betaFeedback

WorksPhysics and Astronomy

Vibrational properties differ between halide and chalcogenide perovskite semiconductors, and it matters for optoelectronic performance

Abstract

We report a comparative study of temperature-dependent photoluminescence and structural dynamics of two perovskite semiconductors, the chalcogenide BaZrS3_3 (BZS) and the halide CsPbBr3_3 (CPB). These materials have similar crystal structures and direct band gaps, but we find that they have quite distinct optoelectronic and vibrational properties. Both materials exhibit thermally-activated non-radiative recombination, but the non-radiative recombination rate in BZS is between two and four orders of magnitude faster than in CPB. Raman spectroscopy reveals that the effects of phonon anharmonicity are far more pronounced in CPB than in BZS. Further, although both materials feature a large dielectric response due to low-energy polar optical phonons, the phonons in CPB are substantially lower in energy than in BZS. Our results suggest that electron-phonon coupling in BZS is more effective at non-radiative recombination than in CPB, and that BZS may also have a substantially higher concentration of non-radiative recombination centers than CPB. The low defect concentration in CPB may be related to the ease of lattice reconfiguration, typified by anharmonic bonding. It remains to be seen to what extent these differences are inherent to the chalcogenide and halide perovskites and to what extent they can be affected by materials processing; comparing BZS single-crystals and thin films provides reason for optimism.

Introduction

Halide perovskites exhibit outstanding optoelectronic properties and are under intensive development for photovoltaic (PV) applications [1, 2, 3]. They can be synthesized at near-room temperature and feature sharp band-edge optical absorption and long excited-state charge carrier lifetimes [4, 5, 6, 7, 8]. They also suffer from the drawbacks of thermal and chemical instability and lead toxicity. For these reasons, research continues into alternative materials for thin-film photovoltaics. Chalcogenide perovskites are an intriguing alternative. They feature similar crystal structure and range of direct band gap as their halide counterparts [9, 10, 11]. They also share a very large dielectric constant, which suggests similar lattice dynamics and screening of charged-defects [13, 31]. However, chalcogenide perovskites are, in other ways, quite different. The materials are made of earth-abundant elements with little toxicity concerns. Synthesis is challenging and typically requires very high temperature and air-free processing. Once formed, they are very stable [9, 15, 16, 17, 24]. The optoelectronic properties of chalcogenide perovskites have been little reported. Existing published results suggest that photoluminescence (PL) is not strong and that defect control is a challenge, but long PL transient decay times have been measured [9, 16, 17, 19, 20, 21, 25]. No chalcogenide perovskite thin-film solar cells have been reported, likely due to the challenging conditions required for film synthesis.

Many researchers, including ourselves, have argued that the outstanding optoelectronic properties of halide perovskites are intertwined with their strongly anharmonic structural dynamics [22, 23, 24, 25, 26, 27, 28, 29, 30, 31]. This poses questions of how the optoelectronic properties and lattice dynamics of chalcogenide and halide perovskites compare and what we can learn about photovoltaic performance from this comparison.

Here, we employ temperature-dependent PL and Raman scattering spectroscopy, alongside first-principles self-consistent phonons simulations, to compare two exemplary single-crystal samples: the chalcogenide perovskite BaZrS3_3 and the halide perovskite CsPbBr3_3. BaZrS3_3 and CsPbBr3_3 share the same distorted perovskite, corner-sharing, orthorhombic crystal structure (space group PnmaPnma) within the studied temperature range [32, 33]. Between 80 and 300 K, the PL emission intensity of both materials decreases due to thermally-activated non-radiative recombination, with electron-phonon interactions broadening the PL linewidth in both cases. However, CsPbBr3_3 emission intensity surpasses that of single-crystal BaZrS3_3 by over four orders of magnitude, suggesting a significantly higher concentration of very-low-barrier non-radiative recombination centers in BaZrS3_3.

We use Raman spectroscopy and first-principles simulations to show that the phonon spectrum in BaZrS3_3 extends to nearly double the frequency in CsPbBr3_3. Raman analysis demonstrates that CsPbBr3_3 exhibits pronounced anharmonic characteristics, even at low temperatures. In contrast, BaZrS3_3 exhibits predominantly harmonic vibrational behavior akin to traditional tetrahedrally-bonded semiconductors. Our findings are consistent with the hypotheses that the lower-frequency and strongly anharmonic structural dynamics in CsPbBr3_3 may reduce carrier capture cross sections due to Franck-Condon effects [34]. They may also decrease defect concentration through annealing, even at room temperature [35]. Definitive tests of these hypotheses await detailed calculations of rate coefficients supported by first-principles quantum calculations and defect spectroscopy to experimentally identify and quantify recombination-active defects.

Methods

We grew single crystals of BaZrS3_3 with dimensions of approximately 100 μ\mum using a flux method as reported previously [9]. We grew single crystals of CsPbBr3_3 with dimensions on the order of 2 mm using the vapor saturation of an antisolvent (VSA) method as reported previously [36].

We performed Raman scattering in a home-built system in back-scattering geometry. We used a 1.58 eV CW pump-diode laser (Toptica Inc., USA), which is below the band gap of BaZrS3_3 (1.9 eV) [9] [37] and CsPbBr3_3 (2.32 eV) [38]. We performed PL spectroscopy using the same optical system and a 2.54 eV CW sapphire-pumped diode laser (Coherent Inc., USA). The signal was measured in a 1 m long spectrometer (FHR1000, Horiba Inc.) equipped with a holographic grating (1800 gr/mm for Raman and 150 gr/mm for PL) and a Synapse Plus CCD detector (Horiba Inc.).

The sample temperature was controlled using a liquid-N2_2 or liquid He-cooled high-vacuum optical cryostat (Janis Inc., USA) for T≤400T \le400 K. For higher temperature measurements, the temperature was controlled in an optical furnace (Linkam TS1000) under N2_2 flow. We performed measurements for a series of increasing temperatures. We took precautions to isolate the effect of sample temperature on PL intensity by maintaining stable measurement conditions, refocusing the laser before every measurement, and taking repeated measurements across the whole temperature range. The measurement system and protocols are further described in Refs. [21] [39, 40, 41, 42].

We calculate normalized PL yield (YY) by analyzing the integrated PL intensity, including both the band-to-band peaks and the shoulders. The PL data is measured in counts, and we compute intensity by dividing by the acquisition time. We define YY as the ratio between the PL intensity emitted from the sample (IPLI_{\mathrm{PL}}, measured in counts/sec) and the excitation laser intensity that reaches the interior of the sample (IinI_{\mathrm{in}}):

Y=IPLIin=IPLI0(1−Rs).(1)Y = \frac{I_{\mathrm{PL}}}{I_{\mathrm{in}}} = \frac{I_{\mathrm{PL}}}{I_0}(1-R_s). \tag*{(1)}

We measure the laser intensity I0I_0 (in photons/sec) at the microscope entrance. We compute IinI_{\mathrm{in}} from I0I_0 by accounting for losses due to reflection from the cryostat windows and the sample surface. The same cryostat window reflection losses affect the collected PL intensity and cancel in the expression for YY, leaving only a dependence on the sample surface power reflection coefficient (RsR_s). Further details are presented in Sec. S1 in the Supplemental Material (SM) [44].

Results

Photoluminescence and Non-Radiative Recombination

In Figure 1, we present temperature-dependent PL spectra of BaZrS3_3 and CsPbBr3_3 measured between 80–300 K. At low temperatures, spectra are dominated by band-edge recombination at Eg∼1.96E_g\mathrel{\sim}1.96 eV and 2.34 eV for BaZrS3_3 and CsPbBr3_3, respectively. The low-temperature spectra for both samples feature low-energy shoulders, which are quenched at elevated temperatures. The band-to-band peak and the shoulder are much broader in BaZrS3_3 than in CsPbBr3_3. In both materials, PL is quenched with increasing temperature, and for BaZrS3_3, it becomes too weak to measure above 170 K. To eliminate external features in the PL spectra, the high-temperature spectra were subtracted from those in Figure 1.

Figure 1. Temperature-dependent PL spectra. PL spectra measured between 80 K and 300 K for (a) BaZrS3 and (b) CsPbBr3, after subtraction of the background spectra at 170 K and 300 K, respectively. The line color indicates temperature. BaZrS3 and CsPbBr3 were measured with different pump fluence: 4.3 MJ/cm2 for BaZrS3, and 2.7 × 10^{-4} MJ/cm2 for CsPbBr3.

