1. Introduction

Microcavity semiconductor lasers employing high spontaneous emission coupling exhibit interesting and potentially useful peculiarities such as “thresholdless” operation [1,2]. In particular, microlasers utilizing high Q-factor cavities of very-small mode volumes allow for exploiting cavity quantum-electrodynamics (QED) effects that can improve greatly the laser performance [3] yielding devices with very low (<µW) power consumption. Yet, it becomes only possible if the amount of gain material is sufficiently small [2]. In this context, semiconductor nanostructures such as quantum wires (QWRs) and quantum dots (QDs) are very attractive active media for microcavity lasers since they can also produce high optical gain, confine charge carriers into very small volumes and provide better spatial and spectral matching with the cavity modes. This, however, crucially depends on fabrication technologies that could provide site- and spectrally-controlled wires and dots with high optical quality.

Since the proposal of the concept [4], the first microlasers were based on Bragg-reflector [5,6] and microdisk [7] optical cavities employing quantum-well (QW) [5,7] or array-QWR [6] active regions, showing threshold power of 0.2-1 mW and relatively low spontaneous-emission coupling factors (β~0.01-0.0001). Only rather small-diameter (~2µm) microdisk devices demonstrated laser thresholds at ~60 μW with β~0.1 [7]. More recently, photonic crystal (PhC) structures have been identified as the most promising microcavities for achieving ultimate semiconductor laser performance [3], due to their substantial potential in spontaneous-emission control. Demonstrated within the past decade [8–12], PhC microlasers based on QWs and randomly-nucleated QDs have yielded high-speed performance [9] and threshold powers below 1 μW with β as large as ~0.7 [10,12]. Most recently, single-QD lasers using PhC [13] and micropillar cavities [14] have also been successfully attempted.

Clearly, for given optical cavity losses, the volume of the active region needs to be adjusted in order to achieve a desired level of optical output power. Moreover, as the light-matter coupling efficiency depends on the spectral and spatial matching of the emitter to the confined optical field [15], good spectral and spatial control of the gain elements is mandatory. Although microlasers incorporating a single QD represent in some sense an “ultimate limit” of semiconductor laser miniaturization, QD ensembles are needed in order to achieve useful output power. Semiconductor QWRs appear attractive, in this context, since they also may offer more efficient carrier capture routes and polarization selectivity [16].

Here we demonstrate optically pumped single-mode PhC microcavity lasers utilizing site-controlled QWR active medium, and analyze their static and dynamic lasing characteristics. By comparing temporal, spectral and input-output behavior versus pump level, we derive criteria for establishing lasing and threshold. QWR-PhC lasers with <1µm long wires, a threshold of ~240nW in absorbed power and β~0.3 are presented.

2. The integrated quantum-wire/photonic-crystal system

Our QWR-PhC laser system consists of an L3m [17] or L6 PhC cavities in a 265nm-thick GaAs membrane incorporating ~1μm-long InGaAs/GaAs V-groove QWRs as the gain medium (Fig. 1 ). The structure is fabricated using electron-beam lithography and metal-organic vapor-phase epitaxy (MOVPE) as described in detail in [18,19]. The active region comprises three vertically-stacked wires [Fig. 1(b)]. Lateral alignment precision of ~40 nm with respect to the optical mode is achieved.

 

Fig. 1 (a) 3D-rendered AFM image of an L6 PhC cavity. (b) TEM cross-sectional image of the embedded stack of 3 InGaAs/GaAs QWRs. (c) Low-temperature µPL spectrum of the “bare” QWRs (without cavity). (d) Top-view SEM image of an L3m PhC cavity. (e) Computed optical-field distribution of the fundamental cavity mode, E_t = (E_x² + E_y²)^1/2.

Download Full Size | PDF

The laser structures were characterized at low temperature using micro-photoluminescence (µPL) spectroscopy (excitation at 700-730 nm, ~1.5µm spot, ~3ps pulses of Ti:Sapphire laser at ~80 MHz). A characteristic spectrum of the bare QWRs (before embedding in cavities) shows wire emission near 890 nm, which can be tuned by the growth parameters. The narrow spectral linewidth (~7 nm) attests to the good uniformity of the QWR potential [Fig. 1(c)]. The PhC patterns were designed (by 2D finite-difference and 3D FDTD methods) such that a fundamental cavity mode within the TE0-like PhC bandgap matches the QWR emission, yielding typical pitch of ≈198 nm with radius/pitch ratios of ~0.24-0.26.

