Articles

https://doi.org/10.1038/s41566-020-00732-4

Comprehensive defect suppression in perovskite

nanocrystals for high-efficiency light-emitting
diodes
Young-Hoon Kim   1,2,10, Sungjin Kim1,2,10, Arvin Kakekhani   3,10, Jinwoo Park   1,2, Jaehyeok Park   4,
Yong-Hee Lee1, Hengxing Xu5, Satyawan Nagane6, Robert B. Wexler   3, Dong-Hyeok Kim1,2,
Seung Hyeon Jo1,2, Laura Martínez-Sarti   7, Peng Tan   3,8, Aditya Sadhanala6,9, Gyeong-Su Park1,
Young-Woon Kim1, Bin Hu   5, Henk J. Bolink   7, Seunghyup Yoo4, Richard H. Friend   6,
Andrew M. Rappe   3 ✉ and Tae-Woo Lee   1,2 ✉

Electroluminescence efficiencies of metal halide perovskite nanocrystals (PNCs) are limited by a lack of material strategies that
can both suppress the formation of defects and enhance the charge carrier confinement. Here we report a one-dopant alloying
strategy that generates smaller, monodisperse colloidal particles (confining electrons and holes, and boosting radiative recom-
bination) with fewer surface defects (reducing non-radiative recombination). Doping of guanidinium into formamidinium lead
bromide PNCs yields limited bulk solubility while creating an entropy-stabilized phase in the PNCs and leading to smaller PNCs
with more carrier confinement. The extra guanidinium segregates to the surface and stabilizes the undercoordinated sites.
Furthermore, a surface-stabilizing 1,3,5-tris(bromomethyl)-2,4,6-triethylbenzene was applied as a bromide vacancy healing
agent. The result is highly efficient PNC-based light-emitting diodes that have current efficiency of 108 cd A−1 (external quan-
tum efficiency of 23.4%), which rises to 205 cd A−1 (external quantum efficiency of 45.5%) with a hemispherical lens.

M
etal halide perovskites (MHPs) with the general ABX3 recombination and defects are reduced by surface-binding ligands;
formula (where A is an organic or inorganic cation, B is however, this strategy has limitations: (1) the dynamic nature
a metal cation and X is a halide anion) have narrow emis- of ligand–surface bonds and (2) steric hindrance effects that can
sion spectra (full-width at half-maximum ≈ 20 nm), allowing them leave undercoordinated sites unpassivated and prone to defect for-
to achieve high colour purity, a tunable emission wavelength range mation8,9. B-site cation engineering10, X-site anion exchange11 and
(400 nm ≤ λ ≤ 780 nm) and low-cost solution processability1–6. As surface-binding ligand engineering12 have also been used, leading to
a result, MHPs have been regarded as promising light emitters5–12. a maximum current efficiency of 76.8 cd A−1 (with an EQE of 17.1%
Such properties have also led to perovskite light-emitting diodes ph/el) in green emission24 and current efficiency of 10.6 cd A−1 (EQE
(PeLEDs), which have exhibited a tremendous increase in electrolu- of 21.3% ph/el considering angular electroluminescence distribu-
minescence efficiency13–15. These improvements have been obtained tion) in red emission11 in PeLEDs based on CsPbX3.
by (1) increasing the radiative recombination rate16 and (2) lower- The majority of research into PeLEDs is based on all-inorganic
ing the non-radiative recombination rate17. The radiative recom- PNCs. It has been shown that the orientational freedom of the
bination rate of the charge carriers has been increased by spatially liquid-like dipoles associated with the organic cations can reduce the
confining the electrons and holes in small perovskite polycrystalline charge recombination rates25–28. Although beneficial for photovoltaic
nanograins3, low-dimensional crystals18–20 or colloidal perovskite applications, this is undesirable for LEDs29. The all-inorganic MHPs
nanocrystals (PNCs)4,5. More effective strategies for suppression of contain spherical atomic A-site cations with zero dipole moment,
defects and associated non-radiative recombination are required to leading to enhancement of charge recombination. Nevertheless,
surpass the state of the art and further increase the electrolumines- there are shortcomings associated with the atomic A-site cations:
cence efficiency. a small phase space for tunability and a lack of lattice-stabilizing
In perovskite polycrystalline bulk films, stoichiometric control of directional hydrogen bonds. At the same time, there has been a lack
precursors3, use of passivation agents18–21 and post-treatments14,22,23 of comprehensive material design strategies to passivate the surface
were shown to lead to a maximum current efficiency of 78 cd A−1 defects and confine charge carriers inside of the nanocrystals.
(with an external quantum efficiency (EQE) of 20.3% ph/el based Here we propose a simple and rational PNC design to simul-
on Lambertian assumption)18. In colloidal PNCs, non-radiative taneously stabilize the undercoordinated sites at the surface while

1
Department of Materials Science and Engineering, Seoul National University, Seoul, Republic of Korea. 2School of Chemical and Biological Engineering,
Institute of Engineering Research, Research Institute of Advanced Materials, Nano Systems Institute, Seoul National University, Seoul, Republic of
Korea. 3Department of Chemistry, University of Pennsylvania, Philadelphia, PA, USA. 4School of Electrical Engineering, Korea Advanced Institute of
Science and Technology, Daejeon, Republic of Korea. 5Department of Materials Science and Engineering, University of Tennessee, Knoxville, TN, USA.
6
Cavendish Laboratory, University of Cambridge, Cambridge, UK. 7Instituto de Ciencia Molecular, Universidad de Valencia, Paterna, Spain. 8Department of
Physics, Harbin Institute of Technology, Harbin, China. 9Clarendon Laboratory, Department of Physics, University of Oxford, Oxford, UK. 10These authors
contributed equally: Young-Hoon Kim, Sungjin Kim, Arvin Kakekhani. ✉e-mail: [email protected]; [email protected]

Nature Photonics | www.nature.com/naturephotonics

Articles Nature Photonics

a b
FA1–xGAx GA

Pb

Br

C
N
H

c d 100

80

60
Intensity (a.u.)

Energy (meV)
40

20

–20
10 15 20 25 30 35 40 45 50 0 0.2 0.4 0.6 0.8 1.0
2θ (°) GA content

Fig. 1 | Structure of FA1–xGAxPbBr3 PNCs. a, A schematic illustration of the ligand-stabilized FA1–xGAxPbBr3 PNC structure. b, The modelled bulk crystal
structure of FAPbBr3. c, Measured (black) and DFT-simulated (red) XRD patterns of FAPbBr3. d, Density functional theory formation energy (blue, relative
to the precursors), internal (enthalpic) energy (red) and configurational entropic stabilization energy (black) of FA1–xGAxPbBr3 crystals (0 ≤ x ≤ 1).

