Mid-infrared astrophotonics: study of ultrafast laser induced index change in compatible materials

1. Introduction

The mid-infrared (MIR), 5–20 μm, represents a fascinating wavelength range for astronomy for a variety of reasons, particularly for exoplanet hunting. Celestial objects emit infrared radiation at wavelengths strongly related to their temperature [1]. Relatively warm objects (> 300 K) such as the Earth are much brighter in the wavelength region known as the astronomical L’ band (3.5 – 4.1μm) than in the near-infrared. At the same time, the brightness of the host star is generally still rising to peak at much shorter wavelengths (often in the visible). As a consequence, the brightness contrast between the star and planet presents a less daunting observational challege in the L’ band [2]. In addition, the mid-IR comprises the “chemical fingerprint” region, where biomarkers such as water, methane and carbon dioxide can be found through their spectroscopic bands.

Research projects and publications during the last decade [3–8] have demonstrated that photonic devices can greatly simplify ground-based and space-borne astronomical instruments and improve their performance [9, 10]. These findings facilitated the birth of a new research discipline: termed astrophotonics [11], where photonic technologies are expected to have an important role in the next generation of Extremely Large Telescopes (ELTs) [1] due to the ability to integrate components such as beam combiners [12–17] and spectrographs [18] in palm-size devices. Their small size makes them easy to temperature stabilise and to implement within the limited instrument volumes available in current telescope facilities. Interferometry is one of the main applications of astrophotonics, however it requires reformatting the telescope pupil (2D) into a linear 1D array of single mode waveguides before recombining them for interferometric processing. This reformatting needs a 3D device configuration based on optical fibres or arrays of 3D waveguides. The latter option adds robustness to the system and makes it more insensitive to the environment. However, one of the most mature fabrication techniques for optical waveguides, lithography, is poorly suited to the creation of arbitrary 3D configurations that not only reformat the pupil the pupil plane of a telescope but do so via a set of waveguides that share the same optical path length in order to preserve the spatial coherence of the guided wave. More details about the design requirements of these reformatting devices can be found in [19]. A leading technology that offers the possibility to fabricate full 3D, compact photonic components is ultrafast laser inscription (ULI) [20]. Over the last two decades there has been a lot of research on the interaction between femtosecond laser pulses and transparent glasses (silicates [21–23], phosphates [24], fluorides [25], etc) to fabricate waveguides tailored for different wavelengths and applications. However, the transparency of common silica glasses rapidly decays beyond 2–2.2 μm. In order to achieve good transparency at longer wavelengths exotic materials (i.e. chalcogenides, fluorides, etc) are required (featured in a special journal issue [26]). This class of glasses presents new challenges that stem from their properties such as large nonlinearities [27], brittleness and sensitivity to UV radiation or hygroscopicity [28]. Nonlinearities, for example, exponentially increase from the fluoride family of glasses [29], to oxides [30], sulfides [31] and selenides [32]. The ultrafast laser inscription of waveguides requires high peak power pulses in order to modify the material at the focal volume. The high nonlinearities of these materials can distort the beam and change the focusing conditions of the femtosecond laser beam during the inscription process.

Fluorides [29] are a family of materials based on fluorine as the main component and can range from a two element crystal such as CaF₂ [33] to more complicated multicomponent compositions like ZBLAN (ZrF₄-BaF₂-LaF₃-AlF₃-NaF) [14, 34, 35]. They benefit from their low refractive index (around 1.5), which reduces the influence of spherical aberrations during ULI, avoids the requirement for complicated multilayer antireflection coatings as can be the case for chalcogenides, whose refractive index is usually above 2 [14].

Chalcogenide glasses represent an important class of amorphous semiconductor consisting of at least one or more of the chalcogen elements: sulphur, selenium and tellurium [36]. These elements are covalently bonded to network former such as As, Ge, Sb, Ga, Si or P [36] enabling a wide variety of possible stoichiometries.

