Nanostructured optical fibre for use as miniature surface-enhanced Raman scattering sensors

Surface-enhanced Raman scattering (SERS) is a powerful spectroscopic tool for detecting low concentrations of many substances. The SERS effect occurs when a Raman-active molecule interacts with the electromagnetic (EM) field generated by surface plasmon excitations in a metal surface having nanoscale structural features. As a result of the interaction, the typically weak Raman scattering is amplified by many orders of magnitude. Despite its usefulness, one of the factors that has so far prevented SERS from moving out of the laboratory and into the marketplace is a lack of affordable and reliable nanostructured substrates. Since the discovery of the SERS effect over 30 years ago, many different techniques have been developed for producing the necessary substrate structures. Common techniques used include mechanical surface roughening, colloidal metal solutions, metal island films and lithographic techniques. While all produce SERS capable surfaces they either rely on random processes (which limit sample-to-sample reproducibility) or expensive manufacturing equipment. Both of these issues prevent cost effective commercialisation. A novel alternative, which is investigated in this thesis, is to use an optical fibre based system. The fibres presented in this work are based on highly modified medical imaging fibres. These fibres consist of a fused bundle of single-mode fibres ('pixels') drawn in such a way that each fibre maintains its position along the length of the bundle. Previous works have shown that a selective chemical etchant can be used to erode the cores of these fibres leaving an array of wells. The work presented in this thesis shows that, by drawing the imaging fibre until the spacing between the cores isof the order of hundreds of nanometres, the structures formed after etching are of a suitable scale for SERS. The glass itself is incapable of producing surface plasmons, so in order to generate the SERS functionality, the structures are coated with a thin layer of metal. Initial fibre SERS sensors were produced by manually drawing commercially available imaging fibres to smaller diameters. As a result of the draw process, the initial inter-core spacing (the distance between two adjacent pixels) was reduced from approximately 4 μm to 150 - 650 nm. The cleaved tips of the fibres were then etched using a modified buffered hydro°uoric acid solution which resulted in the formation of nanostructures. The structures were then coated with silver and the fibre immersed in a solution of thiophenol for SERS testing. Due to di±culties in determining exactly how many molecules contributed to the SERS signal, an accurate measure of the SERS enhancement factor cannot be made. However, by estimating the molecular coverage density, an enhancement figure of 10 5.3 to 10 5.5 can be produced. Although the manual drawing technique proved successful, lack of reproducibility meant that it was not suitable for large-scale commercial manufacturing. Before an attempt at manufacturing could be conducted, the parameters for the draw process (in particular temperature) needed to be determined. As the fibre is drawn by heating it to a high temperature in a furnace, it was necessary to understand the effects of the resultant diffusion of dopants within the fibre structure. To do this, a computer simulation was performed. The simulation package models a unit cell within the fibre structure and calculates firstly how the temperature affects the distribution of dopants and secondly, what the structures produced after etching would look like. It was found that in order to produce suitable nanostructures on the tips of the fibres, the diffusion length of the dopants must be less than 25% of the inter-core spacing. Knowing the maximum allowed diffusion length, means the maximum temperatures and heating times for manufacturing can be calculated. After determining the desirable manufacturing conditions, a number of production attempts using commercial fibre drawing towers were made. Ultimately, a short length of suitable fibre was produced allowing a full characterisation of the SERS capabilities of the fibre to be conducted. It was found that despite the high reproducibility of the structures, for an optimised sensor, the SERS signal counts still varied by approximately 40% from sample to sample. Careful investigations of the surfaces indicated this variability is most likely due to the nature of the metal coating applied rather than the fibre structures themselves. As such, it was found that to use sensors such as this for quantitative analysis, a form of internal SERS calibration is required. This was accomplished by exposing the sensor to a molecule of known concentration and measuring the signal intensity. After calibration, it was found the sample to sample variability dropped to approximately 8%. As a practical demonstration of the use of the SERS fibres, experiments on the detection of glucose, CEES (a simulant for the nerve agent HD mustard) and chlorsulphuron (a commonly used herbicide), were conducted. To test the procedures for gathering quantitative results, the sensors used in the experiment were calibrated by applying a monolayer of 1-decanethiol. In addition to providing the calibration spectrum, this molecule also aids in bringing the target analytes close to the active surface. Characteristic spectral lines could be easily identified from solutions of 5,000 ppm glucose, 549 ppm CEES and 10,000 ppm chlorsulfuron. Results of these experiments also indicate that the SERS fibres are reversible, that is, capable of tracking both increasing and decreasing concentrations. While the sensitivity is, at this stage, low when compared to other substrate formats, the speed and ease with which measurements can be taken and the potential for further optimisation of the fibre's surface both indicate the significant potential of this technique.