Colloidal Nanocrystals with Molecular Metal Chalcogenide Surface Ligands

See allHide authors and affiliations

Science  12 Jun 2009:
Vol. 324, Issue 5933, pp. 1417-1420
DOI: 10.1126/science.1170524


Similar to the way that atoms bond to form molecules and crystalline structures, colloidal nanocrystals can be combined together to form larger assemblies. The properties of these structures are determined by the properties of individual nanocrystals and by their interactions. The insulating nature of organic ligands typically used in nanocrystal synthesis results in very poor interparticle coupling. We found that various molecular metal chalcogenide complexes can serve as convenient ligands for colloidal nanocrystals and nanowires. These ligands can be converted into semiconducting phases upon gentle heat treatment, generating inorganic nanocrystal solids. The utility of the inorganic ligands is demonstrated for model systems, including highly conductive arrays of gold nanocrystals capped with Sn2S64– ions and field-effect transistors on cadmium selenide nanocrystals.

Inorganic colloidal nanocrystals (NCs) with precisely controlled compositions and morphologies have shown useful physical and chemical properties (1) and have found applications in light-emitting devices (2), photodetectors (3), solar cells (4), and other devices. Despite progress in NC synthesis, fabrication of competitive solid-state devices from solution-processed colloidal building blocks remains challenging. In a NC solid, where each individual NC carries size-dependent properties of the respective metal, semiconductor, or magnet, the transport of charges is dominated by the interparticle medium. The most successful synthetic methodologies developed for colloidal nanomaterials use surface ligands to stabilize the particles with long (C8 to C18) hydrocarbon chains (5) or bulky organometallic molecules (6). These large molecules create highly insulating barriers around each NC. Complete removal of surface ligands has proven to be difficult and can create surface dangling bonds and charge-trapping centers (7). In a few successful examples, ligand exchange by mild chemical treatment of PbSe NC films with dilute hydrazine solutions (8, 9) or by linking CdSe NCs with 1,4-phenylenediamine (10) yielded conductive NC solids with carrier mobilities comparable to those of solution-processed organic semiconductors. At the same time, dynamic changes of volatile, easily oxidizable small linking molecules impart instabilities in the electronic properties of conductive NC solids.

It would be beneficial to design surface ligands for colloidal nanostructures that (i) adhere to the NC surface and provide colloidal stabilization, (ii) provide stable and facile electronic communication between the NCs, and (iii) constructively supplement the properties of the NC solid. We propose a generalized approach that is compatible with existing methodology for NC synthesis and is based on exchange of the original organic ligands with molecular metal chalcogenide complexes (MCCs). MCCs provide colloidal stabilization of various nanostructures while enabling strong electronic coupling in the NC solids.

Some molecular MCCs are known as precursors for mesoporous metal chalcogenides (1113) and semiconducting films with high carrier mobility (14, 15). Many MCCs can be synthesized by dissolution of bulk main group or transition metal chalcogenides in hydrazine (14). Excess chalcogen is usually added to form soluble anionic species such as Sn2S64– (14) with hydrazinium (N2H5+) as the counterion. Various structures including covalent (N2H4)2ZnTe (16), layered N4H9Cu7S4 (17), and mixed-metal (N2H4)5SnS4Mn2 (18) are accessible by this approach. Hydrazine-stabilized MCCs, whose exact molecular structure in soluble form has yet to be identified, can be prepared for many other metal chalcogenides including Ga2Se3, Sb2Se3, Sb2Te3, CuInSe2, CuInxGa1–xSe2, and HgSe (19).

