Ag9GaSe6: high-pressure-induced Ag migration causes thermoelectric performance irreproducibility and elimination of such instability

The argyrodite Ag9GaSe6 is a newly recognized high-efficiency thermoelectric material with an ultralow thermal conductivity; however, liquid-like Ag atoms are believed to cause poor stability and performance irreproducibility, which was evidenced even after the 1st measurement run. Herein, we demonstrate the abovementioned instability and irreproducibility are caused by standard thermoelectric sample hot-pressing procedure, during which high pressure promotes the 3-fold-coordinated Ag atoms migrate to 4-fold-coordinated sites with higher-chemical potentials. Such instability can be eliminated by a simple annealing treatment, driving the metastable Ag atoms back to the original sites with lower-chemical potentials as revealed by the valence band X-ray photoelectron chemical potential spectra and single crystal X-ray diffraction data. Furthermore, the hot-pressed-annealed samples exhibit great stability and TE property repeatability. Such a stability and repeatability has never been reported before. This discovery will give liquid-like materials great application potential.

1. Authors have claimed that they have collected single crystal diffraction data of as-prepared, hotpressed-only, and hot-pressed-annealed samples. But how the authors have obtained single crystal just after hot pressing the powder sample or after the annealing of the hot-pressed sample? Also, authors have not given any single crystal X-ray diffraction pattern.
2. Authors have also mentioned that 15% of the Ag1 atoms shift from the 3-fold coordination to the 4fold coordination environment whereas occupancy variation on the Ag2 site during different condensation treatments changes is less. Then it is expected that thermal displacement parameter will be high for Ag1 than Ag2. But the refinement data of both as-synthesized single crystal and hot-pressedonly crystal (in the supplementary information) indicates the reverse fact (for hot-pressed-only crystal Ueq = 0.104 (Ag1) and 0.227 (Ag2)) 3. Also, only isotropic thermal displacement parameters are provided. Do all the atoms behave isotropically?
4. For many chalcogenide systems it is reported that prolonged annealing leads to volatilization of S or Se (ACS Appl. Mater. Interfaces 2021, 13, 45736-45743). Authors must perform the microscopy to find out the exact composition of the samples after annealing.
5. Authors must mention the density of the samples in comparison to the theoretical density.
6. It is recommended to use the measured Cp for the estimation of thermal conductivity and zT, which will reduce the error in zT estimation. Also, authors have compared their data with ref. 15 where Chen et al. have used the measured Cp (higher than Dulong-Petit Cp) for thermal conductivity and zT estimation of Ag9GaSe6. During comparison authors must maintain similar experimental parameters.

