We report for the first time Erbium doped Tellurium dioxide single-mode planar rib waveguide amplifiers with net fiber to fiber gain and wide bandwidth operation. Peak internal gains of up to 14dB have been achieved in 5cm long rib waveguides (2.8dB/cm) fabricated by co-sputtering of Tellurium and Erbium in an oxygen ambient and reactive ion etching.
©2010 Optical Society of America
Tellurite glasses (based on Tellurium dioxide, TeO2) are well known to have a range of interesting properties for fiber and planar waveguide devices, especially for optical amplification . Erbium Doped Waveguide Amplifiers (EDWAs) have potential to provide high gain from a very short device which is advantageous for short pulse amplification, low cost amplifiers, as gain sections for integrated on-chip mode locked or CW lasers, and the provision of on-chip gain to compensate loss in complex integrated optics devices. They are therefore key components in integrated optics. EDWAs have previously been realized by differing techniques in various materials , most successfully until now in Al2O3  or multicomponent phosphate glass . Devices are now commercially available, on-chip gains as high as 5.3 dB/cm  have been attained along a 3.1cm long waveguide with ~5nm FWHM gain bandwidth. In one case, 4.1dB in a 0.3cm long (13.7dB/cm) very highly doped device was reported with a ~35nm FWHM bandwidth  though whether this can be scaled to technologically useful gains (>15dB) remains unclear as a longer device reported by the same authors in the same paper exhibited lower gain per unit length. However there has seemingly been no progress in simultaneously achieving high gains and attaining wider bandwidth operation as is considered desirable for future transmission systems and has been demonstrated for fiber amplifiers . The reasons are many and complex and are related mostly to the intrinsic glass host environment and effects involving the high Erbium concentrations required leading to clustering, concentration quenching, co-operative upconversion, excited state absorption, or other such effects.
Tellurites offer a number of advantages as emission hosts for EDWAs over other materials because of their high refractive index (larger emission cross section and more compact devices), large emission bandwidth, low ion to ion cross relaxation, relative independence of the 1550nm Erbium lifetime on concentration, and high Erbium solubility as has been demonstrated in Tellurite glass and fiber amplifiers [7–10]. Whilst there has until now been no demonstration of Tellurite based EDWAs with net fiber to fiber gain, it is clear that Tellurite based devices have the potential to deliver higher ultimate gain per unit length and bandwidth than previous demonstrations in other materials . The low phonon energy of around 600-800cm−1 in Tellurite also enables the potential use of transitions not possible in other materials commonly used in EDWAs, (for example with Pr3+ gain at 1.3μm).
To date the only prior demonstration of a Tellurite EDWA was a device fabricated by femtosecond laser irradiation . The device produced very limited amounts of internal gain (1.2dB over 2.5cm) due mainly to the high propagation losses in the waveguide of 1.3dB/cm at the signal wavelength, this representing the state of the art in Tellurite waveguides at the time of that article’s preparation. This high loss is the main reason for the lack of progress on Tellurite EDWAs, and this constraint itself has only recently been released .
This work reports Tellurite EDWAs that exhibit net fiber to fiber gain fabricated using reactive RF magnetron co-sputtered Er doped TeO2 films. Rib waveguides were fabricated using standard lithography and Reactive Ion Etching (RIE) using a Hydrogen/Methane/Argon gas mix  in a layer of pure TeO2 deposited on top of the Er doped film.
2.1 Thin film deposition
Erbium doped Tellurium dioxide films were deposited by co-sputtering 3” diameter Te and Er targets (AJA International ATC 2400-V system with A330 guns) in an Oxygen/Argon gas mix. The Te target was run at 148W, the chamber pressure was 5mTorr, and the total flow of Ar and O2 was kept at 15sccm. Varying the O2/Ar mix allows tuning of the Oxygen content in the films, higher O2 flow leading to higher Oxygen content in the deposited films but also a slower deposition rate. Typical conditions were 6sccm O2, 9sccm Ar. The Erbium concentration was controlled via the Er gun power, and was measured post deposition by a laser ablation ICP mass spectrometer. Films with an Er/Te ratio of up to 3% were fabricated by this method. A 1% Er/Te ratio film was achieved with 72W of RF power on the Er gun.
