1. Introduction

Direct optical waveguide writing using focused femtosecond laser pulses is rapidly becoming a reliable technique for manufacturing integrated photonic devices in various materials [1–5]. In particular, the fabrication of optical amplifiers for Metropolitan Area Network (MAN) and Local Area Network (LAN) applications has attracted substantial research interest. At present, the majority of this research is centered on the Erbium Doped Waveguide Amplifier (EDWA), which amplifies signals in the C (1525-1565 nm) and L (1565-1605 nm) telecommunications bands [6–8]. The development of new integrated optical amplifiers operating in different regions of the spectrum is therefore of high interest to increase the useable bandwidth. With this aim in mind, novel materials incorporating other dopants have been under investigation, with Bi-doped glasses showing particular promise.

Recent works published both by us and others have demonstrated broadband emission from Bi-doped glasses spanning approximately 1200 to 1600 nm [10–15]. Significantly, optical amplification has been shown in a Bi-doped bulk glass sample [16], and recently a fiber laser has also been constructed from Bi-doped silicate glass [17], both of which are promising indications that practical devices can be fabricated using the Bi-ion as an active dopant. In order to use Bi-doped glass as a substrate for waveguide amplifiers, a suitable waveguide manufacturing technique must be found. Research in this area is limited thus far however, and a suitable technique is yet to be demonstrated.

In this paper we report on the application of femtosecond laser waveguide inscription to Bi-ion doped silicate glass and demonstrate the fabrication of low insertion loss waveguides using both the single-scan technique and multi-scan technique, which allows waveguides of arbitrary cross-section to be fabricated [8, 18]. The waveguides were manufactured by a diode-pumped cavity-dumped femtosecond Yb-glass laser operating at 1040 nm, close to the two-photon electronic absorption edge of Bi-doped glass [10]. We also demonstrate broadband fluorescence emission from these waveguides under optical pumping.

2. Experimental

2.1. Waveguide fabrication

The precursor composition of the substrate glass used for the waveguide fabrication was 50.9SiO₂ - 25.5Al₂O₃ - 17Li₂O - 5.7ZnO - 0.9Bi₂O₃ (mol %). The fabrication of the substrate was similar to that described extensively in [10]. The laser used to inscribe the waveguides was a Yb:glass mode-locked cavity-dumped laser oscillator emitting 350 fs pulses at a repetition rate of 600 kHz. The wavelength of the laser was 1040 nm. Figure 1 shows a schematic diagram of the waveguide fabrication setup. To fabricate waveguides, the pulse train was focused inside the glass sample to a depth of approximately 150 μm below the sample surface using a ×50, 0.6 NA microscope objective. The sample was then translated in the direction perpendicular to both the laser polarization and propagation direction. Both single-scan and multi-scan waveguide writing were investigated to assess which is the most suitable technique for this substrate material and experimental setup. Since the optimum fabrication parameters for this material were unknown, a large range of pulse energies from 92 to 492 nJ and writing velocities from 3 to 12 mm/s were investigated.

As described in [8,18], the multi-scan writing technique allows the cross sectional size of the waveguide to be controlled by constructing it from many overlapping passes of the fabrication laser, each successive pass having a small incremental translation in the y-axis. The number of scans and scan spacing for our multi-scan waveguides were chosen in order to fabricate a waveguide 8.2 μm wide in the y-axis, the same diameter as a standard telecom single-mode fiber (Corning SMF-28). Using the following well known formula for calculating the Gaussian beam waist of the focussed fabrication beam [9]: $w_y=2M^2.\frac{{\lambda }_{w}n}{\mathit{\pi NA}}$, with M² = 1.15, λ_w = 1040 nm, n ≈ 1.5, and NA = 0.6, we estimate the beam waist of the laser beam inside the substrate material to be 2.0 μm. This value is almost 5 times the maximum scan separation used for multiscan fabrication, therefore ensuring a high degree of overlap between modified regions induced by each successive scan. For each pulse energy and writing velocity, waveguides were fabricated using 40 scans with a 0.21 μm separation, 20 scans with a 0.42 μm separation, and a solitary single scan of the same parameters. This enables us to study the modification building block that is sequentially written in order to build up the waveguide. After fabrication, the sample was diced and polished perpendicular to the x-axis giving a final waveguide length of 1.7 cm.

