Vertical external cavity surface emitting lasers (VECSELs) have attracted considerable interest as versatile platform for a wide range of applications [1–3]. Compared to edge emitting lasers, VECSELs provide many advantages such as an excellent beam quality with almost circular emission and a small divergence as well as a scalable output power [4]. Moreover, the external cavity allows simple optical pumping using readily available high power diode lasers and provides easy control of the emission properties through intracavity elements such as birefringent filters, gratings, nonlinear crystals, and saturable absorbers [1–3]. Different materials have been employed as gain media in VECSELs. For example, III-V semiconductors have been used for emission in the visible and near infrared range ($\lambda <2.x\mathrm{\mu m}$) [2,5], whereas lead chalcogenides (PbTe, PbSe and PbS) were used for mid-infrared VECSELs [6–8]. For lasers, quantum dots (QDs) provide important advantages such as low thresholds, high temperature stability, and broad gain bandwidths [1,9]. Recently, QD-VECSELs have been demonstrated for the 655–1300 nm region [10–12]. The active regions were based on InGaAs or InGaP QDs produced by the Stranski–Krastanov (SK) growth mode, in which case, however, the emission is limited to wavelengths below 2 μm [13]. Alternative narrow gap III-V antimonide QDs exhibit a type II band alignment to the surrounding barriers and thus, rather weak optical transitions [14]. As a result, it has been impossible up to now to the best of our knowledge to obtain QD lasers with longer wavelength emission.

In the present work, we employ a novel material system for realization of QD-VECSELs, combining narrow gap IV-VI and wide gap II-VI semiconductors, to extend the emission well into the mid-infrared region. Mid-IR lasers are important for many sensing applications because of the strong “fingerprint” absorption lines of molecular species in this wavelength region. VECSELs are particular attractive for these applications since their emission wavelength can be tuned easily using intracavity elements or piezo-control of the cavity length [4,8]. In addition, optical pumping of VECSELs eliminates the need for doping of the active region and hence, and reduces free carrier absorption, which is an important limiting factor for mid-IR devices. Using a modular design [7], we demonstrate the first QD-VECSELs for the mid-IR 3.5 to 4.3 μm wavelength region with output powers up to $50{\mathrm{mW}}_{\mathrm{P}}$. This is the longest wavelength achieved not only for QD-VECSELs but also for all QD lasers.

The PbTe/CdTe QDs were produced by a two-step synthesis method, where first few-nanometer thick 2D PbTe layers are epitaxially grown on CdTe at low substrate temperatures, followed by an annealing step at higher temperatures around 350°C [15]. During annealing, the 2D PbTe layers break up into isolated QDs to minimize the interface energy as described in detail in our previous works [15,16]. This transition is driven by phase separation and nanoprecipitation, caused by the large miscibility gap between PbTe and CdTe. As a result, highly symmetric PbTe QDs with nearly spherical shape and atomically sharp interfaces are obtained [15]. Because of the large bandgap difference between CdTe ($E_g=1.46\mathrm{eV}$) and PbTe ($E_g=0.32\mathrm{eV}$), a strong carrier confinement is achieved despite the asymmetric band alignment [17]. The QD size can be tailored by the growth and annealing conditions through which the emission wavelength can be tuned over a very wide range from 1.6 to 5 μm [18,19]. Since the lattice constants of PbTe ($a_0=6.46\AA$) and CdTe ($a_0=6.48\AA$) are nearly equal, strain-free multilayer QD structures with many periods can be grown without defect formation—contrary to the case of highly strained Stranski–Krastanov QD systems.

