Silver nanoclusters prepared in water-in-oil emulsions

Stable silver clusters can be prepared by a simple electroless reduction reaction taking place in water-in-oil emulsions. An emulsion containing AgNO3 in the water droplets was mixed with a similar emulsion containing aqueous NaBH4 droplets. The droplet diameter, based on Rayleigh scattering, was 41 nm and the mean number of Ag+ ions in each droplet varied from 2.0 to 21.7 as the concentration increased from 90 μM to 1 mM AgNO3. The low number of Ag+ ions in each droplet inhibits the growth of large nanoparticles and these emulsions do not show the large plasmon band observed for Ag nanoparticles obtained by the analogous reaction in bulk solution at the same Ag+ concentrations. Atomic force microscopy provides evidence of small Ag nanoclusters and a much lower number of larger nanoparticles. Electrospray mass spectrometry suggests that the clusters are mainly Ag4 species coordinated to water and BH4 −. The Ag nanocluster-containing emulsions are fluorescent and show an emission band with a peak wavelength of 427 nm and a Stokes shift of 81 nm from the first peak at 346 nm in the excitation spectrum. The intensity of fluorescence decreased as the [Ag(I)] increased and our most fluorescent samples were prepared from 90 μM AgNO3 because at higher concentrations more Ag nanoparticles are formed. DFT calculations on Agn clusters indicated that Ag4 species favour a planar rhombic geometry even in the presence of coordinating water molecules or BH4 −. However calculations of vertical excitation energies for Ag4 species do not match the experimental excitation spectra and this suggests the fluorescence arises from bright AgNCs of different nuclearity present at lower abundance in the mixture of species produced by the emulsion reaction. Calculated excitation energies for Ag6 give the best fit to the available data.


Introduction
Fluorescent Ag nanoclusters (AgNCs) have been the subject of much interest, because they exhibit properties, e.g. luminescence, that are unexpected for metals [1,2]. They have promise for applications in sensing and imaging because of their photostability, small size compared to quantum dots, and high quantum yields [2,3]. AgNCs are distinct from nanoparticles (AgNPs) which possess a plasmon resonance owing to the metallic nature of the valence electron structure [4]. There is no sharp demarcation between NCs and NPs in terms of size, but luminescence is usually reported for clusters of 30 Ag atoms or fewer [5,6]. The optical properties of AgNCs have been well-studied using vacuum techniques, involving bombardment of silver targets with noble metal ions [7] and also trapping clusters in rare gas matrices [8][9][10][11]. Fluorescent silver nanoclusters have also been formed photochemically in AgO films [12] and laser-writing of AgNCs in a range of solid matrices has been reviewed [13]. However, many desired applications require stable aqueous dispersions of AgNCs and these require different preparation techniques.
NPs are easily prepared by electroless reduction of the metal salt in the presence of a suitable ligand to decrease the NP surface tension and slow the growth of the nanoparticles. Possibly the simplest method is the reduction of AgNO 3 by aqueous reducing agents (citrate, NaBH 4 ) [14]. On the other hand, NCs require more complex preparation techniques. AgNCs have been prepared and studied using pulsed radiolysis of dilute aqueous AgClO 4 [15]. The γ-rays produce solvated electrons and these react with Ag + (aq) to form solvated silver atoms, Ag 0 (aq). Subsequent aggregation of Ag 0 and Ag + species produces the AgNCs. The evolution of the optical absorption spectra with time after the radiolytic pulse yielded information on the sprectroscopic and electrochemical properties of Ag n (aq) clusters for 1 < n < 16 [16]. However, the pulse radiolysis method does not provide a convenient technique for the preparation of stable aqueous AgNC dispersions.
More recently, stable dispersions of AgNCs have been prepared by electroless reduction in the presence of multi-dentate ligands, including poly(methylmethacrylates) [17,18], dendrimers [19] and oligonucleotides [20]. Etching of mercaptosuccinic acid -protected AgNPs has been used to prepare Ag 7 and Ag 8 clusters [21]. The single-stranded deoxyribonucleic acids (ssDNA) possess a wide variety of metal ion binding sites (nitrogeneous bases, phosphate groups) which are able to prevent the aggregation of the AgNCs into larger AgNPs. These solutions show stable luminescence and have interesting sequence-dependent effects [4], but their composition is complex and has hindered their structural characterisation. A combination of chromatography and mass spectrometry has been used for purification and to deduce AgNC/ssDNA structures [22,23]. Recently single crystal x-ray diffraction data was obtained [24,25]. An electrochemical method for the production of small AgNCs and their interaction with DNA has also been demonstrated [26,27]. These preparations are considered to contain a mixture of species which were assigned as Ag 2 , Ag 3 + and Ag 4 2+ using a combination of TD-DFT calculations and experimental excitation-emission spectra [28].