Reviewer 3 comments:
This paper presents some results about the improved stability and reproducibility of argyrodite Ag9GaSe6 via annealing treatment. The authors claim that high pressure promotes the migration of Ag-sublattice to sites with high chemical potentials, and annealing would drive the metastable Ag atoms to migrate back to the original energetically more stable crystallographic sites. The finding is interesting, and I thus recommend major revision based on the following reasons.
1. This paper only shows the transport properties until 800 K. However, the important references cited in this paper studied the thermoelectric properties up to 823 K. It is necessary to do the comparison only if you give the thermoelectric data and reproducibility until 850 K.
2. How about the stability related to the current density? The stability depends on the applied electronic field should be provided, especially the critical voltage.
3. The theoretical calculation should be carried out to clearly show the migration paths of Ag ions. 4. TEM are strongly recommended to observe the possible precipitation of Ag nanoparticle.
2 / 12 2. The authors claimed the Ag migration was caused by the high pressure, but the evidences are not solid enough. I suggest pressing the samples at different pressures.
Response: Thank you for your comments. In the revision, we have supplied 16 sets of single crystal diffraction data on samples that are obtained under various pressures: 0,30,40,50,52.5,55, in the Supporting Information) Note that the 90 MPa is the accurate pressure threshold of our hot-pressing furnace. Altogether, the 43 sets of data demonstrate a trend in the Ag9GaSe6 unit cell, in which when the pressure is below 60MPa, all Ag atoms remain relative constant occupancies that are irrelevant to the pressure indicating the Ag migration is not initiated yet. (Fig. S8a) As the pressure increases beyond 60 MPa (note that 60MPa is the required pressure to condense the sample during the standard hot-press process), the Ag atoms tend to migrate away from the 3-fold coordinated sites to the 4-fold coordinated sites. As Fig. S8 shown, the occupancies of Ag1 and Ag3 begin to decrease, and those of Ag4 and Ag5 start to increase, whereas that of Ag2 remains almost unchanged.
Eventually, under pressure higher than 60MPa, the population of the 3-fold-coordinated Ag atoms decreases, whereas that of the 4-fold-coordinated Ag atoms increases. (Fig. S8) What is more illustrative is that the net decrease number of the 3-fold-coordinated-Ag atoms exactly equals to the net increase of the 4-fold-coordinated-Ag, which means that the 3-fold-coordinated Ag atoms begin to migrate to the 4-fold-coordinated sites when the pressure is higher than 60MPa. In another word, under higher pressure, Ag tends to adopt a higher coordination number. Similar phenomena were observed by the Xe-F coordination number increasing from 2 to 4 under 23 GPa, then to 8 under 70 GPa, 46 (Nat. Rev. Mater. 2010, 2, 784); as well as by the Cs-F coordination number increasing from 1 to 6 under high pressure. 47 (Nat. Chem. 2013, 5, 846.) Faithfully following your suggestions, the following sentences are added on p.7, l.33: "… … To further establish the correlation between the Ag migration and the hot-press pressure, the hot-pressing procedures were carried out under different pressures (30-90 MPa), during which high quality small single crystals were obtained. (Tables S10-12) Together with the high quality small single crystals that were as-grown inside the corresponding ingots (Tables S1-9), we confirmed the pressure-dependent Ag migration routes that are revealed by the occupancy variation of the corresponding Ag atoms. (Fig. S8) When the pressure is below 60 MPa, the Ag occupancy is nearly constant, indicating the relative stationaries of all the Ag atoms; when the pressure increases beyond 60 MPa, the population of the 4-fold coordinated Ag atoms increases, which equals to the decrease amount of the 3-foldcoordinated Ag atoms. Such a variation proves straightforwardly the Ag-migration driven by the high pressure, which on the other hand illustrates that Ag atoms tend to adopt a higher-coordination environment under high pressure. Similar phenomena were observed by the Xe-F coordination number increasing from 2 to 4 under 23 GPa, then to 8 under 70 GPa, 46 as well as by the Cs-F coordination number increasing from 1 to 6 under high pressure. 47 " The new Fig. S8 and new Tables S10-12 are supplied in the Supporting Information.

Fig. S8.
Hot-press pressure-dependent a) average occupancy of each Ag atom and b) total numbers per unit cell of the 3-fold-(Ag1 + Ag2 + Ag3) and 4-fold-coordinated (Ag4 + Ag5) Ag atoms, respectively. Points are experimental data listed in Tables S10-12 and the dotted line is the fitting curve.

/ 12
3. Why atoms tend to adopt a higher-coordination environment under high pressure?

Response:
Thanks. See also the answers in R1, Q2 above. Here, we also explain this issue in a more general way as below. It's a common phenomenon that atoms tend to adopt a higher-coordination environment under high pressure. The most straightforward effect of pressure on materials is the volume decrease, which leads to a reduction of the interatomic distances and the space around atoms. (Nat. Rev. Mater. 2017, 2, 17005) Once the atoms are compressed, there is a tendency to migrate to larger space. Normally, site with higher-coordination numbers always has larger surrounding spaces than the one with lower-coordination numbers, thus, under high pressure atoms tend to adopt a higher-coordination environment. For example, some binaries (e.g., AlN, AlP, AlAs; GaN; InN, InP, InAs) and ZnQ, CdQ (Q = S, Se, Te) transform from the 4-fold coordinated zincblende or wurtzite structures to the 6-fold coordinated NiAs or NaCl structures under high pressure. 4. How did the authors grow the single crystals? I also don't understand how to get single crystals after hot pressing or hot-pressingannealing processes.
Response: Thanks a lot. The single crystals reported herein are not the large-sized single-crystals grown by a special crystal-growth method, rather, they are as-grown inside the as-synthesized ingots. When we crushed the as-synthesized ingots into pieces, a large number of tiny crystals with sizes of about tens of microns are observed. Then, the suitable single crystals were hand-picked under a microscope and used to collect the single crystal diffraction data.
Following your suggestions, the following sentences are added in the "Sample preparation" section under Methods on P. 9, l.13-20, "… …To pick up the suitable single crystals for the crystallographic structure determination, the above-mentioned assynthesized ingots were crushed into pieces, within which a large number of high-quality small single crystals are obtained. These small crystals are black and shining, with sizes about tens of microns. (Fig. S1) The suitable single crystals were then hand-picked, and used to collect the single-crystal diffraction data. (Tables S1-12) " A new Fig. S1 is supplied (copied below).

