1. Introduction

Femtosecond laser direct writing of waveguide devices is a significant enabling technology in the fields of sensing, communications, defence and integrated quantum optics [1, 2]. Waveguides can be directly inscribed inside many glass materials including the silicates [3] and phosphates [4, 5, 6]. By doping the glass with rare earth ions such as erbium (Er) and ytterbium (Yb), gain structures can be realised [7]. In order to make a laser however, optical feedback provided by external mirrors or gratings is required. We recently demonstrated distributed feedback (DFB) monolithic waveguide lasers (WGLs) fabricated in Yb:phosphate [8] by directly writing a waveguide and grating in a single processing step. In this case we obtained output powers of 100 mW and optical conversion efficiencies of 17% with stable laser operation for periods of approximately 50 hours. However, initial changes in output power and performance were observed in this architecture before the period of stable operation. In the case of the Er, Yb WGL structure [9], laser operation was sustained for only one hour. We believe that these changes in operation are due to annealing of the waveguide Bragg grating (WBG) structure akin to that observed in photosensitive fibres [10]. To better understand this phenomenon and in order to create WGLs with long term stability we investigate the effects of photo- and thermal annealing of ultrafast laser written waveguide and WBG structures.

2. Background

Thermal annealing properties of fibre Bragg gratings (FBGs) have been extensively studied. The reflectivity of a UV written FBG decays rapidly with a decreasing rate over time, when exposed to raised temperatures. The decay follows a stretched exponential or power law function. It has been suggested that this decay is due to the removal of colour centres that were generated by the UV writing beam [10]. In order to produce FBGs with stable long term characteristics pre-annealing can be used. By subjecting a FBG to temperatures around 200 °C colour centres that contribute to the rapidly erasable component of the refractive index can be removed. After pre-annealing however, the grating still continues to decay albeit with a much reduced rate.

The colour centres generated during the UV FBG writing process can also cause an increase in the transmission loss of the fibre in the visible, with a tail extending into the near IR region. This photodarkening phenomenon can result in absorption losses of up to 10000 times that of pristine fibre. This attenuation can be significantly reduced upon exposure of the fibre to visible wavelengths from 400–650 nm. Photobleaching of the colour centres and the corresponding decay in attenuation follows a stretched exponential or power law of a similar form to that observed when thermal annealing [11]. In glasses doped with rare earth ions, such as Yb, the colour centre dynamics become more complicated because the rare earth ion provides a pathway to both create and annihilate colour centres. In Yb-doped silica fibre lasers it has been reported that photodarkening results from a stepwise process starting with the simultaneous de-excitation of 7 inverted Yb ions, the generation of a UV photon, its absorption by the glass lattice and the breaking of a Si-O bond to form a colour centre [12]. Whether the colour centres are created during grating writing or laser operation, it is known that they can in part be bleached by exposure to visible wavelengths. For example, Chávez et al. showed that in a Yb-doped silica fibre the photodarkening losses resulting from exposure to 977 nm pump light could be halved by irradiating the fibre with 1 mW at 543 nm for 1 hour [13].

Femtosecond laser created colour centres in phosphate glass hosts have been shown to be photobleachable in a similar manner to the photobleaching processes observed in UV written doped silica. Chan et al. [14, 15] observed that prolonged exposure to 488 nm light could erase phosphorus-oxygen-hole-centres (POHC) formed during exposure to femtosecond radiation. The POHC has broad overlapping absorption bands at 234 nm, 400 nm, 496 nm and 564 nm [16].

In the work presented here we use femtosecond laser inscribed WBGs as sensors to characterise the photo- and thermal annealing properties of femtosecond laser written waveguides in Kigre ‘QX’ phosphate glass. This glass is different from the Schott ‘IOG-1’ glass used by Chan et al. [15], firstly as the femtosecond modified regions result in positive index changes (typically 2×10^-3) and secondly, as the glass used here is doped with Yb. Our work has found that the grating depth (typically 3×10^-4) reduced over time, the Bragg wavelength blue shifted, and the mode field diameter (MFD) of the WBG increased. Both photo- and thermal annealing result in a lower effective refractive index of the guiding region, however during normal waveguide laser operation, as described in [8], we show that only photo-annealing takes place. We show that the photo-annealing is a result of the visible light generated in the pumped Yb waveguide region caused by co-operative luminescence. It is not directly related to the optical fields of either the pump (at 976 nm) or generated laser light at 1033 nm.

