1. Introduction

The ability to focus ultrafast laser pulses in bulk optical glasses and to create embedded changes of the dielectric function with optical functions, notably refractive index changes (RIC), is the base for a significant range of applications in photonics. The direct three-dimensional access creates an undeniable advantage and ultrafast laser photoinscription has led to the successful fabrication of complex integrated optical devices while, at the same time, has fostered fundamental investigations with respect to the nature of the refractive index change [1, 2].

The strong spatial energy localization characteristic to ultrafast excitation is the key prerequisite for precise material fabrication. This builds up therefore on a unique range of fast, usually non-equilibrium phenomena, that involve confinement conditions, structural changes, and rapid heating and quenching rates with the development of stress fields. The underlying structural modification process is in most cases, at the typical wavelengths of ultrafast lasers (0.8–1 μm), intermediated by nonlinear electronic excitation and its fast relaxation, whereas a debate exists as to the nature of excitation; multiphoton ionization, tunnel-assisted generation of carriers, or collisional multiplication. Equally, questions exist with respect to the particular channels of electronic relaxation as localized defects, structural rearrangements, incoherent phonon modes, and heat. The latter may then involve generation of pressure waves and transitions into viscosity states, up to a gas-phase tranformation. Many studies were carried out in fused silica (a-SiO₂), a model material for ultrafast laser interactions [3–6] that shows typically denser molecular packaging after laser excitation, accompanied by broken molecular Si-O bonds. The fundamental questions can be expanded to a large range of glasses of optical interest (e.g. laser media) where the laser action translates into different types of structures, ranging from slight colorations [7–9], positive and negative alterations of the refractive index [10–16], subwavelength modulation patterns [17, 18], down to micrometer and submicrometer-scaled voids and bubbles [19]. Several scenarios of material modification are currently proposed for glassy materials that involve strained bond restructuring, molecular rearrangement, and kinetics of melting and solidification [3, 4, 10, 20, 21], and were verified using structure-sensitive investigation methods involving photoluminescence and Raman spectroscopy [13, 22, 23].

Currently a practical interest is noticed in the development of price competitive applications, particularly in microfluidics and optical sensing, involving low cost glasses with good optical performances. A scientific concern exists therefore in verifying to which extent the excitation and relaxation scenarios developed for a-SiO₂ are extendable to other glasses (e.g. silicate family). This imply putting into evidence specific material dependent changes of the dielectric functions as well as general features that can be assigned to the material class. Particularly intrinsic relaxation paths and electronic and structural changes associated with either increasing or decreasing the refractive index under light exposure [3, 4, 10, 11] are of interest.

If typically fused silica, due to its densification characteristics based on Si–O ring statistics and its relaxation behavior involving abnormal density increase upon heating and fast cooling (fictive temperature), shows a large processing window for positive RIC, other types of glasses are more difficult to process. We focus here on a glass material still based on a silicate structure but with a fundamentally different character under femtosecond laser irradiation, the Schott borosilicate crown BK7 glass. BK7 is widely used as a low cost technical optical glass for applications in the visible region. The potential use of femtosecond laser pulses for structuring bulk BK7 glass with the aim of inducing RIC and subsequent generation of optical functions was previously demonstrated [15, 24–26], and indicated a certain flexibility in the index change. The optical functions were based on various types of index variations ranging from soft index changes in low energetic regimes in a very narrow processing window, to dominantly local axial and off-axis densification [15, 24–27]. These domains are somehow similar to fused silica despite the completely different thermo-mechanical and relaxation behavior of the BK7 glass [15, 28] (expansive character and low thermal points), putting forward the question of local factors triggering non-uniform densification or rarefaction regions. Aiming to shine light on these issues, the present studies were carried out on N-BK7 glass samples, transparent in the visible domain up to 280 nm (E_g = 4.2 eV). We performed different laser-driven modifications of the BK7 glass, resulting in either positive or negative RIC. Static and dynamic (scanned) regimes in various conditions of irradiation were explored using phase microscopy, surveying the transition between soft positive (type I) and strong modulated (positive or negative; type II) domains. Equally, we have carried complementary photoluminescence (PL) and Raman studies in a space-resolved manner and indicated corresponding electronic and structural material changes associated with the index change and its spatial distribution. The results are analyzed and compared with general features from previously reported observations in fused silica. Summarizing guidelines for processing borosilicate crown glasses are indicated, emphasizing possible ways of exploiting different photoinscription regimes.

2. Experimental setup

Ultrashort 800 nm light pulses from two laser systems were employed to irradiate bulk BK7 glass samples; a regeneratively amplified Ti:sapphire ultrafast laser system delivering 300 mW of power at a repetition rate of 100 kHz and a nominal output pulse duration of 130 fs, and a regeneratively amplified 1 kHz Ti:sapphire laser system providing 120 fs laser pulses at pulse energies of up to 1 mJ. Repetition rates and exposure doses were controlled using externally-driven Pockels cells and electromechanical shutters. Polished BK7 parallelepipedic samples were employed, mounted on a XYZ motion stage that allows translation parallel or perpendicular to the laser propagation axis. The beam was focused inside the target by various focusing optics including long working distance high and low numerical aperture objectives (OB1 with NA_eff =0.09, OB2 NA_eff =0.42; note that the as-defined NA_eff includes the effect of beam truncation at the objective pupil). Both longitudinal writing configurations with translation parallel to the laser propagation axis and in the direction of the laser source, and transverse, perpendicular scan were used. The latter provides elliptical profiles of the trace as the longer confocal dimension upon focusing prevails in photowriting. In addition, for the transverse scan, a slit shaping technique [29–31] was used in combination with 1 kHz repetition rates. This allows to increase the transverse dimension to sizes comparable to the confocal length producing a flat disk of light that enables by perpendicular scanning to obtain circular symmetries in other-wise elliptical transverse forms. An objective OB3 with NA_eff =0.26 and working at a focusing depth of 1.5 mm, with a 350 μm slit situated 20 cm before the pupil was used in this case. An Olympus BX 41 positive optical phase-contrast microscope was employed to image the interaction region in a side-view geometry. The relative positive index changes are appearing dark on a gray background, while white zones indicate negative index variations or scattering centers. The phase contrast microscopy (PCM) and the optical transmission microscopy (OTM) were accompanied by Photoluminescence (PL) and Raman confocal micro-spectrometry. Additionally, the guiding properties were verified upon injection with IR and visible light.

