G.
Grancini‡
*a,
V.
D'Innocenzo
ab,
E. R.
Dohner
c,
N.
Martino
ab,
A. R.
Srimath Kandada
a,
E.
Mosconi
d,
F.
De Angelis
d,
H. I.
Karunadasa
c,
E. T.
Hoke
e and
A.
Petrozza
*a
aCenter for Nano Science and Technology@Polimi, Istituto Italiano di Tecnologia, via Giovanni Pascoli 70/3, 20133, Milan, Italy. E-mail: giulia.grancini@epfl.ch; annamaria.petrozza@iit.it
bDipartimento di Fisica, Politecnico di Milano, Piazza L. da Vinci, 32, 20133 Milano, Italy
cDepartment of Chemistry, Stanford University, 337 Campus Drive, Stanford, California 94305, USA
dComputational Laboratory for Hybrid/Organic Photovoltaics (CLHYO), CNR-ISTM, Via Elce di Sotto 8, I-06123, Perugia, Italy
eDepartment of Materials Science and Engineering, Stanford University, 476 Lomita Mall, Stanford, California 94305, USA
First published on 17th September 2015
Here we identify structural inhomogeneity on a micrometer scale across the surface of a CH3NH3PbI3 perovskite single crystal. At the crystal edge a local distortion of the crystal lattice is responsible for a widening of the optical bandgap and faster photo-carrier recombination. These effects are inherently present at the edge of the crystal, and further enhanced upon water intercalation, as a preliminary step in the hydration of the perovskite material.
Fig. 1 (a) Top view SEM image of a mm-sized MAPbI3 single crystal (see Fig. S2† for XRD data); (b) enlarged view of the SEM in the top-left area of (a); (c) higher-resolution image of the SEM in the top-left area of (b) where spot A (towards the centre of the crystal face) and B (at the edge of the crystal face) have been measured, respectively; (d) PL spectra on spot A and B upon excitation at 690 nm (fluence ∼ 10 μJ cm−2) under a N2 atmosphere; (e) PL time decays at 770 nm (peak of point B PL spectrum) and 800 nm (peak of point A PL spectrum), solid lines represent the fits (y = 0.5exp(−t/τ1) + 0.2exp(−t/τ2), τ1 = 630 ns, τ2 = 40 ns fits the decay at 800 nm; y = 0.2exp(−t/τ1) + 0.44exp(−t/τ2), τ1 = 500 ns, τ2 = 38 ns fits the decay at 770 nm). (f) Raman spectra from point B (blue squares) and point A (red dots) along with fitting results (solid lines). The measurement has been done under a N2 atmosphere. |
Considering that the excitation wavelength of 690 nm has an absorption depth of approximately 250 nm, surface effects (e.g., dangling bonds, under-coordinated atoms) will dominate the micro-PL measurement, certainly significant when probing the surface edges. We suggest that these edge surface effects are responsible for the blue-shifted, faster decaying component of the PL. In polycrystalline perovskite thin films it has been demonstrated11,12 that deformation of the inorganic cage through interactions with the organic cations can compress the lattice, resulting in an increased bandgap.11–15 Specific vibrational modes have been assigned as evidence for these lattice structural distortions.11,16,17
To better understand the origin of the spatially inhomogeneous PL, we measured the micro-Raman spectra at a ∼1 μm scale resolution on exactly the same points at which the micro-PL spectra were measured (Fig. 1f; see the Experimental section for details on the set-up). Excitation at the centre (point A) and the edge (point B) of the crystal face leads to different spectra, suggesting a restructuring of the crystal surface at the edges. The former shows a peak at 109 cm−1, assigned to the Pb–I stretching mode17 and a broad band at 250 cm−1, previously assigned to the torsional mode of the MA cation.11,17 When moving to the edge, we observe that the frequency of the Pb–I stretching mode is blue-shifted (towards 120 cm−1) and the peak gains intensity, while the band at 250 cm−1 slightly shifts and gets more intense. A similar trend has been previously observed for MAPbI3 nanocrystals grown within a mesoporous oxide scaffold, which were reported to be more distorted with respect to micrometer-size MAPbI3 crystallites deposited on a flat glass substrate.11 This