The optical cavity of our VECSELs is formed by two high-reflectivity Bragg mirrors grown by molecular beam epitaxy onto optically transparent ${\mathrm{BaF}}_2$ (111) substrates. The mirrors consist of quarter-wavelength thick ${\mathrm{BaF}}_2/{\mathrm{Pb}}_{0.92}{\mathrm{Sr}}_{0.08}\mathrm{Te}$ layer pairs and were designed for a center wavelength of 3.5 μm. Owing to the extremely large refractive index contrast between ${\mathrm{Pb}}_{0.92}{\mathrm{Sr}}_{0.08}\mathrm{Te}$ ($n=5.25$) and ${\mathrm{BaF}}_2$ ($n=1.46$), ultrabroad stop bands with very high reflectivities are already obtained with a few Bragg pairs. In our VECSELs (see Fig. 1), the planar bottom mirror consists of 4.5 pairs, leading to a reflectivity above 99.9%, whereas 2.5 pairs were used for the external curved mirror to act as output coupler (R $\sim 99.8\%$). The active region with the PbTe/CdTe QDs was grown on a separate substrate clamped on top of the bottom mirror (see Fig. 1). This modular design [7] allows us to test different active region designs with already existing mirrors, which facilitates the comparison of different active regions on the VECSEL performance.

 

Fig. 1. Schematic illustration of the modular VECSEL setup. The bottom Bragg mirror and the active structure are grown on two different substrates. The active region on GaAs consists of PbTe QDs embedded in CdTe and is optically in-dot pumped with a 1.064 μm pump laser.

Download Full Size | PDF

The active PbTe/CdTe QD structures were grown on standard GaAs (001) substrates. After native oxide desorption at 600°C, the substrate temperature was lowered to 350°C and a thin ZnTe layer ($<1\mathrm{nm}$) was deposited to bridge the 12% CdTe/GaAs lattice mismatch. This was followed by the growth of a 200 nm CdTe buffer layer during which the reflection high-energy electron diffraction pattern becomes 2D streaked, revealing nice layer-by-layer growth. Multilayer stacks of up to 100 PbTe and CdTe layers were subsequently deposited at a temperature lowered to $\sim 275°\mathrm{C}$ to suppress CdTe desorption and vertical PbTe segregation to obtain layers with precisely defined thicknesses required for the optical design of the laser structure. The thickness of the PbTe layers was fixed to 5 nm, whereas the thickness of the CdTe barrier and spacer layers was varied between 35 and 360 nm. Eventually, QD formation was induced by post-growth annealing at 350°C for 20 min with a thin Si (5 nm) cap layer deposited on top to prevent CdTe desorption during annealing. This procedure results in PbTe QDs with an average diameter of $\sim 22\mathrm{nm}$ and size dispersion of around $\pm 12\%$, as determined by transmission electron microscopy (TEM) (see Fig. 2). The QD density per layer is $\sim {10}^{11}{\mathrm{cm}}^{-2}$, which is comparable to that of III-V SK QDs used for NIR lasers.

 

Fig. 2. Cross-sectional TEM images at different magnification of the active region of the VECSELs, consisting of 24 PbTe QD layers divided in five groups placed at the antinodes of the laser intensity in the active region (see Fig. 3). The as-grown 2D PbTe layers were 5 nm thick and were separated by 35 nm CdTe barrier layers, which after the annealing step, yields QDs with an average diameter of 22 nm. The spacers between the QD layer groups vary between 320 and 360 nm.

Download Full Size | PDF

For self-assembled QDs, the material gain is generally reduced by the dot size distribution, and the gain of the ground state transition easily saturates when the energy levels are filled with electrons or holes. Consequently, a sufficiently large number of QD layers with a suitable optical confinement is needed to compensate the losses in the device. To this end, we have fabricated and tested active regions with the numbers of QD layers varying between 10 and 100. Figure 3 shows the actual active region design of our VECSELs with in this case 24 QD layers placed near the antinodes of the laser intensity distribution (solid line). To meet this condition, the QDs were grouped into five closely spaced QD layers, and each group was separated by thicker CdTe spacer layers of 320–360 nm (see Figs. 2 and 3).

 

Fig. 3. Active region design of the VECSEL containing 24 PbTe QD layers divided in five groups. The laser intensity profile is represented by the solid line. It is enhanced at the center of the active region because the high average refractive index of the outer QD groups leads to an effect similar to a Bragg mirror.

