1. Introduction

The mid-infrared spectral region is of great interest for a variety of applications such as spectroscopy, medicine and astronomy [1–3 ]. Many of these applications benefit from the intrinsic advantages of integrated optics like miniaturization, robustness, mass producibility and the absence of any alignment requirements once the light is injected into the integrated optical circuit. A promising platform for mid-infrared integrated optical circuits are glasses based on the chalcogenide elements S, Se and Te [4]. This is not only due to their excellent optical transparency in the mid-infrared but also the ability to tailor their properties such as refractive index, transmission and optical nonlinearity through variation of the glass composition [5, 6].

Ultrafast laser inscription [7, 8] has established itself as a viable alternative to photolithography for the fabrication of integrated optical circuits, in particular for rapid prototyping and the inscription of 3-dimensional integrated optical circuits [9–11 ]. However, ultrafast laser inscription of low loss waveguides into chalcogenide glasses has proven to be challenging. While in silicate glasses propagation losses as low as 0.1 dB/cm are readily achievable [12–14 ], the lowest loss reported to date in chalcogenide glass is 0.65 dB/cm [15]. Chalcogenide glasses show a multitude of different photoresponses when irradiated with light. Effects such as photocrystallization and photoamorphization, photopolymerization and photodecomposition as well as photodarkening and photobleaching have been observed [16]. The complexity of the underlying processes means that for instance in As₂S₃ glass, small changes in the inscription parameters can flip the sign of refractive index change from positive to negative [17]. Thus in As₂S₃ glass, ultrafast laser inscribed waveguides based on positive [18, 19] as well negative index modifications [20] have been demonstrated. However, with the addition of Ge to the glass matrix the parameter range for positive index change can be significantly extended [17].

Ultrafast laser inscription provides a multitude of parameters, such as pulse energy, pulse duration, repetition rate, laser wavelength, translation speed and focusing geometry that can be individually adjusted in order to optimize the waveguide inscription. This paper provides a roadmap to the inscription of low loss single-mode waveguides in gallium lanthanum sulfide (GLS) chalcogenide glass by contrasting different waveguide inscription regimes. Thermal as well as athermal fabrication is explored. Three different techniques are employed to inscribe circular waveguides, namely cumulative heating, slit beam shaping as well as the multiscan method. The inscribed waveguides are characterized in terms of physical size, mode-field diameter, index contrast and propagation losses at 1550 nm. Furthermore, guiding properties in the mid-infrared at 3.39 μm wavelength are evaluated.

GLS was chosen because it is a non-toxic, commercially available chalcogenide glass with a wide optical transparency range (0.55–10 μm) and high nonlinearity as well as high photosensitivity [21,22]. This makes the glass an attractive substrate for ultrafast laser inscription, which has led to the demonstration of spectral broadening [23], supercontinuum generation [24] and nonlinear optical switching [25, 26] in laser written GLS circuits.

2. Experimental methods

The GLS glass samples were purchased from ChG Southampton Ltd. in sizes of 10 × 10 × 2 mm and 10 × 20 × 2 mm with all sides polished. For inscribing waveguides in the thermal regime, i.e. cumulative heating [27], as well as for the inscription of multiscan waveguides in the athermal regime, a 5.1 MHz Ti:sapphire chirped pulse femtosecond oscillator (Femtosource XL500, Femtolasers GmbH) was used. The laser emits up to 550 nJ pulses at 800 nm wavelength with a pulse duration of <50 fs. The circularly polarized pulses were directed by an optical setup to an Olympus Plan N 100× oil immersion microscope objective (NA ≈ 1.25, NA_eff ≈ 0.66). Oil immersion reduces the refractive index mismatch and thus mitigates spherical aberrations compared to air objectives [28]. The samples were placed on a set of computer controlled Aerotech air-bearing stages. In the thermal regime optical waveguides were fabricated in a single pass at a writing depth of 170 μm. The samples were translated at feedrates of 100 to 3000 mm/min and the pulse energy was varied between 1 and 9 nJ.

