1. Introduction

Erbium-doped devices have become a cornerstone in the telecommunications industry, first by seeing their spotlight in erbium doped fiber amplifiers and lasers [1], and now by continuing their development into photonic integrated circuits (PICs). Recently, on-chip rare-earth doped waveguide amplifiers and lasers (REDWAs and REDWLs) have been shown to offer advantages when it comes to power output and stability [2] in comparison to other silicon compatible optical amplifiers and lasers including those based on III-V materials or other processes such as stimulated Raman scattering. When integrated into a silicon platform, many doors are opened for REDWAs and REDWLs in chip scale applications needing compact optical amplifiers and lasers [2–4] for not only telecommunications, but also biological and environmental sensing [5].

Various glass rare-earth host materials are available with suitable lanthanide solubilities which can be patterned and integrated with a silicon substrate. Among these, aluminum oxide has demonstrated itself as a platform for rare earth amplifiers and lasers [2–4,6–9] based on its high transparency, high rare earth solubility, high chemical and mechanical stability and moderate refractive index contrast for compact devices. Waveguide fabrication methods have been well-established using reactive co-sputtering deposition and reactive ion etching [6,10]. Net optical gain of 2.0 dB/cm has been demonstrated in Al₂O₃:Er³⁺ waveguide amplifiers [8], as well as 20 dB net gain for 12.9 and 24.4 cm long Al₂O₃:Er³⁺ spiral waveguides [3] when pumped at 980 and 976 nm respectively. To achieve higher gain per unit length various strategies such as slot waveguides and atomic layer deposition have been applied [9] which have shown 20.1 dB/cm optical gain in a 250 µm long waveguide showing promise for ultra-compact devices.

To improve the pump absorption and efficiency in compact amplifier and lasers while likewise allowing for low-loss waveguide structures, Yb³⁺ ions have been suggested as a co-dopant in the optical host matrix. Yb³⁺ ions have a peak absorption cross section one order of magnitude larger than Er³⁺ ions and absorption spectrum spanning from 850–1000 nm, providing increased likelihood of absorption for pump light [11,12]. The ²F_5/2 excited state of the Yb³⁺ ions resonantly provides energy to the ⁴I_11/2 level of the Er³⁺ ions increasing the pumping efficiency by sensitizing the energy conversion process. The enhanced pump absorption can lead to higher signal output powers [13], and ytterbium’s broad absorption around 940 nm has been proposed as a route to temperature insensitive pumping [14]. Additionally, increasing the Er³⁺ dopant concentration can lead to higher optical gain, but due to clustering and fast quenching [15] of the ion emissions, optical gain in singly-doped Al₂O₃:Er³⁺ amplifiers is limited. It is therefore of interest to fabricate Al₂O₃:Er³⁺:Yb³⁺ waveguide amplifiers with low material losses which from fiber simulations and film work also suggest to potentially mitigate the effects from rare-earth clustering sites [12,13–17]. This reduction in clustering sites provides a pathway to increase the overall doping concentration of Er³⁺, thus increasing the net gain achievable.

