Photodarkening experiments are performed on ytterbium-doped silicate glass samples. A strong charge-transfer (CT) absorption band near 230nm in aluminosilicate glass is found to be correlated to the mechanism of induced color center formation. Excitation into the CT-absorption band generates similar color centers as observed in ytterbium-doped fiber lasers under 915nm high power diode pumping. The position of the CT-absorption band is compositional dependent and is shifted to shorter wavelengths in ytterbium doped phosphosilicate glass. Very low levels of photodarkening is observed for the ytterbium doped phosphosilicate glass composition under 915nm high power diode pumping. Possible excitation routes to reach the CT-absorption band under 915nm pumping are discussed.
©2008 Optical Society of America
Induced optical losses, also called photodarkening, is known to be a limiting factor for the lifetime of high power fiber lasers. Koponen et al. have previously shown that the photodarkening rate is correlated to the number density of ytterbium (Yb) ions in the excited state . For cw-fiber lasers working at moderate powers, the level of inversion can be kept low and photodarkening is of minor concern. For pulsed fiber lasers operated under high inversion, on the other hand, the problem of induced optical losses is still an issue. The underlying mechanisms responsible for the induced absorption losses at visible and near infrared wavelengths are still under debate. Koponen et al. have observed a seventh order dependence on inversion and suggested that Yb-clusters containing up to seven ytterbium ions are involved to bridge the bandgap of the silicate host. Generation of UV-photons by the simultaneous deexcitation of three (or more) Yb-ions were suggested by Morasse et al. . We have recently shown that the valence state of the ytterbium ion is unstable in the silicate glass matrix and suggested this to be part of the photodarkening process . We have also shown that ytterbium doped aluminosilicate glass has a charge-transfer (CT) band located near 230nm. Excitation into this CT-band corresponds to the formation of an Yb2+-ion  and a delocalized hole on the surrounding anions, bound to the ytterbium ion. Recombination of this hole results in characteristic Yb 3+ luminescence near 1.0µm through a relaxed CT-transition. A simple model was presented suggesting that excitation into higher lying CT-states would generate mobile charges (free holes) resulting in induced optical losses. In this paper we will show that the CT-band is indeed part of the photodarkening mechanism and excitation into the CT-bands generates similar color centers as under 915nm high power diode pumping. We will also show that the induced losses can be reduced to very low levels by changing the glass composition such that the CT-absorption band is shifted to shorter wavelengths.
All samples were fabricated as optical-fiber preforms using a commercial MCVD-system (Nextrom OFC-12). Porous SiO2-layers were deposited inside synthetic silica substrate tubes (Heraeus Suprasil F300) using SiCl4 (FO Optipur) and O2 (99.9995%). The SiO2-layers were subsequently impregnated with solutions of YbCl3 (99.998%) and Al(NO3)3 (99.997%) in high-purity water (18.2MΩcm) followed by post-processing into a solid preform in the MCVD-system. The investigated samples are summarized in table I. The absorption cross-section near 1µm for ytterbium is known to be smaller in phosphosilicate glass compared to aluminosilicate glass. Thus, for the samples to have comparable absorption coefficients, the ytterbium concentration for the phosphosilicate glass was increased by a factor of two. A 30W deuterium lamp was used in the photodarkening experiments by UV-irradiation. The UV-irradiation profile of the deuterium lamp is shown in Fig. 2. The intensity of the UV-radiation was in the order of 1mW/mm2. A monochromator was coupled to the deuterium lamp for measuring the wavelength dependence of the induced loss. Typical irradiation intensities in the wavelength dependence experiments were in the order of 1-10µW/mm2. All optical characterization was made with a Jobin Yvon Triax 180 monochromator and a UV-enhanced silicon photo-detector connected to a Signal Recovery 7265 lock-in amplifier. For the accelerated photodarkening experiments, the Yb/Al- and Yb/P-doped preforms were drawn to fibers with core and cladding diameters of 20µm and 150µm respectively.
