Open Access Article
Sean
Nations
a,
Lavrenty
Gutsev
*ab,
Bala
Ramachandran
a,
Sergey
Aldoshin
b,
Yuhua
Duan
c and
Shengnian
Wang
*a
aInstitute for Micromanufacturing, Louisiana Tech University, Ruston, LA 71272, USA. E-mail: lgutsev@outlook.com; swang@latech.edu
bInstitute of Problems of Chemical Physics of Russian Academy of Sciences, 1, Acad. Semenov Av., Chernogolovka, Moscow Region 142432, Russian Federation
cNational Energy Technology Laboratory, United States Department of Energy, Pittsburgh, Pennsylvania 15236, USA
First published on 1st April 2022
With promising solar cell applications, organic–inorganic lead halide perovskites belong to a novel and rapidly developing class of semiconducting materials. Unlike its well-investigated iodide counterparts, formamidinium lead chloride (FAPbCl3) is a much less studied perovskite that shows superior stability and surprisingly high selectivity and sensitivity towards certain gas analytes. The origin of the sensing ability of this perovskite material can be traced to its high defect tolerance and the existence of some defects which act as the “lock” to a certain gas analyte's “key”. In this work, we deepen the understanding of this proven perovskite sensor through first-principles computational study of its defect formation energies, charge transition energies, and optical properties. These studies are done with the widely used GGA exchange–correlation functional PBE and the newer meta-GGA functional SCAN, both incorporating the Grimme's D3 dispersion correction. Additionally, we also consider experimental conditions and the effect of temperature on electron screening in the simulations, which demonstrates that screening is particularly important when considering high oxidation states of the defects. We also note the importance of antisites in the Fermi Pinning as well as the importance of FA disorder, which are previously unreported insights.
One of the most influential factors on the properties of semiconductors in general are the types and concentrations of various defects present.24 For example, the majority charge carrier identity of the well-studied case of MAPbI3 is influenced by varying chemical potential of the growing environment due to the particulars of point defect physics.25 Point defects in the bulk of the material are a major driver, with the lifetime, mobility and recombination rate of charge carriers all affected by them, as well as the doping limit of the material. According to Shockley-Read-Hall theory, so-called “deep” or “trap” defects located deep in the band gap can lower the PVE by allowing increased nonradiative recombination rate of charge carriers via balancing the ease of hole and electron trapping from the respective band edges.26,27 On the other hand, “shallow” defects near the band edges with energies of the order of the thermal energy of the lattice can facilitate carrier generation and are likely to emit electrons or holes before recombination can occur. Shallow defects tend to be self-regulating in a process known as Fermi level pinning, in which a low formation energy donor–acceptor pair maintains the Fermi level by forming more of the appropriate donor or acceptor defect as the Fermi energy varies from the pinned level. This phenomenon is a likely contributor to the high defect-tolerance of OIHPs.5,23 It is necessary to have a detailed understanding of the nature of point defects and the relevance to the electronic structure of a given OIHP material.