improving the confinement of charge carriers inside the nano- Beyond concentrations of ~12.5%, enthalpy wins over entropy and
crystals by employing a zero-dipole guanidinium cation (CH6N3+; drives the surplus GA to the surface of the PNCs.
GA+)30,31, which provides lattice-stabilizing effect via its hydrogen As x is increased from 0 to 1, the size of FA1–xGAxPbBr3 PNCs
bonds. We exploit fine substitutional doping of formamidinium decreases from ~10 nm to ~5.1 nm (Fig. 2a and Supplementary Fig.
(CH5N2+; FA+) lead bromide (FAPbBr3) by single GA+ cations in 2). The decrease in size is also confirmed by observing a general
colloidal PNCs instead of cation alloying approaches in perovskite trend for the relative increase of the peak intensity ratio (of the
polycrystalline bulk films that have been used in solar cells32,33. The organic ligands to inorganic perovskite components) in the X-ray
increased surface stability is driven by the extra amino group in the photoelectron spectroscopy measurements (see Supplementary
GA+ due to its extra hydrogen bonds34 and more uniformly distrib- Fig. 3 and the associated discussion for more details). Increasing
uted positive charges35. Decylamine and oleic acid are also used as the GA concentration increases the enthalpy of the bulk of the
surface-binding ligands (Fig. 1a), providing an additional level of PNCs, thereby destabilizing them. To compensate for this energy
surface stabilization. We show that although adding guanidinium penalty, the PNCs increase their surface-to-bulk ratio by decreasing
(GA) to the particle is penalized from an internal energy point of their size. The PNCs are stabilized by the surrounding acid/amine
view due to its larger size, the gain in configurational entropy36 sta- ligands. In samples that lack GA, the optimal surface-to-bulk ratio
bilizes low concentrations of GA. We further reduce non-radiative is determined by the balance between surface formation energy
recombination by applying 1,3,5-tris(bromomethyl)-2,4,6-tri- cost and the stabilization imparted by surface-binding ligands. As
ethylbenzene (TBTB) overcoat, which can heal the leftover halide can be seen in Fig. 2b,c, the latter is a result of both chemical point
vacancies. We ultimately achieved a current efficiency of 108 cd A−1, interactions and dispersion interactions (involving alkyl and alke-
further increased to 205 cd A−1 by employing a hemispherical lens. nyl groups). The point interactions involve the acid (–COOH) or
amine (–NH2) groups and consist of both lone-pair interactions
Results (accompanied by lone-pair polarization)35,40 and hydrogen bonds.
Structural properties. We begin by studying how GA doping influ- Guanidinium alters this balance and drives a tendency to increase
ences the structure of FAPbBr3. We construct a computational model the surface-to-bulk ratio, leading to an increased charge carrier con-
of FAPbBr3 PNCs (see Supplementary Fig. 1 and the Methods for finement without inducing more defects on the PNC surfaces (as
details). The fully ab initio-extended bulk model (Fig. 1b) can rep- explained in the next section).
resent the FAPbBr3 PNCs and reproduce their experimental X-ray Our density functional theory (DFT) calculations indicate that,
diffraction (XRD) patterns (Fig. 1c). We calculate the formation free beyond low concentrations of GA, these cations have an energetic
energy (from precursors) of FA1−xGAxPbBr3 as x increases from 0 to preference to migrate to the surface (Fig. 2d). As the concentration
1, taking into account both enthalpic and configurational entropic of GA increases, the increase in the surface-to-bulk ratio ensures
contributions (Fig. 1d). Guanidinium is larger than FA and beyond the presence of enough surface sites, at which most of the GA cat-
the tolerance of the (lead-bromide-based) perovskite structure37–39, ions can reside (Supplementary Table 1). The inference that GA can
therefore enthalpy does not preferentially drive GA to the inside of only be incorporated into the bulk of the perovskite in low con-
the perovskite. Nonetheless, small concentrations (~12.5%) of GA centrations is supported by the GA concentration-dependent bulk
can still be dissolved in the structure due to entropy stabilization36. moduli (calculated via elastic constants using the ab initio stress–

Nature Photonics | www.nature.com/naturephotonics

Nature Photonics Articles
a b
0

0.05

Number of particles
0.1

0.2

0.3

0.6

0.8

0 2 4 6 8 10 12
Size (nm)

c d

e 0
f 0
1.0 0.05 0.05
0.1 0.1
Photoluminescence

0.8 0.15 0.15

Intensity (a.u.)
intensity (a.u.)

0.2
0.2
0.6 0.25
0.25
0.3
0.3
0.35
0.4 0.35
0.6
0.8 0.6
0.2 1 0.8
1
0
460 480 500 520 540 560 580 600 10 15 20 25 30 35 40 45 50
Wavelength (nm) 2θ (°)

Fig. 2 | Structural and photophysical effects of GA on FA1–xGAxPbBr3 PNCs. a, Experimentally measured size distribution of FA1–xGAxPbBr3 PNCs at
0 ≤ x ≤ 1. b, Three-dimensional charge density difference plots, showing the point chemical interactions between the perovskite surface and acid (left)
or amine (right) ligands. The hydrogen and lone-pair bonds35,40 can be noted by the magenta lobes along the yellow dashed lines. Cyan and magenta
show regions of space that are depopulated or populated by electrons, respectively (as a result of the surface–ligand bonding). c, Yellow regions depict
dispersion interactions among the surface and surface-bound ligands. This image was generated using the Nelson complexity index (NCI; isovalue = 0.6)42.
d, The simulated crystal structure of a GA-rich surface of FA1–xGAxPbBr3 crystals is shown from the side (left) and top views (top-right). At the bottom
right, a two-dimensional charge density difference plot (in greyscale)—projected on the top bromide-layer plane—shows formation of hydrogen bonds
between GA and bromide (white lobes along the yellow dashed lines). e,f, A photoluminescence spectrum (e) and XRD patterns (f) of FA1–xGAxPbBr3
PNCs at 0 ≤ x ≤ 1.

strain methodology), which shows a sudden softening at higher to an amorphous-like (or with low crystallinity) phase in which
GA concentrations (see Supplementary Fig. 4 and Supplementary small core perovskite crystals are fully covered by GA and then to
Discussion 1 for more details). On the surface, the cages formed by a non-perovskite guanidinium lead bromide phase (see the discus-
the PbBr6 octahedra are broken and lack a confining cap, the GA sion below the Supplementary Table 1).
can therefore more favourably fit into the surface sites. The extra
amino group can effectively stay above the surface and favourably Defect suppression via GA doping. The surface GA population
interact with the surface bromide groups (Fig. 2d). leads to effective defect passivation. The photoluminescence quan-
The photoluminescence spectrum experiments corrobo- tum efficiency (PLQE) improves from 79.7% for FAPbBr3 to 93.3%
rate these inferences. The photoluminescence spectrum of the for FA0.9GA0.1PbBr3 PNCs (in solution) (Fig. 3a). FA0.9GA0.1PbBr3
FA0.95GA0.05PbBr3 PNCs (Fig. 2e, Supplementary Fig. 5a) is slightly PNC films also show a PLQE of 92.45% even without using core–
red-shifted compared with that of FAPbBr3 PNCs. This change shell structures both in film and solution. Defect suppression is
confirms that small concentrations of GA incorporate into FA1– directly confirmed by deep-level transient spectroscopy (DLTS).
xGAxPbBr3 crystals and expand the lattice. This is also confirmed Here, the device architecture includes indium tin oxide (ITO)/
by XRD patterns (Supplementary Fig. 6). As x increases (≥0.1), buffer hole-injection layer (Buf-HIL (ref. 2))/FA1–xGAxPbBr3 PNCs
photoluminescence spectra of FA1–xGAxPbBr3 gradually blue-shift (here, x = 0, 0.1)/1,3,5-tris(N-phenyl benzimidazole-2-yl)benzene
while the intensity of the XRD peaks drops due to the size decrease (TPBI)/LiF/Al (Fig. 3b). The DLTS spectrum shows a minimum
(Fig. 2f). At x > 0.6, the structure of the PNCs begins to change, first at 250 K, which corresponds to a defect energy transition level of

Nature Photonics | www.nature.com/naturephotonics

Articles Nature Photonics

a 100 b 0.10
0
0.1
80

DLTS intensity (ΔC/C)

0.05

60
PLQE (%)
FA0.9GA0.1PbBr3/FAxGA1–x FA0.9GA0.1PbBr3/GA 0
FAPbBr3
core/shell core/shell
40

–0.05
20

0 –0.10
0 0.2 0.4 0.6 0.8 1.0 50 100 150 200 250 300
GA concentration (x) Temperature (K)

c d
Photoluminescence intensity (a.u.)

