Silicon-based optoelectronics aims to heterogeneously integrate photonic, electronic, and optoelectronic devices on a single silicon substrate using complementary metal oxide semiconductor (CMOS)-compatible processes. This technology emphasizes the co-optimization of photonic and electronic functionalities, with a particular focus on addressing the critical challenge of large-scale heterogeneous integration between silicon and other functional materials.1 Silicon-based optoelectronic (SBO) chips possess the advantages of both integrated circuit and optoelectronic chips simultaneously,1^–4 and they are the core solution for enabling the miniaturization of information systems under the background of high computing power requirements.5^,6 However, as the scale of SBO chips continues to expand and their functions become increasingly complex, the transmission loss of optical signals within the chips has become a key issue that urgently needs to be solved.7 The integration characteristics of erbium-doped/erbium-ytterbium co-doped waveguide amplifiers (EDWAs/EYCDWAs) enable them to be integrated with other components in SBO chips and can effectively compensate for the energy attenuation of optical signals during the transmission process within the chips, greatly enhancing the intensity and transmission distance of optical signals.8^–11 By amplifying weak optical signals, the signal-to-noise ratio and stability can be improved, providing a powerful guarantee for achieving high-precision optical information processing and transmission. Without significantly increasing the chip area and power consumption, they can greatly improve the overall performance of the system. This is of great significance for promoting the development of SBO chips toward higher integration, smaller size, and lower power consumption. Moreover, a high-gain EDWA/EYCDWA combined with an appropriate resonant cavity structure can generate laser output, and it is one of the effective means for fabricating on-chip integrated communication-band lasers.12^–17

In the realm of academic research, EDWAs/EYCDWAs have achieved a series of crucial progress.18^–25 In the currently reported research, the small-signal gain has exceeded 30 dB, and the maximum output power has even reached the level of hundreds of milliwatts.18^,22 In the field of material research, many host materials, such as silicon nitride (Si3N4),12^,18^,26^–29 thin film lithium niobate (TFLN),13^,15^,22^–25^,30^–46 aluminum oxide (Al2O3),16^,19^–21^,47^–65 tantalum oxide (Ta2O5),14^,66^–74 tellurite oxide (TeO2),75^–80 and polymers,81^–92 have been experimentally used for doping rare-earth ions, and the optical net gain in the corresponding waveguides has been measured. In terms of the design of the waveguide structures, novel waveguide structures have emerged in an endless stream, such as slot waveguide,19 double-layer waveguide,21^,57^,66^,79 ultra-low-loss waveguide (ULLW),27^,93 and large-mode-area (LMA) waveguide structures.22^,94^–96 The purpose of these different structural designs is to strengthen the interaction between light and doped ions, reduce the transmission loss of waveguides, and improve the optical gain of the amplifiers. In addition, regarding performance characterization and testing, many studies have extended beyond simply measuring the net gain of EDWAs/EYCDWAs to more comprehensive evaluations of amplifier performance, encompassing gain characteristics, noise figure, and polarization-dependent properties.31^,37 Although EDWAs/EYCDWAs have already demonstrated promising results in academic research, how to further improve amplifier performance within limited chip areas remains challenging. Meanwhile, investigations into the long-term stability and reliability of the amplifiers under harsh environments such as high temperature and high humidity remain ongoing.

In terms of industrial progress, the commercial products of EDWAs/EYCDWAs can be traced back to the early 21st century. Companies such as Teem Photonics, Inplane, and Molecular Optoelectronics Corporation were the pioneers that launched commercial EDWA products on a small scale.97^–106 They can provide optical signal amplification with certain gain and bandwidth, as well as reasonable stability. However, due to the relatively high technical thresholds and costs, there are very few companies that can currently launch commercialized EDWA products. To achieve large-scale industrial application of EDWAs/EYCDWAs, there are still many challenges, such as high production costs, complicated manufacturing processes, and difficulties in ensuring product consistency and reliability. Currently, the industrial applications of EDWAs/EYCDWAs are mostly stuck at the laboratory stage or the small-batch trial production stage, but some research results have also demonstrated their potential in practical applications. For example, the team led by professor Tobias J. Kippenberg achieved an on-chip continuous optical output power of over 145 mW, which is two orders of magnitude higher than that of the previously reported devices.18 Dawson B. Bonneville and others from the University of Twente demonstrated a well-packaged EDWA prototype, in which the high-gain Er:Al2O3 waveguide was perfectly integrated with the Si3N4 planar lightwave circuit (PLC), providing a brand-new idea for the design of future EDWA/EYCDWA products.107

This review comprehensively sorts out the research progress of EDWAs/EYCDWAs in the field of SBO chips, covering their current development status, challenges, and future development directions. The focus is on the breakthrough achievements in recent years in aspects such as host materials, structural design, integration technologies, and performance optimization. In addition, this review explores the potential research hotspots and application prospects in this field in the future. Figure 1 presents an overview of the research field of on-chip EDWAs/EYCDWAs,which represents the core content of this article. In Sec. 2, a comprehensive and detailed introduction to the latest advancements of EDWAs/EYCDWAs is provided, and an in-depth analysis of the influencing factors of waveguide amplifier performance is conducted from two different perspectives: host materials and waveguide structures. In Sec. 3, a comprehensive exploration of the fabrication processes and applications of EDWAs/EYCDWAs is presented. The fabrication techniques involve two main aspects: the preparation of gain media and the fabrication of waveguide structures. Here, a systematic review and a summary of the well-established process protocols that have emerged in recent years are given. Regarding applications, a thorough analysis of the various application scenarios and commercial products relevant to EDWAs/EYCDWAs is carried out. Finally, the challenges faced by EDWAs/EYCDWAs are summarized, and a forward-looking perspective on their future development directions is provided. The aim is to offer valuable reference bases for research work and industrial development in related fields.

The selection of host materials and the design of waveguide structures are critical factors in determining the gain performance of the amplifiers. Specifically, the solid solubility of Er3+ and Yb3+ in the host material is an important internal factor affecting the performance of the amplifier. At the same time, the intrinsic absorption of the material also constitutes a key part of the waveguide loss. From the perspective of the waveguide structure, its design scheme exerts a significant influence on the scattering losses and the level of interaction between the signal and the pump light within the waveguide. Both of them are directly related to the final performance of the waveguide amplifier. In recent research, many teams have carried out extensive and in-depth studies on host materials and special waveguide structure designs, constantly exploring new types of host materials and attempting innovative waveguide structure designs, aiming to effectively improve the gain level and output power of the EDWAs/EYCDWAs. In this section, we will systematically review the research progress of waveguide amplifiers composed of several major gain media materials prepared from different host materials, as shown in Fig. 1, and conduct a discussion on the special waveguide structure design of the EDWAs/EYCDWAs.

Si3N4 materials possess advantages such as a wide transparent window, low two-photon absorption effect, low loss, and compatibility with CMOS processes. Thus, they are widely utilized in the fabrication of various SBO devices. However, Si3N4 has a relatively low solubility for rare-earth ions such as Er3+ and Yb3+, making it difficult to directly deposit high-quality erbium-doped films. Moreover, thick Si3N4 films suffer from severe stress. When the deposition thickness of the film exceeds 400 nm, it becomes easily cracked.

Liu et al.18 demonstrated an Er:Si3N4 waveguide amplifier with high amplification gain. On a small chip area of 1.2mm×3.6mm, an on-chip continuous optical net gain exceeds 30 dB and an output power of >145mW, being the highest reported on-chip output power to date, as shown in Fig. 2. The core of this technology lies in the breakthrough use of high-energy erbium ion implantation technology combined with the damascene process to fabricate Er:Si3N4 waveguide. By precisely controlling the ion implantation process, a Gaussian distribution of Er3+ in the waveguide was achieved, and the overlap factor between the Er3+ distribution and the optical mode field was as high as 50%. This uniform spatial distribution significantly improved the excitation efficiency of the pump light on Er3+ and effectively reduced the interaction among the Er3+ ions, thus avoiding gain saturation and non-ideal gain behavior caused by ion aggregation. The scattering loss on the waveguide sidewalls and surface is the main source of transmission loss in EDWA. The silicon oxide trench fabricated by the damascene process has a smooth surface, and during the subsequent Er3+ implantation process, the silicon oxide substrate around the waveguide provides good support for the sidewalls of the silicon nitride waveguide, so there is no severe deformation of the waveguide that can result in an increase in waveguide loss during the ion implantation process. Ion implantation allows the doping of other rare-earth ions in different host materials. This research provides new ideas for the preparation methods of various on-chip integrated light source gain media.