Figure 1 shows the absorption spectra for ytterbium doped alumino- and phosphosilicate preforms in the range 180–1100nm. The absorption spectra for non-Yb-doped preforms are also shown for comparison. A strong CT-absorption band is observed for the Yb/Al-doped preform near 230nm, corresponding to the formation of an Yb2+-ion and a delocalized hole, hbound, on nearby oxygen ligands still bound to the ytterbium ion. For the Yb/P-preform, the CT-band is shifted to shorter wavelengths, thus making the glass more transparent in the UV-range. The onset of the CT-band for Yb-doped phosphosilicate glass is shifted nearly 8000cm-1 compared to the situation in the Yb-doped aluminosilicate glass. The host lattice absorption edge for the non-Yb-doped phosphosilicate glass, however, is at somewhat longer wavelengths compared to the case of the non-Yb-doped aluminosilicate glass (compare the red dashed- and red solid lines in Fig. 1). The bandgap of phosphosilicate glass is thus smaller compared to the bandgap of aluminosilicate glass.
The CT-band of rare-earth (RE) ions is known to vary greatly for oxide based host materials . The shift of the CT-band to shorter wavelengths for the Yb/P glass is most likely related to a change in covalency between the ytterbium ion and the oxygen ligands in the phosphosilicate glass matrix. A decrease in covalency at the ytterbium site is expected for the phosphosilicate glass matrix, because the coordination shell around the ytterbium ion is different, ie Yb-O-P instead of Yb-O-Al. The small and highly charged P5+-ion results in a polarization of the oxygen ligands and a lower degree of covalency at the Yb-site . Other factors that determine the position of the CT-band in oxide based hosts are the coordination number and the RE-O distance .
Characteristic Yb3+ luminescence near 1µm is observed upon excitation into the CT-absorption band. Figure 2 displays the excitation spectra for the Yb-doped alumino- and phosphosilicate glasses monitored at 980nm. A good correlation is found between the excitation-and absorption spectra, showing that the strong absorption bands in the UV-range are related to charge-transfer transitions. From the excitation spectrum it is also found that the peak of the CT-band for the Yb/P-preform appears to be centered around 192nm. This is not observed in the absorption spectrum due to overlap with the host lattice absorption. The relevant configurational coordinate diagram for the CT-process is shown in Fig. 3 for the Yb/Al- and Yb/P glasses. Excitation into the CT-band is followed by non-radiative relaxation until a new equilibrium position (R’) is reached. CT-luminescence is characterized by two broad bands with an energy separation of about 10000cm-1 corresponding to a transition from the CT-state (CTS) to the 2F5/2 and 2F7/2 levels. CT-luminescence, however, is usually only observed at low temperatures and has been studied in detail for several Yb-doped crystalline hosts . No reports of CT-luminescence has, to the best of our knowledge, been made for glass materials. This is probably because the CT-state usually has a larger offset in glass materials, such that the lowest point nearly crosses the 4f13-parabolas . Non-radiative transitions to the 2F5/2 level can then quench the CT-luminescence and only emission near 1µm is observed. For ytterbium doped phosphosilicate glass, see Fig. 3b, the CT-state is located at a higher energy. This is, as will eventually be shown, a very important difference when it comes to the induced optical losses observed for high power fiber lasers.
To investigate the role of the CT-band in the photodarkening process, all preform samples in table 1 were irradiated by a deuterium lamp focused by an uncoated quartz lens. The irradiating profile of the D2-lamp and quartz lens has a maximum close to the CT-absorption peak for the Yb/Al-preform sample, see Fig. 2. Figure 4 shows the induced loss after 10 hours for the Yb/Al- and Al samples. A high induced loss is observed at visible wavelengths with a long tale into near infrared wavelengths. No measurable induced loss is observed for the non-Yb-doped Al-sample. A typical photodarkening spectrum for a Yb/Al-doped fiber after about 1 hour, by using 915nm high power diode pumping, is also shown for comparison in Fig. 4. There is a good correlation between the UV-irradiated Yb/Al-preform and the 915nm pumped photodarkened Yb/Al-fiber, which shows that the same type of color centers are formed. This also shows that the strong charge-transfer band plays a major role in photodarkening process. The formed color centers are most likely hole-related defects in the aluminosilicate glass matrix, see e.g. Cohen and Makar . A deeper investigation on the origin of the formed color centers is beyond the scope of this paper. We instead focus on the role of the CT-band as a generator of mobile charges in the glass matrix and on the excitation mechanisms to reach the CT-band under 915nm high power pumping. The deviation in the region 350–400nm for the fiber is due to a decreasing dynamical range of the optical spectrum analyzer (OSA) for wavelengths <400nm.
