1. Introduction

Silicon photonics is a versatile technology platform for high-performance integrated optical circuits, which offer high-speed operation, low energy consumption, dense integration, and low-cost volume manufacturing via mature complementary metal-oxide-semiconductor (CMOS) fabrication technology [1–3]. Recent advances in silicon photonics have gained interest in a myriad of applications spanning from telecommunication to medicine [4], including high-speed modulations [5], nonlinear optical data conversion [6], sensing [7], supercontinuum generation [8], filtering [9], and artificial intelligence [10].

The development of on-chip light sources is required as a crucial component of photonic circuits and systems, especially as they scale up in integration density and complexity [11]. However, due to the indirect bandgap of silicon as a group IV material, efficient light sources in silicon remain a challenge. To overcome these issues, numerous research efforts have been undertaken to propose solutions for silicon photonic on-chip laser sources including electrically and optically pumped lasers [12–17]. However, due to the expensive and complex methods, materials, and fabrication infrastructure, such processes are not entirely embraced in all silicon photonics foundries and multi-project wafer services [18]. Alternative low-cost gain materials and light sources which can be integrated directly on silicon can complement the existing approaches and are also of interest for a variety of prospective applications [19].

On-chip lasers based on optically pumped rare earth gain media have wide gain bandwidths, are low-cost, have straightforward designs, allow for monolithic integration, have high thermal stability, and provide broadband optical gain and lasing in important near-infrared bands [20,21]. Of the prospective rare earth dopants, thulium is of particular interest on silicon platforms because of optical pumping and emission in the silicon transparency range, ultra-broad gain, high efficiency, emission at silicon’s low nonlinear absorption edge, and eye-safe emission in the emerging thulium-doped fiber amplifier (TDFA) band at wavelengths ranging from 1.7 to 2.2 µm [22]. However, roundtrip net optical gain and lasing in on-chip rare earth lasers integrated with silicon have been difficult to achieve because of the combination of relatively low rare-earth gain and high optical loss in silicon waveguides. We have recently demonstrated a single-mode 1906-nm hybrid silicon-thulium microdisk laser integrated on a chip [23]. The laser cavity and output were demonstrated directly in the silicon layer with a 40-µm diameter microdisk design that is compatible with standard passive and active silicon photonic devices. The laser exhibited 580 µW single-sided on-chip output power with a threshold pump power of 16 mW and 2.5 mW with respect to the power coupled into the silicon bus waveguide and microdisk, respectively. Lower threshold lasing is desirable for improved device efficiency and the potential for effectively pumping many lasers on the same chip for applications such as multi-wavelength sources in the emerging TDFA band.

In this work, we demonstrate low-threshold thulium-doped tellurium oxide coated silicon microdisk lasers emitting at wavelengths across the TDFA band. The lasers are monolithically integrated on a silicon-on-insulator (SOI) wafer using a standard foundry process and low-temperature post-processing deposition of the gain medium. We show sub-milliwatt lasing threshold by applying a pulley-coupled bus waveguide design to tune the pump and laser coupling strength. Laser emission is demonstrated at wavelengths spanning from 1825 to 1939nm. Low threshold lasing in a silicon-thulium hybrid micro resonator opens the door to highly efficient and compact monolithic optical circuits for the TDFA band on silicon photonics platforms.

2. Design and fabrication

We designed two sets of silicon microdisks with radii of 20 and 30 µm and pulley-coupled single-mode strip Si waveguides around the resonator creating 20.8-µm effective coupling lengths. The width of the silicon bus waveguides were chosen as 0.45 and 0.4 µm for the 20- and 30-µm radius microdisk resonators, respectively. The gap between the outer walls of the silicon disk and bus waveguide were selected varying from 0.3 to 1.0 µm. The chips were then fabricated on the Advanced Micro Foundry (AMF) wafer-scale SOI platform. Deep ultraviolet 193-nm lithography was used to pattern the silicon microdisk resonators into a 0.22-µm-thick silicon waveguide layer on a 2-µm-thick SiO₂ buried oxide (BOX) layer, without SiO₂ top cladding, allowing for post-processing deposition of the TeO₂:Tm³⁺ gain layer. The silicon waveguides were inversely tapered from 0.45 to 0.18 µm width at the edge of the chip and deep trenches were etched into the BOX and silicon handle wafer to allow for low-loss fiber-chip coupling. The diced silicon chips were then transferred from the foundry for post-processing.

