In-situ resonant band engineering of solution-processed semiconductors generates high performance n-type thermoelectric nano-inks

Thermoelectric devices possess enormous potential to reshape the global energy landscape by converting waste heat into electricity, yet their commercial implementation has been limited by their high cost to output power ratio. No single “champion” thermoelectric material exists due to a broad range of material-dependent thermal and electrical property optimization challenges. While the advent of nanostructuring provided a general design paradigm for reducing material thermal conductivities, there exists no analogous strategy for homogeneous, precise doping of materials. Here, we demonstrate a nanoscale interface-engineering approach that harnesses the large chemically accessible surface areas of nanomaterials to yield massive, finely-controlled, and stable changes in the Seebeck coefficient, switching a poor nonconventional p-type thermoelectric material, tellurium, into a robust n-type material exhibiting stable properties over months of testing. These remodeled, n-type nanowires display extremely high power factors (~500 µW m−1K−2) that are orders of magnitude higher than their bulk p-type counterparts.

. Colloidal stability using Zeta Potential measurements. The averaged (n=102) Zeta Potential and phase distribution for sulfur capped tellurium nanowires (~2.4% sulfur concentration) dispersed in water. Sample concentration was ~10 microgram/mL. The sample was transferred to a zeta cell (Malvern Instruments) and measured at 25 °C. The zeta potential was -27.5 + -10.5 mV (n=102). A viscosity of 0.8872 cP, a dielectric constant of 78.5 and Henry function of 1.5 were used for the calculations. Figure 7. Colloidal stability using Zeta Potential measurements. The averaged (n=72) Zeta Potential and phase distribution for sulfur capped tellurium nanowires (~2.2% sulfur concentration) dispersed in water. Sample concentration was ~10 microgram/mL. The sample was transferred to a zeta cell (Malvern Instruments) and measured at 25 °C. The zeta potential was -18.1 + -10.1 mV (n=72). A negative zeta potential value suggests that the surface is negatively charged. A viscosity of 0.8872 cP, a dielectric constant of 78.5 and Henry function of 1.5 were used for the calculations. Figure 8. Colloidal stability using Zeta Potential measurements. The averaged (n=72) Zeta Potential and phase distribution for sulfur capped tellurium nanowires (~1.8% sulfur concentration) dispersed in water. Sample concentration was ~10 microgram/mL. The sample was transferred to a zeta cell (Malvern Instruments) and measured at 25 °C. The zeta potential was -16.8 + -10.2 mV (n=72). A negative zeta potential value suggests that the surface is negatively charged. A viscosity of 0.8872 cP, a dielectric constant of 78.5 and Henry function of 1.5 were used for the calculations. Figure 9. Colloidal stability using Zeta Potential measurements. The averaged (n=72) Zeta Potential and phase distribution for sulfur capped tellurium nanowires (~1.5% sulfur concentration) dispersed in water. Sample concentration was ~10 microgram/mL. The sample was transferred to a zeta cell (Malvern Instruments) and measured at 25 °C. The zeta potential was -12.3 + -12.8 mV (n=72). A negative zeta potential value suggests that the surface is negatively charged. A viscosity of 0.8872 cP, a dielectric constant of 78.5 and Henry function of 1.5 were used for the calculations. Figure 10. Colloidal stability using Zeta Potential measurements. The averaged (n=72) Zeta Potential and phase distribution for sulfur capped tellurium nanowires (~1.3% sulfur concentration) dispersed in water. Sample concentration was ~10 microgram/mL. The sample was transferred to a zeta cell (Malvern Instruments) and measured at 25 °C. The zeta potential was -10.6 + -10.1 mV (n=72). A negative zeta potential value suggests that the surface is negatively charged. A viscosity of 0.8872 cP, a dielectric constant of 78.5 and Henry function of 1.5 were used for the calculations. Figure 11. Thermopower measurements on thin films of nanowires. Open circuit voltage versus applied temperature gradient for (a) polyvinylpyrollidone-capped (Te-PVP) and (b) fully S 2--exchanged Te NW films (Te-S 2-). The Seebeck coefficient is derived from the slope of the linear fit (R 2 values of 0.9999 and 0.9996 respectively). Error bars representing the standard deviation from averaging 10 readings for each temperature gradient are captured within the data marker.  (001) surfaces, which are denoted as surfaces A, B and C, respectively. The unit cell of tellurium crystal is shown as the tetragonal red dashed box in each plot. As exhibited by (c) and (d), the nanorod grows along the (001) direction with a 5.93Å spacing between two lattice planes. The most stable surface A drives the nanorod to form the hexagonal shape. Supplementary Figure 19. X-ray photoelectron spectrasulfur edge. High-resolution normalized XPS spectra of undoped Te nanowires (Te-PVP) in black and lightly and heavily doped nanowires (Te-PVP-S 2-) in red and blue respectively. In the Te-PVP nanowires there is no peak in the spectrum around 162-162.5 eV which suggests the lack of sulfide species. The presence of a peak around 169-170 eV in undoped Te nanowires corresponds to the Te 4s peak. The apparent red shifting of the peak at 169 eV with increased doping is due to presence of oxidized sulfur species such as sulfates and sulfites which demonstrate S 2p peaks around 169 eV and 166.5 eV respectively leading to a convolution of the Te 4s and S 2p peaks. Figure 20. X-ray photoelectron spectratellurium edge (a, b) High-resolution XPS spectra of the Te-3d peak of undoped Te nanowires (Te-PVP) in black and doped nanowires (Te-PVP-S 2-) in red for two different samples. In both samples, the TeO2 peak is suppressed with doping proving that S 2dopants passivate the nanowire surface effectively thus reducing oxidation. A slight blue shift to higher energies or an increase in binding energies can be observed in the zoomed-in images (c, d) for the doped samples which is suggestive of p-type doping. ) at room temperature. The red curve denotes the simplified analytical expression for when Te is degenerately doped n-type or the degenerate limit, while the blue curve depicts a similar result for the non-degenerate limit but still unipolar doping. At the extreme case of degenerate doping the red curve merges with the dotted black line, while in and around the band edge with non-degenerate doping, the black line matches well with the blue curve. When the effect of the valence band is also taken into consideration for ambipolar transport, the black line deviates from the blue line when the Fermi level is sufficiently far from the conduction band edge (EC) and near the middle of the band gap.

