Partially replacing Pb2+ by Mn2+ in hybrid metal halide perovskites

Tailoring the physical properties of hybrid lead metal halide APbX 3 perovskites by means of compositional engineering is one of the key factors contributing to the development of highly efﬁcient and stable perovskite solar cells. While the beneﬁcial effects of partial ionic replacement at the A- and X-sites are largely demonstrated, partial replacement of Pb 2+ is less explored. Here, we developed a solution-based procedure to prepare thin ﬁlms of mixed-metal MAPb 1- a Mn a I 3 perovskites. Although Mn 2+ ions have a size that can potentially ﬁt in the B-sites of MAPbI 3 , using a combination of structural and chemical analysis, we show that only less than 10% of Pb 2+ can be replaced by Mn 2+ . A 3% replacement of Pb 2+ by Mn 2+ leads to an elongation of the charge carrier lifetimes as concluded from time-resolved PL measurements. However, by analysis of the time-resolved microwave conductance data, we show that the charge carrier mobilities are largely unbalanced, which is in accordance with density functional theory (DFT) calculations indicating that the effective mass of the hole is much higher than that of the electron. Increasing the concentration of Mn 2+ in the precursor solution above 10% results in formation of amorphous

Tailoring the physical properties of hybrid lead metal halide APbX 3 perovskites by means of compositional engineering is one of the key factors contributing to the development of highly efficient and stable perovskite solar cells. While the beneficial effects of partial ionic replacement at the A-and X-sites are largely demonstrated, partial replacement of Pb 2+ is less explored. Here, we developed a solution-based procedure to prepare thin films of mixed-metal MAPb 1-a Mn a I 3 perovskites. Although Mn 2+ ions have a size that can potentially fit in the B-sites of MAPbI 3 , using a combination of structural and chemical analysis, we show that only less than 10% of Pb 2+ can be replaced by Mn 2+ . A 3% replacement of Pb 2+ by Mn 2+ leads to an elongation of the charge carrier lifetimes as concluded from time-resolved PL measurements. However, by analysis of the time-resolved microwave conductance data, we show that the charge carrier mobilities are largely unbalanced, which is in accordance with density functional theory (DFT) calculations indicating that the effective mass of the hole is much higher than that of the electron. Increasing the concentration of Mn 2+ in the precursor solution above 10% results in formation of amorphous Mn-rich domains in the film, while the perovskite lattice becomes depleted of Mn 2+ . These domains negatively affect the charge carrier mobilities and shorten the lifetime of photogenerated carriers. The resulting reduction in charge carrier diffusion lengths will severely limit the photovoltaic properties of solar cells prepared from these mixed metal halide perovskites. © 2018 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/1.5060953 The extraordinarily rapid rise in the power conversion efficiency of metal halide perovskite solar cells (PSCs) 1,2 has in the last years been accompanied by improvements in their stability. 3 While the first breakthroughs in the field of PSCs were due to changes in the device architecture 4,5 and processing conditions, 6 compositional engineering of the perovskite ABX 3 crystal structure has recently played a crucial role in the achievement of highly stable and efficient PSCs. 7 It has been shown that the properties of hybrid halide perovskites can be tailored and optimized by (partially) replacing the ions at the A-sites and X-sites of the perovskite lattice, 7-9 and the most efficient PSCs are currently realized using mixed-cation, mixed-halide perovskites. 10 By contrast, blending the divalent metal cations at the B-sites to obtain mixed-metal perovskites is a relatively less explored path. Most of the efforts in this direction involved the partial replacement of Pb 2+ with Sn 2+ and were initially motivated by the necessity of mitigating the toxicity of lead. [11][12][13][14] MAPb 1-a Sn a X 3 perovskites (where a Authors to whom correspondence should be addressed: d.bartesaghi@tudelft.nl and t.j.savenije@tudelft.nl 2166-532X/2018/6(12)/121106/11 6, 121106-1 © Author(s) 2018 MA = CH 3 NH 3 and X = I, Br, or Cl) have been synthesized for the full compositional range 0 ≤ a ≤ 1 11 and used in photovoltaic devices with efficiencies as high as 17.6%, 15 although the ease with which Sn 2+ oxidizes to Sn 4+ severely limits the processing of this class of perovskites. 