0 500 1,000 1,500 2,000

Lifetime (ns)

Fig. 3 | Defect analysis of FA1–xGAxPbBr3 PNCs. a, Photoluminescence quantum efficiency and schematic crystal structures (inset) of FA1–xGAxPbBr3 PNCs.
b, Deep-level transient spectroscopy data for PeLEDs based on FA1–xGAxPbBr3 PNCs at x = 0 (black) and x = 0.1 (red). c, Photoluminescence lifetimes for
FA1–xGAxPbBr3 PNCs at x = 0 (black), 0.1 (blue), 0.6 (gold) and 1 (grey). d, A DFT-derived FA1–xGAxPbBr3 quantum dot structure with a GA shell and FA core.

0.3 to 0.4 eV. This signal attenuates at x = 0.1 and the calculated (chain-length-dependent) dispersion interactions (Fig. 2c). These
defect density Nt declines from 4.93 × 1012 cm−3 to 3.10 × 1012 cm−3 interactions are the root of the surface-stabilizing effect of these
(Supplementary Table 2). ligands (see Methods for details). Guanidinium doping slightly
Guanidinium doping extends the photoluminescence lifetime in weakens the binding of these ligands (see Supplementary Fig.
FA0.9GA0.1PbBr3 (~113 ns) compared with FAPbBr3 (~88 ns) (Fig. 3c, 9, Supplementary Fig. 10 and the associated discussion, and the
Supplementary Fig. 5b). Our defect suppression scheme increases: Methods); this is another indication that incorporating GA stabi-
(1) the critical bias Vc, as measured by the magnetophotocurrent in lizes the surface internally and decreases its reactivity toward exter-
our PeLEDs (Supplementary Fig. 7a,b); (2) the thermal stability, as nal agents. At x = 0.1, the PNCs are relatively large (Fig. 2a), the GA
determined by thermogravimetric analysis (Supplementary Fig. 7c); induces both bulk and surface stabilization, and is uniformly dis-
(3) the photostability, with a smaller degree of attenuation in the tributed. Most of the surface is still covered by FA and, due to steric
shoulder of the photoluminescence spectrum peak at lower-energy hindrance, ligands can only bind to a limited number of surface sites
states (Supplementary Fig. 7d–f); and (4) the exciton binding and thus selectively bind to FA-sites, unaffected by the added GA
energy, as calculated from the temperature-dependent photolumi- (Supplementary Table 1). The positive effect of 10% GA is due to
nescence (Supplementary Fig. 7g–j and Supplementary Fig. 8). increased surface stability, reduced defects and increased electron–
Guanidinium stabilizes the PNC surfaces by increasing the hole confinement, in addition to the bulk entropy stabilization.
number of hydrogen bonds (as a result of the extra amino group By increasing the GA content and the surface-to-bulk ratio, the
relative to FA) with the undercoordinated surface bromide. We bulk-stabilizing effect of GA attenuates; nevertheless, as the PNC
have determined that the surface of the PNC has an AX termination surfaces become dominantly GA terminated, its surface-stabilizing
(see Supplementary Fig. 1 and Supplementary Discussion 2). We effect increases (Supplementary Table 1). Consequently, at x = 0.6
define a cohesive energy metric for the top surface AX layer (see (with a fully GA-covered surface), the PNC still showed a high
Supplementary Discussion 2) and use it to evaluate the increase in PLQE of 92.4% in solutions and 91.55% in films. At x > 0.6, the crys-
GA-driven surface stability. The change in this quantity (relative to tallinity of the sample decreased, the perovskite phase was desta-
no-GA surfaces) is around −350 meV (per GA) for extended sur- bilized, new diffraction peaks appeared and the PLQE dropped
faces (as in the basal planes of large PNCs) and around −390 meV (Figs. 2f and 3a). Such a departure from the perovskite structure
for the very small quantum dots that we have directly modelled (Fig. on increasing GA content is related to the tendency of guanidin-
3d), due to their greater undercoordination (for example, in corners ium halide perovskites to crystalize in non-perovskite structures,
and edges) relative to extended surfaces. Here, a negative value for a which was previously addressed in the literature for iodoplum-
change in cohesive energy means greater surface stabilization. bates39. Our calculations also confirm this for the bromoplumbates
Density functional theory simulations of surface-binding (see Supplementary Fig. 11 and Methods for details). Our PNC
ligands reveal relatively strong point interactions (−0.6 to materials design strategy for light-emitting diodes (LEDs) can be
−0.7 eV) with the surface (Fig. 2b), accompanied by additional extended to halide perovskites with other halides and cations; to

Nature Photonics | www.nature.com/naturephotonics

Nature Photonics Articles
a b
102 105
0
0.05 0.3
101 0.1 0.35

Current density (mA cm–2)

0.15 0.6 103

Luminance (cd m–2)

100 0.2 0.8
0.25 1

10–1 101

10–2
10–1
10–3

10–4 10–3
0 1 2 3 4 5 6 2 3 4 5 6
Voltage (V) Voltage (V)

c 100 d 25

80 20
Current efficiency (cd A–1)

60 15

EQE (%)
40 10

20 5

0 0
2.4 2.8 3.2 3.6 4.0 2.4 2.8 3.2 3.6 4.0
Voltage (V) Voltage (V)

e 100 f
Electroluminescence intensity (a.u.)

1.0
80
P o w e r e f f i c i e n c y (lm W–1)

0.8

60
0.6

40
0.4

20 0.2

0 0
2.4 2.8 3.2 3.6 4.0 400 440 480 520 560 600
Voltage (V) Wavelength (nm)

Fig. 4 | Characteristics of PeLEDs based on FA1–xGAxPbBr3 PNCs. a–f, Current density (a), luminance (b), current efficiency (c), external quantum
efficiency (d), power efficiency (e) and electroluminescence spectrum (f) of PeLEDs based on FA1–xGAxPbBr3 PNCs (0 ≤ x ≤ 1).

showcase this, we have experimentally studied methylammonium GA can suppress both the bulk and surface defects, with minimal
lead bromide (MAPbBr3) PNCs. About 10% GA can be incorpo- changes to the dimensionality and composition, and can simulta-
rated into MAPbBr3 crystals, inducing red-shifted photolumi- neously increase the radiative recombination of the colloidal PNCs.
nescence. Beyond ~10%, GA is driven to the surface and reduces The use of GA to passivate the defects in MAPbI3-based poly-
the size of the PNCs, inducing blue-shifted photoluminescence crystalline bulk films for solar cells has been reported32,33. In our
(Supplementary Fig. 12). The GA leads to effective defect passiv- work we: (1) synthesize colloidal PNCs rather than perovskite
ation and improved PLQE in MAPbBr3 PNCs. To further indicate polycrystalline bulk films, thus the GA—in addition to passivating
the novel effects induced by the doped GA cations on colloidal defects—simultaneously controls the size of the PNCs and enhances
PNCs, we experimentally studied FAPbBr3 PNCs with other types radiative recombination; (2) understand why GA can be incorpo-
of organic cations such as butylammonium bromide, octylammo- rated into the PNC in small concentrations (~10%), above which
nium bromide, benzylammonium bromide and phenethylammo- it migrates to the surface outermost layer of PNC; and (3) illustrate
nium bromide (Supplementary Fig. 13). As the content of these competing enthalpic and entropic effects for GA incorporation in
organic cations increases (>0.2), the PNCs show limited improve- the bulk and show that such a mixed phase is entropy stabilized.
ment in PLQE (<86%) and several different shoulder peaks in pho-
toluminescence spectrum due to various sizes or compositions of Achieving highest LED efficiencies. The electroluminescence
crystals. These results indicate that GA—featuring only one extra efficiency of our PeLEDs (structure: ITO/Buf-HIL/FA1–xGAxPbBr3
amino group and mainly residing on the surface of the colloidal PNCs/TPBI/LiF/Al) increased from 61.3 to 95.7 cd A−1 (Fig. 4) as x
PNCs—is a well-controlled materials alloying concept. The doped increased from 0 to 0.1. As x increases above 0.1, the current densi-

