1. Introduction

Bismuth germanate (Bi₄Ge₃O₁₂ or BGO), is a crystalline material with interesting optical properties [1]. The cubic crystal is non-hygroscopic, mechanically strong, has a high refractive index of n = 2.15 and is optically transparent down to the long-wavelength ultraviolet (UV) range (cutoff wavelength 320 nm). In addition, BGO features a very large electro-optic coefficient and does not exhibit a natural linear birefringence or optical activity, making BGO an excellent choice for voltage sensing applications where long time stability and precision are of high priority [2, 3]. In addition, BGO crystals are widely used as scintillation material for high-energy radiation detection [4]. Arrays of such scintillator crystals are used for 2-dimensional gamma ray detection for high-energy physics and in positron emission tomography (PET) systems, a functional imaging technique that produces a three-dimensional image of functional processes in the body [5]. Ultimately, the size of each individual crystal within a detector array limits the resolution of the system and the production of many small crystals is time consuming and thus expensive. The drawing of BGO fibers has been investigated to utilize fiber bundles instead of crystal arrays [6]. An alternative approach is to fabricate integrated optical circuits with buried waveguides that are directly inscribed into a bulk crystal [7] as schematically shown in Fig. 1.

 

Fig. 1 Possible implementation schemes for the fabrication of scintillator arrays. Left: Array of discrete crystals. Right: Single crystal with waveguide array embedded.

Download Full Size | PDF

While ion-implantation has been suggested as a possible technique for the formation of planar optical waveguides [8], the femtosecond-laser direct-write method is a highly flexible technique that can be used to fabricate fully 3-dimensional optical circuits into virtually any kind of transparent dielectric material [9, 10]. The standard technique of optical waveguide inscription exploits a local (i.e. limited to the focal volume) increase of the refractive index that is directly induced by a train of focused femtosecond laser pulses. This is commonly referred to as type-I modification and schematically shown in Fig. 2(a).However, in the case of crystalline materials, the index of refraction can normally not directly be increased by the writing process as the crystal lattice is a densely packed arrangement of atoms and further densification is hard to achieve.

 

Fig. 2 Schematic of different types of optical waveguides that can be inscribed via the femtosecond laser direct-write technique. n_b: refractive index of the bulk material. n_c: refractive index of the laser-inscribed core (n_c > n_b). n_d: refractive index of the laser-inscribed depressed cladding structure (n_d < n_b). n_s: refractive index caused by the formation of a stress field (n_s>n_b).

Download Full Size | PDF

Thus, a different technique is typically used: By inscribing small damage lines in close distance to each other, a localized stress field can be induced [11]. Via the elasto-optic coefficient, these stressed regions exhibit a higher index of refraction and light can thus be guided along those tracks, commonly referred to as type-II modification based waveguides as shown in Fig. 2(c). It is thus not surprising that the fabrication of optical waveguides in BGO was demonstrated for the first time by inscribing damage lines into the material using an amplified femtosecond laser system operating at kilohertz (kHz) repetition rates [12]. In this case, the stress field created by the damage lines locally increased the index of refraction and a waveguide was formed. Compared to type-I modifications, these type-II structures exhibit higher losses and a less confined mode profile, as the mode leaks into the damaged regions, resulting in fairly high scattering losses. More recently, it has been shown that thermal post-processing can lead to a substantial reduction of the losses of similarly inscribed waveguides. In addition, the possibility of fabricating depressed-cladding waveguides, as schematically shown in Fig. 2(b), has also been discussed in the same publication [13]. However, so far there are no reports on the feasibility of using Megahertz (MHz) repetition rate femtosecond lasers for waveguide inscription in BGO. Not only does the use of high-repetition rate laser systems lead to a dramatic reduction in fabrication times, it has also been shown that the onset of cumulative heating effects [14] can result in fundamentally different waveguide-formation mechanisms when the laser repetition rate exceeds about 1 MHz [15].