Early work on femtosecond laser inscription in chalcogenides started with the simplest forms of glass: arsenic selenides (As₂Se₃) [32] and arsenic sulphides As₂S₃ [21,31,37–44], the latter glass featuring the lowest nonlinearity by an order of magnitude among the chalcogenide glasses. This makes sulphides in general more suitable when tightly focusing ultrashort laser pulses. The majority of these glasses produce a negative refractive index when exposed to a femtosecond laser. Adding a controlled amount of germanium to these chalcogenide glasses tailors the refractive index and can change the coordination number of the glass, and tune physical properties such as photodarkening, the cut-off wavelength or density of the glass [45–47]. Moreover, the inclusion of Germanium to form Ge-As-S glasses has been shown to have an effect on the sign and magnitude of the refractive index change generated during ultrafast laser inscription [44,48–51]. Replacing arsenic with gallium leads to a highly stable and non toxic chalcogenide glass [52–55] called Gallium Lanthanum Sulphide (GLS). GLS has become a widely studied glass due to its favourable compromise between nonlinearity and transparency [13,56–64].

A lot of the work on femtosecond laser interaction with MIR transparent chalcogenide glasses has been done in thin films [40], bulk materials [27, 31, 32, 37, 42, 45, 46] and in waveguide fabrication [38, 39, 43, 65]. The rise in interest for the MIR over the last few years has led to elevated activity in the development of mid-IR transparent glasses [26]. Some of these glasses may have potential in MIR astrophotonics. Consequently it is important and timely to perform an ultrafast laser inscription parametric study of these glasses.

2. Summary of the investigated materials

For a glass to be suitable for MIR astronomical science and compatible with the ultrafast laser inscription technique for waveguide fabrication it needs to fulfil certain requirements. Firstly, the material needs to be transparent at both the MIR (∼4μm) and at the inscription wavelength (800 nm, in this case), as the technique is based on nonlinear processes including multiphoton ionisation (see section 3).

Materials with low refractive indices are preferred. This reduces the Fresnel losses at the input/output of the waveguides due to the refractive index mismatch between the glass and the air. Therefore, dielectric antireflection coatings can be avoided and a lower refractive index reduces spherical aberrations during ultrafast laser inscription. However, most of the common low refractive index glasses are not transparent in the MIR (except for fluorides), meaning that this condition may not be met. On the other hand, materials with nonlinearities as low as possible are also preferred as the distortion of the beam while being focused is directly related to the magnitude of the nonlinear refractive index of the material. Low refractive index materials also have in general low nonlinearity (see Miller’s rule [66]).

Taking these restrictions into account a selection of available glasses was sourced, these listed in Table 1. All the materials are transparent at both the laser wavelength and the MIR. The term IZSBGC represents a family of fluoride glasses with the following fundamental chemical composition: 32InF₃-20ZnF₂-SrF₂-18BaF₂-8GaF₃-2CaF₂ [67].

Table 1. List of glasses utilised for femtosecond laser fabrication of MIR waveguides. The table also shows the type of glass and their origin. The term IZSBGC represents a family of fluoride glasses with the following fundamental chemical composition: 32InF₃-20ZnF₂-SrF₂-18BaF₂-8GaF₃-2CaF₂.

View Table | View all tables in this article

The transparency of the glasses was measured using a UV-VIS-NIR spectrophotometer (Cary 5000) from 175 to 3300 nm. Fig. 1 shows the absorption coefficient curve of the glasses where the Fresnel losses have already been substracted from the recorded data. The wavelength dependent Fresnel losses were calculated using the Sellmeier equation for the refractive index of the materials. The absorption coefficient determines the minimum loss for waveguides inscribed in these materials. It can be observed that in this case the absorption coefficient is well under 0.25 at long wavelengths, which is equivalent to the 1 dB/cm upper limit set for efficient astrophotonic devices. Table 2 shows the refractive index values of the studied materials at both the laser wavelength (as it can impact the inscription process) and the MIR (as it will determine the Fresnel losses at the interfaces).

 

Fig. 1 Absorption coefficient curve for the selection of potential MIR glasses compatible with the ultrafast laser inscription technique.

Download Full Size | PPT Slide | PDF

Table 2. Refractive index value table for the selection of potential MIR glasses compatible with the ultrafast laser inscription technique..

