1. Introduction

In the last few years, a considerable effort has been devoted to the research activity on erbium-doped waveguide amplifiers (EDWAs), because of the great potential impact of these devices on new system architectures in metropolitan and local-access optical networks [1]. Several fabrication processes, such as silica-on-silicon, ion diffusion and sol-gel, are nowadays available for high quality mass production, but novel approaches, recently adopted, have given proof of good reliability as well as new capabilities.

Direct femtosecond laser writing of waveguides and photonic devices is a very promising technique [2–7], allowing for totally new potentialities, like simple fabrication of 3-D structures, fast prototyping and flexible piece-by-piece fabrication. Up to now, most studies have concerned the understanding of the writing process in various dielectric materials and using different femtosecond laser sources; only recently passive devices with performance close to telecom standards have been presented [8–10]. Femtosecond laser waveguide writing has also been applied to active glasses, and internal gain [11,12], net gain [13], and laser action [14] have been reported.

In this paper we demonstrate a femtosecond laser written EDWA providing net gain in the whole C-band, with a peak internal gain of 9.2 dB at 1535 nm and minimum internal gain of 5.2 dB at 1565 nm. The amplifier has been characterized in terms of insertion losses, gain spectrum, saturation power and noise figure at different wavelengths. Our results approach those attainable with standard techniques and represent an important step towards the industrial application of the femtosecond writing technique for low-cost, mass production of photonic devices.

2. Experimental setup

2.1 Waveguide writing setup

The femtosecond laser used in the experiments is a diode-pumped, cavity-dumped Yb:KYW oscillator at 1040 nm [15,16]. The laser cavity is a z-fold configuration stretched by curved mirrors, resulting in a 22-MHz repetition rate. A semiconductor saturable-absorber mirror, together with chirped mirrors, provides passive mode locking in the soliton regime. The cavity includes an electro-optic cavity dumper with high speed driving electronics, allowing the extraction of 350-fs pulses with energy up to 1 μJ at frequencies exceeding 1 MHz. This laser offers two important advantages: (i) compactness and reliability thanks to diode pumping; (ii) high output energy without amplification stages.

The active material used for fabricating the amplifier is a phosphate glass base (Kigre Inc.) which has a suitable composition for femtosecond waveguide inscription [12–14]. Dopant concentrations (2% wt of Er₂O₃ and 4% wt of Yb2O3) have been optimized to obtain high gain per unit length in order to fabricate compact devices.

To reach intensities high enough to cause material modification and to obtain waveguides with good matching to optical fibers [13], a very high numerical aperture objective (100x oil-immersion Zeiss Plan-Apochromat, 1.4 NA) is used which focuses the femtosecond pulses 170 μm inside the glass. A transverse writing configuration, in which the sample is translated by motorized stages (M.511.DD, Physik Instrumente) in a direction perpendicular to the laser beam, is adopted. Both end-facets of the waveguides are polished after laser inscription.

2.2 Amplifier characterization setup

In Fig. 1 a schematic of the EDWA configuration and characterization set-up is presented. The 37-mm-long active waveguide is butt-coupled on both sides to standard telecom single-mode-fibers (SMFs), using an index-matching fluid able to support high power density at 980 nm. A couple of 975-nm wavelength InGaAs laser diodes in a bi-propagating pumping scheme supply up to 460 mW (260 mW + 200 mW) incident pump power through 980/1550 nm wavelength-division-multiplexers (WDMs). Signal radiation at 1.5 μm is provided by a tunable laser-diode or a broad band source, and the signal level is finely adjusted by means of a variable attenuator.

 

Fig.1. Optical waveguide amplifier configuration and characterization set-up.

Download Full Size | PDF

3. Experimental results and discussion

3.1 Waveguide characterization

Waveguides were manufactured for different values of pulse energy, repetition rate and translation speed. A set of parameters giving good waveguide performance was identified: repetition rates between 500 kHz and 1 MHz, pulse energies between 200 nJ and 500 nJ, and translation speeds between 50 μm/s and 300 μm/s. Pulse energies lower than ~200 nJ and translation speeds higher than ~300 μm/s led to refractive index changes too low (< 10^-3) for good mode matching with single mode fibers. Energies higher than ~500 nJ led to the formation of voidlike structures. At repetition rates higher than ~ 1 MHz, the occurrence of unstable thermal effects spoils the waveguide uniformity. The waveguide investigated in this paper was written at a repetition rate of 885 kHz, with a 250 nJ pulse energy and a translation speed of 50 μm/s. Total insertion losses, comprising coupling and propagation losses, were measured comparing the signal output level at 1600 nm wavelength, outside the erbium absorption band, with and without the waveguide inserted, and turned out to be ~1.9 dB. Assuming ~0.25 dB/facet coupling loss between SMF and waveguide (calculated from the overlapping integral of the measured mode-field profiles), a propagation loss of 0.4 dB/cm was estimated. This figure is improved by a factor of two with respect to our previous study [13,14] employing a laser with 166 kHz repetition rate. This advance is attributed to cumulative heating effects, arising at a higher repetition rate, which increase the waveguide homogeneity, in agreement with recent results by Eaton et al. [8]. We expect that the propagation losses can be further improved by a better control of the uniformity in the translation speed and by minimizing mechanical vibrations.