Fabricating Al₂O₃:Er³⁺:Yb³⁺ waveguide amplifiers in a low-cost, low-temperature and low environmental impact manner is of interest as it continues to grow as an emerging technology. Al₂O₃:Er³⁺:Yb³⁺ waveguides on silicon have been demonstrated using different deposition methods, including ion implantation of sputtered Al₂O₃ films [13,14] and middle frequency sputtering, which achieved 5.2 dB/cm net gain in a 10.5 mm long waveguide [18]. However, these methods both required high temperature annealing to obtain high optical quality thin films. Although a fine control of morphology is available with atomic layer deposition (ALD) [19,20] its lower deposition rates, higher cost, and lack of process flexibility are less attractive for deposition of thick gain layers (1.0 μm). For these reasons, we have utilized reactive magnetron co-sputtering where individual Er, Yb and Al sputtering targets are placed in a chamber with ambient O₂ and Ar, as used for the singly-doped Al₂O₃:Er³⁺ waveguides demonstrated in [3] and [8]. In this manner, high purity Al and rare-earth targets can be used to minimize impurities, doping of the gain layer can be carried out in-situ at relatively low temperature, relatively fast deposition rates can lead to thick films with homogenous doping profiles, and the Er³⁺ and Yb³⁺ doping concentrations can be finely controlled by separately adjusting the sputtering power on each target. Deposition parameters and conditions were based on studies with a similar system [4] where like here, substrate bias is included to further reduce surface roughness of deposited films. To structure the films, we utilize wet chemical etching in order to avoid the use of chlorine-bromine gas chemistries [10]. Phosphoric acid (85 m.%) was selected for its low cost, availability, and relatively safe nature when diluted [21]. This process allows for the use of conventional contact lithography and is well-suited to typical dimensions employed in rare-earth-doped Al₂O₃ amplifiers and lasers [3,22]. To optimize the relative rare earth concentration and maximize gain, parallel work is also on-going to quantify the photoluminescence, luminescent lifetime, Yb³⁺-Er³⁺ energy transfer efficiency and quenching fraction in co-sputtered Al₂O₃:Er³⁺:Yb³⁺ films [17]. Here, our fabrication process is reported and characterized as well as the material properties and optical gain results for co-doped Al₂O₃:Er³⁺:Yb³⁺ waveguides.

2. Al₂O₃:Er³⁺:Yb³⁺ waveguide fabrication

2.1 Al₂O₃:Er³⁺:Yb³⁺ film deposition and etching

We deposited Al₂O₃:Er³⁺:Yb³⁺ films via reactive magnetron sputtering on three-inch silicon substrates with 6 µm of thermally grown silicon dioxide. Three-inch metallic aluminum, erbium and ytterbium targets of 99.999, 99.9 and 99.9% purity, respectively, were co-sputtered in an argon/oxygen ambient. To improve the quality and optical loss of the films the substrate is heated during depositions by a heating element and is backed by an RF substrate bias to provide thermal/kinetic energy into the films. The dopant concentrations were controlled by selecting the sputtering power applied to the erbium and ytterbium targets and measured using Rutherford backscattering spectrometry (RBS) at Western University’s Tandetron facilities. The concentrations were determined based on RBS measurements in singly-doped calibration films deposited using identical Er or Yb sputtering powers because it was difficult to resolve the separate Er and Yb RBS signals in the co-doped films. It is also noted there is a variation in parameters such as aluminium sputtering power, deposition temperature, and oxygen flow which also affect the rare earth concentrations of the films shown in Table 1. The concentrations in the co-doped films were also verified in absorption measurements and based on the known Er and Yb absorption cross sections [15]. A design of experiments was carried out to obtain high-quality optical films based on the work of Magden et al. on Al₂O₃:Er³⁺ films using a similar deposition system [4]. We note that further optimization of the deposition process can lead to better film properties and losses, particularly at lower wavelengths, and relative rare earth concentration control and work in this direction is on-going. The deposition parameters for the three Al₂O₃:Er³⁺:Yb³⁺ films specifically selected for waveguide experiments are outlined in Table 1 as well as the measured Er³⁺ and Yb³⁺ ion concentrations from singly-doped calibration RBS and absorption measurements. A diagram of the deposition chamber can be seen in Fig. 1(a), with the subsequent patterning steps shown in Fig. 1(b).

 

Fig. 1. (a) Diagram of sputter deposition chamber showing 3 RF magnetron sputtering guns with metallic Al, Er, and Yb targets, argon and oxygen gas inlets, a sample heating element, and a substrate bias plasma. (b) Processing steps for fabrication of Al₂O₃:Er³⁺:Yb³⁺ waveguides.

