Vous êtes sur la page 1sur 9

J. Phys. Chem.

B 1997, 101, 189-197

189

Pressure/Temperature Phase Diagrams and Superlattices of Organically Functionalized Metal Nanocrystal Monolayers: The Influence of Particle Size, Size Distribution, and Surface Passivant
James R. Heath,* Charles M. Knobler, and Daniel V. Leff
UCLA Department of Chemistry and Biochemistry, 405 Hilgard AVenue, Los Angeles, California 90095-156905 ReceiVed: April 22, 1996; In Final Form: August 23, 1996X

The phase behavior of organically passivated 20-75 diameter Ag and Au nanocrystals is investigated by examining surface-area isotherms of Langmuir monolayers and transmission electron micrographs of Langmuir-Blodgett (LB) films. The effects of temperature, organic passivant chain length, and nanocrystal size and composition are studied. Three distinct types of phase behavior are observed and may be classified in terms of the extra (conical) volume (Ve) available to the alkyl capping group as it extends from a nearly spherical metal core. For Ve > 350 3, the phase diagram is dominated by extended, low-dimensional structures that, at high pressures, compress into a two-dimensional foamlike phase. This behavior is rationalized as originating from the interpenetration of the ligand shells of adjacent particles. For Ve < 350 3, dispersion attractions between the metal cores dominate particle condensation. For 350 3 > Ve > 150 3, the particles condense to form closest packed structures, which, for sufficiently narrow particle size distributions, are characterized by crystalline phases. For Ve 30 3, the particles irreversibly aggregate into structures similar to those expected from a diffusion-limited-aggregation (DLA) model. Optical properties of certain LB films of the closest packed phases are reported.

Introduction Chemical techniques for the preparation and size separation of organically soluble semiconductor and metal nanocrystals have advanced rapidly over the past several years. Narrow size distributions for particles with diameters in the range of 1-20 nm of a wide variety of metal and semiconductor particles may now be prepared.1-3 Randomly dispersed semiconductor, metal, and metal oxide nanoparticles have historically found applications in areas such as reprography, photocatalysis, catalysis, and optical materials.4 Recent advances in directing the selfassembly of (disordered) semiconductor nanocrystal thin films5 have enabled researchers to electrically address certain nanocrystals,6 making them viable materials for optoelectronics applications as well. Even more recent, however, has been the development of techniques for directing the self-assembly of nanocrystals into ordered aggregates, or quantum dot superlattices.7 These assemblies present some very exciting possibilities. In principle, interparticle separations, particle size, and particle stoichiometry may be individually manipulated to produce a macroscopic solid with a tailored band structure, similar to the well-known case of one-dimensional quantumwell superlattices. For metal particles in particular, quantum dot superlattices might be engineered to exhibit unique magnetic, electronic, and/or superconducting behavior. In addition, they should provide nearly ideal models for the well-studied granular metals. For the case of II-IV semiconductor quantum crystallites, Bawendis group has explored various techniques to investigate the formation of quantum dot superlattice structures.8 In one study, they utilized the Langmuir-Blodgett procedure to prepare ordered arrays of trioctylphosphine oxide-capped CdSe nanocrystals.9 These particles may be prepared with extremely narrow ( < 4%) size distributions,10 and experiments with five different size particles between 25 and 53 were performed at
* Author to whom correspondence should be addressed. X Abstract published in AdVance ACS Abstracts, December 15, 1996.

a single temperature. In all cases, the particles self-assembled upon compression to form closest packed two-dimensional arrays. The only size-dependence observed was related to the pressure per particle area necessary to compress the particles into an array. An alternative technique for the fabrication of ordered monolayers of ligand-stabilized cluster compounds has recently been reported by Schmids group.11 In that work, they utilize strong attractive interactions between polyelectrolytes, in which imino groups on a substrate and sulfonic groups on cluster surfaces were used to promote the self-assembly of the acid of [Au55{PPh2(m-C6H4SO3Na)12}Cl6] on a mica surface coated with poly(ethylenimine). We have also recently reported on the formation of metal nanocrystal superlattices, or opals,12 from an initially broad size distribution of dodecanethiol-capped 20-55 diameter Au particles.13 In that paper, we demonstrated that dispersion interactions between particles can lead to size-dependent phase separations followed by size-selective 2-D and 3-D opal formation. For any nanocrystal system, interparticle dispersion attractions are expected to scale geometrically with increasing particle size.14 For metal nanocrystals in particular, these attractions may be quite strong15 and exhibit a profound and often organizing influence on particle aggregation. The nature of these interactions implies that, for metal particles, superlattice formation should depend on particle size and stoichiometry, the nature and size of the organic passivants on the particle surfaces, and the particle size distribution. Such expectations contrast sharply with what is observed in the fabrication of more traditional two-dimensional colloidal arrays, e.g. arrays of polystyrene, latex, or silica spheres.16 Such particles are typically designed to behave as hard spheres, and superlattice formation and/or size separation is thought to proceed through either hydrodynamic or capillary forces17 or through entropic depletion interactions.18 In this paper, we investigate the role that the variables of particle size, size distribution, and size of the passivating ligand 1997 American Chemical Society

