Synthesis of alkaline-earth Zintl phosphides MZn₂P₂ (M = Ca, Sr, Ba) from Sn solutions

1 Introduction

Towards the multi-terawatt scale implementation of photovoltaic power systems to society, considerable efforts have been devoted to exploring materials for photovoltaic devices (PVs) in the past four decades. While these efforts have led to the development of thin-film PVs with conversion efficiencies over 20%, such as Cu(In,Ga)Se₂ (CIGS) [1], CdTe [2], and halide perovskites [3], the use of scarce and/or toxic elements in these devices is considered as a potential risk for large-scale utilization. Various earth-abundant light-absorbing materials have been studied to overcome this issue, but the conversion efficiencies of emerging PVs based on them are still 13% at most [4,5,6,7]. One of the bottlenecks for the improvement of earth-abundant thin-film PVs would be the limited choices of partner materials to form a p-n junction with light-absorbing materials (so-called buffer layer or n-type emitter). Emerging PVs are usually constructed by referring to the existing devices; thus, II–VI compounds and TiO₂ are the commonly-used partner materials as we can notice from the recent review article [4]. Probably, for this reason, most of the emerging PVs with relatively high efficiency are based on light-absorbing materials chemically similar to the existing ones (e.g., Cu₂ZnSnS₄ as an alternative to CIGS). In order to practicalize diverse emerging PVs, it is therefore important to explore partner materials.

Zn₃P₂-based PV is an attractive example of emerging PVs because the device with the highest conversion efficiency of 6% was composed of the “Schottky” junction of Zn₃P₂ and Mg, which is different from those of any other PVs [8]. The detailed nature of this device had been controversial for a long period [9,10,11], but we recently revealed that this is a heterojunction of semiconductors between Zn₃P₂ and Mg(Mg _x Zn_1−x )₂P₂, which is formed through the reaction at the Mg/Zn₃P₂ interface [12]. The lattice mismatch between Mg(Mg _x Zn_1−x )₂P₂ and Zn₃P₂ is 0.5% at most, and thus, is favorable for photovoltaic applications. On the other hand, Zn₃P₂ has a face-centered cubic (fcc) sublattice of phosphorus, which is often observed in pnictide semiconductors such as InP and ZnSnP₂ [13,14]. Therefore, Mg(Mg _x Zn_1−x )₂P₂ would be intriguing as a partner material in pnictide-based PVs. Inspired by this discovery, we focused on a related series of compounds, MZn₂P₂ (M: IIA elements such as Ca, Sr, and Ba), in this study.

In order to investigate the properties of MZn₂P₂, we need to deal with the difficulties in their synthesis, i.e., high reactivity and high vapor pressure of each element. As a solution to these problems, we recently proposed the use of Sn as a solvent and successfully investigated the phase equilibria of the Mg-P-Zn system [15]. The solubility of each constituent element in the Sn melt is remarkable [16], which would lead to the reduction of the activities of the elements. Considering also that Sn has a low melting point of 232°C, it can be a suitable solvent for the synthesis of MZn₂P₂. In the present study, we accordingly attempted to prepare MZn₂P₂ bulk crystals using Sn as a solvent and investigate their optoelectronic properties. We then discuss the applicability of MZn₂P₂ as partner materials with pnictide absorbers from the viewpoints of lattice and band parameters.

2 Materials and methods

The starting materials for the growth of MZn₂P₂ crystals were Ba (99% up, chunk), Ca (99%, grain), P (99.9999%, flake), Sn (99.99%, grain), Sr (99%, chunk), and Zn (99.99%, grain). All reagents were purchased from Kojundo Chemical Laboratory Co. Ltd., Japan. Prior to weighing, Sn and Zn were chemically etched in 1/10 diluted hydrochloric acid for 5 min to remove surface oxide layers. They were then sequentially washed with ultrapure water and 2-propanol in an ultrasonic bath for 5 min for each step. Ba and Sr were washed with hexane just before weighing and introducing to the vacuum encapsulation system because they were stored in mineral oil to prevent oxidation. The other materials were used without any preprocessing. The starting materials with the compositions listed in Table 1 were loaded in a carbon (Sankyou Carbon Co., Ltd, Japan) or alumina (99.6%, SSA-S grade, Nikkato Corporation, Japan) crucible and encapsulated in a carbon-coated quartz glass ampule under the pressure of 10⁻² Pa. Here, a chunk of B₂O₃ was put on the top of the carbon crucible to suppress evaporation of each element by referring to the liquid-encapsulated Czochralski process for the growth of III-V bulk single crystals [17,18,19]. The samples were denominated as CZP-Sn (M = Ca), SZP-Sn (M = Sr), and BZP-Sn (M = Ba) as shown in Table 1. The ampules were then placed in furnaces, schematically shown in Figure 1(a), for CZP-Sn or 1(c) for SZP-Sn and BZP-Sn, and annealed by the procedures described as follows. The temperature history for CZP-Sn is shown in Figure 1(b). Except for the water quenching steps, the ampules were allowed to stand at a fixed position in the furnace. On the other hand, the furnace for the annealing of SZP-Sn and BZP-Sn was raised with a speed of 5 mm·day⁻¹ after homogenization of the samples at 900°C. The temperature profile in the furnace and the initial position of the sample in the furnace are shown in Figure 1(d).