In this report, we investigate an alternative route to stable preparations of AgNCs. Instead of employing a multifunctional ligand (ssDNA) to stablise the NCs, we restrict the available number of Ag atoms by carrying out the electroless reduction in the aqueous droplets of a water-in-oil microemulsion. Microemulsions have often been used to prepare NPs [29,30], but less frequently to prepare NCs. Previous researchers have used control of the stoichiometry of metal ion : reducing agent in microemulsion systems [31] or a method of kinetic control in microemulsions [32] to prepare AgNCs. In this work, we restrict the concentration of metal ions such that each droplet of the emulsion contains only a few metal ions. These AgNC preparations are stable and we provide evidence that they contain Ag 4 species by mass spectrometry. The optical spectroscopy of the emulsions is investigated and compared to quantum chemical simulations.
Deionized water was obtained from a Barnstead Nanopure TM purification train (model DH 931), Barnstead International, Dubuque, Iowa, USA) with a nominal resistivity of 18.2 MΩ cm.
Resprep SPE solid phase extraction cartridges (Restek, Pennsylvania, USA) with a bonded reverse phase hydrophobic silica-based adsorbent were used to separate the water phase from water-in-oil emulsions. Each cartridge has a tube size of 1 ml, C18 (high load, endcapped), with a bed weight of 100 mg.

Synthesis of AgNCs in water-in-oil emulsions
Preparation of 90 μM -1 mM AgNCs by chemical reduction. AOT (0.2223 g, 0.500 mmol) was dissolved in 5 ml isooctane in a 50 ml beaker with gentle sonication and the resulting solution was divided into two equal portions in separate 15 ml glass vials. Separate equimolar aqueous solutions of AgNO 3 and NaBH 4 were prepared at the desired concentration. Next, 45 μl of the AgNO 3 solution was added with stirring (450-500 rpm) to one portion of the AOT/isooctane solution to form an AgNO 3 (aq)/AOT/isooctane microemulsion. 45 μl of the NaBH 4 solution was added with stirring (450-500 rpm) to the other portion of the AOT/isooctane solution to form an NaBH 4 (aq)/AOT/isooctane microemulsion. A clean magnetic stirrer was inserted in the AgNO 3 microemulsion and the NaBH 4 microemulsion was added within 30 seconds. The solution was kept on the stirrer for another 5 min at 450-500 rpm until the sample colour became stable.
Preparation of 90 μM AgNCs by photochemical reduction. In a few experiments, photolysis (5 h duration) of AgNO 3 emulsions instead of reduction by BH 4 − was employed. The photochemical experiments were performed in a Rayonet ultraviolet photoreactor (The South New England Ultraviolet Company, Connecticut, USA). It has an A.C power source of 120 V, 50/60 Hz, and a mercury light sourcewith a major emission wavelength of 254 nm.
2.3. Extraction of the aqueous phase from the microemulsions Electrospray mass spectrometry of the microemulsions was not possible directly on the as-prepared samples, because of the large concentrations of isooctane and AOT. The aqueous phase was therefore separated using a 1 ml solid phase extraction chromatography cartridge (Resprep SPE, Restek Corporation, USA). To separate the aqueous from the organic phase in the microemulsion, 500 μlof the sample was pipetted into a clean separating funnel mounted on a retort stand and another 500 μl deionized water was added to it to facilitate phase separation. This was allowed to stand for 3-5 minutes before withdrawing the aqueous phase (more dense of the two) at the bottom of the flask into a clean glass vial. The cartridge was fitted onto a Büchner funnel connected to a vacuum pump. Then 1 ml deionized water was eluted with the help of the vacuum pump to condition the cartridge. Next,the aqueous phase (more dense fraction) from the separation flask was pipetted into the funnel and allowed to elute slowly under vacuum (one drop at a time) until no more clear droplets appeared and without generating white surfactant froths.