Nature Photonics | www.nature.com/naturephotonics

Articles Nature Photonics

Br vacancy filling

Absorption TBTB debromination Hydrogenation

ΔE = –1.89 eV ΔE = –1.70 eV

b c 1,000

Current efficiency (cd A–1)

LiF/Al
100
TPBI
TBTB
FA0.9 GA Pb 10
0.1 Br3 NCs

Buf-HIL 1

ITO
0.1
0 5,000 10,000 15,000 20,000 25,000
Luminance (cd m–2)

d 100 e 1,000
Power efficiency (lm W–1)

100
10
EQE (%)

10

1
Without TBTB 1
With TBTB
With TBTB and HSL
0.1 0.1
0 5,000 10,000 15,000 20,000 25,000 0 5,000 10,000 15,000 20,000 25,000
Luminance (cd m–2) Luminance (cd m–2)

f 0° g
20
Simulation
30°
Measured
16
Lambertian
Number of devices

12
60°
8

0
90° 80 90 100 110
Maximum CE (cd A–1)
Fig. 5 | Characteristics of PeLEDs with a TBTB interlayer. a, DFT-derived mechanism for bromide vacancy healing process driven by TBTB molecules on
GA-terminated FAPbBr3 surface. The debrominated radical can then achieve greater stability via further hydrogenation. The reaction energies (negative
means exothermic) are shown by ΔE. b, Device structure of (x = 0.1) PeLEDs with TBTB interlayer. c–e, Current efficiency (c), external quantum efficiency
(d) and power efficiency (e) of PeLED devices (x = 0.1) with or without a TBTB interlayer and a hemispherical lens (HSL). The insets in c are photographs
of operating devices with TBTB, without (left) and with (right) a hemispherical lens (scale bar, 1 mm). f, Angular intensity profiles for simulated emission,
measured electroluminescence emission and Lambertian emission according to the viewing angles. g, Current efficiency histogram and photograph (inset;
scale bar, 1 cm) of an operating large-area PeLEDs (3 cm × 3 cm pixel) with TBTB interlayer (x = 0.1).

ties, luminance and electroluminescence efficiencies of the PeLEDs increased surface-to-bulk ratio (Fig. 2a). The increased number of
gradually decrease; this trend can be ascribed to an increased ratio these insulating ligands hinders charge transport to the PNCs and
of insulating organic ligands surrounding the PNCs due to the thereby reduces the electroluminescence efficiency41.

Nature Photonics | www.nature.com/naturephotonics

Nature Photonics Articles
a 160 b 160 4.0
FA0.9GA0.1PbBr3 PNCs
FAPbBr3 PNCs 3.8
140 140

Voltage (V)
3.6

3.4
120 120

L /L0 (%)

L /L0 (%)
3.2

100 100
3.0
0 20 40 60 80 100 120 140 160
Time (min)

–2
@100 cd m (0.009 mA)
80 80 –2
@200 cd m (0.017 mA)
–2
@500 cd m (0.027 mA)
60 60 –2
@1,000 cd m (0.053 mA)

0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160

Time (min) Time (min)

c 160 d 160
Encapsulated Without annealing
In air (relative humidity: 10%) Annealing at 40 °C
140 140
In air (relative humidity: 50%) Annealing at 60 °C
Annealing at 80 °C
120 120
L /L0 (%)

L /L0 (%)
100 100

80 80

60 60

0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160

Time (min) Time (min)

Fig. 6 | Device lifetime of PeLEDs. a, The device lifetimes of PeLEDs based on FA0.9GA0.1PbBr3 PNCs and FAPbBr3 PNCs at an initial luminance (L0) of 100
cd m–2. b, The device lifetimes and operating voltages (inset) of PeLEDs based on FA0.9GA0.1PbBr3 PNCs under different initial measurement luminance
(applied current). c, The device lifetimes of PeLEDs based on FA0.9GA0.1PbBr3 PNCs at an initial luminance of 100 cd m–2 under various environmental
conditions. d, The device lifetimes of PeLEDs based on FA0.9GA0.1PbBr3 PNCs at an initial luminance of 100 cd m–2 with different thermal annealing
temperatures of devices.

We extended our strategy to further suppress surface defects and devices—without and with an outcoupling hemispherical lens—
minimize non-radiative recombination losses in the FA0.9GA0.1PbBr3 can reach EQEs of 30.2% and 55.61%, respectively (provided that
PNC films by introducing TBTB interlayers (~5 nm) between the the PLQE of PNC films is ~1 and absorption loss of the emitting
PNC films and the TPBI layers, which can act as a bromide vacancy layer is negligible) (Supplementary Fig. 15). These are the highest
healing agent and suppress surface defects in the PNC films. As efficiencies so far reported among PeLEDs based on either poly-
can be seen in Fig. 5a, a TBTB molecule can readily deposit its crystalline bulk films or PNCs and surpass even the highest current
bromine into a surface bromide vacancy. The debrominated radi- efficiency of conventional III–V and II–VI inorganic quantum dot
cal can then become hydrogenated in the environment containing LEDs (Supplementary Fig. 16 and Supplementary Tables 3 and 4).
acid ligands surrounding the PNCs and gets further stabilized (see We also fabricated large-area PeLEDs (3 cm × 3 cm pixel) that used
Supplementary Discussion 2 for more details). These defect healing FA0.9GA0.1PbBr3 PNCs (Fig. 5g inset) to demonstrate the promise
effects are confirmed by a decreased defect peak intensity and Nt in the of using organic–inorganic hybrid colloidal PNCs in solid-state
DLTS measurements, by an increased photoluminescence lifetime, lighting devices and displays. Device lifetime of our PeLEDs were
and by a blue-shifted photoluminescence spectrum (Supplementary measured under various operational and environmental conditions
Fig. 14a–c, Table 2). The insulating TBTB interlayers on the PNC (Fig. 6, Supplementary Fig. 17). PeLEDs based on FA0.9GA0.1PbBr3
films can also improve the charge balance in PeLEDs by retarding PNCs and TBTB interlayers showed a lifetime (when the initial
electron injection into emitting layers. This is confirmed by mea- luminance drops to 50%) of 132 min with an initial luminance of
suring the electron current densities in electron-only devices (ITO/ 100 cd m–2, which is improved by a factor of ~5.3 relative to PeLEDs
polyethylenimine ethoxylated (10 nm)/FA0.9GA0.1PbBr3 PNCs with based on FAPbBr3 PNCs (lifetime = 25 min) and ~8 relative to the
or without TBTB/TPBI/LiF/Al), by measuring hole current densi- previously reported highest-efficiency PeLEDs based on FAPbBr3
ties in hole-only devices (ITO/Buf-HIL/FA0.9GA0.1PbBr3 PNC films PNCs (lifetime = 18 min)24.
with or without TBTB/Tris(4-carbazoyl-9-ylphenyl)amine (50 nm)/
MoO3/Al) (Supplementary Fig. 14d,e). Furthermore, capacitance– Conclusions
voltage characteristics in PeLEDs corroborate the delay in electron We developed a comprehensive strategy that leads to a consider-
injection on inserting the TBTB interlayer (Supplementary Fig. 14f). able increase in the luminescence efficiency of PNCs. This strategy
PeLEDs based on FA0.9GA0.1PbBr3 PNCs and TBTB interlayers reach makes it possible to simultaneously (1) decrease the non-radiative
a maximum current efficiency of 108 cd A−1 and an EQE = 23.4%, charge recombination by comprehensive defect suppression and
calculated by using the full angular electroluminescence distribution bulk entropy stabilization and (2) increase the radiative recombi-
(Fig. 5b–f). The distribution of current efficiency obtained among nation of charge carriers due to increased excitonic confinement.
53 devices shows great reproducibility (Fig. 5g). We further improve First, we used substitutional doping of GA into FAPbBr3 PNCs to
the efficiency of our PeLEDs (to current efficiency = 203 cd A−1 and incorporate an optimal proportion of GA cations into the structure.
EQE = 45.5%) by placing a light outcoupling hemispherical lens on The GA cations can reside in the bulk of the PNC in low concen-
the emitting glass substrate. Optical simulations also verify that our trations (~10%). The surplus GA then accumulates on the surface

