1. Introduction

Photodarkening in ytterbium-doped fibers (YDFs) appears as a time-dependent broadband loss centered at the visible wavelengths with the tail of the loss extending to near-infrared (NIR) wavelengths [1–4]. Inversion-dependent kinetics of this color center formation process have been studied in detail, resulting in derivation of the photodarkening rate dependence on the number density of excited ytterbium ions [4–8]. The reverse process of annealing or bleaching of photodarkening has also been studied, but to a much lesser extent [2,5,9,10]. Bleaching by brief exposure to 355 nm UV light was shown to recover the sample absorption to nearly the pre-photodarkened state, with no apparent effect on the slope efficiency of the fiber [2]. Thermal bleaching of photodarkening was demonstrated by slowly ramping up the temperature while monitoring the sample transmission/absorption in situ [5,9]. Complete recovery of the (room temperature) absorption spectrum between 450 and 1600 nm at temperature of ~500 °C was demonstrated, with the associated activation energy estimated as 0.07 eV [9].

Thermodynamic analysis of the photodarkening process has not yet been exploited to a significant extent, even though it provides several advantages over kinetic analysis. Thermodynamic analysis can be used to identify different mechanisms and their equilibrium states by their activation temperature or energy. In the case of kinetic data, multiple photochemical reaction mechanisms result in a multi-exponential decay of absorption as a function of time, which is difficult to analyze and cannot be tied to specific physical mechanisms. Furthermore, thermodynamic data can be put on an absolute scale (such as eV) without detailed information about fiber parameters, such as rare-earth dopant density.

In this paper, we study thermal bleaching of photodarkening-induced losses by subjecting an YDF sample to multiple thermal ramp-up/down cycles. We show the existence of a post-irradiation heat-driven darkening mechanism, which is operative before recovery of the sample absorption coefficient initiates. We believe this darkening to be evidence of an intermediate energy state in the NIR (e.g. 915 nm) excitation route. We also show that the induced color center losses exhibit temperature-dependent spectral broadening, which may distort the absorption decay curves recorded in the course of photodarkening rate measurements.

2. Experimental

The investigated fiber is a technologically important, commercial 20-µm core diameter large-mode-area (LMA) fiber. This fiber has 0.5 dB/m cladding absorption at 915 nm and core numerical aperture of 0.06. The cladding diameter is 400 µm. For the purpose of this study, the cladding diameter was reduced (by etching) to ~125 µm. Although the etching procedure leaves the cladding surface somewhat rough, and therefore increases pump light loss due to scattering (by ~5%), it is not expected to have any effect on the properties of the core or on the thermal cycling process. To characterize photodarkening, the sample fiber absorption coefficient change, Δα (in units of dB/m), normalized to the pristine fiber absorption, is monitored in situ during cladding pumping at 915 nm. This measurement method is similar to previously reported approaches [6,7] and is described in detail in [10]. A lock-in amplifier is used to detect a modulated signal at 600 nm or 670 nm that passes through the core of the fiber. A reference arm is added to monitor changes in the source power, allowing very long duration measurements where drift of the source power would otherwise be a concern. A 2 cm long sample fiber is spliced between two matching passive fibers and held in air, uncoated. For thermal cycling experiments, a miniature furnace (MHI Inc model FIBHEAT) is moved over the sample fiber so that the fiber passes through a narrow slit into the hot zone of the furnace. The length of the hot zone is ~2.2 cm. A K-type thermocouple is placed next to the fiber within the furnace to measure the temperature during the cycling experiments. The fiber temperature is ramped up/down as a triangle wave over a time of several tens of minutes with the rate of temperature increase/decrease (dT/dt) ranging between 1 and 3 °C/s. The temperature distribution of the heated fiber sample is estimated to be uniform to within 30 °C. The spectral shape of the induced changes, Δα(λ), is also recorded by measuring the full transmitted source spectrum (without a bandpass filter) at relevant points during the course of the experiment and by normalizing this spectrum to the transmission spectrum of the pristine fiber. As discussed in previous work [10], a key feature of such thermal bleaching measurements is the ability to fully recover a given fiber sample to its pre-photodarkened state, thus allowing one to work on the same undisturbed sample through a series of thermodynamic/kinetic measurements.

