Yu
Liu
a,
Doris
Cadavid
*a,
Maria
Ibáñez
bc,
Jonathan
De Roo
d,
Silvia
Ortega
a,
Oleksandr
Dobrozhan
a,
Maksym
V. Kovalenko
bc and
Andreu
Cabot
*ae
aCatalonia Institute for Energy Research-IREC, Jardins de les Dones de Negre 1, 08930, Sant Adrià del Besos, Spain. E-mail: dcadavid@irec.cat; acabot@irec.cat
bInstitute of Inorganic Chemistry, Department of Chemistry and Applied Biosciences, ETH Zürich, CH-8093, Switzerland
cEmpa-Swiss Federal Laboratories for Materials Science and Technology, Dübendorf, CH-8600, Switzerland
dGhent University, Department of Inorganic and Physical Chemistry, Krijgslaan 281 S3, 9000 Gent, Belgium
eInstitució Catalana de Recerca i Estudis Avançats – ICREA, 08010 Barcelona, Spain
First published on 13th April 2016
We present a high-yield and scalable colloidal synthesis to produce monodisperse AgSbSe2 nanocrystals (NCs). Using nuclear magnetic resonance (NMR) spectroscopy, we characterized the NC surface chemistry and demonstrate the presence of surfactants in dynamic exchange, which controls the NC growth mechanism. In addition, these NCs were electronically doped by introducing small amounts of bismuth. To demonstrate the technological potential of such processed material, after ligand removal by means of NaNH2, AgSbSe2 NCs were used as building blocks to produce thermoelectric (TE) nanomaterials. A preliminary optimization of the doping concentration resulted in a thermoelectric figure of merit (ZT) of 1.1 at 640 K, which is comparable to the best ZT values obtained with a Pb- and Te-free material in this middle temperature range, with the additional advantage of the high versatility and low cost associated with solution processing technologies.
AgSbSe2 is a narrow, indirect band-gap semiconductor (0.03–0.10 eV) showing p-type conductivity.15 However, an apparent optical bandgap of 0.6–1.1 eV2,15–18 and a high optical absorption coefficient (104 cm−1) have also motivated the use of AgSbSe2 for photovoltaic applications.2,17,19–21
The AgSbSe2 band structure is characterized by a multi-peak valence band maximum, which potentially results in a high effective mass for holes and thus a high Seebeck coefficient. In addition, AgSbSe2 features a strong anharmonic bonding arrangement associated with the Sb 5s2 lone pair, which translates into strong phonon–phonon interactions that reduce the lattice thermal conductivity to values close to the amorphous limit. This intrinsically low thermal conductivity and the appropriate electronic band structure make AgSbSe2 and AgSbTe2 promising candidates for TE applications in the intermediate temperature range (500–700 K).16,22–28 While AgSbTe2 exhibits higher electrical conductivity, and its nanostructured alloys with GeTe (TAGS)29 and PbTe (LAST-m)30 are well known for their remarkable ZT values, AgSbSe2 is advantageous in terms of thermal stability, abundance of constituting elements and cost. Bulk AgSbSe2 is typically produced by the solid state reaction of its highly purified elements at temperatures above 1000 K. A relatively large variety of extrinsic electronic dopants, introduced as pure elements in the reaction mixture, have been tested, including the substitution of Sb3+ by monovalent Na+,28 divalent Pb2+,16 Zn2+,23 Sn2+,24 Cd2+,26 Mg2+,27 Ba2+,27 and peculiarly by trivalent Bi3+.16 Extrinsic dopants can also introduce additional disorder and point defect scattering reducing the material thermal conductivity. In some cases, e.g. Na+, Zn2+ and Ba2+ can also introduce nanoprecipitates with modified stoichiometry, e.g. Na-rich AgSbSe2 or even of secondary phases, e.g. ZnSe and BaSe3, which further reduce thermal conductivity and allow reaching ZT values of up to ∼1.1.23,27 Besides, intrinsic doping strategies based on tuning the material stoichiometry by modifying, for instance, the Sb content25 and reaching ZT values of up to ∼1 have also been tested by solid state reaction methods.