In this paper, to the best of our knowledge, we report for the first time on the fabrication of type-II as well as type-I (smooth index change, non-damage) modification based waveguides in BGO using high-repetition rate (MHz) femtosecond laser pulses. We have identified a processing window where the laser pulses initiate a local transformation of the crystalline structure into an amorphous state with an increased refractive index. In contrast to other crystalline materials, where it has been shown in the past that high-repetition rate laser pulses can be used to induce a negative refractive index change (Δn < 0) [16], we show here that in BGO a positive index change (Δn > 0) can be achieved. This allowed us to fabricate low loss (< 1dB/cm) type-I waveguides in BGO without damaging the surrounding material and without the need of creating a stress field inside the material. In addition, we discuss the fabrication of stress-induced waveguides in BGO by inscribing multiple damage-lines (4 damage lines in total instead of 2 as was demonstrated previously by Bin et al. in [12] and by He et al. in [13]), and show that this can lead to an improved control over the spatial dimensions of the resulting waveguide.

2. Experiment

The laser source used in our experiments was an extended cavity Ti:Sapphire chirped pulse oscillator (CPO), operating at a 5.1 MHz repetition rate, emitting sub-50 fs laser pulses at a center wavelength of 800 nm (FEMTOSOURCE scientific XL 500, Femtolasers GmbH). The laser beam was focused to an average depth of 170 µm below the sample surface by a 100 × 1.25NA oil immersion objective (refractive of the immersion oil used was n_D = 1.518). The objective used had a working distance of 200 µm, which limits the maximum depth at which waveguides could be fabricated. However, long-working-distance objectives could be used to extend that range if required. A fast RTP Pockels cell enabled us to reduce the repetition rate of the laser to selectively work in either the cumulative heating regime (5.1 MHz) or the repetitive modification regime (500 kHz). The sample itself was mounted on a set of 3D high precision air-bearing translation stages (Aerotech).

First we reduced the repetition rate of the laser to 500 kHz with the aim of fabricating type-II damage based waveguides. Due to the high-NA focusing used we were able to inscribe damage lines at much lower pulse energy than previously reported, i.e. with 300 nJ per pulse compared to several μJ as reported by Bin et al. in [12] and by He et al. in [13], respectively. Moreover, due to the still much higher repetition rate (500 kHz vs. 1 kHz) we were also able to write at much higher translation speeds (17 mm/s as compared to 20-50 μm/s). In addition, in order to better confine the stress-induced waveguiding region, we have inscribed a total of 4 damage lines in two rows as summarized in Fig. 3.This resulted in the formation of a stress-field that was enhanced in the central region between the two rows and the resulting waveguides supported a highly symmetrical optical mode at 1550 nm with a diameter of about 10 μm, see Fig. 3. In agreement to the results that have been reported in [13], guiding was observed for both the TE and the TM modes, respectively. Due to the improved confinement provided by the 4 damage lines (and thus an increase in the strength of the stress-induced refractive index change), the propagation losses of our damage waveguides were measured to be relatively low with a minimum of 3.9 dB/cm even without further post-annealing process steps.

 

Fig. 3 Optical microscope image (left) and measured near-field mode profile for the TM mode (right) of the inscribed type-II damage based waveguides. Writing parameters: repetition rate: 500 kHz, translation speed: 1000 mm/min, pulse energy: 300 nJ.

Download Full Size | PDF

Next, we increased the repetition rate of our laser to its maximum value of 5.1 MHz to investigate the possibility of utilizing cumulative heating effects for the waveguide inscription process. At a reduced pulse energy of 160 nJ and an increased translation speed of 2000 mm/min, we could observe the formation of waveguiding structures without the formation of damage lines, i.e. true type-I modifications. Figure 4 shows a DIC (difference-interference contrast) micrograph of the cross section of such a laser-written waveguide. Due to the high refractive index of BGO (n = 2.15) and the resulting high refractive index mismatch (refractive index of the immersion oil = 1.52), spherical aberrations distort the laser focus within the BGO sample. Thus, the core of the waveguide is highly elliptical and supports multimode guiding at a wavelength of 1.5 µm. As it turned out, those novel type-I structures in BGO feature extremely low propagation loss values. Coupling light from a standard single-mode fiber into the waveguide and measuring the total transmitted power, we were able to determine an upper limit for the propagation losses. Taking into account Fresnel losses at both end facets and assuming negligible coupling loss, propagation losses of < 0.8 dB/cm have been obtained.

 

Fig. 4 Difference-interference contrast (DIC) microscope image (left) and measured near-field mode profile (right) of the inscribed type-I structures. Writing parameters: repetition rate: 5.1 MHz, translation speed: 2000 mm/min, pulse energy: 160 nJ.