result suggests that lattice strain at the crystal edges can be responsible for the blue-shifted PL. The above observations indicate that the edges of the MAPbI3 single crystal differ optically, energetically and structurally from the centre of the crystal face. This is not surprising, considering that defects and lattice reconstructions are common at crystal edges. Note also that the micro-PL spectra we measure at the crystal edges, rather than the crystal centre PL, more closely resemble the usual PL spectra observed for MAPbI3 polycrystalline thin films.10,11,19 Thus, we believe that the distorted crystal surface provides a model for the properties of polycrystalline thin films with small grains where surface effects at high defect densities, grain boundaries and active interfaces dominate and dramatically impact the device operation.18
Because the surface is exposed to the surrounding atmosphere, we investigated how the optical properties of the single crystal surface are sensitive to environmental interaction, in particular to the presence of moisture. The interaction between MAPbI3 and water is of particular concern as moisture has been shown to dramatically impact not only the perovskite film growth and crystallization dynamics20–25 but also accelerate material degradation.25–28Fig. 2 shows the macro-PL spectra measured on the whole crystal surface (a) under vacuum, (b) under low humidity (dry air with water content <10 ppm mol−1) and (c) under high humidity (ambient conditions, ∼50% relative humidity) (see Fig. S3† for a cartoon of the excitation–collection geometry).
Fig. 2 Normalized PL spectra (a–c) and dynamics (d and e, respectively) at the wavelengths indicated by the arrows in (a–c), collected under vacuum (a), under a low humidity atmosphere (b) and under high humidity (c); fluence ∼ 10 μJ cm−2. The solid lines in (2d and 2e) represent exponential fits: (y = 0.9exp(−t/τ1) + 0.1exp(−t/τ2)), τ1 = 35 ns τ2 = 240 ns fits the decay in vacuum; y = 0.95exp(−t/τ1) + 0.05exp(−t/τ2), τ1 = 26 ns, τ2 = 200 ns fits the decay under low humidity and y = exp(−t/τ), τ = 22 ns fits the decay under high humidity. (f) Comparison of the Raman spectra at point B, under dry N2versus under a high humidity environment. |
Here we are probing the signal from the whole crystal surface, including the edges. As shown, when under vacuum (Fig. 2a), the PL spectrum appears broad and asymmetric. Upon increasing exposure to humidity, the blue component becomes dominant (Fig. 2b and c) resulting in a sharp spectrum (FWHM ∼ 22 nm) peaking at 770 nm and the PL decay dynamics are shortened (Fig. 2d and e) showing a complete quenching of the longer-wavelength emission under ambient humidity. The PL spectra were taken within a few minutes of changing the test chamber atmosphere, indicating that the spectral changes are rapid. Such behaviour indicates that under ambient conditions, the reconstruction of the crystal surface induced by lattice distortion becomes more significant.
We hypothesize that the strained surface region grows and extends deeper into the crystal upon moisture exposure. Note that this phenomenon is partially reversible as shown by the recovery of the PL decay dynamics after storing the sample back under vacuum (Fig. S4†). To validate our hypothesis that exposure to the environment leads to a local distortion at the crystal surface we measured the micro-Raman spectra at the edge of the crystal in the presence of ambient humidity (Fig. 2f). In particular, the peak at 109 cm−1 shifts to lower energy and increases in intensity giving evidence for local distortion.11,17 Concomitantly, the MA cation mode at 250 cm-1 red shifts as well. Overall, this behaviour can be the result of water molecule interactions through hydrogen bonding with the organic cations and the halide ions. This is particularly relevant at the crystal edge. The centre of the crystal face, indeed, is less affected by interaction with humidity (see Fig. S5 in the ESI†).