Download Full Size | PDF

Table 1. Al₂O₃:Er³⁺:Yb³⁺ film deposition parameters

View Table | View all tables in this article

Before etching the Al₂O₃:Er³⁺:Yb³⁺ films used to fabricate waveguides, we performed etch tests on several Al₂O₃:Er³⁺:Yb³⁺ films to verify etch rates, and ensure the photoresist would not degrade in the heated phosphoric acid. We defined large mesa features (2 cm × 2 cm) using contact lithography in an 800-nm-thick S1808 photoresist layer spun on at 3500 rpm with a soft bake of 110°C for 1.5 minutes. We performed exposures using 5.7 mW/cm² 365-nm UV light for 9 seconds, after which the resist was soft baked again, before development in pure MF-319 developer for ∼1 minute. We hard baked the sample at 130°C for 10 minutes to reflow the resist sidewalls. After many samples were prepared in a similar manner, we immersed them in ∼250 mL of 85% pure H₃PO₄ at 60 ± 3°C with a stirring magnet at 1.5 Hz for various amounts of time and rinsed them in de-ionized water after removal. We then assessed the etch depths using a stylus profilometer after removing the resist layer with acetone. We observed repeatable etch rates for films of similar composition, including 27 ± 1 nm/min for a sample Al₂O₃:Er³⁺:Yb³⁺ co-doped film denoted Film 3 as shown in Table 1.

We patterned the Al₂O₃:Er³⁺:Yb³⁺ films using an S1808 photoresist mask and wet etching in phosphoric acid (H₃PO₄). The phosphoric acid isotropically etches the Al₂O₃:Er³⁺:Yb³⁺ films, leading to smooth sidewalls but undercut of the photoresist mask, which is an acceptable tradeoff given the typical etch depths and widths required for rare-earth-doped ridge waveguide amplifiers [8]. Further, it is straightforward, low-cost, and avoids etching in hazardous chlorine and bromine-based gases alternatively used to etch Al₂O₃ waveguides [10].

We then used the contact photolithography and wet etching process to define straight waveguide features of varying widths in an Al₂O₃:Er³⁺:Yb³⁺ film. We formed the waveguide end facets by cleaving, and then removed the photoresist mask with acetone and brief immersion in nitric acid (HNO₃). We inspected the photoresist mask and waveguide profile via scanning electron microscopy (SEM), as shown in Fig. 2(a–c) for select widths of 2.0–5.0 μm. As seen in Fig. 2(d) the sidewall roughness is visibly low, with slight curtaining of the waveguide sidewalls laterally to the propagation direction. For each film, a section intended for optical waveguide measurements underwent an additional fabrication step, where a 1.0-µm-thick glass fluoropolymer (CYTOP) was spun on to act as a top-cladding layer. We performed this spin at 1750 rpm for 40 seconds with post bakes of 50, 80, and 180°C for 10, 30, and 30 minutes respectively.

 

Fig. 2. (a) Photoresist mask on an Al₂O₃:Er³⁺:Yb³⁺ film on oxidized silicon wafer prior to etching. (b) Waveguide facet profile after etching and before photoresist removal showing curved sidewalls. (c) Image of an Al₂O₃:Er³⁺:Yb³⁺ waveguide after photoresist removal. (d) Close-up view of the waveguide sidewall.

Download Full Size | PDF

Here ridge waveguide structures are utilized which also allows for polarization insensitive pump and signal operation [22] around 980 and 1550 nm. After the etch process was validated, we designed and fabricated single mode waveguides and ring resonator structures to be used in characterizing the waveguide losses, polarization dependence, and optical gain. We designed a lithography mask aimed at waveguide loss measurements for transverse electric (TE) and transverse magnetic (TM) modes at 1640 nm as well as gain measurements from 1510–1640 nm. Using previous works as a starting point [3,22], the single mode conditions for TE and TM modes at 980 and 1550 nm were investigated and obtained for a 1.0 μm thick film, 250 nm etch depth and 2.2 μm wide waveguide resulting in a simulated minimum bend radius of ∼2.4 mm. We used the above processing steps to fabricate a series of straight 2.2-μm-wide waveguides with lengths up to 5 cm and ring resonator structures with nominal gaps (i.e. gaps on the contact mask) varying from 0.8–5.0 μm and radii of 3000 μm.

