Nanoparticle-host interactions in natural systems

Скачать 194.04 Kb.
НазваниеNanoparticle-host interactions in natural systems
Дата конвертации19.04.2013
Размер194.04 Kb.
1   2   3   4   5   6   7   8   9   ...   12

Experimental results of metal cluster deposition on molybdenite surfaces

When sputtering metal islands onto mineral surfaces, the fastest way to analyze the amount of material sputtered on these surfaces is using x-ray photoelectron spectroscopy (XPS). Quick estimates can be made on how much material was finally adsorbed to the surface (e.g., in units of percentage of monolayer coverage) and on the amount of contaminants present in the mineral host or in the adsorbates. In the experiments by Becker et al. (2003), adsorbate coverages amounted to 5-60% of monolayer coverage. Furthermore, no contaminants (e.g., oxygen or adventitious carbon) were found on the sample. This was due to the sample being cleaved right before introduction into the vacuum chamber, the slow oxidation kinetics of molybdenite, and the fact that metal deposition and analysis took place in the same UHV chamber.

Molybdenite consists of S-Mo-S “sandwiches” parallel to the (001) surface that are bonded to each other by van-der-Waals forces (Fig. 6, space group P63/mmc, a = 3.15 Å, c = 12.3 Å). Within the sandwiches, S and Mo atoms are held together by covalent bonds. Calculations using Crystal98 (Dovesi et al., 2000) show that bulk molybdenite is a low spin system, and the Mulliken charges on S and Mo are –0.35 and +0.7, respectively.

The STM image of a freshly cleaved MoS2 (001) surface at positive or negative sample bias voltage shows a hexagonal array of high tunnelling currents (bright spots) reflecting the hexagonal arrangement of surface atoms. By only interpreting the hexagonal geometry of the image pattern, it cannot be determined if the spots of high tunnelling current are located above S atoms or Mo atoms. As described in the literature (see, e.g., Coley et al., 1991; Magonov & Whangho, 1994; Altibelli et al. 1996), the interpretation of what positions the bright spots in the STM images refer to can depend on the tip structure and tip-sample separation.

The bright spots in the STM image are located above surface S atoms or above Mo atoms that are just below the uppermost S atoms depending on the arrangement of the last atoms of the tungsten tip (molybdenite matrix part of Fig. 7, around the silver islands). Even though orbitals with Mo 4dz2 character dominate the top of the valence band (and the bottom of the conduction band), they fall off more rapidly with distance than the s or p orbitals of the S atoms. Therefore, calculations using Crystal98 in combination with a specific software to generate theoretical STM images (Becker, 1995) reveal bright spots above Mo atoms for tip-sample separations of less than about 3 Å. For larger distances between the tip and the sample, this type of calculations is to a certain degree dependent on the basis functions chosen for the most diffuse orbitals on Mo and S. In addition, for the simulation of these images, an ideal tip structure was chosen. This means that the local density of states (LDOS) above the tip was assumed to be featureless (flat as a function of electron binding energy) and that the distribution of the state density can be approximated to be spherical about the tip apex. Such a calculated image is shown in Fig. 8. The image shows that the bright spots located above Mo atoms are due to the overlap of Mo 4dz2 orbitals and S 3p orbitals that point towards the threefold axis that goes through the Mo atoms and is parallel to the (001) direction. For larger distances between the tip and the sample, calculations using Castep (a quantum-mechanical code that uses planewaves to construct the wavefunction rather than atomic basis function as in Crystal98, MSI, 1998; Milman et al., 2000) indicate that the highest occupied orbitals are located above S atoms with mainly S 3p character. This knowledge is important to interpret, which states are involved in the formation of adsorbates.