S1089-5647(96)01158-3 CCC: $14.00

190 J. Phys. Chem. B, Vol. 101, No. 2, 1997 play in determining the structures resulting from metal particle condensation in two dimensions. Surface-pressure isotherms of films spread at the air/water interface were utilized to determine the phase diagrams corresponding to the organization of various organically passivated Au and Ag nanocrystal systems. The structural nature of certain phases was interrogated by transmission electron microscopy (TEM) of LangmuirBlodgett (LB) films. Certain of the LB films were also optically characterized using UV/vis absorption spectroscopy. We find that the phase diagrams depend strongly upon the amount of excess (conical) volume (Ve) available to the passivating ligands as they extend from the nearly spherical metal nanocrystal surface. For Ve > 350 3 (case I), the phase diagrams are dominated by extended, 1-D structures that, at high pressures, compress into a 2-D foamlike phase. These structures result from the interpenetration of the ligand shells of adjacent particles. For Ve < 350 3 (case II and case III), attractive dispersion forces between the metal cores dominate particle condensation. For Ve > 150 3, these forces are sufficiently weak that, upon compression, the particles condense to form a closest packed 2-D solid. For Ve 30 3 (case III), the particles irreversibly aggregate into nonequilibrium structures similar to those expected from a diffusion-limited-aggregation model. Experimental Section A. Particle Synthesis. All particles were synthesized using chemicals obtained from Aldrich Chemical Co. without further purification except where noted. Particles were structurally and optically characterized by some or all of the following: TEM, X-ray powder diffraction (Cu KR radiation), and UV/vis absorption spectroscopy. More extensive characterizations of these particles are described elsewhere.19 A.1. Synthesis of Dodecanethiol and Nonanethiol-Capped Au Particles. These particles were prepared in the manner developed by Brust et al.20 Briefly, a weighed amount of HAuCl43H2O is dissolved in distilled H2O. The gold salt is then quantitatively transferred into toluene using (octyl)4N+Bras a phase transfer reagent. A measured amount of alkylthiol is added to the mixture. In a separate beaker, a measured amount of NaBH4 reducing agent is dissolved in H2O. While the Au-containing mixture is rapidly stirred, the reducing agent is added, and the reaction is allowed to proceed for several hours. Typical molar ratios for HAuCl43H2O/(octyl)4N+Br-/alkylthiol/ NaBH4 are 1:3:3:10, and a typical initial weight of the gold salt is 200 mg. Upon completion of the reaction, the aqueous layer is removed and the toluene solution is rotary evaporated to a volume of 5 mL. The particles are precipitated with either methanol, ethanol, or acetone. The initial size distribution is broad and may be arbitrarily narrowed (depending on the amount of product available) using the technique of size-selective precipitation.10 All particles used here were selectively precipitated up to six times. A.2. Synthesis of Oleylamine-Capped Au Nanocrystals. The amine-capped Au nanocrystals represent an unusual chemical species since the amine-Au interaction is considered to be quite weak. Nevertheless, the amine-Au interaction in these particles is charge neutral, and the particles are actually kinetically, rather than thermodynamically, stabilized. Other than the fact that these particles possess an 18-carbon atom surface cap (rather than a 12 or 9 atom one), their chemical behavior and stability are identical to that of their thiol-capped counterparts. The synthesis and chemical characterization of these particles is presented elsewhere.21 Their preparation is similar to that for dodecanethiol- or nonanethiol-capped Au nanocrystals, except that the (octyl)4N+Br- phase transfer reagent is not included in the reaction. This has the effect of lowering the product yield.

Heath et al. A.3. Synthesis of Dodecanethiol-Capped Ag Nanocrystals. The dodecanethiol-capped Ag nanocrystals are prepared in a manner similar to the dodecanethiol-capped Au nanocrystals. AgNO3 or AgClO4 is used as a substitute for HAuCl43H2O. B. Measurement of Pressure/Area Isotherms. Pressure/ area isotherms were measured using a Nima Technology Type 611 Langmuir Trough. All measurements were done with 18 M water (pH 5.7), purified with a Millipore Milli-Q UV Plus system. A variable-temperature water circulator cycled water through the body of the trough and allowed for the determination of phase diagrams through the temperature range 15-38 C. For each isotherm, the temperature of the trough was held constant to (0.2 C. The trough was cleaned after each measurement, and fresh material was deposited. Particles were prepared for Langmuir isotherm measurements in the following manner. After synthesis and size selection, a powder of the desired particles was dispersed by sonication in an acetone/ethanol solution and filtered in order to remove any residual organic material. The resulting dry powder was weighed and then dissolved in a known amount of chromatographed hexane to a concentration of 1 mg/mL. The solubility limit of the particles in hexane is somewhere between 10 and 30 mg/mL, depending on particle size and surface passivant. The solution was passed through a 0.2 m pore size filter and stored in clean glassware. The best results are obtained on very fresh particles. Apparently, the particles ripen over time (a few days), resulting in a slight change in the particle size distribution and the release of some surfactant molecules into solution. The result is that isotherms from freshly prepared particles tend to exhibit sharper features than do isotherms from older particles. Practical considerations mandated that some measurements were made on solutions that had been prepared 2-3 days earlier. In that case, the particles were stored in a refrigerator (-20 C) during the period between preparation and measurements. All phase diagrams, however, were measured over the course of single 3-5 h time periods. A glass 100 L syringe was utilized to disperse a known amount of the particles uniformly across the water surface. Depending on the nature of the particles and the concentration of the solution, between 5 and 200 L of material was used. Pressure/area isotherm measurements were carried out using double-barrier compression at a compression rate of 50 mm/ min. C. TEM and UV/vis Characterization. TEM analysis was performed on an Akashi EM002 TEM operating at 200 keV with either 0.17 or 0.30 nm point-to-point resolution (depending on the installed tilt-axis) at the University of Southern Californias Center for Electron Microscopy. UV/vis spectra were measured on a Perkin-Elmer Lambda 3B spectrophotometer. Langmuir monolayers were transferred to TEM grids in two different ways. For certain of the low-density particle films, the edge of a grid was clamped with tweezers and lowered through the clean water surface at right angles to the surface. The LB nanoparticle monolayers were transferred onto the grid by vertical lifting through the interface at constant surface pressure and at a constant speed of 1 mm/min. TEM grids were purchased from Ladd Research Industries, Inc. and used as supplied. Transfers were attempted to Formvar substrates on carbon-coated 200 mesh copper grids and bare-carbon-coated 200 mesh copper grids. The transfers to the carbon-coated grids proved to be the most successful. A second method for transferring the monolayers onto grids and optical (glass) substrates was employed for the high-density films only. In this procedure, which is similar to the LangmuirSchaeffer technique, a grid (carbon coat down) or glass coverslip

