Rare-earth ion doped Al2O3 for active integrated photonics

ABSTRACT Aluminum oxide (Al2O3) is an emerging material in integrated photonics. It exhibits a very broad transparency window from the UV to the mid-IR, very low propagation losses and a high solubility for rare-earth ions leading to optical gain in different spectral ranges. Al2O3 can be deposited by different wafer-level deposition techniques, including atomic layer deposition and reactive magnetron sputtering, being compatible with the monolithic integration onto passive integrated photonics platforms, such as Si3N4, to which it provides optical amplification and lasing. When deposited at low temperatures, it is also compatible with integration onto CMOS chips. In this review, the state-of-the-art on the deposition, integration and device development in this photonic platform is described. Graphical Abstract Graphical abstract

In this review, we first give an overview of the most used techniques for the deposition of rare-earth ion doped Al 2 O 3 . We then review the approaches that have been explored for the integration of Al 2 O 3 to both SOI and silicon nitride. An overview of developed devices realized both completely on Al 2 O 3 as well as on Al 2 O 3 integrated onto a passive platform, will be given. We divide the devices section into amplifiers and lasers, independently of the application field and of the rare-earth ion used as dopant. The review focuses on the developments undertaken during the last decade.

Atomic layer deposition of optical quality Al 2 O 3 layers
Atomic layer deposition (ALD) is a chemical vapor deposition technique in which alternating gas precursors are flown over the substrate reacting with their surface in a self-terminating way, thereby growing a thin film [11,[68][69][70][71][72][73]. In between gases, a purging step is applied to eliminate any unreacted species before the start of the next cycle. The precursors utilized in ALD are always flown independently, in contrast with other chemical vapor deposition techniques. An ALD cycle consists of a series of sequential reactant-purge steps. Each reactant-purge step constitutes a half-cycle. One of the advantages of ALD is the control over the deposition rate, which can be as accurate as the thickness of a half-cycle. The sub-monolayer control over the deposition allows engineering the deposited material to, for example, control the separation of rare-earth ion dopants to limit luminescence quenching effects that limit the achievable gain [27,73]. The great control over the deposition permits to achieve a great degree of conformality, allowing to fill in very small gaps, which has been exploited in slot-waveguide configurations such as the one shown in Figure 1. However, since each cycle typically produces a thickness on the order of 1-1.2, deposition of micrometer thick layers requires very long deposition times. (Figure 1) The deposition of Al 2 O 3 films by atomic layer deposition is one of the most well-known processes in the semiconductor industry. The most used precursors for this process are trimethylaluminum (TMA) and water (H 2 O). Nitrogen purging is usually reported for up to 8 sec per purge [66]. Deposition temperatures between 150 and 300 C have been investigated and the effect of the deposition temperature on the quality, i.e., refractive index and crystallinity, of the deposited layers has been reported [11]. The refractive index increased from 1.629 to 1.639 at a wavelength of 1550 nm for layers deposited at 150 C and 300 C, respectively. All layers were predominantly amorphous with only the layers deposited at 300 C exhibiting the signature of amorphous α-and γ-Al 2 O 3 in its XRD spectrum. Losses below 0.1 dB/cm were measured in slab waveguides from 632.8 nm till 1550 nm for the film deposited at 300 C. Both Alsan et al. [66] and West et al. [2] deposited their Al 2 O 3 films using ALD at 300 C, reporting similar refractive indices and slab losses. Channel waveguide losses below 3 dB/cm at 400 nm were reported by West et al. [2] showing the potential of this material for broad bandwidth applications down to the UV. Rönn et al. showed that annealing above 600 C increases the luminescence of the films when doped with erbium. However, as the annealing temperature is increased above 800 C, a polycrystalline γ -Al 2 O 3 phase starts forming, which leads to high propagation losses (>20 dB/cm) [2]. An annealing temperature of 750 C was the optimum selected by Rönn et al. to achieve the maximum photoluminescence of the rare-earth ions with minimum losses [27,43].