3. Results

Figure 1(a) plots the photodarkening-induced absorption coefficient change Δα versus time with ~15 W of launched pump power at 915 nm. Actually, –Δα is plotted versus time (and later, versus temperature); this form was chosen to consistently indicate degradation (i.e. loss of transmission due to photodarkening) by downward change, and recovery (bleaching) by upward change. Data was measured separately for 600 nm and 670 nm wavelengths to compare the wavelength dependence of photodarkening parameters τ⁻¹ (photodarkening rate constant) and A (saturated photodarkening-induced loss), derived from a stretched-exponential fit to respective data sets. Absorption data measured at 600 nm was converted to 670 nm by using the ratio 1.98, derived from the room temperature color center loss spectrum shown later in Fig. 3 . These data sets do not overlap, and subsequently, different photodarkening parameters are derived with the 670 nm wavelength featuring faster rate (by 28 %) and deeper saturation level (by 8 %). However, when the pump laser is turned off at the end of the measurement(s), the 670 nm curve exhibits a step-like decrease of absorption coefficient up to the same level as the 600 nm curve. No change at 600 nm is seen upon pump turn-off. Therefore the 670/600 nm loss ratio of 1.98 appears to hold well when pump is off, but is reduced by some mechanism when the fiber is being pumped.

 

Fig. 1 (a) Photodarkening-induced fiber absorption coefficient change Δα at 670 nm (red circles) and 600 nm data converted to the same scale as the 670 nm by dividing by 1.98 (blue squares). Data is fitted using stretched-exponential function with fit parameters indicated next to the respective data. (b) Thermal bleaching (from room temperature to ~650 °C and back) of induced losses.

Download Full Size | PDF

 

Fig. 3 Absorption coefficient change spectra Δα(λ) measured after photodarkening, after thermal cycling (from RT to 325 °C and back), and at 146 °C and 304 °C (after completing the first thermal cycling sequence). Dashed lines show a Gaussian fit to the respective data sets.

Download Full Size | PDF

Following photodarkening of the sample fiber, a thermal bleaching cycle between room temperature (RT, about 25 °C) and ~650 °C was done to recover the sample to its pre-photodarkened state. Figure 1(b) presents –Δα plotted versus temperature, measured separately at 600 nm and 670 nm, with the 600 nm data again divided by 1.98. Both curves have been normalized with the Δα temperature-dependency of the pristine fiber, measured to

be ≤ 2 dB/m over the temperature range of interest. From the 600 nm data, recovery of the absorption can be seen to initiate at about 350 °C, with the increase of temperature up to 625 °C returning the fiber fully to the pre-photodarkened state. Interestingly, between 100 °C and 325 °C, a previously unreported heat-induced increase of Δα (i.e. darkening) can be seen to take place. This effect is even more pronounced in the 670 nm data, which together with the pump-dependent behavior seen in Fig. 1(a), suggests that some additional temperature-dependent loss mechanism is occurring. To study this darkening mechanism in more detail, the same sample was photodarkened again and then gently thermally cycled between RT and 325 °C, in order to stay below the temperature where recovery initiates. The results of this measurement at 600 nm are shown in Fig. 2 . The absorption coefficient starts to increase at ~100 °C, and continues to increase nearly exponentially with the temperature until 325 °C is reached. Upon cooling the fiber to room temperature, absorption coefficient change reaches roughly 570 dB/m. This value appears to be quite close to the maximum change caused by heat-induced darkening, as a further increase of temperature does not result in further increase of absorption. The post-irradiation heat-induced darkening is therefore estimated to constitute roughly 20% of total induced losses. The same percentage change can be derived from a spectral measurement of the absorption coefficient change before and after the thermal cycling, as presented in Fig. 3. Furthermore, the similarity of the spectral shapes indicates that the heat-induced darkening originates from the same type of color center as is created through the NIR pump excitation route.

 

Fig. 2 Absorption coefficient change Δα during thermal cycling between RT and 325 °C, followed by cycle between RT and 650 °C. Heat-induced change of Δα at RT is indicated by an arrow between “start” and second “turning point”.