Although much attention has been placed on their use in photovoltaics, OIHPs also find applications in fields such as organic–inorganic light-emitting diodes,28,29 photodetectors2,9,30 and lasers.8,31 The same instability in the presence of foreign species or when subjected to illumination, which is a drawback for PVE, can actually function as a strength for these alternate applications.17,19,20 In addition to these optoelectronic applications, metal-halide perovskites are also investigated experimentally as sensors for their readily tunable optical and chemical properties such as band-gap and emission intensity/wavelength.32 For example, Gu et al. demonstrated the feasibility of organometal halide perovskite nanowires arrays as image sensing systems33 and another group utilized the known sensitivity of ionic perovskite crystals to water to manufacture reusable humidity sensors which were able to out-perform commercial hygrometers.15 Zhuang et al. achieved good response times and high selectivity towards detection of NO2 and acetone using an organometal halide perovskite doped with SCN− anions as the sensing material in an electrochemical sensor.34
Recently, we published a work reporting on a FAPbCl3-based sensor which shows a highly sensitive and selective signal for ammonia while being insensitive towards most other gasses regardless of polarity, such as CO, MeOH, and H2S, which often confound ammonia sensing efforts through signal obfuscation.35 Although evidence has been provided to substantiate the proposed lock-and-key healing mechanism regarding the VCl defect, no comprehensive defect analysis was performed.7 The present work endeavours to increase our understanding of this relatively unexplored perovskite and its sensing mechanism of applications. First principles-based density functional theory (DFT) allows for direct probing of material properties which are difficult to experimentally observe, such as the formation energies of the various potential defects from the pristine crystal cell.36,37 For this reason, DFT is a standard technique for investigating the nature of these critical defects and the effects that they have on the electronic structure of the OIHPs and is a critical tool in the realization of intelligently engineered materials. In this study, we use DFT calculations to elucidate the defects in FAPbCl3 that control the Fermi pinning behavior, which self-regulates the Fermi level of the material by forming charge carriers thus mitigating deviations from the pinned Fermi level. Structural optimizations are carried out using a widely used generalized gradient approximation (GGA) functional. Subsequently, single point energy and property evaluation are done using a more recently (2015) introduced meta-GGA functional, namely the Strongly Constrained and Appropriately Normed (SCAN) functional of Sun et al.38 The latter functional is particularly useful for thermodynamic calculations since it was recently demonstrated that this functional is better than GGA or local density approximation (LDA) functionals at predicting the lattice parameters and formation enthalpies of CsPbI3 phases; it also did not show ferroelectric instability in the phonon spectra, unlike LDA which erroneously reported the material as ferroelectric.39 SCAN has also been shown to yield reliable results for semiconductor bandgaps comparable to the widely used HSE hybrid functional40 but at reduced computational cost.41–44 By analyzing the formation energies of the various electron donating and accepting defects as a function of the Fermi energy, we found antisite defects, which are not previously studied, are in fact thermodynamically quite significant.
Defect energetics can be studied by introducing the defect into a supercell and comparing the energy of that cell with the pure case. The defect formation energies (DFEs) are calculated as:
![]() | (1) |
![]() | (2) |
For eqn (1), the limits of the potentials must be established. This is done by imposing the bulk potentials of the various species in the ABX3 formulations as energetic ceilings (otherwise the bulk material would be formed).
| μFA + μPb + 3μCl = ΔHf(FAPbCl3) | (3) |
| μPb + 2μCl = ΔHf(PbCl2) | (4) |
| μFA + μCl = ΔHf(FACl) | (5) |
Finally, the transition energies, also referred to as thermodynamic ionization levels, were calculated as:
![]() | (6) |
is the transition energy of defect X with charge q to charge state q′.
| DFT Functional | ΔHf(FAPbCl3) | ΔHf(PbCl2) | ΔHf(FACl) |
|---|---|---|---|
| PBE + D3 | −6.13 | −3.44 | −2.78 |
| SCAN + D3 | −6.25 | −3.44 | −2.95 |
The defect energy plots and defect transition energies in Fig. 2 and 3 respectively highlight the differences between the results of the PBE + D3 and SCAN + D3 calculations. The vertical dashed line in Fig. 2 indicates the value of the Fermi level caused by the defect pair that controls the pinning. The two functionals result in broadly similar DFE function shapes, although the point at which the lowest energy DFE charge state for each defect transitions varies significantly, as does the magnitude of the DFE. The case for PBE + D3 with intermediate component ion potentials [purple dot in Fig. 1(a)] at 0 K is shown in Fig. S1(a) in the ESI.