BaZrS3_3 and CsPbBr3_3 are direct band-gap semiconductors and strongly absorb the pump laser. At the pump laser energy of 2.54 eV, the absorption coefficient of BaZrS3_3 is 1.68×1051.68 \times10^5 cm−1^{-1}, and that of CsPbBr3_3 is 4.04×1054.04 \times10^5 cm−1^{-1} [3, 15]. Therefore, the pump laser absorption depth is much smaller than the thickness for both single-crystal samples, and all the light entering the samples is absorbed. However, the PL emission intensity of CsPbBr3 dwarfs that of BaZrS3.

In Figure 2(a), we present an Arrhenius plot of the normalized PL yield (YY). We fit the data to a minimal model of thermally-activated non-radiative recombination [46, 47]:

Figure 2. Model-based analysis of normalized PL yield (YY, Eq. (2)) and linewidth (ΓPL\Gamma_{\mathrm{PL}}). (a) YY as a function of the inverse temperature. (b) Fit to a model with a second non-radiative recombination pathway. Eb/kBE_b/k_{\mathrm{B}} fall below our measurement temperature window. (c) ΓPL\Gamma_{\mathrm{PL}} of the band-to-band peak as a function of the temperature. All fitted values can be found in Sec. S1 of the SM [44]. Purple (circles) and orange (diamonds) traces (markers) are for CsPbBr3 and BaZrS3, respectively.

Y(T)=11+Ae−Ea/kBT+Be−Eb/kBT(2)Y(T) = \frac{1}{1 + A e^{-E_a/k_{\mathrm{B}}T} + B e^{-E_b/k_{\mathrm{B}}T}} \tag*{(2)}

EaE_a and EbE_b are the activation energies of two non-radiative recombination mechanisms, and A=knr,a0/krA = k_{\mathrm{nr,a}}^0/k_{\mathrm{r}} and B=knr,b0/krB = k_{\mathrm{nr,b}}^0/k_{\mathrm{r}} are ratios of exponential prefactors that compare non-radiative to radiative recombination rates. As discussed below, two non-radiative recombination mechanisms are required for a minimal model of the data. This model assumes that all radiative recombination has a sufficiently small activation energy that appears barrierless within the measurement window. The effect of this assumption is that the PL internal quantum efficiency (IQE) approaches unity at low temperatures. The best-fit model parameters are presented in Table I.

We find that BaZrS3 and CsPbBr3 feature a non-radiative recombination process with similar activation energy Ea=68±7E_a = 68 \pm7 and 86±586 \pm5 meV, respectively. This is apparent by eye in the Arrhenius plot (Figure 2(a)), as both data sets have similar curvature and a knee at approximately the same temperature. However, the PL yield of CsPbBr3 is over four orders of magnitude larger than that of BaZrS3, and the data do not extrapolate towards the same yield in the low-temperature limit. The presence of non-radiative recombination mechanisms with similar activation energy belies the difference in PL intensity.

To explain the divergence in PL intensity, we invoke a second non-radiative recombination mechanism in BaZrS3, with activation energy EbE_b too low to be determined in our measurement window. We also assume a similar second mechanism in CsPbBr3 for consistency in the model function. By construction, the thermal scale Eb/kBE_b/k_{\mathrm{B}} falls well below our experimental temperature window, and therefore, EbE_b cannot be determined by regression. This low-temperature process is saturated within the measured temperature range, and e−Eb/kBT≈1e^{-E_b/k_{\mathrm{B}}T} \approx1. With this assumption, we fix the temperature scale Eb/kBE_b/k_{\mathrm{B}} at a low value and fit the data to determine BB. In Figure 2(b), we present a family of model functions with EbE_b varied manually; for EbE_b sufficiently small, its value has an insignificant impact on the fit results and our conclusions. The ratio BBZS/BCPBB_{\mathrm{BZS}}/B_{\mathrm{CPB}} is meaningful, as it determines the measured difference in low-temperature PL yield between BaZrS3 (BZS) and CsPbBr3 (CPB). We find that:

TABLE I. Best-fit parameters for PL yield, Eq. (2). Results are presented here for Eb/kB=0E_b/k_{\mathrm{B}} = 0 K. In Table S3 in the SM [44], we present results for other choices of Eb/kBE_b/k_{\mathrm{B}}.

MaterialI0I_0 (photons/sec)EaE_a (meV)Eb/kBE_b/k_{\mathrm{B}} (K)AABB
BaZrS3_36.7×10156.7 \times10^{15}68±768 \pm70(6.9±3.8)×1017(6.9 \pm3.8) \times10^{17}(2.8±0.5)×1014(2.8 \pm0.5) \times10^{14}
CsPbBr3_34.3×10114.3 \times10^{11}86±586 \pm50(2.8±0.8)×1013(2.8 \pm0.8) \times10^{13}(9.6±1.4)×109(9.6 \pm1.4) \times10^9

Table 1.

BBZSBCPB=(knr,b0/kr)BZS(knr,b0/kr)CPB=28,181.(3)\frac{B_{\mathrm{BZS}}}{B_{\mathrm{CPB}}} = \frac{\left(k_{\mathrm{nr,b}}^0/k_{\mathrm{r}}\right)_{\mathrm{BZS}}}{\left(k_{\mathrm{nr,b}}^0/k_{\mathrm{r}}\right)_{\mathrm{CPB}}} = 28{,}181 . \tag*{(3)}

Both materials have comparable optical absorption coefficients near their band edge; therefore, we can approximate (kr,CPB/kr,BZS)≈O(1)(k_{\mathrm{r,CPB}}/k_{\mathrm{r,BZS}}) \approx O(1). We therefore conclude that the rate knr,b0k_{\mathrm{nr,b}}^0 of low- (or zero-) barrier non-radiative recombination is O(104)O(10^4) times faster in BaZrS3_3 than in CsPbBr3_3.

We analyze the PL linewidth (ΓPL\Gamma_{\mathrm{PL}}) for insight into how excited states couple to lattice vibrations. In Figure 2(b), we show the temperature dependence of ΓPL\Gamma_{\mathrm{PL}} of the main PL peak at the EgE_g for each material. We use multiple Gaussian peaks to fit the data, three for CsPbBr3_3 and four for BaZrS3_3. For both materials, the main PL peak is well modeled by two Gaussian peaks, and we determine ΓPL\Gamma_{\mathrm{PL}} numerically from the best-fit models. We fit ΓPL\Gamma_{\mathrm{PL}} to a relation describing thermal and non-thermal broadening, previously applied to halide perovskites [38]:

ΓPL(T)=Γ0+γeEph/kBT−1,(4)\Gamma_{\mathrm{PL}}(T) = \Gamma_0 + \frac{\gamma}{e^{E_{\mathrm{ph}}/k_{\mathrm{B}}T} - 1} , \tag*{(4)}

Γ0\Gamma_0 represents temperature-independent contributions to ΓPL\Gamma_{\mathrm{PL}}, such as heterogeneous broadening due to quenched structural disorder. The second term describes broadening due to coupling between excited states and lattice vibrations with characteristic energy EphE_{\mathrm{ph}} and electron-phonon coupling constant γ\gamma. These lattice vibrations are likely longitudinal optical phonons (see below), and the electron-phonon interactions can be described through deformation potentials.

We find that luminescent band-edge states couple to vibrations with characteristic energy that is similar in both materials: Eph=24±7E_{\mathrm{ph}} = 24 \pm7 meV for BaZrS3_3, and 17±417 \pm4 meV for CsPbBr3_3 (see below). The electron-phonon coupling constant γ\gamma is larger for BaZrS3_3 (200±100200 \pm100 meV) than for CsPbBr3_3 (50±1050 \pm10 meV). γ\gamma is related to deformation potentials and polar-optical coupling. The higher value in BaZrS3_3 may be related to the stronger covalent nature of the Zr-S bonds compared to Pb-Br bonds: 32% covalent for Zr-S, compared to 26% for Pb-Br, according to the Pauling scale. This implies enhanced sensitivity of electron energies to bond angles in the chalcogenide, as we have studied previously [13] [48]. We also see that BaZrS3_3 has larger heterogeneous broadening, Γ0\Gamma_0, than CsPbBr3_3, reflecting a higher concentration of structural defects and other forms of non-thermal disorder.

In addition to affecting the PL spectra, these factors affect how steeply optical absorption rises at the band gap, the width of the Urbach tail, and the phonon-limited charge transport mobility, and are therefore relevant to photovoltaic performance [49–51].

