1. Introduction

Femtosecond laser material processing for the fabrication of both passive and active photonic devices, in various forms, for potential applications ranging from telecommunications to optical fibre sensing has received significant attention in recent years [1–7] driven by the advent of reliable and ever decreasing size, usually near IR, femtosecond lasers. Highly localised multiphoton absorption leads to a number of phenomena. Focusing on the surface, for example, can lead to ablation allowing the ultra-precise cutting of materials on the micrometer (sometimes sub-micron) scale, whereas focusing inside the glass can give rise to 3D glass structural changes and subsequent refractive index changes, both isotropic and anisotropic. In the case of plasma generation, these changes can extend well beyond the volume where laser interaction takes place through propagating shock waves and densification. It is this change in index that forms the basis of planar waveguide inscription, in various material regimes, using femtosecond laser pulses. Below the plasma threshold these changes are often achieved through densification of the glass [6] analogous to that used, for example, to write waveguides with two-photon UV absorption in silica structured optical fibres [8], through multiphoton absorption into common defects or hydrogen groups in doped glasses [9]. Above the damage threshold, as well as changes analogous to type II UV written damage gratings, densification and defect generation are observed around the processed region.

One of the main advantages of using femtosecond pulses to induce index changes is that energy can be rapidly deposited before thermal build-up can occur – this restricts thermal spreading so simple relaxation kinetics associated with quenched glass do not easily explain changes beyond the damage region when low repetition rate lasers are used. These attributes allow compact 2-D and 3-D multi-component photonic devices to be fabricated in a single step within a variety of transparent materials [1–7]. For many applications, silica is the preferred material, providing excellent physical and chemical properties such as optical transparency from IR to UV, a low thermal expansion coefficient, long term stability and a high resistance to laser induced damage. As well as NIR wavelengths ~(780-1100) nm, generated for example, from Ti:Sapphire oscillator-amplifier systems and a new generation of Yb³⁺ doped fibre lasers, significant work has also been carried using frequency-doubling to generate pulses in the visible spectral region [10]. Depending on the pulse parameters a variety of different laser-material regimes may be defined. Typically, the key parameters that can be tuned are the pulse train repetition rate, pulse energy and polarization, with the laser pulse duration and wavelength usually being fixed.

In terms of repetition rate, two regimes are categorised: (1) the low repetition rate regime is where the material changes are caused by individual pulses [11]; and (2) the high repetition rate regime where changes arise due to cumulative thermal effects [12] since the time between each pulse is less than the thermal diffusion time of silica. The border between these regimes occurs around a repetition rate, RR = 1 MHz although this is not easily defined since thermal dissipation depends on volume and surface area. The investigations of this paper will focus in the low repetition rate regime. Depending on the exposure parameters, three qualitatively different types of structural changes can be induced in fused silica: (1) an isotropic positive refractive index change (type-I); (2) a form birefringence with negative index change [13] (type-II), and (3) voids (type-III). Type-I modifications occur above an energy of ~0.1 μJ (λ = 800 nm, Δt_p = 160 fs, RR = 200 kHz and NA = 0.5) where the index change is permanent and isotropic [14,15]. The maximum index change is Δn ~(3-6) × 10⁻³ in fused silica [16] which is relatively large when compared with the index change achievable with nanosecond lasers [17,18]. In contrast, Type-II modifications are significantly different and happen above a higher energy threshold of 0.31 μJ (λ = 800 nm, Δt_p = 160 fs, RR = 200 kHz and NA = 0.5) [15]. The index change can be as large as Δn ~10⁻² [16] and exhibits impressive thermal durability, exhibiting no signs of decay after two hours at 1000°C [19]. The most striking features of type-II modifications is that the index change is highly anisotropic [15]. The origins of this anisotropy lie in the formation of sub-wavelength features of so-called “nanogratings” or “nanoplanes” [20]. Recently, it has been shown that these nanoplanes consist of porous matter most likely produced as a result of decomposition of SiO₂ into SiO_2(1-x) + x∙O₂ under intense plasma generation [20,21]. These structures and the phenomena related to them are analogous to those previously observed on laser processing of surfaces [22] except constrained within the volume. Arguably, they are not peculiar to femtosecond lasers – on the surface, thermal transport is avoided with longer pulsed lasers whereas within the volume the short pulse timescales help to mitigate similar thermal buildup.