Figure 5 displays the induced loss spectra for the Yb/P- and P-doped preform samples after 10 hours of UV-irradiation. As can be seen, the induced loss for the Yb/P- and Yb/Al-preforms have comparable intensities (see the black solid lines in Fig. 4 and 5), however the spectral shape is different. The UV-irradiated Yb/P- and P-doped preforms have a more pronounced peak near 540nm, which probably is related to phosphorus oxygen hole centers (POHCs) as reported by e.g. Girard et al. . A small peak around 330nm is also observed for the Yb/P-doped preform, which may also originate from a hole-related defect . The green dashed line in Fig. 5 shows the differential spectrum between the Yb/P- and P-preforms. It is clear that excitation into the CT-band accounts for a large part of the induced absorption. Furthermore, this also shows that most of the 540nm absorption band comes from direct excitation across the bandgap of the phosphosilicate glass matrix. The peak near 330nm, on the other hand, seems to originate from excitation into the CT-band. Very low levels of photodarkening is observed for the Yb/P-doped fiber after 1 hour of 915nm high power diode pumping, see blue solid line in Fig. 5. In fact, the induced core loss maintains below 10dB/m even after 46 hours of high power diode pumping (see Fig. 6). As very low levels of photodarkening is observed for the Yb/P-fiber, it is difficult to compare the spectral features to that of the Yb/P-preform. For a Yb/P-fiber doped with 0.91at% ytterbium, however, a higher induced loss is observed after 46 hours under 915nm diode pumping, see purple line in Fig. 5. It is interesting to find the tendency for a POHC-related absorption band near 540nm, as well as an increased induced loss for shorter wavelenghts (<400nm). This again shows that same type of color centers are formed upon excitation into the CT-band as under 915nm high power diode pumping. We thus conclude that the CT-band plays a major role in the photodarkening of Yb-doped high power fiber lasers. Figure 6 compares the time dependent induced core loss under 915nm pumping for the Yb/Al-and the Yb/P-doped fibers, clearly showing the superior PD-resistance of the Yb/P-doped fiber.
Figure 7 shows a deconvolution of the strong UV-absorption bands for the Yb/Al-doped preform by using the PeakFit software. Gaussian line shapes are used for all bands and at least three bands are necessary for a band separation. An r2-value of 0.99990 was achieved for the deconvolution, see dashed curves in Fig. 7. The two lower bands are associated with CT-transitions (see black dashed curves). The green dashed curve is assigned to the onset of the 4f-5d transitions for Yb3+, which is expected to appear at these energies. Furthermore, it is also known that the CT-transitions are located at lower energies compared with 4f-5d transitions in the case of Yb3+ . The deconvoluted CT-band at 51300cm-1 most likely corresponds to a higher lying CT-state (CTS* in Fig. 3), which is associated with the formation of an Yb2+-ion and a free hole, hfree, as described by C. Pedrini .
The blue circles in Fig. 7 show the integrated induced loss in the range 300–1000nm after 10 hours of irradiation as a function of irradiation wavelength. Seven different irradiation wavelengths were selected. The integrated induced loss was normalized to the power of the irradiation wavelength and plotted at the center wavelength of the irradiation profile. A good correlation is found between the deconvoluted CT-absorption band at 51300cm-1 and the integrated induced loss. This favors the assignment of an excited CT-state (CTS*) as this state comprises a free hole, which can result in the formation of color centers. The lower CT-band is apparently not directly responsible for the formation of mobile charges in the glass matrix. Instead, the upper CT-band at 51300cm-1 appears to generate mobile charges (holes) resulting in the formation of color centers in the aluminosilicate glass matrix. The lower CT-state (CTS), on the other hand, can be indirectly involved by acting as an intermediate energy state in the excitation route under 915nm high power diode pumping. A higher lying CT-state is also expected to be found for the Yb/P-doped glass, but the overlap of the CT-bands with the absorption of the host lattice makes such identification difficult.
