Boosting the thermoelectric performance of p-type heavily Cu-doped polycrystalline SnSe via inducing intensive crystal imperfections and defect phonon scattering

In this study, we, for the first time, report a high Cu solubility of 11.8% in single crystal SnSe microbelts synthesized via a facile solvothermal route.


Introduction
With the capability of directly converting between heat and electricity, thermoelectric materials provide a promising alternative energy supplement in applications by collecting the waste-heat and assisting in nding new energy solutions. 1,2 To evaluate the converting efficiency, the unitless gure of merit ZT is dened as ZT ¼ S 2 sT/k and k ¼ k e + k l , where s, S, k, k l , k e , and T are the electrical conductivity, the Seebeck coefficient, the thermal conductivity, the lattice thermal conductivity, the electrical thermal conductivity, and the absolute temperature, [3][4][5] respectively. A high ZT needs a low k and a high power factor (S 2 s). Since S, s and k e are strongly coupled through the carrier concentration (n), achieving high ZT values has been historically difficult. It is therefore essential to explore favourable electrical transport properties to strengthen the energy conversion efficiency, and to realize a low thermal transport speed to relieve the heat loss at the same time. To achieve this goal, with a narrow band-gap of $0.9 eV, 2,6,7 tin selenide (SnSe) has received great attention for applications in low-cost thermoelectrics. [8][9][10][11] A remarkably high peak ZT of $2.6 has been reported along the b-axis of p-type SnSe crystals, 8 where the performance benets from the crystals' reasonable s and low k values at 923 K. 12 However, as they suffer from potentially high production costs and poor mechanical properties, SnSe crystals are difficult to use in thermoelectric devices, and their critical crystal-growth techniques have considerable limitations for industrial scale-up. 13 Meanwhile, there is strong controversy over the high ZT of SnSe crystals due to the fact that the k values determined in these crystals are not their intrinsic values, 14,15 and the reinvestigation of single crystals has demonstrated much higher k values. 15 To overcome these challenges, polycrystalline SnSe has been considered as an alternative approach. 16 However, due to the low s values derived from low n (<10 18 cm À3 ), the ZT values (<0.3) have been found to be undesirable for un-doped polycrystalline SnSe. 8 As indicated from previous calculations, 17,18 the optimised n value of p-type SnSe is $3 Â 10 19 cm À3 to reach an enhanced ZT value, so that there is a great potential to enhance these values through effective engineering.
Doping and/or alloying have been widely used for tuning n to achieve desired s values. 19,20 Various elements, such as alkali metals (Na and K), 21-28 I-B group metals (Cu and Ag), [29][30][31][32][33][34][35][36] and halogens (Cl, Br and I), [37][38][39][40][41] have been used as dopants in either p-type or n-type SnSe. 16 As a typical I-B group metal and its abundant availability in earth, Cu, each atom having one valence electron (similar to alkali metals), becomes a good candidate to for tuning n, 29 and in turn for improving s. 31 However, the fundamental mechanisms, such as the Cu doping limit and its valence state in SnSe, are still unclear. Recent studies have shown that to achieve homogeneous Cu doping in SnSe is a challenge, 29 and the secondary phase (such as Cu 2 Se) generated during the synthesis is difficult to remove from the system via the post-melting route. 29 Furthermore, there is no direct structural evidence to demonstrate the doping behaviours of Cu in SnSe crystals. Therefore, urgent attention is needed to clarify these fundamentals via critical structural and chemical characterizations, which will illustrate the doping behaviours, and effectively improve s to benet the energy conversion efficiency.
To explore these fundamental mechanisms and achieve a high thermoelectric performance at both low and high temperatures, in this study we fabricated Cu-doped SnSe microbelts via a simple solvothermal method as illustrated in Fig. 1(a), from which a high doping limit of Cu (11.8%) in SnSe microbelts was achieved for the rst time. The secondary phase (Cu 2 Se) in the synthesized products can be found when excessive Cu is doped in SnSe, but this was effectively removed through sonic separation and centrifuging aer the solvothermal synthesis. Through detailed structural characterization as illustrated in Fig. 1(b), it was found that with increasing the Cu doping level, the morphology of Sn 1Àx Cu x Se (x is from 0 to 0.118) can be tuned from rectangular plates to microbelts. Both Cu + and Cu 2+ valence states were conrmed in the synthesized Sn 1Àx Cu x Se via XPS analysis. The observed lattice distortion plays a dominant role in keeping the heavily doped SnSe microbelts in the orthorhombic structure. Aer being sintered into pellets as illustrated in Fig. 1(c), the comprehensive thermoelectric properties, such as carrier mobility (m), n, s, S, S 2 s, and k, were measured and calculated, which led to a high ZT of $1.41 at 823 K when x ¼ 0.118, as shown in Fig. 1(d), indicating that our heavily Cu-doped SnSe has full potential for applications in high temperature thermoelectric devices.

