1. Introduction

Processing of glass with ultrashort laser pulses generated numerous fields of research. Within the last two decades multiple groups worked on the inscription of waveguides based on laser induced changes of the refractive index [1–3]. Furthermore, under certain irradiation conditions it is possible to induce birefringent domains [4, 5]. Here, the irradiation with multiple ultrashort laser pulses results in the formation of hollow nanocavities - typically with a feature size of 30 nm - whose sub-wavelength arrangement causes birefringence [6–8]. As these cavities are typically periodically aligned, these structures are called nanogratings. Their period depends on the inscription wavelength λ and scales roughly with λ/2n, with n the refractive index of the material. The formation mechanism of nanogratings is still not completely understood. So far, the favored model considers the formation of nanoplanes induced by nanoplasmonic effects. A three-step evolution is predicted: (1) inhomogeneous dielectric breakdown and generation of nanoplasmas, (2) growth of tiny modifications (size of a few nanometers) and evolution to nanoplanes and (3) self organization of these nanoplanes due to their interaction with the electrical field of the laser pulses [7, 9, 10]. In addition to waveguides and nanogratings it is also possible to induce microexplosions in the glass resulting in hollow voids with a diameter of several hundreds of nm [11–13]. Furthermore, if the time between the subsequent laser pulses is shorter than the heat dissipation time, heat accumulation in the processed volume can occur [14, 15]. Under these circumstances the material may be locally molten [16, 17].

To investigate the formation of these modifications, most often fused silica is used. However, waveguides and nanogratings can be inscribed in different glasses ranging from borosilicate to fluoride glass systems [18–22]. Addition of network modifier oxides to a pristine SiO₂ glass matrix, allows for the formation of a complex variety of laser induced modifications due to the potential generation of different defects and color centers in the glass network [23–25]. Furthermore, when using other types of glasses than vitreous SiO₂, higher amounts of photosensitive agents may be added, for example for the generation of tiny nanoparticle domains with strong plasmon resonances or near-infrared photoluminescence in phosphate [26] or heavy metal oxide glasses [27].

However, with only slight modifications, even the simple SiO₂ network offers tremendous potential for different applications. In this paper we concentrate on the Ultra Low Expansion glass (ULE) produced by Corning (Corning code 7972). This titanium silicate glass has a TiO₂ fraction of about 7.5 wt% [28] leading to a thermal expansion coefficient close to zero over a temperature range from 5–35°C. This property has made ULE an interesting material for substrates, supports, and blanks in which ultra-high geometrical stability is required during operation. Recently, we were able to show that birefringent domains can be induced in ULE [18]. Relevant to laser writing and potential optical applications is the well known fact, that TiO₂ agents can induce darkening - meaning an increased absorption over the entire visual spectral range - of the sample after fs pulse treatment [29]. In this paper we analyzed the formation mechanism of nanogratings in ULE meticulously. Small Angle X-ray Scattering (SAXS) and Focused Ion Beam (FIB) milling techniques were used in the determination of the laser induced nanostructures. In addition, the laser induced darkening process was studied by Raman, optical, and Electron Spin Resonance (ESR) spectroscopy in order to identify the underlying reaction mechanisms.

In addition to its favorable low thermal expansion coefficient at room temperature, the potential of ULE is given by the fact that both modifications, nanogratings, and darkening, can be induced simultaneously. When inscribing nanogratings in the material the orientation of the slow optical axis and the strength of birefringence itself can be tuned, offering the capability for polarization multiplexed writing [30–32]. Recently, Zhang et al. demonstrated multilevel encoding of nanograting-based data storage in fused silica [33]. The option to tune not only the refractive index but also the absorption of laser processed ULE leads to an additional degree of freedom and a further increase of data multiplexing capability.

2. Methodology

A femtosecond oscillator providing pulses at a wavelength of 1030 nm, a repetition rate of 9.4 MHz, an average output power of 5 W, and pulse duration of 450 fs (Amplitude Systems, t-Pulse 500) was utilized for all the experiments presented in this study. The repetition rates and pulse energies were varied by an external acousto-optic modulator and a halfwave plate followed by a polarizer, respectively. An LBO crystal was used to generate the second harmonic (515 nm). The laser pulses were focused with an aspheric lens (f = 4.5 mm, NA of 0.55) into a depth of 100 µm inside the ULE samples. The spot diameter was about one micron. In order to inscribe continuous lines, the samples were translated with respect to the laser focus. By varying the repetition rate R of the laser and the translation velocity v the number of pulses N incident on one particular spot can be changed: $N=\frac{2{\omega }_{0}R}{v}$ (ω₀ is the beam waist). The maximal pulse energy applied (200 nJ) is well below the threshold for self focusing [34].

To realize large areas of laser modified glass, parallel line were inscribed. The line spacing was varied during the experiments. For low repetition rates, the size of a modification is basically given by the focal volume and only tiny modifications are observed. Therefore, a line spacing of about one µm is required. However, for large repetition rates (≥ 2 MHz) heat accumulation occurs which in turn results in large modifications and the spacing can be increased by a factor of 10 in this case.

3. Experimental results

3.1. Processing parameters

Laser processing of ULE causes different kinds of modifications: darkening, nanograting formation, and local melting. Initially, a detailed parameter analysis was conducted to study the influence of the processing parameters (repetition rate and pulse energy) on the resulting modifications. Exemplary images of these modifications are shown in Fig. 1(a). The impact of the processing parameters is depicted in Fig. 1(b).

 

Fig. 1 (a) Exemplary top views on different laser induced modifications in ULE. (b) Overview of the processing parameters (pulse energy and repetition rate) for the inscription of local modifications in ULE (for details see text).

Download Full Size | PDF

In principal, laser darkening can be induced for almost all laser parameters investigated -even for very low repetition rates (dark shaded area in Fig. 1(a)). Mostly, to induce darkening at least pulse energies of 50 nJ are required. For very low repetition rates (≤ 10 kHz) the pulse energy has to be increased to induce verifiably darkening. In contrast, at the highest available repetition rate of 9.4 MHz darkening is already achieved at 25 nJ. However, the total degree of darkening depends on the laser power applied and generally increases with the laser power. Measurements and results concerning the darkening of ULE are presented in Chap. 3.3.

In addition, two other laser induced modifications can be distinguished in ULE: local melting of the glass due to heat accumulation of successive pulses at repetition rates above 2.3 MHz and the formation of nanogratings. These birefringent structures were identified with a polarization contrast microscope. It was shown previously, that numerous pulses are required for the formation of nanogratings [35]. Thus, to realize nanogratings even for very low pulse energies (25 nJ) and repetition rates (10 kHz) we reduced the translation velocity (≈ 0.01 mm/s). For repetition rates of above 100 kHz we worked with a translation velocity of 3.3mm/s. In general, we detected birefringent modifications for all processing parameters investigated (red crosses within Fig. 1(b)) only limited by the onset of heat accumulation and subsequent melting of the glass (yellow dashed line). After melting, the modification within the material is homogeneous and isotropic showing only stress induced birefringence. Thus, heat accumulation poses an upper power limit for nanograting formation. Consequently, the maximum usable pulse energy is inversely proportional to the repetition rate. Here, we deduced a maximum average power of 230 mW for the formation of nanogratings. This finding agrees well with the results obtained earlier for pure fused silica [6].

Heat accumulation was observed for large repetition rates above 2.3 MHz and depends also on the pulse energy (right area in Fig. 1(b)). Local melting can be identified by generation of bright fringes of molten and resolidified material surrounding the initial focal volume. In ULE, within the traces of the molten material large bubbles are formed (lower right picture in Fig. 1(a)) while no birefringence can be detected. Here, even for a line spacing of 20 µm the modified region appears as a continuous black area.

3.2. Formation of nanogratings

A laser processed ULE sample was polished and etched with HF acid (1%) for 120 s in order to visualize the grating structure. A repetition rate of 100 kHz and a pulse energy of 100 nJ was utilized for the inscription. The detected structure shown in Fig. 2 is perpendicular to the laser polarization as expected for nanograting formation [5]. In general, the orientation of the nanogratings (and thus the orientation of the slow axis) can be tuned by the orientation of the laser polarization [33, 36]. For this set of parameters, the grating period is with 250 nm significantly smaller than the laser wavelength (λ = 515 nm). In order to investigate the influence of the processing parameters (number of irradiating pulses and pulse energy) on the structure of the nanogratings, Small Angle X-ray Scattering (SAXS) was applied, using the beamline cSAXS at the Swiss Light Source (PSI Villigen, CH). For illumination an 11.2 keV X-ray beam was employed. The scattered rays were recorded by a large direct-converting pixel detector (Pilatus 2M) after the main beam was blocked. For further details, see [8, 21]. Generally speaking, the formation of nanogratings in ULE is similar to the results obtained in fused silica [6, 35]: The grating period decreases with increasing number of applied laser pulses, as can be seen in Fig. 3(a). For example, when using about 1,000 pulses per spot the period is roughly 140 nm, which is well below the expected value of λ/2 n, with n the refractive index of the material (n = 1.48). The SAXS measurements confirmed that nanogratings in ULE consist of tiny pores with a diameter of 30 nm. The total pore size is independent of the number of laser pulses (see the lower graph of Fig. 3(a)). However, the absolute number of nanopores depends, contrary to the size of the nanopores, on the number of the applied laser pulses. It was shown by Zimmermann et al. that the Porod Invariant (which is proportional to the scattering contrast of the modified material) increases with the number of laser pulses applied at the same pace as the retardance of the birefingence glass [21]. This indicates, that a growing number of nanopores is generated by irradiation with an increasing number of laser pulses. The same mechansim could be identified in ULE, as well. Both, the Porod Invariant and the optical retardance increase with increasing number of pulses applied. Please note that the pulse energy also defines the retardance, as can be seen in Fig. 3(b) for pulse energies of 75 nJ and 150 nJ. Thus, different processing parameters define the value of optical retardance: the pulse energy and the number of pulses targeting the sample.

 

Fig. 2 Top view (SEM image) of laser induced nanogratings in ULE. After laser inscription the sample was polished and etched for 120s with HF acid.

Download Full Size | PDF

 

Fig. 3 (a) Period and diameter of laser induced nanogratings. (b) Retardance and Porod Invariant of nanogratings in ULE using different pulse energies with respect to the number of applied pulses.

Download Full Size | PDF

To conclude this subsection, we deconvoluted a 3-dimensional reconstruction of the pore structure from layerwise SEM data. For this, we employed Focused Ion Beam (FIB) milling to dissect the modified material containing nanogratings which were fabricated with 2,000 pulses per spot and a pulse energy of 150 nJ. Figure 4(a) shows a representative SEM micrograph. The dark lines beneath the nanopores are imaging artifacts due to FIB milling. For the reconstruction in Fig. 4(b) these were removed by standard image processing methods. The nanopores are very small, exhibiting a width of only a few tens of nanometers. As the pores are devoid of glassy material this material has to be moved into the shell surrounding the cavities. However, the resolution and particularly the contrast between densified and pure material is too small to verify this. The SEM image (see magnification of Fig. 4(a)) show some bright colors surrounding the nanopores which may be due to the densified material.

 

Fig. 4 (a) Exemplary SEM micrograph of the nanopores in ULE shown in the y–z plane. (b) 3D reconstruction of 81 SEM images of nanogratings (2000 pulses per spot at a pulse energy of 150 nJ) prepared by FIB milling.

Download Full Size | PDF

The reconstruction (based on 81 SEM images with a distance of 28 nm) of a volume of 1 µm³ is shown in Fig. 4(b)). Hollow structures are depicted as dark blue regions. Their characteristic width along the y direction matches the 30 nm from the SAXS results. The transverse dimensions (along x, and z direction) of the nanopores are of comparable size which is significantly larger (a few hundred nanometer) than their thickness (along y direction). The individual pores form elongated sheets which are periodically aligned with a period of roughly 150 nm.

3.3. Laser induced darkening

The results obtained for the formation of nanogratings in ULE are in good agreement with results obtained in fused silica. However, contrary to the findings concerning the formation of nanogratings, a completely different mechanism has to be responsible for the laser induced darkening which occurs for a very large range of parameters in ULE and which was not analyzed so far. For the following results, high repetition rates (in the MHz range) were selected since the induced modifications are much more pronounced in this case and thus easier to analyze. At first, we used optical microscopy to study the laser induced modifications in ULE. To this end, we inscribed large areas (of 4 × 4 mm²) with a line spacing of 5 µm at a laser repetition rate of 4.7 MHz at different pulse energies. The general trend is already apparent in the images of Fig. 5. The darkening intensifies with increasing laser pulse energy.

 

Fig. 5 Exemplary images of laser induced darkening in ULE, from left to right with decreasing pulse energies. Parallel lines (along y direction) were inscribed into the material at a repetition rate of 4.7 MHz and a spacing of 5 µm.

Download Full Size | PDF

In order to discern the origin of the darkening, the modified material was studied in more detail by scanning electron microscopy. The microscopic top view on the laser written lines (Fig. 6(a)) reveals multiple bubbles which are formed within the molten material. It can be seen that the total number of bubbles increases with the repetition rate leading to a higher bubble density in the material. Microscopy imaging was followed by a FIB cut from the top of the modification. Figure 6(b) proves that the bubbles are almost circular hollow cavities with a diameter of less than 10 µm. The cavities are aligned next to each other forming a dense line. In addition, enlargement of the SEM image exposes bright fringes surrounding the cavities, which is most likely due to the densification of the material as a consequence of cavity formation [13].

 

Fig. 6 Microscopic (a) and SEM (b) top views of laser modified ULE. Continuous lines were inscribed using a pulse energy of 200 nJ. (b) To reveal the inner shape of the modification the sample was dissected using FIB.

Download Full Size | PDF

In a next step we investigated the absorption spectra of the laser modified samples, presented in Fig. 5. To do so, we measured the transmission of the modified samples with an UV-VIS spectrometer (Perkin Elmer Lambda 950). We also measured the reflectance of the sample in order to separate defect induced absorption bands from scattering of laser induced structures. In agreement with the images of Fig. 5 the absorption of laser darkened ULE samples increases evenly without any distinctive peaks leading to a blueish-gray appearance of the modified samples (see Fig. 7(a)). In general, the overall absorption increases with increasing applied pulse energy. The maximal absorption was 0.4 when using a pulse energy of 175 nJ. For comparison pristine ULE is included in Fig. 7(a) as well.

 

Fig. 7 (a) Absorption spectra of laser modified samples. (b) Difference in the absorption spectra between the processed ULE glass and the pristine material.

Download Full Size | PDF

To highlight small variations in the absorption spectra of modified ULE, the spectrum of the pristine glass was subtracted from the spectra of irradiated samples, resulting in the difference or irradiation induced optical spectra (see Fig 7(b)). The differences of the spectra are negligible in the range of the UV absorption edge below 300 nm, which is defined by the charge transfer transition between Ti⁴⁺ - O²⁻ [37]. However, from 300 to 2000 nm new details are now apparent, including a pronounced absorption peak with a maximum around 300 nm and a much broader band with a maximum near 790 nm, shown in the inset of Fig 7(b). The strong decrease of the absorption peak below 300 nm, is due to the combined effect of a shift of the absorption edge to lower wavelengths, indicating a decrease of Ti⁴⁺-CT transitions, which is overlaid by the bands caused by the formation of intrinsic defects, such as the well-known 4.8 eV or ≈260 nm band in irradiated silica [23, 38, 39]. During the further experiments, no corresponding ESR signals were observed in the laser treated ULE glass (see Fig. 8), indicating that the observed absorption bands must arise from diamagnetic defects, thus ruling out the Non Bridging Oxygen Hole Center (NBOHC) or peroxy radicals (POR) [38]. Remaining absorption center might be the oxygen deficiency center ODC(II) which equals a divalent Si atom (=Si:) in laser treated glass [38–41]. Another assignment for the absorption in the UV to visible range is the formation of ozone (O₃). The strong absorption around 300 nm becomes more dominant for higher pulse energies pointing towards an increased generation of defects at large pulse energies.

 

Fig. 8 ESR signal of pristine and laser modified ULE.

Download Full Size | PDF

A broad band with a maximum near 790 nm forms after irradiation by very high laser energies. Such a band is unknown from irradiated silica, but might be associated with photoreduced titanium ions instead. Weyl and Bates describe spectra of Ti³⁺ with a band at 570 and a shoulder at 750 nm, for a Jahn-Teller distorted Ti³⁺ ion (d¹) in octahedral coordination [42, 43]. Bishay describes a similar band in irradiated silicate glass for a photoreduced (Ti⁴⁺)- species [44]. EX-AFS studies on xTiO₂-(100-x)SiO₂ glasses showed, that titanium ions prefer the tetrahehedral coordination in the region of extreme low expansion, that is for 3.5<x<8 [45]. It is not clear how much the d-d transition of Ti³⁺ would shift for a tetrahedrally coordinated Ti³⁺ ion in comparison to an octahedrally coordinated Ti³⁺. Reduction of TiO₂ creating trivalent titanium is a popular technique to realize solar harvesting surfaces for plasmonic applications. The appropriate absorption spectrum has a peak between 750 and 800 nm with additional features into the near IR [46]. Blue colored titanium complexes are described to exhibit a band near 790 nm, but this band is most often assigned to an inter-valence charge transfer (IV-CT) transition between Ti³⁺ ions and iron ions [47, 48] or in analogy even to an IV-CT between Ti³⁺ and Ti⁴⁺ [47]. Even though glass modifications could lead to selective precipitation of Ti₂O₃/TiO₄ clusters, resulting from the strong enthalpic asymmetry of the Si-O-Ti bond linkage, and the close proximity between differently charged titanium ions would give IV-CT transitions, these are often observed together with strong cation to oxygen CT transitions at higher energies [48, 49]. In the current study, we do not observe such a parallel increase of the CT bands. The occurrence of a Fe-Ti IV-CT is even less probable, since the present glass is practically free of iron ions. Therefore, a red shifted d-d transition for low coordinated Ti³⁺ might be the better explanation for the origin of the 790 nm band. A red shift can be expected when moving from octahedral (O_h) to tetrahedral (T_d) complexes, since the number of ligands and the crystal field strength decreases. As an estimation, Dq(T_d)=−4/9 Dq(O_h) (where Dq denotes the splitting energy of the d-orbital) if the bond lengths do not vary [50, 51]. Gruen and McBeth studied the optical spectra of chloride melts at different temperatures and did indeed observe a coordination equilibrium between the octahedral low temperature and the tetrahedral high temperature complex. And as expected, the d-d transitions of [Ti⁽ⁱⁱⁱ⁾Cl₄]⁻ complexes were red shifted in comparison to the transitions of [Ti⁽ⁱⁱⁱ⁾Cl₆]³⁻ complexes [52]. Interestingly, many papers on Ti₂O₃ films and alloys identify the formation of this reduced titan-oxide by its blue color [53–55].

At this point we can safely attribute the 790 nm band to the generation of photoreduced Ti³⁺ species. The photo-reduction of tetravalent titanium (Ti⁴⁺, d⁰) to trivalent titanium (Ti³⁺, d¹) is most likely the reason for the observed laser induced darkening. Ti⁴⁺ with its fully occupied valence orbital is colorless but Ti³⁺, a d¹ ion, gives rise to a broad absorption band with a maximum near 800 nm, thus rendering a blue color to the glass. In combination with the absorption bands of the glass matrix, which absorb below 500 nm (resulting in yellow to red glasses) a darkening is perceived due to the broad absorption over the entire visual spectrum.

To verify the presence of trivalent titanium in laser treated ULE Electron Spin Resonance (ESR) measurements were performed. The ESR spectrometer employed operated at the X-band frequency of 9.78 GHz and contained an internal calibration system (Bruker). The ESR spectra of an untreated ULE sample and one modified sample (repetition rate of 9.4 MHz, pulse energy of 200 nJ, line spacing of 5 µm) are depicted in Fig. 8. The laser processed sample (red curve) shows a broad signal centered around g = 1.93. This signal can be assigned to Ti³⁺ ions [56]. In contrast, this signal is very weak in the pristine ULE sample (black curve). No signals due to defect formation in the vitreous network are observable. Such signals are usually centered around the g-value of the free electron with g ≈2 [38, 40, 41]. Melting of glass does fully rearrange the structure and no defect bands or radicals are left in significant numbers for detection by ESR spectroscopy. As mentioned before, no ESR signal of Fe³⁺ is seen in the untreated or in the modified samples. ESR measurements average over the whole sample and do not allow the spatial resolution of the line pattern. However, ESR spectroscopy allows distinguishing different Ti³⁺ species and since Ti³⁺ is a d¹ ion, all signals observed belong to different bonding sites. Anpo et al. and Yamashita at al. were able to distinguish not only tetrahedral and octahedral Ti³⁺ sites in a titanium catalyst, but also found evidence for Ti₂O₃ cluster [57, 58]. The ESR spectra in their studies show a strong resemblance to the spectra in Fig. 8. The broad central resonance with smaller g-values was assigned to a tetrahedral Ti³⁺ site unsaturated in regard to its coordination, and the sharper signal at higher g-values to distorted octahedral coordinated Ti³⁺ ions. A very sharp band signal with high g-values is assigned to aggregated octahedral Ti³⁺ species, such as found in amorphous Ti₂O₃. The sharp signal at g=1.998 in the irradiated ULE sample appears in repeated ESR measurements and we can therefore assume that it is indeed a genuine signal. The asymmetry of the broader resonance confirms further the presence of differently coordinated, isolated Ti³⁺ species in addition to some Ti₂O₃ cluster. Furthermore, we want to add, that Ti³⁺ in very symmetric environments is known to give only a very weak ESR signal at room temperature [58].

Optical and ESR spectroscopy reveal the presence of Ti³⁺ but no hints for charge compensating intrinsic defects were found. This is intriguing since more questions arise from this observation. How does the glass network adapt to a change from Ti⁴⁺ to Ti³⁺, not only in regard to the type and number of bonds, but also regarding charge compensation?

A useful tool for the study of the glass network is Raman spectroscopy. The vibrational stretching and bending modes of the differently connected SiO₄ and TiO₄ tetrahedra result in Raman spectra with several distinct bands [59]. It is expected, that the comparison of the Raman spectra before and after laser processing will reveal changes in the connectivity of the SiO₄ and TiO₄ tetrahedra, e.g. by a change in the band intensities of ring breathing modes or of Ti-O-Si related bands. Figure 9(a) shows the Raman spectra of pristine and laser modified ULE. The exact location of their acquisition is marked in the micrograph (Fig. 9 (b)). To highlight the effect of laser treatment, all spectra were normalized on the silica band at 430 cm⁻¹ and were treated to have the same baseline above 1300 cm⁻¹. The Raman spectra of not yet modified ULE (black line in Fig. 9(a)) is characterized by a broad band around 430 cm⁻¹ which is due to mixed stretching-bending modes of Si-O-Si bridging bonds in large silicate rings, like the 5- and 6-membered rings in the silica glass network [60]. The defect bands, or ring breathing modes of three- and four- membered silica rings are seen as weak peaks at 602 (D₂) and 483 cm⁻¹ (D₁), respectively [61]. The band near 800 cm⁻¹ is due to Si-O-Si bending modes [62]. All these bands are also known from vitreous silica [61, 62].

 

Fig. 9 (a) Raman spectra of modified and unmodified ULE. (b) Microscope image of laser modified ULE. (c) Integrated intensity of characteristic Raman peaks.

Download Full Size | PDF

Furthermore, ULE shows additional peaks at 390 cm⁻¹ and 690 cm⁻¹ and two very intense Raman peaks at 937 cm⁻¹ and 1110 cm⁻¹. The latter Raman peak was formerly assigned to the asymmetric stretching mode between Si-atoms and non-bridging oxygen ions, i.e. to Si-O⁻ bonds [63, 64]. While the presence of some Si-O⁻ bonds cannot be entirely excluded, formation of non-bridging oxygen atoms - to such an extend as suggested by the intensity of the 1110 cm⁻¹ peak - would modify considerably also the 200–600 cm⁻¹ Raman region [60, 65]. Since this is not observed in Fig. 9(b), we assign both peaks at 937 cm⁻¹ and 1110 cm⁻¹ to internal vibrations of TiO₄ tetrahedral units, i.e. the asymmetric stretch of TiO₄ (937 cm⁻¹) and the symmetric stretch of TiO₄ (1110 cm⁻¹). The large Raman intensity of these peaks can be explained by the high polarizabilities α of Ti⁴⁺: ${\text{Ti}}^{4+}:{\alpha }_{T{i}^{4+}}=0.185{\text{Å}}^3$ with respect to ${\text{Si}}^{4+}:{\alpha }_{S{i}^{4+}}=0.0165{\text{Å}}^3$ [66] leading to large differences in the Raman cross section for stretching of Ti-O and Si-O bonds. As for the remaining peaks at 390 and 690 cm⁻¹, a possible assignment would be to symmetric stretching and bending of mixed Ti-O-Si bridges, respectively.

We also plotted the Raman spectrum of molten ULE (green curve) and one spectrum originating from a cavity (red curve) in Fig. 9(a). Both spectra depict a relative intensity increase at 602 cm⁻¹ and intensity reduction at 390 cm⁻¹, 690 cm⁻¹, 937 cm⁻¹, and 1110 cm⁻¹ with respect to unmodified ULE. Furthermore, we detected an additional peak at 1555 cm⁻¹ in the Raman spectrum of the cavity. Laser processing of ULE shifts the broad peak at 430 cm⁻¹ towards higher wavenumbers. The same behavior can be observed when applying mechanical stress to a pristine ULE sample, e.g. by micro indentation. We want to point out, that a weak increase in Raman intensity around 200 cm⁻¹ can be found in the laser processed sample. This is the only region where amorphous Ti₂O₃ is Raman active [54].

To assign the appropriate network changes to certain kinds of laser modifications, we mapped a section of laser modified ULE. For this sample, the lines were inscribed with a repetition rate of 4.7 MHz, pulse energy of 200 nJ, and a line spacing of 20 µm. Afterwards, each spectrum was baseline corrected and normalized. The relative peak intensities in comparison to pristine ULE are shown in Fig. 9(c) for selected relevant Raman peaks. Over the entire modified area the relative intensity of the 602 cm⁻¹ peak (first picture) increases providing clear evidence for the densification of the silicate network through formation of additional three-membered silicate rings in the areas exposed to the laser beam [59, 65]. Thus, ULE is an anomalous glass exhibiting densification after laser processing [67]. For these glasses, the rapid heating of the glass and subsequent very fast quenching results in a higher fictive temperature resulting in a densified state [68].

In addition, the modified area exhibits a reduced peak intensity at 1110 cm⁻¹. Thus, laser irradiation appears to break Ti-O-Si bridges in ULE and to reduce the number of TiO₄ tetrahedra. Interestingly, cavities seem to exhibit a slightly higher peak intensity (cyan color) at 1110 cm⁻¹ than the surrounding molten material, meaning that less Ti-O-Si bridges are broken. This might be due to the densified shell surrounding the cavities. Furthermore, we detected an additional Raman peak centered at 1555 cm⁻¹ directly in the middle of the cavities, shown in the lowest graph of Fig 9(c). This peak is due to O-O stretching of trapped molecular oxygen [68–70] which forms during the laser inscription and is confined within the cavities. This peak reveals that under femtosecond irradiation molecular oxygen is generated in ULE. Interestingly, we could also identify a weak but distinct oxygen peak in the Raman spectrum of ULE containing laser inscribed nanogratings. Lancry et al., already reported the generation of free and dissolved O₂ within the network of SiO₂ after femtosecond irradation [9]. They concluded, that ultrashort pulse irradiation result in atomic O dissociation from the SiO₂ network, which can recombine with each other to produce molecular oxygen [9]. The oxygen molecules can be released at the point of the nanopores but also remain trapped within the glass. These findings agree with the assignment of the high energy absorption bands to ozone.

4. Discussion

The results obtained can be used to explain the laser induced darkening of ULE. The laser processing of ULE leads to the formation of hollow cavities, whereas the processing parameters define the size of the cavities. When working in the nanograting regime (low laser powers) the cavities are rather nanopores with a diameter of several tens of nanometers. In contrast, when using high repetition rates heat accumulation and local melting results in the formation of large cavities with a size of a few micrometers. Regardless of their size, the cavities contain free molecular oxygen. The formation of cavities compresses the surrounding glass matrix creating densified shells enclosing the cavities. This densification is also evident in the Raman spectra, in analogy to mechanically densified silica glass [65, 71]. Raman measurements verified the laser induced breaking of Si-O-Ti bridges and a decrease in the number of TiO₄ tetrahedra. The broad absorption band at 790 nm is due to the generation of Ti³⁺ species in different phases as it was shown by the ESR results. Therefore, we can deduce, that besides the presence of Ti⁴⁺ in tetrahedral sites dispersed in the silicate matrix, irradiation with ultrashort laser pulses induces the formation and co-existence of (a) Ti³⁺ in isolated tetrahedral sites, and (b) Ti³⁺ in aggregated disordered octahedral sites or ’Ti₂O₃ clusters’. These results can be expressed by the following reaction (Fig. 10):

 

Fig. 10 Photoreduction of Ti⁴⁺ to Ti³⁺ under the formation of molecular oxygen (photoxidation of oxide anions O⁽⁻ⁱⁱ⁾ to ${\mathrm{O}}_2^0$) and subsequent reordering of the vitreous network structure is reflected in an increase of Si-O-Si bonds, eventually forming Ti₂O₃ cluster and even more Si-O-Si bonds.

Download Full Size | PDF

Note that Ti is present as Ti⁴⁺ and Ti³⁺ in the left and right hand side of Fig. 10, respectively. Thus, some oxygen ions of the glass network are oxidized and some of the colorless tetravalent titanium ions (Ti⁴⁺, d⁰) are reduced to trivalent titanium (Ti³⁺, d¹, I_max ≈ 800 nm) which is responsible for the darkening. Formation of Ti₂O₃ has been observed also by electron beam irradiation of TiO₂, a process causing reduction of Ti⁴⁺ to Ti³⁺ [72]. Figure 10 explains not only the observed formation of molecular oxygen, but also the increased formation of Si-O-Si bridges at the cost of mixed Ti-O-Si bonds. The densification of the silica network is reflected in the formation of more 3- and 4-membered rings [71]. As the ESR measurements did not allow for spatial resolution, averaging instead over the whole sample, the following material distribution, as sketched in Fig. 11, is supposed to occur after laser processing. The cavity containing free molecular oxygen is surrounded by a shell of compacted ULE glass containing trivalent titanium. This encompassing layer leads to the darkening of the sample. The cavities are mainly located in the upper part of the laser modification. The modification itself consists of previously molten and resolidified glass, due to the application of high repetition rates. When considering the nanograting regime the situation will be quite similar. With the SAXS experiments we could verify the formation of tiny nanopores. Thus, the large cavity of Fig. 11 is replaced by multiple nanopores. Here, we were not able to detect a densified shell due to the limited resolution of the SEM and the low contrast between pristine and compacted ULE. However, we also found the oxygen peak in ULE when working in the nanograting regime.

 

Fig. 11 Sketch of the inner region of laser modified ULE. Processing with multiple laser pulses at high repetition rates leads to melting and subsequent resolidification of the material. In the upper part a cavity is formed, containing molecular oxygen and being surrounded by a shell of compacted glass.

Download Full Size | PDF

The remaining question concerns the formation of the cavities. Numerous groups investigated the formation of tiny voids (diameter of a few hundred nanometers) for single pulse operation and strong focusing conditions [12, 13, 73]. Under these circumstances, very high intensities in the focal volume create microexplosions in the material leading to expanding shock waves [73]. The spherical shock compresses the material and the simultaneously propagating rarefraction wave creates a void in the central volume. Thus, after the microexplosion the whole material initially located in the central area is deposited into a shell of densified material surrounding a hollow sphere. This theory may be useful to explain the formation of nanopores induced during the nanograting inscription. Irradiation with single laser pulses induces tiny voids surrounded by densified glass. The subsequent irradiation with multiple laser pulses facilitates the formation of periodically aligned nanopores which are surrounded by densified material created by the fast quenching and resulting higher fictive temperature [67, 74].

The generation of large cavities (several µm) in ULE, when working in the heat accumulation regime, is not fully understood. As the radius of the expanding shock wave is defined by the pulse energy applied, pulse energies of a few hundreds of nJ are not sufficient to generate these large cavities. However, the theory of void formation was developed for single pulse operation. In contrast when working in the heat accumulation regime, the material undergoes a much larger temperature and internal pressure leading to completely different material properties. Recently, for fused silica two mechanisms were proposed explaining the formation of large cavities: (1) the perturbation and subsequent interruption of the laser beam due to a self induced refractive index change caused by the high temperature of the processed material. The subsequent fast quenching of the material results in an inhomogeneous resolidification and formation of large disruptions [75]. In contrast, (2) Cvecek et al. proposed that gas bubbles are generated during the laser heating of the glass due to an ionization process and thermal dissociation [76]. The gas bubble can grow until they scatter enough laser light to interrupt the laser heating. However, the latter theory cannot explain the formation of a densified shell surrounding the cavity. To resolve the issue about the cavity formation in situ measurements during the process are required and further measurements have to be conducted.

The capability to realize nanogratings as well as to tune the absorption spectra of laser modified ULE, makes ULE a very promising material for data storage applications. Thus, five degrees of freedom (three spatial degrees of freedom, orientation of the optical axis, and the retardance) are potentially available for inscribing nanogratings. Zhang et al. have already demonstrated the benefits for data multiplexing [33]. In addition, with the capability to tune the absorption of the glass one can add one additional degree of freedom resulting in an increase in the data storage capability of at least a factor of 10.

5. Summary

We analyzed the generation of laser induced modifications in ULE glass. This silicate glass has a significant fraction of TiO₂ (7.5 wt%) to ensure a low thermal expansion. Nanograting formation can be observed for all processing parameters limited only by the onset of heat accumulation. The spatial dimensions of nanogratings in ULE are similar to fused silica, consisting of tiny pores with a diameter of about 30 nm and a length of about 300 nm. These pores are aligned into periodic sheets, oriented perpendicular to the laser polarization.

In addition, for pulse energies above 50 nJ the processed glass tends toward darkening. In general, the laser induced darkening increases with the laser power applied. We analyzed the induced structures with UV-VIS, Raman and ESR spectroscopy. We identified the formation of hollow cavities filled with molecular oxygen surrounded by a compressed shell of glass. Laser processing of ULE appears to break Ti-O-Si bridges in ULE and to reduce the number of TiO₄ tetrahedra. Moreover, additional Si-O-Si bridges are generated, as evident from the increased Raman peaks, resulting in a silicate densification in the laser exposed regions of ULE.

Eventually, the photo-reduction of tetravalent titanium (Ti⁴⁺, d⁰) to trivalent titanium (Ti³⁺, d¹) is the reason for the observed laser induced darkening. Laser induced Ti³⁺, a d¹ ion, gives rise to a broad absorption band with a maximum near 800 nm, thus rendering a blue color to the glass. In combination with the absorption bands of the glass matrix, a darkening is perceived due to the broad absorption over the entire visual spectrum. By combining the inscription of nanogratings with laser induced darkening it is easily possible to locally tune the type of the modification by three independent degrees of freedom (retardance, orientation of the optical axis, amount of darkening) rendering ULE ideally suited for future data storage applications.