 

Fig. 1. Experimental arrangement for waveguide fabrication

Download Full Size | PDF

2.2. Waveguide characterization

Insertion Loss (IL) measurements were carried out on all the fabricated waveguides by butt-coupling SMF-28 fibers to the waveguides using index matching liquid and x-y-z translation stages with pitch and yaw adjustment. The IL was then measured by comparing the transmitted signal through the waveguide to a fusion splice of the input and output fibers. It was found that for single scan waveguides in general, the IL decreased as both the velocity and pulse energy increased. As a result, it was found that the optimal single-scan waveguide with the lowest IL was fabricated using a translation velocity of 12 mm/s and a pulse energy of 358 nJ. This waveguide exhibited a total IL of 1.9 dB at 1320 nm. For multi-scan fabricated waveguides, the trend was reversed, with the IL increasing as the pulse energy and translation velocity were increased. The multi-scan waveguide with the lowest IL was fabricated using a velocity of 3 mm/s, a pulse energy of 92 nJ, and 40 fabrication scans. This waveguide exhibited a total insertion loss of 2.0 dB at 1320 nm. The possible error in the IL measurements is estimated to be ≈ 0.3 dB due to alignment accuracy. Due to the nature of the measurement, the IL cannot be higher than measured.

The waveguide-fiber coupling loss (CL) was measured at 1320 nm by comparing the signal throughput when butt-coupling the output facet of the waveguide to either an SMF-28 or highly multimode fiber. Since the CL between the waveguide and the multimode fiber is assumed to be zero, the CL to SMF-28 can be measured in this manner. The CL’s to both the optimal single-scan and multi-scan waveguides were measured to be within error of each other at 0.6 dB/facet. The error in the CL measurement is estimated to be ≈ ± 0.2 dB due to alignment accuracy. The numerical aperture (NA) of the fabricated waveguides was calculated using the well known relationship for the NA of an optical waveguide, NA = √2n∆n , where n is the refractive index of the cladding material, and ∆n is the absolute difference between the core and cladding refractive indices. For our purposes, ∆n was measured using a commercially available refractive index profiler (as discussed later) and found to be 4.3 × 10^-3 and 2.6 × 10^-3 at 658 nm for the optimum single and multiscan waveguides respectively. The refractive index of the substrate material was measured using a commercially available prism coupler (Metricon 2010) and found to be 1.545 ± 0.001 at 633 nm. Using this experimentally determined refractive index data the NA of the fabricated waveguides was calculated to be 0.12 and 0.09 for the optimal single-scan and multi-scan waveguides respectively. The NA of the multimode fiber for comparison was 0.275 ± 0.015. The origin of the CL is the difference in the MFD (Mode Field Diameter) of the waveguide and fiber modes.

The propagation losses were estimated by subtracting the total measured CL’s from the total measured insertion loss. The propagation losses for both the optimal single and multi-scan waveguides were again within error of each other at 0.4 dB/cm. The error is estimated to be ≈ ± 0.2 dB/cm. Such values of propagation losses are comparable to the state of the art of femtosecond pulse inscribed waveguides and can be attributed partly to glass inhomogeneities and partly to non-uniformity in the sample translation movement [2,3,6]. These values of propagation losses also includes any material absorption. By measuring the bulk sample transmission spectrum using a Bruker IFS 66V/S FT-IR spectrometer, it was observed that the transmission was relatively flat between 1 μm and 2 μm, but increased sharply below 1 μm due to the tail of the UVband edge, as shown in [10]. Due to the flatness of the transmission between 1 μm and 2 μm we conclude that the loss due to ground state absorption in the 1.3 μm region is negligible.

The Polarization dependant loss (PDL) was measured for the optimal multi-scan and single-scan waveguides, using a polarization controller (HP 11896A) and a linearly polarized 1320 nm laser source. The PDL was measured to be 0.4 dB and 0.6 dB for the optimal single-scan and multi-scan waveguides respectively. The PDL values are small enough to indicate that the waveguide birefringence is insignificant.

