ll

Article

Improving Efficiency and Stability of PerovskiteSolar Cells Enabled by A Near-Infrared-Absorbing Moisture Barrier

Qin Hu, Wei Chen, Wenqiang

Yang, ..., Zhubing He, Rui Zhu,

Thomas P. Russell

fengliu82@sjtu.edu.cn (F.L.)

hezb@sustech.edu.cn (Z.H.)

iamzhurui@pku.edu.cn (R.Z.)

tom.p.russell@gmail.com (T.P.R.)

HIGHLIGHTS

A new strategy to enhance the

device performance while

simplifying the device structure

New design rules to rationally

screen multi-functional interface

layer

Long-term stability without

encapsulation upon exposure to

moisture, heat, and light

Correlations between molecular

orientation or passivation and

device performance

A multi-functional interface layer with integrated roles of (1) electron transport, (2)

moisture barrier, (3) near-infrared photocurrent enhancement, (4) trap passivation,

and (5) ion migration suppression to simultaneously enhance the device efficiency

and stability is demonstrated. The narrow-band-gap non-fullerene acceptor, Y6,

was screened out to replace themost commonly used PCBM in inverted perovskite

solar cells. A significantly improved efficiency of 21.0% was achieved along with

the remarkable device stability (up to 1,700 h) without encapsulation upon

exposure to moisture, heat, and light.

Hu et al., Joule 4, 1575–1593

July 15, 2020 ª 2020 Elsevier Inc.

https://doi.org/10.1016/j.joule.2020.06.007

ll

Article

Improving Efficiency and Stabilityof Perovskite Solar Cells Enabledby A Near-Infrared-Absorbing Moisture Barrier

Qin Hu,1,2,12 Wei Chen,3,6,12 Wenqiang Yang,4,12 Yu Li,2,5 Yecheng Zhou,7 Bryon W. Larson,8

Justin C. Johnson,8 Yi-Hsien Lu,2 Wenkai Zhong,9 Jinqiu Xu,5 Liana Klivansky,10 Cheng Wang,9

Miquel Salmeron,2 Aleksandra B. Djuri�si�c,3 Feng Liu,5,* Zhubing He,6,* Rui Zhu,4,11,*

and Thomas P. Russell1,2,13,*

Context & Scale

Perovskite solar cells (PSCs) have

attracted tremendous attention

because of the high efficiencies,

ease of fabrication, and low cost

of production. However, further

enhancement of device efficiency

has been a bottleneck, and the

instability of the PSCs hampers

their commercialization. In this

work, we strategically introduce a

new multi-functional interface

layer that integrates five different

functions to improve the device

efficiency and long-term stability

of PSCs, pushing forward the

development of the PSC

technology. A significantly

improved power conversion

SUMMARY

Simultaneously improving device efficiency and stability is the mostimportant issue in perovskite solar cell (PSC) research. Here, westrategically introduce a multi-functional interface layer (MFIL)with integrated roles of: (1) electron transport, (2) moisture barrier,(3) near-infrared photocurrent enhancement, (4) trap passivation,and (5) ion migration suppression to enhance the device perfor-mance. The narrow-band-gap non-fullerene acceptor, Y6, wasscreened out to replace the most commonly used PCBM in the in-verted PSCs. A significantly improved power conversion efficiencyof 21.0% was achieved, along with a remarkable stability (up to1,700 h) without encapsulation under various external stimuli (light,heat, and moisture). Furthermore, systematic studies of the molec-ular orientation or passivation and the charge carrier dynamics atthe interface between perovskite and MFIL were presented. Theseresults offer deep insights for designing advanced interlayers andestablish the correlations between molecular orientation, interfacemolecular bonding, trap state density, non-radiation recombina-tion, and the device performance.

efficiency of 21.0% was achieved

along with the remarkable stability

(up to 1,700 h) without

encapsulation under various

external stimuli (light, heat, and

moisture). These results open new

avenues to design advanced

interlayers, simplifying the device

structure, and enhancing

efficiency and stability, that can

accelerate the market readiness of

perovskite-based

optoelectronics.

INTRODUCTION

Organic-inorganic hybrid perovskite thin-film solar cells have emerged as an effi-

cient solar-energy technology with excellent power conversion efficiencies (PCEs),

ease of fabrication, and low production cost, proving to be a game changer in pho-

tovoltaics.1–3 Significant effort has been devoted to optimizing device efficiencies4

and to further understand inherent material properties.5 Some strategies, including

crystallization and morphology optimization of perovskite thin films,6 composition-

tunable alloying7 for increasing light absorption, and interface engineering8 within

the device structure, have been adapted to enhance the PCEs. Despite these efforts

to improve device performance, the lifespan of perovskite solar cells (PSCs) is still

too short for practical use.9,10 Thus, simultaneously improving device efficiency

and stability has become the most important issue at present. The inherent ‘‘soft’’

crystal lattice of perovskite solids is one of the key reasons for poor stability, which

makes PSCs vulnerable to aging stresses, such as UV-light, moisture, electric field,

and thermal annealing. In addition, structural defects in the bulk and on the surfaces

of perovskite polycrystalline thin films, ion migration, hygroscopic additives, and the

thermal instability of charge transporting layers, are also contributors to PSCs’

Joule 4, 1575–1593, July 15, 2020 ª 2020 Elsevier Inc. 1575

1Department of Polymer Science andEngineering, University of Massachusetts,Amherst, MA 01003, USA

2Materials Sciences Division, Lawrence BerkeleyNational Laboratory, Berkeley, CA 94720, USA

3Department of Physics, The University of HongKong, Pokfulam, Hong Kong

4State Key Laboratory for Artificial Microstructureand Mesoscopic Physics, School of Physics,Frontiers Science Center forNano-optoelectronics & CollaborativeInnovation Center of Quantum Matter, PekingUniversity, Beijing 100871, China

5Frontiers Science Center for TransformativeMolecules and in situ Center for physicalSciences, School of Chemistry and ChemicalEngineering, Shanghai Jiao Tong University,Shanghai 200240, China

6Department of Electrical and ElectronicEngineering, Shenzhen Key Laboratory of FullSpectral Solar Electricity Generation (FSSEG),Southern University of Science and Technology,No. 1088, Xueyuan Rd., Shenzhen, Guangdong518055, China

7School of Materials Science & Engineering, SunYat-sen University, Guangzhou, Guangdong510275, China

8National Renewable Energy Laboratory, Golden,CO 80401, USA

9Adavanced Light Sources, Lawrence BerkeleyNational Laboratory, Berkeley, CA 94720, USA

10Molecular Foundry, Lawrence BerkeleyNational Laboratory, Berkeley, CA 94720, USA

11Collaborative Innovation Center of ExtremeOptics, Shanxi University, Taiyuan, Shanxi 030006,P.R. China

12These authors contributed equally

13Lead Contact

*Correspondence: fengliu82@sjtu.edu.cn (F.L.),hezb@sustech.edu.cn (Z.H.),iamzhurui@pku.edu.cn (R.Z.),tom.p.russell@gmail.com (T.P.R.)

https://doi.org/10.1016/j.joule.2020.06.007

llArticle

performance deterioration.11 Various approaches have been applied to improve the

perovskite quality and critical interfaces, and PSCs can now survive under environ-

ment stresses from hours to months. For instance, to selectively oxidize Pb0 and

reduce I0 defects, Europium ion pairs, Eu3+-Eu2+, can be added as the ‘‘redox shut-

tle,’’ increasing device efficiency and longevity.9 Ionic liquid additives12 have also

been shown to improve the surface energetic alignment and suppress ion migration,

resulting in longer-lasting output. Moreover, a double-layer perovskite structure

with in situ reaction of hydrophobic chemicals (n-hexyl trimethyl ammonium bro-

mide13 or lead sulfate14) can passivate the perovskite surface, yielding higher

open-circuit voltages (Voc) and enhanced moisture resistance.

In addition to the optimization of the perovskite active layer, significant contribu-

tions have also been made to reduce surface disorder, improve charge extraction,

and enhance the chemical stability of the electron or hole transporting layer (ETL

or HTL) toward the goal of maximizing device stability. Compared with the regular

n-i-p structure, the p-i-n (aka inverted) structure shows great potential to realize

long-term operational stability, due to the use of stable metal oxide semiconductors

or undoped conjugated polymers or small molecules as the interface layers.15–17

Having a robust ETL in p-i-n PSCs is also needed for implementing a perovskite as

the front cell in perovskite:silicon or all-perovskite tandem solar cells.18,19 Further-

more, inverted PSCs are also compatible with plastic substrates that require low-

temperature fabrication for flexible devices.20 Unfortunately, p-i-n PSCs have not

kept pace with n-i-p PSCs in device efficiencies due to lower short current densities

(Jsc)21 and/or larger Voc losses induced by non-radiative recombination.22 On top of

poorer performance, inverted PSCs also usually exhibit modest stability against wa-

ter and light, mainly due to the use of fullerenes with low-crystallinity and large ag-

gregation as the ETL.23 Therefore, new strategies to realize both high efficiencies

and high stabilities for inverted PSCs are urgently needed.

