Solution-processed semiconductor NCs have attracted considerable attention since their discovery due to their broadband absorption and tunable emission. So far, many colloidal NCs (including CdSe, ZnS, InP, CuInS₂, AgS) that exhibit good optical, electrical, and magnetic properties have been synthesized^35-40. Among them, CdSe NCs have been extensively investigated due to their advantages such as high PLQY, good stability, facile synthesis methods and emission in the visible band^13,41,42. In terms of morphology, CdSe NCs are divided into QDs, NRs and NPLs, as shown in the transmission electron microscope (TEM) image in Fig. 1a-c. Due to the quantum confinement effect, their absorption and emission can be tuned from the ultraviolet to the infrared regions, as shown in Fig. 1d-f. In addition, their different morphologies contribute to obvious differences in optical properties. For example, the full width at half maximum (FWHM) and Stokes shift of QDs, NRs, and NPLs gradually decrease, while the absorption cross-section gradually increases. The PLQY and optical stability of CdSe NCs will be difficult to be compared due to the different crystal quality and experimental conditions. In addition, the optical properties of CdSe NCs are affected by surface dangling bonds. In order to improve the optical properties and stability of CdSe NCs, these 0D QDs, 1D NRs, and 2D NPLs are usually coated with another semiconductor with a wider bandgap to passivate the surface defects, such as CdS, ZnS, ZnSe and CdTe^15,43-45. These nano-heterojunctions (CdSe/CdS, CdSe/ZnS, CdSe/ZnSe core/shell etc.) exhibit superior optical properties compared to bare CdSe in terms of PLQY, fluorescence lifetime, absorption cross-section, stability, optical gain, etc.

The optical gain of CdSe NCs is a crucial prerequisite for realizing NC PICs. Optical gain, which is also referred as population inversion, is related to the properties of the material itself. Optical gain is usually characterized with the adoption of a power-dependent pump-probe transient absorption (TA) technique⁴⁶. Specifically, the optical density of the sample is ΔA + A₀, where, ΔA denotes the absorbance change caused by light-induced absorption, and A₀ is the steady-state absorbance of the sample. NCs could produce optical gain when ΔA + A₀ < 0 or the absolute value of ΔA/A₀ is greater than 1⁴⁶. At low power excitation, the TA spectrum of CdSe NCs only exhibits exciton bleaching. At this time, ΔA + A₀ > 0, and no optical gain occurs, as shown in Fig. 2a. With the increase of power to its maximum value, an additional broadband negative peak is exhibited on the low-energy side of the exciton-bleaching signal, as shown in Fig. 2b, which can be attributed to the optical gain of the NCs²⁰. The optical gain threshold of the NCs can be obtained by extracting the signal of the broadband peak and then subtracting the steady-state absorbance, as shown in Fig. 2c. Different from optical gain, ASE is the amplification of spontaneously emitted photons by stimulated emission when propagating in the medium. The ASE of CdSe NCs is usually characterized by power-dependent fluorescence. However, it is worth noting that spontaneously emitted photons suffer optical losses when propagating through the medium. Therefore, at threshold power, the optical gain is greater than the optical loss, and an additional sharp peak appears on the spectrum, as shown in Fig. 2d ⁴⁷. The optical loss in NCs may be waveguide loss due to reabsorption. Recently, Ahn et al. found that the indium tin oxide (ITO) substrate would hinder the light amplification of NCs due to strong free-carrier absorption in the visible band when building a laser with an light-emitting diode (LED) structure^48-50. Therefore, some materials can exhibit optical gain but fail to produce ASE once the optical loss is greater than the optical gain. Therefore, the gain capability, substrate selection, and reabsorption effect of the materials play a decisive role in the realization of ASE from NCs. Lasers are constructed by combining optical amplification of an optical gain medium with an optical resonant cavity. Therefore, the optical gain and optical loss of the laser depend on the ASE of the NCs. In addition to the induced optical loss of NCs, the substrate also affects the optical mode of NCs lasers. Recently, Ahn et al. found that the refractive index of the substrate will cause the NCs laser light field to be ‘pulled’, thereby increasing the waveguide loss⁵⁰. In addition, the optical loss of the resonator also affects the threshold of the laser^51,52. Compared with the ASE threshold, a high-quality resonator can lower the excitation threshold. Therefore, appropriate substrate and cavity design are also important for the NC lasers.

