https://doi.org/10.

1038/s41586-024-07189-3

Accelerated Article Preview

W
Double-side 2-dimensional/3-dimensional
E
I
heterojunctions for inverted perovskite
solar cells
EV
PR
Received: 22 May 2023 Randi Azmi, Drajad Satrio Utomo, Badri V­is­ha­l­, S­hy­ng­gy­s Zhumagali, Pia D­­a­l­­ly­­­­,

E
A­­n­d­­i Muhammad Risqi, Adi Prasetio, Esma Ugur, Fangfang Cao, Imil Fadli Imran,
Accepted: 9 February 2024
Ahmed Ali Said, Anil Reddy P­in­in­ti­, A­na­nd Selvin Subbiah, Erkan Aydin, Chuanxiao Xiao,

L
Accelerated Article Preview Sang Il Seok & Stefaan De Wolf
Published online xx xx xxxx

C
Cite this article as: Azmi, R. et al.

I
This is a PDF file of a peer-reviewed paper that has been accepted for publication.
Double-side 2-dimensional/3-dimensional
Although unedited, the content has been subjected to preliminary formatting. Nature

T
heterojunctions for inverted perovskite
solar cells. Nature https://doi.org/10.1038/ is providing this early version of the typeset paper as a service to our authors and

R
s41586-024-07189-3 (2024) readers. The text and figures will undergo copyediting and a proof review before the
paper is published in its final form. Please note that during the production process

A
errors may be discovered which could affect the content, and all legal disclaimers
apply.

ED
AT
R
E
EL
C
C
A

Nature | www.nature.com
 1 Double-side 2-dimensional/3-dimensional heterojunctions for inverted
2 perovskite solar cells
Randi Azmi,1* Drajad Satrio Utomo,1 Badri Vishal,1 Shynggys Zhumagali,1 Pia Dally,1 Andi

W
3
4 Muhammad Risqi,2 Adi Prasetio,1 Esma Ugur,1 Fangfang Cao,3 Imil Fadli Imran,1 Ahmed Ali
5 Said,1 Anil Reddy Pininti,1 Anand Selvin Subbiah,1 Erkan Aydin,1 Chuanxiao Xiao,3,4 Sang Il

IE
6 Seok,2 Stefaan De Wolf1*
7

EV
1
8 King Abdullah University of Science and Technology (KAUST), KAUST Solar Center (KSC),
9 Physical Sciences and Engineering Division (PSE), Material Science and Engineering Program
10 (MSE), Thuwal, 23955-6900, Kingdom of Saudi Arabia

PR
2
11 School of Energy and Chemical Engineering, Ulsan National Institute of Science and Technology
12 (UNIST), 50 UNIST-gil, Eonyang-eup, Ulju-gun, Ulsan, 44919 Republic of Korea
3
13 Ningbo Institute of Materials Technology and Engineering, Chinese Academy of Sciences,

LE
14 Ningbo City, Zhejiang Province, 315201, China
4
15 Ningbo New Materials Testing and Evaluation Center CO., Ltd, Ningbo City, Zhejiang Province,
16 315201, China
17
C
TI
18 *Corresponding authors: [email protected]; [email protected]
19
AR

20 Abstract

21 Defects at the top and bottom interfaces of three-dimensional (3D) perovskite photo-absorbers
D

22 diminish the performance and operational stability of perovskite solar cells (PSCs) due to charge
E

23 recombination, ion migration, and electric-field inhomogeneities.1-5 Here, we demonstrate that

AT

24 long alkyl-amine ligands can generate near-phase pure two-dimensional (2D) perovskites at the
25 top and bottom 3D perovskite interfaces and effectively resolves these issues. At the rear-contact
26 side, we find that the employed alkyl-amine ligand strengthens the interactions with the substrate
ER

27 through acid-base reactions with the phosphonic acid group from the employed organic hole-
28 transporting self-assembled monolayer molecule, thus regulating the 2D perovskite formation.
EL

29 With this, inverted PSCs with double-side 2D/3D heterojunctions achieved a power conversion
30 efficiency (PCE) of 25.6% (certified 25.0%), retaining 95% of their initial PCE after 1000 hours
C

31 of 1-sun illumination at 85 degrees Celsius in air.

AC

1
 32 Main

33 Engineering of perovskite interfaces is essential to enhance the performance and stability of

34 perovskite solar cells (PSCs). Two-/three-dimensional (2D/3D) perovskite heterojunctions have

W
35 shown particular promise towards this end.6-16 Usually, such heterojunctions are formed by post-

IE
36 treating an as-deposited 3D perovskite film with a solution that contains 2D ligands to reconstruct
37 through cation exchange the 3D perovskite surface into a 2D perovskite.7-9,12,14 However, this

EV
38 strategy is only applicable at the top surface of 3D perovskite films. Moreover, it usually results
39 in a 2D perovskite consisting of a mixture of multiple dimensionalities such as n = 1, 2, 3, etc. with

PR
40 random crystal orientations, where n refers to the number of subsequent corner-sharing octahedral
41 [PbI6]4- inorganic slabs.8,12,14 Unfortunately, such mixed dimensionality may induce interfacial
42 energetic inhomogeneity and could impede charge transfer through band misalignment when such

LE
43 a 2D/3D heterojunction is integrated into a charge-selective contact stack.7,8,12,14 In contrast, phase-
44 pure 2D perovskite passivation can substantially reduce the charge-trap density and ion migration,
45
C
leading to significantly improved device performance and stability.10,11 Hence, it is imperative to
TI
46 narrow the distribution of the different n layers in 2D perovskites.10,11,16 Yet, achieving phase-pure
AR

47 2D perovskite passivating contacts with a proper crystal orientation is experimentally

48 challenging.10,11,16

49 Integrating 2D/3D heterojunctions at the buried bottom interface is desirable too, but has rarely
D

50 been reported to date.17-19 The primary difficulty here is to avoid the dissolution of the pre-
E

51 deposited 2D layer or the organic ammonium-based ligands during the subsequent 3D perovskite
AT

52 deposition.17-19 Indeed, most organic ammonium ligands are readily dissolved into the highly polar
53 aprotic solvents typically used in 3D perovskite precursor inks such as dimethyl formamide (DMF)
ER

54 and dimethyl sulfoxide (DMSO).11,20,21 Hence, one approach for minimizing the dissolution of 2D
55 ligands during perovskite solution casting is to strengthen their interaction with the substrate. This
EL

56 allows for immobilizing 2D ligands prior to perovskite deposition, and forming a 2D perovskite
57 underneath the 3D perovskite film through cation exchange and intercalation of the 2D ligands
C

58 into the perovskite film, post deposition.

AC

59 2D/3D heterojunction at the bottom interface

60 We developed inverted (p–i–n) PSCs with a hole-collecting contact (p–type) that is formed by
61 anchoring 2PACz, a self-assembled monolayer (SAM), onto the transparent bottom electrode,

2
 62 consisting of indium-tin oxide (ITO). To resolve the challenge of 2D/3D heterojunction formation
63 at the bottom of the perovskite film, we mixed the HBzA ligand into the 2PACz SAM solution,
64 which was then coated onto the ITO bottom electrode. Here, we hypothesized that the amine head

W
65 of the HBzA molecule would react with the phosphonic acid (–PO(OH)2) group of the 2PACz,

IE
66 forming an ionic bond, resilient against subsequent processing. This acid-base reaction is favorable
67 with the lower values of the negative log of the acid dissociation constant for the first proton (pKa1)

EV
68 of the phosphonic acid (~2.5) and the negative log of the base dissociation constant (pKb) of alkyl-
69 amines (~3.6).22,23 We note that only the first proton transfer is favorable between 2PACz and

PR
70 HBzA, owing to a higher pKa2 value of the second proton (~8.5).22 We experimentally verified that
71 the abovementioned reaction takes place in the mixture of 2PACz and HBzA, as discussed in
72 Extended Data Fig. 1a–b. Moreover, the –OH group from HBzA can form hydrogen bonds with

LE
73 –PO(OH)2 and ITO, thus strengthening the attachment further.24 The detailed explanation of all
74 these mechanisms is in Fig. 1a.

75
C
We then investigated the chemical properties of HBzA and 2PACz molecules on ITO using X-ray
TI
76 photoelectron spectroscopy (XPS). The high-resolution spectrum of the carbon region (C1s) of
AR

77 HBzA shows the presence of C=C, C–N, and C–OH bonds that correspond to the aromatic ring,
78 amine head, and the phenol group of HBzA, respectively (Extended Data Fig. 1c). Further, C1s
79 spectra of 2PACz feature the signature of C–P bonds, corresponding to its –PO(OH)2 group. These
D

80 identical features of 2PACz and HBzA were also found after their mixture was coated on ITO,
E

81 indicating the presence of both molecules. To strengthen our hypothesis, we also did a washing
AT

82 test on the ITO/HBzA and ITO/(HBzA+2PACz) samples with a highly polar aprotic DMF/DMSO
83 solvent mixture that can easily solvate HBzA molecules by forming strong hydrogen bonding. We
ER

84 observed that if only HBzA layers are anchored on ITO, they could be easily washed away.
85 However, the presence of 2PACz significantly enhances the attachment of HBzA onto the ITO
86 surface, as indicated by a qualitative analysis of the marginal reduction of the atomic ratio of
EL

87 nitrogen to indium (N/In) after washing (Supplementary Table 1). This supports our hypothesis
88 that HBzA and 2PACz form an ionic bonding through proton transfer from phosphonic acid to the
C

89 amine group. This bonding facilitates their adherence to the substrate during perovskite processing
AC

90 and assists in the formation of a 2D perovskite layer beneath the 3D perovskite layer.