Fig. 9 shows an experimental STS spectrum taken above a bright spot of an STM image. The width of the low base in the centre of the spectrum ( 0.3 eV left and right of the Fermi level) can be roughly interpreted as the band gap if tunnelling-specific processes such as band bending are ignored. The nature of the band gap can be best described by molybdenum disulphide being an indirect optical d-d band gap semiconductor (Kertesz & Hoffmann., 1984). The indirect transition is from the  point to the K point in the notation of the band structure in reciprocal space with an approximate band gap of 0.60 eV (Coehoorn et al., 1987a, b). These earlier studies indicate that the top of the valence band has mainly Mo 4dz2 character. This nonbonding band overlaps slightly with the binding part of the valence band, which is predominantly of S 3p character with mixed-in Mo 4d contributions (Brändle et al., 1995).

STM images of adsorbed silver

The amount of silver deposited on the MoS2 surface can be easily controlled by varying the deposition time of the sputtering process. The STM image of one such deposition experiment is shown in Fig. 7a. As long as the silver coverage of the surface is less than about 20 % of one monolayer (as measured using XPS), increasing the deposited silver leads to an increase in the number of islands per unit area but not to an increase in size of each individual island. The island size is uniformly 2 nm in diameter, which is 7 Ag atoms in diameter or an average of 37 Ag atoms per island (Fig. 7b). For the image in Fig. 7a, a coverage of  1 % can be estimated from the image as well as from the corresponding XPS spectrum. The images in Fig. 7 show dark areas around the islands. Since the image was taken at positive bias voltage, this indicates depletion in conduction-band state density around the islands, which will be discussed in more detail with the ab initio calculations of the band structure. The effect of the dark shadows around the Ag islands may be enhanced by topographic effects where the islands cause a slight indentation of the MoS2 surface and by the limitations of the tip to resolve the edge of surface islands.

UPS spectra and the nature of the nanoparticle-mineral bond

In order to get an impression of the changes of the spatially averaged electronic structure near the top of the valence band when silver is adsorbed to the MoS2 surface, UPS spectra were collected of the molybdenite (001) surface, of pure metallic silver (fcc, no specific surface), and of the molybdenite surface with different degrees of silver coverage (Fig. 10). While the pure silver metal (Fig. 10a) and the pristine molybdenite surface (Fig. 10b) are fairly featureless, the electronic structure of the adsorbate shows a significant peak corresponding to an increase of the density of states and/or the sensitivity towards excitation with UPS at about 4.5 eV into the valence band. The height of this distinct peak increases with the concentration of the silver on the MoS2 surface. An attempt to clarify if these electronic states are bonding orbitals between the silver islands and the MoS2 surface or if they are caused by differences in the electronic structure between silver islands and metallic bulk silver is made in the calculation section.

Simulation of Ag adsorbed to molybdenite

In order to get an impression on the characteristics of the Ag to MoS2 bonds, quantum-mechanical calculations were performed using Crystal98. The model setup for these calculations is a slab, two MoS2 sandwich layers thick with an infinite extension perpendicular to the surface vector (Fig. 11). Such a slab is thick enough because atomic and electronic interactions between two sandwiches are weak due to the van-der-Waals character of the interaction. A complete monolayer of Ag atoms or just ¼ of a monolayer (Fig. 12) was adsorbed to a molybdenite slab containing two MoS2 sandwich layers. The energetically most favourable Ag position is located 3.79 Å above Mo atoms which results in an Ag-S bond distance of 2.81 Å (with one Ag at equal distance to three S). Due to the polarization of the Ag atoms on the surface by the uppermost molybdenite layer, the Ag atoms become positive (+0.51 Mulliken charges). This polarization effect is compensated by a more negative charge of the uppermost sulphur layer (-1.07 vs. –0.35 in the bulk). Surprisingly, the Mulliken charges of the Mo and lower S atoms in the uppermost MoS2 sandwich are more positive (less negative) than in bulk MoS2. The Mulliken charges of the second MoS2 sandwich layer are bulk-like. Due to the unpaired spins of the Ag atoms, the most energetically favourable spin configuration is antiferromagnetic in the uppermost two MoS2 layers. This is in contrast to the low-spin (S=0) configuration of the bulk and of the molybdenite surface slab without Ag adsorbed to it. It is interesting to note that the Mulliken charges are bulk-like whereas the Mulliken spin densities are not. The limited thickness of the slab does not allow evaluating how deep into the bulk the antiferromagnetic spin configuration progresses. The energetics of Ag-Ag bond formation and Ag-sulphide adsorption are summarized in Becker et al. (2003) and for the adsorption energies of different cluster sizes in Fig. 13. When a flat hexagonal Ag layer is adsorbed to MoS2 as in Fig. 11, 2.15 eV are gained which results in a total gain of 3.75 eV per Ag atom from a single Ag atom to a complete adsorbed monolayer.