For the fabrication of the multiscan waveguides [12] in the athermal regime, an external electro-optic pulse picker (BME Bergmann Messtechnik KG) was used to reduce the repetition rate to 255 kHz, 510 kHz, 728 kHz and 1020 kHz, respectively. Four different feedrates where tested for each repetition rate. The feedrates were scaled in order to keep the number of pulses per unit length approximately equal for the different repetition rates. At 255 kHz the sample was translated at 60, 120, 240 and 375 mm/min, at 510 kHz – 120, 250, 500 and 750 mm/min, at 728 kHz – 180, 400, 800 and 1125 mm/min and at 1020 kHz repetition rate, feedrates of 240, 500, 1000 and 1500 mm/min were chosen. For all repetition rates and feedrates, pulse energies of 10 to 90 nJ were tested. The laser focus was transversely moved in 400 nm increments to create 6 μm wide waveguides. Therefore a total of 15 scans was necessary to write a single waveguide. The translation direction was alternated between the scans in order to increase the fabrication speed. In order to limit non-reciprocal writing effects due to pulse front tilt [29], the writing direction was chosen to be orthogonal to the axis of the pulse compressor.

For the fabrication of single-scan waveguides in the athermal regime, a Spectra Physics Hurricane Ti:sapphire laser with a repetition rate of 1 kHz and pulse duration ≈120 fs at 800 nm wavelength was used. Three different Olympus air objectives (LUCPlanFL N 20×, LUCPlanFL N 40×, LMPlanFL N 100×) were tested in order to vary the effective numerical aperture between 0.28, 0.55 and 0.8, respectively. The 20× and 40× objective are corrected for spherical aberrations at 170 μm focal depth when focusing into glass with a refractive index of n_D=1.518. Before entering the focusing objective, the beam was astigmatically shaped by slits with widths of 320 μm, 480 μm and 520 μm according to the effective numerical aperture of the objective in order to create waveguides of circular cross-section [30]. The slit reduces the effective numerical aperture in the direction perpendicular to the long axis of the slit. Pulse durations of 112 fs, 400 fs and 1500 fs and pulse energies from 50 to 500 nJ (measured after the slit) provided a wide variation of different peak powers at the focal spot. All waveguides were inscribed at a depth of 170 μm, a translation speed of 1.5 mm/min, and with circularly polarized light. In addition to single passes, 4 and 8 overpasses were investigated. All inscription parameters are summarized in Table 1.

Table 1. Summary of the tested inscription parameters.

View Table | View all tables in this article

The mode-field diameters (MFD) were measured by injecting light into the waveguides using a single-mode optical fiber (Corning SMF-28e) connected to a 1550 nm laser source. At the output of the sample, an Olympus MSPlan air objective with a 50× magnification and a NA of 0.8 was used to image the guided mode onto a Spiricon SP-1550M camera. The Spiricon camera software LBA-PC was utilized in order to process the mode-field profiles and calculate the MFD according to the 4σ-method. All quoted MFD values are the average between the vertical and horizontal dimension. In order to determine the MFDs in the mid-infrared, a dual-line HeNe infrared laser (REO R-40138) emitting at 3390 nm was employed. The light was launched using a free-space setup similar to [31]. The waveguide output was imaged by a 20 mm CaF₂ lens onto an indium antimonide (InSb) focal plane array (FLIR SC7000) with a spectral response covering 1.5 to 5.1 μm. A MATLAB script was used to post-process the data and calculate the 1/e² MFD based on a Gaussian fit.

The physical dimensions of the waveguides were determined using an optical microscope. The refractive index contrast was estimated using the Marcuse formula based on the measured physical size and MFD under the approximation of a step-index refractive index profile [32]. The retrieved index contrast was averaged between the horizontal and vertical dimension. The Marcuse formula is accurate for single-mode waveguides with a normalized frequency (V-number) above 1, i.e. a well confined guided mode.