View Table | View all tables in this article

3. Fabrication

In this study, two Ti:sapphire femtosecond laser systems emitting at 800 nm were used. A regeneratively amplified, low-repetition rate femtosecond laser (Hurricane, Spectra-Physics) and an ultrafast Ti:sapphire oscillator (FEMTOSOURCE XL 500, Femtolasers GmbH). The first system produces 1 kHz pulse trains with a pulse duration of <120 fs, while the high repetition rate laser produces <50 fs pulse duration with a repetition rate of 5.1 MHz (this repetition rate can also be scaled down using a Pockel cell). The pulse duration of the lasers can be stretched to a few picoseconds by adjusting the internal compressor of the systems.

As a general rule, when fabricating waveguide structures in the low-repetition rate regime (up to a few hundred kHz) low numerical aperture (NA<0.6) microscope objectives were used (i.e. 20× Olympus UMPlanFL, NA 0.46 or 40× NA 0.6). However, when working in the high-repetition rate regime high NA microscope objectives were utilised (i.e. a.- Olympus 100× 1.25NA, b.- Zeiss N-Achroplan 100× 1.25NA and c.- Nikon 50× 0.9NA). The latter one is also known as the thermal regime, as opposed to the athermal regime, when fabricating structures at low repetition rates [62].

The refractive index change induced by the tightly focussed femtosecond pulses can be a result of one or several of the following mechanisms: densification [68,69], stress [70], colour centres [71] or elemental diffusion [72]. This refractive index change can be positive, negative or a combination of both and the shape, sign and magnitude of this change will determine the approach taken to create a waveguide. An explanation of the different types of modifications can be found in [73]. In this paper, we present a more extensive classification of ways to create waveguides depending on specific modifications. If the refractive index modification is positive (top row of Fig. 2), the waveguide can be fabricated by inscribing a single track (P1), an overlap of modifications in the horizontal plane (P2, also known as multiscan [74]) or by grouping individual tracks around a central modification (P3). The size and magnitude of the induced refractive index will determine the overlap needed between consecutive single modification tracks. Fig. 2 shows a representation of the different modification types (and shapes) and how to create a waveguide by combining the modification tracks. This last variety can be divided into 2 kinds of waveguides depending on whether the modifications are separated or overlapped. The first version is termed expanded-core or large-mode area waveguide [44,55,61,75], while the latter one will be referred as Rosette pattern waveguide in this manuscript. If the refractive index of the modification is negative there are 3 approaches to creating a waveguide: type I negative modifications (N1) [73], depressed cladding (N2) [76] or a combination of P2 and N2 to create a Multiscan Depressed Cladding waveguide (N3). In the case of N2 and N3, the overlap between modification tracks and the number of modification rings/layers around the unmodified region will determine the propagation loss properties of both the fundamental and the higher order modes. Positive refractive index modifications are usually preferred over depressed cladding structures as the fabrication times are much shorter and the design complexity of interferometric devices increases exponentially for depressed cladding waveguides. Also, depressed cladding structures are formed by many more modifications per unit area, which can lead to stress fractures.

 

Fig. 2 Investigated waveguide geometries depending on the sign and shape of the induced refractive index change.

Download Full Size | PPT Slide | PDF

When analysing the tracks created in the different glasses, a wide variety of modification types can be observed. Table 3 lists the shape and sign of the modification, together with the most suitable waveguide geometry. For some of these glasses the sign of the refractive index modification is fluence dependent. For instance the sign of refractive index change in Ge-As-S glasses can be tailored by changing the Ge/S ratio. Ge acts as a glass network former, modifying the coordination number of the glass which leads to a change in glass density and consequently to a change in refractive index of the bulk material. This phenomenon has already been shown in [44,50]. Table 3 shows how for small concentrations of Ge, the refractive index of the modification is negative (Ge_11.5As₂₄S_64.5 or Ge₁₅Sb₂₀S₆₅), while for larger concentrations (i.e. Ge₂₅As₁₀S₆₅) the sign of the refractive index modification can be either positive or negative.

Table 3. Analysis of the different modification types obtained with a femtosecond laser including the shape of the modification and its sign. See Fig. 2 for a graphic representation of the suitable waveguide (wg) geometry.

View Table | View all tables in this article

4. Waveguide type & morphology

As shown in the previous section, some MIR glasses exhibit a positive refractive index change while others show a negative index change when exposed to femtosecond laser radiation. This section outlines how to generate structures that support guided modes for the three material families studied: chalcogenides, fluorides and germanates.