Metal Nanocrystal Monolayers was attached to the base of a small flat surface and brought parallel to the surface of the trough by hand. The surface of the grid or substrate was briefly contacted with the particle layer and lifted off. These films are not properly LB films. Nevertheless, for simplicitys sake, we will refer to these films as LB films throughout this article. Because of the high optical density of small Au and Ag particles, the high-density films are quite easily observed by eye. Thus, when this second technique was utilized to transfer particle phases to glass substrates, the efficiency of the transfer is readily apparent. Cracks, fissures, holes, and other density gradients in the Langmuir monolayer are transferred intact. The vertical lifting technique described above did not successfully transfer such features. The compressed, transferred films are apparently unstable over extended time periods (many hours or days). Thus, all TEM data were collected 1-4 h after film preparation, and all UV/vis spectra were taken within minutes of film preparation. Surface coverages were estimated using TEM. A hexane solution of particles that was used to prepare a particular isotherm was diluted by a known amount, and a 3.5 L drop of that solution was evaporated onto the carbon-coated side of a 3 mm diameter TEM grid. Two-dimensional particle densities were measured directly and used to calibrate isotherm surface coverages. The particle sizes used for the coverage estimates include both the metallic cores and the ligand shells. We estimate the accuracy of these calibrations to be between 20 and 30%. Results and Discussion In previous work, we demonstrated through both simulations and experiment the critical size-dependent role that attractive dispersion (van der Waals) forces can play in particle aggregation.13 Of particular importance for the formation of nanocrystal superlattices is the case for which such forces are strong enough to cause particle aggregation and yet sufficiently weak that particle aggregates can anneal under ambient conditions to form ordered structures. The important physical parameter that determines the nature (power law scaling) of the dispersion forces between two identical nanocrystals is the ratio of the size of the metal core to the interparticle separation distance. For the present case of nonionic, organically functionalized metal nanocrystals, the separation distance is largely determined by the length of the organic surfactants. This treatment of the dispersion attractions implies that the organic surfactants constitute a spatially uniform dielectric about the metallic core. Such an approximation is, in general, not valid. For the case of linear alkylthiols and alkylamines, the volume available to a surfactant molecule increases as the ligand extends from the metal surface. This is demonstrated graphically in Figure 1. In this illustration the excess volume (Ve) available to the ligand as it extends from the surface of the particle is shown as a hollow cone surrounding the ligand. The volume of the cone is determined by the length of the ligand (L), the footprint of the ligand (f), and the size of the metal core (R). The included angle, /2, is equal to tan-1 (f/2R). For thiols on Au, the area of the ligand footprint is 21.4 2.22 Since f is a constant, the dominant term in Ve scales as L3/R2.23 Clearly, the lack of facets is going to also play a large role in determining Ve. All of the particles discussed here are assumed to be spherical. Although this is certainly an approximation, most of our particles do appear spherical by TEM, even when imaged at high resolution. The magnitude of Ve gives some indication of the amount that a ligand shell associated with one particle may interpenetrate into the ligand shell of a second particle. Such interpenetration

J. Phys. Chem. B, Vol. 101, No. 2, 1997 191

Figure 1. Illustration of the excess volume (referred to as Ve in the text) available to a ligand as it extends out from the surface of a particle.23 The included angle contained in the cone is determined by the footprint size of each ligand (assumed to be independent of particle size) and the radius of the metal core. This ligand is shown as an extended structure for the purposes of this diagram only. The actual ligands are likely to sample many geometries, both compact and extended in nature.