The Fabry-Perot method was used as an accurate way for determining the propagation losses at 1550 nm since it is independent of coupling losses [33]. The accuracy of the retrieved loss values benefited from the high refractive index contrast between the air and GLS glass interface resulting in a high Fresnel reflectivity (17%) and thus strong spectral fringes. The loss measurements were performed by launching into the waveguides free-space using two aspheric lenses (f = 11 mm) to collimate and focus the light from a single-mode fiber. Light from the waveguide output was collected using an identical optical setup and coupled back into a single-mode fiber. The use of free-space optics avoids parasitic Fabry-Perot cavities that otherwise would occur between the fiber tip and sample even when using AR-coated fibers. These parasitic Fabry-Perot cavities effectively introduce a wavelength dependent reflectivity of the end-faces instead of the static Fresnel reflections. This results in a non-uniform fringe pattern, i.e. the fringe contrast changes as a function of wavelength. The small gap between sample and fiber, usually a few tens of microns, results in frequencies of the non-uniform fringe pattern of several nanometers to several tens of nanometers and thus are difficult to resolve if Fabry-Perot loss measurements are performed by scanning the laser only over a few hundred picometers. Immersion oil with a refractive index of 1.5 can be used to fill the gap and avoid the wavelength dependent fringe contrast. However, this reduces the Fresnel reflectivity of the sample’s end-faces to ≈5% and thus increases the measurement error [33]. In order to obtain the most accurate loss measurements, the fringes were acquired across a broad wavelength band using a JDS Uniphase swept wavelength system (SWS15100). The fully automated system sweeps across a wavelength range of 1520 to 1570 nm with 0.003 nm resolution. Thus the 60 pm fringes for a ≈1 cm sample are well resolved by the system with 20 data points per period. The propagation losses are determined based on the minima and maxima of the fringes and averaged across the scanned wavelength band.

3. Thermal fabrication - cumulative heating

The typical cross-section of waveguides inscribed in the thermal regime at 5.1 MHz repetition rate are shown in Fig. 1. With the laser incident from the top, the shapes resemble an upside down tear drop. The focal spot of the laser was located at the bottom of the modifications. This is similar to the results obtained when irradiating GLS with 250 kHz repetition femtosecond pulses [22]. The highly elongated structures are a result of the elongated focal spot and thus elongated heat source which is caused by spherical aberrations [28] and aberrations induced by the thermal lens of the laser heated material [34]. The waveguides were observed to increase in size away from the focal spot towards the incident laser beam. At large pulse energies, significantly above the critical power for self-focusing and beyond the inscription regime for single-mode guiding at 1550 nm, short filaments at the bottom of the modifications became apparent.

 

Fig. 1 Cross-section of waveguides inscribed in the thermal regime at 5.1 MHz repetition rate. Writing parameters: NA_eff = 0.66, 9 nJ pulse energy and translation speeds of 500, 250 and 100 mm/min (left to right). The inscription laser was incident from the top.

Download Full Size | PDF

The physical sizes along the vertical and horizontal axis are plotted in Fig. 2(a) and 2(b), respectively. Pulse energies from 1–3 nJ were below the modification threshold and did not result in a material index change. The structures between 4 and 9 nJ exhibit an elliptical aspect ratio of 0.34–0.43. The smallest structure, observed at the onset of cumulative heating, was 5.6 × 2 μm and the largest waveguide within the parameter range was 18.8 × 7 μm. In general, the size of the waveguides increased with increasing pulse energy and decreasing translation speed. It should be noted that it was possible to create structures with sizes in excess of 50 μm in the vertical and 20 μm in the horizontal direction when using larger pulse energies. This enables tailoring of the inscribed structures to be single-mode at long wavelengths by simply increasing the pulse energy. Because of the large ellipticity of the modifications, inscribing multiple structures next to each other can be an effective means of obtaining circular symmetric guides modes at long wavelengths. However, fracturing of the glass occurred at pulse energies larger than 70 nJ due to the large physical size of the inscribed modifications. This effect can be mitigated by inscribing the waveguides deeper into the sample.

 

Fig. 2 (a) Vertical and (b) horizontal physical sizes of waveguides inscribed with a NA_eff = 0.66, pulse energies of 4 to 9 nJ in 1 nJ steps, 9 different translation speeds between 100 and 3000 mm/min at a writing depth of 170 μm.

Download Full Size | PDF

A large number of the waveguides were found to guide single-mode at 1550 nm wavelength. The smallest measured mode-field with an average diameter of 9.3 ± 0.8 μm (MFD_h/MFD_v = 0.87) was found for a waveguide inscribed with 6 nJ pulse energy and 250 mm/min translation speed. With its physical size of 11.6 × 4.8 μm the index contrast was estimated to be Δn = 0.005 ± 0.001 (0.21%). This refractive index contrast agrees with the reported value of Δn = 0.0045 by Hughes et al. for waveguides inscribed in the same regime [22]. Figure 3(a) provides a summary of the MFDs as a function of pulse energy and translation speed. The white area in the top left corner of the surface plots indicates structures that were not guiding at 1550 nm. This is due to a weak index change of the material and a small physical size. The white area in the bottom right corner of the surface plot indicates multimode guiding at 1550 nm wavelength. Multimodedness was determined by launching into the waveguides off-axis and observing a change in waveguide mode-shape. Despite the asymmetrical aspect ratios of the inscribed structures, circular modes with ellipticities (MFD_h/MFD_v) between 0.83 and 1.15 were observed.