The results presented in this paper clearly show that the CT-bands are part of the induced color center formation in ytterbium doped high power fiber lasers. The excitation mechanism, however, for reaching the CT-states under 915nm high power pumping needs to be addressed. It is known that the photodarkening rate increases with increasing Yb-concentration. An increased Yb-concentration also results in smaller inter atomic distances between the Yb-ions and a larger number of Yb-ion pairs. These Yb-ion pairs have a slightly different energy level structure compared with single Yb-ions . Beside the usual 4f-4f transitions around 10000cm-1, there are also transitions around 20000cm-1, corresponding to transitions from the [φ(4f7/2)φ(4f7/2)]- to the [φ(4f5/2)φ(4f5/2)] configuration. The oscillator strength for these cooperative absorptions are typically a factor ~10-6 less compared with transitions between the 2F7/2 and 2F5/2 multiplets. Nevertheless, the radiative decay (cooperative luminescence) is easily observed as a blue-green luminescence along the fiber, even for relatively low intensities (low inversion levels) and low Yb-concentrations. Thus, it is not unlikely that Yb-ion pairs may constitute intermediate energy states needed to reach the CT-states under 915nm pumping. The energy needed to reach the CT-states is equivalent to the combined energy of 3–4 pump photons. The shift of the CT-states to higher energies observed for the Yb-doped phosphosilicate glass composition is consistent with the much lower induced loss observed under 915nm high power pumping. The probability for the generation of free holes is of course much less as it requires the combined energy of at least 4 pump photons for the Yb/P composition.
Another excitation mechanism, not to be overlooked, is the probability for multi-photon absorption (MPA). MPA has been reported for several rare-earth ions under near infrared irradiation and involves the simultaneous absorption of two (or more) photons from the ground, or excited intermediary, states to higher excited states of the rare-earth ions [14, 15]. Yoo et al.  recently reported induced optical losses by a two-photon absorption (TPA) process in an ytterbium-doped aluminosilicate optical fiber. They used a 488nm Ar+ ion laser and showed that a TPA-process, to the 230nm CT-band, was responsible for the observed induced absorption. This is a very interesting result, which strongly supports the ideas that multi-photon processes under 915nm pumping can be a possible excitation mechanism for reaching the CT-state. Under 915nm high power diode pumping, excitation to the CT-states most likely originates from the excited 2F5/2-level rather then the 2F7/2 ground state level, as this is consistent with the observed inversion dependence for the photodarkening rate. A two-photon absorption process from the excited 2F5/2 level would be enough to reach the onset of the first CT-state for the Yb/Al-glass. A three-photon absorption process from the 2F5/2-level is enough to reach the onset of the first CT-state in Yb/P-doped glass. The lower probability for a three-photon absorption process could account for the much lower induced loss observed for the Yb/P-doped fiber. The probability for a MPA processes, however, is much lower compared to a one-photon absorption. Our results, on the other hand, show that intensities in the order of a few µW/mm2 are sufficient to observe induced optical losses after just a few hours of UV-irradiation. Thus, it seems reasonable that for fiber lasers operated under high powers, the suggested excitation routes could very well account for the induced optical losses observed under 915nm pumping. Clearly, more investigations are needed to fully understand the mechanisms involved in color center formation and further studies addressing the excitation route are in progress.
To conclude, we have shown that photodarkening in Yb-doped high power fiber lasers is correlated to the presence of a charge-transfer absorption band in the ytterbium-doped silicate glass matrix. The charge-transfer band is shifted to shorter wavelengths for an Yb-doped phosphosilicate glass compared to the situation in aluminosilicate glass. This results in a much lower induced loss under 915nm high power diode pumping. A major drawback of the phosphosilicate glass matrix, on the other hand, is the smaller absorption cross-section. This has to be compensated for by an increased Yb-concentration or longer active fiber lengths. A higher Yb-concentration will result in a larger core numerical aperture (NA), which will reduce the performance of the fiber laser. Longer active fiber lengths will result in lower thresholds for nonlinear effects such as stimulated brillouin scattering (SBS) and stimulated raman scattering (SRS). Thus, a trade-off has to be made for each specific application.