TABLE 1: Alkyl Group Chain Length


C18 metal core size Ve (3), case distbn width metal core size Ve (3), case distbn width metal core size Ve (3), case distbn width metal core size Ve (3), case distbn width metal core size Ve (3), case distbn width metal core size Ve (3), case distbn width C12 C9 comment Au cores Ag cores Au cores Ag cores Ag cores Au cores 20 20 20 1550; case Ia 560; case Ic 270; case IId 20% 20% 20% 18 670; case Id 10% 40 40 460; case Ib 150; case IIa 20% 20% 35 200; case IIb <10% 28 200; case IIc <10% 75 30; case III <10%

should maximize dispersion attractions between the ligand shells. In addition, it may lead to activation barriers to the annealing of various particle aggregate structures and/or to distortions of the ligand shell about a given particle. Murrays group has recently employed ideas related to ligand shell interpenetration to explain electron transport behavior through disordered 3-D aggregates of alkylthiol-capped Au nanocrystalssparticles that are, essentially, chemically identical to the Au particles used in the present work.24 Interpenetration effects should also play some role in determining the phase diagrams corresponding to 2-D particle aggregation. Indeed, we find that this is so. Phase diagrams, determined from isotherms, were measured for seven of the particle size/ligand combinations listed in Table 1. Included in the table are the widths of the size distributions (determined from TEM) and Ve for each system. Three distinct cases of phase behavior were observed, with each case being defined by the amount of excess volume/ligand available to a given particle. A. Case I Particles (Ve > 350 3). The general phase behavior discussed in this section was observed for four types of particles, although the details of the phase diagrams vary, of

192 J. Phys. Chem. B, Vol. 101, No. 2, 1997

Heath et al.

Figure 2. Raw isotherm data plotted as pressure () vs % surface coverage (estimated from TEM data) of case Ia particles (top). Only the low-pressure regime of this isotherm is shown, but the slope of the isotherm continues nearly unchanged up to pressures above 20 mN/m. Several isotherms taken at various temperatures and represented as /area derivative data (bottom). The phase boundaries in Figure 3 were taken from the inflection points (minimas) in the derivative data.

Figure 3. /temperature phase diagrams for case Ia (2 nm diameter; C18 capped) and Ib (4 nm diameter; C18 capped) particles (see Table 1 for descriptions). Experimental points are noted on the graph, and above 37 C the boundaries were extended from the lower temperature data. Note that d/dT is negative for all phase boundaries, indicating that the higher pressure phases are increasingly disordered. The gas phase region constitutes the region of the phase diagram in which the measured pressure is not different from that of a clean water surface.

course, from system to system. The particles were small (20 ) Au cores capped with oleylamine (C18H35NH2) (case Ia), large (40 ) Au cores capped with oleylamine (case Ib), and small (20 ) Au cores (case Ic) and (18 ) Ag cores (case Id) capped with dodecanethiol. In Figure 2 are shown two representations of data used for constructing the phase diagrams. At top is a pressure()/area isotherm for case Ia particles. At the bottom of Figure 2 are shown several isotherms for case Ia particles in the temperature range 20-35 C and displayed as derivative data (d/dA vs vs T). Similar data were collected for case Ib particles. Phase boundaries were determined from the local minima in the isotherm derivative plots and were used to construct the /T phase diagram in Figure 3. All isotherms utilized to construct Figure 3 were reversible, as measured by comparing the compression and expansion isotherms. Prior to discussing Figures 2 and 3, it is important to understand the general features expected from a traditional Langmuir monolayer isotherm. Typical Langmuir monolayers, which are composed of amphiphiles whose hydrophilic head groups are in the water surface and whose hydrophobic tails are directed toward the air, have isotherms that can show evidence of many phases. When such monolayers are compressed from large molecular areas, they pass from a highly compressible gaseous phase to a somewhat less compressible liquid phase (often called liquid expanded) and then to one or more relatively incompressible phases in which the molecules are closely packed.

All particles discussed here form nontraditional monolayers since their surfaces are chemically isotropic, and they must present hydrophobic groups toward the water surface. Thus, one might expect their isotherms to differ from those observed for amphiphiles. Nevertheless, Figure 2 (top) is surprising in many ways. As one expects, at low coverages the monolayer is highly compressible, which is consistent with a gas phase. All of the particles that we examined displayed such lowcoverage behavior. It is striking, however, that the phase that extends from 3 to 5% surface coverage is substantially less compressible than the one that extends above 8% surface coverage. Between these two phases is a highly compressible region. Finally, the surface coverages corresponding to the various phases are very low compared to those found for typical amphiphiles. The phase diagrams constructed from the isotherms (Figure 3) are also rather unusual in that, for all phase boundaries, d/ dT is negative. This is counterinituitive, implying that higher temperature leads to more compact phases. In thermodynamic terms, the fact that d/dT is negative implies an increase in entropy (S > 0) between the low-density and high-density phase. This implies that the various phase boundaries correspond to transitions to successively more disordered phases rather than, for example, between a liquid phase and a solid phase. Similar transitions have been recently observed in certain amphiphilic systems.25 Two regions of the phase diagram (at 5 and 18 mN/m, both at 25 C) of case Ib particles were

Metal Nanocrystal Monolayers

J. Phys. Chem. B, Vol. 101, No. 2, 1997 193

Figure 5. Three particle interaction angles, measured from two linear and four cyclic structures similar to those shown at the top of Figure 4. Note that the preferred orientation is for to be 90 or more. Closest packed structures would be characterized by sharply peaked at 60.

Figure 4. TEM micrographs of case Ib particles taken from two points in the phase diagram of Figure 3. The particles in the top micrograph were lifted at 25 C and 5 mN/m pressure. The particles in the bottom micrograph were lifted at 25 C and 18 mN/m pressure (above the y-axis scale of Figure 3).

transferred to a TEM grid as a LB film for TEM analysis. Micrographs of those two films are shown in Figure 4. Figure 4 (top), the micrograph of the low-pressure LB film of case Ib particles, shows a surprising amount of structure. Both annular rings and linear aggregates were quite common, with as much as 50% of the total number of particles involved in one of these structures. We will refer to both the linear and the circular structures collectively as 1-D structures. Upon further compression to a pressure of 18 mN/m, the particles form a continuous, percolating network across the grid (Figure 4 (bottom)). This high-pressure phase is foamlike in structure. Note that, in Figure 4 (bottom), there is no evidence for any long-ranged ordered structures observed in Figure 4 (top), consistent with an increased disorder in the higher density phase. Although clear examples of foamlike phases have been previously observed in Langmuir films of micrometer-scale polymeric spheres,26 we believe that the origin of both of these extended phases originates from large Ve effects. Consider three hydrophobic particles on an aqueous surface. If two of the particles are brought into contact with each other such that their ligands interpenetrate, their ligand shells will distort. The extent of this distortion will be related to the amount of interpenetration, which, in turn, is related to Ve. This dimer now presents two distinct types of binding sites to the third particle. The binding site that maximizes dispersion attractions between the metal cores corresponds to a closest packed arrangement, i.e. the formation of an equilateral triangle. However, the dimer ligand shell near that binding site is already

tied up. Thus the sites that maximize the dispersion attractions resulting from ligand interpenetration will tend to favor a more extended trimer structure. The relative importance of these effects should be manifested in the interaction angles for three adjacent particles. A set of such measurements, put together by analyzing the angles from several (cyclic and linear) 1-D structures, is presented in Figure 5. For a closest packed arrangement, the distribution of angles would be peaked near 60. However, the observations indicate that the preferred threebody orientation angles are between 90 and 180. Thus, although the dispersion attractions between the metal cores are certainly important, this evidence indicates that ligand interpenetration, coupled with large Ve, plays a dominant role in determining the aggregation structures observed in these phase diagrams. In light of these observations, the isotherm and the phase diagram in Figures 2 and 3 may now be assessed. The low compressibility of the phase from 5 to 8% coverage results from compressing large, extended structures into one another. The situation is analogous to filling a container with tree branches. It is difficult to put much material into the vessel without breaking apart the branches. When these structures break, the compressibility becomes quite high. As the area available to the broken branches is decreased, a percolating network of extended structures (i.e. the foamlike phase) is formed, at which point the compressibility again decreases. Thus, the phase boundaries of Figure 3 apparently do correspond to various structured phases. These arguments should apply to other particles provided that Ve is large. We have investigated 18 Ag nanocrystals capped with dodecanethiol, which has Ve similar to the 40 Au nanocrystals capped with oleylamine. A Langmuir isotherm at 25 C was measured for these particles. The isotherm exhibited two inflection points, the first at 0 mN/m and 40% coverage and the second at 23 mN/m and 55% coverage. The phase was transferred to a TEM grid at ) 20 mN/m and 50% surface coverage, and a micrograph of this phase is shown in Figure 6. The phase shown in Figure 6 is very similar

194 J. Phys. Chem. B, Vol. 101, No. 2, 1997

Heath et al.

Figure 6. Compressed phase (20 mN/m; 50% coverage; 25 C) of case Id particles (1.8 nm diameter Ag; C12 cap). Note the similarity of this 2-D foam phase to that shown in Figure 4, bottom. Although the case Id particles are only half as large as the case Ib particles, the calculated Ves are very similar, and, according to our model, the phase behavior should therefore be very similar.

to the 25 C, 18 mN/m phase of case Id particles shown in Figure 4b. In addition, the measured isotherms for all four case I systems are reversible (prior to monolayer collapse), and this is not true for case II or case III particles. As a final note, it is possible to prepare dense, twodimensional phases of all of the case I particles by simply adding sufficient material to the Langmuir trough. Such phases were prepared by compressing a low-density phase to high pressure and subsequently adding sufficient material to achieve a >70% coverage. These phases were sampled as LB films using TEM. B. Case II Particles (150 3 < Ve < 350 3). Particles that fall into case II are clearly going to be the most important particles for preparing ordered two-dimensional metal quantum dot superlattice structures. The Langmuir isotherms of case II particles are very similar to those reported by Dabbousi and co-workers for CdSe quantum dots. Only two inflection points are observed, corresponding to the transition from the gas phase to the 2-D closest packed phase, and from the 2-D phase to the collapsed 2-D (i.e., 3-D) phase. In addition, the gas phase to closest packed phase transition occurs at very nearly 100% surface coverage. A phase diagram constructed from compression isotherms of monolayers of 40 Au nanocrystals capped with dodecanethiol is shown in Figure 7. LB films were transferred onto TEM grids for all case IItype particles listed in Table 1. The major difference observed between the various case II particles was in the degree of crystallinity of the closest packed 2-D phase, and this was related to the both the width of the size distribution and the nature of the solvent. Narrower distributions obviously lead to higher crystallinity. In addition, more slowly evaporating solvents (such as heptane vs hexane) also increase the crystallinity of the phase (presumably an annealing affect). All phases shown here were prepared from hexane solutions. The most crystalline