 

Fig. 3 MFDs (a) and propagation losses (b) versus pulse energy and translation speed for waveguides inscribed in the cumulative heating regime using 5.1 MHz repetition rate pulses.

Download Full Size | PDF

The waveguide with the lowest propagation loss of 1.82±0.07 dB/cm was inscribed with 6 nJ pulse energy and 100 mm/min translation speed, as illustrated in Fig. 3(b). In the same thermal fabrication regime using 0.36 μJ pulses at 250 kHz repetition rate, similar propagation losses of 1.47 dB/cm at 1550 nm have been reported [22], however, at a significantly slower feedrate of 3 mm/min. The highest propagation losses were observed for weakly guiding waveguides inscribed at low pulse energies of 4 and 5 nJ. The propagation losses decreased with decreasing MFD, i.e. increasing index contrast and physical size. Just before the waveguides become multimoded, the propagation losses were almost independent of the pulse energy and translation speed.

Thermal annealing was pursued to create high index contrast waveguides with low propagation and bend losses similar to those observed in boro-aluminosilicate glass [35]. Thus a sample containing large multimode waveguides inscribed with pulse energies ranging from 10 to 60 nJ was thermally annealed by heating it to 580°C within 8 h, followed by slowly cooling it to 430°C over a period of 16 h and then naturally cooling it back down to room temperature by turning off the furnace. This resulted in annealing out the majority of the waveguides in the sample except for waveguides inscribed at low pulse energies of 10 and 20 nJ translated at slow feedrates of 100 and 250 mm/min. Even though the thermal annealing resulted in a reduction of physical size of the remaining waveguides, they were still multimode at 1550 nm. The fact that the majority of the waveguides disappeared indicates that the origin of refractive index change in the cumulative heating regime in GLS is not due to ion migration like observed in silicate [36] and phosphate glasses [37]. Indeed, Hughes et al. did not observe any compositional changes within the 1% detection limit of electron dispersive x-ray measurements when irradiating GLS with 250 kHz repetition rate femtosecond pulses [22].

4. Athermal fabrication

4.1. Low repetition rate using slit beam shaping

Unlike in the cumulative heating regime, lower NA focusing and either slit beam shaping [30] or cylindrical focusing [38] is necessary in the athermal regime in order to inscribe circular symmetric waveguides of appropriate size for single-mode guiding at 1550 nm. Microscope objectives with different numerical apertures were tested for their applicability to inscribe low loss waveguides using 1 kHz repetition rate pulses as well as different pulse durations and pulse energies of 50 to 500 nJ in 50 nJ steps. The structures written with a high NA_eff = 0.8 objective (Olympus LMPlanFL N 100×) and a 520 μm slit showed strong self-focusing with filaments in excess of 100 μm length at the highest pulse energy when using 112 fs long pulses. Similarly, the structures inscribed with the 40× and 20× microscope objectives (NA_eff = 0.55 and NA_eff = 0.28) and a 320 μm slit showed evidence of self-focusing. However, the characteristic triangular shape for waveguides inscribed with a slit in the beam path was clearly evident. Following the calculations described in [39], the critical power for self-focusing in GLS (n₂=2.16⁻¹⁸ m²/W [21]) for a Gaussian, circularly polarized, astigmatically shaped beam with an aspect ratio of 10:1 is 72 kW. Depending on the objective’s transmission at 800 nm, 72 kW corresponds to the peak power of a sech ² pulse of 112 fs FWHM duration with a pulse energy between 13 and 15 nJ, which is 3 times lower than the lowest tested pulse energy. No material modification was observed with pulse energies less than 100 nJ for any of the tested objectives. Stretching the pulse duration to 400 fs reduced the amount of self-focusing, however, the magnitude of index change was still too weak to support a guided mode at 1550 nm wavelength. Using the 20× objective and increasing the pulse energy up to 3.5 μJ did not lead to an increase in index contrast. The strongest index change was found for a modification inscribed with the 40× objective using 350 nJ pulse energy, 400 fs pulse duration and overpassing the structure 4 times. The 14.5 × 18.5 μm large modification (vertical and horizontal dimension, respectively) supported a weakly guided mode at 635 nm, but no guiding was observed at 1550 nm due to the low index contrast.