In order to estimate the electronic structure of a surface with single Ag atoms adsorbed, a 22 surface supercell can be chosen for the calculation with one Ag atom adsorbed per supercell corresponding to a surface coverage of ¼ of a monolayer (Fig. 12). This setup is a compromise between computational expense and the calculation of the adsorption of truly isolated single Ag atoms, because the perturbation due to the adsorption of an Ag atom on the electronic structure is more far-reaching than the distance between the Ag atoms (Becker et al., 2001). This is also reflected by the fact that three out of four surface S atoms are bonded to Ag atoms (Fig. 12) and only one surface S atom per supercell is not. Nevertheless, these calculations show the trend of the electronic structure changes when compared with the calculations of full coverage (Fig. 11). Fig. 12 shows that the adsorbed Ag atoms are now closer to the surface by 0.14 Å than in the case of full Ag coverage. In the case of ¼ coverage, the Ag loses its valence electron completely, resulting in a Mulliken charge of +0.93 and a spin density of  0. Since ¾ of the surface S atoms are now bonded to one Ag instead of three as in Fig. 11, these are less negatively polarized (with –0.65 charges instead of –1.07 as in the case of full coverage). The S atoms without bonds to Ag only have a charge of –0.42, a value that is similar to S in bulk MoS2. The differences are less pronounced for the Mo and S underneath, except for the differences in spin density between the two types of S atoms in the lower S layer. Thus, one can expect that an Ag atom in the centre of a two-dimensional nanoparticle behaves like one that was calculated for full coverage, Ag atoms at the edges may have charges and spin densities that resemble those of the ¼ coverage calculations.

For the simulation of clusters on the surface of non-periodic overlayers, one has to further compromise on the computational rigor and use empirical potentials for calculations using a 2020 supercell. Upon adsorption of a complete Ag layer on MoS2, the silver layer gets slightly warped. The positively polarized Ag atoms are further apart than the S atoms on the surface. The polarization causes longer Ag-Ag metal bonds than in bulk fcc silver. The resulting stress is minimized by the limited size of the Ag islands. Thus, at low coverage, the size of silver islands shown in Fig. 7 is limited such that the lattice mismatch does not cause significant strain.

We can also quantify the energy of adsorption as a function of cluster size. In order to do this, we first go back to the quantum-mechanical approach. 1.60 eV are gained during the formation of a (theoretical) hexagonal monolayer compared to 2.66 eV in the formation of bulk fcc silver. This can be related to the lower coordination of Ag in a two-dimensional layer. The adsorption of such a flat layer results in an additional gain of 2.15 eV. This value can be compared with the desorption energy of 1.99 eV (46 kcal/mol, Li et al., 1998), using temperature dependent desorption measurements. With these calculations in mind, one can calculate the adsorption energy as a function of cluster size (Fig. 13). The figure shows the adsorption energy of different clusters with the Ag3 triangle being the least favourable configuration and the Ag37 cluster the most stable one. There are two effects that cause the energy minimum in Fig. 13b. The edge and corner energies (per Ag atom) decrease with increasing cluster size but the strain increases. Thus, the Ag37 cluster can be considered as the best compromise between these two opposing effects.

The adsorption energies of the different clusters were obtained from a combination of quantum-mechanical and empirical calculations. All clusters were calculated empirically with the potential set listed in Becker et al. (2003). The degree of computational sophistication can be taken one step further if different spin configurations are considered within the silver islands. This was tested for the Ag7 cluster model with total spin numbers of 1, 3, 5, and 7 unpaired spins. The most energetically favourable spin configuration is the one with 1 total spin (b) followed by 7 spins (Fig. 14a) where all spins point in the same direction. In the spin=1 system, the symmetry is lowered, and the bond distance between the central Ag atom and the surrounding Ag atoms is also lowered.