The preparation of AgSbSe2 with controlled properties by means of solution processing methods is especially appealing to investigate the structural and compositional dependent functional properties, optimize the material for relevant applications and develop high performance cost-effective products. In particular, the availability of AgSbSe2 NCs with tuned size, shape, composition and phase would allow high density data storage systems, cost-effective photovoltaic devices and high performance thermoelectric modules to be produced by highly versatile, low-cost, high-throughput and high-yield solution-based bottom-up technologies.31–36 However, to the best of our knowledge, no synthesis protocol to produce colloidal AgSbSe2 NCs is currently available.
Here, we detail the first colloidal synthesis of monodisperse AgSbSe2 NCs on the gram scale, determine the NC surface composition by NMR analysis and demonstrate the possibility of electronically doping this material with controlled amounts of bismuth. Finally, we elaborate upon a procedure to remove organic ligands and demonstrate the suitability of these NCs to produce high performance thermoelectric nanomaterials.
![]() | ||
Fig. 1 (a) TEM micrograph of the AgSbSe2 NCs. (b) NCs dispersed in chloroform. (c) The amount of NCs produced per batch. (d) The XRD pattern of AgSbSe2. |
The presence of OLA and OA in the precursor solution was required to dissolve AgNO3 and SbCl3, respectively.39 The absence of OA prevented the incorporation of Sb and thus the formation of stoichiometric AgSbSe2 NCs. Likewise, the absence of OLA resulted in a polydisperse mixture of AgSbSe2 and Ag2Se (Fig. S2 and S3, ESI†).
Although the reaction mixture already turned dark brown upon injection of the selenium precursor, relatively long reaction times were necessary to produce monodisperse NCs (Fig. 2 and Fig. S4, ESI†). Within the first 15 min, NCs with a bimodal size distribution were obtained. This bimodal distribution possibly originates from an extended nucleation time due to a relatively low reactivity of the precursor complexes, which translates into a slow NC growth that prevents a rapid monomer reduction in the supersaturation.40 A focusing of the size distribution, most probably mediated by an Ostwald ripening process, was observed up to 30 min of reaction time, when the narrowest size distribution was obtained. Larger reaction times resulted in a non-uniform growth of the NCs, increasing the size distribution width.
Consistent with this growth mechanism, an increase of the injection and reaction temperature and thus of the monomer reactivity resulted in larger NCs. Therefore, fixing the reaction time, the AgSbSe2 NC size could be easily tailored in the range from 7 ± 1 nm to around 17 ± 2 nm by just adjusting the reaction temperature in the range from 180 to 240 °C (Fig. 2 and Fig. S5, ESI†). Reaction temperatures below 180 °C did not allow the growth of AgSbSe2 NCs. Higher reaction temperatures resulted in highly polydisperse NCs.
In the 1H nuclear magnetic resonance (NMR) spectrum of the resulting NC dispersion in deuteroform, slightly broadened resonances are observed which correspond to an oleyl chain (Fig. 3a). Since the integrations from the alkene resonance (5) and the CH3 resonance (6) fit the expected 2:
3 ratio we conclude that fully saturated molecules are not present. However, it is difficult to distinguish OA from OLA since the characteristic CH2 resonances (1 and 2) next to the functional group are not detected, presumably due to the interaction of the ligand with the NC surface. The nuclear Overhauser effect spectrum (NOESY) (Fig. 3b) indeed confirms that the oleyl species interact with the surface since negative (black) nOe cross peaks are observed.41 The total ligand concentration was determined to be 4.5 mM (based on resonance 6) and with the NC concentration and NC size, a rather low ligand density of 1.2 nm−2 was calculated. The diffusion ordered NMR spectrum (DOSY, Fig. 3c) features two sets of resonances (apart from acetone), with a different diffusion coefficient, D. This indicates that two species are present which diffuse with a different speed. However, none of the two species diffuses slow enough to correspond to a 15 nm nanocrystal plus a tightly bound ligand shell, which would yield a solvodynamic diameter, ds = 19 nm. We thus infer that the ligands are not tightly bound but are in a dynamic exchange regime between a free state and a bound state. Therefore, the ligand density that was calculated earlier is a maximum value and, in reality, the ligands only spend part of their time on the surface. A low bond strength of the ligands to the NC surface and the observed ligand dynamic exchange are consistent with the NC growth not being limited by the surfactant concentration, but by the growth kinetics.