Thanks to the excellent electro-optic, acousto-optic, and nonlinear properties of lithium niobate materials, in recent years, a wide variety of active and passive photonic devices based on the TFLN platform have emerged in an endless stream. Among them, EDWAs/EYCDWAs based on TFLN have also achieved numerous research results. In particular, recently, through the LMA waveguide structure, an output power exceeding 100 mW has been achieved with a 7-cm erbium-doped lithium niobate waveguide.22 This is the second EDWA to achieve more than a hundred-milliwatt-level output following the Er:Si3N4 waveguide amplifier, and this structure will be analyzed in detail later. The successful development of waveguide amplifiers based on TFLN has greatly inspired research interest in developing active lithium niobate on insulator (LNOI) devices and active–passive hybrid integrated LNOI devices. However, lithium niobate crystals are very hard, and high-quality device fabrication based on TFLN remains challenging. Currently, there are mainly two schemes for processing devices based on Er:TFLN. One is the photo-lithography-assisted chemical mechanical polishing (PLACE) technology developed by Cheng’s team at East China Normal University, and the other is the electron beam lithography combined with reactive ion etching (EBL + RIE) technology, which is the most widely used technology for lithium niobate processing at present.

Figure 3 shows the representative achievements of EDWAs/EYCDWAs fabricated using the PLACE process. Zhou et al.23 achieved an amplifier with an internal net gain of 18 dB and a gain per unit length of 5 dB/cm on an Er:LNOI waveguide that was 3.6 cm long. This was the first EDWA fabricated using the PLACE process, as shown in Fig. 3(a). The waveguides fabricated by this technique have smooth surfaces and sidewalls, which significantly reduces the scattering loss of the erbium-doped lithium niobate waveguide and greatly improves the net gain of the EDWA. To achieve the monolithic integration of active and passive devices on the LNOI platform, in 2023, the team demonstrated a four-channel waveguide amplifier array, as depicted in Fig. 3(b). At wavelengths of 1550 and 1530 nm, the net gains of each erbium-doped waveguide were 5 and 8 dB, respectively.35 This waveguide amplifier achieved the splicing of active and passive TFLN films in a butt-joint way, and finally, the waveguide structure was fabricated through the PLACE process. The active and passive regions were highly aligned, and the difference in refractive index was small, with only 0.26 dB of coupling loss at the splicing point.

In the same year, the team achieved an EYCDWA with a small-signal gain of 27 dB on the LNOI platform for the first time by co-doping erbium and ytterbium,25 as illustrated in Fig. 3(c). Thanks to the waveguide transmission loss of 0.36dB/cm, an erbium-ytterbium co-doped waveguide as long as 15 cm was fabricated. However, the end-face losses of this device were relatively large, at 8.5 and 9.1 dB, leaving much room for improvement. As we can see from Fig. 3(d), they used the coupling of the 1480-nm pump light and the 1530-nm signal light through a 2×2 multi-mode interferometer, and a TFLN electro-optic Mach–Zehnder interferometer modulator was used to adjust the relative phases of the two paths; then, they achieved a total power of up to 12.9 mW by means of coherent beam combination in 2023.34 The method of coherent beam combination is very helpful for increasing the output power of the EDWA, but it makes the fabrication of the entire device more difficult.

Undoubtedly, the PLACE process occupies an important position in the fabrication of TFLN devices. Through this process, waveguide devices with ultra-low loss characteristics, such as delay lines and microcavities, can be successfully fabricated. However, this process also has certain limitations. First, in terms of the waveguide structure, as can be clearly seen from Fig. 3, its cross-section presents a trapezoidal shape with a narrow upper base and a wide lower base, and the waveguide is relatively wide overall with a cross-sectional area reaching several square micrometers. When fabricating higher-precision functional devices such as 980-/1550-nm wavelength division multiplexers (WDMs), directional couplers, and microrings, how to further reduce the waveguide width and the spacing in the coupling region has become an urgent problem that needs to be solved. Second, the PLACE process relies on chemical–mechanical polishing to form a low-loss waveguide structure. However, during the polishing process, the waveguide may be damaged. The yield rate of devices fabricated using this process will probably become a key issue that needs to be seriously considered in future industrial development. Furthermore, it is hard to integrate the devices fabricated through this process with optoelectronic devices on other platforms. Crucially, the coupling issues between fibers and devices, as well as among the devices themselves, demand effective solutions. This is especially true in the context of large-scale chip integration, where the fabrication of more multifunctional devices and the associated coupling problems will emerge as major hurdles for this particular process.

Figure 4 presents the representative outcomes of TFLN-based EDWAs/EYCDWAs fabricated via the combination of electron beam lithography (EBL) and reactive ion etching (RIE) techniques. In this approach, the well-established EBL process and argon ion etching technology are employed to manufacture TFLN photonic devices. The etched waveguide surfaces exhibit a low roughness of less than 5 nm and a high degree of surface smoothness, ensuring high reproducibility and a favorable device yield. Presently, this fabrication process has been adopted by numerous research teams. The team led by Cheng Wang at City University of Hong Kong fabricated an erbium-doped straight waveguide structure with a total length of 5 mm on a 0.5% (mole fraction) Er:TFLN wafer, achieving a total net gain of >5dB and a unit net gain exceeding 10dB/cm under the wavelength of 1530 nm,40 as shown in Fig. 4(a). In the same year, Chen’s team at Shanghai Jiao Tong University fabricated a high-gain spiral waveguide amplifier with a total length of 5.3 mm and a footprint of ∼0.06mm2 on a 1% (mole fraction) erbium-doped LNOI wafer, realizing a maximum internal net gain of 8.3 dB and a maximum net gain per unit length of up to 15.6dB/cm under the wavelength of 1530 nm,36 as shown in Fig. 4(b). Wu’s team fabricated a 2.58 cm EDWA on the Er:LNOI platform, achieving a signal enhancement of 27.94 dB, an internal net gain of 16.0 dB, and a unit net gain of 6.20dB/cm. In addition, this on-chip optical waveguide amplifier had a saturation power of −8.84dBm, a power conversion efficiency of 4.59dB/mW, and a noise figure of 4.49 dB (at 1531.6 nm).37 In this research achievement, the team members conducted an in-depth study on the pump wavelength, pump scheme, output power, and noise figure in the optical waveguide amplifier for the first time. This work will contribute to a comprehensive understanding of the waveguide amplifier, as shown in Fig. 4(c). Based on this work, in 2024, the team fabricated an Er:TFLN waveguide amplifier with a longer length, achieving higher gain and output power.31Figure 4(d) shows the EYCDWA device fabricated on TFLN substrate via EBL and RIE process, exhibiting an ultralow pump threshold of 0.1 mW. The 5-mm-long waveguide achieves 16.52dB/cm gain under 974-nm pumping. Enabled by efficient Yb3+−Er3+ energy transfer, the device reaches 10% conversion efficiency.24

In addition to the research achievements mentioned above, some other results are also worthy of attention.24^,32^,39 Compared with PLACE, the combination of EBL and RIE technology has more advantages in fabricating high-precision and large-scale integrated devices, and the device yield is more guaranteed, which is very advantageous in the production of small-batch photonic devices. However, the exposure speed of EBL is very slow, making it difficult to meet the needs of large-scale production, and the cost is relatively high. In future large-scale production, it may be necessary to combine the use of UV/DUV lithography and only use EBL to fabricate the parts with the highest precision requirements in the entire chip to improve production efficiency and reduce production costs.

