Thermoelectric performance enhancement of eco-friendly Cu2Se through incorporating CB4

We have prepared Cu2Se + x wt% CB4 composites with x = 0, 0.1, 0.3, 0.5, and 0.7 by a hydrothermal method and hot-pressing technique. The structural and compositional analysis indicates that pure phase Cu2Se powders were synthesized and the densified layered bulk samples were obtained. Electrical properties testing showed that the sample with x = 0.5 has the high power factor of 0.886 mW m−1 K−2 due to its high Seebeck coefficient. Meanwhile, the thermal conductivity was suppressed to 0.6 W m−1 K−1 at 773 K. As a result, the final optimized ZT value of 1.46 at 773 K was achieved. These results suggest that CB4 could be an alternative inclusion to improve effectively the thermoelectric performance of Cu2Se.


Introduction
With the wide demand for environment-friendly energy, thermoelectric (TE) materials have received signicant attention during the past two decades. 1-6 TE materials can realize the conversion directly between thermal and electrical energy. [7][8][9] High-efficiency TE materials require a high dimensionless gure of merit (ZT), ZT ¼ S 2 sT/k, where S is the Seebeck coef-cient, s is the electrical conductivity, T is the absolute temperature, and k is the total thermal conductivity respectively. [10][11][12][13] Numerous exploratory efforts tried to optimize these parameters, i.e., enhance the power factor (PF ¼ S 2 s) and reduce the thermal conductivity simultaneously, to achieve high TE performance. 14,15 However, these parameters are strongly coupled together, making it challenging to improve the ZTs. 16,17 Chalcogenide TE materials have attracted widespread interests and have been intensively studied because of their intrinsically low lattice thermal conductivity (k L ). [18][19][20][21] Very recently, Chung and co-workers reported 22 the ultralow k L of 0.07 W m À1 K À1 obtained in polycrystal SnSe and thereby got a high ZT of 3.1, which breaks thermoelectric performance limits. 23 Cubased chalcogenide also received much attention for environmental-friendly TE application and Cu 2 Se is one of the strongest candidates. 24,25 To improve the TE performance of Cu 2 Se, numerous strategies have been developed, such as nanoengineering, 26,27 solid solution alloying, 28,29 element doping, [30][31][32] and introducing highly dispersed nanoparticles. [33][34][35][36] Among them, adding nanoparticles to Cu 2 Se matrix has been demonstrated to be an effective and prevailing approach to improve TE performance. The nanoparticles in composites, such as TiO 2 , 37 graphene, 38 nano-boron, 39  In this work, we try to introduce boron carbide (CB 4 ) as a second phase in the Cu 2 Se matrix because nano-CB 4 has good thermal stability, high temperature electrical conductivity and relatively large Seebeck coefficient. 43 For preparing Cu 2 Se samples with different contents of CB 4 , the hydrothermal method and hot-pressing technique are employed. The results show that CB 4 inclusions do reduce k signicantly and meanwhile enhance PF slightly over a wide range of temperature. As a result, the highest ZT reaches 1.46 at 773 K, which is much higher than the value of undoped sample.