Optical absorption and emission spectroscopy
UV-Vis absorption spectra of the AgNCs were obtained using a Varian Cary 100 Bio spectrophotometer with a tungsten halogen visible source and a deuterium arc ultraviolet lamp. The maximum scan range of the instrument was 250-800 nm and a 500 μl quartz microcuvette (Hellma UK Ltd) was used. Absorbance readings for the samples were obtained after synthesis by withdrawing 500 μl aliquots into the 500 μlquartz microcuvette and run against isooctane as background. Isooctane itself was observed to have a strong absorption below 220 nm and AOT to have a strong absorption below 240 nm. Neither substance has significant absorption in the range of the data presented (λ > 250 nm). Samples of AgNPs prepared in water were measured with deionized water as the background. Rayleigh scattering measurements were also performed using the Varian Cary 100 Bio spectrophotometer. The intensity ratio of light that is scattered by the emulsion I to incident light I 0 is given by equation (1). is the ratio of refractive indices of the dispersed phase and the medium, L is the pathlength =1 cm, d is the droplet diameter in the emulsion and λ is the wavelength. The droplet number density ρ was calculated from the volume fraction of water used and the volume of a droplet, assumed spherical: V d 6 3 = p . The fraction of scattered intensity can be converted to an effective absorbance A using equation (2) and fitted to the data to determine the droplet diameter using the method of least squares. Fluorescence emission and excitation spectra were recorded with a SPEX FluoroMax TM spectrofluorimeter instrument, New Jersey, USA. The excitation source was an ozone-free xenon lamp. A photomultiplier served as emission detector and there was also a photodiode reference detector. The maximum emission/excitation range of the instrument was 200-950 nm. The spectra shown have not been subjected to background subtraction. Instead, the control sample (an isooctane/AOT/water emulsion) is indicated in the caption to figure 13. No emission was detected from the isooctane or AOT under the conditions employed.

Mass spectrometry
A Waters LCT Premier Electrospray ionization mass spectrometer with Masslynx TM 4.1 software was used. The instrument has a time of flight (TOF) analyzer and a microchannel plate (MCP) detector. The reference sample was leucine enkephalin. Spectral simulation was carried out using Gabedit [33].
1 μl of the extracted aqueous phase of the 90 μM microemulsion was diluted in 199.0 μl water at a dilution factor of 1:200 before analysis. A dilution of 1:1000 was used for the 1 mM samples. Spectra were obtained in negative ion mode with deionized water as the background.

Atomic Force Microscopy (AFM)
Silicon wafer was cut into approximately square chips of area 1 cm 2 . The chips were cleaned by soaking for 15 minutes in 3:1 v/v conc H 2 SO 4 : 30 vol H 2 O 2 , then rinsed thoroughly with copious amounts of deionized water and blown dry with N 2 gas. The aqueous phase of the AgNC samples was separated by solid phase extraction (see section 2.3) and 1 μl was spotted on a Si chip . The samples were allowed to dry overnight in a laminar flow hood and then imaged in air.
Atomic force microscopy (AFM) images were acquired on a Bruker MultiMode 8 instrument using ScanAsyst in Air mode. Silicon tips on silicon nitride cantilevers (ScanAsyst, Bruker) were used for imaging. The nominal tip radius was approximately 2 nm, the resonant frequency was 150 kHz, and the spring constant was k ; 0.7nm −1 . The image data was analyzed using Gwyddion 2.41 software (http://gwyddion.net/) [34].

Quantum chemical simulations
The structures of Ag 4 clusters and their derivatives were optimised using density functional theory with a pseudopotential basis set (SBKJC) [35][36][37] using the Firefly quantum chemistry program (Alex A. Granovsky, http://classic.chem.msu.su/gran/firefly/index.html) which is partially based on the GAMESS (US) [38] source code. The same code was used to calculate vertical excitation energies at the equilibrium geometry of the ground state using time-dependent density functional theory (TD-DFT) and the B3LYP functional. Some calculations requiring UHF determinants were carried out using GAMESS (US, version 12 Jan 2009, R3) [38,39] which allows for TD-DFT calculations on ground states described by unrestricted determinants. Finally, coupled cluster calculations and TD-DFT calculations using the PBE0 [40,41] and M11 [42] functionals were carried out using Gaussian 09 [43].

Results and discussion
A well-known procedure for the formation of AgNPs involves the addition of an aqueous solution of an Ag(I) salt to a solution of a reducing agent [14]. Small Ag clusters rapidly form and grow to produce larger, metallic nanoparticles (AgNPs) with radii on the scale of tens of nm. Dispersions of roughly spherical metallic AgNPs prepared by this method have a characteristic yellow colour owing to the strong plasmon resonance near 400nm. In this report we are interested in preparing stable nanoclusters (AgNCs) which are much smaller (30 Ag atoms) and show very different, non-metallic electronic and spectroscopic properties. Water-in-oil emulsion techniques have previously been used to prepare AgNCs [31,32]. The approach taken in this work is to restrict the aggregation number by controlling the metal ion concentration in the aqueous droplets of a microemulsion and therefore the number of available metal atoms in each droplet. Figure 1 shows the concept employed to form the AgNCs and to restrict subsequent growth to minimise the formation of AgNPs.
