corresponded to the initial maximum power point voltage. As shown in Fig. 5c, the power output of the PSM maintained 90% of its initial value after continuous working under light illumination after 800 h. The RHCVD process shows a significantly shorter processing time and longer operation stability and is a promising method for upscalable fabrication of PSMs with larger areas, smaller PCE decay and longer operational lifetime (Table S3†).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ta09007g |
‡ L. Qiu and S. He contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2020 |