Download Full Size | PDF

Figure 3 also presents two spectra measured at 146 °C and 304 °C after completing the thermal cycling between RT and 325 °C. Vertical dotted line indicates the 670 nm measurement wavelength. The previously discussed temperature-dependent additional loss mechanism at 670 nm can now be identified as heat-induced spectral broadening of the absorption spectrum. Losses increase from ~230 dB/m to ~360 dB/m going from RT to 304 °C, in good agreement with the 670 nm results shown in Fig. 1(b). Meanwhile, the 600 nm measurement wavelength is close to the cross-over wavelength where the observed spectral broadening is negligible. Therefore, neither photodarkening nor bleaching experiments (e.g. Figure 1(a) and (b)) at this wavelength are expected to have significant contribution from spectral broadening.

A more rigorous study of the temperature dependence of Δα for 612 nm, 635 nm and 670 nm measurement wavelengths is presented in Fig. 4 . These data points are derived from the least-squares Gaussian fits to data in Fig. 3 with additional spectral measurements taken at roughly 50 °C intervals. The fit to the ‘photodarkened & T-cycled to 325 C’ spectra measured at RT (shown in Fig. 3) indicates a loss peak Δα_peak centered at ~430 nm, which was then used to force the center wavelength of the fits on data measured at higher temperatures. All fits have an r²-value better than 0.997, and the assumption that the center wavelength is constant as a function of temperature has a negligible effect on the quality of the fit. The observed temperature-dependent changes in the absorption spectrum shape are therefore well described by spectral broadening of a single absorbing species, but contribution from other absorbing species cannot be fully ruled out. Data between 850 nm and 1100 nm is excluded from the

 

Fig. 4 Temperature dependence of Δα measured at 612 nm, 635 nm and 670 nm (after completing the first thermal cycling sequence). Secondary y-axis plots the full width half maximum (FWHM) of the loss peak centered at 430 nm.

Download Full Size | PDF

data-fit because (temperature-dependent) ytterbium absorption distorts the loss spectrum. The secondary y-axis plots the full width half maximum (FWHM) of the Gaussian fit versus temperature. All of the data sets show a linear dependence on temperature. The spectral broadening induced absorption coefficient change slopes (d(Δα)/dT), derived from the best fits lines in Fig. 4, are summarized in Table 1 . Spectral broadening is minimized in the vicinity of 612 nm (λ₀). Wavelengths shorter than λ₀ exhibit a negative slope (reduced losses versus temperature) and longer wavelengths exhibit a positive slope. The magnitude of the slope increases as a function of |λ-λ₀|. Therefore, both 635 nm (used in many previous photodarkening rate measurements) and 670 nm wavelengths are prone to measurement artifacts associated with such temperature-dependent photodarkening parameters through spectral broadening. The measurement wavelength of 600 nm used in our experiments has d(Δα)/dT of −0.06 (dB/m)/°C which explains the decrease in absorption observed in Fig. 2 in going from room temperature to 325 °C (after completing the first thermal cycle). Therefore, if the 600 nm data in Fig. 2 were corrected for the spectral broadening, Δα would be observed to have a constant value of ~570 dB/m between room temperature and 325 °C.

Table 1. Temperature dependence of absorption coefficient due to spectral loss broadening.

View Table

Spectral broadening may also explain some or all of the observed step-like decrease of absorption coefficient shown in Fig. 1(a). For example, the measured decrease of ~20 dB/m evident in the 670 nm data could be explained by temperature-dependent spectral broadening if the fiber core temperature rises from room temperature to ~120 °C when the pump is turned on. But based on an estimate of the amount of pump light absorbed by the fiber sample (~270 mW), the quantum defect for pumping at 915 nm, and temperature measurements made in the vicinity of the fiber sample, we would not predict an increase in the core temperature of this magnitude. Further studies are being undertaken to determine the actual fiber core temperature during pumping. This is relevant not only with regard to the study of heat-induced darkening (which is likely to be occurring at the estimated fiber core temperature), but also with regard to understanding what other possibly detrimental effects the underlying heat-generating mechanism may be causing.