ALD permits engineering the deposited layers by controlling the composition of each ALD cycle. In that way, the optical, electronic and morphological properties can be controlled by adjusting either the bilayer thickness or the relative thicknesses of the layers in the stack [70]. Al 2 O 3 -Y 2 O 3 multilayers were investigated for integrated optical waveguides with tailorable bandgap [70]. Al 2 O 3 -Er 2 O 3 multilayers were studied to maximize optical gain for onchip amplification in the C-band [27,28,73]. Al 2 O 3 -ZnO multilayers were exploited to produce highly non-linear optical waveguides [85,86].

Reactive magnetron sputtering of optical quality Al 2 O 3 films
Reactive magnetron sputter deposition of Al 2 O 3 is based on the sputtering of metallic aluminum (Al) from a target that subsequently undergoes a chemical reaction with the reactive gas, in this case O 2 , on the substrate and walls of the reaction chamber to deposit an Al 2 O 3 layer. Inert argon ions are generally used to sputter the aluminum atoms from the target. During the sputtering process not only aluminum atoms are released from the target but also secondary electrons. An electric potential (i.e., bias voltage) between the target and the chamber wall, which acts, respectively, as cathode and anode, attracts the argon ions to the target. The emitted electrons become energized and are accelerated away from the target ionizing the neutral argon gas. A strong bended magnetic field on the target surface traps the electrons close to the target to enhance the overall sputtering yield. DC reactive sputtering suffers from continued localized charge build up when the target is partially oxidized, with the consequent reduction in optical quality of the deposited layers due to arcing [74]. RF reactive sputtering has been demonstrated to lead to high optical quality layers with low optical propagation losses. Optical propagation losses of undoped Al 2 O 3 slab waveguides below 0.4 dB/cm at 633 nm have been reported [12]. After etching, channel waveguides with typically less than 0.2 dB/cm at 1550 nm have been fabricated [6,9].
The optical and morphological properties of the sputtered layers depend largely on the deposition parameters, including the substrate temperature, oxidation state of the sputtering target and impact of energetic species on the substrate. The bias voltage can be measured as a function of oxygen flow or oxygen partial pressure in the chamber to determine the oxidation state of the target [87,88]. The sputtering mass deposition rate is directly related to the oxidation state of the target since metallic aluminum has a higher sputtering yield than oxidized aluminum (i.e., alumina) [88]. An oxidized target exhibits also higher secondary electron emission and sputtering of negative species such as O − , which are accelerated towards the substrate and add energy to the deposited material affecting its morphology [89,90]. Figure 2 shows a typical bias curve for RF reactive sputtering of Al 2 O 3 . For low oxygen flows, the target is fully metallic (i.e., zone 1). The sputtered aluminum reacts with the O 2 on the chamber walls and substrate to produce Al 2 O 3 (i.e., 'gettering'). The argon bombardment on the target keeps it 'clean' from formation of an Al 2 O 3 layer. As the O 2 flow increases within Zone 1, the bias voltage stays stable and the oxygen partial pressure remains stable for increasing oxygen flow into the chamber. After a certain O 2 flow, known as the 'knee point' or 'sputtering knee point', the gettering capacity of the walls and substrate saturates and the oxygen partial pressure starts increasing with an increased oxygen flow. The excess O 2 in the chamber contributes to an increase of the oxidation of the target. Since the argon flow is constant, the argon bombardment is not sufficient to keep a metallic target by removing the oxidation on the target surface. Partial oxidation of the target starts, with an increase of secondary electron emission due to the higher secondary emission yield of oxidized aluminum. Keeping a constant power results in a drop of the bias voltage (i.e., zones 2 and 3 in Figure 2). The deposition rate drops due to a lower sputtering yield of Al 2 O 3 versus Al and a decreased discharge voltage, reducing the incident energy of the argon ions on the target. This is a self-reinforcing loop that converges to the right side of the bias curve (i.e., zone 4), in which the target is fully oxidized ('poisoned') and the deposition rate is significantly decreased. This process can exhibit hysteresis [91], which prevents achieving an optimal and stable process in the transition region (i.e., zones 2-3) of the bias curve [90]. Bobzin et al. reported the different material structures expected for depositions in the different zones of the bias curve. A full description of morphological zones as a function of deposition parameters can be found in [89]. By controlling the oxidation state of the sputtering target, Al 2 O 3 layers with the desired morphology could be deposited in a reproducible manner with high optical quality [75]. Cathode voltage (also known as bias voltage) versus oxygen flow in the reaction chamber (left axis) and deposition rate as a function of oxygen flow (right axis). Different zones are indicated, namely the zone 1, which corresponds to a metallic target and deposition of not fully oxidized aluminum, zones 2 and 3, corresponding to a metallic target and fully oxidized layers, and zone 4, which corresponds to an oxidized ('poisoned') target. Reprinted from [90] with permission from Elsevier.