16,17 Recently, Klug and co-workers demonstrated that partial replacement of lead is possible using transition or alkaline earth metals. 18 In those cases, the photovoltaic performance significantly dropped if more than a few atomic percent of lead was replaced. Very recently, positive effects on the stability and performance were reported on replacing 3% of the Pb 2+ by Zn 2+ , Mn 2+ , or Ni 2+ . 19 In this paper, we investigate how partial replacement of Pb 2+ with Mn 2+ affects the optoelectronic properties of MAPbI 3 . Manganese is one of the most abundant metals in the Earth's crust and has a low toxicity. 20 Besides the possibility of using the mixed metal perovskite MAPb 1-a Mn a I 3 in photovoltaic devices, 18,19,21 its magnetic properties can be used in low-power, light-controlling applications such as magneto-optical data storage devices. 22 Recently, partial replacement of Pb 2+ with Mn 2+ has been demonstrated for lead halide perovskite nanocrystals and nanoplatelets. [23][24][25] We prepared thin films of MAPb 1-a Mn a I 3 from solution, and we characterized the chemical composition, crystal structure, and optoelectronic properties. Although the elemental analysis confirms the presence of Mn 2+ and the material retains the perovskite crystal structure even up to 50% replacement, we demonstrate that only less than 10% of Pb 2+ can be replaced by Mn 2+ at the B-sites. The excess manganese is located in amorphous, Mn-rich domains.
We characterized the optoelectronic properties of the mixed metal perovskite thin films by means of UV-vis absorption spectroscopy, time-resolved photoluminescence (TRPL), and time-resolved microwave conductance (TRMC). If less than 10% of Pb 2+ is replaced by Mn 2+ , charge carrier lifetimes increase; however, the charge transport becomes largely unbalanced: The mobility of holes is more than one order of magnitude lower than the mobility of electrons, which is in accordance with density functional theory (DFT) calculations revealing the effective masses. Segregation of Mn 2+ in domains outside the perovskite crystals restores balanced transport; however, the mobilities are significantly lower and the recombination rate is faster than in MAPbI 3 .
Details regarding the experiments are summarised below. Equimolar amounts of hydroiodic acid (HI, 57 vol% in water) and methylamine (CH 3 NH 2 , 33 wt. % in ethanol) were added in a roundbottom flask and let to react for 2 h in an ice bath while stirring. The precipitate of the solution was collected by rotary evaporation at 50 • C, then washed with diethyl ether, and dried at 60 • C in a vacuum oven for at least 24 h before use.
MAI, PbAc 2 , and MnCl 2 were dissolved in dimethylformamide (DMF) inside of a nitrogen-filled glovebox. Various solutions with different PbAc 2 :MnCl 2 molar ratios have been prepared, keeping the molar ratio of MAI to the total concentration of inorganic salt constant (3:1). The solutions were heated on a hot plate at 70 • C and magnetically stirred for 1 h. Subsequently, the solutions were allowed to cool down to room temperature and filtered (polypropylene filters, 0.45 µm) before use.
Samples for XRD, UV-vis absorption, TRPL, and TRMC measurements have been deposited on quartz substrates; samples for SEM and EDX have been deposited on FTO-coated glass substrates. The quartz substrates were cleaned by O 2 plasma; for FTO-coated glass substrates, a thorough cleaning by using an ultrasonic bath in acetone and isopropanol has been performed prior to O 2 plasma cleaning. The cleaned samples have been transferred in the glovebox, and perovskite films have been deposited on them by spin coating (2000 RPM, 2000 RPM/s, 45 s). The deposited films were annealed at 100 • C (15 min for MAPbI 3 films, 2 h for films containing Mn 2+ ) with a thickness of about 200 nm. XRD measurements were performed using a Bruker D8 Advance diffractometer in the Bragg-Brentano geometry. The monochromatic Co Kα X-ray source was operated at 40 kV and 40 mA.