Results and discussion
To understand the extraordinary thermoelectric performance found in our heavily Cu-doped SnSe, we rst investigated the solubility of Cu in SnSe via X-ray diffraction (XRD) analysis and electron probe micro-analysis (EPMA), and then studied the valence state of Cu in SnSe via X-ray photoelectron spectroscopy (XPS). Detailed characterizations by scanning electron microscopy (SEM), high resolution transmission electron microscopy (HR-TEM), spherical aberration corrected scanning transmission electron microscopy (Cs-STEM) with high-angle annular dark-eld (HAADF) imaging and energy dispersive spectroscopy (EDS) are presented and discussed to explain the fundamental reasons for the obtained high thermoelectric performance.
In this study, we use Na 2 SeO 3 as the Se source, SnCl 2 $2H 2 O as the Sn source, and CuCl 2 as the Cu doping source. To study the solubility of Cu in SnSe, we dene the molar percentage r of CuCl 2 in the total amount of CuCl 2 and SnCl 2 $2H 2 O. The selected r values in this study were 0% (no CuCl 2 added), 1%, 2%, 5%, 7.5%, 10%, 20%, and 30%, respectively. Through detailed EPMA studies, the Cu doping level (dened as x for Sn 1Àx Cu x Se) from different r values was found as 0%, 1%, 2%, 5%, 7.5%, 10%, 11.8%, and 11.8%, respectively, indicating that the solubility of Cu in the SnSe system is 11.8%. In the cases of r ¼ 20% and 30%, an obvious secondary phase of Cu 2 Se can be identied when the doping concentration is beyond the solubility (11.8%) of Cu in the SnSe system. However, the secondary phase can be effectively removed through ultrasonic separation and centrifuging techniques aer the solvothermal synthesis; a detailed discussion is shown in the ESI, Fig. S1(a-c). † Therefore, the obtained nal synthesized products with r ¼ 20% and 30% are almost single-phase Sn 0.882 Cu 0.118 Se microbelts.
Investigating the structural characteristics of our synthesized products, Fig. 2(a) shows their XRD patterns. All diffraction peaks for all products can be exclusively indexed as the orthorhombic-structured SnSe, and a Pnma space group (Standard Identication Card, JCPDS 48-1224). As can be seen in Fig. 2(a), the strongest peak is the 400* peak for all products, suggesting that all products should possess signicant {100} surfaces. Because the 400* peak is much more signicant than the other peaks, it is hard to see most of the peaks in detail. To solve this problem, we magnied one of our XRD patterns (r ¼ 20%) as shown in Fig. S1(c), † from which all peaks can be exclusively indexed as the orthorhombic-structured SnSe, and no secondary phase can be found. Fig. 2(b) shows detailed 400* diffraction peaks for different r values, indicating that for r > 0%, all 400* peaks deviate from the standard value at 2q ¼ 31.081 . Even for r ¼ 0%, the slightly right-shied 400* peak indicates the Sn vacancies exist in the SnSe structure. 10 Our extensive EPMA studies found that the true atomic ratio of Sn : Se is $0.996 : 1. With an increase of the Cu doping level, the 400* peaks shi towards a higher 2q, indicating that Cu atoms are incorporated into the SnSe lattice. Because the size of Cu ions is smaller than Sn ions, the Cu-doping leads to a decrease of the lattice parameter a. 42 However, for r > $10%, no further observable shi of the 400* peak suggests that the doping limit of Cu in SnSe is reached, agreeing with the EMPA results, which is a surprising value. To doubly conrm this, we also synthesized products with r ¼ 11.8%, as shown in the yellow-highlighted regions in both Fig. 2(a) and (b). It is clear to see that the peak shi from r ¼ 11.8% is same as that from r ¼ 20% and r ¼ 30%, indicating that the solubility of Cu in the SnSe structure is 11.8%. A detailed discussion about the variations of the calculated lattice parameters (a, b, and c) and unit cell volume can be seen in Fig. S2 in the ESI. † Cu doping has been reported to contribute to a morphology and/or facet change for many materials during their single crystal growth via various solution methods. [43][44][45] For the case of single-crystal SnSe synthesized via our solvothermal route, morphological evolution in Cu-doped SnSe was also observed. Fig. 2(c-e) show typical SEM images of the synthesized products for x ¼ 0, 0.05, and 0.118 (r ¼ 20%), respectively. For x ¼ 0, as shown in Fig. 2(c), the synthesized products have a typical rectangular plate-like morphology, and their lateral dimensions vary between 30 and 200 mm, similar to the reported morphology. 