In other highly nonlinear materials, such as LiNbO₃ [40] and polycrystalline ZnSe [41], the use of sub-picosecond pulses for the inscription of waveguides has proven to be challenging. Indeed, in the athermal regime in GLS glass pulse durations of a few hundred femtoseconds have been reported to only lead to a minor material modification that is insufficient for supporting a tightly guided mode at 1560 nm when using a femtosecond laser operating at 250 kHz repetition rate in conjunction with slit beam shaping [15]. However, by stretching the pulses to picosecond durations, a significantly stronger material modification and thus well confined guided modes was achieved [15]. Similarly, in this body of work when increasing the pulse duration to 1.5 ps, strong modifications became apparent as shown in Fig. 4. This is a result of reduced nonlinear propagation effects, contribution from avalanche ionization and a spatially stronger confined energy deposition when using picosecond pulse durations as compared to femtosecond [42]. To improve the waveguide circularity, two different slit widths (320 μm and 520 μm) were tested with the 40× Olympus objective. Multiple overpasses of the modified region were examined in order to maximize the index contrast. Single-pass and 4 overpasses were evaluated and for the case of the 320 μm slit, additional waveguides with 8 overpasses were inscribed. Figure 4 shows the cross-section of the waveguides fabricated using 200, 350 and 500 nJ pulse energy and 4 overpasses. The structures written with the 520 μm slit resemble the shape of the structures written with shorter pulse durations, exhibiting a triangular to rectangular shape. The cross-section of the waveguides fabricated with the 320 μm slit are almost circular at low pulse energies and become increasingly elliptical at larger pulse energies. A horizontal black line is apparent in the microscope image of the waveguides fabricated with 350 and 500 nJ pulse energy using the 320 μm wide slit. This is caused by fracture due to internal stress relief after grinding and polishing the sample’s end-face.

 

Fig. 4 End on view of the inscribed waveguides using 1.5 ps pulses and a slit width of 520 μm (top) and 320 μm (bottom). The structures were written with the following parameters: 40× objective, pulse energy 200, 350 and 500 nJ (left to right), translation speed 1.5 mm/min, 4 overpasses. Small stress fractures are apparent in the waveguides inscribed with 350 and 500 nJ pulse energy and the 320 μm wide slit.

Download Full Size | PDF

The graph in Fig. 5 illustrates the MFD at 1550 nm as a function of pulse energy, number of overpasses and slit width. For the structures written with the 520 μm slit (solid black lines), the mode size increases dramatically for pulse energies below 200 nJ due to the lower index contrast and reduced physical size. A stronger material change is achieved when overpassing the same structure multiple times, resulting in a smaller MFD when using the 520 μm wide slit. Correspondingly, using the 320 μm slit the MFD gradually decreases from single-pass to 8 overpasses while the physical size stays constant. The smallest MFD of 8.3±0.8 μm was achieved using a pulse energy of 200 nJ and 8 overpasses, resulting in a index contrast of Δn = 0.006 ± 0.001 (0.25%). The mode-profiles at 1550 nm of a single pass and 8 overpasses waveguide written with 200 nJ pulse energy, as well as the mode-profile of a Corning SMF-28e single-mode fiber for comparison, are shown in Fig. 6.

 

Fig. 5 MFDs in microns at 1550 nm as function of pulse energy for waveguides written in GLS with 1.5 ps pulse duration at 1 kHz repetition rate using a 40× objective and a translation speed of 1.5 mm/min.

Download Full Size | PDF

 

Fig. 6 Mode-field profiles at 1550 nm of waveguides inscribed with 200 nJ pulse energy using a 320 μm slit for (a) single and (b) 8 overpasses and (c) Corning SMF-28e.