Raman Scattering and Phonon Dynamics

The lower luminescence yield and larger PL broadening in BaZrS3_3, compared to CsPbBr3_3, are related to differences in lattice vibrations. To better understand these relationships, we turn to theoretical and experimental studies of the phonon spectra. In Figure 3, we present the total (black) and partial (orange, green, and blue) vibrational density of states of BaZrS3_3 (top) and CsPbBr3_3 (bottom), calculated ab initio using the temperature-dependent effective potential (TDEP) method at 300 K (details in Sec. S2 the SM [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19] [44]). The optical vibrations in BaZrS3_3 span frequencies up to 430 cm−1^{-1} (53 meV), while those in CsPbBr3_3 peak at 160 cm−1^{-1} (20 meV). This is due to the higher mass of the atoms in CsPbBr3_3 and the more covalent character of the bonds in BaZrS3_3 [68].

Figure 3. Calculated room-temperature phonon density of states (DOS, right axis) of BaZrS3_3 and CsPbBr3_3. The colors indicate the partial contribution of the A (blue), M (green), and X (orange) ions in the AMX3_3 formula. The red lines are the measure (left axis) of the LO character.

Longitudinal optical (LO) phonons couple to electronic excitations and therefore likely contribute most to the temperature-dependent PL linewidth broadening. In Figure 3, the vertical red lines provide a measure of the phonon LO character. As detailed in Sec. S2 in the SM [44], the LO character is estimated by computing the longitudinal charges for optical vibrations from the mode eigenvectors and atomic Born charge tensors projected on the [101] excitation direction and divided by the vibration frequency to account for the average mode amplitudes. We see that BaZrS3_3 has many phonons with prominent LO character along the [101] excitation direction, whereas CsPbBr3_3 only shows one mode of appreciable LO character at ≈125\approx125 cm−1^{-1}. This factor likely contributes to the stronger temperature-dependent PL broadening in BaZrS3_3 in addition to the higher deformation potentials.

The characteristic energy of lattice vibrations responsible for PL broadening in CsPbBr3_3, Eph≃17E_{\mathrm{ph}} \simeq17 meV (≃140\simeq140 cm−1^{-1}), is comparable to the most prominent LO phonon in the calculated spectrum. For BaZrS3_3, Eph≃24E_{\mathrm{ph}} \simeq24 meV (≃190\simeq190 cm−1^{-1}) falls roughly in the middle of the calculated spectrum. Likely, band-edge states couple to a diversity of lattice vibrations in both materials through polar-optical coupling and deformation potentials, and the characteristic energy determined by PL broadening represents a distribution of processes. Similar conclusions have been reached in studies of ultra-fast lattice rearrangement following optical excitation of halide perovskites, including CsPbBr3_3 [69].

In Figure 4, we present the low-frequency Raman spectra of BaZrS3_3 and CsPbBr3_3 measured in the temperature range 80–350 K. The spectra of BaZrS3_3 and CsPbBr3_3 are consistent with previous reports [23] [42]. In § 7 in the SM [44], we present the Raman spectra of BaZrS3_3 measured over a wider temperature range, from 10–875 K, showing no indication of a phase transition [24, 26, 27].

Figure 4. Temperature-dependent Raman scattering spectra of BaZrS3_3 and CsPbBr3_3. Temperature is indicated by color. The individual data sets are shifted vertically for clarity; the dashed horizontal lines indicate the zero level for each data set.

The Raman spectra of CsPbBr3_3 exhibit sharp first-order peaks (one photon scattered by one vibration) [22] at low temperatures, which shift and broaden significantly with increasing temperature. These structural dynamics are characteristic of halide and certain oxide perovskites [20] [75, 76, 77]. The broad, featureless spectra at high temperatures, combined with a rapid decrease in vibrational lifetime (evidenced by peak broadening), indicate the impact of anharmonic vibrational dynamics [23] [38].

At low temperatures, the Raman spectra of BaZrS3_3 resemble those of CsPbBr3_3 and other halide perovskites [39, 78], with a similar number of prominent peaks divided into doublets and triplets. The vibrations in BaZrS3_3 are higher in frequency than those in CsPbBr3_3. Unlike CsPbBr3_3, the Raman peaks in BaZrS3_3 remain sharp and distinct with increasing temperature, and even at 350 K, there is scant signature of anharmonic dynamics. The harmonic dynamics in BaZrS3_3 recall those in tetrahedrally-bonded semiconductors (e.g. GaAs), in which vibrations along the stiff covalent bonds do not interact and change little with temperature [79, 80].

We can more quantitatively compare the structural dynamics of BaZrS3_3 and CsPbBr3_3 by considering the temperature dependence of the phonon frequencies and linewidths. We model the Raman spectra as a convolution of pseudo-Voigt peaks multiplied by the Bose-Einstein occupation factor [28, 29, 30] (see § 7 in the SM [44]). We focus on the prominent peaks because they yield the most confident fits. In Figure 5(a) we present the vibrational frequency (Ω\Omega) and linewidth (ΓRj\Gamma_{Rj}) of the prominent first-order peaks of BaZrS3_3 (left panel, orange) and CsPbBr3_3 (right panel, purple). The frequency and linewidth data are divided by their values at 80 K for ease of comparison (for original values, see § 7 in the SM [44]). The Raman peaks of BaZrS3_3 red shift and broaden linearly with increasing temperature, as is expected from perturbation theory for weak anharmonic effects [84, 85]. A slight change in slope near 200 K may relate to the observed increase in the low-frequency dielectric susceptibility in this same temperature range [31].

Figure 5. Analysis of Raman Spectra: (a) Frequencies (Ω\Omega - top) and linewidths (ΓRj\Gamma_{\mathrm{R}j} - bottom) of the first-order peaks of BaZrS3_3 (left) and CsPbBr3_3 (right) as a function of temperature, relative to the value at 80 K. Different markers mark different peaks. (b) Linear regression results (R2R^2 - top, and slopes α\alpha - bottom) of the temperature-dependent frequencies (left) and linewidths (right) of BaZrS3_3 and CsPbBr3_3, between 80-280 K. The results are plotted as a function of peak frequency at 80 K. Orange and purple markers are for BaZrS3_3 and CsPbBr3_3, respectively.

CsPbBr3_3 exhibits a similar monotonic and linear trend in frequency and linewidth with increasing temperature up to 280 K, above which the data deviate substantially, including sharp blue shifts and rapidly increasing linewidth (see Figure S6 is the SM [44]). We attribute these changes to two causes: (1) the thermal population of all the optical phonons above TDebye≈250T_{\mathrm{Debye}} \approx250 K, and (2) the incipient orthorhombic-to-tetragonal phase transition at T≈360T \approx360 K. Peak broadening with increasing temperature is attributed to the shortening of vibrational lifetime due to phonon-phonon scattering [29] [84, 85]. It is apparent that the structural dynamics of BaZrS3_3 exhibit minor thermal effects and are less impacted by anharmonic vibrational interactions than the dynamics of CsPbBr3_3 [84, 85].

Figure S6. Analysis of Raman Spectra: (a) Ωj\Omega_j (top) and ΓRj\Gamma_{Rj} (bottom) of the first-order peaks of BaZrS3_3 (left) and CsPbBr3_3 (right) as a function of temperature. (b) Linear regression results (R2^2 - top, and slopes, alpha - bottom) of Ωj(T)\Omega_j(T) (left) and ΓRj(T)\Gamma_{Rj}(T) (right) of BaZrS3_3 (diamonds) and CsPbBr3_3 (circles), between 80–280 K. A different color represents each peak.

The extent of anharmonicity in the phonon dynamics can be further quantified by analyzing the deviation from linearity of the temperature dependence of the phonon frequencies and linewidths. A linear temperature dependence is expected within perturbation theory for weak phonon-phonon interactions. Therefore, strong and anharmonic interactions are indicated by deviations from linearity. In Figure 5(b), we present the R2R^2 parameter and the absolute value of the slope (∣α∣\lvert\alpha\rvert) of linear regression to the Ω(T)\Omega(T) and ΓRj(T)\Gamma_{\mathrm{R}j}(T) (for mean values, see § 7 in the SM [44]). For CsPbBr3_3, the R2R^2 values are far from unity, and the slopes are much larger than for BaZrS3_3. This analysis further supports the conclusion that anharmonic effects in BaZrS3_3 can be modeled as a weak perturbation, whereas in CsPbBr3_3, they substantially change the lattice vibrational properties.

Discussion

Lead halide perovskites excel as PV absorbers largely due to their slow rates of defect-assisted Shockley-Read-Hall (SRH) non-radiative free energy loss. Electron-phonon interactions mediate SRH recombination processes, converting excited-state energy into phonons and heat [25, 27, 31] [86]. Therefore, by focusing on vibrational properties and electron-phonon interactions, we can better understand the origin of the excellent PV performance of the halides and identify promising new semiconductors for thin-film PV.

