Effect of Zn(TFSI)2 on the performance-aging time of perovskite solar cells

Hole transport layers (HTLs) are one of the essential layers of perovskite solar cells (PSCs). Generally, 2,2ʹ,7,7ʹ-Tetrakis [N,N-di(4-methoxyphenyl)amino]-9,9ʹ-spirobifluorene (spiro-MeOTAD) doped by lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) is used as the HTL in PSCs. PSCs employing spiro-MeOTAD require an additional aging process to reach an optimized point of photovoltaic performance due to doping and energy alignment. However, LiTFSI is responsible for low thermal stability and has a hygroscopic nature; therefore, Zinc(II) bis(trifluoromethanesulfonyl)imide (Zn(TFSI)2) has been reported as an outstanding candidate to replace LiTFSI. Nevertheless, utilization of Zn(TFSI)2 as a dopant for PSCs has rarely been reported, which is likely due to the difficulty in achieving high device performances comparable to that with LiTFSI. Herein, we investigate the effect of Zn(TFSI)2 on the doping kinetics of spiro-MeOTAD and correlate it with the time-dependent photovoltaic performance of PSCs employing Zn(TFSI)2. Devices with Zn(TFSI)2 require a considerably longer aging time (∼270 h) to reach the optimized performance, while LiTFSI takes only ∼20 h due to the different doping kinetics of spiro-MeOTAD depending on the dopant. Remarkably, engineering at the interface of the perovskite/HTL can effectively shorten the device aging time by manipulating the recombination rate, leading to a comparable aging time to LiTFSI.


Introduction
In 2012, halide perovskite was remarkably issued due to its outstanding optoelectronic properties, particularly for solid-state photovoltaics, exhibiting about 10% power conversion efficiency (PCE) [1,2].The PCE of perovskite solar cells (PSCs) has rapidly increased over the last decade to 26.1%, competing with the most conventional crystalline silicon solar cells showing 26.1% [3].
Briefly, PSCs consist of an electron transport layer (ETL), a perovskite layer, and a hole transport layer (HTL), which is essential for hole extraction and transfer.The HTL performs various roles such as hole extraction, hole transfer to an electrode, as a barrier for electron transfer, and as an internal protection layer for the moisture-sensitive perovskite layer.Furthermore, an open-circuit voltage (V OC ) can be controlled by tuning the highest occupied molecular orbital (HOMO) level of the HTL [4,5].
Meanwhile, zinc (II) bis(trifluoromethanesulfonyl)imide (Zn(TFSI) 2 ) has been found to be considerably suitable for the spiro-MeOTAD doping, resulting in improved long-term device stability as well as increased PCE [16].This was noticeable because all the other dopant candidates adopted to enhance device stability, except Zn(TFSI) 2 , unavoidably bring a significant PCE reduction, which in turn loses the best advantage of PSCs, the high PCE, over other photovoltaics, making LiTFSI unrivaled as a dopant for the spiro-MeOTAD in PSCs.
Nevertheless, utilization of Zn(TFSI) 2 as a dopant for spiro-MeOTAD has rarely been reported for PSCs since the first paper [16], which is most likely because achieving a PCE as high as the reported value is challenging, raising the reproducibility issue.In this study, we investigate the effect of Zn(TFIS) 2 on the doping kinetics of spiro-MeOTAD as well as on the time-dependent PCE of PSCs for the first time, pointing out that the required aging time to reach the optimized performance of PSCs is quite different between LiTFSI and Zn(TFSI) 2 .Therefore, we reveal that the initial poor performance by Zn(TFSI) 2 is an improper evaluation standard for optimized performance.Furthermore, the deposition of an interlayer between the perovskite and spiro-MeOTAD is attempted to overcome the much longer aging time for Zn(TFSI) 2 -based PSCs by tuning the interface energy alignment.