Considering the QWR gain bandwidth represented by the spectrum in Fig. 1(c), PhC L6 or L3m cavities are effectively single-mode. Before analyzing the QWR-PhC lasers, it is hence interesting to explore possible QED effects in this system. To this end, time-resolved µPL spectra of the QWRs inside or outside the PhC cavities were studied using time-correlated photon counting (temporal resolution ~60 ps) detection. Figure 2 compares the measured PL transients for the bare QWRs and for those embedded in a non-lasing L3m cavity, either resonant or non-resonant (blueshifted by 13 nm) with respect to the QWR peak. All transients were detected with a 0.11-nm wide spectral window.

 

Fig. 2 Transient µPL spectra of the bare QWRs (grey) and the QWRs embedded in a PhC L3m, resonant (non-lasing) cavity (green) and non-resonant one (red). Inset shows the emission spectrum. Pump power is 11.5 µW. (T = 20K).

Download Full Size | PDF

The bare QWR emission decay is rather slow (τ_d~0.9ns) suggesting pronounced disorder-induced exciton localization effects [20]. The QWR emission coupled into the cavity mode is enhanced by a factor of ~2-2.5 (as measured from several L3m, L6 structures). The “non-resonant” QWR emission rate is, on the contrary, inhibited by a factor of ~2-3 due to the reduced photonic density of states. Note that on resonance the decay contains a second, slower component (τ_d2~1.9 ns), having the same decay rate as that of QWR background (detected off the resonance). The enhancement factor, if compared to the potential cavity Purcell effect, is somewhat low due to ensemble averaging in the measured data [15]. However, the QWR emission into the cavity mode is enhanced at least by a factor of 5-6 as compared with the off-resonant (background) QWR emission, which is a favorable condition for a microlaser [2].

4. Observation of stimulated emission and lasing

Figure 3 presents the emission spectra and light input-output curves for several characteristic L3m and L6 devices. The power-dependent emission spectra are dominated by a well-defined cavity mode above the pump level of ~1µW [Fig. 3(a) and (b)]. However, the measured input-output [Fig. 3(c)] and the derived differential efficiency [Fig. 3(d)] allow for distinguishing two different types of devices, labeled here “lasing” and “non-lasing”. In particular, whereas for the lasing devices the slope efficiency increases with power tending to a constant value – as would be expected for semiconductor lasers, for the non-lasing ones it rather declines. Interestingly, such non-lasing structures that may have laser-like spectral (e.g. narrow linewidth) properties can also be interesting for applications as single-mode light-emitting diodes [21].

 

Fig. 3 Spectra of (a) lasing QWR-PhC device (L3m cavity), and (b) non-lasing device (L6 cavity, inset illustrating QWR emission background, log scale). T = 50K. (c) Comparison of characteristic input-output curves (both cavities are L6). (d) Slope efficiency derived from (c).

Download Full Size | PDF

Further indications of stimulated emission and lasing are evident in the evolution of the transient PL signal as a function of pump power. Figure 4 compares the transient µPL signals from two L3m devices pumped at various power levels. A non-lasing device (“identified” from the input-output characteristics) exhibits only a single-exponential decay with time constant of τ_d~400 ps [Fig. 4(a)]. In contrast, for a similar but lasing structure [Fig. 4(b)], a faster component (τ_d<50 ps) is detected “on top” of a slower one (τ_d~350 ps) above a certain power level. As the estimated Purcell factor (compared to “bare” QWR) is only ~2-3, the faster component cannot stem from enhanced spontaneous-emission rate, it thus appears due to stimulated emission. Moreover, the delay time τ_delay between the excitation pulse (measured via IRF, the instrumental-response function) and the cavity response increases and then decreases with increasing pump level [Fig. 4(c)]. Both dynamic effects occur near a soft turn-on in the input-output curve, ~2 μW of average pump power for this particular lasing device.

 

Fig. 4 Transient micro-PL at different excitation levels collected from a non-lasing (a) and lasing (b) L3m device. (T = 20K). (c) Input-output curve and delay time (see inset); based on that [22] (see text), lasing threshold is indicated.

Download Full Size | PDF

The observed dynamic effects were interpreted by solving the carrier-photon coupled rate equations [2] assuming pulsed excitation [22]. The calculations of the time-dependent mean photon number (Nph) reproduce the fast decay component due to the dominant role of the stimulated emission above a certain threshold, as well as the local maximum in the delay time around threshold. Similar characteristics have also been observed in microdisk and in QD PhC-cavity lasers [23,24]. Using the threshold definition of Nph = 1 [2], this analysis yields ~2 μW threshold for these particular L3m devices. Note that such threshold is below the local maximum in delay (τdelay)max, whereas the conventional definition (based on the extrapolation from the input-output curve) is consistent with threshold exactly where τdelay = (τdelay)max.