The physical mechanism that gives rise to these modifications therefore involves multiphoton ionization inside the glass and the generation of highly localised plasma. Optical excitation ends before the surrounding lattice is perturbed, leading to highly localized “damage” in the material [21]. Secondary generation of a range of point defects with associated polarisability changes also occurs. Whilst many of the processes are identical to those observed in UV-induced changes in glass, including damage through two photon absorption (for an overall review see, for example [23], ), it is the control of thermal equilibration, and therefore localized relaxation [24], and highly defined multiphoton spatial localisation that allows an unprecedented level of control over glass properties.

Most work in pure silica has been in the so-called type-II regime, above the threshold required for plasma generation. In this paper, we investigate in detail the make-up of a waveguide in this regime written using femtosecond 1 μm light from an Yb³⁺ femtosecond fibre laser with polarisation parallel to the waveguide direction. Waveguiding properties reveal a larger than expected attenuation in the visible and polarisation studies show only a small linear birefringence whilst at 45° substantial elongation of the LP₀₁ optical mode is observed. SEM images reveal that the orientation of the nanostructure is closer to this axis. The repetition rate of the inscription beam is kept low to exclude a simple thermal relaxation as the origin of densification that accounts for the higher index region supporting modes. The overall objective of this work is to show how seemingly disparate observations between fundamental material studies and applied waveguides are linked, producing a larger picture of the processes that are taking place. This opens up opportunities for new directions in both material studies and in novel device areas.

2. Experimental Approach

The direct writing procedure using infrared femtosecond laser pulses has been already described extensively in other work [1–5]. In this work, silica glass (Heraeus Suprasil Type I and Ted Pella 26013) plates of 1 mm thickness are used. Processing is undertaken with a femtosecond fibre laser (λ = 1030 nm, Δt_p = 250 fs, RR = 200 kHz). The single mode output is focused below the surface of the silica plate using a 0.6 NA aspheric lens with the $\overrightarrow{k}$ vector of the beam being perpendicular to the surface of the plate. The sample can then be moved in three dimensions using computer-controlled stages. The linear polarisation of the laser output was usually kept parallel to the sample translation direction (parallel configuration) although the polarization transverse configuration state was also explored: those results will be reported elsewhere. The laser pulse energy was varied over (0.15 – 1.0) µJ; i.e. above the second damage threshold where nanostructures are formed. The scanning speed was fixed at 200 μm/s.

3. Results

(a) Femtosecond induced damage waveguides: To confirm the reproducibility of previous work, results are for typical structures formed in the type II regime where the most stable index changes and also the formation of features such as periodic nanoscale structures are expected to occur. After irradiation and in order to observe the intimate structure of the nanoplanes, the laser tracks are originally characterised using scanning electronic microscopy (SEM). Planar optical waveguides were inscribed using the following conditions: λ = 1030 nm, E = 0.7 μJ, Δt_p = 300 fs, RR = 200 kHz and a scanning speed v = 200 μm/s. The waveguides were written in 1 mm-thick, (25 × 25) mm² silica square (Ted Pella, product # 26013), 22 μm below the surface. To be consistent with the largest anisotropy observed in previous results, the polarisation of the inscription pulsed beam was parallel to the scanning direction. This result also leads to a densified region that can support light guiding.