3
 91 After successfully attaching HBzA ligands on the ITO/2PACz surface, we deposited the 3D
92 perovskite ink via spin-coating. The HBzA ligands are liberated after 3D perovskite deposition,
93 initiating the 2D perovskite formation at the bottom interface through cation exchanges between

W
94 HBzA+ and formamidinium (FA+) or caesium (Cs+) of the 3D perovskite ink. This cation exchange

IE
95 process is completed during the thermal annealing of perovskite films at 120 °C for 30 min to
96 crystallize the 2D/3D perovskite film. To prove this mechanism, we completely removed the

EV
97 resulting perovskite layers from the ITO/2PACz/2D-HBzA/3D-perovskite structure by washing
98 the samples with a DMF/DMSO solution and measured the work function (WF) of the resulting

PR
99 structure using kelvin-probe-force microscopy (KPFM). We noted that the WF of bare
100 ITO/(2PACz+HBzA) sample remained similar before and after washing with DMF/DMSO (5.08
101 eV, Extended Data Fig. 2a–c). However, after the perovskite film deposition and subsequent

LE
102 removal of the HBzA ligands, the WF of ITO/(2PACz+HBzA) is shifted to 5.42 eV, closer to the
103 WF of pristine ITO/2PACz (5.40 eV) with identical conditions (Extended Data Fig. 2d–e). This
104
C
indicates that DMF/DMSO alone was not sufficient to remove HBzA molecules. Rather, the
TI
105 removal was facilitated by the 3D perovskite deposition, corroborating our hypothesis that 2D
AR

106 ligands are released only after perovskite film formation.

107 The cross-sectional high-resolution scanning transmission electron microscopy (HR-STEM)

108 images in Fig. 1b reveal the formation of a discontinuous 2D perovskite layer at the bottom
D

109 interface between ITO/2PACz and 3D perovskite layers. The discontinuity of 2D perovskite layers
E

110 might be related to their complex formation mechanism, such as random distribution of HBzA
AT

111 molecules and their release during cation exchange. We further analyzed the 2D/3D perovskite
112 heterojunction at the bottom-side interface by photoluminescence (PL) spectroscopy and observed
ER

113 distinct PL emission peaks for 2D perovskite at ~570 nm and ~520 nm corresponding to n = 2 and
114 n = 1, respectively (Fig. 1c).12 The significantly high PL emission intensity of n = 2 indicates the
115 dominant formation of near-phase pure 2D perovskite layers. In addition, we performed PL
EL

116 mapping at ~570 nm with different concentrations of HBzA ligand and observed discontinuities
117 for 2D perovskite crystal PL emission at the bottom interface (Supplementary Fig. 1), in
C

118 accordance with HR-STEM results.

AC

119 We further analyzed the structural properties and chemical composition of 2D perovskite at the
120 bottom-side interface by gently separating the bottom-side 2D/3D perovskite film from the
121 ITO/hole-selective surface using the lift-off technique following previous report.25 This approach

4
 122 preserves the chemical and structural integrity of the film (see the schematics of the process in
123 Supplementary Fig. 2a). We then characterized the bottom-side perovskite film by scanning
124 electron microscopy (SEM) and XPS measurements. SEM top-view images of the lift-off samples

W
125 show no distinctive morphological damage to the control and 2D/3D heterojunction samples

IE
126 (Supplementary Fig. 2b). However, we note that the crystal grain-sizes of perovskite of 2D/3D
127 samples are larger (1.64 ± 0.31 µm) compared to control perovskite (0.93 ± 0.20 µm), indicating

EV
128 that 2D perovskite formation at the bottom interface also regulates the crystallization of the
129 overlying 3D perovskite film.26 Further, the XPS analysis of 2D/3D samples revealed the presence

PR
130 of C–O and C–N bonds at 286.5 and 285.6 eV, associated with HBzA ligands (Supplementary
131 Fig. 2c).

132

LE
133 3D/2D heterojunction at the top interface

C
134 For the 3D/2D heterojunction at the top interface, we developed a simple two-step hybrid method
135 to form phase-pure 2D perovskite layers. This process initially involves depositing the PbI2 layer
TI
136 with controlled thickness via vacuum evaporation on top of the as-prepared 3D perovskite film
AR

137 (see Fig. 2a, step (i)). This is then followed by dropping the HBzA salt or mixture of HBzA
138 salt+FAI solutions on top of perovskite/PbI2 film in step (ii). Finally, the samples are thermally
139 annealed to form a 2D perovskite layer with controllable dimensionality and phase purity in step
D

140 (iii).
E

141 Fig. 2b shows the normalized PL spectra of 3D/2D perovskite films for two conditions. First, when
AT

142 coating the HBzA salt cation-only solution, the 3D/2D perovskite sample shows a single emission
143 peak at a low wavelength ~520 nm (corresponding to n = 1).8,12,27 Second, the sample based on the
ER

144 mixture of HBzA salt+FAI also shows a single emission peak but at a higher wavelength (~570
145 nm), which corresponds to n = 2.8,12 These results confirm the success of our method to control
EL

146 the phase-purity and dimensionality of 2D perovskites, which to date remained a challenge.7,8,12,14
147 In contrast, the conventional method by direct coating of HBzA salt solution on top of 3D
C

148 perovskite followed only steps (ii) and (iii) (with the absence of PbI2 deposition) forms a mixture
of multiple dimensionalities of 2D-perovskite (n = 1, 2, and 3; Supplementary Fig. 3), consistent
AC

149
150 with previous works.8,12 We note that a higher dimensionality of the 2D perovskite (n >1)

5
 151 passivation layer is critical, owing to a higher conductivity by having less bulky-organic cations
152 interfaces.8,11,12

153 Next, we identified the crystal orientation and dimensionality of the 2D perovskite layers through

W
154 grazing-incidence wide-angle X-ray scattering (GIWAXS) measurements. Fig. 2c shows the

IE
155 diffraction map of GIWAXS of the 3D/2D sample with strong diffraction peaks at lower qz (0.31
156 Å-1 and 0.64 Å-1) representing the (002) and (004) planes of 2D perovskite (n = 2), respectively,

EV
157 with a layer spacing d of ~20.3 Å-1 and ~9.8 Å-1.13,14,28,29 Interestingly, strong and more discrete
158 Bragg spots at a higher qz (~1.5 Å-1 and ~1.7 Å-1) were observed in the 3D/2D sample. This

PR
159 indicates the improved orientation of the 2D perovskite corresponding to the (282), (121), and
160 (082) planes, formed with perpendicular orientation to the 3D perovskite underneath, as
161 highlighted with oval dot lines.30,31 In contrast, we do not observe the diffraction peaks at lower qz

LE
162 and discrete Bragg spots in control perovskite samples (Supplementary Fig. 4a–b). This result is
163 also consistent with the X-ray diffraction pattern of phase-pure 2D perovskite thin film, as shown
164
C
in Supplementary Fig. 4c–d. Further, we visualize the 2D perovskite formation and orientation
TI
165 at the top-contact using cross-sectional HR-STEM in Fig. 2d, with a predominantly perpendicular
AR

166 orientation. In addition, a thin layer of 2D perovskite does not change the morphology of 3D
167 perovskite underneath (Supplementary Fig. 5), suggesting the robustness of our hybrid method
168 to form perovskite heterostructures.
D

169
E

170 Device performance and characterizations

AT

171 After successfully completing the near-phase pure 2D perovskite passivation layers at the bottom
172 and top 3D perovskite interfaces, we adopted them to PSCs with a structure of ITO/2PACz/2D-
ER

173 perovskite/3D-perovskite/2D-perovskite/C60/BCP/Ag (Fig. 3a). We first tested various alkyl-

174 amine-based 2D ligands with different concentrations with respect to the ratio between 2PACz
EL

175 SAM and 2D ligands at the bottom-side 2D/3D passivation, as shown from the statistical analysis
176 of devices performance in Extended Data Fig. 3a–b. The device results revealed that the optimum
C

177 concentration of the 2D ligand at the bottom contact is around a 1:1 molar ratio with 2PACz SAM.
AC

178 The higher concentration of 2D ligand at the bottom interface leads to a thicker 2D perovskite
179 layer, lowering the FF of devices. This is likely due to the nature of higher resistance of long-alkyl
180 ligands in parallel orientation to the 3D perovskite layer, which can inhibit charge transport.32

6
 181 Interestingly, among the five alkyl-amine ligands tested, the HBzA ligand gives the highest
182 performance, potentially due to the addition of an –OH group that could form a hydrogen bond
183 with –PO(OH)2 and ITO or coordinate with 3D perovskite at the 2D/3D interface.33,34 The longer

W
184 2D ligand (i.e., hexylmethylamine) results in the lower performance, potentially due to a higher

IE
185 charge transfer resistivity induced by long-alkyl ligand. This suggests the critical roles of
186 additional functional groups and the design of the proper size of 2D ligands for bottom

EV
187 passivation.8,16

188 Next, we fabricated the top-side 3D/2D heterojunction using the abovementioned hybrid method.

PR
189 Here, we deposited PbI2 with different thicknesses (7, 15, and 25 nm) and found the thickness of
190 2D perovskite will depend on the initial thickness of the inorganic PbI2 layer. Based on the device
191 results, it is found that 15 nm-thick PbI2 film results in 35 nm-thick 2D perovskite layer, which is

LE
192 optimal for top-side passivation (Extended Data Fig. 3c). In contrast, a 25 nm PbI2 film forms
193 thicker 2D perovskite (around 55 nm, see cross-sectional SEM image in Supplementary Fig. 6),
194
C
which slows down the charge transfer/extraction rate at the perovskite/C60 interface. It is consistent
TI
195 with the decreased FF values from these devices, as indicated by the transient PL measurement
AR

196 and differential lifetime analysis in Supplementary Fig. 7.