Water-in-isooctane (2,2,4-trimethylpentane) emulsions were prepared by dissolving AOT in isooctane to a concentration of 1 mol dm −3 and then injecting small volumes (45 μl) of aqueous media into 2.5 ml isooctane solution with rapid stirring. Emulsions containing AgNO 3 (aq) and NaBH 4 (aq) were prepared separately and then mixed as emulsions. The small volumes of the droplets of aqueous solution in these emulsions restrict the number of metal ions available to react. When an AgNO 3 emulsion is mixed with an NaBH 4 emulsion, reduction takes place between droplets of each type, but with a much more limited amount of metal than when the reaction is carried out directly in aqueous solution. Below we present evidence that the products (referred to as AgNC emulsions) contain a mixture of AgNC species and also contain a small number of AgNPs. We characterise the optical properties of these emulsions, then provide evidence that they contain small Ag clusters and finally we investigate the fluorescence spectra of the AgNC emulsions.
3.1. Preparation and characterisation of silver nanoclusters in water-in-oil emulsions 3.1.1. Optical spectra of nanoclusters prepared in emulsions and comparison with aqueous preparations of nanoparticles figure 2(a) shows optical absorption spectra of AgNC emulsions prepared from AgNO 3 solutions of concentrations between 0.09 mmol dm −3 and 1 mmol dm −3 . The absorption spectrum of the AgNC emulsions was observed to be unchanged for at least 96 h, indicating the stability of the products during this period. For comparison, figure 2(c) shows optical absorption spectra of AgNP preparations from bulk solutions of matching concentrations of aqueous AgNO 3 . The spectra of the AgNPs show the characteristic plasmon band in the optical spectra at 394 nm (1 mmol dm −3 sample) and 405 nm (0.09 mmol dm −3 sample). The sharp plasmon band is notably absent from the AgNC emulsions, which are pale pink in contrast to the yellow colour of the AgNP dispersions and have broad absorption extending across the visible spectral range until about 750 nm. A similar pink colour was reported upon formation of AgNCs by photochemical reaction of AgNO 3 in the presence of poly(methylmethacrylic acid) [17].
The sharply rising absorbance near λ = 250 nm in figure 2(a) is due to the onset of the absorption band of AOT which peaks below 250 nm and sets a limit on the useful wavelength range. The absorption spectrum of the AgNCs also overlaps with the region of the spectrum in which strong elastic scattering occurs. In figure 2(a) there is, however, one clear peak at about 275 nm that has previously been assigned to Ag 4 2+ species [44,45]. This assignment is supported by TD-DFT calculations on Ag 4 2+ collected in table 1 using a pseudopotential basis and the PBE0 and M11 functionals. The optimised geometry of Ag 4 2+ was observed to be a tetrahedron with bond lengths of 0.28-0.29 nm. The computed spectra show a dominant peak from a triply-degenerate singlet state at 267 nm (vacuum) and 299-305 nm (in aqueous media) in agreement with previous calculations [46]. The DFT methods were supplemented with a coupled-cluster calculation on a geometry that was constrained to have exact tetrahedral symmetry; this calculation supports the TD-DFT results.
The obvious difference between the absorption spectra of bulk solution preparations and of AgNC emulsions indicates that different species are formed in these two cases. Figure 2(b) shows the variation of the absorbance of the AgNC emulsion (figure 2(a)) in the region of the plasmon band at λ = 400 nm with concentration of AgNO 3 in the preparation. At low concentrations, the absorbance rises linearly with concentration, but above about 0.5 mmol dm −3 there is a change in slope. We interpret this observation as evidence for the formation of a small number of AgNPs, which possess a surface plasmon resonance, in the AgNC emulsion system as the AgNO 3 concentration is increased and the number of metal ions per droplet rises. The presence of AgNPs is confirmed below using microscopy.
In order to quantify the droplet volume considerations above, we carried out Rayleigh scattering measurements of the AgNO 3 emulsions to estimate the droplet size. The apparent absorption of light by an emulsion is shown in figure 3. A droplet diameter of 41 nm was estimated from this data by least squares using equation (2) and a linear background as the regression model. The only parameter related to the emulsion which is floated is the droplet diameter d.
