1. Introduction

Direct UV writing is a fabrication technique where a planar sample is scanned under a focused UV laser beam to induce channel waveguides through a permanent refractive index change [1]. The technique does not require photolithography or etching and is becoming more widely adopted in the research community for increasingly advanced applications [2–4]. However, since the details of the refractive index change reactions are not well understood, quantitative knowledge of the UV induced index distribution does not follow directly from the applied processing parameters. This lack of knowledge is a limitation for more advanced applications of UV writing which would benefit significantly from the use of beam propagation methods for detailed modeling.

Germanium doped silica is commonly used for UV writing and contains several optically active defects which emit luminescence bands centered around either 410 nm (blue) or 650 nm (red). Blue luminescence has previously been associated with germanium related oxygen deficient defect centers (GODC’s) [5–7] while red luminescence has been associated with non bridging oxygen hole centers (NBOHC’s) [8,9], however these associations and the molecular structure remains under discussion. The concentration of these defects and the luminescence band intensities change significantly during UV irradiation and are intimately related to the photolytical reactions underlying photosensitivity. Luminescence band imaging has earlier been applied in fiber preforms [9–13], however to our knowledge others have not applied this technique for silica thin films.

In this paper we use a characterization method [14] based on luminescence microscopy to demonstrate that the luminescence profile of UV written waveguides in our samples closely follows the profile of UV induced index changes. This method features higher spatial resolution, better signal-to-noise ratio and is cheaper to implement than conventional index profiling techniques. Due to the detailed nature of the measurements this method can also be very useful for testing candidate theories for photosensitivity.

2. Experiment

The samples used for this work consists of a buffer/core/cladding structure of silica-based glass, fabricated by plasma enhanced chemical vapor deposition (PECVD) on a silicon wafer. After deposition the samples are annealed at 1100 °C in an oxygen-rich atmosphere to remove OH centers. The core layer is 5.5 µm thick and doped with germanium and boron in amounts so that the refractive index matches the buffer and cladding layers [15]. Annealing in oxygen also removes the intrinsic population of GODC’s. The intrinsic photosensitivity is therefore very low and the sample is photosensitized using room temperature deuterium loading at 190 bar prior to UV writing [16]. Channel waveguides are written in the core layer by scanning the sample under a 257 nm continuous wave laser beam focused to a spotsize of 3.1 µm. After UV writing the sample is annealed at 80 °C for 24 hours to outdiffuse residual deuterium. The waveguides typically have low propagation loss (<0.05 dB/cm), excellent mode matching to optical fibers and a peak index step of 10^-3 to 10^-2.

 

Fig. 1. Luminescence microscopy imaging setup. A sample is illuminated with a UV beam producing luminescence which is imaged onto a CCD camera. A filter in the microscope can be used to discriminate between various luminescence bands.

Download Full Size | PDF

 

Fig. 2. Measured blue and red luminescence profiles and refractive index profile of a waveguide written with a beam power of 44 mW and a scan velocity of 110 µm/s. Each image measures 13×13 µm².

Download Full Size | PDF

 

Fig. 3. Point-by-point correlation between blue luminescence intensity and refractive index in the waveguide core (same waveguide as shown in Fig. 2). Within the measurement accuracy, illustrated by error bars on a single datapoint, the data displays a linear corelation indicated by the red fitted line.

Download Full Size | PDF

The luminescence measurement setup [14] is illustrated schematically in Fig. 1. A diced sample with a polished end-facet is mounted vertically under a compound microscope. The microscope objective (×50, N.A.=0.7) projects an image of the end-facet through a blue (392–508 nm passband) or red filter (612–670 nm passband) onto a cooled CCD still camera. The spatial resolution is measured to be 0.4 µm in blue light and 0.6 µm in red light. The UV laser, which is normally used for direct writing, is now used to produce a probe field incident horizontally on the sample. Care has been taken to ensure that the beam profile is uniform across the field of view and to attenuate the beam to a low intensity (0.01 W/cm²), so that it does not alter the luminescence signal during the measurement [17]. An inexpensive UV lamp could also be used for this purpose. By masking the sample with a sheet of aluminum foil, only the uppermost ~50 µm is exposed to the UV field, thereby minimizing the formation of a weak halo of luminescence guided by the waveguide structure. Illuminating shorter sections of the waveguide was difficult and did not yield significant improvements in image quality. The CCD exposure time required to obtain an excellent signal-to-noise ratio is 1–10 seconds.

3. Luminescence and index profiles

The blue and red luminescence profile of a waveguide written with an incident UV power of 44 mW and a scan velocity of 110 µm/s is shown in Fig. 2(A) and Fig. 2(B). The profiles are very similar, this will be discussed further in section 5. The shape of the profiles is trapezoidal-like, with the widest end oriented towards the top cladding and decreasing in width by roughly a factor of two towards the buffer layer. The vertical extent of the luminescence profile corresponds very well to the core layer thickness, which is expected since Ge is instrumental in providing photosensitivity. In some samples, where diffusion of Ge had occurred, the luminescence profile could be seen to extend slightly into the upper cladding layer (Fig. 4, inset). The upper part of the core profile is ~7 µm wide, which is significantly wider than the UV beam diameter, while the width of the lower part is more similar to the UV beam diameter. The internal structure is surprisingly complex, with a bright bar on the top end which is intersected by a darker region in the central top. The sloping sides and the lower edge are ~5–10% brighter than the central area. The area outside the UV written core is very dark, as expected from the reducing anneal applied after the glass deposition. The weak signal in these regions has been determined to arise mainly from scattered luminescence emitted from the UV written waveguides and may be removed by image processing.