197 We then investigated the relation between the 2D perovskite orientation and charge transfer and
198 extraction rates. For this, we compared the charge transfer and extraction rates between the 3D
D

199 perovskite and C60 when a 35 nm-thick 2D perovskite layer prepared by hybrid and solution post-
E

200 treatment methods is inserted between them. It should be noted that the 2D perovskite layer
AT

201 prepared with solution post-treatment method has a parallel orientation to the 3D perovskite
202 surface, which significantly hinders charge transport, and limits the thickness of the 2D
ER

203 layer.8,11,12,16,35

204 The charge transfer and extraction at the interfaces also depend on the energetic alignment between
EL

205 3D perovskite and charge-selective contacts (Extended Data Fig. 4a). With the two-step hybrid
206 method, the secondary electron-cut-off of 2D perovskite shifted to lower binding energy
C

207 (Extended Data Fig. 4b). As a result, the conduction band minimum (CBM) of top-side phase-
AC

208 pure 2D perovskite (n = 2) prepared by the hybrid method was deeper compared to the previously
209 reported mix-phase of 2D perovskite top-contact passivation using the conventional solution post-
210 treatment method, which resulted in dramatically minimizing the electron barrier.8,12,16 Further,

7
 211 the slightly deeper valence band maximum (VBM) of the bottom-side 2D perovskite compared to
212 the control 3D perovskite is favourable for hole transfer at the ITO/2PACz contact interface
213 (Extended Data Fig. 4c–d).

W
214 We then adopted double-side 2D/3D heterojunction passivation to our inverted PSCs and

IE
215 demonstrated a maximum PCE of 25.63% (VOC = 1.19 V, JSC = 24.94 mA cm−2, and FF = 85.9%,
216 see Fig. 3b) under reverse scan and stabilized PCE of 25.1% using triple-cation

EV
217 Cs0.025MA0.075FA0.90PbI3 perovskite with a bandgap of 1.53 eV (Supplementary Fig. 8). This
218 result represents an absolute ~2.3% PCE enhancement compared to control devices (PCE around

PR
219 23.3%) and >1% to only top- or only bottom-side 2D/3D heterojunction with PCE around 24%.8,10-
12
220 Fig. 3c shows the statistical analysis results of the PCE, VOC, and FF values from 25 devices for
221 each condition. We demonstrated the universality of our approach by testing various hole-

LE
222 transporting SAM molecules (i.e., 2PACz, MeO-2PACz, and 4PADCB) and other perovskite
223 compositions of Cs0.05FA0.95PbI3 (~1.54 eV) with PCEs reaching up to 25% for each condition, as
224
C
shown from the statistical analysis of devices in Supplementary Figs. 9 and Extended Data Fig.
TI
225 5, respectively. We also confirmed the high process robustness of our approaches by showing less
AR

226 than 1% PCE deviation from person-to-person, performed by three different researchers
227 (Supplementary Fig. 10). The performance of the double-side 2D/3D heterojunction passivated
228 solar cells were verified by an independently accredited testing center by presenting a PCE of
D

229 25.00% (VOC = 1.17 V, JSC = 25.00 mA cm−2, and FF = 85.7% under reverse current-voltage scan,
E

230 Supplementary Fig. 11). This result represents one of the highest certified efficiencies in inverted
AT

231 PSCs in current literature (Supplementary Table 2).

232 Notably, the VOC × FF value of our double-side 2D/3D heterojunctions-based device was around
ER

233 91% with respect to the Shockley–Queisser limit, one of the highest values among PSCs with only
234 top-side or bulk passivation (Supplementary Fig. 12). The high VOC and FF values of the devices
EL

235 can be associated with the low charge recombination and traps at the top and bottom interfaces.15,36
236 First, we measured the PL quantum yield (PLQY) to demonstrate the quality of perovskite film
C

237 before and after 2D perovskite passivation. The substantially improved PLQY of the double-side
AC

238 2D/3D passivation (6.67% ± 0.58%) confirms the high quality of 2D/3D/2D perovskite film
239 compared to control perovskite sample (1.84% ± 0.63%) and one-side 2D/3D passivation sample
240 only (Extended Data Fig. 6a). Even after adding the C60 layer, the maintained a high PLQY value
241 of double-side 2D/3D perovskite (5.78% ± 0.33%) versus the control sample (significantly

8
 242 dropped around 0.19% ± 0.06%), demonstrating the improved optoelectronic quality of perovskite
243 layer with the full stack of devices. This suggests that the 2D perovskite passivation layer can
244 efficiently suppress the charge recombination at these interfaces. These results are also consistent

W
245 with the longer Shockley-Read-Hall lifetime of double-side 2D/3D heterojunctions (2.69 µs),

IE
246 around three and two times longer versus control (0.97 µs), and only one-side passivation (1.34 µs
247 and 1.73 µs for the bottom- and top-side only passivation, respectively), as shown in Extended

EV
248 Data Fig. 6b.

249 We also determined the trap states for control and double-side 2D/3D passivation of perovskite

PR
250 films using space-charge-limited-current (SCLC) and thermal admittance spectroscopy (TAS)
251 analyses.12,37 For the SCLC, we fabricated the electron-only (ITO/SnO2/perovskite/C60/BCP/Ag)
252 and hole-only (ITO/SAM/perovskite/PTAA/MoOx/Ag) devices to estimate the trap densities of

LE
253 perovskite films. The trap-filled limited voltage (VTFL) of double-side 2D/3D passivation samples
254 was more than 2-fold lower than control samples for both electron- and hole-only devices
255
C
(Extended Data Fig. 6c–d), indicating reduced trap density states (Nt) for both electron- and hole-
TI
256 only devices (see Supplementary Note 1). This result suggests reduced negative and positive
AR

257 defects at both interfaces with 2D perovskite passivation. It should be noted that the SCLC analysis
258 has limitations in extracting the exact value of Nt due to the presence of mobile ions in perovskite
259 materials.38 Therefore, we also performed TAS analysis to quantify the Nt of perovskite films
D

260 relative to their energetic defect levels.12,37 Overall, the Nt of double-side 2D/3D perovskite
E

261 passivation-based devices has lower Nt values in all energetic defect levels, suggesting sufficient
AT

262 passivation of shallow and deep levels of defect in perovskite film by 2D perovskite passivation
263 (Extended Data Fig. 6e).
ER

264 Further, the defects at the interfaces of perovskite films can affect to charge accumulation and
265 electric-field inhomogeneity of perovskite devices.1,3-5,37 We then analyzed the capacitance–
EL

266 voltage under various light illumination intensities to evaluate charge dynamics at the interfaces
267 (Supplementary Fig. 13). The shift in peak-potential (Vpeak) under reverse bias with respect to the
C

268 illumination intensity reflected the degree of interfacial charge accumulation, which was due to
interfacial charge trapping.37 Marginal shifting of the Vpeak in the double-side 2D/3D device
AC

269
270 indicated a reduced charge accumulation (less charge trapping) compared to the control device
271 (larger Vpeak shifting), consistent with the above results. Interestingly, from the cross-sectional
272 KPFM mapping (Fig. 3d and 3e), stronger electric-field differences after 2D perovskite insertion

9
 273 at the perovskite/C60 contact further prove a lower interfacial defect density.39 We note that the
274 electric-field extended further into the perovskite film, more effectively repelling holes and thus
275 suppressing recombination at the perovskite/C60 interface.40 We note that a similar electric-field

W
276 strength difference is observed at the 2PACz/perovskite contact for both control and double-side

IE
277 passivation devices, presumably due to the higher dipole of 2PACz SAMs.39 However, a uniform
278 electric-field is observed after 2D perovskite insertion at 2PACz/perovskite contact, as indicated

EV
279 by a single peak. Consequently, this aids in charge transfer/extraction and leads to enhance VOC
280 and FF values of devices.

PR
281
282 Stability and ion migration analyses
283 We evaluated the long-term stability of encapsulated PSCs under light and thermal stresses. First,

LE
284 Fig. 3f shows the stability under the open-circuit condition at 1-sun and 85 °C of the devices in air
285 (with a relative humidity of ~50%). The double-side 2D/3D heterojunction devices delivered a
286
C
relative PCE loss of only 5% after more than 1000 h. In contrast, the bottom- and top-side
TI
287 passivation-based devices retain around 75% and 86% from their initial PCEs, respectively, versus
AR

288 the control devices (without any 2D/3D heterojunction) retain around only 35% under identical
289 conditions. This result indicates the importance of passivating perovskite film's top and bottom
290 interfaces to improve their stability. Further, we measured the operational stability of double-side
D

291 2D/3D heterojunctions device under maximum-power-point tracking (MPPT). Around 90% of
E

292 their original PCE was retained after the PCE tracking for over 1000 h under 1-sun at 40 °C (Fig.
AT

293 3g).