10,46 Interestingly, with increasing x, the morphology of SnSe gradually transfers from rectangular platelike into long belt-like morphology. More evidences of the morphology transition are shown in Fig. S3(a-f) in the ESI. † To determine the preferred facets for different x values, detailed SEM investigations were performed. Fig. 2(f) shows a SEM image of the synthesized SnSe plates with x ¼ 0, from which the circled area is magnied as shown in Fig. 2(g), in which the (100) surface is labelled. It is of interest to note that, compared with other surfaces, the SnSe microplates possess signicant {100} surfaces, which explain why 400* is the strongest peak. To illustrate the potential surfaces of our SnSe microplates, we simulated the single crystal microplate of SnSe using soware (WinXMorph), 47 and the corresponding crystal model is shown in Fig. S4(a) of the ESI. † On the other hand, Fig. 2(h) shows the SEM image taken from a typical Sn 0.882 Cu 0.118 Se microbelt, from which the circled area is also magnied as shown in Fig. 2(i) with the labelled (100) surface. {100} are still the most signicant surfaces on the microbelts. Besides, Fig. 2(i) shows many surface steps parallel to the axial direction of the belt, which is likely to be caused by the irregular stacking of Sn-Se thinner belts. To illustrate the facets of our heavily Cu-doped SnSe, we also simulates the single-crystal microbelts using soware (WinXMorph), 47 and the corresponding crystal model is shown in Fig. S4(b) of the ESI. † Fig. 3(a) shows a TEM image taken from a typical SnSe microplate, in which the electron beam is parallel to the normal direction of the plate. Fig. 3(b) and (c) are the HRTEM image and selected area electron diffraction (SAED) pattern taken from the thin corner area of the plate, and show that the plate has the orthorhombic structure and has a {100} surface. Fig. 3(d) is a TEM image taken from a section of a typical Sn 0.882 Cu 0.118 Se microbelt with a width of $300 nm. The inset is the SAED pattern taken along the d zone-axis, showing that the axial direction of the belt is parallel to the direction. Fig. 3(e) is the corresponding HRTEM image, showing the typical orthorhombic structure. Dislocations are oen found through our HRTEM investigations, and an example is shown in the inset in Fig. 3(e). Fig. 3(f) is an HRTEM image taken from a relatively larger area in a belt, and shows a signicant strain contrast. Such a strain contrast could be caused by the local nonuniformity of Cu doping and a possible mixture of Cu + and Cu 2+ . To conrm this, energy dispersive spectroscopy (EDS) mapping was performed. We used a Mo grid rather than a Cu grid to avoid Cu impact from the grid. Fig. 3(g) shows respective the EDS maps for Se, Sn, and Cu, and overlapped images from a typical microbelt. All of the elements are well distributed, indicating the successful doping of Cu in the SnSe system. The local non-uniformity of Cu can also be seen. Besides, extensive EDS measurements are used to analyse the Cu concentration, and an example is shown in Fig. 3(g), which agrees with our EPMA analysis. To understand the detailed structural characteristics of the Sn(Cu)/Se slabs stack, Cs-corrected STEM-HAADF investigations were performed. Fig. 4(a) is a STEM-HAADF image taken from a typical Sn 0.882 Cu 0.118 Se microbelt viewed along the a-axis, which also shows non-uniform contrast and varied structural patterns, suggesting the local elemental variation. This explains the strain contrast observed in Fig. 3(f). In fact, such local compositional variation and dislocations cause lattice distortions, which in turn enhance the phonon scatterings. Fig. 4(b) and (c) show HR-STEM HAADF images of Area-1 and Area-2 indicated in Fig. 4(a), respectively. For Area-1, the overlays in Fig. 4(b) show lattice parameters, axes, and Sn/Cu atoms in purple and Se atoms in green (shown in the dashed rectangle). The dotted white rectangle in the centre of the overlay indicates the projected unit cell, and the theoretical values of c and b are 4.439 A and 4.186 A, respectively. 