Supplementary Note 1. Colloidal Stability using Zeta Potential Measurements
In the fully exchanged nanowires, the observation that the nanowires do not aggregate and are still dispersible in polar solutions in the absence of any polymer suggests that there exists sufficient electrostatic repulsion between adjacent nanowires to stabilize the colloidal dispersion. In order to confirm this, we perform Zeta potential measurements to establish the charge on the Te-surface for a range of different sulfur concentrations. The binding of the negatively charged S 2ions resulted in a negative ζ-potential in all cases as expected.

Supplementary Note 2. Band structure using density functional theory calculations origin of n-type transport
The crystal structure of tellurium (Te) contains two types of bonds which have distinct bond lengths, one is ~2.9Å and another is ~3.4Å. As shown in Supplementary Figure 12, the short bonds fasten the Te atoms together to form an atomic chain coiled up with 3fold symmetry in (001) direction, whereas the longer bonds bind the atom chains together by van-der-Waals interactions to form the Te crystal structures in space. To understand the bonding mechanism more fully, first, we elucidate the origin of the hexagonal shape of pure Te nanowires by calculating the energies of the three prominent surfaces namely (010), (110), and (001), denoted as A, B, and C respectively. The surface energy calculation can be formulized as = ( − )/2 . Here, denotes the surface energy, is the total free energy of the slab supercell, n is the number of atoms in the supercell, is the energy of a Te atom in bulk form, and is the surface area of the surface of the slab on each side. From our calculations, the surface energies are 18 meV/Å 2 , 21 meV/Å 2 , and 36 meV/Å 2 respectively for the A, B and C surfaces. During the growth process, the nanowire tries to minimize the total configuration energy, so that surface A (010) is always preferred, leading to the hexagonal shape of the nanowire. In all further calculations, we shall focus our analysis only on the (010) surface, since it is the most stable and exposed surface to the S-adatoms. The nanorod grows along (001) direction with the inter-atomic-plane spacing of 5.93Å. 6 The surface C has much higher energy amongst all three surfaces, proving that the ~2.9Å bond is much stronger than that of ~3.4Å bond.
Allotropes of sulfur occur in various polymorphs with a complex phase diagram due to co-existence of weak S-S VdW bonds and relatively stronger S-S bonds with flexible bond geometries. On the other hand, the cohesive energy of Te-Te is weak (small surface energy proves this). Hence, the surface-adsorbed sulfur on tellurium blends the nature of tellurium and sulfur bonding properties and can thus exhibit a variety of surface/interface atomistic geometries in reality. The two extreme cases of the sulfur-surface adsorptions, therefore, can be categorized as chemical-adsorption (or chemisorption) and physicaladsorption (or physisorption) with almost identical formation energy (difference of nearly 60 meV between chemisorbed and physisorbed S). At around room temperature, due to perturbations from thermal energy, in all likelihood, the real physical system might manifest as an intermediate case incorporating both chemically and physically adsorbed sulfur. S adatoms can choose to form bond with either other S-atoms or with Te atoms. In chemical adsorption, S-Te bonding dominates over S-S bonding, while in physical adsorption S-S bonding is more pronounced. S-S chains in the latter case enable conductive surface-states as shown in the corresponding band structure for physical adsorption.