294 The degradation of perovskites under illumination and thermal stress can be related to defect-
ER

295 induced ion migrations.1,25,37,41 First, the defects related to iodine generation and migration were
296 investigated under 1-sun illumination at 85 °C. It has been reported that the significant release of
297 iodine compound from perovskite film during light soaking at elevated temperature is mainly due
EL

298 to interstitial iodide or PbI2 defects.25,41 Passivating those defects by double-side 2D/3D
299 passivation suppresses the iodine generation/migration in perovskite films. This is proved by the
C

300 lowering intensity of UV-vis absorption spectrum of iodine at the maximum peak of wavelength
AC

301 around 500 nm after immersing the perovskite films into toluene solution to light and heat
302 (Extended Data Fig. 7a–b).25,41

10
 303 Finally, the activation energy for ion migration was obtained to reveal the ion-induced migration
304 character in 2D/3D and control 3D perovskite films using temperature-dependent conductivity
305 measurement.37 The activation energy value of the control 3D perovskite was fitted to be 0.27 eV,

W
306 in agreement with a previously reported value. In contrast, the activation energy value for the

IE
307 double-side 2D/3D heterojunctions film is increased to 0.47 eV (Extended Data Fig. 7c). This
308 result indicates that the double-side 2D/3D heterojunctions have a significantly boosted energy

EV
309 barrier for ion migration, which might also improve the perovskite crystal stability.

310

PR
311 References
312 1 Ni, Z. et al. Evolution of defects during the degradation of metal halide perovskite solar
313 cells under reverse bias and illumination. Nature Energy 7, 65–73, doi:10.1038/s41560-

LE
314 021-00949-9 (2022).
315 2 Ni, Z. et al. Resolving spatial and energetic distributions of trap states in metal halide

C
316 perovskite solar cells. Science 367, 1352–1358, doi:10.1126/science.aba0893 (2020).
317 3 Chen, H. et al. Regulating surface potential maximizes voltage in all-perovskite tandems.
TI
318 Nature 613, 676–681, doi:10.1038/s41586-022-05541-z (2023).
AR

319 4 Jiang, Q. et al. Surface reaction for efficient and stable inverted perovskite solar cells.
320 Nature 611, 278–283, doi:10.1038/s41586-022-05268-x (2022).
321 5 Tan, S. et al. Stability-limiting heterointerfaces of perovskite photovoltaics. Nature 605,
322 268–273, doi:10.1038/s41586-022-04604-5 (2022).
D

323 6 Jung, E. H. et al. Efficient, stable and scalable perovskite solar cells using poly(3-
E

324 hexylthiophene). Nature 567, 511–515, doi:10.1038/s41586-019-1036-3 (2019).

325 7 Yang, G. et al. Stable and low-photovoltage-loss perovskite solar cells by multifunctional
AT

326 passivation. Nature Photonics 15, 681–689, doi:10.1038/s41566-021-00829-4 (2021).

327 8 Chen, H. et al. Quantum-size-tuned heterostructures enable efficient and stable inverted
ER

328 perovskite solar cells. Nature Photonics 16, 352–358, doi:10.1038/s41566-022-00985-1

329 (2022).
330 9 Luo, L. et al. Stabilization of 3D/2D perovskite heterostructures via inhibition of ion
EL

331 diffusion by cross-linked polymers for solar cells with improved performance. Nature
332 Energy 8, 294–303, doi:10.1038/s41560-023-01205-y (2023).
333 10 Jang, Y.-W. et al. Intact 2D/3D halide junction perovskite solar cells via solid-phase in-
C

334 plane growth. Nature Energy 6, 63–71, doi:10.1038/s41560-020-00749-7 (2021).

AC

335 11 Sidhik, S. et al. Deterministic fabrication of 3D/2D perovskite bilayer stacks for durable
336 and efficient solar cells. Science 377, 1425–1430, doi:10.1126/science.abq7652 (2022).
337 12 Azmi, R. et al. Damp heat-stable perovskite solar cells with tailored-dimensionality 2D/3D
338 heterojunctions. Science 376, 73–77, doi:10.1126/science.abm5784 (2022).

11
 339 13 Niu, T. et al. Spacer engineering of diammonium-based 2D perovskites toward efficient
340 and stable 2D/3D heterostructure perovskite solar cells. Advanced Energy Materials 12,
341 2102973, https://doi.org/10.1002/aenm.202102973 (2022).
342 14 Proppe, A. H. et al. Multication perovskite 2D/3D interfaces form via progressive

W
343 dimensional reduction. Nature Communications 12, 3472, doi:10.1038/s41467-021-
344 23616-9 (2021).

IE
345 15 Azmi, R. et al. Moisture-resilient perovskite solar cells for enhanced stability. Advanced
346 Materials 2211317, https://doi.org/10.1002/adma.202211317 (2023).

EV
347 16 Gu, H. et al. Phase-pure two-dimensional layered perovskite thin films. Nature Reviews
348 Materials 8, 533–551, doi:10.1038/s41578-023-00560-2 (2023).

PR
349 17 Yang, X. et al. Buried interfaces in halide perovskite photovoltaics. Advanced Materials
350 33, 2006435, https://doi.org/10.1002/adma.202006435 (2021).
351 18 Mahmud, M. A. et al. Cation-diffusion-based simultaneous bulk and surface passivations
352 for high bandgap inverted perovskite solar cell producing record fill factor and efficiency.

LE
353 Advanced Energy Materials 12, 2201672, https://doi.org/10.1002/aenm.202201672
354 (2022).

C
355 19 Mahmud, M. A. et al. Double-sided surface passivation of 3D perovskite film for high-
356 efficiency mixed-dimensional perovskite solar cells. Advanced Functional Materials 30,
TI
357 1907962, https://doi.org/10.1002/adfm.201907962 (2020).
358 20 Wu, S. et al. Modulation of defects and interfaces through alkylammonium interlayer for
AR

359 efficient inverted perovskite solar cells. Joule 4, 1248–1262,

360 https://doi.org/10.1016/j.joule.2020.04.001 (2020).
361 21 Degani, M. et al. 23.7% efficient inverted perovskite solar cells by dual interfacial
362 modification. Science Advances 7, eabj7930, doi:10.1126/sciadv.abj7930 (2021).
D

363 22 Kyuji, O. Prediction of pKa values of alkylphosphonic acids. Bulletin of the Chemical
E

364 Society of Japan 65, 2543–2545, doi:10.1246/bcsj.65.2543 (1992).

AT

365 23 Dorner, R. W., Deifallah, M., Coombes, D. S., Catlow, C. R. A. & Corà, F. Synthesis and
366 structure determination of a novel layered aluminophosphate material templated with 1-
367 phenylethylamine: [AlPO4(OH)](NH3C2H4C6H5). Chemistry of Materials 19, 2261–2268,
ER

368 doi:10.1021/cm070106u (2007).

369 24 Szatyłowicz, H. Structural aspects of the intermolecular hydrogen bond strength: H-bonded
370 complexes of aniline, phenol and pyridine derivatives. Journal of Physical Organic
EL

371 Chemistry 21, 897–914, https://doi.org/10.1002/poc.1394 (2008).

372 25 Chen, S. et al. Stabilizing perovskite-substrate interfaces for high-performance perovskite
373 modules. Science 373, 902–907, doi:10.1126/science.abi6323 (2021).
C

374 26 Li, H. et al. 2D/3D heterojunction engineering at the buried interface towards high-
AC

375 performance inverted methylammonium-free perovskite solar cells. Nature Energy 8, 946–
376 955, doi:10.1038/s41560-023-01295-8 (2023).

12
 377 27 Stolterfoht, M. et al. The impact of energy alignment and interfacial recombination on the
378 internal and external open-circuit voltage of perovskite solar cells. Energy &
379 Environmental Science 12, 2778–2788, doi:10.1039/C9EE02020A (2019).
380 28 Alanazi, A. Q. et al. Benzylammonium-mediated formamidinium lead iodide perovskite

W
381 phase stabilization for photovoltaics. Advanced Functional Materials 31, 2101163,
382 https://doi.org/10.1002/adfm.202101163 (2021).

IE
383 29 Cao, D. H., Stoumpos, C. C., Farha, O. K., Hupp, J. T. & Kanatzidis, M. G. 2D homologous
384 perovskites as light-absorbing materials for solar cell applications. Journal of the American

EV
385 Chemical Society 137, 7843–7850, doi:10.1021/jacs.5b03796 (2015).
386 30 Quintero-Bermudez, R. et al. Compositional and orientational control in metal halide
387 perovskites of reduced dimensionality. Nature Materials 17, 900–907,

PR
388 doi:10.1038/s41563-018-0154-x (2018).
389 31 Tsai, H. et al. High-efficiency two-dimensional Ruddlesden–Popper perovskite solar cells.
390 Nature 536, 312–316, doi:10.1038/nature18306 (2016).