(b) Waveguide fabrication and characterization: The spontaneous emission output of an erbium-doped fibre amplifier (EDFA) was used to test the guidance properties of the waveguide. This light was butt-coupled to the waveguide using standard SMF-28 fibre. A × 20 microscope objective was used to image the output of the waveguide onto a Vidicon VIS-IR camera. The resultant waveguide channel was found to be single mode in the 1550nm region, slightly elliptical in shape as shown in Fig. 1(a) . The 1/e² dimensions of the elliptical mode were approximately 7.2 µm × 15.0 µm and appear consistent with the elliptical nature of the induced laser track close to the surface. The total insertion loss of the waveguide, including coupling to input and output SMF 28 optical fibres, is found to be around 24 dB. The performance of the waveguide was also tested using the 632.8 nm output of a He-Ne laser; this was found to scatter significantly as might be expected at shorter wavelengths. Figure 1(b) shows the 632.8 nm output of the waveguide for the same coupling alignment as Fig. 1(a). It can be seen that the light is heavily scattered and there is no guided mode in the region where the 1550 nm mode was observed (illustrated by the white dotted oval). The nanopores observed in the SEM images in the following section, are too small to explain this scattering, which instead can be attributed to Mie-like scattering from the nanostructured plates. However, movement of the input fibre horizontally either side of the inscribed track can lead to weak guidance in densified areas of high index; this is shown in Figs. 1(c) and 1(d). Figure 1(e) is an optical microscope image of the end facet taken under white light illumination in reflection. The laser scanning direction, S, laser propagation direction, k, and linear polarization direction direction, E. This image illustrates iridescent scattering, consistent with periodic underlying structure and surface roughness, in the region above that supporting the 1550nm mode. By launching outside of any coupling into the slab itself, we were able to use the observable distortion arising from top and bottom induced-interference to obtain a phase interference image of the region supporting the optical mode shown in Fig. 1(f).

 

Fig. 1 (a) 1550 nm guided mode; (b) 632.8 nm injection for the same coupling alignment as (a); (c) 632.8 nm injection with butt-coupling fibre shifted to the right; (d) 632.8 nm injection with butt-coupling fibre shifted to the left. White dotted oval indicated location of 1550 nm mode; (e) Optical microscope image of waveguide end face. Note the damage region is clearly visible, scattering white light into a multitude of colours; (f) Phase interference image of the region supporting the optical mode.

Download Full Size | PDF

Figure 2 shows the output mode profiles for the following input waveguide polarisation states: vertical, horizontal and 45°. It can be seen that there is a negligible difference in mode profiles between the cases for vertical and horizontal input polarizations. However, for an input polarization that lies at 45°, an elongation of the mode along the y direction can be observed with an ellipticity of approximately 2:1 (long dimension: short dimension). The birefringence of the inscribed waveguides was measured using an optical microscope to analyse the polarization state of white light that is transmitted orthogonally to the waveguide track at 1550 nm. The waveguides were inscribed using the following conditions: λ = 1030 nm, E = (0.2 −1.0) μJ, Δt_p = 300 fs, f = 200 kHz and a scanning speed v = 200 μm. The effect of overlaying multiple scan lines on top of each other was investigated by inscribing up to 8 lines that are each separated by 1 μm in height. This technique has proven to be an effective method to overcome form based, high-asymmetry of waveguides that are inscribed using low energy pulses below the damage threshold [9], allowing the shape of the guiding region to be “built-up”. However, here we demonstrate how to use it to generate very large form birefringence by increasing energy and the number of scans. Figure 3 shows the degree of birefringence for various pulse energies and for increasing number of inscription scans. The birefringence is found to increase with increasing energy until it saturates above 0.7 μJ, approaching as high as 10⁻².

 

Fig. 2 Mode profiles and cross-sections for the 1550 nm guided mode for different input polarization states.

Download Full Size | PDF

 

Fig. 3 Birefringence for various pulse energies and scan numbers.

Download Full Size | PDF

(c) Femtosecond induced damage and imaging: The initial results are presented for the type II modification regime where the most stable index changes, and also the formation of features such as periodic nanoscale structures, are expected to occur.

