Liquid state properties and solidification features of the pseudo binary BaS-La2S3

The high temperature thermodynamic properties of chalcogenides materials based on BaS remain elusive. Herein the pseudo binary BaS-La2S3 is investigated above 1573 K. The liquid properties of BaS-La2S3 are measured by means of high resolution in-situ visualization coupled with thermal arrest measurements in a thermal imaging furnace. This enables to report the first observation of such melts in a container-less setting. The melting points of BaS and La2S3 are revisited at 2454 K and 2004 K respectively. La2S3 demonstrates a high stability in its liquid state, in strike difference with the sublimation observed for BaS. BaS is however partially stabilized with the addition of few percents of La2S3. The remarkable chemical and thermal stability of La2S3-rich samples contrasts with the partial decomposition and high vapor pressure observed for BaS-rich samples. Observations and analysis of the solidified samples suggest three different solid solutions. Solid and liquid densities are investigated along the different compositions, supporting a first estimate of the volumetric thermal expansion coefficient for La2S3.

while BaS is an insulating, ionic compound. It is therefore of interest to evaluate if the electronic nature of the second compound added to BaS affects the melting behavior. This requires to investigate a pseudo-binary system with an ionic second compound, and lanthanum (+3) sesquisulfide (La 2 S 3 ) is a good candidate.
For La 2 S 3 and other ceric (+ 3) sesquisulfides, different solid phases, α , β , γ , have been reported 26 . α-La 2 S 3 , with an orthorhombic structure, is stable until 1173 K. β-La 2 S 3 has a stability domain from 1173 K to 1573 K and exhibits a tetragonal structure, similar to the oxysulfide La 10 S 14+x O 1−x . The high-temperature stable phase, γ-La 2 S 3 has a cubic Th 3 P 4 -type structure, with composition La 3−x V x S 4 (0 ≤ x ≤ 1/3, ranging between La 3 S 4 and La 2 S 3 ), where V represents La vacancies. The β and γ phases are unstable at room temperature, however γ -La 2 S 3 can be stabilized by introducing alkali ions 27 or europium 28 . Furthermore Kumta et al. found a process to generate and stabilize fine β-La 2 S 3 and γ-La 2 S 3 powder at high temperature 29 . Two other phases δ (monoclinic) and ǫ (rhombohedral) exist for some sesquisulfides but have not been reported for La 2 S 3 1 . Prior studies of La 2 S 3 intended to determine its chemical stability with respect to oxygen and sulfur as a function of temperature. In 1981, Kamarzin et al. studied lanthanum (+3) sesquisulfides from LaS to La 2 S 3 30 . In 1984, Kay et al. proposed a phase stability diagram for lanthanum versus the vapor pressure of O 2 and S 2 at 1100 K 31 . Vasilyeva described the thermodynamic of the La 2 S 3 -LaS 2 system in 2010 32 .
The melting point of La 2 S 3 has been investigated several times 10,15,30,33,34 , reported between 2133 and 2350 K. The potential reactivity with crucibles might have affected the accuracy of the results. In addition, the vaporization of La 2 S 3 before its melting has been reported at around 2000 K. Flahaut and Picon investigated the lanthanum oxide ( La 2 O 3 ) and oxysulfides ( La 2 O 2 S) and concluded their formation from sulfide was kinetically slow 35,36 . Sulfates La 2 (SO 4 ) 3 and oxysulfates (LaO) 2 SO 4 require specific conditions to be formed from sulfides, unattained herein 37,38 . La 2 S 3 has found applications in certain glasses, β-La 2 S 3 potentially exhibits properties of phosphorous material 39 . La 2 S 3 also found usage as an electrode in its α phase for its pseudo-capacitive behavior when immersed into a Na 2 SO 4 electrolyte 40,41 . From the common crystal structure between La 2 S 3 and La 3 S 4 at low temperature ( α phase), superconducting properties has been observed with a Curie point varying with the metal vacancy concentration 42 . The transformation of La 2 S 3 to La 3 S 4 leads to drastic changes in the sulfide electrical properties; Wood et al. reported an insulator behavior of La 2 S 3 , compared to La 3 S 4 which acts as a semi-metal 43 .