3. Optical characterization

To optically characterize the films and waveguides, three main measurements were performed to extract and confirm the relative rare-earth concentrations, polarization dependent signal background losses, and optical gain from 1510–1640 nm. To do so, a single film was prioritized for each measurement to accommodate the varying waveguide structures required, namely short waveguides for optical absorption measurements, ring resonators of varying gaps for loss measurements, and longer waveguides for gain measurements using 970 nm pumping. Table 2 summarizes these measurements and their results for the three films shown in Table 1.

Table 2. Al₂O₃:Er³⁺:Yb³⁺ film and waveguide characterization summary

View Table | View all tables in this article

3.1 Film and waveguide absorption transmission measurements – Film 1

We investigated the thickness, refractive index, and propagation loss in all three Al₂O₃:Er³⁺:Yb³⁺ films using the prism coupling method for pump and signal wavelengths. Optical techniques were used to measure dopant concentrations for short waveguides patterned into Film 1 to avoid the difficulties in differentiating between the signals of Er and Yb co-dopants, which have similar atomic mass, using RBS [17]. The thickness of Film 1 was determined to be 1070 nm, and the refractive indices were found to be 1.616 and 1.604 at 638 and 1550 nm, respectively. The optical loss of the fundamental transverse-electric (TE) polarized mode of the film was first measured from 1510–1640 nm, where erbium absorption, but no ytterbium absorption, is present, to characterize the erbium dopant concentration. We measured the loss at different wavelengths within this range and compared it to the known erbium absorption cross section of Al₂O₃:Er³⁺ [8], and accounting for the overlap of the optical mode with the Al₂O₃:Er³⁺:Yb³⁺ layer (80%), used the fit to extract the erbium dopant concentration. These optical measurements were then used to verify the concentrations determined from RBS calibration measurements using singly-doped films as shown in Table 1. Figure 3(a) shows the loss measured at each wavelength, while Fig. 3(b) shows the loss versus absorption cross section, giving an erbium ion concentration of 1.7×10²⁰ ions/cm³. Losses measured around 980 nm are the result of both erbium and ytterbium absorption combined. The previously calculated erbium concentration and known absorption cross sections in Al₂O₃:Er³⁺ [15] can be used to calculate the contribution of erbium atoms to the absorption in this range. By subtracting the estimated erbium absorption loss from the total measured loss, the remaining absorption loss is considered the result of ytterbium absorption and background loss. That loss is similarly fit compared to the known Al₂O₃:Yb³⁺ absorption cross sections [15] to extract an ytterbium ion concentration. Here, we measured the loss through a short 0.5-cm-long ridge waveguide (etch depth of 250 nm and width of 2.2 µm), because the large input powers and > 1 cm propagation length required to obtain measurable signal on the prism coupling system resulted in significant Yb³⁺-ion excitation and absorption saturation in the film. Figure 3(c) shows the measured loss and the contribution of the erbium and ytterbium ions to the total loss. Figure 3(d) shows the measured ytterbium-absorption related loss versus cross section, which is fit to extract an Yb³⁺ ion concentration of 1.8×10²⁰ ions/cm³. Using the same fittings (and film loss measurements in the range 950–970 nm where saturation does not occur), the background loss of the film was measured to be 2.3 and 0.6 dB/cm at 980 and 1550 nm, respectively. From these fittings we can conclude a slight underestimation of Yb concentration (∼0.4–0.6 ions/cm³) is obtained from the RBS singly-doped film calibration measurements for Film 1 which is considered for future depositions and relative dopant level optimization.

 

Fig. 3. (a) Measured Al₂O₃:Er³⁺:Yb³⁺ film absorption loss versus wavelength and (b) Er³⁺ absorption cross section to extract an erbium ion concentration of 1.7 × 10²⁰ ions/cm³. (c) The combined total Al₂O₃:Er³⁺:Yb³⁺ waveguide absorption loss measured around 975 nm. The calculated erbium absorption loss (red line) based on the samples measured ion concentration and measured absorption cross sections in [15] is subtracted from the total loss (black dots), leaving the ytterbium-related absorption loss (blue line). (c) The measured ytterbium-related absorption loss in the waveguide fit against the Yb³⁺ absorption cross section at each wavelength, used to determine a 1.8 × 10²⁰ ions/cm³ ytterbium concentration.