 

Fig. 4. Calculated effective index versus the peak index change of a scaled luminescence profile (inset, 13×13 µm²). The scaled luminescence profile reproduces the measured effective index (dotted line) for a peak index change of 0.0128 which is in excellent agreement with the value of 0.0129 measured using the RNF technique.

Download Full Size | PDF

To relate the observed luminescence profiles with the UV induced refractive index changes, the same end-facet was measured by the refracted near field (RNF) technique at a wavelength of 656 nm [18]. The spatial resolution of this measurement is 0.6 µm and the point-to-point measurement noise is ±5×10^-4. The resulting image is shown in Fig. 2(C), where the darkest areas correspond to a refractive index of 1.449 and the brightest to 1.466. The RNF measurement reveals a small index variation intrinsic to the sample (dark band in lowest part of core layer) which may be due to diffusion of dopants during the high temperature annealing. The peak magnitude of the UV induced index change relative to this background distribution is 0.010. The RNF signal-to-noise ratio for the core profile is therefore ±5%, which is roughly 10 times worse than that of the luminescence profiles. The index profile is strikingly similar to the luminescence profiles, containing the same structural features. In the top-center part of the waveguide there is a depression in the index profile, corresponding very well to a similar feature in the luminescence profile. This existence of this feature is in good agreement with the behaviour seen in optical fibers, where a continued exposure to UV light can lead to a reversal in the sign of the index change rate [7,19]. In fibers this only occurs for the largest fluences, which corresponds well with the fact that the depression occurs at the upper center part of the waveguide core where the UV irradiation has been most intense. A point-by-point comparison of the index and luminescence profiles show that the blue profile matches the index slightly better than the red profile, however this may be a measurement artifact due to resolution and sampling effects. A linear relationship between the blue luminescence and the induced index change was confirmed for five different waveguides written with scan velocities ranging from 50 µm/s to 400 µm/s, an example of which is shown in Fig. 3. This correlation seems to be an integral part of the index change reactions in our samples and can be used as an important benchmark when evaluating theoretical models of the UV induced formation process.

 

Fig. 5. Peak intensity of blue (circles) and red (squares) luminescence profile versus the characteristic expsoure time. The data has been corrected for the varying spectral response of the imaging setup. The blue luminescence profile is displayed for three selected waveguides. Each image is 7 µm wide.

Download Full Size | PDF

It should be pointed out that the luminescence-index change correlation observed here is not a general feature of photosensitivity reactions. For instance, if the glass had been treated with a non-reducing anneal after deposition there would be copious amounts of blue luminescence from the non-UV exposed areas outside the waveguide core. A point-by-point comparison with a RNF measurement under these circumstances could easily produce a different result than reported here. In addition, luminescence profiles in our samples are only correlated with UV induced index changes; i.e. index contrast due to compositional and stochiometric variations outside the UV written waveguide core do not appear in the luminescence measurements. However, for waveguides written in oxygen annealed germanosilica samples luminescence microscopy is a simple and accurate method for obtaining the relative index profile.

The luminescence profile can easily be converted to an absolute index change profile if the waveguide effective index, n_eff, and the index, n₀, of the surrounding medium is known. From ellipsometer measurements of monitor samples made in the same PECVD run we have n₀=1.4450 for λ=1.53 µm. The effective index can be measured by UV writing a weak Bragg grating with a known period, Λ, into the waveguide and measuring the Bragg wavelength, λ_B, i.e. n_eff=λ_B/2Λ. The absolute index change profile is obtained by scaling the luminescence profile in a mode field calculation until the calculated effective index matches the measured value. This is shown in Fig. 4, where commercial software was used to calculate the effective index versus the peak value of the index change distribution. The measured effective index is plotted as a dashed line. The two curves intersect for a peak index change Δn=0.0128, which agrees very well with the peak value of 0.0129 measured with the RNF technique. The peak index change obtained in this way consistently matched the RNF value to within ±5% for all sampled waveguides, i.e. similar to the RNF measurement accuracy.

 

Fig. 6. Temporal development of the normalized blue luminescence profile, measured along a horisontal line in the core center. Before ‘turn-on’, at a characteristic exposure time of ~0.015 s, the profile is approximately Gaussian in shape and does not change in width. After turn-on the profile assumes a more rectangular shape and increases gradually in width.