Experimental section
Cu 2 Se nanopowders were prepared by hydrothermal method and CB 4 nanopowders were added during the hydrothermal process. We rst mixed 20 mmol copper chloride dehydrate (CuCl 2 $2H 2 O, 99.9%), 10 mmol selenium dioxide (SeO 2 , 99.9%) and a recommended amount of CB 4 (99%) into about 60 milliliters deionized water, and then magnetic stirred continuously for 12 h. The amount of CB 4 is determined by the mass ratio of Cu 2 Se and CB 4 . In this work, we consider ve Cu 2 Se + x wt% CB 4 (x ¼ 0, 0.1, 0.3, 0.5, 0.7) samples, which are named as S0, S1, S3, S5 and S7 respectively for the convenience of description. Aer the components were completely dissolved, we added 20 milliliters hydrazine hydrate (N 2 H 4 $H 2 O) into the mixture solution. The solution was then loaded into a Teonlined stainless steel autoclave and sealed. The autoclave was heated at 453 K for 24 h and cooled down to room temperature naturally. The product was collected by centrifuging and washed by deionized water and anhydrous ethanol for several times. We nally dried the product in vacuum at 333 K for at least 12 h. For getting the nal samples, the synthesized powders were sintered into pallets with a diameter of 20 mm by hot-pressing method at 973 K and under a pressure of 40 MPa in the vacuum for 30 minutes.
The phase compositions of the prepared samples were identied by X-ray diffraction measurements (XRD, PANalytical, Netherlands) with Cu Ka radiation. The fracture surface microstructure was observed by eld emission scanning electron microscopy (SEM, TM4000) and Energy-dispersive X-ray spectroscopy (EDS, X-MAX N 20, England). Seebeck coefficients  and electrical conductivity simultaneously were measured by using a CTA-3 system (Cryoall, China) from 323 K to 773 K in a helium atmosphere. k was calculated according to the formula k ¼ rDC p , where r is the density of the samples, D is the thermal diffusivity, and C p is the heat capacity respectively. D was measured using the laser ash method (DFL-1600, Netzsch) on a square piece specimen with 10 mm in side and 1.5 mm in thickness. C p was measured by differential scanning calorimetry (DSCQ-2000, TA). The density of the samples r was obtained from Archimedes' method. Fig. 1(a) shows the XRD patterns of the synthesized Cu 2 Se + x wt% CB 4 (x ¼ 0, 0.1, 0.3, 0.5, 0.7) nanopowders. The main diffraction peaks are all in consistent with the ICDD le No. 00-006-0680, indicating b-phase Cu 2 Se is prepared successfully. 40,[44][45][46] No CB 4 phase is to be detected, which might be due to introducing only a small amount into the main b-phase Cu 2 Se. Fig. 1(b) shows the XRD patterns of the sintered bulk samples. The results match with the ICDD le No. 00-019-0401 very well, indicating that the as-prepared b-Cu 2 Se nanostructures has transferred to pure a-Cu 2 Se aer sintering. 40,46,47 The typical morphology of the synthesized Cu 2 Se precursor nanopowders is shown in Fig. 2(a). Nanoplates with a lateral size of 100-500 nm, oen aggregating together, can be clearly observed. Fig. 2(b-f) display the cross-sectional microstructures of the hot-pressed Cu 2 Se + x wt% CB 4 bulk samples. Similar lamellar structure is observed for each sample. Such structure is conducive to reducing the thermal conductivity due to the increased scattering of phonons between layered interfaces. 48 A few pores on the surface can be spotted which might be due to Se volatilization during the sintering process and hence the corresponding density is slightly lower than the pure Cu 2 Se. The samples S0, S1, S3, S5, and S7 have densities of 5.75, 5.65, 5.38, 5.29, 5.20 g cm À3 respectively (theoretical density of Cu 2 Se is 6.74 g cm À3 ). Fig. 3 shows the surface EDS mapping of the sample S5. These maps suggest that copper, selenium, carbon and boron signals overlap, showing CB 4 is uniformly distributed in the sample.