The combination BaS-La 2 S 3 , despite very limited prior art, supported the extension of molten sulfides electrolyte based on BaS, in particular for copper electrowinning from Cu 2 S 44 . BaS-La 2 S 3 was postulated to bring two essential properties for the electrolysis of Cu 2 S into Cu (l) and S 2(g) : the wide band-gap of BaS allows ionic conduction and the addition of La 2 S 3 decreases the melting point of the electrolyte.
Herein, the liquid state properties and solid phases found upon solidification of the pseudo binary BaS-La 2 S 3 are investigated. The use of a container-less thermal imaging furnace allows to observe unique features such as molten state stability, melt density and evaporation processes. The liquidus line and melting points of BaS and La 2 S 3 are reported and compared to the literature. Mass loss, porosity and liquid state behavior are investigated to highlight the differences between BaS-rich samples and La 2 S 3 -rich samples. Following solidification, different solid solutions are observed and compared to the literature, allowing to propose a preliminary version of the pseudo binary BaS-La 2 S 3 phase diagram. Some hypothesis regarding the oxidation processes of alkaline sulfides (BaS) and rare-earth sulfides (La 2 S 3 ) are proposed to compare the analyzed phases with the literature. Preliminary results for densities of molten and solidified samples are reported. In closing, an attempt to determine the volumetric isobaric thermal expansion coefficient of La 2 S 3 is presented.

Experimental and analysis
Samples preparation. The melting behavior is investigated along the pseudo-binary system (BaS) x - BaS and La 2 S 3 powders come from Alfa Aesar with a respective purity of 99.97 % and 99.99 % (metal basis). Each sample is prepared in a glove box with atmosphere control using argon, indicating a maximum oxygen content of 10 ppm. Each powder is weighed, mixed and milled into a mortar and pestle. The homogeneous mix is poured into a cleaned rubber balloon. The balloon is stretched and compressed to obtain a rod-like shape of homogeneous density and diameter, then a knot is tight at both ends. With both ends attached to a stainless steel support to maintain a rod-like shape, the balloon is placed into an hydrostatic press and compressed to 30000 psi (2068 bar). After compression, the rod is removed with care by cutting along the rubber balloon with a stainless-steel scalpel in a fume hood. The rod is maintained at the end of a moveable and rotating central shaft of a thermal imaging furnace by means of a tungsten or molybdenum wire. The time required to assemble the rod and set it up in the furnace involves contact with ambient atmosphere for around 1 hour. The compacted rods are slightly friable prior to melting, influencing to a certain extend the repeatability of mass loss measurements.
Thermal imaging furnace. Both BaS and La 2 S 3 are compounds with melting temperature above 1900 K, requiring special attention with respect to materials compatibility, few refractory materials being compatible. Instead herein a container-less configuration is used in a thermal imaging furnace (Crystal Systems Corp., model TX-12000-I-MIT-PC). It is equipped with four Xenon lamps of 3 kW of power each. The lamps irradiate Visual access to the samples is possible between the lamps on the 4 sides of the thermal imaging furnace. A camera (Watec, WAT-233 1/3", Extender EX2C, Computar Co., Fujinon TF4XA-1) equipped with a UV filter is fixed to the front door and used to observe samples live. A second camera (Canon Inc., EOS Rebel T5i DSLR) equipped with a zoom lens (Canon Inc., EF-S 18-135 mm) is set up on a tripod orthogonal to the front door. Melting and solidification processes are observable, either in live mode or recorded. A feed-through at the bottom of the quartz tube allows to insert a type C thermocouple (W-Re 5%, W-Re 26%) for direct temperature measurement of the melt. The thermocouple signal is converted into temperature following the Reference Tables N.I.S.T. Monograph 175-Revised to ITS-90. The reference junction temperature is corrected by 17 K to consider the actual room temperature. The thermocouple signal limits of error is 1 % of the read value in the range 273 K to 2600 K. A frequency acquisition of 3 Hz is typically used, increased to 1000 Hz when monitoring rapid cooling. Simple thermometry 45 is used to study the liquidus temperature as illustrated in Fig. 2. Other methods were not found yet compatible with the thermal imagining furnace. For a given sample, the liquidus temperature is crossed several times from visual observation, and several traces are then used to provide an average value.  