3. Waveguide device fabrication/general characterisation

The femtosecond laser direct-write technique was used to create the waveguide and the WBG simultaneously and in a single processing step [17]. The laser used to fabricate this device was a circularly polarised, 1 kHz repetition rate, 120 fs pulse length, 800 nm regeneratively amplified Ti:sapphire laser that was focused into the glass sample using a 20× (0.45 NA) microscope objective. Waveguide devices were written 170 μm below the surface of 3.4 mm long Kigre ‘QX’ phosphate glass samples with nominal doping concentrations (by weight) of 5% Yb, 9% Yb as well as an undoped sample. WBGs were written in the C-band, outside of the Yb absorption/emission bands which enabled probing using a high resolution swept C-band source/detector. To write a first-order WBG near 1540 nm, the glass sample was translated at 25 μm/s through the focused writing beam which was 100% intensity modulated at approximately 50 Hz and with a 50:50 mark space ratio to create the required ≈500 nm period refractive index perturbation. The writing power was varied from 0.7 mW to 1.5 mW resulting in WBGs with varying grating strengths and effective MFDs. Further details pertaining to device fabrication can be found in [8, 18, 19].

The experimental arrangement for device annealing and characterisation is shown in Fig. 1(a). On the input side a WDM allowed pumping of the waveguide device at 976 nm whilst probing with a swept C-band source. A circulator on the input side allowed reflection measurements to be taken while a WDM on the output side provided the ability to separate the transmitted C-band light and transmitted 976 nm pump light. To ‘pump’ the device at 1064 nm or 532 nm a 99/1 coupler was inserted between the input WDM and the waveguide device such that the incident ‘pump’ light was maximised while still coupling sufficient C-band light to measure the grating strength.

 

Fig. 1. (a) Experimental setup for annealing waveguide devices. WDM-Wavelength division multiplexor, T-Transmission, R-Reflection. Components inside the dotted blue box are removed from the setup for 976 nm ‘pumping’. (b) Sketch of the relative refractive index of the bulk, a waveguide and the grating contrast of a WBG. (c) Example of transmission data through a WBG before and after photo-annealing.

Download Full Size | PDF

Analysis of each WBG was made in terms of its transmitted and reflected light. From measurements of the wavelength of the reflection peak, λ_B, and knowing the period of the grating, Λ, the underlying effective index, n _eff, of the sample was calculated (Fig. 1(b)). This corresponds to the index of the guided mode, Eq. (1).

To calculate the grating strength, κL, we measured the depth of the grating from the transmission data (example shown in Fig. 1(c)) and used the relationship given in Eq. (2), where R is the grating reflectivity, κ, the coupling coefficient and L the grating length.

The grating index contrast, ∆n _grating, can be calculated knowing the coupling coefficient, κ, and the Bragg wavelength, λ_B, using Eq. (3).

By monitoring changes in both the WBG reflectivity and Bragg wavelength during photo- or thermal annealing, any change in the refractive index contrast of the grating or the effective index of the guided mode can be measured. An example of such changes can clearly be seen in Fig. 1(c) which shows C-band transmission spectra of a WBG that has been photo-annealed for 2 hours using the 976 nm pump source. An explanation of these changes will follow in later sections. It should be noted that such a shift is not due to thermal loading (an increase in sample temperature, and hence Bragg wavelength, of the device during the pumping process) or thermal annealing (a permanent index change by changing the structure of the glass). The absence of thermal loading was ensured by allowing the WBG to cool for several minutes after pumping and the discounting of thermal annealing during pumping experiments was deduced by using a similar WBG as a temperature sensor. In this case by measuring the shift in Bragg grating wavelength in a temperature controlled oven, we found a tuning rate of 11.9 pm/°C. This value accounts for both the change in dn/dT and the coefficient of linear thermal expansion and is comparable (10.4 pm/°C) to that reported by Zhang et al. [20] who used fused silica WBGs as temperature sensors. When pumping WBGs at 976 nm we find the maximum wavelength shift to be ≈ 150 pm (Fig. 1(c)) which from our calibration measurements suggests a pump induced temperature rise of only 12.6 °C. In Section 5 we show temperatures in excess of 70 °C are needed to affect any refractive index change to femtosecond written WBGs and so we can discount thermal annealing as the reason behind the decay of our photo irradiated gratings.

