Realizing telecom-band lasing in GaAs-based nanowires (NW) with low bandgap gain media has proven to be notoriously difficult due to the high compressive strain built up in the active regions. Here, we demonstrate an advanced coaxial GaAs-InGaAs multi-quantum well (MQW) nanowire laser that solves previous limitations by the introduction of a strain compensating InAlGaAs buffer layer between the GaAs core and the MQW active region. Using a buffer layer thickness comparable to the core diameter applies a significant tensile strain to the GaAs core which efficiently minimizes the compressive strain in the InGaAs MQW and enables large In-content without plastic relaxation. Experimental verification is shown for NW-lasers with an In-content of up to 40% in the MQW, evidencing a clear strain-relieved redshift of the lasing emission and a strong reduction of the lasing threshold compared to highly strained MQWs in state-of-the-art GaAs NW-lasers. This way we achieve optically pumped room temperature lasing operation with a threshold below 50 μJ cm−2 in the telecom O-band close to 1.3 μm.
III–V based semiconductor nanowire (NW) lasers hold considerable potential for telecom-band optical data communication and sensing applications due to their ultracompact size and unique properties.1–3 As single-mode Fabry–Pérot type resonator cavities with sub-μm2 footprint, NW-lasers offer routes for ultrahigh-density integration and low power consumption on silicon (Si) based photonic circuits.4 Although important performance metrics, such as ultrafast external modulation (200 GHz), large spontaneous emission coupling factors (∼0.2), as well as coupling of single, site-selective III–V-based NW-lasers to Si waveguides have been demonstrated,5–7 an outstanding challenge is to realize low-threshold lasing emission in the relevant telecom spectral bandwidth (1.3–1.55 μm). Most III–V NW-lasers reported to date emit in the near-infrared (NIR) range at wavelengths below the Si transparency region (<1.15 μm),6–14 where realizing high optical gain is comparatively easy. Recently, bandgap tunable InGaAs and InAsP NWs were proposed to push the lasing wavelength toward the telecom band; however, lasing required external cavities using one- or two-dimensional photonic crystal arrays.15,16 For single vertical-cavity NW-lasers, a more encouraging route exploits quantum-confined gain media [multi-quantum dots/disks (MQD) and quantum wells (QW)], with efforts focusing on the promising InP and GaAs NW platform. Successes were obtained for InP NWs tuning lasing emission across the O-, E-, and C- telecom-bands, using, e.g., near lattice-matched InGaAs/InP QWs radially grown around needle-shaped InP pillars (on Si),4 or axial InAs/InP multi-quantum disk (MQD) NW-lasers (on InP substrates).17 However, the site-selective growth of gold (Au)-free telecom-band InP-based NW-lasers, having well-defined resonator dimensions and good homogeneity, is very challenging on Si substrates.18 In comparison, the GaAs platform offers fairly mature Au-free NW growth on Si, with NW growth yield > 90%,19 accurate control of NW length distribution,20 and crystal phase,21 but it suffers from severe lattice-mismatch problems. GaAs NW-lasers with stacks of GaAsSb/GaAs13 and InGaAs/GaAs22 MQDs were reported, but tuning lasing wavelengths beyond 1 μm proved to be difficult due to the cumulative compressive strain in the stacks. Similarly, GaAs NW-lasers with coaxial InGaAs QWs were attempted but plastic strain relaxation with maximum attainable In-content of 30% for pseudomorphic QW growth limited laser operation to low temperature and emission wavelengths below 1.1 μm.23 In this Letter, we overcome these limitations by proposing and implementing a strain compensating In0.3Al0.3Ga0.4 As buffer layer in advanced coaxial GaAs/InGaAs multi-quantum well (MQW) NW-lasers. We demonstrate low-threshold, room temperature lasing at ∼1.3 μm.
Figure 1(a) shows a cross section of the proposed NW core-multishell heterostructure. The center consists of a GaAs core (diameter 106 nm) used for the vertical NW growth, which is surrounded by a shell (buffer) layer allowing good confinement of the TE01 guided optical mode.9,23 We investigated two different buffer layers, each having a thickness of 90 nm: (i) a conventional Al0.3Ga0.7As buffer lattice-matched to the GaAs core (no strain compensation)23 and (ii) a quaternary In0.3Al0.3Ga0.4As buffer highly strained to the GaAs core (strain compensation case). The consecutive MQW region consists of seven InxGa1-xAs QWs (8 nm) and Al0.3Ga0.7As or InAlGaAs barriers (8 nm), respectively, which are separated by 2 nm thick GaAs interlayers. Using a 2D freestanding strain model, we calculated the strain state of this structure by minimizing the total elastic energy density in the cross section of the NW using nextnano3.24 Assuming pseudomorphic growth, the calculated strain is used to determine the band edges and the effective mass Schrödinger equation is solved for the lowest energy states in the MQW. Figure 1(b) compares the computed elastic energy density at 10 K for QWs with fixed In-content of 30% for the two situations using the Al0.3Ga0.7As and In0.3Al0.3Ga0.4As buffers, respectively. For the AlGaAs buffer case, the core and buffer are only weakly strained, as expected, because most of the NW material consists of the nearly lattice-matched GaAs and AlGaAs. This leads to highly compressively strained In0.3Ga0.7As QWs (hydrostatic strain −2.1%). For an InAlGaAs buffer and barrier, the core and interlayers are highly strained, while the QWs experience a reduced hydrostatic strain of −0.8%. The corresponding low-temperature bandgap energy along the radial -direction is shown in Fig. 1(c) for both cases. Since the quaternary In0.3Al0.3Ga0.4As buffer induces a tensile strain on the GaAs core,25 we find that the bandgap of the GaAs core is significantly reduced to ∼1.3 eV. This value is only weakly affected by variations in core diameter typically observed in the experiment. Albeit the bandgap of InAlGaAs is also smaller than that of AlGaAs in the buffer and barriers, it has no adverse effect on the optically pumped system as shown below. We note, further, that the bandgap of the In0.3Al0.3Ga0.4As buffer increases slightly with radius due to the 2D NW cross section and the associated radial change of the strain. Most importantly, the structure employing the In0.3Al0.3Ga0.As buffer demonstrates the desired reduction of the bandgap of the QWs, as a result of the reduced compressive strain.
Figure 1(d) summarizes these combined effects on the lowest energy transition of both the GaAs core and InGaAs QWs for a wide range of In-contents in the QWs. The data show two prominent effects: First, the transition energy of the GaAs core decreases with increasing In-content in the QW similar to previous observations,25,26 and the effect is much stronger in the case of the InAlGaAs buffer (see dashed lines). Second, the transition energy of the InGaAs QWs in the strain-compensated InAlGaAs buffer structure is red-shifted by ∼60–70 meV compared to the case using the AlGaAs buffer structure (solid lines). The reduced compressive strain in the InGaAs/InAlGaAs MQWs is further expected to allow for enhanced In incorporation, as opposed to InGaAs/AlGaAs MQWs where the plotted transition energies for In-contents > 30% are rather hypothetical due to the confirmed plastic relaxation and suppressed luminescence.23 According to Fig. 1(d), NW-lasers using strain-compensated In0.3Al0.3Ga0.4As buffer and In-contents of ∼40% in the InGaAs/InAlGaAs MQW active region should, thus, yield room temperature emission near the telecom O-band (1.3 μm).
To test this prediction, NW-lasers were grown by molecular beam epitaxy (MBE) on SiOx-masked Si (111) wafers prepatterned by electron beam lithography and reactive ion etching.27 Similar to Refs. 23 and 27, vertical GaAs NW cores were realized via self-catalyzed vapor-liquid-solid (VLS) growth at a substrate temperature of 650 °C and a Ga flux of 0.5 Ås−1. After 5 min of prewetting under steady Ga flux, the As cell was opened for 15 min providing an As flux of 0.81 Ås−1. Subsequently, the As flux was increased to 1.63 Ås−1 to continue growth for another 75 min to obtain long GaAs NWs. After the core growth, the Ga shutter was closed to consume the liquid Ga droplet and the growth temperature reduced to 420 °C for the consecutive shell growth. Planar growth on the side facets of the NWs was facilitated, using a higher As flux of 24.4 Ås−1 and a total group-III flux of the quaternary buffer and MQW of 1 Ås−1 and 0.9 Ås−1, respectively. All fluxes are given as equivalent planar growth rates on GaAs (100).28 A scanning electron microscopy (SEM) image of a typical NW-laser is shown in Fig. 2(a) for a sample grown with In-content of 30% in the InGaAs MQW. Several other NWs measured from this sample show that the length is 12±1 μm and the diameter 500–600 nm, where size variations mainly stem from changes in GaAs core dimensions across the NW array. The NWs have smooth sidewall facets, which indicates that plastic strain relaxation did not take place.29
We use high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) and energy dispersive x-ray spectroscopy (STEM-EDXS) to confirm the expected material composition and thickness of each layer. For this, a thin lamella containing a radial cross-sectional cut from the middle of the NW was prepared using focused ion beam milling (FIB). As visible in Fig. 2(b), the NW shows a symmetric hexagonal shape and all individual layers can be clearly distinguished. Figure 2(c) shows the EDXS-measured alloy composition along the major facet of the NW [blue arrow in (b)]. The buffer layer has a uniform composition of In0.31Al0.30Ga0.39 As over its entire radius and in different directions, except for the six corner facets. Along these 1.7 nm thin facets, the Al molar fraction is nearly doubled (i.e., In0.19Al0.55Ga0.26As), an observation frequently encountered in AlGaAs-type shell systems.30,31 Opposing the trend seen for Al, both In and Ga species are consequently deficient in the corner facets, in agreement with recent InAlGaAs shell growth studies by Francaviglia et al.26 The MQW heterostructure is also quantitatively resolved by the EDXS scan, yielding uniform QW composition of In0.29Ga0.71As and barrier composition In0.23Al0.20Ga0.57As across the whole MQW structure. All layer compositions agree with the nominal values within the experimental error (±5%) of the STEM-EDXS analysis. In addition, the 2 nm thin GaAs interlayers between QWs and barriers are also resolved as peaks in the Ga molar fraction. This suggests that intermixing effects between QWs and barriers can be neglected.
