Si-based photonic integrated circuits (PICs) have been receiving extensive attention in optoelectronics, due to their ability to achieve high functionality and low-cost chips through mature CMOS processes [1]. An epitaxially-grown III-V light source is a desired building block for compact, scalable Si-based PICs [2,3]. Tunable on-chip lasers are of particular interest due to their potential applications in wavelength-division multiplexing (WDM) and realizing broadband PICs [4–6]. However, choices in up-to-date wide-range emitters are still very limited, especially in monolithic and compact near-infrared laser sources [7].

One strategy to control the wavelength of III-V lasers is to design laser cavities for precise mode selection. In one approach, the laser cavity’s optical feedback comes from periodically arranged structures in the active region (distributed feedback (DFB) lasers), and the lasing wavelength is adjusted by tailoring the DFB laser parameters [5,8,9]. In another approach, the emission peak can be precisely manipulated by defining photonic crystal cavities [4]. Other tuning schemes based on selecting resonance modes include integrating double-ring resonators [10] and distributed Bragg reflectors (DBRs) [11] into the laser cavities. Although the lasing peak is precisely controlled under these schemes, the tunable range of the lasing wavelength is narrow (less than 70 nm). Moreover, the footprint and volume of these lasers are relatively large because of the introduction of external cavities. Another scheme to alter the lasing wavelength is to change the chemical composition or structure of the active region. For ternary III-V alloys such as InGaAs, the band gap can be changed from 0.36 eV to 1.4 eV by adjusting the indium composition. Wide-band lasing over 300 nm has been demonstrated in InGaAs/InGaP nano-beam lasers with different indium fractions [4]. Additionally, for III-V heterostructures such as InP/InGaAs, the emission peak also depends on the thickness of the InGaAs quantum well (QW) [12,13]. A broad lasing range of over 400 nm has recently been demonstrated in InP/InAs nanowire lasers with different InAs QW thicknesses [14]. However, tuning the lasing peak of this type of laser source requires different active regions and thus requires multiple epitaxy processes. Therefore, achieving a widely tunable telecom-range nanolaser with a single epitaxial structure is still a challenge.

Recently, we discovered a wide-band lasing behavior from InP/InGaAs nano-ridge lasers with identical InP/InGaAs QW active regions [15,16]. By defining the InP/InGaAs nano-ridge laser cavity into different lengths, the emission peak of the nano-ridge lasers spans over 200 nm from the O-band to the C-band. However, the origin of this behavior is still unclear. In this paper, we present a study of the relationship between the lasing wavelength and the nano-ridge laser cavity length, and unveil the underlying mechanism through extensive mode and gain analyses.

Figure 1(a) and (b) show the structural schematic and scanning electron microscopy (SEM) image of an InP/InGaAs nano-ridge laser array directly grown on (001) silicon-on-insulator (SOI) wafers using metal organic chemical vapor deposition. The growth and fabrication methods are detailed in our earlier works [15,17,18]. Five uniform and symmetric InGaAs QWs were embedded inside the InP nano-ridge as the active gain medium (see the inset of Fig. 1(b)). The average thickness of the InGaAs QWs was measured as 6 nm using cross-sectional transmission electron microscopy. The length of the nano-ridge lasers, defined using focused-ion beam (FIB) milling, varied from 25 µm to 80 µm. Room-temperature lasing was observed from the nano-ridges under optical pumping conditions. The nano-ridge laser features a broad spontaneous emission with evenly spaced Fabry-Pérot (FP) resonance peaks below the lasing threshold, and a strong peak at 1494 nm above the threshold (see Fig. 1(c)). Additionally, most of the measured nano-ridge lasers exhibited clean lasing spectra together with low lasing thresholds, suggesting that only a single transverse mode is amplified in our nano-ridge lasers.

 

Fig. 1. (a) Schematic of the cross-section of the designed InP/InGaAs nano-ridge laser array grown on SOI substrate. (b) 70° tilted SEM image of nano-ridge lasers on SOI; the inset shows the zoomed-in image on the cross-section of the nano-ridge laser. (c) Room-temperature emission spectra of the 60 µm nano-ridge laser below (blue, magnified ×20000) and above (red) the lasing threshold (41.8 µJ/cm²).