Oxide materials usually possess excellent optical transparency within a relatively wide spectral range. The thin film preparation process of oxide is mostly CMOS-compatible and relatively mature and simple, with relatively low costs. Moreover, oxide materials have a high solubility for rare-earth elements and low intrinsic losses, making them highly suitable to be used as host materials for doping with rare-earth ions such as Er3+ and Yb3+. Currently, among the erbium-doped/erbium-ytterbium co-doped oxide materials, Er:Al2O3 has been the most widely studied. In addition, Er:Ta2O5 and Er:TeO2 also exhibit considerable research potentials.

Al2O3 materials have relatively low intrinsic losses and exhibit small optical field absorption losses within the wavelength range of 1−6μm. Moreover, Al2O3 has a relatively high solubility for rare-earth ions, which can reach the order of magnitude of 1020−1021cm−3. Er/Er‐Yb:Al2O3 gain media can be prepared by means of magnetron sputtering, ion implantation, atomic layer deposition (ALD), and so on. Currently, there are mainly two types of EDWAs/EYCDWAs based on the Er:Al2O3 platform. One is to form the waveguide structure by directly etching the gain medium, and the other is to construct the amplifier by integrating with passive waveguides instead of directly etching the gain materials. Herein, we summarize some representative achievements of these two forms.

Figure 5 presents the relevant achievements of directly etched Er/Er‐Yb:Al2O3 waveguide amplifiers. Vázquez-Córdova et al.48 prepared Er:Al2O3 with different erbium doping concentrations by means of magnetron sputtering. They fabricated ridge waveguide structures with a ridge thickness of 0.35μm through standard lithographic techniques and chlorine-based reactive ion etching, as shown in Fig. 5(a). This structure has a good ability to confine the optical field and exhibits relatively low transmission losses of only 0.2dB/cm. A small-signal gain as high as 20 dB was achieved under an erbium doping concentration of 1.92×1020cm−3 and a waveguide length of 12.9 cm. Bonneville et al.58 deposited the films at an elevated wafer temperature during magnetron sputtering based on the previous work, which significantly improved the quality of the Er:Al2O3 film and reduced the OH− content within the film, thereby enhancing the performance of the device. Eventually, under 1480-nm pumping, a small-signal gain exceeding 30 dB and an on-chip output power of ∼10mW were achieved. Figure 5(b) shows the waveguide cross-section and the excited fluorescence image of this device. This team also fabricated an Er‐Yb:Al2O3 waveguide amplifier through the reactive co-sputtering and wet etching processes in 2020,49 as shown in Fig. 5(c). This waveguide amplifier obtained a net gain per unit of 4.3±0.9dB at the central wavelength of 1533 nm. Although the EDWAs/EYCDWAs shown in Fig. 5 all exhibit positive gains, the waveguide sidewalls formed by this processing method of directly etching the gain materials are relatively rough, which could largely lead to an increase in transmission losses and affect the overall gain of the device.

To avoid directly etching the gain medium and reduce transmission losses, an effective approach is to form a hybrid waveguide structure with other easily etchable passive materials and process the undoped passive materials in other layers into waveguide structures for guiding modes. Figures 6(a)–6(d) respectively show the hybrid structures with Si/Si3N4/HSQ/GeSbS materials. Agazzi et al.20 from the University of Twente achieved the monolithic integration of Si waveguides and Er:Al2O3 waveguides, obtaining a signal enhancement of over 7 dB at 1533 nm. The optical field transition between the passive and active waveguides was realized through an inverted taper coupler, as shown in Fig. 6(a). However, due to mode leakage at the coupling position, this coupling structure had an insertion loss of 2.5 dB, which had a significant impact on the gain. Mu et al.21 designed a Si3N4‐Er:Al2O3 monolithically integrated waveguide amplifier, achieving a gain of 18.1±0.9dB at 1532 nm through a 10-cm-long spiral waveguide. This work realized a smooth transition of the optical field through a vertical adiabatic taper coupler, with an insertion loss of only 0.49 dB, which was a great improvement compared with previous ones, as shown in Fig. 6(b).

To enhance the fluorescence performance of Er:Al2O3, many teams have explored the Er:Al2O3 gain medium using the ALD technology.19^,50^,60^–63 Zhang et al.62 optimized the doping concentration of erbium ions based on previous work and prepared strip-loaded waveguides composed of hydrogen silsesquioxane polymers and Er:Al2O3. Through a 3.55-cm-long S-shaped waveguide, a signal enhancement of 30.4 dB and an internal net gain of 8.4 dB were achieved at 1531.6 nm, as shown in Fig. 6(c). Based on this work, the team fabricated an EDWA with broadband net gain exceeding 9 dB across the C-band in 2024, which has superior performance observed for both transverse electric (TE) and transverse magnetic (TM) modes. Wang et al.50 adopted a similar structure, except that GeSbS was used as the loading strip. Eventually, a net gain of 6.27±0.618dB was achieved at 1533 nm through a 4.6-cm-long waveguide, as shown in Fig. 6(d). The above research works based on passive/active hybrid waveguide structures have reduced the transmission losses of Er:Al2O3 waveguides to some extent. However, the end-face coupling is rather difficult. In future work, special structures need to be designed to reduce the coupling losses and achieve efficient coupling between passive waveguides and gain media.

Another interesting example is Er3+-doped Ga2O3 waveguide amplifiers because Ga belongs to the same family as Al in the periodic table, and thus, its oxide can be expected to have similar physical and chemical properties to Al2O3. It is believed that Ga2O3 has its own unique advantages. For example, Ga itself can promote the solubility of the element Er and avoid the quenching of the photoluminescence due to the formation of Er clusters in the glass. Photoluminescent spectrum of Er-doped Ga2O3 exhibits a broad band of 71 nm at its full width at half-maximum around 1.55μm and a high refractive index of 1.9 which is larger than that of Al2O3 (1.65). An impressive result of the device demonstration is that, in a 7-mm-long waveguide with an optical loss of 4.5dB/cm, a maximum net internal net gain of ∼4.7dB (∼6.7dB/cm) at 1545 nm is obtained when the signal power is −42.4dBm and the pump power is 48.6 mW (∼16.9dBm).108 Such a large net gain is unusual considering its larger waveguide loss. Nevertheless, the use of amorphous Ga2O3 as photonics material is still in its dawn, and further improvement in its performance via material preparation and waveguide fabrication by the photonic society can be expected in the future.

TeO2 has a higher refractive index (2.08 at 1550 nm) than Al2O3, which means that TeO2 has a higher emission cross-section and can be used to fabricate devices with a higher level of integration. Existing research results have demonstrated that the background loss of TeO2 waveguides is as low as 0.1dB/cm. Moreover, TeO2 has a relatively high solubility for Er3+ and a large emission bandwidth, indicating that it is very promising to utilize TeO2 as the host material for Er3+. Compared with Er/Er‐Yb:Al2O3, there has been relatively less research on Er/Er‐Yb:TeO2 at present.

Figure 7 shows the representative works based on the Er:TeO2 platform. Vu and Madden79 prepared Er:TeO2 thin films by co-sputtering with Te and Er targets. They then used a mixture of gases of hydrogen, methane, and argon to etch the films to fabricate TeO2 passive waveguides and Er:TeO2 active waveguides, as presented in Fig. 7(a). It can be seen from the figure that the passive waveguides have relatively smooth surfaces and sidewalls, whereas the Er:TeO2 waveguides are very rough, accompanied by columnar structures and grassing effects. To address the difficulty in etching Er:TeO2, the team prepared a bilayer thin film integrating pure TeO2 and Er:TeO2 and formed the guiding mode by etching pure TeO2. Eventually, a total gain of 14 dB was achieved through a 5-cm-long rib waveguide, with a gain per unit length of 2.8dB/cm. In fact, there is still room for optimizing this gain. For example, in the work of this team, the thicknesses of pure TeO2 and Er:TeO2 are the same. If the thickness of pure TeO2 is appropriately reduced, more optical fields will be confined in the active layer, which can improve the final gain of the device to some extent.