We then used low-temperature reactive radio frequency (RF) co-sputtering to coat the entire structure on six separate chips with thulium-doped tellurium oxide (TeO₂:Tm³⁺) films. Three-inch metallic tellurium and thulium targets with 99.999 and 99.9% purity were sputtered with RF powers of 120 and 60–90 W, respectively, and Ar and O₂ flow rates of 12 and 8.0 sccm, respectively, at 20°C. We measured refractive indices of 2.11 ± 0.03 at 638 nm and 2.05 ± 0.02 at 1550 nm wavelengths of the TeO₂:Tm³⁺ films using spectroscopic ellipsometry. We used the prism coupling method and witness films deposited on thermally oxidized silicon chips to determine thin-film propagation losses of ≤ 0.6 dB/cm at 638 nm and ≤ 0.3 dB/cm at 1510 nm and thicknesses ranging from 320–385 nm. The thickness uniformity of ±2% variability was measured across the wafer. We measured thulium ion dopant concentrations varying from 2.5 × 10²⁰cm⁻³ to 4.9 × 10²⁰cm⁻³ using Rutherford backscattering spectrometry. A cross-section drawing of the hybrid microdisk structure is shown in Fig. 1(a). A top-view scanning electron microscope (SEM) image, zoomed-in SEM view of the coupling region, and focused-ion-beam (FIB) cross-section of the TeO₂:Tm³⁺-coated Si microdisk laser structure, are displayed in Figs. 1(b), 1(c), and 1d, respectively.

 

Fig. 1. a) Cross-section profile of the TeO₂:Tm³⁺-coated Si laser showing the hybrid microdisk and bus waveguide structure. b) Top-view SEM image of microdisk lasers with radii of 20 and 30 µm. c) SEM image of the coupling region showing the pulley coupled design. d) FIB-milled SEM cross-section image of the coupling region between the TeO₂:Tm³⁺-coated Si microdisk and bus waveguide.

Download Full Size | PDF

We calculated the optical properties of the hybrid TeO₂:Tm³⁺-Si resonator structures using a finite-element method mode solver. The calculated microdisk transverse electric- (TE-) polarized fundamental mode profiles are shown in Fig. 2(a). For the microdisk resonator, 15.3% and 19% of the optical power is confined in the TeO₂:Tm³⁺ coating at the 1620-nm pump and 1930-nm laser wavelengths, respectively, while 74.4% and 67.6% of the optical power are confined to the silicon layer, respectively. We selected radii of 20 and 30 µm for compact devices while ensuring minimal bending radiation loss, and exploring the influence of cavity length on lasing behavior close to the minimum bend radius.

 

Fig. 2. a) Calculated electric field profile of the fundamental transverse-electric- (TE-) polarized mode using a finite-element method mode solver for the TeO₂:Tm³⁺-coated silicon microdisk and the strip waveguide at the 1620 nm pump wavelength and 1930nm lasing wavelength. b) Effective material index profile of microdisk and bus waveguide for a 20.8 µm long pulley coupler, with inset showing calculated 2D planar effective indices versus wavelength for the different regions. c) Simulated electric field propagation along microdisk and bus waveguide at 1850nm wavelength for a 10.4 µm long coupler. d) Simulated cross-over ratio between bus waveguide and microdisk versus wavelength for different pulley coupler lengths with inset showing cross-over ratio versus coupler length for potential pump and signal wavelengths of 1620 and 1850nm, respectively.