Individual NW-lasers were optically pumped under pulsed excitation using confocal micro-photoluminescence (μ-PL) spectroscopy. Hereby, NWs from the same sample as in Fig. 2 were mechanically transferred to a sapphire substrate and excited in lying geometry with a mode-locked titanium-sapphire laser in a He-flow cryostat.9,23 Excitation was from the top using an elliptical spot to illuminate the entire NW uniformly (∼200 fs pulse length, 82 MHz repetition rate, 1.59 eV excitation energy). The emission was collected from one end-facet through the same objective.
Figure 3(a) shows a pump fluence dependent μ-PL measurement at 10 K of a single NW. At low pump fluence, the spectrum is dominated by two broad spontaneous emission peaks centered around 1.12 eV and 1.28 eV, respectively. In good accord with the simulated data [Fig. 1(d)], we attribute these peaks to the MQW active region (1.12 eV) and the highly tensile strained GaAs core (1.28 eV). Around the threshold pump fluence (Pth) of 4.0 ± 0.3 μJ cm−2, a sharp peak at 1.12 eV appears with a much larger intensity than the spontaneous emission background, characteristic for transition into lasing. For higher pump fluence still evenly spaced peaks appear that can be attributed to the same transversal mode. The lasing intensity is further analyzed by extracting the integrated peak intensities (areas) shown in Fig. 3(b) for the two strongest peaks (1.12 eV and 1.16 eV). Additionally, the total peak area of all Fabry–Pérot modes is also presented in the input–output characteristics. As expected, above the lasing threshold (4 μJ cm−2) the total peak area increases linearly and is dominated by the peak at 1.12 eV. Above 13 μJ cm−2, additional peaks contribute to the lasing process, which is evidenced by an increase in the slope of the total peak area. In fact, this unusual increase is governed by the modes above 1.16 eV, and we believe this behavior is caused by two distinct gain media. Moreover, the data directly illustrate the anticipated strain tuning effect, namely that the low-threshold lasing mode at 1.12 eV is well red-shifted (by ∼70 meV) with respect to the lasing energy of equivalent NW-lasers grown without a strain compensation buffer layer (see Ref. 23 and Fig. 4). Indeed, this shift agrees quantitatively remarkably well with the simulated data [cf. Fig. 1(d)], which predicts a shift of the QW ground state energy from 1.181 eV (AlGaAs buffer) to 1.118 eV (InAlGaAs buffer) for a QW In-content of 30% under the assumption of coherent, pseudomorphic growth. It is quite remarkable that such coherent growth can prevail throughout the NW-laser structure, considering that the InAlGaAs buffer thickness (90 nm) exceeds by far the equivalent planar critical thickness for strain relaxation on GaAs.32 We attribute this to the alleviated strain formation behavior seen in similar InGaAs/GaAs core-shell NW systems25,33 as well as kinetic effects, e.g., the employed low-temperature shell growth which increases the critical thickness.34–36 From our experiments, we cannot fully exclude the existence of dislocations in the buffer, but the observed shift in QW and core emission confirms that the strain state is close to the pseudomorphic model.
The strain compensation effects are further exploited to realize high-quality MQW active regions with larger In-content (In0.4Ga0.6 As MQW) under otherwise identical structure and growth conditions. Compared to the former case, the strain in the QW is thus slightly increased (hydrostatic strain of −1.35%), while the bandgap is lowered. NWs from this sample show clear lasing behavior, both at 10 K and room temperature, which is qualitatively very similar to the sample with In0.3Ga0.7As MQW. The pump fluence dependent μ-PL spectra in Fig. 4(a) at 300 K evidence a single broad peak around 1.06 eV at low pump fluence, while at higher pump fluence a second broad peak around 0.96 eV (1.292 μm) appears that evolves into a distinct lasing peak. Emission from the strained GaAs core is outside of the plotted spectral range and is located at 1.20 eV (simulation: 1.19 eV). We attribute both broad peaks to the MQW as already discussed for the previous sample in Fig. 3. Likewise, the total peak area shows the cumulative behavior of both MQW gain media, while peak areas of the individual input–output characteristics of the low-threshold, low-energy (0.96 eV), and a high-energy lasing peak are also depicted in Fig. 4(b).