www.nature.com/scientificreports/ Method for (BaS) x -La 2 S 3 . When the furnace is turned on at the minimum attainable power (0 % power), the Xenon lamps are already emitting light leading to a minimum temperature of around 1550 K to 1650 K for the samples investigated herein. This value is dependent on the specific sample materials, its geometry, and the furnace configuration. Herein, 1% of power increment leads to 30 to 40 K increase when the sample is solid. A first slow increase at around 1 % power min −1 is used to reach the melting point and not overheat the sample. When interested in the solid-phases, the power is decreased gradually to 0 % power and samples solidify up to 400 K.s −1 when the furnace is turned off. For the investigation of the liquidus, the ramp rates range from 0.2 to 4% min −1 . The melting point or liquidus temperature are measured by repeating heating and cooling processes across the phase transition until the thermal trace shows a reproducible profile. For all samples, a gaseous volume is formed inside the liquid droplet during a first melting of a sample. This gaseous volume has several origins, from the initial porosity consequence of the rod processing, to the chemical generation of a sulfuric gaseous phase composed of elemental sulfur S and including traces of metallic elements. The sulfuric gas phase has a yellowish color and is mainly observed for BaS-rich samples. Primary vacuum is applied several times during each experiments to extract this gas phase and ensure a fully liquid composition of the droplet for liquid density measurements. In the case of BaS-rich samples, the formation of a neck is observed at the solid/liquid interface when the sample stay in its liquid state for more than half an hour. The formed neck can be re-melted by slightly lowering the sample into the hot zone, for example for liquid density measurements. The droplet is considered deprived of gaseous phase when no visible effect is observable during the application of primary vacuum.
Density measurements. Pictures from camera recordings along with pixel size determination are used to calculate the equivalent volume of liquid, based on the method detailed by Wu et al. 46 . Four pictures per sample are analyzed to average the volume and minimize the possible departure from axi-symmetry. After the experiment, the solidified droplet is separated from the rest of the sample with a stainless-steel scalpel, and weighed. The distinction between the melted and non-melted part is clear on the La 2 S 3 -rich samples. The neck formation with the BaS-rich samples makes this distinction more tedious. The weighted mass of solidified droplet and the liquid volume from the pictures provide an estimate of the liquid density. Archimedes' law is used to evaluate the density of the solidified droplet (specific gravity kit from Mineralab, using air and ethanol). Oxidation. BaS-La 2 S 3 samples are unstable in air, leading to potential oxidation after exposure to atmosphere, and some oxygen content was found with EDS or WDS. As found from the literature, oxygen contamination of lanthanum sulfides is mainly due to the formation of A small quantity of powderous material from condensation on the quartz tubes is recovered and analyzed. Its difficult acquisition leads to high uncertainty in its composition. The presence or absence of metallic compound can however be discussed.

Results
Lanthanum (+3) sulfide and barium (+2) sulfide. Pure La 2 S 3 generates a small volume of visible gas during the first heating process. A mass loss of 1.5 wt% is observed. S, and La in a minor extent are detected with EDS on the condensate recovered on the quartz tube. After this first melting, La 2 S 3 demonstrates a stunning stability as a liquid phase, never reported before in the literature. Pure BaS generates a non-quantifiable volume of visible gas both as solid upon heating or as liquid. BaS decomposes before melting, producing barium and sulfur vapors that condensate on the quartz tube. No metallic barium is found inside the sample droplet, indicative of a vaporization of both Ba and S, in agreement with the composition of the condensates. As described in the literature, BaS (l) cannot be stabilized under argon at atmospheric pressure.  Fig. 3a-c. The gaseous halo is of mild intensity and observed for a short amount of time (from a minute at 1889K to twenty minutes for 2133 K). The mass loss is low, and moderately sensitive to temperature, with 3 % at 1903 K and 4 % at 2123 K. Traces of Ba and La have been collected on the quartz tube for all four experiments. The shape of the stabilized liquid droplets, as shown in Fig. 3a-c is very similar to that of pure La 2 S 3 shown in Fig. 1. Figure 4a,b shows the secondary electron images of the cross section of the solidified droplet observed in the SEM. Macroscopically, a single phase as well as a macropore of few millimeters are observed for all La 2 S 3 -rich compositions, regardless of the temperature. This macropore is not found on solidified samples if the droplet is exposed to several cycles of controlled vacuum/atmospheric melting. The small porosity observed at the bottom of the pictures is always present and attributed to the initial solid-rod porosity.