Experimental
ETL substrate.For the deposition of a compact (cp-) TiO 2 layer, an approximately 10 nm thick Ti layer was deposited on patterned flurine-doped tin oxide (FTO) glass (Pilkington, TEC-8) using an radio frequencies magnetron sputtering system (A-Tech system, Korea), followed by oxidation at 500 • C for 30 min in air.A mesoporous (mp-) TiO 2 layer with ∼140 nm thickness was prepared by spin-coating at 5000 rpm for 30 s using TiO 2 paste (SC-HT040, Sharechem), followed by annealing at 500 • C for 30 min.The prepared FTO/cp-TiO 2 /mp-TiO 2 substrates were placed in a TiCl 4 aqueous solution (20 mM) at 90 • C for 20 min and washed with deionized water before annealing at 500 • C for 30 min.
Perovskite layer.FAPbI 3 (δ-phase, Sharechem) and MAPbBr 3 (Sharechem) were predissolved in a mixture of dimethylformamide (DMF; 99.8%, Sigma-Aldich) and dimethyl sulfoxide (DMSO; >99.9%,Sigma-Aldrich) (DMF:DMSO = 4:1 v/v) to prepare 1.5 M stock solutions.Additionally, 1.5 M CsI (99.9%, metal basis, Alfa-Aesar) was also predissolved in DMSO as a stock solution.For the preparation of a precursor solution of [CsI] 0.05 [FA 0.90 MA 0.10 Pb(I 0.90 Br 0.10 ) 3 ] 0.95 , two stock solutions were mixed with a volume ratio of 95:5.A 50 µl perovskite precursor solution was added to the prepared ETL substrates and then spun at 6000 rpm for 15 s.After 5 s of spinning, 100 µl of chlorobenzene was added dropwise on the spun substrate, followed by annealing on a hot plate at 100 • C for 1 h.For the surface-engineered perovskite film, 0.06 M of FABr (>98%, Sigma-Aldrich) and 0.06 M of TEAI, respectively, were dissolved in 2-propanol (⩾95%, Sigma-Aldrich), followed by mixing the two solutions with a 1:1 volume ratio.Then, 100 µl of the prepared solution containing the mixed salt was dropped on the annealed perovskite film and spun at 4000 rpm for 30 s, followed by annealing at 80 • C for 10 min.
Top electrode.A 50 nm-thick Au layer was deposited by using a thermal evaporator (Korea Vacuum Tech.).

Characterization
I-V measurements were performed using a source measurement unit (Model 2400, Keithley), where a solar simulator (Model K730, McScience) with an air mass (AM) 1.5 G intensity was used as a light source after a light calibration using a National Renewable Energy Laboratory (NREL)-calibrated Si solar cell equipped with a KG-5 filter.UV-vis absorbance spectra were obtained using a UV-2600 (Shimadzu, Japan).UPS measurements were performed by Thermo scientific Nexsa.IS measurements were conducted by a Potentiostat/galvanostat board (SP-300, Bio-Logic), where a 20 mV of alternating current (AC) perturbation ranging from 1 MHz to 1 Hz was placed on the direct current (DC) V app .

