Embedding semiconductor lasers with Bragg gratings as wavelength-selective feedback mechanisms is a well-established approach to achieving high-quality single-frequency lasing. In conjunction with the distinctive properties of various compound semiconductor gain materials, distributed feedback (DFB) and distributed Bragg reflector (DBR) lasers are finding a wide range of applications in both classical and quantum domains, such as with InGaAs emitting in the near infrared for optical communication [1], with GaAs or GaAsP in the red spectral range [2] for atomic clocks [3], atom interferometry [4], and efficient optical pumping [5], with GaSb or InAs/AlSb in the middle to far infrared [6] for trace-gas sensing [7], and with nitride semiconductors in the green to ultraviolet for absorption spectroscopy [8] and high-density data storage [9].

For optical data interconnect applications, the 1310 nm band is of particular interest for low-cost wavelength division multiplexing (WDM) systems [10], 5G and 6G optical networks [11], as well as for LiDAR [12,13] and sensing [14,15]. Compared with conventional InGaAs/InGaAlAs quantum well (QW) materials emitting in the same wavelength range, self-assembled InAs/GaAs quantum dots (QDs) have achieved superior performance, including lower threshold currents [16] and higher-temperature stability [17] due to the zero-dimensional (0D) carrier confinement in QDs. A highly desirable feature of QD lasers is their ultralow relative intensity noise (RIN) [18–20] originating from their very low linewidth enhancement factor $\alpha$, which affords great advantages in analog transmission [such as radio over fiber (RoF)] and sensing [21]. The low $\alpha$ also underpins their tolerance to high levels of external optical feedback [22–24], making QD lasers very promising light sources for reliable, scalable, and isolator-free photonic integrated circuits (PICs) [25]. Furthermore, InAs/GaAs QD materials have also been proved to be highly tolerant to epitaxial defects [26–28], yielding high-performance Fabry–Perot (FP)-type laser diodes epi-grown on silicon substrate [28–30]. Similar advantages have also been observed in other QD laser systems such as InAs/InP [31,32].

For DFB lasers, the buried Bragg gratings, as first realized in InP-based 1550 nm lasers, are conventionally placed on top of the active layer and fabricated through a regrowth process after grating definition and etching. The proximity to the waveguide allows the grating to intercept the optical field with a high coupling coefficient $\kappa$. This conventional approach has also been attempted in other materials and wavelengths such as GaAs-based QD lasers. In 2011, Tanaka et al. demonstrated a 1293 nm InAs/GaAs QD DFB laser with a GaAs grating buried in a metal organic vapor phase epitaxy (MOVPE) regrown InGaP cladding, achieving a $\kappa$ of $40{\mathrm{cm}}^{-1}$ and sidemode suppression ratio (SMSR) of 45 dB [33]. In 2020, Wan et al. demonstrated a 1310 nm QD DFB laser epitaxially on silicon by molecular beam epitaxy (MBE) regrowth. Using a GaAs grating buried in an ${\mathrm{Al}}_{0.4}{\mathrm{Ga}}_{0.6}\mathrm{As}$ upper cladding, a $\kappa$ of $45{\mathrm{cm}}^{-1}$ and an SMSR of $>50\mathrm{dB}$ were achieved [34].

However, the regrowth process represents poor productivity for many material systems. For GaAs-based laser devices, an InGaP upper cladding layer requires wafer transfer from MBE to MOCVD systems, while an AlGaAs upper cladding layer requires rigorous pretreatment before regrowth via an ultrahigh-vacuum MBE chamber. Harder still, the regrowth process for GaN(Sb)-based lasers suffers from lack of a contamination-free AlGaN(Sb) regrowth process or other latticed-matched cladding materials with sufficient bandgap and refractive index contrasts.

An alternative, regrowth-free approach uses Bragg gratings etched alongside a ridge waveguide to form a laterally coupled DFB (LC-DFB) laser. For InP-based laser structures, the gratings can be fabricated simultaneously during the waveguide etching process. Using an aluminum-containing stop-etch layer and a chemically selective recipe [35,36], the grating penetration depth can be precisely delimited to just above the active layer to achieve a precise $\kappa$ value. However, translating this approach to GaAs- or GaSb-based DFB laser structures has proven to be challenging due to the lack of a suitable selective etch-stop layer. Previously demonstrated GaAs [37], GaSb [38], or GaN [39] ridge waveguides with LC gratings fabricated using such a one-step reactive ion etching (RIE) approach suffer from the fact that, due to local chemical transportation and reaction rate variations caused by the etched ridge and the very narrow grating gaps, it is very difficult to control the etch depth at the foot of the ridge waveguide where the grating intercepts the optical mode. An undesirable feature known as “footing,” which is a gradual increase in etch depth away from the foot of the etched waveguide, and another feature known as “RIE-lag,” which is a decreased etch depth in narrow gaps, result in significant uncertainties in the grating $\kappa$ value.