/ 12
5. Ag9GaSe6 is a n-type semiconductor, why did the authors measure the edge of the valence band maximum? What is valence band X-ray photoelectron spectroscopy (VBXPS)? Is this method really able to derive the information about valence band maximum? It seems the VBM energy change a lot after annealing, why?

Response:
The VBXPS is actually X-ray photoelectron spectroscopy, but it stresses the low binding energy region (usually 0 to tens eV) close to the Fermi level. Electrons excited by photons of such low energy are no longer the core level electrons, but valence electrons. In general, due to the valence electron interactions of atoms in the solid and the limitation of the line width of the X-ray source, the lines of the valence band spectrum will be close together, and a broad plateau will appear near the Fermi level on the spectrum (e.g., the region of 1.5-4 eV in Fig. 6b), which reflects the electronic structure information of a solid.
By the XPS measurement data, the valence band maximum (VBM) is determined by the intersection of the linearly fitting of the leading edge of the valence band and the flat energy baseline (Phys. Rev. Lett. 1980Lett. , 44, 1620. The VBM represents the difference between the valence band edge energy and the Fermi level, which has nothing to do with the semiconductor nature whether it is a p-type or an n-type. Meanwhile, such a method is also widely used for the n-type materials. Literature reports that with/without the F-doping, the VBM values of n-type TiO2 films are the same indicating that the F-doping does not affect the density of states in the valence band (J.
Mater. Chem. A 2014, 2, 3513); in the n-type InGaAs nanowires, the separation from the Fermi level to the surface VBM verifies a composition-dependent band bending that is associated the increase of the Ga-content (Nano Lett. 2016, 16, 5135−5142); even, the conduction type conversion from n-to p-type is detected by the VBM change after doping Zn in α-Fe2O3 (J. Appl. Electrochem.

2021, 51, 521).
In our work, the VBM was measured to compare the energy differences among the hot-pressed-only, hot-pressed-annealed (1day) and hot-pressed-annealed (5 days) samples. The principle of XPS is to irradiate the sample with the X-rays so that the inner electrons or the valence electrons are excited and emitted. The low binding energy indicates the corresponding electron is easily to be excited, that is, the electron itself has high energy and is relatively unstable. Herein, we take the advantage of this point to indirectly compare the stability of the valence electrons by comparing the corresponding VBM binding energy, and eventually to illustrate the sample stability.
Regarding your last concerns on "the VBM energy change a lot after annealing", the reason is that after annealing, the Ag atoms migrate back to the 3-coordinated sites that are thermodynamically more stable,(as shown in Fig. S8) The higher population of the 3-coordinated Ag atom that are more thermodynamically stable in the structure causes the higher excited energy threshold of the valence electrons.

