Crowning lithium ions in hole transport layer toward stable perovskite solar cells


 State-of-art perovskite solar cells exhibit comparable power conversion efficiency to silicon photovoltaics. However, the device stability remains a major obstacle that restricts widespread application. Doping hole transport layer induced hygroscopicity, ion diffusion, and use of polar solvent are detrimental factors for performance degradation of perovskite solar cells. Here, we report phase transfer catalyzed LiTFSI doping in Spiro-OMeTAD to address these negative impacts. 12-crown-4 as an efficient phase transfer catalyst promotes the dissolution of LiTFSI without requiring acetonitrile. Crowning Li+ ions by forming more stable and less diffusive crown ether-Li+ complexes retards the generation of hygroscopic lithium oxides and mitigates Li+ ion migration. Optimized solar cells deliver enhanced power conversion efficiency and significantly improved stability under humid and thermal conditions compared with the control device. This method can also be applied to dope π-conjugated polymer. Our findings provide a facile avenue to improve the long-term stability of perovskite solar cells.


Introduction
Despite an encouraging power conversion e ciency (PCE) exceeding 25% for state-of-art perovskite solar cells (PSCs) being su cient for practical applications, 1-6 a huge concern over the device stability remains to be addressed. Except for the vulnerability of perovskite layers toward ambient conditions, 7 the most frequently used small-molecule hole transporting layer (HTL) for highly e cient PSCs, Spiro-OMeTAD [2,2',7,7'tetrakis(N,N-di-p-methoxyphenylamine)-9,9'-spirobi-uorene], shows a poor stability toward humid and thermal conditions. Limited by an intrinsically low mobility of pristine Spiro-OMeTAD, LiTFSI [bis(tri uoromethanesulfonyl)imide] must be added as a p-type dopant to increase the conductivity. 8 LiTFSI, however, has low solubility in chlorobenzene (CB), the solvent for dissolving Spiro-MeOTAD. It is usually pre-dissolved in a polar solvent, typically acetonitrile. 9 The instability of Spiro-MeOTAD mainly arises from the indispensable dopants. On the one hand, the metal halide perovskite are vulnerable to polar solvent and it is desired to avoid or reduce the use of acetonitrile. 10 On the other hand, the hygroscopic lithium salt facilitates moisture penetration and Li + ion migration that is accelerated under high temperature conditions induces lithium intercalation into the perovskite layer with the formation of deleterious defects. 11 Thus far, considerable efforts have been devoted to optimizing the HTL of PSCs for enhanced PCE and stability. [12][13][14] The research community is actively searching for equivalent substitutes [15][16][17][18] to Spiro-OMeTAD for simple synthesis and cost reduction. Notwithstanding the recent success, LiTFSI is essential to these newly synthesized HTLs and Spiro-OMeTAD still represents the best choice for PSCs at the current stage owing to its high compatibility with dopants and energy band matching with perovskite. 19,20 Toward Spiro-OMeTAD based HTL, alternative dopants instead of LiTFSI such as Spiro(TFSI) 2 , 21 lithiumion endohedral fullerene, 22 and lithium-ion-free salts 23,24 have been studied to eliminate the notorious effect of Li + ions. Recently, CO 2 doping was developed to oxidize the Spiro-OMeTAD precursor solution and reduce the density of Li + ions in the HTL. 25 Unfortunately, most of these advances bring the bene t of improved device stability at the cost of sacri cing PCE. The progress in boosting the long-term stability of PSCs lags far behind the rapid increase in PCE. Thus, there is an urgent need to minimize the negative effects of dopants while maintaining device performance. Developing a simple yet e cient method to solve these problems is expected but now it remains challenging.