Download Full Size | PDF

A systematic measurement of the lasing wavelength with respect to the nano-cavity length is summarized in Fig. 2. An S-shaped wavelength versus cavity-length relationship is shown, with a turning point at 45 µm. Surprisingly, with identical InGaAs QWs as the active region, the lasing wavelengths can cover a range of over 200 nm from the O-band to the S-band, which is unexpected and approximately an order of magnitude greater than the tunable range in traditional InGaAs lasers [19]. Generally, a shorter lasing wavelength can be achieved via tailoring the nano-ridge laser to a shorter length. For the lasers shorter than 35 µm (Region A) and longer than 55 µm (Region C), lasing only arose at around 1350 nm and around 1500 nm, respectively. Both the long and short cavity length regions display a quasi-linear relationship with the lasing wavelength. It is noteworthy that a broad range of wavelengths from 1350 nm to 1500 nm can be supported at moderate lengths of about 40 µm to 50 µm (Region B). This unique phenomenon of wavelength dependence in the telecom region demonstrates broadband emission in a single epitaxial structure, which has yet to be observed in other lasers.

 

Fig. 2. Relationship between lasing wavelength and nano-ridge laser cavity length.

Download Full Size | PDF

To understand the origin of the broadband lasing behavior, it is crucial to identify the lasing mode and study the modal reflectivity and confinement factors. Finite-difference eigenmode (FDE) and Finite-difference time-domain (FDTD) simulations were performed to investigate the mode characteristics of the nano-cavities. The first few modes are modeled at different mode wavelengths. The modes are guided in an infinitely long cavity with one end facet. Their group indices, confinement factors, and modal reflectivity are shown in Fig. 3. Note that the confinement factor is defined here as the proportion of electric field strength in the QWs to that of the whole transverse plane, with the energy velocity taken into account [20,21]. Interestingly, the TE₀₁ mode possesses a remarkably high mode confinement factor in the active region and a large end-facet reflectivity across the whole gain spectrum, as illustrated in Figs. 3(b) and 3(c). These results suggest that the TE₀₁ mode is most likely the dominant lasing mode in our nano-ridge lasers because of its uniquely strong feedback and high energy gain. It is worth noting that the fundamental modes HE_11a and HE_11b can also be supported, but they experience a significantly lower confinement and reflectivity, and are therefore unlikely to lase. Furthermore, the group index n_g = 4.36 is obtained at a wavelength of 1475 nm, according to the measurements of the free spectral range (FSR) below the lasing threshold (see Fig. 1(b)). This value corresponds well to the results in the FDE simulation as shown in Fig. 3(a), indicating that the observed mode is indeed the TE₀₁ mode. We attribute the minute deviations in mode group index to the slight unevenness in size and shape of the supporting Si pedestals. Hence, we are confident that the TE₀₁ mode is the only lasing mode supported in our lasers, regardless of the emission wavelength.

 

Fig. 3. (a) FDE simulation of the group index ${\textrm{n}_\textrm{g}}$ of the first three eigenmodes in the nano-ridges at different mode wavelengths; the inset illustrates the profiles of the HE_11a, HE_11b and TE₀₁ modes at 1400 nm. (b) Confinement factors of the modes in QW active layers at different mode wavelengths. (c) FDTD simulation of modal reflectivity at different mode wavelengths of the same three modes.

Download Full Size | PDF

Lasing occurs when loss equates to gain (${g_m} = {\alpha _T}$). Since the well-defined structure of the nano-ridges leads to strong feedback from the end facets, the nano-ridges can be treated as excellent FP resonators. The loss in FP lasers is mainly determined by mirror loss at the end facets, and intrinsic loss from defect sites. The total loss ${\alpha _T}$ in our lasers, hence, can be described as

 

Fig. 4. (a) Relationship between the pumping threshold ${\textrm{F}_{\textrm{th}}}$ and $1/\textrm{length}$ of the nano-ridge lasers. The linear relation is fitted using Eq. (1) according to the fact that the loss of the lasers ${\alpha _T}$ is directly proportional to the lasing threshold ${\textrm{F}_{\textrm{th}}}$. (b) Modal gain spectra of the InP/InGaAs QW active layer under different carrier densities. The carrier densities are indicated on the spectra in the unit of $\textrm{c}{\textrm{m}^{ - 3}}$.