/ 12
Reviewer #2: The manuscript entitled "Ag9GaSe6: High-pressure-induced Ag migration causes thermoelectric performance irreproducibility and elimination of such instability" by Li-Ming Wu and co-workers demonstrate the effect of annealing treatment on the reproducibility of thermoelectric data. As per the authors claim annealing treatment promotes the Ag-atoms to the more stable low chemical potential site and inhibits high-pressure induced Ag-migration after the hot-press. The work lacks providing novel fundamental ideas as it is quite well known in the literature that annealing treatment in most of the cases reduces the irreversibility in both thermal and electrical transport properties. Also, after the thermal annealing there is no overall improvement in thermoelectric performance of Ag9GaSe6. Therefore, I feel the manuscript is not suitable for such a high-impact journal, Nature Communication. Also, there are several other concern authors must take care.
Response: Thank you very much for the insightful evaluation and sincere advice of this work.
1. Authors have claimed that they have collected single crystal diffraction data of as-prepared, hot-pressed-only, and hot-pressedannealed samples. But how the authors have obtained single crystal just after hot pressing the powder sample or after the annealing of the hot-pressed sample? Also, authors have not given any single crystal X-ray diffraction pattern.
Response: Please see answers to R1, Q4. The detailed descriptions are provided in the "sample preparation" section on p.9. Thanks.
2. Authors have also mentioned that 15% of the Ag1 atoms shift from the 3-fold coordination to the 4-fold coordination environment whereas occupancy variation on the Ag2 site during different condensation treatments changes is less. Then it is expected that thermal displacement parameter will be high for Ag1 than Ag2. But the refinement data of both as-synthesized single crystal and hot-pressed-only crystal (in the supplementary information) indicates the reverse fact (for hot-pressed-only crystal Ueq = 0.104 (Ag1) and 0.227 (Ag2)) Response: Thanks for your concerns. In fact, there is no direct relationship between the atom thermal vibration displacement parameter (Ueq) and the possible site that an Ag atom prefers to migrate to under high pressure. If the thermal vibration mode is like simple harmonic vibration, no matter how large the vibration amplitude is, the atoms will always vibrate around the equilibrium position that is lower in energy, and will not migrate to the other sites having higher energy. Just like a spring, whether being stretched or compressed, it will eventually return to the origin position (Δx = 0) with the lowest elastic potential energy. However, the Ag-migration discussed herein is not a simple harmonic vibration, the Ag atoms don't return to their original sites after leaving, but stay at the destination site of the migration which shows lower potential energy. Therefore, the Ueq, representing the magnitude of the thermal vibration amplitude, is not directly related to the tendency of the Ag-migration.
3. Also, only isotropic thermal displacement parameters are provided. Do all the atoms behave iso-tropically?
Response: Yes, all atoms are refined anisotropically, as indicated in the cif files. Following your suggestions, the anisotropic displacement parameters, including U11, U22, U33, U12, U13 and U23 for all atoms are provided in Supporting Information p 25-31 as Tables S7-9, and S12.

/ 12
4. For many chalcogenide systems it is reported that prolonged annealing leads to volatilization of S or Se (ACS Appl. Mater. Interfaces 2021, 13, 45736-45743). Authors must perform the microscopy to find out the exact composition of the samples after annealing.
Response: Following your suggestions, we performed the scanning electron microscopy (SEM) on a polished Ag9GaSe6 pellet before and after the 1d-annealing at 823K (new Fig. S9 is supplied). Compared with what observed in the mentioned literature, no obvious Se-volatilization-induced pores, or metal precipitation were observed in our sample, even the scratches caused by sanding are still clearly visible, indicating the unaffected sample surface after annealing. The elemental distribution mappings on the right also proved that the composition is uniform and homogenous.
Besides, after the annealing treatment, the unit cell parameters calculated by the powder-XRD data merely change by a factor less than 0.09‰ in the a parameter (11.1436 vs. 11.1446 Å) and 0.3‰ in V (1383.80 vs. 1384.17 Å 3 ) (Fig. S5b). These experimental observations strongly support that the phase composition is constant and stable under our experimental conditions. Faithfully following your suggestions, the following sentences are added in the sample preparation as the 2 nd paragraph on p. 9, l.22-27:"……According to the scanning electron microscopy (SEM) observations on the hot-pressed-annealed samples, no obvious metal precipitation or Se violation were detected. (Fig. S9) The corresponding powder XRD and single-crystal diffraction data also indicate the consistent homogenous phase composition. … … " SEM images and elemental distribution mappings of the polished surfaces of Ag9GaSe6 ingots: our work (left, Fig. S9), and the reported work (right, the mentioned literature of ACS Appl. Mater. Interfaces 2021, 13, 45736-45743).