Here, we demonstrate a phase transfer catalyzed method for LiTFSI doping in Spiro-OMeTAD to improve the stability of PSCs. The modi ed composition in Spiro-OMeTAD precursor solution is superior to the conventional recipe in following aspects. First, 12-crown-4 as a lithium ionophore can selectively bond with Li + ions through coordination and promote the dissolution of LiTFSI in CB without requiring acetonitrile. Second, the crowned Li + ions with a larger size and mass become less active and more di cult to migrate, in turn posing a lowered threat to the device stability. Third, Li(12-crown-4) + complexes can passivate the defects and reduce the charge recombination both at the interface and in the bulk. The optimized Spiro-OMeTAD precursor enables PSCs to deliver a PCE approaching 23%, compared to 22.28% for the control device. The stability of PSCs is signi cantly improved due to crowning Li + ions in the HTL with an increased resistance toward moisture and decreased Li + ion migration. The phase transfer catalyzed LiTFSI doping can also be applied on π-conjugated polymer HTL, showing a universal application.

Results
Phase transfer catalyzed LiTFSI doping in Spiro-MeOTAD. The sandwich structure of PSCs with LiTFSI doped Spiro-OMeTAD as the HTL is shown in Fig. 1a. The components in conventional precursor solution include Spiro-OMeTAD, LiTFSI, acetonitrile, 4-tert-Butylpyridine (TBP), and CB, in which TBP assists a uniform dispersion of LiTFSI. The related molecule structures are provided in Fig. 1b. LiTFSI as an ionic compound has low solubility in CB (Fig. 1c). Crown ether possesses a speci c charge distribution and cavity size that can bind alkali-metal cation with a tted size through host-guest assembly. Particularly, 12-crown-4 has an interior cavity diameter that matches well with Li + ion size. 26 In this study, it was chosen to promote the dissociation of LiTFSI and increase the solubility through the following coordination reaction: LiTFSI+12-crown-4→Li(12-crown-4) + +TFSI -. After adding 12-crown-4, LiTFSI can be dissolved quickly in CB and we successfully prepared transparent precursor solution without requiring acetonitrile. For simplicity, the precursor solution (HTL or PSCs) prepared with conventional or modi ed recipe are denoted as control and target precursor (HTL or PSCs), respectively, in the following discussion unless otherwise stated.
According to electrospray ionization-mass spectrometry (ESI-MS), amass/charge ratio (m/e) of 136.1 that is indicative of H(TBP) + can be detected in the control precursor but the signal of Li(TBP) 4 + is not present (Fig. 1d), which implies easy desolvation between Li + ions and TBP. An extra intense peak with a m/e of 183.1 assigned to Li 12-crown-4) + is found in the target precursor, con rming the speci c coordinating reaction. 27 The interaction between 12-crown-4 and Li + ion was further characterized by nuclear magnetic resonance (NMR) and Fourier transform infrared spectroscopy (FTIR). Compared with the mixed solution without LiTFSI (acetonitrile+12-crown-4), a positive 1 H NMR chemical shift of 12crown-4 after being combined with Li + ion is identi ed for the mixture with LiTFSI (Fig. 1e Fig. 1), validating the formation of crown ether-Li + complex.
Notably, the control and target precursor exhibit a slight different color ( Supplementary Fig. 2) and absorption pattern (Fig. 1f).The difference is caused by the change in the Li + ion solvation after modifying the precursor recipe. A schematic drawing that depicts the Li + ion solvation shell is shown in the inset. We studied the solvation energy of Li + ions with rst-principal calculations (Supplementary Fig. 3 and Supplementary Note 1). In the control precursor, four TBP molecules surrounding a Li + ion is most favorable from the perspective of energy reduction and a stable solvation shell consequently forms.
The Li + ion solvation in the target precursor, however, is different. Li + ion can tightly bond with 12-crown-4 with a strong bonding and the crowned Li + ion can further interact with a TBP molecule which will be discussed later. As a consequence of less anchoring sites available, there will be more free TBP molecules that can induce Spiro-OMeTAD de-doping because of the chemical reaction with oxidized Spiro-OMeTAD. 30 Both the absorption spectra and the conductivity measurement ( Fig. 1g and Supplementary Note 2) certify that the modi cation toward the HTL recipe will not greatly affect the generation of oxidized species and the charge transport. We further monitored the time-dependant change in the PL spectra of unoxidized HTLs in ambient air (Fig. 1h). Pristine Spiro-OMeTAD under UV excitation emits intense blue light with an emission peak around 420 nm while p-type doping results in photoluminescence (PL) quenching. 25 The PL intensity of both the HTLs gradually decrease as the exposure time, a phenomenon of doping in progress. The decrease in PL intensity is nearly in step with each other, proving a negligible impact of 12-crown-4 on the oxidation rate of Spiro-OMeTAD. Ultraviolet photoelectron spectroscopy (UPS) of the oxidized HTLs was measured to study the energy band structure ( Supplementary Fig. 4). The two HTLs after oxidation show only a small difference in the energy band position and work function, which suggests a similar p-type doping level regardless of the HTL composition.