Chalcogenide perovskites have an untested potential for PV mainly due to challenges in thin film synthesis. There are few published measurements of the optoelectronic properties of chalcogenide perovskites, and the results are ambiguous. We reported time-resolved PL data that suggest long radiative lifetimes in BaZrS3_3 and the Ruddlesden-Popper compound Ba3_3Zr2_2S7_7, comparable to lead halides and other high-performing PV materials such as CIGS [19]. Several research groups have reported band-edge PL measured from BaZrS3_3 samples of various form factors, including solution-processed nanoparticles and thin films made by sulfurizing oxide precursors [16, 20, 21, 25]. However, quantitative PL measurements (albeit on early-stage powder samples) suggest low quantum efficiency and low quasi-Fermi level splitting (i.e. open-circuit voltage potential) [9]. Our experience of measuring PL on BaZrS3_3 single crystals is that many samples have weak or no measurable band-edge PL and that the band-edge PL can be strongly position-dependent. We have observed similar variability in measuring PL on epitaxial thin films (unpublished). Substantial sample-to-sample variation, combined with challenges in synthesis and defect control, makes it difficult to assess the technological potential of chalcogenide perovskites.

SRH recombination is much faster in BaZrS3_3 than in CsPbBr3_3 due to an unidentified recombination pathway that is active even at low temperature, implying that it has a very low thermal activation energy (Figure 2(b)). The theory of multi-phonon non-radiative recombination highlights the importance of the overlap between the vibrational wavefunctions of the initial and final electronic states, consistent with the Franck-Condon principle [34]. Model calculations for a harmonic lattice indicate a steep dependence of SRH recombination rate on phonon energy [86]. Lower phonon energy means more phonons are needed to participate in non-radiative recombination processes. Within this model, the overlap of vibrational wavefunctions is suppressed with increasing phonon number. Therefore, the lower phonon frequency in CsPbBr3_3 compared to BaZrS3_3 may explain the difference in SRH recombination rates, even if the concentration and properties of defects are similar.

The same harmonic model also predicts that SRH recombination rates increase with the extent of lattice distortion accompanying defect charge capture; this distortion can be parameterized by the Huang-Rhys factor [34, 86]. These lattice distortions often result in long-lived charge trapping and are responsible for persistent photoconductivity in many semiconductors; it is related to DX-center phenomena in III-V compounds, and we termed it defect-level switching in the context of resistive switches [87–90]. There have been many studies of persistent photoconductivity in halide perovskites. Still, there is no clear model for defects in CsPbBr3_3 [91–94]. The timescale of photoconductivity decay in BaZrS3_3 crystals depends on sample polishing, suggesting that defects introduced during plastic deformation (e.g. dislocations) may be effective charge traps with large Huang-Rhys factors [95]. However, these observations may not be directly relevant to the Huang-Rhys factors of recombination-active defects, which may differ from the traps that influence photoconductivity decay. Further, it has been suggested that polaronic effects resulting from strong charge-lattice coupling - including Fröhlich (i.e. large, effective mass states) polarons and defect-localized polarons - may suppress SRH recombination rates by screening dissimilar charges [96, 97]. It is likely that a full quantum calculation of electron-phonon interactions, combined with first-principles modeling of the known crystal structures and the most likely recombination-active defects, will be required to resolve these questions over the role of strong electron-phonon coupling on SRH recombination rates.

The effect of anharmonic phonon-phonon interactions on SRH recombination remains an open question. It would be enlightening to compare a model calculation of vibrational wavefunction overlap during multi-phonon recombination with and without anharmonic lattice vibrations and for varying Huang-Rhys factor for otherwise similar phonon energies. Model system calculations could help to understand the conflated effects of low phonon energy, lattice distortion at defects, and anharmonic vibrations on SRH recombination rates. However, full quantum calculations are likely needed to accurately model the differences in carrier capture rates between halide and chalcogenide perovskites.

The discussion to this point has focused on the effects of lattice vibrations on SRH recombination rates via the carrier capture processes. However, another significant unknown is the concentration of recombination-active defects. Anharmonic vibrations indicate softer bonds, a more pliable crystal lattice, and lower activation energy for ionic diffusion [98–100]. Therefore, anharmonic vibrations in lead halide perovskites may improve optoelectronic performance by accelerating the elimination of point defects by annealing, even at room temperature (termed “self-healing”) [35, 101–104]. BaZrS3_3, in contrast, requires high temperature to form and, therefore, is likely to retain quenched-in defects (e.g. chemical impurities). This suggests that the PL yield and PV performance of chalcogenide perovskites will depend more strongly on defect control during sample processing than the performance of halide perovskites.

Defect characterization by methods such as deep-level transient spectroscopy (DLTS) and related forms of impedance spectroscopy could directly measure defect concentrations and quantify the extent of self-healing [105–111]. Interpretation of DLTS data is challenging for halide perovskites due to mixed ionic-electronic conduction, but progress has been reported [112]. Similar studies have not yet been undertaken for chalcogenide perovskites.

Conclusions

We have presented a detailed spectroscopic comparison between two perovskite semiconductors, the chalcogenide BaZrS3_3 and the well-studied halide CsPbBr3_3. We find that the PL emission yield of CsPbBr3_3 is over four orders of magnitude higher than that of BaZrS3_3 for the single-crystal samples considered here, indicative of a far higher concentration of very-low-barrier, non-radiative recombination centers in BaZrS3_3. Ab initio simulations show that BaZrS3_3 exhibits higher LO vibrational frequencies than CsPbBr3_3; this may increase carrier capture cross sections due to Franck-Condon effects [34]. Raman scattering results suggest that the strongly anharmonic structural dynamics in CsPbBr3_3, relative to BaZrS3_3, may reduce defect concentration through accelerated solid-state diffusion and annealing [35].

We are optimistic that the fast SRH recombination rates observed here for BaZrS3_3 can be reduced through sample processing. Large sample-to-sample variability, and preliminary results on BaZrS3_3 epitaxial thin films (not shown), suggest that low PL yield may not be intrinsic to BaZrS3_3The usefulness of BaZrS3_3 (and likely other chalcogenide perovskites) for optoelectronic applications will hinge on continued progress in defect control during synthesis and processing - as has been historically required for other semiconductors.

Acknowledgements

We acknowledge the support of Dr. I. Pinkas (WIS) for help designing the experimental setup, Dr. L. Segev (WIS) for developing the Raman software, and Dr. S. Aharon (Princeton) for help in the synthesis of CsPbBr3_3 crystals. OY acknowledges funding from the European Research Council starting grant (850041 - ANHARMONIC). We acknowledge support from the MIT-Israel Zuckerman STEM Fund and the Sagol Weizmann-MIT Bridge Program. We acknowledge support from the United States-Israel Binational Science Foundation, grant no. 2020270. We acknowledge support from the National Science Foundation (NSF) under grant no. 1751736. We acknowledge support from the Skolkovo Institute of Science and Technology and the MIT-Skoltech Next Generation Program. KY acknowledges support from the NSF Graduate Research Fellowship, grant no. 1745302. B.Z. and J.R. acknowledge support from an ARO MURI with award number W911NF-21-1-0327, an NSF grant with award number DMR-2122071, and an ONR grant with award number N00014-23-1-2818. F.K. acknowledges support from the Swedish Research Council (VR) program 2020-04630, and the Swedish e-Science Research Centre (SeRC). The computations were enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at NSC and PDC partially funded by the Swedish Research Council through grant agreement no. 2022-06725.

Supplemental Material

Vibrational properties differ between halide and chalcogenide perovskite semiconductors, and it matters for optoelectronic performance

Photoluminescence Data and Analysis

PL Data Processing

Figure S1 shows the as-measured photoluminescence (PL) spectra of BaZrS3_3 (left) and CsPbBr3_3 (right) in the temperature range of 80–300 K. The measurement parameters are listed in Table S2.

Figure S1. Photoluminescence Raw Spectra of BaZrS3_3 (left) and CsPbBr3_3 (right) as measured, from 80–300 K.

ParameterBaZrS3_3CsPbBr3_3
Excitation Wavelength [nm]488488
Excitation Energy [eV]2.542.54
Excitation Power [mW]2.741.74×10−41.74 \times10^{-4}
Acquisition Time [s]155
Spot Size [μ\mum]1.11.1
Excitation Fluence [MJ/cm2^2]4.32.7×10−42.7 \times10^{-4}

Table S2. Photoluminescence Measurement Parameters

We first subtract a wavelength-independent constant background from each spectrum from electronic noise to ensure that the tails of the spectra approach zero. In the case of CsPbBr3_3, this was the only pre-processing step performed before integrating the entire spectra. In the case of BaZrS3_3, due to its low PL intensity, we found a small but non-negligible contribution from the excitation laser, reaching a maximum near 2.2 eV. The PL signal from the

Determination of Normalized PL Yield

sample is quenched at ≈170\approx170 K, leaving only the temperature-independent laser contribution. Therefore, we remove it by subtracting the spectrum measured at high temperature from all lower-temperature data.