Download Full Size | PDF

In order to compare the performance and characteristics of those two different types of waveguides in BGO, we have measured the effective electro-optic coefficient for the different waveguides and have compared those results with the values for the bulk crystal. For this, we have directly contacted the top and bottom surfaces of the crystal with metal electrodes and have launched light from a polarised laser at a wavelength of 632 nm into the waveguides. Figure 5 shows the signal measured by a photodiode placed behind the analyser (a polarising beamsplitter mounted on a rotation stage) as a function of the voltage applied to the electrodes. It can clearly be seen that the type-II damage based waveguides still exhibit electro-optic properties while the type-I structures are characterised by a total loss of any electric field-induced polarisation rotation.

 

Fig. 5 Electro-optic switching comparison between the type-II (left) and the type-I modification based waveguides (right).

Download Full Size | PDF

We then collimated the output of our probe laser and measured the half-wave voltage (voltage required for a full 90 degree of polarization rotation) of the bulk crystal and compared this value to the type-II based waveguides. For the given crystal dimensions we have measured a half-wave voltage of V_π = 80kV for the bulk crystal (in accordance with the literature) and V_π = 126 kV for the waveguide, respectively. This shows that the waveguide inscription process resulted in a reduction of the electro-optic coefficient to a value, which is only 63.5% of its bulk value. We explain this behavior by a competition between the laser-induced stress in the waveguide and the field induced birefringence.

In order to investigate the reason for the total loss of any electro-optic properties in the type-I structures, and also to investigate the possible mechanisms responsible for the formation of the waveguides themselves, we have performed Raman spectroscopy measurements (Renishaw Ramascope) of the bulk material and the laser-modified region, respectively. For this the sample/waveguide was illuminated by 442 nm He–Cd laser focused using a 50 × microscope objective, yielding an estimated spot size of about 1 μm. The Raman spectrum was obtained by collecting the backscattered light, filtering the laser line and then reflecting the remaining signal off a diffraction grating and onto a charge coupled device (CCD) array. Figure 6 compares the Raman spectrum of the bulk crystal (left) with that of the waveguide (right). It can clearly be seen that the waveguide has a much broader Raman response, which is indicative of a local transformation of the crystalline structure into an amorphous state [17]. All the elements that form the BGO crystal are common glass-forming elements and we thus believe that the femtosecond-laser inscription process drives a local phase change from a perfectly ordered crystalline state to a less-ordered glass-like state. More research is necessary on this topic but we believe that it is highly likely that a novel non-equilibrium material state (embedded inside the BGO crystal) is generated during waveguide inscription, similar to what has recently been observed in the context of lithium-niobo-phosphate glasses [18].

 

Fig. 6 Raman spectrum of the bulk crystal (left) and the type-I modification (right).

Download Full Size | PDF

3. Conclusions and outlook

We have demonstrated the possibility of utilizing high-repetition rate femtosecond laser pulses for the inscription of waveguides into Bismuth Germanate (BGO) crystals. In contrast to previous results, we have shown that femtosecond laser pulses can be used to not only fabricate type-II damage based waveguides, but also true type-I structures of smooth refractive index change by utilizing cumulative heating effects at MHz repetition rates. For the type-II based waveguides, we have shown that the inscription of additional damage lines can lead to a better-confined stress region and thus to lower propagation losses as compared to the standard “parallel-line” technique. In addition, we have compared the effective electro-optic coefficient within those waveguides with the properties of the bulk crystal and have discovered an almost 40% reduction that is a result of the femtosecond laser inscription process.

The type-I structures on the other hand feature extraordinarily low propagation losses but are characterized by a complete loss of the electro-optic properties. We have shown that this can be explained by the fact that the waveguides are the result of a local (femtosecond laser-induced) transformation of the crystalline structure into an amorphous glass state. In addition, the waveguides are multimode at 1550 nm and support highly elliptical optical modes; an asymmetry that is mainly caused by spherical aberrations that are the results of the high refractive index mismatch between the BGO sample and immersion oil. However, the use of sophisticated beam-shaping techniques should make the fabrication of symmetric waveguides possible [19]. This, in turn would enable the realization of complex BGO ARROW structures [20] that could potentially be utilized to fabricate waveguides that feature both, very low propagation losses and at the same time high electro-optic coefficients as light in such structures is mainly confined to non-laser modified regions of the crystal.