Download Full Size | PDF

In addition, the wavelength-dependent peak modal gain from the QW active layer also affects the lasing peak. The modal gain spectrum of the TE₀₁ mode in the InP/InGaAs nano-ridge lasers is calculated according to

The modal gain spectra are calculated from the radiative transition model constructed according to Eqs. (2) and (3). Figure 4(b) displays the modal gain spectra of the InP/InGaAs QW, with a thickness of 6 nm, under various carrier densities. Owing to the consistently high confinement factors of the TE₀₁ mode at different wavelengths, the modal gain spectrum shares a similar shape with the material gain spectrum. The modal gain spectra indicate that there is only a slight blue-shift at the relative low carrier densities ($\textrm{n} \le 1.4 \times {10^{19}}\; \textrm{cm}^{ - 3}$), but at moderate to high carrier densities, the gain peak dramatically shifts to a shorter wavelength. The redistribution of energy states resulting from the quantum confinement effect forms discrete energy sub-bands in the QWs. With the reduced density of states in each sub-band, the band filling effect can be drastically enhanced [23]. As a result, the transitions between different energy sub-bands can bring the multiple peaks feature to the gain spectrum under reasonably low pumping levels (i.e. $\textrm{n}\;\sim \; 2 \times {10^{19}}\; \textrm{cm}^{ - 3}$). By slightly increasing the carrier density in the QWs, the peak gain position shifts from around 1460 nm to 1360 nm. This suggests the possibility of adjusting emission wavelength according to carrier density in nano-ridge lasers with low-dimensional active layers, such as the InP/InGaAs QWs described here. Rearranging the energy band structure in QW is crucial in gaining access to higher transition energy channels and thus covering a wider wavelength range, which bulk nanolasers cannot achieve.

To predict the lasing behavior of the nano-ridge lasers with various cavity lengths, the relationship between the emission wavelength and the length of the nano-ridge lasers is theoretically modeled. Figure 5 illustrates the normalized gain spectrum as a function of nano-ridge laser length. Since the length of the laser is implicitly included in the expression of the pumping threshold (see Fig. 4(a)), by extracting the relation between the carrier density and the peak wavelength from the gain spectra in Fig. 4(b), the peak emission wavelength and the laser cavity length can be correlated using

 

Fig. 5. Contour plot of the normalized modal gain spectrum of InP/InGaAs QW at 6 nm thickness.

Download Full Size | PDF

Although a wide wavelength range of over 150 nm was expected in our calculations (assuming 90% or higher of the peak gain could compensate for the mirror loss, and thus lase), the experimentally observed emission range over 200 nm was still beyond our anticipation. We attribute the unexpectedly wide tuning range of the nano-ridge lasers to the inhomogeneity of the QW active layers. The slight variations in QW thickness from the epitaxy process induce additional gain spectrum broadening, resulting in a larger lasing wavelength range than expected. The wider distribution of transition energies from the QW thickness discrepancies also extends the overall lasing band. Hence, the emission range is outstretched and allows the lasing wavelength to span across the O-band to the S-band.

In conclusion, we have demonstrated a broadband lasing behavior via controlling the cavity length of InP/InGaAs nano-ridge lasers directly grown on SOI wafers and explained the underlying mechanism through detailed mode and gain calculations. The TE₀₁ mode was identified to be the lasing mode due to the large confinement factor and strong feedback from the end facets. By simulating the modal gain and mirror loss of the TE₀₁ mode, we predicted an S-shaped relationship between the emission peak and the cavity length that agrees well with our experimental data. This work suggests an interesting scheme for realizing broadband emission in laser sources with low-dimensional active regions, and provides a new perspective on constructing wide-range on-chip infrared emitters for data-communication applications.