Results and discussion
When using Zn(TFSI) 2 as a dopant, a different doping mechanism of spiro-MeOTAD is observed compared to LiTFSI [17].While LiTFSI requires oxygen to oxidate the spiro-MeOTAD [18,19], a doping process of spiro-MeOTAD spontaneously occurs in the presence of Zn(TFSI) 2 even in an inert atmosphere by forming zinc complexation [17].The doping kinetics were investigated by monitoring the time-dependent absorbance of spiro-MeOTAD films employing additives, a dopant (LiTFSI or Zn(TFSI) 2 ) and tert-butylpyridine (tBP), with the same composition being used for PSCs.
Figures 1(a) and (b) represent the ultraviolet-visible (UV-vis) absorption spectra of the spiro-MeOTAD films doped with LiTFSI and Zn(TFSI) 2 , respectively.UV-vis absorbance measurements were consecutively conducted for the spiro-MeOTAD films being aged under ambient air over 500 h.The absorbance peak at 377 nm indicates pristine spiro-MeOTAD, while the peak at 520 nm is responsible for the oxidized spiro-MeOTAD (spiro-MeOTAD + ) formed by doping [19,20].In figures 1(a) and (b), the peak intensity of 377 nm gradually decreases under aging due to a consecutive conversion of pristine spiro-MeOTAD into spiro-MeOTAD + with the aid of a dopant, which consequently leads to a steady increase in peak intensity at 520 nm.It is noticeable that the absorbance peak at 520 nm is grown more rapidly and gets saturated with LiTFSI compared to Zn(TFSI) 2 .
Figure 1(d) represents the generating rate of spiro-MeOTAD + by first deriving [spiro-MeOTAD + ]/[spiro-MeOTAD] 0 .The generating rate of spiro-MeOTAD + by LiTFSI rapidly decreases by an order of magnitude within 400 h, which indicates an early arrival of the saturation point.On the other hand, Zn(TFSI) 2 tends to oxidize the spiro-MeOTAD more steadily in the early stage (up to 200 h), showing constant differential values, followed by a gradual reduction in the generating rate of spiro-MeOTAD + .Unfortunately, [spiro-MeOTAD] t is not enough to simply determine the rate law, as discussed in figure S1, yet the distinct difference in dopant-dependent oxidation kinetics of spiro-MeOTAD underlies a different aging time to reach the best device performance as well.The doping saturation time, however, cannot be exactly the same as the device aging time for optimized performance because of the different extent of negative influences under aging (e.g. the hygroscopicity of LiTFSI, mobile Li cation, etc).
PSCs were prepared to compare the effects of dopants (LiTFSI and Zn(TFSI) 2 ) on the time-dependent photovoltaic performance based on a normal structure, consisting of fluorine-doped tin oxide as a transparent conducting oxide, compact-and mesoporous-TiO 2 as the ETL, perovskite, spiro-MeOTAD as the HTL, and Au as a top electrode.The prepared devices employing either LiTFSI or Zn(TFSI) 2 as a dopant for spiro-MeOTAD were aged in ambient air, in the same manner as the spiro-MeOTAD films for UV-vis measurements, but different light exposure was inevitably applied due to the exposure of 1 sun illumination for measuring the photovoltaic performance.
The photovoltaic parameters of the short-circuit current density (J SC ), V OC , fill factor (FF), and PCE were consecutively measured and are shown in figures 2(a)-(d), respectively.For the case of LiTFSI, all photovoltaic parameters readily closely approach the highest values being monitored after 20 h of aging.More specifically, the J SC , V OC , and FF exhibit almost 100%, 95%, and 92% of the optimized values at t = 20 h, respectively, as soon as the devices are prepared (t = 0, without any intended aging).Meanwhile, devices with Zn(TFSI) 2 require a much longer aging time to approach the optimized performance point (t ⩾ 270 h), where the J SC and FF particularly demand a longer aging time (only 64% and 46% of the optimized J SC and FF, respectively, are observed at t = 0) compared to theV OC (92% of the optimized V OC is shown at t = 0), leading to a poor PCE in the beginning but with a continuous increase over 270 h.This is because the initial low hole mobility of spiro-MeOTAD under the slow doping process by Zn(TFIS) 2 (as shown in figures 1(c) and (d)) is detrimental to the J SC and FF.Photogenerated holes are prone to accumulate in the HTL owing to its low hole mobility, which accelerates the carrier recombination at the perovskite/HTL interface and thus prevents the photogenerated holes from being extracted into an external circuit [21].Therefore, the J SC and FF are directly affected by immature hole mobility in the beginning.
On the other hand, the V OC is less mobility-dependent because the open-circuit condition tends to form a spatial flat band across the perovskite bulk, where the photogenerated carriers are more dominantly governed by electron and hole quasi-Fermi levels in the perovskite bulk than any optoelectronic properties of the HTL and ETL [22].Likewise, the slow performance approach to the optimized point by Zn(TFIS) 2 makes an early estimation difficult, providing only ∼27% of the best PCE immediately after device preparation (t = 0), whereas ∼86% of the optimized PCE can be promptly examined from the beginning (t = 0) by using LiTFSI.
Such a long aging time for Zn(TFSI) 2 to optimize device performance would be troublesome in many aspects (e.g.unreliable initial performance, time-inefficiency, strict atmosphere control during the doping process, etc), which could mean avoiding the use of Zn(TFSI) 2 in spite of its strong advantage of long-term stable high performance [10,16].Therefore, control of energy alignment was attempted by interface engineering at the perovskite/HTL interface to shorten the device aging time for performance optimization.The interface energy alignment was deliberately tuned by reconstructing the top surface of the perovskite with mixed salt of formamidinium bromide (FABr) and CF 3 CH 2 NH 3 I (TEAI).It has been reported that FABr initiates surface reconstruction by reacting with excess PbI 2 , where TEAI indirectly controls the recrystallization by interacting with FABr [23].TEAI, meanwhile, is effective in passivating the perovskite surface to a lower defect density [23].