![]() | ||
Fig. 3 (a) 1H NMR spectrum of AgSbSe2 15 ± 2 nm NCs (4 μM) and the corresponding (b) NOESY and (c) DOSY spectra. |
To elucidate the nature of the two species in DOSY, we stripped the ligands from the NC surface by the addition of TFA. After removal of the NCs by centrifugation, the supernatant was again measured using NMR (Fig. S6, ESI†) and the characteristic resonances of both OA and OLA were observed with a relative abundance of 18% and 82% respectively.
Controlled amounts of bismuth were incorporated within the AgSbSe2 NCs by replacing a small amount of SbCl3 by an equivalent amount of Bi(CH3COO)3 in the precursor solution. At relatively low Bi concentrations, [Bi] < 5%, EDX analysis showed the final Bi composition in the NCs to match, within the experimental error, the nominal concentration of introduced Bi(CH3COO)3. The incorporation of Bi ions within the AgSbSe2 lattice was confirmed by a slight peak shift toward lower angles in the XRD pattern (Fig. 4b), pointing towards an increase of the lattice parameters. Bi-doping had also an evident influence on the size and shape distribution of the produced NCs (Fig. 4a).
![]() | ||
Fig. 4 (a) TEM micrograph of AgSb0.98Bi0.02Se2 NCs and (b) their XRD pattern including the reference JCPDS 00-012-0379. The inset shows details of the (200) peak. |
To use AgSbSe2 NCs in solid-state devices that require an efficient transport of charge carriers, original insulating organic ligands have to be removed. Among the different ligand displacement agents tested, NaNH2 was the most effective one, as characterized by Fourier-transform infrared spectroscopy (FTIR). After ligand removal with a 0.02 M NaNH2 solution, AgSbSe2 could not be re-dispersed in non-polar solvents and the strong C–H vibration modes (2850–3000 cm−1) and the bands corresponding to C–C, C–N and NH2 (700–1650 cm−1) had completely disappeared from the FTIR spectrum (Fig. 5), proving the effective removal of OLA and OA from the NC surface.
![]() | ||
Fig. 5 From top to down, FTIR spectra of pure OA, DDT, OLA, as-produced AgSbSe2 NCs (AgSbSe2-OL) and AgSbSe2 NC after ligand removal with NaNH2 (AgSbSe2-LD). |
After ligand displacement, dried 15 nm AgSbSe2 NCs were hot-pressed into 10 mm diameter and 1.5 mm thick pellets at 350 °C under 70 MPa pressure for 30 min. All pellets produced had a metallic lustre, were mechanically robust and had relative densities above 93% as measured by Archimedes' method. No secondary phases or changes in composition of the AgSbSe2 and AgSb0.98Bi0.02Se2 materials during the hot press process were detected by EDX and XRD (Fig. S7 and S8, ESI†). SEM and TEM characterization showed a very low porosity and large grains, up to several hundred nanometers, during the thermal processes (Fig. 6a and Fig. S9, ESI†). However, HRTEM characterization allowed discerning crystallographic order domains in the nanometer size regime (Fig. 6b).