After irradiation and in order to observe the intimate structure of the nanoplanes, the laser tracks have been observed using a Field-Emission Gun Scanning Electron Microscope (FEG-SEM ZEISS SUPRA 55 VP). Some of these FEG-SEMs allow examination of uncoated insulating or dielectric specimens using a low accelerating voltage (typ. in the range of 1 kV) and very low current (a few pA) because they can keep an image resolution good enough even in these extreme conditions and thus the original characteristics of the samples may be preserved for further testing or manipulation (no conductive coating in particular).

Using the femtosecond laser linear beam polarization perpendicular to the laser scanning direction and E = 0.5 μJ, tracks of index changes are inscribed 200 μm below the surface of the silica glass plate (Heraeus Suprasil Type I). With secondary electron imaging of a cleaved facet (unpolished), contrast nanogratings are observed as shown in Figs. 4(a) and 4(c); these are nanoscale quasi-periodic structures that likely correspond to areas of density contrast and thus refractive index modulation. In Fig. 4(b), where the femtosecond laser polarisation is now along the waveguide writing direction (as used during writing in the previous section), it can be seen that the porous head (close-up shown in Fig. 4(d)) and the tail of the modified region contain the nanogratings. In between, where wave guidance can occur, there is no observable structure. This phenomenon of nanogratings has been explained as arising from the interference between the inscription light and the electric field of the induced bulk electron plasma wave, resulting in periodic modulation of electron plasma concentration in the direction of polarization that records permanent structural changes in glass through electron trapping and plasma energy absorption [20,25]. Other explanations involve the generation of nanoplasma and/or self-organised material ripples arising from imperfections triggered by the plasma distribution [25–27]. The obvious directionality seen in Figs. 4(a) and 4(c) demonstrates preferential directionality of the “interferogram”, or periodic pattern, which should give rise to significant anisotropic properties [25] such as birefringence as shown in Fig. 3.

 

Fig. 4 FEG-SEM, Secondary electrons images of laser tracks cross-section for each writing laser polarisation. The laser parameters were: 0.5 μJ/pulse, 1030 nm, 300 fs, 200 kHz, 200 μm/s. A focusing lens of 0.5 NA was used. With the laser polarisation perpendicular (a) and parallel (b) to the scanning direction. (c) and (d) show close-ups of the nanograting and nanoplane regions respectively.

Download Full Size | PDF

There are observable differences between the structures written with different writing beam polarization directions, suggesting different structural configurations and orientations can be obtained. When the polarization is parallel to the scanning direction (Figs. 4(b) and 4(d)), porosity is observed within the nanoplanes. The nanopores exhibit a typical diameter range between 10 and 30 nm. In Fig. 4 it is clear that these nanogratings are in fact self-aligned, mesoporous nanoplanes and are at the root of the strong refractive index contrast of self-organized nanogratings. Nanopore formation is thought to arise from oxygen rejection [28] during silica glass dissociation that occurs above a threshold laser-induced temperature. In addition, this laser-induced porosity can also explain the enhanced contrast observed by Auger spectroscopy [20] that can be attributed to the nano-cavities.

4. Discussion

When the waveguides were fabricated using femtosecond light with the polarisation aligned parallel to the direction of writing, the porous “damage” regions (i.e. laser tracks head and tail) were not able to support light; rather, this was supported below in the densified glass suggesting that the average index above and below is less than the densified region, perhaps as a result of porosity. Images of the damage, and an intrinsic periodic structure, are revealed in both the iridescent scattering observed in Fig. 1(e) and the SEM damage in Fig. 4(b). For the waveguides written with a fs polarization perpendicular to the scanning direction, no intermediate zone is observed. The origin of the densified glass is likely due to a pressure or shock front compacting the glass beyond the damage region which expands slightly, as evidenced by the mode profile at 1550 nm shown in Fig. 1(a), which is not as elliptical as might be expected from the SEM images of Fig. 4(b). This volume of glass is able to support a higher order 632.8 nm visible mode with intensity on either side of where a 1550 nm mode can be supported, indicating visible light is attenuated strongly immediately below the damage region. A model simply based on melting and quenching appears inconsistent with the low repetition rate and the highly localised excitation process. The observed nanopores of Fig. 4(d) are too small to account for the observed scattering in the red; on the other hand, although Mie scattering off the nanoplanes can explain this in principle, supported by the iridescent scattering observed in Fig. 1(e), it is not clear that it can account for the significant overall attenuation, especially at increasingly shorter wavelengths. Alternatively, if silica-oxygen desorption has occurred then photodarkening through either silicon-rich (strongly oxygen deficient) glass or through the formation of silicon above the waveguide can account for this. More work is needed to verify and quantify these contributions.