Download Full Size | PDF

The intensity distribution inside the active region was calculated using the transfer matrix method for a design wavelength of ${\lambda }_D=3.6\mathrm{\mu m}$, corresponding to the QD emission at $\sim 250\mathrm{K}$. Contrary to III-V VECSELs where the QWs and barrier layers are made of similar materials, in our VECSELs there is a large refractive index contrast between PbTe and CdTe of $n_{\mathrm{PbTe}}\sim 6$ versus $n_{\mathrm{CdTe}}\sim 2.7$. Consequently, the average refractive index of the QD groups is high compared to that of the CdTe spacer layers. Therefore, the exact positions of the QD groups must be tailored to achieve a resonant gain structure. In our current design, the position of the QD groups was adjusted to have an effect similar to a Bragg mirror, leading to laser intensity enhancement at the center of the active region, as indicated by the intensity profile (solid line) in Fig. 3. The QDs were optically pumped with a 1.064 μm laser focused to a spot with a diameter of 100 μm. Since the CdTe bandgap of 1.46 eV is larger than the photon energy of the pump laser (1.17 eV), the pump power is exclusively absorbed by the QDs (in-dot pumping scheme), i.e., no carriers are excited in the barriers.

Normalized laser spectra of the QD-VECSEL with 24 QD layers are shown in Fig. 4 at different heat sink temperatures. The spectra were measured in pulsed mode with 120 nsec pulses and 10 kHz repetition rate. The multimode emission reflects the broad gain bandwidth of the QDs because of the size dispersion [19]. The $4{\mathrm{cm}}^{-1}$ mode spacing is caused by interference effects because of the 380 μm thick intracavity GaAs substrate [7]. The laser emission shifts from 4.3 μm at 85 K to 3.6 μm at 230 K because of the increase of the PbTe bandgap with temperature, $d{E}_g/dT\sim 0.4\mathrm{meV}/K$ [19]. The absorbed threshold power, which in our case is $\sim 50\%$ of the incident pump power, is shown in Fig. 5(a) as a function of temperature. At 85 K, the absorbed threshold is below $100{\mathrm{mW}}_{\mathrm{P}}$ peak power (${\mathrm{W}}_{\mathrm{P}}$), which increases to $2{\mathrm{W}}_{\mathrm{P}}$ at 240 K with a characteristic temperature of $T_0\sim 50\mathrm{K}$ (solid line). It is to be noted that because of the QD size variation, only few QDs with a specific size actually contribute to the gain of a single laser mode. If we assume an inhomogeneous gain broadening of 30 meV, as derived from photoluminescence measurements, and a homogenous gain broadening of the order of 0.01–1 meV for the individual dots, the effective threshold power of one single laser line is actually about 30–3000 times lower than the threshold power of the whole device. This reveals the high potential of QDs over QW systems and indicates that the threshold can be significantly reduced by decreasing the size fluctuations.

 

Fig. 4. Normalized laser emission spectra of the PbTe QD-VECSEL at different heat sink temperatures between 86 and 230 K. The laser emission covers a wavelength range from 3.6 to 4.2 μm.

Download Full Size | PDF

 

Fig. 5. (a) Threshold power of the PbTe QD-VECSEL containing 24 QD layers plotted as a function of the heat sink temperature. The characteristic temperature T₀ of the VECSEL is around 50 K. (b) Output power of the PbTe QD-VECSEL in dependence of the absorbed pump power at different heat sink temperatures.

Download Full Size | PDF

Figure 5(b) shows the output power in dependence of the absorbed pump power at different heat sink temperatures. At 140 K, the threshold is around $400{\mathrm{mW}}_{\mathrm{P}}$ and the pulsed output power more $50{\mathrm{mW}}_{\mathrm{P}}$. At higher operation temperatures, the power characteristics show a thermal rollover, explaining the highest operation temperature of 240 K of the VECSEL. It is important to note that under our excitation conditions, the quantum deficit between the VECSEL wavelength ($h\nu \sim 0.35\mathrm{eV}$) and the pump wavelength ($h\nu \sim 1.17\mathrm{eV}$) is rather large, which strongly contributes to undesired heating of the device because of the limited thermal conductivity of CdTe of only around $1–10\mathrm{W}/\mathrm{mK}$. Thus, laser operation up to room temperature might be reachable by an improved thermal management using SiC or diamond heat spreaders on top of the active chip. On the other hand, optical losses caused by free carrier absorption in the CdTe barriers are negligible because of the in-dot pump scheme and the semi-insulating properties of CdTe.