Figure 4(c) shows an overview of the lasing energies (lowest threshold peak energy) at 10 K and 300 K for the different samples as well as their statistical distribution. For comparison, the horizontal lines denote the simulated ground state transition of the investigated MQWs. The map also includes recent experimental data for equivalent MQW-NW-lasers using Al0.3Ga0.7As buffer/barriers in the structure (green data, Ref. 23). As noted above, at 10 K, the change from Al0.3Ga0.7As to In0.3Al0.3Ga0.4As buffer leads to a clear redshift of the average lasing energy in In0.3Ga0.7As MQWs from 1.193 eV to 1.127 eV. The increase in In-content in the MQW to 40% reduces the average lasing energy further to 1.023 eV (simulated: 1.018 eV) due to the associated reduction in bandgap. The spread in lasing energy at 10 K is fairly low among the individual NW-lasers. Based on the bandgap narrowing with increasing temperature, the lasing energies at 300 K are further red-shifted, while the distribution of lasing energies gets wider, especially for the In0.4Ga0.6As MQW. Importantly, the quaternary buffer layer allows tuning the emission energy close to 1.3 μm, while simultaneously enabling room temperature operation, constituting a large extension of the operating regime of GaAs/InGaAs NW-lasers.
Figure 4(d) benchmarks the lasing threshold fluences for the same set of NW-lasers as in (c), all measured under similar experimental conditions. The different samples show distinctly different thresholds, indicating that their characteristics reflect the nature of their intrinsic structure. In contrast, any slight variations in the threshold seen within a given sample are usually attributed to random variations of the NW cavity, e.g., varying NW length and nature of the end-facet.37 Specifically, NW-lasers using quaternary InAlGaAs buffer/barriers exhibit overall very low lasing threshold at 10 K. The threshold increases on average from 4.4 μJ cm−2 to 10 μJ cm−2 upon increasing the In-content in the MQW from 30% to 40%, concurrent with lengthening of the lasing wavelength. Such an increase in lasing threshold with increasing wavelength is commonly observed in other strained InGaAs-MQW lasers and attributed to degrading optical quality of the InGaAs QWs.36 Likewise, we also observe a 5–10-fold increase in the threshold from 10 K to 300 K, which results from the faster non-radiative recombination and reduced peak gain at higher temperature (thermal broadening of the gain spectrum). More remarkable is the comparison with In0.3Ga0.7As MQW-NW-lasers using Al0.3Ga0.7As (green data), where the low-temperature values of the average threshold fluence are about two orders of magnitude larger. Here, the NW-laser structure and growth conditions were largely identical. (The average NW-cavity length was about 1.5× shorter for the NWs with AlGaAs buffer.) In addition, a smaller excitation spot was used to probe the sample with AlGaAs buffer, which increases the threshold pump fluence.23 Furthermore, pump absorption might be increased due to additional absorption in the barrier for NWs with an InAlGaAs barrier. We performed low excitation μ-PL measurements of all presented types of NWs standing on their native growth substrate, which suggest partial strain relaxation of the MQW for NWs with AlGaAs buffer (see the supplementary material). This reduced quality of the gain medium would explain an increased threshold and why no room temperature lasing has been observed for those NWs.23
To summarize, we have developed an advanced generation of GaAs-based multi-quantum well NW-lasers that exploit a strain-engineered In0.3Al0.3Ga0.4As buffer. Concerted experimental and modeling efforts showed that the quaternary buffer serves as a strain-compensating layer to lower the compressive strain and to enable large In-content in the respective InGaAs MQW. Thereby, room temperature lasing under optical pumping was achieved in the telecom O-band (∼1.3 μm) with threshold as low as 44 μJ cm−2, outperforming previous InGaAs/AlGaAs-based MQW-NW-lasers.
See the supplementary material for low-excitation μ-PL measurements of NWs standing on the native growth substrate.
The authors sincerely thank H. Riedl for experimental support and Dr. S. Schmidt for help with the preparation of the STEM lamella.
This work was supported financially by the ERC project QUANtIC (ID: 771747) funded by the European Research Council. Further support was provided by the Deutsche Forschungsgemeinschaft (DFG) via Project Grant Nos. KO-4005/6-1, KO-4005/7-1, and FI-947/4-1 and via Germany's Excellence Strategy-EXC2089/1-390776260 (e-conversion).
The data that support the findings of this study are available from the corresponding author upon reasonable request.