Analysis of the structure of the nanoparticle film
Figure 2a shows the TEM image of Fe nanoparticle arrays prepared by depositing the nanoparticle beam vertically on the surface of the amorphous carbon film with a deposition rate of 0.1Å∙s− 1. The nanoparticle array is composed of densely-packed 2D monolayer domains ordered over 100-200nm scale. A high magnification TEM image of a well-ordered defect-free monolayer domain is shown in Fig. 2b. Typically, the nanoparticle monolayers may contain some defects, such as lattice distortions, dislocations, vacancies or voids, as well as size variations of the nanoparticles. The average nanoparticle size is 6.1 ± 1.6nm, as determined using a minimum of 300 nanoparticles in the arrays (Fig. S1). The size dispersion is significantly larger than those in the self-assembled superlattices of thiol-passivated nanoparticles [26, 27]. A high-resolution transmission electron microscopy (HRTEM) image given in Fig. 2d shows the individual Fe nanoparticles are mainly single crystals with spherical shapes. They are randomly oriented on the substrate surface. Prior to the observation, the nanoparticles have been exposed to air for a significantly long time so that their surfaces are sufficiently oxidized, as can be distinguished in the HRTEM image. The existence of the oxidation layer on the nanoparticle surface can be further confirmed by EDX and XPS. As shown in Fig. S2, O elements are always observed together with the Fe nanoparticles in the EDX elemental mapping images. XPS measurements also show the evidence of oxidation of Fe nanoparticles. As shown in Fig. S3, both metallic Fe and Fe oxide can be distinguished from the photoemission data of the Fe 2p core levels. After the nanoparticle specimens are cleaned with Ar ion sputtering, the XPS peaks corresponding to the 2p core levels of pure Fe are greatly enhanced, indicating that Fe oxides only present on the nanoparticle surfaces. Therefore, the nanoparticle arrays can be looked as a compact packing of closely contacted core/shell nanoparticles. The crystalline metal cores are separated from each other with amorphous oxide shells. The oxide shell acts as a passivation layer preventing further oxidation of the Fe nanoparticles. The mean edge-to-edge distance separating well-aligned nanoparticles in the closely packed lattice is measured at 1.7 ± 0.6nm. Correspondingly, the thickness of the amorphous oxide shell is about 0.85 nm on average. It should be noted that the oxide shells are formed after the deposition process is finished. Since the nanoparticle deposition is performed under high vacuum condition, the self-assembling occurs in the pure metal nanoparticles, rather than the surface-oxidized nanoparticles. The amorphous oxide shells play no role in the organization of the nanoparticles.
The fast Fourier transform (FFT) of the densely-packed monolayers is shown in Fig. 2c. Well-defined spots arranged in hexagon are shown, attesting to a densely packed nanoparticle lattice ordered over a long range. However, only one hexagon related to the first order is distinguishable, indicating the scale of the ordered monolayer domains is limited.
Tailoring the assembling morphology with nanoparticle deposition conditions
We have found the deposition rate of the nanoparticles plays a definite role in the formation of the ordered densely packed monolayers. In Fig. 3a-c, TEM images of Fe nanoparticle arrays prepared with deposition rate ranging from 0.3–0.7Å∙s− 1 are shown. The FFT of each image is shown as the insets. The deposition time of each specimen is such controlled that a constant nanoparticle coverage (i. e. total deposition mass) on the substrate is maintained. In each image, the average nanoparticle size and distribution (Fig. S1), is almost identical (the average diameter is measured to be 6.0 ± 1.4nm, 6.1 ± 1.3nm and 6.1 ± 1.7nm respectively). From Fig. 2a and Fig. 3a-c, we can find that with the increase of the deposition rate, the range scale of the ordered monolayer domains becomes smaller and smaller. We analyse the TEM images by counting the nanoparticle number contained in each monolayer domain. The sizes of the monolayer domains can be compared quantitatively with the nanoparticle numbers they contain. The histograms of the counted nanoparticle numbers are shown in Fig. 3d. The maximum of the distribution tends to smaller nanoparticle numbers with the increase of deposition rate. The average nanoparticle number contained in an individual monolayer domain decreases from 77 at a deposition rate of 0.1Å∙s− 1 to 27 at a deposition rate of 0.7Å∙s− 1. Meanwhile, the spots present in the FFT pattern become more and more diffuse. With a deposition rate of 0.7Å∙s− 1, only a diffuse ring without any hexagonal symmetry can be seen in the FFT pattern. The nanoparticles in the TEM image display a random distribution on the whole.