An important requirement for the application of CdSe NCs in PICs is ASE since the information transfer of photonic chips depends on lasers with high monochromaticity and coherence. Over the past two decades, extensive breakthroughs and significant progress have been made in understanding the ASE properties of CdSe NCs^53,54. In general, the realization of the ASE in CdSe QDs requires that the stimulated emission should overcome optical absorption^53,54. The structural scheme of CdSe QDs ASE can be described by a two-level system. As shown in Fig. 3a, in the absence of optical pumping, two electrons with opposite spins are distributed in the ground state. Spontaneous emission is characterized by an electron being excited to the conduction band and then recombining with a hole to release a photon (Fig. 3b). This process does not produce stimulated emission due to the equal amount of spontaneous emission and optical absorption⁵³. For CdSe QDs, due to the non-uniform degeneracy of the band-edge states, ASE is only possible when two electrons are excited into the conduction band to form a multi-exciton (Fig. 3c)⁵³. Indeed, numerous experiments have been reported to demonstrate that the ASE of CdSe QDs is originated from multi-exciton recombination^53,54. Moreover, the ASE is generally red-shifted relative to the spontaneous emission, which further confirms that the ASE of CdSe QDs is derived from multi-exciton recombination. Similar to CdSe QDs, the ASE mechanism of CdSe NRs and dot-in-rod (DIR) is also from biexciton^48,52,54. However, the ASE mechanism of CdSe NPLs is complex since it depends on the size and excitation density. Specifically, when the size of the CdSe NPLs is smaller than that of the Coulomb bound exciton, the optical gain of the CdSe NPLs is similar to that of the CdSe QDs, which is biexciton gain. However, when the lateral size of CdSe NPLs is larger than that of the Coulomb bound excitons, multiple excitons exist simultaneously in the same NPLs⁴⁶. In this case, the maximum exciton number of CdSe NPLs depends on the size of NPLs rather than the state filling, which results in higher ASE threshold for larger-sized NPLs³¹. In addition to exciton states, the Coulomb-bound biexcitons may lead to optical gains of CdSe NPLs. This is ascribed to the fact that the biexciton binding energy of NPLs (30 to 40 meV) is greater than the room temperature thermal energy (25 meV), which allows the existence of biexcitons stably^27,55. It is worth noting that since 2D NPLs could reduce dielectric shielding, the ASE mechanism of both large- and small-sized NPLs may be originated from Coulomb-bound biexcitons. In addition, when the NPLs are excited by high power density, the lattice heating generated by Auger decay and the increase in charge-carrier screening may dissociate the Coulomb-bound space, which may lead to stimulated emission from e-h plasma. In the study conducted by Tomar et al., it was found that CdSe NPLs exhibit Coulomb-bound biexciton gain when excited at low power⁵⁶. With the increase of excitation power, an additional blue-shifted gain peak appears at the high-energy edge, which is the optical gain induced by e-h plasma. Although exciton-exciton scattering is also a common optical gain mechanism⁵⁷, this mechanism has not been observed in CdSe NCs. Although multiple mechanisms may act on the optical gain of CdSe NCs, the optical gain of the NCs regardless of the gain mechanism is hindered by the nonradiative Auger recombination (AR)⁵⁸. Nonradiative AR can be explained by the fact that the energy of electron and hole recombination is transferred to a third particle (electron or hole), which is extremely disadvantageous for achieving ASE and lasing. In the last 20 years, many methods to suppress AR have been demonstrated, such as increasing the volume^19,54, changing the morphology of NCs⁵⁹, coating with wide bandgap materials, and smoothing the interface potential³². A more detailed discussion can be found in these reviews.