Figure 1

(a) Schematic illustration of the experimental setup and (b) the heat treatment process for the Sn-solution synthesis of CaZn₂P₂. (c) Schematic illustration of the Bridgman-type apparatus and (d) temperature profile in the furnace for the Sn-solution synthesis of SrZn₂P₂ and BaZn₂P₂.

The samples after the heat treatment were cut into plates in the direction perpendicular to the longitudinal direction by a low-speed diamond wheel saw. We here used isoparaffin oil (Lubricant Q purchased from Refine Tec Ltd., Japan) as a lubricant for cutting to suppress decomposition of alkaline-earth compounds via the reaction with water. The sample plates were then mechanically polished by water-resistant abrasive papers while pouring hexane for scanning electron microscopy (SEM) observation and energy-dispersive X-ray spectroscopy (EDS) analysis. Some of the samples were ground into a powder with an agate mortar and pestle for X-ray diffraction (XRD) measurements and optoelectronic analysis.

The compositions of the phases in the ingots were evaluated by SEM-EDS (JCM-6000 Plus equipped with MP-05030-EDK, JEOL Ltd., Japan). The crystal structures of the samples were analyzed by XRD (X’Pert Pro Alpha-1, PANalytical) with a Bragg-Brentano geometry using Cu Kα₁ incident X-ray from a Johansson-type monochromator. The optical bandgap energies of MZn₂P₂ were evaluated from the Tauc plot of the Kubelka–Munk-transformed diffuse reflectance spectra of powdered specimens [20,21]. The diffuse reflectance spectra were measured by an ultraviolet-visible (UV-Vis) spectrophotometer (UV2600, Shimadzu Corporation, Japan). The ionization potentials (IPs), namely the energy difference between the vacuum level and the valence band maximum of semiconductors, were estimated by photoelectron yield spectroscopy (PYS). The PYS system (BIP-KV201, Bunkoukeiki Co. Ltd., Japan) was calibrated using a Au film as a standard sample right before the measurements.

3 Results and discussion

Figure 2 shows the SEM backscattered electron detector-compositional (BED-C) images and the corresponding EDS mappings of the sample plates containing phosphide phases? Also, the results of EDS quantitative analysis of the regions labeled as A–D in the SEM-BED-C images are summarized in Table 2. Two or more regions with different chemical compositions were observed in each sample. The regions denoted as A possess chemical compositions close to the stoichiometry of MZn₂P₂. Here, note that the EDS analysis conducted in this study was semi-quantitative because it was difficult to prepare standard samples for alkaline-earth elements. This would be the reason for the several percent of deviations from the stoichiometry seen in this study, and it remains as future work to investigate off-stoichiometry of MZn₂P₂ phases as in the case for Mg(Mg _x Zn_1−x )₂P₂. Another region with the composition with a certain amount of phosphorus was detected in the CZP-Sn (region B), whereas MZn₂P₂ is the only phosphide region in SZP-Sn and BZP-Sn. Zn and Sn are not detected in the EDS profiles of region B in CZP-Sn. The standard deviation in the composition is much larger compared to the other results and the estimated composition is between the stoichiometries of already-known calcium phosphides such as Ca₃P₂ and CaP. We thus assume that this region would be a mixture of them. The other regions should be the solidified flux because they are mainly composed of Sn and Zn.

Figure 2

SEM BED-C images and corresponding EDS mappings of the region containing MZn₂P₂ crystals in (a) CZP-Sn, (b) SZP-Sn, and (c) BZP-Sn. The contrast observed in area A at the bottom right of (c) is caused by the inclined shape of the inclined sample surface.