48,49 The clear atomic structure of SnSe with no atom disarrangement was observed. Fig. 4(d) and (e) are the intensity line prole-1 (dashed orange line) taken along the c-axis and prole-2 (dashed blue line) taken along the b-axis in Fig. 4(b), respectively. As can be seen, the measured cell parameter for c was $0.44 nm, which is close to the calculated value (4.44 A). Similarly, the measured cell parameter of b in Fig. 4(e) is $0.41 nm, which also is close to the calculated value (4.13 A). All of these evidences demonstrate the nature of the orthorhombic structure of SnSe. Considering the slight difference between peak intensities shown in Fig. 4(d) and (e), it is predicted that Cu 2+ substitutes the position of Sn 2+ , resulting in weakened peaks. For Area-2, the yellow dashed circles in Fig. 4(c) show the areas with a disordered arrangement of atoms. (dashed red line) along the c-axis, from which the measured disordered arrangement of atoms possesses a symmetry line, indicating the potential existence of Cu + illustrated by the inserted crystal structure in Fig. 4(f).
To conrm the co-existence of Cu + and Cu 2+ in our Cu-doped SnSe, XPS analysis was performed. Fig. 5(a) shows the survey scan for synthesized Sn 0.882 Cu 0.118 Se microbelts, indicating the presence of Sn 3d, Se 3d, and Cu 2p energy states, without any energy states of other elements except for O and C. To analyse the detailed information of Sn, Se, and Cu, Fig. 5(b-d) respectively show high-resolution scans of XPS spectra for Sn 3d, Se 3d, and Cu 2p, from which both Sn and Se atoms present single valence states. For Sn, the peaks corresponding to Sn 3d 3/2 and Sn 3d 5/2 are singlets, and no accessorial binding energy peaks can be found, indicating the divalent characteristic of the Sn ions. For Se, a binding energy peak at 53.7 eV corresponds to Se 3d. 46 For Cu, as shown in Fig. 5(d), strong peaks corresponding to Cu 2p 3/2 were observed at $933 eV, indicating the successful doping in SnSe. Interestingly, there were two valence states for the Cu ions (Cu + for the peak at 932 eV and Cu 2+ for the peak at 935 eV) in SnSe, which is a new nding in the doping behaviour of Cu. The quantied at% of Cu agreed with the proposed 11.8% of Cu.
To understand the thermoelectric properties of our Cudoped SnSe microbelts, we sintered as-synthesized products (with x ¼ 0, 0.01, 0.02, 0.05, 0.075, 0.1 and 0.118, respectively) into pellets, and cut the pellets into rectangular chips to measure and calculate the key properties (s, S, S 2 s and k) between 300 and 873 K. Considering that all properties except S measured along the t directions (perpendicular to the sintering pressure) are higher than those measured along the k directions (parallel to the sintering pressure) due to the anisotropy (shown in Fig. S5 in the ESI †), 8,10,50 we chose the t direction as the main measured direction in the following discussions. Fig. 6(a) shows the measured temperaturedependent s parameters for pellets with different x values. Aer doping with Cu, the s values were greatly enhanced at low temperature (from 300 to 450 K) and high temperature (above 773 K) when x ¼ 0.118. As can be seen, two regions for s exist. From 323 to 573 K (the rst region), a typical metallic transport behaviour can be observed. Aer being heavily doped with Cu, the metal cations (especially Cu + ) increased. In this situation, with increasing the temperature, the vibration of the metal cations becomes more intensive than their un-doped counterparts, which severely impede the carrier transport, resulting in a drastic drop in s. 8 From 573 to 873 K (the second region), typical thermally activated semiconducting behaviour derived from the thermal excitation of the carriers is seen, which is similar to the case of single crystals. 8 Besides, the strong bipolar effect, 51 arising between 500 and 600 K, can produce additional holes, leading to a rapid n increase, and in turn increasing s. 18 These results indicate that the doped Cu (mainly Cu + ) can signicantly improve the s of pure SnSe at high temperature by strengthening the thermal excitation of the carriers, even though it results in a slight reduction of s at medium temperature, which is why the pure SnSe sample outperforms most of the Cu-doped samples in this temperature range.