We model the band-to-band PL intensity by fitting four Gaussian peaks to the spectra for BaZrS3_3 and three in the case of CsPbBr3_3. Figure S2 shows an example of a fit at 80 K.

Figure S2. Photoluminescence intensity modeling. Gaussian deconvolution results for the photoluminescence spectra of BaZrS3_3 (left) and CsPbBr3_3 (right) at 80 K. The blue and green traces are the data and fit, respectively. The shaded traces are individual peaks – purple and orange for the main transition and other transitions, respectively.

Determination of Normalized PL Yield

To convert the measured spectra in counts per second to normalized PL yield (YY, Eq. (1) [1]), we start by considering the incident excitation laser power measured upstream of the microscope (I0I_0). The cryostat windows attenuate the incident laser light and the sample PL emission, and these effects cancel in the expression for YY. We calculate sample reflection loss using the Fresnel equation; with ϵ=n\sqrt{\epsilon}=n, we obtain n=3.1n=3.1 for BaZrS3_3 and n=2.3n=2.3 for CsPbBr3_3 [2, 3].

We fit the normalized PL yield data to Eq. (2) for fixed values of EbE_b, as described in the main text [1]. In Table S3, we report the excitation power I0I_0 and the best-fit values.

MaterialI0I_0 (photons/sec)EaE_a (meV)Eb/kBE_b/k_B (K)AB
BaZrS3_36.7×10156.7 \times10^{15}68±768 \pm70(6.9±3.8)×1017(6.9 \pm3.8) \times10^{17}(2.8±0.5)×1014(2.8 \pm0.5) \times10^{14}
BaZrS3_36.7×10156.7 \times10^{15}68±768 \pm72(6.9±3.8)×1017(6.9 \pm3.8) \times10^{17}(2.9±0.5)×1014(2.9 \pm0.5) \times10^{14}
BaZrS3_36.7×10156.7 \times10^{15}68±768 \pm74(7.0±3.9)×1017(7.0 \pm3.9) \times10^{17}(3.0±0.5)×1014(3.0 \pm0.5) \times10^{14}
BaZrS3_36.7×10156.7 \times10^{15}68±768 \pm78(7.2±4.0)×1017(7.2 \pm4.0) \times10^{17}(3.1±0.5)×1014(3.1 \pm0.5) \times10^{14}
CsPbBr3_34.3×10114.3 \times10^{11}86±586 \pm50(2.8±0.8)×1013(2.8 \pm0.8) \times10^{13}(9.6±1.4)×109(9.6 \pm1.4) \times10^9
CsPbBr3_34.3×10114.3 \times10^{11}86±586 \pm52(2.8±0.8)×1013(2.8 \pm0.8) \times10^{13}(9.9±1.5)×109(9.9 \pm1.5) \times10^9
CsPbBr3_34.3×10114.3 \times10^{11}86±586 \pm54(2.9±0.8)×1013(2.9 \pm0.8) \times10^{13}(1.0±0.2)×1010(1.0 \pm0.2) \times10^{10}
CsPbBr3_34.3×10114.3 \times10^{11}86±586 \pm58(2.9±0.8)×1013(2.9 \pm0.8) \times10^{13}(1.1±0.2)×1010(1.1 \pm0.2) \times10^{10}

Table S3. Normalized PL Yield Fitted Parameters

Determination of PL Linewidth

We require two considerably overlapping Gaussian peaks to model the band-to-band PL lineshape of both materials. Therefore, we numerically calculate the linewidth (full-width-half-maximum, ΓPL\Gamma_{\mathrm{PL}}) of these combined peaks. We fit the results using Eq. (4) [1], and the best-fit parameters are presented in Table S4.

MaterialEphE_{ph} (meV)γ\gamma (meV)Γ0\Gamma_{0} (meV)
BaZrS3\mathrm{BaZrS}_{3}24±724 \pm7205±116205 \pm11657±457 \pm4
CsPbBr3\mathrm{CsPbBr}_{3}18±418 \pm448±1248 \pm129±29 \pm2

Table S4. PL Linewidth Fitted Parameters

Phonon density of states and LO character

Self-consistent phonon scheme

We use the stochastic temperature-dependent effective potential (sTDEP) scheme to compute temperature-dependent phonons described by the force constants {Φij(T)}\{\Phi_{ij}(T)\} which depend on temperature TT. This scheme is formally equivalent to the QSCAILD method described in Ref. [4] and relies on minimizing the mean-square error of the harmonic model forces,

S[Φ,T]≡∑i⟨(fi−fΦ;i)2⟩Φ,T,(5)S[\Phi,T] \equiv\sum_i \left\langle(f_i-f_{\Phi;i})^2 \right\rangle_{\Phi,T}, \tag*{(5)}

where ⟨⋅⟩Φ,T\langle\cdot\rangle_{\Phi,T} denotes a thermal average with respect to the force constants Φ\Phi at temperature TT, i,ji,j denote atomic and Cartesian indices, and the harmonic force fΦ;if_{\Phi;i} is given by fΦ;i=−∑jΦijujf_{\Phi;i}=-\sum_j\Phi_{ij}u_j, where uju_j is a vector with atomic displacements from the thermal average positions.

As shown and discussed in Ref. [4, 5], the least-squares solution to Eq. (5) is given by

Φij(T)=−⟨uiuj⟩Φ,T−1⟨fiuj⟩Φ,T≡⟨∂2V∂ui∂uj⟩Φ,T,(6)\Phi_{ij}(T)=-\left\langle u_i u_j\right\rangle_{\Phi,T}^{-1}\left\langle f_i u_j\right\rangle_{\Phi,T}\equiv\left\langle\frac{\partial^2 V}{\partial u_i\partial u_j}\right\rangle_{\Phi,T}, \tag*{(6)}

i.e., the configuration average of the second derivative of the potential energy VV with respect to the effective harmonic distribution of atomic positions at target temperature TT. This solution is equivalent to a Gibbs-Bogoliubov free energy minimization and yields the best trial free energy of harmonic form. Further details are given in [6, 7, 8].

After obtaining the temperature-dependent force constants at room temperature, we can use them to compute the dynamical matrix

Dij(q)=1mimj∑Leiq⋅RLΦi0,jL,(7)D_{ij}(\mathbf{q})=\frac{1}{\sqrt{m_i m_j}}\sum_L e^{i\mathbf{q}\cdot\mathbf{R}_L}\Phi_{i0,jL}, \tag*{(7)}

via mass weighting with the atomic masses mim_i and Fourier-transforming to q-space by summing over lattice vectors RL\mathbf{R}_L, from which temperature-renormalized dynamical properties can be inferred. In particular, we define the eigenvalue equation diagonalizing the atomic basis i,ji,j to mode indices ss,

∑jDij(q)ϵjqs=ωqs2ϵiqs,(8)\sum_j D_{ij}(\mathbf{q})\epsilon_{j\mathbf{q}s}=\omega_{\mathbf{q}s}^2\epsilon_{i\mathbf{q}s}, \tag*{(8)}

with eigenfrequencies ωqs\omega_{qs} and eigenvectors ϵiqs\epsilon_{iqs}.

Without going into details, we note that these frequencies do not correspond to the quasiparticle poles of phonon excitations. For physical spectral properties, phonon-phonon interactions need to be taken into account, which can be done in the TDEP framework by obtaining third-order force constants [9]. We denote the fully thermally shifted quasiparticle frequencies as ω~qs\widetilde{\omega}_{qs} in the following.

Density of states

From the phonon quasiparticle spectrum {ω~qs}\{\widetilde{\omega}_{qs}\} at room temperature, we can compute the density of states (DOS)

DOS⁡(ω)=∑qsδ(ω−ω~qs).(9)\operatorname{DOS}(\omega) = \sum_{qs} \delta\left(\omega-\widetilde{\omega}_{qs}\right). \tag*{(9)}
LO character

Longitudinal optical (LO) modes are characterized by a longitudinal charge,

ZksLO=kk⋅Zks,(10)Z^{\mathrm{LO}}_{ks} = \frac{\mathbf{k}}{k} \cdot\mathbf{Z}_{ks}, \tag*{(10)}

with

Zks=∑iZiϵiks/mi,(11)\mathbf{Z}_{ks} = \sum_i Z_i\epsilon_{iks}/\sqrt{m_i}, \tag*{(11)}

where ZiZ_i is the Born effective charge tensor for atom ii, and k→0\mathbf{k} \to0 denotes a small but finite wave vector characteristic for the probing laser in an optical experiment.

