Forum Article pubs.acs.org/IC
Chlorine Incorporation in the CH3NH3PbI3 Perovskite: Small Concentration, Big Effect Claudio Quarti,†,‡ Edoardo Mosconi,*,†,∥ Paolo Umari,§,⊥ and Filippo De Angelis*,†,∥ †
Computational Laboratory for Hybrid/Organic Photovoltaics, CNR-ISTM, via Elce di Sotto 8, I-06123 Perugia, Italy Dipartimento di Fisica e Astronomia, Università di Padova, via Marzolo 8, I-35131 Padova, Italy ⊥ Theory@Elettra Group, c/o Sincrotrone Trieste, Area Science Park,CNR-IOM DEMOCRITOS, I-34012 Basovizza, Trieste, Italy ∥ CompuNet, Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genova, Italy §
S Supporting Information *
ABSTRACT: The role of chlorine doping in CH3NH3PbI3 represents an important open issue in the use of hybrid perovskites for photovoltaic applications. In particular, even if a positive role of chlorine doping on perovskite film formation and on material morphology has been demonstrated, an inherent positive effect on the electronic and photovoltaic properties cannot be excluded. Here we carried out periodic density functional theory and Car− Parrinello molecular dynamics simulations, going down to ∼1% doping, to investigate the effect of chlorine on CH3NH3PbI3. We found that such a small doping has important effects on the dynamics of the crystalline structure, both with respect to the inorganic framework and with respect to the cation libration motion. Together, we observe a dynamic spatial localization of the valence and conduction states in separated spatial material regions, which takes place in the 10−1 ps time scale and which could be the key to ease of exciton dissociation and, likely, to small charge recombination in hybrid perovskites. Moreover, such localization is enhanced by chlorine doping, demonstrating an inherent positive role of chlorine doping on the electronic properties of this class of materials. the field. Meso-superstructured and planar heterojunction solar cells were successfully implemented with MAPbI3−xClx,3,13 while the same devices based on the prototype MAPbI3 perovskite showed comparably lower performances. This behavior, initially interpreted on the basis of the improved carrier mobility of MAPbI3−xClx compared to MAPbI3,14 was later ascribed to a reduced carrier recombination.15 A possibly related phenomenon is the observation of improved charge transport in dye-sensitized solar cells based on MAPbI3−xClx and MAPbI3−xBrx,16,17 leading to enhanced photocurrent and/ or open-circuit voltage compared to MAPbI3. Moreover, recently energy-dispersive X-ray spectroscopy showed a positive correlation between the chlorine concentration and regions of brighter photoluminescence (PL), whereas PL imaging revealed that chemical treatment with pyridine could activate previously dark grains.18 However, this issue presented a great number of fundamental open questions. For instance, the actual concentration of the chlorine in MAPbI3−xClx has been unknown for a long time, being estimated around 33% in the seminal work by Lee et al.13 and then decreased to an estimation of less than ∼4%16 and ∼1%.19 The role of chlorine doping at the perovskite/TiO2
1. INTRODUCTION Hybrid lead halide perovskites are revolutionizing the landscape of emerging photovoltaic technologies. From their first application in 2009 by Kojima et al. as solar cell sensitizers,1 photovoltaic devices based on these materials showed a fast and continuous increase in their efficiency,2−8 with recent certified efficiency exceeding 22%.7,8 Organohalide lead perovskites can be solution-processed at low temperature9 and vapordeposited,3 realistically holding the promise to reach efficiencies comparable to those of typical thin-film photovoltaic technologies. Furthermore, these materials can be combined with organic electron acceptors/donors to deliver flexible photovoltaic devices.10,11 Perhaps the most intriguing feature of hybrid perovskites is the fact that they can support both electron and hole transport.9,12,13 Thus, on the one hand, these materials can be employed as solar cell sensitizers when deposited on ntransporting mesoporous TiO2, as in traditional dye-sensitized solar cells, and, on the other hand, meso-superstructured and planar heterojunction solar cells can be fabricated, in which the perovskite serves both as a light absorber and as an electron transporter. A depleted perovskite layer deposited on mesoporous TiO2 was also shown to transport holes directly to a gold cathode evaporated on top of the perovskite.12 Methylammonium lead iodide, hereafter MAPbI3, and the related mixed-halide MAPbI3−xClx analogue have dominated © XXXX American Chemical Society
Special Issue: Halide Perovskites Received: July 21, 2016
A
DOI: 10.1021/acs.inorgchem.6b01681 Inorg. Chem. XXXX, XXX, XXX−XXX
Forum Article
Inorganic Chemistry
Figure 1. (a) Structural model for the pristine MAPbI3 perovskite and for the 8% chlorine-doped species, with chlorine in the apical position. (b) Models for the pristine MAPbI and for the 1% chlorine-doped perovskite, with chlorine at both the apical (MAPbCl-ap) and equatorial (MAPbICleq) positions.