Download Full Size | PDF

The measured propagation losses in the athermal regime using slit beam shaping are summarized in Table 2. The lowest propagation loss of 2.01 ± 0.12 dB/cm for waveguides inscribed with the 320 μm slit was found using 200 nJ pulse energy and 4 overpasses. Increasing the number of overpasses to 8 but leaving the pulse energy unchanged increased the propagation losses to 2.58 ± 0.18 dB/cm. Therefore, unlike in the cumulative heating regime, the lowest propagation loss does not correlate with the smallest MFD when using slit beam shaping and 1 kHz pulses. This indicates that the stronger modification of the glass network observed using 8 overpasses is associated with the generation of defects that increase the propagation losses. Correspondingly, raising the pulse energy from to 200 to 250 nJ increased the index contrast as well as the propagation losses. The minimum propagation loss of 2.01 ± 0.12 dB/cm is significantly higher than the lowest losses reported in GLS in the athermal regime by McMillen et al. [15]. McMillen et al. used slit beam shaping, 400 nJ pulse energy, 250 kHz repetition rate pulses with 1.5 ps duration to create waveguides with 11.3 μm MFD and 0.65 dB/cm propagation loss at a wavelength of 1560 nm. Compared to this work, the higher repetition rate of 250 kHz and faster feedrate of 15 mm/min meant that the sample was irradiated with a 2.5× larger number of pulses per unit length. This suggests, that even though the pulse duration, pulse energy and resulting MFD are comparable, the propagation losses of GLS waveguides are highly sensitive to the irradiation parameters.

Table 2. Summary of the propagation losses for waveguides written in GLS with 1.5 ps pulse duration using a 40× objective, 1.5 mm/min translation speed and a 320 μm slit.

View Table | View all tables in this article

4.2. Multiscan waveguides

Multiscan waveguides in the athermal regime were fabricated using an Olympus 100× Plan N immersion oil objective with NA = 1.25 (NA_eff = 0.66) at a depth of 170 μm. The waveguide width was set to 6 μm by writing 15 adjacent lines with 400 nm spacing while translating the sample back and forth. The focal diameter of ≈1 μm ensured adequate overlap between the individual lines. Figure 7 shows the cross-section of three waveguides written with 400 mm/min translation speed and a repetition rate of 728 kHz. Structure (a) was fabricated with 20 nJ pulse energy and is square-like in shape. Structure (b) was inscribed with 40 nJ and has a rectangular shape while structure (c) was inscribed using 90 nJ pulse energy which led to an elongated tear drop shape of the waveguide as a result of strong heat diffusion. This change in shape was used to distinguish between athermal and thermal fabrication. Unlike when using slit beam shaping, as a result of the high NA focusing no evidence of self-focusing was observed for the tested pulse energies even for 50 fs short pulses.

 

Fig. 7 End on microscope images of multiscan waveguides inscribed using 20, 30 and 90 nJ (a, b, c) pulse energy at a repetition rate of 728 kHz with a pulse duration <50 fs and a translation speed of 400 mm/min.

Download Full Size | PDF

An increase in pulse energy dominantly led to an increase in the vertical size of the inscribed waveguides while the width stayed constant. This illustrates the flexibility of the multiscan technique in tailoring the waveguide size. Lower pulse repetition rates enabled the use of larger pulse energies without strong thermal diffusion or heat accumulation effects. Hence lower repetition rates feature greater flexibility in tailoring the waveguide size by just using the pulse energy without having to change the focusing optics to a lower numerical aperture in order to increase the vertical size. For instance at 255 kHz, the vertical size could be tailored from 6 μm to 18 μm by increasing the pulse energy from 20 to 90 nJ. This can be of importance for inscribing waveguide tapers, however, using lower repetition rates comes at the expense of slower fabrication speeds. In general, the influence of the translation speed on the physical size increased with increasing pulse energy, although the translation speed had less influence on the physical size compared to the pulse energy. Pulse energies of more than 30, 40 and 50 nJ at 1020, 728 and 510 kHz, respectively, led to fabrication in the thermal regime while at 255 kHz repetition all investigated pulse energies (maximum pulse energy 90 nJ) resulted in athermally inscribed structures.