Here, we introduce amulti-functional interface layer (MFIL) approach to combine the

roles of (1) electron transport, (2) moisture barrier, (3) near-infrared (NIR) photocur-

rent enhancement, (4) trap passivation, and (5) ion migration suppression in inverted

PSCs, and to simultaneously improve the device efficiency and long-term stability of

PSCs. Compared with the fullerenes, the non-fullerenes can be easily chemically

tuned to obtain superior physical properties, such as light absorption, energy level,

and charge generation.24–27 Thus, non-fullerenes have the great potential to realize

these functions of MFILs. A new small-band-gap non-fullerene acceptor (NFA), Y6 (as

shown in Figure 1), a popular BT-core-based fused-unit dithienothiophen[3,2-b]-pyr-

rolobenzothiadiazole (TPBT) derivative, was screened from a series of narrow-band-

gap acceptor materials to replace the commonly used PCBMwithout any additive or

donor materials in the solid thin film. The molecular orientation, the electronic states

of local structures, and the surface morphology of the Y6 layer were systematically

investigated for the full device stack. The high mobility and suitable energy levels

of Y6 matched the perovskite active layer quite well,24 and the narrow optical

band gap extended the photoresponse cutoff from �775 to �940 nm, resulting in

an improved integrated Jsc. Interfacial bonding interactions between the perovskite

and Y6 were further analyzed using advanced surface spectroscopies: X-ray photo-

electron spectroscopy (XPS) and synchrotron-based X-ray total fluorescence yield

(TFY). We find that chemical complexing of electron-rich C–S–C, C=O, ChN, and

C–F functional groups in Y6 with the empty orbitals in Pb2+ or other electron traps

at the perovskite interface, led to reductions in trap state density and non-radiative

recombination, enhanced electron extraction, and suppressed ion migration in

operation, resulting in the enhanced Voc, fill factor (FF) and a champion PCE of

1576 Joule 4, 1575–1593, July 15, 2020

Figure 1. The Illustration of Device Configuration and Multi-functional Interface Layer Screening

(A) Schematic diagram of an inverted planar heterojunction PSC based on the MFIL, and the

functions marked in the diagram are (1) electron transporting, (2) moisture barrier, (3) near-infrared

photocurrent enhancement, (4) trap passivation, and (5) ion migration suppression.

(B) Normalized UV-vis absorption spectra of perovskite and different ETLs.

(C) The energy level schema of the perovskite and the charge transport layers in this work.

(D and E) (D) The J-V curves of PSCs based on three different narrow-band-gap NFAs and (E) the

corresponding EQE spectra.

llArticle

21.0%. We also used steady-state and time-resolved photoluminescence (TRPL)

spectra, steady-state microwave conductivity (SSMC) measurements, and ultrafast

transient absorption spectroscopy (TAS) to elucidate the underlying charge carrier

dynamics. Furthermore, in combination with the hydrophobic properties of Y6,

which impedes water penetration, the surface chemical stability, intrinsic photo-sta-

bility at standard AM1.5G conditions, ambient stability with humidity of 60%–65%,

and thermal stability at 85�C were all increased significantly for Y6-based devices in

contrast to PCBM-based devices. This MFIL strategy opens up new pathways for ETL

design and optimization for inverted PSCs, since it concomitantly enhances both de-

vice efficiency and stability, thus more broadly accelerating the market readiness of

perovskite-based optoelectronics.

Joule 4, 1575–1593, July 15, 2020 1577

llArticle

RESULTS AND DISCUSSION

MFIL Screening and Molecular Orientation

Figure 1A shows the device architecture of the inverted planar heterojunction

PSCs studied in this work. A cross-sectional scanning transmission electron

microscopy (STEM) image is shown in Figure S1. Poly(triarylamine) (PTAA) is

used as the HTL, and the tri-cation CsFAMA (FA, formamidinium; MA,

methylammonium) perovskite with a band gap of �1.6 eV was fabricated as photo-

active layer (see details in Experimental Procedures). We observed a uniform,

compact morphology of the perovskite layer from the SEM image (Figure S1B). As

a control, PCBM was used as the ETL to contrast a series of narrow-band-gap

NFAs24–28 (Y6, IEICO-4F and IOIC-2Cl). The UV-visible (UV-vis) absorption spectra

of tri-cation CsFAMA perovskite and the ETLs are shown in Figure 1B. All of the

NFAs studied here extend the light-absorption from the visible (�775 nm) to the

NIR region (�940 nm). However, when we fabricated full devices to screen these

ETLs, only the Y6-based device gave a higher efficiency (20.5%, Figure 1D) than

the PCBM control, a result of a Voc of 1.11 V, FF of 0.78, and Jsc of 23.7 mA$cm�2,

with contribution from the extended NIR photoresponse as shown in the external

quantum efficiency (EQE) spectrum (Figure 1E). In contrast, the IOIC-2Cl and

IEICO-4F devices both decreased the relative device performance with PCEs of

13.6% (Voc = 1.03 V, FF = 0.54, Jsc = 24.5 mA$cm�2) and 10.8% (Voc = 1.04 V,

FF = 0.46, Jsc = 22.5 mA$cm�2), respectively. The very poor FF and Voc results

were likely caused by mismatched energy level alignments and high interfacial

charge recombination as shown in Figure 1C. Thus, we propose the first guidelines

to screen effective narrow-band-gap NFAs as candidates for a MFIL: (1) complemen-

tary light harvesting, (2) a favorable surface energetic alignment (deeper lowest un-

occupied molecular orbital [LUMO] and highest occupied molecular orbital [HOMO]

energy than perovskite), (3) suitable charge mobility, and (4) electron donating func-

tional groups to passivate the perovskite surface. Of the three NFAs here, Y6 is the

only one that meets the criteria set for MFILs, though we also expect that inorganic

semiconductor materials, such as NIR-absorbing quantum dots, could do so as well,

as long as the energetic and functional group requirements are also fulfilled.

In organic semiconductors, exciton dissociation and charge transport are sensitive

to molecular order. We found that the crystallization and aggregation of Y6 can

be precisely controlled in different solvent environments. The p-p stacking of Y6

films for efficient charge extraction was investigated by grazing incident X-ray

diffraction (GIXD), as shown in Figure 2A. Three typical solvents with different boiling

points, chloroform (CF), toluene (Tol), and chlorobenzene (CB), were used to process

the Y6 films. The CF-processed Y6 (Y6-CF) film has a dominant p-p stacking diffrac-

tion peak at q � 1.77 A�1 in the out of plane (OOP) direction with a d-spacing of

3.55 A and a crystalline coherence length (CCL) of 21.36 A (calculated by Scherrer

equation29 and the full width at half maximum intensity was calculated fromGaussian

fitting), forming a typical face-on orientation. The (110) peak and (11-1) peaks, simu-

lated from single crystal structure with the unique 2D packing with polymer-like con-

jugated backbone (as shown in Figure S2), in the in-plane (IP) direction are located at

q � 0.29 and q � 0.42 A�1 with d-spacing of 21.65 and 14.95 A, respectively. The

GIXD scattering results and the related single crystal simulation indicate that the ba-

nana-like Y6 molecules grow or stack normal to substrate interface resulting in more

efficient charge transfer by p-p stacking.30 The diffraction profiles from the Tol-pro-

cessed Y6 (Y6-Tol) and CB-processed Y6 (Y6-CB) films are quite similar, but the p-p

stacking reflection is much weaker and less oriented in comparison to Y6-CF, sug-

gesting reduced crystallinity and a more random orientation. The enhanced

1578 Joule 4, 1575–1593, July 15, 2020

Figure 2. The Crystallization, Morphology, and Electronic Properties of Y6 Interface Layer

(A and B) (A) The 2D GIXD files and (B) corresponding intergrated line-files for Y6 films processed with different solvents.

(C) Molecular engineering and charge transfer in Y6 films with or without orientation.

(D)The d-spacing, peak area, and the CCL of p-p stacking diffraction in OOP direction.

(E) The peak ratio of p-p stacking diffraction in OOP to the diffraction at in plane.

(F) The surface potential distribution of perovskite/Y6 films measured by KPFM (the scale bars are 500 nm).

llArticle

scattering at q � 1.20 A�1 indicates an increased amorphous content in Y6-CB and

Y6-Tol. The corresponding integrated profiles in the IP and OOP directions are

shown in Figure 2B. The detailed CCL and areas of the characteristic peaks are sum-

marized in Figure 2D (OOP) and Figure S3 (IP). The Y6-CF has a greater peak area

and CCL for the p-p stacking and (11-1) reflections, in comparison with those of

Y6-CB and Y6-Tol, which have close d-spacings. In addition, the peak areas of p-p

Joule 4, 1575–1593, July 15, 2020 1579

llArticle

stacking reflections in OOP and IP are summarized in Figure 2E. The peak area ratio

of the p-p stacking reflection in OOP to that in IP for Y6-CF is �3.5, while the Y6-CB

and Y6-Tol values are close to 1.0, indicating that the p-p stacking for the Y6-CF film

is significantly more oriented normal to the substrate, reducing the surface disorder

at the perovskite and Y6 interface and enhancing efficient electron extraction and

transport from the perovskite active layers, as shown in Figure 2C.

The morphology and electronic properties of the surface of Y6 films were further

investigated by the Kelvin probe force microscope (KPFM). Figure S4 shows the sur-

face topography of the perovskite/Y6 films. The root mean square (RMS) roughness

values are 5.49 (Y6-CF), 5.95 (Y6-Tol), and 8.28 nm (Y6-CB), which could correspond

to effects of different solvent vapor pressures during the film drying process. The sol-

vent-rich films (with high-boiling point solvent) have insufficient time to heal surface

roughness caused by Marangoni instabilities, reducing the film roughness.31 The

redshift of the longest-wavelength light absorption peak (�840 nm) in the UV-vis

spectra, as seen in Figure S5, suggest the enhanced molecular packing with larger

aggregations in Y6-Tol and Y6-CB films, this is consistent with the recent report in

Y6-based organic solar cells.30 The contact potential difference (CPD) images of

perovskite/Y6 films are shown in Figure 2F. In agreement with the surface topog-

raphy, the variation of the CPD value for Y6-CF is 0.016 V and Y6-Tol is 0.022 V

whereas Y6-CB is 0.024 V. The similarity of the CPD values suggests that different

molecular packing or surface morphology in Y6 films has little influence on the sur-

face potential. In summary, compared with Y6-CB and Y6-Tol, the Y6-CF film shows

the dominant face-on orientation with the highest degree of crystallinity (GIXD re-

sults), as well as the smoothest surface topography and surface potential distribution

(KPFM). Therefore, the best device performance with Y6-CF was expected.