The fraction of absorbed light was measured with a Perkin-Elmer UV/VIS/NIR Lambda 1050S spectrometer containing a 150 mm Spectralon integrating sphere. The perovskite thin films were measured in the wavelength range 400-850 nm.
TRPL measurements were carried out using an Edinburgh LifeSpec spectrometer equipped with a single-photon counter. The perovskite films were excited at 405 nm with a picosecond pulsed diode laser. A high pass filter was used to prevent photons with wavelength below 450 nm to reach the detector. PL emission spectra of all samples were measured in the range 650-850 nm. PL lifetimes were recorded at the maximum emission wavelength.
Samples were sealed in a microwave cavity within a nitrogen-filled glovebox. The fraction of absorbed microwave (frequency ca. 8.5 GHz) power (∆P/P) was measured after pulsed optical excitation (repetition rate 10 Hz, λ = 550 nm). The rise of ∆P/P is limited by the width of the laser pulses (3.5 ns) and by the response time of the microwave cavity (18 ns). A series of neutral density filters was used to perform light-intensity dependent TRMC measurements.
Top-view SEM secondary electron images were taken with a JEOL-JSM-610LA microscope using 10 kV voltage energy, 10 mm working distance, and ×3000 magnification. Backscattered electron images were taken using 20 kV voltage, 10 mm working distance, and ×200 magnification.
EDX elemental maps and point analysis were carried out using the EDX detector of the JEOL-JSM-610LA microscope, with settings of 20 kV voltage, 10 mm working distance, and ×200 magnification.
The structural optimization of MAPbI 3 , MAPb 0.9375 Mn 0.0625 I 3 , and MAPb 0.875 Mn 0.125 I 3 using 12 (unit cell), 192, and 96 atoms (super cells) was performed using DFT within the Perdew, Burke, and Ernzerhof (PBE) 26 functional implemented in the Vienna ab initio simulation package (VASP). 27,28 The outermost s, p, and d (in the case of Pb and Mn) electrons are treated as valence electrons whose interactions with the remaining ions are modeled by pseudopotentials generated within the projectoraugmented wave (PAW) method. 29,30 An energy cutoff of 500 eV and 6 × 6 × 6, 2 × 3 × 3, and 3 × 3 × 3 k-point meshes were used to achieve energy and force convergence of 0.1 meV and 2 meV/Å for 12,192, and 96 atoms cells, respectively. The spin-orbit coupling effect was taken into account for the band structure calculations.
Recently, the lead-free compound methylammonium manganese iodide (MAMnI 3 ) has been synthesized and characterized; such a compound is amorphous and does not form a cubic perovskite structure. 31 We started our study by investigating how much Pb 2+ can be replaced by Mn 2+ before disrupting the perovskite crystal structure. The most commonly used predictor for the perovskite phase is the Goldschmidt tolerance factor, 32 where r A , r B , and r X are the ionic radii of monovalent cation (A), divalent cation (B), and halogen anion (X), respectively [ Fig. 1(a)]. It has been empirically observed that the perovskite structure can be formed if the tolerance factor is in the range 0.875 < t < 1.06. 33 Besides the tolerance factor, the octahedral factor is used to assess if the divalent cation B has the right size to fit in the X 6 octahedron, 34 If µ < 0.41, the BX 6 octahedra cannot form, which excludes the formation of the perovskite structure. 34 By combining the perovskite formation criteria for t and µ, one can define a perovskite formation region on the t-µ structure map [ Fig. 1(b)]. We calculated t and µ for MAPb 1-a Mn a I 3 with 0 < a < 1; for the mixed-metal compounds, we approximated the radius of the divalent cation as the weighted average of the cationic radius of Pb 2+ (r Pb ) and of Mn 2+ (r Mn ), For each composition considered, a point is placed on the t-µ structure map in Fig. 1(b). The replacement of Pb 2+ with the smaller Mn 2+ cations results in a reduction in µ, which drops below the empirical limit for a > 0.5. The above analysis suggests that it is thus possible to replace half of the Pb 2+ with Mn 2+ without disrupting the perovskite structure. To verify this prediction, we prepared mixed-metal compounds using a one-step solution process: methylammonium iodide (MAI), lead acetate (PbAc 2 ), and manganese chloride (MnCl 2 ) were dissolved in DMF and subsequently spin cast to form thin films. The molar ratio of MAI to the total amount of inorganic salt content was kept constant at 3:1. We define the amount of Mn 2+ in the precursor solution (a sol ) as the percentage of MnCl 2 with respect to the total amount of bivalent metal salts, where n i is the number of moles of the salt i in the precursor solution. For the remaining part of this paper, a film with a sol = 10% denotes a film that is prepared from a precursor solution with a fraction Mn 2+ of 0.1. After spin casting the films, the elemental composition was measured by means of energydispersive X-ray spectroscopy (EDX). Quantitative analysis of the data provides the atomic fraction of each element in the sample (Table SI in the supplementary material); the percentage of Mn 2+ with respect to the total amount of bivalent metals is given by where f i is the atomic fraction of the element i in the film. Except for the film with a sol = 50%, all percentages found for a film are somewhat smaller than those for a sol (see Table I). Moreover, from the backscattered electron images and elemental mapping of Mn (Fig. S1), it is evident that for films with a sol > 30%, the films are not uniform and that the lateral distributions of Pb and Mn are not homogenous.