LE
391 32 Chen, Y. et al. Tailoring organic cation of 2D air-stable organometal halide perovskites for
392 highly efficient planar solar cells. Advanced Energy Materials 7, 1700162,
393 https://doi.org/10.1002/aenm.201700162 (2017).
394 33
C
Wang, Y. et al. Stabilizing heterostructures of soft perovskite semiconductors. Science 365,
TI
395 687–691, doi:10.1126/science.aax8018 (2019).
396 34 Cohen, B.-E. et al. Hydroxyl functional groups in two-dimensional dion–jacobson
AR

397 perovskite solar cells. ACS Energy Letters 7, 217–225, doi:10.1021/acsenergylett.1c01990

398 (2022).
399 35 Zhao, W. et al. Orientation engineering via 2D seeding for stable 24.83% efficiency
400 perovskite solar cells. Advanced Energy Materials 13, 2204260,
D

401 https://doi.org/10.1002/aenm.202204260 (2023).

E

402 36 Zhang, Z. et al. Rationalization of passivation strategies toward high-performance

403 perovskite solar cells. Chemical Society Reviews 52, 163–195, doi:10.1039/D2CS00217E
AT

404 (2023).
405 37 Azmi, R. et al. Shallow and deep trap state passivation for low-temperature processed
ER

406 perovskite solar cells. ACS Energy Letters 5, 1396–1403,

407 doi:10.1021/acsenergylett.0c00596 (2020).
408 38 Le Corre, V. M. et al. Revealing charge carrier mobility and defect densities in metal halide
EL

409 perovskites via space-charge-limited current measurements. ACS Energy Letters 6, 1087–
410 1094, doi:10.1021/acsenergylett.0c02599 (2021).
411 39 Zheng, X. et al. Co-deposition of hole-selective contact and absorber for improving the
C

412 processability of perovskite solar cells. Nature Energy 8, 462–472, doi:10.1038/s41560-

413 023-01227-6 (2023).
AC

414 40 Hou, Y. et al. Efficient tandem solar cells with solution-processed perovskite on textured
415 crystalline silicon. Science 367, 1135–1140, doi:10.1126/science.aaz3691 (2020).

13
 416 41 Lin, Y.-H. et al. A piperidinium salt stabilizes efficient metal-halide perovskite solar cells.
417 Science 369, 96–102, doi:10.1126/science.aba1628 (2020).
418

W
419 Figure legends

420 Fig. 1: 2D/3D heterojunction at hole-selective bottom interface | a, Scheme showing the

IE
421 mechanism of 3D-perovskite formation on ITO/2PACz+HBzA and cation exchange at the bottom
422 interface (left-hand-side); 2D/3D heterojunction on ITO/2PACz bottom-contact (right-hand-side).

EV
423 b, Cross-sectional High-angle annular dark-field (HAADF)-STEM image of near bottom HTL
424 contact with zoom-in of the area showing the parallel 2D perovskite layers. Fast Fourier transform
425 (FFT) shows 2D inter-planar spots (002) and (004) are matched with the dominant phase of n = 2.

PR
426 c, Normalized PL spectra of bottom-side 2D/3D from low to high wavelengths with different
427 excitation directions from the glass-side and top perovskite surface film using 405 nm excitation.

428 Fig. 2: 3D/2D heterojunction at electron-selective top interface | a, Sketch explaining two-step

LE
429 hybrid process to form 3D/2D heterojunction perovskites. b, Normalized PL spectra of each film
430 with light direction from the top perovskite surface film. c, GIWAXS maps of top-side 3D/2D
431 heterojunction at a low incidence angle (0.2°). d, Cross-sectional HAADF-STEM image of near

C
432 top C60-contact with zoom-in of the area showing the perpendicular orientation of 2D-perovskite.
TI
433 Fig. 3: Device performance and stability analyses | a, Sketch of the fabricated single-junction
434 PSC. b, Representative J–V curve for control, bottom, top, and double-side 2D/3D heterojunctions
AR

435 PSC. c, Statistics parameters of PCE, VOC, and FF (from top to bottom) values of PSCs. The box
436 and whisker plots indicate the statistical distribution: the center line represents the average value;
437 the bottom and top of the box represent the 25% to 75% percentiles; the small square represents
438 the mean; the whiskers represent outliers. d, e, Electric field distribution through both interfaces
439 from C60-ETL to glass/ITO side acquired via KPFM cross-section scans. We mapped the photo-
D

440 voltage of the cells illuminated by an LED light. Note that we break the hole- and electron-selective
E

441 peaks for better visualization. f, Stability of encapsulated control bottom, top, and double-side
442 2D/3D passivation-based devices under 1-sun illumination at 85 °C and open-circuit conditions.
AT

443 Five cells were used to construct the statistic, with the average represented by the symbol and the
444 standard deviation by the error bar. g, MPPT of an encapsulated double-side 2D/3D passivation-
445 based device under 1-sun illumination at 40 °C.
ER

446
447 Methods
EL

448 Materials

449 Dimethylformamide (DMF, 99.8%, anhydrous), dimethylsulfoxide (DMSO, 99.9%, anhydrous),

C

450 potassium chloride (KCl, 99.9%), 4-Hydroxybenzylamine (HBzA, C7H9NO), benzylamine,

AC

451 phenethylamine, butylamine, hexylmethylamine, and cesium iodide (CsI, 99.9%) were all
452 purchased from Sigma Aldrich. Formamidinium iodide (FAI) and methylammonium chloride
453 (MACl) were purcahsed from Xian-Polymer. Lead iodide (PbI2, ultra-dry) was purchased from

14
 454 Alfa Aesar. 4-Hydroxybenzylammonium salts were purchased from Alfa-chemical. 2PACz ([2-
455 (9H-carbazol-9-yl)ethyl]phosphonic acid, >98.0%) and MeO-2PACz were purchased from TCI.
456 4PADCB (4-(7H-dibenzo[c,g]carbazol-7-yl)butyl)phosphonic acid) was purchased from

W
457 Dyenamo. C60 (>99.5% purity) and bathocuproine (BCP, >99%) were purchased from Ossila Ltd.

IE
458 Device fabrication

EV
459 Glass/ITO substrates were washed in first acetone and then isopropyl alcohol for 20 min of each.
460 After washing, the substrates were dried and treated with UVO prior to use for 10-20 min. The
461 hole-transporting SAM (1 – 2 M) or mixed SAMs+2D-ligands (1:0.5, 1:1, 1:1.5, and 1:2 molar

PR
462 ratio) were dissolved in ethanol. Then, the solution was spin-coated onto glass/ITO at 5000 rpm
463 for 30 s in a nitrogen-filled glovebox. After thermal annealing at 100 °C for 10 min, the hole-

LE
464 selective layer coated ITO was cooled down and washed by ethanol by dynamic coating at 5000
465 rpm for 30 s to remove unbounded molecules.

466
C
A perovskite precursor solution (~1.5 M) of Cs0.025MA0.075FA0.90PbI3 was prepared by mixing FAI,
TI
467 MAI, CsI, and PbI2 with a molar ratio of 0.90:0.075:0.025:1 and then dissolving them in a mixed
468 solvent of DMF and DMSO (4:1 vol. ratio). 15 mol% of MACl was also added. For
AR

469 Cs0.05FA0.95PbI3 was prepared by mixing FAI, CsI, and PbI2 with a molar ratio of 0.95:0.05:1 and
470 then dissolving them in a mixed solvent of DMF and DMSO (4:1 vol. ratio) to get ~1.5 M. Then,
D

471 around 10 mol% of MACl and 2 mol% of KCl were also added in the mixed perovskite solution.
472 The mixture of precursor solutions was shaken by a vortex shaker overnight at room temperature
E

473 until fully dissolved.

AT

474 The perovskite precursor solution was spin-coated at 2000 rpm for 40 s and 6000 rpm for 10 s. At
ER

475 the 10th s, 200 mL of anisole was dropped onto the films by the end of the spin-coating process.
476 Then, the films were annealed at 120 °C for 30 min to form the perovskite layer. For 2D perovskite
477 passivation at the top side, at the first step, PbI2 beads were thermally evaporated (10−6 Torr) at an
EL

478 evaporation rate of 0.5 Å s−1 to get thickness around 7, 15, and 25 nm. The second step is dropping
479 HBzA salt solution or mixture HBzA salt: FAI solution with 1:1 molar ratio (in mixture IPA and
C

480 DMF, 98:2 vol%) and then spin-coated at 2000 rpm for 30 s, followed by thermal annealing at
AC

481 100 °C for 10 min. After this process, the perovskite films were stored overnight in vacuum
482 condition. All the process is done under N2-glovebox with O2 level was <10 ppm and H2O level
483 was <0.1 ppm. Then the samples were transferred into a thermal evaporator for the C60 (25 nm)

15
 484 and BCP (3 nm) deposition. For the final step, a 120 nm thick Ag layer was evaporated at low
485 pressure (<10-6 Torr) with an area of ~0.1 cm2. A ~120 nm MgF2 was also evaporated at the glass
486 side to minimize the reflection losses from the glass substrates.

