Complex electronic structure and compositing effect in high performance thermoelectric BiCuSeO

BiCuSeO oxyselenides are promising thermoelectric materials, yet further thermoelectric figure of merit ZT improvement is largely limited by the inferior electrical transport properties. The established literature on these materials shows only one power factor maximum upon carrier concentration optimization, which is typical for most thermoelectric semiconductors. Surprisingly, we found three power factor maxima when doping Bi with Pb. Based on our first-principles calculations, numerical modeling, and experimental investigation, we attribute the three maxima to the Fermi energy optimization, band convergence, and compositing effect due to in situ formed PbSe precipitates. Consequently, three ZT peaks of 0.9, 1.1, and 1.3 at 873 K are achieved for 4, 10, and 14 at.% Pb-doped samples, respectively, revealing the significance of complex electronic structure and multiple roles of Pb in BiCuSeO. The results establish an accurate band structure characterization for BiCuSeO and identify the role of band convergence and nanoprecipitation as the driving mechanism for high ZT.


Responses to the referees' comments Referee #1:
Comments to the Author The authors investigated the thermoelectric properties of BiCuSeO, and showed that Pbdoping would introduce three power factor maxima rather than typical one upon carrier concentration optimization for most TE semiconductors, which can be further attributed the Fermi energy optimization, band convergence, and compositing effects due to in-situ formed PbSe precipitates based on sufficient analysis. This work established an accurate band structure characterization for BiCuSeO and identify the role of band convergence and nanoprecipitation as the driving mechanism for achieving high ZT, further three peaks of 0.9, 1.1, and 1.3 at 873 K have been achieved for 4 at.%, 10 at.%, and 14 at.% Pb-doped samples, respectively. It is an innovative work and the paper is well written, this paper is publishable in Nature Communications after addressing follow comments。 Response: We appreciate referee's positive assessment of our manuscript as being an "innovative work and the paper is well written". 2) For the "Introduction" part and following sections, the authors have demonstrated the band structure of BiCuSeO is nonparabolic, and many conduction valleys have given rise to the electrical transport. Nevertheless, a single parabolic band model has been used to estimate the dependence of the effective mass and Lorenz number on the concentration of carriers, which is confusing. We agree that considering the effects of nonparabolicity and multiband the obtained results are convoluted, however, the adjustable parameters related to nonparabolicity, band offset, etc. are typically difficult to be accurately determined. Especially when multiple bands are being considered, the calculations will be rather complex. This, we believe, is also beyond the scope of this paper, and will be studied in our future work.
Thanks very much for your valuable comment! 3) In Fig. S3a and Fig. 4, the evidence of XRD and TEM both indicated that other than PbSe precipitates, Cu 2 Se δ nanodots are also formed. For the third maximum of PF, the authors have attributed it to the appearance of PbSe, how about the increasing amount of Cu 2 Se δ ? Will it play more important role than that of PbSe for this compositing system? A little more analysis of the comparison between two impurities that cause the third PF/ZT maximum may be helpful.
Response: This is a good point. Cu 2 Se δ nanodots, primarily generated during the rapid First of all, we didn't find any Cu 2 Se δ nanodots amount dependence on the Pb-content.
Secondly, as shown in Fig. 2d, the hole mobility of the x > 12 at.% samples with PbSe precipitates, increases noticeably as compared with those of 10 at.% and 12 at.% Pbdoped BiCuSeO. This anomalous mobility increase is the primary origin of the third PF maximum, and can't be explained by the band structure and transport theory without influence on hole transport, considering their small volume fraction. Though, the Cu 2 Se δ nanodots play an important role in phonon scattering and thus the lattice thermal conductivity reduction, which, combined with the third PF maximum resulted from PbSe precipitates, leads to the third ZT maximum. Therefore, we believe that Cu 2 Se δ nanodots play a minor effect on the electrical transport and the third PF peak. To clarify these, we have modified the manuscript accordingly.
Page 10-"Compositing effects of PbSe for the x > 0.12 samples" section: By scrutinizing the transport data, we find that further increase in PF for the x > 0.12 samples is predominantly resulted from the increases in μ Η and thus σ. This μ Η increase is primarily attributed to the appearance of PbSe precipitates with high hole mobility (~ 1000 cm 2 V -1 s -1 for lightly doped p-type PbSe at 300 K 43,44 ), rather than the Cu 2 Se δ nanodots with small volume fraction and low mobility (~11.1 cm 2 V -1 s -1 for the SHS-SPSed Cu 2 Se at 300 K). 25 4) As shown in Fig.2, the "Pb content" is actual or nominal? Please address and mark it on the X axis.
Response: Thanks for the suggestion. In Fig. 2, all Pb contents shown are nominal, and thus we have changed all x-axis into "Nominal Pb content" in the revised manuscript per referee's suggestion.
Page 7-"Exploring the Complex Band Structure" section: 6) BiCuSeO has strong anisotropy, especially in polycrystalline. In calculation, I am wondering whether the anisotropy was considered in electrical conductivity (Fig. 3e) and lattice thermal conduction (Fig. 5b)