The Al 2 O 3 can be doped by using a second target [7,9]. As already mentioned before, Yb 3+ , Nd 3+ , Er 3+ , Tm 3+ and Ho 3+ have already been utilized in practical devices. The influence of doping concentration, in particular of thulium, on the refractive index of RF reactive co-sputtered Al 2 O 3 has been reported by Loiko et al. [92].
Typically, reported deposition temperatures are relatively elevated (i.e., above 450 C). The high temperature limits the type of substrates over which the material can be deposited. By applying a voltage to the substrate, excellent guiding (i.e., losses ~0.1 dB/cm at 1550 nm) Al 2 O 3 slab waveguides were deposited at 250 C [37]. In that study, 90 V was the maximum voltage that could be applied to the substrate. The combination of 90 V bias to the substrate and 250 C led to the best guiding layers. Higher or lower temperatures both drastically increased the propagation losses, likely due to nano-crystallinity and cluster-void formation in the layer, respectively. By controlling the energy of the aluminum atoms arriving to the surface of the substrate by a combination of bias voltage and temperature, different layer morphologies were obtained [89].

Monolithic integration of Al 2 O 3 onto passive photonic platforms
The complexity of the integrated photonic chips realized in silicon photonics technology, including the silicon-on-insulator (SOI) and the silicon nitride (Si 3 N 4 ) platforms, is steadily increasing. However, these photonic integration platforms do not possess the capability of light generation/ amplification (i.e., optical gain). To achieve active devices such as lasers and amplifiers, active gain materials are normally integrated into these passive platforms, typically via hybrid or heterogeneous integration schemes [51]. Monolithic integration of rare-earth ion doped Al 2 O 3 onto these platforms shows promising potential towards manufacturable integration without additional non-scalable complex assembly steps. Furthermore, the integration of the laser pump source by butt coupling [51], grating coupling (i.e., VCELS flip-chip bonding [93]) or by heterogeneous integration [94] does not impose stringent requirements as it does not form part of the device cavity. In the following subsections, the different integration schemes of rare-earth ion doped Al 2 O 3 onto both the SOI and silicon nitride platforms will be reviewed.

Integration of rare-earth ion doped Al 2 O 3 onto SOI
The monolithic integration of Er 3+ :Al 2 O 3 onto the SOI platform towards the realization of on-chip optical amplifiers in this platform was first demonstrated by sputtering an Er 3+ :Al 2 O 3 layer on the top of SOI waveguides [48]. Adiabatic waveguide width tapers were used to couple the light from the silicon waveguides to the ridge erbium doped Al 2 O 3 waveguides, as shown in. A signal enhancement of 7 dB (i.e., ratio of the output power through the device with the pump on and off) was experimentally demonstrated after pumping with 1480 nm light. In that study, optical gain was not achieved due to the very elevated propagation losses of the device [48] Figure 3(a).
In a theoretical study, Pintus et al. suggested inserting heavily Er 3+ doped Al 2 O 3 into a silicon slot waveguide to achieve emission at 2.8 µm, which would be very useful for optical sensing and spectroscopy [42]. Recently, erbium-doped Al 2 O 3 was utilized as cladding to silicon microring resonators [49] and silicon photonic molecules [50] to reduce the losses and therefore increase the Q-factors (Figure 3(b)).

