ARTICLE

https://doi.org/10.1038/s41467-020-20745-5 OPEN

Preserving a robust CsPbI3 perovskite phase via

pressure-directed octahedral tilt
Feng Ke1,2,5, Chenxu Wang 2,5, Chunjing Jia 1,5, Nathan R. Wolf 3, Jiejuan Yan2, Shanyuan Niu 1,2,

Thomas P. Devereaux 1,4, Hemamala I. Karunadasa1,3, Wendy L. Mao 1,2 & Yu Lin 1 ✉
1234567890():,;

Functional CsPbI3 perovskite phases are not stable at ambient conditions and spontaneously
convert to a non-perovskite δ phase, limiting their applications as solar cell materials. We
demonstrate the preservation of a black CsPbI3 perovskite structure to room temperature by
subjecting the δ phase to pressures of 0.1 – 0.6 GPa followed by heating and rapid cooling.
Synchrotron X-ray diffraction and Raman spectroscopy indicate that this perovskite phase is
consistent with orthorhombic γ-CsPbI3. Once formed, γ-CsPbI3 could be then retained after
releasing pressure to ambient conditions and shows substantial stability at 35% relative
humidity. First-principles density functional theory calculations indicate that compression
directs the out-of-phase and in-phase tilt between the [PbI6]4− octahedra which in turn tune
the energy difference between δ- and γ-CsPbI3, leading to the preservation of γ-CsPbI3. Here,
we present a high-pressure strategy for manipulating the (meta)stability of halide perovskites
for the synthesis of desirable phases with enhanced materials functionality.

1 Stanford Institute for Materials and Energy Sciences, SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA. 2 Department of Geological

Sciences, Stanford University, Stanford, CA 94305, USA. 3 Department of Chemistry, Stanford University, Stanford, CA 94305, USA. 4 Department of
Materials Science and Engineering, Stanford University, Stanford, CA 94305, USA. 5These authors contributed equally: Feng Ke, Chenxu Wang, Chunjing Jia.
✉email: [email protected]

NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5

L
ead halide perovskite solar cells based on (MA)PbI3 (MA =
CH3NH3+) have achieved power-conversion efficiencies
comparable to commercial Si-based solar cells in recent
years1–3. However, the volatile organic MA cation is largely
responsible for this material’s instability to humidity and heat, a
critical issue that hinders its large-scale implementation. Sub-
stitution of an inorganic cation such as Cs+ has been explored as
a way to improve the robustness of lead halide perovskites4–7.
There are four known polymorphs of CsPbI38–13: a room-
temperature non-perovskite phase (δ), and three high-
temperature perovskite-related phases with cubic (α), tetragonal
(β), and orthorhombic (γ) structures. Perovskite-structured
CsPbI3 has a bandgap of 1.6–1.8 eV that is favorable for photo-
voltaic applications4–9. However, a major challenge to realizing
CsPbI3-based solar cells is that these black perovskite phases are
not thermodynamically stable at ambient conditions and spon-
taneously convert to the non-functional δ phase10–13. A recent
study reported that perovskite solar cells made from surface- Fig. 1 Experimental setup and pressure-induced preservation of a
treated β-CsPbI3 achieved power-conversion efficiencies >18%, metastable γ-CsPbI3 perovskite to ambient conditions. a Illustration of a
but only at high temperature4. It is crucial to find ways to stabilize cross-sectional view of the high-pressure and high-temperature setup. The
CsPbI3 perovskites at room temperature for practical device white scale bar is 1 cm. Inset: Schematic of a diamond–anvil cell and an
operation. external resistive heater used in our experiments. b The observed
Previous studies have offered a few strategies for stabilizing CsPbI3 structures of CsPbI3 at varying P–T cycles. The red and blue arrows
perovskite phases to room temperature, including thermal engi- represent the most viable heating and cooling pathways to preserve a γ-
neering9–13, compositional tuning14–17, nanocrystal growth5,18,19, CsPbI3 perovskite phase, where the applied pressure (P) is 0.1–0.6 GPa and
solvent and surface treatments6–8,20–28, and strain engineering29. temperature (T) is up to 450 °C. Atoms in the structures are shown in gray
However, these approaches present several drawbacks and (Pb), purple (I), and cyan (Cs). c Representative P–T phase transformations
limitations30. For instance, thermal engineering based on a solid-state observed in the study. RT and Amb indicate room temperature and ambient
method required rigorously anhydrous reagents, a moisture-free pressure, respectively. The yellow δ and gray γ circles around RT are in fact
environment, a melt state at high temperature, and extremely rapid both at RT and plotted with a slight offset for clarity. Once γ-CsPbI3 is
cooling. The as-preserved bulk γ-CsPbI3 also showed severe moisture formed by temperature quenching at high pressure, it can be preserved
sensitivity13. The strain-stabilized γ-CsPbI3 film was rendered back to ambient conditions after fully releasing pressure.
by the use of a combined substrate clamping and biaxial strain in
an inert atmosphere and returned back to the δ phase within minutes and its PL intensity remains above 80% of the initial intensity for
in air at 27% relative humidity (RH)29. Compositional tuning more than 30 days and up to 10 days at 20% and 35% RH,
introduced undesirable changes in the electronic structure20,21, respectively.
and nanomaterials showed an increase in grain boundaries
which inhibited charge transport and caused considerable recombi- Results
nation loss6,15. These drawbacks motivate us to explore other P–T phase diagram of CsPbI3. In situ XRD and Raman mea-
approaches. surements were conducted to study the structural evolution of
Previous calculations suggested that the [PbI6]4− inter- CsPbI3 as a function of pressure along heating and cooling routes.
octahedral tilt had a large influence on the formation energies of All the diamond–anvil–cell preparation, sample loading, and P–T
the three perovskite phases (α, β, and γ) of CsPbI3, among which treatments were done in air without special handling. The sam-
the γ phase has the lowest energy owing to its largest octahedral ples were first compressed to a target pressure at room tem-
tilt9. Hence, tuning the octahedral tilt of the high-temperature perature followed by heating and cooling from high temperature.
perovskite phase(s) using external stimuli such as pressure will be The pressure may shift slightly as the temperature is varied
favorable for preserving the desired structure back to ambient during the cycle, but the final pressure after cooling back to room
conditions. Many halide perovskite systems have been studied temperature is nearly identical to the initial pressure before
at high pressure. Pressure has proven to be an effective and clean heating.
tool for modifying their structures and creating exotic physical At ambient pressure, XRD indicates that the starting δ-CsPbI3
properties31–35, such as the bandgap modulation31,32 and closure phase transforms to the cubic α-CsPbI3 phase when heated to
in compressed (MA)PbI335,36 and emission enhancement in 320 °C (Fig. 2a), consistent with previous observations10–12. With
compressed halide double perovskites37. The pressure was also increasing pressure, the δ-to-α phase transition occurs at higher
found to markedly tilt the [PbI6]4− octahedra in CsPbI3 perovskite temperatures. Specifically, the transition temperature increases to
nanocrystals38,39. However, none of the desired high-pressure 370 °C at 0.6 GPa (Fig. 2b) and exceeds 500 °C at 1.1 GPa
phases has been preserved back to ambient conditions after (Fig. 2c). Raman measurements also agree with the XRD results.
releasing pressure. At ambient pressure and with heating up to 315 °C, the Raman
In this work, we study the solid-to-solid structural evolution of modes of δ-CsPbI3 remain nearly invariant except for subtle
CsPbI3 at high-pressure and high-temperature conditions in a frequency shifts, peak broadening, and changes in relative peak
resistive-heated diamond–anvil cell (Fig. 1a) using synchrotron intensities (Fig. 2d). As the temperature reaches 315 °C, the color
X-ray diffraction (XRD), Raman spectroscopy, photo- of the sample changes dramatically from yellow to black
luminescence (PL) measurements, and first-principles density (Supplementary Fig. 1) and all of the Raman modes suddenly
functional theory (DFT) calculations. We identify viable disappear, suggesting that CsPbI3 converts to the cubic α phase
pressure–temperature (P–T) pathways to access and quench a which is Raman inactive based on the group theory analysis on a
black CsPbI3 perovskite phase back to ambient conditions. The cubic perovskite structure40,41. With increasing pressure, the
preserved CsPbI3 phase shows substantial stability to moisture disappearance of the Raman modes (Fig. 2e, f) and the color