Device Performance and Interface Characterizations

Inverted PSC devices incorporating Y6 layers with different molecular orientations

(based on casting solvent) were fabricated and characterized. The device parame-

ters measured under a standard AM1.5G solar illumination (100 mW$cm�2) are sum-

marized in Table S1. All the Y6-based devices have a very similar Jsc and Voc. How-

ever, the Y6-CF device had the highest FF of 0.78 and the highest PCE of 20.5%,

while the Y6-Tol and Y6-CB had lower FF of 0.70 and 0.68, resulting in PCEs of

17.9% and 17.7%, respectively. A sharp decrease in the FFs of Y6-Tol- and Y6-CB-

based devices is ascribed to the increased series resistance (Rs) and decreased shunt

resistance (Rsh) (shown in Table S1), which could be caused by increased surface

roughness and less efficient interfacial charge transport. These results further argue

that the face-on orientation and p-p stacking in the Y6-CF film is responsible for

more efficient extraction and transport of electrons from the perovskite layer. More-

over, the device performance also confirms that the photoexcited excitons from the

Y6-CF layer can be dissociated at the interface and contributes to a higher Jsc. This is

further evidenced by the enhanced EQE in the NIR region of Y6-CF devices (Fig-

ure S6). The Y6-CF devices were also optimized by varying the film thickness using

different solution concentrations (Figure S7), a 9.5 mg/mL solution giving a thickness

of 46.3 nm (Table S2) yielded the best performance. When a very thin layer of C60

(20 nm) was added onto the Y6-CF film to further enhance the charge extraction,32

both the FF and Voc increased, yielding a champion PCE of 21.0%. As a result, this

approach was used for all devices.

The device performance with Y6 and traditional PCBM was further examined.

Compared with those of the PCBM-based device, the Voc (from 1.10 to 1.12 V)

and FF (from 0.77 to 0.79) of the Y6-based device are slightly enhanced while the

1580 Joule 4, 1575–1593, July 15, 2020

Figure 3. The Device Performance

(A) The J-V curves of PSCs based with Y6 and PCBM, separately.

(B) The corresponding EQE spectra.

(C and D) (C) Steady-state and (D) transient-time photoluminescence spectra.

(E and F) (E) The t-DOS characteristics and (F) Mott–Schottky characteristic results.

llArticle

Jsc increased significantly from 22.64 to 23.68 mA$cm�2, leading to the enhance-

ment in the PCE from 19.2% to 21.0%, as shown in Figure 3A. The EQE spectrum

of the Y6 device (Figure 3B) showed an extended photoreponse in the NIR region

(�940 nm), demonstrating the synergistic functions of Y6, both as a photoactive

layer and a charge transport medium. The integrated current densities from the

EQEs are 21.86 mA$cm�2 (PCBM) and 22.85 mA$cm�2 (Y6), which are consistent

with the enhanced Jsc from the J-V curves. It should be noted that the devices

have negligible hysteresis under a range of different scanning conditions. In addi-

tion, the stabilized power output (SPO) at maximum power point (MPP) were also

tracked to confirm the efficiency reliability. As shown in Figure S8, the Y6-based de-

vices had a stabilized efficiency of 20.9%, while the PCBM-based device had an

output of 19.1%. In addition, the Y6-based devices also showed excellent reproduc-

ibility with small standard deviations, as shown in Table S3 based on 20 devices.

To explore the origins for the enhanced device performance of Y6-based devices,

the carrier transport properties were determined. Both steady-state and TRPL

spectra were measured (Figures 3C and 3D). The PL peaks of perovskite (PVSK)

with Y6 and PVSK with PCBM films are both located at �767 nm. Compared with

that of the pure perovskite sample, the PL intensities of both the PVSK with PCBM

and PVSK with Y6 samples decreased markedly, indicating a comparable charge

extraction for Y6 in comparison to PCBM. The charge recombination kinetics were

measured by TRPL spectra, which could be described with a double exponential

decay function, suggesting two decay processes, a non-radiative recombination

process and a radiative recombination process. The charge-carrier lifetimes are

shown in Table S4. The average lifetime for the pure perovskite is 602.0 ns, but drop-

ped to 30.1 ns for the PVSK with PCBM sample, due to charge-carrier transfer from

PVSK to PCBM, giving rise to the TRPL intensity quenching. The further decrease in

Joule 4, 1575–1593, July 15, 2020 1581

Figure 4. The Molecular Interactions and Reduced Trap States

(A–D) (A–C) XPS spectra and (D) TFY results of PVSK with Y6 interfaces.

(E) The schema of the molecular bonding between Y6 and perovskites.

(F) The diagram of the reduced trap states for the Y6 based device compared to the PCBM device.

llArticle

the TRPL intensity and shorter lifetime of 14.5 ns for PVSK with Y6 suggests a faster

charge transfer for the PVSK with Y6 than the PVSK with PCBM. This is also in agree-

ment with the reduction in trap density of states (t-DOS) in the complete devices, as

shown in Figure 3E. The t-DOS distribution can be obtained from the frequency-

dependent capacitance using DOS ðEuÞ= � VbiqW

dcdu

uKT , where Vbi, q ,W, C, u, K,

and T correspond to the following: the built-in potential, elementary charge, deple-

tion width, capacitance, angular frequency, Boltzmann constant, and temperature,

respectively.33 Vbi can be obtained from the 1/C2-V Mott-Schottky plots shown in

Figure 3F, where Vbi is given by the intersection on the bias axis.34 As shown in Fig-

ure 3E, the device with Y6 had a lower t-DOS over the entire trap depth region. The t-

DOS with the shallow trap region of �0.30–0.42 eV is assigned to traps at grain

boundaries or interfaces, while the deeper region of 0.50–0.60 eV refers to the

bulk defects inside the perovskite films.33,35 It is well known that solution-processed

perovskites often have defects,36 such as impurities, vacancies and interstitials both

in the bulk and at the surface, and dangling bonds at boundaries and surfaces,

which, in addition to the energy disorder of ETLs, can result in nonradiative recom-

bination of photo-carriers, preventing the devices from hitting their theoretical effi-

ciency limit.22 The t-DOS results in this study show that the non-radiative recombi-

nation in the Y6 device is significantly reduced, which is consistent with the

enhanced Voc and FF.

We used XPS37 and the TFY spectra38 near the N-edge to explore possible molecular

bonding or chemical interactions between the perovskite and Y6, which are possible

origins of decreased t-DOS and operational device stability with external stress. Fig-

ures 4A–4D and S9 show the XPS spectra of characteristic C–F, C=O, C–S–C, and

ChN groups and Pb 4f at the PVSK with Y6 interface. The control samples are

pure Y6 and pure perovskite films on indium tin oxide (ITO) substrates. As seen in

Figure 4A, the F 1s peak shifted by�0.2 eV to a higher energy and O 1s peak shifted

by �1.3 eV, originating from a higher oxidation state arising from the additional

1582 Joule 4, 1575–1593, July 15, 2020

llArticle

Coulombic interaction between the photoemitted electrons and the ion core. The

additional interactions could result from new Pb–F and Pb–O binding.39–42 New S

2p peaks of PbS appearing in Figure 4C confirmed the chemical interaction between

Pb2+ in perovskite and C–S–C in Y6 directly. However, no obvious peak shifting of Pb

4f (Figure S9) was observed, which could be due to the large amount of Pb-I (perov-

skite) bindings on the surface of the PVSK with Y6 film, and only a small part of Pb

defects or PbI2 residue were passivated by Y6. There are three different C–N bind-

ings in the perovskite (MA+ (C–NH2+), FA+(C=NH2

+, C–NH2)), while there are 4

different C–N bindings in Y6 (C–N–C, C=N–S, and two kinds of C=ChN). No fea-

tures can be distinguished from the XPS N 1s spectrum due to significant overlap

of possible peaks. High-resolution synchrotron-based TFY measurements were per-

formed to track the evolution of the p*ChN resonances in Y6. Figure 4D shows the

TFY p*ChN transition peaks emerged at 398–400 eV. The near disappearance of

the peak located at 398–399 eV in the PVSK with Y6 sample indicates an electron

loss from ChN in Y6, probably transfered to perovskite (Pb2+). The molecular inter-

actions were further simulated by density functional theory (DFT) calculations (Fig-

ures 4E and S10).43 The Y6 molecule initially lays down on the PVSK surface without

specific connections. The geometry was then optimized by Vienna Ab initio Simula-

tion Package (VASP). After optimization, ChNwas blended and connected with sur-

face Pb atoms, which suggests that ChN could be the most effective electron-

donating group to passivate the Pb traps on the perovskite surface. This is further

confirmed by a very recent report on the passivation of a small molecular, 6TIC-

4F, on the inorganic perovskite by comparing the formation of F–Pb, N–Pb, and

S–Pb, where the N–Pb bonding motif has the largest formation energy.44 Here, in

our simulation, the binding energy (Ebinding = Ecombined � Esurface � Emolecule) of Y6

with perovskites is calculated to be �0.64 eV, which is much higher than that for

PC61BM (�0.11 eV), indicating stronger interactions at the PVSK with Y6 interfaces.