Download Full Size | PDF

A series of 2D finite-difference time-domain (FDTD) simulations of the bus/microdisk region were carried out to estimate the pump and signal wavelength coupling strength, where the increased wavelength dispersion created by the pulley coupler design’s longer coupling length [24,25] can improve the independent optimization of stronger pump coupling with weaker signal coupling to improve laser performance. To create a 2D simulation model the chip surface was broken into three distinct planar-waveguide regions. Firstly, areas away from the silicon layer where a 385-nm-thick layer of TeO₂:Tm³⁺ lies directly on the buried oxide. Secondly, the areas with silicon where the same thickness of TeO₂:Tm³⁺ lies above a 220-nm-thick layer of silicon on the buried oxide. Lastly, the transition region between the two, where a 605-nm-thick layer of TeO₂:Tm³⁺ lies directly above the buried oxide along the waveguide and disk edges. For each of these regions a calculation of the effective index for the fundamental transverse-electric polarized mode in the planar layer stack was made for wavelengths from 1400 to 2000nm, with the effective index for each region found in the inset of Fig. 2(b). The simulation window was set out to replicate the bus/microdisk coupler, with each layer stack region having a material region based on the planar mode calculation, as shown in Fig. 2(b). A simulation was then run with light initially sent through the bus waveguide and propagating through the system as shown in Fig. 2(c). The far side of the microdisk below the simulation window was set to be a perfect absorption region, such that only the coupling strength into the disk from the bus waveguide and no round-trip interference conditions were considered. The power confined in the bus waveguide was monitored before and after the microdisk coupler region of the simulation, with any difference before and after the microdisk simulation assumed to be power coupled into the microdisk, used to calculate the power coupling cross-over ratio of the coupler. The cross-over ratio simulation method was then carried out for a 20 µm radius microdisk with pulley couplers between 0 to 26 µm in length at a coupling gap of 0.4 µm for wavelengths from 1400 to 2000nm and the results are shown in Fig. 2(d). From the data, it is seen that for short coupling lengths longer wavelengths have higher cross-over ratios compared to shorter wavelengths, which is due to the larger evanescent field at longer wavelengths increasing the coupling coefficient. However, once the pulley-coupler reaches a length of 20.8 µm for this respective gap, a significant increase in the coupling strength around the 1620 nm pump wavelength can be achieved, while reducing the coupling strength of the lasing signal around 1900nm. This increases the pump coupling efficiency while reducing the roundtrip loss at the lasing wavelength to lower the lasing threshold. The variation in coupling strength and gain with wavelength across thulium’s emission band can also be used to shift the lasing wavelength for different pulley coupler designs. We note that the double peak for 26.0 µm a pulley coupler length of 26.0 µm might be in part due to coupling to higher order microdisk modes and/or limitations in the simulation resolution and accuracy and bears further investigation for optimized designs. Here, we simply used the calculated pulley coupler properties for a length of 20.8 µm as a guideline for the expected trend and choosing our initial designs. Overall, these simulations motivate the use of a pulley-coupled design for this lasing structure to increase coupling length with the potential for improved thresholds and lasing over a wider wavelength range.

3. Experimental setup

Schematics of the experimental setups used for the characterization of the passive and active microdisk laser properties are shown in Figs. 3(a) and 3(b), respectively. For passive transmission measurements, the light was coupled to the chip from a tunable 1510–1640 nm or 1860–2000nm laser via a polarization controller and 2-µm spot size tapered optical fiber at 1550-nm wavelength mounted on an xyz stage. The output was also coupled from the chip using a tapered fiber with a 2-µm 1550-nm spot size and coupled to a photodetector to measure and characterize the transmitted light in the two wavelength ranges. For active laser measurements, we set the tunable Agilent laser at 1620 nm and connected it to a high-power L-band erbium-ytterbium-co-doped fiber amplifier (L-band EYDFA) to amplify the pump light. The pump light was then coupled to the chip through a polarizer paddle to select TE-polarization, a 1600/1900nm fiber wavelength division multiplexer (WDM), and a 2-µm spot-size tapered fiber via an xyz alignment stage. The emitted signal of the microdisk laser was coupled from the chip via another tapered fiber and directed through a second 1600/1900nm WDM to an optical spectrum analyzer (OSA) with a resolution of 0.1 nm to capture the spectrum and measure the microdisk laser output power. The power meter was used during the experiment to monitor and optimize the pump coupling condition by adjusting the tunable laser wavelength onto resonance.

 

Fig. 3. a) Passive and b) active experimental setups used for measuring the on-chip silicon-thulium microdisk lasers.

Download Full Size | PDF

4. Results

4.1 Passive microdisk characterization

We carried out transmission measurements to characterize the passive properties of the microdisk resonators. Figures 4(a) and 4(b) show sample transmission measurements from 1520–1620 nm with resonances corresponding to several quasi-TE modes for 20- and 30-µm radius microdisks, respectively, with a TeO₂:Tm³⁺ layer thickness of 385 nm, gap of 0.4 µm and Tm concentration of 4.3 × 10²⁰cm⁻³. The fiber-chip coupling loss in this range was determined to be 7.0 ± 0.2 dB per facet, influenced by mode mismatch, Fresnel reflections, and scattering due to the conformal TeO₂:Tm³⁺ coating on the facet.

 

Fig. 4. Passive TE-polarization transmission measurements near the thulium ∼1.6 µm absorption edge in a) 20- and b) 30-µm radius microdisk resonators showing resonances corresponding to multiple microdisk modes. Transmission measurements over a narrower range for the same c) 20- and d) 30-µm radius resonators. Insets: Fine scans showing for the indicated resonance showing fitted Lorentzian function and internal and loaded Q factors.