One key challenge in realizing Tellurite waveguides has been etching technology. Plasma etching is highly favored as it provides high quality etched surfaces with high anisotropy and accurate etch rate monitoring . Previous work on plasma etching of Tellurium dioxide employed physical etching using Argon resulting in propagation losses of 6dB/cm . High quality Tellurite waveguides made by RIE with Hydrogen/Methane/Argon were recently reported with propagation losses below 0.1dB/cm . Figure 1(a) shows a typical profile of a waveguide etched by this method.
However, when Erbium doped thin films were etched using the Hydrogen/Methane/Argon gas mix, the surfaces became very rough with columnar structures and grassing effects as seen in Fig. 1(b). This resulted from re-deposition of low volatility Erbium Hydride compounds . Whilst there may be some prospect for preventing the grassing by raising the etch temperature and lowering the pressure , in the interests of simplicity we instead adopted a strip loaded waveguide design . Here the strip was formed from a pure TeO2 layer etched using the established process to within 100nm of the Er doped TeO2 surface.
2.3 Waveguide fabrication and passive properties
A bilayer film was first deposited using the conditions described above on a thermally oxidized 100mm diameter <100> oriented Silicon wafer (2.0µm thick oxide). Layer 1 was Er doped TeO2 with a thickness of ~0.9µm, followed by a 0.9µm thick pure TeO2 layer. This process was completed in a single 120 minute sputtering run. The obtained film had an equivalent measured refractive index of 2.03 at 1550nm (SCI Filmtek 4000). Erbium doped films were generally observed to have refractive index 0.01-0.04 lower than undoped material, depending on Er concentration and O/Te ratios. The bilayer film was therefore slightly Oxygen rich as undoped stoichiometric films have refractive indices around 2.08 and the expected index for the bilayer would then be 2.06 for stoichiometric material .
An etch mask was then patterned using standard I-line contact photolithography methods utilizing 0.9μm thick Clarient AZ 701 MiR photoresist and 150nm thick Brewer Science XHRiC-16 bottom anti reflective coating (BARC). The BARC layer also crucially acts as a protective layer during the photoresist development process to avoid wet etching of the film . After photoresist development, the exposed BARC layer was removed via Oxygen plasma in an Oxford ICP 100 plasma etching system.
Ridge waveguides were formed by etching the unmasked Tellurium dioxide area 0.7µm deep. The etching was performed in RIE mode at 30mT, 200W RF power, and Hydrogen/Methane/Argon gas flows of 30sccm, 10sccm, and 30sccm, respectively, giving an etch rate of 50nm/min. The photoresist mask was then Oxygen plasma stripped. Finally, the waveguides were top clad with 3µm of PMMA. Ridge waveguides 1.8µm thick, 0.7µm rib depth and widths of 1.5, 2.5 and 3.5 µm were obtained.
To illustrate the quality of the doped waveguides obtained by this process, Fig. 2 presents the insertion loss spectrum of a 3.5μm wide 7cm long waveguide doped with 2.7% Er/Te (~7x10−20 ions/cm3). The minimum insertion loss is ~2.5dB, almost identical to that achieved in undoped waveguides . The Erbium appears to be well dispersed in the host despite the high concentration, with little impact on the passive properties of the waveguide.
3. Er doped film and waveguide photoluminescence lifetime
For the photoluminescence experiments a waveguide with 1.3% Er/Te was first fabricated. The excited state lifetime of this waveguide was measured using the setup shown in Fig. 3 .