/ 12
5. Authors must mention the density of the samples in comparison to the theoretical density.
Response: Done as suggested. The density and the comparison to the theoretical density are provided in Supporting Information as Table S13. (copied below) Table S13. Experimental density measured by Archimedes method and relative density before and after annealing.

Sample
Before/after annealing Density * (g/cm 3 ) Relative density  The κ and ZT of Ag9GaSe6 calculated by both the measured-and Dulong-Petit-Cp are presented in Fig. 2. Due to the measured Cp is higher than the Dulong-Petit estimated value, the corresponding κ is also higher with an average difference about 10%.
Accordingly, the overall ZT using the experimental measured-Cp (Fig. 2) is averagely about 10% lower than that utilizing Dulong-Petit-Cp.

/ 12
Reviewer #3: This paper presents some results about the improved stability and reproducibility of argyrodite Ag9GaSe6 via annealing treatment.
The authors claim that high pressure promotes the migration of Ag-sublattice to sites with high chemical potentials, and annealing would drive the metastable Ag atoms to migrate back to the original energetically more stable crystallographic sites. The finding is interesting, and I thus recommend major revision based on the following reasons.
Response: Thank you for your great supporting.
1. This paper only shows the transport properties until 800 K. However, the important references cited in this paper studied the thermoelectric properties up to 823 K. It is necessary to do the comparison only if you give the thermoelectric data and reproducibility until 850 K.
Response: Thanks for your suggestion. We also tried higher temperature (above 823K), under which, the samples show signs of the Se volatilization after the measurement, we observe the thermocouple of ZEM-3 turned black indicating contamination. Thus, to protect our facility, all property measurements were executed below 800K.
The following sentence has been added in the revision, on p.9, l. 53-57 under the "Electrical and thermal transport property measurements" section: "Note that all the property measurements were executed below 800 K to protect the facilities since at the higher temperature the samples may undergo Se volatilization which will contaminate the thermocouple probe." 2. How about the stability related to the current density? The stability depends on the applied electronic field should be provided, especially the critical voltage.
Response: Thanks. Faithfully following your suggestions, we tested the current density (J) dependence of the relative electrical resistance variation (R/R0) for both the hot-pressed-only-and the hot-pressed-annealed-Ag9GaSe6, Ag8.3Cu0.7GaSe6 and Ag9GaSe5.5Te0.5 at 750K. A new section "critical voltage determination" was added in the Supporting Information. As new Fig.   S10 reveals, the R/R0 declines almost linearly as J increases for the hot-pressed-only-samples, but shows firstly a plateau and then decreases sharply for the hot-pressed-annealed Ag9GaSe6 and Ag9GaSe5.5Te0.5 samples indicating the existence of a critical voltage (Vc). (Vc is first defined in Nat. Commun. 2018Commun. , 9, 2910 The appearance of Vc after the annealing demonstrates a stability enhancement. However, even after the annealing treatment, the Cu-doped Ag8.3Cu0.7GaSe6 sample doesn't show a Vc (Fig. S10), this is probably because that 1): Ag8.3Cu0.7GaSe6 undergoes a phase transition at 330K as reported in Ref. 27 (Fig. S2); and 2): Cu atom is smaller than Ag, which is more mobile that aggravates the sample instability, so that the simple annealing treatment is not as helpful and effective as those observed in samples of Ag9GaSe6 and Ag9GaSe5.5Te0.5. Nevertheless, the reproducibility of TE properties of the Cu-doped sample is improved as shown in the main text.
Moreover, in Fig. S11, the Vc of the annealed Ag9GaSe6 (0.046V) and Ag9GaSe5.5Te0.5 (0.072V) lie between those of Cu2S (~ 0.02V) and Cu2Se (~ 0.11V), indicating the stability of our sample is comparable to that of the Cu2Q-based liquid-like TE materials (Nat. Commun. 2018, 9, 2910. The following sentences are added in the revision on p. 5, l. 30: "Moreover, the critical voltage (Vc) was measured to be 0.05 and 0.07 V on the hot-pressed-annealed-Ag9GaSe6 and the hot-pressed-annealed-Ag9GaSe5.5Te0.5, respectively, (Fig. S10) which lies between those of the state-of-the-art liquid-like TE materials, Cu2S (0.02 V) and Cu2Se (0.11 V), 42 indicating that the subsequent annealing treatment enhances the sample stability. (Fig. S11)" 11 / 12  3. The theoretical calculation should be carried out to clearly show the migration paths of Ag ions.