Figure 7. /temperature phase diagram for case IIa particles (4 nm diameter Au cores with C12 caps). Although the slope of the phase boundaries resembles those presented in Figure 3, the particle coverage is much greater, and the phases are closest packed, rather than extended. Data above 35 C were extrapolated from lower temperature data. The gas phase region constitutes the region of the phase diagram in which the measured pressure is not different from that of a clean water surface.

phases were observed for dodecanethiol-capped 35 and 28 Ag nanocrystals (case IIb and IIc), both of which were characterized by a fwhm of the size distribution of <10%, while the Au case IIa nanocrystals are characterized by a fwhm of the size distribution of 15-20%. TEM micrographs of a LB film of the 28 Ag nanocrystals (case IIc) are shown in Figure 8A,B. This film was prepared at an isotherm pressure (14 mN/m, 90% coverage) just below the (2-D)-(collapsed 2D) phase boundary (16 mN/m) at 25 C. The Figure 8A micrograph is a relatively high resolution image, which has been cropped to highlight the crystallographic structure of the phase. Figure 8B is a lower resolution image over the same phase and shows a single (although not perfect) domain extending over 0.3 m. Note that, in the phase diagram of Figure 7, the 2-D to collapsed 2-D phase boundary is characterized by a negative d/dT, similar to the situation observed for the case I particle phase diagrams. As discussed above, this implies that the higher pressure phase is more disordered than the lower pressure phase. This increased disorder is easily observed for the case II particles. Figure 9 is a micrograph of the case IIc particles taken at 25 C and 45 mN/m (180% coverage). This micrograph shows both the 2-D phase and the collapsed 2-D phase, and it is readily apparent that the particles are more highly ordered prior to collapse. The collapsed 2-D phase regions exhibit a certain amount of order, although with domains extending only over a few particle diameters. It is our experience that the collapsed 2-D phase may be annealed to varying degrees by addition of small amounts of solvent.

Metal Nanocrystal Monolayers

J. Phys. Chem. B, Vol. 101, No. 2, 1997 195

Figure 9. TEM micrograph of case IIc (2.8 nm diameter Ag; C12 cap). The particles were extracted from the Langmuir trough at nearly 180% surface coverage, or well above monolayer collapse. Note that the collapsed phase is much less ordered than the 2-D phase. This is consistent with a phase boundary described by a negative d/dT.

Figure 8. TEM micrographs of case IIc (2.8 nm diameter Ag; C12 cap). The particles were extracted from the Langmuir trough at a pressure just below that required to collapse the two-dimensional film into a three-dimensional film. The top figure is a relatively high resolution image of this phase, presented in a way so that the crystallinity of the phase is apparent. At the bottom is a lower resolution image, showing crystalline domains extended for a few tenths of a micrometer. Lattice vacancies and irregular particle sizes contribute defects to the domain.

There exist both advantages and disadvantages of using a Langmuir trough to prepare ordered arrays of nanocrystals. Ordered 2- and 3-D arrays, similar to those shown in Figure 8, can be prepared by simply evaporating a drop of the nanocrystal solution onto some substrate.13 Semiconductor quantum dot superlattices with 2- and 3-D order over micrometer length scales have been fabricated via traditional crystal growth techniques by Murray and co-workers.8 Clearly the Langmuir

trough does not lend itself to the production of 3-D ordered arrays. However, the Langmuir trough naturally produces thin film materials at thermodynamic equilibrium. Any particular phase, whether it covers a 0.3 m square such as that shown in Figure 8 or whether it covers a 12 in. Si wafer, is still that same phase. Furthermore, the Langmuir trough lends the additional variable of pressure. If an array of Ag particles is formed by evaporating an Ag nanocrystal solution onto a TEM grid, then the separation between the Ag cores is controlled only by the length of the surfactant chain and, for dodecanethiol caps, is near 15-20 . Arrays formed as an ordered phase on a Langmuir trough may be compressed such that the interparticle separation distance is substantially less than 15 . In Figure 8, the measured distance between metal cores is between 10 and 12 . Such control is going to be absolutely critical for the investigation of electronic transport properties. C. Case III Particles (Ve 30 3). The isotherms and the phase diagrams of these particles are very similar to those observed for the case II particles. One major experimental difference is related to the compression and relaxation behavior of the case III particle monolayer. Although the phase diagrams for the case II particles are not reversible, the compressed monolayer of those particles did relax to larger area coverages once the trough barriers were opened. However, the monolayer of the case III particles remained compressed, even after the barriers were reopened. The structure of the compressed, 2-D phase, as interrogated by TEM, is not a crystalline closest packed structure, but is instead that expected from the well-known diffusion-limited-aggregation (DLA) model.27 Such structures have been observed previously many times for the case of metal nanocrystals.28 They are clearly dominated by strong dispersion attractions and are predicted by our previous model for opal formation.13 They will not be discussed further here.

196 J. Phys. Chem. B, Vol. 101, No. 2, 1997

Heath et al. Although a detailed analysis of the optical properties of these films is beyond the scope of this paper, casual inspection of the data indicates several interesting trends and features. Figure 10 (top) reveals that the basic signatures of the optical spectra of the LB films remain unchanged with varying film thickness. The plasmon resonance in the LB films of all three particle systems shifts to lower energy as compared to the hexane solutions. For both dodecanethiol-capped particle systems (Figure 10, top and middle) this shift is approximately 0.25 eV and is not accompanied by a change in linewidth. For the nonanethiol-capped particles (Figure 10, bottom), the shift is only 0.15 eV and is accompanied by a decrease in the plasmon linewidth. A simplistic, classical treatment can be employed to qualitatively account for these effects. The red shift exhibited by all the films may be traced to the change in the dielectric constant of the immediate environment surrounding a given particle. The classical plasma frequency p, is proportional to (Ne2/Km 0)1/2, where N is the density of free electrons, e is the unit of charge, m is the mass of the charge carriers, K is the dielectric constant of the medium containing the oscillating charges, and 0 is the permittivity constant. Thus, as the dielectric constant of the surrounding medium increases (as would be expected for metal particles surrounded by other metal particles rather than hexane), the plasmon frequency shifts to lower energies. The linewidth of the plasmon resonance is determined by a characteristic scattering, or collision time, of the oscillating charges. The general trend observed for both alkylthiol-capped Ag and Au nanocrystals is that the plasmon resonance feature broadens with decreasing particle size, indicating that the physical extents of the particle determine the collision time. The observation that the plasmon linewidths in Figure 10 (top, middle) are similar for both the solution phase and the thin film cases implies that the scattering time scales are not a function of environment for dodecanethiol-capped nanocrystals, but rather are determined by the size of the individual particles. However, for the nonanethiol-capped particles, the linewidth decreases, indicating an increase in the scattering time scales. Such an increase argues that some fraction of the free electron density becomes delocalized among two or more metal cores when the particles are compressed into a monolayer. A narrowing of the plasmon resonance would definitely be expected provided that the interparticle separation distance were sufficiently small to effect appreciable delocalization of charge. The fact that sharpening of the plasmon is seen in particles capped with C9 ligands (nonanethiol) but not C12 ligands (dodecanethiol) then is consistent. Clearly, electronic transport measurements need to be carried out to test the validity of these arguments, and we are currently carrying out such experiments. Summary and Conclusions The Langmuir-Blodgett technique has been utilized to prepare various phases of organically functionalized 2-7.5 nm gold and silver nanocrystals. The phase diagrams were investigated as a function of particle size, the length of the organic surfactant, and the nature and size distribution of the metal cores. In general, the phase behavior exhibits various dependencies on all of these variables. However, general trends can be categorized according to the excess conical volume (Ve) available to the ligand as it extends from the surface of the particles. In the limit of large excess volume (case I), the phase diagrams are reversible and are characterized by low-density structures. At low pressures, these structures are dominated by chain and ring morphologies, while at high pressures the

Figure 10. UV/vis absorption (extinction) spectra of case III (7.5 nm Au; C12 cap; top), case IIb (3.5 nm Ag; C12 cap; middle), and case IIc (2 nm Au; C9 cap; bottom) particles. For each case, the spectrum of a hexane solution of the particles is shown, as is the spectra of a compressed LB film. For the case III particles, the spectra of a single monolayer (labeled 1 mL) and of a double layer are both shown. For the other two cases, the LB film is a multilayer. All spectra are normalized to the peak of the plasmon resonance. Note that the resonance is shifted to lower energies for all of the compressed films. For the case IIc particles the resonance is substantially sharpened in the LB film spectrum.

D. Optical Properties of Langmuir-Blodgett Films of Case II and Case III Particles. The compressed, 2-D Langmuir films of case II and case III particles are colored due to the strong surface plasmon resonance characteristic of both the Ag and Au nanocrystals. Upon compression the films became quite reflective. This indicates that the compressed monolayers are characterized by significant real (reflective) and imaginary (absorptive) terms in the refractive index, although no attempts were made to measure this complex refractive index. It was possible to lift off these films onto a glass slide and directly measure the UV/vis absorption (or extinction) spectrum of single monolayer or multilayer films using a standard spectrophotometer. Several optical spectra are shown in Figure 10, and all spectra have been normalized to the absorbance maximum of the plasmon peak. Figure 10 shows the UV/vis spectra of case III particles (75 dodecanethiol-capped Au nanocrystals) in hexane, as a single LB monolayer and as a LB multilayer (two monolayers). Figure 10 also shows the UV/vis spectra of case IIb particles (35 dodecanethiol-capped Ag particles) in hexane and as a LB multilayer and the UV/vis spectrum of case IIc particles (20 nonanethiol-capped Au particles) in hexane and as a LB multilayer.