BaS-rich samples. BaS-rich samples do not exhibit favorable thermal properties and are difficult to melt.
A thermal gradient is present and it proved possible to observe a liquid surface while the thermocouple remains mechanically entrapped in the core solid. BaS-rich liquid behavior is complex and cannot be described only through the analysis of (BaS) 0.75 -(La 2 S 3 ) 0.25 samples, representing long term stability liquid behavior. The mass loss during (BaS) 0.75 -(La 2 S 3 ) 0.25 experiments reaches 10 % for the lowest temperature (2103 K) and increases to 15 % at higher temperature (2233 K). Mass loss observed with other experiments on BaS-rich samples is not representative, droplets often fell or exploded. A larger quantity of powder is deposited on the quartz tube for BaS-rich samples than La 2 S 3 -rich samples. Ba has been rejected on the quartz tube, no La has been found.  Fig. 4c-d, a porosity consisting on a sporadic repartition of pores with diameters of one-tens of a millimeter is observable. Increasing the temperature enhances the volume of visible gas generated. If no vacuum is applied during a long experiment, the porosity seems to decrease with the increase of temperature or time spent in the liquid state. However when BaS-rich samples are melted for few minutes and the neck formation is at its initial stage, the fast solidification without the application of vacuum leads to a macropore similar to the one observed for La 2 S 3 -rich samples. This macropore is not found with repeated application of vacuum.
Composition of the solidified droplets. Three solid phases, named φ 1 , φ 2 and φ 3 are observed, distinguished by their relative barium content presented in Table 1. φ 1 is characterized by a maximum ratio of barium to the other metallic elements of 9 %. In φ 2 this ratio is at minimum 90 %. φ 3 is characterized by a ratio between  (Fig. 4c,d), the neck region is in majority φ 2 with some φ 3 , whereas the molten region is mostly φ 3 solid solution with some φ 2 . Traces of φ 1 are observable mostly in the sintered region as seen in Fig. 6b.
Between 10 and 40 mol La 2 S 3 , the sample is lowered by few millimeters in the hot zone to melt the neck and primary vacuum is applied several times in order to take liquid density measurements. BaS-rich samples exhibit a majority of φ 2 , and φ 3 solid solutions. However at 30 and 40 % La 2 S 3 , φ 1 and φ 2 are largely observed while the presence of φ 3 solid solution tends to decrease. As shown in Fig. 6c the solidification of (BaS) 0.70 -(La 2 S 3 ) 0.30 sample led to a phase separation in φ 1 and φ 2 solid solutions.    Table 2 represents the liquidus temperature measured for the BaS-La 2 S 3 pseudo-binary compounds. The mean value for the melting point of BaS is 2454 K and 2004 K for La 2 S 3 . The addition of few percent of La 2 S 3 into BaS leads to an important decrease of the liquidus temperature. The minimal melting temperatures are around 25 mol% and between 86 mol% La 2 S 3 , representative of two eutectic behaviors. For La 2 S 3 -rich samples, the gas generation is visibly limited, and the final elemental composition is close to the initial one. However, for BaS-rich samples, the important gas generation and sensibility to ambient air may lead to a slight shift of the initial BaS composition. Liquidus measurements are conducted as promptly as possible to stay close to the initial composition. Table 2 reports the solid and liquid density estimates obtained for the different compositions. A slightly closed porosity is observable at the bottom of the droplets, but has not been taken into account in those estimates. La 2 S 3 has a measured solid density of 5.2 g cm −3 , lowered to 3.7 g cm −3 around 2050 K in its liquid state. The solid density of BaS post experiment is 3.3 g cm −3 , however this value is highly affected by the presence of porosity. The liquid density is not reported here due to the decomposition of BaS in temperature.

Densities (Table 2).
From 50 to 100 mol% La 2 S 3 , the solid densities are approximately independent of composition at around 5.2 g cm −3 while the liquid density is around 3.7 g cm −3 . From 10 to 40 mol% La 2 S 3 , the solid density varies from 4.3 to 4.8 g cm −3 , and is around 4.7 g cm −3 at 40 mol% La 2 S 3 . The liquid density in the range 10 to 40 mol% La 2 S 3 fluctuates around 3.5 g cm −3 . However values obtained for solid and liquid densities in the case of BaS-rich samples are impacted by the difficulty to determine the frontier between the solid and the liquid.
