1. Introduction

Femtosecond laser direct writing starts with multiphoton photoionization [1, 2], allowing to deliver a finite quantity of energy in very short time. This results in 3D highly localized breakdown [3, 4] with minimal collateral damages. Up to date no other manufacturing process has the potential to integrate 3D multifunctional devices made in a single monolithic chip and within a variety of transparent materials [5, 6]. In particular, it is well known that the coupling between the electromagnetic field and a plasma density fluctuation wave give rise to self-organized structures. Those nanostructures are subsequently imprinted into the glass through periodic plasma energy absorption and plasma electron trapping [7]. This was the first time that quasi-periodic sub-wavelength structures induced by light were demonstrated within a material volume. These nanogratings have explained observed strong form birefringence [8], a property that can be controlled, allowing the integration of plenty optical devices within a single optical substrate chip. Recent publications report optical rotation in integrated waveguides, waveguides retarders, optical vortex waveplates, achromatic polarization rotator rotated waveplates and optical data storage [9, 10].

Shimotsuma et al. [7] showed contrast nanogratings in back-scattered electron imaging corresponding to atomic density contrast. Chemical analysis by Auger spectroscopy revealed that it could correspond to oxygen depletion and related density modulation [7]. Hnatovsky et al. [11, 12] reported nano-cracks and it is not clear if whether these nanogratings can best be described as highly modified regions of differing materials (e.g. through bond breaking accumulation) or as some nanovoids. Regardless of the precise mechanistic explanation of nanoplanes (nanoplasma [13], photon-plasmons interference [14], plasmon polaritons [15] or complex self-organization similar to a Turing structure), we have shown that nanoplanes are characterized by glass decomposition with oxygen released [16]. This nanoporosity has been confirmed recently by Richter et al. using small angle X-ray scattering [17]. In 2013, the Rui Vilar group [18] has also reported the formation of SiO_x nanocrystals within nanoplans that is in agreement with oxide decomposition. Specifically, porous nanogratings and related anisotropic index changes are a spectacular manifestation of glass modification and have been reported primarily in pure silica and slightly doped silica. Very recently, Richter et al. [19] have reported a significant birefringence in 7TiO₂-93SiO₂ glass and maybe in Borofloat [20] (but not in BK7, a crown borosilicate from Schott) but there are no direct evidence related to the formation of nanoporous structures. In contrast, within BK7, pyrex, 16SnO₂-84SiO₂ or sodalime silicate glasses no nanogratings have been observed although strong damage and permanent refractive index changes occur anyway.

The formation of these nanopores enables a plethora of new applications. In the chemical sciences, for example, controlled pore formation is of great interest in optofluidics and nanofluidics, chromatography, nano and molecular sieves and so on - a top-down laser fabrication based approach offers many attractive features to existing work on bottom-up self-assembly methods [21], including spatial localisation and control that is amenable to integration with lab-in-a-fibre, lab-in-a-microfibre [22–24], and, indeed, more familiar lab-on-a-chip technologies. Understanding the laser and material interactions giving rise to these features is of immense importance.

Recently, some tentative explanations based on fast glass quenching have been put forward to explain that nanopores are expected in anomalous glasses like silica and ULE glass. Due to the anomalous behavior of the silica glass density with cooling rate (final glass having a smaller specific volume) it has been postulated in the first instance, given the short timescales in which energy is imparted into the amorphous network, that there is an enormous stress created during the subsequent, fast relaxation process originating from the surrounding glass resisting the change of volume in the interaction volume. This stress is equivalent to having enormous tensile stresses, on the shrunk glass as explained in [21]. It is intuitively obvious that there must be holes, cracks or nanopores formed for the quenching to occur – such dramatic changes under high effective local temperature and pressure lead to zeosil-like glass formation. Therefore, in the first instance the creation of nanopores are not expected in “normal” glass materials. Taking into account an associated reduction in silicon coordination where extraction of oxygen from the network has occurred, the processes can be further complicated. Here, we explore these ideas to understand the underlying physico-chemistry and the new potential it brings. A set of GeO₂-SiO₂ PCVD optical fiber preforms were manufactured and their density-fictive temperature (ρ-T_fic) dependence measured to spread across both the anomalous and normal glassy regimes. Subsequently, we have analyzed the laser-induced nanostructures using birefringence and scanning electron microscopy.