Metal Nanocrystal Monolayers Langmuir monolayers form a foamlike phase. For very small excess volume (case III), the particles irreversibly aggregate into nonequilibrium structures such as those expected from a diffusion-limited-aggregation model. In between these two limits (case II), however, the particles may be compressed into two-dimensional closest packed phases. The crystalline order within these closest packed phases is limited primarily by the width of the particle size distribution, and domains with order extending over 1 m in length have been observed. UV/vis absorption (extinction) spectra of the closest packed phases were measured and compared with spectra corresponding to the free particles in hexane. Two primary trends were observed. The particles in the closest packed phase exhibit a red shift in the plasmon resonance when compared to the solution phase particles. This is rationalized in terms of the change in the dielectric environment around a given particle. In addition, the compressed phase of the particles with the shortest ligands (smallest interparticle spacings) exhibits a plasmon resonance that is substantially sharper than that observed for the particles in solution. This is rationalized in terms of an increased scattering time of the electrons, implying that some of the charge is delocalized over one or more particles. Thermodynamic control over particle aggregation was shown to be a useful way for preparing various ordered phases of metal nanocrystals. The ease of transferring these phases to almost any substrate should facilitate the fabrication of simple quantum effect electronic devices as well as the measurement of various transport properties. Acknowledgment. We would like to acknowledge many helpful discussions with the following people: Prof. Andrea Liu, Dr. Joseph Shiang, Ms. Pam Ohara, and Prof. Bill Gelbart. Helpful insight into the Langmuir-Blodgett technique was gleaned from Mr. Gary Marshall and Dr. Jiyu Fang. Certain of the particles discussed here were prepared by Ms. Erica DeIonno. J.R.H. and D.V.L. acknowledge primary support from the NSF-NYI program and the David and Lucile Packard Foundation. C.M.K. acknowledges support from the NSF. Certain equipment used here was supported by the Office of Naval Research, Order No. N00014-95-F-0099, under the auspices of the ONR-funded Molecular Design Institute at UC Berkeley. References and Notes
(1) Bawendi, M. G.; Steigerwald, M. C.; Brus, L. E. Annu. ReV. Phys. Chem. 1990, 41, 477. 59.

J. Phys. Chem. B, Vol. 101, No. 2, 1997 197


(2) Eychmuller, A.; Mews, A.; Weller, H. Chem. Phys. Lett. 1993, 208, (3) Micic, O. I. et al. J. Phys. Chem. 1995, 99, 7754. (4) Schmid, G. Chem. ReV. 1992, 92, 1709. (5) Colvin, V. L.; Goldstein, A. N.; Alivisatos, A. P. J. Am. Chem. Soc. 1992, 114, 5221. (6) Colvin, V. L.; Schlamp, M. C.; Alivisatos, A. P. Nature 1994, 370, 354. (7) Bentzon, M. D.; van Wonterghem, J.; Morup, S.; Tholen, A.; Koch, C. J. W. Philos. Mag. B 1989, 60, 169. Bentzon, M. D.; Tholen, A. Ultramicroscopy 1990, 38, 105. (8) Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Science 1995, 270, 1335. (9) Dabbousi, B. O.; Murray, C. B.; Rubner, M. F.; Bawendi, M. G. Chem. Mater. 1994, 6, 216. (10) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706. (11) Peschel, S.; Schmid, G. Angew. Chem., Int. Ed. Engl. 1995, 34, 1442. (12) The term opals is utilized here because of the direct comparison of these nanocrystal superlattice structures to naturally occurring opals, which are also colloidal crystals. However, in naturally occurring opals, the size of the individual particles, and the corresponding d-spacings within the extended structure, are in the size range 0.1-1 m, and hence opalescence is the Bragg diffraction of light. See, for example: Sanders, J. V. Acta Crystallogr. Sect. A 1967, 24, 427. (13) Ohara, P. C.; Leff, D. V.; Heath, J. R.; Gelbart, W. M. Phys. ReV. Lett. 1995, 75, 3466. (14) Hamaker, H. C. Physica (Utrecht) 1937, 4, 1058. (15) Bargeman, D.; Van Voorst Vaden, F. J. Electroanal. Chem. 1972, 37, 45. (16) See, for example: Nagayama, K. Phase Transitions 1993, 45, 185. (17) Alfrey, T.; Bradford, E. B.; Vanderhoff, J. W.; Oster, G. J. Opt. Soc. Am. 1954, 44, 603. (18) See: Bibette, J. J. Colloid Interface Sci. 1991, 147, 474. (19) Leff, D. V.; Ohara, P. C.; Heath, J. R.; Gelbart, W. M. J. Phys. Chem. 1995, 99, 7136. (20) Brust, M. et al. J. Chem. Soc., Chem. Commun. 1994, 801. (21) Leff, D. V.; Brandt, L.; Heath, J. R. Langmuir 1996, 12, 4723. (22) Sellers, H.; Ulman, A.; Schnidman, Y.; Eilers, J. E. J. Am. Chem. Soc. 1990, 112, 570. (23) Ve was calculated by evaluating the integral r2al dl from l ) 0 to l ) L and subtracting the cylindrical volume f2L. Here, L is the length, in , of the ligand; is the included angle of a cone; a is the radius of the particle; f is the radius of the footprint of the ligand. The function r2al ) [(a + l)tan ]2. The angle is a constant, fixed by (f/a ) tan ). (24) Terrill, R. H.; et al. J. Am. Chem. Soc. 1995, 117, 12537. (25) Ibn-Elhaj, M.; Riegler, H.; Mohwald, H. J. Phys. I 1996, 6, 969. (26) Corrales, R. G.; Maestro en Ciencias Tesis, Universidad Autonoma De San Luis Potosi, San Luis Potosi, 1995. (27) See, for example: Ball, R. C. et al. Phys. ReV. Lett. 1987, 58, 275. (28) Fractal structures resulting from DLA are reported in the following references: (a) Schmid, G.; Lehnert, A. Angew. Chem., Int. Ed. Engl. 1989, 28, 180. (b) Ishizuka, H. et al. Colloids Surf. 1992, 63, 337. (c) Hirai, H.; Aizawa, H. J. Colloid Interface Sci. 1993, 161, 471.

Vous aimerez peut-être aussi