Although elemental analysis confirms the presence of Mn 2 in the films (with roughly the same fraction as in the precursor solutions), this does not imply whether Mn 2+ ions occupy the B-sites of the perovskite lattice. The X-ray diffraction (XRD) patterns for films with different concentrations of Mn 2+ are shown in Fig. 2(a). The diffraction pattern of the Mn-free film presents the characteristic peaks of MAPbI 3 . 35 We note here that the (110) peaks of our samples are located at 2ϑ ≈ 16.35 • , in contrast to the more frequently reported value of 2ϑ ≈ 14.20 • . The discrepancy is due to the Co Kα X-ray source used in our experiment, which emits X-ray at longer wavelengths than the more commonly used Cu Kα source. Interestingly, independent of the values of a sol , the perovskite structure is preserved. Replacement of Pb 2+ ions with the smaller Mn 2+ ions is expected to cause a reduction of the perovskite lattice, resulting in a shift of the XRD reflections toward larger angles. For samples  with a low content of Mn, the (110) peak indeed shifts toward larger 2ϑ [Figs. 2(b) and 2(c)], which is in agreement with the recent literature. 19 Remarkably, the shift is not monotonously increasing with higher Mn concentrations. Instead, for films with a sol > 3%, the (110) peak shifts back toward the position of the (110) peak of MAPbI 3 . We explain this as follows: while limited amounts of MnCl 2 in the precursor solution results in the replacement of Pb 2+ with Mn 2+ at the B-sites of the lattice, larger concentrations of MnCl 2 yield the formation of Mn-rich domains, depleting the perovskite lattice with Mn 2+ . Hence, the XRD pattern reverts back to that of MAPbI 3 . We can also rule out the formation of Mn 2+ interstitials as this should result in an expansion of the perovskite lattice and consequently a shift of the XRD peaks toward lower angles, as recently observed by Bai and coworkers for Mn-alloying of CsPbI 2 Br. 36 However, we do not observe any expansion of the MAPbI 3 lattice even at the highest Mn 2+ concentrations (Table SII of the supplementary material). When a sol exceeds 30%, new features appear in the XRD patterns (Fig. S2). These findings agree with the SEM images of these films (Fig. S1), showing the appearance of new, non-amorphous phases (particularly evident for a sol = 40% and 50%). Therefore, we decided to not further investigate films with a sol > 30%.
For films with a sol = 30%, presence of Cl is detected by means of EDX (Table SI). Similarly as discussed above for Mn 2+ , we would expect a shift of the XRD peaks toward larger angles if Cl − replaced the larger I − ions in the perovskite structure; the absence of such a shift indicates that Cl − is also segregated in Mn-rich domains. have a large bandgap. Careful analysis of the PL emission spectra presents a small red-shift of the PL peak position [ Fig. 3(b)] with respect to the PL spectrum of MAPbI 3 . The peak shift reaches a maximum for films with a sol = 3% and decreases with higher concentrations of Mn 2+ , at which the PL peak position is almost identical to that of MAPbI 3 [ Fig. 3(c)]. The dependency is similar to what we observed in Figs. 2(b) and 2(c) for the XRD data, which strengthens the idea that for films with a sol > 3% the Mn 2+ is located in Mn-rich domains and not in the perovskite lattice.