Similarly, Frankis et al.75 fabricated an Er:TeO2-coated Si3N4 waveguide amplifier to avoid directly etching Er:TeO2. This device formed the waveguide structure by etching Si3N4 and depositing the gain medium on it, as shown in Fig. 7(b). Thanks to the low-loss characteristic of Si3N4, the transmission loss of this hybrid waveguide was as low as 0.25dB/cm. This work finally achieved an internal net gain of over 5 dB. Apparently, both of the works mentioned above have made efforts in terms of how to avoid directly etching the gain medium and reducing the transmission losses of erbium-doped waveguides. However, the final output gains are not particularly high. In the future, optimizations can be made in terms of waveguide structures and pumping methods. Besides, there is still room for further exploration regarding the fluorescence intensity and fluorescence lifetime of the thin films themselves, including using other rare-earth ions to enhance the fluorescence intensity and reducing the quenching caused by the concentration of erbium ions.

Besides TeO2, Ta2O5 is another oxide material with great potential as a host material for erbium doping. Ta2O5 has a refractive index (2.1 at 1550 nm) similar to that of Si3N4 and TeO2, and existing research results have confirmed its ultra-low intrinsic loss as low as 3dB/m.109 In addition, Ta2O5 has an extremely wide transparent window from ultraviolet to mid-infrared wavelength, relatively low phonon effects, and relatively high third-order nonlinear effects. Moreover, Ta2O5 has a relatively high solubility for rare-earth ions such as Er3+ and Yb3+. Similar to TeO2, currently, there are not many research teams focusing on erbium-doped waveguide amplifiers based on the Ta2O5 platform. However, several teams, including our own team, have been making efforts in this regard.

Subramanian et al.68 from the University of Southampton prepared Er:Ta2O5 thin films with an erbium doping concentration of ∼2.7×1020cm−3 by means of RF magnetron sputtering. They also fabricated rib waveguides through the use of lithography and argon ion beam milling (Ar IBM) processes. When pumped by a 977-nm laser, a net gain per unit length of 2.1dB/cm was achieved at 1531.5 nm through a 2.3-cm-long Er:Ta2O5 rib waveguide. As can be seen from Fig. 8(a), the use of a 45-deg milling angle significantly reduced the sidewall roughness of the waveguide and largely decreased the transmission losses of the waveguide. This work was the first to fabricate an EDWA based on the Er:Ta2O5 platform and has great guiding significance for the development of subsequent work. Although the waveguides fabricated by the Ar IBM process were relatively smooth, their losses still reached 0.65dB/cm. How to improve the waveguide structure to reduce transmission losses and how to optimize the doping concentration to enhance the fluorescence performance are the key issues that urgently need to be solved.

Zhang et al.66 drew on the bilayer hybrid waveguide structure in Er:Ta2O5 and fabricated a waveguide amplifier integrating pure tantalum oxide and Er:Ta2O5, as shown in Fig. 8(b). This work avoided directly etching the Er:Ta2O5 materials and instead formed the guiding mode by etching the passive Ta2O5. Eventually, this work achieved a net gain per unit length of 4.63dB/cm. Although the net gain per unit length has been greatly improved compared with the achievements of other teams, there is still a large amount of room for optimization in terms of film quality and waveguide fabrication in this work, and it has far from fully exploiting the maximum advantages of Ta2O5 as a host material for erbium doping. Recently, our team has cooperated with this team to fabricate Er/Yb:Ta2O5 gain media, and the overall performance has been greatly improved compared with that of Er:Ta2O5. It can be expected that Ta2O5 has great potentials as a host material for erbium-doped/erbium-ytterbium co-doped systems. It is believed that more teams will join the research in the future.

Compared with other erbium-doped gain media, Er3+ in Er:silicate materials exists in the form of compound cations. As a result, the erbium doping concentration will not be limited by the solid solubility of the host material, and the maximum doping concentration can be increased to the order of 1022cm−3, which is nearly two orders of magnitude higher than that of other host materials. Although high Er3+ doping concentrations can be realized in Er:silicate, the hardness of the material makes it difficult to prepare waveguide structures via direct etching. Moreover, in Er:silicate, high-temperature annealing is required to activate Er3+. However, after annealing, Er:silicate often exhibits a polycrystalline state, leading to very high losses. Figure 9 shows the waveguide amplifiers fabricated with Er:silicate in recent years.

Figure 9(a) presents the Si3N4 hybrid, strip-loaded, and slot waveguide amplifiers based on Er-Yb:silicate films fabricated by Wang’s team at Peking University from 2011 to 2012.110^–112 In the structure of the strip-loaded waveguide amplifier, SiO2 is deposited above the gain material and etched into a waveguide structure. By patterning the SiO2 waveguide, the optical field is confined in the gain layer, avoiding directly etching the hard Er-Yb:silicate. Eventually, a signal enhancement of 5.5 dB was achieved under a pump power of 372 mW.111 However, the optical field confinement ability of this structure is relatively weak, and it has large bending radiation losses, so it is not suitable for fabricating devices with very high integration. In the slot waveguide amplifier structure, the thickness of the gain medium is only 110 nm, and silicon materials with a higher refractive index are used above and below the gain layer, so that the optical field can be more effectively confined in the gain materials in the slot, the optical power density in the slot can be increased, and then the gain of the waveguide amplifier can be improved. Finally, a signal enhancement of 1.7 dB was achieved at 1530 nm.112 In the Si3N4 hybrid waveguide amplifier, first, a Si3N4 film is deposited on the substrate and patterned to form a Si3N4 waveguide structure. Then, Er-Yb:silicate is deposited on the Si3N4 waveguide structure to form a hybrid waveguide structure.110 This structure does not require directly etching the gain material but directly etching the Si3N4 material, which can reduce the transmission losses of the optical field in the waveguide. After the waveguide amplifier undergoes high-temperature annealing at 1000°C, a signal enhancement as high as 10.1dB/cm is finally achieved.

Sun et al.113 first prepared single-crystalline erbium chlorosilicate compound nanowires using the chemical vapor deposition method, as shown in Fig. 9(b). As the nanowire material has a single-crystalline structure, very few defects in the material greatly reduce the transmission loss of the waveguide and improve the stability of the material. The length of this nanowire is 56.2μm, its diameter is 1μm, and the concentration of erbium ions is 1.62×1022cm−3. This concentration is about two orders of magnitude higher than that of erbium-doped films prepared by other methods. Moreover, at such a high doping concentration, the nanowire still exhibits a relatively low concentration quenching and a fluorescence lifetime as long as 0.54 ms. As shown in Fig. 9(b), after end-face coupling through tapered optical fibers, an optical net gain exceeding 100dB/cm was measured near the wavelength of 1530 nm, which is the highest unit net gain among all erbium-doped materials so far. However, although the nanowire has a very high unit net gain, limited by its complicated preparation process, its total length is only dozens of micrometers, which largely restricts its final gain and output power. In addition, it is very difficult to align and couple the nanowire with other devices and optical fibers. Special packaging and connection technologies need to be developed, which to a large extent limits the large-scale application of such devices.