The longitudinal charge creates a macroscopic polarization along k\mathbf{k} when mode ss is excited

pksLO=ZksLOAkskk,(12)\mathbf{p}^{\mathrm{LO}}_{ks} = Z^{\mathrm{LO}}_{ks} A_{ks} \frac{\mathbf{k}}{k}, \tag*{(12)}

where AksA_{\mathbf{k}s} is the mode amplitude. Through this polarization, a mode ss can couple to electronic excitations via Fröhlich coupling. Since the mode amplitude goes with

⟨Aks2⟩∝1/ωks2,(13)\left\langle A_{\mathbf{k}s}^{2}\right\rangle\propto1/\omega_{\mathbf{k}s}^{2}, \tag*{(13)}

we take

ZksLO/ωks(14)Z_{\mathbf{k}s}^{\mathrm{LO}}/\omega_{\mathbf{k}s} \tag*{(14)}

as an estimate for the “longitudinal mode coupling strength” or simply “LO character”. Since this coupling strength is direction-dependent via the incident wave vector k\mathbf{k}, we take k\mathbf{k} to point in [101] direction in line with the experimental setup. However, we have verified that the qualitative picture does not change if k\mathbf{k} is chosen along another direction, e.g., one of the Cartesian axes.

Computational details

For both systems, we use a unit cell with 20 atoms, space group 62 (Pnma) (Materials Project codes mp-540771 for BaZrS3_3 and mp-567681 for CsPbBr3_3 [10]). We use the temperature-dependent effective potential method (TDEP) with the stochastic sampling scheme described above (sTDEP). Details can be found in [7] and references therein.

For force sampling, 3×2×33 \times2 \times3 supercells (360 atoms) and the all-electron DFT code FHI-aims were used, with light default basis sets [11], 2×2×22 \times2 \times2 k-points, and am05 xc functional [12, 13]. To compute the Born effective charge tensor, Quantum Espresso v7.1 was used [14, 15, 16], with 75 Ry energy cutoff, 4×3×44 \times3 \times4 k-points, PBE functional [17], and Schlipf-Gygi norm-conserving pseudopotentials created with the ONCVPSP code [18, 19].

Raman Data and Analysis

BaZrS3_3 Raman Data Over a Wide Temperature Range

In Figure S3, we present the Raman spectra of BaZrS3_3 from 10–875 K, where we see signs of sample degradation in the optical microscope. Each spectrum was normalized between zero and one by reducing a constant background and dividing by the maximum intensity. The Raman spectra below 80 K were measured in the same optical system, using liquid-He as the cryogenic coolant.

Figure S3. Raman Spectra of BaZrS3_3 crystal in the range of 10–875 K. Spectra were normalized and shifted vertically for clarity.

At low temperatures, the Raman spectra are characterized by sharp isolated peaks assigned to first-order scattering from single normal-mode vibrations (one photon scattering by one vibration). As often happens, the first-order peaks slightly red shift and broaden with increasing temperature [20, 21]. The broad, higher-frequency (ω≥300 cm−1\omega\ge300\ \mathrm{cm}^{-1}) features that emerge with increasing temperature are assigned to second-order scattering due to their absence at low temperatures, their rapid increase with temperature [22], and from the calculated phonon density of states. A broad background forms below the low-frequency peaks; this we assign to disorder-enhanced high-order scattering of unknown origin [20]. We can distinguish first-order scattering peaks from the broad features at higher temperatures only by following the peaks at increasing temperatures from cryogenic conditions. Further

assignment using integrated intensity temperature dependence is presented below.

The temperature-dependent Raman data show no evidence of a phase transformation; BaZrS3_3 remains in the orthorhombic phase throughout [23]. This is consistent with reported calorimetric measurements [24] and x-ray studies [25, 26, 27] showing no phase transformation up to ≈1300\approx1300 K, at which point degradation begins.

Deconvolution

To assign and analyze the temperature-dependent trends of the Raman peaks, each spectrum was deconvolved into a product of the Bose-Einstein distribution (nω,Tn_{\omega,T}) and multiple pseudo-Voigt oscillators:

S(ω)=[nω,T+1]×∑jcj[(1−ζj)ω∣Ωj∣ΓRj2ω2ΓRj2+(ω2−Ωj2)2+ζjexp⁡(−(ω−Ωj)2ln⁡16ΓRj2)](15)S(\omega) = [n_{\omega,T} + 1] \times\sum_{j} c_j \left[ (1-\zeta_j) \frac{\omega\lvert\Omega_j\rvert\Gamma_{Rj}^{2}}{\omega^{2}\Gamma_{Rj}^{2} + (\omega^{2}-\Omega_j^{2})^{2}} + \zeta_j \exp\left(-\frac{(\omega-\Omega_j)^{2}\ln16}{\Gamma_{Rj}^{2}}\right) \right] \tag*{(15)}

S(ω)S(\omega) and ω\omega are the Raman intensity and Raman shift, and Ωj\Omega_j, ΓRj\Gamma_{Rj} and cjc_j are the jj’th mode’s vibrational frequency, FWHM, and intensity, respectively. The Bose-Einstein distribution is given by nω,T=(eℏω/kBT−1)−1n_{\omega,T} = (e^{\hbar\omega/k_{\mathrm{B}}T} - 1)^{-1}, where ℏ\hbar, kBk_{\mathrm{B}} and TT are the reduced Planck’s constant, the Boltzmann constant and temperature. The parameter 0≤ζj≤10 \le\zeta_j \le1 is the Gaussian fraction of the pseudo-Voigt line shape, used to capture broadening unrelated to vibrational lifetime, such as higher-order scattering. Overall, ζj≈0\zeta_j \approx0 was used for the sharp first-order peaks.

Figure S4 shows the deconvolution results for the 80 K spectra of BaZrS3_3 (left) and CsPbBr3_3 (right). The blue and green traces are the measured and fitted spectra, while the red traces are individual peaks. The intensity of a Raman peak depends on temperature through the Bose-Einstein distribution according to the scattering order [22]. As the temperature is lowered, first-order peaks are more prominent than higher orders. Therefore, we identify the first-order scattering peaks by their presence in the spectra at 10 K (black traces), which contain primarily first-order peaks. The inset shows the deconvolution of the high-frequency modes of BaZrS3_3, which we assign as second-order scattering. This assignment is corroborated by the calculated vibrational density of states, which shows a high density of optical modes up to 250 cm−1^{-1} and a few optical modes at 300 cm−1^{-1} and 400 cm−1^{-1}, with a large gap between them. Assuming low Raman activity of the 300 cm−1^{-1} and 400 cm−1^{-1} branches, our assignment nicely matches the calculated vibrational density of states.

Figure S4. Raman deconvolution. pseudo-Voigt deconvolution results for the Raman spectra of BaZrS₃ (left) and CsPbBr₃ (right) at 80 K. The blue and green traces are the data and fit, respectively. The red traces are individual peaks. Black traces are the spectra at 10 K that were used to identify the first-order peaks at higher temperatures. Inset - deconvolution of the high-frequency modes of BaZrS₃.

In the case of CsPbBr₃, the sharp peaks at 80 K broaden significantly at higher temperatures, making it possible to determine their scattering order only at low temperatures (see Figure 4 [1]).