† As can be seen, the contributions of the commonly studied vacancies and interstitials are overshadowed by the contributions of the lower formation energy anti-site defects. This more negative DFE indicates that the formation of the defect is more energetically favorable, and thus that the defect plays a more important role in the properties of the material due to the Arrhenius-type relationship between the DFE and the defect concentration.37 The vertical black line highlights the intersection point of the defect pair which dictates the pinning behavior, here being the PbFA cationic defect and the FAPb anionic defect. The induced pinning is at 0.23 eV above the VBM, which indicates that a p-type semiconductor will be formed under stoichiometric synthesis conditions. Fig. S1(b) (ESI†) shows the case of SCAN + D3 with intermediate potentials at 0 K. The same pair of defects is found to be important to the pinning in this case, and the pinned Fermi level is still nearer to the VBM than the CBM, again indicating p-type behavior. We summarize the Fermi pinning behavior for all cases considered in Table 2. Generally, SCAN + D3 and PBE + D3 often predict the same qualitative results regarding p-type vs. n-type character, but in some cases differ significantly in the depth in the band gap of the pinned Fermi level. A notable exception is the Pb-rich case, for which the Makov-Payne (MP) correction is quite important due to the prominent role played by the highly charged PbCl and ClPb defects. Here, the large magnitude of the charges of these defects increases the contribution of the correction to the DFE viaeqn (2). This large difference in pinned Fermi level highlights the importance of seemingly minute variations in the calculated thermodynamics of the defects, demonstrating that the accuracy afforded by the better approximation of the exchange–correlation energy by SCAN + D3 is crucial to prediction of the Fermi level pinning behavior. Fig. 3 contains the intersection state plots at 0 K for both functionals with defects labelled using the Kröger–Vink notation.50 PBE + D3 resulted in a band gap of 2.163 eV versus a much lower 1.665 eV for SCAN + D3. The cationic defects are listed first followed by the anionic defects. When comparing the defect transition energies, we note that the two functionals give relatively similar results (Fig. 3). Generally, the intersections are quite similar in ordering and energy; however, PBE + D3 has fewer intersections between higher oxidation states than SCAN + D3. Notably, PBE + D3 predicts deep traps for PbFA, FAi, Pbi, and VCl while SCAN + D3 does not share this result for VCl, mainly due to the reduction in predicted lowest unoccupied molecular orbital energy level; however, PbFA and FAi are both relatively insignificant due to kinetic hindrance caused by a two-hole (−/+) transition. For PbFA, FAi, and Pbi the traps are shallower due to being shifted towards the conduction band. Another significant difference seen in the anionic defects is that PBE + D3 predicted transitions just inside of the conduction band while SCAN + D3 predicts relatively shallow traps within the gap, e.g. Cli's (1-/0) transition. Finally, ClFA and PbCl are both predicted by SCAN + D3 to have a pair of twin transition levels, whereas PBE + D3 only predicted one of the intersections occurring; the ClFA pair of transitions is quite near the center of the band gap, indicating that it may also act as a deep trap. Overall, a significant driver in the difference in trap types predicted by the two methods is the band gap difference. These results indicate that while the origin of deep trap levels may be in many cases due to the interstitial and vacancy defects normally considered, the antisite defects are sometimes potential deep traps and are also found to be significant in determining the pinned Fermi level and thus the concentrations of the deep trap defects.
| Cl-Rich | Intermediate | Pb-Rich | ||||
|---|---|---|---|---|---|---|
| 0 K | 300 K | 0 K | 300 K | 0 K | 300 K | |
| PBE + D3 FPE (eV) | 0.22 | 0.22 | 0.21 | 0.54 | 1.07 | 1.63 |
| SCAN+D3 FPE (eV) | 0.68 | 0.12 | 0.67 | 0.24 | 1.53 | 1.10 |
Comparing the Cl-rich case of Fig. S2 in ESI† and Pb-rich case of Fig. 4(a) highlights the importance of experimental synthesis conditions. Merely changing the potentials of the constituent ions results in the pinned Fermi level shifting from 0.22 eV to 1.07 eV at 0 K and thereby lower carrier concentrations due to the deep location of the Fermi level in the Pb-rich case.
In Fig. 5 we further illustrate the importance of including the Makov–Payne correction by showing the DFE of the VPb2− and VPb1− with and without the correction included. It should be noted that each of the terms within eqn (2) do not vary between PBE + D3 and SCAN + D3, so the MP correction for a given defect and charge state at a particular temperature is the same for both. While the case where the correction is omitted shows the two DFE series which never intersect within the band thus leaving the −2 series lower in energy, the inclusion of the correction correctly shows a transition within the band to the −1 oxidation state. Introducing the effect of temperature via the dielectric constant in the MP correction and comparing Fig. 4(a) with Fig. 4(b), the correction again is shown to be important, this time via its mitigation with increasing temperature due to a significant increase in the dielectric constant, which is inversely proportional to the MP correction's magnitude. The shifting of the PbCl defect's transition energy levels compared to the 0 K case in Fig. 4(a) results in n-type pinning induced by its interaction with the FAPb defect.