Download Full Size | PDF

3.2 Ring resonator loss measurements – Film 2

We measured ring resonators at varying gaps patterned into Film 2 to extract the internal Q factor and waveguide propagation loss for the fundamental TE and TM modes at 1640 nm. We used edge coupling and manual micrometer-controlled stages to launch 1510–1640 nm laser light from polarization maintaining 2.5 μm spot size lensed fibers into the fabricated Al₂O₃:Er³⁺:Yb³⁺ waveguides. To control the launched polarization to the waveguide facet, we adjusted polarization paddles in between measurements. By fitting resonance spectra as shown in Fig. 4(a–b), waveguide losses were estimated for varying wavelengths and launched polarizations. To confirm the TE and TM nature of the modes we observed shifting resonant wavelengths, and consistent free spectral range differences between TE and TM modes. A summary of the ring resonator measurements is shown as insets in Fig. 4(a–b), with overlapped Lorentzian fits to the spectra for TE and TM modes. Figure 4(c–d) demonstrates the simulated TE modes in the measured waveguide at signal and pump wavelengths respectively, while Fig. 4. (e–f) includes a diagram and image of a ring resonator during measurement.

 

Fig. 4. (a) TE and (b) TM transmission spectra for a ring resonator with a 3.0 μm designed gap and overlaid Lorentzian fits for extracting the quality factor. Simulated TE mode at (c) 1550 nm and (d) 980 nm for a 2.2 μm wide waveguide structure with 250 nm etch depth and 1.0 μm thick film. (e) Diagram of ring resonator waveguide coupled structure demonstrating bus and ring region. (f) Photograph of chip during measure with dual side 970 nm pumping used to illuminate ring with characteristic green erbium emission.

Download Full Size | PDF

Figure 4 demonstrates the results from the ring resonator measurements for Film 2. It can be seen in the insets of Fig. 4(a–b) that we observed relatively polarization insensitive losses at 1639 nm, showing similar results within error with 0.9 and 1.0 ± 0.1 dB/cm of waveguide loss estimated for TE and TM modes. To be certain the resonance can be used for an estimation of waveguide loss the ring is required to be in the under-coupled regime. This was verified by sweeping the mentioned gaps for each wavelength and polarization conditions which arrived at a 3.0 μm gap for TE and TM 1640 measurements. The characteristic green glow of the erbium ions in the waveguide during measurement can be seen in Fig. 4(f), while pumping at 970 nm. From these resonance measurements, an upper-limit estimation of waveguide signal losses can be appropriately concluded for the waveguide fabrication process. These measurements were used to estimate the waveguide losses and quantify the internal net gain in a 3.0-cm-long straight waveguide structured in Film 3.

3.3 Optical gain measurements – Film 3

We measured optical gain by using the setup shown in Fig. 5(a). We used dual side pumping with two 970 nm diode lasers to increase the amount of on-chip pump power and fiber wavelength division multiplexers (WDMs) to combine/separate the pump from the 1510–1640 nm signal. The incident pump and signal powers were measured using an integrating sphere, with estimated –4.5 dB coupling losses per facet of the waveguide based on insertion loss measurements. High launched signal powers were achieved with the use of an external EDFA in the set-up. Besides the WDM, separating and filtering of the pump light was carried out with additional free space filtering before reaching the photodetector. We also utilized lock-in amplification to separate the amplified signal from the amplified spontaneous emission (ASE) during measurement. The waveguide transmission spectra for varying pump powers are shown in Fig. 5(b) for patterned Film 3 in a 3.0-cm-long, 2.2 μm wide waveguide with 250 nm etch depth. Waveguide loss estimates based on ring resonator measurements in Film 2 provide an upper limit on propagation losses at 1640 nm as shown in Fig. 4(a). To extract the internal gain in straight waveguides patterned into Film 3, waveguide losses were determined using a combination of insertion loss measurements and upper limits provided by ring resonator measurements in Film 2. Lower insertion loss measurements for Film 3, and the absence of ring coupling and excess bending losses suggest a waveguide propagation loss range which includes the lower limit background film loss and upper limit estimate from fit resonant data. This results in a waveguide loss estimate of 0.5 ± 0.3 dB/cm for Film 3 which was used to quantify the internal gain and associated error margin. Figure 5(c) summarizes this internal net gain for selected wavelengths at different pump powers. The inset shows the internal net gain measured for varying signal powers at 1533 nm at peak pump power.

 

Fig. 5. (a) Diagram of measurement setup and photograph of sample during measurement. (b) Pumped and un-pumped transmission spectra from 1510–1640 nm in 1.0 μm thick Film 3 with a 2.2 μm wide waveguide and an etch depth of 250 nm at varying internal pump powers demonstrating inverted rare earth absorption dip and internal net gain at low-level signal powers. (c) Internal net gain plotted for various wavelengths and pump powers. Peak internal net gain of 4.3 ± 0.9 dB measured at 1533 nm. Internal net gain measured for varying launched 1533 nm signal powers at 190 mW launched pump power (inset).

Download Full Size | PDF

We measured a peak internal net gain of 4.3 ± 0.9 dB at 1533 nm for a 3.0 cm long waveguide, corresponding to 1.4 ± 0.3 dB/cm. We also demonstrated that net internal gain was achieved up to ∼–8.5 dBm launched signal power. It should be noted that with limited pump power on-chip due to high insertion loss, we did not observe gain saturation, which points toward higher gain being possible with decreased coupling and propagation losses. Compared to previous works for singly-doped Al₂O₃:Er³⁺ waveguide amplifiers on Si which demonstrated 1.6 and 2.0 dB/cm gain and were fabricated using similar methods [3,8], this result shows comparable performance with promise moving forward with decreased losses.

Future work aims to demonstrate higher net gain in longer waveguides by decreasing the optical losses for both pump and signal wavelengths. Material related film losses can be optimized with variations in oxygen flow and temperature during deposition [17]. Improved fiber-chip coupling and optimized rare-earth dopant ratios are also expected to lead to superior performance [12] with lower pump thresholds for gain and higher gain per unit length, with potential for temperature insensitive pumping at 940 nm [14]. Additionally, decreasing Er³⁺ clustering sites which have been shown to lead to fast quenching processes in Al₂O₃:Er³⁺ amplifiers [3], may be possible with Yb³⁺ co-doping [16,17]. This can be verified by modelling the results shown here using a rate equation solver for the co-doped matrix, similar to the quenching considerations shown in [3]. By building in parameters obtained from spectroscopic characterization of the Al₂O₃:Er³⁺:Yb³⁺ films [17], such a model will allow for optimization of the relative doping concentrations and waveguide length to maximize amplifier and laser performance.

4. Conclusion

We have demonstrated a flexible and low-cost process for co-sputtering of Al₂O₃:Er³⁺:Yb³⁺ thin films and their subsequent patterning with contact lithography and H₃PO₄ wet-etching. We used SEM and profilometry to assess the etch rates and sidewall roughness as well as facet profile throughout the waveguide fabrication process. Optical pumping via edge coupling was shown to produce 4.3 ± 0.9 dB internal net gain at 1533 nm for a 3.0 cm long, 2.2 μm wide waveguide with a 250 nm etch depth in a 1.0 μm thick film, with Er³⁺ and Yb³⁺ dopant concentrations of 1.4 × 10²⁰ and 2.1 × 10²⁰ ions/cm³, respectively. This gain compares to previous works for Al₂O₃:Er³⁺ on Si waveguide amplifiers fabricated using similar methods and includes pathways to improve. The material platform and fabrication process show promise for producing waveguide amplifiers and lasers with high gain, efficient pump absorption and high output power for telecommunications, LIDAR, biosensors and photonic integrated circuit applications.