The corresponding droplet volume is 3.6 × 10 −17 ml and the mean number of Ag + ions present in each droplet is 2.0 for a 0.09 mmol dm −3 solution, rising to 21.7 for a 1.0 mmol dm −3 solution of AgNO 3 . It is clear that even for the highest AgNO 3 concentrations employed, AgNPs in the 10 nm size regime cannot be formed by the reduction of Ag(I) in a single droplet. This does not rule out the eventual formation of AgNPs by coalescence of multiple droplets, but it does indicate why the plasmon resonance band which is responsible for the yellow colour of AgNPs is suppressed in these emulsion preparations. It is also worth bearing in mind that the number of Ag + ions per droplet does not necessarily determine the cluster sizes in the emulsions because multiple smaller clusters may be formed if certain species are particularly stable. This is illustrated in figure 1. Below we provide evidence that the clusters are mainly Ag 4 species, but there are other clusters present and some larger AgNPs, especially in the samples prepared at higher concentrations of AgNO 3 .

Atomic force and transmission electron microscopy
To prepare samples for imaging, the AgNC emulsion was mixed with an equal volume of deionized water, the aqueous phase was separated from the organic phase and then subjected to solid-phase extraction to remove the remaining traces of organic components. The remaining aqueous phase was drop-cast onto either oxidised Si chips (AFM) or holey carbon (TEM). Figure 4(a) shows an atomic force microscope image of a 1.0 mmol dm −3 sample prepared in a similar manner to the TEM samples and drop-cast onto an Si chip. The surface is covered in small bright features on the nanometre scale. The height profile along a line is presented in figure 4(b); This shows that the feature height is 0.5 nm for the majority of these objects, although there are a few larger particles with heights of about 1.5-2.0 nm. Similar AFM images have been reported for CuNCs prepared in microemulsion [31] and similar STM images reported for AgNCs prepared in microemulsion [32]. Following these authors, we interpret the small features in figure 4 as AgNCs with a height indicating clusters 1 or 2 Ag atoms high. The lateral dimensions may reflect tip convolution effects rather than real cluster size and are therefore not interpreted in detail. Figure 5 shows a larger field of view of the same sample. In this image, features of height up to about 15 nm are visible. These objects are much less abundant than the small features discussed in figure 4, but they are large    Table 1. DFT calculations on Ag 4 2+ using an effective core potential and basis set [35][36][37]. The polarizable continuum model was used to model water as the solvent [47]. The optimised geometries were all observed to be tetrahedral and the mean and standard deviation of the 6 Ag...Ag bond lengths are reported. The max l was obtained from the major peak in the absorption spectrum calculated using TD-DFT at the equilibrium geometry. The lowest excited state is triply degenerate in an exact tetrahedral geometry and the values quoted are the mean and standard deviation obtained from the three lowest vertical excitation energies. a Tetrahedral symmetry was enforced in the coupled-cluster calculation and the excited states were computed by the equation of motion method. enough to sustain a plasmon resonance and contribute to the increase in absorbance at λ = 400 nm observed in figure 2(b). Transmission electron micrographs of the AgNC emulsions were obtained by drop-casting the aqueous phase onto holey carbon films. Figure 6 shows two images of these samples. The presence of weakly scattering residual surfactant molecules not removed by the solid phase extraction is evident in the images, however small dark particles (examples circled in red) are clearly visible. These are a comparable size to to the larger particles observed in figures 4 and 5 by AFM and larger than the small clusters studied by mass spectrometry below. Similar particles have been reported in samples of poly(methacrylic acid)-stablised AgNCs, which are also mixtures of species of different nuclearity [17]. The TEM images provide indirect evidence that the features are metallic Ag because organic matter or residues comprising low atomic number elements (B, O) would not be expected to scatter electrons so strongly. Below we provide more direct characterisation of the composition of the AgNCs using electrospray mass spectrometry.

Mass spectrometry
Further information on the species produced by reduction of Ag + in microemulsion droplets was obtained by mass spectrometry. The emulsions are stabilised with AOT and in order to apply the technique, the aqueous phase must first be separated from the isooctane by solid phase extraction chromatography. The aqueous phase  was then examined by electrospray ionisation mass spectrometry and the spectra are shown in figures 7 and 8. Both spectra were acquired in negative ion mode; no signals from Ag clusters were obtained in positive ion mode.