W
487 Device Analysis

IE
488 The J–V curves were recorded using a Keithley 2400 source unit under AM 1.5 G (100 mW cm−2)

EV
489 illumination with an Abet Technologies Sun 3000 solar simulator. The spectral mismatch was
490 calibrated using a KG-5 filter-covered mono-silicon standard cell (Newport). All devices were
491 measured in nitrogen filled glovebox for a sweep mode of reverse and forward scans with a

PR
492 scanning rate of 0.1 V s-1. The stabilized output of the devices was acquired by recording the power
493 output of the illuminated device at a constant voltage near the MPP extracted from the J–V curve.

LE
494 External quantum efficiency spectra were obtained by passing the output of a 400 W Xenon lamp
495 through a monochromator and filter. Calibration was performed with a 603621 Calibrated Silicon

C
496 and Germanium Reference Detector. Activation energy measurement of ion migration. Electrical
TI
497 poling was performed by applying different biases to the devices for 1 min in the dark; current-
498 voltage were measured with a Keithley 2401 source. Temperature dependence was measured with
AR

499 a CTI cryogenic probe station with an attached Lakeshore 331 temperature controller under
500 vacuum (1  10-4 Pa). The temperature was monitored using a thermocouple placed in contact with
501 the sample. During the measurement, devices were first cooled to 10 K for 10 min and then heated
D

502 to adjust the temperature. The temperature was stabilized for 5 min prior to recording the current
E

503 for each step. Cross-section KPFM measurements were carried out using a home-built system
AT

504 based on Bruker Dimension Icon atomic force microscope. The sample preparation and
505 measurements were conducted inside an Ar-filled glovebox. We used a Pt-Ir coated silicon probe
ER

506 (PPP-EFM) and operate in tapping mode. The mapping was performed with 1024 pixels in the
507 fast-scan axis and more than 32 lines in the slow-scan axis, at a scan rate of 0.35 Hz. To ensure a
EL

508 flat surface and expose all layers of the device, the cell was directly cleaved without polishing or
509 ion-milling. We then mapped the surface potential at the same location under dark and LED
C

510 illumination, with the LED providing white light of approximately 1 Sun intensity. To reduce the
AC

511 noise level, we averaged the electrical potential profiles from all slow-scan lines. Next, we
512 subtracted the dark potential from the illuminated condition and calculated the first derivative to
513 obtain the electric field distribution across the device.

16
 514 Sample Characterization

515 Field-emission SEM images were obtained with a JSM-7610F device (JEOL Ltd.). The PL
516 emission was spectrally resolved using collection optic. For STEM based study, cross-sectional

W
517 electron-transparent lamella was prepared in focused ion beam (FIB) equipped SEM-FIB Helios

IE
518 G4 DualBeam, FEI with help of an EasyLift nanomanipulator and Ga ion source. First protect the
519 region of interest during FIB, two types of protective coatings were deposited: a 0.5 µm layer of

EV
520 Pt coating deposited by the e-beam, followed by a 3 µm layer of Pt coating deposited by the ion
521 beam for final protection. Step by step ion beam milling procedure with beam current (2.4, 0.44,

PR
522 0.26, 0.045, 0.025 nA for 30 to 5 kV) was carried out by cutting and thinning down lamella to 60
523 nm, while decreasing beam current (till 0.025 nA at 2 kV) to avoid ion beam damage. Additionally,
524 low current cleaning process (5-2 kV, 81-28 pA) was performed to remove any potential

LE
525 contamination. STEM based experiments such as HAADF-STEM were performed in Cs Prob-
526 corrected ThermoFisher Titan 60-300 Cubed TEM microscope operating at 300 kV.TEM data
527
C
processing was performed using Gatan™ Digital Micrograph and Thermo Scientific™ Velox
TI
528 suites. For TEM and PL characterizations, the perovskite film samples preparation used 1:2 molar
AR

529 ratio of SAM and 2D ligand to help observation of 2D perovskites.

530 GIWAXS was conducted in the PLS-II 6D UNIST-PAL beamline of the Pohang Accelerator
531 Laboratory (PAL), Korea. The energy of X-rays from the bending magnet was monochromatized
D

532 to 18.986 keV (0.6530 Å) using a double-crystal monochromator. The 2D charge-coupled device
E

533 detector (MX225-HS, Rayonix, LLC, USA) was used to record the 2D GIWAXS pattern, and the
AT

534 GIWAXS sample chamber was equipped with a 5-axis motorized stage for fine sample alignment.
535 The sample-to-detector distance was 242.58 mm, and the diffraction angles were calibrated using
ER

536 NIST SRM660b, lanthanum hexaboride (LaB6). Incidence angle of ~0.2° was used for the
537 GIWAXS measurements to obtain crystallographic information on the perovskite films used in
EL

538 this work.

539 XPS/UPS measurements were performed in a UHV chamber (ScientaOmicron) operating at a

C

540 pressure of 5 x 10-10 mbar. UPS measurements were carried out using a 21.2 eV vacuum UV
AC

541 source (focus). The sample was biased by 10 eV to observe the low kinetic energy cutoff. The
542 photoelectrons were collected at an angle of 80° between the sample and analyzer, with a normal
543 electron take off angle. The constant analyzer pass energy was 5 eV for the valence band region

17
 544 and for the secondary electron cut-off. XPS was carried out in the same spectrometer, equipped
545 with a monochromatic Al Ka X-ray Omicron XM1000 X-ray source (hv = 1486.6 eV) operating
546 at a power of 390 W. The high-resolution spectra were collected at a CAE of 15 eV. The spectra

W
547 were analyzed with Casa XPS software. The individual peak envelopes were fitted by a Gaussian

IE
548 –Lorentzian (GL30) using a Tougaard based background function. Fourier-transform infrared
549 spectroscopy (FTIR) was conducted using Cary 600 Series FTIR Spectrometer (Agilent Tech.)

EV
550 equipped with PIKE GladiATR geometry using diamond crystal in transmission mode. Samples
551 of powder were first dissolved into ethanol solvent and dried in a vacuum, then annealed at 100 oC

PR
552 under N2-glovebox for 10 min to completely remove the solvent. The dried powder samples are
553 spread uniformly above the crystal surface of FTIR. The spectral resolution was set to 4 cm-1 and
554 obtained by averaging 20 scans of each measurement. Transient PL decay and PLQY were

LE
555 recorded using a spectrofluorometer (Fluoroma-Modular, Horiba Scientific). The 520 nm
556 calibrated pulse laser was generated from a picosecond laser diode (PicoQuant) equipped band-
557
C
pass 650 nm filter to collect 795 nm emitted light. The fluence is estimated to be around ~10
TI
558 nJ/cm2. For PLQY, an integrating sphere (Fluorolog, Horiba JobinYvon) was used and compared
AR

559 with different tools using a JASCO FP-8500 spectrofluorometer equipped with an ILF-835
560 integrating sphere with the same procedure. It is noted that the results were consistent from two
561 different setups and measurements, confirming our reliability results as given in the statistical
D

562 PLQY values with a small deviation standard in Extended Data Fig. 6a.
E

563 Device stability testing

AT

564 For the stability test, ~10 nm layer of tin (IV) oxide and an 80 nm layer of indium zinc oxide were
565 deposited on top of the C60 sequentially, by atomic layer and sputtering depositions, respectively,
ER

566 to replace the BCP layer. The operational stability tests were carried out at the MPPT for the
567 encapsulated devices under 1 sun in ambient air (RH >50 %) at ~40 °C. The voltage at the MPPT
EL

568 was automatically applied, and the power output of the devices was tracked. For the open-circuit
569 light soaking test, the devices were kept under full 1 sun illumination and constant temperature at
C

570 85 °C on hotplate in ambient air (RH above 50 %). The devices then measured periodically until
AC

571 1000 h. The devices were encapsulated with simple glass and epoxy method. All processes were
572 done in N2 glovebox.

573

18
 574 Acknowledgement.

575 This work was supported by the King Abdullah University of Science and Technology (KAUST)
576 Office of Sponsored Research (OSR) under award nos. OSR-2021-4833, OSR-CARF/CCF-3079,

W
577 IED OSR-2020-4611, IED OSR-2019-4580, OSR-CRG2020-4350, OSR-2020-CPF-4519,

IE
578 OSR438 CRG2019-4093, IED OSR-2019-4208, HERO. We acknowledge the use of KAUST
579 Solar Center and Core Lab facilities and the support from its staff. We thank KPV Lab members,

EV
580 Helen Bristow, Maxime Babics, Shruti Sarwade, Asmat Ullah, etc. for their support. A.M.R. and
581 S.I.S. acknowledge financial support from the Basic Science Research Leaders Program of the

PR
582 National Research Foundation of Korea (NRF) (Grant NRF-2018R1A3B1052820)

583 Author contributions

LE
584 R.A. conceived, directed the research, and wrote the original draft. R.A., D.S.U., I.F.I, and A.P.
585 fabricated the solar cells. B.V. conducted FIB and TEM measurements and analyses. A.M.R.
586
C
performed the GIWAXS measurement and analysis. R.A. and A.M.R. performed PLQY
TI
587 measurement and analysis. S.Z. contributed to conceptualizing the idea of 2D perovskite and
588 discussion at the bottom contact. P.D. performed XPS, IPES, and UPS measurements and did data
AR

589 analysis. R.A., D.S.U., and E.U. performed optical spectroscopy measurements and data analysis.
590 D.S.U. performed SEM measurements. A.P. performed the top-surface KPFM measurement and
D

591 did data analysis. R.A. performed and analysed the electronic characterization of the devices. F.C.
592 performed cross-section KPFM and analysis. D.U. and A.A.S. performed the FTIR and analysis.
E

593 I.F.I. helped to synthesized and characterized 2D perovskite layers. I.F.I. and A.A.S. performed
AT

594 the XRD measurement analysis. A.R.P. fabricated the aperture mask and determined the area.
595 A.A.S. and A.S.S. helped to review the manuscript. E.A. contributed to reviewing and writing the
ER

596 manuscript. S.I.S. and C.X. supervised the project. S.D.W. supervised the overall project and
597 secure the funding. All authors contributed to the manuscript.
EL

598 Conflicts of Interest

C

599 The authors declare no competing interests.

AC

600 Data availability

601 The data sets generated during and/or analysed during the current study are available from the
602 corresponding author upon request.