The ASE of CdSe NCs (QDs) was first demonstrated by Klimov et al. in 2000¹⁷. This pioneering work revealed the underlying physical mechanism and conditions for the realization of NCs lasers and provided important guidance for the future exploration of NCs lasers. Following this work, CdSe NCs have been extensively investigated as a laser gain medium. To further lower the ASE threshold of CdSe NCs, wide bandgap shells (ZnS, CdS, CdTe, etc.) were deposited on the CdSe surface to suppress the AR caused by surface defects^32,61. The absorption cross-section of the NCs was also simultaneously enhanced⁵⁴. Due to the energy band misalignment between the CdSe core and the shell, the electron-hole wavefunction overlap is also an important factor for the optical gain of CdSe NCs. In general, the core/shell band alignment is classified into type I (Fig. 3d), quasi II (Fig. 3e), and type II (Fig. 3f). Compared with the type II and quasi II band alignments, the AR effect of the type I band alignment is the most intense since the electron-hole wave function is confined to the core. Although electrons and holes are separated the type II band alignment into different regions (core or shell), the reduced intensity of the optical oscillator strength makes it difficult to achieve optical gain. The CdSe NCs with quasi II band alignment are promising for optical devices because they not only delocalize the electron-hole wave function but also reduce defects²⁸. For example, under the same conditions, the ASE threshold of type I CdSe/ZnS QDs (300 μJ/cm²) under femtosecond laser pumping is an order of magnitude higher than that of quasi-II CdSe/CdS QDs (26 μJ/cm²)⁶². In addition, the AR of CdSe NCs can be further suppressed by smoothing the core/shell interface with the adoption of the core/alloy/shell structure³². In a recent study performed by Klimov et al., electric pump gain was achieved for the first time adopting CdSe/Cd_xZn_1-xSe/ZnSe_0.5S_0.5 core/alloy/shell structure with an Auger lifetime of 2.4 ns (Fig. 4a)²⁰. For a more intuitive comparison, the optical gains of CdSe NCs with different types (morphology and band alignment) are summarized in Table 1, including ASE and laser threshold, emission wavelength and pump source. Obviously, for 0D CdSe QDs, the multi-shell alloy structure has the smallest ASE and laser threshold, which could be attributed to the passivating defects of multi-shell structure and the smoothing of the interface by the alloy structure.

Compared to 0D spherical QDs, CdSe NRs exhibit lower ASE thresholds due to the increase in absorption cross-section with increasing volume⁶⁹. The charge carriers can move freely along the length of the NRs but are confined in the diameter direction^12,70. The unique structure of NRs leads to different carrier relaxation and optical gain compared to QDs^59,70-72. The lower ASE threshold of CdSe/ZnS NRs than that of QDs has been confirmed by Kazes et al. More importantly, the ASE of CdSe/ZnS NRs can be observed at room temperature, while that of QDs with the same diameter can only be observed at 50 K⁷³. At the same time, it was also found that the ASE threshold of the CdSe/ZnS core/shell NRs remained constant below 120 K and increased with increasing temperature above 120 K due to Auger heating effect and phonon-assisted thermal relaxation. In addition to CdSe NRs, DIR nano-heterojunctions composed of CdSe QDs as cores and CdS NRs as shells have attracted more and more attention in recent years^69,74. A lower gain threshold can be found in CdSe/CdS DIR than in CdSe/CdS QDs due to the larger volume of the former⁶⁹. Importantly, the carriers transit from the core and shell states, respectively, leading to the observed two-color ASE in the core and shell of the CdSe/CdS DIR⁶⁶. In addition, Iwan et al. found that the ASE threshold of CdSe/CdS DIR gradually decreased with increasing rod length⁶⁹. At the same time, the ASE threshold of CdSe/CdS DIR is independent of temperature below 350 K, which is extremely important for on-chip lasers. Similarly, the ASE and laser characterization of different types of CdSe NRs are summarized in Table 1.

