1. Introduction

Whispering gallery mode (WGM) resonators are an appealing platform due to their high quality factor (Q) of up to a few billion [1]. Achieving a high-Q resonator on-chip has been possible with novel planar designs such as microdisks [2, 3], which have applications, amongst others, in sensing, signal processing and frequency metrology [4–6]. However, these resonators suffer from unavoidable fabrication errors which cause variations in dimensions that change the resonance frequencies substantially. Of the few recently proposed post-processing techniques to compensate for the fabrication tolerances [7–10], in situ photo-tuning of the material properties has gained significant interest as a non-contact process [9]. For this method to be efficient, nano-devices based on highly photosensitive material are required.

There have been few reports on photoinduced cavity formation based on commonly available silica [11]. This is due, in part, to the large light intensity required when using the photosensitivity of silica. The chalcogenide glass family, on the other hand, exhibits more than a million times higher photosensitive response [12–14], and this drastically lowers the laser intensity required for post-tuning.

Chalcogenide glasses display a wide range of photoinduced effects including photo-darkening, photo-crystallization, photo-contraction and photo-expansion [15]. This has led to applications in fiber bragg gratings [16] as well as photowritten high-Q bottle and photonic crystal cavities [17, 18]. Recently post-trimming of the resonance wavelength of the coupled ring resonators [9] and weak cavity enhanced resonance tuning on a microdisk based on As₂S₃ has been demonstrated [10, 19].

In this work we demonstrate post-tuning of the resonance frequency of an AMTIR-1 (Ge₃₃As₁₂Se₅₅, E_g = 1.95eV) microdisk with a Q factor of 1.2x10⁵. We show for the first time that both positive and negative shift can be produced on a single microresonator. We have achieved 1 nm positive and 0.5 nm negative tuning. This observation opens avenues for tuning of the resonances of an optical cavity with partial reversibility to compensate for the fabrication tolerances. For example, the FSR requirements of the cavity modes for mechanical mode generation from interference of stokes and pump resonant modes can be relaxed [20, 21].

AMTIR-1 glass is a readily available and attractive platform for integrated photonics because of its relatively higher Kerr nonlinearity (3 x As₂S₃, 350 x silica); excellent long wavelength transparency (0.7-12µm), limited by the presence of impurities higher glass transition temperature (362 C) compared with As-Se binary glasses due to its three-dimensional network structure caused by the presence of four-fold coordinated germanium [22, 23].

2. Experimental setup and Results

We fabricated microdisks from a 300nm thick AMTIR-1 film deposited by thermal evaporation onto an oxidized silicon wafer, using photolithograhy and ICP etching. The amorphicity and composition of the sample was analysed with X-ray diffraction (XRD) and energy-dispersive X-ray spectroscopy (EDS) respectively which was identical to the bulk sample, within measurement error of 1% [24]. After patterning the samples were isotropically wet-etched with HF to remove the silica underlayer controllably to yield a ~2µm high pedestal which supported the microdisk. The samples used here were not annealed, and we note that annealing could induce behaviour different to that of an as-deposited sample [25].

To couple light into the microdisk a silica fibre taper was fabricated using CO₂ laser heating and pulling with motorised stages. The final waist diameter was around 1.5µm with an insertion loss of 0.1dB. The TE waves were launched from an SWS (swept wavelength system - C + L band, dynamic range > 70dB and ~3pm resolution), into the fibre taper which was then side-coupled to the microdisk as shown in Fig. 1. The tapered fiber contacted the rim of the microdisk due to the unavoidable flexing caused by electrostatic forces as the taper was brought close to the edge of the disk. A reference spectrum was first taken without the microdisk and this was subtracted from the in–contact spectrum to get the final resonance spectrum shown in Fig. 2(a). The coupled modes were of higher radial order due to a significant phase mismatch between the fundamental mode of silica taper and the coupled resonant mode of the AMTIR-1 microdisk [18], which can be avoided with a chalcogenide taper based coupling. The radial and azimuthal order for the coupled mode (Q_load = 1.2x10⁵) shown in the inset in Fig. 2(a) are 13 and 362, as shown in Fig. 2(b), as calculated using COMSOL multiphysics.

 

Fig. 1 Experimental setup with an SWS (swept wavelength system). A band edge 633 nm laser source is irradiating (with a spot size ~280 µm) a 50 µm radius microdisk sitting over a 2 µm silica pedestal on a 0.5 mm silicon substrate.

Download Full Size | PDF

 

Fig. 2 (a) Resonance spectrum of a 50 µm radius AMTIR-1 microdisk with several higher order modes coupling is shown. In the inset, a lorentzian fitting of the resonance dip of the higher order mode gives a loaded Q ~1.2x10⁵. (b) Cross-sectional view of the disk with the estimated coupled mode intensity profile of the resonant mode, radially spanning about ~12 µm from the rim of the disk, having radial and azimuthal order 13 and 362 respectively. (c) Optical image of the microdisk evanescently coupling to the silica taper.

Download Full Size | PDF

To study the photoinduced effects, the whole microdisk was illuminated with band-edge light (at 633nm), as shown in Fig. 1. The microdisk was exposed constantly for up to 150 mins, and the position of the resonance monitored periodically, gave rise to the evolution in spectral shift as shown in Fig. 3.

 

Fig. 3 Resonance wavelength shift of the microdisk (measured at 1534.5nm of radial and azimuthal order 13 and 361 respectively), with 633nm light exposure. Red shift (1nm) and blue shift (0.5nm) are observed for high power case 10mW/mm², while mainly red shift (0.4nm) is observed for low power case, 1mW/mm². Note the dashed line is shown to discern the trend.