19
 603 Extended Data Figure legends
604 Extended Data Fig. 1: Chemical analysis of SAMs and HBzA ligand. a, Photograph of the
605 mixture 2PACz:HBzA (1:1 molar ratio) compound after drying under vacuum for a few days and
then annealed at 100 oC in a nitrogen glovebox to completely remove the solvent. The yellow

W
606
607 precipitated powder indicates the protonation between the first P–OH of the phosphonic group of
608 2PACz and the amine of HBzA molecules, forming an ionic bond. b, FTIR peaks of HBzA,

IE
609 2PACz, and mixture (2PACz:HBzA, 1:1 molar ratio) in solid-state powder. The shifting peaks of
610 P–OH at ~950 cm-1 and ~1033 cm-1 to the higher wavenumber of the mixture compound indicate

EV
611 stronger bonding due to deprotonation of the first P–OH group of the phosphonic acid of 2PACz.
612 Consequently, the additional new peak associated with the P–O bond at 1105 cm-1 is formed. In
613 contrast, N–H bending peaks of the amine group (HBzA) at ~1530 cm-1 and 1620 cm-1 are shifted

PR
614 to the lower wavenumber with enhanced intensity in the mixture compound, presumably due to
615 the protonation of –NH2 to –NH3+. These results successfully confirm the acid-base reaction
616 between the pKa1 of the phosphonic acid and the pKb of alkyl amines, as illustrated in Fig. 1a. c,
617 XPS analysis of 2PACz on ITO film. High-resolution spectra of C1s of 2PACz, HBzA, and a

LE
618 mixture of 2PACz+HBzA onto ITO after washing with DMF/DMSO.

619 Extended Data Fig. 2: SKPM mapping images and distributions of WF. a, ITO/2PACz+HBzA
620 sample before and after washing. The DMF/DMSO solvent mixture was used to wash the
621
C
ITO/2PACz+HBzA sample. The minor WF shifting of the ITO/2PACz+HBzA sample before and
TI
622 after washing with DMF/DMSO indicates a strong attachment of HBzA molecules onto 2PACz
623 molecules, consistent with XPS results. b, ITO/2PACz sample after washing. c, Bare ITO. d,
624 SKPM mapping images of ITO/2PACz/perovskite and ITO/2PACz+HBzA/perovskite samples
AR

625 after washing with DMF/DMSO. e, Distributions of WF of both samples from KPFM
626 measurement. After the perovskite film deposition and subsequent removal of the HBzA ligands,
627 the WF of ITO/2PACz+HBzA is shifted back to the WF of pristine ITO/2PACz (~5.40 eV),
628 confirming our hypothesis.
D

629 Extended Data Fig. 3: Statistics of PCE, VOC, and FF of PSCs. a, Bottom-side 2D/3D with
E

630 different molar ratios of 2PACz: HBzA. b, Bottom-side 2D/3D with various 2D ligands
AT

631 (benzylamine, BzA; phenethylamine, PEA; butylamine, BA; hexylmethylamine, HMA) for
632 bottom-side 2D/3D heterojunction with a 1:1 molar ratio of 2PACz. c, Top-side 3D/2D with PbI2
633 thickness variation for top-side 3D/2D heterojunction. The statistics are obtained from 10 to 20
ER

634 cells for each condition from different batches. It is noted that the perovskite used here is based on
635 Cs0.05FA0.95PbI3.

636 Extended Data Fig. 4: Energetic alignment analysis. a, Energy level scheme for 2PACz
EL

637 anchored on ITO, 3D-perovskites, 2D-perovskites on top and bottom interfaces, and C60 layers
638 extracted from UPS data. The VBM was obtained as hν – (Ecutoff – EVB, min). The position of the
639 CBM with respect to the VBM was defined by extracting Eg (~1.53 eV) from EQE in
C

640 Supplementary Fig. 8 for 3D-perovskite and by IPES for 2D-perovskite with n = 2 (Eg = ~2.1 eV).
641 b, c, Secondary electron cut-off and valence spectra plots from UPS measurement. d, IPES spectra
AC

642 of the 2D perovskite film with n = 2 using the hybrid method.

643 Extended Data Fig. 5: Device performance with double-cation perovskite. a, Representative
644 J–V curve for control and double-side 2D/3D heterojunctions perovskite solar cells with

20
 645 Cs0.05FA0.95PbI3 perovskite, achieving a maximum PCE of 25.2% under reverse scan. Statistical
646 parameters of (b) PCE, (c) VOC, and (d) FF values of PSCs. The statistical parameters are obtained
647 from 50 cells for each condition from different batches.

W
648 Extended Data Fig. 6: Passivation and trap analysis. a, PLQY results, and (b) transient PL
649 decays of perovskite films on various 2D/3D heterojunctions. c, d, SCLC plots for control and

IE
650 double-side 2D/3D heterojunction-based devices of (c) electron-only and (d) hole-only devices. e,
651 TAS analysis of control, bottom-side, and double-side 2D/3D passivation-based devices.

EV
652 Extended Data Fig. 7: Stability and ion migration analysis. a, b, UV-vis absorbance spectra
653 evolution of iodine in toluene solution for the control (a) and double-side passivation of perovskite
654 films (b) at different aging times. Inset is a photo of the sealed vials with the control and 2D/3D

PR
655 perovskite films, taken after 100 h of light and heat exposures. c, Temperature-dependent
656 conductivity of control and double-side 2D/3D perovskite films.

657

LE
C
TI
AR
E D
AT
ER
EL
C
AC

21
 W
a Mechanism of 2D ligand formation and release from the ITO/2PACz surface b b Bottom-side 2D/3D perovskite
OH

IE
pKb = ~8.2 ionic bonding OH ionic bonding OH

H2N H OH

EV
N N N NH2 X N
H2N H2N H +
H3N H3N 3D perovskite
EtOH (abs.) Glass/ITO DMF/DMSO NH2
P P P O P O

PR
OH O H3N X
HO O HO O O O O O
2D ligand
In In In In with 2D No 2D
pKa1 = ~2.5 pKa2 = ~8.2
O O
ITO binding site ITO binding site

LE
50 nm ITO
3D perovskite formation on 2PACz+HBzA 2D/3D perovskite heterojunction on 2PACz

C
3D perovskite 004 002

TI
[PbI6]4-
/nm

0. µm

AR
2D perovskite
FA+/Cs+

2PACz
ED
Oxygen-Pi 10 nm ITO
stacking
AT

–PO(OH)2 cc Bottom-side 2D/3D perovskite

1.0

Normalized PL Intensity
R

HBzA+

–NH3+
E

cation exchange From perovskite side

0.5
EL

From glass side

–OH n=1 n=2

C

0.0
Glass/ITO
Glass/ITO Glass/ITO
Glass/ITO 500 600 700 800 900
AC

Wavelength (nm)
 W
IE
a c Top-side 3D/2D perovskite
HBzA salt + FAI mixture 2.5

EV
Top-side
or HBzA salt only in 3D/2D heterojunction
IPA/DMF (98:2 vol%)

PR qz (Å-1)
LE
2D (004)
2D (002)
0.0

C
-2.0 0.0 2.0
(i) Thermal evaporation (ii) Spin-coating 2D ligands (iii) Annealing
qxy (Å-1)
of PbI2 (yellow layer)

TI
b Top-side 3D/2D perovskite d Top-side 3D/2D perovskite

AR
C60/BCP C60/BCP
1.0 004
HBzA salt+FAI mixture
HBzA salt only
Normalized PL Intensity

ED
Control n=2
2D perovskite
002
AT

0.5 1/nm

2D perovskite Perpendicularly
n=1 aligned
R

n=2
E

3D perovskite
EL

0.0
500 600 700 800 900 50 nm 10 nm
Wavelength (nm)
C
AC
 W
a b c
26

IE
Ag 25 25

PCE (%)
C60/BCP 24

EV
20

PCE (%)
25

Current density (mA cm-2)

2D Perovskite 20 Stabilized PCE 25.1% 23
15
3D Perovskite 15 10 22
0 100 200 300 400

PR
Time (s)
21
2D Perovskite 10 Control, Reverese scan
Control, Forward scan 1200
ITO/2PACz Bottom-side, Reverese scan
5 Bottom-side, Forward scan 1175

LE
Glass

VOC (mV)
Top-side, Reverese scan
0 Top-side, Forward scan
1150
Double-side, Reverese scan
Double-side, Forward scan

C
-5 1125
0.0 0.2 0.4 0.6 0.8 1.0 1.2
1100

TI
Voltage (V)

d Glass/ITO 2PACz/2D 3D perovskite 2D/C60 e Glass/ITO 2PACz 3D perovskite C60 87

AR
Δelectric-field (a.u.)
Δelectric-field (a.u.)