Thermodynamic approach of the volumetric isobaric thermal expansion coefficient α p for La 2 S 3 . Figure 7 shows La 2 S 3 sample at four different temperatures above the melting point. The temperatures considered are only reaching 60 K above the melting point. In order to maintain a quantity of matter constant at different temperatures, the origin of the considered droplet is set up at the solid/liquid interface of the highest measured temperature. This involves the consideration of a small solidified part for the lower temperature as observable in Fig. 7b-d. The results indicates a volumetric isobaric thermal coefficient of 3.4x10 −3 K −1 with a standard deviation of 3x10 −4 K −1 .

Discussion
To the authors' best knowledge,for the first time this experimental design is used in the study of liquid state properties of high temperature sulfide compounds. Herein is demonstrated the range of possibilities of the thermal imaging furnace in the study of liquid state properties of high temperature compounds hardly attainable with other apparatus.
BaS and La 2 S 3 . The melting point of BaS reported here is in good agreement with the range of values reported in the literature: 2508, 2480, 2475, 2470 and 2430 K (respectively [6][7][8][9][10] ). The small shift found in the literature can be due to the uncertainty of measurements, the high instability of BaS (l) and the possible presence of impurities. Ba (l) and S 2(l) vaporize at 2170 K and 713 K respectively; BaS (s) therefore decomposes directly into Ba (g) and S 2(g) above 2454 K. The absence of liquid Ba (l) into the remaining sample in the present study confirms this.
Regarding La 2 S 3 , the observed melting point here at 2004 K is a few hundred Kelvins lower than the values indicated in the literature 10,33,34 . Notwithstanding, Bolgar et al. reported in 1986 the change of enthalpy with temperature for La 2 S 3 47 . A change of slope is noticeable at 2000 K, matching with the melting point reported herein. Unfortunately Bolgar et al. did not address melting or boiling points. The few volume of gas generated and the marginal mass loss indicate a boiling point higher than the melting point and a low partial pressure around 2000 K. The small volume of gas generated herein could be the consequence of the unknown saturation vapor pressure or the presence of impurities. La 2 S 3 demonstrates a high stability in its liquid phase not described yet in the literature. From the conclusions presented by Flahaut et al. 12 regarding the strong similarities of ceric sulfides, it can be supposed that all ceric (+3) sulfides exhibit high stability in their liquid state. www.nature.com/scientificreports/ BaS-rich and La 2 S 3 -rich liquid behavior. The addition of BaS into La 2 S 3 , from 10 to 40 mol% leads to remarkably stable La 2 S 3 -rich melts, similar to pure La 2 S 3 . The thin frontier between the solid and liquid, and the fast homogenization of the temperature inside the droplet indicate good thermal conduction. Added to a mass loss of a few percent, it can be concluded that La 2 S 3 could be used as a liquid host to stabilize less stable compounds such as BaS.
The addition of few percent of La 2 S 3 into BaS leads to an important decrease of the liquidus. Several studies 3,6,7 make the case for a similar phenomenon for other additions to BaS. The addition of La 2 S 3 favorably stabilizes BaS with temperature, however the presence of two immiscible liquids is not observed as in the case of BaS-Cu 2 S 6 . The liquid phase has a longer lifetime than pure BaS, nevertheless the formation of a neck at the solid-liquid frontier is visible over time.
A strong thermal gradient is observed in the case of BaS-rich samples. Figure 4 demonstrates different regions as a function of the distance from the hot zone. A important sintering process is also observed at the top, likely a consequence of the thermal gradient. The neck formation over time could be directly linked to the low thermal diffusion of BaS-rich samples 4[c-d]. Gravity might also play a key role in the non-homogeneous melting. Complementary studies are required to undertake the complex liquid behavior of BaS-rich samples.

Porosity.
A macropore is observed for La 2 S 3 -rich samples and regardless of the temperature. During the first heat up, the chemical formation of a sulfuric gaseous compound lead to the formation of microbubbles in addition to the micropores. These microbubbles can reach the surface and leave the system or agglomerate in the center of the molten droplet. The surface tension being too high, the trapped bubble cannot escape and result in a macropore.