Nature Photonics | www.nature.com/naturephotonics

Articles Nature Photonics

of the PNC. Guanidinium incorporation provides bulk entropy 18. Lin, K. et al. Perovskite light-emitting diodes with external quantum
stabilization, surface stabilization (by additional hydrogen bonding efficiency exceeding 20 per cent. Nature 562, 245–248 (2018).
19. Xiao, Z. et al. Efficient perovskite light-emitting diodes featuring
contributed by their extra amino group), and better electron–hole nanometre-sized crystallites. Nat. Photon. 11, 108–115 (2017).
confinement. A GA to FA ratio of 0.1 was shown to maximize the 20. Zhao, B. et al. High-efficiency perovskite–polymer bulk heterostructure
electroluminescence efficiency. Moreover, a TBTB overcoat was light-emitting diodes. Nat. Photon. 12, 783–789 (2018).
introduced to heal the halide vacancy defects in FA1–xGAxPbBr3 21. Cao, Y. et al. Perovskite light-emitting diodes based on spontaneously formed
submicrometre-scale structures. Nature 562, 249–253 (2018).
PNCs and to improve the charge balance in PeLEDs. These syn-
22. Yang, X. et al. Efficient green light-emitting diodes based on
ergistic strategies yielded the highest electroluminescence efficien- quasi-two-dimensional composition and phase engineered perovskite with
cies in PNC-based PeLEDs (CE = 108 cd A−1 and EQE = 23.4%; surface passivation. Nat. Commun. 9, 570 (2018).
CE = 203 cd A−1 and EQE = 45.5% with hemispherical lens). Our 23. Zhao, L. et al. Electrical stress influences the efficiency of CH3NH3PbI3
efficiencies demonstrate that this design strategy provides a clear perovskite light emitting devices. Adv. Mater. 29, 1605317 (2017).
24. Chen, H. et al. High-efficiency formamidinium lead bromide perovskite
pathway to translate PNCs into PeLEDs for a new generation of nanocrystal-based light-emitting diodes fabricated via a surface defect self‐
high-efficiency display applications. passivation strategy. Adv. Opt. Mater. 8, 1901390 (2020).
25. Zhu, H. et al. Screening in crystalline liquids protects energetic carriers in
Online content hybrid perovskites. Science 353, 1409–1413 (2016).
Any methods, additional references, Nature Research report- 26. Yang, Y. et al. Observation of a hot-phonon bottleneck in lead-iodide
perovskites. Nat. Photon. 10, 53–59 (2016).
ing summaries, source data, extended data, supplementary infor- 27. Zheng, F., Tan, L. Z., Liu, S. & Rappe, A. M. Rashba spin–orbit
mation, acknowledgements, peer review information; details of coupling enhanced carrier lifetime in CH3NH3PbI3. Nano Lett. 15,
author contributions and competing interests; and statements of 7794–7800 (2015).
data and code availability are available at https://doi.org/10.1038/ 28. Rappe, A. M., Grinberg, I. & Spanier, J. E. Getting a charge out of hybrid
s41566-020-00732-4. perovskites. Proc. Natl Acad. Sci. USA 114, 7191–7193 (2017).
29. Swarnkar, A. et al. Quantum dot-induced phase stabilization of α-CsPbI3
perovskite for high-efficiency photovoltaics. Science 354, 92–95 (2016).
Received: 1 February 2020; Accepted: 28 October 2020; 30. Banerjee, A., Chakraborty, S. & Ahuja, R. Bromination-induced stability
Published: xx xx xxxx enhancement with a multivalley optical response signature in guanidinium
[C(NH2)3]+-based hybrid perovskite solar cells. J. Mater. Chem. A 5,
References 18561–18568 (2017).
1. Tan, Z.-K. et al. Bright light-emitting diodes based on organometal halide 31. Giorgi, G., Fujisawa, J.-I., Segawa, H. & Yamashita, K. Organic–inorganic
perovskite. Nat. Nanotechnol. 9, 687–692 (2014). hybrid lead iodide perovskite featuring zero dipole moment guanidinium
2. Kim, Y.-H. et al. Multicolored organic/inorganic hybrid perovskite cations: a theoretical analysis. J. Phys. Chem. C 119, 4694–4701 (2015).
light-emitting diodes. Adv. Mater. 27, 1248–1254 (2015). 32. Jodlowski, A. D. et al. Large guanidinium cation mixed with
3. Cho, H. et al. Overcoming the electroluminescence efficiency limitations of methylammonium in lead iodide perovskites for 19% efficient solar cells. Nat.
perovskite light-emitting diodes. Science 350, 1222–1225 (2015). Energy 2, 972–979 (2017).
4. Protesescu, L. et al. Nanocrystals of cesium lead halide perovskites (CsPbX3, 33. Marco, N. D. et al. Guanidinium: a route to enhanced carrier lifetime and
X = Cl, Br, and I): novel optoelectronic materials showing bright emission open-circuit voltage in hybrid perovskite solar cells. Nano Lett. 16, 1009–1016
with wide color gamut. Nano Lett. 15, 3692–3696 (2015). (2016).
5. Schmidt, L. C. et al. Nontemplate synthesis of CH3NH3PbBr3 perovskite 34. Kanno, S., Imamura, Y. & Hada, M. First-principles calculations of the
nanoparticles. J. Am. Chem. Soc. 136, 850–853 (2014). rotational motion and hydrogen bond capability of large organic cations in
6. Kim, Y.-H., Cho, H. & Lee, T.-W. Metal halide perovskite light emitters. Proc. hybrid perovskites. J. Phys. Chem. C 122, 15966–15972 (2018).
Natl Acad. Sci. USA 113, 11694–11702 (2016). 35. Kakekhani, A., Katti, R. N. & Rappe, A. M. Water in hybrid perovskites: bulk
7. Kim, Y.-H., Kim, J. S. & Lee, T.-W. Strategies to improve luminescence MAPbI3 degradation via super-hydrous state. APL Mater. 7, 041112 (2019).
efficiency of metal‐halide perovskites and light‐emitting diodes. Adv. Mater. 36. Yi, C. et al. Entropic stabilization of mixed A-cation ABX3 metal halide
31, 1804595 (2019). perovskites for high performance perovskite solar cells. Energy Environ. Sci. 9,
8. Koscher, B. A., Swabeck, J. K., Bronstein, N. D. & Alivisatos, A. P. Essentially 656–662 (2016).
trap-free CsPbBr3 colloidal nanocrystals by postsynthetic thiocyanate surface 37. Travis, W., Glover, E. N. K., Bronstein, H., Scanlon, D. O. & Palgrave, R. G.
treatment. J. Am. Chem. Soc. 139, 6566–6569 (2017). On the application of the tolerance factor to inorganic and hybrid halide
9. De Roo, J. et al. Highly dynamic ligand binding and light absorption perovskites: a revised system. Chem. Sci. 7, 4548–4556 (2016).
coefficient of cesium lead bromide perovskite nanocrystals. ACS Nano 10, 38. Pham, N. D. et al. Tailoring crystal structure of FA0.83Cs0.17PbI3 perovskite
2071–2081 (2016). through guanidinium doping for enhanced performance and tunable
10. Wang, H.-C. et al. High-performance CsPb1−xSnxBr3 perovskite quantum dots hysteresis of planar perovskite solar cells. Adv. Funct. Mater. 29, 1806479
for light-emitting diodes. Angew. Chem. Int. Ed. 129, 13838–13842 (2017). (2019).
11. Chiba, T. et al. Anion-exchange red perovskite quantum dots with 39. Deng, Z., Kieslich, G., Bristowe, P. D., Cheetham, A. K. & Sun, S. Octahedral
ammonium iodine salts for highly efficient light-emitting devices. Nat. connectivity and its role in determining the phase stabilities and electronic
Photon. 12, 681–687 (2018). structures of low-dimensional, perovskite-related iodoplumbates. APL Mater.
12. Song, J. et al. Organic–inorganic hybrid passivation enables perovskite QLEDs 6, 114202 (2018).
with an EQE of 16.48%. Adv. Mater. 30, 1805409 (2018). 40. Kakekhani, A. et al. Nature of lone-pair–surface bonds and their scaling
13. Park, M.-H. et al. Boosting efficiency in polycrystalline metal halide relations. Inorg. Chem. 57, 7222–7238 (2018).
perovskite light-emitting diodes. ACS Energy Lett. 4, 1134–1149 (2019). 41. Kim, Y.-H. et al. Highly efficient light-emitting diodes of colloidal metal–
14. Xu, W. et al. Rational molecular passivation for high-performance perovskite halide perovskite nanocrystals beyond quantum size. ACS Nano 11,
light-emitting diodes. Nat. Photon. 13, 418–424 (2019). 6586–6593 (2017).
15. Zhao, X. & Tan, Z.-K. Large-area near-infrared perovskite light-emitting 42. Contreras-García, J. et al. NCIPLOT: a program for plotting noncovalent
diodes. Nat. Photon. 14, 215–218 (2020). interaction regions. J. Chem. Theory Comput. 7, 625–632 (2011).
16. Yang, Y. et al. Comparison of recombination dynamics in CH3NH3PbBr3 and
CH3NH3PbI3 perovskite films: influence of exciton binding energy. J. Phys.
Chem. Lett. 6, 4688–4692 (2015). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
17. Stranks, S. D. et al. Recombination kinetics in organic-inorganic perovskites: published maps and institutional affiliations.
excitons, free charge, and subgap states. Phys. Rev. Appl. 2, 034007 (2014). © The Author(s), under exclusive licence to Springer Nature Limited 2021