significantly modify the electronic band gap of the perovskite. Ab initio molecular dynamics simulations performed at 330 K to include the inherent effect of structural disorder on the electronic properties of the perovskite have also been carried out. Car−Parrinello molecular dynamics (CPMD), performed on pure and ∼1% chlorine-doped MAPbI3 perovskites, show the important effects of chlorine both on the crystalline structure of the material and on the reorientational motion of the methylammonium (hereafter MA) cations. Consistently, a spatial localization of the valence and conduction electronic states is observed, which can be reasonably associated with the ease of charge separation and maybe the reduction of charge recombination. Moreover, we observe that such an effect is enhanced in the presence of chlorine doping. In this perspective, an inherent positive role of a small chlorine doping in hybrid perovskite can be invoked, consisting of an easier exciton separation and lower charge recombination, in line with the experimental evidence.15
interface has also been investigated, demonstrating that the presence of interfacial chlorine induces an improvement of carrier transport across the heterojunction interfaces20 and a higher charge-separation efficiency upon photoexcitation.19,21 Only recently were X-ray photoelectron spectroscopy investigations allowed to set an upper chlorine content of ∼0.1%,22,23 which conceptually makes the chlorine a dopant or an impurity. These investigations demonstrated also a preferential segregation of the chlorine-rich phase at the perovskite surface, with the chlorine concentration ranging from 0.2%22 to 0.7%,23 in line with previous studies that predicted the formation of a solid MAPbI3/MAPbCl3 solution being thermodynamically unfavorable.24,25 Recent papers highlighted also a positive role of chlorine in perovskite film formation, resulting in improved control over the sample morphology, thickness, and coverage.26−28 Morover, previous calculations demonstrated that the carrier relaxation time can be considerably increased by mixing halogen atoms in the perovskite material,29 and the presence of chlorine doping reduces electron−hole recombination.30 An inherent effect due to small chlorine incorporation on the bulk properties of the hybrid perovskite can thus not be excluded, which could possibly contribute to the fundamental electronic processes taking place in this class of materials. Here, we perform a theoretical investigation to address the possible effect arising from chlorine doping on the electronic properties of the MAPbI3 perovskite. At zero temperature, state-of-the-art many-body perturbation theory calculations, including spin−orbit coupling (SOC),31 reveal slightly higher band gaps and increased carrier effective masses for ∼8% levels of chlorine doping, compared to the pure MAPbI3 perovskite, suggesting that this doping level could be detrimental for solar cell operation. On the contrary, ∼1% chlorine doping does not
2. MODELS Our models originated from an optimized MAPbI3 structure,32 closely related to the tetragonal I4/mcm or I4/cm crystal structures experimentally reported.33,34 Here, we considered two sets of structural models to investigate different doping levels: (i) the first set of models consists of one tetragonal crystal cell (a = b = 8.8556 Å and c = 12.66 Å),35 containing four MAPbI3, allowing for a doping level of ∼8% (see Figure 1a); (ii) the second set of models consists of a 2 × 2 × 2 supercell of the tetragonal crystalline cell, containing 32 MAPbI3 units and 384 atoms, allowing for a doping level of ∼1% (see Figure 1b). We considered three types of systems, in relation to chlorine doping. The first is the pristine compound, hereafter named MAPbI, which is taken as the reference B
DOI: 10.1021/acs.inorgchem.6b01681 Inorg. Chem. XXXX, XXX, XXX−XXX
Forum Article
Inorganic Chemistry
Table 1. Band Gaps (Eg, in eV) and Effective Holes and Electronic Masses (mh and me) of the Investigated Systems SOC-GW
SR-DFT (SOC-DFT)
expt cella
opt cell
expt cella,b
opt cell
a = b = 8.86 and c = 12.66
a = 8.78, b = 8.75, and c = 12.70
a = b = 8.86 and c = 12.66
a = 8.76, b = 8.80, and c = 12.50
Eg (eV) SOC-GW model 4 MAPbI3 units
system
expt
MAPbI3 8% Cl ap
1.6 −1.7 c
expt cell d
a
SR-DFT (SOC-DFT)
opt cell
1.67
f
model 4 MAPbI3 units 32 MAPbI3 units
a
system
expt
MAPbI3 8% Cl ap MAPbI3 8% Cl ap 1% Cl eq 1% Cl ap
μ = 0.09−0.15
expt cell e
opt cell
1.56 1.65f (0.60)f 1.77 mh/me/μ (m0)
SOC-GW a
expt cella,b
1.55 (0.57) 1.72 (0.64)
SR-DFT (SOC-DFT) expt cella,b
opt cell
0.36/0.73/0.23 (0.28/0.17/0.10)
− (0.27/0.18/0.11) − (0.28/0.19/0.11) 1.60 (0.65) 1.74 1.52 1.61b (0.64)
opt cell
0.25/0.19/0.11 0.29/0.21/0.12
1.6c−1.7d
1.67b 1.70 1.68
Reference 35. bReference 40. cReference 41. dReference 42. eReference 43. fReference 31.