Here a tapered lensed fiber producing a 2.5μm 1/e2 diameter spot was aligned to the edge of the film and the 980nm pump laser modulated with ~10ms long pulses. The fluorescence was then coupled back into the fiber and the 1550nm component extracted and the pump rejected by a chain of three 980/1550nm WDMs providing in excess of 90dB pump rejection. Detection was accomplished with a connectorised 150μm diameter InGaAs diode (Fermionics) and a fast, very low noise Signal Recovery Inc. transimpedance amplifier. Data was captured on a PC equipped with a 16 bit analog to digital converter, and custom software enabled arbitrary amounts of trace averaging to remove noise. The system had an intrinsic rise time of ~8μs and could pump the sample with intensities from levels approaching those experienced in an actual amplifier down to its lowest detection limit, thereby enabling the intensity dependence of lifetime and the various non-linear effects to be captured. Additionally the short focal length of the detection means the probed path length (the measurement region) is very short thereby enabling the elimination of absorption/re-emission issues previously observed [9,16].
The lifetime of the 1.3% Er/Te film was investigated with this setup and the results shown in Fig. 4 were obtained at minimum detectable and maximum available (~300mW) pump powers (curves a) and b) respectively). At low powers the decay was a single exponential function as expected with a lifetime of 0.62ms. This is much shorter than the 4-8ms lifetimes reported in some multicomponent Tellurite glasses  though the actual lifetime in pure TeO2 does not seem to have been previously reported. We also measured a film with 0.3% Er/Te (shown as curve c) in Fig. 4) which showed a lifetime of 2.1ms. Such an increase in lifetime with Er concentration decrease does not align with previously observed results in bulk glasses unless substantial amounts of OH groupings are present in the glass . Examining Fig. 2 there is clear evidence of an absorption dip centered at 1290nm, which is in fact a combination stretch/bend overtone of OH in Tellurite glass [17,18]. Looking at a number of waveguides fabricated from different films we found an inverse linear relationship between the 1290nm absorption per unit length and the measured second decay lifetime (ie the underlying fundamental radiative lifetime) as shown in Fig. 5 .
Figure 5 clearly indicates an issue with OH contamination in the films. To put a perspective on the quantity of OH versus work done on bulk glass and fibers, then the amount of OH in the films can be estimated based on the measured ratio of the 1290nm to 1480nm water dips and the measured 1480nm to 3μm water dip ratio . The former ratio of losses in dB/cm is ~4.5, and the latter is ~667 giving a net ratio of losses at 1290nm to 3μm of about 3000x.
Amongst others, Dai et al. (2008)  studied the lifetimes of Er doped Tellurite glasses with different Er and OH concentrations. Using the scaling figures above, the 3μm loss of the data in Fig. 6 ranges from ~15 to ~300cm−1, substantially above the figures typically quoted in glasses of which  is typical. To determine the impact of these higher OH concentrations, Dai’s data  were fitted and recast as per Fig. 6.
The original data in Fig. 6(a) show a concentration independent lifetime of ~5ms, and the Er concentration dependence of the OH sensitivity of the inverse lifetime is plotted in Fig. 6(b) revealing a linear trend. Applying the range of OH concentrations in the present work (estimated at 15cm−1 and 300cm−1) to the 5ms base lifetime and taking a 1.3% Er/Te ratio, the lifetime is predicted to be shortened from 4.1ms in a good bulk glass to 1.2ms and 75μs respectively for the best and worst films. This is approximately the range of lifetime values displayed in Fig. 5 and is a large change from the OH free lifetime indicating the severity of the OH problem in the films. The origins and removal of this OH will be the subject of further research but it must come from either physical water absorption on the targets/vacuum system, a leak in the vacuum system, or from Erbium Hydrides/Hydroxides on the Erbium target.
The lifetime was also studied at higher powers as per Fig. 4 curve b), where upconversion, pairing, and other effects may come into play. A slightly shortened lifetime was observed. A measurement technique dependence for the high power lifetime was seen. The role of the waveguide structure has been previously identified  through its modification of the emission modes and so the high power lifetime was measured in three geometries; at the unstructured film edge, perpendicular to the film, and at the facet of two different waveguide lengths (4mm and 5cm). The results for the 1/e lifetime are presented in Table 1 .