Elucidation of doping mechanism. To deepen our perception about the role of 12-crown-4 in the doping process, cyclic voltammetry (CV) was conducted to study the oxygen redox couple in LiTFSI-containing electrolyte without or with 12-crown-4. The CV curve without 12-crown-4 has an obvious oxidation peak of O 2 2at 0.01 V while no oxidation peak of O 2 can be found ( Fig. 2a 7 Li NMR spectra of the precursor solutions before and after conducting O 2 -bubbling ( Fig. 2b) were measured to further clarify the difference between the two HTLs in the oxidation process. The O 2bubbling process is an effective method to pre-oxidize HTL precursor solution. The control and target HTL precursor without pre-oxidization has a peak around -0.85 ppm and -0.78 ppm, respectively. A positive chemical shift in the 7 Li NMR spectra suggests a different coordination condition for Li + ion due to the formation of crown ether-Li + complexes in the target HTL precursor solution. 26 This phenomenon is in agreement with the result of 1 H NMR as discussed above. After O 2 -bubbling, the dominating peak position remains almost the same for the control precursor and the signal at the positive side slightly increases, which is due to the formation of Li 2 O 2 that contains Li vacancies. 33 The 7 Li NMR signal of the target precursor shifts to -0.97 ppm after pre-oxidization. We speculate that this change is attributed to the replacement of TBP with O 2 − as the ligand for the crowned Li + ion. In lithium oxides, the 7 Li NMR peak position follows the order: LiO 2 <Li 2 O 2 <Li 2 O and the peaks at lower frequencies can be assigned to Li species with a higher coordination number. 35,36 The oxidation induced chemical shift is consistent with the enhanced coordination number for Li + ions after the bonding between O 2 − and the crowned Li + ion.
To get an in-depth understanding of the molecular interaction at the nanoscale, we calculated the binding energy of involved con gurations (Fig. 2 9 Lithium oxides are extremely hygroscopic and can easily react with moisture to form lithium hydroxide. 37,38 It is one of the main reasons for the instability of Spiro-OMeTAD based HTL and will be discussed later. In the target HTL, Li + ions are crowned by 12-crown-4 and the direct reaction with oxygen-reduction species is postponed, leaving less opportunity for the generation of lithium oxides. The crowned Li + ion with a decreased Lewis acidity has a strong interaction with the moderately soft base O 2 − as con rmed by the calculations and C-V analyses, so they can be stabilized in the nal HTL. At this point, the oxidation mechanism of the p-type doping in the target HTL is straightforward and the involved oxidation process is presented in Fig. 2g. Charge dynamics and photovoltaic performance. Both the control and target HTLs deposited on perovskite lms are compact according to the scanning electron microscope (SEM) and atomic force microscope (AFM) images ( Supplementary Fig. 5). The perovskite/HTL lms have quenched PL intensity compared with the pristine perovskite lm and the one with a target HTL display a little stronger PL intensity than that with a control HTL (Fig. 3a). The time-resolved PL (TRPL) spectra ( Supplementary   Fig. 6) can be divided and tted into two parts (Supplementary Table 1). With the existence of HTL, the fast process (τ 1 ) relates to the charge extraction and the slow one (τ 2 ) represents the radiative recombination. 39 The perovskite lm with a target HTL has a lower τ 1 and higher τ 2 . We conclude that the target HTL possesses a better capability in extracting charge carriers from perovskite 40  can passivate the perovskite and reduce charge recombination. 29 The defect passivation will exert a positive in uence on the device performance. 41 In addition, we measured the space charge limited current (SCLC) of the HTLs (Fig. 3b and Supplementary Note 6) and the defect state density in the target HTL is lower than that in the control HTL.