The atom-projected band structure reveals conductive surface states that originate from the S-S chains along the surface in the physical adsorbed scenario due to the sulfur band crossing the Fermi level (Supplementary Figure 15). Hence, the charge mobility should be increased greatly around the Fermi level. While, the chemical adsorbed structure does not have the conductive band crossing the Fermi level, it does introduce a new dopant band close to the conduction band edge which can explain the surprising ntype behavior of the sulfur-doped tellurium nanowires (Supplementary Figure 16). If we compare the calculated Seebeck coefficients from the DOS for the three different cases cited aboveundoped Te and sulfur-doped Te with sulfur adsorbed either physically or chemically (Supplementary Figure 17), we can observe very distinct behavior for the variation in Seebeck coefficient as a function of the Fermi level in each of the three scenarios. In the case where sulfur is physically adsorbed on the surface of tellurium, the S-S chains introduce a lot of surface states which result in switching to negative Seebeck coefficients by shifting the Fermi level to about 25 meV above the valence band (VB) edge of tellurium. Similarly, for the chemically adsorbed case, a Fermi level that is about 190 meV above the VB edge results in negative Seebeck coefficients as compared to 240 meV for undoped tellurium. What these results point out is that, if we assume that the Fermi level in the three different scenarios does not change (i.e. no extra charge carriers are added to the system by doping), then with a fixed Fermi level, it is possible to obtain n-type doping or n-type transport (negative Seebeck coefficients) simply by modifying the local band structure around the Fermi level. For example, if we pin the Fermi level for all three systems at say 230 meV above the VB edge, while we would obtain a Seebeck coefficient of 440 µV/K for the undoped Te (p-type), we would obtain Seebeck coefficients of -76 µV/K and -680 µV/K (n-type) for the physi-sorbed and chemi-sorbed cases respectively. As discussed before, the two extreme cases of the sulfur-surface adsorptions, categorized as chemical-adsorption (or chemisorption) and physicaladsorption (or physisorption) exist with almost identical formation energy. At around room temperature, due to perturbations from thermal energy, in all likelihood, the real physical system might manifest as an intermediate case incorporating both chemically and physically adsorbed sulfur and thus, the Seebeck coefficients that we observe would be some intermediate value.
Traditional electronic doping strategies employ aliovalent atoms, i.e. they use an impurity atom with a different valence compared to the host (e.g. boron or phosphorus in silicon) in order to introduce extra charge carriers. A more recent approach to improve the electronic properties of thermoelectrics has been to manipulate the local density of states (resonant levels) in around the Fermi level of the host material (e.g. Tl in PbTe, Heremans et al. Science 321, 554, 2008, and Sn in Bi2Te3, Jaworski and Heremans et al. Phys. Rev. B, 80, 233201, 2009). To the best of our knowledge there exists only one other report from Jin and Heremans et al., (Energy Environ. Sci. 2015) that reports isoelectronic doping in materials (indium and gallium in bismuth). Our approach also uses iso-electronic doping (sulfur in tellurium) to modify the band structure and local density of states in tellurium and introduces an impurity band close to the conduction band of tellurium.

Supplementary Note 3. Energy Dispersive Spectroscopy and X-Ray Photoelectron Spectroscopy.
Since the average diameter of our Te nanowires is nearly 80-nm, only about 2.5% of the total Te atoms constitute the surface. Assuming 100% coverage of the surface Te atoms with S 2atoms would give us only about 2.5% S-species in the samples. While it remains a challenge to accurately quantify the S 2incorporation at low concentrations, qualitatively we are able to use electron dispersive spectroscopy (EDS) and X-ray photoelectron spectroscopy (XPS) to observe a general increase in sulfur concentration with increasing dopant addition. For fully surface-exchanged samples, while quantifying by EDS gives us nearly 2.4% sulfur, quantification by XPS gives us nearly 2.2% sulfur. These values are pretty close to what one would expect with complete surface exchange (2.5%).