Ultraviolet photoelectron spectroscopy (UPS) measurements were carried out to understand the energetic offsets (∆E offset ) between the perovskite bulk and Zn(TFSI) 2 -doped spiro-MeOTAD.Figure 3(a) shows E cutoff (left) and E onset (right) from the UPS measurements, where the band energy level was determined by −E HOMO/VBM = 21.21eV − (E cutoff − E onset ) [24].The valence band maximum (VBM) of perovskite bulk shows −5.65 eV, being close to reported values [25], which is aligned with the HOMO of spiro-MeOTAD.The spiro-MeOTAD doped by Zn(TFSI) 2 shows a −5.40 eV HOMO level, with a further lowering by aging (∆E HOMO = −0.08 eV after 30 h), implying the effect of doping on the downshift of E HOMO [26][27][28].The initial ∆E offset between the E VBM of perovskite and E HOMO of spiro-MeOTAD was found to be 0.25 eV.The large ∆E offset between perovskite and the HTL encourages the photogenerated holes to transfer to the HTL even at open-circuit conditions and results in a large hole population in the HTL, where an obvious energy loss reluctantly occurs under the exponential increase in charge recombination rate as a function of ∆E offset [22].The downshift of E HOMO by aging, leading to a reduced ∆E offset at the perovskite/HTL interface, is well reflected in the marginal but gradual increase in the V OC shown in figure 2(b).
When measuring the UPS of the surface-engineered perovskite film, the E VBM was raised to −5.56 eV from −5.65 eV of the control film without surface treatment.In figure 3(b), an overall energy diagram is illustrated by reflecting the surface engineering, where the reduced ∆E offset makes the perovskite/HTL interface get closer to an ideal energy alignment with an insignificant ∆E offset [22].The resultant energy level alignment by surface reforming would be particularly favorable for energetically and spatially separating photogenerated holes and electrons and decreasing the chance of recombination [29,30].Furthermore, TEAI can serve solely as a passivating agent to reduce the carrier recombination by decreasing the defect density [23], as well as preventing electrons in both the conduction band and valence band from interacting with holes [29,30].Therefore, the surface engineering of perovskite films is likely to shorten the aging time of Zn(TFSI) 2 -based devices by controlling the charge recombination rate at the interface.
Photovoltaic parameters of PSCs employing Zn(TFSI) 2 -doped spiro-MeOTAD are compared in figure 4 to confirm the effect of the interface engineering on aging time.In figure 4, the initial photovoltaic parameters (black symbols) were measured after 10 h from device preparation while the average value (red dashed line) at the best performance point was measured at 270 h for control devices and at 20 h for devices with surface reconstruction of perovskite, respectively.When the control devices without surface engineering were measured after 10 h from preparation (t = 10 h), the J SC , V OC FF, and PCE of the control devices achieved 88.9%, 93.2%, 90.2%, and 74.9%, respectively, of the saturated parameters monitored at t = 270 h.On the other hand, devices with surface engineering showed a remarkably early saturation even at 10 h by reaching almost 100% of the best performance observed at t = 20 h.The swift increase specifically in the J SC and V OC (figures 4(a) and (c)), being mainly responsible for the long aging time (t = 270 h) for control devices, eventually leads to 99.4% acquisition, at t = 10 h, of the best PCE by interface engineering  (figure 4(d)).Therefore, it is evidenced that energy alignment control via interface engineering can effectively manipulate the device aging time, resulting in a distinct reduction in aging time from 270 h to 10-20 h for Zn(TFSI) 2 -based PSCs.
It is assumed that the prominent effect of surface reconstruction on reducing device aging time is closely related to a change in charge recombination behavior with the aid of the controlled energy alignment at the perovskite/HTL interface [21,22].Therefore, impedance spectroscopy (IS) measurements were carried out for the devices at an early stage of aging to investigate the recombination resistance (R rec ) at the perovskite/HTL interface.Figure 5(a) shows current-voltage (I-V) curves of devices being utilized to obtain Nyquist plots under 1 sun illumination.As noted in figure 4, the device with an engineered surface demonstrates obviously higher parameters (J SC = 21.6 mA cm −2 , V OC = 1.174V, FF = 0.772, and PCE = 19.6%)compared to the control device (J SC = 21.2 mA cm −2 , V OC = 1.024V, FF = 0.715, and PCE = 15.5%) at t = 10 h.The obtained Nyquist plot by measuring IS was fitted by using a simple equivalent circuit, as shown in figure S2, where the last R-C component at low frequency is dominated by recombination behavior specifically at the interface and is thus used for R rec estimation [31].Because the opposite interface (ETL/perovskite) is identical for both devices, R rec is dominantly governed by the perovskite/HTL interface.Figure 5(b) shows R rec as a function of applied voltage (V app ), where the R rec of the surface-engineered device is an order of magnitude higher than that of the control device over a whole range of V app .The R rec at 0.6 < V app < 1.0 V reflects the recombination behavior near the maximum power point, being closely related to the FF, while the R rec at V app = 1.0 V responds to near the open-circuit condition, correlating with the V OC .Therefore, the overall higher R rec in figure 5(b) is in good accordance with the photovoltaic parameters in figure 5(a).The photogenerated hole tends to recombine at the interface more easily once it is transferred to the HTL, having lower mobility than the case where the hole remains in the perovskite bulk because the bulk recombination in perovskite is negligible based on its high defect tolerance [32,33].In other words, the suppressed recombination at the perovskite/HTL interface by decreasing ∆E offset as well as passivating surface defects effectively counterbalances the low hole mobility of the HTL at the early stage of device aging.