We then tried to retrieve MZn₂P₂ powder from the sample plates for further analyses because single-phase plates of MZn₂P₂ were unfortunately not obtained as described above. The color of Ca₃P₂ is known to be reddish-brown or gray. This indicates that it might have a bandgap in the near infrared-visible range; therefore, it should be separated as possible from the powder of CZP-Sn to investigate the optoelectronic properties of CaZn₂P₂. Fortunately, calcium phosphides are known to be highly reactive with water and oxygen in the atmosphere, and thus, they could be removed from the plates through weathering in the atmosphere for several days. The metallic phases observed in all the samples are relatively ductile and not easily pulverized compared to MZn₂P₂. Utilizing this difference in ductility, we attempted to separate metallic phases and MZn₂P₂ by pounding the sample plates with an agate mortar and pestle. Figure 3 shows the XRD profiles of the powders extracted from CZP-Sn, SZP-Sn, and BZP-Sn through the above-described procedure. The diffraction lines are well indexed with the reference patterns of MZn₂P₂ and Sn. The CaZn₂P₂ and the SrZn₂P₂ crystals obtained in this study have the trigonal CaAl₂Si₂-type structure, which is the same for Mg(Mg _x Zn_1−x )₂P₂, and the BaZn₂P₂ crystals have the tetragonal ThCr₂Si₂-type structure. In addition to the signals from these phases, several weak diffraction lines are detected in the XRD profile of the powder from CZP–Sn. These diffractions are probably from decomposition products of calcium phosphides. The intensities of the diffraction lines from MZn₂P₂ are much stronger than those from Sn and the unidentified phase in CZP-Sn; therefore, we could obtain powders mainly composed of MZn₂P₂ from the sample plates.

Figure 3

XRD profiles of the powders extracted by pounding and grinding (a) CZP-Sn, (b) SZP-Sn, and (c) BZP-Sn. The diffraction lines denoted by a question mark in (a) could not be identified.

Subsequently, we evaluated the optoelectronic properties of MZn₂P₂ using the powders extracted for the XRD analysis. Figure 4 summarizes the results of the UV-vis diffuse reflectance analyses of the powders. The data for the powders from CZP–Sn and SZP–Sn are plotted as Tauc plots of the Kubelka–Munk-transformed profiles because they have absorption edges in the measured range of wavelength. According to the profiles, it is revealed that CaZn₂P₂ has an indirect fundamental bandgap of 1.85 eV and a direct bandgap of 2.05 eV, and SrZn₂P₂ has an indirect fundamental bandgap of 1.70 eV and a direct bandgap of 1.89 eV. In contrast, the diffuse reflectance profile from the BZP-Sn powder does not show any absorption edges. We thus assume that BaZn₂P₂ has a metallic or a semimetallic band structure, or a bandgap narrower than 0.9 eV. The above evaluations qualitatively correspond to the calculated band structures in the Materials Project database [22], where CaZn₂P₂ and SrZn₂P₂ are semiconductors with indirect fundamental bandgaps and slightly wider direct gaps (ID: mp-9569 and mp-8276), and BaZn₂P₂ is a semimetal (ID: mp-7426). The characteristics of the optical inter-band transitions in CaZn₂P₂ and SrZn₂P₂ also agree with the recent computational studies by Murtaza et al. based on the Perdew–Burke–Ernzerhoff generalized gradient approximation (PBE-GGA) with the modified Becke–Johnson (mBJ) potential [23,24], whereas the calculated bandgap of SrZn₂P₂ was 0.1 eV larger than that of CaZn₂P₂ in contrast to the results in this study. Besides, they reported that the conduction bands of CaZn₂P₂ and SrZn₂P₂ are mainly composed of the Ca- or the Sr-d orbital and P-p orbital. However, the GGA-mBJ approach is known to give unreliable d-state binding energy [25,26]. This might be the reason for the difference in the bandgap values. In the case of Zintl arsenides (CaZn₂As₂ and SrZn₂As₂), the density functional theory calculations using a more accurate Heyd–Scuseria–Ernzerhof (HSE06) hybrid functionals were conducted by Xiao et al. [27]. They reported that the bandgap of CaZn₂As₂ is larger than that of SrZn₂P₂, which coincides with the trend in the phosphides in this study. Also, our result for BaZn₂P₂ is consistent with the recent tight-binding linear Muffin-Tin orbital (TB-LMTO) calculation by Balvanz et al. [28]. Here, we should again note that the crystal structure of BaZn₂P₂ samples in this study is the well-known ThCr₂Si₂-type structure, whereas they reported that there is another polymorph for BaZn₂P₂, i.e., the α-BaCu₂S₂-type structure. They suggested that the stable polymorph in the temperature range below 850°C should be the α-BaCu₂S₂-type one, but we could only obtain BaZn₂P₂ crystals with the ThCr₂Si₂-type structure with the experimental protocol in this study. We thus suggest further studies are required to draw conclusions on the stability of the polymorphs and phase transformation in BaZn₂P₂.