Nature Photonics | www.nature.com/naturephotonics

Nature Photonics Articles
Methods Transmission electron microscopy measurement. To measure the transmission
Synthesis of PNCs. All FA1−xGAxPbBr3 PNCs were synthesized in air at room electron microscopy images of the PNCs, 0.05 ml of FA1–xGAxPbBr3 PNC solutions
temperature. Precursor solutions were prepared by dissolving FABr (Dyesol), were dropped onto a copper grid and measured using a JEOL-JEM 2100F with an
GABr (Dyesol) and PbBr2 (Aldrich, 99.999%), with the FABr:GABr ratio changing acceleration voltage of 200 kV.
from 1:0 to 0:1 while maintaining the molarity of FABr:GABr at 0.1 mmol and the
molarity of PbBr2 at 0.2 mmol, in 0.5 ml of anhydrous N,N-dimethylformamide DLTS. Deep-level transient spectroscopy measurements were performed with a
(Aldrich, 99.8%); 0.15 ml of precursor solution was then dropped into a PhysTech FT1030 DLTS system. The capacitance was measured using a modified
crystallization-inducing solution comprising 5 ml of toluene, 2 ml of 1-butanol, Boonton 72B capacitance meter with a 1 MHz. Temperature scans were performed
0.3 ml of oleic acid and 24.2 μl of n-decylamine, which was mixed for 10 min between 20 and 300 K at a heating rate of 2 K min–1. Samples were placed in the
under vigorous stirring. The resulting colloidal PNCs were washed by sequential helium-contact gas of a liquid helium cryostat. The pulse height, filling pulse width
centrifugation and then collected in toluene. and pulse period width were 0.4 V, 10 ms and 10 ms, respectively. The activation
energy, capture cross-section and concentration of traps were calculated using an
PeLED fabrication. Indium tin oxide patterned glasses were cleaned by sequential Arrhenius plot.
sonication in acetone and 2-isopropanol for 15 min each. The glasses were boiled
in 2-isopropanol for 20 min, then transferred into an ultraviolet-ozone cleaner and DFT calculations of structures. We used a fully vc-relaxed, self-consistent
treated with ultraviolet-ozone for 10 min to remove the residual organics. Buf-HILs DFT-derived structure to model the hybrid perovskite systems we have studied
composed of poly(3,4-ethylenedioxythiophene)/poly(styrene sulfonate) and in this work. This means no constraint was assumed and implemented on the
perfluorinated ionomer were then spin-coated in air to make a 40-nm-thick layer, atomic positions or lattice vectors. In hybrid perovskite materials, the higher
which was then annealed at 150 °C for 30 min. The samples were then transferred temperatures derive the structures towards higher effective symmetries due to
into a glove box, in which emitting layers were formed by spin-coating-appropriate thermal and spatial averaging of the local structural symmetries. The nominal
FA1–xGAxPbBr3 PNC solutions at 1,000 r.p.m. for 60 s. To fabricate the 5-nm-thick higher symmetry (cubic) perovskite structures deduced from XRD, which
TBTB layer, TBTB solutions, which were dissolved in toluene, were spincoated at are macroscopically averaged, should not be used to model the electronic
3,000 r.p.m. for 60 sec. The samples were transferred to a vacuum evaporator, and properties (naturally including formation and binding energies), which are
then a 50-nm-thick layer of TPBI, a 1-nm-thick layer of LiF and a 100-nm-thick sensitive to the order and coordination of atoms in angstrom scale rather than a
layer of aluminium were sequentially deposited. macroscopic average. FAPbBr3 is experimentally known to have a low temperature
orthorhombic structure and a room temperature cubic phase. Both of these are
General information on our DFT simulations. We performed DFT actually an averaged order and we find a (technically speaking) triclinic structure
calculations with plane-wave basis sets using the Quantum Espresso software to be the fully self-consistent DFT structure. This structure is greatly similar to the
package43. Ultrasoft GBRV pseudopotentials44 and Perdew–Burke–Ernzerhof orthorhombic phase (only very small deviations in cell parameters from the ideal
exchange correlations were used45. Dispersion interactions were accounted orthorhombic phase) and is depicted in Fig. 1b. The volume of such a structure
for using the Grimme DFT-D3 method46. The inclusion of dispersion is ~852 Å3, which was within ~2% error of the experimentally measured value for
interactions to accurately describe hybrid perovskite systems has been shown this phase. This corresponds to an error of ~0.5% in lattice constants. Moreover,
to be important47. As mentioned in the main text (in the context of Fig. 1c), as mentioned in the main text, such a structure was able to reproduce the
the DFT-D3 dispersion correction can accurately describe the experimental experimental XRD patterns virtually perfectly (Fig. 1c). As has been pointed out
data. We used a k-mesh sampling density greater than 15,000 points per Å3. experimentally, such a phase has a c(2 × 2) × 2 structure, in which the FA cations
The plane-wave energy cutoff (Ecut) was set to ~550 eV, whereas the charge have an alternative orientational pattern, that is, the long direction of the molecule
cutoff was set to ten times this value. We set the force convergence criteria alternates its orientation (Fig. 1c). We also perform DFT calculations of surfaces
for the relaxation calculations as very small: each force component on each and quantum dots (see Supplementary Discussion 2 for more details).
atom needed to be smaller than 10−4 Rydberg per Bohr radius. Relative to the
regular solid-state structures, we have molecular groups as part of the structure; DFT calculations of bulk energies. We perform the procedure explained below