4.1. Chalcogenides

The easiest way to inscribe a waveguide in any material using ULI occurs when the sign of the refractive index modification is positive. If, at the same time, the size and shape of the modification is proportional to the fluence applied to the glass, then fabricating waveguides at different wavelengths can be achieved by varying the irradiation conditions to obtain waveguides of the right dimensions. Unfortunately, this case does not match what happens in reality, especially with chalcogenide glasses as they present with high-nonlinearities, which result in distortion of the focused beam. This was observed, for example, during the fabrication of waveguides in GLS in the thermal regime. The shape of the modification changed from circular to a tear-drop shape when increasing the fluence. In the case of 75(GeS₂)-15(Ga₂S₃)-10(CsCl) in the thermal regime, low pulse energies were enough to fabricate round waveguides (P1) that were big enough to guide a highly confined mode at 1550 nm (MFD ∼8μm) with a refractive index contrast of 7×10⁻³, using Marcuse’s equation [78, 79]. However, due to the shape dependency with the applied fluence, in order to scale up these waveguides to the MIR more complicated structures need to be built as the cross section of these modifications show regions of positive as well as negative index change. One option that can be used when the shape of a single modification is vertically elongated is the so-called multiscan technique (P2), where a series of tracks are overlapped horizontally to create a rectangular cross-section waveguide (see Fig. 3b). For GLS in the athermal regime, this resulted in single-mode waveguides at 1550 nm with a refractive index change of 3.5×10⁻³. This structure could be easily scaled for guiding at longer wavelengths.

 

Fig. 3 Bright-field microscope image, mode field diameter and estimated refractive index contrast of (left to right): As₃₉S₆₁, GLS and Ge₂₅As₁₀S₆₅.

Download Full Size | PPT Slide | PDF

Other glasses like As₃₉S₆₁ and some of the Ge-As-S types are well suited to the inscription of stress waveguides (N1) in the thermal regime due to the modification shape and refractive index contrast (usually below 4×10⁻³). Ge₁₅Sb₂₀S₆₅ and Ge_11.5As₂₄S_64.5 showed a refractive index change of around ∼2–3×10⁻³, while for As₃₉S₆₁ the refractive index contrast can reach up to ∼4×10⁻³. These kind of structures show strong light confinement in the horizontal axis but are relatively lossy along the vertical axis due to the lack of any refractive index change. This can be solved by adding a set of extra tracks above and below the original pair or modifications, but most of these soft glasses, because of their high thermal coefficient, suffer from stress fractures that can propagate across the whole waveguide structure. One glass well suited to this kind of structure in the athermal regime is Ge₂₅As₁₀S₆₅. The higher concentration of Ge (compared to the other three glasses above) is likely to improve the mechanical properties of the glass by reducing the network reticulation. This results in less mechanical stress when redensification occurs, thus reducing the susceptibility to fracture. It must be stressed that fractures can still manifest in this glass (as illustrated in Fig. 3c). This material can be used to create multiscan depressed cladding structures (N3). The refractive index contrast obtained for this chalcogenide is the largest of all these glasses (∼7×10⁻³). The multiscan depressed cladding geometry also has the benefit of straight forward scalability in size to tailor the waveguide to longer wavelengths. Fig. 3 shows three of the largest refractive indices obtained for some of the chalcogenide glasses.

Four kinds of waveguides were formed (P1, P2, N1 and N3). For glasses such as As₃₉S₆₁, Ge₁₅Sb₂₀S₆₅ and Ge_11.5As₂₄S_64.5 stress waveguides (N1) were created as the superposition of several modification led stress induced fractures in the glass. For the same reason, no single mode waveguide could be fabricated for 1550 nm guidance (see Fig. 3). However, bigger waveguide structures were characterised at longer wavelengths in As₃₉S₆₁, resulting in mode field diameters around 33 μm when injecting light at 3.39 μm and an estimated refractive index change of −4×10⁻³. The cross-sectional shape obtained for Ge₂₅As₁₀S₆₅ was very similar to the previous case yet the thermal expansion coefficient of this glass is lower, which allows for the superposition of multiple tracks (N2) without suffering from catastrophic damage. Some fractures can be observed on Fig. 3c, but the structure still guided light. The refractive index contrast obtained for this glass was the largest of any of the glasses with negative modification (−7×10⁻³), leading to the smallest MFD. The same order of magnitude (but opposite sign) refractive index contrast was obtained for 75(GeS₂)-15(Ga₂S₃)-10(CsCl).