Download Full Size | PDF

4. Luminescence dynamics

The temporal evolution of the refractive index and luminescence during the waveguide formation process can be observed indirectly by measuring luminescence profiles of a series of waveguides written with different scan velocities. For a given scan velocity, v, a characteristic exposure time ω/v can then be defined where ω is the 1/e² diameter of the UV spot. A series of waveguides were fabricated using an incident power of 32 mW and scan velocity ranging from 20 µm/s to 280 µm/s. Figure 5 illustrates the measured peak luminescence band intensity versus the characteristic exposure time. Also shown is the blue luminescence profile for three selected waveguides. Even though the red and blue profiles generally resemble each other spatially, their dynamics are qualitatively different [7]. For characteristic exposure times less than 1×10^-2 s there is essentially no detectable luminescence and no index change (confirmed by visual inspection in a differential interference contrast microscope). Around 1.5×10^-2 s both the blue and red luminescence intensity increases sharply along with the index change, we call this the ‘turn-on’ time. The luminescence profile before ‘turn-on’ is highly elongated in the vertical direction, as illustrated by the profile measured for v=236 µm/s. The horizontal profile is roughly Gaussian in shape with a width close to that of the UV beam, as shown in Fig. 6 for characteristic exposure times of 1.1×10^-2 s and 1.3×10^-2 s. These curves also show that during this phase the profile width remains constant. This profile closely resembles what would be expected for a non-saturated index change reaction that only depended on the local UV intensity or accumulated fluence; i.e. the width corresponds fairly well to the beam diameter and the strength decreases from top to bottom as the UV beam is attenuated down through the core layer. Just after ‘turn-on’, in less than 3 milliseconds, the profile changes to be fairly flat and starts to increase steadily in width. For larger exposure times the blue luminescence band intensity starts to decrease while the red band remains fairly constant; a behavior also seen for fiber illumination [7,19]. During this phase the induced index change increases monotonically, eventually reaching 7.2×10^-3 for the largest sampled exposure time. The profile gradually assumes a trapezoidal shape; for longer exposure times than shown here it would attain a slight depression in the upper, central part, similar to that shown in Fig. 2. The ‘turn-on’ time described here appears to be a general feature for our material system; we have observed it in various samples for many different UV powers and deuterium loading conditions. For larger incident UV powers the ‘turn-on’ time becomes smaller, for example, when going from 32 mW to 51 mW it decreases by a factor of four from 1.5×10^-2 s to 3.5×10^-3 s.

5. Discussion

The fact that the blue and red luminescence profiles are so similar is surprising considering that blue luminescence has most often been associated with positive index changes occurring for low- to moderate fluences while red luminescence is associated with negative index changes occurring for high fluences [7,19]. In our results both luminescence profiles closely follow the induced index profile. The similarity may be related to previous findings [8] where red luminescence was seen to arise from NBOHC’s that were excited via a transfer of energy from UV excited GODC’s. In this situation spatial variations in the induced GODC population would strongly affect the measured distribution of red luminescence, a behavior also seen for fiber preforms [9]. Hence, variations in the observed red luminescence would not be indicative of the actual NBOHC profile.

The complicated core profile morphologies and dynamics reported here strongly suggest that the index change process is not a simple function of local fluence. Instead it seems that the UV illumination can affect the nature and magnitude of the UV/glass interaction. Such an induced change in the interaction seems to take place at the ‘turn-on’ time, where the dynamics and spatial profiles undergo a qualitative change. The fact that this transition occurs only after several milliseconds of UV exposure implies that some reactions in addition to the direct electronic excitations [7] commonly conjured may be important for understanding the index change process around ‘turn-on’. A localized heating by the intense UV beam might explain the observed behavior, if a threshold temperature existed for which the UV/glass interaction changed. Such a mechanism could be the previously described [20,21] thermal activation of molecular hydrogen or deuterium in germanosilica from which GODC’s are formed. In this picture, for characteristic exposure times below the ‘turn-on’ time the local heating does not reach a value sufficient for activation. In addition, the rapid heating and cooling (simple calculations suggest several hundreds of degrees over times of a few milliseconds) due to the UV irradiation may trigger structural changes due to stress relaxation. Since direct UV writing of waveguides involves an average power density ~10³–10⁵ times larger than that used for fiber Bragg grating inscription it is plausible that local heating could play a much more significant role and lead to added complexity. The interplay of UV exposure, absorbtion and heating also make it difficult to define laser power or exposure time as the main driving parameters for the reactions.

6. Conclusion

A simple characterization technique involving microscopic imaging of luminescence profiles of UV written waveguides has been presented. The luminescence microscopy technique is simple to implement and can provide highly detailed information about the formation dynamics and core morphology of UV written waveguides. We have shown that the luminescence profile of UV written waveguides in oxygen annealed germanosilica closely resembles the UV induced refractive index profile, as verified by direct comparison with refracted near-field measurements. Our measurements show that an effective exposure time of several milliseconds is required before strong waveguides are formed. In addition, the core profile undergoes significant changes in shape during the brief interaction time with the writing UV beam. These observations suggest that the index change dynamics are not governed by optical excitation of GODC’s alone and that additional interactions, possibly associated with localized material heating, are required to understand the waveguide formation process.