Compared with CdSe QDs and NRs, CdSe NPLs are the most promising in the application of ASE and lasers due to their extremely large absorption cross-section, giant oscillator strength, narrow emission band, and long Auger lifetime^31,47,76-78. According to the unique layered structure, CdSe NPL heterojunctions can be divided into core/shell and core/crown structures^79,80. In general, the CdSe core/shell NPL structure can achieve continuous regulation of the emission wavelength⁶⁸. However, the CdSe core/shell NPLs structure also changes the emission FWHM and the quantum confinement effect in the vertical direction. For the CdSe core/crown NPL structure, the growth of the crown not only passivates the lateral defects of the NPLs, but also preserves the strong quantum confinement effect. Therefore, it is difficult to continuously tune the emission wavelength of the core/crown NPLs^34,81. This difficulty has recently been addressed by Wu et al., which is beneficial for realizing continuous modulation of the on-chip laser peak⁸¹. For the same volume as CdSe QDs and NRs, NPLs exhibit lower ASE threshold, which may be related to the localized field and large absorption cross-section⁵⁴. Taking advantage of the unique layered structure, CdSe NPLs can be stacked by special methods. As shown in Fig. 4b, the ASE threshold of CdSe NPLs can be gradually decreased with the increasing number of stacked layers⁷⁵. Compared with bare CdSe NPLs, the ASE threshold can be lowered by growing a crown on the lateral surfaces to passivate the surface defects, as shown in Fig. 4c ⁵¹. For the CdSe/Cd_xZn_1-xS core/shell structure, the ASE peak position can be continuously regulated by adjusting the ratio of Zn and Cd in the shell. The minimum ASE threshold was found in the study by Volkan et al., when x was equal to 0.75⁶⁸. For the CdSe_xS_1-x core, CdSe_xS_1-x/CdS core/crown, and CdSe_xS_1-x/CdS core/shell NPLs, the CdSe_xS_1-x/CdS core/shell structure exhibits a low ASE threshold due to surface passivation and reduced AR rate (Fig. 4d-e). Unlike CdSe QDs and NRs, the ASE threshold of 2D NPLs tends to decrease with decreasing temperature due to the reduction of band-edge exciton degeneracy by exciton coherence region extension⁸². Exciton coherence extension does not occur in QDs and NRs^46,82. Furthermore, Lian et al. found that the exciton coherence region of the CdSe NPLs decreases with increasing thickness of the NPLs^46,82. Therefore, thin CdSe NPLs are potential candidates for the realization of on-chip lasers⁸³. Similarly, the ASE and laser characterization of different types of CdSe NPLs are summarized in Table 1.

As mentioned earlier, the ASE threshold of CdSe NCs could be determined by AR due to the degeneracy of the band-edge energy levels. How to suppress non-radiative AR and explore new optical gain methods has become a hot topic in the researches of optical gain in CdSe NCs. With the advancement of CdSe NCs synthesis methods and the design of advanced devices, some new ASE concepts have been demonstrated, such as single-exciton ASE, continuous-wave ASE, multi-photon and electric pump ASE. These advances provide important references for both lowering the ASE threshold of CdSe NCs and achieving the PICs.

As the heart of PICs, lasers with high monochromaticity and coherence are essential. Transferring good, cheap lasers to tiny photonic chips is a goal currently pursued by the scientific community. Since CdSe NCs were demonstrated to exhibit ASE properties, many lasers based on CdSe NCs, including WGM, DFB, DBR, random lasers and photonic crystals have been demonstrated^{19,21,22,65,95,96}. For example, a CdSe/CdS/ZnS QDs multimode WGM laser was constructed by Wang et al. by combining NCs with optical fibers (Fig. 6a)⁹⁷. The lasing behavior is maintained at 312 K due to passivation by ZnS. Similarly, blue CdZnS/ZnS QDs WGM lasers have been confirmed by Sun et al. (Fig. 6b and c)⁹⁸. CdSe NCs random laser is also a common laser, which is caused by the repeated scattering of light in the gain medium⁶⁵. CdSe NCs random lasing usually consists of multiple peaks with different heights and positions. Although random lasers are efficient and inexpensive light sources, the poor quality and lack of directionality of the output light are detrimental to PICs. An efficient single mode laser can be constructed by coupling CdSe NCs and gratings to form DFB lasers. For example, single- and two-photon-pumped single-mode lasers have been realized by Signorini et al. by combining CdS/Cd_0.5Zn_0.5S/ZnS QDs with periodic gratings (Fig. 6d)⁹⁹. Furthermore, ideal coherent single-mode lasing can be realized by confining CdSe NCs as laser gain materials to two highly reflective Bragg mirrors. As shown in Fig. 6e, CdSe/CdS-ZnO with FWHM 0.9-nm single-mode laser output is detected with the increase of pump intensity⁶³. The same as QDs, CdSe NRs and NPLs, lasers with a lower lasing threshold can also be realized by combining gain materials with the aforementioned optical microcavities (Fig. 6f and g)^{21,22,31,50,51,67,100,101}. Their laser characterization, cavity type, pump-light type, gain material, and other properties are summarized in Table 1. Combined with the advantages of easy preparation, low cost, solution treatment, and low substrate requirements, CdSe NCs can be used as potential candidates for on-chip light sources.