Quantum-mechanical calculations on the density of states

The electronic density of states (DOS) was calculated using Castep for MoS2, silver in both bulk (fcc) and layer (hexagonal) forms and MoS2 with a monolayer of Ag (Fig. 15). The valence electron wavefunctions were projected onto pseudo-atomic basis orbitals (Sanchez-Portal et al., 1995) to determine the dominant contributions to each part of the band structure. MoS2 was confirmed as having Mo 4d contributions at the top of the valence band, and bands with S 3p, Mo 4d, and Mo 5s character below, down to approximately 7 eV below the Fermi level, EF. All peaks in the density of states (DOS) for MoS2 are of similar size in this region, consistent with the lack of strong features in the UPS spectrum. S 3s states appear at –13 eV.

The DOS of a hexagonal layer of silver is significantly different from the DOS of the bulk fcc phase, although the valence bands of both are comprised mainly of Ag 4d states. Ag 5s states are at the very top of the valence band. It is noticeable that the hexagonal silver layer has sharper DOS peaks than the fcc phase.

The density of states of MoS2 with an Ag overlayer has no clear band gap (highly metallic character), and has three distinguishable regions below EF. Two are similar to the MoS2 layer: a peak around –6.5 eV has S 3p + Mo 5s, 4d character, and another at –1.5 eV is mostly Mo 4d. A larger peak between –3.5 eV (FWHM≈2 eV) is dominated by Ag 4d character. Since this peak is at an energy level similar to 4d states in the hexagonal silver level without MoS2, it suggests that the change in lattice from fcc to a hexagonal plane must contribute significantly to the change in electronic structure.

The total Ag-MoS2 DOS is much less uniform than MoS2 or Ag alone and is consistent with the dominant peak in the UPS spectrum. Since the UPS sensitivity of the calculated band structure is not yet known, more detailed calculations would be necessary for a definite assignment of that peak to internal Ag 4d states in the overlayer. However, our calculations show no clear evidence that the main peak in Fig. 10c can be related to a strong interaction between Ag and MoS2 electronic states so the hexagonal Ag lattice is assumed to be mainly responsible for the UPS observations.

Surface diffusion of copper nanoparticles

The copper islands that were found after initial deposition (Fig. 16a) are larger (diameter of  8 nm, height of  1.5 nm or about three atomic copper layers thick) than the described silver islands (Fig. 7). Despite their size, these clusters undergo significant surface diffusion such that after about one day after deposition, very few islands can be found on flat surfaces. Instead, they get attached to perturbations on the molybdenite surface (Fig. 16b). The remaining islands on the surface tend to coagulate to aggregates (Fig. 16c), thereby keeping their original shape and not fusing to a single island. This together with the observation of a uniform islands size (Fig. 16) may be an indication that this is a preferred and stable island size. No atomistic calculations on the larger copper and gold islands have been performed yet because the size of the large metal clusters is beyond the capabilities of proper atomistic, especially quantum-mechanical treatment. The cluster size may also point future work in a direction to describe this behaviour by using macroscopic properties such as surface tension instead or in addition to a purely quantum-mechanical approach to explain the size of the copper islands. However, the high surface diffusivity and the more spherical Cu island shape can be interpreted to be due to a weaker Cu-S bond than in the case of Ag. Especially the fact that large Cu clusters (about 3000 atoms) undergo such extensive surface diffusions indicates a weak MoS2-Cu adsorption energy.

