Exploiting the Lability of Metal Halide Perovskites for Doping Semiconductor Nanocomposites

Cesium lead halides have intrinsically unstable crystal lattices and easily transform within perovskite and nonperovskite structures. In this work, we explore the conversion of the perovskite CsPbBr3 into Cs4PbBr6 in the presence of PbS at 450 °C to produce doped nanocrystal-based composites with embedded Cs4PbBr6 nanoprecipitates. We show that PbBr2 is extracted from CsPbBr3 and diffuses into the PbS lattice with a consequent increase in the concentration of free charge carriers. This new doping strategy enables the adjustment of the density of charge carriers between 1019 and 1020 cm–3, and it may serve as a general strategy for doping other nanocrystal-based semiconductors.

R esearch on metal halide perovskites is advancing rapidly, owing to the compelling electronic, optical and structural properties of these ionic semiconductors, such as long diffusion lengths and carrier lifetimes, 1−3 low exciton binding energies, 4 low number of trap states despite the high concentration of vacancies, 1,5 composition-tunable bandgap, 6,7 and ease of processability. 8−11 Because of the relatively labile crystal structure of metal halide perovskites, the understanding and control of the chemical and structural transformations that these compounds readily undergo are some of the most pressing questions. 12 Cesium lead halides can adopt perovskite and nonperovskite structures with different dimensionalities. Perovskites are composed of Cs + cations stabilizing [PbX 6 ] 4 − octahedra in a cubic or, upon slight distortion, tetragonal or orthorhombic phases, where all the corners of the [PbX 6 ] 4 − octahedra are shared. 13 All these compounds adopt regular CsPbX 3 stoichiometry (3D). Nonperovskite structures include polymorphs with various stoichiometries, but all of them lose the corner-sharing motif in the lattice. The first example of these structures is a polymorph that despite having the same CsPbX 3 stoichiometry crystallizes in an orthorhombic phase (δ-phase) with chains of edge-sharing octahedra (1D). 14 Another example is the lead-depleted Cs 4 PbX 6 structure where the [PbX 6 ] 4 − octahedra are isolated (0D). 13,15−17 A third closely related nonperovskite structure that has lower Cs + content and does not contain [PbX 6 ] 4 − octahedra, CsPb 2 X 5 , can be described as layers (2D) of [Pb 2 X 5 ] ̅ clusters separated by Cs + ions. 13,15−17 Because of the large difference in the involved atoms' electronegativities, all these cesium lead halides exhibit mixed bonding nature. The lead halide framework is dominated by covalent bonds and balanced by ionically bound Cs + cations. The lack of covalency between Cs + and the anionic units enables crystal lability. 18 As a direct consequence, the ions move easily and allow transformation between the different structuresprovided that the stoichiometry is compensated.
Very common are the transformations between CsPbBr 3 and Cs 4 PbBr 6 structures especially in nanocrystals (NCs). The transformation of Cs 4 PbBr 6 into CsPbBr 3 is achieved by adding PbBr 2 to the structure, 19 extracting CsBr by chemical complexation or intercalation, 20 or diffusion and dissolution of CsBr in water. 21 The chemical transformation from CsPbBr 3 to Cs 4 PbBr 6 can be induced by removing PbBr 2 from the crystal lattice by its complexation with thiols or amines. 22−24 In all these reactions purification of the NCs is required to separate the undesired byproducts: either CsBr or PbBr 2 .
PbS NC-based solids find application in various fields including transistors, 25−28 solar cells, 29−32 photodetectors, 31,33,34 and thermoelectrics. 35−37 In most of these applications adjusting the number of free charge carriers to control charge transport is crucial. The conventional n-type dopants for bulk lead chalcogenides are halides. Incorporation of halide ions (X= Cl − , Br − , I − ) in the chalcogenide (Y= S 2− , Se 2− , Te 2− ) sublattice results in the addition of one electron in the conduction band per halide to compensate for the different valency of halides and chalcogenides. 38 Strategies to dope bulk semiconductors require control over the composition at the impurity level; the straightforward translation into doping bottom-up assembled NC solids is to use doped NCs. However, impurity doping of NCs is energetically 39 and kinetically unfavorable, because the diffusion path of impurity atoms to the surface is short. 40 Although tuning NCs composition has been attempted, 26,41 the introduction of a controlled amount of impurities in small structures is problematic for the preparation of heavily doped semiconductors. 42−44 Alternative approaches to dope NC-based solids have focused on changing the NC surface chemistry, 30,36,38,45−48 inducing partial cation exchange, 49,50 and blending with other NCs. 35,51,52 PbY−perovskite nanocomposites have been used in photodetectors, 53,54 LEDs, 55,56 and photovoltaics. 57,58 Remarkably, PbS NC inclusions can provide stability to the cubic phase of perovskites suppressing the cubic−orthorhombic phase transition, 59 and perovskite passivation layers enhance charge carrier separation in PbS NCs by providing adequate energy alignment. 60 Furthermore, PbY nanocomposites with secondary phases might be of great importance for thermoelectrics. 35,61 Herein we propose a strategy to produce doped PbS nanocomposites that makes use of the byproduct of the transformation of CsPbBr 3 into Cs 4 PbBr 6 and PbBr 2 , upon heating a mixture of CsPbBr 3 and PbS NCs (Scheme 1). Our approach simultaneously introduces dopant ions (Br ̅ ) and a secondary phase (Cs 4 PbBr 6 ) in PbS nanocomposites ( Figure  1).