Figure 6 shows the highest operation temperature of our VECSELs as a function of the number of QD layers in the active region. We find that at least 20 QD layers are required for lasing at temperatures above 100 K. For 24 or more QD layers, the VECSELs operate up to 240 K with little change when the number of QD layers is further increased. Even for 100 QD layers, the highest operation temperature is still 200 K, which is quite surprising since the first 20–30 QD layers at the front absorb more than 80% of the pump power entering the active region. Thus, QDs at the far end of the active region are only weakly pumped and hence, are expected to act rather as absorbers reducing the laser performance than as a gain medium. However, the absorption of unpumped QDs seems to be weak because of the Burstein–Moss shift caused by residual n-doping because of Cd incorporation [20]. Indeed, Hall measurements of annealed PbTe layers ($\sim 20\mathrm{nm}$) in CdTe reveal an n-doping of $\sim 1\times {10}^{18}{\mathrm{cm}}^{-3}$, which supports the assumption of a Burstein–Moss shift in PbTe QDs.

 

Fig. 6. Highest operation temperature of the QD-VECSELs in dependence of the number of QD layers in the active region. For lasing above 100 K, at least 20 QD layers are required and the highest operation temperature of around 240 K is reached for 24 or more QD layers.

Download Full Size | PDF

Since our PbTe QDs are embedded in a material providing very high barriers for the charge carriers, detrimental non-radiative Auger recombination processes similar to those in colloidal lead salt nanocrystal systems are expected. For such nanocrystals a quite universal size dependence of the Auger coefficient was reported [21], where the Auger coefficient $C_A\sim 1\times {10}^{-9}{\mathrm{cm}}^3{\mathrm{s}}^{-1}\times {\mathrm{R}}^{-3}$ increases proportionally to the cube of the QD radius $R[\mathrm{cm}]$. Extrapolating the Auger coefficient to our PbTe QD size yields an Auger coefficient of around $1\times {10}^{-27}{\mathrm{cm}}^6{\mathrm{s}}^{-1}$, which is about a factor 10 larger than for PbTe bulk [22]. This indicates that Auger losses are mainly responsible for the moderate characteristic temperature of our devices.

This conclusion is supported by our photoluminescence measurements which reveal a strong monotonic decrease of the quantum yield of the QDs with increasing pump power. In fact, our VECSELs realized with smaller PbTe QD sizes $<22\mathrm{nm}$ also show a reduced laser performance and in the case of QDs smaller than 14 nm no lasing was obtained for our devices. For small QDs, however, the pump laser absorption also decreases because of our in-dot pumping scheme, i.e., for small QDs barrier pumping could significantly improve the performance of the device. Since Auger losses $C_A\times {(N_{e-h}/V_{\mathrm{QD}})}^3\sim 1\times {10}^{-11}{\mathrm{cm}}^3{\mathrm{s}}^{-1}\times N_{e-h}^3/R^6$ are actually inversely proportional to $R^6$ (where $V_{\mathrm{QD}}$ are the QD volume and $N_{e-h}$ the average number of excited electron-hole pairs), tuning of the laser emission to shorter wavelengths by decreasing the QD size would result in increased Auger losses that strongly degrade the laser performance. Thus, tuning of the laser emission by changing the composition of rather large QDs is probably more effective. This can be easily achieved by alloying the PbTe QDs with SrTe or SnTe, which shifts the QD emission to shorter, respectively, longer wavelengths [18,23].

In summary, we have demonstrated the first QD VECSELs operating in the mid-IR. The active regions consist of PbTe QDs embedded in a CdTe host matrix. By optimizing the layer structure, number of QD layers as well as the optical design, operation temperatures as high as 240 K have been achieved, which allow for thermoelectric cooling of the devices. The laser emission covers a wavelength range from 4.3 to 3.6 μm in dependence of the operation temperature. At 85 K, the threshold power is as low as $60{\mathrm{mW}}_{\mathrm{P}}$ and the output power more than $50{\mathrm{mW}}_{\mathrm{P}}$. While the threshold power could be decreased by reducing the QD size dispersion, the broad gain bandwidth of the QDs opens the way for a large tuning range of the laser emission by modulation of the cavity length or intracavity elements, which is of great interest for molecular spectroscopy applications.