Integration of rare-earth ion doped Al 2 O 3 onto Si 3 N 4
Recent efforts have been directed towards the integration of rare-earth ion doped Al 2 O 3 to the passive silicon nitride platform. Bradley et al. proposed an integration scheme in which the doped Al 2 O 3 material was sputtered into trenches etched on the SiO 2 cladding deposited on top of segmented Si 3 N 4 waveguides [12], as shown in Figure 4(a). The advantage of this integration scheme is that it eliminates the need for performing reactive ion etching of the Al 2 O 3 material, which together with the Si 3 N 4 waveguides underneath forms a hybrid mode. The concept was further developed in [14,44] by depositing Al 2 O 3 into a trench etched into the SiO 2 cladding without silicon nitride waveguides underneath, Figure 4(b). The Al 2 O 3 rings are coupled to Si 3 N 4 bus waveguides. This integration scheme has been applied mostly to devices where bends are needed. In those devices, only the microring  [48]. (b) Use of erbium-doped Al 2 O 3 to reduce the losses, thereby increasing the Q-factor, of SOI ring resonators and photonic molecules. Taken from [49].
contains the doped Al 2 O 3 gain material while the bus waveguide is fully passive. In [39], an adiabatic taper is designed to make a low loss transition, ~0.3 dB/cm loss, between the Si 3 N 4 and doped Al 2 O 3 waveguides. For straight devices, such as distributed feedback lasers (DFB) and distributed Bragg reflector (DBR) lasers, the doped Al 2 O 3 layer can be simply deposited on top of the Si 3 N 4 waveguides both multi- [8,35,36,45] or single-striped [33,40,41] (Figure 5(a)). A hybrid mode is then formed with close to 90% mode overlap with the doped Al 2 O 3 oxide region and between pump and laser modes ( Figure 5(b)). The reported devices to date are fully doped. The integration with passive optical functions in silicon nitride are enabled thanks to the development of suitable couplers to produce a low-loss transition between the hybrid mode and the Si 3 N 4 mode [76].
A double-layer photonic platform has been proposed, in which Si 3 N 4 waveguides and rare-earth ion-doped Al 2 O 3 waveguides lie in two different photonic layers separated by a SiO 2 spacer layer. The transition between the bottom Si 3 N 4 and top Al 2 O 3 waveguides is done by means of low-loss adiabatic couplers in which the thickness of the Si 3 N 4 waveguides is vertically tapered to circa 30 nm. Figure 6(a) shows the 3D, top view and side  view schematics of the couplers as well as the cross-sections at different positions along the length of the coupler. Transition losses below 0.2 dB/ coupler were experimentally demonstrated [77,78]. The couplers exhibited very high tolerance to lateral misalignment (~±2 µm for 1 dB penalty for wavelengths in the C-band) during the lithography step of the top photonic layer. On-chip amplifiers [79] and microring lasers [80] have been demonstrated using this approach. Figure 6(b) shows and example of a zigzag Si 3 N 4 waveguide integrated with several sections of Er 3+ :Al 2 O 3 , where the green light is generated by energy transfer upconversion under 980 nm wavelength pumping.
A single-layer integration scheme has also been proposed and experimentally demonstrated for active on passive Al 2 O 3 integration [47]. The process is illustrated in Figure 7(a). An ytterbium-doped Al 2 O 3 layer is sputtered through a shadow mask followed by the deposition of undoped Al 2 O 3 and subsequent planarization by chemical mechanical polishing (CMP). A single lithographic step defines the passive-active devices simultaneously. Unfortunately, no amplifiers or lasers have been yet demonstrated using this approach, only the demonstration of the proofof-concept of the integration process.
Recently, erbium-doped Al 2 O 3 lasers integrated with Si 3 N 4 waveguides were introduced onto a CMOS-compatible silicon photonic platform to produce a fully integrated system-on-a-chip [76,95].

On-chip amplifiers in Al 2 O 3
Since the invention of the erbium-doped fiber amplifier in the late 1980s [96], EDFAs have become key components in many telecommunication systems. Fiber amplifiers doped with different rare-earth ions including Yb 3+ (YDFA) [97], Nd 3+ [98], Er 3+ [99] and Tm 3+ [100] can enable applications covering the wavelength ranges near the 1 µm, 1.3 µm, 1.5 μm and 2 µm bands. Advantages of rare-earth ion doped amplifiers include the broad gain bandwidth provided by these ions, the intrinsically low noise and the capability to amplify high-bit-rate signals. This is mainly due to their long excited state lifetime and their compatibility with the rest of the fibers in the system.