To circumvent this problem, in a previous work, the authors chose to etch the lateral grating deeply through the active region so that the grating etch depth no longer affects $\kappa$ [40]. However, a deep-etched active waveguide suffers from increased surface recombination and optical scattering loss, and a waveguide with practical widths can support more than one transverse mode, with the unwanted high-order modes as favored lasing modes due to their higher $\kappa$ values [40]. For GaSb lasers, metal gratings deposited after waveguide etching [41] were also used. While providing strong optical coupling, metal gratings can introduce significant additional absorption loss in the laser cavity.

In this paper, we demonstrate a novel dielectric grating structure placed alongside single transverse mode ridge waveguides that have a precisely controlled trapezoid cross-sectional profile etched to a depth just above the active layer. Fabricated in an amorphous silicon ($\alpha$-Si) layer deposited after the formation of the ridge waveguide, the grating corrugations, plasma-etched into the $\alpha$-Si, are precisely stopped at an underlying etch stop layer of ${\mathrm{Al}}_2{\mathrm{O}}_3$ deposited after the waveguide etching and before the $\alpha$-Si layer. A high-contrast grating (with a refractive index difference of $\mathrm{\Delta }n\sim 2$) is formed between the $\alpha$-Si corrugations and a subsequently deposited silicon dioxide (${\mathrm{SiO}}_2$) cladding layer, producing an LC grating with significantly enhanced and precisely controllable coupling coefficient $\kappa$.

We implemented the novel structure on an InAs/GaAs QD gain material, producing LC-DFB laser arrays emitting across the 1300 nm band on a 0.8 THz local area network wavelength division multiplexing (LWDM) grid. The devices emit more than 26.6 mW of single-mode output power at room temperature and a typical SMSR greater than 52.7 dB. They also demonstrate ultralow RIN of $<-165\mathrm{dB}/\mathrm{Hz}$ in the range of 2.5–20 GHz and isolator-free operation under external feedback levels of up to $-12.3\mathrm{dB}$ (5.9%). The output power, SMSR, and RIN values are the best of the reported values of InAs/GaAs QD LC-DFB lasers, as far as we are aware of. These superior performances of the devices validate the novel LC grating as an effective regrowth-free approach to grating-based laser fabrication. In addition to the elimination of regrowth, the deployment of the novel LC grating structure decouples its fabrication from specific laser materials, and therefore the scheme could serve as a universal alternative approach for high-performance semiconductor laser devices employing grating structures as optical feedback mechanisms.

The InAs/GaAs QD laser structure, in a typical p-i-n configuration [Fig. 1(a)], was grown on 3-inch (1 inch = 2.54 cm) semi-insulating GaAs (001) substrates in a solid-source MBE chamber. First, a 500 nm Si-doped GaAs buffer layer and a 1.8 μm ${\mathrm{Al}}_{0.4}{\mathrm{Ga}}_{0.6}\mathrm{As}$ cladding layer were grown, followed by an active region that contains a five-layer InAs/GaAs QD separated by 35 nm GaAs barriers. Each QD layer comprises 2.4 ML InAs covered with a 3.5 nm ${\mathrm{In}}_{0.15}{\mathrm{Ga}}_{0.85}\mathrm{As}$ strain-reducing layer. Modulation p-doping with Be was implemented in a 6 nm GaAs layer located 10 nm beneath each QD layer to obtain a concentration of 20 acceptors per dot. Afterwards, 1.8 μm Be-doped ${\mathrm{Al}}_{0.4}{\mathrm{Ga}}_{0.6}\mathrm{As}$ and 100 nm GaAs layers were grown as p-cladding and p-contact layers, respectively. The InAs/GaAs QD gain materials with a density of $5.5\times {10}^{10}{\mathrm{cm}}^{-2}$ were achieved, as indicated by the $1\mathrm{\mu m}\times 1\mathrm{\mu m}$ atomic force microscope (AFM) inset of Fig. 1(b) from an uncapped QD sample grown on a GaAs substrate under the same QD growth conditions. Room-temperature photoluminescence (PL) emission peaking at 1308 nm was observed with a narrow full-width at half-maximum (FWHM) of 30.9 meV. Additionally, large quantized-energy separation of 80 meV between the ground-state and the first exited state effectively suppressed carrier overflow at high temperatures. Traditional multi-mode ridge waveguide FP lasers (20 μm ridge width) with cleaved facets were fabricated to characterize the properties of the QD materials. Light–current–voltage ($L–I–V$) characteristics of the fabricated laser with a length of 2000 μm and its temperature dependence under a continuous-wave (CW) condition are shown in Fig. 1(c). The threshold currents are as low as 80 mA ($200{\text{A/}\mathrm{cm}}^2$) at 25°C. Reduced temperature sensitivity is achieved with the characteristic temperature $T_0$ as high as 98 K in the range of 25°C–55°C and 53 K in the range of 65°C–115°C. Since the measurements were carried out under CW mode, the extracted $T_0$ should be an underestimate of the true value due to junction heating.