Finally, we note that in photodarkening measurements of two other commercially available 20-µm core diameter LMA fibers, while both fibers exhibited significantly different photodarkening behavior (having different photodarkening rates and saturation levels), they did all feature the same post-irradiation heat-induced darkening, spectral broadening, and bleaching behavior described above.

4. Discussion

The results presented on post-irradiation heat-induced darkening indicate that ~20% of the color centers are not directly created through the NIR pump (e.g. 915 nm) excitation route. Instead, we hypothesize the existence of one or more NIR-induced darkening precursor species that do not contribute to absorption losses prior to thermal activation. Upon thermal activation, these precursor species convert to one or more absorbing species with the same spectral characteristics as those created directly by the NIR excitation route. Engholm et al. have attributed the color center formation mechanism to a charge-transfer (CT) process between the Yb³⁺-ion and the surrounding ligands [11]. Excitation into this CT absorption band at ~230 nm results in temporary formation of an Yb²⁺-ion (i.e. change of Yb³⁺ valence state to Yb²⁺) and a delocalized hole, bound to an ytterbium ion. In this proposed mechanism, the bound hole and electron associated with the Yb ion recombine, and characteristic Yb³⁺ fluorescence with a peak wavelength of 976 nm is observed. Higher excitation energy increases the possibility of creating free charges (holes), which can then become trapped, forming color centers [12]. Our results suggest an additional formation mechanism, the possibility of permanent (at least on a time scale ~10² hours) separation of the charges, without formation of color centers. Derivation of the thermal activation energy related to this additional darkening mechanism is in progress and is expected to help in identification of the darkening mechanism. It should finally be mentioned, that previous publications on thermal recovery of photodarkening, while reporting similar bleaching temperatures, do not report seeing any post-irradiation heat-induced darkening [5,9]. This discrepancy is not well understood, but may be explained by differences in the studied fibers or experimental details.

We have also explained how temperature-dependent changes in the photodarkening absorption spectrum can distort accelerated photodarkening rate measurements. Pumping an YDF results in generation of heat, and if the sample is not actively cooled, this temperature increase will result in (what appears to be) pump power (and therefore inversion) dependent saturation behavior, as clearly shown by the two decay curves measured at different wavelengths in Fig. 1(a). However, the magnitude and sign of the absorption coefficient change due to spectral broadening is likely to vary between fibers of different composition. The effect of spectral broadening can be mitigated either by using probe light centered at the least temperature-dependent wavelength λ₀ (e.g. 612 nm for the fiber used in this study) or by making sure the YDF sample temperature is maintained at constant temperature. Preferably both methods should be used for precaution, as there may be also other temperature-dependent mechanisms at work. Further work on this topic is underway and will be reported later. Finally, spectral broadening of losses may also play a role in practical fiber laser and amplifier operating parameters, as the temperature-dependence d(Δα)/dT is more pronounced in the 1.0 µm wavelength region.

5. Conclusions

Thermal bleaching studies made on a photodarkened, commercially available, 20-µm core LMA Yb-fiber have revealed two new mechanisms related to photodarkening; post-irradiation heat-induced darkening and spectral loss broadening. Absorption measurements made at room temperature on a photodarkened YDF fiber sample showed a further increase of absorption (i.e. darkening) after the temperature of the fiber sample was ramped up from 25 °C to 325 °C and then ramped back down to 25 °C. We hypothesize that this post-irradiation heat-induced darkening indicates the presence of an intermediate energy state in the NIR (e.g. 915 nm) excitation route, from which a charge can be thermally activated to form a color center.

Photodarkening-induced losses were found to exhibit heat-induced spectral broadening. The spectral width of the photodarkened fiber absorption spectrum exhibited a linear dependence on temperature. For absorption measurements conducted at a single wavelength (or within a narrow range of wavelengths), the temperature dependent change in absorption is proportional to λ−λ₀. λ₀ was found to be ~612 nm for the fiber used in this study. To avoid temperature-dependent measurement artifacts, photodarkening measurements should preferably be made at the cross-over wavelength λ₀ and using a temperature controlled sample.