Download Full Size | PDF

Figures 4(c) and 4(d) show narrower scans at the edge of the thulium absorption range. The insets show zoomed in views of the indicated under-coupled resonances, determined to have the highest Q factor and corresponds to the fundamental quasi-TE mode. To estimate the waveguide propagation loss from the measured resonance, we fitted a Lorentzian curve to obtain the Q factor. For the devices shown in Figs. 4(c) and 4(d), the internal Q factors were fitted as ${Q_i} = 9.6 \times {10^5}$ at 1530.43 nm and $9.4 \times {10^5}$ at 1526.11 nm for the 20 µm and 30 µm radius microdisks, respectively, corresponding to propagation losses of 0.60 dB/cm and 0.54 dB/cm, respectively, using [26]:

Figures 5(a) and 5(b) show sample transmission measurements around 1900nm near the thulium laser emission wavelength. The fiber-chip coupling losses in this wavelength range were determined to be 6.7 ± 0.1 dB. The insets show fine sweeps of the highest Q factor resonances and fitted Lorentzian curves to obtain the Q factor and loss. As shown in Fig. 5, internal quality factors of 4.6 × 10⁵ and 4.8 × 10⁵ were measured at 1900nm for 20 µm and 30 µm radii of microdisk resonators, respectively, corresponding to losses of 0.97 and 0.90 dB/cm at the laser wavelength.

 

Fig. 5. Passive TE-polarization transmission measurements near the thulium ∼1.9 µm laser emission wavelength in a) 20- and b) 30-µm radius microdisk resonators. Insets: Fine scans showing for the indicated resonance showing fitted Lorentzian function and internal and loaded Q factors.

Download Full Size | PDF

In Figs. 6(a) and 6(b), we present the FSR and internal quality factor versus wavelength from 1510–1630 nm, respectively, for microdisk lasers with 20 µm and 30 µm radii. The inset in Fig. 6(a) shows similar FSR data measured around 1900nm (Q vs. wavelength data was not taken around 1900nm due to the significantly longer fine scan times for the 1860–2000nm tunable laser). The FSRs were measured as 5.0 nm and 3.53 nm around 1510 nm increasing to 8.6 nm and 5.9 nm at the ∼1900nm laser wavelength for 20 and 30 µm radius microdisk resonators, respectively. The quality factor decreases at longer wavelengths with the onset of stronger thulium absorption around 1550 nm. As shown in Fig. 5, lower Q factors are observed around 1900nm which can be attributed in part to the Tm³⁺ ion absorption tail [27] and less confined modes resulting in stronger interaction with the microdisk sidewalls [28].

 

Fig. 6. Passive properties of 20 and 30 µm radius silicon microdisk lasers with 0.4 µm gap sizes. a) Free spectral ranges (FSRs) of the microdisk modes at pump and laser (inset) wavelengths. b) Microdisk intrinsic quality factor versus wavelength for pump and laser (inset) wavelengths.

Download Full Size | PDF

4.2 Laser measurements

The TeO₂:Tm³⁺-coated Si microdisk resonators were resonantly pumped to investigate their lasing potential under 1620-nm pumping at room temperature with no external thermal stabilization of the laser chip. A maximum of 19.5 mW pump power was launched into the bus waveguides to avoid damaging the facets. We typically observed single-mode and multi-mode lasing in 20 and 30 µm radius resonators, respectively. As shown in Figs. 7(a) and 7(b), respectively, single-mode lasing was observed at 1930nm with up to 2 mW double-sided on-chip power in a 20 µm radius microdisk resonator and multi-mode lasing was observed at 1872nm with up to 2.6 mW double-sided (1.3 mW per facet) on-chip power in a 30 µm radius microdisk resonator, with both devices having a gap of 0.4 µm and Tm concentration of 4.3 × 10²⁰cm⁻³. The lasers were measured at room temperature with continuous pumping at ∼20 mW power for more than 5 hours, confirming their stability.

 

Fig. 7. Single- and multi-mode laser emission spectra of TeO₂:Tm³⁺-coated Si microdisk resonators under 1620-nm pumping at a microdisk-waveguide gap of 0.4 µm and a) radius of 20 µm obtained with 18.4 mW on-chip pump power with lasing at 1930nm and b) radius of 30 µm obtained with 19.5 mW on-chip pump power with lasing at 1872nm. Insets in a) and b) show the residual pump and laser emission.