Er:polymer possesses several advantages. For example, they are easy to process, have relatively low costs, and the wrap of rare-earth ions by organic ligands can reduce concentration quenching. In addition, polymers also have good mechanical properties and flexibility and can be used to fabricate flexible optoelectronic devices. However, Er:polymer also has some drawbacks. For instance, they have poor thermal stability and are prone to deformation and degradation at high temperatures, so heat dissipation measures need to be taken. Their gain stability is insufficient, and they are easily affected by environmental factors. For scenarios with high requirements for stability, strict environmental control and monitoring are necessary. Relatively low optical damage threshold makes them easily optical damaged under high-power pumping, which largely limits the output power and application range of Er:polymer waveguide amplifiers. Currently, Er/Er-Yb:polymer waveguide amplifiers mainly adopt two types of waveguide structures. One is the embedded waveguide structure, that is, waveguide grooves are prepared using SiO2 or polymethyl methacrylate (PMMA), and then Er/Er-Yb:polymer is spin-coated to fill the grooves to form waveguides. The other is the Er/Er-Yb:polymer cladding structure. After waveguides are prepared using SU-8 photoresist, the active gain medium is directly spin-coated on them.

The team led by Daming Zhang is one of the earlier teams that studied Er/Er-Yb:polymer waveguide amplifiers. In 2024, this team synthesized Er-Yb co-doped sodium lutetium fluoride nanocrystals through the high-temperature thermal decomposition method. Then, they copolymerized these nanocrystals with methyl methacrylate to form a nanocomposite material and fabricated a 0.5-cm-long strip-loaded waveguide amplifier. When the on-chip pumping power was 77 mW, the internal net gain of the EYCDWA was as high as 17.7 dB.88 The fabrication process of this EYCDWA was relatively simple. Standard lithography and inductively coupled plasma (ICP) etching techniques were used to accurately form a rectangular structure composed of grooves with a depth of 2μm and a width of 3μm on the surface of an 8-μm-thick SiO2 film, as shown in Fig. 10(a). Subsequently, the grooves were filled with commercial SU-8 2002 photoresist using the spin-coating technique to fabricate the embedded strip waveguide. Finally, a 2-μm-thick nanoparticle-polymethyl methacrylate nanocomposite film was spin-coated onto the silica surface to complete the fabrication of the strip-loaded structure optical waveguide amplifier. Using the undoped SU-8-loaded waveguide, the transmission loss of the polymer waveguide amplifier was significantly reduced, reaching as low as 1.8dB/cm at 1530 nm. In 2016, this team adopted the same process to fabricate a 1.3-cm-long EYCDWA,114 as shown in Fig. 10(b). With an input signal power of 0.1 mW and a pumping power of 400 mW, a net gain of 15.1 dB at 1530 nm was achieved, which was the highest gain value reported at that time. The difference between this achievement and that shown in Fig. 10(a) lies in the fact that PMMA material instead of SiO2 was used to prepare the grooves, and the waveguide structure was an inverted ridge structure. The transmission loss of this structure was higher than that of the embedded SU-8 waveguide structure. The team’s previous research work is also worthy of attention. They mostly adopted the inverted ridge waveguide structure and used erbium-doped nanocrystal composite materials as the waveguide core layer.

Another approach is to first fabricate the passive waveguide structure of SU-8 and then directly spin-coat the gain medium onto the waveguide to form a waveguide amplifier. Compared with the method of preparing grooves first, this structure has a simpler fabrication process. Moreover, it also uses undoped SU-8 waveguides to transmit the signal light, and the up cladding gain medium is used to amplify the signal light, which can reduce the transmission loss of the polymer waveguide amplifier. Shi et al.84 developed a silicon-based polymer optical waveguide amplifier doped with one-dimensional erbium coordination chains [Er(DBTTA)3(FDPO)]n, as shown in Fig. 11(a). They improved the local concentration of Er3+ and the radiation efficiency through the bridging of phosphorus oxide ligands, as well as the efficiency of intramolecular energy transfer from ligands to Er3+. By adjusting the cross-sectional dimensions of the waveguide to increase the overlap between the 1.5-μm signal light and the 365-nm light emitting diode (LED) pumping light fields, in a waveguide with a cross-sectional dimension of 2μm×3μm and a length of 1 cm, unit gains of 10.5 and 8.5dB/cm at wavelengths of 1.53 and 1.55μm were achieved, respectively. This achievement adopted the top LED pumping method. The pumping light was distributed more evenly along the entire length of the waveguide, and the pumping light could be effectively supplemented. However, the overall efficiency of top LED pumping was relatively low. In the same year, this team constructed a mononuclear erbium complex Er(DBTTA)3(DBFDPO) and fabricated an EDWA based on this material. In a waveguide with a cross-sectional dimension of 4×4μm2, a relative gain of 8.2dB/cm in the C-band was achieved.92 As can be seen from Fig. 11(b), by adding an aluminum mirror, the absorption efficiency of the device for LED pumping light was further improved, and the relative and internal gains were increased to 11.6 and 7.4dB/cm, respectively. The top pumping method provides new ideas for the pumping mechanism of EDWAs/EYCDWAs. Especially for erbium-doped organic ligands, the absorption efficiency of erbium complexes for pumping light can be improved by changing the ligands. However, this scheme may have limited effects in inorganic host materials. In addition, the idea of adding an aluminum mirror is beneficial for better confining the light field within the gain medium, which is helpful for the performance of the device. However, as discussed earlier, the thermal stability issue is the most critical that limits the practical application of Er:polymer waveguide amplifiers.

Currently, the vast majority of on-chip waveguide amplifiers adopt strip or ridge waveguide structures. In fact, the waveguide structure is of great importance for reducing the transmission loss of the waveguide, which will, in turn, affect the actual gain of the waveguide amplifier. The strip waveguide is the most common waveguide structure. It has a relatively strong ability to confine the light field, but it is easily affected by the roughness of the waveguide surface and sidewalls. The ridge waveguide adds a ridge structure on the basis of the strip waveguide. Compared with the strip waveguide, the ridge waveguide is less affected by the roughness of the waveguide sidewalls and has lower transmission losses. However, its ability to confine the light field is relatively weak. Especially when the waveguide is bent, light is prone to leak into the cladding, causing radiation losses. For waveguide amplifiers, to obtain higher gains and output powers, simple strip and ridge waveguide structures can hardly meet the requirements of practical applications. Therefore, it is of research significance to design special waveguide structures to enable the signal light and the pumping light to fully interact in the gain medium. In this section, we will specifically analyze several waveguide structures that are very suitable for the design of EDWAs/EYCDWAs, such as low-confinement waveguides,27^,93 large-mode-area waveguides,22^,94^–96 slot waveguides,19 and double-layer waveguides.21^,57^,66^,79

Both high-confinement and low-confinement waveguide structures can achieve low-loss characteristics. Especially for the low-confinement waveguide, due to its thinner and wider core layer, it is less affected by the roughness of the waveguide sidewalls and has extremely low waveguide transmission losses. In the published works so far, ultra-low transmission losses have been achieved at low-confinement waveguide structures on both Si3N4 and Ta2O5 platforms.27^,109 Blumenthal’s93 team put forward the concept of the ULLW integration platform, in which passive structures can be integrated with active gain media and serve as gain modules for on-chip lasers and amplifiers. However, this structure had not been experimentally verified until 2024 when Qiu et al.27 at the Optical Fiber Communication Conference published relevant work on multi-lane hybrid integrated Si3N4 waveguide amplifiers, as shown in Fig. 12(a). In this work, they fabricated an erbium-doped silicon nitride waveguide with a width of 5μm and a thickness of 200 nm, achieving a net gain of 15 dB and an output power of 22 mW at a length of 17 cm. Compared with the team’s previous work, the 200-nm-thin platform reduced the thickness of Si3N4 film deposition. Therefore, the energy of Er3+ implantation was decreased from 2 MeV to 500 keV, which greatly reduced the difficulty in preparing the gain media. On the basis of this work, they also fabricated a tunable erbium-doped laser covering the C-band,12^,26^,115 shown in Figs. 12(b) and 12(c). The low-confinement waveguide structure has certain advantages in reducing waveguide transmission losses. However, the coupling between the fiber and the waveguide end face is rather difficult, and the mode field mismatch is relatively serious. In the future, it may be necessary to design special coupling structures to reduce the end face coupling losses.