2. Experimental details

Several kinds of glasses produced by Plasma Chemical Vapor Deposition (PCVD) were studied in this work: a pure silica glass, xGeO₂-(1-x)SiO₂ binary glasses with various x = GeO₂ content ranging from 1.9 up to 17 mol %. Those samples are optical fiber preforms with standard compositions corresponding to key materials for optical fiber manufacturing e.g. standard optical fiber, photosensitive optical fibers, highly doped fiber for gain tlattening filters or dispersion compensation and multimode fibers. The characterization of the response of these key materials for optical telecommunication is thus of interest since femtosecond laser processing in optical fibers is an increasingly attractive topic of interest as a new approach to fabricating real components (despite the lack of understanding of the mechanisms that we are contributing here).

In order to fix our sample’s fictive temperatures, thermal treatments with temperatures ranging from 950 °C up to 1400 °C were performed during times exceeding their structural relaxation times [22]. The samples were then quenched in water at room temperature. All the samples were then optically polished to reach the same surface quality (typ. < λ/10) for each of them. The fictive temperatures, T_fic, were then verified using the frequency of Si-O-Si asymmetric stretching vibration band located υ ~1120 cm⁻¹, measured by FTIR reflection spectroscopy as described in Ref [22, 23]. Calibration curves were then built, plotting the reflection band at υ ~1120 cm⁻¹ wavenumbers versus fictive temperatures for the samples treated at temperatures between 950 and 1200 °C. These curves allow recalculation (when necessary) of the fictive temperature of any heat treated samples exhibiting the same composition [23]. Densities of the thermal treated samples were estimated with the ‘sink-float method’ based on the comparison of their weights in air and in toluene taking into account the solvent temperature. The relation between these weights and density and temperature of toluene is well known. For each fictive temperature, measurements were performed on a large number of samples (typ. 20) with weights varying from 10 to 1000 mg, resulting in a density error of approximately ± 10⁻³.

The laser radiation, in a Gaussian mode, was produced by a femtosecond laser system (Satsuma, Amplitude Systèmes Ltd.) operating at 1030nm and delivering pulses of 300fs at 100 kHz. Considering that the propagation vector $\overrightarrow{k}$ is along z direction, the beam was focused from 100 to 200μm. The writing configurations defined by writing and polarization directions are shown in Fig. 1. Therefore, when the laser was moving along X and the polarization was lying along x, we define it as “Xx writing” (a parallel configuration). Then moving the sample along X-axis with various pulse energies ranging within a few fractions of μJ/pulse and varying the polarization direction (along x or y-axis). The lines were written in paying attention to their direction (referenced to the laboratory) and orientation (e.g. left to right and right to left). Then, we observed samples in transmission using a polarized optical microscope for detecting neutral axes, determining slow axis orientation and for measuring the linear birefringence by using the Senarmont method. Nanogratings and porous nanoplanes were then analyzed by cleaving the samples across the laser tracks. The cross sections were then imaged at 1kV using a Field-Emission Gun Scanning Electron Microscope (FEG-SEM, ZEISS SUPRA 55 VP).

 

Fig. 1 Densities, ρ, of heat-treated PCVD germanosilicate glasses according to their fictive temperature, T_fic. Black straight lines are linear fitting functions obtained by χ² minimization method.

Download Full Size | PDF

3. Results

The density according to fictive temperature for pure silica glass and xGeO₂-(1-x)SiO₂ glasses (with x = 1.9, 5, 11.8 and 16.9 mol% in GeO₂) are presented in Fig. 1. The errors bars about density are estimated to be equal to 10⁻³ and reported in this figure as well. The density of pure silica samples increases monotonically with increasing T_fic, meaning that the anomalous behavior of silica is observed as expected [22, 24, 25]. The same trend is observed for 1.9 mol% GeO₂ doped silica. In contrast, the opposite trend is highlighted for the germanosilicate glasses that contain more than 5 mol% in GeO₂: the higher the density, the lower the fictive temperature. This indicates that the anomalous trend of silica disappears for a high germanium concentration and so we have a suite of glasses with which to test whether or not we can fabricate porous nanostructures in the normal glassy regime using femtosecond laser processing. Based on results already reported in literature [26, 27], we assume that the local dependence of ρ on T_fic is linear in the studied temperature range. This provides the following linear equations in the set of parameters (T = 25 °C, P = 1 bar), whose straight lines are plotted in Fig. 1. The slope values (volume, V, change attached to T_fic) for the various glasses differ significantly: positive for silica and 1.9 mol% GeO₂ silica but negative in silica glasses containing more than 5 mol% GeO₂. The question is then to understand if and how this behavior impacts the formation of porous nanogratings.

For all samples, we wrote series of straight lines with various femtosecond laser pulse energies ranging from E ~0.05 to E ~1.2 μJ/pulse. For each series, at least 6 lines were produced with a linear polarization oriented either along the y-axis (parallel configuration) or x-axis (perpendicular configuration). Then the retardance (R is proportional to the birefringence, B) of samples was measured using the Sénarmont compensator technique. Figure 2 displays the amplitude of the retardance according to the pulse energy for all the four glasses. As can be seen, the retardance increases and then saturates when the pulse energy increases up to 1 μJ/pulse. At higher energy, the retardance has been observed to strongly decrease [24]. As shown in Fig. 2, the maximum retardance amplitude depends on the amount of germanium oxide within the silicate glass. In our experimental conditions (500 kHz repetition rate), the higher the GeO₂ content, the lower the retardance at saturation. To explain the origin of the retardance dependence on the chemical composition, we have imaged the cross section of the cleaved samples using FEG-SEM in order to analyze the nanogratings and the porosity-filling factor.

 

Fig. 2 Retardance, R, according the laser pulse energy, E, in various germanosilicate glass samples. The laser parameters were 1030 nm, 300 fs, 0.6 NA, 500 kHz, 500 μm/s, Xx configuration. The scanning direction was horizontal and oriented at 36° off the pulse front tilt. The linear polarisation was parallel to the writing direction.

Download Full Size | PDF

In Fig. 3, where the femtosecond laser polarisation is perpendicular to the scanning direction, the observed electronic contrast in the nanogratings arises from atomic density contrast. Figures 3 and 4 are taken at 0.5 μJ/pulse and after 10³ pulses/µm - under these exposure conditions, the nanogratings appears quasi-periodic and they are oriented along the laser propagation axis. The width of the interaction volume is on the order of 3 to 5 μm. The average spacing of the nanoplanes is found to be around (300 – 350) nm. Increasing the GeO₂ content, we observe a slight decrease of the nanoplanes average spacing, or pitch, from Λ ~350 nm for SiO₂ down to Λ ~310 nm for 11.8 mol% GeO₂.

 

Fig. 3 Secondary electrons images of the sample’s cross-section for Xy writing (orthogonal configuration). The laser parameters were: E = 0.5 μJ/pulse, 1030 nm, 300 fs, rep. rate = 500 kHz, v = 500 μm/s; i.e. 10³ pulses/µm. A focusing aspheric lens of 0.6 NA was used.

Download Full Size | PDF

 

Fig. 4 Secondary electrons images of laser track cross-section for Xx writing (parallel configuration). The laser parameters were: E = 0.5 μJ/pulse, 1030 nm, 300 fs, rep. rate = 500 kHz, v = 500 μm/s i.e. 10³ pulses/µm and using a 0.6 NA aspheric lens.

Download Full Size | PDF

To investigate the internal nanostructure of the nanoplanes, we flipped the laser polarization 90° towards Xx configuration (i.e. linear polarization is oriented parallel to the writing direction). As shown in Fig. 4, nanoporosity is not homogeneously distributed since we observed “white” layer arising from surface topography variations that corresponds to material between the nanolayers. It seems that nanogratings could be slightly tilted allowing visualization of both layers and the matter in-between. This will be further studied using atomic force microscopy (AFM). In all our samples, we observed formation of nanopores at the same time as the appearance of nanogratings. After processing SEM images, a mean mesopore size of φ ~50 nm and a filling-factor around ff ~38% within porous nanoplanes is measured in SiO₂. Increasing the GeO₂ content, we observe that the porosity-filling factor has decreased from ff ~38% for SiO₂ down to ff ~22% for 11.8 mol% GeO₂. At the same time, we also observed a significant decrease of the nanopores average size from φ ~ 55nm down to φ ~ 25nm.

4. Discussion

The main question treated here is related to the creation of the nanopores themselves in the investigated glass samples. In the literature, oxide decomposition may occur by various mechanisms including fragmentation, nucleation and growth, or “spinodal” decomposition. The dominant mechanism will depend on where the pathway of the fluid relaxation and where it intersects the liquid-vapor dome but also on the relative timescales for each process, where the fastest mechanism wins. In these cases, quasi-equilibrium chemistry required that the temperature should exceed the vaporization temperature (3177 K at PO₂ = 0.21 [25]) for at least a few seconds [26] for a thermal dissociation of SiO₂. However, as we have reported earlier, neither the time nor the temperature are sufficient for thermal dissociation of the oxide. In addition, using such temperature-time profiles, the accumulated thermal diffusion length of network oxygen within the experiments is computed to be less than one nanometer as shown in Ref [16]. We can conclude that no average relaxation, or thermal equilibration, processes associated with longer timescale heating and cooling, are expected to be significantly involved in the initial formation of the nanopores.

Based on the results reported in the literature, nanopore formation may happen because the irradiated glass volume is constrained and under intense stress (that accumulates from pulse to pulse) due to the glass expansion immediately after ionization where a shock front is generated. In the first instance, one would expect in the anomalous glassy regime that this process is enhanced substantially because silica in the liquid state is more dense than the glass state [21]. Indeed, due to the smaller specific volume of the irradiated silica when compared to the pristine glass, a singular feature of SiO₂ glass is quickly realized i.e. an enormous tensile stress created during the relaxation process originating from the surrounding glass resisting the change in volume. This process is described in Fig. 5. Supporting this, Champion et al. deduced a generated stress as large as 18 GPa in pure silica [27], and it is conceivable that there may be a much higher dynamic stress involved. Due to the involved geometry and scale of nanogratings, it is immediately obvious under these conditions that there must be voids, nanocracks or nanopores formed for the quenching to occur, particularly when there is a reduced glass resulting in a strong amount of oxygen deficient centers that is estimated to be around a few 10²⁰/cm³ compared to 2.2x10²² SiO₂/cm³ [28]. This should lead to a porous structure on condensation, further encouraged by the reduction of the average numbers of O in the first Si or Ge coordination shell. In the case of silica, this prevents tetrahedral structure from reforming; it is even possible with sufficiently released oxygen that solid silicon aggregation may occur. Supporting this proposal, evidence for silicon nanocluster formation was recently reported [18]. The exact nature of this under-stoichiometric SiO_2-x structure and the surrounding structure needs further clarification. For example, we can expect an increase in the average coordination number of silicon like stishovite molecules or stishovite nanocrystals formation in-between the nanoplanes. Alternatively, it may be plausible that under such extreme conditions, after the number of O neighbors is reduced, during cooling, more complex structures involving silicon aggregation and novel amorphous Si; O coordination of four and, speculatively, even higher if frozen in before separation can occur. Furthermore, despite the amorphous nature of the condensation process, there is likely to be preferential structural alignment given the orientation of the plasmon-photon interference that gives rise to the oriented nanoplanes and therefore the likely presence of an axial stress variation within the volume.

 

Fig. 5 Simplify scheme of differential volume changes in the high energy regions of the focal volume. The focal volume may expand rapidly under initial excitation. The preferred condensation at the troughs and peaks of the plasmon interference fringes, along with the laser direction, give rise to nanoplanar structure and will affect the stress gradients, dynamic expansion and structural orientation around the various axes of the processed volume.

Download Full Size | PDF

In this model, in the first instance, the porous oxide formation appears like a natural SiO₂ zeolite polymorph in response to this tensile stress mechanism. In the simplest consideration, in normal glasses the stress should be compressive so it is unconceivable this is sufficient to create a shredding of the silica glass; rather the converse. Therefore in this framework, porous nanoplanes should not be seen in normal glasses because the specific volume of the liquid state is higher and expands with the “explosive” process involved with multiphoton ionization. Without a decrease in glass Si coordination, one could not expect nanopore formation to occur except perhaps in the extreme case where volume expansion arising from very rapid ionization is faster than the normal volume expansion of the glass. The evidence to date is that nanoporous oxide has been reported mainly in anomalous glasses like SiO₂ and slightly doped silica (typically 1.9 mol% in GeO₂ or 7TiO₂-93SiO₂) but not in most normal glasses (BK7, SnO₂-SiO₂, or CaO-Na₂O-SiO₂ glasses) until very recently in GeO₂ [29, 30] and borofloat [20].

However, we have shown in this work that nanogratings and nanopores can indeed be formed in both anomalous and normal glassy regimes: i.e. whatever the sign of the density changes versus T_fic may be. This is also in agreement with recents results reported in GeO₂ glass in refs [29, 30]. A preliminary conclusion might be that Coulombian nano-explosion could create the pores themselves if indeed the rate of volume expansion is faster than the liquid volume expansion of normal glass in order to create anomalous like conditions (as schematically summarized in Fig. 5). A more likely alternative (or combination thereof) is the reduced density of the final solid arising from a significant coordination reduction in structure when so much oxygen is removed. The reduction of the average numbers of O in the first Si coordination shell in silica glasses is clearly supported by the experimental observation of silicon, or silicon rich, formation within silica glass [18] as well as the generation of free molecular oxygen within the pores and dissolved within the glass, well after condensation [16]. To satisfy the lower Si coordination of the final solid, a porous structure will form even for the normal glassy regime (as for Ge-doped silica or GeO₂) because processes begin to resemble the dynamics expected from the anomalous regime. Notice that in our irradiation conditions at 500kHz, the filling factor and the pore size decrease with increasing GeO₂ content and in GeO₂. This likely reflects the total number of pores, which are erased due to viscoelastic flow of glass that is obtained for temperatures close to the softening point.

In summary, nanopores formation is related to breaking of Si-O bonds initiated in a few 100’s femtoseconds. During fast relaxation of under-stoichiometric silica, O atoms recombines to produce O₂ both within the pores and dissolved in the silica network [16]. This prevents any possible recombination of the glassy network that freezes in a mesoporous oxyde, ODC defects [28] and/or amorphous Si and even crystalline SiO_2-x aggregates, as shown recently by Vilar et al. [18]. Under such extreme conditions it is very likely that for the mixed normal glasses, phase separation will occur – if pure Si can form, then some Ge rich analogues might be anticipated as well. All these factors suggest the conditions resemble more the anomalous regime rather than normal one.

5. Conclusion

To sum-up, the evidence indicates that porous nanogratings, which were not reported in glasses other than SiO₂ and slightly doped silica, can also occur in normal glass samples like 17 mol% GeO₂ silica glass. The proposed theory using anomalous behaviour alone needs to consider the reduction of the average numbers of O in the first Si coordination shell, including solid amorphous or crystalline silicon and germanium formation under some conditions, as well as likely phase separation of components in mixed glass systems. All these contributing factors create anomalous-like conditions that lead to pore formation, a feature that reflects the creation of extraordinary environments possible using high intensity femtosecond laser. It will be interesting to see if such regimes can be found in nature where analogous extreme environments may exist and whether femtosecond lasers, and as they come on line attosecond lasers, may even offer a tool for creating and studying such environments more broadly in the laboratory. Consistent with our work here, we expect that porous nanogratings can in principle be formed in most glassy systems including various borosilicate and alumino-borosilicate glasses such as BK7, Borofloat or Gorilla glasses provide that heat accumulation can be avoided and very fast negative pressures and subsequent condensation occur. Any reduction in the average number of O first neighbors will further aid these processes.

The nanopores average size and porosity filling factor decrease when increasing the GeO₂ content up to 17 mol% resulting in a lower birefringence, consistent with a complex mixture that may be undergoing phase separation in part because the constituents have different softening points. By playing with the glass chemical composition, it is possible to effectively master glass porosity and related anisotropic optical properties but also to reduce the scattering light of the birefringent optical components. The use of these multiphoton-triggered processes allow true glass milling on the nanoscale – it is possible to envisage the techniques and aesthetics of glass artists being applied on unprecedented scales.

From a practical perspective, such control allows the fabrication of nanoporous layers that can be patterned in a micro-array, leading to many novel applications not only for birefringent devices (such as waveguides or bulk retarders [31], 3D space variant waveplates, polarization converters, long lifetime 3D optical memories) but we can also expect composites, encapsulants, catalysts and molecular sieves. Assuming interconnected nanopores as it is the case for zeolites, such mesoporous glass can also be used for filtration and separation of compounds. By mastering the nanopores size, the filling factor, the resulting glass network opens the door towards size-selective permeability and can be integrated with high precision into waveguides and patterned components, a potentially more robust and efficient top-down alternative to methods based on bottom-up self-assembly.