We prepared nanocomposites by mixing PbS NCs with a controlled amount of CsPbBr 3 NCs in toluene. This NCs blend was dried under vacuum and annealed with forming gas (5% H 2 in N 2, 1 bar) to yield a powder that was then pressed into pellets with relative densities of ∼92% (Table S1) using spark plasma sintering ( Figure 1).
To follow the chemical transformation of the perovskites, we performed in situ temperature-dependent X-ray diffraction (XRD) measurements of a mixture of PbS NCs and 30% wt. CsPbBr 3 NCs (Figures 2 and S1). This mixture was heated to 450°C and kept at that temperature for 60 min, mimicking the annealing step. Upon heating to 150°C, CsPbBr 3 NCs show strong sharpening of the reflections explained by accelerated ion migration between particles causing grain growth. 62−64 After reaching 450°C, CsPbBr 3 reflections progressively lose intensity, and peaks corresponding to Cs 4 PbBr 6 become visible, shifted to lower angles because of thermal expansion of the lattice. This experiment demonstrates that CsPbBr 3 converts to Cs 4 PbBr 6 , necessarily releasing PbBr 2 . This reaction could be enhanced by the binding affinity of PbBr 2 to PbS surface dangling bonds. 65,66 Control experiments with CsPbBr 3 and Cs 4 PbBr 6 pure phases showed that both phases are recovered after heating to 500°C and cooling to room temperature and do not undergo any transformation on their own ( Figures S9 and S10). Besides, annealing induces crystal grain growth in pure PbS NCs, evidenced by the narrowing of the XRD reflections and scanning electron microscopy (SEM) images ( Figures S2 and S5). The addition of CsPbBr 3 enhances the growth of the PbS crystal domains leading to even narrower reflections (Figures S3 and S4 and Tables S3  and S4). This enhancement can be explained by the presence of PbBr 2 , a phase with high solubility in PbS 67 and a large difference in melting point, which lowers the activation energy for diffusion. 68−70 Because no reflections associated with PbBr 2 are visible, we considered the possibility of other phases in the PbS−PbBr 2 phase diagram being formed. Still, Pb 7 S 2 Br 10 , the only stable phase, was not observed in the diffraction patterns. 67 Recently, the metastable phase Pb 4 S 3 Br 2 was reported in NCs; this phase is not present in the XRD patterns, and we disregarded it because of the high temperature and long reaction time, which are not compatible with such a kinetically stabilized phase. 71 Previous studies showed pure CsPbBr 3 NCs undergo partial transformation to the lead-rich CsPb 2 Br 5 ; this phase is not present in our experiments. 64 These observations suggest that bromide substitutes sulfide in PbS, doping the matrix as expressed in the chemical reactions in Scheme 1.
To test the doping efficiency, we prepared pellets with the proper nominal CsPbBr 3 concentrations to achieve charge carrier concentrations between 10 19 and 10 20 cm −3 , which is close to the optimal charge carrier concentration for thermoelectrics. 72,73 The molar amount of Br introduced is referred to the amount of S, X% = 0.25, 0.50, 1, 2, 3 (see the Supporting Information, page S1).