Conclusions
A much longer aging time (∼270 h) of PSCs employing Zn(TFSI) 2 as a dopant for spiro-MeOTAD is required to reach the best device performance compared to that of LiTFSI (∼20 h) since Zn(TFSI) 2 reveals comparably slow doping kinetics of spiro-MeOTAD.In this respect, interface engineering between perovskite and spiro-MeOTAD is found to be an effective strategy to reduce device aging time to 10-20 h, which is comparable with LiTFSI.The surface reconstruction of perovskite with mixed salts leads to a beneficial energy alignment at the perovskite/spiro-MeOTAD interface as well as a defect passivation on the perovskite surface, which effectively prevents increased hole population in the spiro-MeOTAD from recombining owing to its low hole mobility at the early stage of aging.Therefore, the low hole mobility of spiro-MeOTAD can be sufficiently offset by retarding the hole recombination rate under a carefully tailored interface.This study provides insight not only into the performance evaluation of HTLs, employing a new dopant, but also into an important interface design strategy for HTLs.

Figure 2 .
Figure 2. Photovoltaic parameters of PSCs employing LiTFSI or Zn(TFSI)2 as a dopant for spiro-MeOTAD.(a) JSC, (b) VOC (c) FF, and (d) PCE are shown as a function of aging time.Black and red symbols represent the data point obtained from LiTFSI and Zn(TFSI)2, respectively.

Figure 3 .
Figure 3. (a) UPS spectra for perovskite film (black line), perovskite film with an engineered surface (red line), Zn(TFSI)2-doped spiro-MeOTAD film as prepared (blue line), and aged Zn(TFSI)2-doped spiro-MeOTAD film (green line).Dashed lines indicate the fitting.The aging for spiro-MeOTAD was carried out for 30 h.(b) Schematic illustration of an energy diagram at the interface between perovskite and Zn(TFSI)2-doped spiro-MeOTAD.The engineered surface was formed by FABr and TEAI while the passivation molecules indicate TEAI.

Figure 4 .
Figure 4. Photovoltaic parameters of PSCs employing Zn(TFSI)2 as a dopant for spiro-MeOTAD.(a) JSC, (b) VOC (c) FF, and (d) PCE are compared depending on the surface engineering of perovskite film.The black data points were obtained after 10 h from device preparation.The red dashed line indicates the average value of each parameter when reaching the best performance.The control devices and devices with engineered surfaces reached the optimized point at 270 h and 20 h, respectively.

Figure 5 .
Figure 5. (a) I-V curves of the control device and the device with an engineered surface (red).(b) Rrec of the control device (black) and the device with an engineered surface (red) as a function of Vapp.The measurements were performed at t = 10 h.