Important indications of the laser phase-coherence build-up and threshold can be gained from the dependence of linewidth on the excitation level. In semiconductor lasers, this trend is complicated by effects of absorption saturation below threshold and because spontaneous emission and intensity fluctuations are coupled to phase variations via the linewidth enhancement factor α [25], leading to non-monotonous variations of linewidth with pump level. Observed experimentally [26,27], this served as a good indication of lasing threshold for high-β QD-PhC microlasers [10–12]. Figure 5(a) presents the linewidth measured in our QWR-PhC L3m, L6-cavity microlasers confirming expectations, namely, narrowing at very low powers due to absorption saturation, broadening because of α [25], and further narrowing at high pump powers where intensity fluctuations saturate [28] and the stimulated emission prevails. Note that at very low powers (as measured at ~0.005 μW, for devices not shown here) the Q-factor reduces to ~2000. It’s also worth noting that, as the α-factor is spectral-position [26] and medium-dimensionality [29] dependent, and since in high-β lasers the line broadening is less pronounced [25], we sometimes observed a plateau rather than the line broadening within the threshold region (as also in [10,12]).

 

Fig. 5 (a) Linewidth (inverse normalized units) vs pump power for lasing and non-lasing L3m, L6 QWR-PhC devices (see also Fig. 3). (b) Estimated β-factor derived from fitting to the input-output curve of an L3m device [Fig. 3(a)]. Inset shows the input-output curve on a linear scale.

Download Full Size | PDF

The input-output curves of lasing devices were fitted applying the rate equation model of [2] using β as a fitting parameter. The following parameters were used: active-medium volume V_a = 3 × (5 × 5 × 1000) nm³ (3 stacked QWRs), non-radiative recombination time constant τ_nr = 1.25 ns (inferred from QWR luminescence measured at 70K from non-resonant cavities), spontaneous-emission lifetime τ_sp = 300 ps, Q = 5000, transparency carrier density N_tr = 10¹⁸cm⁻³, and a linear gain model [2]. The results are shown in Fig. 5(b), where the best-fit parameter is β = 0.3. Similar analysis of several L3m, L6 QWR-PhC cavities yields β~0.3 ± 0.1, inferred as a conservative estimate. The fits are reliable over about two decades of the input-output curve. The divergence at high pump level is due to emission saturation [30]; at low power it is possibly owing to reservoir effect, as the QWRs are pumped via bulk-GaAs and QW barriers.

For the studied QWR-PhC lasing devices, different criteria – input-output curves, photon dynamics and linewidth trends – lead to consistent threshold observations, with average pump levels of ~0.2-2µW in incident power, or ~10000-80000 photons per pulse. Considering the GaAs membrane thickness of 265nm, α_A = 10⁴ cm⁻¹ for the GaAs absorption coefficient and reflectivity of R = 0.3 at the air-GaAs interface, we estimate that ~22% of the incident power is delivered to the QWRs, yielding threshold, on average, as low as 240 nW. Cavities similar to this work but incorporating high-density, self-assembled QDs exhibited laser thresholds of ~1 μW (incident power), with ~100 QDs contributing to the emission [12]. In our QWR-PhC structures, although localized excitons are observed at low pump powers (with an estimated localization length of ~20nm), it is nevertheless expected that the gain is due to Coulomb-correlated electron-hole plasma [31], since “bare”-QWR emission peak blueshifts at high pump powers and since lasing was observed more often in cases of the cavity mode blueshifted with respect to the low-power QWR-emission peak. The total QWR active volume in our L3m, L6 devices, however, is equivalent to some 100 localization sites.

5. Summary and Conclusions

We have demonstrated PhC-based microcavity lasers that employ site-controlled QWR active medium. Comparing the emission spectra, input-output characteristics and transient photon dynamics of lasing and non-lasing devices, we establish lasing at low temperatures with threshold (absorbed) pump powers as low as ≈240 nW and β-factors of ~0.3. Preliminary results at 70K point to faster laser dynamics, however the performance decreases due to non-radiative recombination, most likely arising from QWR exciton diffusion [32]. Prospectively, surface passivation [9] is envisaged to yield laser operation at higher temperatures.