The as-grown wafers were then processed into LC-DFB laser arrays. As shown in Fig. 2(a), a waveguide width of 2.1 μm and a depth of 1.7 μm are used to support only the lowest-order transverse mode (TE00). The ridge waveguides were patterned using electron beam lithography (EBL) and etched using an optimized chlorine-based GaAs inductively coupled plasma reactive-ion etching (ICP-RIE) process, with a trapezoid cross section, sidewall slope of $\theta =76°$, and near-zero footing [Fig. 2(b)]. The near-ideal trapezoid waveguide profile is key to a deterministic grating coupling coefficient $\kappa$ and low scattering loss, both very important for improved DFB laser performance. In addition, a small footing gives rise to a larger $\kappa$, which results from the increased evanescent field into the grating region [42].

A 10 nm ${\mathrm{Al}}_2{\mathrm{O}}_3$ passivation layer and a 150 nm thick $\alpha$-Si ($n\sim 3.495$) layer were deposited by atomic layer deposition (ALD) and ICP-CVD, respectively, with both layers covering the entire sample surface terrain, including the sloped sidewalls. First-order gratings with a period $\mathrm{\Lambda }$ in a range of 194.5–199.7 nm, grating duty cycle of 1:1, and extrusion of 8 μm from the waveguide foot were designed and patterned by an EBL resist (ARP6200) alongside the ridge. The grating duty cycle is defined as the ratio of the grating width to the groove width. A $\lambda /4$ phase shift was placed in the middle of the gratings to force lasing in the defect mode. LWDM-compatible laser arrays were achieved by adjusting the grating period $\mathrm{\Lambda }$ of adjacent lasers on the same bar, with a change of $\mathrm{\Delta }\mathrm{\Lambda }=0.727\mathrm{nm}$ resulting in a wavelength increment of 0.8 THz. The grating corrugations were etched through the $\alpha$-Si using a fluoride-based RIE process. The SEM image of Fig. 2(c) reveals a high-quality LC grating with very sharp and smooth sidewalls. With near-zero footing and a naturally chemically selective etch recipe, this dielectric grating structure is less sensitive to processing variations and, thus, more manufacturable than a buried heterostructure DFB structure. Grating coupling coefficient $\kappa$ is calculated from the lateral electric field distribution and effective index of the fundamental transverse electric mode via coupled-mode theory [42–44], detailed in Appendix A. For the designed devices, a first-order grating with a duty cycle of 1:1 produces a calculated $\kappa$ of $3.24{\mathrm{mm}}^{-1}$.

The corrugations were subsequently covered with a 200 nm layer of ${\mathrm{SiO}}_2$ ($n\sim 1.46$) to form a high-contrast grating prior to contact window opening and Ti/Pt/Au p-type contact deposition [Fig. 2(d)]. An AuGe/Ni/Au n-type contact was deposited in the back after thinning the GaAs substrate down to 200 μm. After being cleaved into bars, the two facets were covered with a six-layer ${\mathrm{SiO}}_2/{\mathrm{TiO}}_2$ high-reflection (HR, 96.8%) coating and a one-pair ${\mathrm{SiO}}_2/{\mathrm{TiO}}_2$ anti-reflection (AR, 1.7%) coating to suppress the facet backreflections and increase output power. Finally, the laser arrays were mounted epitaxy-side-up on gold-coated copper heat sinks.