We also performed simulations on the PVSK/ETL interface to compare the charge in-

jection efficiency. The density of states and corresponding orbitals of a PCBM ad-

sorbed on a perovskite surface are shown in Figure S10A. The conduction band

lays about 0.4 eV below the Fermi level. The HOMOof PCBM is located near a defect

level of adsorption. Due to the limited thickness layer, a lot of the surface states are

presented in the range of (�0.4 to �0.1 eV), which are mixed with the conduction

band edge. The LUMO and LUMO-1 of PCBM are�0.2 eV lower than the conduction

band. The electron injection driving force is 0.2 eV, which suggests that the electron

injection could be very efficient. One should note that the existence of the large

number of surface states is due to the thin slab model, which cannot represent the

real surface of a bulk perovskite. Figure S10B gives the density of states of the model

of a Y6 molecule adsorbed on a perovskite surface and the corresponding orbitals.

The valence band of perovskite is about 0.2 eV below the Fermi level, the HOMO of

the Y6 molecule is slightly higher than the valence band. Since the layer is not thick

enough, surface states are seen between 0.3 to 0.9 eV. The LUMO of the Y6 mole-

cules aligns at 1.0 eV. Hence, the gap of the Y6 molecule is 1.1 eV in our calculation.

The conduction band of the perovskite is 0.2 eV above LUMO, just as seen for PCBM,

meaning they should show similar charge injection efficiencies. These molecule-

scale chemical interactions studied by XPS, TFY, and DFT unveil the underlying

reasons of the reduction in trap states (as shown in Figure 4F) at the PVSK with Y6

interface, suppressing the non-radiative recombination and delivering the higher

efficiencies and better stabilities.

Charge Carrier Dynamics at the Perovskite/Y6 Interface

In order to understand whether excitons generated in Y6 dissociate at the PVSK with

Y6 interface akin to an organic photovoltaic (OPV) bulk-heterojunction (BHJ) system

Joule 4, 1575–1593, July 15, 2020 1583

llArticle

(like PM6:Y6 BHJ), we conducted SSMC45,46 and ultrafast TAS measurements on

neat films of Y6, PVSK, and the PVSK with Y6 bilayer. We compared the results

with those of a control OPV BHJ system (neat polymer donor PM6 and the BHJ

with Y6). The SSMC measurement is essentially an internal quantum efficiency

(IQE) of equilibrium-state photoinduced charge generation, where the absolute

signal is a product of the yield of carriers generated, weighted by their respective

motilities, and their charge-separated lifetime (4 3 Sm 3 t), recorded here over a

monochromated excitation range. A compliment to the EQE measurement, SSMC

contactlessly probes the sample with GHz frequency microwaves, meaning that

with both techniques, we can compare the quantum efficiency (QE) response of

the PVSK with Y6 interface either with or without contacts to help understand the

mechanism and origin of the observed NIR photocurrent (in Figure 1E). We investi-

gated whether the PVSK with Y6 interface mimics the charge generation interface

found in an OPV blend, where a donor-acceptor energy offset provides sufficient

driving force to split the exciton. As shown in Figure 5A, the PM6:Y6 BHJ sample

shows that long-lived free carriers are generated when Y6 is excited (PM6 does

not absorb at wavelengths longer than 680 nm, Figure 5B) as evidenced by the

SSMC signal at wavelengths of 680 to 920 nm, perfectly tracking the absorption con-

tributions of Y6. In contrast, the SSMC signal for the PVSK with Y6 samples only ap-

pears once the excitation wavelength meets the band edge of the perovskite. These

results clearly indicate that the mechanism responsible for NIR photocurrent at the

PVSK with Y6 interface is not the same as in OPVs.

This notion is further supported on a different timescale in the TAS data (Figures 5C–

5F and S11). There is a slightly faster decay of the PVSK bleach kinetics after its direct

excitation in the PVSK with Y6 film compared with the neat PVSK film (Figure S11A),

supporting the PL result of some charge-transfer across the interface with Y6 in Fig-

ure 3C. However, the TAS data show no bleach of the PVSK band edge after excita-

tion of the Y6 component (Figure 5D), and the kinetics of the Y6 bleach in neat versus

PVSK with Y6 films is identical when it is directly excited at 890 nm (Figure S11B),

meaning electron transfer into Y6 may be slightly more favorable than splitting Y6

excitons at the PVSK interface. In contrast, the PM6:Y6 BHJ show signatures of

polaron bands at 780 nm (black arrow in Figure 5E and red arrow in Figure 5F)

regardless of excitation of the PM6 or Y6 component. The mirror-image kinetics of

polaron rise versus bleach decay after photoexcitation of each component of the

PM6:Y6 BHJ (Figures S11C–S11D) are a strong indicator of direct and efficient inter-

facial charge transfer. The fact that electron injection from PVSK to Y6 occurs when

the PVSK is excited, but exciton splitting and hole injection from Y6 to PVSK is not

observed when Y6 is excited is a thought-provoking result, especially considering

that Y6 by itself produces as much photocurrent as when traditional BHJ is

used.47,48 As such, we do not know the exact mechanism responsible for photocur-

rent from the Y6 layer, but rather conclude from the SSMC and TAS data that charge

generation at the PVSK with Y6 interface is not the same as at the PM6:Y6 BHJ inter-

face. One explanation could be that the built-in potential of the device, where metal

contacts are present, provides sufficient additional driving force to split the exciton,

but certainly more studies are warranted to further understand the origin of, and

possibly improve, photocurrent due to absorption exclusively by the MFIL, which

will also provide guidance on the PVSK with BHJ research.

Device Stability and Ion Migration Suppression

Stability is currently the biggest challenge for PSCs in practical application. Among

the severe work environment (heating, light, mechanical stress, etc.), humidity has

been shown to be a major stressor causing perovskite-based device degradation.

1584 Joule 4, 1575–1593, July 15, 2020

Figure 5. The Charge Carrier Dynamics of PVSK with Y6 Interface

(A) SSMC spectra and UV-vis absorption spectra of the PM6:Y6 BHJ film and the PVSK with Y6 film

deposited on quartz substrates.

(B) UV-vis absorption spectra of the samples for transient absorption (TA) measurements.

(C) Photoinduced absorption spectra collected at delay times shown in the legend for the PVSK

with Y6 sample photoexcited at 550 nm (2.5 nJ pulse energy).

(D) 890 nm (10 nJ pulse energy). Red and black arrows indicate positions chosen to monitor Y6

bleach and polaron absorption, respectively.

(E and F) (E) Photoinduced absorption spectra collected at delay times shown in the legend for the

PM6:Y6 BHJ sample photoexcited at 550 nm (2.5 nJ pulse energy) and (F) 890 nm (10 nJ pulse

energy). Red and black arrows indicate positions chosen to monitor PM6 bleach and polaron

absorption, respectively.

llArticle

Integrating a hydrophobic barrier layer has been shown to decrease moisture corro-

sion. However, previous reports have mostly focused on a post modification either

on the perovskite surface14 or charger transfer layer49 to enhance the hydrophobic-

ity, while a MFIL that works as a complementary light absorber, a charge carrier

transport medium, as well as a hydrophobic moisture barrier has never been re-

ported. The Y6-based device not only shows improved device efficiency, but also ex-

hibits superior device stability after light and moisture exposure as well as thermal

annealing. The water contact angle (shown in Figure 6A) for the PVSK with PCBM

film is 74.0�, while that for the PVSK with Y6 film was 86.7�. The increased contact

Joule 4, 1575–1593, July 15, 2020 1585

Figure 6. The Enahnced Device Stability

(A) The water contact angle of perovskite with different ETL films (left) and the corresponding

photoimages of water testing results (right).

(B and C) (B) ToF-SIMS results of the aged PCBM based and (C) Y6 based whole devices.

(D) The MPP tracking of perovkite devices under AM1.5G 1 sun illumination in ambient air with a

relativite humidity of 60%~65% for ~1,400 min.

(E) The device ambient stability test upon a relativity humidity of 60%~65% at room temperature

under dark for ~1,700 h.

(F) The device thermal stabity test upon 85�C in an inert enviroment (glove box) under dark for 1,000 h.

llArticle

angle suggests an increased hydrophobicity. This was also confirmed by depositing

water droplets directly on the PVSK/ETL films for 5 min, simulating a rainfall atmo-

sphere, as shown in Figure 6A. The PVSK with PCBM film began to degrade imme-

diately and became totally yellow in 1 min, implying that the water quickly pene-

trates into the perovskite through PCBM. In contrast, the PVSK with Y6 film

showed a much higher water tolerance, and only minor spots emerged at the film

edges while most of the perovskite covered with Y6 retained black.

To demonstrate the stability of devices with different ETLs under operational condi-

tions, non-encapsulated solar cells were aged under AM 1.5G 100 mW/cm�2 solar

irradiation, ambient stability with relativity humidity of 60%–65% (Figure 6D). The

Y6-based devices maintained over 95% of their initial efficiencies (T95 lifetime) after

1,300 min of accelerated aging in ambient environment, whereas the devices with

the PCBM layer degraded to 95% in 60 min, and to 90% within 150 min. The T95 life-

time of the Y6 device is 20 times longer than the PCBM-based device. In addition,

the Y6 device showed substantial long-term stability uponmoisture (the relativity hu-

midity of 60%–65%), and retained 72% of its initial PCE after 1,700 h (�70 days)

without encapsulation, while the PCBM devices almost fully degraded (Figure 6E).