LMA structures are commonly used in fiber amplifiers and lasers to increase the gain and output power of the devices. However, the application of similar structures in planar waveguides is still relatively rare at present. Singh et al.95 first proposed the concept of the LMA waveguide and achieved on-chip output power at the watt level. When the light field propagates in this structure, it has a very large mode field area, which means that higher pumping power can be used and more rare-earth ions can be activated. They used Tm:Al2O3 as the gain medium, which is instructive for the realization of EDWAs/EYCDWAs with LMA structures. Bao et al.22 first realized the EDWA based on the LMA structure. A net gain of 16 dB was achieved at a length of 7 cm, and the output power reached 113 mW, as shown in Fig. 13.

This is the second EDWA that has achieved an output power at the hundred-mW level since Er:Si3N4. In this work, the LMA structure is simpler. A very wide waveguide structure is directly fabricated in the non-bending area, whereas a narrower waveguide is fabricated in the bent waveguide area to meet the single-mode condition and avoid the loss of higher-order modes at the bends. The straight waveguide and the bent waveguide achieve a smooth transition through a tapered structure. Compared with single-mode EDWA, LMA-EDWA has a higher pumping energy conversion efficiency. Due to its larger mode field area, it can withstand higher power without easily suffering from nonlinear effects and optical damage issues. However, the LMA is not suitable for all scenarios. Its size is larger compared with that of the small mode area structure. In some application scenarios with very strict requirements for spatial size, it may be subject to certain limitations. In addition, the radiation loss of the bent waveguide in the LMA structure is very large. To reduce the loss in the bent part, more complicated designs may be needed to achieve lower radiation losses and mode mismatch losses, which increases the difficulty in the fabrication process of LMA-EDWA to some extent. In fact, depending on different application scenarios, EDWA with LMA/SMA structures can be used in combination to achieve a compromise among output power, fabrication difficulty, integration level, and so on.

The slot waveguide structure usually has a relatively small size and a compact design, which enables it to amplify optical signals within a small space and fabricate devices with a relatively high level of integration. Rönn et al.19 used the ALD technique to deposit Er:Al2O3 into Si3N4 slot waveguides and fabricated a slot-EDWA with a unit net gain as high as 20.1±7.31dB/cm,19 as shown in Fig. 14. However, the fabrication process of slot waveguides is extremely complex and the transmission loss is greatly affected by the waveguide sidewalls, the length of the slot waveguides fabricated in this work is limited, so the final output power is not high.

Moreover, the slot waveguide structure is rather special. When integrating it with other optical devices in SBO chips, such as waveguide splitters and modulators, it is necessary to design special coupling structures to achieve low-loss coupling among different devices. Also, the width and height of the slot need to be strictly controlled to avoid a decrease in the gain performance in the slot waveguide amplifier. In addition, the high power density characteristic of the slot waveguide also limits the final output power of the device to some extent. When the power is too high, extremely high heat may be generated in the slot area. If the heat cannot be effectively dissipated, it will lead to a decline in device performance or even damage to the device. In the practical application process of slot-EDWA, the design of appropriate thermal management solutions is necessary, which further increases the complexity and cost of the system. Therefore, although slot-EDWA has a relatively high unit net gain, there are still many issues that need to be improved to fabricate devices with high output power.

The double-layer waveguide consists of the passive and active layers with similar refractive indices, where the waveguide structure is formed by the passive layer without direct etching of the active layer, and thus, its structure is somewhat similar to the ridge structure. Sometimes, the passive and active layers could be different materials, but their refractive indices are similar. The bilayer Er:TeO2-EDWA,79 double-layer Ta2O5-EDWA,66Er:TeO2-coated EDWA,75 and monolithic integrated double-layer EDWA20 mentioned previously can all be classified as double-layer EDWA, as shown in Figs. 6(b), 7(a), 7(b), and 8(b), respectively. The greatest advantage of the structures shown in Figs. 7(a) and 7(b) is that low-loss materials with similar refractive indices are used as the core layer for guiding modes to reduce the transmission losses of double-layer EDWA/EYCDWA. Although these structures are less affected by scattering losses, they have relatively large radiation losses at the bends. To minimize the bending radiation losses as much as possible, more research can be conducted on the design of bent waveguides in the future. The structure shown in Fig. 6(b) prepares the passive and active waveguides in separate layers, providing a good idea for the monolithic integration of active and passive components. However, the overall fabrication process is more complicated, and the alignment issue between the passive and active waveguides needs to be taken into consideration.

In Sec. 2, we provide a comprehensive review of several gain media and waveguide structures that are most widely studied at present. Due to space limitations, we do not discuss in detail all the work carried out on various materials. To provide readers with a more direct understanding of the relevant research progress, we present an overview of the representative works in Table 1. The overview includes gain material and waveguide preparation processes, waveguide geometries, propagation losses, and final gains, enabling readers to better trace the development of waveguide amplifiers based on different host materials.

High-quality and strongly luminescent Er-doped/Er-Yb co-doped gain media are the key to the preparation of EDWAs/EYCDWAs. At present, gain media mostly are prepared via the following methods, including ions implantation,116 ALD,19^,63 magnetron sputtering,49 and the ion slicing combined with bonding process44 used for the preparation of Er:TFLN, corresponding to Figs. 15(a)–15(d). Ion implantation technology can control the doping concentration by adjusting the energy and fluence of implanted Er3+ ions. Selective doping in local areas can be achieved through processes such as using masking layers and multiple implantations. A relatively uniform doping distribution to some extent can be achieved via careful design of the dose and beam energy of ion implantation. However, to obtain a greater implantation depth, high-energy ion implantation is often required, which will cause damage to the material surface. High-temperature annealing is needed to eliminate damages and avoid introducing large waveguide losses. Besides, due to the complex process and high cost, there are currently no cases of using this process to achieve Er-Yb co-doping. ALD can achieve atomic-level control of the film thickness, ensuring the uniformity and accuracy of doping. Moreover, the films deposited by ALD have excellent step coverage and can achieve uniform deposition on slot waveguides and other complex three-dimensional waveguide structures. However, the biggest drawback of ALD is its extremely slow deposition rate. It is more suitable for preparing thinner gain media. Compared with ALD, although the film quality prepared by magnetron sputtering is poor, its deposition rate is higher, and it can quickly achieve large-area doping. Besides, magnetron sputtering is the most studied process in the preparation of Er-Yb co-doped gain media at present. It has certain advantages in the preparation of large-area erbium-doped gain media with certain requirements for cost and production efficiency. Figure 15(c) shows the schematic diagram of using magnetron sputtering to prepare Er-Yb co-doped alumina gain media. The ion slicing combined with the bonding process is also a very effective process for preparing erbium-doped gain media. At present, it is mostly applied to the preparation of Er:TFLN wafers. The greatest advantage of this process is that it can achieve the combination of different materials, especially crystalline materials, through slicing and bonding. However, the whole process is very complex.

The sidewall and surface roughness play an important role in determining waveguide transmission losses. Therefore, how to prepare smooth waveguide sidewalls and surfaces is a crucial factor influencing EDWAs/EYCDWAs. Figures 16(a)–16(c) respectively display the process flowcharts of the damascene process,117 EBL combined with ICP-RIE,118 and PLACE.119 Currently, in the aspect of Er-doped waveguide processing, all three techniques have demonstrated considerable potential. In the damascene process, the SiO2 grooves are first etched, and then the waveguide structure is formed through thin film deposition. This avoids directly etching the gain medium, where a micromask is easily formed by the aggregated rare-earth ions, leading to a rough sidewall.