The waveguide geometries considerably shorten the observed lifetime. The longer waveguide produces lower lifetime than the shorter waveguide. We believe this is due to the depletion of the excited Er ions in the measurement region by stimulated emission with photons generated elsewhere in the waveguide. Clearly the rate of stimulated depopulation depends on the number of incident photons which is a function of both the length of the waveguide and the pump power. Consequently a longer waveguide generates more photons outside the measurement region which propagate to the measurement region and depopulate the excited Er ions faster thereby reducing the observed lifetime further. Measuring perpendicular to the film surface also has a shorter lifetime than measuring through the film edge. Here, although no photons are generated outside the measurement region to depopulate the excited Er ions by stimulated emission as in the waveguide case, the measurement region is in fact contained inside a Fabry-Perot cavity comprising the film (high index) and air/the silica substrate (both low index). Thus some photons initially generated by spontaneous emission in the film are reflected at the surfaces and produce repeated stimulated emission events as they “bounce” inside the film thereby again reducing the number of excited ions by a non-spontaneous decay process thereby shortening the observed lifetime. In contrast to the two prior cases, measuring through the film edge results in the pump diffracting away quickly in the plane of the film as it is unguided there, so little spontaneous emission occurs outside the measurement zone and any that is generated is not guided back into the measurement zone. Thus the number of stimulated emission results is considerably reduced and the measured results are expected to be far more representative of the true spontaneous decay rate of the film than the other measurement modalities. However the results of Table 1 pose the question of whether this measurement modality remains completely unaffected by backward propagating spontaneous emission. The definitive measurement to clarify this would be to deposit the Er doped Tellurite film on an index matched substrate with top side anti-reflection coating and then measure perpendicular to the film. In this case there is by definition no spontaneous emission outside the measurement region, and no reflections of spontaneous photons generated inside the measurement region back into it for stimulated emission to occur. Therefore it would be expected that this geometry would measure only the spontaneous decay rate. Unfortunately no suitable substrates were available to try this.
Lifetime reductions due to high pump intensity may also have resulted from impurities in the sputtering system (from the sputtering of other targets) or to pairing effects inherent to the high Er doped concentration or the sputtering process itself  but further investigations are required to quantify and understand the potential effects of such processes in these sputtered films.
4. Gain experiments
A bilayer film was fabricated with a 1.0% Er/Te (~2.2x1020ions/cm3) 1350nm Er layer and index at 1550nm of 2.075, and a pure TeO2 layer 530nm thick. A 5cm long waveguide was prepared as previously described to yield a rib waveguide with 400nm etch depth. The lifetime of the film measured from the edge was 1.3ms. The loss spectrum showed 1290nm loss of ~0.02dB/cm indicating 3μm OH loss of ~15cm−1. A 2μm wide waveguide was single moded with polarization dependent loss 0.3dB due to mode size variation . The loss spectrum with the 2.5μm waist tapered lensed fibres indicated a total fibre to fibre insertion loss of 6dB at 1400nm, comprising 3dB of mode overlap and facet reflection and 3dB of background loss for 5cm (~0.6dB/cm) due to imperfect etching. The high reflectivity of the AR coatings on the lensed fibers made pumping at 980nm impractical so gain measurements were performed using the setup shown in Fig. 7 with bi-directional 1475nm, 250mW multiple longitudinal mode pumps.
An external cavity tunable CW laser and a supercontinuum source were used as probe signals. The pumps and signal were combined in a broadband 1420-1490/1520-1620 WDM and the signal power at the WDM input was tuned to −20dBm. The internal gain was measured as the signal enhancement (power at output with pumps on minus power at output with pumps off) minus the absorption loss. The enhancement factor was measured using an optical spectrum analyzer (OSA) to reject the amplifier’s broadband ASE, and the absorption profile of the doped waveguide was measured with a low power (<-20dBm integrated across the spectrum) supercontinuum source and the OSA. The measured gain results are shown in Fig. 8 .