For BaS-rich samples, a macropore is also observed for short term experiments and has the same origin as for   φ 2 potentially describes BaS solid-solution with addition of La 2 S 3 within a maximum concentration of 5 mol% La 2 S 3 . Andreev and Khritohin also confirmed the presence of a solid solution of BaS containing few percents of Lu 2 S 3 , Pr 2 S 3 , Sm 2 S 3 , Tb 2 S 3 , Y 2 S 3 or Nd 2 S 4 : rare earth sesquisulfides with properties similar to La 2 S 3 8,24 . φ 3 is a third solid solution with a composition ranging from 41 to 63 mol% La 2 S 3 . The median composition of this solid solution is around 50 mol% La 2 S 3 , leading to a potential φ 3 solid solution built around BaLa 2 S 4 . However, BaLa 2 S 4 has not been reported and is not observed here. The phase diagrams of BaS-Sm 2 S 3 and SrS-Tb 2 S 3 also show a third solid solution. But these phases have a shorter stability domain than observed herein and occur where MLn 2 S 4 (M = Ba, Sr; Ln = Sm, Tb) represents a limit and not the median composition 24,25 . Figure 8 represents the liquidus temperature over the BaS-La 2 S 3 pseudo-binary composition range. φ 1 , φ 2 and φ 3 solid solutions are represented in their stability domain observed though the experiments. Traces of φ 1 in the BaS-rich samples and those of φ 3 at 90 mol% La 2 S 3 are not considered. These traces may be the consequence of a fast cooling process, leading to thermodynamic unstable phases. Two eutectic points are hypothesized at 25 mol% and 86 mol% La 2 S 3 . Solid density. The solid density of La 2 S 3 , herein 5.2 g cm −3 , is higher than the 5.0 g cm −3 reported literature value 17 . An experiment with another density measurement apparatus and a study of the crystal structure by XRD would be able to confirm or deny the result. A possible explanation to this difference could be the generation of a new crystal structure for La 2 S 3 : the solidification process from its liquid state does not crystallize in the γ , β or α -La 2 S 3 form, but into another unreported crystal structure denser than the ones reported so far. The solid density of BaS post experiment reaches 3.3 g cm −3 , a value highly affected by the porosity hence discarded.
Except for La 2 S 3 and BaS, the presence of at least two solid solutions involves an apparent density function of the proportion of each solid solution and their respective composition. The solid density along the La 2 S 3 -rich samples seems constant around 5.2 g cm −3 . The values calculated on the BaS-rich samples are impacted by the presence of open and closed porosity. In addition the thermal insulator behavior of BaS-rich samples makes the frontier between the molten part and the rest of the rod difficult to determine.

Liquid density.
No literature discussing the density of rare earth sulfide in temperature has been found. As in the case of metallic sulfide 48 , BaS-La 2 S 3 density decreases with the increase of temperature. The difference between solid and liquid densities fluctuate between 1 and 1.5 g cm −3 , which are typical values reported for metallic sulfides. The small sample size, imperfect radial homogeneity, neck formation and porosity affect the results. A study with larger samples may be beneficial to increase the precision in the liquid density estimation.
Volumetric isobaric thermal expansion coefficient α p for La 2 S 3 . Using two different samples, the isobaric thermal expansion coefficient was estimated around 3.4x10 −3 K −1 with a standard deviation of 3x10 −4 K −1 . The authors did not find any existing literature discussing this coefficient for liquid rare-earth sulfides.

Conclusion
High temperature liquid sulfide compounds were analyzed using a container-less thermal imaging furnace. Direct visualization of melting and solidification processes allowed to verify the thermal trace obtained from simple thermometry. Liquid properties such as stability, evaporation rate and liquid density were investigated. The melting point of BaS and La 2 S 3 were re-evaluated to 2454 K and 2004 K respectively. The high stability of liquid La 2 S 3 was observed while the sublimation of BaS was visually confirmed. BaS, unstable in its liquid state, partially stabilized with the addition of La 2 S 3 . Degassing and neck formation created difficulties in the study of BaS-rich samples. La 2 S 3 -rich liquids demonstrated for their part a stunning stability.