Integrated photonic waveguides increase the mode field intensity, leading to a more efficient interaction between the pump photons and active ions. Furthermore, the higher refractive index contrast in integrated photonic platforms with respect to optical fibers leads to smaller devices. Rare-earth ion doped waveguide amplifiers have been the subject of much research over the last few decades with many host materials for the rare-earth ions being investigated. A very complete review can be found in [15]. The requirements for a good material platform for the realization of rare-earth ion doped waveguide amplifiers include a high rare-earth ion solubility, deposition at the wafer level, and low passive propagation loss. From all the materials studied [15], Al 2 O 3 appears as a good candidate due to the high solubility for rare-earth ions, a refractive index of 1.65-1.73 varied by doping, and passive losses of less than 0.1 dB/cm at 1550 nm of wavelength. The spectroscopy of rare-earth ions and, in particular, of rare-earth ions doped in Al 2 O 3 has been the subject of many previous reports [101][102][103][104]. In the following paragraphs, the rare-earth-ion doped Al 2 O 3 amplifiers and its integration with other passive platforms in the last decade will be reviewed. Their achieved parameters will be introduced, such as maximum gain per unit length, maximum total net gain, amplification bit-rate, noise figure and amplification bandwidth.
In  , with an erbium concentration of 1.7 × 10 20 cm −3 . Such waveguide amplifier was integrated onto the Si 3 N 4 integrated photonic platform via double-layer monolithic integration technology using adiabatic vertical tapers [77]. A net Si 3 N 4 -Si 3 N 4 peak gain of 18 dB was achieved with a 10 cm long spiral amplifier at 1532 nm with a bidirectional pumping scheme with total incident pump power of 50 mW (estimated coupling losses ~12 dB for 976 nm wavelength) and incident signal power of -20 dBm [108].
The achievable gain depends on various parameters including the available pump power as well as the dopant concentration of rare-earth-ions that can be introduced into the material without clustering, which leads to luminescence quenching. To achieve high ion concentration, Rönn et al. alternated monolayers of Al 2 O 3 and Er 2 O 3 at different ratios to control the average erbium concentration [27]. The layers were deposited with atomic layer deposition (Section 2.1) in the gap of Si 3 N 4 slots waveguides with slot width of ~100 nm and strip size of ~460 × 460 nm 2 . Using this technique, an erbium concentration of 4.1 × 10 21 cm −3 was achieved. They demonstrated a maximum net gain of 1.98 dB for a 250 µm long amplifier with 4.5 mW launched pump power at 980 nm. Similar approach was utilized by Demirtas et al. [73]. The researchers alternated 5 monolayers of Al 2 O 3 and 5 monolayers of Er 2 O 3 for a final thickness of ~0.72 µm with total erbium concentration of 1 × 10 21 cm −3 . Channel waveguides had a trapezoidal cross-section with a base 3.95 µm wide and a top width of 2.36 µm. An internal net peak gain per unit length of 13.72 dB/cm was measured for a 2 mm long waveguide amplifier (i.e., 2.74 dB total internal net peak gain) when pumped at 980 nm with 30 mW of incident pump power. Furthermore, optical amplification at 880 nm was demonstrated in a Nd 3+ doped Al 2 O 3 waveguide amplifier. A ridge waveguide cross-section of 3 µm thickness, 2.5 µm width and 1 µm etch depth and Nd 3+ ion concentration of 0.5 × 10 20 cm −3 delivered 2.42 dB of net internal gain for a launched pump power of 55 mW at 802 nm of wavelength [81].

On-chip lasers in Al 2 O 3
The combination of low propagation losses and relatively high optical gain enables the use of rare-earth ion doped Al 2 O 3 for the realization of on-chip lasers. Furthermore, the very small linewidth enhancement factor of rareearth ion gain material, due to the lack of coupling between intensity and phase fluctuations leads to lasers with very narrow linewidths. In the following, we have classified the on-chip lasers developed in this material in two categories, namely lasers with cavities based on microring resonators and cavities based on distributed Bragg reflectors, both distributed feedback lasers (DFBs) and distributed Bragg reflector lasers (DBRs). An overview of the lasers reported in the literature with their performance parameters is given in Table 1.

Microring lasers on rare-earth ion-doped Al 2 O 3
From the development of the first layers with optical gain in rare-earth ion doped Al 2 O 3 , it still took over a decade before the first integrated laser in Er 3+ :Al 2 O 3 was demonstrated by Bradley et al. [30]. The demonstrated laser cavity consisted of a ring resonator, where the pump and signal coupling was achieved using two consecutive directional couplers, as shown in Figure 8(a). The couplers were designed for high coupling of the pump into the cavity while keeping the coupling at the laser wavelength low to ensure a high-Q factor for the cavity. The output power versus launched pump power of this device is shown in Figure 8(b). Varying the coupler and cavity length the output wavelength varied between 1530 and 1557 nm, showing the possibility for selecting the emission wavelength of the Er 3+ :Al 2 O 3 devices (Figure 8(c)). Due to the multi-longitudinal mode nature of the ring cavity (i.e., cavity length ranging between 3 and 5.5 cm) and the used coupler design, the output was inherently multimode.
A more compact design based on an Yb 3+ :Al 2 O 3 microdisk/microring coupled to a bus waveguide is reported in [3]. The radius of the microdisk is 100 μm and the bus waveguide has a width of 1.4 μm and a coupling gap of 0.6 µm. The thickness of the Al 2 O 3 core is 0.55 μm. The bottom cladding is thermal SiO 2 8 μm thick and the top cladding is water or other liquid delivered through a PDMS microfluidics channel, such as urine (Figure 9). This device exhibited an output power of ~25 μW with a ~ 2% on-chip slope   Figure 4 [12]. An output power of ~2.5 μW with double-side slope efficiency of 0.3% was measured. An on-chip laser output power of ~100 μW and double-side slope efficiency of 8.4% have been measured for an Yb 3+ -doped microring laser coupled to a SiN x bus waveguide [12]. Based on the same technology, Su et al. demonstrated a thulium-doped Al 2 O 3 microring laser [44]. This laser exhibited an on-chip output power of ~220 μW and double-side slope efficiency of 24% with 1.6 μm resonant pumping and lasing in the wavelength range of 1.8-1.9 μm.
In order to optimize pump absorption in the microcavity while reducing coupling losses at the signal wavelength, which would increase the laser threshold, a wavelength division multiplexer (WDM) can be implemented. Such WDM permits tuning the coupling strength of the pump and laser wavelengths independently [82]. The design of the laser cavity is shown in Figure 10(a). This design was first demonstrated in Yb 3+ doped Al 2 O 3 for a biosensing application with a H 2 O top cladding. A 2 mW multimode laser power was measured at the output fiber, corresponding to ~6 mW on-chip laser output power, which is the largest reported so far for a microring laser cavity in this material. A single side fiber-to-fiber slope efficiency of 1.2% has been measured, which corresponds to ~11% single side on-chip slope efficiency.
The free spectral range (FSR) of a typical RE 3+ :Al 2 O 3 ring laser is much smaller than the gain bandwidth of the rare-earth ion. This, in general, leads to a multi-longitudinal mode behavior of these ring lasers. In order to achieve a single-mode tunable laser, Li et al. demonstrated a hybrid ring laser cavity with an Er 3+ :Al 2 O 3 waveguide as the gain section and two Si 3 N 4 rings as intra-cavity tunable Vernier filter as shown in Figure 11 [39]. An on-chip laser output power of 1.6 mW and 2.2% on-chip slope efficiency were measured. The Si 3 N 4 Vernier ring filters are thermally tuned resulting in a 46 nm wavelength tuning range.