The data of figure 7 shows the mass spectrum for a sample prepared by mixing 0.09 mM emulsions of AgNO 3 and NaBH 4   is therefore expected for a singly-charged anion. There are weaker peaks at m/z values intermediate because B has two isotopes with m = 10, 11 of approximately 20% and 80% abundance. All the features in figure 7 are assigned to Ag 4 and not to other Ag clusters. It is not possible to completely exclude the possibility of fragmentation of larger Ag n clusters during the electrospray ionisation process, but no Ag clusters were detected at lower or higher m/z outside the range shown. The Ag 4 cluster and derivatives appear to be the major species in these samples and quantum chemical calculations were carried out to investigate the likely structures. Figure 8 shows a sample prepared by photoreduction of a 0.09 mM AgNO 3 emulsion using UV light (λ = 254 nm). These samples provide controls to support the role of BH 4 − anion in the data of figure 7. Again, groups of peaks with a characteristic isotope pattern for Ag 4 are observed; these groups are shifted to lower m/z than in figure 7 because of the absence of BH 4 − . The structure of each group is slightly more complicated because there is evidence of doubly-charged anions formed by aggregation of two identical Ag 4 species in each group. Figure 8 compares the experimental and simulated spectra for the cluster assigned to Ag 4 (H 2 O)(OH) − . The  relative amounts of that cluster and its dimer were determined by least squares regression to be 51%:49% dimer: monomer. Other Ag 2 and Ag 3 species were observed at lower m/z.

Quantum chemical calculations of Ag n clusters
The optimised geometries of the Ag n species for 1 < n 6 using the B3LYP functional and SBKJC basis set were all observed to be planar, with the exception of Ag 5 . In contrast to tetrahedral Ag 4 2+ , the neutral Ag 4 cluster is a rhombus with internal angles of 56.7°and 123.3°. Such a structure has been reported in previous DFT calculations [48] and its optical spectrum is consistent with experimental data obtained by trapping Ag n clusters in noble gas matrices [9,11]. The Ag-Ag bond lengths defining the sides of the rhombus are 0.28nm and those defining the diagonals that bisect the internal angles are 0.266 nm and 0.49 nm. The short bond length of 0.266 nm is notably less than twice the atomic radius (0.144 nm) calculated from the lattice parameter of FCC bulk Ag of 0.4085nm [49]. This indicates a strong interaction between the two Ag atoms involved. The Ag 5 cluster consisted of a prism obtained by adding the 5 th Ag atom above the midpoint of the rhombus and some relaxation of the other 4 Ag atoms. The neutral Ag n clusters have been well-studied in noble gas matrices [9,11], but these are unlikely to remain uncoordinated in our aqueous system. Structures of several Ag 4 species relevant to the mass spectral data, optimised at the B3LYP/SBKJC level of theory, are shown in figure 9 and the frontier orbitals are shown in figure 10. The Ag 4 unit is approximately planar and the optimised geometry of Ag 4 species showed the same rhombus motif in the presence of coordinated water molecules (figures 9(c,d)) and for Ag 4  As expected from the molecular geometry, the bound water molecule in Ag 4 H 2 O has relatively little effect on the HOMO, but there is substantial distortion of the LUMO. For the symmetrical Ag 4 (H 2 O) 2 species, the HOMO is again similar to that of Ag 4 because the water molecules lie near the nodes, however the LUMO is affected by the water molecules.

Fluorescence spectra of nanoclusters and TD-DFT calculations
Metal nanoparticles typically show a plasmon resonance and weak or no fluorescence owing to efficient nonradiative decay channels. However, sufficiently small metal clusters are known to show strong emission which is attributed to the opening of a gap in the electronic spectrum [12,[50][51][52][53]. Figure 11 shows an excitationemission map for a 0.09mmol dm −3 AgNC emulsion. The emission near 430 nm has a peak position independent of excitation wavelength, but its intensity varies and is a maximum for an excitation wavelength of about 340 nm. The presence of this feature only in mixed AgNO 3 and BH 4 − emulsions is the reason it is assigned  Figure 11. Fluorescence excitation-emission map for a sample prepared from 0.09 mmol dm −3 AgNO 3 . Raman peaks and features due to second order scattering show as black pixels.
to fluorescent AgNCs. An additional, weaker emission feature is observed near 600 nm with an excitation wavelength near 280 nm. Examples of red-emitting AgNCs include poly(methacrylic acid)-stabilised AgNCs (λ em ; 625 nm, λ exc ; 525 nm) [17] and DNA-stabilised AgNCs (λ em = 721 nm, λ exc = 640 nm) [54]. Red emission from DNA-stabilised AgNCs with similar large, apparent Stokes shifts, which likely arise from energy transfer, is known [20]. Based on the excitation wavelength, it might have been assigned to the Ag 4 2+ species, but that has been reported to emit near 340 nm [28]. The origin of this 600 nm emission peak in our samples remains unclear.