Assuming decomposition of silica occurs, to explain the generation of O₂ and under-stoichiometric SiO₂ in more depth, the following mechanism is offered. The major contrast in our case when compared to the glass destabilization results reported in the literature is the different timescales of these processes. Indeed, it is commonly considered that a time of minutes or hours is required for matter to reorganize and for phases to be separated on the basis of diffusion alone. However, in our case, the silica destabilization is initiated by a laser field that is present for only ≈250 fs. Self-trapped excitons (STE) are formed from trapping of the free electrons (which are released as plasma over only a few fs); this plays an important part in the structural changes that follow and the trapping occurs over a few hundred femtoseconds. Therefore, at least ten pulses in the one area are required (e.g. this corresponds to a scanning velocity V < 100 μm/s at 1 kHz). Besides the formation of self-trapped-excitons (STE) in SiO₂ through radiative recombination they may relax into permanent defects such as SiE’ and NBOHC, in a few nanoseconds or even slower. These defects are characteristic of a changed local structure. The total subsequent thermal dissipation lasts a few μs.

To clarify how this can occur, we recall that our experiments (pulse repetition rate f = 200 kHz, sample velocity v = 200 μm/s) were performed below the so-called thermal accumulation regime [29,30]. In the experimental conditions employed, the material temperature is not determined by the cumulative effect of many laser pulses. That is, we know from the observed pore diameter, d = 20 nm, and the oxygen thermal diffusivity, D = 3.10⁻¹¹ m²/s at 2000 K, that nanopore formation must occur over a time of τ_c = d²/D = 20 μs. In contrast, the heat dissipation is around 1 μs and the time interval between laser pulses is 5 μs. Thus, nanopores cannot be formed under these conditions.

On the other hand, if we consider that the silica glass remains hot enough over 1 μs, a diffusion coefficient 20 times higher, corresponding to a temperature rise ~3600 K is necessary - this appears unlikely. In fact, the diffusion coefficient may not be considered appropriate in such a case where it is no longer a chemical diffusion under weak gradient but under very intense optical field (laser intensity around 10¹³ W/cm² and the related electric field of around 10⁹ V/m). Indeed the energy deposited by 1 pulse in the focal volume is of the same order of magnitude as the formation energy of the glass; we may therefore discuss the origins in terms of an optically induced mobility of the oxygen and local defects. The light is coupled to an atom providing a momentum that can be significantly higher than that of diffusion. By equivalence, mv = hk. From this we can calculate the displacement possible by the optical field as high as 50 nm/ps which is consistent with our observations.

From this argument, we tentatively propose the following mechanism illustrated in Fig. 5 to explain the laser-induced silica oxide decomposition. Initially after excitation and plasma formation, “hot spots” are present and are likely due to inhomogeneous multiphoton ionization. These areas correspond to regions of high plasma density which, by the ponderomotive effect described above, will explode. Indeed, tight focusing yields high peak intensities but also results in large-intensity gradients and, therefore, large ponderomotive forces. These large intensities will also lead to localized self-focusing of the beam due to the optical Kerr effect, thus compounding any ionization inhomogeneities in the nanoplane region. From Fig. 4(d) if O₂ is present in the pores, it appears that it is in the nanoplanes, in contrast to the orthogonal writing beam polarisation where, from Fig. 4(c), it appears to be in between the nanoplanes.

 

Fig. 5 Tentative mechanism and related phenomena together with their energy scale and timescale.