2.1. Photoluminescence measurements

The PL spectra of the photoinscribed traces were recorded with a confocal photoluminescence and Raman microspectrometer in a backscattering configuration (Horiba Jobin Yvon Aramis). Three laser sources were used for excitation, offering four excitation wavelengths: HeNe laser light at 633 nm (1.96 eV), Argon ion (Ar⁺) laser radiation at 488 nm (2.54 eV), and coherent photons from a HeCd laser at 442 (2.81 eV) and 325 nm (3.82 eV). These wavelengths usually cover the excitation bands of nonbridging oxygen hole centers (NBOHC), non-bonding oxygens (NBO), metallic impurities, and oxygen deficiency centers (ODC) as markers of bond breaking. The PL detection is made off-axis or in trace cross-section with a lateral spatial resolution of approximately 1.5 μm. Axial resolution lies at around 3 μm, a reasonable compromise between precision and high signal yields. For the interpretation of the spectra, two issues were further considered, spectral deconvolution of PL signals and defect bleaching. Spectral deconvolution using known band positions was performed in the energetic domain, where the bands have symmetrical forms. Defect bleaching and fluorescence losses in damage traces limits the ability to measure stable defects, and are particularly observed during excitation with Ar⁺ and HeCd laser. Therefore, usually low energies of laser radiation were used. To diminish the role of a time bleaching artefact, the spectra were recorded in the same day of laser irradiation.

2.2. Raman measurements

The Raman spectra were collected with the Raman spectrometer in a backscattering configuration using HeCd laser light at 442 nm for excitation. The spatial resolution is approximately 1 μm. As the detection was carried on cross-sections of longitudinal traces with constant properties along the axis, the depth resolution is not important. Thereby this was adjusted for at 8 μm for type I traces and 4 μm for type II traces respectively, as a trade off between efficient signal collection, reasonable accumulation time (1 min). The spectral resolution lies below 1 cm⁻¹.

3. Borosilicate crown (BK7) glass structure

Borosilicate Schott BK7 is a silicate glass where the chemical composition in percentage by weight reads as follows: silica (70), boron oxide (11.5), sodium oxide (9.5), potassium oxide (7.5), barium oxide (1), titanium oxide, calcium oxide and impurities in small quantities. This leads to some peculiarities in the electronic structure, lower softening temperature and, notably, a high expansion coefficient (approximately ten times higher than for fused silica). Considering the influence of network modifiers such as boron and metal alkali atoms, we note that alkali atoms dramatically change the silica network since they lead to bond breaking of the polymerized silica structure according the following reaction:

Additionally, the presence of alkali metals influences the boron network. Normally, boron has a valence of three in B₂O₃ glass, where the network consists of BO₃ units. However, the addition of alkali modifiers in borate and borosilicate glasses leads to a change of its normal valence. A valence of four imply that BO₄ tetrahedral units appear in the network. In silicate crystals these tetrahedral units could be incorporated in the silicate networks forming types of 4-membered rings consisting of one or two atoms of boron [32]. These are danburite units, R₂B₂Si₂O₈, consisting of two boron atoms in the ring, and reedmergnerite units, RBSi₃O₈, that have only one atom of boron.

Nuclear magnetic resonance studies [33] indicate a possible structural model of reedmergnerite and danburite units composing borosilicate networks. When the molar ratio between alkali, boron, and silica is close to that of BK7, mixing of silica and boron subnetworks occurs. However, a characterization procedure which only uses these units has some difficulties. For example, not all the boron atoms can change their valence. It is known that borosilicate glasses undergo phase separation between boron and silica networks resulting in inhomogeneities with the size of the order of hundred angstroms. According to Konijnendijk [34,35], efficient incorporation of B atoms into the silica network needs high pressures. The presence of non-incorporated B was equally found [33]. Therefore we assume the presence of boron atoms not incorporated in silica network but in the form of piroborate ${\text{B}}_2{\text{O}}_5^{4-}$ or metaborate ${\text{B}}_3{\text{O}}_6^{3-}$ rings.

4. Ultrafast laser processing of BK7 and modification types; static and dynamic regimes

We have firstly performed a parametric scan of various irradiation conditions (mainly above the single pulse modification threshold) in order to define the limits of the processing window in a static, non-scanned situation. These conditions encompass the use of various laser repetition rates, pulse energies, number of pulse per site, and pulse durations The results are shown in Fig. 1 in terms of PCM images. Figure 1 (a–c) indicates the material transformation as a function of the energy, number of pulses and pulse duration for an irradiation source with a low repetition rate (166 Hz). Figure 1 (d–f) depicts the corresponding behavior for a higher repetition rate source (100 kHz), in a regime of stress rather than thermal accumulation [15]. Light focusing was made using the OB2 with NA_eff =0.42.

 

Fig. 1 Typical ultrafast laser-induced static transformations in BK7. Top row: The evolution of the static index change with energy (a) (for single pulse exposures), number of pulses at constant energy per pulse (b), and pulse duration (c) for low repetition rates (166 Hz). The pulse duration in (a,b) is 160 fs and the input energy for (b,c) is 1 μJ. Bottom row: The evolution of the static index change with the energy (d), number of pulses (e) and pulse duration (f) for high repetition rates (100 kHz). The exposure dose corresponds to 50000 pulses in (d,f). The pulse duration in (d,e) is 160 fs. The input power in (e) and (f) is 65 mW, 17 mW (0,65 and 0,17 μJ), respectively. Other experimental conditions are marked on the figure. All structures are made at a depth of 250 μm below the input surface using the OB2 with NA_eff =0.42.

Download Full Size | PDF

The images show as a rule dominant negative index changes (white color) with signs of lateral compaction (black colors) in more energetic conditions, generally denoted as type II, and a contrast that is slightly dependent on the irradiation dose. Previous modeling results that coupled the nonlinear propagation of femtosecond light pulses in BK7, simulation of spatio-temporal excitation footprints and of the subsequent thermo-elasto plastic response of the material [15] indicated that this typical appearance is related to the formation via electron excitation of typically above 10²⁰ cm⁻³ of a heat source of several hundred degrees leading to material expansion and redistribution of stress [36, 37]. This behavior is facilitated by a relatively strong expansion coefficient (α=86×10⁻⁷°C⁻¹), indicating that expansion above a certain excitation threshold is a standard relaxation path delivering rarefaction and negative index changes with shallow compaction rings around (from mass conservation considerations). The extension and the strength of the lateral stress as well as the level of rarefaction depend on the magnitude and the geometry of the heat source which are controllable by the focusing strength, input energy (Fig. 1(a,d)), irradiation dose (Fig. 1(b,e)), and pulse duration (Fig. 1(c,f)) [15]. The later control factor is particularly strong as a longer pulse duration, provided that the excitation is still sufficient, allows to diminish plasma defocusing effects during the nonlinear energy deposition [38] by delaying the carrier generation with respect to light delivery. The consequence is a stronger energy concentration for pulse durations around 2–4 ps (Fig. 1(c,f)), with a characteristic axial compaction pattern for high repetition rate excitation (Fig. 1(f)).

However, as long as ultrashort multipulse sequences are involved in a less tighter focusing geometry and at very low energy densities, well below the single pulse modification threshold, the BK7 material shows a positive index change accompanied by a slight coloration of the sample. This is indicated in Fig. 2(a). These RIC can be made either using low NAs or with the slit-shaping technique that preserves a low energy concentration and short pulse duration, a necessary condition to achieve low excitation levels (typically below 10²⁰ cm⁻³ with parameters, i.e. three photon absorption coefficients in the range given in Refs. [15, 39]). The present photoinscription conditions involved the use of OB1 with NA_eff =0.09 to focus 0.5 μJ pulses of 150 fs at 100 kHz and with the translation speed of 100 μm/s. The structure resistance to heat is low, the traces being erasable upon annealing at above 300 °C. Interestingly, this low energy density photosensitivity is often encountered in fused silica (denoted type I), where it serves as the main base for low-loss optical guiding (<1 dB/cm). Its origin is often associated to an augmentation of the fictive temperature in silica glass (a point still under debate), which, we recall, upon heating and fast cooling, shows an increase of its mass density with a frozen more compacted structure [6, 40, 41]. However, the BK7 borosilicate has a standard normal thermodynamic glass transition volume-temperature dependence, where a decrease of density is expected upon heating and fast cooling. This leads to a behavior that requests further investigation. If the energy is further increased, a situation already anticipated by the static low index traces in Fig. 1 is encountered, where a modulation of the refractive index and the onset of negative RIC zones are observed.

 

Fig. 2 Various regimes of modification induced by longitudinally-scanned (LW) femtosecond (160 fs) laser pulses at 100 kHz repetition rate in BK7 glass as a function of the writing conditions. These are: (a) PCM image of type I weak positive RIC, (b) PCM image of type II-NGN with negative RIC, and (c) strong positive RIC type II-GP region in the core surrounded by negative RIC cladding type II-GN. The insert shows the corresponding static traces. The irradiation conditions are given on the figure.

Download Full Size | PDF

Hereafter we will focus for simplicity only on dynamic scanned traces (via sample translation). A series of characteristic types of damages displayed in Fig. 2 are discussed below since they encompass a large range of physical situations. Taking into account that different exposure conditions produce various types of damage, as in the case of fused silica [18], we will separate the resulting traces traces according to the focusing conditions and the optical properties of the generated structures in two groups; type I and type II traces (by analogy with fused silica). We can conclude that at very low laser fluences shallow smooth positive index changes (Δn ≈10⁻⁴) can be produced, denoted type I (Fig. 2(a)). Increasing the laser fluence towards the onset of nonlinear effects, negative RIC can be produced, indicated here as type II-NGN traces (see Fig. 2(b). The traces were generated by focusing 1 μJ pulses via OB2 NA_eff =0.42 at 100 kHz with the translation speed of 100 μm/s. The reason for the negative RIC can be associated with thermomechanical expansion of the heated region followed by quenching in the low density phase [15], as indicated in the previous section. Increasing the pulse energy and the effective number of acting pulses (via e.g. scan velocity) even more, positive refractive index changes could be induced in the core again, which we will denote as type II-GP (Δn > 5 ×10⁻³). This can be seen as a replication of the compression shell appearing around the interaction region that creates a compaction trace in the center upon scanning [38] (see insert in Fig. 2(c)). This region is surrounded by a zone of negative refractive index changes denoted as type II-GN, remnant of the expansion. These modifications were generated by focusing 3 μJ, 150 fs ultra-short laser pulses at 100 kHz with the translation speed of 10 μm/s via the objective OB2 of NA_eff =0.42. As a side note, no nanogratings were observed in case of type II in BK7 glass [18].