To prepare MCC-capped NCs (Fig. 1A), we developed a simple ligand-exchange procedure based on phase transfer of NCs from a nonpolar organic medium into a polar solvent such as hydrazine or dimethyl sulfoxide (DMSO) (19). Typically, a solution of MCC in anhydrous hydrazine (~1 to 5 μmol/ml) was stirred with NCs dissolved in hexane (1 to 20 mg/ml) until the organic phase turned colorless and a stable colloidal solution of NCs in hydrazine was formed. The ligand-exchange process was greatly facilitated by the nucleophilic nature of MCCs, the electrophilicity of undercoordinated metal atoms at the NC surface, and the ability of hydrazine to efficiently strip off the hydrocarbon ligands (8). Infrared (IR) spectroscopy, elemental analysis, and nuclear magnetic resonance spectroscopy revealed the absence of hydrocarbon species in MCC-capped NCs, suggesting complete exchange of original ligands during the phase transfer (19). The generality of this approach is illustrated by capping 3.6-nm CdSe NCs with various MCCs (Fig. 1B) as well as by combining a given MCC ligand, Sn2S64–, with various NCs (Fig. 1C). Notably, nanomaterials of virtually any shape (dots, rods, tetrapods, nanowires) and size (from dots 1.5 nm in diameter to wires 25 nm in diameter and 5 μm in length) could be solubilized by MCCs to form stable colloidal solutions. Below we report characteristic features of MCC-capped NCs, using Au and CdSe as representative examples of metal and semiconductor NCs.

Fig. 1

(A) Sketch of a CdSe NC capped with Sn2S64– ions. (B) Stable colloidal solutions of 3.6-nm CdSe NCs capped with various metal chalcogenide complexes in hydrazine. (C) From left to right: stable colloidal solutions obtained by combining (N2H5)4Sn2S6 with 3.6- and 5.8-nm CdSe NCs, 9-nm CdTe NCs, CdS nanorods (NRods; ~60 nm wide, 5 nm long), CdSe nanowires (NWires; ~25 nm wide, 2 to 5 μm long), Bi2S3 NRods (~25 nm wide, 6 nm long), 5-nm Au NCs, and 3-nm Pd NCs. (D) From left to right: CdSe-Sn2S64– NCs dispersed in DMSO, ethanolamine (EA), formamide (FA), and water. (E) Photograph of NC photoluminescence excited by an ultraviolet lamp. From left to right: 2.9-nm CdSe-Sn2S64– NCs in ethanolamine [green emission, quantum yield (QY) ≈ 2%], 3.6-nm CdSe-Sn2S64– in DMSO (yellow emission, QY ≈ 3%), and Sn2S64–-capped CdSe-CdS core-shell NRods in hydrazine (red emission, QY = 10%). (F) ζ-Potential measured for Sn2S64–-capped 8-nm CdSe NCs in hydrazine. (G) Size histogram obtained from dynamic light scattering for 8-nm CdSe-Sn2S64– NCs in hydrazine.

Measurements of electrophoretic mobility and ζ-potential indicated that in the presence of (N2H5)4Sn2S6, both CdSe and Au NCs are negatively charged (Fig. 1F and fig. S4) as a result of bonding of negatively charged MCC ions to the NC surface. The charging prevents aggregation of NCs and stabilizes the colloidal solutions. Dynamic light scattering revealed single-particle populations in colloidal solutions (Fig. 1G and fig. S4). Solvents other than hydrazine can be used for handling MCC-capped NCs in applications where chemical compatibility or toxicity of hydrazine is a concern. Figure 1D shows colloidal solutions of Sn2S64–-capped CdSe NCs in water, DMSO, formamide, or ethanolamine (19). ζ-Potential measurements carried out for various NCs dispersed in different solvents and stabilized with different MCCs suggest that negative charging of the NC surface is a general phenomenon responsible for colloidal stabilization of MCC-capped NCs (fig. S4).

Transmission electron microscopy (TEM) showed that MCC-capped NCs retain their size and shape in hydrazine phase (Fig. 2 and fig. S5). The ability of monodisperse NCs to form periodic structures (superlattices) was also preserved, as shown for two- and three-dimensional superlattices of 5-nm Au NCs capped with Sn2S64– (Fig. 2B and fig. S6). Typically, the assembly of MCC-capped particles requires rigorous control of the surface charges through adjustment of ionic strength and pH and may potentially lead to the design of binary superlattices stabilized by electrostatic interactions (20).

Fig. 2

TEM images of selected examples of NCs capped with Sn2S64– ligands: (A) 8.1-nm CdSe NCs, (B) face-centered cubic superlattice of ~5-nm Au NCs, and (C) CdSe nanowires ~25 nm wide and 2 to 5 μm long. Inset shows a magnified view of a single bundle of CdSe nanowires. (D) CdSe-CdS nanotetrapods.

Both CdSe and Au NCs capped with Sn2S64– preserved their size-dependent optical absorption features (figs. S7 and S8). In polar organic solvents such as DMSO, formamide, and ethanolamine, Sn2S64–-capped CdSe NCs exhibited size-dependent excitonic photoluminescence with a quantum efficiency of several percent, comparable to that of as-prepared organics-capped NCs (Fig. 1E and figs. S9 and S10), indicative of a good passivation of surface states by MCC ligands. In neat hydrazine, the luminescence of CdSe NCs was quenched, whereas CdSe-CdS core-shell nanorods and tetrapods exhibited ~10% quantum yields (Fig. 1E).

An important feature of MCCs is their transformation into amorphous or crystalline metal chalcogenides upon mild thermal treatment (1418). For example, Sn2S64– neutralized by volatile hydrazinium counterions decomposes at 180°C: (N2H5)4Sn2S6 → SnS2 + 4N2H4 + 2H2S (14). Because of the absence of any nonvolatile or carbon-containing impurities, the SnS2 phase crystallizes into pure electronic-grade semiconductor (14). The thermogravimetric scans for 3.6-nm CdSe NCs capped with Sn2S64– ions generally follow thermogravimetric scans for pure (N2H5)4Sn2S6 but reach the plateau at slightly lower temperatures, about 160°C (Fig. 3A). The ligand conversion process involves a total weight loss of only 3.8%, favorable for producing continuous crack-free NC solids. The elemental analysis (fig. S11) is consistent with thermogravimetric data and confirms that typically 2 to 10 weight % of MCCs is sufficient to stabilize colloidal NCs in solution (19). After heating to 180°C, we observed complete disappearance of absorption features in IR spectra, confirming the absence of entities containing C–H, S–H, and N–H bonds (Fig. 3B and figs. S12 and S13). In control experiments on Au and CdSe NCs prepared using dodecanethiol and ODPA-HDA-TOPO (octadecylphosphonic acid, hexadecylamine, and trioctylphosphine oxide), correspondingly, the IR spectra remained nearly unaltered after 180°C annealing (fig. S14). X-ray diffraction and TEM studies revealed substantial stability of MCC-capped CdSe and Au NCs against sintering into a bulk phase (fig. S15).

Fig. 3

(A) Thermogravimetric scans for (N2H5)4Sn2S6 (dashed line) and for 3.6-nm (N2H5)4Sn2S6-capped CdSe NCs (solid line). (B) Fourier-transform infrared spectra for 3.6-nm CdSe NCs capped with long-chain organic ligands (red line) and with (N2H5)4Sn2S6 ligands before (blue line) and after (green line) annealing at 180°C. The IR spectra were normalized to the amount of absorbing material (19) and are vertically shifted for clarity. The assignment of peaks is given in (19). (C) Absorption spectra collected using an integrating sphere for thin films composed of 4.6-nm CdSe NCs capped with original organic ligands (red line) and with (N2H5)4Sn2S6 ligands before (blue line) and after (green line) annealing at 180°C.

When two or more metallic or semiconducting nanoparticles are in close proximity to each other, their wave functions can couple forming states delocalized over several NCs or propagating throughout the entire NC solid. The quantum mechanical coupling energy can be approximated as β ≈ hΓ ≈ exp[–(2mE/ℏ2)1/2Δx], where h is Planck’s constant, Γ is the tunneling rate between two NC neighbors, m* is the carrier effective mass, ℏ is Planck's constant divided by 2π, and ΔE and Δx are the height of the tunneling barrier and the shortest edge-to-edge distance between the NCs, respectively (21, 22). Replacement of bulky insulating hydrocarbon chains with much smaller and more conductive MCCs should strongly reduce both ΔE and Δx, thereby facilitating electronic communication between the NCs. We probed the coupling between MCC-capped NCs by optical absorption and charge transport measurements. Thus, for 4.6-nm CdSe NCs capped with conventional hydrocarbon ligands, the absorption spectra of colloidal solution and close-packed films are very similar to each other, indicative of strong localization of electron and hole wave functions on individual NCs (Fig. 3C and fig. S16). In contrast, the excitonic features in close-packed films of CdSe NCs capped with Sn2S64– showed a pronounced 46-meV red shift relative to individual NCs in solution (Fig. 3C and fig. S16). Such a shift implies partial leakage of the wave functions into neighboring NCs that relaxes the quantum confinement. Conversion of the Sn2S64– ligands to SnS2 at 180°C lowers the barrier height ΔE, leading to an additional 34-meV red shift of the optical transitions in the NC solid while maintaining the excitonic features and quantum confinement (Fig. 3C). Similarly, Au NCs capped with Sn2S64– showed a pronounced plasmonic absorption peak around 535 nm in solution, which completely disappeared in NC films (fig. S17A), indicating strong delocalization of the electronic states (23). Note that all observed spectral changes were reversible; the plasmonic peak appeared again after redissolution of the NC film. The reference samples of dodecanethiol-capped Au NCs showed strong plasmonic absorption in solutions and in close-packed films (fig. S17B).

For charge transport studies, we used highly doped Si wafers with a 110-nm layer of thermal oxide and lithographically patterned Ti-Au electrode structures (19). The Si substrate was used as the back gate electrode for field-effect transistor (FET) measurements. Close-packed NC films were deposited on these substrates by spin coating or dropcasting. The film thickness was measured using atomic force microscopy profiles and cross-sectional scanning electron microscopy (SEM) studies. The amount of MCC ligands was kept below 10 weight %, sufficient to provide colloidal stabilization but insufficient to form any continuous conductive channels of phase-separated metal chalcogenide, as discussed below.

The original organic ligands rendered NC films highly insulating, with conductivities (σ) on the order of ~10−9 S cm−1 for 5-nm Au NCs (Fig. 4D) and less than 10−12 S cm−1 for 5.5-nm CdSe NCs. Replacing dodecanethiol ligands with Sn2S64– increased the conductivity of Au NC solids by ~11 orders of magnitude, approaching σ values of ~200 S cm−1 (Fig. 4D). (N2H5)4Sn2S6 films could not support efficient charge transport (fig. S18), indicating that high conductivity resulted from the NCs. After electrical measurements, the film could be easily dissolved in hydrazine or H2O, disassembling into individual NCs. Comparison of TEM images before and after electrical measurements revealed no changes in NC size and shape, ruling out sintering as a possible reason for the observed very high conductivity. To the best of our knowledge, the highest conductivity reported for Au NC solids with short-chain organic capping (e.g., n-ethyl or n-butanethiol) was less than 10−1 S cm−1 (24). The disappearance of the plasmonic absorption peak strongly supports the metallic nature of Sn2S64–-capped Au NC solids. A Mott-type metal-insulator transition was predicted to occur in the Au NC solids when the interparticle spacing is reduced below ~1.1 nm (24). Our TEM studies revealed a strong decrease in the mean interparticle distance from ~1.6 nm for dodecanethiol-capped Au NCs to less than 0.5 nm for Sn2S64–-capped Au NCs (Fig. 4, A to C). Relative to other prospective solution-processable materials, the conductivity of Sn2S64–-capped Au NC solids is much higher than the conductivities of solution-cast conducting polymers (25) and graphene-based composites (26) and is lower than the conductivity reported for a dense network of single-walled carbon nanotubes (27).

Fig. 4

(A) TEM image of an array of ~5-nm Au NCs capped with dodecanethiol. (B) TEM image of a layer of ~5-nm Au NCs capped with (N2H5)4Sn2S6. (C) TEM image of a three-dimensional superlattice of ~5-nm Au NCs capped with (N2H5)4Sn2S6. (D) Current-voltage (I-V) scans for a film of dodecanethiol-capped 5-nm Au NCs (open circles) and for a film of the same Au NCs capped with (N2H5)4Sn2S6 (black squares). Dashed arrows show the voltage scan direction. (E) Plot of drain current ID versus drain-source voltage VDS for a NC FET with a channel composed of 4.5-nm CdSe NCs capped with (N2H5)4Sn2S6 and annealed at 200°C (channel length 10 μm, width 3800 μm, 110-nm-thick SiO2 gate dielectric).

MCC capping of colloidal NCs is a promising approach to designing solution-processed inorganic semiconductors. Figure 4E and fig. S19 show characteristics of a field-effect transistor with a channel assembled of 4.5-nm CdSe NCs capped with Sn2S64– and annealed for a short time at 200°C. We observed an n-type gate effect with current modulation Ion/Ioff ≈ 105 and electron mobility μ ≈ 3 × 10−2 cm2 V−1 s−1 in the saturation regime (19). Similar results were obtained for CdSe NCs capped by Sn2Se64– ions (fig. S20). Higher mobilities should be achievable by replacing Au electrodes by a metal with lower work function and by using FETs with top-electrode configuration (14, 19). For reference, μ ≈ 10−2 cm2 V−1 s−1 has been previously achieved for CdSe NC films at very high doping densities by means of electrochemical gating (10, 28). Melting and sintering pyridine-capped CdSe NCs into a polycrystalline film by high-temperature anneal yielded μ ≈ 1 cm2 V−1 s−1 (29). Relative to those results, MCC-capped CdSe NCs enabled good performance of solid-state FETs while retaining optical and electronic tunability provided by the quantum confinement. Illumination of MCC-capped CdSe NCs increased their conductivity by several orders of magnitude, as shown for Sn2Se64–-capped CdSe NCs (fig. S21). We used small concentrations of MCCs to observe transport through “pure” quantum dot solids. At high concentrations of MCC component, highly conductive continuous channels of metal chalcogenide can form between the NCs. Indeed, we observed an abrupt increase of conductivity in composite materials made of CdSe NCs and (N2H5)4Sn2Se6 when the CdSe-to-SnSe2 ratio approached 1:1 by weight (fig. S20). Complementary x-ray diffraction studies revealed the appearance of the SnSe2 phase at this composition (fig. S22).

In the previous examples, MCC ligands behaved as electronically transparent “glue” for NCs. However, they can be used for creating composite materials where the properties of MCC and NC components complement each other. Thermal decomposition of the hydrazinium-based MCCs can generate various chalcogenide phases with n- and p-type conductivity, phase-change properties (15), etc. (fig. S23). For example, combining electron-conducting nanowires (e.g., CdS) with hole-conducting hosts (e.g., CuIn1–xGaxSe2) will form materials with distributed networks of p-n junctions. Future work can certainly benefit from rational design and optimization of MCC-NC pairs. The possibility of all-solution device fabrication makes this approach appealing for large-area, roll-to-roll applications such as manufacture of thin-film solar cells.

Supporting Online Material

Materials and Methods

Figs. S1 to S23

Tables S1 and S2


References and Notes

  1. See supporting material on Science Online.
  2. We thank D. B. Mitzi and P. Guyot-Sionnest for stimulating discussions, M. Bodnarchuk and J. Huang for synthesis of Au nanoparticles and CdSe/CdS tetrapods, E. Shevchenko and I. Steele for help with energy-dispersive x-ray and SEM measurements, J.-S. Lee and R. Divan for help with electrical measurements, the Analytical Chemistry Laboratory at Argonne National Laboratory for elemental analysis, and F. Stafford for reading the manuscript. Supported by Deutsche Studienstiftung (M.S.), American Chemical Society Petroleum Research Fund grant 48636-G10, the Chicago Energy Initiative, and Evident Technologies Inc. Work at the Center for Nanoscale Materials, Argonne National Laboratory, was supported by the U.S. Department of Energy under contract DE-AC02-05CH11357.
View Abstract

Stay Connected to Science

Navigate This Article