Liu et al.18 first utilized this process to fabricate Er-doped waveguides with losses as low as 5 dB/m. The whole process is CMOS-compatible, which is beneficial for large-scale production and the fabrication of complex integrated optical structures. However, the overall process involves numerous and rather complicated steps. The technique of EBL combined with ICP, as shown in Fig. 16(b), was proposed by the research group at Harvard University in 2014 and is currently widely used in the fabrication of LNOI devices.118 EBL has an extremely high resolution, and ICP has a good etching rate and excellent perpendicularity, which can meet the requirements for processing waveguide structures with high precision and good sidewall perpendicularity. Nevertheless, EBL has a low efficiency, and large-area processing consumes time and money. Thus, it is suitable for scientific research and small-batch high-precision processing. Wu et al.120 proposed the PLACE technique combining chemical mechanical polishing with femtosecond laser direct writing technology. After depositing a layer of metal chromium (Cr) mask with much higher hardness than lithium niobate on the surface of LNOI, the selective etching is realized in LNOI. This process can achieve the fabrication of three-dimensional waveguide structures, and combined with polishing, it can improve the surface quality. However, limited by the spot size of the femtosecond laser, this process may encounter some difficulties when it comes to the requirement for higher precision processing. In addition, although the chemical mechanical polishing technology can smooth the waveguide surface, waveguide breakage may easily occur during polishing, which may lead to device damage. How to improve the yield of devices is an important issue that this process needs to consider. More critically, this process is not CMOS-compatible. How to integrate it with mainstream processing platforms and how to integrate it with other devices for large-scale processing of complex devices are all problems that need to be solved, although it can achieve the customized processing of some simple devices.

EDWAs/EYCDWAs, as important components of SBO chips, demonstrate considerable application potential in various scenarios such as optical communication, optical signal processing, and optical sensing. Chips with waveguide amplifiers are expected to be directly applied in different fields such as optical interconnections, medical and sensing devices, and quantum optics in the future. Figure 17 summarizes the different application scenarios of EDWAs/EYCDWAs. Among them, Fig. 17(a) shows the first terabit-class coherent optical transmission system achieved using EDWAs, realizing single-channel 1.6-Tb/s coherent transmission in the C-band and 16-channel 25.6-Tb/s WDM transmission over an 81-km optical fiber.29 In this system, EDWAs are used as booster amplifiers for coherent transmitters, which undoubtedly proves the application potential of EDWAs in high-speed coherent transmission. In fact, even earlier, EDWAs had already been used as local amplification modules at different nodes of optical switches and optical routers. By amplifying optical signals, they can compensate for the continuous attenuation of optical signals during processing and transmission, ensuring that optical signals can be accurately transmitted along the required paths and avoiding communication errors caused by signal loss.

Figure 17(b) presents another important application of EDWAs/EYCDWAs, which is on-chip integrated lasers. By designing appropriate resonant cavities for high-gain waveguide amplifiers, high-performance optically pumped laser outputs can be generated. Apparently, for on-chip integrated Er-doped lasers, apart from the gain medium, the key factor that determines their performance is the design of the waveguide resonant cavity. On the one hand, the resonant cavity can confine the photons generated by stimulated emission within the cavity, making them be reflected back and forth inside the cavity, thereby enhancing the intensity of stimulated emission, increasing the gain output. On the other hand, the resonant cavity can play a role in selecting the wavelength of stimulated emission. The wavelengths of light waves oscillating inside the cavity need to meet the standing wave condition to achieve coherent superposition, ensuring a stable laser output frequency. Currently, the mainstream waveguide resonant cavity structures mainly include Fabry–Pérot (FP) resonant cavities, distributed feedback resonant cavities, distributed Bragg reflector (DBR) resonant cavities, and micro-ring resonant cavities.

Figure 17(b) shows the FP-type,12 DBR-type,17 and micro-cavity-type resonant cavities13 from left to right in sequence. Among them, the FP type has achieved a fully integrated Er-doped laser with a narrow linewidth of 50 Hz, a high output power of 17 mW, and a wavelength tuning range exceeding 40 nm. The realization of Er-doped lasers makes on-chip high-speed, low-noise, and high-output-power optical engines possible, and these engines can be directly applied in different fields such as LiDAR and high-speed coherent communication links. Figure 17(c) shows the solutions for the frequency-modulated continuous-wave (FMCW) source in LiDAR.121^,122 It can be seen that EDWAs, as the core components of these solutions, play an important role in laser generation, pre-amplification, and power amplification.

The applications of EDWAs/EYCDWAs go far beyond the scenarios mainly mentioned above. For example, in the field of communication, thanks to their ability to uniformly amplify optical signals in different modes, in addition to traditional long-distance and high-speed optical communications, EDWAs/EYCDWAs may play an important role in emerging technologies such as space division multiplexing and mode division multiplexing in the future. Besides, due to their high integration and stable performance, these amplifiers are also expected to be applied in fields such as visible light communication and free space optical communication, providing support for achieving broader wireless communication coverage and higher data transmission rates. In the field of computing, as the scale of optoelectronic computing continues to expand, EDWAs/EYCDWAs can be integrated with other optical devices as local amplification modules to compensate for the optical transmission losses among chips in real time, enabling signal parallel processing and computing, and thus improving the computing speed and efficiency. In the sensing domain, EDWAs and EYCDWAs play a pivotal role by amplifying feeble sensing signals, thereby enhancing both the detection capabilities and the measurement range of the overall system. Take biomedical sensing systems as an illustrative example. Biosensing signals derived from optical detection techniques such as fluorescence and Raman scattering are inherently weak. Here, EDWAs and EYCDWAs step in to amplify these faint optical signals. By doing so, they not only boost the signal intensity but also significantly improve the signal-to-noise ratio. This dual improvement makes the subsequent processes of signal processing and analysis far more accurate and reliable, enabling more precise diagnoses and in-depth biological research.

In terms of the commercial applications, due to high technological requirements and expenses, currently, there are relatively few companies capable of launching commercial EDWAs/EYCDWAs and related products. The vast majority of research is still at the laboratory stage. In fact, as early as the beginning of the 21st century, companies such as Teem Photonics in France and Inplane Photonics in the United States had already demonstrated related products at the Optical Fiber Communication (OFC) conference.97^–106 However, in the early years, compared with EDFAs, EDWAs did not have obvious advantages in terms of output power and research and development costs. Coupled with the limitations of micro-nano processing technologies at that time, which led to difficulties in research and development, EDWAs were not as widely used as EDFAs in fields such as optical communication and sensing. Figure 18 shows some packaged EDWA products. Among them, Figs. 18(a) and 18(b) are the schematic diagrams of the principle and the physical pictures after packaging of early commercial products, whereas Figs. 18(c) and 18(d) are the physical pictures of the packaged prototypes of EDWAs that may be commercialized in recent years.

Figure 18(a) presents the Metro module launched by Teem Photonics. This module integrates a 980-nm uncooled pump source. After packaging, its size is 81mm×35mm×12mm. It can amplify the signal light in the wavelength range of 1530 to 1560 nm for 1−8 channels. When the input optical signal power is −15dBm, the gain is 27 dB.101 This is a landmark product that represents the real application of EDWAs in actual optical networks. Figure 18(b) shows the EDWA product launched by Inplane Photonics during the same period. The upper part of the figure presents the highly integrated lossless bus interface (LBIC) modules introduced by the company for avionic fiber-optic communication networks.103 This module supports functions such as channel selection, signal sampling, signal addition, power detection, and system health diagnosis. It can operate in fixed power, automatic power control, or automatic gain control modes, and the insertion losses on the optical link can be compensated by the internal EDWA. The lower part of the figure shows the metropolitan area network EDWA amplifier product GEM-A2000 launched by the company at OFC 2003. This product adopts the pump light sharing method and integrates four EDWAs.104 Each amplifier can independently control the optical gain. A total of 32 functional optical devices, including splitters, wavelength division multiplexers, variable optical attenuators, and pump light band-stop filters, are integrated on a 9mm×25mm PLC chip. These two products of Inplane Photonics achieve a relatively high level of integration in a rather compact package and possess good functional extensibility, providing a technical reference for the research and development of related products in the future.