As discussed above, the greatly enhanced s aer being heavily doped with Cu should come from the n enhancement in the Cu-doped SnSe, as suggested in our measured n and m values (see Fig. 6(b) and (c), respectively). To clearly present the key properties, Table 1 summarizes the measured n, m, s, S, S 2 s, C p , and k values of Cu-doped SnSe at both room temperature (300 K) and high temperature (873 K). As can be seen, with increasing the Cu doping level, n is drastically enhanced by roughly one order of magnitude from 1.82 Â 10 17 to 3.44 Â 10 18 cm À3 at room temperature, resulting in an obvious s enhancement. This is because with an increase in the Cu doping level, the proportion of Cu + in the SnSe system is increased, resulting in the rise in n. For m, Fig. 6(c) indicates the relationship with T. In fact, the relation of the power law (m f T d ) governs the variation of m as a function of T. 46,52 There are two regions for m. In the rst region, from 300-673 K, m decreases with T roughly following the curves related to m f T À1.5 , even though the curves uctuate more for Cu-doped SnSe, indicating that the scattering mechanism should still be Table 1 The r, n, m, s, S, S 2 s, C p , and k of Cu-doped SnSe for x ¼ 0, 0.01, 0.02, 0.05, 0.075, 0.1 and 0.118 at both room temperature (300 K) and high temperature (873 K) r (g cm À3 ) 6.084 6.068 6.089 6.112 6.125 6.14 6.142 n (cm À3 ) at 300 K  46,52 In the second region, at high temperatures of 673-873 K, m increases with T roughly following the curves related to m f T 2.3 , which contributes to higher electrical transport properties above 673 K, indicating that an additional scattering mechanism should exist. 46,52 Previous studies have shown that potential barrier scattering at grain boundaries and/or crystal defects combined with phonon scattering may cause such a special m f T d relationship. 46,53,54 Considering that our Cu-doped SnSe has intensive crystal defects, these results are reasonable. Meanwhile, an increase in the Cu doping level, m decreases gradually, which should be derived from the lattice distortion in SnSe, which scatters the transport of carriers. Fig. 6(d) shows the measured temperature-dependent S values for pellets with different x values, in which giant S values can be observed within the moderate temperature range (from 450 to 700 K), similar to the case for SnSe single crystals. 8 The peak S value found in single crystals ($600 mV K À1 at 525 K along the a-axis) 8 is slightly lower than our peak S value ($700 mV K À1 at 523 K with x ¼ 0.01). Such peak S values come from the bipolar transport. 51 With increasing x, the bipolar transport occurring shis slightly to a higher temperature, indicating the increase of n. Fig. 6(e) shows the determined temperature-dependent S 2 s data for pellets with different x values. It is clear to see that the s values play a dominant role in determining S 2 s, and the peak S 2 s value of 5.57 mW cm À1 K À2 can be found at high temperature (823 K) in the Sn 0.882 Cu 0.118 Se pellet.
To further understand the electrical transport properties of our heavily Cu-doped SnSe, we performed density function theory (DFT) calculations to illustrate the evolution of the band structure of SnSe aer Cu-doping. Fig. 7(a) and (b) show the calculated band structures of SnSe before and aer heavy Cudoping, respectively, and the valence band maxima are both pinned to 0 eV in energy. For pure SnSe, as shown in Fig. 7(a), two distinct conduction band minima can be observed around Y and G points of the Brillouin zone, which are denoted as CB 1 and CB 2 , respectively. For the valence band, six maxima can be clearly depicted, with two principal ones lying along the G-Z line. For the heavily Cu-doped SnSe, as shown in Fig. 7(b), there are also two distinct conduction band minima around the Y and G points of the Brillouin zone, denoted as CB 1 and CB 2 . However, for the valence band, different from the pure SnSe, the maxima is not as sharp as pure SnSe, and obvious band convergence of multiple-valences can be observed aer heavy Cu-doping, which is responsible for the enhanced S 2 s.  Cu (mainly by Cu_d) enhances the DOS at the valence bands, indicating the increase of n, agreeing with the experimental results. Overall, the heavy Cu-doping can signicantly improve the hole concentration in SnSe and result in an enhanced S 2 s. By using k ¼ DC p r, 10 the temperature-dependent k values for pellets with different x values can be calculated and plotted, as in Fig. 6(f). The D values are plotted in the inset of Fig. 6(f) as a reference, and the measured C p and r values are also listed in Table 1. With increasing x, k decreases gradually, which could be derived from increased lattice distortions in SnSe, which contribute to effective phonon scatterings. A low k of 0.32 W m À1 K À1 is achieved at 823 K in the Sn 0.882 Cu 0.118 Se pellet.