Bimodal size distributions of gold nanoparticles

The adsorption of gold seems to combine the observations found for silver and copper. Fig. 17a shows two distinct island sizes, one that is on the order of the size shown for Ag in Fig. 7 and a larger one that is very similar to the one described for copper in Fig. 16. For the smaller islands, more surface diffusion is observed and they are finally “swallowed” by the larger islands such that a few hours after deposition, no small islands can be found any more on the surface (Fig. 17b). Even though no atomistic calculations have been performed for Au on MoS2, it is expected that the chemistry and atomic configuration of Au and Ag islands are similar. Thus, similar arguments may explain the relative and temporary stability of the small islands. In addition, the UPS spectra of Au on MoS2 exhibit the same features that are described for Ag in Fig. 10. From the increased surface diffusivity of Au with respect to Ag and the lower diffusivity with respect to Cu, we conclude that the adsorption energy of Au to MoS2 is lower than the one for Ag and higher than the one for Cu. This is supported by the fact that Au islands show morphology features of both Ag and Cu islands.

Summary of observations and calculations on two-dimensional metal clusters

In this section, we have explained the shape and electronic structure of different metal clusters on molybdenite surfaces as a model substrate for sulphide surfaces. Atomistic calculations at the quantum-mechanical and empirical level agree well with island size and mobility observations (STM), and measurements of the surface chemistry (XPS) and electronic structure (UPS and STS). The simulations help to gain additional information on the nature of the metal to sulphide bond, changes in local magnetic properties and local charge distribution in the near surface region. The combination of these tools are valuable in comparing the relative stability of various metal clusters on sulphide surfaces and can, therefore, be used to evaluate adsorption properties of other metal-sulphide combinations in future studies.

Specifically, we have learned that the adsorption of Ag to MoS2 forms a strong bond due to polarization of the Ag atoms by the molybdenite surface. Especially at low concentrations of the adsorbate, where Ag can form small islands on the MoS2 surface, the adsorption of Ag to MoS2 is not hindered from a chemical point of view. The adsorption of Au and especially Cu is expected to be weaker. Thus, from a purely chemical point of view, we expect a slightly higher association of Ag and Au with MoS2 than Cu. However, this study does neither consider to what degree the respective metal does actually co-occur with MoS2, e.g., in a hydrothermal solution, nor does it take into account the concentration of Ag, Au, or Cu.

Future studies may show how strongly Ag can be incorporated in MoS2 interlayers. In addition, the reductive power of MoS2 to adsorb/reduce Ag+ from solution can be studied, similar to the work on the reactions paths from Au3+ to metallic Au on PbS (Becker et al., 1997b).
1   2   3   4   5   6   7   8   9   ...   12


Nanoparticle-host interactions in natural systems iconNanoparticle-host interactions in natural systems

Nanoparticle-host interactions in natural systems iconChapter 25 Greenhouse Gases Emissions from Natural Systems: Mechanisms and Control Strategies

Nanoparticle-host interactions in natural systems iconBuilding Energy Management Systems: Applications to loenergy hvac and natural ventilation control, 2nd Edition

Nanoparticle-host interactions in natural systems iconAchieving sustainable urban water systems: identifying pathways and obstacles in implementing sustainable water systems using a systems thinking approach

Nanoparticle-host interactions in natural systems iconPrecise sub-nm plasmonic junctions for sers within gold nanoparticle assemblies using cucurbit[n]uril ‘glue’

Nanoparticle-host interactions in natural systems iconReliefweb: unep and Sweden host conference on climate impact on Kenya

Nanoparticle-host interactions in natural systems iconApproved Queueing systems and their application to telecommunication systems and dams

Nanoparticle-host interactions in natural systems icon1. Please suggest resources that would be particularly useful for those new to the systems field (and please indicate category to which they would belong). For example, can you think of the first reading/conversation/website that introduced you to systems science?

Nanoparticle-host interactions in natural systems iconKeywords: Turing, Natural Philosophy, Natural Computing, Morphological Computing, Computational Universe, Info-computationalism, Computing and Philosophy

Nanoparticle-host interactions in natural systems iconKeywords: Turing, Natural Philosophy, Natural Computing, Morphological Computing, Computational Universe, Info-computationalism, Computing and Philosophy

Разместите кнопку на своём сайте:

База данных защищена авторским правом © 2012
обратиться к администрации
Главная страница