Assignment of Scattering Order

To substantiate our assignment of the modes of BaZrS₃, we compare the temperature dependence of the integrated intensity to the expected dependence of first- and second-order scattering [20]. The intensity of first-order Raman scattering is proportional to the product of the Bose-Einstein distribution and the mode spectral function (Jj\mathcal{J}_j) [28]. Therefore, for Stokes scattering, it follows that [29, 30]

Sj(ω)=Ξ∫0∞e−iωt⟨Uj∗,Uj⟩ dt=Ξ[nω,T+1]Jj(Ωj,ΓRj),(16)S_j(\omega) = \Xi\int_{0}^{\infty} e^{-i\omega t} \langle U_j^*, U_j \rangle\,\mathrm{d}t = \Xi[n_{\omega,T} + 1] \mathcal{J}_j(\Omega_j,\Gamma_{Rj}), \tag*{(16)}

where Ξ\Xi is the proportionality factor that contains the Raman tensor and properties of the material affecting the scattering, and Jj\mathcal{J}_j is defined as the temperature-independent Fourier transform of the pair displacement correlation function ⟨Uj∗,Uj⟩\langle U_j^*, U_j\rangle such that

∫0∞Jj dω=1.(17)\int_{0}^{\infty} \mathcal{J}_j\,\mathrm{d}\omega= 1 . \tag*{(17)}

Therefore, we expect the integrated intensity of any pseudo-Voigt peak originating from first-order scattering to remain constant at all temperatures. The intensity of second-order Raman scattering (one photon scattered by two vibrations of opposite momentum) is more complicated to handle because, at each frequency, there are contributions from many combinations of modes. Therefore, we can not describe the expected dependence of a specific mode and must calculate the dependence of an extensive spectral range. To do that, we model second-order scattering as the convolution of mode spectra functions [20],

S(ω)=Ξ∫0∞e−iωt∑j,j′⟨Uj∗,Uj′∗,Uj,Uj′⟩ dt=Ξ×∑j,j′[nω,T+1]Jj(Ωj,ΓRj)∗[nω,T+1]Jj′(Ωj′,ΓRj′).(18)\begin{aligned} S(\omega) &= \Xi\int_{0}^{\infty} e^{-i\omega t}\sum_{j,j'} \langle U_j^*, U_{j'}^*, U_j, U_{j'}\rangle\,\mathrm{d}t \\ &= \Xi\times\sum_{j,j'} [n_{\omega,T}+1]\mathcal{J}_j(\Omega_j,\Gamma_{Rj}) * [n_{\omega,T}+1]\mathcal{J}_{j'}(\Omega_{j'},\Gamma_{Rj'}) . \tag*{(18)} \end{aligned}

following the same notation and conditions from Eq. (16) and Eq. (17). The set of Jj\mathcal{J}_j that reproduces the experimental line shape is not unique; However, if we find one solution that can reproduce the line shape and integrated intensity, we prove that second-order scattering dominates in the spectral range in question. To that end, we estimate that combination modes (j≠j′j \ne j') have negligible contribution compared to overtones (j=j′j=j'), and simplify Eq. (18) to a single sum,

S(ω)=Ξ×∑j[nω,T+1]Jj(Ωj,ΓRj)∗[nω,T+1]Jj(Ωj,ΓRj).(19)S(\omega) = \Xi\times\sum_j [n_{\omega,T}+1]\mathcal{J}_j(\Omega_j,\Gamma_{Rj}) * [n_{\omega,T}+1]\mathcal{J}_j(\Omega_j,\Gamma_{Rj}) . \tag*{(19)}

The simplified version allows us to solve each Jj\mathcal{J}_j separately and then sum the contributions. As mentioned, we assign all the high-frequency peaks (ω≥300 cm−1\omega\ge300\ \mathrm{cm}^{-1}) to second-order Raman scattering. Therefore, we tried to find a set of Jj\mathcal{J}_j to reproduce the line shape of the high-frequency spectral range and compare the temperature dependence of the integrated intensities, as follows:

  1. Draw the high-frequency pseudo-Voigt peak using the fitted parameters at T0T_0.

  1. Find a Jj\mathcal{J}_j such that SpV(ω)=[nω,T0+1]Jj(Ωj,ΓRj)∗[nω,T0+1]Jj(Ωj,ΓRj)[nω,T0+1]S_{pV}(\omega) = \frac{[n_{\omega,T_0}+1]\mathcal{J}_j(\Omega_j,\Gamma_{Rj}) * [n_{\omega,T_0}+1]\mathcal{J}_j(\Omega_j,\Gamma_{Rj})}{[n_{\omega,T_0}+1]} and ∫0∞Jj dω=1\int_{0}^{\infty}\mathcal{J}_j\,\mathrm{d}\omega= 1.

  1. Integrate the fitted pseudo-Voigt peaks and the sum of reproduced line shapes separately.

  1. Assuming the set of Jj\mathcal{J}_j is temperature-independent, recalculate the expected line shape at T≠T0T \ne T_0 for the given set through the Bose-Einstein distribution.

  1. Compare the temperature-dependent integrated intensity of the fitted peaks and the recalculated line shapes.

Figure S5 shows the relative integrated intensities of the low-frequency sharp peaks (a) assigned as first-order scattering, the high-frequency peaks (b) assigned as second-order scattering, and the low-frequency broad background peaks (ΓRj>10 cm−1\Gamma_{Rj} > 10\ \mathrm{cm}^{-1}) compared to the expected trends for first- and second-order scattering (c). Dashed lines are to guide the eyes, and grey-dotted lines are expectation curves.

Figure S5. pseuo-Voigt relative integrated intensities compared to expected trends. (a) low frequency (ω<300 cm−1\omega< 300\ \mathrm{cm}^{-1}) sharp peak compared to expectation for first-order scattering. (b) - high-frequency (ω≥300 cm−1\omega\ge300\ \mathrm{cm}^{-1}) peaks compared to expectation for second-order scattering. (c) - low-frequency broad features (ΓRj>10 cm−1\Gamma_{Rj} > 10\ \mathrm{cm}^{-1}) compared to expectation for first- and second-order scattering. Colored markers represent the different fitted peaks with a dashed line to guide the eye, and the dotted gray lines are expectation curves.

The integrated intensity of the low-frequency sharp modes varies by ≈50%\approx50\% within the temperature range, which is within our allowed error range, considering the complexity of a real crystal and the experiment. Furthermore, according to our analysis of second-order Raman scattering, there is an excellent correlation between the integrated

intensity of the high-frequency peaks and the expectation. The expected curve was drawn using fitting parameters from the 80 K fit. Nevertheless, using the fitting parameters of other temperatures produced similar results (data not shown). The case of the broad low-frequency features presented in the right panel is more complicated - their integrated intensity remains constant for T≤200T \le200 K, then increases dramatically with temperature, almost following a second-order scattering trend. The dramatic change correlates well with a reported increase in dielectric response at kHz frequencies [31]. Therefore, we faithfully assign the sharp ω<300\omega< 300 cm−1^{-1} modes and the ω≥300\omega\ge300 cm−1^{-1} peaks to first- and second-order scattering, respectively. Following previous work [20], we assign these features to high-order scattering due to symmetry-breaking.

Figure S6 shows the Ωj\Omega_j (top) and ΓRj\Gamma_{Rj} (bottom) of the prominent first-order peaks of BaZrS3_3 (left) and CsPbBr3_3 (right) as a function of temperature, resulting from the pseudo-Voigt fit. Figure S6 shows the results of the linear

regression of Ωj(T)\Omega_j(T) (left) and ΓRj(T)\Gamma_{\mathrm{R}j}(T) (right) of BaZrS3_3 (diamonds) and CsPbBr3_3 (circles) as a function of the peak index jj. The R2^2 value and slope are presented in the top and bottom panels, respectively.

Table S5 shows the average R2^2 values and slopes of the linear regression of Ωj(T)\Omega_j(T) and ΓRj(T)\Gamma_{\mathrm{R}j}(T) of the peaks of BaZrS3_3 and CsPbBr3_3 in the range of 80–280 K. The average R2^2 values and their standard deviation (STD) indicate that the frequencies and linewidths of BaZrS3_3 are more linear than those of CsPbBr3_3, showing average values closer to 1 and lower STD. Moreover, the average slope values indicate that the peaks of CsPbBr3_3 shift and broaden more than the peaks of BaZrS3_3. The drastic change in the structural dynamics and the anharmonicity of CsPbBr3_3 compared to BaZrS3_3 are indicated by the larger STDs.

MaterialPropertyParameterMean ±\pm STD
BaZrS3_3Ωj(T)\Omega_j(T)R2^20.98 ±\pm 0.01
BaZrS3_3Ωj(T)\Omega_j(T)Slope [cm K]−1^{-1}0.020 ±\pm 0.008
BaZrS3_3ΓRj(T)\Gamma_{\mathrm{R}j}(T)R2^20.9 ±\pm 0.1
BaZrS3_3ΓRj(T)\Gamma_{\mathrm{R}j}(T)Slope [cm K]−1^{-1}0.015 ±\pm 0.006
CsPbBr3_3Ωj(T)\Omega_j(T)R2^20.8 ±\pm 0.3
CsPbBr3_3Ωj(T)\Omega_j(T)Slope [cm K]−1^{-1}0.03 ±\pm 0.02
CsPbBr3_3ΓRj(T)\Gamma_{\mathrm{R}j}(T)R2^20.83 ±\pm 0.09
CsPbBr3_3ΓRj(T)\Gamma_{\mathrm{R}j}(T)Slope [cm K]−1^{-1}0.05 ±\pm 0.03