Single-mode guiding at 1550 nm was achieved for waveguides with vertical sizes between 9 and 13 μm. The smallest MFD was 10.1 ± 0.8 μm, matching well the injected mode-profile of 10.4 ± 0.8 μm diameter of a Corning SMF-28e optical fiber. The insets in Fig. 8(c) show the mode-profile of the waveguide as well as the fiber for comparison. This waveguide with a physical size of 6.1 × 12.3 μm was inscribed using 728 kHz repetition rate, 25 nJ pulse energy and 800 mm/min translation speed. The physical size and MFD corresponds to an estimated index contrast of Δn = 0.004 ± 0.001 (0.17%). As illustrated by Fig. 8, the MFDs progressively decrease with increasing repetition rate within the tested parameter range. The amount of index change is repetition rate dependent. This is evidenced by the fact that the MFD and thus index contrast of waveguides with similar physical sizes, is less when written at low repetition rates compared to that of waveguides inscribed at high repetition rates. This indicates an increasing contribution to the total index change by thermally induced modifications to the glass network at higher repetition rates.

 

Fig. 8 MFDs of waveguides inscribed with a range of pulse energies, translation speeds and repetition rates of (a) 255 kHz, (b) 510 kHz, (c) 728 kHz and (d) 1020 kHz. The insets in (c) show the smallest mode-profile (left) of 10.1 ± 0.8 μm obtained using the multiscan technique (728 kHz, 25 nJ, 800 mm/min) in comparison to the mode-profile of a Corning SMF-28e fiber (right).

Download Full Size | PDF

The propagation losses at 1550 nm, measured using the Fabry-Perot technique, are presented in Fig. 9. Waveguides written at 255 kHz not only feature the lowest index contrast thus largest MFD, but also exhibit the largest propagations losses. This is similar to the observations in the thermal regime. With increasing repetition rate the propagation losses decrease, with the lowest propagation losses observed for waveguides written at pulse repetition rates of 728 kHz and 1020 kHz. Similar to LiNbO₃ [43], this suggests a rapid thermal annealing process taking place at higher pulse repetition rates whereby defects are removed and the associated propagation losses are reduced. The lowest propagation loss was measured to be 2.56 ± 0.09 dB/cm for a waveguide inscribed at 1020 kHz repetition rate, 240 mm/min translation speed and 15 nJ pulse energy. For comparison, multiscan waveguides inscribed using picosecond pulses at 2 MHz repetition rate into mid-infrared transparent polycrystalline ZnSe with propagation losses as low as 1.07 ± 0.03 dB/cm at 1550 nm have been reported [41]. In contrast, at a wavelength of 3.39 μm propagation losses of 0.8 dB/cm have been achieved in multiscan GLS glass waveguides [44].

 

Fig. 9 Comparison of the propagation losses of multiscan waveguides inscribed with a range of pulse energies, translation speeds and repetition rates of (a) 255 kHz, (b) 510 kHz, (c) 728 kHz and (d) 1020 kHz.

Download Full Size | PDF

A large number of the inscribed waveguides were multimode at 1550 nm. Thus their guiding properties were investigated in the mid-infrared wavelength range at 3.39 μm. Figure 10 illustrates the MFDs as a function of translation speed and pulse energy. All waveguides written at 255 kHz repetition rate failed to guide light at 3.39 μm due to their small size and insufficient index contrast. At 510 kHz even at the largest pulse energy of 90 nJ the waveguides were still single-mode at 3.39 μm. For the cases of 728 kHz and 1020 kHz repetition rate, multimode guiding was observed at low feedrates for pulse energies larger than 60 and 40 nJ, respectively, due to the strong contribution of heat diffusion to the physical size of the waveguide. The waveguide with the smallest measured MFD of 25 μm, as shown in Fig. 10(d), was inscribed at 1020 kHz repetition rate, 50 nJ pulse energy and 1000 mm/min translation speed. For comparison, a mode-field diameter of 22 × 27 μm at 3.39 μm was reported for multiscan waveguides in GLS glass inscribed using 400 fs long pulses at 500 kHz repetition rate and a laser wavelength of 1030 nm [44]. The similarity in MFD indicates that the pulse duration (50 versus 400 fs) and laser wavelength (800 versus 1030 nm) has limited influence on the obtainable index contrast in GLS glass. Using the multiscan technique waveguides with a MFD as low as 20 μm at 3.39 μm were demonstrated in the borate crystal YCOB [45].

 

Fig. 10 MFD at 3.39 μm over translation speed and pulse energy at repetition rate of (a) 510 kHz, (b) 728 kHz, (c) 1020 kHz and (d) smallest mode found for 3.39 μm wavelength (25 μm). The waveguide (11.3×20.1 μ) was inscribed with 1020 kHz repetition rate, 50 nJ pulse energy and 1000 mm/min translation speed.