Download Full Size | PDF

With high power exposure, 10mW/mm², the material shows a rapid red shift, δλ~1nm, saturating around 40 mins which is followed by a slow blue shift, δλ~0.5nm. Whilst with the 1 mW/mm² exposure a red shift of δλ~0.4nm is observed. The observed existence of red and blue shifts can be caused by processes such as a change in index and/or a change in volume since δλ/λ = (δr/r, δn/n), where r and n are the radius and index of the disk. The observed shifts were stable over the monitoring period of 24 hrs.

To identify the dominant processes, we measured the thin film response to the band edge illumination. The film was prepared from high purity element using conventional melt-quenching technique which was used as starting material for thermal evaporation. Prior to deposition the chamber was evacuated to 10⁻⁶ Torr, and then the substrates were irradiated by a 50eV, 1A Ar⁺ beam from an ion gun for 3 minutes to improve adhesion of the AMTIR-1 film to the oxidised silicon substrate. A film of ~960nm thickness was deposited at a rate of 2-5Ǻ/sec. The chemical composition and amorphicity of the film was analysed with EDX and XRD in a 2θ scan and no evidence of crystallization was found [24].

The film was then exposed with a red laser to monitor evolution of refractive index as a function of fluence using spectroscopic reflectometer (SCI FilmTek 4000). The measured data on index is shown for comparison with the microdisk resonance shift in Fig. 4. The thickness could also be monitored and showed a gradual expansion but due to a high measurement error a reliable absolute value was hard to determine.

 

Fig. 4 The measured refractive index change (measured at 1550nm) of the film (a thickness of 960nm was chosen for an accurate measurement by the spectroscopic reflectometer) and resonance shift (as taken from Fig. 3) of the microdisk as a function of fluence is shown. Here we observe a positive shift in the resonance of the microdisk for the same amount of fluence that causes a decrease in the film index. This indicates the expansion of the microdisk dominates the initial positive shift in the microdisk resonance, which is followed by a negative resonance shift mainly governed by the reduction in refractive index. The dashed line is shown to discern the trend. Inset: Stretch exponential function fitting.

Download Full Size | PDF

We observed a gradual drop in the refractive index of the film by up to 2% due to photobleaching effect for the same fluence for which we observed a 1nm red shift in the resonance, as shown in Fig. 4. This suggest that for small exposure the volume expansion in the material likely dominates [26], especially since the freely standing microdisk on a pedestal is unrestrained, compared with the film on a substrate. For an index driven resonance shift, a 0.1% positive shift would be required, which is not consistent with the thin film measurements. In Fig. 3, we see after 2400 seconds of irradiation of the disk the expansion slows down and is followed by a blue shift due to the photobleaching process becoming dominant. Photobleaching in germanium based chalcogenide materials is caused by the creation of coordination defects which involves breaking the existing chemical bonds by light followed by rebonding in a different arrangement as the defects relax. This allows the system to evolve towards a state which represent the lowest potential energy associated with the chemical bonds [24]. In the low power case however, only the initial expansion causes the resonance shift as the power is too low for the photobleaching process [27].We have also calculated the temperature increase in the sample to be < 2 C for the high power case which has a negligible effect on the change in refractive index [28]. For a disorder matrix of glass such as chalcogenide, a phenomenological description of the kinetics of the photobleaching processes can be defined by the stretched exponential function [29],

Where δn, C_b (2.65), τ_b(1.4x10⁹), t and β_b (0.44) are the index shift, initial index, effective response time, illumination time and stretched exponential parameter respectively which were used in the fitting shown in the inset in Fig. 4.

In Table 1 the calculated and measured parameters of the microdisk is presented where the change in band gap energy, E_g was estimated through an empirical relation for amorphous semiconductors (Ravindra’s model, an approximation of Penn’s model was applied) [30].

Table 1. Calculated parameters of the thin film and the microdisk with the measured loaded Q obtained in this study.

View Table

In the experiment the input signal was low enough that we didn’t observe any photoinduced effects due to the resonating modes, which we monitored by comparing several spectra (without an external illumination) taken at different time intervals while the input signal was on. However, we expect the presence of thermal effects as well as photoxidation as the experiments were performed just below the room temperature in air in a dark room [31, 32]. The contact length of the tapered fibre waist and the disk was roughly between 10 and 15 µm, measured from an optical microscope, which would cause a negligible change in the illumination area (0.06-0.09%). This would hardly have any effects considering the higher order mode coupling into the microdisk, whose energy lies well away from the rim, close to the centre of the disk.

3. Conclusion

In summary, we have demonstrated for the first time an efficient way of positive and negative tuning of the resonance of the cavity modes of a microdisk (or any resonator to the best of our knowledge) based on AMTIR-1, with Q ~120,000 (of a higher radial order mode, which was limited by the measuring instrument). We observe that these positive and negative shifts in the resonance are occurring due to the photo-expansion and photo-bleaching effects of the material in accordance with the change in refractive index of the thin film exposed with a band-edge laser. Also, we observe that shift is directly dependent on the applied intensity of the incident laser light, and for the low power exposure of the microdisk we don’t observe a negative shift (Photobleaching effect). In addition, a good fit between the observed refractive index change of the film and calculated stretch exponential model is shown. The proposed technique can be useful for in situ non-contact low power compensation of the fabrication tolerances.