![]() | ||
Fig. 6 (a) SEM micrograph of the fractured surface of a AgSbSe2 pellet (inset); (b) HRTEM micrograph of the same AgSbSe2 pellet showing several crystal nanodomains. |
Fig. 7 shows the electrical conductivity (σ), Seebeck coefficient (S), thermal conductivity (κ), and thermoelectric figure of merit (ZT = σS2T/κ) of AgSbSe2 and AgSb0.98Bi0.02Se2 pellets. Both materials showed a p-type electronic character. AgSbSe2 was characterized by a relatively low σ, which increased from 1380 S m−1 at ambient temperature up to 1700 S m−1 at 600 K. It showed high Seebeck coefficients reaching up to 462 μV K−1 at 695 K, which relates with its flat valence band maximum and multipeak valence band structure.16 The introduction of an atomic 0.5% Bi significantly enhanced the electrical conductivity, which increased from 4520 S m−1 at room temperature to 5970 S m−1 at 618 K in AgSb0.98Bi0.02Se2. These values are slightly above those previously reported for bulk AgSb0.98Bi0.02Se2 produced by solid state methods.16 With Bi doping, the Seebeck coefficient decreased to values below 360 μV K−1, which is consistent with an increase of carrier concentration. Overall, the AgSb0.98Bi0.02Se2 power factor (PF) was higher than that of AgSbSe2, up to ∼0.74 mW m−1 K−2 at 580 K (Fig. S10, ESI†), and slightly higher than previously reported for bulk AgSb0.98Bi0.02Se2 (Fig. S12, ESI†).16
Room temperature Hall measurements provided hole concentrations p = 4 ± 2 × 1019 cm−3 in AgSbSe2 and 5-fold higher for AgSb0.98Bi0.02Se2, p = 2 ± 1 × 1020 cm−3. As in related Cu2SbSe342,43 and AgSbTe244–47 compounds, the origin of the p-type conductivity in AgSbSe2 is found in its defect structure and non-stoichiometry, and it is generally associated with Ag vacancies. The introduction of a small amount of an impurity may perturb the AgSbSe2 lattice and modify this defect concentration. In the particular case of substituting Sb3+ ions by the larger Bi3+ ions a compressive strain is introduced, which we hypothesize can in part be alleviated by a larger density of Ag vacancies and thus an increase of the hole concentration. The confirmation of this hypothesis needs in all cases a careful analysis of the defect formation energies within this compound in the presence of Bi, which is out of the scope of the present work. Besides, the relatively low Hall mobilities measured, 2.5 cm2 V−1 s−1 and 1.9 cm2 V−1 s−1 for AgSbSe2 and AgSb0.98Bi0.02Se2, respectively, are consistent with the large degree of disorder, the nanoscale precipitates and the intrinsically large effective mass of holes in this system. The mobility decrease with the introduction of Bi ions is also consistent with the presence of additional point defects scattering charge carriers.
Very low thermal conductivities (κtotal = λCpρ < 0.5 W m−1 K−1) were measured for both AgSbSe2 and AgSb0.98Bi0.02Se2 compounds in the whole studied temperature range of 300 to 700 K (Fig. 7c). These low thermal conductivities can be explained by an efficient multi-level phonon scattering; that is by nanometer scale precipitates, by the highly disordered lattice and by the strong phonon–phonon interactions from the high degree of anharmonicity of the Sb–Se bonds in this material.18 In spite of the higher density of defects associated with the presence of Bi, thermal conductivities were slightly higher for AgSb0.98Bi0.02Se2 due to the larger contribution from the electronic thermal conductivity (Fig. 7c inset).
As a result, ZT values of up to 1.1 at 640 K were obtained for AgSb0.98Bi0.02Se2 (Fig. 7d), which represents a two-fold increase over pristine AgSbSe2, and is among the best values obtained for a Pb- and Te-free material at this temperature (Table S1, ESI†).
Footnote |
† Electronic supplementary information (ESI) available: Additional experimental details and TEM, XRD, EDX, and NMR data. See DOI: 10.1039/c6tc00893c |
This journal is © The Royal Society of Chemistry 2016 |