5. Type I modifications in BK7

5.1. Slit shaping

The type I domain represents the transformation range where low-loss guiding without strong mode confinement can be achieved [24, 27]. As indicated before, this is facilitated by low excitation doses and, subsequently, reduced excitation levels, making the slit-shaping method ideal for this regime. We will therefore focus on traces created by this method. Slit-shaping experiments were performed at low repetition rate (1 kHz), as the conditions allow for low energy exposure, below the onset of strong nonlinearities, while maintaining a circular symmetry of the trace. The results are given in Fig. 3. Figure 3(a) indicates the side PCM and OTM images of the type I traces, pointing to the existence of a coloration region in the positive RIC zone. Figure 3(b) shows the axial transillumination microscopy images for traces written at different input energies, indicating the spatial distribution of the excitation within the trace. Note the increase of symmetry and the half-moon characteristic imprint of the positive index change and the high symmetry of the guided mode at 633 nm. A typical simulated footprint of the nonlinear irradiance distribution in the exposure region is given in Fig. 3(c). The spatial distribution of the PL signal given Fig. 3(d), indicated here for the higher energy trace cross-section in Fig. 3(b), matches the excitation and index distribution and will be discussed in the next sections. The corresponding index changes measured via the far-field mode characteristics are low, in the range of 10⁻⁴, and scale linearly with the input energy (Fig. 3(e)). Most of the traces are multimode at 633 nm and do not guide at 1550 nm. A slightly lower index is noted for traces photoinscribed by circular polarization as compared to those written by linear polarization.

 

Fig. 3 (a) Side PCM and OTM images of type I traces made by slit-shaping transverse scanning (200 fs). Note the coloration of the glass in OTM. (b) Axial transillumination microscopy images for traces written at different energies corresponding to those shown in (e) and corresponding guided modes at 633 nm, (c) Simulated footprint of the nonlinear irradiance distribution in the exposure region, (d) Red PL (650 nm) signal distribution after 633 nm excitation (upper part) and green PL (550 nm) after 488 nm excitation (bottom) in highest energy traces (b). (e) Dependence of the positive index change on the input energy.

Download Full Size | PDF

6. Type II modifications in BK7

In conditions of direct and tighter focusing with moderate input energies, the electronic excitation level superior to 10²⁰ cm⁻³ and the local heat deposit of several hundred degrees in BK7 trigger local pressures in the MPa range [15]. Thermo-mechanical expansion induced by a single pulse leads to a rarefied region in the focal zone where the density falls down by 10⁻³ and a compression shell appears around the region (Fig. 1(a)). This yielding is facilitated also by a low softening temperature. The development of mechanical stresses, their release following heat flow, and the associated density redistribution [15, 37] are the main characteristics of the type II range of thermo-mechanical origin. Depending then on the photoinscription geometry, guiding traces can be written (type II-G) with GP and GN parts, taking advantage on the stress-induced compaction. The compaction can replicate a positive index change on axis, erasing the preceding low density modification in case of longitudinal writing [15,38], or it can set off axis, in the neighborhood of the interaction region [26]. Several examples are given in Fig. 4(a–c).

 

Fig. 4 Guiding properties of type II traces in different ultrafast photoinscription settings. Depending on the photo-writing geometry, guiding compacted regions are either central or off axis. (a) Side image of longitudinal trace type II G (with GP, GN parts) written at 100 kHz, (b) Side image of a transverse trace at 100 kHz, (c) Axial image of a transverse trace at 1 kHz. Typical guided modes are indicated at the right side, showing either central or lateral guiding in stressed regions. The direction of the laser beam is marked.

Download Full Size | PDF

7. PL spectroscopy from laser-modified regions

In order to get insights into the electronic and structural changes of the different photoinscription domains, we have tried to characterize these specific regions via PL and Raman spectroscopy. In view of the complexity of the spectral information, several possible scenarios are described below, either based on defects in Si and B networks or relying on metallic impurities. The objective was to indicate in a space-resolved manner specific transitions characterizing type I and type II structures as markers of the photoinscription regimes.

In a first perspective, defect excitation occurs via several absorption bands attributed to borosilicate glasses [42, 43]. Following optical studies on alkali borosilicates [42] we will take into account ${\text{H}}_1^{+}$ ${\text{H}}_2^{+}$, ${\text{H}}_3^{+}$, ${\text{H}}_4^{+}$ centers, i.e. non-bonding oxygens. These hole trap centers are found in alkali-silicate classes [8, 43]; however, due to certain similarities in the chemical composition and network structure, we expect the appearance of these centers in BK7. In addition to these hole-trapping centers placed near the valence band, one could account equally for electrons traps placed near the conduction band. Here we will consider only hole-trapping centers since the used excitation sources cover their characteristic absorption energies (Tab. 1).

Table 1. Absorption and PL bands of point defects reported for borosilicate glasses.

View Table

In terms of PL we have observed the 540 nm (2.3 eV), the 650 nm (1.9 eV), and the 775 nm (1.6 eV) bands under different excitation sources. An intrinsic emission is observed in the pristine sample. The bands are indicated in Fig. 5(a,b) for unmodified and type I modified samples. The assignment of these PL bands, though complex and not without ambiguities, could be related to intrinsic point defects but also to the luminescence of metallic ion impurities (Fe, Ti) included in the network. For example, numerous works reported broad luminescence bands in the red part of the visible spectrum produced by Fe³⁺ ions in alkali-silica and borosilicates. Photoluminescence was reported in these host materials centered at 625–725 nm for tetrahedral Fe³⁺[44] and in the near infrared at 875 nm for octahedral Fe²⁺[45, 46].

 

Fig. 5 (a) PL spectra of the pristine BK7 glass for different excitation wavelengths indicated respectively by different colors. The red line is the HeNe 633 nm (2 eV) excitation, light blue - Ar⁺ 488 nm (2.54 eV), blue – HeCd 442 nm (2.81 eV), and magenta – HeCd 325 nm (3.8 eV). (b) PL spectra observed for type I structures, showing bands at 540 nm, 650 nm, and in the 750–775 nm range, with the potential assignment given in the text. The PCM image of type I modification is given in the insert.

Download Full Size | PDF

The 650 nm band observed under Ar⁺ 488 nm (2.54 eV) excitation and the 775 nm band appearing under HeCd 442 nm (2.81 eV) excitation are partly similar to those usually formed in fused silica. There they can relate to NBOHCs [5]. Additionally, as will be seen below, the intensities of these two bands show a certain correlation for the different modified regions. Thus, we firstly assume that the PL of these bands and the PL at 775 nm under HeNe 633 nm (2 eV) could be caused by NBOHCs. It is important to note that not only PL patterns are important to understand the defect structure, but equally excitations used, since relaxation could happen with the assistance of third centers. If 633 nm directly excites potential NBOHCs and ${\text{H}}_3^{+}$ centers, the 442 nm and the 488 nm do not interact resonantly with those levels, particularly NBOHCs, while they can excite ${\text{H}}_2^{+}$ centers, Fe³⁺, and eventual oxygen deficiencies. We note also that the 650 nm and the 775 nm PL bands are not excited similarly-effective by 442 nm and 488 nm excitation. Regarding ${\text{H}}_2^{+}$, ${\text{H}}_3^{+}$, and NBOHC, all centers related to non-bridging oxygen (NBO) in different environments, luminescence at 650 nm and to a lesser extent at 775 nm characterizes the appearance of broken bonds in that regions. We may presume therefore that the indirect excitation of NBOHCs involve a third intermediate center that, upon excitation, modifies the environment and facilitates population transfer. Concerning the 540 nm (2.3 eV) band, there is no clear assignment, but recombination in the B subnetwork [47], Ba impurities [44], and oxygen deficiencies can be involved, though the intrinsic PL in pristine samples can derive from alkali-modified glass structures and dangling oxygens [48]. This region requires further identification.

The next paragraphs will summarize PL results for different types of modifications. Typical PL spectra of type I traces (off-axis geometry) displayed in Fig. 5(b) with respect to the pristine sample (Fig. 5(a)) indicate a slight increase of the 775 nm band under HeNe laser light (633 nm) and HeCd laser light (442 nm). Also an increase of the 650 nm band under Ar⁺ 488 nm light is especially visible. This appears to indicate an increase in the number of NBOHC states as previously indicated in Refs. [27]. Little et al. [27] correlated this with breaking B-O links upon light exposure (see equally the Raman subsection). However, for the 442/488 nm excited 650 nm PL it is equally possible that the concentration of some third center involved in the reaction increases since, with Ar⁺ (488 nm) and HeCd (442 nm) radiation, the NBOHC levels are not resonantly accessible and the population of NBOHCs occurs in a modified charge environment after energy transfer via potential intermediate defect states.

In case of type II traces, the appearance of index modulation makes a space-resolved spectral study challenging. Since type II (NGN, GP, GN) traces show radial inhomogeneities distributed across the section of the longitudinal traces, the samples were cut and polished to study the cross-section of these traces. The corresponding spectral maps are plotted in Figs. 6 and 7 where Fig. 6(a) and Fig. 7(a) show the PCM images and the geometry of investigation, Fig. 6(b) and Fig. 7(b,c) show the spectral PL features for excitation at several wavelengths, and Fig. 6(c–e) and (f,g) and Fig. 7(d–f) and (g,h) depict the spatial distribution and deformation of PL upon blue (442 nm) and red (633 nm) excitation. HeCd 442 nm light excitation reveals the 540 nm luminescence band characterizing unknown defects and the 775 nm band characterizing NBO-HCs and Fe³⁺, while the 633 nm excitation indicate solely NBOHC and Fe³⁺ PL.

 

Fig. 6 (a) PCM image of type II-NGN modification in BK7 glass showing schematically the geometry of cross-section map spectroscopy. (b) The associated PL spectra measured in the trace cross-section. Color assignment of different excitation sources is the same as in Fig. 5. Bottom: 2D spectral maps of (c) the 540 nm band. (d) 650–775 nm band intensities and (e) 775 nm band shift upon excitation with HeCd (442 nm) light. (f) 775 nm band intensities and (g) 775 nm band shift upon excitation with HeNe (633 nm) light.

Download Full Size | PDF

 

Fig. 7 (a) PCM image of type II-G modification in BK7 glass showing schematically the geometry of cross-section map spectroscopy. PL spectra of (b) type II-GP and (c) type II-GN regions. Color assignment of different excitation sources is the same as in Fig. 5. Bottom: 2D spectral maps of (d) the 540 nm band. (e) 650–775 nm band intensities and (f) 775 nm band shift upon excitation with HeCd (442 nm) light. (g) 775 nm band intensities and (h) 775 nm band shift upon excitation with HeNe (633 nm) light.

Download Full Size | PDF

In case of type II traces, 2D spectral PL maps show already significant modifications in both 775 nm and 540 nm bands, with visible increase of the bands intensities and a shift of the 775 nm band position, augmenting its low-wavelength part. We will see below how the different index modulations in type II regime are individualized. Considering the type II-NG traces (Fig. 6) and corroborating with the PCM images (Fig. 6(a)) and luminescence bands upon several excitation wavelengths from UV to the red part of the visible spectrum (Fig. 6(b)), two regions of interest can be identified, with characteristics appearing for blue and red excitation where main PL bands are well-defined. The central dominant region of the negative RIC is the region where the 540 nm band and the 650 nm band intensities increase upon 442 nm excitation (Fig. 6(c–e)). The 775 nm band intensity augments, being accompanied by the blueshift of the band position. Same pattern was obtained for 633 nm excitation (Fig. 6(f,g)). Additionally, on the periphery of the excitation region, another region with different spectral patterns can be identified. There the blue-triggered 650 nm and 775 nm bands exhibit the most visible changes, but the 540 nm band rests unmodified. The spectral signature of this region is similar to the type I region.

In a similar manner, the spectroscopy map of the type II-G guiding trace (Fig. 7) shows two different spectral patterns. These, corresponding to the PCM zones (Fig. 7(a)) are given for several excitation bands in Fig. 7(b,c) and we concentrate the discussion for blue and red excitation. In the type II-GP core region of positive RIC, a strong increase of the blue (442 nm) excited 540 nm and the 775 nm PL bands was observed (Fig. 7(d,e)) with respect to pristine. Additionally, the 775 nm PL band is strongly distorted (Fig. 7(f)). In the surrounding tubular type II-GN region of negative RIC an even stronger increase of the 540 nm band was detected. As the excitation does not resonantly activate NBOHCs, we may assume that this is the distribution of a precursor center, particularly attached to the negative RIC zones. In addition, when directly excited by 633 nm (Fig. 7(g,h)), the 650 nm and the 775 nm bands are not so intense as in the region of the core, indicating a smaller concentration of NBOHCs in that region.

Summarizing the observations, a pattern can be identified. In regions of positive RIC the 650 nm and the 775 nm bands showing possible presence of NBOHC are always well defined, indicating a dominant presence of these centers. The strongest PL pattern appears following red absorption. However, in the region of negative RIC the band at 540 nm related to yet unknown intermediate centers starts to play a visible, even dominant role, suggesting a different chemical and structural environment, with an enhanced blue-absorption-triggered PL. This is particularly reinforced by non-resonant and intermediate activation of red NBOHC emission.

An alternative scenario concerning both type I and type II traces may involve excitation of Fe³⁺ via quasi-resonant blue excitation at 442 nm and 488 nm [45], and its luminescence at 775 nm. This PL competes with NBOHC relaxation. The augmentation of Fe³⁺ may derive from potential laser oxidation of octahedral Fe²⁺ in the presence of alkali impurities and the generation of non-terminated oxygen bonds. A maximal emission would mark regions exposed to higher field intensities. We note that electron traps and Fe³⁺ relaxation are competitive [47], therefore other factors should play a role. The red band distortion with the augmentation of the high-wavelength part (towards 700 nm) suggests structural changes in the glassy matrix.

8. Raman spectroscopy from laser-modified regions

We give below a complementary view of the ultrafast laser-modified structures using Raman spectroscopy. A recent report [27] already suggested, repetition rate dependent, an important role of bond-breaking upon ultrafast exposure, particularly at low repetition rates. We offer here a spatially-resolved study in correlation with PCM and PL. Typical Raman spectra of BK7 [27, 34, 35, 49] are displayed in Fig. 8(a). Since BK7 is composed from silica in a proportion of 70 %, it is natural to consider vibrational bands associated with silicate and boron subnetworks including possible vibrations of Si–O–B units. Assuming this, the following assignment of the bands could be given. Firstly, vibrations of the silica network are in many details similar to those observed in fused silica [50]. The exact band positions however may change due to the environment. The observed features are summarized below: a main band centered at 520 cm⁻¹ which is a result of Si–O–Si bending vibration, and a 790 cm⁻¹ band arising from vibrations of the silica network with substantial movement of Si atoms. The strong peak in the high frequency region appears due to the vibrations of the nonbridging oxygens. It consists of three bands at 950 cm⁻¹, 1080 cm⁻¹ and 1160 cm⁻¹ indicating stretching vibrations of Si–O bonds in the Q₂, Q₃ and Q₄ units, i.e. SiO₄ tetrahedra with two and one nonbridging oxygens, and fully connected tetrahedra [51].

 

Fig. 8 (a) Raman spectrum of the pristine BK7 glass. (b) Raman spectrum of type I positive RIC longitudinal trace (made by slit-shaping with circular polarization, index contrast 1.76×10⁻⁴). (c) Raman spectrum of nonguiding longitudinal trace of negative RIC (type II-NGN). (d) Raman spectrum of the negative RIC cladding region for the guiding longitudinal trace (type II-GN). (e) Raman spectrum of the positive RIC core region of the guiding longitudinal trace (type II-GP). For type II, Raman spectroscopy was realized in cross-section.

Download Full Size | PDF

Considering the vibrations of the boron subnetwork, we take into account B incorporated in the silica network and pure B network. Several indications based on the spectroscopic observations point out to the validity of this argument. The characteristic peak at 628 cm⁻¹ was assigned in different works either to danburite-like units or to metaborate units which are 3-membered boron rings ( ${\text{B}}_3{\text{O}}_6^{3-}$). Reedmergnerite and danburite ring units vibrations at 580 and 614 cm⁻¹ respectively were already indicated [32]. These were observed in crystals, thus, narrow lines were reported. However, assuming a wide distribution of Si–O–B interbonding angles incorporated into the silicate network, as characteristic to amorphous networks, wide bands should be equally observed for these units. A more plausible explanation will be the existence of small rings by analogy with D₂ peak characteristic to the formation of 3-member rings in SiO₂. These are metaborate units ${\text{B}}_3{\text{O}}_6^{3-}$ reported for potassium (sodium) metaborate K(Na)₂O·B₂O₃ crystalline networks. The position of the Raman peak can be tuned in a certain range depending on the chemical composition of the glass. In such a way crystals have Raman lines placed either at 610 cm⁻¹ (for K) or 630 cm⁻¹ (for Na). These vibrations correspond to symmetric bending vibrations of all three boron atoms in the ring [35,52]. Nevertheless, we will not exclude wide bands in the region of 600 cm⁻¹ which could possibly derive from bending vibrations of the Si–O–B bridge. Finally, in the high-frequency region at 1470 cm⁻¹, a band associated to the stretching vibrations of boron nonbridging oxygens (>B–O) is placed [50].

As in the case of PL, Raman spectroscopic measurements were performed on the cross-sections of the longitudinal traces. After background subtraction, the spectra were normalized at the main band intensity of the unmodified BK7 glass sample. For further interpretation, differential Raman spectra were calculated, indicating the difference between the spectra of normalized modified and pristine glass regions.

8.1. Type I traces

Since only small positive variations in the refractive index were measured for such traces (Δn ∼10⁻⁵–10⁻⁴), equally, minimal changes in the network reorganization were detected. These are illustrated on the differential Raman spectrum depicted in Fig. 8(b). Negative changes in the band intensities occur at 450, 630, and 1080 cm⁻¹ which are the positions of the main spectral maxima of the pristine BK7 sample. At the same time, the spectral signal rises up in the range of 800–900 cm⁻¹. This can be interpreted either by increasing of a particular spectral band in that region or by the broadening of main and high frequency bands. The latter means introducing a certain degree of disorder in the lattice.

One question relies to the mechanism of positive soft RIC in type I BK7 and we believe that this kind of structural modifications shows significant differences as compared to the positive RIC induced in fused silica, despite their optical resemblance. It is usually invoked that in the fused silica case the optical properties are connected with changes in material density and the degree of structural compactness and could, arguably, be deduced from the fictive temperature model [22]. The non-standard behavior leads to higher density upon fast cooling (see. e.g. Fig.4 in [6], and is often considered a strong contribution to positive index changes among other densification factors. In the case of BK7 glass the thermodynamic situation is not similar. We expect a standard cooling behavior with a density decrease upon cooling due to the domination of the thermal expansion of the BK7 glass network. Thus, the thermodynamical approach should give negative refractive index changes. We observe nevertheless a RIC apparently similar from an optical perspective to the detected type I in fused silica, which puts forward the question of the existence of a more general reaction type or, on the contrary, particular photosensitivity paths in glasses. In addition the type I regime in BK7 shows a slight coloration (note the OTM and PCM images in Fig. 4(a)) and may completely disappear after annealing above 300 °C.

The presence of defects, NBOHC and possible oxygen deficiencies or activated impurities (e.g. Fig. 5) and negligible Raman signature indicates that, contrary to fused silica, type I modification in BK7 glass is characterized by bond breaking with little of network reorganization. Such kind of rearrangements can induce refractive index changes via variations in atomic polarizability. The role of defects, notably NBOHC, was recently suggested [27].

8.2. Non-guiding type II longitudinal traces

We recall that type II-NGN show a negative index change caused by a decrease of density presumably related to material thermal expansion. At the microscopic level, a certain class of particularities were detected for vibrations characterizing silica subnetwork in type II-NGN traces (Fig. 8(c)). To this extent we observe slight changes in the main band. We assume that some vibrational acoustic modes of around 300–400 cm⁻¹ decrease in strength. In the high-frequency region changes in the spectral form of the 1080 cm⁻¹ band were equally measured. The changes can be attributed to shifts of the 1080 cm⁻¹ band towards higher frequencies with the slight decrease of 1160⁻¹ bands indicating some degree of bond breaking in the silica subnetwork. Mechanical density changes occur in this case without strong bond rearrangements.

The boron subnetwork undergoes on the contrary significant changes. The 630 cm⁻¹ peak decreases in contrast and becomes broader indicating the breaking of bonds in the boron subnetwork (and the formation of nonbridging oxygen bonds). The detected augmentation of the 1470 cm⁻¹ band indicating the increase of B–O broken bonds confirms the observation, and should lead to specific B defect centers and related non-bonding oxygens [53]. This is the main particularity of this regime, accompanying the mechanical density reduction.

Finally, a small peak centered at 1540 cm⁻¹ signaling molecular oxygen appears. This confirms a bond breaking scenario accompanying rarefaction/cavitation behavior similarly observed in β-irradiated borosilicate glasses [54].

8.3. Longitudinal traces, guiding type II traces

The type II guiding traces present equally compressive and rarefaction regions. The silica subnetwork undergoes the greatest changes in this writing regime (Fig. 8(d,e)). In the positive RIC region (type II-GP), the main peak (520 cm⁻¹ range) shifts approximately 7 cm⁻¹ to higher frequencies (Fig. 8(e)). We consider this as a sign of densification via the reduction of the interbonding angles in the silica network. At the same time considerable changes happen in the high-frequency spectral region. The 950 cm⁻¹ band slightly increases, the 1080 cm⁻¹ increases observably, while the 1160 cm⁻¹ band decreases. This involves the rise in the number of SiO₄ tetrahedra with one and two nonbridging O atoms, with a number decrease of fully connected tetrahedra, indicating efficient processes of bond breaking. In the boron subnetwork the 630 cm⁻¹ peak slightly decreases, which means changes in the metaborate ring structure. The background level under this band goes up, which could be a result of the increase of the 650 cm⁻¹ region associated with the changes in the main band. The 1470 cm⁻¹ band becomes higher, indicating the presence of B–O nonbridging oxygens. The presence of molecular oxygen is signaled via the small peak at 1540 cm⁻¹[55]. We recall here that the GP region onsets upon scanning via a erasure of a previous central GN structure during the translation of the static trace. The axial positive RIC region is also sustained by the bond breaking in the whole network, accompanying the hydrodynamical elastoplastic mechanisms that lead to material redistribution and subsequent compaction in this region [15].

Considering the negative RIC region (type II-GN), no significant shifts in the main band were detected; however, its form changes. Equally, almost no changes occur with the 630 cm⁻¹ peak. Observing the high-frequency region, the intensities of the 950, 1080 and 1470 cm⁻¹ bands are increasing; however, the character of the band changes, becoming different as if compared to the type II-GP region. For example, the 1080 cm⁻¹ peak is less intense. The signal at 1540 cm⁻¹ characterizing molecular oxygen is scarcely detectable in type II-GN region.

These observations allow us to conclude the following: the negative RIC observed in this GN case despite its mechanical origin has some specific features as compared to the ones observed in type II-NGN regime. Firstly, important spectral changes occur in the bands characterizing silica subnetwork. Secondly, it is situated on the periphery of the traces, where the intensity of laser field was the lowest; therefore, structural changes are possibly caused by the incubation effect or due to thermomechanical mechanisms and subsequent stress accumulation effects.

For a comparative study, in the case of the type II-NGN traces, the 630 cm⁻¹ and 1470 cm⁻¹ bands, characterizing the boron subnetwork undergoes greater changes. In addition, the modified zone is placed on the axis of the laser beam. Thus, rupture of the bonds mainly in the boron/boron-incorporated network is caused by strong interaction with the light and relaxation of the photoionized material. This effect is accompanied by oxygen dissociation, which rests stable in the form of O₂ long time after exposure. The latter was previously detected in β-irradiated silicate glasses [54], where oxygen can be stored in the generated pores (small bubbles of 50 Å diameter), but also in GeO₂[55]. Therefore, it indicates that negative index changes of type II-NGN traces could be explained by density changes produced due to material expansion, less strong for type II GN.

Overall the supposed mechanical scenario where expansion leads to rarefaction and to the appearance of a shell of compacted material around the focal zone with positive RIC (that is being replicated upon translation leading to mechanical compaction on the guide axis) seems to be validated. For the rarefaction region (where naturally the Raman signal is lower and can be contaminated by the surroundings) the spectroscopy results do not allow to indicate if the density has continuously decreased (via. e.g. less packaging) or a mesoporous nature is in place.

9. Conclusions

We have investigated both positive and negative refractive index changes induced in borosilicate crown BK7 glass irradiated by femtosecond laser pulses. The produced modifications can be divided in the two classes: type I shallow positive refractive index changes and strong contrast, stable to heat type II modifications. The latter could be equally divided in positive or negative RIC domains. We demonstrated that under low energetic density, smooth type I positive RIC are induced. However, these kind of structural modifications have a significant difference with positive RIC induced in silicate a-SiO₂. Type I modification in BK7 glass are shown to be characterized by bond breaking, with only slight network reorganization. Thus, in this case, positive RIC appear to be induced by changes in the atomic and molecular polarizabilities. Type II modifications generated under high laser energies are characterized by substantial bond breaking and defect formation accompanying the mechanical deformation and redistribution of material. PL spectroscopy shows specific defect distributions in spatially-modulated index areas, with a preferential red absorption and accumulation of possible NBOHC in positive RIC zones, while a different chemical environment is present in the negative RIC domains, with enhanced luminescence triggered by blue light absorption. A spatial selectivity in the Fe state may equally be present. The Raman spectra associated with positive index changes indicate therefore specific transformations, notably a main band shift to higher wavelengths, sign of densification characterized by the reduction of Si–O–Si or Si–O–B angles. Some chemical redistribution can also occur. In the rarefaction region, laser pulses induce negative refractive index changes characterized by bond breaking and changes mainly in the B subnetwork. Additionally, we were able to indicate generation of molecular oxygen in these regimes of interaction.