Download Full Size | PDF

The output power of the laser versus launched and absorbed pump power was measured and displayed in Fig. 8. The on-chip laser output power is considered to be the power in the hybrid silicon bus waveguide, by taking the power measured at the OSA and accounting for the output WDM insertion loss and fiber-chip coupling loss at the laser wavelength. Low threshold lasing is observed, and the laser output signal is highly stable at room temperature without thermal control of the substrate, or any adjustment of the pump wavelength required, even though the device is pumped on a narrow resonance. The launched pump power was determined by measuring the incident power from the input fiber using an integrating sphere photodiode power monitor and accounting for fiber-chip coupling loss at the pump wavelength. The extinction ratio at the laser wavelength is 3.2 dB and 3.5 dB in the microdisk resonator with the radii of 20 µm and 30 µm, respectively, which means that 52% and 55% of the power in the bus waveguide are absorbed. The single-sided laser slope efficiency versus launched and absorbed pump power are 5.9% and 11.3%, respectively, with the threshold pump power of 997 µW and 531 µW in the bus waveguide and microdisk resonators, respectively, in the 20 µm radius microdisk resonator, as shown in Fig. 8(a).As shown in Fig. 8(b), a threshold pump power of 918 µW launched into the bus waveguide and 505 µW coupled into the microdisk resonator with 30 µm radius was observed. The single-sided laser slope efficiency versus launched and absorbed pump power are 6.7 and 12.2%, respectively. The laser output is bidirectional, and similar output power was observed at the pump input side of the chip. A maximum double-sided output power of ∼2.6 mW was obtained for a launched pump power of ∼19.5 mW, for a 30 µm radius.

 

Fig. 8. On-chip laser power vs. launched pump power for a) 20 and b) 30 µm radius microdisks showing sub-milliwatt threshold with respect to bus-waveguide-coupled (launched) and microdisk-coupled (absorbed) pump power. Insets: zoomed in view of the lasing threshold region.

Download Full Size | PDF

We investigated microdisk lasers with different gaps and thulium concentrations. The maximum output power and threshold pump power for microdisks with gaps of 0.3, 0.4, 0.6, and 0.8 µm are shown in Fig. 9(a). The threshold pump power is lowest, and the output power is highest for smaller gaps of 0.3 µm and 0.4 µm. The inset shows the maximum single-sided on-chip output power for microdisks with different thulium ion dopant concentrations ranging from 2.5 × 10²⁰cm⁻³ to 4.9 × 10²⁰cm⁻³. The highest laser output power and lowest thresholds were observed for a disk-bus gap of 0.4 µm and pumping at the fundamental TE mode around 1620 nm with the thulium concentration of 4.3 × 10²⁰cm⁻³. We show the laser spectra obtained for microdisk resonators with different gaps and a Tm concentration of 4.3 × 10²⁰cm⁻³ in Fig. 9(b). Lasing is observed at wavelengths spanning from 1825 to 1939nm. Low threshold lasing requires strong pump coupling and relatively low laser output coupling and the low threshold observed at a gap of 0.4 µm agrees with the coupling simulations in Fig. 2(d). For the peak launched pump power used in our experiments, we also see the highest output powers for the lowest threshold devices. As illustrated by the sample calculated coupling curves in Fig. 2(d), for a particular pulley-coupler length and gap the ring resonator modes can incur greater coupling losses at longer wavelengths, which favors lasing closer to the thulium gain peak around 1850nm. For a different design with a weaker coupling condition, the laser output can shift to longer wavelengths where the Tm³⁺ absorption is lower, and population inversion and the threshold condition are more easily achieved. Besides the coupling loss, the total roundtrip gain also depends on the resonator radius and volume which thus also influences the laser threshold, output power, and optimal lasing wavelength. Therefore, by pumping resonators with different radii and pulley-coupler gaps, different lasing thresholds, powers, and wavelengths can be achieved.

 

Fig. 9. a) Maximum on-chip laser output power and launched threshold power versus bus-microdisk gap size for microdisks with a thulium ion dopant concentration of 4.3 × 10²⁰cm⁻³. Inset: on-chip laser output power versus thulium ion concentration for a gap of 0.4 µm. b) Laser emission spectra of TeO₂:Tm³⁺-Si microdisk resonators with a concentration of 4.3 × 10²⁰cm⁻³ and varying microdisk-waveguide gaps and radii, showing emission from 1825 to 1939nm.

Download Full Size | PDF

5. Conclusions

In this study, we demonstrate ultra-compact hybrid silicon-thulium microdisk lasers integrated on an SOI platform. We observe bi-directional slope efficiencies of up to 13.4% versus power launched into the bus waveguide and on-chip output powers of up to 2.6 mW. On account of their high Q factors and engineered pump and laser coupling via a pulley-coupled bus waveguide design, the lasers exhibit threshold pump powers as low as 505 and 918 µW with respect to the power coupled into the silicon microdisk and bus waveguide, respectively. In addition, we observe lasing across thulium’s broad gain spectrum at wavelengths from 1825–1939nm in microdisks with different gaps and radii. These results are promising for broadband amplification and devices such as low-threshold WDM light sources and widely tunable lasers on SOI chips in the emerging TDFA band.