Table S5. Mean values of the linear regression presented in Fig. S6

References

References

  1. [1]See the main text of this supplemental material.
  2. [2]J. W. Bennett, I. Grinberg, and A. M. Rappe, Effect of substituting of S for O: The sulfide perovskite Bazrs 3 investigated with density functional theory, Physical Review B 79, 235115 (2009).DOI
  3. [3]G. Ermolaev, A. P. Pushkarev, A. Zhizhhchenko, A. A. Kuchmizhak, I. Iorsh, I. Kruglov, A. Mazitov, A. Ishteev, K. Konstantinova, D. Saranin, A. Slavich, D. Stosic, E. S. Zhukova, G. Tselikov, A. Di Carlo, A. Arsenin, K. S. Novoselov, S. V. Makarov, and V. S. Volkov, Giant and Tunable Excitonic Optical Anisotropy in Single-Crystal Halide Perovskites, Nano Letters 23, 2570 (2023).DOI
  4. [4]A. v. Roekeghen, J. Carrete, and N. Mingo, Quantum Self-Consistent Ab-Initio Lattice Dynamics, Computer Physics Communications 263, 107945 (2021).
  5. [5]R. Bianco, I. Errea, L. Paulatto, M. Calandra, and F. Mauri, Second-order structural phase transitions, free energy curvature, and temperature-dependent anharmonic phonons in the self-consistent harmonic approximation: Theory and stochastic implementation, Physical Review B 96, 014111 (2017).DOI
  6. [6]N. Benshalom, G. Reuveni, R. Korobko, O. Yaffe, and O. Hellman, Dielectric response of rock-salt crystals at finite temperatures from first principles, Physical Review Materials 6, 033607 (2022).DOI
  7. [7]F. Knoop, N. Shulumba, A. Castellano, J. P. A. Batista, R. Farris, M. J. Verstraete, M. Heine, D. Broido, D. S. Kim, J. Klarbring, I. A. Abrikosov, S. I. Simak, and O. Hellman, TDEP: Temperature Dependent Effective Potentials, Journal of Open Source Software 9, 6150 (2024).DOI
  8. [8]N. Shulumba, O. Hellman, and A. J. Minnich, Intrinsic localized mode and low thermal conductivity of PbSe, Physical Review B 95, 014302 (2017).DOI
  9. [9]O. Hellman and I. A. Abrikosov, Temperature-dependent effective third-order interatomic force constants from first principles, Physical Review B 88, 144301 (2013).DOI
  10. [10]A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, and K. A. Persson, Commentary: The Materials Project: A materials genome approach to accelerating materials innovation, APL Materials 1, 011002 (2013).DOI
  11. [11]V. Blum, R. Gehrke, F. Hanke, P. Havu, V. Havu, X. Ren, K. Reuter, and M. Scheffler, Ab initio molecular simulations with numeric atom-centered orbitals, Computer Physics Communications 180, 2175 (2009).DOI
  12. [12]R. Armiento and A. E. Mattsson, Functional designed to include surface effects in self-consistent density functional theory, Physical Review B 72, 085108 (2005).DOI
  13. [13]A. Mattsson, R. Armiento, J. Paier, G. Kresse, J. M. Wills, and T. R. Mattsson, The AM05 density functional applied to solids, The Journal of Chemical Physics 128, 084714 (2008).DOI
  14. [14]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. D. Corso, S. d. Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials, Journal of Physics: Condensed Matter 21, 395502 (2009).DOI
  15. [15]P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. B. Nardelli, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, M. Cococcioni, N. Colonna, I. Carnimeo, A. D. Corso, S. d. Gironcoli, P. Delugas, R. A. D. Jr, A. Ferretti, A. Floris, G. F. Frugallo, R. Gebauer, U. Gerstmann, F. Giustino, T. Gorni, J. Jia, M. Kawamura, H.-Y. Ko, A. Kokalj, E. Kuç̌ıkbenli, M. Lazzeri, M. Marsili, N. Marzari, F. Mauri, N. L. Nguyen, H.-V. Nguyen, A. Otero-de-la Roza, L. Paulatto, S. Poncé, D. Rocca, R. Sabatini, B. Santra, M. Schlipf, A. P. Seitsonen, A. Smogunov, I. Timrov, T. Thonhauser, P. Umari, N. Vast, X. Wu, and S. Baroni, Advanced capabilities for materials modelling with Quantum ESPRESSO, Journal of Physics: Condensed Matter 29, 465901 (2017).
  16. [16]P. Giannozzi, O. Basseggio, P. Bonfà, D. Brunato, R. Car, I. Carnimeo, C. Cavazzoni, S. d. Gironcoli, P. Delugas, F. F. Ruffino, A. Ferretti, N. Marzari, I. Timrov, A. Urru, and S. Baroni, Quantum ESPRESSO toward the exascale, The Journal of Chemical Physics 152, 154105 (2020).DOI
  17. [17]J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Physical Review Letters 77, 3865 (1996), pBE.pubmed.ncbi.nlm.nih.gov/10062328
  18. [18]D. R. Hamann, Optimized norm-conserving Vanderbilt pseudopotentials, Physical Review B 88, 085117 (2013).DOI
  19. [19]M. Schlipf and F. Gygi, Optimization algorithm for the generation of ONCV pseudopotentials, Computer Physics Communications 196, 36 (2015).DOI
  20. [20]M. Menahem, N. Benshalom, M. Asher, S. Aharon, R. Korobko, O. Hellman, and O. Yaffe, Disorder origin of Raman scattering in perovskite single crystals, Physical Review Materials 7, 044602 (2023).DOI
  21. [21]M. Asher, M. Bardini, L. Catalano, R. Jouclas, G. Schweicher, J. Liu, R. Korobko, A. Cohen, Y. Geerts, D. Beljonne, and O. Yaffe, Mechanistic View on the Order-Disorder Phase Transition in Amphidynamic Crystals, The Journal of Physical Chemistry Letters 14, 1570 (2023).DOI
  22. [22]M. Cardona and G. Guntherodt, Light Scattering in Solids II - Basic Concepts and Instrumentation (Springer-Verlag, 1982) pp. 1–252.
  23. [23]N. Gross, Y.-Y. Sun, S. Perera, H. Hui, X. Wei, S. Zhang, H. Zeng, and B. A. Weinstein, Stability and Band-Gap Tuning of the Chalcogenide Perovskite BaZrS3 in Raman and Optical Investigations at High Pressures, Physical Review Applied 8, 044014 (2017).DOI
  24. [24]S. Niu, J. Milam-Guerrero, Y. Zhou, K. Ye, B. Zhao, B. C. Melot, and J. Ravichandran, Thermal stability study of transition metal perovskite sulfides, Journal of Materials Research 33, 4135 (2018).DOI
  25. [25]C. Comparatto, A. Davydova, T. Ericson, L. Riekher, M. V. Moro, T. Kubart, and J. Scragg, Chalcogenide Perovskite BaZrS3 : Thin Film Growth by Sputtering and Rapid Thermal Processing, ACS Applied Energy Materials 3, 2762 (2020).
  26. [26]J. Xu, Y. Fan, W. Tian, L. Ye, Y. Zhang, Y. Tian, Y. Han, and Z. Shi, Enhancing the optical absorption of chalcogenide perovskite BaZrS3 by optimizing the synthesis and post-processing conditions, Journal of Solid State Chemistry 307, 122872 (2022).DOI
  27. [27]Z. Yu, X. Wei, Y. Zheng, H. Hui, M. Bian, S. Dhole, J.-H. Seo, Y.-Y. Sun, Q. Jia, S. Zhang, S. Yang, and H. Zeng, Chalcogenide perovskite BaZrS3 thin-film electronic and optoelectronic devices by low temperature processing, Nano Energy 85, 105959 (2021).
  28. [28]P. C. Kwok, Green’s Function Method in Lattice Dynamics, in Solid State Phys. (Academic Press, Cambridge, MA, USA, 1968) pp. 213–303.DOI
  29. [29]A. A. Maradudin, E. W. Montroll, G. H. Weiss, and I. P. Ipatova, Theory of Lattice Dynamics in the Harmonic Approximation (Academic Press, New York, USA, 1971) pp. 1–67, 2nd ed.
  30. [30]S. A. Safran, G. Dresselhaus, and B. Lax, Theory of spin-disorder Raman scattering in magnetic semiconductors, Phys. Rev. B 16, 2749 (1977).DOI
  31. [31]S. Filippone, B. Zhao, S. Niu, N. Z. Koocher, D. Silevitch, I. Pina, J. M. Rondinelli, J. Ravichandran, and R. Jaramillo, Discovery of highly polarizable semiconductors BaZrS3 and Ba3Zr2S7, Physical Review Materials 4, 10.1103/PhysRevMaterials.4.091601 (2020).DOI

Paper details

vlnc:W_x3wqtjmt
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents