1. Introduction

One of the most attractive features of femtosecond (fs) lasers is their ability to produce internal structural modifications far below the surface in transparent materials [1,2]. This space-selectivity is due to the high intensities these ultrashort pulsed lasers reach when tightly focused, resulting in nonlinear multiphoton absorption at the focal point [1,3]. At high repetition rates of 200 kHz or more [4], this absorption can lead to heat accumulation and a variety of internal structural and optical changes (i.e. ‘heat modification’) including local melting [4], elemental migration [5], refractive index change [6], and crystallization of functional phases with high second-order nonlinearity [1,3]. Such crystals are important components of modern optics because their nonlinear properties are necessary for processes like wavelength conversion, signal modulation, and optical switching [3]. Local crystallization of glass by fs laser irradiation thus offers the possibility of creating compact composites in which optically active crystalline architectures are integrated directly into a glass matrix in three dimensions.

Recently we demonstrated proof of concept for creating 3D single crystal architectures in glass [7,8]. This initial work was performed on LaBGeO₅ glass, a congruent model system in which the glass and ferroelectric crystal share the same composition. Heat modifications could be produced at all focal depths investigated and under a variety of irradiation conditions, confirming the space-selectivity of local heating. When crystallization did occur, ferroelectric LaBGeO₅ was often preceded by an intermediate phase [7] (likely La₂Ge₂O₇ [9]). Once a LaBGeO₅ crystal was established, it would expand rapidly and could be grown into single-crystal lines at all focal depths by moving the sample stage. However, despite the ease with which continuous structures could be grown from a seed crystal, the initiation of new seed crystals in pristine glass was surprisingly sensitive to focal depth [7]. Crystallization occurred within seconds when depth was optimized, but was not observed at all after more than thirty minutes of irradiation when depth was varied. This depth-dependence of crystallization seems to defy a fundamental expectation of fs laser processing. Therefore, in the present work we have examined the various stages of heat modification and crystal initiation at different depths.

2. Experiment

Stoichiometric LaBGeO₅ glass samples were prepared by conventional melt-quenching. Reagent-grade batch materials of La₂O₃, H₃BO₃, and GeO₂ were weighed and mixed for 12 hours, with additional H₃BO₃ included to compensate for 1.9 wt. % B₂O₃ volatilization [10]. The batch was melted for 30 minutes at 1250 °C, poured and pressed between stainless steel plates preheated to 500 °C, and annealed at 640 °C for two hours to relieve thermal stress. The glass was then cut into small rectangular prisms of approximately 1.2 x 1.2 x 5 mm and polished with diamond suspensions and CeO₂ slurry on all four large faces. The samples could then be irradiated from above while viewing the resulting modifications from the side.

Irradiations were performed using a regeneratively amplified Ti:sapphire laser (Coherent Inc.) with 800 nm wavelength, 250 kHz repetition rate, 60 fs pulse width, and 2.5 μJ pulse energy. The beam was focused through an optical microscope equipped with a 50x, 0.55 N.A. objective lens and digital camera to observe and record the focal point during irradiation. A second camera was used for in-situ viewing of modifications from the side. Irradiations were performed with samples placed inside a halogen lamp heating stage to allow irradiation at elevated temperatures, which required focusing the beam through a sapphire window.

The laser focal point was positioned at particular depths by first obtaining a sharp image of the sample surface, then raising the sample stage by a fixed distance, effectively moving the focal point inside the glass. The final position of the focal point was approximately twice as deep as the stage displacement due to refraction and optical aberrations. References to specific values of focal depth that follow will refer to this operationally-defined stage displacement distance. In-situ videos of irradiation and crystallization were collected, and modifications were examined with optical microscopy and Raman spectroscopy. The micrographs that follow will be shown in profile view, beam incident from above, at 100 μm depth unless otherwise noted.

3. Results and discussion

Heat-induced refractive index modifications were seen at all depths, but with significant differences in their size and shape (Fig. 1). A focal depth of 100 μm or less was necessary for fast and consistent crystallization (i.e. < 1 minute of irradiation) when focusing through the sapphire window of the heating stage. Interestingly, growth rates of an established crystal were fairly independent of focal depth, suggesting that only the nucleation process is greatly affected.

 

Fig. 1 Optical micrographs of heat modifications produced by 10 seconds of irradiation at increasing focal depths from 100 to 400 μm, exhibiting structural defects (arrows).

Download Full Size | PDF

Figure 1 also reveals clusters of defects, typically at the base of the modification and disappearing as focal depth increases. Whenever crystallization occurred, as shown in Fig. 2, the crystal initiated from the defects at the bottom of the heat modification and expanded upward. The radial extent of crystal growth appears bounded by an additional interface inside the heat modification. The distribution of defects also seems to follow the contour of this inner boundary, but inspected at higher magnification (Fig. 3 inset), they are found to lie just outside it. Since crystallization originated consistently from these defect sites, it was considered that they could be crystallites, but they remain dark when imaged under crossed polarizers and appear to be voids instead (Fig. 3). In contrast, after crystallization occurs, a strong birefringence is seen.

 

Fig. 2 Optical micrographs of crystallization time progression. Defects (solid black arrows) appear before the crystal (solid white arrows), an intermediate phase [7] which grows upward inside the region enclosed by an inner boundary (white dashed arrows).

Download Full Size | PDF

 

Fig. 3 Polarized optical micrographs before (left) and after crystallization (right). In the highlighted region, the defects (solid black arrows) are seen to occur just outside the inner boundary (dashed white arrows). The bright circles (dashed black arrow) are due to nearby cracks. After crystallization, strong birefringence is seen (solid white arrow).

Download Full Size | PDF

We performed Raman spectroscopy to clarify the structural changes which precede crystallization. Figure 4 shows normalized spectra obtained from various positions around the heat modification. Some fluctuations occur in relative band intensities, but all the spectra show typical features of LaBGeO₅ glass: a large band near 800 cm⁻¹ attributed to BO₄ and GeO₄ stretching modes, a band near 560 cm⁻¹ attributed to GeO₆ units and BO₄ bending modes, and a band near 300 cm⁻¹ attributed to GeO₄ and BO₄ bending modes [9,11,12]. In some regions optical scattering caused global variations in collected intensity, although this did not affect spectral features. The top inset of Fig. 4 effectively maps this optical scattering, which notably occurred not only at the modification base, but at the inner boundary indicated in Fig. 2 as well.

 

Fig. 4 Raman spectra collected from 7 different regions of the heat modification and scaled to normalize their intensity at high frequency. Spectral windows from 100–440 and 440–660 cm⁻¹ indicate ranges used for mapping. Inset (top): Map of average intensity in Window 2 before normalizing spectra. Darker pixels indicate lower average intensity, numbers indicate where spectra were collected. Inset (bottom): Map of position of the centroid of the area under the curve in Window 1. Darker pixels indicate stronger low frequency peaks.

Download Full Size | PDF

Despite all spectra appearing predominantly glassy, Spectra 1, 2, and 3 show two sharp peaks at very low frequency which suggest the presence of crystals. Surprisingly, these spectra do not come from the defects at the base of the modification, but from the inner boundary. The absence of higher frequency peaks is unusual, although an additional peak appears to emerge near 260 cm⁻¹ in Spectrum 1. If crystallites are responsible, they are likely on the order of nanometers in size based on the suppression of higher frequency peaks [13]. Finely dispersed crystallites could also explain the weak optical scattering seen in this region. However, these Raman peaks are not easily identified with crystalline LaBGeO₅ modes. The 119 cm⁻¹ peak could perhaps be attributed to a GeO₄ wagging mode, and the faint peak at 260 cm⁻¹ to a GeO₄ or BO₄ bending mode, but the 134 cm⁻¹ peak has no clear assignment [11,12].

Further experiments are needed to confirm the source of the features below 200 cm⁻¹ in the Raman spectra, but we note from the centroid map in Fig. 4 that these features do not overlap with the cluster of voids. Whatever structural change is responsible, it fades away toward the bottom of the heat modification and is effectively absent around the void cluster. Thus, while these Raman results raise new questions about the nature of the inner boundary, they may be tangential to the issue of crystal nucleation and focal depth. Since the large crystals seen in Figs. 2 and 3 emerge consistently from the void region, these voids seem to influence crystallization independently of whatever mechanism is responsible for the unusual Raman peaks.

The growth rate of an established LaBGeO₅ crystal is very high under fs laser irradiation regardless of depth (∼45 μm/s [7]), suggesting that nucleation is the rate-limiting factor. Bubbles have often been found to act as preferential crystallization sites in glasses [14], so the presence of voids in the heat modification could accelerate crystal nucleation rates and shorten the time required for laser crystallization. However, as seen in Fig. 1, the range of depths in which voids occur (up to 300 μm) well exceeds the range conducive to crystallization (up to 100 μm). Thus, the presence of voids may be necessary to quickly crystallize this glass, but it is not always sufficient. The local conditions at the surface of the voids must also be considered.

Femtosecond laser heating produces a steep temperature gradient with peak temperatures that can reach thousands of °C at the center [15]. This is a much different situation than furnace heating, in that a wide range of temperatures is sustained simultaneously across the gradient. This range should easily contain the optimum nucleation and growth temperatures, even if optical aberrations cause the size and shape of the gradient to vary. The depth-dependent nucleation behavior seen in LaBGeO₅ glass makes little sense in this context if homogeneous bulk nucleation dominates. In contrast, a heterogeneous nucleation model substantially restricts the potential nucleation sites by its reliance on structural defects. In this case, nucleation temperatures must coincide with the location of voids or other defects in order for nucleation to occur. If the correlation between void distribution and temperature gradient is weak, variations in the size and shape of the temperature gradient with focal depth could cause the local temperature at the void surface and the corresponding nucleation rate to change.

This correlation can be investigated by raising the sample temperature during irradiation, effectively shifting the laser-induced gradient to higher temperatures. The effects of this heating are seen in Fig. 5 as an increase in the size of the heat modified zone, as the outer boundary has a strong temperature correlation. In contrast, the void distribution does not expand to follow a particular isotherm as temperature is raised, but gradually disappears instead. Thus, although the size and number density of the voids is affected by temperature changes, their position is largely independent. This means heterogeneous nucleation rates could vary substantially as changes in focal depth cause the temperature at the void surfaces to vary. This plausible mechanism explains the influence of focal depth on fs laser crystallization of LaBGeO₅.

 

Fig. 5 Optical micrographs of heat modifications produced by 10 seconds of irradiation at 100 μm depth with increasing sample temperatures.

Download Full Size | PDF

As seen in Fig. 2, the voids themselves appear to originate at the very tip of the heat modification. Beam damage initially persists for several microns beyond the edge of heat modification, creating structural defects which are no longer erased by local melting. Heat accumulation gradually extends downward, but rather than simply relaxing or melting these features away, the molten glass apparently acts as a wedge which pushes them outward toward lower temperatures in an extrusion-like process. We expect that their final position is determined by local stress fields and viscosity considerations, but this remains to be confirmed.

4. Conclusion

A model of heterogeneous nucleation at void surfaces has been proposed to explain the unexpected influence of focal depth on single crystal formation in LaBGeO₅ glass by fs laser. We have shown that voids form at the base of the heat modification, and that crystallization subsequently initiates from this region. Variations in the size and shape of heat modifications at different focal depths suggest that corresponding changes in the temperature distribution occur, particularly at the surface of the voids, which consequently impact the heterogeneous nucleation rate. The expected 3D space-selectivity of fs laser crystallization should still be possible; whether by choosing compositions that crystallize readily by homogeneous bulk nucleation, by seeding the glass with nucleating agents, or by more carefully controlling the beam profile to compensate for optical aberrations at different depths.