Apart from the shift in the PL peak position, the TRPL lifetime [ Fig. 3(d)] is influenced by the presence of Mn 2+ . We performed mono-exponential fits on the TRPL traces (Fig. S3) and collected the TRPL lifetimes in Table SIII. The TRPL lifetime for the MAPbI 3 sample is 70 ns. For the film with a sol = 3%, the TRPL lifetime is a factor 2 larger (140 ns). This enhancement is in agreement with that measured in Ref. 19 by means of transient absorption experiments. On the contrary, the formation of Mn-rich domains causes a reduction in the TRPL lifetime. This occurs to some extent in the film with a sol = 10% and becomes particularly evident for even higher Mn 2+ concentrations. Furthermore, the PL emission of the film with a sol = 30% is significantly quenched in comparison with the other samples (Fig. S4). These results indicate that for large Mn 2+ concentrations non-radiative recombination pathways dominate the charge dynamics, which will be discussed in more detail below.
We further investigated the charge dynamics of light induced excess carriers by TRMC. 37 In short, photoexcitation of the film by means of a nano-second laser pulse at 550 nm results in the generation of free charges, which leads to an increase in the conductance of the sample. The photo-conductance (∆G) is probed by measuring the reduction in reflected microwave power in time.
After an initial rise corresponding to photogeneration of free charges, ∆G decays as a result of charge recombination and/or immobilization of charges in deep trap states. The TRMC traces have been corrected for the fraction of absorbed photons, Fa, which eliminates the effect of optical inhomogeneities.
Intensity-normalized TRMC traces are displayed in Figs. 4(a)-4(d) for films with different concentrations of Mn. For the film with a sol = 3%, the TRMC lifetimes are slightly longer [compare Fig. 4(a) with 4(b)], in line with the TRPL observations. However, for films with a sol > 3%, the charge carrier lifetimes drop significantly.
Hutter et al. modelled TRMC traces of MAPbI 3 films using a kinetic model representing a ptype semiconductor including band-to-band recombination and recombination via deep trap states [ Fig. 4(e)]. 38 The trapping process in this model is assumed to be a second-order process. More details about the method and the set of coupled differential equations which describe the time evolution of the populations of free and trapped charges are given in the supplementary material. We used the same model to reproduce the TRMC data presented in Figs. 4(a)-4(d). Note that this fully mathematical model also holds for the opposite situation, i.e., an n-type perovskite with trap states for holes and additional dark CB electrons.  Band-to-band recombination rate 6.9 6.4 7.2 15.0 k T (× 10 10 cm 3 s 1 ) Trap filling rate 1.5 3.6 6.0 7.0 k R (× 10 11 cm 3 s 1 ) Trap recombination rate 1. . We note that for MAPbI 3 without Mn 2+ we obtain similar kinetic values as reported earlier. 38 The film with a sol = 3% is characterized by a slightly smaller k 2 as compared to MAPbI 3 . This is well in line with the longer TRPL lifetime found for this sample. In addition, the presence of Mn 2+ in the perovskite lattice has a large effect on the mobilities of both charges, which become greatly unbalanced. We note here that the TRMC does not allow distinguishing between electrons and holes since mobile carriers interact with the microwave radiation independently on the sign of their charge. Nevertheless, the kinetic model reported in the supplementary material makes a distinction between the majority and the minority charges. In order to reproduce the TRMC data for the sample with a sol = 3%, we have to set the mobility of majority charges one order of magnitude larger than the mobility of minority charges.