consequently, the energy landscape can have regions with very small energy to calculate the internal and free energies reported in Fig. 1d. For calculations
gradients (mostly corresponding to the rotations of the molecular groups), thus regarding 0, 25, 50, 75 and 100% GA content, we use the c(2 × 2) × 2 structure
precise calculations of forces and low enough force convergence thresholds were that was previously mentioned. We dope the aforementioned structure with the
necessary to properly explore the real local minima of such structures. Most of needed GA, make a number of initial guesses and then fully vc-relax the structures
our calculations (except some molecule-in-box calculations) were variable-cell with no constraints. For 12.5% concentration, we use a 2 × 2 × 2 supercell. The
relaxations (vc-relax), in which both the atomic coordinates and the cell vectors relative internal energies reported in Fig. 1d are relative to the precursors, and
were optimized to find the zero-force, zero-stress structures with minimum the 0 has been chosen to be that of FAPbBr3. The precursors are DFT-derived
internal energy. The smearing scheme for the occupations of the Kohn−Sham PbBr2, GABr and FABr. The free energy was derived by adding the configurational
orbitals was the cold smearing of Marzari and Vanderbilt48, with a temperature entropy and (zero-temperature DFT-derived) internal energy contributions. The
equal to 10 mRy/kB (where kB is the Boltzmann constant). The computational configurational entropy (per f.u.c. for FA1–xGAxPbBr3) was calculated using the
XRD patterns were generated using the Mercury software49, with fully relaxed analytic formula for ideal alloys, which has been shown to be an effective model for
atomic coordinates and lattice constants as input. Using the aforementioned entropy stabilization of hybrid perovskite systems: S = kB[xln x + (1 − x)ln(1 − x)]36.
convergence parameters, we successfully achieved the precision (~1 meV per
formula unit cell (f.u.c.)) we needed in formation energy plots (Fig. 1d), and a DFT calculations of acid or amine binding energies. An important part of
virtually perfect structural agreement with the experiment (Fig. 1c). We used our experiments was the stabilizing role of the long-chain acids and amines
norm-conserving pseudopotentials—as opposed to ultrasoft ones—to calculate we synthesize our PNCs in. Specifically, we use oleic acid and decylamine. We
the reduced density gradient and the NCI index42; Ecut was set to ~680 eV for performed DFT calculations to understand how such ligands bind to our PNC’s
these calculations. The charge density differences calculations for the binding surfaces and how and what extent they stabilize these surfaces. Our calculations
of ligands were also performed using norm-conserving pseudopotentials to be indicate that our perovskite samples posses an AX-termination (see Supplementary
consistent with the NCI plots. Section 2 for more details). Consequently, we study the interaction of acids and
bases to such surfaces, and we investigate how the presence of surface GA groups
Photoluminescence and PLQE measurement. Photoluminescence spectra of changes the binding.
FA1–xGAxPbBr3 PNCs in solution states were measured using a JASCO FP8500 We performed a set of numerical experiments to understand the interaction
spectrofluorometer. Photoluminescence quantum efficiencies were detected between the ligands and the surface. We study the binding energy of the acid and
using a 100 nm integrating sphere (ILF-835) that was mounted on the same amines as a function of their chain length (Supplementary Fig. 9). The surface
spectrofluorometer, and calculated using JASCO SpectraManager II Software. binding energy of the ligand was sizable and was in the root of their stabilizing
The PLQE of the thin film samples was measured using an integrating sphere effect on our PNCs. The amines interact via their nitrogen 2p lone-pair orbital,
method. A continuous-wave 405 nm blue diode laser with an excitation power of with the proton-like nitrogen-bound hydrogen of the surface organic cation,
50 µW and a focused beam spot of ~0.3 mm2 was used to photoexcite the samples. while an amine hydrogen forms a hydrogen bond with the surface bromide (Fig.
Emission was measured using an Andor iDus Silicon detector. The samples were 2b). The acid interacts from the oxygen end with the proton-like nitrogen-bound
encapsulated between two glass cover slips before measurements. hydrogen of the surface organic cation, whereas the acid OH group interacts via
its hydrogen with the surface bromide (Fig. 2b). It can be seen that the interaction
Time-correlated single-photon counting measurement. Photoluminescence between the acids and amines can be decomposed into two parts: (1) a strong
lifetimes of FA1–xGAxPbBr3 PNCs were measured using a FluoTime 300. FA1– point interaction that is present for smaller-chain acids and amines, (2) a weaker
xGAxPbBr3 PNCs were photoexcited by a picosecond-pulse laser head (excitation
non-local van der Waals interaction (Fig. 2c) that is a function of the chain length,
wavelength = 405 nm, LDH-P-C-405B, PicoQuant) and their photoluminescence the folding of the chain relative to the surface and to the neighbouring ligands.
lifetimes were detected using a photon counting detector (PMA Hybrid 07) and The non-local part of the interaction closely depends on the specific orientation
a time-correlated single-photon counting measurement module (PicoHarp, and folding of the ligand relative to the surface and neighbouring ligands, which is
PicoQuant). naturally affected by the thermal motions and steric hindrance. Nevertheless, the

