Modulation Doping Enables Ultrahigh Power Factor and Thermoelectric ZT in n‐Type Bi2Te2.7Se0.3

Abstract Bismuth telluride‐based thermoelectric (TE) materials are historically recognized as the best p‐type (ZT = 1.8) TE materials at room temperature. However, the poor performance of n‐type (ZT≈1.0) counterparts seriously reduces the efficiency of the device. Such performance imbalance severely impedes its TE applications either in electrical generation or refrigeration. Here, a strategy to boost n‐type Bi2Te2.7Se0.3 crystals up to ZT = 1.42 near room temperature by a two‐stage process is reported, that is, step 1: stabilizing Seebeck coefficient by CuI doping; step 2: boosting power factor (PF) by synergistically optimizing phonon and carrier transport via thermal‐driven Cu intercalation in the van der Waals (vdW) gaps. Theoretical ab initio calculations disclose that these intercalated Cu atoms act as modulation doping and contribute conduction electrons of wavefunction spatially separated from the Cu atoms themselves, which simultaneously lead to large carrier concentration and high mobility. As a result, an ultra‐high PF ≈63.5 µW cm−1 K−2 at 300 K and a highest average ZT = 1.36 at 300–450 K are realized, which outperform all n‐type bismuth telluride materials ever reported. The work offers a new approach to improving n‐type layered TE materials.


Structural characterization in (CuI) x Bi 2 Te 2.7 Se 0.3 (x=0 ~ 0.004) crystals
The powder x-ray diffraction patterns of CuI doping samples are shown in Figure S1a.
All the patterns are indexed to the rhombohedral lattice structure of Bi 2 Te 3 with the space group of R3m with slight peak shifts that caused by CuI doping. The lattice constants as a function of dopant content is shown in Figure S1b. For all the specimens in the series, the change in lattice parameter a and c is not discernible. It is known that Iand Te 2have almost identical ionic radii, and the change in lattice parameters caused by the substitution of I for Te is negligible. As for Cu atoms, in addition to entering the van der Waals gap and occupying the interstitial site between the two quintet. They may also form copper nano-precipitates and disperse in the matrix. From the TEM images of (CuI) 0.003 Bi 2 Te 2.7 Se 0.3 ( Figure S1c), we observed several nano-precipitation of approximately 3-7 nm along the c-axis. More importantly, these Cu-rich nanoscale precipitates show a semi-coherent boundary between the matrix and precipitate. Figure S2a shows the Raman spectra of (CuI) x Bi 2 Te 2.7 Se 0.3 (x=0 ~ 0.004) crystals. Two prominent peaks at 103 cm -1 and 136 cm -1 are assigned to 2 and 1 2 modes, respectively.  Figure S2b shows the x-ray photoelectron spectrum (XPS) of (  It is well known that the best thermoelectric properties of Bi 2 Te 3 -based materials are along the basal plane, so all following thermoelectric properties in this work were performed along the basal plane in order to obtain the best thermoelectric transport performance. Figure S3 presents the temperature dependence of thermoelectric properties for (CuI) x Bi 2 Te 2.7 Se 0.3 (x=0 ~ 0.004). The electrical conductivity (σ) of the samples gradually increases with the increase of CuI concentration, and changes from a non-degenerate semiconductor to a highly degenerate semiconductor ( Figure S3a), which is mainly due to the increase in Hall carrier concentration (n H ). The measured σ, n H and mobility (μ) for all samples at 300 K are summarized in Table S1. We also quantify the number of free electrons donated by each CuI by analyzing the relationship between Hall carrier concentration and CuI doping concentration. Each CuI can contribute about 1.4 electrons in (CuI) x Bi 2 Te 2.7 Se 0.3 crystals. Figure S3b shows the temperature-dependent Seebeck coefficients of all samples. The result is very consistent with the increase in carrier concentration (Table S1).
We found that the Seebeck coefficient of bismuth telluride is very sensitive to the donor-like defects caused by the cooling rate in preparation process. Five (CuI) 0.002 Bi 2 Te 2.7 Se 0.3 crystal samples prepared with the same procedure were examined, and the reproducibility of single crystal preparation was confirmed. CuI doping is essential for accurately controlling the thermoelectric performance of n-type Bi 2 Te 2.7 Se 0.3 materials during the crystal growth. Figure S3c illustrates the temperature dependence of power factor for all samples. Obviously, CuI doping remarkably improves the power factor of Bi 2 Te 2.7 Se 0.3 in the entire measurement temperature range. The power factors of CuI doped crystals with x = 0.002, 0.003 and 0.004 peak at 41 μW cm -1 K -2 around 300 K, and decrease with increasing temperature. The total thermal conductivities (κ) as a function of temperature for all crystals are shown in Figure S3d. The doping of CuI raises κ at 300 K. As the CuI doping level increases from x= 0.001 to 0.004, the corresponding κ also gradually rises from 1.2 to 1.6 W m -1 K -1 at 300K. Generally, κ is the sum of the lattice contribution (κ lat ) and electronic contribution (κ ele ). κ ele is proportional to the σ and can be calculated by the semiconductor, the contribution of bipolar thermal conductivity (κ b ) to the κ will become more and more obvious as temperature rises ( Figure S3e). When the doping content is controlled to x= 0.002~0.003 it has the lowest κ around 325 ~ 400 K. Such reduced κ and enhanced power factor simultaneously result in a significant ZT enhancement in 300 ~ 400 K. As illustrated in Figure S3f, a maximum ZT of 0    Figure S5: Charge transfer analysis of Cu-intercalated Bi 2 Te 3 crystal structure.

Inter-valley and Intra-valley scattering
As shown in Figures. 5e-5h, the energy difference between the 2 nd CBM and the CBM for doped samples is slightly smaller than that for the undoped sample ( Figure. 5h).
For the case of a given electron concentration, a smaller energy difference between 2 nd CBM and CBM indicates that more portion of electrons lies in the 2 nd conduction band valley. This would reduce the intra-valley scattering because fewer electrons lie in the same valley although the inter-valley scattering might be strengthened. The mobility probably can be improved because generally, the inter-valley scattering is much weaker than intra-valley scattering.

Change in wave function of conduction elelctrons due to Cu/I impurities
As shown in Figures. 5a-5d, the charge density of the electronic states inside the quintuple layer (QL) closest to the intercalated Cu ( Figure. 5a) and that inside the QL whose Te atom is substituted by I ( Figures. 5b-c) are higher than that of pure sample ( Figure 5d). The reason would be that both I and Cu in our specimens act as donor, and thus they should contribute donor levels around the conduction band minimum (CBM). If we can calculate a unit cell of nearly infinite volume with a single intercalated Cu (or a substituted I) to model the condition of an isolated impurity, we would see some non-dispersive bands lower than the CBM corresponding to the donor levels and some Cu states merging with Bi 2 Te 3 state above the CBM corresponding to the resonant states. In a practical calculation of a finite and even quite small unit cell, which would be more relevant for high doping concentration of our specimens, all the donor (Cu or I) states should merge with the conduction (Bi 2 Te 3 ) states (note that generally the ionization energy of impurities is only about 10 meV). Therefore, the electronic states around the CBM, to some extent, will show the properties of impurity states. i.e., the wave function becomes more localized around the QL which the impurity is closest to as shown in Figures. 5a-