We remark here that the widening of the band gap observed upon moisture exposure along with structural distortion happening at the crystal edges are distinct from the optical and structural features related to perovskite conversion into its hydrated phases.21,23,26 Hydrated forms of MAPbI3, such as the (CH3NH3)4PbI6·2H2O compound consisting of isolated PbI64− octahedra,21,23,26,29 are indeed pale yellow in colour, and exhibit a bandgap of over 3 eV and a zero-dimensional structure.21,23,26 We therefore believe that what we observe is likely an initial step of structural reorganization towards the conversion to a hydrated phase.
As evidenced by ab initio molecular dynamics simulations,30 water molecules can permeate across the perovskite structure to form a partly hydrated phase. Based on this preliminary observation, we further computationally simulated the possible phases obtained from stepwise perovskite hydration. To investigate how water and oxygen molecules can interact with the hybrid perovskite we calculated the formation energies for 4MAPbI3·nH2O and 4MAPbI3·O2, (Fig. 3a and b, respectively), by scalar-relativistic Perdew–Burke–Ernzerhof (PBE) calculations employing van der Waals corrections. Starting from a tetragonal unit cell, consisting of four MAPbI3 units, we sequentially add either one oxygen or up to four water molecules. The calculated relaxed atomic positions for 4MAPbI3·H2O and 4MAPbI3·O2 are shown in Fig. 3. Both structures exhibit an increase in cell volume of ∼11–12 Å3 with respect to MAPbI3. While the incorporation of one water molecule preserves the original tetragonal cell, with a ≈ b < c cell parameters, O2 interaction with MAPbI3 produces a sizable elongation/contraction of the a/b cell parameters (a = 8.61, b = 8.81, c = 12.82), respectively. Interestingly, as shown in Table S1 in the ESI,† the formation energy of 4MAPbI3·1H2O is negative, (−0.45 eV) indicating spontaneous water incorporation into the perovskite lattice. On the other hand, the incorporation of molecular oxygen is unfavourable, with a positive ΔEform of 0.17 eV.
Fig. 3 Optimized geometry structures for 4MAPbI3·1H2O (a) and 4MAPbI3·1O2 (b). Pb = light blue, I = violet, green = carbon, blue = nitrogen, red = oxygen, white = hydrogen. |
The driving force for water incorporation into the perovskite lattice is primarily the formation of hydrogen bonds between the water molecules and lattice iodides, as well as the tight hydrogen bonding interaction between one of the MA cations and the water oxygen atom, as signalled by a ∼1.6 Å separation distance between oxygen and one hydrogen atom of the MA cation (Fig. 3a). Notably, O2 interacts only through weaker hydrogen bonds with two vicinal MA cations, reflecting its lower tendency to be intercalated into the perovskite lattice. Water as well as oxygen (although less likely) incorporation into MAPbI3 increases the band gap by 50 and 90 meV, respectively, estimated here at the scalar-relativistic level of theory used for geometry optimization (see Table S1† for further details). While this level of theory is only able to accurately calculate the absolute bandgap values of lead halide perovskites due to a fortuitous cancelation of spin–orbit coupling effects,12 the calculated variations in bandgap are still valid. Subsequent incorporation of water molecules into MAPbI3 was also investigated up to 4:4 stoichiometries (see Table S1 and Fig. S6†). The results show that incorporation of a second water molecule has an essentially additive effect on the perovskite properties, with a slight additional volume and bandgap increase as a result of lattice distortion. However, adding four water molecules destabilizes the perovskite lattice, resulting in a large band gap increase. This is consistent with reports of a large band gap hydrated MAPbI3 phase with a non-perovskite structure.21,23,26,29 These findings support our experimental observations: the crystal edge surface, with a certain degree of distortion with respect to the bulk structure, is prone to water intercalation, resulting in a widening of the bandgap.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5sc02542g |
‡ Present address: EPFL Valais Wallis, Rue de l'Industrie 17, CH-1951 Sion, Switzerland. |
This journal is © The Royal Society of Chemistry 2015 |