While comparable refractive index contrast to those reported in the literature was also obtained for GLS [13,58,62], for other chalcogenides such as the Ge-As-S family (i.e. Ge₂₅As₁₀S₆₅), the refractive index contrast was around 10× the value reported by Damico et al. in their work based on Ge₁₅As₁₅S₇₀ [44]. Refractive index contrasts in the order of mid 10⁻³ are needed to keep the modes tightly confined inside the waveguide structure and to be able to design bends in the waveguides without major losses.

4.2. Fluorides

Crystals based on two atomic elements by the ionic bond between fluorine and an alkaline earth metal (group IIA) have a low refractive index and show remarkably high transparency from the visible to long infrared wavelengths. For these reasons. some of these crystals such as BaF₂ and CaF₂ are commonly used for the fabrication of MIR lenses. Being crystalline materials the interaction with femtosecond laser pulses for the fabrication of waveguides in the thermal regime is far from ideal as the modifications tend to show a pearl chain structure [77], making it impossible to generate a low-loss waveguide. Here, the fabrication of waveguides in CaF₂ was investigated with no successful waveguiding results (see Fig. 4b).

 

Fig. 4 Some examples obtained for 3 fluoride materials (left to right): Pr : IZSBGC, CaF₂ and IZSBGC. Bright-field microscope image, mode field diameter and estimated refractive index contrast.

Download Full Size | PPT Slide | PDF

In the case of fluorides from the IZSBGC family, the problem was not the nonlinearity but a series of positive and negative substructures generated when trying to scale up the waveguide size by increasing the fluence in the thermal fabrication process. At low pulse energies, the waveguides show a close-to-homogenous refractive index profile (although low refractive index contrast 9×10⁻⁴), but this configuration is not maintained when increasing the pulse energy. A solution for this is to overlap multiple tracks in a circular configuration around a central modification in order to create larger waveguide structures. However, this approach did not work as expected for IZSBGC glasses as the refractive index contrast was too small, thus the mode confinement was weak and the light leaked into the bulk material, making the waveguides highly lossy. The refractive index change values obtained during these experiments for fluoride glasses agree with those obtained previously in the literature for ZBLAN [80]. Lancaster et al. reported a refractive index contrast of the same order of magnitude (but opposite in sign) for their Tm:ZBLAN waveguide lasers [80]. Fig. 4 shows the best results obtained for fluoride materials.

4.3. Germanate glass

When the shape of the modification is close to a circle (either circular or oval) and the sign of the refractive index of the modified region is negative, the best waveguide format is to surround an unmodified region of material by a superposition of negative modifications to form a depressed cladding. The mode will then travel though the unmodified region, while the depressed cladding arrangement will confine the light inside the structure. One of the main benefits of this technique is that it is extremely easy to scale with wavelengths, by controlling the size of the waveguide. This waveguide composition has been successfully shown in fluorozirconate glass materials such as ZBLAN [14] with low propagation losses in the MIR. When fabricating interferometric devices, the small refractive index contrast (∼10⁻³, low confinement) obtained in ZBLAN would result in large bend losses or extremely long waveguides, which would consequently mean large overall losses. Materials like Schott IRG-2 can manifest in a refractive index contrast up to 10× larger than ZBLAN, which leads to a higher mode confinement and, therefore, to low bend losses. For the fabrication of depressed cladding waveguides (N3) the best results in the thermal regime were obtained for IRG-2 with a refractive index contrast of around −5×10⁻³ and a mode field diameter of 8.3 μm when injecting light at 1550 nm. One of the benefits of IRG-2 and its moderate/large refractive index contrast is the ease of scaling the waveguide core size for single-mode propagation at different wavelengths. Fig. 5 shows the best waveguide result obtained for IRG-2 germanate glass.

 

Fig. 5 Bright-field microscope image, mode field diameter and estimated refractive index contrast of IRG2.

