In the experiment, nanoisland films were formed on the glass surface after processing the sample in hydrogen. AFM images of the sample surface before and after 20-min thermal processing in hydrogen at 200°C are presented in Figure 1a,b. In Figure 1b, one can observe densely packed nanoparticles with diameters of about 60 nm on the glass surface. The shape of the nanoparticles is close to hemispherical or hemiellipsoidal, with longer axis perpendicular to the glass surface and wetting angle of about 90° (see Figure 1c).

Figure 1b,d shows that the centers of nanoparticles are aligned. Such an alignment is due to the microscratches created on the surface in the process of glass polishing. The surface defects associated with these microscratches become the nucleation sites giving rise to the long range order in the created island film 7 . It is worth noting that this phenomenon can be used to manufacture island films with prescribed geometry via formation of the required surface defects pattern on the glass surface.

The island film can be easily removed from the glass surface with e.g., water-wet cotton. It can be also transferred from the glass to a flexible substrate (e.g., scotch tape) that could be of interest for biomedical applications. After removal of the island film from the glass surface, e.g., by cleaning or transferring to a scotch tape), it can be easily restored by the thermal treatment of the sample in hydrogen atmosphere. Figure 1e shows the AFM image of the sample surface after 9 removal-anneal cycles. In this case we used mechanical cleaning of the glass surface with wet cotton.

Each annealing process at 200°C was 20 min long. Although particle size dispersion and packing density change after several annealing cycles (compare Figure 1d and e), our results demonstrate that the same silver-enriched glass substrate can be used in multiple SERS experiments. Figure 2 shows optical absorption spectra after several ‘anneal-removal’ cycles. One can readily observe from Figure 2 that the surface plasmon resonance (SPR) in silver nanoparticles 8 in the region of 450 nm dominates the spectra. The developed technique enables producing films of different morphology via varying both processing conditions (temperature and duration of the annealing) and number of removal-anneal cycles. Detailed studies of the dependence of the island film morphology on the processing mode will be presented in our forthcoming paper.

The process responsible for the formation of the MIF on the surface of metal-enriched glass is often referred to as the reactive diffusion. In this process, the diffusion of hydrogen ions into the glass is followed by the chemical reaction, in which silver ions bonded to non-bridging oxygen atoms are replaced by hydrogen 9 :

-O⁻-Ag⁺ + 1/2H₂ → -O⁻-H⁺ + Ag⁰,

where O⁻ is non-bridging oxygen atom. This reaction results in changing of the glass composition in the subsurface region. The neutral silver atoms created in this reaction move towards the glass surface, nucleate, and finally form nanoislands of metal silver on the surface of the sample.

Optical absorption spectra of the silver nanoisland films formed using different number of removal-anneal cycles are presented in Figure 2. One can observe that the magnitude of SPR peak decreases with the number of annealing cycle and that the SPR disappears after cleaning of the glass surface. This experimental finding indicates that after the first thermal processing, silver nanoparticles are created on the surface of the substrate rather than in the bulk.

One can also observe from Figure 2 that the SPR ‘survives’ after 9 removal-anneal cycles (curve 3 in Figure 2). However, we found that after the final cleaning of this sample (curve 4 in Figure 2), the SPR peak remains in the absorption spectrum. This shows that the increasing reactive diffusion time results in the accumulation of silver nanoparticles not only on the glass surface, but also in the subsurface region of the phosphate glass. The modeling of the mechanism of the island film formation is presently in progress and it will be presented soon in our next paper.

It is necessary to note that in silicate glasses, the formation of nanoparticles on the surface was also both predicted 9 and experimentally observed 10 . However, in silicate glasses, the silver-sodium ion exchange and annealing in hydrogen result in silver nanoparticles growth predominantly in the subsurface layer 8 . In contrary, our results show that in phosphate glass, the main sink for neutral silver atoms is the glass surface. Thus, the formation of nanoisland silver films in the course of reactive diffusion is more efficient in phosphate rather than in silicate glass. Such a difference may originate from a higher solubility of neutral silver in phosphate matrixes in comparison with silicate ones.

Figure 3 shows absorption spectra of the silver nanoisland films before (curve 1) and after the R6G film deposition (curves 2 and 3).

One can see from Figure 3 that at 10⁻⁷ M concentration of R6G, the R6G absorption resonance is not observable, while the SPR peak is red-shifted (compare curves 1 and 2). At 10⁻³ M R6G concentration, the R6G absorption resonance dominates the spectrum in 500 to 600 nm range, while the SPR peak remains at its initial position (compare curves 1 and 3). The reverse shift of the SPR may be due to the change in the R6G film composition and hence, the position of the electron resonance when the R6G concentration increases. Specifically, the detailed investigation of the silver MIF absorption spectra for different concentrations of deposited R6G performed by Zhao et al. 11 has shown that the thickness of the deposited R6G layer and its composition (fraction of monomers, dimers, and more complex aggregates) influences the shift, width, and amplitude of the SPR peak.

To study the performance of the manufactured MIF in SERRS, we performed Raman mapping of the R6G 10⁻⁷ M film with spatial resolution of 10 μm. Figure 4 shows the intensity distribution of the 1,651-cm⁻¹ Raman line of R6G (with background subtracted) over 60 × 70 μm² sample region. One can observe that the intensity of the SERRS signal is nearly the same over the whole scanned area except in one point, in which the intensity exceeds 15 times the average value. The regions where signal essentially exceeds the average one are called ‘hot spots’ 1 . The higher signal from the vicinity of the hot spot is due to the stronger enhancement of the local electromagnetic field by closely placed metal nanoparticles in comparison with an isolated metal nanoparticle. However, we have no evidence to suggest that the recorded signal is due to the hot spots because of the large area of the probing beam (10 μm) that should provide the signal about the same magnitude over the entire area of our scan 12 . It is more likely that the anomalously high signal is due to a surface defect where the R6G coating is thicker than the average. In Figure 4, we present SERRS spectra of the 10⁻⁷ M film measured in the regions of the scanned area shown in the inset. Curves 1, 2, and 3 represent Raman spectra measured in areas where 1,651-cm⁻¹ line intensity was below, above, and anomalously high in comparison with the average value, respectively. One can see here that even without anomalous spot enhancement (curves 1 and 2), the signal/noise ratio in the SERRS spectra is high enough to detect 10⁻⁷ M R6G concentration.