The understanding of the physical mechanism involved in thermal poling is crucial for the identification of the critical parameters affecting the process. Such a knowledge could lead towards engineering of the glass system and/or of the poling procedure for enhancing the induced second-order nonlinearity. Early on it was realized that progress was hindered by the difficulty in accessing the depletion region and obtaining information from its thickness, profile and dynamics. Hence, the improvement of characterization techniques and the development of new methods has never stopped from the early days
of poling. The Maker’s fringe technique (MFT), which will be described in detail in Chapter 4, was the first method to be employed for the measurement of χ(2) [7], but,
unless the thickness of the nonlinear region is measured independently, the MFT is not suitable for the characterization of poled glasses when the width of the nonlinear region is thinner than one coherence length. Typically, the silica plates were etched in hydrofluoric acid solution in steps separated by a few seconds or minutes. At each step the thickness of the etched glass was measured with a profilometer and the SH signal recorded. The width of the nonlinear region was extrapolated to be the thickness for which the SH signal had disappeared [7]. Lesche et al. observed that the presence of Edc affects the etching rate. Namely the etching rate of poled silica is 0.7 times slower than unpoled silica. Based on this discovery they developed an interferometric etching technique that enabled them to monitor, in real time, the evolution of the depletion region by looking at the sudden change in the etching rate [42]. In 2003 Kudlinski and co-workers proposed a very accurate method for the determination of the χ(2) profile. The evolution of the SH signal is monitored as the poled sample is etched with hydrofluoric acid. The sample thickness is measured in real time using an interferometric method. The nonlinear profile was reconstructed from the full set of data by means of a layer peeling method [43]. Given that the spatial resolution is below 1µm and that the reconstructed profile does not depend on a priori assumptions, this technique is a very powerful tool for characterizing the distribution of the nonlinearity in the poled glass and so gain knowledge about the physical processes involved in thermal poling. The nonlinear profile that was inferred for a 10 min poling of InfrasilTMat 290◦C
was compatible with a multiple carrier model but not with the single carrier model. Sign reversal of the nonlinearity in the first 2µm beneath the anodic surface was observed during poling of pre-annealed samples [87].
The aforementioned technique has the only disadvantage that samples are unusable after having been characterized. Another direction followed by researchers in order to obtain the distribution of the nonlinearity in the glass was the improving of the MFT technique. The limitation in the MFT is caused by total internal reflection of the SH light beam at the output surface of the sample being tested. Pureur et al. demonstrated a prism- assisted MFT, in which the sample was sandwiched between two prisms in order to avoid the total internal reflection. For the first time the profile of the nonlinearity, and not only its thickness, could be estimated. Best fit to the experimental data was obtained for a truncated Gaussian profile, 8µm wide when the nonlinear coefficient decreased by a factor 1/e [88]. Quiquempois et al. suggested to use hemispherical lenses instead of the prisms [89]. The main advantage was that this configuration allowed for easier measurements as it was no longer necessary to adjust the focus of the beam on the sample for each angle. The use of cylindrical lenses in place of the hemispherical ones is a further technical improvement of the same initial idea [90]. A different approach was followed by Faccio and co-workers where a Non-Collinear MFT (NCMFT) was demonstrated [91]. In this technique two beams at the fundamental frequency radiation
are focused on the poled sample and SH light generated in the non-collinear direction is collected. The coherence length of such a process is about 2µm in silica compared to the 24µm of the collinear interaction, meaning that the NCMFT is able to resolve the spatial distribution of the nonlinearity. Best fit to the data gave a truncated Gaussian profile for poling of HerasilTM1 glass plates. Guillet de Chatellus et al., also used a NCMFT but they also vary the angle between the two pump beams to increase the sensitivity to the nonlinear profile [92]. The nonlinearity profile that was inferred from the data gave credit, once more, to the existence of a double charged layer close to the anodic surface in poled glass.
In contrast to the layer peeling method mentioned before, all these techniques derived from the MFT are non-destructive. On the other hand the profile of the nonlinearity is inferred from the Maker’s fringe patterns under a priori hypothesis of having a sin- gle maximum and no change in sign. This last assumption might be too crude as in some poling conditions the existence of double layer, positive (H+, ionized glass) and
negatively charged depletion region was shown [30, 74].
In parallel to the development of new and accurate characterization techniques, a con- spicuous amount of effort was devoted to the fabrication of novel glasses for poling. As the characterization techniques became more and more sophisticated there was the need for glass scientists to have a simple and practical tool for evaluating the glass sam- ples and have an immediate feedback for improving the glass composition. Bearing this idea in mind, in this thesis we proposed a variation of the MFT technique in which a sandwich structure formed by two identically poled samples stacked together with their nonlinear regions facing each other. The thickness of the nonlinearity is inferred from interference pattern between the SH light beams generated by the two nonlinear layers. The major advantage of such a technique is that it is non-destructive and can measure thicknesses of the depletion layer as small as 4µm, with a resolution of 1µm. The approximation of a square nonlinear profile is made and therefore this technique provides the same information as a classical MFT plus an etching measurement of the thickness of the nonlinear region. A beautiful technique was proposed by Ozcan et al. at Stanford, that makes use of the same sandwich structure, together with a cylindrical lens assisted MFT, to determine the nonlinear profile [40, 93]. The method they sug- gested is based on an inverse Fourier transform algorithm. Their technique was refined in subsequent publications [94, 95], also employing a Fineup algorithm [96] and applied to the characterization of thermally poled germano-silicate thin-films [44]. Very recently Tr´eanton and co-workers proposed to use a quartz plate, of known nonlinearity, instead of one of the poled samples in the stack so that theχ(2)(z) spatial distribution could be determined by an interferometric method.