system. In addition, we consider two different kinds of doping in relation to the fact that chlorine substitution in the PbI6 octahedra takes place in the apical position, which is along the unique axis (the c axis), or in the equatorial position, that is, orthogonal with respect to the symmetry unique c axis. The former and latter models are respectively named MAPbICl-ap and MAPbICl-eq hereafter (see Figure 1b).
As previously noted, SR-DFT fortuitously reproduces SOCGW and experimental band gaps, while SOC-DFT delivers underestimated band gaps.31,38,39 The difference between the SR-DFT and SOC-DFT values is almost constant (∼1 eV); thus, while SOC effects can surely modulate the band gap, SRDFT results can be employed for qualitative comparative analyses. This is important for the largest systems, for which converged SOC-GW calculations are impractical. Upon going down to 1% chlorine doping, we observe that the SR-DFT band gaps of MAPbI3−xClx and MAPbI3 are comparable, with a general trend of only slightly higher band gaps for the 1% chlorine-doped structures, in line with recent low-temperature UV−vis spectra.41 For this reason, models with 1% chlorine doping can be considered to be reliable models to investigate the effect of chlorine doping at room temperature (section 3.3). It is also worth mentioning that recent experimental investigations demonstrated larger chlorine doping at the interfacial region,22,44 with an estimated doping of around 0.7%.23 Thus, the 1% chlorine-doping models investigated here can be considered as realistic models for the material region close to the interfaces. 3.2. Effect of Chlorine on the Thin-Film Properties. As in ref 45, we studied the effect of chlorine doping on the electronic properties of perovskite thin films at zero temperature. To simulate thin films, we considered finite models in the direction of the film thickness (with a length of six pseudocubic unities) and periodic in the two perpendicular directions. For MAPbI, MAPbICl-ap, and MAPbICl-eq, we consider film models exposing both the [001] and [110] surfaces. The layered electronic DOSs of the these film models are reported in Figure 2. For all of the systems along both [001] and [110], we observe a clear gradient of the valence band maximum (VBM) and conduction band minimum (CBM) levels with respect to the film thickness, which is reasonably associated with the orientation of the MA cations. A similar result has been found in ref 45, where this bending of the electronic levels was definitely associated with the partial alignment of all of the MA cations in the same direction of the thickness. In Table 2, we report the computed gradient of the electronic VBM and CBM states, for the pristine MAPbI compound and for the
3. RESULTS 3.1. Electronic Properties at Zero Temperature. The smallest considered systems, reported in Figure 1a, are employed to extract reliable band-gap information and effective carrier masses by GW calculations36,37 and to calibrate the density functional theory (DFT) calculations employed for the larger models. While the structural properties of the investigated systems are nicely reproduced by scalar-relativistic (SR) DFT,24,31 the inclusion of SOC is crucial for a correct electronic structure description of lead perovskites.31,38,39 A survey of the electronic structure results is reported in Table 1, while calculated cell parameters, band structures, and densities of states (DOSs) are reported in the Supporting Information. As can be noticed, our SOC-GW-calculated band-gap values are in good agreement with experimentally available data for MAPbI3 (1.6−1.7 eV).41,42 Calculated reduced masses for MAPbI3 are also in good agreement with the measured values in the range 0.09−0.15,43 suggesting an accurate description of the electronic properties of these materials. Our results clearly suggest that a 8% chlorine doping should provide a ∼0.2 eV blue shift in the band gap compared to MAPbI3 (1.77 vs 1.56 eV), while experimentally the MAPbI3‑xClx perovskite shows roughly the same absorption onset as MAPbI3.41 Carrier effective masses follow the band-gap trends and show slightly increased values for 8% chlorine doping compared to MAPbI3. Thus, on the basis of the calculated electronic structure, 8% chlorine doping should impart transport and light-harvesting properties inferior to those of the mixed halide perovskite compared to MAPbI3, which is not consistent with what was experimentally observed, further confirming that the actual chlorine concentration in real samples is much smaller.22,23 C
DOI: 10.1021/acs.inorgchem.6b01681 Inorg. Chem. XXXX, XXX, XXX−XXX
Forum Article
Inorganic Chemistry
Figure 2. Local DOS computed for finite slabs originating from the bulk MAPbI, MAPbICl-ap, and MAPbICl-eq compounds, along both [001] and [110]. The structural models employed for the DOS calculations are also reported.