Response: Thanks for your suggestion. We have tried the corresponding calculations, but run into the following difficulties that hinders the theoretical simulation at present: 1) The Ag atoms are highly disordered in the cubic Ag9GaSe6, the complicated occupancies of Ag atoms lead to a great difficulty in modeling. To date, the calculations reported in literature, merely consider the case where the Ag atoms are fully ordered in the low-temperature phase (space group P213). (Joule 2017, 1, 816;Mater. Today Phys. 2018, 5, 20) 2) In terms of the calculation method, the ab initio static calculation can only simulate the influence of the pressure, but not the temperature. However, our actual experimental conditions are under the high temperature and simultaneously under the high pressure.
Therefore, to accurately calculate and simulate the experimental conditions, needs extra efforts, which will be done with cooperation with other experts in the future. Therefore, the positive effect of annealing to stabilize the Ag9GaSe6 is reconfirmed.
As you suggested, the following sentences are added in the main text on p. 4, l. 31-35: "… … Besides, the scanning electron microscopy observation reveals that the Ag-rich precipitation is observed on the hot-pressed-only-Ag9GaSe6 after the TE-property measurement (Figs. S7a vs b), manifesting the sample instability as similar as that observed by Luo. 27 … …" and on p.4, l.57-61: "… … Moreover, the SEM observation reveals no obvious Ag precipitation in the hot-pressed-annealed sample before and after the TE property measurements, indicating the phase and composition stability of the samples. (Fig. S7c vs d) Response: Firstly, our aims is to study the intrinsic mechanism of the stability of Ag9GaSe6, which has never been deeply understand before. Therefore, we purposely chose the best reported Ag9GaSe6-based materials, which are mentioned above, as cited in our manuscript ref. 15, 27, respectively, to prove that within the identical materials, we not only can realize the high ZT values as reported, but also the reproducibility and stability, which has never been reported.
For the liquid-like thermoelectric materials, the common problem is the most unwanted poorstability and thermoelectric performance irreproducibility, which were reported to be evidenced by metal particles that are as-decomposed and deposited on the sample pellet even after the 1st measurement run.
In this manuscript, we demonstrate such a poor-stability and irreproducibility are caused by the pressure-driven Ag migration to the crystallographic sites with higher-chemical potentials. Using the previous reported pristine-and doped-Ag9GaSe6 based-materials as examples, we show such instability can be eliminated by a simple annealing treatment.
Q3: Eventually Qi, X. et al. also did the annealing treatment as reported in the present manuscript and mentioned "The annealing treatment can be utilized to recover the composition change and make the sample homogeneous again." Therefore, the work presented in this manuscript does not show any improvement in thermoelectric figure of merit in comparison to the earlier reported literature as well as no new idea is presented. In terms of novelty it is not suitable for the publication in Nature Communications.

Response:
We provided 43 sets of single crystal diffraction data collected on the as-obtained Ag9GaSe6 samples to support our discussion and conclusion. Single crystal diffraction is a useful and wildly applied technique are commonly utilized in MOF, coordination chemistry, NLO material, or battery material fields, etc., however, in the thermoelectric community, which has been less used.
In the mentioned literature, (Qi, X. et al., Chem. Eng. J. 2019, 374, 494-501, also the ref. 27 in our manuscript) indeed reported the electrical properties of their "annealed sample". As shown in CoverlettFig. 2, their annealed sample still exhibits poor reproducibility indicated by the large deviations between the 1 st and 2 nd run of tests of the electrical conductivity and Seebeck coefficient, respectively. Moreover, Qi,X. et al. in ref. 27 only report that the annealing treatment makes the sample return to the original state, but doesn't address why the unrepeatability and instability is still observed, as they have written "It is found that the annealed sample is very close to the as-prepared one because of the similar the electrical conductivity and Seebeck coefficients", which indicates in their sample