The excitation spectrum of the λ = 430 nm emission peak is shown in figure 12 as the red line. The green line is the emission spectrum at a fixed excitation wavelength of 340 nm. The sharp features in these spectra near λ = 380 nm arise from Raman scattering from isooctane. The Stokes shift is 81 nm measured from the first peak in the excitation spectra at λ = 346 nm to the emission maximum at λ = 427 nm. The black lines indicate the wavelengths corresponding to the vertical transitions to and from the first excited state of Ag 6 computed by TD-DFT and discussed below. A second peak in the excitation spectrum is also observed at λ = 318 nm. Unfortunately we cannot find a clear candidate for this transition in the TD-DFT calculations on AgNCs. Figure 13(a) shows a set of emission spectra as a function of the concentration of Ag + employed in the preparation. There is a broad, unstructured emission band with a maximum at about 430 nm in the Agcontaining samples that is absent in the blank emulsion; this control confirms the emission is not due to isooctane or AOT. This feature is also absent from spectra of aqueous solutions of NaBH 4 or AgNO 3 . We therefore attribute the λ = 430 nm emission to AgNCs present in the aqueous droplets of the emulsion. The sharp feature near 380 nm is due to Raman scattering from the isooctane, which is the dominant component of the sample by mass. Interestingly, the emission intensity decreases as the concentration of Ag + increases despite the absorbance values at the excitation wavelength being no more than 0.1 ( figure 2(a)). The fluorescence intensity drops by a factor of more than 2 as the concentration increases by a factor of 11. However, the inner filter effect can only account for at most a 100 10 1 25% 0.1 ( ) -= reduction of the 11-fold increase due to the concentration. The data therefore presents evidence that as [Ag + ] increases, the Ag n species present in the emulsion droplets change and the species responsible for the 430 nm emission is favoured at low [Ag + ]. Based on the observation of AgNPs by microscopy in figure 5 and the non-linear concentration-dependence of the absorbance at the wavelength of the plasmon resonance in figure 2), we suggest that AgNPs are responsible for the decline in fluorescence at high [Ag + ]. Quenching of AgNC emission by AgNPs has also been reported for AgNCs prepared by laser ablation of AgNPs [46].

Time-Dependent DFT calculations of Vertical Excitation Energies
In order to further investigate the nature of the transitions responsible for the optical spectra, we carried out time-dependent density functional theory (TD-DFT) calculations on a range of Ag n clusters to estimate the excited state energies. Figure 14 shows the absorption spectra of Ag 2 -Ag 6 clusters calculated using the SBKJC pseudopotential basis set [35][36][37] and the B3LYP functional. In all cases the ground-state geometry was Figure 12. Normalised fluorescence excitation and emission spectra. The sample was prepared from 0.09 mmol dm −3 AgNO 3 . The black lines indicate the vertical excitation energies in Ag 6 from TD-DFT calculations using the M11 functional and discussed in the text. The line at λ = 347 nm is calculated at the optimised geometry of the ground state and that at λ = 441 nm from the optimised geometry of the first excited state. optimised first at the same level of theory and the vertical ionization energies were calculated from this geometry. In the case of the even numbered clusters, the ground states are spin singlets and the relevant excited states for absorption transitions are also singlets because of the spin selection rule. Ag 3 and Ag 5 clusters were assigned as  doublets. Ignoring vibronic effects, the vertical excitation energies ought to approximate the energies corresponding to the wavelengths of the peaks in the absorption spectrum or the excitation spectrum of the fluorescence.