At room temperature (25°C), a typical $2.1\mathrm{\mu m}\times 1.5\mathrm{mm}$ DFB laser device [Fig. 3(a)] has a turn-on voltage of 0.93 V and a differential series resistance of $10\mathrm{\Omega }$. The measured CW threshold current of 18 mA corresponds to a current density of $571\mathrm{A}{\mathrm{cm}}^{-2}$. Above the threshold, the output power follows a near-linear curve with a slope efficiency of $0.2\mathrm{W}{\mathrm{A}}^{-1}$. The AR facet output power of 26.6 mW was obtained at an injection current of 150 mA or $\sim 8.3\times I_{\mathrm{th}}$. Figure 3(b) shows CW lasing up to 55°C with output power of $>6\mathrm{mW}$, and it is believed that the operation temperature can be further increased by more effective junction heat dissipation, by either further thinning the substrate or flip-chip bonding.

The experimental value of $\kappa$ is estimated from the photonic bandgap width ${\lambda }_s=0.436\mathrm{nm}$ observable from the below-threshold amplified spontaneous emission (ASE) spectrum of Fig. 3(c). The total coupling strength $\kappa L$ for this 1.5 mm long device was estimated to be $\sim 3.0$ ($\kappa =2.0$), which is consistent with the theoretical value when taking into consideration the deviations of rating duty cycle (1:1.9 as shown in the inset of Fig. 2(c)]. When increasing the current to 80 mA ($4.4\times I_{\mathrm{th}}$), the central defect mode seen in the ASE spectrum rises to be the dominant longitudinal mode with an SMSR of 52.7 dB [Fig. 3(d)].

Stable single-mode operation was observed at CW currents up to 150 mA in the temperature range of 20°C–50°C, with a linear wavelength–temperature tuning rate of 0.12 nm/°C and a quadratic wavelength–current tuning curve [Figs. 4(a) and 4(b)]. It is noteworthy that the laser maintains single-mode operation across the entire current range. This high single-mode quality is credited to the novel $\alpha$-Si grating along the trapezoid waveguide, which affords reliable deterministic optical coupling.

Across each eight-device bar, channel spacing of $0.80\pm 0.10\mathrm{THz}$ was measured, producing wavelengths ranging from 1300.05 to 1332.41 nm [Figs. 4(c) and 4(d)], matching well with the standard LWDM grid. This wavelength range was limited by the range of grating periods we fabricated. The high-precision EBL process together with the broad gain bandwidths of QD materials affords very promising applications for both LWDM and coarse wavelength division multiplexing within the O-band.

The RIN across the frequency range of 2.5–20 GHz is assessed using a commercial system (SYCATUS A0010A), measuring $<-155\mathrm{dB}/\mathrm{Hz}$ at $4\times I_{\mathrm{th}}$ and reducing to a saturated minimum level of $\sim -165\mathrm{dB}/\mathrm{Hz}$ at $9\times I_{\mathrm{th}}$, as shown in Fig. 5. The relaxation oscillation (RO) of this device is suppressed due to the large damping factor of the QD as reported [45]. To the best of our knowledge, this ultralow RIN is the best-reported result among QD DFB lasers and is in good agreement with reported values in both GaAs [20] and silicon [19] based QD-FP lasers. We consider that the high output power decreases the proportion of spontaneous emission, while the suitable grating $\kappa$ value ensures single-mode operation without mode hopping or spatial hole burning at high injection currents. Both can effectively suppress the disturbance of photons and carriers and thus yield an ultralow RIN.