Gain existed from below 1520nm (limited by pump WDM) to above 1630nm with a 3dB bandwidth of ~40nm. Peak internal gain of 14dB was obtained at 1530nm as shown on Fig. 8(a) for a gain per unit length of 2.8dB/cm. Even factoring in the total coupling and propagation losses of ~7.6dB (3.0dB mode overlap losses/facet reflections, 2.5dB lensed fiber loss, 2.1dB WDM & connectors) this results in net fiber to fiber gain. Most of the off chip loss can ultimately be eliminated by using existing techniques for low loss mode matching  and by integrating the WDMs. The amplifier gain appeared close to saturation at maximum available pump power as shown in Fig. 8(b), the signal gain saturation in Fig. 8(c) showing a saturation input power of 0.4mW and an output power exceeding 10dBm. The measured gain curves appear limited by inversion clamping at about 60-70% as gauged by the shape of the gain curve  due to the resonant pumping limiting inversion to a theoretical maximum of ~75% . Along with the measured gain data in Fig. 8(a) are plotted theoretical curves for the gain. These were obtained from the infinite pump approximation  at 100% and 70% inversion, for the latter with the 1480nm pump effective cross-section calculated as from Becker, et al., 1997  using the measured absorption and McCumber emission theory. This simple theoretical estimation gives good agreement with the experiment and predicts that the peak gain at 1530nm would be considerably improved to 35dB by using non-resonant 980nm pumping to achieve higher inversion. However, there is a penalty in the bandwidth as the 1530nm peak is much narrower (10nm width at 3dB down). Higher Er doping and gain flattening could however be used to extend the gain bandwidth, the data presented in Fig. 2 showing that higher doping should be possible without penalties from waveguide effects.
We have demonstrated a single-mode Tellurium Dioxide based Erbium doped planar rib waveguide amplifier with peak internal gain of 2.8dB/cm and net overall gain. More than 30dB of gain could be achieved by doping with a higher Er concentration, improving the lifetime through improved film fabrication to reduce OH contamination, by better waveguide structure, and by using a non-resonant pumping scheme to attain higher inversion levels.
This work was supported by the Australian Research Council under Discovery Project Grant DP070333.
References and Links
1. R. El-Mallawany, Tellurite Glasses Handbook: physical properties and data (CRC Press, 2002).
2. S. Berneschi, G. Nunzi Conti, and G. Righini, “Planar Waveguide Ampflifiers,” Ceramist: Jnl Korean Ceram. Soc. 10, 75–85 (2007).
3. G. van den Hoven, R. Koper, A. Polman, C. van Dam, J. van Uffelen, and M. Smit, “Net optical gain at 1.53 m in Er-doped Al2O3 waveguides on silicon,” Appl. Phys. Lett. 68(14), 1886–1888 (1996). [CrossRef]
4. Y. C. Yan, A. J. Faber, H. de Waal, P. G. Kik, and A. Polman, “Erbium-doped phosphate glass waveguide on silicon with 4.1 dB/cm gain at 1.535 m,” Appl. Phys. Lett. 71(20), 2922 (1997). [CrossRef]
5. G. Della Valle, S. Taccheo, P. Laporta, G. Sorbello, E. Cianci, and V. Foglietti, “Compact high gain erbium-ytterbium doped waveguide amplifier fabricated by Ag-Na ion exchange,” Electron. Lett. 42(11), 632–633 (2006). [CrossRef]