The long-term stability of characteristic photovoltaic performance parameters is

shown in Figure S12, and all the parameters retained�90% of the initial value, espe-

cially for the Jsc, which only suffered 5% loss in 2 months. Furthermore, the Y6 device

also showed superior thermal stability. Upon aging at 85�C in a N2 atmosphere, 95%

of the initial efficiency was retained after 1,000 h (Figures 6F and S13), while the

PCBM devices decreased to 75% of output after 1,000 h of continuous thermal an-

nealing. These results demonstrate superior moisture, light, and thermal stability of

Y6-based devices in comparison to the PCBM device. These are among the top sta-

bility results reported for PSCs with inverted structures.

Time-of-flight secondary ion mass spectrometry (ToF-SIMS) was used to probe the

ion distribution as a function of depth. The ToF-SIMS profiles of selected different

1586 Joule 4, 1575–1593, July 15, 2020

llArticle

elements are shown in Figures 6B and 6C. Strikingly, a I� peak signal is seen in the

PCBM/Ag interface (Figure 6B), whereas no I� accumulation is found in the Y6/Ag

interface (Figure 6C), confirming the I� migration into the PCBM during operational

aging (MPP tracking with light illumination and humidity of 60%–65%). The I�migra-

tion is a recognized phenomenon in lead halide perovskites due to the low activation

energies required for formation and/or transport of the ion migration.50 As there are

high densities of ionic defects in polycrystalline perovskites, the migration and accu-

mulation of I� become more severe under operational conditions. These results in

Figure 6B are also consistent with previous reports, where ion diffusion from the

perovskite to the silver electrode was directly observed by ToF-SIMS with different

aging protocols, and the generation of an AgI barrier accelerated the device degra-

dation.51 In addition to the I� migration, there is obvious O� signal in the Ag layer

and an enhanced O� intensity in the PCBM layer (Figure S14), reflecting the oxida-

tion of the Ag electrode and possible moisture invasion throughout the devices.

These results unveil that the Y6 layer can efficiently reduce the trap states in perov-

skite, and suppress I� migration and moisture infiltration, preventing the perovskite

degradation and boosting the long-term device stability.

Conclusion

In summary, we have demonstrated a new strategy to utilize a multifunctional inter-

face layer to simultaneously enhance the device efficiency and operational stability

in p-i-n PSCs. A series of narrow-band-gap NFAs were studied to further explore the

criteria for screening the MFIL materials. A molecular narrow-band-gap organic

semiconductor, Y6, was selected as a MFIL here, instead of the traditional PCBM

ETL in inverted PSCs, which provides extended light absorption, moisture barrier,

and enhanced electron collection, leading to an increased PCE of 21.0% and

enhanced operational stability upon exposure to light, heat, and moisture as long

as �1,700 h. We showed that the solvents for processing Y6 molecules have signif-

icant impact on the molecular orientation and crystallization of Y6. Compared with

Y6-CB and Y6-Tol, the Y6-CF with unique 2D packing and p-p stacking predominat-

ing the charge carrier extraction and transport, achieved the highest photocurrent

enhancement and champion device efficiency. Meanwhile, the charge carrier dy-

namics at NIR was preliminary explored by SSMC and TAS, and the built-in potential

of the device was proposed to provide sufficient additional driving force to split the

excitons produced in Y6. The reduced trap density states contributed to the

decreased non-radiative recombination and faster charge transfer in Y6-based de-

vices in comparison with PCBM-based devices, leading to the increased PCE from

� 19% to � 21%. Detailed molecular bindings at the critical PVSK with Y6 interface

were further analyzed by XPS and TFY measurements, which could explain the

reduced trap states and suppressed ion migration and moisture infiltration, evi-

denced by Tof-SIMS measurements. Correlations between molecular orientation,

interface molecular bonding, trap state density, non-radiation recombination, and

device performance were established. These results present new insights for the

design of novel interface layers, open a new window to simplify the device structure

with MFIL in PSCs, and will also significantly impact the use of perovskite-based op-

toelectronics for future applications.

EXPERIMENTAL PROCEDURES

Resource Availability

Lead Contact

Further information and requests for resources and materials should be directed

to and will be fulfilled by the Lead Contact, Thomas P. Russell (tom.p.russell@

gmail.com)

Joule 4, 1575–1593, July 15, 2020 1587

llArticle

Materials Availability

This study did not generate new unique materials.

Data and Code Availability

This study did not generate any unique datasets or code

Materials

Anhydrous solvents (DMF, DMSO, isopropanol [IPA], CB, CF, and 1-chloronaphtha-

lene [1-CN]) were purchased from Sigma-Aldrich and used as received. Lead (II) io-

dide (PbI2), lead (II) bromide (PbBr2), cesium iodide (CsI), and buckminsterfullerene

(C60) were also obtained from Sigma-Adrich and used as received. Methylammo-

nium bromide (MABr) and formamidinium iodide (FAI) were purchased from Great-

Cell Solar Ltd. (Australia). PC61BM was purchased fromMerck Company, and Y6 was

ordered from WEIZU Chemical Company (Shanghai, China).

Device Fabrication

Inverted PSCs were fabricated with a configuration of ITO/PTAA/perovskite/ETL/

C60/BCP/Ag. ITO glass was cleaned by sequentially washing with detergent, deion-

ized water, acetone, and IPA. The substrates were dried with N2 and cleaned by UV

ozone for 20 min. PTAA HTLs were prepared by spin coating the Tol-solution of

PTAA (1.5 mg/mL) at 6,000 rpm for 30 s and annealing at 100�C for 10 min. The

CsFAMA mixed perovskite films were fabricated according to our previously re-

ported one-step antisolvent method.52 Briefly, precursor solution was made by mix-

ing PbI2, PbBr2, FAI, and MAI in DMF/DMSO (v/v 4/1). The mole concentration was

kept at 1.3 M with 0.1 M PbI2 excess. The I/Br and FA/MA mole ratios were main-

tained at 0.85/0.15. Additional 35 mL of CsI (2 M in DMSO) was added into the pre-

cursor solution. The solution was stirred 2 h at 65�C. Perovskite films were prepared

by spin coating the precursor solution (5,000/2,000 rpm, 35 s). In the last 25 s of the

procedure, films were quickly treated with 300 mL CB and annealed at 100�C for

60 min. After the perovskite growth, for the control device, PCBM (2 wt % in CB) films

were spin-coated with 1,000 rpm 30 s, followed by annealing at 100�C for 30min. For

Y6 devices, Y6 was dissolved in CF, CB, and Tol, respectively, with the addition of

0.5% 1-CN as an additive, and stirred on a hotplate in a nitrogen-filled glove box

for 2 h. The blend solution was spin-cast on the top of the perovskite layer

(3,000 rpm, 30 s), and the film was thermally annealed at 80�C for 5 min to optimize

the blend morphology. For Y6/C60 devices, C60 (20 nm) layers were thermal evap-

orated on top of Y6. After that, BCP (0.5 wt % in IPA) was deposited as cathode buffer

layer at 4,000 rpm spinning speed. Finally, an Ag back electrode was deposited by

thermal evaporation at high vacuum. The device active area was 10 mm2 and all de-

vices were measured with the mask area of 8.0 mm2.

Film and Device Characterizations

J-V measurements were carried out using a Keithley 2400 source meter in ambient

environment at �25�C and �60%–65% relative humidity (RH). The devices were

measured both in reverse scan (1.2 /�0.2, step 0.01 V) and forward

scan (�0.2 /1.2, step 0.01 V) with 10 ms delay time. Illumination was provided

by an Oriel Sol3A solar simulator with AM1.5G spectrum and light intensity of

100 mW/cm2 calibrated by a standard KG-5 Si diode. The aperture area of our de-

vice is 8.0 mm2, calibrated with a shadow mask during the measurements. The

UV-vis-NIR spectra were recorded by a Cary 5000 UV-Vis-NIR spectrometer (Agi-

lent). EQE measurements for devices were conducted with an Enli-Tech (Taiwan)

EQE measurement system. A FEI Helios Nanolab 600i dual beam focus ion beam

and field emission gun-scanning electron microscope (FIB/FEGSEM) was used to

1588 Joule 4, 1575–1593, July 15, 2020

llArticle

prepare cross-section for STEM imaging and analysis. FEI Talos transmission elec-

tron microscope (TEM) with Super-X EDX was employed to acquire the STEM image

with high-angle annular dark field (STEM-HAADF) mode. PL and time resolved PL

spectra were carried out by Spectrofluorometer (FS5, Edinburgh instruments).

Xenon lamp and 405 nm pulsed laser were used as excitation sources for the PL

(excited from glass side) and TRPL measurement (excited from Y6 or PCBM side),

respectively. The MS and t-DOS spectra for the devices were measured by a Zahner

IM6e electrochemical station (Zahner, Germany) in ambient environment of 23�Cand 37% RH. KPFMmeasurements were performed in a single-pass frequency-mod-

ulation mode using Cypher ES (Asylum Research) and HF2LI lock-in amplifier (Zurich

Instrument) with PtIr coated tips (PPP-EFM, nanosensors).53 XPS measurement was

conducted on a Thermo Scientific K-alpha XPS apparatus equipped with a mono-

chromatic Al K(alpha) source and food gun for charge compensation. For the XPS

samples of PVSK with Y6 films, diluted Y6 solution (0.5 mg/mL in CF) was spin-cast

on the top of the perovskite layer at 3,000 rpm, 30 s, to control thickness of Y6

film below 10 nm. Film thicknesses were measured with a surface profilometer

(KLA-Tencor D-120).