As discussed in Section 2.2, considerable progress has been achieved by CdSe NCs in the field of ASE and lasers since the first demonstration of NCs optical gain. At the same time, CdSe NCs have also attracted special attention due to their potential to be used as light sources for PICs. According to the timeline, the on-chip laser formed by the integration of CdSe NCs and microcavities has undergone a transition from the uncontrollable stage to the controllable one³. At the early stage of on-chip laser-based NCs, Min et al. fabricated an on-chip WGM laser by covering CdSe/ZnS QDs on a ring-shaped fiber¹⁰⁵. The pump pulse was efficiently coupled to the ring resonator through the tapered fiber (Fig. 8a), thus demonstrating the extremely low on-chip lasing threshold (9.9 fJ)¹⁰⁵. However, complexity of the fibers for the output laser is not conducive to the realization of PICs. Therefore, the integration of microcavity and gain medium on the same substrate is beneficial for PICs. In the work of Wang et al., a WGM laser was confirmed by covering Cd_(1−x)Zn_xSe_(1−y)S_y/ZnS QDs to hemispherical resonator. The hemispherical resonator can be well mounted on the substrate to simplify the fabrication of the on-chip laser (Fig. 8b)⁶⁴. In the construction of on-chip lasers, CdSe NCs were adopted as gain materials and resonators, which not only simplify the construction of on-chip light sources, but also reduce the cost of resonator construction. For example, CdSe/CdS QDs are drop-cast to form on-chip coffee rings as not only gain media, but also laser resonators. This coffee-ring effect can occur in CdSe QDs, NRs, NPLs, etc^66,108. When the NC coffee ring is pumped, both sides can be used as an FP cavity to realize on-chip laser. In the work of Zavelani-Rossi et al., on-chip single mode lasers based on the coffee-ring effect of CdSe/CdS NIRs have also been demonstrated⁶⁶. However, some disadvantages of coffee rings must be taken into consideration, such as their uncontrollable size and position. Therefore, some high-precision control of the position, size, and laser peak position of on-chip NCs light sources have been realized in the recent years²⁶. For example, Xie et al. designed a SiN/QDs/SiN (100/55/100 nm) planar sandwich-stack structure by e-beam lithography (EBL) etching method. A 5-μm-diameter micro-stage combined with CdSe/CdS QDs exhibited a low threshold fluence of 27 μJ/cm² WGM laser (Fig. 8c and d) under picosecond pulse pumping¹⁰⁶. Subsequently, the transverse electric (TE)- and transverse magnetic (TM)-mode lasing of the QDs micro-stage was confirmed. In addition, single-mode lasers can also be observed by reducing the diameter of the microstage¹⁰⁶. Similarly, DFB resonators in SiN/QDs/SiN plane waveguide were fabricated by their group with the adoption of the same etching method, as shown in Fig. 8e-h ¹⁰⁷. Single-mode lasing can be achieved by this SiN/QDs/SiN microstage at 270 μJ/cm² nanosecond-pulsed pumping (Fig. 8i). Such single-mode lasers show more potential in PICs due to fewer modes and higher optical purity. Moreover, these SiN/QDs/SiN sandwich structures with good compatibility with CMOS compatibility confirm that NCs are potential candidates for PICs⁶. However, some shortcomings still remain to be addressed. For example, the complex construction of these SiN/QDs/SiN and the fluorescence quenching caused by the high-temperature processing of NCs are not favorable for their future development. In addition, it should be noted that the current pump sources for on-chip lasers are all based on short-pulse lasers. More efforts are required to realize electrically pumped PICs so as to meet the current needs of high-speed information transmission.