Figure 18(c) depicts the physical picture of the packaged erbium doped waveguide laser (EDWL) prototype provided by EDWATEC. EDWATEC is an integrated device manufacturer that offers devices based on rare-earth-ion-doped photonic integrated circuits, including erbium-doped waveguide amplifiers and lasers,12^,18 as well as customized photonic integrated circuit (PIC) design, fabrication, doping, and packaging services. The products of this company are typical cases of the transformation of scientific and technological achievements. The core of its technology lies in the ion implantation technique and the wafer-level damascene process for fabricating low-loss Si3N4 PICs mentioned previously. Figure 18(d) displays the packaged EDWA solution proposed by Bonneville.107 Unlike Er:Si3N4, this solution does not directly implant rare-earth ions into the low-loss Si3N4 waveguide. Instead, it adopts a technical solution that combines LioniX’s TriPleX low-loss Si3N4 platform and gain platform. Although this solution is simpler in the preparation process of the gain medium and has relatively lower costs, there is still significant room for optimization in the currently measured maximum power output. Both the prototypes shown in Figs. 18(c) and 18(d) exhibit highly integrated characteristics. It can be predicted that as information systems develop toward miniaturization and integration, more optoelectronic functional devices will be integrated on the same chip with the continuous progress of micro-nano processing. As power compensation units of the entire chip, EDWAs/EYCDWAs are bound to play a more important role in the future.

In Sec. 3.2, we discussed the applications of EDWAs/EYCDWAs in several typical areas such as optical transmission networks, lasers, and LiDAR. In fact, as crucial components of SBO chips, EDWAs/EYCDWAs can be placed almost anywhere within SBO chips with great flexibility. Figure 19 presents a schematic diagram of an SBO chip integrated with EDWAs/EYCDWAs, with the potential application scenarios of such chips shown on the right. First, in the field of optical communication, high-gain EDWAs have the potential to replace EDFAs.123 Specifically, EDWAs with an output power reaching the hundred-mW level can perfectly assume the roles of pre-amplifiers, post-amplifiers, and in-line amplifiers that are typically fulfilled by EDFAs. Moreover, EDWAs/EYCDWAs can also be potentially applied in SBO computing124 and quantum computing.125 In the era of artificial intelligence (AI), the volume of data operations is growing at an astonishing rate. Relying solely on traditional all-electrical chips makes it difficult to meet the ever-increasing demand for data operations. To this end, SBO chips, which combine the advantages of microelectronics and optoelectronics, offer an effective solution as AI accelerators. However, SBO chips face a significant issue: due to waveguide transmission losses, as the computing network expands, substantial inter-chip losses emerge within the network. Thanks to the high integration of EDWAs, they can be employed as signal-compensating elements in large-scale computing chip networks that traditional EDFAs cannot do.

Optical sensing has also become a hot research area in the SBO field in recent years, especially in areas such as silicon-based on-chip LiDAR, silicon-based image sensing, and biochemical sensing. These application areas share a common feature: they require a significant amount of signal light energy to complete the processes of laser emission or detection.126 Taking FMCW as an example, a key technical challenge restricting its application lies in the limited output power of most on-chip C-band light sources, which limits the detection range of the LiDAR. In contrast, high-gain EDWAs/EYCDWAs can directly fabricate on-chip lasers with an output power of up to the hundred-mW level. In addition, in the subsequent stages, even greater output can be achieved by cascading multiple sets of EDWAs. Evidently, EDWAs and their derivative applications represent highly significant devices for SBO chips and even information systems. In the future, they are expected to be directly applied to every link in information technology.

As a critical component of the SBO chips, EDWAs/EYCDWAs have demonstrated impressive capabilities in amplifying optical signals on-chip within the communication band and are gaining attention for potential applications in other areas. The comprehensive analysis provided in this paper aims to serve as a valuable reference for both research and industrial development in related fields, ensuring that EDWAs/EYCDWAs can play a pivotal role in a broader range of applications. This paper provides a comprehensive review of the research progress of EDWAs/EYCDWAs in the SBO chip field, focusing on the development status, challenges, and future directions. It specifically analyzes recent breakthroughs in materials, structural design, integration technology, and performance optimization of EDWAs/EYCDWAs. A thorough review of the latest advancements in EDWAs/EYCDWAs, composed of various host materials, is presented first. In particular, the factors influencing the gain limitations of waveguide amplifiers, such as host materials and structural design, are examined in detail. The paper also explores the fabrication processes and applications of EDWAs/EYCDWAs, with particular attention given to the fabrication of the gain medium and waveguide structure, which are critical to the technology’s performance. The main fabrication protocols are systematically reviewed and summarized. In terms of applications, the paper discusses various scenarios and commercial products that incorporate EDWAs/EYCDWAs. Finally, it points out the challenges that EDWAs/EYCDWAs may face in the future, as well as the potential applications and development prospects within the SBO chip field. It is foreseeable that in the near future, EDWAs/EYCDWAs, similar to EDFAs, will be directly applied in diverse areas of optical transmission.

Xiwen He received his PhD from Shanghai Institute of Optics and Fine Mechanics (SIOM), Chinese Academy of Sciences (CAS), where he worked in the Department of Aerospace Laser Technology and Systems. His research primarily focuses on integrated erbium-doped waveguide amplifiers (EDWAs), a cutting-edge area in silicon-based optoelectronics that combines erbium doping techniques with silicon-based platforms to develop high-performance on-chip optical amplification.

Zheng Zhang received his PhD from Ningbo University under the supervision of Professor Rongping Wang. His research primarily centers on silicon-integrated EDWAs, on-chip optical nonlinear devices, and optical sensing applications. Throughout his doctoral studies, he designed and fabricated a range of on-chip photonic devices utilizing tantalum oxide waveguides, such as EDWAs, on-chip supercontinuum generation, and low-loss microring/microdisk resonators. His work has contributed to the exploration of novel material platforms in the field of silicon-based optoelectronic integration.

Rongping Wang received his PhD from the Institute of Physics, Chinese Academy of Science, Beijing, China, in 2000. Then he worked at the Isreal Institute of Technology (2001), Institute of Advanced Science and Technology of Japan (2002-2005), and Australia National University (2005-2017). Since 2017, he has been a professor with Ningbo University, Ningbo, China. He has been involved in the research of optical materials and related devices. He is the author of more than 300 research papers with an H-index of 41.

Zhiping Zhou received his PhD from Georgia Institute of Technology (GT), USA, in 1993. From 1993 to 2005, he was with the Microelectronics Research Center at GT, where he engaged research and development in the areas of silicon-based optoelectronics, ultrafast optical communications, integrated optoelectronics, semiconductor devices and sensors, and nanotechnology. He is now a professor at Peking University, China. He has been credited for over 650 technical papers, presentations, and patents. He is a fellow of OPTICA, SPIE, and IET. He serves as honorary director of the Chinese Optical Society and managing director of the Chinese Society for Optical Engineering (CSOE), as was the founding editor-in-chief of Photonics Research. He was also founding chair of the IEEE Wuhan Section (2007–2008) and director of IEEE Atlanta Section (2001–2003). He has also chaired, co-chaired, and served on many program committees for various conferences of IEEE Photonics Society, OPTICA, SPIE, COS, and CSOE.

Weibiao Chen received the BS degree in electronics and information system and the PhD in physical oceanography from the Ocean University of China, Qingdao, Shandong, China, in 1992 and 1997, respectively. He completed his postdoctoral studies at Chiba University, Chiba, Japan, in 2000. He is currently the director of Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai, China. He is also a member of the Laser Committee and the Space Optics Professional Committee at the Chinese Optical Society, vice chairman of the Environmental Optics Committee, and executive director of the Chinese Society for Optical Engineering. He was awarded the titles of Top Ten Outstanding Young Scientists by the Chinese Academy of Sciences and Shanghai Outstanding Academic Leader. He has long been researching the atmospheric and oceanic laser remote sensing sciences and laser technology.

Biographies of the other authors are not available.