Panels a and b of Figure 3 show the electrical conductivity and Seebeck coefficient of the samples between room temperature and 900 K, respectively. At low temperatures, the absolute value of the Seebeck coefficient decreases and the electrical conductivity increases with the starting concentration of CsPbBr 3 , proving the doping effect of the transformation (Supporting Information, page S8). 72 Besides, the negative sign of the Seebeck coefficient confirms the n-type character of the obtained nanocomposites. These results are consistent with the charge carrier concentrations measured by Hall effect (Table S2), which correspond exactly to one-half of the total Br concentration, as expected from the reaction in Scheme 1. The doping effect of CsPbBr 3 is independent of the use of NCs; ground perovskite crystals react with PbS in the same way. However, NCs provide a much shorter ion diffusion path and better mixing with PbS, which lead to faster kinetics (Supporting Information, page S10).
The observed tendencies in the electrical conductivity of nanostructured materials are a trade-off between many effects. On the one hand, the thermal excitation of carriers through energy barriers (grain boundaries) leads to an increase of the conductivity with temperature. This effect dominates at low temperatures in samples with smaller grains. On the other hand, an increase in temperature enhances charge carrier scattering. This is the typical behavior of heavily doped semiconductors and metals. Additionally, as the temperature increases, thermal excitation of minority carriers (bipolar effect) occurs. Consequently, the absolute value of the Seebeck coefficient starts decreasing. The higher the doping level, the higher the temperature at which this effect is perceptible. 61 Figure 2. X-ray diffraction patterns of a mixture of PbS NCs and CsPbBr 3 NCs at different temperatures together with the reference patterns for the different crystal phases. In the 450°C patterns, t indicates the time held at 450°C before measurement. The patterns are overlapped with a 2D intensity plot to help to identify low intensity reflections (PbS PDF 00-002-0699, CsPbBr 3 PDF 00-054-0753, Cs 4 PbBr 6 PDF 01-075-0412). The XRD patterns of the produced pellets show neither CsPbBr 3 nor Cs 4 PbBr 6 reflections ( Figure S6), which we associate with the low content of perovskite required for the doping level targeted. To demonstrate that the pelletized material contains Cs 4 PbBr 6 , we evaluated the microscopic structure of the bulk nanocomposites ( Figure 4) by highresolution transmission electron microscopy (HRTEM). A 0.5% pellet was thinned to electron transparency by Ar + polishing to produce a self-suspended lamella. Figure 4d shows an HRTEM micrograph obtained from a large PbS grain (defined as the matrix) containing multiple precipitates. The power spectrum of the highlighted area shows diffraction spots of the matrix and the precipitates, along with satellite spots of the Moiréfringes, which are formed because of orientation differences between the matrix and precipitates. Inverse Fourier (frequency) filtering of the power spectrum allowed mapping the matrix and the nanodomains. The matrix phase (in red) is identified as FCC PbS (space group Fm3̅ m) with a lattice parameter a = 0.5936 nm visualized along its [011] axis. The nanodomain (in green) corresponds to the same PbS phase visualized along the [001] zone axis. We performed the same analysis on the precipitate marked in Figure 4e. The diffraction spots observed in the power spectrum correspond to the matrix phase (in red) visualized along its [011] axis. The spots of the precipitate correspond to the hexagonal Cs 4 PbBr 6 phase (space group R3̅ c:H) with lattice parameters a = 1.373 nm and c = 1.732 nm, which is visualized along the [11̅ 02] zone axis. Because of the difficulty in analyzing all the precipitates crystallographically, to evaluate their nature we performed energy dispersion X-ray line scans across many precipitates. We found that they show no difference in composition with the matrix (Figure S7), proving that the vast majority of precipitates are PbS. The scans also show a small content of Br evenly distributed through the nanocomposites as expected in its role as a dopant. In short, HRTEM revealed that the doped samples are composed of a PbS matrix with a large amount of PbS nanodomains and randomly distributed Cs 4 PbBr 6 nanoprecipitates (Figure 4), confirming the chemical transformation observed by XRD.
In summary, we showed that CsPbBr 3 and PbS react to give the lead depleted Cs 4 PbBr 6 phase. The byproduct of the transformation, PbBr 2 , dissolves in the PbS matrix, leading to doped nanocomposites and inducing grain growth while Cs 4 PbBr 6 forms nanoprecipitates. We made use of this transformation to simultaneously dope and introduce a secondary phase in n-type PbS nanocomposites. These results validate this new doping strategy for nanocomposites based on NC blending that involves the reaction of one of the components. This approach is not limited to the preparation of bulk samples at high pressures but could also be applied to the preparation of films or other nanocomposites because the reaction that leads to doping is also demonstrated at ambient pressure. Finally, our results reassess the intrinsic instability of metal halide perovskites and are therefore of high relevance for applications where these materials are exposed to high energy densities.