To further realize NCs PICs, on-chip integration of lasers and waveguides will be introduced. In general, on-chip integration of NCs-based active photonic devices and passive photonic devices based on other materials greatly increases the difficulty of fabricating PICs. Nevertheless, some noble progress in PICs based on NCs is worthy to be mentioned. For example, micro-stage multimode and single-mode lasers based on QDs have been demonstrated by Chen et al. adopting a simple and inexpensive approach¹⁰⁹. The micro-stage (40 × 40 μm) was combined with a waveguide strip with a length of 150 μm and a width of 10 μm, as shown in Fig. 9a. When the pump pulse excites the micro-stage, the multimode laser spectra at P₁ and P₂ are collected as shown in Fig. 9b. This similar spectrum indicates that the laser propagates from the micro-stage P₁ to the waveguide P₂. Furthermore, no lasing was detected when the waveguide was pumped, confirming that the previous P₂ lasing is originated from the micro-stage. However, the position, size, and shape of the on-chip QDs laser and waveguide are uncontrollable. In addition, this waveguide composed of CdSe/ZnS QDs shows a large absorption loss because since the laser comes from the same NCs. In their other experiments, small-sized CdSe/ZnS QDs (green) as waveguides and large-sized CdSe/ZnSe/ZnS QDs as high-quality lasers were coupled on glass by pattern-assisted stacking to overcome absorption loss, as shown in Fig. 9c. Here, the lasing peak of CdSe/ZnSe/ZnS was found at 640 nm pumped by 532 nm ps pulses. As shown in Fig. 9d, the same lasing spectra obtained from A and B confirm that the micro-stage lasing from the red QDs can be coupled to the green waveguide. However, the disadvantage of this structure is that the laser and the waveguide are not on the same substrate, which results in an increase in the volume of the PICs and subsequent complex integration.

Position-, size-, and shape-controllable waveguides and lasers are confirmed by the Jeon's group by integrating lasers and SiN planar waveguides on a chip¹¹¹. The schematic structure of these PICs is shown in Fig. 9e, where, Si₃N₄ and SiO₂ were used as the waveguide framework and substrate, respectively. Square lattices were constructed on 140-nm-thick Si₃N₄ by efficient EBL. CdSe/CdS/ZnS QDs were then selectively covered in the square lattice to form an on-chip laser. No etched Si₃N₄ acts as a waveguide due to the low absorption loss of this structure in the visible to NIR. A grating coupler was placed at the other end of the waveguide. Therefore, the waveguide propagates the laser mode only in the vertical direction. When the QD-covered resonator was pumped, the single-mode laser was then coupled into the waveguide. Due to the presence of the grating coupler and the special thickness of the waveguide, the final laser was output in the vertical direction. This structure overcomes the formerly mentioned shortcomings that the position, size, and shape of the laser and the waveguide are not easy to be controled. More importantly, single-mode lasers were easily formed and efficiently coupled into the waveguides. However, the lack of lateral confinement of this planar waveguide may contribute to some crosstalk problems in future applications. Fortunately, this problem was solved in the same year by Dries Van's group¹⁰⁶. As mentioned earlier, SiN/QDs/SiN sandwich structures have been placed on SiO₂ substrates to realize responsive WGM, and DFB lasers have been mentioned. In the current work, the on-chip laser and waveguide coupling were realized by this group. The construction and layout of these PICs is shown in Fig. 9f. In this structure, the structural parameters are carefully chosen to support TE and TM propagation. In addition to the coupling of dielectric waveguides and lasers, the coupling of plasmonic waveguides and lasers has also been demonstrated owing to the simple construction steps and low damage to the gain material. Specifically, Rong et al. adopted high-resolution dark-field optical imaging technology to integrate single or multiple microcavity lasers and silver nanowire waveguides on a chip, including tangential coupling, radial coupling and complex coupling¹¹². As shown in Fig. 10a, the silver wire waveguide is placed in the tangential direction of the QDs micro-ring resonator. When the ring resonator is pumped, the emission spectra of the B and C ends of the silver wire are consistent with those of A, indicating that the laser is effectively coupled into the silver-wire waveguide (Fig. 10b). In addition, radial and tangential coupling of the microcavity laser to single or multiple silver-wire waveguides has also been demonstrated, as shown in the dark-field optical imaging diagram in Fig. 10c. Under picosecond laser pumping, the end of the silver-wire waveguide and the consistent spectrum of the microcavity laser indicate that the laser is coupled into multiple silver-wire waveguides. In addition, the coupling of multiple microcavity lasers to a silver-wire waveguide has also been demonstrated (Fig. 10d)¹¹². This ease of operation is of great significance for realizing high-density assemblies of on-chip lasers, waveguides and functional devices.