The waveguide facets were viewed using a calibrated optical microscope operating in transmission mode. The images were then captured using a CCD camera together with a computer and frame grabber. The guiding properties of the waveguides were then investigated by imaging the waveguide output facet with an Electrophysics-7290A IR Vidicon camera, while coupling 1320 nm light into the waveguide from the opposite end using fiber butt-coupling. Imaging was performed using a ×40, 0.65 NA microscope objective. The optimal single-scan waveguide (lowest insertion-loss) had a 1/e² mode field diameter of 10.9 μm. The mode field of the optimal multi-scan waveguide had a slight square-like resemblance with a MFD of 11.8 μm. Figures 2(a) and 2(b) show the facet images for the optimal single-scan and multi-scan waveguides respectively. Also shown in Fig. 2(e) and 2(f) are the captured near-field mode images at 1320 nm. The MFD’s of the fabricated waveguides compare favorably with the 9.2 μm MFD of SMF-28 fiber at 1320 nm. The errors in the measured MFD’s were estimated to be ≈ ± 0.9 μm due to inaccuracies in the camera focusing. The theoretical coupling losses, estimated from the overlap integral between fiber and waveguide modes, are of about 0.2 dB/facet, not far from the previously measured value.

 

Fig. 2. a) Facet image of the optimal single-scan WG, b) Facet image of the the optimal multi-scan WG, c) Refractive index profile of the optimal single-scan WG, d) Refractive index profile of the optimal multi-scan WG, e) Near-field mode image of the optimal single-scan WG, f) Near-field mode image of the optimal multi-scan WG

Download Full Size | PDF

Refractive index measurements were carried out using a Near-Field Profilometer (Rinck Elektronik) at 658 nm. Images of the 2 dimensional profiles of the optimal single-scan and multi-scan waveguides are shown respectively in Fig. 2(c) and 2(d). Interestingly, the maximum refractive index change was much greater for the optimal single-scan (∆n = 4.3×10^-3) than the multi-scan (∆n = 2.6×10^-3). This difference in magnitude may be due to the large difference in fabrication pulse energy between the optimal single-scan waveguide (358 nJ) and optimal multi-scan waveguide (92nJ). The index contrast for the optimal multi-scan waveguide appears to be saturated, as the insertion loss for the waveguide fabricated using the same pulse energy and translation speed, but with only 20 fabrication scans is merely 0.3 dB greater. We suggest that the small decrease in insertion loss by using 40-scans may be the result of higher traversal homogeneity. This may be the result of either or both increased overlap between successive modified regions or a reduction in scattering centres as a result of increased annealing by successive scans. Also, we suggest that the higher pulse energy used during the single-scan fabrication induces a different index modification regime, whereby the high pulse energy can cause greater modification effects such as considerably stronger thermal gradients.

Another intriguing point is that the optimal multi-scan shows a peculiar shaped profile, with two lobes. The refractive index profile was then measured for the building block of this multi-scan waveguide. Figure 3 shows the obtained index profile, the reason for this double structure is presently unclear and under investigation. This structure does not appear to negatively influence the guided mode profile (see Fig. 2(f)) and thus device performance. The magnitude of the index contrast for the building block is measured to be relatively small (∆n = 9×10^-4), however due to convolution effects of the point spread function on the profilometer, the index contrast may well be greater than measured. The dimensions of the building block are approximately 10 μm in the z-axis, and 2.5 μm in the y-axis. The level of asymmetry in this building block is much greater than for the optimal single-scan waveguide, further evidence of a different index modification mechanism.