4. Photo-annealing

Photo-annealing experiments were performed on a series of identical ≈3.4 mm long WBG samples as a function of incident power, doping concentration and for a range of wavelengths. Initially we illuminated a 9% Yb-doped phosphate glass sample at 976 nm (pumping the Yb absorption) with a range of different pump powers and recorded the grating depth of the WBG over time. The results are shown in Fig. 2. For this experiment the 976 nm pump powers were all above the pump saturation intensity of this material, which at 974 nm is approximately 7–10 kWcm^-2. At powers above the saturation intensity increasing the pump power increases the optical field strength but does not significantly increase the Yb inversion density. From Fig. 2 we see that the refractive index contrast of the grating initially decays rapidly followed by a slower decay rate with time. The results show that despite a 4 fold increase in incident power at 976 nm there is only a marginal change in decay rate or depth. This is analogous with photodarkening in Yb-doped silica fibres where the photodarkening rate depends on the Yb inversion density and material specific Yb ion clustering rather than directly on the optical field strength. However, in the work presented here, photobleaching occurs due to the visible light generated from co-operative luminescence, see inset Fig. 2, which has an intensity dependent on the inversion density, which in turn is clamped at the pump saturation intensity.

To rule out the possibility that the observed photo-annealing is directly due to the 976 nm field we pumped new WBGs at 1064 nm (which was close to laser wavelength of an operating DFB WGL) and at 532 nm (close to the wavelength of the co-operative luminescence emission from the pumped Yb waveguide). As shown in Fig. 3(a) irradiating the WBG at 1064 nm has no effect on the grating contrast while irradiating the WBG at 532 nm results in a similar decay rate to that when pumping at 976 nm. From these results we can conclude that photo-annealing is not related to the optical field at 976 nm directly as the 1064 nm light has no effect. Pumping the WBG at 976 nm however, does result in the generation of significant blue-green light due to the co-operative luminescence effect and as we see the grating decay when illuminated at 532 nm we thus believe that it is the visible light generated from co-operative luminescence which photo-anneals the Yb-doped WBG.

 

Fig. 2. Photo-annealing WBGs (written in 9% Yb-doped phosphate) with various 976 nm pump powers. Inset shows the blue-green co-operative luminescence along the WBG while pumping at 976 nm.

Download Full Size | PDF

To confirm our predictions we prepared another WBG sample made from an undoped Kigre ‘QX’ phosphate glass sample. In this case there is no absorption at 976 nm and so we expect these samples to pass this wavelength un-attenuated. Indeed by irradiating different gratings within this sample at 976 nm and 532 nm we see that this time there is no effect on the grating using the 976 nm light while annealing occurs when irradiating at 532 nm as shown in Fig. 3(b). It should be noted that the pump powers shown in Fig. 3 are those incident on the WBGs. The coupled pump powers into the WBGs would be significantly lower than those quoted.

 

Fig. 3. Photo-annealing at various wavelengths of WBGs written in (a) 9% Yb-doped phosphate and (b) undoped phosphate.

Download Full Size | PDF

The results shown in Fig. 3 clearly demonstrate that WBG annealing is both wavelength and doping dependent. For both the doped and undoped samples photo-annealing arises due to the absorption of visible light, resulting in the annihilation of colour centres induced during the femtosecond inscription process. In terms of our colour centre model this is not surprising as the POHC has broad and overlapping absorptions at 234 nm, 400 nm, 496 nm and 564 nm. It is also interesting to note that when irradiating WBGs at 532 nm we see that the annealing rate and the total annealing depth increase with irradiated power (Fig. 4). At maximum irradiated power 40% of the WBG erased after 2 hours of exposure. Clearly the WBGs fabricated here would not be efficient long term if fabricated for use at this wavelength. More importantly, however, we see that the contribution short term colour centres make to the grating depth (min 40%) is far higher than that observed in UV written fibre lasers. For example Askins and Putnam [11] observed a 20% decrease in Bragg reflectivity with an 80% drop in colour centre absorption after exposure to 300 mW at 650 nm for 2 hours.