Download Full Size | PDF

The physical size of the waveguide with the smallest MFD was 11.3 × 20.1 μm. This corresponds to an estimated index contrast of Δn = 0.003 ± 0.001 (0.13%) which is smaller than the index contrast at 1550 nm wavelength. A reduction in index contrast of ultrafast laser inscribed waveguides at longer wavelength has also been observed in YCOB crystals [45] and ZBLAN glass [31]. Even though the achieved index contrast in GLS supports a well confined guided mode, a four times larger index contrast of Δn = 0.012 ± 0.001 was reported for multi-scan waveguides at 10.6 μm wavelength in 75GeS₂-15Ga₂S₃-4CsI-2Sb₂S₃-4SnS chalcogenide glass [46].

5. Discussion

Table 3 compares the individual inscription regimes against each other and highlights their advantages and disadvantages for the waveguide inscription into GLS. In the athermal regime using slit beam shaping, strong self-focusing due to the high Kerr-nonlinearity of GLS glass was observed. Stretching the pulse duration of the 1 kHz repetition pulse train to 1.5 ps, efficient energy deposition and thereby material modification was achieved. By overpassing the same waveguide up to 8 times resulted in the highest index contrast (Δn = 0.006 ± 0.001) and smallest mode-field diameter (8.3 ± 0.8 μm) out of all inscribed waveguides. However, increasing the number of overpasses beyond 4 increased the propagation losses from 2.01 ± 0.12 dB/cm to 2.58 ± 0.18 dB/cm due to the creation of defects. The size of the waveguides in the athermal regime using slit beam shaping is determined by the focusing conditions. In order to obtain guiding in the mid-infrared wavelength region the cross-sectional size of the waveguides has to be increased. This requires lower NA focusing thereby increasing the detrimental influence of Kerr self-focusing, which makes the inscription of single-scan mid-infrared waveguides challenging using slit beam shaping. In contrast, the multiscan technique in the athermal regime offers great flexibility in waveguide shape and size. Furthermore, the high NA focusing enables the use of 50 fs short pulses without suffering from self-focusing even in a highly nonlinear material such as GLS. It was found that the minimum propagation losses gradually decreased from 5.50 ± 0.23 dB/cm to 2.56 ± 0.09 dB/cm and the maximum index contrast of the inscribed waveguides increased when increasing the laser repetition rate from 255 kHz to 1020 kHz. The smallest MFD using multiscanning was 10.1 ± 0.8 μm corresponding to an index contrast of Δn = 0.004 ± 0.001. This indicates that the additional heat diffusion associated with higher repetition rates not only contributes to the index contrast but also mitigates the creation of defects that otherwise would increase the propagation losses. However, the smallest propagation losses using multiscanning were higher than those for slit-beam shaping and the index contrast was lower. Finally, in the thermal regime using 5.1 MHz repetition rate pulses, tear-drop shaped structures were observed associated with spherical aberrations due to refractive index mismatch between sample and immersion medium. The thermal regime enabled simple control over waveguide size but unlike multiscanning, not over the shape. Nevertheless, circular mode-profiles with diameters as low as 9.3 ± 0.8 μm were observed corresponding to an index contrast of Δn = 0.005 ± 0.001. Furthermore, the lowest loss waveguide exhibited propagation losses of 1.82 ± 0.07 dB/cm.

Table 3. Comparison between the different regimes for the inscription of waveguides into GLS glass.

View Table | View all tables in this article

6. Conclusion

A detailed study on the ultrafast laser inscription of waveguides into commercial chalcogenide glass (GLS) was conducted. Athermal and thermal fabrication by using different laser repetition rates was explored and three different techniques, cumulative heating, slit beam shaping and the multiscan method were employed to inscribe waveguides with circular mode-fields. Within the tested parameter range, inscribing waveguides in the thermal regime, i.e. cumulative heating, resulted in the lowest propagation losses and enabled straight forward control over the waveguide size. Thermal fabrication offered the second highest index contrast only exceeded by using picosecond pulses in conjunction with slit beam shaping. Furthermore, cumulative heating exhibited shorter fabrication times compared to the multiscan method and less susceptibility to Kerr self-focusing in contrast to slit beam shaping.