2 NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5 ARTICLE

Fig. 2 X-ray diffraction (XRD) and Raman spectra showing the effect of pressure on the structural transition of CsPbI3 along with thermal treatments.
a–c XRD results (λ = 0.4959 Å) of CsPbI3 along heating and cooling cycles at ambient pressure (a), 0.6 GPa (b), and varying pressures (c). The sample
remains in a crystalline state up to the highest temperature studied in each cycle. The intensity differences of XRD patterns for α- and γ-CsPbI3 could be a
result of different grain sizes and orientations due to the grain growth to highly textured, coarse grains at high temperatures. d–f Raman spectra of CsPbI3
along heating and cooling cycles at ambient pressure (d), 0.6 GPa (e), and varying pressures (f), confirming the structural transitions observed from XRD
measurements.

change from yellow to black occur at higher temperatures, the non-perovskite δ phase (Fig. 2b, e). The sample also remains
completely consistent with our XRD results. Following slow opaque black throughout cooling (Supplementary Fig. 1). In route
cooling, the high-temperature α phase transforms back to the δ 2’ (Fig. 3a), in which the pressure increases from 0.5 to 1.1 GPa
phase below the transition temperature. Based on the XRD and during heating to 450 °C and goes back to 0.5 GPa after cooling to
Raman results, a P–T phase diagram of CsPbI3 was mapped out room temperature, XRD and Raman measurements indicate that
as shown in Supplementary Fig. 2. while CsPbI3 remains in the yellow δ phase during the entire
heating cycle (Fig. 2c, f), the sample instantaneously transforms to
Pressure effect on preserving γ-CsPbI3 to ambient conditions. the α phase at the onset of cooling, followed by the appearance of
The structural transitions observed upon rapid cooling at a rate of a similar black phase as observed in route 2 which persists down
~90 K/min strongly depend on the pressure applied to the system. to room temperature.
XRD and Raman results indicate that a black perovskite phase The splitting of diffraction peaks indicates that this preserved
can be metastably preserved back to room temperature when the black phase has a lower symmetry compared to the cubic α-
initial δ-CsPbI3 is subjected to a pressure between 0.1 and 0.6 GPa CsPbI3 phase. We indexed the diffraction pattern with the other
followed by rapid cooling from high temperature. Three different known phases of CsPbI3 (β, γ, and δ), and found that all of the
sets of experimental routes were identified according to the diffraction peaks of the preserved black phase in Fig. 2b, c could
applied pressure: <0.1 GPa (route 1), 0.1–0.6 GPa (routes 2 and be assigned to the orthorhombic γ-CsPbI3, but not to β- or δ-
2’), and >0.6 GPa (routes 3, 3’, and 3”), as shown in Fig. 3a. At CsPbI3 (Supplementary Figs. 3 and 4). As shown in Supplemen-
ambient pressure (route 1), when rapidly cooling from 430 °C, the tary Fig. 5, the diffraction pattern changes from smooth
high-temperature α-CsPbI3 returns to the thermodynamically diffraction rings before heating to spots after CsPbI3 undergoes
preferred δ phase below 220 °C (Fig. 2a). transitions to the α and γ structures, indicating significant grain
Interestingly, when the applied pressure is between 0.1 and 0.6 growth in the sample to highly textured, coarse grains. The
GPa (routes 2 and 2’), CsPbI3 undergoes different structural diffraction peak intensities are unreliable as they are mainly
transitions upon rapid cooling. Taking the cooling process at determined by the grain size and the orientation of each grain
~0.6 GPa as an example (Fig. 2b and route 2 in Fig. 3a), as the relative to the incident X-ray beam. While all of the peaks in the
sample cools from 450 °C, the diffraction peaks of the high- XRD patterns of the recovered phase can be indexed to γ-CsPbI3,
temperature α-CsPbI3 split obviously below 150 °C, and these the relative peak intensities look distinct in routes 2 and 2’
changes persist down to room temperature. The preserved (Fig. 2b, c) and are also different from previous studies on powder
diffraction pattern and Raman spectrum differ from those of samples9,10. The apparent differences in the diffraction angle (2θ)

NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5

a b d
2 3 2’ 3’ 3”
1
b

α c
a c
Out-of-phase tilt

δ
α
γ a c
b
In-phase tilt

e P2: metastable γ-CsPbI3

γ δ

Free energy, G
Metastable P1: no γ-CsPbI3

γ δ γ δ
γ δ
ΔG

ΔG = G γ Gδ

27 °C 100 °C 200 °C T

Fig. 3 Representative cooling routes and the effect of [PbI6]4− octahedral tilt on the energy difference between γ- and δ-CsPbI3 with pressure. a The
experimental cooling routes used to access and preserve γ-CsPbI3 to room temperature. The background color shows the P–T phase diagram of CsPbI3 that
includes the δ and α phases. γ-CsPbI3 is not a thermodynamically stable phase and only forms during rapid cooling from high temperature marked as a gray
crossed-hatched region. The numbers inside the black circles indicate different cooling routes. The temperature error bars are determined based on the
temperature fluctuation before and after measurements, which further causes the pressure uncertainty (see “Methods” for details). b The total energy
difference per unit cell between γ- and δ-CsPbI3 (ΔE = Eγ – Eδ) under compression. c The evolution of the calculated out-of-phase tilt along [101] and the in-
phase tilt along [010] as a function of pressure. The open triangles are the previously reported tilt angles of γ-CsPbI3 at room temperature and ambient
pressure10,13. d Ball-stick models visualizing the out-of-phase and in-phase octahedral tilts. e A schematic Gibbs free energy (G) diagram of γ- and δ-CsPbI3
as a function of temperature at pressure P1 (e.g., 0.8 and 1.6 GPa) without the formation of metastable γ-CsPbI3 (black curve) and at pressure P2 (e.g., 1.2
GPa) with the formation of γ-CsPbI3 (red curve). ΔG (= Gγ – Gδ) at 0.8, 1.2, and 1.6 GPa are 0.806, 0.589, and 0.941 eV at 27 °C, 0.934, 0.614, and 1.011 eV
at 100 °C, and 1.128, 0.653, and 1.118 eV at 200 °C, respectively.

values of our diffraction peaks from the previous reports9,10 are pressure, the total energy difference between γ- and δ-CsPbI3
due to the different X-ray wavelengths used in the XRD (ΔE = Eγ – Eδ) reduces with pressure (Fig. 3b), indicating an
measurements. The Raman spectrum of the preserved black increased stabilization of γ-CsPbI3 with respect to δ-CsPbI3. ΔE
phase also agrees well with that of γ-CsPbI3 synthesized by a reaches its minimum at 0.9 GPa and remains almost invariant up
solid-state method (Supplementary Fig. 6). No Raman modes or to 1.4 GPa. With further compression, ΔE shows a sharp rise due
XRD peaks from CsI or PbI2 precursors are observed, indicating to a larger rate at which the total energy of γ-CsPbI3 increases
no decomposition of CsPbI3. relative to that of δ-CsPbI3 (Supplementary Fig. 7). DFT results
Our study further indicates that γ-CsPbI3 can only be accessed suggest that 0.9–1.4 GPa is the most suitable pressure range for
when the final pressure falls between 0.1 and 0.6 GPa. When the favoring the formation of γ-CsPbI3, spanning a small pressure
final pressure is above 0.6 GPa, CsPbI3 transforms back to the δ window of 0.5 GPa as in our experimental results. To further
phase after the P–T treatments or remains in the δ phase in the elucidate the combined effect of pressure and temperature on
entire P–T range studied, as supported by three independent preserving γ-CsPbI3, the Gibbs free energy (G = E + PV – TS)
routes covering varying P–T pathways (routes 3, 3’, and 3”). that includes the vibrational entropy at varying pressures was
Based on the XRD and Raman results, we determined a P–T calculated (Fig. 3e and Supplementary Fig. 8). It is beyond our
window marked by a gray cross-hatched area in Fig. 3a for the computational capacity to survey the entire P–T space studied.
formation and preservation of γ-CsPbI3. The key ingredient for Representative pressures of 0.8, 1.2, and 1.6 GPa were chosen
being able to metastably preserve γ-CsPbI3 to room temperature based on the total energy calculations. The results show that at
is the final pressure applied to the system, i.e., 0.1–0.6 GPa. 1.2 GPa, the Gibbs free energy difference between γ- and δ-
CsPbI3 returns to the non-perovskite δ phase after P–T cycles CsPbI3 (ΔG = Gγ – Gδ) is much smaller than that at 0.8 and 1.6
when the applied pressure falls outside this narrow range. The GPa, and it keeps reducing as the material cools from 200 °C to
cubic α-CsPbI3 phase is found to serve as an intermediate for room temperature, implying a preferred stabilization of meta-
accessing γ-CsPbI3. It is encouraging that once preserved to room stable γ-CsPbI3 towards room temperature as pressure falls in the
temperature, γ-CsPbI3 can be retained even after fully releasing window of 0.9–1.4 GPa.
pressure back to ambient conditions.

Discussion
Relative energetic stability of compressed δ- and γ-CsPbI3. After analyzing the structural details, the tilt of [PbI6]4− octa-
Although γ-CsPbI3 is a kinetically trapped metastable phase and hedra was found to correlate with the stabilization of the γ phase.
does not appear on the P–T phase diagram, energy calculations There are two types of octahedral tilts in γ-CsPbI3: an out-of-
can still shed light on the effect of pressure on the relative stability phase tilt along the [101] direction with adjacent octahedra
of the competing δ and γ phases (Fig. 3). We first performed first- turning in opposite directions and an in-phase tilt along the [010]
principles DFT calculations on these two structures as a function direction with adjacent octahedra rotating towards the same
of pressure at T = 0 K. At 0 GPa, δ-CsPbI3 has a lower total direction (Fig. 3d and Supplementary Fig. 9). The simulated out-
energy compared to γ-CsPbI3, supporting that δ-CsPbI3 is the of-phase and in-phase tilt angles at 0 GPa are comparable with
thermodynamically stable phase. With the application of the values calculated from previously reported structures10,13.

4 NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5 ARTICLE

Fig. 4 PL spectra of re-pressurizing the recovered γ-CsPbI3 perovskite and PL intensity over time of the preserved γ-CsPbI3. a PL spectra collected as a
function of pressure at room temperature. Inset: The pressure dependence of the PL energy. A pseudo-Voigt function is used to fit the PL peak that yields
the peak position and uncertainty. b Normalized PL intensity of the preserved γ-CsPbI3 perovskite over time in a diamond–anvil cell and in air at 20% and
35% RH, respectively. The error bars are defined as the largest deviation from the average value of the normalized PL intensity collected at different sample
positions.

Under compression, the out-of-phase tilt increases rapidly below humidity. The PL intensity of the recovered samples remains
0.4 GPa and continues the upward trend at a reduced rate with above 80% of the initial value for more than 30 days and up to
further compression (0.4–0.8 GPa), followed by a pressure- 10 days in air at 20% and 35% RH, respectively (Fig. 4b). After
invariant behavior of staying at ~10° between 0.9 and 1.4 GPa degradation in air over time, γ-CsPbI3 transforms back to the
(Fig. 3c). On the other hand, the in-phase tilt increases linearly yellow δ-CsPbI3. With reheating to high temperature at ambient
with pressure up to 1.4 GPa. The similar trend between the out- pressure, the preserved γ-CsPbI3 transforms to the yellow δ-
of-phase tilt and ΔE below 1.4 GPa, especially between 0.9 and CsPbI3 phase above 100 ˚C (Supplementary Fig. 11). With
1.4 GPa, suggests that the out-of-phase tilt contributes the most to applying pressure at room temperature, Raman measurements
the reduction of ΔE and consequently the preservation of γ- indicate that the preserved γ-CsPbI3 is metastable to at least 5.0
CsPbI3. At above 1.5 GPa, the in-phase tilt rises sharply, con- GPa (Supplementary Fig. 11).
current with the dramatic increase of ΔE. This indicates that at In summary, a metastable γ-CsPbI3 perovskite phase can be
higher pressures, the in-phase tilt could be the main contributing synthesized and then preserved back to room temperature via
factor for the increase of the total energy of γ-CsPbI3 that again pressure-tuning the tilt of [PbI6]4− octahedra, given a pressure
favors the formation of δ-CsPbI3. between 0.1 and 0.6 GPa is applied to the yellow δ-CsPbI3 fol-
The compression-directed tilt of [PbI6]4− octahedra is also lowed by heating and rapid quenching. The occurrence of α-
reflected from the PL measurements of re-pressurizing the pre- CsPbI3 at high temperatures likely assists the formation of γ-
served γ-CsPbI3. The preserved γ-CsPbI3 exhibits a PL peak at CsPbI3. Once formed, γ-CsPbI3 can be retained after releasing
~1.77 eV at ambient conditions (Fig. 4a), which is comparable pressure to ambient conditions and shows substantial stability in
with the value of the strain-stabilized γ-CsPbI329 and ~ 60 meV air at 35% RH for up to 10 days. First-principles DFT calculations
smaller than that of γ-CsPbI3 synthesized by a nano crystal- suggest that pressure directs the out-of-phase and in-phase
lization method5. With re-compressing the recovered γ-CsPbI3 at octahedral tilt that control the relative energy difference between
room temperature, the PL peak shows a blue shift at a rate of 25 the competing δ- and γ-CsPbI3 and ultimately stabilizes γ-
meV/GPa up to 2.6 GPa beyond which the PL signal disappears CsPbI3. Our study provides key insight into manipulating the
(Fig. 4a, inset). The pressure response is different from that phase (meta)stability of halide perovskites through structural
observed in CsPbI3 perovskite nanocrystals where the PL at first control and opens effective pathways to synthesizing metastable
shifted to lower energies below 0.4 GPa and then to higher phases with improved properties.
energies with further compression38,39. These differences could be
caused by the different structures and particle sizes of the sam- Methods
ples. However, the PL development of a CsPbI3 perovskite Synthesis of δ-CsPbI3 samples. The yellow non-perovskite δ-CsPbI3 samples
structure as a function of pressure is typically controlled by both were synthesized using a solution-based method adapted from literature42. Solid
the intra-octahedral compression and inter-octahedral tilt, with PbI2 (0.46 g, 1.0 mmol) and CsI (0.26 g, 1.0 mmol) were first mixed with aqueous
HI (10.0 mL, 7.58 M, stabilized), and then heated to 130 °C with stirring in a closed
the former resulting in a red shift and the latter causing a blue 20-mL glass vial. The solution was then cooled to room temperature at a rate of
shift39. The exclusive blue shift observed in the PL of our pre- 25–35 °C/h by turning off the power of the hot plate. The yellow δ-CsPbI3 solid
served γ-CsPbI3 supports the inter-octahedral tilt being sig- powders were collected by vacuum filtration, rinsed with large amounts of diethyl
nificantly modulated at high pressure. The band-structure ether, and dried under reduced pressure for 12 h.
calculations confirm that the increase of the out-of-phase and in-
phase tilt both enlarges the bandgap of γ-CsPbI3 with the in- High-pressure and high-temperature apparatus. High pressure was achieved
phase tilt playing a bigger role (Supplementary Fig. 10). using a BX-90 type diamond–anvil cell with an anvil culet size of 500 or 600 μm in
Further study on the functionality of the preserved γ-CsPbI3 diameter. A home-made external coiled resistive heater with a resistance of ~2.5 Ω
made of KA1 alloy wires (Hyndman Industrial Products) was placed around the
shows that this perovskite phase has substantial stability to T301 stainless-steel gasket and the tip of the diamond anvils to heat up the samples.

NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5

The exposed KA1 alloy wires were covered with 940 HT fast cure alumina adhesive expansion was neglected in our calculations, and hence the phonon dispersion and
for electrical insulation. Asbestos thermal insulation layers were used to cover the phonon density of states at 0 K were used for entropy calculation at high tempera-
remaining area of the diamonds, seats, and the resistive heater to ensure tem- tures. The α- and δ-CsPbI3 phases have been found experimentally to have the same
perature uniformity across the sample. volume expansion coefficient of αv = 1.18 × 10−4 K−1 (ref. 53). It is reasonable to
assume γ-CsPbI3 has a similar volume expansion coefficient. Within the temperature
range of interest from 200 to 27 °C where γ-CsPbI3 is metastably preserved during
Sample loading and reaching high P–T conditions. Powdered δ-CsPbI3 samples,
cooling, the volume change is about ΔV/V = αv × ΔT = 2% at ambient pressure, and
together with Au powders or a ruby ball as pressure calibrants, were loaded into
this volume reduction will be smaller at high pressures. Further considering that both
250-μm sample chambers in pre-indented T301 stainless-steel gaskets in air. No
the δ- and γ-CsPbI3 phases experience similar volume changes, we expect our cal-
pressure transmitting medium was used to ensure good contacts between samples
culations are valid for predicting the relative phase stability. Previous computational
and the diamond surface for temperature monitoring. The pressure at the room
work on CsSnI3 that studied the thermal expansion effect on the Gibbs free energy
and the high temperature was calibrated using the equation of state of Au43,44 or
the fluorescence shift from a ruby ball45. The temperature was monitored by a K- further validated our assumption54. With the current computing resources, their
calculations were done by assuming that high temperature only changes the volume
type thermocouple glued on the diamond pavilion close to the culet. The pressure
but does not change other degrees of freedom, such as tilt angles. The Gibbs free
uncertainty at high temperature was determined based on the temperature fluc-
energies of α- and γ-CsSnI3 showed a similar behavior as a function of temperature
tuation before and after the XRD and Raman measurements. A higher heating
with and without considering the thermal expansion effect54. The quasi-harmonic
temperature and a longer heating duration favor the conversion to pure γ-CsPbI3.
approximation can be taken to include the thermal expansion effect. However, it
requires constructing a set of test structures with different expanded unit cell volumes
Synchrotron XRD measurements. High-temperature and high-pressure XRD and performing phonon calculations for all the test structures to find out the structure
measurements were performed at beamline 12.2.2 of the advanced light source with the lowest Gibbs free energy. In the case of γ-CsPbI3, the structure has many
(ALS), Lawrence Berkeley National Laboratory; and beamline 16-BMD of the degrees of freedom, including at least three lattice constants, the Pb–I bond length,
Advanced Photon Source (APS), Argonne National Laboratory (ANL). Two- and the in-phase and out-of-phase octahedral tilt. The [PbI6]4− octahedron also
dimensional Debye-Scherrer diffraction rings were collected at a wavelength of λ = deviates from the ideal geometry in γ-CsPbI3, adding extra degrees of freedom for
0.4959 Å (12.2.2, ALS) and λ = 0.4133 Å (16-BMD, APS) on a Mar345 image plate structural variations. The large number of degrees of freedom in γ-CsPbI3 makes it
detector, and integrated using the Dioptas software package46, yielding intensity extremely complex to construct the test structures. It is also beyond the current
versus 2θ patterns. The intensity has been normalized for better comparison. The computational capacity to perform phonon calculations that consider the thermal
sample-to-detector distance and other parameters of the detector were calibrated expansion due to the astronomically increased number of test structures.
using the CeO2 standard.
Band-structure calculations. The band-structure calculations were performed
Raman spectroscopy measurements. Raman spectra were collected using a using Quantum Espresso with GGA exchange-correlation functional, and a 12 ×
Horiba LabRam HR Evolution Raman system at the Stanford Nano Shared 8 × 12 k-point grid. The evolution of the bandgap as a function of the in-phase/out-
Facilities (SNSF). A laser excitation wavelength of 633 nm was utilized. A threshold of-phase tilt was calculated by changing one of the tilts and fixing the other to the
power of ~0.5 mW was kept throughout the measurements to avoid the potential value at ambient conditions. Through analyzing the high-pressure DFT structures,
laser-induced heating on the samples. Three accumulations with an exposure time we found that the in-phase tilt was directly related to the ratio of the two in-plane
of 10 s per accumulation were done to obtain a good Raman spectrum. Ultra-low- lattice constants, a and c. Therefore, we tuned the in-phase tilt by increasing the
frequency filter kits (ULFK633-17-LR) were used to obtain Raman signals down to lattice constant a by 0.5%, 1.0%, and 2.0% and decreasing c while keeping the unit
10 cm−1. Before the measurements, the Raman system was calibrated using the cell volume and the fractional coordinates of the Pb and I atoms fixed. This way of
Raman mode of a silicon wafer at 520 cm−1. constructing the structures was found to effectively change the in-phase tilt angle
while minimizing the change of the out-of-phase tilt angle by at least an order of
PL measurements. The evolution of PL intensity over time was collected on magnitude smaller. In the case of tuning the out-of-phase tilt, the apical I atoms
quenched samples using the Horiba system at SNSF (λ = 532 nm) and the were found to always move within the bc plane while changing the out-of-phase
Renishaw inVia Raman system in Extreme Environments Laboratory at Stanford tilt. Hence, we tuned the out-of-phase tilt by rotating the octahedron with the
University (λ = 514.5 nm). Laser power as low as 0.1 mW and an exposure time of internal bond lengths and angles of the [PbI6]4− octahedron being fixed and
2 s per measurement were used. The PL peaks were smoothed by using the moving the apical I atoms within the bc plane.
Savitzking Golay filter to remove the noise and ruby signal. Several positions of the
γ-CsPbI3 samples were monitored with time. The PL intensity of each sample Data availability
position at a different time was first normalized to that measured at day 1, and then All data that support the findings of this study are present in the paper and the
the normalized values were averaged to get a normalized PL intensity versus time supplementary information. Additional data related to the study are available from the
curve as shown in Fig. 4b. The PL spectra as a function of pressure were obtained corresponding author upon request.
by re-compressing the recovered samples to high pressure at room temperature.

Humidity tests. Humidity tests were performed by monitoring the PL intensity of Received: 23 June 2020; Accepted: 15 December 2020;
γ-CsPbI3 over time in air at different levels of RH. The preserved γ-CsPbI3, along
with a hygrometer for humidity calibration were sealed in a glass bottle. The
hygrometer was calibrated using the saturated NaOH (~7% RH) and MgCl2 (~35%
RH) solutions at ambient conditions47. A 5% RH uncertainty was observed from
the hygrometer readings. A RH of 20% was obtained by sealing the sample, the
hygrometer, and a small number of desiccants together in a glass bottle. References
1. Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. Organometal halide
First-principles calculations. DFT calculations for the total energies of the δ- and γ- perovskites as visible-light sensitizers for photovoltaic cells. J. Am. Chem. Soc.
CsPbI3 phases under pressure were performed with Quantum Espresso48. The input 131, 6050–6051 (2009).
structures were from experimental configurations10. The structures at ambient pres- 2. Yang, W. S. et al. Iodide management in formamidinium-lead-halide-based
sure for both phases were calculated under variable cell relaxation. The structure and perovskite layers for efficient solar cells. Science 356, 1376–1379 (2017).
related total energy at high pressure were then calculated using the relaxed ambient 3. Jeon, N. J. et al. A fluorene-terminated hole-transporting material for highly
structure as the initial configuration. The pressure was applied by setting the stress to efficient and stable perovskite solar cells. Nat. Energy 3, 682–689 (2018).
a target value based on the method developed in previous studies49,50, which has been 4. Wang, Y. et al. Thermodynamically stabilized β-CsPbI3-based perovskite solar
implemented in Quantum Espresso. PBE exchange-correlation functional was chosen cells with efficiencies >18%. Science 365, 591–595 (2019).
for the exchange and correlation terms51. An 8 × 6 × 8 k-grid for the γ phase and a 5. Swarnkar, A. et al. Quantum dot-induced phase stabilization of α-CsPbI3
6 × 12 × 3 k-grid for the δ phase were used. The kinetic energy and charge density perovskite for high-efficiency photovoltaics. Science 354, 92–95 (2016).
cutoff were 75 and 500 Rydberg, respectively. The Gibbs free energy (G = E + PV – 6. Xiang, S. et al. Highly air-stable carbon-based α-CsPbI3 perovskite solar cells
TS) was calculated, where V is the optimized unit cell volume of δ- and γ-CsPbI3 at with a broadened optical spectrum. ACS Energy Lett. 3, 1824–1831 (2018).
the corresponding pressure and S is the vibrational entropy of each phase. The 7. Zhang, T. et al. Bication lead iodide 2D perovskite component to stabilize
entropy S was calculated using the following formula: inorganic α-CsPbI3 perovskite phase for high-efficiency solar cells. Sci. Adv. 3,
  1700841–1700846 (2017).
∂F 1 X hvj ðqÞ X  
hvj ðqÞ

8. Zhao, B. et al. Thermodynamically stable orthorhombic γ-CsPbI3 thin films for
S¼ ¼ hvj ðqÞcoth  kB ln 2sinh ð1Þ
∂T 2T q;j 2kB T q;j
2kB T high-performance photovoltaics. J. Am. Chem. Soc. 140, 11716–11725 (2018).
9. Sutton, R. J. et al. Cubic or orthorhombic? Revealing the crystal structure of
where vj ðqÞ is the energy of the jth phonon mode at momentum q. Phonon dis- metastable black-phase CsPbI3 by theory and experiment. ACS Energy Lett. 3,
persions were calculated at 0 K using Quantum Espresso48 and Phonopy52. Thermal 1787–1794 (2018).

6 NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5 ARTICLE

10. Marronnier, A. et al. Anharmonicity and disorder in the black phases of 41. Maalej, A. et al. Phase transitions and crystal dynamics in the cubic perovskite
cesium lead iodide used for stable inorganic perovskite solar cells. ACS Nano CH3NH3PbCl3. Solid State Commun. 103, 279–284 (1997).
12, 3477–3486 (2018). 42. Stoumpos, C. C., Malliakas, C. D. & Kanatzidis, M. G. Semiconducting tin and
11. Dastidar, S. et al. Quantitative phase-change thermodynamics and lead iodide perovskites with organic cations: phase transitions, high mobilities,
metastability of perovskite-phase cesium lead iodide. J. Phys. Chem. Lett. 8, and near-infrared photoluminescent properties. Inorg. Chem. 52, 9019–9038
1278–1282 (2017). (2013).
12. Lai, M. et al. Structural, optical, and electrical properties of phase-controlled 43. Ming, L. et al. Gold as a reliable internal pressure calibrant at high
cesium lead iodide nanowires. Nano Res. 10, 1107–1114 (2017). temperatures. J. Appl. Phys. 54, 4390–4397 (1983).
13. Straus, D. B., Guo, S. & Cava, R. J. Kinetically stable single crystals of 44. Heinz, D. L. & Jeanloz, R. The equation of state of the gold calibration
perovskite-phase CsPbI3. J. Am. Chem. Soc. 141, 11435–11439 (2019). standard. J. Appl. Phys. 55, 885–893 (1984).
14. Beal, R. E. et al. Cesium lead halide perovskites with improved stability for 45. Rekhi, S., Dubrovinsky, L. S. & Saxena, S. K. Temperature-induced ruby
tandem solar cells. J. Phys. Chem. Lett. 7, 746–751 (2016). fluorescence shifts up to a pressure of 15 GPa in an externally heated diamond
15. Swarnkar, A., Mir, W. J. & Nag, A. Can B-site doping or alloying improve anvil cell. High. Temp. -High. Press. 31, 299–305 (1999).
thermal-and phase-stability of all-inorganic CsPbX3 (X = Cl, Br, I) 46. Prescher, C. & Prakapenka, V. B. DIOPTAS: a program for reduction of two-
perovskites? ACS Energy Lett. 3, 286–289 (2018). dimensional X-ray diffraction data and data exploration. High. Press. Res. 35,
16. Wang, Y., Zhang, T., Kan, M. & Zhao, Y. Bifunctional stabilization of all- 223–230 (2015).
inorganic α-CsPbI3 perovskite for 17% efficiency photovoltaics. J. Am. Chem. 47. Young, J. F. Humidity control in the laboratory using salt solutions—a review.
Soc. 140, 12345–12348 (2018). J. Appl. Chem. 17, 241–245 (1967).
17. Hu, Y. et al. Bismuth incorporation stabilized α-CsPbI3 for fully inorganic 48. Giannozzi, P. et al. QUANTUM ESPRESSO: a modular and open-source
perovskite solar cells. ACS Energy Lett. 2, 2219–2227 (2017). software project for quantum simulations of materials. J. Phys. Condens.
18. Protesescu, L. et al. Nanocrystals of cesium lead halide perovskites (CsPbX3, X Matter 21, 395502 (2009).
= Cl, Br, and I): novel optoelectronic materials showing bright emission with 49. Nielsen, O. H. & Martin, R. M. First-principles calculation of stress. Phys. Rev.
wide color gamut. Nano Lett. 15, 3692–3696 (2015). Lett. 50, 697 (1983).
19. Hoffman, J. B., Schleper, A. L. & Kamat, P. V. Transformation of sintered 50. Nielsen, O. H. & Martin, R. M. Quantum-mechanical theory of stress and
CsPbBr3 nanocrystals to cubic CsPbI3 and gradient CsPbBrxI3–x through force. Phys. Rev. B 32, 3780 (1985).
halide exchange. J. Am. Chem. Soc. 138, 8603–8611 (2016). 51. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation
20. Luo, P. et al. Solvent engineering for ambient-air-processed, phase-stable made simple. Phys. Rev. Lett. 77, 3865 (1996).
CsPbI3 in perovskite solar cells. J. Phys. Chem. Lett. 7, 3603–3608 (2016). 52. Togo, A. & Tanaka, I. First principles phonon calculations in materials
21. Fu, Y. et al. Stabilization of the metastable lead iodide perovskite phase via science. Scr. Mater. 108, 1–5 (2015).
surface functionalization. Nano Lett. 17, 4405–4414 (2017). 53. Trots, D. M. & Myagkota, S. V. High-temperature structural evolution of
22. Wang, P. et al. Solvent-controlled growth of inorganic perovskite films in dry caesium and rubidium triiodoplumbates. J. Phys. Chem. Solids 69, 2520–2526
environment for efficient and stable solar cells. Nat. Commun. 9, 2225 (2018). (2008).
23. Eperon, G. E. et al. Inorganic caesium lead iodide perovskite solar cells. J. 54. Silva, E. L., Skelton, J. M., Parker, S. C. & Walsh, A. Phase stability and
Mater. Chem. A 3, 19688–19695 (2015). tranformations in halide perovskite CsSnI3. Phys. Rev. B. 91, 144107
24. Li, B. et al. Surface passivation engineering strategy to fully-inorganic cubic (2015).
CsPbI3 perovskites for high-performance solar cells. Nat. Commun. 9, 1076
(2018).
25. Wu, T. et al. Efficient and stable CsPbI3 solar cells via regulating lattice distortion Acknowledgements
with surface organic terminal groups. Adv. Mater. 31, 1900605 (2019). We thank Martin Kunz, Jinyuan Yan, and Andrew Doran of Beamline 12.2.2 at the
26. Wang, K. et al. All-inorganic cesium lead iodide perovskite solar cells with Advanced Light Source, and Matthew Smith at Stanford University for experimental
stabilized efficiency beyond 15%. Nat. Commun. 9, 4544 (2018). assistance. This work was supported by the Department of Energy (DOE), Office of
27. Sun, J.-K. et al. Polar solvent induced lattice distortion of cubic CsPbI3 Science, Basic Energy Sciences, Materials Sciences and Engineering Division (DE-AC02-
nanocubes and hierarchical self-assembly into orthorhombic single-crystalline 76SF00515). Beamline 12.2.2 is a DOE Office of Science User Facility under contract No.
nanowires. J. Am. Chem. Soc. 140, 11705–11715 (2018). DE-AC02-05CH11231. Portions of this work were performed at HPCAT (Sector 16),
28. Wang, Y. et al. Efficient α-CsPbI3 photovoltaics with surface terminated APS, ANL. HPCAT operations are supported by DOE-NNSA’s Office of Experimental
organic cations. Joule 2, 2065–2075 (2018). Sciences. The APS is a DOE Office of Science User Facility operated for the DOE Office
29. Steele, J. A. et al. Thermal unequilibrium of strained black CsPbI3 thin films. of Science by ANL under Contract No. DE-AC02-06CH11357. Portions of this work
Science 365, 679–684 (2019). were performed at the Stanford Nano Shared Facilities, supported by the National Sci-
30. Masi, A., Gualdron-Reyes, A. F. & Mosa-Sero, I. Stabilization of black perovskite ence Foundation under award ECCS-1542152. First-principles DFT calculations were
phase in FAPbI3 and CsPbI3. ACS Energy Lett. 5, 1974–1985 (2020). supported by the resources of the National Energy Research Scientific Computing Center
31. Liu, G. et al. Isothermal pressure-derived metastable states in 2D hybrid (NERSC), a DOE Office of Science User Facility operated under Contract No. DE-AC02-
perovskites showing enduring bandgap narrowing. Proc. Natl Acad. Sci. USA 05CH11231. N.R.W. was partially supported by a Stanford Interdisciplinary Graduate
115, 8076–8081 (2018). Fellowship.
32. Li, Q. et al. High-pressure band-gap engineering in lead-free Cs2AgBiBr6
double perovskite. Angew. Chem. 129, 16185–16189 (2017). Author contributions
33. Wang, Y. et al. Pressure-induced phase transformation, reversible F.K. and Y.L. designed the project and wrote the paper. F.K, C.W., J.Y., S.N, W.L.M., and
amorphization, and anomalous visible light response in organolead bromide Y.L. conducted the experiments and analyzed the data. N.R.W. synthesized the sample
perovskite. J. Am. Chem. Soc. 137, 11144–11149 (2015). under H.I.K.’s supervision. C.J and T.P.D. performed the calculations. All authors con-
34. Jaffe, A. et al. High-pressure single-crystal structures of 3D lead-halide hybrid tributed to the discussion and revision of the paper.
perovskites and pressure effects on their electronic and optical properties. ACS
Cent. Sci. 2, 201–209 (2016).
35. Jaffe, A., Lin, Y. & Karunadasa, H. I. Halide perovskites under pressure: Competing interests
accessing new properties through lattice compression. ACS Energy Lett. 2, The authors declare no competing interests.
1549–1555 (2017).
36. Jaffe, A., Lin, Y., Mao, W. L. & Karunadasa, H. I. Pressure-induced
metallization of the halide perovskite (CH3NH3)PbI3. J. Am. Chem. Soc. 139,
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41467-
4330–4333 (2017).
020-20745-5.
37. Fang, Y. et al. Pressure-induced emission (PIE) and phase transition of a two-
dimensional halide double perovskite (BA)4AgBiBr8 (BA=CH3(CH2)3NH3+).
Correspondence and requests for materials should be addressed to Y.L.
Angew. Chem. Int. Ed. 131, 15393 (2019).
38. Cao, Y. et al. Pressure-tailored band gap engineering and structure evolution
Peer review information Nature Communications thanks the anonymous reviewers for
of cubic cesium lead iodide perovskite nanocrystals. J. Phys. Chem. C. 122,
their contribution to the peer review of this work. Peer reviewer reports are available.
9332–9338 (2018).
39. Beimborn, J. C., Hall, L. M., Tongying, P., Dukovic, G. & Weber, J. M.
Reprints and permission information is available at http://www.nature.com/reprints
Pressure response of photoluminescence in cesium lead iodide perovskite
nanocrystals. J. Phys. Chem. C. 122, 11024–11030 (2018).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
40. Yaffe, O. et al. Local polar fluctuations in lead halide perovskite crystals. Phy.
published maps and institutional affiliations.
Rev. Lett. 118, 136001 (2017).

NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20745-5

Open Access This article is licensed under a Creative Commons

Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.

© The Author(s) 2021

8 NATURE COMMUNICATIONS | (2021)12:461 | https://doi.org/10.1038/s41467-020-20745-5 | www.nature.com/naturecommunications