scale,35,47 has recently been associated with peculiar electronic properties, as a more effective screening of the electron−hole pair,48 a dynamically disorder-driven charge separation,49,50 and the formation of local ferroelectric domains in operative conditions.51 All of these phenomena are expected to have important effects on the photovoltaic performances of this class of materials. Thus, to take into account the inherent structural motion at room temperature, especially the MA cation librations, on the electronic properties of hybrid perovskites, we resorted to CPMD, considering the large models (32 MAPbI3 units) of MAPbI, MAPbICl-ap, and MAPbICl-eq systems (see Figure 1b). Starting from the DFT-optimized structures made by 32 MAPbI3 units (Figure 1b), we considered 12 ps of molecular dynamics at 330 K, after a few picoseconds of thermalization. We start by analyzing the effect of chlorine doping on the material structure, considering both the inorganic framework and the librations and reorientations of the MA cations. In Figure 3, we compare the theoretical radial distribution function (RDF) associated with lead and iodine atoms for
Table 2. Spatial Gradient (meV/Å) of the VBM and CBM Electronic Levels, Computed on the Finite Thin Films for the MAPbI, MAPbICl-ap, and MAPbICl-eq Compounds MAPbI MAPbICl-ap MAPbICl-eq
[001]
[110]
41.4 41.7 45.8
44.4 31.3 44.6
chlorine-doped MAPbI-ap and MAPbI-eq compounds, along both [001] and [110]. The gradients in all cases are very similar, except for MAPbICl-ap along [110], which is slightly smaller than the other cases. The present results thus point out a very small contribution of the chlorine doping on the spatial gradient of the VBM and CBM levels in this class of materials. 3.3. Material Structure and Electronic Properties at Room Temperature. Hybrid perovskites are inherently complex systems. In particular, it is well-known that the MA cations are disordered at room temperature,35,46 and their reorientational motion, taking place in the picosecond time D
DOI: 10.1021/acs.inorgchem.6b01681 Inorg. Chem. XXXX, XXX, XXX−XXX
Forum Article
Inorganic Chemistry
Figure 3. Theoretical RDF associated with the iodine and lead atoms of the pristine MAPbI and of the chlorine-doped MAPbICl-ap and MAPbICl-eq. The experimental data from ref 52 are also reported.
the pristine MAPbI and the chlorine-doped MAPbICl-ap and MAPbICl-eq. The experimental data for the pristine compound are also reported.52 The theoretical data of all three structures agree quite well with the experiment. Also, the peak broadening, which is useful information directly provided by molecular dynamics simulation, is well reproduced. A small signal in the experimental RDF is found around 5.6 Å, which is not paralleled by theory. Note, however, that such a signal falls between two well-resolved signals of the inorganic framework, at 4.5 and at 6.4 Å, assigned respectively to the first I−I coordination shell and to the second Pb−I and I−I coordination shells. Thus, it is not unreasonable to associate the signal at 5.6 Å with some nonideality with respect to the crystalline structure, as structural defects or contributions from the interface. In particular, the experimental data in ref 52 have been measured on a mesostructured perovskite sample, thus making the latter interpretation quite probable. Dealing with the effect of chlorine doping on the material structure, we note that the RDF of MAPbICl-ap is very similar to that of the pristine MAPbI. The RDF of MAPbICl-eq instead shows broader peaks, especially for distances larger than 6 Å. Such a broadening is informative of a more disordered structure of MAPbICl-eq with respect to the pristine MAPbI and/or to a large degree of dynamical disorder associated with the thermal vibrations of the inorganic framework. To understand the type of structural deformation induced by chlorine doping in the equatorial position, we proceeded as follows. Perovskites, in general, are characterized by a specific pattern of rotations of the PbI6 octahedral along the three pseudocubic axes, as discussed in detail by Glazer.53 An ideal cubic perovskite is characterized by an a0a0a0 structure, without rotations of the PbX6 octahedral (X = halogen). This situation corresponds to a zero value of the X−Pb−Pb−X dihedral angle (see Figure 4a). MAPbI3 presents an a0a0c− pattern for the octahedral tilting,33,34 i.e., with alternation of positive and negative I−Pb−Pb−I dihedral angles only along the c axis (see Figure 4a). The values assumed by all of the X−Pb−Pb−X (X = Cl, I) dihedral angles within the crystalline cell, during the 12 ps CPMD simulation, have been evaluated for the three investigated systems, and they have been reported in Figure 4b for the three pseudocubic axes [110], [1−10], and [001]. For the pristine MAPbI perovskite, the X−Pb−Pb−X dihedral angle along the [110] and [1−10] directions oscillates around an average value of zero, demonstrating that, on average, the structure does not show octahedral tilting along [110] and [1−
Figure 4. (a) Perovskite structures with negative (0) values of the I−Pb−Pb−I dihedral angle. (b) Distribution of the values assumed by all of the X−Pb−Pb−X (X = Cl, I) dihedral angles, within the reference cell, by the MAPbI, MAPbICl-ap, and MAPbICl-eq models, during the 12 ps CPMD simulations.