 

Fig. 4. Photo-annealing WBGs (written in 9% Yb-doped phosphate) with various 532 nm ‘pump’ powers.

Download Full Size | PDF

5. Thermal annealing

Thermal annealing is the basis for accelerated aging of FBGs used in telecommunications. This effect was not observed by our group during characterisation of our DFB waveguide lasers. As detailed in Section 3 this is not unexpected as the average WBG temperature obtained when irradiating the 9% Yb phosphate glass at 976 nm using 250 mW of incident pump power, and without lasing, is only approximately 12 °C above ambient. In order to characterise the thermal stability of WBGs we incrementally heated a 3.4 mm, 9% Yb WBG sample in a temperature controlled tube furnace. To ensure consistent measurement we pigtailed input and output fibres to the fabricated WBG using UV curing epoxy. This was done to ensure that we probed the same region of the waveguide at each temperature and also so that we could probe the WBG while at elevated temperatures in real time. At each temperature (increasing in 10 °C increments from ambient) we let the grating ‘anneal’ for approximately 30 minutes after which the decay rate would approximately ‘stabilise’. The results of this experiment are shown in Fig. 5. In terms of our colour centre model ‘stabilisation’ suggests that at any given temperature above ≈70 °C we can only anneal those centres which have activation energies at or less than the lattice temperature. It should be noted that the grating reflectivity at the end of the experiment was 6.5% which was less than 1/10^th of its initial reflectivity (95.7%). The results demonstrate that the WBGs used in our DFB waveguide lasers are thermally robust up to 70 °C and could be ‘aged’ in the same way as FBGs used in the telecommunications industry to provide long term stability.

 

Fig. 5. (a) Thermal annealing of WBGs (written in 9% Yb-doped phosphate). At each temperature there are multiple data points taken over a 30 minute period. This is shown for temperatures of 145 °C and 265 °C in (b).

Download Full Size | PDF

6. White light transmission spectra/discussion

In Sections 4 and 5 we show that both photobleaching and thermal annealing can affect the performance of femtosecond laser written WBGs in doped and undoped Kigre ‘QX’ phosphate glass. We believe that the underlying reason for this behaviour is the annihilation of colour centres formed during the femtosecond laser writing process.

In this section we look more closely at the absorption spectra of the bulk versus modified regions in both waveguides and WBGs in order to determine the extent and strength of the visible absorption resulting from the induced colour centres. In particular, we measured white light transmission spectra before and after both photo- and thermal annealing of the structures. Measurement of the absorption spectra of the bulk is relatively simple and was done using a high resolution spectrophotometer. Measuring the waveguide absorption at visible wavelengths, and in particular to see changes during either photo- or thermal annealing, requires longer sample lengths than those used previously. Consequently we fabricated 70 mm long WBGs, with a Bragg wavelength at 1534 nm, in a 5% Yb-doped Kigre ‘QX’ phosphate glass sample using a writing power of 0.8 mW.

White light transmission spectra including the bulk glass, a WBG, a photobleached WBG and the same WBG after thermal annealing at 150 °C and 250 °C are shown in Fig. 6(a). The WBG clearly has a much reduced visible transmission, over the bulk, with the ‘UV’ cut-off extending past 500 nm. At our probe wavelength at 532 nm the WBG absorbs 10 times that of the bulk. This change in visible transmission is consistent with colour centre formation either obtained during UV grating writing or during femtosecond inscription. It is also worth noting the 2^nd order grating effect at 767 nm.