Grazing Incidence X-Ray Scattering (GIXD Measurements

GIXD measurements were conducted on beamline 7.3.3 at Advanced Light Source

(ALS) at LBNL. The wavelength of the X-ray was 1.240 A, and the scattering intensity

was detected by a PILATUS 2M detector. All the samples were prepared on Si sub-

strates. The 2D GIXD images were circular averaged using Nika software package.

The integrated peak area, the crystal size, and the calculation of peak ratio were per-

formed using IGOR software, a commercially available software product. The single

crystal simulation of Y6 was analyzed by the software of Mercury, developed by the

Cambridge Crystallographic Data Centre.

Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS)

The ToF-SIMS spectra were acquired using a ToF-SIMS V spectrometer from ION-

TOF GmbH, Germany. The analysis was conducted with the Bi3+ beam operating

at 25 keV with the pulse current of 0.3 pA.

Density Functional Theory (DFT) Caculations

The binding energy of molecules and perovskite surface was calculated by

VASP,54,55 which was implemented with Projector Augment Wave (PAW) based on

plane-wave basis. The energy cut off of plane wave basis is 550 eV. The exchange

functional was described by the Perdew-burke-Ernzerh (PBE) of generalized gradient

approximation (GGA). We replaced the all-carbon chains of Y6 molecules with CH4

to reduce the size of model and speed up calculations. The surface model used to

adsorb the Y6 molecule has a surface area of 25.75 3 32.46 A2, which includes

two layers of perovskite with the bottom layer fixed and PbI2 (001) terminated sur-

face, separated by 15 A of vacuum. Due to the large cells of our surface models,

only G point was used to k-point mesh. The Y6 molecule initially lay down on the

perovskite surface. Calculation of perovskite with PC61BM with similar model was

also performed for comparison. Geometry structures of our models are optimized

until the force on every atom is less than 0.05 eV/A. The binding energy was calcu-

lated by Ebinding = Ecombined � Esurface � Emolecule.

Total Fluorescence Yield (TFY) Spectroscopy Measurements

TFY measurments were performed at beamline 11.0.1.2 of ALS, LBNL. Sample prep-

aration was the same as the XPS measurment. The TFY signal was recorded with a

Princeton Instrument PI-MTE CCD detector with a pixel size of 0.027 by 0.027 mm.

Joule 4, 1575–1593, July 15, 2020 1589

llArticle

The Steady-State Microwave Conductivity Spectra (SSMC) Measurement

This technique is new, but builds upon the principles of the more well-established

time-resolved microwave conductivity (TRMC) measurements in which transient sig-

nals are analyzed after laser excitation pulse and provides complimentary informa-

tion to TRMC.56,57 In a SSMC experiment, the excitation is chopped at a lock-in fre-

quency and operates at a photon intensity about a factor of 4 lower than solar flux;

the technique, instrument, and analytical methods are described explicitly for the

first time elsewhere.45 Here, we use the technique to probe the photoconductivity

action spectrum of the perovskite/Y6 composite sample as well as the PM6:Y6

BHJ sample using a sealed N2-filled cavity with a lower K factor than the measure-

ments by Blackburn et al., (�29,000 versus �200,000). All samples were deposited

identical to device films on 11 3 25 mm quartz substrates.

fs-TAS Experimental

The transient absorption data were collected with a Helios Fire spectrometer (Ultra-

fast Systems), using a Pharos amplified laser (Light Conversion) as the excitation

source. The Pharos was operated at 6 kHz and 6 W power, with a small portion

directed to a sapphire crystal to generate the supercontinuum probe (400–

1,650 nm). The pump was converted to 550 or 890 nm with an Orpheus optical para-

metric amplifier and attenuated with neutral density filters before being directed to

the sample with a spot size of approximately 0.5 mm. The probe was overlapped

with the pump at the sample position and detected with either a Si or InGaAs photo-

diode array, and photoinduced transmission changes were determined by modu-

lating the pump at 3 kHz and subtracting ‘‘pump off’’ from ‘‘pump on’’ probe spectra.

The data were corrected for chirp and background scatter using Surface Explorer. All

the film samples are prepared on glass substrates.

Absorption Spectroscopy

The absorption of films for TRMC and TAS was measured using a Cary 7000 optical

spectrophotometer with the sample mounted inside an integrating sphere DRA

module. After baselining with the bare substrate, the absorption spectra were cor-

rected for both specular and diffuse reflectance during the transmission

measurement.

Device Stability Test

TheMPP tracking test was used to evaluate the device operational stability. The non-

encapsulated solar cells were fixed at the Vmpp and the current density variation un-

der ambient environment (�25�C, �60%–65% RH) without temperature control re-

cord. For thermal stability and ambient long-term stability tests, the devices were

stored in an inert environment (O2 < 0.1 ppm and H2O < 0.1 ppm at 85�C and

dark) and ambient environment (�25�C, dark condition, �60%–65% RH), respec-

tively. All devices are measured with a mask area of 8.0 mm2.

SUPPLEMENTAL INFORMATION

Supplemental Information can be found online at https://doi.org/10.1016/j.joule.

2020.06.007.

ACKNOWLEDGMENTS

Q.H., W.C., and W.Y. contributed equally to this work. Q.H. and T.P.R. were sup-

ported by the US Office of Naval Research under contract N00014-17-1-2241. This

work was also financially supported by the National Natural Science Foundation of

China (61775091, 51573076, 61722501, and 61805138). F.L. was supported by

1590 Joule 4, 1575–1593, July 15, 2020

llArticle

Beijing National Laboratory for Molecular Sciences (BNLMS201902). W.C. and

A.B.D. are grateful for support from the Seed Funding for Strategic Interdisciplinary

Research Scheme of the University of Hong Kong and RGCCRF grant 5037/18G. The

authors thank Dr. Tao Liu and Prof. He Yan for the material of IOIC-2Cl and IEICO-4F;

Dr. Xinle Li and Dr. Chongqing Yang for the discussion of XPS results; Prof. Benzhong

Tang, Prof. Lu-tao Weng, and Zaiyu Wang for the ToF-SIMS analysis; and Dr. Wanli

Yang for the discussions of TFY studies. TFY and GIXD were performed at beamline

11.0.1.2 and 7.3.3 at Advanced Light Source, Lawrence Berkeley National Labora-

tory, which was supported by US Department of Energy, Office of Science, and

Office of Basic Energy Sciences. We thank the support for sample preparation and

device fabrication at Molecular Foundry, LBNL. Work at Molecular Foundry was sup-

ported by the Office of Science, Office of Basic Energy Sciences, the US Department

of Energy under contract no. DE-AC02-05CH11231. SSMC and TAS work were sup-

ported by the US Department of Energy, Office of Basic Energy Sciences, Division of

Chemical Sciences, Biosciences, and Geosciences under contract no. DE-AC36-

08GO328308 with the National Renewable Energy Laboratory.

AUTHOR CONTRIBUTION

Q.H., T.P.R., and F.L. conceived the idea; T.P.R., F.L., R.Z., and Z.H. supervised the

project; Q.H., W.C., and W.Y. fabricated and optimized the PSCs as well as the

related device characterizations, including the UV-vis absorption, PL and TRPL,

t-DOS, EQE, STEM and SEM, and so on; Q.H. and Y.L. performed GIWAXS; W.Z.,

Q.H., and C.W. conducted TFY measurements; Y.Z. and J.X. performed the theoret-

ical calculation; Y.-H. L. and M.S. conducted the KPFMmeasurements; Q.H. and L.K.

conducted the XPS experiment; B.W. L., J.C.J., and Q.H. performed the SSMC and

TAS measurements; Q.H. wrote the first draft of the paper; W.C., W.Y, Y.L., Y.Z.,

B.W.L., and J.C.J. revised the paper; all the authors discussed the results and edited

the manuscript.

DECLARATION OF INTERESTS

The authors declare no competing interests.

Received: January 16, 2020

Revised: April 10, 2020

Accepted: June 4, 2020

Published: July 1, 2020

REFERENCES

1. Li, Z., Klein, T.R., Kim, D.H., Yang, M., Berry,J.J., van Hest, M.F.A.M., and Zhu, K. (2018).Scalable fabrication of perovskite solar cells.Nat. Rev. Mater. 3, 18017.

2. Wang, R., Xue, J., Wang, K.L., Wang, Z.K., Luo,Y., Fenning, D., Xu, G., Nuryyeva, S., Huang, T.,Zhao, Y., et al. (2019). Constructive molecularconfigurations for surface-defect passivation ofperovskite photovoltaics. Science 366, 1509–1513.

3. Zheng, Y., Su, R., Xu, Z., Luo, D., Dong, H., Jiao,B., Wu, Z., Gong, Q., and Zhu, R. (2019).Perovskite solar cell towards lower toxicity: Atheoretical study of physical lead reductionstrategy. Science Bulletin 64, 1255–1261.

4. NREL Best research-cell efficiency chart.https://www.nrel.gov/pv/cell-efficiency.html,Accessed June 2020.

5. Jena, A.K., Kulkarni, A., andMiyasaka, T. (2019).Halide perovskite photovoltaics: background,status, and future prospects. Chem. Rev. 119,3036–3103.

6. Hu, Q., Zhao, L., Wu, J., Gao, K., Luo, D., Jiang,Y., Zhang, Z., Zhu, C., Schaible, E., Hexemer, A.,et al. (2017). In situ dynamic observations ofperovskite crystallisation and microstructureevolution intermediated from [PbI6]4� cagenanoparticles. Nat. Commun. 8, 15688.