10], as found experimentally. It is also worth mentioning that the X−Pb−Pb−X dihedral angles assume values comprised within a 30° range, thus demonstrating that the material structure can instantaneously be very different from the average geometry, probed in X-ray diffraction experiments.54 Along the [001] direction, two symmetric distributions with averages at ±27° are found, which correspond to the aforementioned outof-phase rotation of the PbX6 octahedral. Thus, the theoretical simulations reproduce the a0a0c− structure. Note that the same a0a0c− structure is found for MAPbI and MAPbICl-ap. In the case of MAPbICl-eq instead, a clear difference is found along the [001] direction, in which the two maxima of the distributions fall around ±24° and the X−Pb−Pb−X dihedral angle can assume a value of 0°. An in-depth analysis of the evolution of the X−Pb-Pb-X dihedral angle along the [001] direction of MAPbICl-eq shows that some dihedrals drift toward 0°, thus going to a locally “pseudocubic” structure. An example of this phenomenon is depicted in Figure SI3. This difference in the average and in the dynamics of the inorganic framework in the presence of just 1% chlorine doping in the equatorial position has significant effects on the reorientational dynamics of the MA cation. As in our previous works,45,55,56 we used the Tait−Bryan ϕ and θ angles (depicted in Figure 5a) to describe the orientation of the CN vector of each one of the 32 MA cations in the unit cell. As shown in Figure 5a, positive/negative values of the ϕ angle correspond to the nitrogen atom pointing in the same/opposite direction of E
DOI: 10.1021/acs.inorgchem.6b01681 Inorg. Chem. XXXX, XXX, XXX−XXX
Forum Article
Inorganic Chemistry
to investigate the electronic properties of our models at room temperature, thus considering the inherent vibrational motion of the material explicitly. We extracted one geometry each 0.06 ps from the CPMD simulations, and we performed a singlepoint calculation, at the same theoretical level of the CPMD simulation (SR-DFT). It is well-known that the band gap predicted at the SR-DFT level of theory is only in fortuitous agreement with the experiment for this class of materials because of an error cancelation between the electronic correlation and SOC.31,38,39 Thus, a more reliable estimation of the band-gap variation associated with structural dynamics would require at least the introduction of SOC effects in the electronic structure calculations. On the other hand, in an our previous work, we demonstrated that SR-DFT calculation predicts a band-gap variation, associated with the structural dynamics, in good agreement with the more demanding SOCDFT calculations, and thus we can safely rely on the first method.49 The evolution of the electronic band gap during the CPMD simulations for the three investigated systems is reported in Figure 6. For the pristine MAPbI, the electronic Figure 5. (a) Tait−Bryan θ and ϕ angles for the description of the orientation of the MA cations. (b) Distributions the values assumed by the ϕ angles of all of the MA cations during the CPMD simulations, for MAPbI, MAPbICl-ap, and MAPbICl-eq.
the c axis, respectively, while a null value of the ϕ angle corresponds to the MA molecule lying in the ab plane. In Figure 5, we have reported the distribution of values assumed by the ϕ angle during the 12 ps CPMD simulations, for the three systems of interest. For MAPbI, the molecules are slightly tilted, by ∼30° with respect to the ab plane, with the nitrogen pointing in the opposite direction of the c axis, consistent with our previous investigations.45,49 Such a spatial orientation allows for an optimized hydrogen bond between the inorganic framework and the organic cations.49 For MAPbIClap, the distribution of the ϕ angle is practically the same as that of the pristine MAPbI. In contrast, for MAPbICl-eq, the distribution of the ϕ angle differs significantly with respect to the other systems, with organic cations that frequently explore orientations with ϕ values