84
100 100

FF (%)
81
0 0
ED
78

-100 -100 75
200 nm 200 nm
AT

Control Bottom Top Double

gg
ff 24
24
R

20 20
PCE (%)

PCE (%)

T90
16 16
E

12 Control 3D 12
EL

MPPT
Bottom-side 8
8 Top-side Light +Thermal ITO/2PACz/2D/3D-CsFAPbI3/2D/C60/SnO2/IZO/Cu
Double-side 2 2/IZO/Cu
ITO/2PACz/CsFAPbI3/C60/SnO 4
Double-side
4
C

0
0 200 400 600 800 1000 0 200 400 600 800 1000
AC

Time (h) Time (h)

 W
IE
EV
a b

PR
N-H

N-H
C=C
C=C

C-O

C-OH

C-O
LE
Carbazole
Intensity (a.u)

C=C

C=C

P=O
P-OH

P-OH

C C-O
C=C

C-OH
C-O

P=O
P-OH
C=C

NH3+

NH3+
P-OH

P-O

N-H

N-H
TI
HBzA
2PACz
Mixture
AR

900 1000 1100 1500 1600 1700

c Wavenumber (cm -1)

C 1s C 1s C 1s
2000 C - H, C = C C - H, C = C C - H, C = C
C-N 600 C-N 600 C-O
D

C-P C-O C-N

1500 C-P
Intensity (Cps)

Intensity (Cps)

Intensity (Cps)

ITO/2PACz after ITO/2HBzA after

washing with ITO/2PACz+HBzA
E

washing with after washing with

DMF/DSMO DMF/DSMO
1000 DMF/DSMO
300 300
AT

500

0
ER

290 288 286 284 282 280 292 290 288 286 284 282 292 290 288 286 284 282
Binding energy (eV) Binding energy (eV) Binding energy (eV)
EL

Extended Data Fig. 1

C
AC
 W
IE
ITO/2PACz+HBzA ITO/2PACz+HBzA (washing)
10
ITO/2PACz+HBzA
Relative frequency (%)

ITO/2PACz+HBzA washing

EV
5

0
5.00 5.05 5.10 5.15 5.20
Work Function (eV)

ITO/2PACz ITO
10
ITO/2PACz 10

PR
ITO only
Relative frequency (%)

Relative frequency (%)

5
5

0
5.30 5.35 5.40 5.45 5.50 0
4.85 4.90 4.95 5.00 5.05
Work function (eV)
Work Function (eV)

ITO/2PACz/3D- ITO/2PACz/2D-HBzA/3D-
perovskite (washing) perovskite (washing)

LE
10
ITO/2PACz washing (after perovskite)
Relative frequency (%)

0
5.30 5.35 5.40 5.45 5.50
Work Function (eV)

C
TI
AR
E D
AT

a
ITO/2PACz+HBzA ITO/2PACz+HBzA (washing)
ER

10
Relative frequency (%)

ITO/2PACz+HBzA
ITO/2PACz+HBzA washing

0
5.00 5.05 5.10 5.15 5.20
Work Function (eV)

b c
ITO/2PACz ITO
10
ITO/2PACz 10
Relative frequency (%)

Relative frequency (%)

ITO only

5
5
EL

0
5.30 5.35 5.40 5.45 5.50 0
4.85 4.90 4.95 5.00 5.05
Work function (eV)
Work Function (eV)

d e
ITO/2PACz/3D- ITO/2PACz/2D-HBzA/3D-
perovskite (washing) perovskite (washing)
10
Relative frequency (%)

ITO/2PACz washing (after perovskite)

ITO/2PACz+2DHBzA washing (after perovskite)

0
5.30 5.35 5.40 5.45 5.50
C

Work Function (eV)

Extended Data Fig. 2

AC
 W
IE
EV
a Bottom-side 2D/3D heterojunction

25
1170 84

PR
24
1160
82
VOC (mV)
PCE (%)

23 1150

FF (%)
80
22 1140
78
1130
21

LE
1120 76
20
1:0.5 1:1 1:2 1:0.5 1:1 1:2 1:0.5 1:1 1:2

2PACz:HBzA ratio 2PACz:HBzA ratio 2PACz:HBzA ratio

C
b 2D/3D at bottom with various ligands
24
1160
TI 84

23 82
1150
VOC (mV)
PCE (%)

FF (%)
AR

80
22 1140
78
1130
21
76
1120
20 74
BzA PEA BA HMA BzA PEA BA HMA BzA PEA BA HMA
D

1:1 ratio with 2PACz 1:1 ratio with 2PACz 1:1 ratio with 2PACz

c Top-side 3D/2D heterojunction

E

25
1180 84
AT

24
1160
23 81
VOC (mV)
PCE (%)

FF (%)

1140
22
78
ER

21 1120

20 1100 75

0 nm 7 nm 15 nm 25 nm 0 nm 7 nm 15 nm 25 nm 0 nm 7 nm 15 nm 25 nm
EL

PbI2 thickness PbI2 thickness PbI2 thickness

Extended Data Fig. 3

C
AC
 W
IE
EV
a

PR
-3.66 eV

LE
-4.04 eV
Energy (eV)

-4.05 eV -4.10 eV
-4.87 eV -4.69 eV
5.17eV
EF

C
- .57 eV
5
-5.80 eV -5.76 eV
-6.15 eV
TI
-6.40 eV
AR

2PACz 2D perovskite 3D perovskite 2D perovskite C60

Hybrid

b c d

3D perovskite 3D perovskite
D

UPS IPES
2D perovskite (top) 2D perovskite (top)
2D perovskite (bottom) 2D perovskite (bottom)
Normalized Intensity (a.u)
Normalized Intensity (a.u)

Counts (a.u)

Eg = 2.10 eV
AT

VB-WF = 0.88 eV
VB-WF = 0.95 eV 0.95 eV 1.15 eV
16.53 eV

16.35 eV

16.05 eV

VB-WF = 0.89 eV
ER

2D perovskite
n=2

18.0 17.5 17.0 16.5 16.0 6 5 4 3 2 1 0 -1 6 4 2 0 -2 -4

Binding Energy (eV) Binding Energy (eV) Binding energy w.r.t Ef (eV)
EL

Extended Data Fig. 4

C
AC
 W
IE
EV
a b
25

PR
25
20 Double-side
20
Control

Count
15 15

LE
10
10
5
5
Control, Reverese scan
C 0
19 20 21 22 23 24 25 26
TI
0 Control, Forward scan PCE (%)
Double-side, Reverese scan
Double-side, Forward scan
AR

-5
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Voltage (V)
c d
25 25
D

20 20
E
Count
Count

15 15
AT

10 10
5 5
0 0
ER

1.08 1.11 1.14 1.17 1.20 1.23 72 74 76 78 80 82 84 86

VOC (V) FF (%)
EL

Extended Data Fig. 5

C
AC
 W
IE
EV
a b
10

PR
PL intensity (a.u.)
PLQY (%)

LE
2D/3D/2D (ave = 2.69 s)
2D/3D(ave = 1.34 s)

C
3D/2D (ave = 1.73 s)
Control 3D (ave = 0.97 s)
0.1
TI Fitting
D

��

��
l3

/3

/2

/2

0 1 2 3 4 5 6
/C

/C
2D

3D

D
tro

D
/3

l3

/2
on

2D

Time (s)
tro
C

/3

AR
on

2D
C

c 102 d 102 e
Control SCLC region Control SCLC region
Double-side 2D/3D Double-side 2D/3D
Current density (mA cm )
Current density (mA cm-2)

-2

100 100
tDOS (cm³ eV-¹)

1017
E

VTFL = 0.71 V
VTFL = 0.75 V
-2
10 10-2
AT

VTFL = 1.80 V VTFL = 1.78 V

Trap-filling Trap-filling
Ohmic region limited region Ohmic region limited region
1016 Control 3D
10-4 10-4
Bottom-side 2D/3D
ER

Electron-only device Hole-only device Double-side 2D/3D

0.01 0.1 1 10 0.01 0.1 1 10 0.20 0.25 0.30 0.35 0.40 0.45
Applied Voltage (V) Applied Voltage (V) E (eV)
EL

Extended Data Fig. 6

C
AC
 W
IE
EV
PR
LE
a 2.0
b c
2.0

C
100 hrs -4
100 hrs Control 3D perovskite Double-side 2D/3D
50 hrs 50 hrs -5
Absorbance (a.u.)

Absorbance (a.u.)

1.5 1.5 Ea = 0.27 eV

TI
20 hrs
20 hrs
0 hrs 0 hrs -6

ln ( T)
1.0 1.0 -7
AR

-8
Ea = 0.47 eV
0.5 0.5 -9
Control
-10
Double-side
0.0 0.0 -11
D

300 350 400 450 500 550 600 300 350 400 450 500 550 600 36 38 40 42 44 46 48
Wavelength (nm) Wavelength (nm) 1/kT (eV)
E
AT

Extended Data Fig. 7

ER
EL
C
AC