7. Wang, K., Subhani, W.S., Wang, Y., Zuo, X.,Wang, H., Duan, L., and Liu, S.F. (2019). Metalcations in efficient perovskite solar cells:progress and perspective. Adv. Mater. 31,e1902037.

8. Deng, W., Liang, X., Kubiak, P.S., andCameron, P.J. (2018). Molecular interlayers in

hybrid perovskite solar cells. Adv. EnergyMater. 8, 1701544.

9. Wang, L., Zhou, H., Hu, J., Huang, B., Sun, M.,Dong, B., Zheng, G., Huang, Y., Chen, Y., Li, L.,et al. (2019). A Eu3+-Eu2+ ion redox shuttleimparts operational durability to Pb-Iperovskite solar cells. Science 363, 265–270.

10. Tu, Y., Xu, G., Yang, X., Zhang, Y., Li, Z., Su, R.,Luo, D., Yang, W., Miao, Y., Cai, R., et al. (2019).Mixed-cation perovskite solar cells in space.Sci. China Phys. Mech. Astron. 62, 974221.

11. Meng, L., You, J., and Yang, Y. (2018).Addressing the stability issue of perovskitesolar cells for commercial applications. Nat.Commun. 9, 5265.

12. Bai, S., Da, P., Li, C., Wang, Z., Yuan, Z., Fu, F.,Kawecki, M., Liu, X., Sakai, N., Wang, J.T.-W.,

Joule 4, 1575–1593, July 15, 2020 1591

llArticle

et al. (2019). Planar perovskite solar cells withlong-term stability using ionic liquid additives.Nature 571, 245–250.

13. Jung, E.H., Jeon, N.J., Park, E.Y., Moon, C.S.,Shin, T.J., Yang, T.Y., Noh, J.H., and Seo, J.(2019). Efficient, stable and scalable perovskitesolar cells using poly(3-hexylthiophene).Nature 567, 511–515.

14. Yang, S., Chen, S., Mosconi, E., Fang, Y., Xiao,X., Wang, C., Zhou, Y., Yu, Z., Zhao, J., Gao, Y.,et al. (2019). Stabilizing halide perovskitesurfaces for solar cell operation with wide-bandgap lead oxysalts. Science 365, 473–478.

15. Luo, D., Yang, W., Wang, Z., Sadhanala, A., Hu,Q., Su, R., Shivanna, R., Trindade, G.F., Watts,J.F., Xu, Z., et al. (2018). Enhancedphotovoltage for inverted planarheterojunction perovskite solar cells. Science360, 1442–1446.

16. Liu, T., Chen, K., Hu, Q., Zhu, R., and Gong, Q.(2016). Inverted perovskite solar cells:progresses and perspectives. Adv. EnergyMater. 6, 1600457.

17. Chen, H., Wei, Q., Saidaminov, M.I., Wang,F., Johnston, A., Hou, Y., Peng, Z., Xu, K.,Zhou, W., Liu, Z., et al. (2019). Efficient andstable inverted perovskite solar cellsincorporating secondary amines. Adv. Mater.31, e1903559.

18. Tong, J., Song, Z., Kim, D.H., Chen, X., Chen,C., Palmstrom, A.F., Ndione, P.F., Reese, M.O.,Dunfield, S.P., Reid, O.G., et al. (2019). Carrierlifetimes of >1 ms in Sn-Pb perovskites enableefficient all-perovskite tandem solar cells.Science 364, 475–479.

19. Jo�st, M., Kohnen, E., Morales-Vilches, A.B.,Lipov�sek, B., Jager, K., Macco, B., Al-Ashouri,A., Kr�c, J., Korte, L., Rech, B., et al. (2018).Textured interfaces in monolithic perovskite/silicon tandem solar cells: advanced lightmanagement for improved efficiency andenergy yield. Energy Environ. Sci. 11, 3511–3523.

20. Yang, D., Yang, R., Priya, S., and Liu, S.F. (2019).Recent advances in flexible perovskite solarcells: fabrication and applications. Angew.Chem. Int. Ed. Engl. 58, 4466–4483.

21. Chen,W., Sun, H., Hu, Q., Djuri�si�c, A.B., Russell,T.P., Guo, X., and He, Z. (2019). High short-circuit current density via integrating theperovskite and ternary organic bulkheterojunction. ACS Energy Lett 4, 2535–2536.

22. Luo, D., Su, R., Zhang, W., Gong, Q., and Zhu,R. (2020). Minimizing non-radiativerecombination losses in perovskite solar cells.Nat. Rev. Mater. 5, 44–60.

23. Akbulatov, A.F., Frolova, L.A., Griffin, M.P.,Gearba, I.R., Dolocan, A., Vanden Bout, D.A.,Tsarev, S., Katz, E.A., Shestakov, A.F.,Stevenson, K.J., and Trosin, P.A. (2017). Effectof electron-transport material on light-induceddegradation of inverted planar junctionperovskite solar cells. Adv. Energy Mater. 7,1700476.

24. Yuan, J., Zhang, Y., Zhou, L., Zhang, G., Yip,H.-L., Lau, T.-K., Lu, X., Zhu, C., Peng, H.,Johnson, P.A., et al. (2019). Single-junctionorganic solar cell with over 15% efficiency usingfused-ring acceptor with electron-deficientcore. Joule 3, 1140–1151.

1592 Joule 4, 1575–1593, July 15, 2020

25. Yao, H., Cui, Y., Yu, R., Gao, B., Zhang, H., andHou, J. (2017). Design, synthesis, andphotovoltaic characterization of a smallmolecular acceptor with an ultra-narrow bandgap. Angew. Chem. Int. Ed. Engl. 56, 3045–3049.

26. Yan, C., Liu, T., Chen, Y., Ma, R., Tang, H., Li, G.,Li, T., Xiao, Y., Yang, T., Lu, X., et al. (2020). ITC-2Cl: a versatile middle-bandgap nonfullereneacceptor for high-efficiency panchromaticternary organic solar cells. Sol. RRL 4, 1900377.

27. Chen, Y., Liu, T., Hu, H., Ma, T., Lai, J.Y.L.,Zhang, J., Ade, H., and Yan, H. (2018).Modulation of end groups for low-bandgapnonfullerene acceptors enabling high-performance organic solar cells. Adv. EnergyMater. 8, 1801203.

28. Liu, T., Luo, Z., Chen, Y., Yang, T., Xiao, Y.,Zhang, G., Ma, R., Lu, X., Zhan, C., Zhang, M.,et al. (2019). A nonfullerene acceptor with a1000 nm absorption edge enables ternaryorganic solar cells with improved optical andmorphological properties and efficiencies over15%. Energy Environ. Sci. 12, 2529–2536.

29. Lilliu, S., Agostinelli, T., Pires, E., Hampton, M.,Nelson, J., and Macdonald, J.E. (2011).Dynamics of crystallization and disorder duringannealing of P3HT/PCBM bulkheterojunctions. Macromolecules 44, 2725–2734.

30. Zhu, L., Zhang, M., Zhou, G., Hao, T., Xu, J.,Wang, J., Qiu, C., Prine, N., Ali, J., Feng, W.,et al. (2020). Efficient organic solar cell with16.88% efficiency enabled by refined acceptorcrystallization and morphology with improvedcharge transfer and transport properties. Adv.Energy Mater. 10, 1904234.

31. Strawhecker, K.E., Kumar, S.K., Douglas, J.F.,and Karim, A. (2001). The critical role of solventevaporation on the roughness of spin-castpolymer films. Macromolecules 34, 4669–4672.

32. Golubev, T., Liu, D., Lunt, R., and Duxbury, P.(2019). Understanding the impact of C60 at theinterface of perovskite solar cells via drift-diffusion modeling. AIP Adv 9, 035026.

33. Zheng, X., Chen, B., Dai, J., Fang, Y., Bai, Y.,Lin, Y., Wei, H., Zeng, X.C., and Huang, J.(2017). Defect passivation in hybridperovskite solar cells using quaternaryammonium halide anions and cations. Nat.Energy 2, 17102.

34. Landi, G., Neitzert, H.C., Barone, C., Mauro, C.,Lang, F., Albrecht, S., Rech, B., and Pagano, S.(2017). Correlation between electronic defectstates distribution and device performance ofperovskite solar cells. Adv Sci (Weinh) 4,1700183.

35. Walter, T., Herberholz, R., Muller, C., andSchock, H.W. (1996). Determination of defectdistributions from admittance measurementsand application to Cu(In,Ga)Se2 basedheterojunctions. J. Appl. Phys. 80, 4411–4420.

36. Aydin, E., De Bastiani, M., and De Wolf, S.(2019). Defect and contact passivation forperovskite solar cells. Adv. Mater. 31,e1900428.

37. Chun-Ren Ke, J., Walton, A.S., Lewis, D.J.,Tedstone, A., O’Brien, P., Thomas, A.G., andFlavell, W.R. (2017). In situ investigation ofdegradation at organometal halide perovskite

surfaces by X-ray photoelectron spectroscopyat realistic water vapour pressure. Chem.Commun. (Camb.) 53, 5231–5234.

38. Lin, F., Liu, Y.J., Yu, X.Q., Cheng, L., Singer, A.,Shpyrko, O.G., Xin, H.L.L., Tamura, N., Tian,C.X., Weng, T.-C., et al. (2017). Synchrotron X-ray analytical techniques for studying materialselectrochemistry in rechargeable batteries.Chem. Rev. 117, 13123–13186.