The formation of Mn-rich domains for films with a sol > 3% leads to faster decay kinetics, which translates into an increase in k 2 . In addition, the TRMC data clearly show that alloying Mn 2+ in MAPbI 3 negatively affects the mobilities. To verify how the presence of Mn 2+ changes the charge dynamics for electron and holes, we carried out DFT calculations including spin-orbit coupling to reveal the band structure of MAPbI 3 , MAPb 0.9375 Mn 0.0625 I 3 , and MAPb 0.875 Mn 0.125 I 3 (Fig. 5). The choice of these compositions is motivated by the computational constrains of the DFT simulations. We have calculated the effective masses of electrons and holes, listed in Table III by   of s and d states of Mn with s state of Pb and p state of I). This means that the effective mass of the holes is much larger than the effective mass of the electrons (Table III), and hence the electrons are more mobile than holes. Another interesting point to note is that the effective mass of holes of the MAPb 0.9375 Mn 0.0625 I 3 is significantly larger than that of MAPbI 3 , while the effective mass of the electrons is similar in the two samples. Given the effective masses calculated by DFT, we argue that the electrons in the mixed metal perovskite have a substantial larger mobility than the holes. Combining TRMC and DFT allows us to conclude that, in the sample with a sol = 3%, the electrons are the majority carriers and that therefore the material is most likely n-type. For samples with a larger concentration of Mn 2+ , we keep the assumption of the n-type material.
The effective masses of electrons and holes in MAPb 0.875 Mn 0.125 I 3 are larger than those of MAPbI 3 . This is in agreement with the incorporation of a transition metal, such as Ag, Cu, or Au in MA 2 BB I 6 (B and B are metal cations). The corresponding effective masses are generally larger than those of lead halide perovskites. 41 However, in view of the XRD data previously shown and the conclusions about the segregation of Mn 2+ out of the perovskite structure for films with a sol > 3%, we do not consider the calculated DFT effective masses of MAPb 0.875 Mn 0.125 I 3 to be related to the lower TRMC mobilities observed in Figs. 4(c) and 4(d). Instead, at higher Mn 2+ concentrations, it seems likely that the reduction in µ is, at least partially, caused by a reduction in the size of crystalline domains in the film and the formation of more grain boundaries that would slow down the motion of free charges. Secondary electron SEM images (Fig. S5) show that the perovskite crystalline domains are smaller for the film with a sol = 30% than those for the film with a sol = 10%. Thus, the change in domain size is likely to lead to a reduction in the charge mobilities when the Mn 2+ concentration is increased from a sol = 10% to a sol = 30%. In addition, the amorphous Mn 2+ rich phases formed in films with a sol > 10% are expected to enhance to charge decay as observed in the TRPL and TRMC measurements.
In this work, we partially replaced Pb 2+ with Mn 2+ in MAPbI 3 and studied the change in structural and electronic properties of the resulting films. From structural and elemental analysis, we conclude that only a limited amount of the Pb 2+ atoms at the B position of the perovskite structure can be replaced by Mn 2+ . Using precursor solutions with higher Mn 2+ concentrations leads to the formation of additional amorphous Mn-rich domains. At 3% replacement, the band-to-band recombination rate slightly reduces yielding somewhat longer charge carrier lifetimes as determined from TRPL and TRMC measurements. However, this is accompanied by a substantial increase in the effective mass of the holes, resulting in a strongly unbalanced charge transport. According to the recent results of Zheng et al., the increase in charge carrier lifetimes still positively affects the photovoltaic performance of devices made from this composition. 19 For higher concentrations of Mn 2+ , formation of Mn-rich domains restores the balanced transport, although both electron mobility and hole mobility become ca. 1 order of magnitude smaller than in MAPbI 3 due to enhanced charge scattering. Besides the lowering of the mobilities, a reduction in the charge lifetime is found. At higher concentrations of Mn 2+ , the reduced charge carrier diffusion lengths will be severely limiting the photovoltaic properties of solar cells prepared from these mixed metal halide perovskites. These findings are in line with the lower efficiencies observed in MAPbI 3 solar cells in which more than 5% of the Pb 2+ was replaced by Mn 2+ . 18,19 Combination of experimental and theoretical data indicates that alloying MAPbI 3 with small amounts of Mn 2+ is a unique handle to manipulate the dynamics of the carriers.
See supplementary material for the differential equations used for modeling the TRMC data and for additional data.