In Fig. 3e, the radial distribution functions (RDFs) calculated from the TEM images are shown. For the nanoparticle arrays formed at 0.1Å∙s− 1 and 0.3Å∙s− 1 deposition rates, the RDF curves display sharp and clear first and second peaks, corresponding to the nearest and subnearest neighbors with average particle-particle intervals of 8nm and 17nm, and a distinguishable third peak corresponding to the third neighbors with an average interval of 24nm, indicating that the nanoparticle arrays are well-ordered with superlattice periodicities. For the nanoparticle arrays formed at 0.5Å∙s− 1 deposition rate, the second peak in the RDF curve becomes much reduced and the third peak is completely indistinguishable, indicating a decreasing organization and reduced lattice periodicity. With a deposition rate of 0.7Å∙s− 1, the nanoparticle arrays only display a weak first peak in the RDF curve, which strongly reflects the loss of the lattice periodicity and short-range order. It is clear that a low deposition rate is an important parameter dominating the well-ordered monolayer formation.
We have also found the structure of the nanoparticle arrays is correlated with the feature of the substrate surface. Different assembling patterns are obtained with different substrates. Figure 4 shows a TEM image of the Fe nanoparticle arrays deposited on a Formvar film. The deposition is performed with a deposition rate of 0.1Å∙s− 1. Although the operation parameter of the cluster source and the deposition mass is identical to that used for the sample shown in Fig. 2, ordered densely-packed morphology could no more be observed. The distribution of the nanoparticles on the surface is completely random. No evidence of organization could be observed. In some area homogeneous coalescences of the nanoparticles form larger particles. It is known the mobility of nanoparticles softly landing on solid surface is strongly dependent on the nature of the surface [20], especially its defect state and binding ability with the deposits. It has been shown that metal nanoparticles have high mobilities on the surface of carbon materials [22, 23], on the contrary, no diffusive aggregation of metal nanoparticles has been observed on the Formvar film surface [28]. Instead, the nanoparticles are mostly pinned where they are deposited. It is difficult for them to diffuse and aggregate on the substrate. Coalescence takes place locally as a fusion process under particle-particle collision in the deposition process. These results suggest that certain mobility is needed once the nanoparticles are deposited on the surface in order to form densely-packed monolayers.
To understand the formation mechanism of the densely-packed nanoparticle monolayers in the gas phase cluster deposition, we have to consider the competition between the diffusion rate of the nanoparticles and the filling speed of the nanoparticles deposited on the substrate surface, which is dependent on the deposition rate. This is similar to the situation that happens in the spontaneous organization process occurring at the liquid/substrate interface to form periodic 2D arrays of thiolate encapsulated nanoparticles upon solvent evaporation from a droplet of colloidal solution depositing on the substrate. Previously, experiments [29, 30] showed that when a droplet of nanoparticle solution was deposited onto a substrate and dried shortly, amorphous nanoparticle aggregates with little uniformity and symmetry were formed. As the droplet was dried more and more slowly, increasing uniformity was observed and finally closely-packed nanoparticle superlattices were formed. With a slow solvent evaporation rate, the nanoparticles benefit from more time to diffuse on the substrate and adjust their sites attached to the nanoparticle assembly, giving rise to a higher level of ordering. Similarly, the nanoparticles deposited on the carbon substrate from gas phase can diffuse on the free surface with a high mobility. If the arrival rate of nanoparticles to the surface is too high, the motion of the nanoparticles on the surface will be limited by each other, and the free area available for each nanoparticle will be soon exhausted. The nanoparticles cannot sufficiently adjust their positions on the surface, resulting in randomly packed aggregates. Moreover, if the sticking coefficient between the nanoparticles remains high, low-density factal aggregates are formed [22, 24]. However, with a mild deposition rate, the arrival time of the nanoparticles is controlled such that the nanoparticles have enough time to diffuse on the free surface and find equilibrium lattice sites on the growing structure. As a result, ordered densely-packed monolayers are formed. As the flux of nanoparticles adding to the surface is increased by increasing the deposition rate, the arrival rate exceeds the surface mobility of nanoparticles and the formation of an inhomogeneous disordered aggregate occurs.