Results and discussion
The temperature dependence of the electrical conductivity s of our samples is shown in Fig. 4(a). Within the temperature range of 400-773 K, the electrical conductivity s of all samples decreases monotonously with temperature increasing, which shows the typical behaviors of metallic-like characteristic. However, we can clearly nd that the electrical conductivity s varies abnormally around 400 K, which have been proved to be due to the phase transition from a-Cu 2 Se to b-Cu 2 Se. 49 In addition, the electrical conductivity s decreases with the increasing of CB 4 components at each testing temperature. For example, compared to the sample S0, the electrical conductivity s is about 3 times smaller around 400 K and 2 times smaller at 773 K for the sample S7. This trend in electrical conductivity s is very different from the report of Cu 2 Se/SiC composite, 27 but similar to the report of Cu 2 S/SiC composite. 17 For clarifying the underlying mechanism of the electrical transport behaviors, the room temperature carrier concentration (n) and carrier mobility (m) are measured, as shown in Table 1. Clearly, there is a reduction in carrier concentration and an increase in carrier mobility with increasing of the CB 4 composite content. For the superionic conductor Cu 2 Se, unstable copper ions show a liquid-like behavior and hence cause the unrepeatable electrical transport properties. The carrier concentration n contains two parts: the majority carrier (n hole ) and the ion carrier (n ion ), and therefore, the electrical conductivity s ¼ s hole + s ion . In our work, the introduction of CB 4 might effectively block the migration of copper ions, which causes the reduction in n ion and hence in the whole carrier concentration n. In addition, the migration restrictions of copper ions can also effectively enhance the stability of Cu 2 Se chemical structure, which is conrmed by the cyclic testing shown in Fig. 4. The increase in mobility m might be due to the grain growth during the process of hot pressing, as can be clearly seen in Fig. 2. Because of the difficulty for measuring the transport of ions, more related discussions do not be present here.
In our work, the samples S5 and S7 show a quite low electrical conductivity s. This fact would not be very good for optimizing the thermoelectric properties of Cu 2 Se. However, thankfully, the reduction in electrical conductivity s is oen accompanied by enhancement of the Seebeck coefficient. As we can see from Fig. 4(b), the samples S5 and S7 exhibit a larger  Seebeck coefficient than the others within the entire testing temperature range and the optimal value reaches 137 mV K À1 at 773 K for the sample S7. Additionally, the positive Seebeck coefficient for all samples suggests the p-type electrical transport in Cu 2 Se. The calculated PFs of the samples are shown in Fig. 4(c). Very clearly, PFs curves of the samples S0, S1, S3 and S5 show a valley around 400 K due to the phase transition from b-Cu 2 Se to a-Cu 2 Se, 50,51 while the sample S7 has the abnormal high PF because of its very high S and low s. However, we cannot evaluate precisely these parameters around 400 K because they are very sensitive to temperature change during the process of phase transformation. This might be the reason that some previous studies did not show the related data close to 400 K. As T >400 K, all PFs increase rapidly as temperature increases. However, the sample S7 does not show the superior PF value, which might be due to its very low electrical conductivity. In contrast, the sample S5 has the optimal PF as T >400 K (PF max ¼ 0.886 mW m À1 K À2 at 773 K), indicating that a moderate content of CB 4 composites may result in a better electrical performance of Cu 2 Se.
Another important parameter to determine the TE properties is the thermal conductivity k. The thermal conductivity can be divided into two parts: k ¼ k L + k carrier , where k L comes from heat phonons travelling through the crystal lattice and k carrier arises from heat carriers' movement. Generally, at high temperature, k L dominates in Cu 2 Se matrix. 39 The measured thermal conductivity is shown in Fig. 4(d). Clearly, all of the thermal conductivity k decrease with temperature increasing, which is a reasonable trend because of the increasing in phonon scattering and the reduction in carrier concentration. For a xed temperature, as CB 4 component increases from 0.0 to 0.7%, the thermal conductivity decreases very rapidly. To take a step further, carrier thermal conductivity (k carrier ) is calculated by using the Wiedemann-Franz relationship, k carrier ¼ LsT, where L is the Lorenz number. When the detailed band structure and scattering mechanism are not known, the Lorenz number can not be viewed as a constant but can be calculated by L ¼ 1.5 + exp [ÀjSj/116], where L is in unit of 10 À8 W U K À2 and S (Seebeck coefficient) in unit of mV KÀ 1 . 52 The calculated L and k carrier are showed in Fig. 5. However, it seems that the measured total thermal conductivity of S5 and S7 are lower than the calculated values. The similar phenomenon happened to Cu 2 Se/SiC system 27 and Cu-S system. 17,53 It might be due to the different thermal transport properties between electron/hole carriers and ion carriers. The equation k carrier ¼ LsT may not be suitable for Cu 2 Se/CB 4 composite. The ion carriers have a non-negligible contribution to total thermal transport, so the modied equation can be taken as k carrier ¼ L (s À s ion )T + k ion , 27,30,53 where k ion is the ionic thermal conductivity, s ion is ionic electrical conductivity. However, s ion and k ion cannot be measured precisely now and hence the carrier thermal conductivity and lattice thermal conductivity are not offered here.
Considering the total thermal conductivity, the samples S5 and S7 remain very low values and very close to each other as T >400 K. The lowest k value is 0.59 W m À1 K À1 at 773 K for S7 sample. This result indicates that the higher concentration of CB 4 inclusions could not improve further the thermal conductivity. Moreover, as shown in Fig. 4(c), the PF of sample S7 is also smaller than sample S5. Therefore, in this work, the S5 sample would show the optimal thermoelectric performance. This can be conrmed positively by the relationship between ZT values and temperature, as shown in Fig. 6. Sample S5 exhibits  the largest ZT than the others as T >400 K and the largest value is of 1.46 at 773 K, which is much larger than that of undoped sample S0 (ZT max ¼ 0.55 at 773 K). The average ZT value of S5 (ZT average ¼ 1.034) is also larger than one within the temperature range of 400-773 K. These results suggest that Cu 2 Se compositing a certain amount of CB 4 is very promising for TE application.

Conclusions
We have prepared the pure phase Cu 2 Se + x wt% CB 4 (x ¼ 0, 0.1, 0.3, 0.5, 0.7) nanopowders by hydrothermal method and have sintered the corresponding bulk samples by hot-pressing technique. With the content of CB 4 increasing, the electrical conductivities of our samples decrease very rapidly, while the Seebeck coefficients steadily increase. The nal PFs show that the sample S5 possesses the optimal value within the testing temperature range. The thermal conductivity testing conrms that the CB 4 nano-inclusions can drastically reduce the k of our samples. The lowest k value is 0.59 W m À1 K À1 at 773 K for S7 sample. However, the k of sample S5 is very close to that of sample S7 as T >400 K. Finally, the optimized ZT value of 1.46 at 773 K is achieved for S5 sample, which is about 165% larger than the undoped sample.

Conflicts of interest
The authors have no conicts to declare.