Although the aforementioned studies show that the absorption and propagation losses of NC waveguides can be overcome by unique methods, potential losses still exist. Therefore, optical amplifiers are also crucial for NC PICs, especially for high-energy excitation of on-chip functional devices. The on-chip optical amplifier can directly amplify the optical signal without converting it to an electrical signal. Based on this situation, some designs of NC optical amplifiers have been proposed. The basic principle of the currently reported optical amplifier is that the optical signal generated by the on-chip laser propagates through the waveguide and is coupled to the optical amplifier to achieve signal amplification. In order to achieve optical signal amplification, high optical gain of NCs is a prerequisite. In 2017, Noriss et al. coupled NCs with surface plasmon systems to build on-chip lasers and optical amplifiers¹¹³. Specifically, two Ag reflective blocks with widths of 400 and 600 nm were extracted from the ultra-smooth silver surface to form a resonant cavity. Subsequently, a tapered waveguide was constructed by the template-stripping method, as shown in Fig. 11a. On-chip lasing was observed when CdSe/CdS/ZnS QDs covering the resonator were pumped by femtosecond pulses (Fig. 11b). After passing through the tapered waveguide, the on-chip laser is focused at the tip (Fig. 11c). Spectral characterization confirmed that the optical signal is extracted, propagated, and focused from the microcavity to the tip, as shown in Fig. 11d. However, the optical signal is not amplified due to propagation loss. When CdSe/CdS/ZnS QDs are simultaneously covered in the microcavity and the tapered waveguide, the on-chip laser can be amplified. The laser intensity of the waveguide with QDs covered is 1000-times higher than that of the waveguide without QDs, as shown in Fig. 11e and f. However, at low pump densities, the spontaneous emission noise is amplified by almost the same amount. In practical applications where pure signal amplification is required, the noise of these amplifications is undesirable. Recently, a noise-free optical amplifier has been proposed by Chen et al.¹¹⁴. Firstly, a micro-ring QD laser is constructed. Under picosecond pulse pumping, the on-chip laser is equally coupled to the left and right waveguides of different lengths (L_L = 15 μm, L_R = 40 μm) for signal transmission and optical signal amplification (Fig. 11g). Under small spot (17 μm) pumping, the laser intensity on the left is 1.3-times that of the right since part of the waveguide at the right end is not pumped (Fig. 11h). However, when the light spot size is increased to 80 μm, the light signal intensity on the right is 1.8-times that of the left (Fig. 11i). Importantly, the stimulated emission background signals on the left and right sides are the same sicne the laser signal is well coupled into the waveguide, while the stimulated emission is undirected. At the same time, this low-noise optical amplifier also prevents ASE generation along the NC waveguides.

According to the aforementioned introduction, it can be known that PICs include light sources, waveguides, amplifiers, filters, detectors, and other devices. However, when these photonic devices are composed of different materials, the complexity, compatibility, and uncontrollability of photonic chips must be taken into consideration. Although many high-quality microcavity lasers and waveguides have been demonstrated and can be integrated on a chip as mentioned earlier, the integration of multifunctional devices, lasers, and waveguides remains a challenge. To solve the difficulty of integrating multiple functional devices on a chip, Rong et al. adopted the same QD material to construct multiple photonic devices on a chip by pattern-assisted stacking¹¹⁴. Firstly, a high-precision trench pattern is etched in the polymethyl methacrylate (PMMA) film with MgF₂ as the substrate. Subsequently, an appropriate QD solution is drop cast into the trenches to form on-chip lasers, waveguides, optical amplifiers, Y-splitters, and Mach-Zehnder interferometers, as shown in Fig. 12a-d. The same material is used to overcome the difficulty of building multiple photonic devices on a single chip. By combining the bottom-up and top-down approaches, the damage to the QDs could be negligible. However, absorption losses resulting from the use of the same material need to be considered. Although the current research on the integration of semiconductor microcavity lasers and functional devices is scarce, the continuous development of NC lasers and excellent photonic devices will promote the development of future photonic chips.