 

Fig. 3. Refractive index profile of the building block scan used to fabricate the optimal multi-scan WG

Download Full Size | PDF

To test the suitability of the fabricated waveguides as broadband near-IR optical amplifiers, we pumped them at either 810 nm with up to 60 mW of power, or 980 nm with up to 220 mW of power; in both cases, direct fiber-waveguide butt-coupling was used. Upon pumping, a clear orange-yellow upconversion , the mechanism of which is still unclear at this time, was visible along the the length of the waveguide. Figure 4 shows a photograph of the lowest insertion loss multi-scan waveguide pumped from the right hand side with 220 mW at 980 nm. From Fig. 4, it is clear that the waveguide is of high quality, as the upconversion emission is clean and homogeneous along the length of the waveguide.

Fluorescence characterization experiments were conducted by optically pumping the waveguide under test with either the 980 nm light or the 810 nm light, using direct fiber-waveguide butt-coupling. The generated fluorescence was then collected at the opposite end of the waveguide using an SMF-28 fiber directly butt-coupled to the waveguide facet. The collected light was then collimated and passed through two silicon pump blocking filters and refocused back into an SMF-28 fiber, which was then connected to an Advantest Q8384 optical spectrum analyzer with a resolution of 0.5 nm. The two pump blocking filters were necessary to completely block the pump light. Figure 5 shows the fluorescence spectra for the optimal single-scan waveguide when pumped with 220 mW of 980 nm light or 60 mW of 810 nm light. When the waveguide was pumped with 220 mW of 980 nm light, broad fluorescence centered at 1320 nm was detected, with a bandwidth (FWHM) of approximately 500 nm. When pumped with 60 mW of 810 nm light the fluorescence spectrum is narrower, with a bandwidth (FWHM) of approximately 200 nm. The magnitude of the fluorescence when pumped with 60 mW of 810nm is also approximately 5 dB’s higher than when pumped with 220 mW of 980 nm, thus indicating that 810 nm may be the more desirable pump wavelength. It is also worth noting that the bismuth fiber laser constructed by Dianov et al in [17] was pumped using an Nd:YAG laser operating at 1064 nm. As such, other pump wavelengths may be more efficient.

 

Fig. 4. Upconversion observed when pumping the optimal multi-scan waveguide with 220 mW at 980 nm. The pump light enters the sample from the right hand side.

Download Full Size | PDF

 

Fig. 5. Fluorescence spectra of the optimal single-scan waveguide pumped at 810 nm and 980 nm.

Download Full Size | PDF

Currently, the origin of the fluorescence broadening is not fully understood, and is the topic of considerable debate. The shape, width and spectral position of the fluorescence spectra we have measured are all considerably different to that observed in other work. The peak position of the fluorescence has also been shown to change for different pump wavelengths [12, 14, 15].

For example, Suzuki et al in [15] observed a shift in the peak from 1250 nm when pumped at 800 nm to 1100 nm when pumped at 974 nm, and a decrease in the bandwidth from 440 nm under 800 nm pumping, to 290 nm under 974 nm pumping. In stark contrast to this report, our fluorescence spectra are unchanged in the peak position of 1320 nm for both 810 nm and 980 nm pump wavelengths, and we have observed an increase in bandwidth from 200 nm under 810 nm pumping, to 500 nm under 980 nm pumping.

3. Conclusion

In this paper we have demonstrated the fabrication of high quality, low insertion loss optical waveguides in a Bi-doped silicate glass using both single-scan and multi-scan femtosecond laser inscription. Large refractive index changes have been observed with values up to ∆n = 4.3×10^-3. Broadband fluorescence with a bandwidth (FWHM) of up to 500 nm centered on 1.3 μm has been measured from these waveguides under optical pumping at 810 and 980 nm.

The future prospects for creating Bi-doped waveguide amplifiers using femtosecond waveguide inscription are, we feel, promising given the successful application of this technique to fabricating Er-doped waveguide amplifiers [6]. Demonstrations such as net gain in a bulk sample and a fiber laser have already proven the potential of Bi-doped glass as a gain material for near infrared wavelengths [16, 17].

This work was funded the UK Engineering and Physical Sciences Research Council (EP-SRC). N.D. Psaila, R.R. Thomson and A.K. Kar acknowledge support from the European Community Access to Research Infrastructure action, contract RII3-CT-2003-506350 (Centre for Ultrafast Science and Biomedical Optics).