NKT Research, Denmark and Fiber Optic Valley, Sweden are gratefully acknowledged for the financial support. Crystal Fibre A/S, Denmark is gratefully acknowledged for providing fiber samples and measurement assistance.
References and links
1. J. Koponen, M. Söderlund, H. Hoffman, D. Kliner, and J. Koplow, “Photodarkening measurements in large mode are fibers,” Proc. SPIE 6453, 64531E (2006). [CrossRef]
2. B. Morasse, S. Chatigny, E. Gagnon, C. Hovington, J.-P. Martin, and J.-P. de Sandro, “Low photodarkening single cladding ytterbium fibre amplifier,” Proc. SPIE 6453 (2007). [CrossRef]
3. M. Engholm, L. Norin, and D. Åberg, “Strong UV-absorption and visible luminescence in ytterbium-doped aluminosilicate glass under UV-excitation,” Opt. Lett. 32, 3352–3354 (2007). [CrossRef] [PubMed]
4. G. Blasse, “The ultraviolet absorption bands of Bi3+ and Eu3+ in oxides,” J. Solid State Chem. 4, 52 (1972). [CrossRef]
6. L. van Pieterson, M. Heeroma, E. de Heer, and A. Meijerink, “Charge transfer luminescence of Yb3+,” J. Lumin. 91, 177–193 (2000). [CrossRef]
7. G. Blasse and B. Grabmaier, eds., Luminescent Materials (Springer-Verlag, Berlin Heidelberg, 1994), p. 83.
8. M. Cohen and L. Makar, “Models for color centers in smokey quartz,” Phys. Status Solidi A 73, 593–596 (1982). [CrossRef]
9. S. Girard, E. Régnier, A. Boukenter, Y. Ouerdane, J.-P. Meunier, and D. Hamdani, “Gamma and UV radiation-induced color centers in optical fibers,” Mat. Sci. Forum 480–481, 323–328 (2005).
10. P. Ebeling, D. Ehrt, and M. Friedrich, “X-ray induced effects in phosphate glasses,” Opt. Mater. 20, 101–111 (2002). [CrossRef]
11. J. Hölsä, M. Lastusaari, M. Marysko, and M. Tukia, “A few remarks on the simulation and use of crystal field energy level schemes of the rare earth ions,” J. Solid State Chem. 178, 435–440 (2005). [CrossRef]
12. C. Pedrini, “Electronic processes in rare earth activated wide gap materials,” Phys. Status Solidi A 2, 185–194 (2005). [CrossRef]
14. H. You and M. Nogami, “Three-photon-excited fluorescence of Al2O3-SiO2 glass containing Eu3+ ions by femtosecond laser irradiation,” Appl. Phys. Lett. 84, 2076–2078 (2004). [CrossRef]
15. H. You, T. Hayakawa, and M. Nogami, “Upconversion luminescence of Al2O3-SiO2:Ce3+ glass by femtosecond laser irradiation,” Appl. Phys. Lett. 85, 3432–3434 (2004). [CrossRef]
16. S. Yoo, C. Basu, A. Boyland, C. Sones, J. Nilsson, J. Sahu, and D. Payne, “Photodarkening in Yb-doped aluminosilicate fibers induced by 488nm irradiation,” Opt. Lett. 32, 1626–1628 (2007). [CrossRef] [PubMed]
17. A changed oxidation state of an Mn+ rare-earth-ion caused by ionizing irradiation is often denoted as (Mn+)+ or (Mn+)- instead of M(n+1)+ or M(n-1)+ respectively. This is because the local environment may be different compared to that formed initially in the glass under e.g. heat treatment. This may result in small shifts of the observed absorption bands. The formal charge of the RE-ion is nevertheless the same and we will use the notation M(n+1)+ or M(n-1)+ for an ion with a changed oxidation state.