The surface mobility of the nanoparticles is dependent on the interaction between the nanoparticle and the surface. On the surface of organic materials metal nanoparticles are mostly pinned where they are deposited. Although they have high mobility on the perfect surface of carbon substrate, their diffusion may also be limited by the particle diffusion barriers on the surface, such as the defects. If the thermal energy of a nanoparticle is low compared with the binding energy, it may be arrested on the diffusion barriers. It is possible to increase the diffusion length of the nanoparticles either by increasing the temperature or by increasing their lateral migration energies when they land on the surface. Increasing temperature is not preferred in the nanoparticle assembling since it induces sufficient coalescences among nanoparticles, we therefore try to increase the lateral migration energies of the nanoparticles by increasing their momentums along the surface when they impact on the substrate. This is achieved by depositing nanoparticles with glance incidence relative to the substrate surface. Generally, the initial kinetic energy of the nanoparticles generated from a cluster source is several eV on average. With a glance incidence, a partial of the kinetic energy transfers to the migration energy of the nanoparticle on the surface. This will enhance the abilities of the nanoparticles to escape from the diffusion barriers where they are arrested, so as to increase the migration length of the nanoparticles. In Fig. 5a, a TEM image of the Fe nanoparticle arrays prepared with a 45° glance incidence angle is shown. The equivalent deposition rate is 0.1Å∙s-1. Comparing with the nanoparticle arrays prepared with same deposition parameters under normal incidence (Fig. 2a), we find the range scale of the ordered monolayer domains is significantly increased, and the hexagonally arranged FFT spots become more sharp, clear and scattered. From the RDF curve shown in Fig. 5b, we can see the subnearest peak is notablely enhanced and sharpened. Especially, the third neighbor peak which is indistinct in the case of the normally deposited samples becomes sharp and clear now, indicating a significant improvement on the organization length and lattice periodicity. It therefore demonstrates a simple way to increase the diffusion length of the nanoparticles so as to realize larger scale ordered monolayers.
It should be noted that the balance between the diffusion rate and the arrival time of the nanoparticles on the surface is not the only condition sufficient for the ordered nanoparticle monolayer formation. The ordering is driven by the interparticle attractive forces. Unlike in the case of self-assembled superlattices of thiolate encapsulated nanoparticles, in which the main contribution to the interaction comes from the surfactant molecules, which produce a soft structure [27], in the present study the interaction dominates the ordered nanoparticle array formation comes from the metal nanoparticles themselves, which produces a rigid hard structure. In the 2D densely-packed monolayer, a nanoparticle falls on the equilibrium site since it receives the maximum attractive interactions from the identical nearest neighbors. With a sufficiently long time for free diffusion, the individual nanoparticles can sufficiently modify their positions to find the equilibrium lattice positions. A challenge comes from that Fe is a ferromagnetic material. The dipolar magnetic interactions between magnetic nanoparticles increase with the particle volumes and oppose any 2D long range ordering. Previous studies showed that Co nanoparticles larger than 16 nm tended to form one-dimensional chains and a variety of linear structures [31]. Therefore, in the present case, magnetic interactions play no role in the self-assembling of the 2D densely-packed monolayers of Fe nanoparticles. In fact, magnetization measurements on the Fe nanoparticle deposits display no ferromagnetic hysteresis loops and remnant magnetizations around room temperature, as shown in Fig. 6a, indicating that the Fe nanoparticles are in the superparamagnetic states. It is more likely that attractive van der Waals interactions or dipolar interactions arisen from polarizations dominate the self-assembling of the ordered 2D densely-packed monolayers. In fact, we can also obtain ordered densely-packed monolayers from nanoparticles of nonmagnetic materials. TiN nanoparticles are generated in the gas aggregation cluster source and deposited on the amorphous carbon with similar deposition conditions. From the TEM image shown in Fig. 6b, we can see most of the TiN nanoparticles are involved in a number of ordered monolayers with 2D densely-packing superlattice structures. Similar to the case of Fe nanoparticles, the TiN nanoparticle superlattices can spread over hundreds nm scales.
Regarding the size of the ordered densely-packed 2D monolayer structures that can be achieved with the gas phase cluster deposition, we show in Fig. 7a TEM image of Fe nanoparticle film with a coverage approaching 100% (i.e., a complete monolayer). By controlling the deposition mass, the densely-packed 2D monolayer structure spreads over the whole substrate surface covered by the deposition spot (at least at the centimeter scale). The monodispersed nanoparticles show a perfect homogeneous distribution in the wide range. Only a few several tens of nanometers sized voids distribute in a very low density. The FFT of the monolayer (inset in Fig. 7) shows two rings of hexagonally arranged spots, related to the first and second orders, attesting to a well-defined hexagonal network ordered over a sufficiently long range. Even though the large scale assembling structure contains domains of ~ 100 nm in size, with a number of packing arrays or orientations of the same structure, it is difficult to find any well-defined boundaries between the ordered domains. This result demonstrates the gas phase cluster deposition may provide an efficient way for the fabrication of well-defined patterned superstructures assembled from nanoparticle building blocks on a sufficient large scale.