Download Full Size | PPT Slide | PDF

5. Discusion

One of the main challenges for astrophotonic devices is the need for low intrinsic losses (< 1 dB/cm), meaning that the glass absorption will always need to be below that value for these devices to be efficient. Any extra loss would mean that less photons would arrive at the detector after travelling through the instrument. This problem can be mitigated with one or more of the following solutions: using a larger telescope, lower readout noise cameras and longer integration times; all of them increase the cost of the experiment and/or reduce the efficiency of the system. Access to larger telescopes is expensive and very limited unless the instrument has already been tested in smaller telescopes with high success. Increasing the camera’s integration time not only reduces the efficiency of the already expensive telescope time, but can also lead to other problems such as an increase in the noise level, which can mask interesting results. This problem can be addressed using low readout noise cameras/detectors which may not be accessible (and the price increases exponentially as the readout noise decreases). All these solutions are therefore either expensive, inefficient or both, meaning that special care needs to be taken on the reduction of glass absorption and waveguide intrinsic losses.

Depending on the sign and shape of the refractive index modification obtained when exposing mid-infrared materials to femtosecond laser irradiation, different approaches can be used when creating waveguides. When the refractive index sign of a single modification is negative, the guided mode will travel through unmodified material surrounded by several negative tracks. When the refractive index change is positive, the waveguide can be formed from a single modification or an overlap of several depending on the shape of the single track cross-section. Table 4 summarises the results obtained for each glass including preferred geometry, the horizontal MFD and the induced refractive index contrast.

Table 4. Summary of the results obtained for each glass including the preferred waveguide geometry, the horizontal mode field diameter (MFD) and the refractive index contrast.

View Table | View all tables in this article

It is important to note that low refractive index materials (i.e. fluorides) that a priori seem ideal for the fabrication of low-loss waveguides, due to their lower Fresnel losses, tend to exhibit a refractive index contrast (∼ 10⁻⁴) that is too low for tight mode confinement. This results in large bend-losses associated with light leaking out of the waveguide structure. Other materials like chalcogenide glasses possess a larger refractive index (usually > 2), but they also have larger nonlinearities. As the focused beam travels through the glass it suffers from aberrations, which affects the focal volume shape and is intensity dependent. A simple way to create bigger structures that do not suffer large aberrations is the superposition of small modifications to form waveguides. One of the benefits of the superposition approach is the ability to arbitrarily scale and choose the shape of the waveguides depending on the application needs. The best three chalcogenides to fabricate waveguides using the superposition of multiple tracks under the optimized exposure conditions were GLS, 75(GeS₂)-15(Ga₂S₃)-10(CsCl) and Ge₂₅As₁₀S₆₅. For the latter one, the addition of the right amount of Germanium to the As-S family of glasses helps tailor the mechanical properties and refractive index of the modification.

Finally, one glass that offers easy scalability while having an intermediate refractive index (∼1.89) and nonlinearity is Schott IRG-2. Under the tested conditions, this germanate glass exhibited oval-shaped highly-negative refractive index modifications (∼ −5 × 10⁻³) that allowed for the fabrication of depressed cladding waveguides in a relatively simple manner.

For all the tested materials, the easiest way to generate single-mode waveguides in the L’ band was the overlap of individual modifications to create a larger structure, whether it was by generating a depressed cladding waveguide or a positive refractive index core surrounded by unmodified bulk glass. The superposition of individual modification tracks adds the possibility of tailoring the glass refractive index modification and the size of the guided mode.

The characterisation of the propagation losses falls out of the scope of this paper and will be reported in a follow up manuscript.

6. Conclusion

This paper reports a study of laser writing results and outlines the best strategy for successful writing of waveguides and discusses the role of the optical material properties on this process. We believe the waveguide refractive index contrast results obtained for some of the glasses explored in this manuscript show evidence that low-loss astrophotonic devices for the L’-band (3.5 – 4.1μm) will be possible in the next few years. These glasses include chalcogenides such as GLS, Ge₂₅As₁₀S₆₅ and 75(GeS₂)-15(Ga₂S₃)-10(CsCl), and germanate glasses such as IRG-2. Furthermore, the transparency of some of these glasses, such as 75(GeS₂)-15(Ga₂S₃)-10(CsCl) and Ge₁₅Sb₂₀S₆₅, expands up to the N-band, an atmospheric transmission window centered at 10 μm. The N-band is of special interest in the study of molecular clouds and protostars, allowing advanced photonic instrumentation to be constructed for these, and a handful of other astronomical themes based in this waveband.