DFB and DBR on-chip lasers in rare-earth ion doped Al 2 O 3
A single longitudinal mode operation laser in Er 3+ :Al 2 O 3 was first shown by Bernhardi et al. [31] in a distributed feedback (DFB) waveguide cavity. The single-frequency output with a linewidth of 1.7 kHz has, to the best of our knowledge, not yet been surpassed in Al 2 O 3 based integrated lasers. The erbium-doped waveguide, with a doping concentration of 3.0 × 10 20 cm −3 and a length of 1 cm, was pumped at 1480 nm to obtain laser action with a threshold of 15 mW of launched pump power and 3 mW maximum offchip output power with a slope efficiency of 6.2% with respect to launched pump power. The high slope efficiency is a consequence of the relatively strong mode confinement in the small cross-section of the Er 3+ :Al 2 O 3 waveguide. The Er 3+ :Al 2 O 3 was deposited using a reactive sputter process, to produce a 1 μm thick layer onto which a 2.2 μm wide ridge was etched 100 nm deep using reactive ion etching.
Using the same material and waveguide geometry, a highly efficient monolithic distributed-Bragg-reflector laser doped with ytterbium was demonstrated. An ytterbium doping concentration of 5.8 × 10 20 cm −3 and an effective cavity length of 4.13 mm was utilized. The lasers exhibited a threshold with respect to the launched pump power of 10 mW with a double side combined slope efficiency of 67% and a maximum combined double side off-chip output power of 47 mW [21]. By introducing a second phase shift region within the DFB laser cavity, a dual wavelength laser was demonstrated [22], which was used to produce a 9 kHz wide microwave signal at 15 GHz with a temporal stability of ±2.5 MHz.
As described in section 3, integration of rare-earth ion doped Al 2 O 3 onto passive photonic platforms might pave the road to fully monolithically integrated photonic chips. In that direction, Purnawirman et al. [33]   A similar design was reported by Belt et al., with the cross-section shown in Figure 12 [40]. A DFB cavity was fabricated with a sidewall corrugation grating. The authors fabricated an array of DFB cavities designed for different wavelengths, thereby creating a multiwavelength source with emission at 1531, 1534, 1537, 1540, and 1543 nm. The on-chip pump laser power threshold of the system was 81 mW and the internal slope efficiency was 5.7 × 10 −3 %. This is likely a result of the short cavity length limiting the pump absorption. The low efficiency and therefore low output power resulted in a relatively broad linewidth of 501 kHz.
An improved pumping efficiency was demonstrated by Sing et al. [35]. They enclosed the DFB laser in a DBR cavity for the pump wavelength. When the cavity is pumped on resonance, the slope efficiency was improved by a factor of 1.8. In addition to resonance pumping, the waveguide crosssection was optimized to maximize overlap of both the pump and signal wavelength with the rare-earth ion-doped Al 2 O 3 , thereby increasing the pump efficiency. The efficiency is further increased in this configuration by the increased overlap between the pump and signal mode. The waveguide architecture utilized consists of a multi-stripe waveguide design, similar to the one proposed by Bradley et al. [12], which lowers the effective refractive index of the Si 3 N 4 waveguide section, while maintaining significant thickness. A schematic of the resonant cavity as well as the waveguide crosssection is shown in Figure 13.
As discussed in Section 2, Magden et al. developed a fully CMOScompatible reactive sputtering deposition process of Al 2 O 3 by reducing the temperature during deposition to 250 C. This was possible by applying a DC bias to the substrate to increase the kinetic energy of the aluminum and erbium ions arriving to the substrate, therefore achieving similar morphology at a lower temperature [37]. The lower substrate temperature during deposition allows integration of the Al 2 O 3 layers with active components in Si 3 N 4 , since the required metal contacts do not melt at this temperature. This technology is also compatible with integration on a CMOS platform. The achieved on-chip pump threshold and slope efficiency of a quarter wave shift DFB cavity for a doping concentration of 1.9 × 10 20 cm −3 were 24.9 mW and 1.3%, respectively, for an emission wavelength of 1553 nm. An on-chip laser output of 2.6 mW at 1552.98 nm was obtained for an on-chip pump power of 220 mW at 976 nm of wavelength. Another improvement in the design of DFB laser cavities is introduced by Purnawirman et al. [38]. where a curvature is added to the waveguides comprising the DFB cavity to compensate for the nonuniformity in the active Al 2 O 3 layer thickness. Using this technique, a 6-fold reduction in the on-chip threshold power (i.e., 16 mW) is reported with respect to the control straight waveguide DFB laser (i.e., 105 mW). An on-chip laser power of 1.2 mW was demonstrated with this architecture.