We fabricated PSCs with a regular-type structure of ITO/SnO 2 /perovskite/HTL/Ag to investigate the photovoltaic performance (Fig. 3c). Crowning lithium ions improves device stability. To evaluate the stability of the perovskite lms at the presence of HTL, as-prepared perovskite lms with different HTLs were stored in an ambient condition with a humid of around 40%. The fresh samples are dark in color (Fig. 4a). After aging for 550 h, the perovskite lm with a control HTL turns yellow while the one with a target HTL still keeps a dark color. To obtain more insights into the degradation, we measured the XRD patterns of the sample before and after aging (Fig. 4b). Regarding the perovskite lm with a control HTL, the signal from PbI 2 increases markedly and an extra peak appears at 11.6 o resulting from δ-phase perovskite, 42 which indicates a severe degradation of perovskite caused by the moisture. Conversely, the PbI 2 content of the perovskite lm with a target HTL holds constant and the δ-phase has not formed during aging, suggesting a better resistance toward the moisture. The change in nanoscale morphology by comparing the cross-sectional SEM images between the fresh and aged samples 43 and the water contact angle of the HTLs ( Supplementary  Fig. 10, 11) further illustrate the better moisture stability of the target HTL.
We then explored the Li + ion distribution in the PSCs by measuring time-of-ight secondary ion mass spectrometry (ToF-SIMS). ToF-SIMS depth pro les and corresponding 3D mapping are provided in Fig. 4c-4f. In both the PSCs Li + ion concentration does not obey a simple Fickian diffusion distribution and there is a higher concentration in the SnO 2 layer than that in the perovskite layer. This suggests a lower a nity of Li + ions to perovskite than SnO 2 . The same trend of Li + ion distribution had been reported in previous literature. 44,45 For the control device, a much higher Li + ion signal at the SnO 2 layer than that at the HTL can be found, suggesting a rather serious Li + ion migration. In sharp contrast, Li + ions primarily located within the target HTL and a relatively small portion of Li + ions are diffused out and detected at the SnO 2 layer, indicating the ion migration has been greatly suppressed. 17 The Li + ion signal intensity over the perovskite range in the target device is a little higher than that for the control, which is due to the migration of Li(12-crown-4) + complexes into the inner part of perovskite lm where the 12crown-4 has a high a nity to Li + ions.
To assess the moisture stability of the resultant PSCs, we traced the photovoltaic performance during storing the unencapsulated devices in the dark at 25 o C and 30% relative humidity. The target device maintains 95% of initial PCE after aging over 2000 h, whereas a signi cant degradation in PCE with only 67% initial value being preserved after 1400 h occurs for the control device (Fig. 4g). The thermal stability of control and target PSCs were compared with unencapsulated devices aged at 60 o C in a glove box. The target device shows a much slower decrease in PCE than the control device and 90% of initial PCE can be maintained after thermal aging for 700 h ( Supplementary Fig. 12). The lowered Li + ion diffusivity in the target HTL due to a strong complexation of the Li + ion with 12-crown-4 contributes to a higher degree of thermal stability. These results fully demonstrate that the stability of the target solar cell is better than that of the control device, which is ascribed to the use of high-quality HTL.
Phase transfer catalyzed doping in π-conjugated polymer. π-conjugated polymers can be also used as HTL for PSCs in spite of less satisfactory performance compared with Spiro-OMeTAD. Finally, we performed phase transfer catalyzed LiTFSI doping in poly(3-hexylthiophene-2,5-diyl) (P3HT), a wellknown π-conjugated conductive polymer. The addition of 12-crown-4 into the P3HT based HTL precursor also avoid the use of acetonitrile and convert the mobile Li + ions into the crown ether-Li + complexes. PSCs fabricated using P3HT as the HTL with a phase transfer catalyzed LiTFSI doping exhibit comparable PCE compared with conventional P3HT HTL based device (Supplementary Fig. 13). The P3HT layer with phase transfer catalyzed LiTFSI doping protects the perovskite lm from moisture attacks in humid conditions (Supplementary Fig. 14). The improvement in stability is primarily attributed to the formation of the crown ether-Li + complexes in the HTL with an enhanced resistance against the moisture.