The calculated absorption spectra of Ag 2 and Ag 4 show peaks at 411 nm (3.02 eV) and 415 nm (2.99 eV) respectively. A coupled cluster (EOM-CCSD) calculation on Ag 4 using the same basis set gave a major peak at 425 nm (2.92 eV) and another calculation, using the range-corrected M11 functional [55], indicated a peak at 433nm (2.86 eV). These values are the same, within the typical accuracy achieved by time-dependent DFT, of the values previously computed and measured for Ag 2 and Ag 4 in noble gas matrices [9,11,48]. The Ag 3 species show an absorption at 523nm which corresponds to excited states that are much lower in energy than suggested by the experimental spectra (figures 11 and 12). Ag 5 has a major peak at 362 nm (3.42 eV). The calculated spectrum of the Ag 6 cluster shows a very strong absorption at 375 nm (3.31 eV) in agreement with a previous calculation for an Ag 6 cluster of D 3h symmetry [9]. Reported experimental data in noble gas matrices for Ag 6 shows an absorption near 359nm (3.45 eV) [11] or 342 nm (3.63 eV) [9]. In additional calculations, using the range-corrected M11 functional, we found a vertical excitation energy of 347nm (3.57 eV) for this state at the equilibrium geometry of the ground state. At the optimised geometry of the first singlet excited state, the vertical excitation energy, which ought to correspond to the emission by Kasha's rule, was 441 nm (2.81 eV). This calculation is in reasonable agreement with the peak in the emission spectrum at 430 nm (figures 12, 13), although it should be noted that the transition is dipole-forbidden. Nevertheless, the best candidates to explain the experimental spectra, based simply on comparing the calculated vertical excitation energies to the experimental spectra ( figure 12), are the Ag 6 clusters.
Although the mass spectrometric data ( figure 7) indicates that Ag 4 species predominate in the emulsions, the excitation spectrum of the AgNC emulsion ( figure 12) is in disagreement with the computed vertical excitation energies (figure 14) of bare Ag 4 . We therefore carried out TD-DFT calculations on the various Ag 4 species of figure 9 to determine if the difference could be explained by coordination of water and BH 4 − to Ag 4 . The corresponding absorption spectra simulated from the vertical excitation energies obtained in a time-dependent DFT calculation are shown in figure 15. The coordination of two H 2 O causes a small redshift of the first absorption peak ( figure 15) but the band shape is unchanged. In the case of Ag 4 BH 4 − , another excited state appears at low energy (;560 nm), but this is absent in the solvated Ag 4 (H 2 O) 2 BH 4 − . Overall, the effect of coordination of water and BH 4 − on the absorption spectra of Ag 4 is small, the major peak in the spectrum is expected between λ = 415 − 427 nm and these species cannot account for the experimentally observed peak at 346 nm in the excitation spectrum of the fluorescence. We conclude that the fluorescent species are most likely to be clusters of different nuclearity, present in lower abundance, but with brighter fluorescence.

Conclusions
Fluorescent silver nanoclusters (AgNCs) are conveniently prepared by mixing water-in-oil emulsions that contain droplets of aqueous AgNO 3 and of NaBH 4 . Restriction of the Ag(I) numbers in the small droplets of the ). Time-dependent DFT was used to compute vertical excitation energies to excited singlet states from the geometry-optimized ground states of figure 9. The experimental excitation spectrum is shown in black for comparison.
aqueous phase restrains growth of the Ag(0) species and favours small Ag clusters. Equivalent reactions in bulk solution produce dominantly AgNPs with their characteristic plasmon resonance near λ = 400 nm. In contrast the AgNC emulsions produced by this method contain mixtures of AgNC species and a few larger AgNPs because the number of Ag + ions in a droplet is insufficient to form a NP. Optical absorption spectra and quantum chemical calculations provide evidence for the presence of the Ag 4 2+ ion. No fluorescence is observed in the absence of either AgNO 3 or NaBH 4 and therefore we also assign the fluorescence to transitions in Ag n clusters. Direct evidence for both silver nanoclusters and nanoparticles was obtained by atomic force and electron microscopy. Electrospray mass spectrometry provided evidence that the majority of the AgNC species are based on a Ag 4 core surrounded by complexed water and borohydride, such as Ag 4 (H 2 O) 2 BH 4 − , although fragmentation of larger Ag n clusters cannot be excluded. Optimised geometries from DFT calculations indicate the rhombic (D 2h ) geometry of bare Ag 4 is retained by the silver core of the cluster after coordination by water and borohydride. However, excitation spectra of the emulsions show overlapped peaks at 318 nm and 346 nm, which are at energies significantly higher than calculated for any of the Ag 4 species observed by mass spectrometry. This suggests that the predominant species observed in mass spectrometry are not responsible for the fluorescence. The calculated absorption and emission data for Ag 6 is in better agreement with experiment, but it is acknowledged that these species have not been confirmed independently of the optical spectroscopy. Finally, as the concentration of Ag(I) in the emulsion increases, the fluorescence intensity decreases which may be attributed to the formation of larger, non-emissive particles such as those observed in the atomic force microscopy images.