Nature Photonics | www.nature.com/naturephotonics

Articles Nature Photonics
property we were more interested in was the change in the ligand binding upon GA Data availability
doping. Such a quantity is more robust and less dependent on the non-local part The data that support the plots within this paper and other findings of this study
of the interaction, as the change in surface cation (FA to GA) mostly changes the are available from the corresponding author on reasonable request.
local part of the interaction. Consequently, we quantify the change in the binding
energy by introducing the GA via investigating the binding of short-chain acids
and amines (acetic acid and ethylamine), which are dominated by the local point References
interactions (Supplementary Fig. 9a). Probing the trends in adsorption energies 43. Giannozzi, P. et al. QUANTUM ESPRESSO: a modular and open-source
of some adsorbates, using smaller adsorbates with identical chemical groups software project for quantum simulations of materials. J. Phys. Condens.
interacting with the surface, is similar in philosophy to the idea of scaling relations Matter 21, 395502 (2009).
in surface chemistry and catalysis. As mentioned in the main text, the GA doping 44. Garrity, K. F., Bennett, J. W., Rabe, K. M. & Vanderbilt, D. Pseudopotentials
slightly weakens the binding of these ligands (Supplementary Fig. 9b,c); this result for high-throughput DFT calculations. Comput. Mater. Sci. 81, 446–452
was yet another indication that incorporating GA stabilizes the surface internally (2014).
and decreases its reactivity toward external agents. 45. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation
As a side note, the data used for Supplementary Fig. 9a to decompose the made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
interaction of acids and amines as well as Figs. 2b and 2c, have been generated 46. Grimme, S., Antony, J., Ehrlich, S. & Krieg, H. A consistent and accurate ab
based on an earlier set of calculations performed with a slightly higher energy bulk initio parametrization of density functional dispersion correction (DFT-D)
structure with an effective 1 × 1 × 1 reconstruction for FAPbBr3, as opposed to for the 94 elements H–Pu. J. Chem. Phys. 132, 154104 (2010).
c(2 × 2) × 2. Nevertheless, the conclusions of these figures are robust, as we extract 47. Wang, Y. et al. Density functional theory analysis of structural and electronic
the trends among binding energies and the nature of chemical bonds, as opposed properties of orthorhombic perovskite CH3NH3PbI3. Phys. Chem. Chem. Phys.
to absolute value of binding energies. Even the absolute values vary insignificantly 16, 1424–1429 (2014).
when compared with the numbers that correspond to the ground-state c(2 × 2) × 2 48. Marzari, N., Vanderbilt, D., De Vita, A. & Payne, M. C. Thermal contraction
reconstruction of the bulk FAPbBr3. For example, the binding energies for acetic and disordering of the Al(110) surface. Phys. Rev. Lett. 82, 3296–3299 (1999).
acid and ethylamine change by 0.02 eV, which do not change the conclusions 49. Macrae, C. F. et al. Mercury CSD 2.0—new features for the visualization and
reached on the trend of binding energies in Supplementary Fig. 9a. investigation of crystal structures. J. Appl. Cryst. 41, 466–470 (2008).
50. Jeong, S.-H. et al. Characterizing the efficiency of perovskite solar cells and
light-emitting diodes. Joule 4, 1206–1235 (2020).
Perovskite phase instability of GAPbBr3. As mentioned in the main text, in the
context of Fig. 1d, the perovskite phase of GAPbBr3 (with 3D connectivity and
corner-sharing structure) was destabilized due to the large size of GA, which was Acknowledgements
beyond the tolerance of the perovskite structure. The phase transition of GAPbI3 This work was supported by the National Research Foundation of Korea (NRF) grant
to an edge-sharing phase has been experimentally observed and theoretically funded by the Korea government (MSIT) (NRF-2016R1A3B1908431). A.K., R.B.W.,
verified in a recent work. Both our DFT and experimental results also confirm and A.M.R. acknowledge the support of the US Department of Energy, Office of
that the perovskite structure is not a relevant phase for guanidinium lead bromide. Basic Energy Sciences, under grant no. DE-SC0019281 and also the computational
Our DFT results show that a one-dimensional edge-sharing phase for GAPbBr3 support from NERSC of the DOE. P.T. acknowledges the scholarship from Chinese
(Supplementary Fig. 11c), was enthalpically favoured over the three-dimensional Scholarship Council (CSC). H.J.B. acknowledges funding from the European Research
perovskite phase by −0.11 eV per f.u.c. Furthermore, as shown in Supplementary Council (ERC) under the European Union’s Horizon 2020 research and innovation
Fig. 4, the edge-sharing structure was even softer than the perovskite and thus programme (grant agreement no. 834431) and the Spanish Ministry of Economy and
was more entropically favoured. Consequently, the one-dimensional edge-sharing Competitiveness (MINECO) via the Unidad de Excelencia María de Maeztu CEX2019-
structure should possess a lower free energy than the three-dimensional perovskite 000919-M and MAT2017-88821-R. A.S., S.N. and R.H.F. acknowledge support from the
structure. Density functional theory calculations show that the face-sharing phase UKRI Global Challenge Research Fund project, SUNRISE (EP/P032591/1) and UKIERI
does not have the enthalpy drive and was almost equal in internal energy to the projects. S.N. acknowledges funding and support from Royal Society-SERB Newton
perovskite phase. Moreover, one should note that our experimentally synthesized International Fellowship.
guanidinium lead bromide PNCs do not have to be in GAPbBr3 stoichiometry.
Another possibly relevant phase (as shown for the case of iodoplumbates) can be Author contributions
the GA-rich corner-sharing two-dimensional GA2PbBr4 structure (Supplementary Y.-H.K., S.K. and A.K. equally contributed to this work. Y.-H.K., S.K. and T.-W. L.
Fig. 11b). Our experimental XRD patterns also confirm the deviation of initiated and designed the study. Y.-H.K. and S.K. performed experiments and analysed
the guanidinium lead bromide PNCs (x = 1) from the perovskite structure data. A.K., R.B.W. and P.T. performed the simulations, analysed data, helped understand
(Supplementary Fig. 11d). Although we have not been able to theoretically match the structures and mechanisms behind the great efficiency. J.P., D.-H.K. and S.H.J. helped
the exact phase of guanidinium lead bromide (Supplementary Fig. 11e), it was clear to analyse the data. H.X. and B.H. performed and analysed magnetic field-dependent
that, unlike lower values of x, the structure was no longer in the corner-sharing characteristics. Y.-H.L. and Y.-W.K. measured and analysed TEM. L.M.-S. and H.J.B.
perovskite phase. commented on the synthesis of nanocrystals. J.P. (KAIST) and S.Y. performed optical
simulation of the devices. A.S., S.N. and R.H.F. performed the PLQE measurements and
Characterization and optical simulation of PeLEDs. Electroluminescence analysis. A.M.R. and T.-W.L. supervised the study. All authors discussed the results and
efficiencies of the fabricated PeLEDs were measured using a Keithley 236 source commented on the manuscript.
measurement unit and a Minolta CS 2000 spectroradiometer. External quantum
efficiency of PeLEDs was calculated by measuring full angular electroluminescence Competing interests
distribution50. In analysing PeLEDs, the normalized power spectral densities were The authors declare no competing interests.
calculated using custom MATLAB codes based on the classical oscillating dipole
approach. To separate the outcoupling effect, the internal quantum efficiency
was set to unity. As absorption loss of the emission layer (EML) was negligible in Additional information
the emission spectra, lossless EML was assumed in the simulation for numerical Supplementary information is available for this paper at https://doi.org/10.1038/
stability. The refractive indices of aluminium, TPBi, EML (= FA0.9GA0.1PbBr3), s41566-020-00732-4.
Buf-HIL, ITO and glass used in the simulation were measured using spectroscopic Correspondence and requests for materials should be addressed to A.M.R. or T.-W.L.
ellipsometry (Woollam M2000D). For the refractive indices of TBTB layer, the
refractive index of TPBI was applied because of its thin thickness (~5 nm). The Peer review information Nature Photonics thanks the anonymous reviewers for their
emission zone was set to be located at the centre of EML, which was supported by contribution to the peer review of this work.
the electron/hole-only device experiments. Reprints and permissions information is available at www.nature.com/reprints.

Nature Photonics | www.nature.com/naturephotonics