39. Naumkin, A.V., Kraut-Vass Gaarenstroom A,S.W., and Powell, C.J. (2012). NIST X-rayphotoelectron spectroscopy database, version4.1. http://srdata.nist.gov/xps/.

40. Rondon, S., and Sherwood, P.M.A. (1998). Corelevel and valence band spectra of PbO by XPS.Surf. Sci. Spectra 5, 97–103.

41. Foreman, K., Echeverria, E., Koten, M.A.,Lindsay, R.M., Hong, N., Shield, J., andAdenwalla, S. (2016). The metal/organicinterface in cobalt/vinylidene fluorideheterostructures. Mater. Res. Express 3, 116403.

42. Lin, Y., Shen, L., Dai, J., Deng, Y., Wu, Y., Bai, Y.,Zheng, X., Wang, J., Fang, Y., Wei, H., et al.(2017). p-conjugated lewis base: efficient trap-passivation and charge-extraction for hybridperovskite solar cells. Adv. Mater. 29, 1604545.

43. Butler, K.T., Sai Gautam, G., and Canepa, P.(2019). Designing interfaces in energymaterialsapplications with first-principles calculations.npj Comput. Mater. 5, 19.

44. Wang, J., Zhang, J., Zhou, Y., Liu, H., Xue, Q., Li,X., Chueh, C.C., Yip, H.L., Zhu, Z., and Jen,A.K.Y. (2020). Highly efficient all-inorganicperovskite solar cells with suppressed non-radiative recombination by a lewis base. Nat.Commun. 11, 177.

45. Blackburn, J.L., Zhang, H., Myers, A.R., Dunklin,J.R., Coffey, D.C., Hirsch, R.N., Vigil-Fowler, D.,Yun, S.J., Cho, B.W., Lee, Y.H., et al. (2020).Measuring photoexcited free charge carriers inmono- to few-layer transition-metaldichalcogenides with steady-state microwaveconductivity. J. Phys. Chem. Lett. 11, 99–107.

46. Labram, J.G., Perry, E.E., Venkatesan, N.R., andChabinyc, M.L. (2018). Steady-state microwaveconductivity reveals mobility-lifetime productin methylammonium lead iodide. Appl. Phys.Lett. 113, 153902.

47. Guo, Q., Bai, Y., Lang, K., Yu, Z.Z., Hayat, T.,Alsaedi, A., Zhou, E., and Tan, Z.A. (2019).Expanding the light harvesting of CsPbI2Br tonear infrared by integrating with organic bulkheterojunction for efficient and stable solarcells. ACS Appl. Mater. Inter. 11, 37991–37998.

48. Daboczi, M., Kim, J., Lee, J., Kang, H.,Hamilton, I., Lin, C.-T., Dimitrov, S.D.,McLachlan, M.A., Lee, K., Durrant, J.R., andKim, J.-S. (2020). Towards efficient integratedperovskite/organic bulk heterojunction solarcells: interfacial energetic requirement toreduce charge carrier recombination losses.Adv. Funct. Mater. https://doi.org/10.1002/adfm.202001482.

49. Wu, C., Wang, K., Feng, X., Jiang, Y., Yang, D.,Hou, Y., Yan, Y., Sanghadasa, M., and Priya, S.(2019). Ultrahigh durability perovskite solarcells. Nano Lett 19, 1251–1259.

50. Lee, J.-W., Kim, S.-G., Yang, J.-M., Yang, Y.,and Park, N.-G. (2019). Verification and

llArticle

mitigation of ion migration in perovskite solarcells. APL Mater 7, 041111.

51. Li, J., Dong, Q., Li, N., and Wang, L. (2017).Direct evidence of ion diffusion for the silver-electrode-induced thermal degradation ofinverted perovskite solar cells. Adv. EnergyMater. 7, 1602922.

52. Chen, W., Wu, Y., Fan, J., Djuri�si�c, A.B., Liu, F.,Tam, H.W., Ng, A., Surya, C., Chan, W.K.,Wang, D., and He, Z.-B. (2018). Understandingthe doping effect on NiO: Toward high-performance inverted perovskite solar cells.Adv. Energy Mater. 8, 1703519.

53. Zhang, Y.J., Ziegler, D., and Salmeron, M.(2013). Charge trapping states at theSiO2-oligothiophene monolayer interface infield effect transistors studied by kelvin probeforce microscopy. ACS Nano 7, 8258–8265.

54. Kresse, G., and Furthmuller, J. (1996).Efficiency of ab-initio total energy calculationsfor metals and semiconductors using a plane-wave basis set. Computational MaterialsScience 6, 15–50.

55. Kresse, G., and Furthmuller, J. (1996). Efficientiterative schemes for ab initio total-energycalculations using a plane-wave basis set.

Phys. Rev. B Condens. Matter 54, 11169–11186.

56. Reid, O.G., Moore, D.T., Li, Z., Zhao, D., Yan, Y.,Zhu, K., and Rumbles, G. (2017). Quantitativeanalysis of time-resolved microwaveconductivity data. J. Phys. D: Appl. Phys. 50,493002.

57. Savenije, T.J., Ferguson, A.J., Kopidakis, N.,and Rumbles, G. (2013). Revealing thedynamics of charge carriers inpolymer:fullerene blends using photoinducedtime-resolved microwave conductivity. J. Phys.Chem. C 117, 24085–24103.

Joule 4, 1575–1593, July 15, 2020 1593

JOUL, Volume 4

Supplemental Information

Improving Efficiency and Stability

of Perovskite Solar Cells Enabled

by A Near-Infrared-Absorbing Moisture Barrier

Qin Hu, Wei Chen, Wenqiang Yang, Yu Li, Yecheng Zhou, Bryon W. Larson, Justin C.Johnson, Yi-Hsien Lu, Wenkai Zhong, Jinqiu Xu, Liana Klivansky, Cheng Wang, MiquelSalmeron, Aleksandra B. Djuri�si�c, Feng Liu, Zhubing He, Rui Zhu, and Thomas P. Russell

Figure S1. a) The cross-sectional scanning transmission electron microscopy (STEM) image of inverted perovskite solar cells studies in this work. b) The SEM image of perovskite thin film.

Figure S2. The simulation of single-crystal structure of Y6.

Figure S3. The summary of d-spacing, peak area and the crystalline coherence length (CCL) of a) π-π stacking, b) (110) facet, and c) (11-1) facet scatterings at the in-plane direction.

Figure S4. The topography of Y6 films processed with different solvents. The root mean square (rms) roughness are 5.49 nm (CF), 5.95 nm (Tol) and 8.28 nm (CB). All the scale bars are 500 nm.

Figure S5. The UV-vis spectra of Y6 films processed with different solvents.

Figure S6. The EQE spectra of devices based on Y6 films processed with different solvents.

Figure S7. The film thickness optimization of Y6-CF film for perovskite solar cells.

Figure S8. The stabilized power output (SPO) at maximum power points (MPP) of devices with different electron transport layers.

Figure S9. The XPS analysis of the perovskite/Y6 interface.

Figure S10. Density of states of a PCBM a) and Y6 b) adsorbed on a perovskite surface and corresponding orbitals.

Figure S11. Kinetic profiles corresponding to red and black arrows at spectral positions indicated in Figure 5 for PVSK/Y6 sample photoexcited at a) 550 nm and b) 890 nm, and for PM6/Y6 BHJ sample photoexcited at c) 550 nm and d) 890 nm.

Figure S12. The stability of devices based on different ETLs in ambient atmosphere with the humidity of 60%-65%.

Figure S13. The thermal stability of devices with different ETLs upon 85 °C aging condition.

Figure S14. a) ToF-SIMS results of aged a) PCBM and b) Y6 based whole devices.

Table S1. Summaries of champion device performance based different electron transport layer.

Device Voc (V) Jsc (mA/cm2) FF PCE (%) RS (W) RSh (kW) Y6-CF 1.11 23.7 0.78 20.5 2.9 2.31 Y6-Tol 1.10 23.3 0.70 17.9 7.3 0.84 Y6-CB 1.11 23.4 0.68 17.7 7.2 0.92

Y6-CF/C60 1.12 23.7 0.79 21.0 2.3 5.05 PCBM/C60 1.10 22.6 0.77 19.2 2.5 4.21

Table S2. Thickness summary of the Y6 films prepared from the CF solution with different concentrations.

Concentration (mg/mL) 8 9 9.5 10 10.5 Average thickness (nm) 39.4 43.2 46.3 50.7 56.2

Table S3. The reproductivity of device performance based on different electron transport layer. The standard deviations were calculated with 20 devices.

Device Voc (mV) Jsc (mA/cm2) FF (%) PCE (%) Y6-CF 1091±10.1 23.6±0.12 74.9±0.77 19.3±0.32 Y6-Tol 1088±8.2 23.3±0.09 67.6±1.22 17.1±0.39 Y6-CB 1092±7.8 23.4±0.11 66.5±0.97 17.0±0.35

Y6-CF/C60 1110±10 23.52±0.081 77.4±1.3 20.2±0.32 PCBM/C60 1092±7.9 22.59±0.095 76.2±1.2 18.7±0.46

Table S4. Summaries of the decay lifetime for perovskite films with different ETLs.

Samples A1 (%) t1 [ns] A2 (%) t2 [ns] Weighted average t2 [ns] Glass/PVSK 32.9 66.6 67.1 629.8 602.0 PVSK/PCBM 64.9 1.9 35.1 33.1 30.1

PVSK/Y6 66.2 4.4 33.8 19.0 14.5 A τ"#$ refers to weighted average lifetime which is calculated as following: 𝜏&'( =

*+,+-.*-,--

*+,+.*-,- .