6. F. Patel, S. DiCarolis, P. Lum, S. Venkatesh, and J. Miller, “A compact high-performance optical waveguide amplifier,” IEEE Photon. Technol. Lett. 16(12), 2607–2609 (2004). [CrossRef]
7. A. Mori, “Tellurite-based fibers and their applications to optical communication networks,” J. Ceram. Soc. Jpn. 116(1358), 1040–1051 (2008). [CrossRef]
8. C. Chryssou, F. Di Pasquale, and C. Pitt, “Er3+ doped channel waveguide amplifiers for WDM systems: A comparison of tellurite, alumina and Al/P silicate materials,” IEEE J. Sel. Top. Quantum Electron. 6(1), 114–121 (2000). [CrossRef]
9. S. Dai, C. Yu, G. Zhou, J. Zhang, G. Wang, and L. Hu, “Concentration quenching in erbium-doped tellurite glasses,” J. Lumin. 117(1), 39–45 (2006). [CrossRef]
10. S. Shen, A. Jha, X. Liu, M. Naftaly, K. Bindra, H. J. Bookey, and A. K. Kar, “Tellurite glasses for broadband amplifiers and integrated optics,” J. Am. Ceram. Soc. 85(6), 1391–1395 (2002). [CrossRef]
11. T. T. Fernandez, G. Della Valle, R. Osellame, G. Jose, N. Chiodo, A. Jha, and P. Laporta, “Active waveguides written by femtosecond laser irradiation in an erbium-doped phospho-tellurite glass,” Opt. Express 16(19), 15198–15205 (2008). [CrossRef] [PubMed]
12. S. J. Madden and K. T. Vu, “Very low loss reactively ion etched Tellurium Dioxide planar rib waveguides for linear and non-linear optics,” Opt. Express 17(20), 17645–17651 (2009). [CrossRef] [PubMed]
14. S. M. Pietralunga, M. Lanata, M. Ferè, D. Piccinin, G. Cusmai, M. Torregiani, and M. Martinelli, “High-contrast waveguides in sputtered pure TeO2 glass thin films,” Opt. Express 16(26), 21662–21670 (2008). [CrossRef] [PubMed]
15. C. Tewell and S. King, “Observation of metastable erbium trihydride,” Appl. Surf. Sci. 253(5), 2597–2602 (2006). [CrossRef]
16. R. Grew, H. Ebendorff-Heidepriem, P. Veitch, and T. Monro, “Concentration effects in erbium doped tellurite glass,” in Conference on Optoelectronic and Microelectronic Materials and Devices, Proceedings, COMMAD(Institute of Electrical and Electronics Engineers Inc., Perth, Australia, 2006), pp. 196–199.
17. J. Wang, E. Vogel, and E. Snitzer, “Tellurite glass: a new candidate for fiber devices,” Opt. Mater. 3(3), 187–203 (1994). [CrossRef]
18. X. Feng, W. H. Loh, J. C. Flanagan, A. Camerlingo, S. Dasgupta, P. Petropoulos, P. Horak, K. E. Frampton, N. M. White, J. H. Price, H. N. Rutt, and D. J. Richardson, “Single-mode tellurite glass holey fiber with extremely large mode area for infrared nonlinear applications,” Opt. Express 16(18), 13651–13656 (2008). [CrossRef] [PubMed]
19. E. Taylor, L. Ng, J. Nilsson, R. Caponi, A. Pagano, M. Potenza, and B. Sordo, “Thulium-doped Tellurite Fiber Amplifier,” IEEE Photon. Technol. Lett. 16(3), 777–779 (2004). [CrossRef]
20. S. Dai, C. Yu, G. Zhou, J. Zhang, and G. Wang, “Effect of OH- content on emission properties in Er3+-doped tellurite glasses,” J. Non-Cryst. Solids 354(12-13), 1357–1360 (2008). [CrossRef]
22. A. Wucher, Sputtering: Experiment (Det Kongelige Danske Videnskabernes Selskab, Copenhagen, 2006).
24. W. Miniscalco, “Erbium-doped glasses for fiber amplifiers at 1500nm,” J. Lightwave Technol. 9(2), 234–250 (1991). [CrossRef]
25. P. Becker, N. Olsson, and J. R. Simpson, Erbium-Doped Fiber Amplifiers: Fundamentals and technology (Academic Press, San Diego, CA, 1997).