The material is best compared to FAPbI3. The most immediate difference when comparing FAPbCl3 to FAPbI3 is that the latter is not stable in the cubic perovskite phase at room temperature. Meanwhile, FAPbCl3 is experimentally known to be stable in the α phase.35 This can be explained by the difference in the Goldschmidt tolerance factor t:
![]() | (7) |
The tolerance factor can be thought of as a measure of strain inside of the perovskite unit cell. A proposed reason for the ease with which iodine vacancies form in FAPbI3 is that such point defects release strain,53 the resulting interstitial iodides can then have an undesirable tendency to rotate the PbI6 octahedra, which is what causes the phase instability.54 Since FAPbCl3 is not strained it does not stabilize such vacancies.
However, it's notable that the antisites we have calculated appear to be abnormally stable when compared to other perovskite materials. If, in the case of eqn (1) we replace E(bulk) with the value of the disordered material then we observe an upward shift of the DFE values of about 8 eV with nearly none of the defects retaining a negative DFE. However, if the material was to be considered totally disordered this would imply a very low concentration of vacancy defects. This would imply that it is not a good lock-and-key optical sensor, which is contrary to our previous work.35 There we observed the sensitivity originating from the localized disorder of the cubic phase, rather than the disorder of the whole material. As such, we believe that further experimental guidance is needed to fully understand the point defect concentrations. Since disorder so strongly affect the DFE's, it's likely that the real material is in a range between full order and full disorder. Regardless of this issue, it is clear that this perovskite material is very unique in its strong preference towards antisites.
Strong absorption peaks are shown in the UV-Vis optical spectrum of this FAPbCl3 perovskite material.35 To further investigate the impact of defects on the optical properties of FAPbCl3, spin-orbital coupling (SOC), which is known to be important when calculating optoelectronic properties of heavy elements,55 is included in the calculations derived in this section. This was investigated by obtaining the frequency-dependent dielectric matrix in the long wavelength limit with a sum over states approach.56 The procedure is available in the ESI,† as a more complete description is available in our previously published papers.57–60 With the real and imaginary portions of the dielectric constant calculated, it is straightforward to evaluate optical absorption, α, for a particular wavelength, λ:
![]() | (8) |
![]() | ||
| Fig. 6 Absorption spectrums of the pristine perovskite with the two pinning defects for (a) PBE + D3 and (b) SCAN + D3 with intermediate potential at 300 K. Calculated spectra are compared with experimental spectrum from Parfenov et al.35Fig. 2(d) in shape and characteristic peak gap, although being shifted to lower wavelengths. Note that the experimental data's scaling relative to the other three series is arbitrary for both PBE + D3 and SCAN + D3. | ||
Finally, to understand the role of order and disorder in the material we plotted the top 3 occupied orbitals (HOMO) and top 3 unoccupied orbitals (LUMO) (Fig. 7), we took a weighted sample of k points for these plots. Beneath these plots we also plot the density of states (DOS). The ordered phase of the material is quite ordinary for a perovskite material – the valence band is mostly represented by Cl(p) while the conduction band is mostly represented by Pb(p). We can see; however, that the disordered phase of the material is highly unusual in that the random rotations of the FA creates traps inside of the band gap. Both acceptor and donor deep defects are visible. The HOMO and LUMO indicate that these defects are localized to the formamidinium ions both in the HOMO and the LUMO. This highly unusual result implies that the formamidinium orientation will strongly influence photovoltaic properties of the material. Previously, we proposed that ammonia will temporarily heal the chloride vacancy which would temporarily restore the material's electrical conductivity; however, we can also now propose that the healing also temporarily lowers the disorder of the material.
![]() | ||
| Fig. 7 SCAN + D3 plots of (A) HOMO, (B) LUMO and (C) DOS of the ordered and disordered phases of cubic FAPbCl3. | ||
Footnote |
| † Electronic supplementary information (ESI) available. See DOI: 10.1039/d2ma00087c |
| This journal is © The Royal Society of Chemistry 2022 |