Because the densities of our sintered pellets are relatively high (all >98.2%), 8,14 these k values are close to the intrinsic value of isotropic SnSe. To understand the observed low k in our pellets, we investigated the lattice contributions (k l ) and electrical contributions (k e ). k e and k l are determined by k e ¼ LsT and k l ¼ k À k e according to the Wiedemann-Franz law, 56 where L is the Lorenz number and L z 1.5 Â 10 À8 V 2 K À2 is used in this study, as calculated using the single parabolic band model 57-59 as shown with calculation details in Section 6 of the ESI. † In fact, L ¼ 1.5 Â 10 À8 V 2 K À2 has been widely used previously since, for SnSe, the k signicantly depends on phonon scattering. 2,8,10,16 Fig. 6(g) shows plots of the determined temperature-dependent k e for pellets with different x values, in which the obtained s shown in Fig. 6(a) were used for determining k e . Our obtained k e values possess the same trend as for s, but the values are very low (all <0.05 W m À1 K À1 over the entire temperature range). Fig. 6(h) plots k l using k l ¼ k À k e for pellets with different x values, where all of the k l values are signicantly low, in particular, only $0.25 W m À1 K À1 at 823 K for x ¼ 0.118. It should also be noticed that our achieved k l value is close to the calculated minimum k l (k l min ) via a classical Debye-Cahill model, 60 from which the calculated k l min were 0.26, 0.36 and 0.33 W m À1 K À1 along the a-, b-and c-axis, 8,17 respectively. In fact, because this calculation is based on the intrinsic SnSe without doping and an ideal relative density of 100%, our achieved k l values are slightly lower than the calculated k l min , which is reasonable. The inset of Fig. 6(h) shows the plots of k l as a function of 1000/T for pellets with different x values and all show a linear relationship, indicating that the phonon scatterings are dominated by the Umklapp phonon scattering. 61,62 Such low k l values are attributed to the strongly anharmonic bonding, 8,62-67 as well as crystal imperfections such as the lattice distortions caused by local non-uniform doping and dislocations and grain boundaries (or interfaces). 68,69 The calculated k l /k ratio for our pellets are all greater than 80%, indicating that the phonon transport dominates the k values, as shown in Fig. S6(b) in the ESI. † Fig. 6(i) shows a comparison of experimental ZT values with predicted values by calculation at 823 K, where the calculation was based on a single parabolic band model (detailed calculations can be seen in Section 6 of the ESI †). [57][58][59]70 It is clear to see that our measured n value (2.04 Â 10 19 cm À3 ) is very close to the predicted value ($3 Â 10 19 cm À3 ), which can result in a peak ZT of $1.5, indicating that there is still scope for achieving a higher ZT.