Download Full Size | PDF

In such a case, the strongest gradients point radially inward, so the ponderomotive force pushes electrons outward, directly away from the regions of high intensity. It produces behind a centrifugal electrostatic force that can extract oxygen atoms from the destabilised network. This is similar to the Knotek-Feibelman electron-stimulated mechanism [31] that involves the ejection of positively charged anions (O⁺) by repulsive electrostatic forces, which can cause the creation of anion and cation vacancies and interstitials. In more detail, if a positively charged oxygen ion is formed, the electrostatic term in the lattice potential becomes negative and the ion is forced to move to an interstitial site and thus to create a Frenkel pair within the bulk of the material or to pre-existing defects within the bulk. The damage mechanisms in electron beam irradiated a-SiO₂ are thus described as a combination of ionization and field-induced migration processes. As a consequence of aggregation, these atoms/ions occur in an over-stoichiometric ratio where they cannot be diluted so they later combine to stabilize in the form of bubbles. The back-migration of oxygen towards the point of emission cannot be done by the electrostatic field but as a result of a concentration gradient, and it takes several 10’s microseconds, long after the light is extinguished although the temperature was still quite high. Notice that in the case of silica glass, as noted earlier the heat dissipation (effective cooling time of a few μs) is too rapid to allow oxygen to migrate back. These phenomena together with their energy scale and timescale are summarised in Fig. 5 As a result of the anisotropic light-plasma interference, the intense fields are likely to align defects that also form through recombination processes. Lancry et al. previously confirmed such anisotropy in the UV-induced luminescence from a common defect, the oxygen deficient centre ODC [32].

The observed waveguide mode distortion at 45° polarisation is not fully understood but we note from Fig. 4(b) that the quasi-periodic nanoplanes are not aligned orthogonal to the track direction but roughly 45°, highlighted by the red line; thus both eigenstates are scattered similarly so the waveguide birefringence is small. On the other hand, when light is launched at 45° polarisation, this polarisation is aligned with the nanoplanes and the field extends further away from the damage region, as shown in Fig. 2, perhaps as a result of increased scattering. We note that the conventional waveguide reference for the definition of the polarisation eigenstates is for a uniform waveguide with orthogonal states perfectly aligned to the substrate or surface of the plate in which a waveguide is present. Overall, the results point to opportunities in controlling the polarisation properties, including polarisation beating and polarisation rotation, by controlling the orientation of the nanoplanes and other structures. This in combination with porous silicon-rich (or strongly oxygen deficient) glass may offer very novel devices that are not possible to fabricate by other means.

5. Conclusion

In conclusion, we have carried out a systematic study of the laser induced damage of type II induced changes in silica associated with silica waveguide formation below and around the damage region. The induced anisotropy gives rise to form birefringence of the waveguide when probed orthogonally. This can be increased both by using higher energies and writing multiple waveguides on top of each other, allowing a simple way to tune the birefringence. Remarkably, this process does not erase or photo-anneal any of the original induced nanostructure. The observation of nanopores and the large waveguide attenuation at shorter wavelengths is further circumstantial evidence of significant molecular decomposition above the waveguide region. By following the introduced anisotropy from the periodic nanoplanes and waveguide properties, it is shown that there are direct correlations between the properties induced by laser and the final waveguides. Consequently, it is possible to alter not only material properties and extent, but also the orientation of the nanostructured periodic plates and other structural anisotropies to impact on waveguides and other devices in potentially useful ways. In the first instance, it is possible that the scattering giving rise to some of the attenuation in these waveguides can be greatly reduced. In combination with many of the other properties that now can be feasibly controlled, this promises to open up a new dimension in device fabrication and material processing. Perhaps the most significant implications of these results is the impact of induced silicon-rich regions on waveguide properties including absorption and attenuation. In addition to some applications already mentioned, this could have benefits in making custom absorption filters, Raman generators and generating other nonlinear effects and devices within silica, including silicon lasers, for example. In doped glasses, similar molecular decompositions may be used to fabricate other material rich systems, such as Ge waveguides and devices. Further investigation is required to establish the extent of the material changes.