To see if these changes to the glass transmission could be reversed in a similar fashion to colour centre bleaching in FBGs, we irradiated the 70 mm long sample with 532 nm light for approximately 2 hours. The maximum incident green power was approximately 200 mW although the coupled power into the waveguide was approximately 50 mW. The results confirm our previous observation in that the grating strength is reduced by 40% during the photo-annealing process, however, the visible transmission is substantially restored suggesting that the colour centres induced during the femtosecond writing process are no longer active. It is worth noting (shown in Fig. 6(b)) that the MFD of the guided mode has increased from 14 μm to 19 μm. This has multiple implications. Firstly, it reduces the coupling into the WBG which in turn makes scaling of the transmission spectra difficult but also demonstrates that whilst the grating strength is reduced so too is the n _eff of the guide. In terms of designing/constructing waveguide lasers this implies that devices would need to be made with a MFD considerably smaller in order to offset annealing effects or look to materials or writing conditions that do not result in colour centre formation [21]. In Fig. 6(a) we have also plotted the transmission spectra after thermal annealing at 150 °C and 250 °C. These results were taken using the already photobleached sample. Whilst there is little difference to the effective UV cut-off wavelength there is a reduced overall transmission which is biased towards longer wavelengths. This arises, we believe, due to the increasing MFD (MFD at 150 °C ≈21 μm, at 250 °C ≈25 μm) which not only reduces the coupling but affects the longer wavelength disproportionately due to the mode extending further out of the core at these wavelengths and hence not being gathered by the collecting fibre. This could be clearly seen by putting a small bend in the coupling fibres and observing reduced transmission at the longer wavelengths only. It is interesting to note that if we do not photobleach our sample we obtain an equivalent effect in terms of increased MFD and reduced grating depth by annealing the WBG at approximately 125 °C. This allows us to thermally anneal waveguide devices so that they are photo stable over long periods of time assuming a maximum incident visible photon flux. We repeated these experiments at lower writing powers and with a waveguide only and found that the response of the waveguide to the photo- and thermal annealing effects was identical to the WBG sample.

 

Fig. 6. (a) White light transmission through a 70 mm long WBG (written in 5% Yb-doped phosphate) subject to photo- and thermal annealing. Note that the initial WBG shows a 2nd order grating effect at 767 nm. (b) MFD of the guided mode at 800 nm, shown in white, increases post annealing.

Download Full Size | PDF

Colour centre formation and photobleaching have been observed previously by Chan et al. in both Schott ‘IOG-1’ and phosphate (P₂O₅, La₂O₃) glass [14] with the POHC the most likely colour centre candidate. Similarly, colour centre formation has also been observed in borosilicate and alkali silicate glasses [22]. Clearly the presence of colour centres forms an important part of waveguide fabrication in many glass hosts and whilst these colour centres can tolerate temperatures up to 70 °C in Kigre ‘QX’ phosphate glass they are susceptible to photobleaching. In terms of building infrared DFB lasers [9, 8] the observed levels of photobleaching arising from co-operative luminescence can be factored into grating design such that for long term DFB laser operation, longer gratings and lower doping concentrations glasses would be used. It is important to note that the total refractive index contrast obtained in the waveguides characterised here is not solely due to colour centre formation. While the grating contrast is mostly erased after annealing at 250 °C (∆n = 1 × 10^-5), the n _eff of the waveguide is still 10^-3 (with a MFD of 25 μm). For the Yb-doped phosphate glass samples used here we believe the contribution colour centres make to the overall induced index change, n _eff, is approximately 15% (max, ∆n =3×10^-4, n ^eff =2×10^-3). The remainder is mostly likely to be due to densification which has been reported when femtosecond processing phosphate glass in the heat accumulation regime [21].

7. Conclusion

We have observed colour centre formation in waveguides written into Kigre ‘QX’ phosphate glass using infrared femtosecond laser inscription. Using waveguide Bragg gratings as a diagnostic tool we demonstrated that in the case of doped phosphate, photobleaching of induced colour centres arises from the visible light generated from co-operative luminescence between neighbouring Yb ions. We also show that thermal annealing is possible above 70 °C although would not typically be seen in a waveguide laser such as those reported in [8]. The results reported here suggest that colour centre formation forms an important part of femtosecond waveguide device fabrication in many glasses and that ultrafast laser processed samples need to be ‘aged’ in a similar way to FBGs used in the telecommunications industry in order to obtain long lifetime stable devices.