Figure 4

Tauc plots of the Kubelka–Munk-transformed diffuse reflectance of the powders extracted from (a) CZP-Sn (CaZn₂P₂) and (b) SZP-Sn (SrZn₂P₂). (c) The diffuse reflectance spectrum of the powder extracted from BZP-Sn (BaZn₂P₂).

Figure 5 shows the PYS profiles of the powders extracted from CZP-Sn, SZP-Sn, and BZP-Sn. According to Fowler and Kane [29,30], photoelectron yield (Y) is proportional to (hν – IP) ⁿ near the threshold, where hν is the photon energy and n is the parameter depending on the production and scattering process of photoelectrons. n takes a value in the range from 1 to 3 in the case of semiconductors and it is two for metals. All the PYS spectra in this study show linear dependence on hν (n = 1) as we can see in Figure 5. This indicates that BaZn₂P₂ is not a metallic conductor and that the thresholds observed in the spectra correspond to the IPs of MZn₂P₂. Hence, the IPs are evaluated to be 5.0 eV for CaZn₂P₂ and SrZn₂P₂, and 4.9 eV for BaZn₂P₂. Here, we should note that the work function of Sn is 4.42 eV [31], and thus, the PYS spectra are not affected by the secondary phase in powders. These values are close to the IPs of Zn₃P₂ (5.0 eV) calculated from the bandgap (1.5 eV) and the electron affinity (3.5 eV) reported by Nelson et al. [32] and that of ZnSnP₂ (5.2 eV) evaluated by our group [33]. Accordingly, the VBM of MZn₂P₂, Zn₃P₂, and ZnSnP₂ are aligned within 0.2 eV.

Figure 5

PYS profiles of the powders extracted from (a) CaZn₂P₂, (b) SrZn₂P₂, and (c) BaZn₂P₂.

Finally, we discuss the lattice and band matching between MZn₂P₂ and pnictide absorbers such as Zn₃P₂ and ZnSnP₂. Figure 6 shows the relationship between nearest-neighbor pnictogen–pnictogen distance, d _pnictogen, in pnictogen sublattice in pnictide semiconductors and fundamental bandgap of various pnictide semiconductors. By and large, a downward trend is observed in Figure 6 as is also the case for the bandgap vs lattice constant relationship in semiconductor solid solutions. Considering the bandgap energies and the IPs discussed above, the conduction band minimum of CaZn₂P₂ is 0–0.2 eV shallower relative to those of Zn₃P₂ and ZnSnP₂. According to the device simulation for the sulfide/CIGS type solar cells by Minemoto et al. [34,35] and Liu and Sites [36], such an offset in the conduction band minimum is in the optimum range for PV applications. On the other hand, pnictogen sublattice in pnictide semiconductors is typically face-centered cubic (fcc, e.g., in InP and GaAs) or hexagonal closed packed (hcp, e.g., GaN) structure; that in MZn₂P₂ is also hcp except for BaZn₂P₂. d _pnictgen is, therefore, a measure to consider epitaxial lattice mismatch in the orientation relationship where the closest packed planes of the pnictogen sublattice ((111) in fcc and (0001) in hcp) are parallel to each other. Such an orientation relationship was also observed in the Mg(Mg _x Zn_1−x )₂P₂/Zn₃P₂ interface in our previous study [12]. As shown in Figure 6, CaZn₂P₂ has lattice mismatches of about 1% at most with Zn₃P₂ and ZnSnP₂. Therefore, assuming that the discussion for chalcogenide PVs is directly applicable to other types of devices, CaZn₂P₂ has a desirable lattice constant and band positions as a partner material in PVs based on Zn₃P₂ and ZnSnP₂.

Figure 6

Relationship between the nearest-neighbor pnictogen–pnictogen distance and the fundamental bandgap of various pnictide semiconductors.

4 Conclusion

We have proven that bulk crystals of MZn₂P₂ (M = Ca, Sr, Ba) can be synthesized from Sn-based solutions, whereas the crystals obtained in this study were embedded in solidified Sn flux. The MZn₂P₂ crystals could be separated from Sn flux sufficiently for optoelectronic characterization by pounding the samples with an agate mortar and pestle. Through the diffuse reflectance and PYS analyses, we revealed that CaZn₂P₂ and SrZn₂P₂ are semiconductors with indirect bandgaps of 1.85 and 1.70 eV, respectively, and an ionization potential of around 5.0 eV. BaZn₂P₂ might be a semiconductor with a narrow bandgap less than 0.9 eV or a semimetal, which corresponds well with the recent calculations. Especially, we found that CaZn₂P₂ would be an appropriate partner material in Zn₃P₂- and ZnSnP₂-based PVs from the viewpoints of lattice and band matching. As demonstrated in this study, Sn is an adequate solvent to explore and synthesize compounds including alkaline-earth elements and pnictogens.