To further understand the low k l data, we analysed our sintered pellets by XRD, SEM and TEM characterizations, and the results are shown in Fig. 8. Fig. 8(a) shows typical XRD results for both pure SnSe and Sn 0.882 Cu 0.118 Se pellets; here, all diffraction peaks for all sintered pellets can be exclusively indexed as the orthorhombic structured SnSe, and a space group of Pnma (Standard Identication Card, JCPDS 48-1224), indicating that the compositional features were successfully retained aer sintering and no other phase was observed. Fig. 8(b) shows the magnied XRD patterns and demonstrates the peak deviation at 111* and 400*, from which the samples cut along the t direction show a strong 400* peak, and the samples cut along the k direction shows a strong 111* peak. Comparing the XRD results of the two pellets, it is clear that the 111* peak of Sn 0.882 Cu 0.118 Se is much stronger than that of pure SnSe along the t direction, and the 400* peak of Sn 0.882 -Cu 0.118 Se is also stronger than that of pure SnSe along the k direction, both indicating that Sn 0.882 Cu 0.118 Se pellets possess a much weaker anisotropy than pure SnSe pellets. Besides, Table 2 Comprehensive summary of the thermoelectric performance of p-type doped polycrystalline SnSe. Here, solvothermal is abbreviated as ST, hydrothermal is abbreviated as HT, melting is abbreviated as M, zone-melting is abbreviated as ZM, annealing is abbreviated as A, solidstate solution is abbreviated as SSR, mechanical alloying is abbreviated as MA, hot-pressing is abbreviated as HP, and spark plasma sintering is abbreviated as SPS. The * means that the n values were measured at room temperature  Fig. 8(f) is a magnied TEM image taken from a laminar TEM specimen sliced using an ultramicrotome (inset TEM image in Fig. 8(f)), in which cracks (fractured during ultramicrotome processing) can be seen due to the weak van der Waals force between the Sn-Se layers. Nevertheless, crystals can be seen between the cracks, which can be used to evaluate the structural characteristics of the sintered pellets. Fig. 8(g) is a [100] zone-axis HRTEM image with inset the fast Fourier transform (FFT) pattern, where strain contrast is observed. Fig. 8(h) is another HRTEM image taken from a typical grain boundary. Such local structural variations cause lattice distortions, which in turn enhance the phonon scatterings and contribute to the low k l values. All these results demonstrate that the compositional and structural features have been successfully retained during the sintering. In fact, the "intensive crystal imperfections" were derived from the synthesis, which were shown in Fig. 3 and 4.
To compare the thermoelectric properties in more detail, Table 2 summarizes the main thermoelectric properties, including ZT, s, S, S 2 s, k, n and r with the similar studies of p-type doped SnSe. As can been seen, the low k and moderate s values play the dominant role in achieving the competitive high ZT gures in our heavily Cu-doped SnSe. Section 7 in the ESI † also summarizes both average and peak ZT values with the similar studies of p-type doped SnSe, indicating that our heavily Cu-doped SnSe is very competitive.

Conclusions
In conclusion, a high doping limit of Cu at 11.8% has been achieved in single-crystal Cu-doped SnSe microbelts for the rst time synthesized via a facile solvothermal method. Through detailed structural and chemical characterizations, with increasing the Cu doping level, the morphology of Cu-doped SnSe transfers from rectangular plates to microbelts. Both Cu + and Cu 2+ co-exist in the microbelts. Lattice distortions are observed, which play a dominant role in keeping the heavily doped SnSe microbelts as an orthorhombic structure. Besides, the pellets sintered from such heavily Cu-doped microbelts demonstrate a high thermoelectric performance. The high ZT value of $1.41 at 823 K was achieved, coming from the high power factor and low thermal conductivity. This study lls in the gaps of the existing knowledge concerning the doping mechanisms of Cu in SnSe systems, and provides a new strategy for achieving a high thermoelectric performance in SnSe-based thermoelectric materials.
Here NaOH (99.99%) was used to adjust the environment of the solvent, and ethylene glycol (EG, 45 ml) acted as both the solvent and the reducing agent, which beneted the ion reaction. 46,80,81 The solution was kept stirring for 10 min at room temperature, before being sealed in a polytetrauoroethylenelined stainless steel autoclave (125 ml). The autoclave was heated at 230 C for 36 h in an oven, followed by furnace cooling to room temperature. The synthesized products were collected by centrifugation, and the secondary phase (Cu 2 Se) was removed via ultrasonic-assisted sedimentation. The puried products were then washed using ethanol and deionized water several times, before drying in the oven at 60 C for 15 h.

Instruments
The synthesized products were characterized by XRD (Bruker-D8) to determine their crystal structures, and by XPS (Kratos Axis Ultra) to determine the valence state of Cu in the SnSe (the energy scale was calibrated by carbon). Lattice parameters were obtained by analysing the diffraction patterns using the JADE soware package. EPMA (JEOL JXA-8200) was used to determine their compositions. SEM (JSM-6610, JEOL Ltd.) was used to obtain the morphological characteristics of the synthesized products, and HR-TEM (TECNAI-F20) and Cs-corrected HR-STEM (Titan-G2) were used to characterize their structural and chemical features. The TEM specimens of the sintered