Despite the many improvements on the initial DFB lasers integrated with Si 3 N 4 , the threshold power, slope efficiency and linewidth of the DFB laser by Bernhardi et al. [31] have not yet been achieved by a quarter phase shifted (QPS) cavity design ( Figure 13). An alternative phase shift approach is proposed and demonstrated by Purnawirman et al. [36]. The quarter phase shift is now generated by tapering the thickness of one of the waveguide ridges segments. The distributed phase shift (DPS) element approach is similar to the varying waveguide width proposed in the DFB laser by Bernhardi et al. [31], also resulting in very similar behavior. The structure of the laser as well as its performance are shown in Figure 14. The achieved on-chip pump threshold was 14 mW with a slope efficiency of 2.9% and maximal on-chip output power of 5.4 mW. The linewidth of the DPS-DFB laser was measured to be 5.3 kHz, showing an almost 6-fold improvement compared to the QPS-DFB laser [36].

Applications of RE 3+ :Al 2 O 3 on-chip lasers on a passive photonic platform
Passive ring resonators are popular for sensing applications due to their high-Q and high sensitivity to environmental changes [109][110][111][112][113]. A lasing ring cavity has a much higher Q than its cold cavity, which leads to a lower intrinsic limit of detection of the sensor [114,115].
An Yb 3+ :Al 2 O 3 microdisk laser has been recently demonstrated for the label-free detection of cancer biomarkers directly from undiluted urine [3]. The top cladding of this laser is the liquid to be sensed, in this case urine, which is delivered through a microfluidic channel. Attachment of the biomolecules onto the Yb 3+ :Al 2 O 3 waveguide leads to a change of the effective refractive index, which subsequently changes the optical path length of the laser cavity and therefore the lasing frequency. The frequency shift is accurately detected by heterodyning the output of the microdisk laser with that of an external reference laser. This type of sensor is not only sensitive to biomolecules binding on the surface of the waveguide but it is also sensitive to temperature and bulk refractive index changes. These sensitivities have been characterized (Figure 15) to be 1.7 GHz/K (6 pm/K) and 5.7 THz/RIU (20 nm/RIU). A limit-of-detection of 300 pM (3.6 ng/ml) of the protein rhS100A4 (12 kDa) in urine was experimentally demonstrated, being mostly limited by thermal noise. This result shows the potential of integrated active microdisk lasers in Al 2 O 3 for biosensing.
Benefiting from all the developments discussed above in this review paper, namely the development of low-loss Al 2 O 3 deposited at low temperature [37], integration with Si 3 N 4 using a trench configuration [12,39,44], DFB laser design [35,36,116] and adiabatic transitions from the hybrid active gain region to the passive Si 3 N 4 waveguides [39,77,78], complex systems can be built. An impressive demonstration of such system for LIDAR application was recently presented by Notaros et al. [76]. A first proof-of-concept of an electrically steerable integrated phased array activated by an on-chip erbium-doped DFB laser was demonstrated. A steering angle of 0.85° x 0.20° as well as 30°/W electrical steering efficiency were experimentally shown (Figure 16).
A silicon photonics data link integrating an Er 3+ :Al 2 O 3 -Si 3 N 4 DBR laser similar to the one reported in [35] with a silicon photonics circuit integrating modulators and germanium photodetectors has also been recently demonstrated [95]. These demonstrations show the potential of this material for real-life applications.

Final remark
Rare-earth ion doped Al 2 O 3 is an active optical material that has gained growing interest in recent years owed to the excellent optical properties of the host, Al 2 O 3, in combination with the gain performance provided by the rare-earth ions. Recent developed monolithic integration technologies of this material onto the passive silicon on insulator and silicon nitride technologies permit the development of complex circuits, where the properties of the on-chip generated light can be controlled by the passive circuitry. This technology opens the road to potentially disruptive devices that can find interesting applications in emerging fields including LiDAR, quantum technology, optical sensing/metrology and artificial intelligence.

Disclosure statement
No potential conflict of interest was reported by the authors.