Response:
We thank the referee for the valuable suggestion and have cited the corresponding reference.

Referee #2:
Comments to the Author BiCuSeO has received more and more attentions as a high-performance thermoelectric material. As usual, the thermoelectric materials only have one power factor maximum upon carrier concentration optimization. Interestingly, the authors reported a novel phenomenon in BiCuSeO system that the three power factor maxima were observed when doping Bi with Pb. Based on DFT calculations, numerical modeling, and experimental investigation, the three maxima were attributed to the Fermi energy optimization, band convergence, and compositing effect due to in-situ formed PbSe precipitates. The experiments were well designed and elaborately conducted, and the results are interesting and reasonably discussed. This work is important for the thermoelectric community and of high interest to the materials scientists. In general, I feel that this manuscript should be accepted for publication in Nat. Comm. with a minor revision.
Response: Thanks very much for the referee's comments on our work being "important for the thermoelectric community and of high interest to the materials scientists".
1. Why the calculated C p was used for calculating thermal conductivity?
Response: This is a good question. To obtain the total thermal conductivity, the formula κ = DC p ρ is used, where thermal diffusion coefficient D was measured by a laser flash method (LFA-457, Netzsch, Germany) and the mass density ρ was measured by the Archimedes method. The heat capacity C p is normally measured by DSC, however, the C p of BiCuSeO measured by DSC shows a large variation with temperature or even negative temperature dependence, which would cause large errors in calculating thermal conductivity. This may be related to a slight sublimation of Se or minute endo-or exothermal reactions or processes upon heating-up.
In general, C p (Fig. S1b)  etc.]. This is primarily due to the difficulties in accurately measuring C p using DSC.
2. In supporting information, XRD patterns, there is an impunity peak around 40 o for the pristine sample and Pb0.08 sample, but without any explanation and indexation.

Response:
Thanks for the question. The peak at 40.3 o could be indexed to the (004) pattern of BiCuSeO according to PDF#45-0296, as shown in Fig. S3a, rather from an impurity phase. We have labeled all peaks in the XRD pattern for clarification.
3. Fig. 4 a, the elemental mapping, the makers of Bi, Se, O, Cu are hard to be seen.

Response:
Thanks for the suggestion. We have optimized the figure in the revised manuscript, as shown below.
Page 11-"Compositing effects of PbSe for the x > 0.12 samples" section:  With the increasing Pb content over the solubility limit, the formation of PbSe precipitates will cause non-stoichiometry and thus indirectly increases the Cu 2 Se δ volume fraction, as shown in the XRD patterns in Fig. S3a.
5. Is the sample thermally stable? The repeat of thermoelectric properties measurement is required.
Response: Thank you for the good question. Thermal stability is certainly critical for practical applications. We have repeated the electrical property measurements for these samples, as shown in Fig. R1 and Fig. S9. It is obvious that our SHS-SPSed samples are thermally stable, at least in our repeatable measurements. However, more thorough and systematic studies, including long-time annealing in air or vacuum or heating-cooling cycling, are required to fully characterize the thermal stability of our SHS-SPSed samples.
This is however beyond the scoop of the present study and will be in our future work. Other revisions in the revised paper are shown as below: