Letter pubs.acs.org/NanoLett
Chemically Tailored Dielectric-to-Metal Transition for the Design of Metamaterials from Nanoimprinted Colloidal Nanocrystals Aaron T. Fafarman,†,∥ Sung-Hoon Hong,†,∥ Humeyra Caglayan,† Xingchen Ye,‡ Benjamin T. Diroll,‡ Taejong Paik,‡ Nader Engheta,† Christopher B. Murray,§,‡ and Cherie R. Kagan*,†,‡,§ †
Department of Electrical and Systems Engineering, ‡Department of Chemistry, and §Department of Materials Science and Engineering, University of Pennsylvania, Philadelphia, Pennsylvania 19104, United States S Supporting Information *
ABSTRACT: We demonstrate optical metamaterial design using colloidal gold nanocrystal building blocks. In the solid state, chemically exchanging the nanocrystals’ surface-capping molecules provides a tailorable dielectric-to-metal transition exhibiting a 1010 range in DC conductivity and dielectric permittivity ranging from everywhere positive to everywhere negative throughout the visible-to-near-IR. Direct, wide-area nanoimprinting of subwavelength superstructures at room temperature, on plastic and glass substrates, affords plasmonic resonances ranging from 660 to 1070 nm, in agreement with numerical simulations. KEYWORDS: Plasmonics, soft lithography, gold nanoparticles, ammonium thiocyanate, ligand exchange, dielectric function
S
size, shape, and stability of the NCs, remain present on their surface once assembled in the solid state. These ligands enforce a well-defined and large separation (1−2 nm) between NCs, and consequently NC solids derived from colloids are highly resistive insulators in electrical transport measurements.15 In pioneering work, Heath and co-workers observed evidence for a reversible insulator-to-metal transition upon compression of a film of colloidal metal NCs on a Langmuir−Blodgett trough.16 However, the means to harness this control over interparticle spacing for the fabrication of permanent, optical quality plasmonic structures has been elusive. Chemical exchange processes in the solid state can permanently replace long capping groups with shorter ones,15,17−20 and with the recent discovery by us21 and others22 of very compact inorganic ligands, the range of programmable resistivity values has been dramatically increased. In this work, we fabricate optical quality thin-films and nanoscale superstructures from NC building blocks. By ratcheting down the interparticle spacing within the solid, we realize a dielectric-to-metal transition by design, allowing us to engineer the properties of NC superstructures from nonplasmonic insulators to plasmonic, bulk-like conductors. Colloidal NC materials present tremendous technological opportunities as they can be deposited like inks by simple fabrication techniques such as spincoating, spraycasting, and dipcoating, or patterning techniques such as inkjet printing and nanoimprint lithography. Nanoimprint lithography in particular makes it possible to print these inks into complex,
ynthetic metamaterials offer an unprecedented ability to engineer the light−matter interaction using combinations of plasmonic and nonplasmonic materials, patterned into structures that are smaller than the operative wavelength of light.1−3 Metamaterials can harness light’s energy for photovoltaics,4,5 control its propagation for photolithography,6 biosensing,7 and cloaking8,9 and manage its information content in the form of optical circuits.10,11 Metamaterial properties depend critically on the optical dispersion of the constituent materials. For the plasmonic components the coinage metals, copper, silver, and gold are commonly employed; however, the optical properties of these materials are fixed, restricting the wavelengths of operation and excluding them from many applications where a smaller negative permittivity value is required.12,13 Consequently, new materials are being sought, with optical properties optimized for various existing and emerging applications. This search has begun to take a broad view of the periodic table, evaluating pure metals, alloys, and highly doped semiconductor compounds, or in the case of graphene, exploring the configuration of chemical bonds.14 Here, we present a new and complementary axis of explorationwe synthesize plasmonic materials from colloidal metal nanocrystal (NC) building blocks and intentionally design the optical properties by controlling the electronic coupling between NCs. Colloidal metal NCs are ubiquitous in the field of plasmonics due to their size- and shape-dependent localized surface plasmon resonances (LSPRs). They have been largely unexplored, however, as discrete building blocks for larger, complex, plasmonic superstructures because, without modification, colloidal NC solids act as dielectric materials. Long-chain molecular capping groups, necessary to control the © 2012 American Chemical Society
Received: August 24, 2012 Revised: November 30, 2012 Published: December 10, 2012 350
dx.doi.org/10.1021/nl303161d | Nano Lett. 2013, 13, 350−357
Nano Letters
Letter
Figure 1. Chemically controlled Au NC thin films. Photographs of spincast Au NC thin films on 0.75″ diameter optical flats, cartoons, and TEM images with different molecular capping groups. (a−c) Oleylamine (OLA). (d−f) Ethanedithiol (EDT). (g−j) Ammonium thiocyanate (SCN). Photographs (a, d, g) show 0.75″ diameter quartz rounds viewed perpendicularly (left) or obliquely (right). Note: TEM images (c, f, i) of monolayer samples were chosen to highlight the impact of ligand exchange on interparticle spacing, whereas in thicker samples, infilling of NCs from upper layers prevents the formation of void space as seen in j, maintaining smooth, optical quality thin films upon ligand exchange.
of a majority of metallic and semiconducting NCs use surfactants both for colloidal dispersibility and to direct size and shape monodispersity during synthesis, characteristics seen in the transmission electron microscopy (TEM) image of a monolayer prepared by dropcasting in Figure 1c. However, these bulky ligands, seen only as bright regions of high transmission by TEM, prevent the close approach of the NCs, leading to poor electronic coupling. We performed a postdeposition exchange15 of the OLA ligands with more compact alternatives: the well-studied ethanedithiol (EDT) 17−20 ligand or a new NC ligand, ammonium thiocyanate (SCN), we recently reported.21 Using Fourier transform infrared (FTIR) spectroscopy, the ligand-exchange reaction was observed to go to completion (Supporting Information, Figure S2). Ligand treatment preserved the smooth, optical quality surface exhibited in spincast films, as seen in Figure 1d for an EDT treated NC solid. The cartoon in Figure 1e depicts the decreased interparticle spacing upon exchange with EDT, which was evident by TEM (Figure 1f) in the contraction of the gaps. Cracking of the monolayer films is due to the lost ligand volume. Monolayer regions of the TEM sample were chosen to accentuate the changes in interparticle spacing for the different ligands; however the cracking observed for monolayers contrasts with what is observed for multilayers. For example, for films with an original height of 100 nm, the smooth optical quality of the solid is maintained after exchange due to in-filling with NCs from upper layers, as confirmed by AFM which demonstrated a collapse of the film height but no increase in surface roughness (Supporting Information, Figure S1). When the NC solid was spincast and subsequently treated with NH4SCN, the optical quality of the film was similarly preserved, while the interparticle spacing was even further reduced (Figure 1g−h). In Figure 1i the majority of the NCs showed no discernible separation from their neighbor, and in
subwavelength superstructures over wide areas by simply molding the fluid on the substrate with a patterned stamp.23−26 Nanoimprinting is not constrained by the diffraction limit like traditional photolithography (i.e., in the absence of interference27−29 or superlensing6) and is much less costly than electron- and focused ion-beam lithography.28 It can be used for roll-to-roll fabrication,30 opening the door to widearea patterning on flexible substrates. Here, we unite the technological advantages of direct printing25,26 with precise chemical control over the optical properties of the NC solid to realize complex nanoimprinted plasmonic and nonplasmonic superstructures on glass and plastic substrates. Experimentally determined dielectric functions of these engineered NC solids were critical for accurate classical finite-difference time-domain (FDTD) simulations, which predict the optical properties of nanoimprinted superstructures. Nanoimprinting of NC superstructures and chemical exchange to control NC coupling allows for the realization of plasmonic structures and metamaterials, over wide areas, in a simple, low-cost process without any heat applied, making it compatible with fine feature sizes and a great variety of substrates. To understand the optical and electrical properties of these NC solids as a function of chemically exchanging the surface capping groups, we first fabricated thin films by spincasting a dispersion of as-synthesized 10 nm diameter Au NCs in octane. The resultant NC solids were smooth and uniform, with a surface roughness of 8 nm in a typical 100 μm2 area as characterized by atomic force microscopy (AFM; see Supporting Information, Figure S1). When viewed obliquely, they were highly reflective and gold-colored to the eye; when viewed perpendicularly they were semitransparent (Figure 1a). As synthesized, each NC was capped with a dense coating of the long-chain hydrocarbon surfactant oleylamine (OLA), depicted in the cartoon in Figure 1b. Nonaqueous syntheses 351
dx.doi.org/10.1021/nl303161d | Nano Lett. 2013, 13, 350−357
Nano Letters
Letter
Table 1. Resistivity of Spincast or Thermally Evaporated Au Films conditions resistivity (Ω·cm)
pristine ∼106 (2-point)
EDT 0.3 (±0.2)
SCN 5.1 (±0.8) × 10−5
thermally evaporated 4.8 (±0.4) × 10−6
Figure 2. Optical properties of spincast and thermally evaporated Au films. Top: transmittance spectra at normal incidence (black) and reflectance spectra at 45° (red). Bottom: real (black) and imaginary (red) parts of the dielectric functions versus wavelength. (a−b) Oleylamine (OLA)-capped, (c−d) ethanedithiol (EDT)-capped, and (e−f) thiocyanate (SCN)-capped Au NC solids. (g−h) Thermally evaporated Au thin film.
some cases appeared to have fused into larger aggregates. Small and wide-angle X-ray studies of these solids (Supporting Information, Figure S3) corroborated this picture, indicating a decrease in the interparticle reflection intensity and an approximate doubling in grain size, respectively, for SCNtreated samples. We stress that these samples were processed under strictly room temperature conditions and all changes were due only to chemical treatment. Thicker films of thiocyanate-exchanged NCs are dense and did not exhibit large void spaces (Figure 1j), as in-filling with NCs from upper layers gives rise to a larger collapse in film height than that for EDT exchange while similarly maintaining low surface roughness. The direct current (DC) sheet resistivity of the Au NC solids with different interparticle spacing provided a stark demonstration of the increase in electronic coupling upon exchange of the NC ligands (Table 1). The long OLA ligands present a significant barrier to electronic transport evidenced by large resistivities of ∼106 Ω·cm in these NC solids. After EDT exchange, four-point probe measurements revealed greater than 6 orders of magnitude decrease in resistivity, while X-ray measurements demonstrated that the NCs remained individual particles, that is, the NCs did not fuse (Supporting Information, Figure S3). Although this resistivity value is slightly lower than all examples of EDT-treated Au NC solids in a recent review,19 it is still 5 orders of magnitude more resistive than thermally evaporated Au films. Temperature-dependent DC resistivity measurements have been widely pursued to examine whether EDT treatment of colloidal Au NC solids induces a Mott− Hubbard insulator-to-metal transition. However, for similar temperature ranges, metallic,20 nonmetallic,18 and metallictransitioning-to-nonmetallic17 conduction have all been observed, suggesting that these solids are right at the cusp of the transition, and highlighting the importance of thorough characterization of the structure of the NC solid, such as that obtained here from TEM and X-ray studies. Below we explore the consequences of this interesting resistivity regime, intermediate between insulating and highly conducting, for
the optical properties of this NC solid and its use as a building block for nanoimprinting of optical metamaterials. Remarkably, SCN treatment decreased the resistivity of Au NC solids to within only a factor of 10 from a reference sample of an evaporated Au thin film, a 10 orders-of-magnitude reduction from the OLA-capped precursor from which it was derived (Table 1). Thus, with entirely room temperature, solution-based processing, we were able to fabricate Au NC thin film solids with a resistivity close to that found in thermally evaporated Au thin films. We turned next to spectroscopic measurements to observe how conduction currents induced electrostatically in these NC solids correlate with displacement currents induced optically. The smooth, optical quality of the spincast Au NC solids, preserved by chemical treatment, was critical for wide-area (several square millimeters) ellipsometric, transmittance, and reflectance spectroscopy. For OLA-capped NCs in Figure 2a, the transmittance (black) exhibited a trough at 566 nm, the same wavelength as a peak in the reflectance (red). This resonance is slightly shifted from the value of the characteristic LSPR of the starting colloidal Au dispersion at 519 nm. This can be understood in either of two equivalent frameworks: as an array of discrete NCs whose LSPRs couple weakly with one another in the solid state, leading to a small red-shift;31 or from the perspective of the LSPR of any single NC, which shifts in frequency due to an increase in the average dielectric function of its environment when condensed into a solid.1 An effective medium approximation (EMA)32,33 allows us to derive the average, macroscopic, dielectric environment within the NC solid (the effective medium) given a microscopic structure consisting of gold inclusions (the nanocrystals) embedded in a hydrocarbon dielectric medium (the ligands). By modeling the complex reflectance ratio measured by ellipsometry with an EMA, we extracted the optical dielectric function shown in Figure 2b (details of the modeling are provided in Methods and plots of the complex index of refraction in Supporting Information, Figure S4). The real part (black curve) of the complex dielectric function was everywhere positive through the visible to near-IR, consistent with the insulating nature of 352
dx.doi.org/10.1021/nl303161d | Nano Lett. 2013, 13, 350−357
Nano Letters
Letter
the NC solid, while the imaginary part had a feature coinciding with the LSPR seen in transmittance spectra. Transmittance (black) and reflectance (red) spectra of an EDT-capped Au NC solid in Figure 2c similarly showed evidence of a LSPR associated with individual NCs; however, the resonance was downshifted and broadened substantially, indicating the existence of strong plasmonic coupling between NCs, or equivalently that the dielectric properties of the effective medium are increasingly dominated by those of the metal inclusions, in contrast to the example with OLA, introduced above. As in the previous example, the EMA was applied to the ellipsometric data, yielding the dielectric function (Figure 2d). The imaginary part of the complex dielectric function (red) showed a feature coinciding with the downshifted and broadened LSPR. The real part of the complex dielectric function (black) exhibited an interesting property: a small excursion into negative values in the visible portion of the spectrum. Comparing the optical dielectric function, the intermediate values of DC resistivity (Table 1) and the literature precedent for a Mott−Hubbard insulator−metal transition17−20,34 at or near the interparticle spacing of EDT capped Au NC solids provided a consistent description of the onset of an optical dielectric-to-metal transition. This onset occurred as the interparticle spacings accessed by ligand exchange are