Discussion
In summary, we have demonstrated the feasibility of phase transfer catalyzed doping in small molecule and conjugated polymer as HTL for stable and high-e ciency PSCs. We show that 12-crown-4 as an e cient phase transfer catalyst can ensure a successful p-type doping and capturing Li ions by forming crown ether-Li + complexes in the HTL greatly reduces dopant induced side effects such as ion migration, moisture sensitivity, and perovskite layer degradation. This simple but effective doping process can simultaneously improve the photovoltaic performance and long-term stability, which may help pave the way for the commercialization of PSCs in the future. The phase transfer catalyzed doping for organic semiconductor may also nd application in other optoelectronic devices by serving as e cient and stable building blocks.
Device fabrication. Patterned ITO substrates were cleaned sequentially with ethanol and deionized water in an ultrasonic system each for 15 min and then dried with high purity N 2 gas. After a hydrophilic treatment in a UV cleaner for 10 min, a compact SnO 2 layer was deposited by spin coating diluted SnO 2 colloid aqueous solution (5 wt%) and heated at 150°C for 30 min. Before transferring into a glove box for perovskite lm deposition, the substrates were treated with UV light for 10 min. Perovskite lms were deposited using an anti-solvent method. The perovskite precursor solutions (1.3 mol L −1 ) contained stoichiometric FAI, PbI 2 , MABr, and CsI in a mixed DMF/DMSO solvent (4:1 v/v). The perovskite solution was spin-coated in a two-step program at 1000 r min −1 and 6000 r min −1 for 10 s and 20 s, respectively. During the second step, chlorobenzene (100 µL) was quickly poured on the spinning substrate 5 s before the end of the program. An annealing step at 120°C for 20 min was followed to promote the crystallization of perovskite lm. The as-prepared perovskite lm has a composition of Cs for the aging test. For thermal stability test, the devices were aged at a temperature of 60 o C in a glove box full of N 2 . TOF-SIMS (Tofsims 5, Iontof) was used to obtain depth pro les and ion distribution in 3D space. The ion depth pro les were measured by sputtering PSCs with a 1 kV O 2 gas gun rastered over a 280×280 µm 2 area followed by probing the composition of the exposed surface with a 30 kV Bi + liquid metal ion gun and analyzing positive secondary ions across a 112×112 µm 2 area. The devices were sent out to external labs and the measurements were conducted 48 hours after the fabrication.

Data Availability
The authors declare that the data supporting the ndings of this study are available within the paper and its supplementary materials. Figure 1 Characterization for phase transfer catalyzed LiTFSI doping in Spiro-OMeTAD. a An schematic architecture of regular-type PSCs. b Molecular structures of Spiro-OMeTAD, LiTFSI , and 12-crown-4. c Optical images showing the dissolution of LiTFSI in CB without and with 12-crown-4. d ESI-MS spectra of control and target precursor solutions. e 1H NMR spectra of 12-crown-4 and acetrolline mixture without and with LiTFSI. f Absorption spectra of precursor solutions. Inset schematically depicts the Li+ ion solvation shell in control (left) and target (right) precursor. g Absorption spectra of HTLs. Inset shows the I-V curves for conductivity measurement. h Time-dependant PL intensity of unoxidized HTLs exposed in air.

Figure 2
Page 14/14 Spiro-OMeTAD oxidation mechanism. a CV scan of LiTFSI-DMSO electrolyte without and with 12-crown-4. The oxidation peaks are zoomed in the inset. b 7Li NMR spectra of precursor solution before and after O2-bubbling. Molecular interaction in different con gurations: c crowned Li+ ion, d crowned Li+/TBP, e crowned Li+/O2-. Schematic drawing showing the oxidation mechanism in f control and g target HTL.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.