Finally, the performance of the QD LC-DFB lasers under coherent optical feedback was assessed using the optical measurement setups shown schematically in Fig. 6(a). The emission from the QD laser AR facet is coupled into the feedback test system by a lensed fiber with coupling efficiency of 20%–30%, and divided into feedback and detection paths by a 90/10 fiber coupler. On the feedback path (10% of the coupled power), a fibered optical circulator is used to feed the light back to the laser cavity. Since the external cavity resonance frequency (17.25 MHz in this stage) is much less than the laser RO frequency, the impact of the feedback phase is negligible. The feedback strength $r_{\mathrm{ext}}$ is defined as the ratio of the returning power to the laser free-space output power, and precise returning power can be obtained by calculating the product of the power (detected by a power meter) and the lensed fiber coupling efficiency. The optical feedback intensity is controlled by changing the operating current of a polarization-maintaining boost optical amplifier (BOA, Thorlabs S9FC1132P). A filter with 0.8 nm bandwidth is employed to spectrally suppress the ASE from the BOA. A polarization controller is inserted in the external cavity to compensate for the polarization rotation in the fiber. The insertion losses produced in the lens fiber, BOA, beam splitter, and each connector are carefully calibrated. The remaining 90% of the coupled power is sent to a high-resolution (0.03 nm) optical spectrum analyzer (OSA, Anritsu MS9740A) or RIN measurement system (SYCATUS A0010A) to monitor the evolution of spectra and RIN as the feedback strength $r_{\mathrm{ext}}$ varies. For the whole measurement, the DFB laser is mounted on a thermo-electric cooler (TEC) operated at 25°C.

Figures 6(b) and 6(c) present the evolution of SMSR and RIN with the laser operating at $4\times I_{\mathrm{th}}$. This device shows RO peaks at a low injection current, which move from 2 GHz to about 5.5 GHz with the increasing bias current [inset of Fig. 6(c)]. Little sign of deterioration is observed until the optical feedback levels reach $r_{\mathrm{ext}}=5.9\%$ ($-12.3\mathrm{dB}$). In particular, the SMSR of the laser is found to be still above 50 dB under this feedback strength. RIN levels (at the frequency of 5 GHz) rise only slowly until $r_{\mathrm{ext}}=5.6\%$ ($-12.5\mathrm{dB}$), without any visible periodic or chaotic oscillations in the RIN spectra. A sharp increase in RIN indicates a transition to the coherence collapse regime beyond this critical level of optical feedback $f_{\mathrm{ext},\mathrm{c}}$. For DFB lasers, $f_{\mathrm{ext},\mathrm{c}}$ is strongly associated with the normalized coupling coefficient $\kappa L$ ($L$ is the length of the laser cavity), where an increased $\kappa L$ will lead to an increased $f_{\mathrm{ext},\mathrm{c}}$ [46,47].

Given the rapid development of QD-DFB lasers at 1310 nm on both GaAs and silicon substrates, it is useful to compare the performance of our device with those reported in the literature. Table 1 lists the threshold current (current density), power, highest work temperature, SMSR, RIN, and optical feedback tolerance of reported DFB lasers together with our device. In general, compared with commercial QW lasers, QD lasers exhibit excellent performance in terms of high operating temperature, low RIN, and high optical feedback tolerance owing to their stronger carrier confinement, larger damping factor, and smaller $\alpha$ factor. QD lasers with buried [24,52] gratings show good feedback tolerance, but the RIN needs to be further improved. Our device simultaneously achieves high output power (26.6 mW), ultralow RIN ($-165\mathrm{dB}/\mathrm{Hz}$), and high tolerance to optical feedback ($-12.3\mathrm{dB}$).Table 1.

Comparison of the Performance of Our Device with Reference QD DFB Laser at 1310 nm

To conclude, by implementing a novel first-order $\alpha$-Si Bragg grating LC to a near-ideal trapezoid GaAs ridge waveguide, high-performance 1300 nm InAs/GaAs QD DFB laser arrays have been realized with high power, ultralow RIN, high robustness against optical feedback, and accurate LWDM grid. These excellent features make the device a very attractive candidate for high-performance digital and analog WDM optical transmission systems as well as on-chip sources for PICs where optical isolators are not readily available. In the near future, decreasing uniformities in the size or energy level of QDs while keeping the QD density [17], is expected to further improve the material modal gain and therefore thermal stability.

The novel grating structure affords accurate and deterministic grating coupling coefficient $\kappa$, which can be engineered independent of the laser material epitaxy process. The scheme can therefore be readily implemented on other material systems such as InP-, GaAs/Si- and GaSb-based compound semiconductor lasers, and in other wavelength windows by scaling the size of the grating and using low absorption dielectric materials for those wavelengths. The simplicity and versatility of the regrowth-free scheme make it possible to establish a new paradigm of semiconductor laser manufacturing. In this paradigm, multiple types of grating-based semiconductor lasers (including DFB, DBR, and tunable lasers) can be jointly manufactured on the same fabrication platform, with different active materials and operating wavelengths, which would potentially reduce their manufacturing cost very significantly.