3.2 Small signal gain measurements

Using a broad band source, properly attenuated (about -3 dBm over 100 nm bandwidth) the small signal internal gain of the amplifier as a function of wavelength was measured with an accuracy of about 0.5 dB. The results are shown in Fig. 2: the maximum internal gain is 9.2 dB at 1535 nm, corresponding to 2.5 dB/cm. Taking into account the total insertion losses, indicated by a dashed line in Fig. 2, a net gain is demonstrated from 1530 to 1580 nm, spanning the whole C-band of optical communications with 3.6 dB net gain at 1565 nm, and 7.3 dB peak value at 1535 nm. We note that there is still potential for a further improvement of net gain by both reducing the insertion losses and optimizing the doping concentrations of Er and Yb. The wavelength dependence of the measured internal gain is in fairly good agreement with that obtained in ion-exchanged waveguides produced in the same host glass [17]. This result demonstrates that the femtosecond laser inscription of waveguides, although locally altering the material structure, does not influence significantly the gain spectrum.

 

Fig. 2. Measured internal gain spectrum obtained with an incident pump power of 460 mW in bi-directional pumping configuration. The dashed line indicates the total insertion losses.

Download Full Size | PDF

The small-signal internal gain versus incident pump power is reported in Fig. 3 for two different signal wavelengths, namely 1535 nm and 1550 nm. The small signal (about -20 dBm) was provided by an attenuated tunable laser. In the first part of the curves, between zero and 200 mW of incident pump power, the pump radiation is provided by the counter-propagating pump only. In the second part of the curves the co-propagating pump is operating as well, providing up to 460 mW total power. The pump power at the zero internal gain point is 220 mW and 150 mW at 1535 nm and 1550 nm, respectively. The discontinuity in the curves is mainly due to the fact that each pump inverts about half of the amplifier; therefore turning on the second pump a steep gain increase is observed [18].

 

Fig. 3. Internal gain of the 37-mm long Er-Yb waveguide amplifier as a function of the incident pump power for two different signal wavelengths: 1535 nm (triangles) and 1550 nm (circles).

Download Full Size | PDF

3.3 Gain saturation measurements

To investigate the amplifier performance in the saturated regime, we measured the amplifier gain at different signal power levels. The internal gain as a function of the signal output power for three different wavelengths is shown in Fig. 4. The saturation output powers (corresponding to a 3 dB gain reduction) are 8.4 dBm, 12.9 dBm and 19.0 dBm at 1535 nm, 1550 nm and 1565 nm, respectively. As expected, the observed saturation powers are higher for longer wavelengths, basically because of the lower Er emission cross section. More precisely, the signal saturation power of a three-level system is proportional to (1/σ_E)(1+P_p/P_p0), where σ_E is the erbium emission cross section, P_p and P_p0 are the pump power and the pump power at zero internal gain, respectively [19]. In Kigre QX phosphate glass the emission cross section at 1550 nm is roughly 50% less than at 1535 nm.

 

Fig. 4. Internal gain saturation curves measured at three different wavelengths: 1535 nm (triangles), 1550 nm (circles) 1565 nm (squares). Pump power is 460 mW.

Download Full Size | PDF

The measured saturation power at 1550 nm is 4.5 dB higher than that at 1535 nm, in fairly good agreement with the expected difference of 4.2 dB obtained from the previous formula with the measured pump power at the zero internal gain point. It may be noted that the three saturation curves intersect at about 9.5 dBm output signal power, where the internal gain results to be almost independent of signal wavelength, and equal to about 4.7 dB over the whole C-band (i.e., the gain is almost equalized when the amplifier is fed by an input signal of about 4.8 dBm).

3.4 Noise figure

Figure 5 shows the noise figure of the optical amplifier at different wavelengths. Experimental values (triangles) have been obtained from a standard procedure, based on the measurement of the amplified spontaneous emission noise floor in a 0.5 nm bandwidth around the signal peak at different wavelengths. It is assumed that the signal to noise ratio of the input radiation (from the tunable laser source) is limited by quantum noise [20]. These results are compared with a theoretical curve (dashed line) computed from the approximate relation [20]:

where G is the measured net gain of the amplifier, σ_A and σ_E are the erbium absorption and emission cross sections measured for Kigre QX glass, N₂ and N₁ are the upper and lower energy levels populations of the 1.5 μm erbium transition, corresponding to a population difference of 35%. This value, which confirms that the amplifier was well inverted (N₂~65%, close to erbium population saturation), was estimated [20] from the ratio between the absorption and internal gain values at 1535 nm, equal to 17.6 dB and 9.2 dB, respectively, and taking into account the ratio between σ_A and σ_E at the same wavelength, which is about 0.76 in our glass. We can observe the fall of the noise figure below the 3 dB level for wavelengths longer than 1560 nm, where the gain is noticeably decreased, as expected in short, heavily-doped erbium amplifiers, which are well inverted low gain devices [20].

 

Fig. 5. Waveguide amplifier noise figure at different wavelengths (triangles). Theoretical curve (dashed line) is calculated from the estimated value of the effective inversion level (35%).

Download Full Size | PDF

4. Conclusions

In conclusion, we demonstrated an EDWA fabricated by the femtosecond laser writing technique, operating in the whole C-band of optical communications. The amplifier provided up to 9.2 dB peak internal gain at 1535 nm and 5.2 dB minimum internal gain in the whole C-band. Saturation output powers greater than 8 dBm and full amplifier characterization were also reported. Even if these figures still need to be improved to compete with those of commercially available devices, it is noteworthy that photonic devices fabricated by femtosecond lasers have now reached a performance level sufficient to be considered for telecom applications and directly compared to devices fabricated with conventional techniques.